Variations in the organisation of and outcomes from Early Pregnancy Assessment Units: the VESPA mixed-methods study

Primary aim

The primary aim of our study was to test the hypothesis that the rate of hospital admissions for early pregnancy complications is lower in EPAUs with high consultant presence than in EPAUs with low consultant presence.

Secondary aims

• To test the hypothesis that increased consultant presence in EPAUs improves other clinical outcomes, including the proportion of women having follow-up visits, non-diagnostic ultrasound scans, negative laparoscopies for suspected ectopic pregnancies and ruptured ectopic pregnancies requiring blood transfusion.

• To assess the effect of variations in opening hours and service accessibility on the overall admission rates and other clinical outcomes.

• To determine the optimal skill mix to run an effective and efficient EPAU service.

• To examine the cost-effectiveness of different skill mix models in EPAUs.

• To explore patient satisfaction with the quality of care received in different EPAUs.

• To make evidence-based recommendations about the future configuration of EPAUs in the UK.

Clinical outcomes in EPAUs

All women (aged ≥ 16 years) attending the participating EPAUs because of suspected early pregnancy complications for the first time during the index pregnancy were included in this strand of the study. A gestational age limit was set at 13⁺⁶ weeks’ gestation to standardise recruitment from all EPAUs across the country. There were no exclusion criteria for this strand of the study.

Emergency hospital care audit

Routine data were collected for all women who attended hospital emergency services because of early pregnancy complications over a period of 3 months, following completion of clinical data collection from the EPAU.

Patient satisfaction

All pregnant women (aged ≥ 16 years) attending EPAUs because of suspected early pregnancy complications who agreed to sign a written consent form to participate in the questionnaire arm of the VESPA study were included. Women who were haemodynamically unstable or in severe pain and those who declined consent were excluded.

Staff satisfaction

All members of staff directly involved in providing early pregnancy care were eligible to consent to this data strand. Non-permanent members of staff and locum and agency staff were excluded.

Qualitative interviews

Women who had taken part in the patient satisfaction survey and who had provided consent to being approached later to participate in a telephone interview formed the sampling frame for the qualitative interviews. Using the EPAU location and strata configuration, pregnancy outcome and participants’ 2-week post-discharge satisfaction score [Short Assessment of Patient Satisfaction (SAPS)27], a sampling frame was created to select a maximum variation, purposive sample of women. Please see Report Supplementary Material 1, Table 3, for the strata of each component of the sampling frame.

Health economic evaluation

The health economic evaluation included pregnant women (aged ≥ 16 years) attending EPAUs because of suspected early pregnancy complications who agreed to provide signed consent to participate in the questionnaire arm of the VESPA study. Women who were haemodynamically unstable or in severe pain and those who did not consent to participate in the questionnaire arm of the study were excluded from this data strand.

Workforce analysis

All interactions of women with any member of staff (i.e. administrative, nursing, medical) during their visits to the EPAU, and the duration and type of these interactions, were contemporaneously collected.

Clinical outcomes in EPAUs

All women who attended the participating EPAUs were included in this data strand. Clinical follow-up of women was organised by the local clinical team based on clinical need; no additional clinical follow-up visits were arranged for the VESPA study.

Emergency hospital care audit

All participating sites were asked to provide data for this data strand.

Patient satisfaction

Women who attended the participating EPAUs and had consented to participate in the questionnaire arm of the study were recruited in this data strand.

Staff satisfaction

Staff (including the principal investigators) were approached by the local research teams and provided with a PIL. A limit of 15 consented members of staff was set per site. If staff members had agreed to take part in the survey they were asked to sign the designated consent form with the consenting member of the local research team. Only staff working at the participating EPAUs during the period of patient recruitment were approached.

Qualitative interviews

Women who attended the participating EPAUs and who had consented to participate in the questionnaire arm of the study were recruited in this data strand. Recruitment began by contacting women who had rare pregnancy outcomes (i.e. molar pregnancies, ectopic pregnancies, terminations) as these accounted for the smallest proportion of women in the VESPA study overall, and it was anticipated that these participants would be the most difficult to recruit. Women who had experienced miscarriages were also contacted, as were women who had ongoing pregnancies. The method of recruitment was determined by the contact details we had for each participant and, therefore, a combination of first contact by e-mail (preferred) or letter, followed by telephone calls, was used.

Health economic evaluation

Women who attended the participating EPAUs and who had consented to participate in the questionnaire arm of the study were recruited to this data strand.

Workforce analysis

All women who attended the participating EPAUs were included into this data strand.

Diagrammatic representations of the journeys that women may have followed are shown in Figures 2–4. Figure 2 shows the journey of women at the initial visit, Figure 3 shows the potential journey of women at clinical follow-up visits and Figure 4 shows the journey of women following discharge from the EPAU.

FIGURE 2.

Journey of women at initial visit. CRF, case report form; EQ-5D-5L, EuroQol-5 Dimensions, five-level version; QA, questionnaire arm of the study; VAS-A, visual analogue anxiety scale.

FIGURE 3.

Journey of women at clinical follow-up visits (if applicable). CRF, case report form; QA, questionnaire arm of the study; VAS-A, visual analogue anxiety scale.

FIGURE 4.

Journey of women following discharge from the EPAU. CSRI, Client Service Receipt Inventory; EQ-5D-5L, EuroQol-5 Dimensions, five-level version; LMUP, London Measure of Unplanned Pregnancy.

