Shock-absorbing flooring for fall-related injury prevention in older adults and staff in hospitals and care homes: the SAFEST systematic review

Article history

The research reported in this issue of the journal was funded by the HTA programme as project number 17/148/11. The contractual start date was in February 2019. The draft report began editorial review in August 2020 and was accepted for publication in January 2021. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this report.

Copyright statement

Copyright © 2022 Drahota et al. This work was produced by Drahota et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. This is an Open Access publication distributed under the terms of the Creative Commons Attribution CC BY 4.0 licence, which permits unrestricted use, distribution, reproduction and adaption in any medium and for any purpose provided that it is properly attributed. See: https://creativecommons.org/licenses/by/4.0/. For attribution the title, original author(s), the publication source – NIHR Journals Library, and the DOI of the publication must be cited.

2022 Drahota et al.

Description of the health condition

Falls, commonly defined as ‘an unexpected event in which the participants come to rest on the ground, floor, or lower level’,2 are a serious and growing threat to global trends in healthy life expectancy. 3 They have significant morbidity, mortality and economic impacts, particularly for older adults who are more at risk of both falls and fall-related injuries. 4,5 Falls constitute the most commonly reported safety incident in hospitals,5 with prevalence rates estimated to be two to three times higher in care settings (hospitals and care homes) than in the community. 6 Inpatient falls account for approximately one-quarter of the £2.3B cost of falls to the NHS each year,5,7 with falls in care homes contributing considerably to this cost.

The causes of falls are complex and multifaceted, and may stem from a combination of intrinsic risk factors related to the person (e.g. eyesight, mobility, cognition, comorbidities), extrinsic risk factors associated with the environment (e.g. trip/slip hazards, footwear, clothing, medication, disorientating environment, staffing), and the activity of the individual (e.g. mobilising from bed, going to the toilet). 8–10 Similarly, fall-related injuries are a product of an individual’s susceptibility to injury (e.g. as a result of bone strength, physical and cognitive impairments), the environment in which they fall (e.g. the surfaces they come into contact with, the presence of others) and the dynamics of the fall itself. 11,12

Fall-related injuries can vary from pain and bruising to lacerations, sprains, head injuries and fractures. Approximately 72% of inpatient falls may result in no visible harm;5 it is still vital to prevent these falls whenever possible because of their impact on the subsequent falls risk, behaviour and psychology of the individual. 13,14 Injurious falls have additional consequences, for example affecting independence and ongoing care needs. 15,16 In 2018 in the NHS, 2439 hip fractures occurred in inpatient settings, constituting 3.8% of all hip fractures in England and Wales. 17 In addition to having a higher 30-day mortality risk,17 those sustaining an inpatient hip fracture are at increased risk of moving into nursing homes or residential care to meet their ongoing care needs, and can experience a substantial loss of healthy life-years. 18 Although it is recognised that there is no single solution to prevent all falls, increasingly there is an emphasis placed on finding ways to reduce the number of severe falls.

Description of flooring interventions

Owing to the complex nature of falls, a wide range of interventions exists to try and prevent them. For example, in hospitals and care homes, there is low-quality evidence with uncertain conclusions to support the use of exercise, physiotherapy, sensor alarms and multifactorial interventions tailored to individuals’ risk factors. 19 Shock-absorbing flooring is an additional potential intervention for the prevention of injurious falls; however, although the evidence-base for flooring interventions has been mapped in a scoping review,20 it has not yet been systematically reviewed. By decreasing the stiffness of the ground surface, shock-absorbing (or compliant) flooring aims to reduce the impact forces of falls to lower the risk of injury. In the care sector, commonly used flooring materials may vary in stiffness; for example, some carpets may offer someone who falls a softer landing than a thin vinyl, and wooden subfloors may be less severe to land on than concrete subfloors. Yet the health and care sector is beginning to think beyond these standard flooring types, turning to the sports sector. A multitude of sports floors exists, designed to offer shock-absorbency for the protection and comfort of players in sports halls, which could be repurposed for use in hospitals and care homes. In addition, some specialist flooring manufacturers have recognised the gap in the market for purposefully designed shock-absorbing ‘health’ floors and novel flooring interventions are now available that have been designed with fall-related injury prevention in mind.

‘Shock-absorbing’ floors, therefore, can differ with regard to their intended purpose, thickness, material choice and composition, and will vary with regard to the level of shock-absorbency they offer. Some manufacturers offer shock-absorbing underlays, which are designed to be used in conjunction with a regular overlay material (such as a standard vinyl), and other manufacturers provide a complete flooring system that can be laid on a concrete or wooden subfloor.

Why it is important to do this review

Using the floor as an intervention to prevent injurious falls in care homes and hospitals has certain appeals, as, unlike other injury prevention interventions (such as hip protectors or helmets), it has the potential to ‘treat’ all those that come into contact with it (and whichever body part may come into contact with it), without necessarily requiring any active engagement or compliance from the user (patients/residents or staff). Once installed, a floor may be expected to last up to 20 years, which, if effective at reducing hip fractures and other injuries, offers the potential for a significant return on investment. 21–24 Yet the decision to implement a shock-absorbing floor is not a straightforward one. Softer floors can create greater rolling resistance for wheeled objects, making it harder to initiate and sustain movement when pushing or pulling furniture like beds, trolleys, hoists, etc. 25,26 Considerably greater forces and awkward work positions have been reported for pushing and/or pulling wheeled equipment on soft surfaces like carpet, compared with hard surfaces like linoleum,27,28 and the performance of pushing and pulling activities is strongly associated with the prevalence of musculoskeletal injuries in the health-care sector. 29,30 In this respect, there is a risk that staff may experience more adverse events when working on shock-absorbing floors.

Furthermore, although a potential benefit with regard to reducing musculoskeletal symptoms where workers must stand or walk for long periods has been reported,31 concerns exist as to whether or not a more compliant surface underfoot would introduce more instability in people who are already at higher risk of falling. Thus, in an attempt to decrease the proportion of falls resulting in injury, the implementation of a shock-absorbing floor may inadvertently increase the overall number of falls experienced. In laboratory-based research, a debate is ongoing as to whether or not the gait of older adults (particularly those individuals with complex health needs) may be adversely affected by softer floors,32–43 and yet evidence also suggests that older adults would benefit most from falling on softer floors. 33,44–48 The potential benefits and risks of shock-absorbing floors may differ depending on the individuals using them; however, clinical evidence is required to better inform our understanding of this issue.

A scoping review identified all of the published evidence on shock-absorbing flooring up until May 201620 and since then further studies have emerged. 49–54 We have undertaken this systematic review because, to our knowledge, the growing body of evidence has not yet been critically appraised or synthesised, with an exploration of different care settings and flooring types, to help resolve the current uncertainties, to better inform NHS investment decisions and to clearly identify the next steps for research activity.

Aims and objectives of the research

In this mixed-methods review (including randomised, non-randomised, qualitative and economic studies), we aimed to systematically review the evidence on shock-absorbing flooring use in care settings (hospitals and care homes) for fall-related injury prevention in older adults. The objectives were to:

• assess the potential benefits (fall-related injury prevention) and risks (falls, staff injuries) of different flooring systems in care settings

• assess the extent to which these potential benefits and risks may be modified by different studies/settings, intervention and participant characteristics

• critically appraise and summarise current evidence on the resource use, costs and cost-effectiveness of shock-absorbing flooring in care settings for older adults, compared with standard flooring

• summarise findings on the implementation of flooring interventions in the included studies

• summarise the views and experiences of shock-absorbing flooring use from staff, patients’, residents’ and visitors’ perspectives

• identify gaps in existing evidence.

Eligibility criteria

Theoretical framework

Figure 1 conceptualises the causal pathway between shock-absorbing floor systems and their outcomes (falls, fall-related injuries, adverse events – staff injuries), and potential (often related) moderators of that relationship (effect modifiers). We developed this framework during the initial stages of the review in consultation with our public involvement members. The purpose of this framework was to help direct the review process by informing data collection, risk-of-bias assessment (particularly in relation to confounding), exploration of heterogeneity and analysis of the data.

FIGURE 1.

Theoretical framework of potential effect modifiers.

Outcomes and prioritisation

There is no core outcome set specifically for shock-absorbing flooring interventions. A common outcome data set for fall injury prevention trials has been developed; however, it focuses on community-dwelling populations. 2 In addition, an international consensus statement for trials on hip protectors has also been developed. 65 In developing our outcomes, we considered these related core outcome sets (and their differences in foci with the current review), discussions with our public involvement group, wider stakeholder engagement activities66 and the peer review comments we received on our protocol. 67 Following these considerations, we opted to focus on the following outcome measures, in the following order of priority:

• primary outcomes –

  1. injurious falls rate per 1000 person-days

  2. falls rate per 1000 person-days

• secondary outcomes –

  1. number of falls resulting in injuries (e.g. none, minor, moderate, severe, death)

  2. number of fractures

  3. number of hip fractures

  4. number of fallers

  5. number of adverse events (staff injuries)

  6. number of head injuries

  7. fractures per 1000 person-days

  8. hip fractures per 1000 person-days

  9. qualitative outcomes (e.g. staff’s, patients’/residents’ and visitors’ attitudes, views and experiences)

  10. economic outcomes [to include assessments of quality-adjusted life-years (QALYs)]

  11. process outcomes (e.g. ease of or problems with flooring installation).

The primary outcomes (1 and 2) were selected to assess the potential benefits and harms of flooring interventions for patients/residents, accounting for occupancy levels and follow-up time; the injurious falls rate additionally accounts for variations in the underlying falls rate as a pragmatic measure of effectiveness. For example, if shock-absorbing floors do reduce the proportion of falls resulting in injury, but inadvertently increase the number of falls, then the rate of injurious falls would provide a better reflection in real terms of how many injurious falls are occurring in practice. Rate measures are also considered most suitable when dealing with count data types.

The first seven prioritised outcomes have been incorporated into our summary of findings tables (which headline the findings for each comparison; see Chapter 3), and these outcomes formed the basis of our risk-of-bias and Grading of Recommendations Assessment, Development and Evaluation (GRADE) assessments. The quantitative outcomes (1–10) are reported in Chapter 3, qualitative outcomes (11) are reported in Chapter 4 and economic outcomes (12) are reported in Chapter 5. Process outcomes (13) are reported when available as part of the studies included in Chapters 3 and 4.

Search methods

The search of this systematic review built on a comprehensive search already conducted in a scoping review. 20 The scoping review identified literature relating to shock-absorbing flooring published up to and including 20 May 2016. The clinical effectiveness (n = 20), cost-effectiveness (n = 12), and qualitative (n = 2) records identified by the scoping review were all assessed for eligibility in this systematic review. The search strategy of the scoping review was updated and refined in scope (by AD) to focus on identifying studies of clinical effectiveness, cost-effectiveness and qualitative experiences. We searched the following electronic databases: AgeLine, Cumulative Index to Nursing and Allied Health Literature (CINAHL), MEDLINE, Scopus, Web of Science and the NHS Economic Evaluation Database (NHS EED) (see Appendix 1, Table 22, for search strategies). The MEDLINE search strategy was adapted for the remaining databases by one review author (LMF) and checked by another (AD). With the exception of AgeLine (searched by DCM), all database searches were run by Lambert M Felix, who compiled the search ‘hits’ ready for screening in duplicate. We conducted forward and backward citation searches on included studies, and a hand-search of the journal Age and Ageing. The grey literature search included a review of clinical trial registries, ProQuest Dissertations & Theses, conference proceedings and relevant websites (Table 1). No language restrictions were placed on the search.

Data collection

Risk-of-bias assessment

Risk-of-bias assessments were undertaken at the outcome level, using the Cochrane Risk of Bias tool version 2.0 (RoB 2.0) for randomised trials,72 the Risk Of Bias In Non-randomized Studies – of Interventions (ROBINS-I) tool for non-randomised studies73 and the JBI critical appraisal checklist for qualitative studies. 56 The quantitative studies were all assessed by the same review author (LMF) and assessed independently in duplicate by one of the other team members (AD, CCL, BEK or OO). The qualitative studies were all assessed by the same review author (CM) and independently in duplicate by one other team member (AD or KFS). Disagreements were resolved through discussion (through e-mail and teleconference) and involvement of a third independent review author as necessary. Review authors were not blinded during risk-of-bias assessments; however, care was taken to ensure that no review authors assessed studies that they were involved in (i.e. as a co-author). We sought further information from study authors if there was inadequate information to form a risk-of-bias judgement; we approached study authors with open-ended questions asking them to describe the relevant study processes in more detail to avoid biased answers.

The public involvement members (Margaret Bell, Liz Burden and Joleen Tobias) supported the risk-of-bias assessments through group meetings and electronic liaison, in which they commented on the clarity and transparency of the reporting of our judgements. Feedback from our public involvement members influenced the style and wording of the supporting statements incorporated into the risk-of-bias tables (see Appendix 2).

Data analysis (quantitative studies)

Synthesis of qualitative studies

We used the JBI approach to meta-aggregation of qualitative studies,56 with the support of NVivo software, version 12 (QSR International, Warrington, UK). The analysis followed a three-stage process:

1. The agreed findings from the data collection phase were identified and coded in the study reports using NVivo. Findings were assigned short names (usually the title of the subtheme to which they related), and accompanying illustrations (i.e. participant quotations) and other contextual information (i.e. the authors’ analytical descriptions) from the study reports were coded under each respective finding.

2. One review author (AD) organised the findings into categories (each containing at least two findings). This stage was achieved through reading and re-reading the findings to identify conceptual similarities across the data. Amy Drahota drafted a description for each category, and these were reviewed by other team members. The names for the categories were agreed through discussion between Amy Drahota and Chris Markham.

3. The categories were subsequently combined into a set of synthesised findings (each containing at least two categories), with accompanying descriptions. Amy Drahota created the initial synthesis and this was finalised through discussion with Chris Markham. The aim of this third stage was to develop a comprehensive set of statements (synthesised findings) that are representative of the collated categories and individual findings, and that can be used to inform evidence-based decision-making.

Synthesis of economic evidence

We have tabulated and summarised the results of included economic evaluations narratively in the text. We adjusted all costs to 2019 Great British pounds (GBP) values using gross domestic product (GDP) deflators83 and used relevant exchange rates for international comparisons.

Confidence in cumulative evidence

Triangulation of methods

The review incorporates quantitative, qualitative and economic evidence, with each type of evidence contributing complementary information to our understanding of the use of shock-absorbing flooring in hospitals and care homes. Each type of evidence has its strengths in addressing different dimensions of the research question: clinical effectiveness (quantitative), views and experiences (qualitative), and cost-effectiveness (economic). Therefore, we opted to analyse each type of evidence separately prior to configuring the results in our overall discussion (see Chapter 5), in what is referred to as a convergent segregated approach to mixed-methods syntheses. 60,90 In configuring the findings, we employed a constant comparison approach to determine whether the findings from each type of evidence were supportive or contradictory, what each type of evidence adds (in terms of understanding and explanation), and what the different types of evidence contributed that was missing from the other studies.

Changes from the protocol

We were unable to search the World Health Organization (WHO) Health Evidence Network as planned because of a technical error with the server. We originally planned to conduct all of the data collection in Covidence; however, at the time of conducting this element of the review, we found that the software did not readily support a mixed-methods review, the various risk-of-bias assessment tools we were using or the outcome measures we were collecting. Therefore, we ended up using a mixture of Covidence and Microsoft Excel to undertake data collection, while maintaining the plan to conduct all of the review processes independently in duplicate. During the review process, in consultation with the advisory board members, we agreed on some of the potential confounding variables that then informed our assessment of the ‘confounding’ domain in the ROBINS-I tool. 73

We initially planned to undertake a subgroup analysis to explore the influence of acuity of care (acute, sub-acute, intermediate, long-term care); however, owing to the nature of the evidence included in the review, it was not logical to conduct this analysis. Instead, we analysed the studies according to whether they were based in hospitals or care homes. Some of the sensitivity analyses were not specified in the protocol, as we had not anticipated how to handle a 2 × 2 factorial observational study in the analysis or deal with personally communicated data that differed from the published report. We had also not specified the sensitivity analyses around rarer outcomes (fractures and hip fractures); however, our rationale for these alternative analyses is underpinned by methodological research, which highlights that our main approach to analysis (DerSimonian and Laird random-effects method) can provide biased estimates in the situation of rare events. 81 For the fracture data, the Mantel–Haenszel method had minimal impact and did not influence the review conclusions, compared with the DerSimonian and Laird method; however, we opted to report the Mantel–Haenszel figures as we considered this approach to be the most methodologically sound.

Results of the search

The majority of electronic databases were last searched on 29 September 2019 (see Table 1). The search results are summarised in the PRISMA flow diagram (Figure 2). One hundred records were identified through MEDLINE, 23 through CINAHL and 18 through AgeLine (all searched through EBSCOhost). We also searched Scopus (172 records identified), Web of Science (192 records identified), ProQuest Dissertations & Theses (two records identified), NHS EED (six records identified) and OpenGrey (four records identified). We searched trial registries: 14 records were identified through the WHO’s International Clinical Trials Registry Platform and 626 records were identified through ClinicalTrials.gov. Hand-searching, backwards and forwards citation searching, and review of conference proceedings and websites retrieved a further 2785 records for screening.

FIGURE 2.

The PRISMA flow diagram. Reproduced from Drahota et al. 1 © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reproduced from Drahota et al. 1 © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Description of studies

Assessment of clinical effectiveness

Any shock-absorbing flooring versus standard flooring

Three randomised trials (two hospital based, one long-term care based) and seven observational studies (four hospital based, three long-term care based) compared one or more types of shock-absorbing flooring with ‘standard/rigid flooring’. Table 7 presents a summary of findings for hospital settings, and Table 8 presents for a summary of findings for care home settings. We have summarised the studies for care homes and hospitals separately because of heterogeneity between these settings and populations. Data for different flooring types (novel flooring, sports flooring, carpet and wooden subfloors) are also detailed separately in a subgroup analysis within each outcome, and summarised in Table 9.

TABLE 7 - Summary of findings for shock-absorbing flooring vs. standard/rigid flooring in hospital settings

The risk with shock-absorbing flooring (and its 95% CI) is based on the assumed risk with standard flooring (taken from Drahota et al. 77) and the pooled relative effect of the intervention (and its 95% CI).

See Appendix 3, Table 33, for explanations of GRADE profiles.

These data should be interpreted with caution as the denominator (falls) used in the calculation of RR is count data. All data contributing to this outcome are considered observational.

Note

The GRADE Working Group suggested definitions for grades of evidence have been published elsewhere. 117

Reproduced from Drahota et al. 1 © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Outcomes	Anticipated absolute effectsa (95% CI)		Relative effect (95% CI)	Total sample size (n studies)	Quality of the evidence (GRADE)b	Comments
	Risk with standard/rigid flooring	Risk with shock-absorbing flooring
Injurious falls rate per 1000 person-days
RCTs	3 per 1000	2 per 1000 (1 to 6)	RaR 0.58 (0.18 to 1.91)	9085 person-days (1 RCT)	⊕⊕◯◯ Low	These data (on sports flooring) are too imprecise to offer any certainty for this outcome
All studies	3 per 1000	2 per 1000 (1 to 3)	RaR 0.55 (0.36 to 0.84)	25,989 person-days (2 studies)	⊕◯◯◯ Very low	If three injurious falls per day occur among 1000 inpatients on a standard floor, then very low-quality evidence suggests that there would be one fewer (95% CI two fewer to about the same number) injurious fall per day on a shock-absorbing floor
Falls rate per 1000 person-days
RCTs	7 per 1000	8 per 1000 (5 to 13)	RaR 1.07 (0.64 to 1.81)	9085 person-days (1 RCT)	⊕⊕◯◯ Low	These data (on sports flooring) are too imprecise to offer any certainty for this outcome
All studies	7 per 1000	6 per 1000 (5 to 8)	RaR 0.88 (0.71 to 1.09)	25,989 person-days (2 studies)	⊕◯◯◯ Very low	If seven falls per day occur among 1000 inpatients on a standard floor, then very low-quality evidence suggests that between two fewer falls and one more fall would occur per day on a shock-absorbing floor (there may be no difference)
Number of falls resulting in injury
All studiesc	424 per 1000	165 per 1000 (64 to 433)	RR 0.39 (0.15 to 1.02)	559 falls (3 studies)	⊕◯◯◯ Very low	If 424 out of 1000 inpatient falls resulted in an injury on a standard floor, then very low-quality evidence suggests that 259 fewer (95% CI 360 fewer to 9 more) injurious falls would occur on a shock-absorbing floor. A sensitivity analysis removing a study on carpets judged to be at high risk of bias removes the heterogeneity and increases the precision of the effect for novel/sports floors (RR 0.64, 95% CI 0.44 to 0.93)
Number of fractures
RCTs	9 per 1000	3 per 1000 (0 to 69)	OR 0.33 (0.01 to 8.13)	448 participants (1 RCT)	⊕⊕◯◯ Low	These data (on sports flooring) are too imprecise to offer any certainty for this outcome
All studies	9 per 1000	3 per 1000 (0 to 16)	OR 0.28 (0.04 to 1.77)	626 participants (2 studies)	⊕◯◯◯ Very low	These data (on sports and novel flooring) are too imprecise to offer any certainty for this outcome
Number of hip fractures
RCTs	4 per 1000	1 per 1000 (0 to 32)	OR 0.33 (0.01 to 8.15)	448 participants (1 RCT)	⊕⊕◯◯ Low	These data (on sports flooring) are too imprecise to offer any certainty for this outcome
All studies	4 per 1000	4 per 1000 (1 to 25)	OR 0.88 (0.12 to 6.47)	626 participants (2 studies)	⊕◯◯◯ Very low	These data (on sports and novel flooring) are too imprecise to offer any certainty
Number of fallers
RCTs	99 per 1000	223 per 1000 (56 to 895)	RR 2.25 (0.56 to 9.04)	502 participants (2 RCTs)	⊕◯◯◯ Very low	These data (on sports flooring and carpet) are too imprecise to offer any certainty. Only RCTs contributed data to this outcome
Adverse events
RCTs	Staff raised concerns about moving wheeled equipment on sports floors, and one staff member sustained a pulled lower back on the intervention floor over 12 months’ follow-up			Not reported (1 study)	⊕◯◯◯ Very low	See Chapter 4 for further descriptions of push and pull challenges
Observational studies	No evidence to suggest greater risk of injury on intervention flooring (28 injuries in 30 months), compared with three concurrent control wards (30 injuries per ward) or a post-intervention control site (45 injuries in 30 months)			Not reported (1 study)	⊕◯◯◯ Very low

TABLE 8 - Summary of findings for shock-absorbing flooring vs. standard/rigid flooring in care homes

The risk in the intervention group (and its 95% CI) is based on the assumed risk taken from the comparison group of the RCT data and the pooled relative effect of the intervention (and its 95% CI).

See Appendix 4, Table 34, for explanations behind GRADE profiles.

These data should be interpreted with caution as the denominator (falls) used in the calculation of RR is count data. All data contributing to this outcome are considered observational.

Note

The GRADE Working Group suggested definitions for grades of evidence have been published elsewhere. 117

Reproduced from Drahota et al. 1 © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Outcomes	Anticipated absolute effectsa (95% CI)		Relative effect (95% CI)	Total sample size (n studies)	Quality of the evidence (GRADE)b	Comments
	Risk with standard/rigid flooring	Risk with shock-absorbing flooring
Injurious falls rate per 1000 person-days
RCTs	3 per 1000	3 per 1000 (2 to 4)	RaR 0.91 (0.62 to 1.32)	213,854 person-days (1 RCT)	⊕⊕⊕⊕ High	This study compared a novel underlay with vinyl overlay and concrete subfloor, with a plywood underlay with vinyl overlay and concrete subfloor
All studies	3 per 1000	3 per 1000 (2 to 4)	RaR 0.91 (0.62 to 1.32)	308,981 person-day (2 studies)	⊕◯◯◯ Very low	Data are missing from one observational study (novel vs. standard floors), rated as having a high risk of bias, that did not report on this outcome
Falls rate per 1000 person-days
RCTs	8 per 1000	10 per 1000 (7 to 14)	RaR 1.21 (0.87 to 1.68)	213,854 person-days (1 RCT)	⊕⊕⊕◯ Moderate	This study compared a novel underlay with vinyl overlay and concrete subfloor, with vinyl with a plywood underlay and concrete subfloor
All studies	8 per 1000	7 per 1000 (4 to 13)	RaR 0.87 (0.47 to 1.62)	308,981 person-days (2 studies)	⊕◯◯◯ Very low
Number of falls resulting in injury
All studiesc	330 per 1000	264 per 1000 (231 to 300)	RR 0.80 (0.70 to 0.91)	2800 falls (3 studies)	⊕◯◯◯ Very low	If 330 out of 1000 resident falls resulted in injuries on a rigid floor, then very low-quality evidence suggests that 66 fewer (95% CI 99 fewer to 30 fewer) injurious falls would occur on a shock-absorbing floor
Number of fractures
RCTs	58 per 1000	44 per 1000 (18 to 106)	OR 0.74 (0.29 to 1.92)	357 participants (1 RCT)	⊕⊕◯◯ Low	These data (on novel flooring vs. vinyl on plywood underlay) are too imprecise to offer any certainty for this outcome
All studiesc	11 per 1000	7 per 1000 (3 to 16)	OR 0.61 (0.26 to 1.48)	2074 falls (2 studies)	⊕◯◯◯ Very low	These data are too imprecise to offer any certainty for this outcome
Number of hip fractures
RCTs	12 per 1000	11 per 1000 (2 to 76)	OR 0.94 (0.13 to 6.74)	357 participants (1 RCT)	⊕⊕◯◯ Low	These data (on novel flooring vs. vinyl on plywood underlay) are too imprecise to offer any certainty for this outcome
All studiesc	2 per 1000	3 per 1000 (2 to 4)	OR 1.17 (0.77 to 1.80)	8548 falls (2 studies)	⊕◯◯◯ Very low	These data are too heterogeneous to offer any certainty for this outcome
Number of fallers
RCTs	676 per 1000	697 per 1000 (602 to 798)	RR 1.03 (0.89 to 1.18)	357 participants (1 RCT)	⊕⊕⊕⊕ High	This study compared a novel underlay with vinyl overlay and concrete subfloor, with vinyl with a plywood underlay and concrete subfloor. Only this RCT contributed data to this outcome
Adverse events
All studies	There was no evidence to suggest an increase in force-induced musculoskeletal injuries in care home staff			Not reported (1 study)	⊕◯◯◯ Very low	Personal communication (Dawn C Mackey, Simon Fraser University, 2020). Nested pre–post design in RCT study

TABLE 9 - Summary of findings for different flooring types vs. standard/rigid flooring

The risk with shock-absorbing flooring (and its 95% CI) is based on the assumed risk of vinyl on concrete (taken from Drahota et al. 77) and the pooled relative effect of the intervention (and its 95% CI).

See Appendix 5, Table 35, for explanations behind GRADE profiles.

These data should be interpreted with caution as the denominator (falls) used in the calculation of RR is count data. All data contributing to this outcome are considered observational.

Note

The GRADE Working Group suggested definitions for grades of evidence have been published elsewhere. 117

Outcomes	Anticipated absolute effectsa (95% CI)		Relative effect (95% CI)	Total sample size (n studies)	Quality of the evidence (GRADE)b	Comments
	Risk with standard/rigid flooring	Risk with shock-absorbing flooring
Injurious falls rate per 1000 person-days
Novel floors	3 per 1000	2 per 1000 (2 to 3)	RaR 0.80 (0.59 to 1.09)	225,124 person-days (2 studies)	⊕◯◯◯ Very low	One RCT rated as being at low risk of bias produced a RaR of 0.91 (95% CI 0.62 to 1.32)
Sports floors	3 per 1000	1 per 1000 (1 to 3)	RaR 0.46 (0.23 to 0.92)	14,720 person-days (2 studies)	⊕◯◯◯ Very low	If three injurious falls per day occur among 1000 inpatients on a standard floor, then very low-quality evidence suggests that two fewer (95% CI two fewer to about the same) injurious falls per day would occur on a sports floor
Falls rate per 1000 person-days
Novel floors	7 per 1000	6 per 1000 (5 to 9)	RaR 0.89 (0.65 to 1.24)	320,251 person-days (3 studies)	⊕◯◯◯ Very low	These studies are too heterogeneous to offer any certainty for this outcome. One RCT rated as being at low risk of bias produced a RaR of 1.21 (95% CI 0.87 to 1.68)
Sports floors	7 per 1000	6 per 1000 (4 to 9)	RaR 0.85 (0.56 to 1.28)	14,720 person-days (2 studies)	⊕◯◯◯ Very low	These data are too imprecise to offer any certainty for this outcome
Number of falls resulting in injuryc
Novel floors	424 per 1000	335 per 1000 (297 to 373)	RR 0.79 (0.70 to 0.88)	2871 falls (3 studies)	⊕◯◯◯ Very low	If 424 out of 1000 falls resulted in injury on a rigid floor, then very low-quality evidence suggests that 89 fewer (95% CI 127 fewer to 51 fewer) injurious falls would occur on a novel floor
Sports floors	424 per 1000	246 per 1000 (127 to 475)	RR 0.58 (0.30 to 1.12)	152 falls (2 studies)	⊕◯◯◯ Very low	These data are too imprecise to offer any certainty for this outcome
Carpet	424 per 1000	68 per 1000 (30 to 170)	RR 0.16 (0.07 to 0.40)	213 falls (1 study)	⊕◯◯◯ Very low	If 424 out of 1000 falls resulted in injury on a vinyl floor, then very low-quality evidence suggests that 356 fewer (95% CI 394 fewer to 254 fewer) injurious falls would occur on carpeted floors
Number of fractures
Novel floors	9 per 1000	6 per 1000 (3 to 13)	OR 0.67 (0.28 to 1.59)	482 participants (2 studies)	⊕◯◯◯ Very low	These data are too imprecise to offer any certainty for this outcome
Sports floors	9 per 1000	3 per 1000 (0 to 25)	OR 0.31 (0.03 to 3.07)	289 participants (2 studies)	⊕◯◯◯ Very low	These data are too imprecise to offer any certainty for this outcome
Carpet	Not estimable; see comments			54 participants (1 study)	⊕◯◯◯ Very low	One small study did not observe any fractures
Number of hip fractures
Novel floors	4 per 1000	6 per 1000 (1 to 28)	OR 1.25 (0.24 to 6.47)	482 participants (2 studies)	⊕◯◯◯ Very low	These data are too imprecise to offer any certainty for this outcome
Sports floors	4 per 1000	1 per 1000 (0 to 35)	OR 0.33 (0.01 to 8.15)	447 participants (2 studies)	⊕◯◯◯ Very low	These data are too imprecise to offer any certainty for this outcome
Carpetc	30 per 1000	36 per 1000 (23 to 54)	OR 1.19 (0.77 to 1.84)	6641 falls (1 study)	⊕◯◯◯ Very low
Wooden subfloorc	30 per 1000	18 per 1000 (14 to 24)	OR 0.59 (0.45 to 0.78)	6641 falls (1 study)	⊕◯◯◯ Very low	If 30 out of 1000 falls result in hip fracture on concrete subfloors, then very low-quality evidence suggests that 12 fewer (95% CI 16 fewer to 6 fewer) hip fractures would occur on wooden subfloors
Number of fallers
Novel floors	99 per 1000	102 per 1000 (88 to 116)	RR 1.03 (0.89 to 1.18)	357 participants (1 RCT)	⊕⊕⊕⊕ High
Sports floors	99 per 1000	138 per 1000 (62 to 306)	RR 1.40 (0.63 to 3.10)	448 participants (1 RCT)	⊕⊕⊕◯ Moderate	These data are too imprecise to offer any certainty for this outcome
Carpet	99 per 1000	641 per 1000 (85 to 1000)	RR 6.50 (0.86 to 49.30)	54 participants (1 RCT)	⊕◯◯◯ Very low	These data are too imprecise to offer any certainty for this outcome
Adverse events
Novel floors	No evidence to suggest an increase in force-induced musculoskeletal injuries in care home staff			Not reported (1 study)	⊕◯◯◯ Very low	Personal communication (Dawn C Mackey). Nested pre–post design in RCT study
Sports floors	Staff raised concerns about moving wheeled equipment on sports floor. One staff member pulled lower back on sports floor during 12-month follow-up			Not reported (1 study)	⊕◯◯◯ Very low	See Chapter 4 for further descriptions on push and pull challenges