Clinical outcomes in EPAUs

Demographic and routine clinical data were collected from all women attending the EPAUs [see case report form (CRF), URL: www.journalslibrary.nihr.ac.uk/programmes/hsdr/140441/#/ (accessed 2 June 2020)]. It was designed by the study team based on the CRF used for the pilot study, and modified in accordance with the feedback provided by the researchers and clinical staff who conducted the pilot study. All participating sites were provided with paper CRFs; however, the CRFs were also available online on the secure VESPA study website, hosted by MedSciNet (London, UK), in case sites opted to use the online version. CRFs were uploaded to the secure VESPA study website by a member of the local research team, either concurrently or retrospectively.

Emergency hospital care audit

Data for this data strand were collected following collaboration with the information services departments in participating hospitals to retrieve already collected data (from routine hospital systems). We collected data about the total number of A&E attendances for pregnant women under 14 weeks’ gestation. We also asked for data about the total number of emergency admissions, emergency operations and their outcomes, the number of women receiving blood transfusions and the number of admissions to intensive care units. Wherever possible, all data relating to hospital admissions that were provided by the information services departments were cross-checked by the site principal investigator or the lead local researcher, with support from a member of the central VESPA study research team, against ward admission books/databases, to ensure the accuracy of the data obtained.

The aim of the audit was to capture emergency activities of the hospital with regard to early pregnancy care, looking at A&E attendances for women in early pregnancy and emergency admissions (from A&E and other settings, e.g. outpatient clinics). The time frame covered at each unit was set at 3 months. The data were collected retrospectively, with the end of the 3-month period corresponding to the date when the last woman recruited to the clinical outcome data strand was discharged from EPAU care. The central team contacted the local research teams as soon as the last follow-up visit was established. The teams were provided with a Microsoft Excel^® file and a Microsoft Word^® document (Microsoft Corporation, Redmond, WA, USA). The documents listed the criteria to be included in the data set and showed the model search that was conducted at the host site as an example of the methodology for extracting the data.

Patient satisfaction

Women who consented to this arm of the study were asked to complete the SAPS27 measure, as well as a condition-specific patient satisfaction questionnaire (i.e. the modified Newcastle–Farnworth Questionnaire) [see modified Newcastle–Farnworth questionnaire, URL: www.journalslibrary.nihr.ac.uk/programmes/hsdr/140441/#/ (accessed 2 June 2020)] and the London Measure of Unplanned Pregnancy (LMUP)28 at 2 weeks post discharge from the EPAU. The SAPS questionnaire was chosen as it is a short, validated, well-established measure of patient satisfaction of hospital care. However, as it is generic and applicable across disciplines, it was supplemented by a pregnancy-specific measure. Given that a validated satisfaction questionnaire specific to early pregnancy care was not available, we modified and used, with permission from the key researcher, the ‘Patient Satisfaction Survey’ that was developed by Allison Farnworth and the team at Newcastle University as part of a Knowledge Transfer Partnership review of the Early Pregnancy Care Policy. The original questionnaire is specific to women who have experienced a miscarriage. The modified Newcastle–Farnworth questionnaire was developed by the co-investigators of the VESPA study for women who had been reviewed in EPAUs, irrespective of clinical outcome. As it is not a validated measure, the score was calculated as the average of the individual component scores. The LMUP, a short, validated measure, was used to define unplanned pregnancies and control for patient satisfaction (as there is emerging evidence that women with unplanned pregnancies report different levels of satisfaction with health-care services). Following careful consideration by the VESPA co-investigators and after consultation with women, through our links with the Miscarriage Association and The Ectopic Pregnancy Trust, the timing of the questionnaire was set at 2 weeks.

Staff satisfaction

Eligible members of staff who had consented were asked to complete a confidential and anonymous online shortened version of the standard NHS staff satisfaction survey. 29 We contacted Picker (Oxford, UK) to obtain permission prior to use; however, it was confirmed that, as the survey was intended for use for an NHS study, additional permissions were not required. To the best of our knowledge, the annual NHS staff satisfaction survey is the largest survey of staff opinion in the UK. The survey gathers views on staff experience at work and includes key areas such as (1) appraisal and development, (2) health and well-being, (3) staff engagement and involvement and (4) raising concerns. The VESPA study staff survey was based on the 2015 NHS staff survey [see VESPA staff survey, URL: www.journalslibrary.nihr.ac.uk/programmes/hsdr/140441/#/ (accessed 2 June 2020)]. Once valid consent was obtained and the designated consent form signed, the local research team would securely e-mail the central VESPA study e-mail account with the e-mail address of each consenting member of staff. The central team set up accounts for each staff member to complete the survey online, on the MedSciNet database system, and sent login details to the staff members individually. Sites had the option to complete the surveys on paper and return them by post to the central VESPA study office, where the responses were uploaded by the central team onto the database and the paper forms were disposed of in a confidential manner.

Qualitative interviews

As soon as fully informed consent was obtained, interviews were conducted with 39 participants over the telephone. The interviews lasted between 20 and 78 minutes. Participants were asked a series of questions as per the topic guide [see topic guide URL: www.journalslibrary.nihr.ac.uk/programmes/hsdr/140441/#/ (accessed 2 June 2020)], relating to the beginning of their pregnancy, their experience of attending the EPAU, their care following discharge from the EPAU, and their suggestions to improve quality of care and women’s experiences of the EPAU.