Primary outcomes

Injurious falls rate per 1000 person-days

Two randomised trials rated as being at low risk of bias (one hospital-based pilot study and one care home-based trial) and one observational study rated as having serious concerns for risk of bias (based in a hospital) contributed data to this outcome. 49,50,77 The pooled effect from two RCTs (Figure 3) provides no clear evidence to support shock-absorbing flooring use for reducing the injurious falls rate (RaR 0.87, 95% CI 0.61 to 1.25; p = 0.46; two studies; I² = 0%). 50,77 There was no evidence of a differential effect by settings based on the RCT data [test for subgroup differences: χ² = 0.50, degrees of freedom (df) = 1; p = 0.48; I² = 0%; see Appendix 6, Table 36]; however, the hospital-based study was very imprecise. The observational study reported a non-parametric test because the data were not normally distributed. 49 This study reported ‘a non-significant trend’ of lower injury rates in the shock-absorbing flooring cohort than among the control cohort [p = 0.059; intervention: median of three injurious falls per 1000 person-days (interquartile range 0–6); control: median of five injurious falls per 1000 person-days (interquartile range 0–8)]. 49

FIGURE 3.

Forest plot of comparison: any shock-absorbing flooring vs. rigid flooring; subgroups: settings; outcome: injurious falls rate per 1000 person-days. a, Number of person bed-days (total) in each group provided by the study author; b, data derived from unpublished data. Published analysis reports a non-significant result (p = 0.059). IV, inverse variance; SE, standard error. Reproduced from Drahota et al. 1 © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reproduced from Drahota et al. 1 © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Unpublished data, supplied by the contact author of the observational study, enabled data to be derived for pooling. These data should be interpreted with extreme caution, especially as it is noted that the derived analysis provides a more positive effect estimate for the study. 49 Incorporating evidence from the observational study maintains a reasonable level of statistical homogeneity (τ² = 0.04, χ² = 3.06, df = 2; p = 0.22; I² = 35%), but decreases the effect estimate more in favour of shock-absorbing flooring, still incorporating the potential for no effect (RaR 0.71, 95% CI 0.48 to 1.04; p = 0.08; three studies; see Figure 3). Tests for subgroup differences were non-significant when exploring both setting type (χ² = 3.05, df = 1; p = 0.08; I² = 67.2%; see Appendix 6, Table 36) and study design (χ² = 2.56, df = 1; p = 0.11; I² = 61.0%; see Appendix 6, Table 36).

Hospitals

One RCT conducted in hospital settings was too imprecise to determine the effect of shock-absorbing flooring on injurious falls rate, as the 95% CI incorporates the potential for no effect, for benefit and for harm (RaR 0.58, 95% CI 0.18 to 1.91; p = 0.37; see Figure 3). 77 However, incorporating the derived data, with caution, from an observational study (which has published data indicative of a non-significant effect; p = 0.059),49 suggests that shock-absorbing flooring may reduce the injurious falls rate in hospital settings (RaR 0.55, 95% CI 0.36 to 0.84; p = 0.006; two studies; I² = 0%; very low-quality evidence; see Figure 3). There was no evidence of a differential effect by study design in the data for hospital settings (χ² = 0.01, df = 1; p = 0.92; I² = 0%; see Appendix 6, Table 36).

Care homes

A single RCT conducted in a care home setting found an injurious falls rate on a novel underlay covered in vinyl with a concrete subfloor that was similar to that of a comparison group of vinyl on a plywood underlay with a concrete subfloor (RaR 0.91, 95% CI 0.62 to 1.32; p = 0.62; see Figure 3). 50 Observational studies from care homes did not contribute any data to this outcome.

Subgroup analyses by flooring type

A subgroup analysis limited to RCTs exploring flooring types only (sports flooring and novel shock-absorbing flooring) produces the same analysis as for the subgroup on settings of care (χ² = 0.50, df = 1; p = 0.48; I² = 0%; see Appendix 6, Table 36), providing no evidence to support a difference in treatment effects between flooring types.

A subgroup analysis by flooring type that also incorporates the unpublished and derived data from an observational study deemed to have serious concerns for risk of bias49 showed no evidence of a difference in treatment effects between flooring types for rate of injurious falls (χ² = 2.01, df = 1; p = 0.16; I² = 50.2%; Figure 4). The data on sports flooring were also hospital based and were indicative of positive treatment effects, with large uncertainty around the size of those effects (RaR 0.46, 95% CI 0.23 to 0.92; p = 0.03; I² = 0%; very low-quality evidence). The data for novel shock-absorbing flooring (which were obtained from a hospital and a care home) incorporated the potential of ‘no effect’, as well as of potential benefit (RaR 0.80, 95% CI 0.59 to 1.09; p = 0.15; I² = 0%; very low-quality evidence).

FIGURE 4.

Forest plot of comparison: any shock-absorbing flooring vs. rigid flooring; subgroups: flooring type; outcome: injurious falls rate per 1000 person-days. a, Data for Hanger49 are derived and unpublished; control group data are shared between comparisons; b, number of person bed-days (total) in each group provided by the study author. IV, inverse variance; SE, standard error.

Sensitivity analyses

Analyses removing studies judged to be at serious risk of bias produce the same results as detailed previously for RCT evidence only. The care home-based RCT also reported a multivariable model that adjusted for age, dementia, falls history, anti-anxiety medication and analgesic medication (RaR 1.09, 95% CI 0.75 to 1.58). 50 Incorporating these data into the analyses does not alter the bottom-line conclusions of the analyses (see Appendix 7, Table 37). However, the data do provide more conservative estimates for RCT-based evidence (RaR 1.03, 95% CI 0.72 to 1.47; p = 0.87; I² = 0%) and novel shock-absorbing floors (RaR 0.84, 95% CI 0.55 to 1.27; p = 0.40; I² = 33%), and introduces more statistical heterogeneity into the analysis of all studies combined (RaR 0.75, 95% CI 0.43 to 1.29; p = 0.30; heterogeneity: τ² = 0.14; χ² = 5.67, df = 2; p = 0.06; I² = 65%).

Falls rate per 1000 person-days

Five studies contributed data to this outcome (two RCTs and three observational studies). 49,50,52,77,101 Only four studies provided data suitable for use in a meta-analysis,49,50,52,77 with data from one of these studies obtained via personal communication. 49

The CI of the pooled effect estimate from two RCTs judged to be at low risk of bias (one care home based, one hospital based) incorporates the possibilities that shock-absorbing flooring may increase the rate of falls or make little to no difference (RaR 1.17, 95% CI 0.89 to 1.54; p = 0.27; two studies; I² = 0%; Figure 5). There was no evidence of a differential effect by setting (χ² = 0.14, df = 1; p = 0.71; I² = 0%; see Appendix 6, Table 36).

FIGURE 5.

Forest plot of comparison: any shock-absorbing flooring vs. rigid flooring; subgroups: settings; outcome: falls rate per 1000 person-days. a, Gustavsson et al. 52 is adjusted for age, sex, cognitive ability, walking ability, medications (antidepressants and sedatives) and visual impairment; b, Hanger49 data are derived and obtained via personal communication (Dr Carl Hanger, Burwood and The Princess Margaret Hospitals, Christchurch, New Zealand, 2020). IV, inverse variance; SE, standard error. Reproduced from Drahota et al. 1 © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reproduced from Drahota et al. 1 © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Two hospital-based observational studies (exploring novel flooring, sports flooring and carpet) also reported non-significant differences in falls rates from analyses unsuitable for meta-analysis. 49,101 We obtained personally communicated and derived data from one of these studies (on novel and sports flooring),49 and, alongside a third observational study of novel flooring in a care home setting,52 we were able to assess the suitability of incorporating observational data into the meta-analysis.

Incorporating the two observational studies into the meta-analysis increases the heterogeneity (τ² = 0.06; χ² = 10.44, df = 3; p = 0.02; I² = 71%); although pooling these studies is, therefore, questionable, the probability that any observed effects are due to chance is maintained (RaR 0.89, 95% CI 0.67 to 1.18; p = 0.41; four studies; see Figure 5). The heterogeneity can be somewhat explained by study design (χ² = 5.44, df = 1; p = 0.02; I² = 81.6%; see Appendix 6, Table 36), with the pooled effect of prospective cohort studies deemed to be at serious risk of bias positively favouring the intervention (RaR 0.74, 95% CI 0.57 to 0.97; p = 0.03; two studies; I² = 61%), counter to the RCT evidence that was judged to be at low risk of bias. There is no evidence to suggest a differential effect by settings (χ² = 0.00, df = 1; p = 0.98; I² = 0%; see Appendix 6, Table 36).

Hospitals

One small cluster randomised trial judged to be at low risk of bias was too imprecise to determine the influence of shock-absorbing flooring on falls rate. 77 The CI incorporates the possibility that shock-absorbing floors may increase the risk of falls; however it also suggests that any observed differences may be due to chance alone (RaR 1.07, 95% CI 0.64 to 1.81; p = 0.78; low-quality evidence; see Figure 5). Including data from an observational study judged to have serious concerns for risk of bias improves precision and maintains the possibility that the range of possible intervention effects on falls rates includes no difference (RaR 0.88, 95% CI 0.71 to 1.09; p = 0.25; two studies; I² = 0%; very low-quality evidence; see Figure 5). There was no evidence of a differential effect by study design in the hospital evidence (χ² = 0.67; p = 0.41; I² = 0%; see Appendix 6, Table 36). Not included in this analysis are data from an observational study,101 judged to be at serious risk of bias, comparing carpet with vinyl through an observation of time trends; this study also found no evidence to support an effect on falls rate per 1000 person-days. 101

Care homes

One trial (on SmartCells flooring) deemed to be at low risk of bias incorporated the probability that shock-absorbing flooring may make no difference to falls rates. 50 However, the 95% CI also incorporated the possibility of harm (RaR 1.21, 95% CI 0.87 to 1.68; p = 0.26; one study; see Figure 5). Including data from an observational study (on Kradal flooring)52 deemed to be at serious risk of bias introduced substantial heterogeneity (τ² = 0.18, χ² = 9.28, df = 1; p = 0.002; I² = 89%), with the two studies demonstrating opposite effects (test for subgroup differences: χ² = 9.28, df = 1; p = 0.002; I² = 89.2%; see Appendix 6, Table 36).

Subgroup analyses by flooring type

A subgroup analysis limited to RCTs only, exploring flooring types (sports flooring and novel shock-absorbing flooring), produces the same results as for the subgroup analysis on settings of care (χ² = 0.14, df = 1; p = 0.71; I² = 0%; see Appendix 6, Table 36), providing no evidence to support a difference in falls rates between flooring types.

Exploring the heterogeneous evidence from all studies by flooring type does not resolve the heterogeneity, with the data for novel shock-absorbing floors still providing mixed effects (τ² = 0.08; χ² = 11.38, df = 3; p = 0.010; I² = 74%), and displaying similar combined results to those of sports flooring (χ² = 0.04, df = 1; p = 0.84; I² = 0%; Figure 6). These findings are also compatible with the non-significant observational study comparing carpet with vinyl,101 providing no evidence to suggest that different flooring types affect falls rates differentially.

FIGURE 6.

Forest plot of comparison: any shock-absorbing flooring vs. rigid flooring; subgroups: flooring type; outcome: falls rate per 1000 person-days. a, Gustavsson et al. 52 is adjusted for age, sex, cognitive ability, walking ability, medications (antidepressants and sedatives) and visual impairment; b, Hanger49 data are derived (unpublished); the control arm has been split between comparisons. IV, inverse variance; SE, standard error.

Sensitivity analyses

A sensitivity analysis removing studies judged be at serious risk of bias is identical to the RCT evidence presented previously. One RCT presented a multivariate model for falls rates adjusted for age, falls history, dementia, anti-anxiety medication and analgesic medication (RaR 1.32, 95% CI 0.94 to 1.84). 50 Incorporating these data into the analyses instead of the unadjusted estimates did not alter the overall conclusions (see Appendix 7, Table 37).

Through our personal communications with the study authors of two of the observational studies, potential errors in the published data were revealed. The contact author of the care home study supplied us with new data suggestive of an even more beneficial effect in the shock-absorbing flooring group (RaR 0.60, 95% CI 0.48 to 0.74); however, there was uncertainty around the accuracy of these data, as some of the numbers did not add up as expected. 52 These revised data served to increase the heterogeneity in the review findings only and did not change the over-riding conclusions (see Appendix 7, Table 37). The contact author of the hospital-based observational study identified a coding error for one fall attributed to the wrong group;49 the impact of this was negligible (revised RaR 0.86, 95% CI 0.68 to 1.09 vs. RaR 0.85, 95% CI 0.67 to 1.07) and did not seriously alter the bottom-line results of the analyses (see Appendix 7, Table 37).

Secondary outcomes

Number of falls resulting in injury

Six studies (three hospital based, three in care homes) provided data for this outcome. 49,50,52,77,102,103 We believe that these data are not as robust as the data contributing to the primary outcome of injurious falls rates for a number of reasons and should therefore be interpreted with extreme caution:

• The samples are not independent (i.e. the same individuals may contribute more than once to the risk in each group and, in some cases, may contribute to both groups).

• There is no adjustment for exposure time or ‘at-risk time’.

• For this outcome, we are treating the RCT data as if they are observational data and not reporting them with elevated prominence, as the outcome denominator is based on the observed number of falls and not the number of people randomised to each group.

• Not all of the data have been adjusted for covariates and many of the studies are subject to confounding, thereby putting the data at high risk of bias.

More studies are able to contribute data to this outcome, however, as the outcome lends itself to observational study designs, which take the falls data recorded in the study setting as the starting point for generating the intervention and control groups, and these study types often report this as the primary outcome.

Figure 7 summarises the descriptive data for injurious falls by severity. Heterogeneity exists in the classification systems used for injury severity types, along with the types of floors, settings and study designs employed. Donald et al. 105 did not provide sufficient information to be incorporated into our synthesis, as data were not provided at a group level (one-third of falls resulted in minor injuries; there were no fractures, but there was only a single fall on vinyl).

FIGURE 7.

Studies comparing shock-absorbing flooring with rigid flooring; outcome: proportion of falls resulting in injuries of different severities. a, Fracture data have been inserted for Knoefel et al. 102 under severe injury.

The descriptive data indicate that a smaller proportion of falls resulted in injury in the intervention arms and that (when measured) the injuries that did occur tended to be less severe, more noticeably so in hospital environments (see Figure 7). Healey103 is a clear outlier, with a very high proportion of falls classified as injurious in the control arm. The proportion of falls resulting in injury was also high in Knoefel et al. 102 and neither of these studies classified the injuries by severity. Knoefel et al. 102 did consider various injury types (redness, bruise, abrasion, cut, two or more injuries, fracture, other), some of which were termed ‘significant injuries’; however, the presentation of the data precludes an assessment of severity breakdown, as the independence of data between categories is unclear.

We have cautiously summarised the data according to risk of falls resulting in any injury (Figure 8). The findings of all studies combined, although in favour of shock-absorbing flooring (RR 0.69, 95% CI 0.52 to 0.90; p = 0.006), are heterogeneous (τ² = 0.06, χ² = 14.71, df = 5; p = 0.01; I² = 66%). This heterogeneity appears to be largely explained by Healey’s103 highly positive study exploring carpet versus vinyl (see the subsequent section, Subgroup analyses by flooring type). The pooled point estimate for RCTs (RR 0.77, 95% CI 0.67 to 0.88) was more conservative and precise than those of prospective cohort studies (RR 0.68, 95% CI 0.51 to 0.89) and retrospective cohort studies (RR 0.41, 95% CI 0.07 to 2.35); however, there was no evidence to suggest a differential effect according to study design (χ² = 1.10, df = 2; p = 0.58; I² = 0%; see Appendix 6, Table 36) and, as highlighted previously, the RCT evidence for this outcome should be considered less robust than that of our primary outcome. Similarly, although there was some heterogeneity between the subgroups of hospitals (RR 0.39, 95% CI 0.15 to 1.02; see Figure 8) and care homes (RR 0.80, 95% CI 0.70 to 0.91; see Figure 8), the test for subgroup differences was not significant (χ² = 2.08, df = 1; p = 0.15; I² = 51.9%; see Appendix 6, Table 36).

FIGURE 8.

Forest plot of comparison: any shock-absorbing flooring vs. rigid flooring; subgroups: settings; outcome: number of falls resulting in injury. a, Data have been adjusted for clustering (ICC = 0.046); b, data adjusted for age, sex, body mass index, visual and cognitive impairments, walking ability, hip protectors, fall location (room type), activity when falling and time of day. IV, inverse variance; SE, standard error. Reproduced from Drahota et al. 1 © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reproduced from Drahota et al. 1 © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Subgroup analyses by flooring type

A subgroup analysis including all studies exploring the number of falls resulting in injury suggests that the type of flooring makes a difference to the outcome (χ² = 12.09, df = 2; p = 0.002; I² = 83.5%; Figure 9). However, this result is largely down to the findings for carpet versus vinyl, which, counterintuitively, are much more favourable than the findings for both novel and sports floors versus vinyl. The single study contributing to this subgroup was a retrospective cohort study judged to be at serious risk of bias and with large imbalances between study arms. 103 Removing this subgroup from the analysis indicates that the findings for novel floors and sports floors are similar (χ² = 0.81, df = 1; p = 0.37; I² = 0%). The data for sports floors are very imprecise and indicate a range of possible effects, from improved outcomes with the intervention to no difference (RR 0.58, 95% CI 0.30 to 1.12; p = 0.11; I² = 0%; see Figure 9). The data for novel floors are more precise and are positively in favour of the intervention (RR 0.79, 95% CI 0.70 to 0.88; p < 0.0001; I² = 0%; see Figure 9).

FIGURE 9.

Forest plot of comparison: any shock-absorbing flooring vs. rigid flooring; subgroups: flooring type; outcome: number of falls resulting in injury. a, Data adjusted for age, sex, body mass index, visual and cognitive impairments, walking ability, hip protectors, fall location (room type), activity when falling and time of day; b, Hanger49 data were obtained via personal communication (Dr Carl Hanger); the control arm has been split between comparisons; c, data have been adjusted for clustering (ICC = 0.046). IV, inverse variance; SE, standard error.

Sensitivity analyses

The subgroup of hospital studies was sensitive to the estimated ICC in Drahota et al. ,77 with a smaller ICC (0.023) resulting in CIs that no longer cross the line of no effect (RR 0.40, 95% CI 0.16 to 0.97; p = 0.04; I² = 73%). Removing the controversial carpet study from this subgroup (while keeping the original, more conservative, ICC) also resolves the heterogeneity and positions the effect estimate more precisely in favour of the intervention (RR 0.64, 95% CI 0.44 to 0.93; p = 0.02; I² = 0%). Other adjustments to the ICC did not make any material difference to the findings (see Appendix 7, Table 37).

We assessed the influence of some slightly revised data received via personal communication on Hanger (with Dr Carl Hanger);49 however, this made a negligible difference to the overall results (see Appendix 7, Table 37). A sensitivity analysis based on risk of bias provides the same results as the RCT evidence covered previously.

Number of fractures

Five studies (three RCTs and two observational studies) reported on the outcome of number of fractures. 49,50,77,102,105 Four studies reported on the number of participants who sustained a fracture (three RCTs and one observational study),49,50,77,105 and another retrospective cohort study observed the number of falls that resulted in fracture (data that are also derivable from the first four studies). 102 The number of fracture events in the majority of the studies were very few, resulting in imprecise estimates; three studies had zero events in one or both arms. 77,102,105 One small RCT with zero events in both arms was unable to contribute any information to the analyses. 105 We have analysed the number of fractures as a function of the number of participants and the number of falls to incorporate all of the evidence. Caution needs to be exercised when interpreting data from the latter method because the same individuals may contribute one or more falls to the study groups [and, in one study, it is possible that the same individual(s) contributed to falls counts in both groups]. 102 Therefore, the data are not independent and it is impossible to gauge how much overall influence this has on the findings. However it can be seen that the RCT evidence is biased slightly more in favour of the intervention when using number of falls as the denominator (Figure 10).

FIGURE 10.

Forest plot of comparison: any shock-absorbing flooring vs. rigid flooring; subgroups: settings; outcome: number of fractures. a, Study displayed for information purposes; it was not included in analyses as it had zero events; b, data adjusted for clustering (ICC = 0.013). M–H, Mantel–Haenszel. Reproduced from Drahota et al. 1 © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reproduced from Drahota et al. 1 © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

The pooled effect estimate from the two RCTs that contributed data was imprecise; the 95% CI included the potential for substantial benefit, no effect or possible harm (OR 0.69, 95% CI 0.28 to 1.70; p = 0.42; two studies; I² = 0%; see Figure 10). Incorporating the observational evidence into the pooled effect estimates increases the precision somewhat, without contributing to increased statistical heterogeneity, when participants are the denominator (OR 0.59, 95% CI 0.26 to 1.35; p = 0.21; three studies; I² = 0%; see Figure 10) and when falls are the denominator (OR 0.53, 95% CI 0.24 to 1.17; p = 0.11; four studies; I² = 0%; see Figure 10). However, although the point estimates are in favour of the intervention, the wide CIs mean that there is insufficient information to detect a signal.

With low precision and statistically homogeneous findings, we found no evidence to suggest a differential effect by settings when focusing on the RCT evidence alone (χ² = 0.23, df = 1; p = 0.63; I² = 0%; see Appendix 6, Table 36) or all evidence combined (by participants: χ² = 0.86, df = 1; p = 0.35; I² = 0%; by falls: χ² = 0.48, df = 1; p = 0.49; I² = 0%). There was no indication of a differential effect by study design in these data (χ² = 0.85, df = 2; p = 0.65; I² = 0%; see Appendix 6, Table 36). We have subgrouped the data for hospitals alone and care homes alone in Figure 10; the data are too imprecise to determine whether or not shock-absorbing flooring is beneficial for reducing fractures in either setting.

Subgroup analyses by flooring type

Owing to the levels of uncertainty in the data, there was no evidence that the type of shock-absorbing flooring had a differential effect on the number of fractures when looking at RCTs alone (χ² = 0.23, df = 1; p = 0.63; I² = 0%; see Appendix 6, Table 36), or all available evidence analysed at the participant level (χ² = 0.37, df = 1; p = 0.54; I² = 0%; see Appendix 6, Table 36) or the falls level (χ² = 0.23, df = 1; p = 0.63; I² = 0%; see Appendix 6, Table 36). Although data in each subgroup showed minimal statistical heterogeneity, the studies were too imprecise to detect a signal for this outcome (Figure 11).

FIGURE 11.

Forest plot of comparison: any shock-absorbing flooring vs. rigid flooring; subgroups: flooring type; outcome: number of fractures. a, Unreported data; data for the control arm have been split between the three comparisons; b, study displayed for information; this study contributed no data to the analyses. M–H, Mantel–Haenszel.

Sensitivity analyses

We ran sensitivity analyses on the choice of analysis method, using Peto ORs instead, with negligible impact on the findings (see Appendix 7, Table 37). Our choice of ICC for the cluster RCT also did not alter the overall effect estimates77 and some slightly revised unpublished figures received from Hanger49 were inconsequential to the review findings. Gustavsson et al. 52 provided us with some unpublished data (all of the serious falls in this observational study resulted in one or more fractures) (Johanna Gustavsson, personal communication); including these (unadjusted) data in the analyses did not alter the overall conclusions (see Appendix 7, Table 37). We ran a sensitivity analysis to include the observational data reported in Lange;82 however, this made no material difference to the findings. We did not run any sensitivity analyses on risk of bias, as these would have produced the same findings as looking at RCT evidence alone.

Number of hip fractures

Five studies (three RCTs and two observational studies) reported the number of hip fractures. 49,50,76,77,105 Four studies reported the number of participants who sustained a hip fracture (three RCTs and one observational study)49,50,77,105 and Simpson et al. 76 observed the number of falls that resulted in hip fracture (data that are also derivable from the other four studies). The Simpson et al. 76 study was a factorial study, exploring carpeted and uncarpeted floors, and if the subfloors were wooden or concrete. In Figure 12, we have focused the comparisons on overlay and underlay materials (for Simpson et al. 76 this is carpeted vs. uncarpeted floors), and subsequently, in the assessment of different flooring types, we have also segregated the evidence for wooden versus concrete subfloors. The number of events in the majority of studies was very small, resulting in imprecise estimates; three studies had zero events in one or both arms,49,77,105 with Donald et al. 105 not contributing any information to the analyses.

FIGURE 12.

Forest plot of comparison: any shock-absorbing flooring vs. rigid flooring; subgroups: settings; outcome: number of hip fractures. a, Donald et al. 105 displayed for information purposes; it was not included in analyses as it had zero events; b, Drahota et al. 77 data are adjusted for clustering (ICC = 0.002); c, Mackey et al. 50 events data were obtained via personal communication; d, Simpson et al. 76 data compare carpeted with uncarpeted floors on different subfloors. M–H, Mantel–Haenszel.

For the outcome ‘any fracture’, we have summarised the data with participants as the unit of analysis, and, secondary to this, with falls as the unit of analysis to incorporate the additional observational study. Studies have slightly inflated point estimates when analysed by the number of falls (for RCTs this bias is more in favour of the intervention, for Hanger49 it is more in favour of control). However, owing to the imprecision in the data, none of the analyses is able to detect whether or not shock-absorbing flooring can help reduce the number of hip fractures (see Figure 12).

The Simpson et al. 76 study contributes heterogeneity to the analyses when the data are stratified by subfloor type, as we have presented in Figure 12, although see Sensitivity analyses; the other studies are largely homogeneous. The Simpson et al. 76 data are at serious risk of bias due to confounding. The unbalanced groups in this study are an artefact of the smaller number of falls occurring on uncarpeted floors (likely, in part, due to less exposure to these surfaces). The counterintuitive finding that fewer fractures occurred on uncarpeted concrete floors (the hardest of all surfaces to land on) was deemed by the study authors to be associated with such areas being bathrooms, which may be subject to increased staff vigilance, or more handrails and other items to help break the fall. 76

There is no evidence to suggest a differential effect by setting type (RCT evidence: χ² = 0.30, df = 1; p = 0.59; I² = 0%; all studies’ participant-level data: χ² = 0.00, df = 1; p = 0.96; I² = 0%; all studies’ falls-level data: χ² = 0.06, df = 1; p = 0.81; I² = 0%; see Appendix 6, Table 36) or study design (participant-level data: χ² = 0.46, df = 1; p = 0.50; I² = 0%; falls-level data: χ² = 0.30, df = 1; p = 0.59; I² = 0%; see Appendix 6, Table 36).

Subgroup analyses by flooring type

The data on novel shock-absorbing flooring (OR 1.25, 95% CI 0.24 to 6.47; p = 0.79; two studies; I² = 0%) and sports flooring (OR 0.33, 95% CI 0.01 to 8.15; p = 0.50) were too sparse to detect whether or not these floors are effective at reducing hip fractures (Figure 13). Although Simpson et al. 76 found no evidence to suggest that carpeted floors were more effective than uncarpeted floors (OR 1.19, 95% CI 0.77 to 1.84; p = 0.44), the data for wooden versus concrete subfloors, overall, was indicative of a beneficial effect in favour of wooden subfloors (OR 0.59, 95% CI 0.45 to 0.78; p = 0.008; serious risk of bias). Uncarpeted floors contributed heterogeneity to this study’s findings (χ² = 7.09, df = 1; p = 0.008; I² = 86%), and the data were also confounded by room type (e.g. uncarpeted areas were typically bathrooms, toilets and utility areas). Assuming that four out of 100 falls result in a hip fracture (based on the control arm of Simpson et al. 76), the effect estimate can be re-expressed to suggest that one fewer person will fracture their hip for every 63 falls (95% CI 47 to 118 falls) that occur on wooden, as opposed to concrete, subfloors (very low-quality evidence). Or, if 40 out of 1000 falls result in hip fracture on a concrete subfloor, then 16 (95% CI 9 to 22) fewer falls out of 1000 will result in hip fracture on wooden subfloors (see Table 9 for an alternative re-expression based on a lower assumed risk).

FIGURE 13.

Forest plot of comparison: any shock-absorbing flooring vs. rigid flooring; subgroups: flooring type; outcome: number of hip fractures. a, Hanger49 are unpublished data; data for the control arm have been split between the three comparisons; b, Mackey et al. 50 events data were obtained via personal communication; c, Drahota et al. 77 data are adjusted for clustering (ICC = 0.002); d, Donald et al. 105 is displayed for information; it contributed no data to the analyses. M–H, Mantel–Haenszel. Reproduced from Drahota et al. 1 © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reproduced from Drahota et al. 1 © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Owing to the imprecision in the majority of the data, there was no indication of a differential effect by flooring types (see Appendix 6, Table 36). This was the case when focusing on the participant-level data (χ² = 0.53, df = 1; p = 0.47; I² = 0%); falls-level data comparing novel, sports and carpeted floors (χ² = 0.69, df = 2; p = 0.71; I² = 0%); and falls-level data comparing novel floors, sports floors and wooden subfloors (χ² = 1.03, df = 2; p = 0.60; I² = 0%). We have not run the test for subgroup differences between all four flooring types in one analysis because the Simpson et al. 76 data contribute to two of the subgroups.