Two particular research questions were highlighted for this investigation:

1. How do women’s experiences of EPAU services vary by unit configuration?

2. How do women’s experiences of EPAU services vary by clinical outcome?

Health economic evaluation

Data were collected at up to three time points during the study. Quality-of-life [EuroQol-5 Dimensions, five-level version (EQ-5D-5L)30] data were captured before the consultation at the initial visit and then at two time points post discharge from the EPAU (planned at 2 and 12 weeks, but actually at 4 and 18 weeks; see Chapter 3 for clarification). Anxiety data were collected prior to clinical assessment for every visit (whether initial or clinical follow-up visit) using the visual analogue anxiety scale. 31 Patients also completed the same scale at the end of every visit. Data on resource use during visits (e.g. staff contacts, diagnostic tests) were collected as part of the study CRF. Data on additional resources used after the visit were collected from a sample of patients at a planned time point of 3 months.

Workforce analysis

Data for this data strand were collected as part of the study CRF. All members of staff, including administrative and clinical staff, who had contact with women attending the early pregnancy service were asked to record the type of interaction, the start and end time of their interaction, as well as their staff type. The duration of the interaction was then calculated and recorded on the study database at the time of uploading the CRF information to the VESPA study database.

Summary of data collection methodology

See Report Supplementary Material 1, Table 5, for a summary of the data collection tools used to collect data from the different data strands, the sources and the planned timings of data collection.

Clinical outcomes in EPAUs

We investigated the relationship between outcomes and consultant presence, unit volume and weekend opening hours using appropriate regression models (i.e. linear models for continuous outcomes, logistic models for binary outcomes and Poisson models for count outcomes). Hierarchical models were used when analysing patient-level data. Unit-level data were analysed using standard regression models.

Consultant presence was defined as the percentage of planned hours which consultants were expected to spend in the unit divided by the planned opening hours. We also defined a binary variable indicating planned consultant presence (yes/no), which was used in sensitivity analyses.

Volume was defined as the number of patient visits per year and was estimated using the time taken to obtain data on 150 patients and the average number of visits per patient. We also defined a binary variable indicating whether or not the yearly number of visits was greater than the median of 2500 (yes/no) for sensitivity analyses. This cut-off level was defined based on the data analysis, which showed that half of the EPAUs have a yearly visit volume of < 2500 visits per year.

Weekend opening was defined as the total number of opening hours per weekend. We also defined a binary variable indicating whether or not an EPAU was open at weekends (yes/no). In addition, a three-level weekend opening variable was defined: no/Saturday, only/Saturday and Sunday. The latter variables were used for sensitivity analyses.

Most models were adjusted using two sets of potential confounder variables:

1. final diagnosis (FD) and maternal age at initial visit (MA)

2. FD, MA, deprivation score (DS) and Gestational Age Policy (GAP).

Two sets were used because the DS (10 decile groups) and GAP had a reasonable number of missing data. Analyses based on the first set are described first in each instance.

Most of the statistical models are adjusted for FD (five groups), MA, DS (10 decile groups) and GAP. The FD and GAP groups are defined as follows.

Final diagnosis (five groups):

1. early intrauterine and normal/live intrauterine pregnancy

2. early embryonic demise, incomplete miscarriage, complete miscarriage, retained products of conception

3. ectopic pregnancy

4. inconclusive scan

5. other (molar, twin, not pregnant, etc.).

Gestational Age Policy (two groups):

1. < 6 weeks

2. ≥ 6 weeks.

The main analyses modelled consultant presence, unit volume and weekend opening hours using continuous variables. Sensitivity analyses were performed by replacing these continuous variables with the corresponding binary or categorical variables.

Emergency hospital care audit

The relationship between emergency admissions from A&E and consultant presence, unit volume and weekend opening hours was investigated by fitting multivariable logistic models. ‘Emergency admissions from A&E’ was defined as a binary outcome to indicate whether or not a patient had an emergency admission from A&E. Data from 29 EPAUs were used for this analysis. Models were either unadjusted or adjusted for GAP.

Patient satisfaction

We investigated patient satisfaction by exploring the relationship between the SAPS or the modified Newcastle–Farnworth score and consultant presence, unit volume and weekend opening hours [see modified Newcastle–Farnworth questionnaire; URL: www.journalslibrary.nihr.ac.uk/programmes/hsdr/140441/#/ (accessed 2 June 2020)].

The SAPS questionnaire consists of seven questions, with most of them having the following categorisation: 0 ‘very dissatisfied’, 1 ‘dissatisfied’, 2 ‘neither satisfied nor dissatisfied’, 3 ‘satisfied’ and 4 ‘very satisfied’.

Based on the above, the SAPS score may range between 0 and 28 points. 27 The modified Newcastle–Farnworth score was calculated as the average of the individual component scores.

Models were adjusted using three sets of potential confounder variables, two of which are defined in Clinical outcomes in EPAUs (FD + MA and FD + MA + DS + GAP). The third set of confounders was FD, MA, DS, GAP, ethnicity, parity, gravidity and LMUP score.

Staff satisfaction

The association between staff experience of providing early pregnancy care and consultant presence, unit volume and weekend opening was explored using descriptive statistics.