Sensitivity analyses

The choice of analysis method (Peto vs. Mantel–Haenszel) did not affect the findings, nor did the choice of ICC for Drahota et al. 77 or adjusting Hanger49 for slightly revised figures that were personally communicated to us (see Appendix 7, Table 37). Reanalysing the data without stratifying Simpson et al. 76 did not alter the bottom-line findings in any of the analyses (see Appendix 7, Table 37). This was the case for carpeted versus uncarpeted floors (stratified: OR 1.19, 95% CI 0.77 to 1.84; combined raw data: OR 1.12, 95% CI 0.72 to 1.73) and for wooden versus concrete subfloors (stratified: OR 0.59, 95% CI 0.45 to 0.78; combined raw data: OR 0.59, 95% CI 0.45 to 0.79). When Simpson et al. 76 was not stratified by the other factor explored in the study, the statistical heterogeneity was removed from all of the analyses. For example, in the analysis of care home studies, for any type of shock-absorbing floor versus control (see Figure 12), the statistical heterogeneity went from being substantial (χ² = 7.40, df = 2; p = 0.02; I² = 73%) to unnoteworthy (χ² = 0.05, df = 1; p = 0.82; I² = 0%). We received some unpublished data from Gustavsson et al. 52 on the number of hip fractures observed (Johanna Gustavsson, personal communication); however, including these data did not alter the bottom-line conclusions (see Appendix 7, Table 37).

Number of fallers

Three RCTs contributed data to the outcome of number of fallers. 50,77,105 Although the point estimate is in favour of control floors, the CI for the pooled studies incorporates the possibility that shock-absorbing flooring may make no difference to the risk of being a faller (RR 1.28, 95% CI 0.73 to 2.25; p = 0.40; I² = 46%; Figure 14). Although there was no evidence of a differential effect by setting type (χ² = 1.20, df = 1; p = 0.27; I² = 16.4%), it should be noted that the underlying event rates were much higher in the care home study, in which participants had longer exposure to the floors under investigation (68% of the control arm fell in Mackey et al. ,50 compared with 10% in Drahota et al. 77 and 4% in Donald et al. 105). The hospital studies are, therefore, subject to greater imprecision and it is possible that further evidence may change the findings of this analysis.

FIGURE 14.

Forest plot of comparison: any shock-absorbing flooring vs. rigid flooring; subgroups: settings; outcome: number of fallers. a, Drahota et al. 77 has been adjusted for clustering (assumed ICC = 0.02). IV, inverse variance; SE, standard error. Reproduced from Drahota et al. 1 © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reproduced from Drahota et al. 1 © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Hospitals

Pooled data from two imprecise trials49,77 offer inconclusive evidence on the influence of shock-absorbing flooring on falling risk in hospitals (RR 2.25, 95% CI 0.56 to 9.04; two studies; I² = 48%; very low-quality evidence; see Figure 14). Although the CIs overlap, there is some variation between the two studies (τ² = 0.57), and these studies were quite different with regard to the flooring type under investigation (carpet and sports flooring); Donald et al. 105 was even more prone to random error, as the study was so small, and was also considered to be at higher risk of bias.

Care homes

Evidence from a single trial suggests that vinyl with a novel shock-absorbing underlay has little or no effect on the risk of being a faller in care homes, when compared with a vinyl floor covering with plywood underlay (RR 1.03, 95% CI 0.89 to 1.18; one trial; high-quality evidence; see Figure 14). 50

Subgroup analyses by flooring type

The evidence is too sparse to determine if there is a differential effect by flooring type, although none is indicated (χ² = 3.68, df = 2; p = 0.16; I² = 45.7%; Figure 15); with only one study per subgroup, this analysis is somewhat problematic and should be interpreted with caution.

FIGURE 15.

Forest plot of comparison: any shock-absorbing flooring vs. rigid flooring; subgroups: flooring type; outcome: number of fallers. a, Drahota et al. 77 has been adjusted for clustering (assumed ICC = 0.02). IV, inverse variance; SE, standard error.

Sensitivity analyses

Sensitivity analyses exploring the influence of the ICC adjustment in Drahota et al. 77 did not change the conclusions for this outcome (see Appendix 7, Table 37). Limiting the analysis to the two studies deemed to be at low risk of bias removes the heterogeneity (τ² = 0.00, χ² = 0.55, df = 1; p = 0.46; I² = 0%) and the pooled effect estimate remains centred around the line of no effect (RR 1.04, 95% CI 0.90 to 1.19; p = 0.60; two studies). 50,77

Adverse events (staff injuries)

Two RCTs collected data on staff injuries;50,77 however, because the unit of allocation in Mackey et al. 50 was the resident room, the data pertaining to staff injuries from working in the care home are more akin to a pre–post design and have not been published. The unit of allocation in Drahota et al. 77 was at the facility level, so staff were randomised in clusters to a ward with or without an intervention flooring bay. One further observational study associated with Hanger49 has since been published in 2020. 106 This study compared injury data over 30 months from the intervention ward with those of three concurrent control wards (no baseline period), and also with a 30-month post-intervention period when the ward was moved to a new site without the intervention flooring (no concurrent control). Hanger and Wilkinson106 is analysed using non-parametric tests, with no adjustments for potential confounding, and time-series data are presented in charts.

Neither of the two hospital-based studies were able to determine a denominator population. In both studies, adverse events occurring in the participating wards were reported, and could have related to staff based both internally and externally to the wards, meaning robust data on the number of staff members and exposure time were not obtained. Hanger and Wilkinson106 obtained staff injury data from the electronic Quality Improvement Event Reporting system, whereas Drahota et al. 77 relied on sites completing adverse events forms as part of the study’s standard operating procedures.

Drahota et al. 77 reported one pulled lower back in the intervention group (8.3-mm sports floor) during the 12-month follow-up period, which occurred while moving a patient on a trolley and did not require medical attention. Staff across the intervention sites raised concerns about moving wheeled equipment on the intervention floor, captured in the qualitative interviews (see Chapter 4), and five additional adverse events forms were submitted to raise concerns (received from two of the four intervention sites), but these did not pertain to specific events/injuries; no adverse events were reported by the control group. 77 Over a 30-month follow-up period, Hanger and Wilkinson106 found no evidence to suggest that there was a higher risk of staff injuries on intervention flooring; if anything, the risk appeared to be lower on the intervention flooring (however, this could have been confounded by improved reporting on control floors).

The care home-based study collected data on staff musculoskeletal injuries from routinely completed staff incident reports for 2 years pre intervention and 1 year post intervention. 109 Data were collected for three groups of staff: (1) all staff (food services, housekeeping, laundry, maintenance, resident care, therapies, administration, housing or other), (2) resident care staff and (3) resident care staff who specifically worked in the four units/villages where the trial was occurring. Overall, the analyses suggested that the intervention flooring (a 25.4-mm novel underlay covered with vinyl) did not increase the risk of force-induced musculoskeletal injuries among long-term care staff, or among resident care staff specifically (unpublished data, Dawn C Mackey, 2020, personal communication).

Overall, the data suggest that, although initial concerns of working on a shock-absorbing floor may be raised, there is very low-quality evidence to suggest that, over longer periods of follow-up, there may be no difference in staff injuries (Table 10).

TABLE 10 - Adverse events associated with staff injuries

Reproduced from Drahota et al. 1 © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Study	Main findings	Comments	Risk of bias
Drahota et al. 201377	Concerns raised and one pulled lower back in intervention arm. No adverse events reported in control arm (12-month follow-up)	See Chapter 4, Category 4: push and pull challenges	Low
Hanger and Wilkinson 2020106	There were no statistically significant differences in staff injuries between intervention (28 injuries in 30 months) and concurrent control wards (average 30 injuries per ward), or the post-intervention control ward (45 injuries in 30 months)	Quality of reporting improved post intervention	Serious
Mackey et al. 201950	The intervention did not increase force-induced musculoskeletal injuries (24-month follow-up)	Unpublished data. Based on pre–post nested design	Not assessed

Head injuries

Two studies reported the number of head injuries49,50 and we also obtained personally communicated data pertaining to Gustavsson et al. 52 (Johanna Gustavsson, personal communication) (see Sensitivity analyses). We have analysed the data with both the number of participants and number of falls as the unit of analysis, as it is not clear whether the number of events are independent or could possibly relate to recurrent fallers; this makes a negligible difference to the findings (Figure 16). Although the CIs include the possibility that shock-absorbing flooring may help to reduce the number of head injuries, the data were too imprecise to know this with any certainty and the possibility remains that shock-absorbing flooring makes no meaningful difference when focusing on the RCT data alone (RR 0.60, 95% CI 0.24 to 1.51; p = 0.28) or both studies combined (RR 0.52, 95% CI 0.24 to 1.12; p = 0.10; I² = 0%). The two studies were statistically similar, although they were conducted in different settings and used different study designs (test for subgroup differences: χ² = 0.26, df = 1; p = 0.61; I² = 0%).

FIGURE 16.

Forest plot of comparison: any shock-absorbing flooring vs. rigid flooring; subgroups: unit of analysis; outcome: number of head injuries. a, Events based on location of serious fall-related injuries (head/skull). IV, inverse variance; SE, standard error. Reproduced from Drahota et al. 1 © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reproduced from Drahota et al. 1 © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Subgroup analysis by flooring type

Segregating the data according to flooring type does not reveal any noteworthy differences between novel and sports floors (Figure 17), as the data remain too imprecise to detect any effects (test for subgroup differences at the participant level: χ² = 0.53, df = 1; p = 0.47; I² = 0%).

FIGURE 17.

Forest plot of comparison: any shock-absorbing flooring vs. rigid flooring; subgroups: flooring type; outcome: number of head injuries. a, Hanger49 data unpublished; control group has been split between the three comparisons; b, events based on location of serious fall-related injuries (head/skull). IV, inverse variance; SE, standard error.

Sensitivity analyses

We received unreported data from Hanger49 that were slightly different to those published, and data for Gustavsson et al. 52 that had not been previously published (Johanna Gustavsson, personal communication). The Hanger49 data made no material difference to the findings (see Appendix 7, Table 37). The inclusion of Gustavsson et al. 52 improved the precision of the effect estimate to indicate that shock-absorbing flooring may reduce the number of head injuries (all studies combined), when analysed with falls as the unit of analysis (RR 0.55, 95% CI 0.31 to 0.97; p = 0.04; three studies; I² = 0%). The Gustavsson et al. 52 data also improved the precision of the effect estimate, but not the overall conclusions, when exploring care home findings (RR 0.57, 95% CI 0.31 to 1.07; p = 0.08; two studies; I² = 0%) and data for novel shock-absorbing floors (RR 0.59, 95% CI 0.33 to 1.04; p = 0.07; three studies; I² = 0%). The head injury data personally communicated to us from the Gustavsson et al. 52 study included head injuries of any severity, and should be interpreted with some caution as they are from an observational study and have not been adjusted for confounding (see Appendix 7, Table 37).

Fractures per 1000 person-days

It was possible to derive fracture rates for three of the included quantitative studies (none of these analyses have previously been reported). 49,50,77 Mackey et al. 50 opted not to fit regression models for fractures because of the low event rates and data for the other studies were even fewer. 49,77 These data do not provide any further information above and beyond the analyses we have already presented for the number of fractures. The data were too imprecise to detect whether or not shock-absorbing flooring makes a difference to fracture rates when focusing on RCTs alone (RaR 0.75, 95% CI 0.31 to 1.82; p = 0.52; two studies; I² = 0%; low risk of bias) or all studies (RaR 0.65, 95% CI 0.28 to 1.48; p = 0.30; three studies; I² = 0%; Figure 18), and there were no indications of any differences by setting type (χ² = 1.39, df = 1; p = 0.24; I² = 28.2%; see Figure 18) or study design (χ² = 0.75, df = 1; p = 0.39; I² = 0%; see Appendix 6, Table 36). We opted not to run subgroup analyses for flooring type because of the number of continuity corrections required when using this analysis method; instead, we refer readers to the outcome of number of fractures for this exploration.

FIGURE 18.

Forest plot of comparison: any shock-absorbing flooring vs. rigid flooring; subgroups: settings; outcome: fracture rate per 1000 person-days. a, Adjusted for clustering (assumed ICC = 0.002); continuity correction of 0.5 added to event counts because there were zero events in the intervention arm; b, number of person-days obtained via personal communication. IV, inverse variance; SE, standard error.

Sensitivity analyses

Selecting more or less conservative ICCs for Drahota et al. 77 did not affect the overall conclusions (see Appendix 7, Table 37). We ran a sensitivity analysis to include the data cited in the economic study by Lange;82 however, this only marginally improved the precision, without affecting the overall conclusions (see Appendix 7, Table 37).

Hip fractures per 1000 person-days

Derived hip fracture rates were obtainable from three studies (Figure 19). 49,50,77 The data were sparse and the imprecision means that there is insufficient information to determine whether or not shock-absorbing flooring can influence hip fracture rates (RCTs deemed to be at low risk of bias: RaR 0.73, 95% CI 0.13 to 4.01; p = 0.72; all studies: RaR 0.94, 95% CI 0.21 to 4.23; p = 0.94). There was no evidence to indicate any difference by setting (χ² = 0.02, df = 1; p = 0.88; I² = 0%; see Figure 19) or study design (χ² = 0.38, df = 1; p = 0.54; I² = 0%; see Appendix 6, Table 36). We opted not to run subgroup analyses for flooring type because of the number of continuity corrections required when using this analysis method and splitting Hanger49 down into even smaller pairwise comparisons; we refer readers to the outcome of ‘number of hip fractures’ for this exploration.

FIGURE 19.

Forest plot of comparison: shock-absorbing flooring vs. rigid flooring; subgroups: settings; outcome: hip fractures per 1000 person-days. a, Adjusted for clustering (assumed ICC = 0.002); continuity correction of 0.5 added to event counts because there were zero events in the intervention arm; b, number of person-days obtained via personal communication; continuity correction of 0.5 added to event counts because there were zero events in the control arm; c, number of events and person-days obtained via personal communication. IV, inverse variance; SE, standard error.

Sensitivity analyses

Selecting more or less conservative ICCs for Drahota et al. 77 did not affect the overall conclusions (see Appendix 7, Table 37).

Process outcomes

We did not identify any published process evaluations associated with the included studies; however, installation and maintenance were discussed as part of the qualitative outcomes, as well as rich data on the practical issues of using the floor (see Chapter 4). In addition, two hospital-based quantitative studies reported on practicality issues associated with shock-absorbing flooring, summarised in the following section. 49,77

Implementation and installation

The initial phase of Hanger49 was to determine the practicality issues associated with shock-absorbing flooring; additional costs of shock-absorbing flooring were identified as being one of the factors that influenced decision-making not to invest further in the products. No further details were provided of what these costs were and no mention was made of any issues related to installation. 49 Drahota et al. 77 described how the installation process was managed, with hospital wards planning for a 1-week installation in which intervention bays were gradually ‘run down’ by not admitting new patients or by transferring patients to vacant beds elsewhere in the ward or hospital. The installation process was planned directly between each hospital estate and facilities department and the prime contractors installing the floor. Thresholds between the intervention floors and standard floors in adjoining areas were managed using a transition strip or gradual ‘seamless’ gradient. The installation processes took place in accordance with the study schedule. One out of four of the intervention sites in Drahota et al. 77 reported a 20- to 30-cm split seam in the new floor, attributed to the welding at installation, which was subsequently repaired.

Flooring coverage and protection offered

One of the potential advantages of shock-absorbing floors is that they require no adherence, as such, on the part of patients/residents and staff. However, an overview of the evidence reveals that they are not a panacea in this regard, as not all flooring types are deemed suitable for all areas of an institution and falls do not always occur in the areas in which the shock-absorbing flooring maybe implemented. Table 11 reports data for studies in which we can assess the implementation areas for intervention and control groups. The hospital studies were similar in that they targeted the bedded areas, which are where most falls occurred (weighted average 84%; two studies). 49,77 The care home studies investigated different floor types, which were selected for use in different areas; notably, the SmartCells floor50 is a shock-absorbing underlay, which can be combined with vinyl coverings suitable for use in wet areas (the percentage of falls protected against was not given). Gustavsson et al. 52 included the communal areas alongside resident rooms (excluding bathrooms in the intervention group); 91% of falls occurred on one of the surfaces being compared in this study [note that the majority of the falls excluded (7.6%) occurred on mattresses and fall mats]. However, approximately 118 of the falls in the control group of Gustavsson et al. 52 were in bathrooms (not an area covered by the intervention floor), which, if accounted for, bring the proportion of falls occurring on target areas down to 78% in this study.

TABLE 11 - Floor coverage and proportion of falls protected

Reproduced from Drahota et al. 1 © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Study	Intervention	Areas covered by intervention flooring	Total number of falls	Percentage of falls on target areas
Hospitals
Drahota et al. 201377	Tarkett Omnisports Excel	Hospital bays (bedded areas excluding bathrooms and corridors)	68	75
Hanger 201749	Tarkett Omnisports Excel, Kradal and SmartCells	Hospital bays (bedded areas excluding bathrooms and corridors)	323	86
Care homes
Mackey et al. 201950	SmartCells	Resident rooms (living, bathroom and closet areas) excluding common areas (dining rooms, hallways, lounges, outside areas)	Not described; only bedroom falls reported
Gustavsson et al. 201852	Kradal	Resident apartments, communal dining room, corridor (excluding bathrooms and outdoor areas)	851	78

Impact on the working environment

Both Hanger49 and Drahota et al. 77 referred to issues associated with the increased effort required to move wheeled equipment across the intervention floors. Information documented in the 2011 full report of Drahota et al. 118 highlighted that one of the four intervention sites increased staffing (raised from six staff at baseline to seven staff during follow-up covering the 07.00–15.00 shift), which was reportedly to assist with the manual handling of equipment during the busy morning periods. Another intervention site in the same study altered the shift patterns of staff (maintaining the overall staffing levels), moving one staff member from the morning shift to increase the cover on the night shift. This alteration coincided with a patient group change (from a general care of the elderly to a care of the elderly with fractured neck of femur), but it was reported verbally to the researchers that the change was in response to concerns regarding moving equipment on the intervention floor during the night.

Head-to-head comparisons

Hanger49 was the only study that was able to provide (unpublished) data to support the direct comparison of different types of shock-absorbing floors. In this observational, hospital-based study, the published data from the different intervention floors were pooled into one intervention group to improve the power of the analyses; therefore, the analyses we present here are subject to large degrees of imprecision. The data have not been adjusted for confounding and the outcomes are at serious risk of bias. For these reasons, the quality of the evidence for all head-to-head comparisons has been graded as very low (Tables 12–14).

TABLE 12 - Summary of findings of Kradal vs. SmartCells

The risk with Kradal (and its 95% CI) is based on the assumed risk with SmartCells flooring (taken from Mackey et al. 50) and the relative effect (and its 95% CI).

See Appendix 8, Table 38, for explanations behind GRADE profiles.

These data should be interpreted with caution as the denominator (falls) used in the calculation of RR is count data.

Note

The GRADE Working Group suggested definitions for grades of evidence have been published elsewhere. 117

Outcomes	Anticipated absolute effectsa (95% CI)		Relative effect (95% CI)	Total sample size (n studies)	Quality of the evidence (GRADE)b	Comments
	Risk with SmartCells	Risk with Kradal
Injurious falls rate per 1000 person-days
Observational studies	2 per 1000	2 per 1000 (1 to 5)	RaR 1.00 (0.45 to 2.23)	6382 person-days (1 study)	⊕◯◯◯ Very low	These data are too imprecise to offer any certainty for this outcome
Falls rate per 1000 person-day
Observational studies	9 per 1000	13 per 1000 (9 to 19)	RaR 1.38 (0.94 to 2.02)	6382 person-days (1 study)	⊕◯◯◯ Very low	These data are too imprecise to offer any certainty for this outcome
Number of falls resulting in injury
Observational studiesc	254 per 1000	185 per 1000 (91 to 370)	RR 0.73 (0.36 to 1.46)	107 falls (1 study)	⊕◯◯◯ Very low	These data are too imprecise to offer any certainty for this outcome
Number of fractures
Observational studies	43 per 1000	12 per 1000 (0 to 236)	OR 0.27 (0.01 to 6.79)	73 participants (1 study)	⊕◯◯◯ Very low	These data are too imprecise to offer any certainty for this outcome
Number of hip fractures
Observational studies	11 per 1000	3 per 1000 (0 to 69)	OR 0.27 (0.01 to 6.79)	73 participants (1 study)	⊕◯◯◯ Very low	These data are too imprecise to offer any certainty for this outcome
Number of fallers
Observational studies	No data available
Adverse events
Observational studies	No data available

TABLE 13 - Summary of findings of Kradal vs. Tarkett Omnisports Excel

The risk with Kradal (and its 95% CI) is based on the assumed risk with Tarkett Omnisports Excel flooring (taken from Drahota et al. 77) and the relative effect (and its 95% CI).

See Appendix 8, Table 38, for explanations behind GRADE profiles.

These data should be interpreted with caution as the denominator (falls) used in the calculation of RR is count data.

Note

The GRADE Working Group suggested definitions for grades of evidence have been published elsewhere. 117

Outcomes	Anticipated absolute effectsa (95% CI)		Relative effect (95% CI)	Total sample size (n studies)	Quality of the evidence (GRADE)b	Comments
	Risk with Tarkett Omnisports Excel	Risk with Kradal
Injurious falls rate per 1000 person-days
Observational studies	2 per 1000	3 per 1000 (1 to 7)	RaR 1.50 (0.61 to 3.67)	6382 person-days (1 study)	⊕◯◯◯ Very low	These data are too imprecise to offer any certainty for this outcome
Falls rate per 1000 person-days
Observational studies	8 per 1000	12 per 1000 (8 to 18)	RaR 1.55 1.04 to 2.31)	6382 person-days (1 study)	⊕◯◯◯ Very low	If eight falls occur per day among 1000 inpatients on Tarkett Omnisports Excel (sports flooring), then very low-quality evidence suggests that four more (95% CI 0 to 10 more) falls per day may occur on Kradal (novel flooring)
Number of falls resulting in injury
Observational studiesc	229 per 1000	222 per 1000 (98 to 494)	RR 0.97 (0.43 to 2.16)	102 falls (1 study)	⊕◯◯◯ Very low	These data are too imprecise to offer any certainty for this outcome
Number of fractures
Observational studies	Not estimable, as there were zero events in both groups			68 participants (1 study)	⊕◯◯◯ Very low	There were zero events in both groups
Number of hip fractures
Observational studies	Not estimable. See comments			68 participants (1 study)	⊕◯◯◯ Very low	There were zero events in both groups
Number of fallers
Observational studies	No data available
Adverse events
Observational studies	No data available

TABLE 14 - Summary of findings of SmartCells vs. Tarkett Omnisports Excel

The risk with SmartCells (and its 95% CI) is based on the assumed risk with Tarkett Omnisports Excel flooring (taken from Drahota et al. 77) and the relative effect (and its 95% CI).

See Appendix 8, Table 38, for explanations behind GRADE profiles.

These data should be interpreted with caution as the denominator (falls) used in the calculation of RR is count data.

Note

The GRADE Working Group suggested definitions for grades of evidence have been published elsewhere. 117

Outcomes	Anticipated absolute effectsa (95% CI)		Relative effect (95% CI)	Total sample size (n studies)	Quality of the evidence (GRADE)b	Comments
	Risk with Tarkett Omnisports Excel	Risk with SmartCells
Injurious falls rate per 1000 person-days
Observational studies	2 per 1000	3 per 1000 (1 to 7)	RaR 1.50 (0.61 to 3.67)	6382 person-days (1 study)	⊕◯◯◯ Very low	These data are too imprecise to offer any certainty for this outcome
Falls rate per 1000 person-days
Observational studies	8 per 1000	9 per 1000 (6 to 13)	RaR 1.13 (0.73 to 1.72)	6382 person-days (1 study)	⊕◯◯◯ Very low	These data are too imprecise to offer any certainty for this outcome
Number of falls resulting in injury
Observational studiesc	229 per 1000	304 per 1000 (139 to 670)	RR 1.33 (0.61 to 2.93)	85 falls (1 study)	⊕◯◯◯ Very low	These data are too imprecise to offer any certainty for this outcome
Number of fractures
Observational studies	Not estimable. The risk with Tarkett is currently unknown because there are few data		OR 2.63 (0.10 to 67.17)	61 participants (1 study)	⊕◯◯◯ Very low	These data are too imprecise to offer any certainty for this outcome
Number of hip fractures
Observational studies	Not estimable. The risk with Tarkett is currently unknown because there are few data		OR 2.63 (0.10 to 67.17)	61 participants (1 study)	⊕◯◯◯ Very low	These data are too imprecise to offer any certainty for this outcome
Number of fallers
Observational studies	No data available
Adverse events
Observational studies	No data available

Three different shock-absorbing floors were incorporated into the study: two novel floors specifically designed for fall-related injury prevention among older adults (Kradal and SmartCells), and one originally designed as a sports floor (Tarkett Omnisports Excel).

Primary outcomes

Injurious falls rate per 1000 person-days

For this analysis, the number of person-days in each group was approximated based on the total number of bed-days estimated in the intervention arm of the published study (3191 person-days for each floor type). There were 12 injurious falls on SmartCells flooring and Kradal flooring, and eight injurious falls on Tarkett; the data were too imprecise to detect if any flooring may be superior to another at preventing injurious falls (Figure 20).

FIGURE 20.

Forest plot of head-to-head comparisons; outcome: injurious falls rate per 1000 person-days. a, Data obtained via personal communication for Hanger (with Dr Carl Hanger). 49 IV, inverse variance; SE, standard error.

Falls rate per 1000 person-days

As with the injurious falls rate, the number of bed-days in each group of this analysis were approximated to equal 3191. Fewer falls were experienced on Tarkett flooring (40 events) than on Kradal flooring (62 events), indicative of a positive effect, with uncertain clinical significance due to large imprecision (RaR 1.55, 95% CI 1.04 to 2.31; p = 0.03). These data should be interpreted with caution because any observed difference may be due to confounding. A similar number of falls (n = 45) were experienced on SmartCells as on Tarkett and, owing to the imprecision, it is not clear whether SmartCells may be any better or worse than the other floors (Figure 21).

FIGURE 21.

Forest plot of head-to-head comparisons; outcome: falls rate per 1000 person-days. a, Data obtained via personal communication for Hanger (with Dr Carl Hanger). 49 IV, inverse variance; SE, standard error.

Secondary outcomes

Number of falls resulting in injury

The two serious injuries that occurred in the intervention group in Hanger49 both occurred in a bedroom with the SmartCells flooring (one hip fracture and one head injury with a major laceration, which was caused by hitting the door surrounds rather than the floor). Patients experienced a similar proportion of minor and moderate injuries across the various shock-absorbing flooring types in Hanger49 (Figure 22).

FIGURE 22.

Injury severities on intervention floors in Hanger. 49

When exploring the number of falls that result in any type of injury, the data are too imprecise to determine any meaningful differences (Figure 23); it is possible that there are no differences between the shock-absorbing flooring types observed, but we cannot conclude this with any certainty.

FIGURE 23.

Forest plot of head-to-head comparisons; outcome: number of falls resulting in injury. a, Data obtained via personal communication for Hanger (with Dr Carl Hanger). 49 IV, inverse variance; SE, standard error.

Number of fractures

Only one fracture occurred on shock-absorbing flooring (SmartCells) in Hanger49 out of the 33 people who fell on the SmartCells flooring. Forty people fell on Tarkett flooring and 28 people fell on Kradal flooring, without fractures. 49 There are too few data to detect any suggestion of a difference between shock-absorbing flooring types.

Number of hip fractures

The single fracture that occurred on SmartCells in Hanger49 was to the hip; however, as stated previously, the data are too few to draw any meaningful conclusions as to whether or not certain shock-absorbing floor types may be more successful than others at reducing the risk of hip fracture.

Number of fallers

No data are available on the risk of being a faller from Hanger;49 this is because the data collected in the study were on those who fell only and not the entire patient population.

Number of head injuries

Data on the number of head injuries were also few in Hanger;49 in Figure 24, we present the comparisons using the Mantel–Haenszel method of analysis with ORs; however, the conclusions remain the same if using RRs. Although no head injuries were reported from falls on the Tarkett flooring, one of the head injuries in the SmartCells group was a result of an impact with the door surrounds, rather than the flooring, and only two further moderate head injuries were documented (one in each of the SmartsCells group and the Kradal group). With these data, we cannot conclude if any particular intervention floor is more favourable or confidently state if they have similar effects on head injuries.

FIGURE 24.

Forest plot of head-to-head comparisons; outcome: number of head injuries. a, Data obtained via personal communication for Hanger (with Dr Carl Hanger). 49 M–H, Mantel–Haenszel.

Rate of fractures and hip fractures

Given the sparsity of fractures (one fracture in the SmartCells group only) in Hanger,49 we have not attempted to derive fracture rates to make head-to-head comparisons for this one study.

Process outcomes

We do not have any data from direct comparisons of different shock-absorbing floor types pertaining to process outcomes.

Conclusions

We included 12 quantitative studies; however, only 10 of these contributed data to our prespecified outcomes, ranging from three to seven studies for any one outcome. We identified two well-conducted RCTs from different settings (one being very imprecise),50,77 and another small RCT with concerns around risk of bias;105 each of the RCTs investigated a different floor type. Potential confounding alongside other risks of bias afflicts the observational evidence. The studies were heterogeneous concerning the comparisons explored, settings and areas studied, population sampling, and definition of outcomes.

With these caveats in mind, there is very low-quality hospital-based evidence to suggest that certain types of shock-absorbing flooring may reduce injuries without increasing the rate of falls. The hospital-based data for risk of falling, fracture and head injury are too imprecise to determine effectiveness. The care home-based evidence suggests that novel shock-absorbing flooring (specifically SmartCells underlay with vinyl overlay and concrete subfloor) does not reduce the number of injuries (high-quality evidence), nor increase the number of falls (moderate-quality evidence) or fallers (high-quality evidence), when compared with a vinyl overlay with plywood underlay and concrete subfloor. 50 Incorporating the observational data on care homes suggests that shock-absorbing flooring may reduce the number of falls resulting in injury (very low-quality evidence). The fracture and head injury data in care homes are too imprecise to determine effectiveness.

The data for different types of floor coverings were often imprecise, with few studies in each subgroup. There were also some overlaps between setting type and flooring type, meaning that it is difficult to attribute any signals in the data to one particular factor. Novel floors, sports floors and carpet all demonstrated a reduction in injuries (very low-quality evidence) depending on how the data were analysed (i.e. as a rate or risk), while retaining the probability that they may not increase the risk of falls or of being a faller. The data on fractures were generally too imprecise to be conclusive; however, one observational study judged to be at high risk of bias, which compared wooden with concrete subfloors, indicated that fewer hip fractures were likely to occur on wooden subfloors. 76 The head-to-head comparisons for different flooring types were all personally communicated from the same observational study (unadjusted for confounding), and subject to large imprecision. Only one of the comparisons for one outcome (falls rate) was indicative of a potential difference, with Tarkett sports flooring outperforming Kradal novel flooring (very low-quality evidence); however, owing to the large imprecision and risk of bias for all of the head-to-head comparisons, the evidence remains uncertain.