Qualitative interviews

Thirty-nine interviews were transcribed verbatim and analysis was conducted by members of the qualitative team, in an iterative and consultative manner,36 cross-checking with members of the wider VESPA study team for clinical and other facts related to other data collection strands, when appropriate. The data were analysed using a thematic framework analysis,37,38 focusing on women’s clinical and emotional pathways through their care experience at the EPAU, and how these were influenced by the configuration and practices of the service they used. The thematic framework was devised and agreed for the preliminary analysis. It was produced from the interview guide, the researchers’ knowledge of the content of each interview and the associated memo writings for each interview, as well as notes made during the refamiliarisation process.

In accordance with a thematic framework analysis approach appropriate for multidisciplinary health research,39 the interview transcripts were first read in their entirety to achieve refamiliarisation with the interviews and uploaded to NVivo software (QSR International, Warrington, UK) for management and analytical work up. Coding was initially broad and used text from within the transcript data as the preliminary codes, before the merging or splitting of some codes to form more defined strata. The second coding was more granular and further refined subthemes were generated, leading to a hierarchical structure of themes and subthemes which aligned with the pre-existing thematic framework.

All transcripts were coded by two members of the qualitative research team independently, and regular meetings were held to discuss and revise the coding thematic framework. Any discrepancies between researchers were resolved through explanations, debate and revisiting the data, to ensure that they had been completely coded and that the analysis satisfied a psychological, clinical and public health perspective40 for a dynamic health-care system (as in Zubairu et al. 41). We explored recommendations made by women for improvements to service and drew additional recommendations from our analysis.

Particular attention was paid to women’s emotional experience. These data were first coded to demarcate them as relating to an experience of emotions and then were refined with specific codes to reflect the emotion being expressed (e.g. ‘anxious’, ‘concern’, ‘worried’, ‘upset’, ‘uncertainty’). These codes were then grouped into four strata: (1) anxiety, (2) procedure-related uncertainty; (3) diagnosis-related uncertainty; and (4) upset. We also noted when women expressed feeling ‘guilty’ or ‘vulnerable’, or when women drew positives from a negative situation. Through iterative coding, reworking of the data regarding women’s emotions and comparison between interviews, we produced six robust and distinct emotional typologies which mapped to different care pathways through the EPAU.

For each of the other main themes we undertook a process of charting, creating a matrix for each theme for which text from each transcript was summarised, allowing retention of the original sense and meaning of each transcript. We then looked for patterns within the theme in accordance with strata or emotional typology to see how these factors influenced women’s experiences. If relevant, we linked back to the satisfaction score from the 2 weeks post-discharge SAPS score.

Health economic evaluation

The analysis takes that of an NHS health and social care payer perspective. Personal and productivity costs collected from a subsample of patients were analysed separately.

We analysed costs and outcomes at baseline and each follow-up time point. Costs were analysed, adjusting for the site-level stratification variables, as well as age and FD. We used a multilevel model to estimate adjusted costs. Multilevel models have been recommended for use in health economics as they are able to incorporate the hierarchical structure of data, including patients within centres, and provide more appropriate estimates of patient- and centre-level effects than ordinary least squares models. 32

We examined mean total costs and mean QALYs for each stratum, as well as mean change in anxiety pre and post consultation. We also carried out a probabilistic sensitivity analysis, which reflects uncertainty around the estimates of costs and QALYs. As the probabilistic analysis requires simulated samples from the mean cost and utility estimates, Monte Carlo simulation was performed within Microsoft Excel to obtain 10,000 simulated samples.

For each stratum we analysed the expected total QALYs and expected total cost, averaged over the simulation sample, together with 95% confidence intervals (CIs). We also computed net benefit for a given willingness to pay per QALY, λ (ceiling ratio), where net benefit is defined as:

This converts utilities to a monetary scale, so that the costs and QALYs can be compared directly. Expected net benefit is the average net benefit over the simulation samples. For a given willingness-to-pay threshold, λ, the optimal stratum is that with the highest expected net benefit. We present expected net benefit for λ = £20,000.

We allowed for the uncertainty in the optimal stratum by plotting the probability that each stratum is the most cost-effective (has highest net benefit) against willingness to pay per QALY, using cost-effectiveness acceptability curves (CEACs).

We also analysed mean total costs, QALYs and anxiety change at the unit level.

Workforce analysis

The workforce analysis calculated the time spent with each staff type for each visit and interaction, and used this time to calculate the salary cost for each staff type. The total cost for each staff type for each EPAU was amalgamated into a stratum. In this way, the staff cost profiles (showing each EPAU’s staff type makeup) could be presented by individual unit and stratum per 1000 patients. This also allowed comparisons between salary cost of staff types across EPAUs and strata.

The staff salary costs were further examined by FD, allowing further analyses. The average, median and standard deviations (SDs) of salary cost by diagnosis were computed and presented. Any statistically significant variations were identified using two-way analysis of variance. Other breakdowns/categorisations of staff salary costs, such as the number of patients seen and the time spent per patient, by each staff type, were also studied.

Among the final diagnoses, the number of inconclusive scans [pregnancies of unknown location (PULs)] was determined to be a useful indicator of the quality of an EPAU’s patient care: the higher number of PULs, the lower the level of care, and the lower number of PULs, the better the care. Number of admissions and number of visits were also identified as useful indicators. The analysis focused on the six clinical staff types: (1) consultant, (2) specialist nurse, (3) sonographer, (4) doctor, (5) nurse and (6) midwife. These indicators were investigated using correlation coefficients and ratios to determine if any of the staff types had an effect on them.

Primary outcome measure

• The proportion of emergency hospital admissions for further investigations and treatment, as a proportion of women attending the participating EPAUs (data strand 1).