The data for adverse events suggest that, although initial concerns of staff working on a shock-absorbing floor may be raised, there may be no difference in staff injuries over longer periods of follow-up; this evidence is of very low quality. Process outcomes indicate that shock-absorbing flooring affects the work environment, resulting in adaptations to staffing levels and schedules to accommodate the increased effort required to move wheeled objects on the intervention flooring. Some shock-absorbing floors have limitations as to where they can be laid (e.g. they are not all suitable for wet areas) and other objects can break a fall causing further injury (e.g. doorways and furniture). Therefore, the implementation of shock-absorbing floor coverings will not successfully protect 100% of fallers (even if they offer suitable shock absorbency); however, covering at least the bedded areas captured upwards of 75% of falls in these studies.

In summary, the high-quality quantitative evidence does not support the use of shock-absorbing flooring; however, this relates to a specific comparison (vinyl with novel underlay and concrete subfloor vs. vinyl with plywood underlay and concrete subfloor) in care homes. It is possible that a control group of vinyl on a concrete subfloor and/or a hospital setting with a higher turnover of (acutely unwell) patients would have produced a more pronounced effect. There is very low-quality evidence that shock-absorbing flooring may be beneficial; however, any estimate of effect is very uncertain.

Description of studies

Qualitative results

The five included qualitative studies yielded 69 findings, of which 61 were considered ‘unequivocal’ and eight were judged to be open to challenge, and therefore rated as ‘credible’. These findings were aggregated into 10 categories representing similar concepts (see Report Supplementary Material 2), which were subsequently grouped into three synthesised findings (Figure 25). We have numbered the 69 study findings according to their category (i.e. finding 1.1 is the first finding in category 1; 1.2 is the second finding in category 1, etc.), and tabulated them with illustrative quotations in Report Supplementary Material 2. Six findings contributed to two categories each owing to their breadth in scope; the remaining 63 findings contributed to one category each.

FIGURE 25.

Qualitative synthesis overview flow chart. Reproduced from Drahota et al. 1 © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reproduced from Drahota et al. 1 © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

The three synthesised findings were graded according to an assessment of confidence in the evidence, using the CERQual approach (Table 17). 85 We have moderate confidence in the evidence supporting the first and third synthesised findings, and high confidence in the second synthesised finding; a breakdown of our assessments can be viewed in the evidence profile table (see Appendix 9, Table 39). The findings should be read together to obtain the most coherent understanding of the evidence. The first finding relates largely to positive perspectives towards shock-absorbing flooring and the second finding captures more of the negative consequences of shock-absorbing flooring; the third finding has a smaller number of supporting data, but its congruence with the other findings strengthens its case.

Conclusions

Five qualitative studies of reasonable quality contributed to this synthesis, resulting in three synthesised findings:

1. Shock-absorbing flooring is viewed by many as a potential solution to help protect people from fall-related injuries, with a potential side effect of improving environmental comfort (moderate confidence in evidence.)

2. Changing a floor has consequences for the wider system (e.g. affecting the ease of moving equipment), potentially leading to further adaptations and adjustments in behaviours, attitudes, equipment, processes and staffing (high confidence in evidence).

3. Installation may be an initial concern, but it can be effectively managed; however, cost and funding considerations need to extend beyond the initial purchase and installation to consider potential adaptations in staffing/processes/equipment and potential cost savings from fall-related injury prevention (should the floor be effective) (moderate confidence in the evidence).

The studies predominantly focused on novel and sports floors; however, one study on carpet contributed credible evidence on issues of push and pull tasks. A broad range of views were represented from patients, residents, visitors and different staff roles. Although there may be divergent views on the effects of shock-absorbing flooring on slips, trips and falls risks; mobility; cleanliness; and how demanding it is to walk on, the findings highlighted that beliefs and attitudes towards the flooring are important as they can affect behaviours, routines and experiences. Views were convergent when it came to the increased effort required to push and pull equipment, and it would make sense for organisations to consider the adaptations required for this if they were considering implementing a shock-absorbing floor. Alongside the divergent views were findings expressing the desire for further evidence to support implementation decisions, not least around clinical effectiveness, giving rise to future research directions on the longevity of shock-absorbing floors, cost-effectiveness evidence that incorporates the costs of wider adaptations, effects on staff health and well-being, and the influence of shock-absorbing flooring on patient/resident mobility.

The studies

Five economic studies were included in this review. 82,119–122 Two studies were based in Sweden,82,121 one in New Zealand,122 one in the UK119 and one in the USA. 120 Four studies were in care homes, one was in hospitals (Table 18). We have assessed the studies against the 24-item Consolidated Health Economic Evaluation Reporting Standards (CHEERS) checklist to summarise the quality of reporting (see Report Supplementary Material 3). We have rated the reporting quality as poor (< 50% of items addressed), moderate (50–75% of items addressed) or high (> 75% of items addressed) according to how many items on the CHEERS checklist were addressed (see Table 18).

Research questions, interventions and settings

The shared, general research question was whether or not, and to what extent, absorbent floors were cost-effective in comparison with standard floors. However, the precise meaning of the question depended on the details of the alternatives compared and the assumptions made, both of which are discussed below.

Methods

Four studies used decision trees (of which three82,119,122 were explicit and one120 implicit) and one121 used a Markov model. The Markov model considers the risks and effects of recurrent events using time-dependent transition probabilities. Decision trees show pathways, generally for one-off events, within limited time frames. Longer time frames and the possibility of recurrence would seem desirable, particularly when modelling the effects of floors with long life expectancies in residential care homes. The studies, however, were limited by the data available.

The data used to populate the models varied widely. Only one study, Latimer et al. 119 (associated with Drahota et al. 77), used new data from a randomised comparison; however, when there were very few data, assumptions were based on the literature. 119 Another study82 used data from a small non-randomised experiment (not referenced but based on observing the effects in an extension to an existing care home), but supplemented by best estimates from the literature, which was made necessary by the small size of that study. The other three were desktop exercises using available estimates mainly from the published literature. 120–122 Although questions might be asked about the inclusion of these three studies,120–122 they are included here for completeness.

Results and discussion

Three studies found that absorbent floors dominated standard floors, that is costs were lower and outcomes were better. 120–122 Only Lange82 and Latimer et al. 119 disagreed. Lange82 estimated that the new floors had increased both cost and QALYs, but at a cost per increased QALY well above the accepted threshold level. Only Latimer et al. 119 found reduced QALYs, albeit with reduced costs, which, despite a favourable incremental cost-effectiveness ratio (ICER), was considered a result unlikely to lead to implementation.

What is more striking is the differences in relation to benefits. All of the studies but Latimer et al. 119 showed an increase in benefits. The QALY gains in the relevant studies were a direct result of assuming relatively large QALY losses due to hip fracture. Only Latimer et al. 119 collected data on quality of life.

The quality-of-life difference in Latimer et al. 119 was small, at 0.006 QALYs, well below estimates of the minimally important difference, which is put at around 0.4. Furthermore, the QALY loss was very sensitive to the overall risk of falling. In a sensitivity analysis in which the number of fallers was equalised, but the lower proportion of severe falls in the intervention group maintained, the flooring intervention would become dominant. 119

Overall conclusions on the studies

Overall, the quality of the studies was poor, with Latimer et al. 119 being the only exception. What is striking is that the best study found much less benefit than any of the other studies. 119 Although there was heterogeneity between the floors, settings and population groups assessed, it seems reasonable to conclude that the assumptions made in the other studies were unduly optimistic. Yet, in Latimer et al. ,119 both the cost and QALY differences were very small; neither was statistically significant. Given the differences discussed previously, it is clear that each study attempted to answer different research questions to do with particular comparisons and costs, and with very different assumptions.

Research recommendations for any future study might include: (1) improved specification of the alternatives evaluated; (2) distinguishing falls by severity; (3) specifying the processes by which reductions in type of falls were expected to lead to improved health; (4) using appropriate time frames, particularly when mortality was included; and (5) providing greater details (perhaps in an impact inventory) to enable different definitions of costs to be used in estimated ICERs.

Summary of main results

Triangulation and identifying gaps in existing evidence

The different types of evidence in this review were largely complementary of each other, exploring different elements of the research question; however, there were some exceptions. Qualitative evidence indicated that many people (including patients, residents, visitors and staff) view shock-absorbing flooring as a potential solution for preventing injurious falls. Although the more robust quantitative evidence did not confirm this to be true, very low-quality quantitative evidence was indicative of a positive effect. This poses a quandary, as it cannot be discerned whether the personal belief systems held about the effectiveness of shock-absorbing flooring are contributing to the bias in the low-quality evidence (leading to overly beneficial estimates) or if the qualitative evidence is merely reflecting what the low-quality quantitative evidence has found to be true. It may also be the case, as has been highlighted by senior care home managers54 and wider stakeholder groups,66 that individuals, such as senior managers, hold more guarded views: that they are open to shock-absorbing flooring as a potential intervention if it is first shown to be effective. The qualitative data were not directly linked to the specific comparison assessed in the non-significant trial contributing to the more robust quantitative evidence,50 which may further explain this contradiction. More robust quantitative evidence is required to resolve this impasse, as the current high-quality quantitative evidence relates to only one trial of a specific comparison (i.e. SmartCells underlay vs. plywood underlay, both with vinyl overlays and concrete subfloors) in care homes. 50

The qualitative and quantitative data were aligned in their conclusions that staff found manoeuvring equipment on shock-absorbing floors harder, and this can lead to adaptations being made in the workplace. The very low-quality quantitative data indicated that there were no more staff injuries on shock-absorbing floors than on rigid floors; however, it could not be discerned whether this was because of adaptations made to accommodate increased push and pull forces on the softer floors (e.g. increasing the number of staff involved in manoeuvring wheeled equipment), or despite any adaptations that may have been made. The numbers of adverse events reported in the trial data were much fewer than those from the observational research, despite employing similar definitions. This may be a product of the data collection mechanisms employed by the different types of research (spontaneous report monitoring vs. surveillance systems), in which the trial may have been more likely to capture only those events with higher perceived causality linked to the floor. 77 The observational data, on the other hand, may have been better at ascertaining all events that were possibly linked to the floor, but not necessarily caused by it. 106 The qualitative data, which were linked to the trial data, contradicted the reporting of adverse events somewhat, suggesting that there may have been unreported adverse events in the quantitative analysis of outcomes. 118 Future trials therefore need to incorporate more robust data capture mechanisms with better triangulation of sources to record staff injuries. All future quantitative research should better track and report the spontaneous and planned workplace adaptations that occur during the study, and consider which adaptations can be made in advance as part of the intervention implementation to better facilitate staff acceptance and manage risks.

The qualitative evidence highlighted that senior care home managers were aware of the potential for additional costs associated with shock-absorbing flooring related to the workplace adaptations required to manage risks to staff. These included costs related to staffing levels, training and equipment upgrades. To date, researchers have not incorporated the potential of these additional costs (associated with the intervention) into economic evaluations, which should be considered in future. Based on the existing economic data (of sports flooring),119 we conclude that, if the very low-quality evidence were true (i.e. that shock-absorbing floors decrease injuries without increasing falls), then shock-absorbing flooring would be a dominant strategy (i.e. would be less costly and result in better health outcomes). These conclusions are subject to change, however, with better-quality evidence and a more comprehensive assessment of costs. If the higher-quality evidence (that shock-absorbing flooring makes little difference to injury and falls rates)50 were to be more widely confirmed with other settings and floors, then the costs of the intervention would be likely to outweigh the benefits. The intervention floor (a novel underlay) assessed in Mackey et al. 50 was compared with a plywood underlay to give the study floors comparable thicknesses and provide an element of masking to the study design (both study groups had vinyl overlays and concrete subfloors). It is feasible that an alternative control group (with no underlay and a concrete subfloor) may have produced a different result.

In terms of the overall completeness and applicability of the evidence, there are gaps with regard to high-quality evidence in hospitals, and alternative brands/designs of novel floors, sports floors, carpets and wooden subfloors. Few of the studies measured our primary outcomes, and the way outcomes were defined and measured differed across studies, an issue which has also been highlighted elsewhere. 123–125 Future activities are required to establish a core outcome set for flooring interventions, and then to design studies that can measure these, as well as control for potential confounding.

Certainty of the evidence

We used the GRADE approach to summarising the certainty of the evidence. Following this approach, randomised studies without important limitations provide high-quality evidence, whereas observational studies without special strengths or important limitations provide low-quality evidence. We did not upgrade any of the evidence from observational studies on the grounds of any special strengths. Special strengths could include having a large magnitude of effect, a dose–response gradient, or when the effect of plausible residual confounding works is in the opposite direction to the observed effect; it is rare to upgrade the quality of evidence when important limitations are considered to be present,84 as was the case with the evidence in this review.

The quality of observational evidence tended to be downgraded further owing to risk of bias (with uncontrolled confounding being the predominant problem). Hospital-based RCT evidence was downgraded because of imprecision (i.e. wide CIs expressing uncertainty as to where the intervention effects may lie), as were fracture outcomes. A couple of issues related to inconsistency (i.e. heterogeneous study effects) were present in the care home data for falls rates and hip fractures, and, occasionally, we downgraded the evidence because of concerns around publication bias.

In Gustavsson et al. ,52 the measurement of our primary outcome (injurious falls rate) was considered plausible, but was not analysed or reported. In Donald et al. ,105 the data were not reported fully, precluding inclusion in the meta-analysis (relating to falls resulting in injuries); however, owing to the size of the study, this was unlikely to have a substantive bearing on the findings. For Mackey et al. ,50 the adverse events data had not been published in full, although they were summarised for the purposes of inclusion in this review. We were unable to summarise any of the adverse events data as rates or risks, because of unknown denominators; for this reason, we downgraded this outcome for indirectness. The adverse events data in Mackey et al. 50 were also considered indirect evidence as they did not fully fit our inclusion criteria. Although the adverse events data in this study were collected as part of a randomised trial, it was the residents’ rooms that were individually randomised, rather than the staff, making the observation of staff outcomes more of a before-and-after design nested within the trial. We still included these data because of the importance of the outcome and the known difficulties of obtaining data on adverse events. 126

Owing to all of these limitations, the evidence in favour of shock-absorbing flooring was graded as being of very low quality, whereas some high- and moderate-quality evidence exists based on Mackey et al. ,50 which found one specific shock-absorbing floor (SmartCells underlay) no more beneficial than a plywood underlay in a care home.

Potential biases in the review process

Some of our decision-making processes through the course of this review could be considered contestable. For example, we opted to include very low-quality evidence and incorporated data considered to be at high risk of bias into our analyses. We included unpublished data and derived data, which may have been less trustworthy. We have attempted to guard against these decisions by providing cautionary conclusions, emphasising the findings from more robust evidence when available and undertaking sensitivity analyses to check the influence of these decisions on the findings. We estimated the design effect of the included cluster randomised trial for our secondary outcomes, which were not analysed in the trial report, despite us having access to the original patient data. 77 This approach was taken as re-analysing the original data was considered outside the scope of this review owing to time and resource issues. As a pilot study, it was very small for a cluster randomised trial, contributing minimal data to the analyses, and sensitivity analyses around the choice of ICC demonstrated that this approach was likely to have negligible impact on the review findings.