Secondary outcome measures

• Total number of emergency admissions of women presenting with early pregnancy complications (data strands 1 and 2).

• Ratio of new to follow-up visits (data strand 1).

• Rate of non-diagnostic ultrasound scans (PUL). This has traditionally been used in early pregnancy as an indicator of quality of care (data strand 1).

• Proportion of laparoscopies performed for a suspected ectopic pregnancy with a negative finding (data strand 2).

• Ruptured ectopic pregnancies requiring blood transfusion (data strand 2).

• Estimated blood loss at operation (data strand 2).

• Patient satisfaction with the quality of care received (data strands 3 and 4).

• Staff experience of providing care in EPAUs (data strand 4).

• Proportion of women diagnosed with miscarriage and treated surgically, medically or expectantly (data strand 1).

• Proportion of women diagnosed with ectopic pregnancy and treated surgically, medically or expectantly (data strand 1).

• Visits to A&E departments (data strand 2).

• Admissions to intensive therapy units (data strand 2).

• Duration of admissions (data strand 2).

• Waiting times from referral to assessment (data strand 1).

• Quality-of-life measures before and after assessment of women at the EPAU during initial and follow-up visits (data strand 6).

• Anxiety before and after assessment of women at the EPAU during initial and follow-up visits (data strand 6).

• Cost-effectiveness of different staffing models (data strand 7).

Sample size

EPAU recruitment

A national survey was undertaken with the help of the AEPU. We approached 212 EPAUs and we received responses from 205. Each responding EPAU provided information about the key factors: planned weekly consultant presence, yearly number of attendances and weekend opening hours. Using a random sampling methodology we recruited 44 EPAUs with equal distribution of planned weekly consultant presence (yes/no), volume (< 3000 and ≥ 3000 attendances) and weekend opening (yes/no). The recruitment of the EPAUs was completed between December 2015 and April 2016. Following site initiation visits, four hospitals were moved to different strata as a result of changes in their configuration.

When data collection was complete, we reviewed the level of activity in all EPAUs. The volume of each EPAU was defined as the number of patient visits per year, and was estimated using the time taken to obtain data on 150 patients and the average number of visits per patient. The majority of the EPAUs (37/44) had a unit volume of < 4000 visits per year and 22 EPAUs had a yearly visit volume of < 2500 visits per year. In view of this, we used a cut-off point of 2500 visits to describe the units as high or low volume in the final data analysis.

See Report Supplementary Material 1, Table 6, for all participating EPAUs and their characteristics. Random three-letter codes were generated for each EPAU and used to anonymise data in accordance with our study protocol.

The distribution of planned consultant presence in the recruited units is shown in Figure 5. The majority of the EPAUs (41/44) had a planned presence of < 35% of opening time and 25 EPAUs had no planned consultant presence.

FIGURE 5.

Planned consultant presence at EPAUs, expressed as percentage of opening hours across 44 EPAUs.

The distribution of unit volume is shown in Figure 6.

FIGURE 6.

Estimated number of visits per year across the 44 participating EPAUs.

The distribution of weekend opening hours is shown in Figure 7. Approximately half of the EPAUs (23/44) were open at weekends, but the majority of these EPAUs (12/23) were open for < 10 hours.

FIGURE 7.

Distribution of weekend opening hours across EPAUs.

Patient and staff recruitment

Emergency hospital admissions as a proportion of women attending EPAUs

We first investigated the association between emergency admissions from the EPAU and consultant presence, unit volume and weekend opening hours. We used a binary outcome to indicate whether or not a patient had an emergency admission from the EPAU following any of her visits.

The relationship between emergency admissions and consultant presence is shown in Figure 12. There was only one unit without any emergency admissions from the EPAU.

FIGURE 12.

Emergency admissions from the EPAU vs. consultant presence for each unit.

The relationship between emergency admissions and unit volume is shown in Figure 13. The majority of the units (31/44) had an emergency admission rate between 1% and 5%, and a unit volume of < 5000 visits. There is no obvious trend.

FIGURE 13.

Emergency admissions from the EPAU vs. unit volume.

The relationship between emergency admissions and weekend opening is shown in Figure 14. Again, there is no obvious trend.

FIGURE 14.

Emergency admissions from the EPAU vs. weekend opening.

Fitting hierarchical logistic regression models for ‘emergency admission from the EPAU’ produced the results in Table 10. There is no evidence of an association with consultant time (p = 0.874) or unit volume (p = 0.247). However, there is evidence of an association with weekend opening hours (p = 0.027). That is, a 1-hour increase in weekend opening hours is associated with a 3.0% (95% CI 0.3% to 5.8%) higher odds of an emergency admission from the EPAU. These associations do not change after further confounder adjustment.

See Report Supplementary Material 1, Table 10, for the results of repeating the analyses using categorical variables. There is no evidence of an association when using categorical variables.

A sensitivity analysis was performed by excluding (statistical) outliers from the analyses. Specifically, units with large residuals (absolute value > 1) were removed. The analysis adjusting for FD and MA excludes four units (i.e. BDX, CZX, WWR and WDI), whereas the analysis adjusting for FD, MA, DS and GAP excludes four units with large residuals (i.e. BDX, CZX, WWR and WDI) and a further unit (i.e. SXM) because of missing data (GAP).

See Report Supplementary Material 1, Table 11, for the results of refitting the multivariable logistic models for admissions. There is now strong evidence of an association with weekend opening hours (p = 0.001) and weak evidence of an association with unit volume (p = 0.169). These associations do not change after further confounder adjustment.