Our review primarily focused on older adults as the target population, and staff were included only in relation to adverse events for quantitative outcomes. Limited literature exists around other outcomes in staff, such as comfort, fatigue96 and pain,31 as well as laboratory-style experiments testing the influence of flooring on the forces required to push and pull various items of wheeled equipment relevant to care settings, which have not been included in this review. 25,127,128 The qualitative data highlighted these outcomes; however, we did not consider them priorities in the quantitative outcomes list. We attempted to obtain missing data to support the analysis of our prioritised outcomes and incorporated these when possible. However, some of our communication attempts and/or data requests to study authors were unsuccessful, either because of the time that had passed since the study was conducted76 or non-response. 102,105

Agreement and disagreements with other studies or reviews

Systematic reviews covering flooring materials are sparse. The Cochrane systematic review19 on falls prevention in hospitals and care homes excluded studies on shock-absorbing flooring, with the rationale that the intent of the intervention is to reduce fall injuries, rather than falls. A Cochrane review129 on hospital environments has been conducted; however, it is outdated (the last search was in 2006), and, with different inclusion criteria, subsequently included only two flooring studies. 105,130 A 2019 review131 of floor finishes with a facility management focus and no quality assessment included 71 articles (not all primary research) covering issues such as the influence of flooring finishes on indoor air quality, infection control, recyclability, maintenance, durability, cost, comfort, noise, aesthetics, healing, flame resistance, ease of movement, glare and safety. The section on safety advocates for shock-absorbing flooring (defined as linoleum sheets and carpet only) and is not comprehensive, mentioning two studies82,105 and other literature reviews. 132,133

The present review is based on an update of the clinical effectiveness and cost-effectiveness sections of a previous scoping review. 20 As a scoping review, it was descriptive in nature, in line with standard scoping review methodology,134 and did not contain any quality appraisal or meta-analyses. New studies and fuller reports have also been published since the original search of the scoping review, which have subsequently been included in this systematic review. 49–54,106 Therefore, to our knowledge, this is the first comprehensive systematic review on shock-absorbing flooring interventions for fall-related injury prevention.

The scoping review did, however, include a broader range of literature on shock-absorbing flooring, including laboratory-based and biomechanical studies exploring outcomes such as impact absorption, gait and balance. 32–41,43–48,67,130,135–149 The high-quality data relating to the ineffectiveness of novel flooring in care homes for reducing injury rates50 is counterintuitive to laboratory studies, which often indicate the superior shock absorbency in these floors and their comparability to hip protectors in terms of impact absorption. 26,37,136 Without an overlay, the novel underlay used in the care home trial (SmartCells) has been demonstrated to attenuate peak force by up to 33.7%, compared with 2 mm of rigid flooring, in a simulated fall to the hip,33 and by 80% in a simulated fall to the back of the head. 35 Explanatory factors for the negative findings in this review and their disconnect from the wider laboratory-based research may relate to:

• the underlying assumptions of laboratory-based research associated with (1) the biofidelity of test systems, (2) the types of fall-related impacts simulated and (3) the type of impact required to sustain injury (and, conversely, the degree of impact attenuation required to prevent injury)

• co-interventions (e.g. hip protectors) negating the power of the study to detect a change attributable to flooring

• study populations having relatively low tissue tolerance values such that injuries occurred even for scenarios in which shock-absorbing floors reduced impact forces

• setting characteristics, such as grab bars or staff vigilance, moderating the intervention effect

• fall dynamics and environmental factors resulting in falls with impact sites and injury mechanisms not primarily associated with floors (e.g. walls, furniture, mobility aids).

These hypotheses would require further exploration to confirm or refute their basis. The laboratory research has indicated that shock-absorbing flooring can achieve greater impact reduction to head impacts35 and sensitivity analyses in our review (that included unpublished unadjusted data) identified a possible signal that head injuries may indeed be reduced in practice. This may imply that shock-absorbing floors need to achieve a higher force attenuation to successfully reduce injuries, making certain injuries (such as head injuries) more preventable than others (such as hip fractures). Alternatively, the higher force attenuation that can be achieved for head impacts and a higher baseline risk of head injuries (vs. hip fractures)115 may translate into a louder signal, which is easier to detect in the clinical research. Indeed, the fracture data in this review were generally too imprecise to detect a signal and further studies are likely to improve our understanding of potential effects.

Further biomechanical studies have been conducted on pushing tasks involving floor-based lift devices and wheelchairs. 25,127,150 One of these studies demonstrated increased pushing forces for a novel underlay (Sorbothane™, Sorbothane, Leyland, UK) that was not evaluated by any of the studies in the present review and is not in widespread use as a flooring material. 127 Both carpet and novel underlay (SmartCells) have been shown to increase the pushing forces required to initiate and sustain a manual wheelchair and floor-based lift; however, the addition of carpet has been shown to be more influential than a novel underlay to the measured hand forces. 25,150 Long-term care staff’s subjective ratings of pushing difficulty across different flooring conditions were all low for wheelchairs and motor-driven lifts (< 3 points on a 5-point scale),25,150 and higher for conventional lifts on carpet with or without shock-absorbing underlay. 25 It is considered significantly more difficult to wheel objects across carpet than across vinyl on concrete subfloors,25,150 and, although the introduction of a novel underlay did not significantly influence subjective difficulty ratings for wheelchairs and motor-driven lifts,25,150 it did for conventional lifts. 25 For conventional lifts,25 carpet with or without the novel shock-absorbing underlay exerted forces above recommended tolerance limits. 151 The type of lift used is highly influential, with motor-driven lifts negating the problems experienced with conventional lifts. 25 The qualitative research in the present review covered flooring brands different from the SmartCells underlay; in one study, a sports floor (Tarkett Omnisports Excel) was likened to carpet and staff expressed considerable concerns about pushing tasks. 118 In comparison to the SmartCells underlay, when Kradal, Omnisports Excel and carpet are laid, they are visibly different to a standard vinyl floor covering; we could speculate that this may also influence perceptions. The biomechanical literature complements the qualitative findings of the review by emphasising the important interactions between equipment types, flooring materials and pushing forces required, which indicate the potential for risk mitigation strategies to help alleviate the risks of adverse events and increased staff stress.

One of the concerns surrounding shock-absorbing flooring is whether or not it may inadvertently increase the risk of falling based on evidence that surfaces with extremely low levels of stiffness can influence balance control. 34 Our meta-analyses of falls rates per 1000 person-days and risk of being a faller were all non-significant, supporting the possibility that shock-absorbing flooring does not affect falls risk. Some of these analyses were imprecise, however, particularly in relation to evidence from RCTs in hospital settings, sports floors and carpets. We found very low-quality evidence in hospitals that the rate of falls did not increase with shock-absorbing flooring, and moderate- and high-quality evidence in care homes that falls rates and faller risk were not affected by SmartsCells flooring. These findings are in keeping with indirect evidence from the biomechanical-based literature, which suggests that, in general, individuals are able to maintain their balance on carpet36,138,139 and novel shock-absorbing floors. 33,34,36,38,152,153 Carpets and more compliant surfaces have been contraindicated, however, for their effects on balance in situations when sensory input is affected (e.g. with poor visual cues). 32,138,146 The majority of the biomechanical studies are conducted with healthy adults; however, some evidence exists based in clinical populations. Older inpatients have been observed to walk faster, with longer strides, on carpet than on vinyl,130 but these findings have been contradicted in a study of stroke patients who found carpet more challenging than parquetry. 39 One study has compared the abilities of inpatients (17 with stroke, 10 with Parkinson’s disease and 10 undergoing general geriatric rehabilitation) to perform the Timed Up and Go test on SmartCells, Tarkett Omnisports Excel and hospital-grade vinyl, and found no significant differences between flooring types (precision not reported). 153 This same study found no differences in patient-reported comfort and stability using visual analogue scales. 153 The current direct and indirect evidence, therefore, appears promising with regard to suggesting that falls are not increased on shock-absorbing floors; however, the evidence is imprecise in hospital settings, and biomechanical evidence of clinical populations is sparse.

Implications for practice

The evidence suggests that one type of novel shock-absorbing floor may not be effective in care homes, compared with rigid flooring; however, gaps still exist in the knowledge. There is very low-quality evidence that shock-absorbing flooring may reduce injuries in hospitals and care homes, without increasing falls, and that wooden subfloors may result in fewer hip fractures than concrete subfloors. The economic evidence (based on sports flooring) suggests that, if injurious falls are reduced and the number of falls not increased, then shock-absorbing flooring is likely to be a dominant strategy (i.e. costs would be lower and QALYs increased). The evidence in favour of shock-absorbing flooring is, however, of very low quality, meaning that future research may change our understanding and there is much uncertainty. If future research indicates an increased risk of falling on shock-absorbing flooring, then the economic evidence suggests that this could result in an undesirable reduction in QALYs, even if injurious falls were reduced.

The review findings indicate that introducing shock-absorbing flooring to care settings has wider workplace implications, meaning that adaptations may be required in staffing, equipment and processes. Research confirms that wheeled equipment can be more challenging for staff to manoeuvre on shock-absorbing floors. However, evidence indicates both that adaptations can be made to accommodate this and that there is no overall increased risk of flooring-related staff injuries (very low-quality evidence). The evidence indicates that, if planning to install shock-absorbing flooring, it is important to consider the wider impacts (and related costs) on the workplace and how best to manage these; the current economic evidence has not evaluated these costs.

Implications for research

We have prioritised the following recommendations for research:

• The current evidence base is diverse concerning how outcomes are defined, prioritised, measured, analysed and reported. In addition, there are complexities related to unit of analysis (i.e. individuals may experience multiple falls in different setting locations, and each fall may result in multiple injuries of different severities), which complicate analyses and future syntheses. Therefore, a fundamental step required in the field is the establishment of a clearly defined core outcome set, which includes recommendations for measurement, analysis and reporting. This piece of research should include a discussion around the relative merits and disadvantages of measuring injuries (and falls) as a product of the number of person-days, falls or even ‘active’ exposure time. This would require input from researchers, statisticians, clinicians, health and social care managers, policy-makers, and public and patient members, and should be designed to complement related core outcome sets (e.g. in community-based injurious falls prevention and hip protectors). 2,65

• The majority of quantitative studies included in this review were observational, judged to be at serious risk of bias, and did not address our primary outcomes of injurious falls rate and falls rate. Certain questions (e.g. related to common flooring types, such as carpets and different subfloors) may lend themselves well to observational study designs; however, these should be designed to address the above core outcome set and to comprehensively deal with potential confounding. Other questions (particularly those addressing new flooring interventions) lend themselves more readily to pragmatic RCT designs (of which there is a paucity). Co-interventions, which are broadly implemented in care settings, will need to be carefully considered in the design of future studies.

• There is a dearth of robust research on the effectiveness of shock-absorbing flooring in hospital settings. Hospitals differ from care homes, for example in terms of the population characteristics (e.g. with patients experiencing acute illness or episodes of reduced function), equipment in use, environmental characteristics and patient turnover. Future research should address this gap.

• This review indicates that implementing shock-absorbing flooring leads to adaptations in the workplace. Future research therefore needs to plan for these adaptations as part of the study design. For example, implementation studies should be accompanied by process evaluations and risk management plans to better mitigate, measure, manage and evaluate the risks to staff. Investigators could define the ‘shock-absorbing flooring intervention’ more broadly when devising the study research question to involve the implementation of a package of measures for the protection of patients and staff (such as new equipment more suitable for use on softer surfaces, an additional staff member to support manual handling activities). As part of these considerations, further research and innovation is required to identify how best to adapt the workplace to accommodate shock-absorbing flooring.

• There is currently limited high-quality economic evidence exploring the implementation of different flooring interventions. Future economic evaluations should –

  • provide improved specifications of the alternatives evaluated

  • distinguish falls by severity and type

  • specify the processes by which reductions in type of falls were expected to lead to improved health

  • use appropriate time frames, particularly when mortality is included

  • provide greater details (perhaps in an impact inventory) to enable different definitions of costs to be used in estimated ICERs; consideration should also be given to the costs of additional workplace adaptations.

• With the present uncertainty surrounding current flooring solutions, research and innovation is required to establish the specifications for improved products to support fall-related injury prevention in care settings. A multidisciplinary approach will be required to validate a set of mechanical properties for floors (measuring shock absorbency and push and pull forces) and establish the minimally important differences required to make a meaningful difference in practice.

We are unaware of any ongoing studies in these areas.

Contributions of authors

Amy Drahota (https://orcid.org/0000-0001-9772-0220) (Reader in Health and Social Care Evidence and Evaluation) contributed to the conception and design of the review; the acquisition, analysis and interpretation of the data; and drafting the final report.

Lambert M Felix (https://orcid.org/0000-0001-6517-9089) (Senior Research Associate, systematic reviews) contributed to the design of the review; the acquisition, analysis and interpretation of the data; and drafting the final report.

James Raftery (https://orcid.org/0000-0003-1094-8578) (Professor, Health Economics) contributed to the design of the review; the acquisition, analysis and interpretation of data; and drafting the final report.

Bethany E Keenan (https://orcid.org/0000-0001-7787-2892) (Research Fellow, Medical Engineering) contributed to the conception and design of the review, and the acquisition and interpretation of data.

Chantelle C Lachance (https://orcid.org/0000-0001-5755-5160) (Visiting Researcher, Knowledge Synthesis and Injury Prevention) contributed to the conception and design of the review, and the acquisition of data.

Dawn C Mackey (https://orcid.org/0000-0001-9854-1486) (Associate Professor, Ageing and Population Health) contributed to the conception and design of the review, and the acquisition and interpretation of data.

Chris Markham (https://orcid.org/0000-0001-8398-9333) (Head of School, Qualitative Research) contributed to the acquisition, analysis and interpretation of data.

Andrew C Laing (https://orcid.org/0000-0001-8128-011X) (Associate Professor, Injury Biomechanics and Ageing) contributed to the conception and design of the review.

Kirsten Farrell-Savage (https://orcid.org/0000-0002-2139-9095) (External Promotion and Liaison Lead, Communications and Public Involvement) contributed to the acquisition of data.

Olanrewaju Okunribido (https://orcid.org/0000-0003-4996-4890) (Ergonomics and Human Factors Specialist) contributed to data acquisition.

All authors critically revised the final report for important intellectual content and gave final approval, and agree to be accountable for all aspects of the work.

Publications

Lachance C, Felix L, Drahota A, Farrell-Savage K, Keenan B, Okunribido O, et al. The Shock-Absorbing Flooring Effectiveness SysTematic (SAFEST) review: a protocol. Poster presented at the Canadian Association on Gerontology, 24–26 October 2019, Moncton, NB, Canada.

Drahota AK, Keenan BE, Lachance C, Felix L, Raftery JP, Farrell-Savage KR, Mackey D. The SAFEST review: the Shock-Absorbing Flooring Effectiveness SysTematic review in care settings. Innov Aging 2019;3(Suppl. 1):S492–3. Poster presented at the Gerontological Society of America Annual Scientific Meeting, 13–17 November 2019, Austin, TX, USA.

Drahota A, Felix LM, Keenan BE, Lachance CC, Laing A, Mackey DC, et al. Protocol for the SAFEST review: the Shock-Absorbing Flooring Effectiveness SysTematic review including older adults and staff in hospitals and care homes. BMJ Open 2020;10:e032315.

Drahota A, Felix LM, Raftery J, Keenan BE, Lachance CC, Mackey D, et al. The SAFEST review: a mixed methods systematic review of shock-absorbing flooring for fall-related injury prevention. BMC Geriatrics 2022;22:32.

Data-sharing statement

Most technical data are included in the main report, appendices and supplementary materials. Project files are available on the Open Science Framework (https://osf.io/ev6xs/). Any additional data can be obtained from the corresponding author on request.

Patient data

This work uses data provided by patients and collected by the NHS as part of their care and support. Using patient data is vital to improve health and care for everyone. There is huge potential to make better use of information from people’s patient records, to understand more about disease, develop new treatments, monitor safety, and plan NHS services. Patient data should be kept safe and secure, to protect everyone’s privacy, and it’s important that there are safeguards to make sure that it is stored and used responsibly. Everyone should be able to find out about how patient data are used. #datasaveslives You can find out more about the background to this citation here: https://understandingpatientdata.org.uk/data-citation.

Disclaimers

This report presents independent research funded by the National Institute for Health Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health and Social Care. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health and Social Care.