We also recorded the actual amount of time the consultants spent in the EPAU delivering clinical care and compared it with the planned consultant presence (Figure 15). We found that, in general, the actual amount of time spent by the consultants in the units was less than planned and they were present for > 5% of time in just five units.

FIGURE 15.

Actual vs. planned consultant presence in the EPAU.

We then compared the emergency admissions and actual consultant presence (Figure 16).

FIGURE 16.

Emergency admissions from the EPAU and actual consultant presence.

We then repeated the analysis based on the actual consultant presence in the EPAU. The analyses adjusted for FD and MA were based on 6397 patients (97% of the total), whereas the analyses adjusted for FD, MA, DS and GAP were based on 5841 patients (43 units) (see Report Supplementary Material 1, Table 12).

There was no evidence of an association between the admission rate and consultant presence (p = 0.497). This relationship was consistent across adjustment models and different definitions of consultant presence.

As a sensitivity analysis, we also investigated this association by using a binary variable indicating consultant presence at each unit (see Report Supplementary Material 1, Table 13).

There was little evidence of an association between the admission rate and actual consultant presence (p = 0.376).

Emergency admissions as a proportion of women attending accident and emergency

We then investigated the relationship between emergency admissions from A&E and consultant presence, unit volume and weekend opening hours. We used a binary outcome for whether or not a patient had an emergency admission from A&E. This analysis is based on just 29 units (5464 patients). In total, 1445 (26.5%) patients had an emergency admission from A&E. The percentage of emergency admissions from A&E ranges from 7% to 58%, with the majority of the units having an emergency admission rate of between 10% and 30%.

The relationship between emergency admissions from A&E and consultant presence is shown in Figure 17. There is no obvious trend.

FIGURE 17.

Emergency admissions from A&E vs. consultant presence for each unit.

The relationship between emergency admissions from A&E and unit volume is shown in Figure 18. There is an indication of a decreasing trend.

FIGURE 18.

Emergency admissions from A&E vs. unit volume for each unit.

The relationship between emergency admissions from A&E and weekend opening is shown in Figure 19. There is a suggestion of a negative association between emergency admissions and weekend opening.

FIGURE 19.

Emergency admissions from A&E vs. weekend opening for each unit.

Fitting multivariable logistic models for ‘emergency admissions from A&E’ produced the following results (see also Report Supplementary Material 1, Table 14). These models were either not adjusted for potential confounders or adjusted for the GAP only. There is some evidence of an association between the emergency admissions from A&E and weekend opening (p = 0.037). A 1-hour increase in weekend opening hours is associated with 2.4% (95% CI 0.1% to 4.7%) lower odds of an emergency admission from A&E. However, there is no evidence of an association with unit volume (p = 0.647) or planned consultant time (p = 0.280). Adjusting for GAP does not change the conclusions. Repeating the analysis with categorical variables also leads to similar conclusions.

Ratio of emergency admissions from the EPAU and accident and emergency

The proportion of emergency admissions via the EPAU out of the total number of emergency admissions was calculated for 29 units, after adjusting for the different time periods used for data collection in the EPAU and emergency care audit data sets:

The relationship between the number of emergency admissions from the EPAU and A&E is shown in Figure 20. The number of emergency admissions via A&E is higher than that for the EPAU for the majority of the units (23/29).

FIGURE 20.

Emergency admissions from the EPAU vs. A&E over a period of 3 months.

The distribution of the percentage of emergency admissions via the EPAU out of the total emergency admissions is shown in Figure 21. The distribution is right-skewed and hence we used a log-transformation for our analyses.

FIGURE 21.

Percentage of emergency admissions via the EPAU out of the total number of emergency admissions across units.

The relationship between the percentage of emergency admissions via the EPAU and consultant presence is shown in Figure 22. There is perhaps a suggestion of a negative association.

FIGURE 22.

Percentage of emergency admissions via the EPAU out of the total number of emergency admissions vs. consultant presence for each unit.

The relationship between the percentage of emergency admissions via the EPAU and unit volume is shown in Figure 23. There is a suggestion of a positive association between the variables.

FIGURE 23.

Percentage of emergency admissions via the EPAU out of the total number of emergency admissions vs. unit volume for each unit.

The relationship between the percentage of emergency admissions via the EPAU and weekend opening is shown in Figure 24. There appears to be a positive association.

FIGURE 24.

Percentage of emergency admissions via the EPAU out of the total number of emergency admissions vs. weekend opening for each unit.

Report Supplementary Material 1, Table 15, shows the results of fitting multivariable linear regression models for ‘(log)-proportion of emergency admissions via the EPAU’. There is evidence that the proportion increases as weekend opening hours increase (p = 0.040).

Primary outcome analysis: key findings

• Consultant presence (both planned and actual) had no significant effect on emergency admission rates from EPAUs.

• Unit volume was not significantly associated with emergency admission rates.

• Emergency admissions through EPAUs were significantly higher with weekend opening.

• Rates of emergency admissions from A&E were significantly higher than admissions from the EPAU.

• Emergency admissions from A&E were significantly lower in EPAUs with weekend service, but there was no association with planned consultant time or unit volume.

Ratio of new to follow-up visits

We investigated the association between the ratio of new to follow-up visits and consultant presence, unit volume and weekend opening hours, using a binary outcome to indicate whether or not a patient had three or more visits.

The relationship between this outcome and consultant presence is shown in Figure 25.

FIGURE 25.

The percentage of patients with three or more visits vs. planned consultant presence for each unit.

The relationship between the ratio of new visits to follow-up visits and unit volume is shown in Figure 26. There is a suggestion that the percentage of patients with three or more visits increases as unit volume increases.

FIGURE 26.

The percentage of patients with three or more visits vs. unit volume for each unit.

The relationship between the ratio of new visits to follow-up visits and weekend opening hours is shown in Figure 27.

FIGURE 27.

The percentage of patients with three or more visits vs. weekend opening for each unit.

Report Supplementary Material 1, Table 16, shows the results of fitting multivariable logistic regression models for the ‘proportion of patients with three or more visits’. There is no evidence of an association with planned consultant time (p = 0.281) or weekend opening (p = 0.443). However, there is evidence of an association with unit volume (p = 0.025).

We also considered the relationship between the number of patient visits and consultant presence, unit volume and weekend opening hours.

The relationship between the number of visits and consultant presence is shown in Figure 28. There is little suggestion of an association.

FIGURE 28.

Number of visits vs. planned consultant presence for each unit.

The relationship between the number of visits and unit volume is shown in Figure 29. The number of visits appear to increase as unit volume increases.

FIGURE 29.

Number of visits vs. unit volume for each unit.

The relationship between the number of visits and weekend opening hours is shown in Figure 30.

FIGURE 30.

Number of visits vs. weekend opening for each unit.

Report Supplementary Material 1, Table 17, shows the results of fitting multivariable Poisson regression models for the number of visits. There is evidence that the number of visits increase as unit volume increases (p = 0.002).

Rate of pregnancy of unknown location

We analysed the relationship between the PUL rate at the initial visit and consultant presence, unit volume and weekend opening hours.

The relationships between the PUL rate at first scan and consultant presence, unit volume and weekend openings are shown in Figures 31–33, respectively.

FIGURE 31.

Rate of PUL scans at first scan vs. planned consultant presence for each unit.

FIGURE 32.

Rate of PUL scans at first visit vs. unit volume for each unit.

FIGURE 33.

Rate of PUL scans at first visit vs. weekend opening for each unit.

Report Supplementary Material 1, Table 18, shows the results of fitting multivariable logistic regression models for PULs at the initial visit. There is some evidence of a positive association between the PUL rate at first scan and unit volume (p = 0.07).

We also investigated the relationship between the total inconclusive scan rate (PUL) and consultant presence, unit volume and weekend opening hours. We used a binary outcome to indicate whether or not a patient had an inconclusive scan at any time during their follow-up.

The relationship between the rate of non-diagnostic ultrasound scans and consultant presence is shown in Figure 34. There is little suggestion of an association between PUL rate and consultant presence.

FIGURE 34.

Rate of PUL scans vs. planned consultant presence for each unit.

The relationship between the rate of non-diagnostic ultrasound scans and unit volume is shown in Figure 35. There is a suggestion that the PUL rate increases as unit volume increases.

FIGURE 35.

Rate of PUL scans vs. unit volume for each unit.

The relationship between the rate of non-diagnostic ultrasound scans and weekend opening is shown in Figure 36. There is no clear trend.

FIGURE 36.

Rate of PUL scans vs. weekend opening for each unit.

Report Supplementary Material 1, Table 19, shows the results of fitting multivariable logistic regression models for PUL rate. There is no evidence of an association between the PUL rate and consultant presence (p = 0.955) and weekend opening (p = 0.658). However, there is some evidence of a positive association with unit volume (p = 0.075).

Proportion of laparoscopies negative for ectopic pregnancy

We investigated the relationship between the rate of negative laparoscopies for suspected ectopic pregnancy and consultant presence, unit volume and weekend opening hours. However, only 90 laparoscopies were performed in 21 units, of which 16 (18%) were negative. There were six units at which all laparoscopies were negative and nine units with no negative laparoscopies.

The relationships between negative laparoscopies and consultant presence, unit volume and weekend opening are shown in Figures 37–39, respectively.

FIGURE 37.

The percentage of negative laparoscopies vs. planned consultant presence for each unit. The numbers indicate the number of laparoscopies at each unit (some points and numbers are overlaid).

FIGURE 38.

The percentage of negative laparoscopies vs. unit volume for each unit. The numbers indicate the number of laparoscopies at each unit (some points and numbers are overlaid).

FIGURE 39.

The percentage of negative laparoscopies vs. weekend opening hours for each unit. The numbers indicate the number of laparoscopies at each unit (some points and numbers are overlaid).

We fitted separate univariable logistic regression models for the rate of negative laparoscopies, with no confounder adjustment because of the small sample size (see Report Supplementary Material 1, Table 20). There is perhaps some weak evidence of an association with weekend opening hours (p = 0.112).

Patient satisfaction

Short Assessment of Patient Satisfaction score

The SAPS questionnaire consisted of seven questions, most of which have the following categorisation: 0 (very dissatisfied), 1 (dissatisfied), 2 (neither satisfied nor dissatisfied), 3 (satisfied) or 4 (very satisfied).

We investigated the relationship between SAPS score and consultant presence, unit volume and weekend opening hours.

There were variations in patient satisfaction between units. Some units have satisfaction rates in excess of 95%, whereas in other units the rates could be as low as 66%. The satisfaction scores for all individual units are shown in Table 11.

TABLE 11 - Observed and ranked data (from model) at the unit level for patient satisfaction (SAPS) as binary outcome (0–18 points: dissatisfied; 19–28 points: satisfied)

Ranking adjusted for FD and MA.

Note

Five patients were omitted from this analysis because of missing data.

Unit	n	Satisfied patients, n (%)	Ranka
MJL	41	40 (97.6)	1
ZAR	52	50 (96.2)	2
QSL	44	42 (95.5)	3
YUS	37	36 (97.3)	4
GFY	57	54 (94.7)	5
XQZ	46	44 (95.7)	6
AIS	44	42 (95.5)	7
OVA	28	27 (96.4)	8
RXO	63	59 (93.7)	9
XNL	47	44 (93.6)	10
CXP	27	26 (96.3)	11
ZNI	59	55 (93.2)	12
IWX	25	24 (96)	13
OYN	25	24 (96)	14
JPM	64	58 (90.6)	15
BDX	39	36 (92.3)	16
TCS	20	19 (95)	17
PCO	41	37 (90.2)	18
JNM	40	36 (90)	19
JDG	23	21 (91.3)	20
WDI	23	21 (91.3)	21
SCC	31	28 (90.3)	22
QAR	35	31 (88.6)	23
SXB	32	28 (87.5)	24
NSK	34	30 (88.2)	25
SHS	24	21 (87.5)	26
YLJ	47	41 (87.2)	27
HJZ	39	33 (84.6)	28
ZAM	28	24 (85.7)	29
XQD	21	18 (85.7)	30
SXM	10	8 (80)	31
WYW	50	43 (86)	32
BVU	42	36 (85.7)	33
VXL	36	31 (86.1)	34
FVX	49	42 (85.7)	35
ULV	35	30 (85.7)	36
GLR	35	30 (85.7)	37
JII	39	33 (84.6)	38
WWR	19	15 (78.9)	39
CZX	9	6 (66.7)	40
SDD	17	13 (76.5)	41
UOY	30	24 (80)	42
HYG	52	43 (82.7)	43
RPR	31	23 (74.2)	44

The relationship between the SAPS score and consultant presence is shown in Figure 40. There is some evidence that the SAPS score (satisfaction) decreases as consultant presence at the unit increases.

FIGURE 40.

Short Assessment of Patient Satisfaction score vs. planned consultant presence for each unit.

The relationship between the SAPS score and unit volume is shown in Figure 41. There is a suggestion that satisfaction decreases as unit volume increases.

FIGURE 41.

Short Assessment of Patient Satisfaction score vs. unit volume for each unit.

The relationship between the SAPS score and weekend opening hours is shown in Figure 42. There is no clear trend.

FIGURE 42.

Short Assessment of Patient Satisfaction score vs. weekend opening for each unit.

Fitting multivariable linear regression models for SAPS score produced the results shown in Table 12. A third set of confounders was used for these analyses, with additional adjustment for ethnicity, parity, gravidity and the LMUP score. There is some evidence of a negative association between SAPS score and planned consultant presence (p = 0.075).

TABLE 12 - Coefficients from hierarchical linear regression models for SAPS score with confounder adjustment

E, ethnicity; G, gravidity; L, LMUP score; P, parity.

Variable	Adjustment	Coefficient (95% CI)	p-value
Consultant time (%)	FD + MA (n = 1585)	–0.016 (–0.033 to 0.002)	0.075
Volume (per 100 visits)		–0.007 (–0.024 to 0.009)	0.361
Weekend opening (hours)		0.001 (–0.032 to 0.033)	0.973
Consultant time (%)	FD + MA + DS + GAP (n = 1555)	–0.017 (–0.035 to 0.002)	0.072
Volume (per 100 visits)		–0.008 (–0.025 to 0.009)	0.349
Weekend opening (hours)		0.003 (–0.031 to 0.037)	0.876
Consultant time (%)	FD + MA + DS + GAP + E + P + G + L (n = 1505)	–0.015 (–0.033 to 0.004)	0.117
Volume (per 100 visits)		–0.005 (–0.022 to 0.012)	0.538
Weekend opening (hours)		–0.0004 (–0.034 to 0.033)	0.980
Consultant presence (yes/no)	FD + MA (n = 1585)	–0.12 (–0.78 to 0.54)	0.717
Volume (≥ 2500 visits)		–0.17 (–0.88 to 0.53)	0.629
Weekend opening (yes/no)		–0.04 (–0.74 to 0.66)	0.914
Consultant presence (yes/no)	FD + MA + DS + GAP (n = 1555)	–0.13 (–0.83 to 0.57)	0.718
Volume (≥ 2500 visits)		–0.21 (–0.94 to 0.53)	0.584
Weekend opening (yes/no)		–0.02 (–0.76 to 0.72)	0.962
Consultant presence (yes/no)	FD + MA + DS + GAP + E + P + G + L (n = 1505)	–0.06 (–0.74 to 0.62)	0.859
Volume (≥ 2500 visits)		–0.15 (–0.87 to 0.56)	0.674
Weekend opening (yes/no)		–0.04 (–0.77 to 0.68)	0.905

Satisfaction versus waiting time for appointment

We investigated the relationship between the SAPS score and the waiting time for women’s first appointment (based on number of days from referral, self or other, to point in time seen in the EPAU). This analysis is based on 1576 patients.

The relationship between the average SAPS score and waiting time for the first appointment is shown in Figure 43.

FIGURE 43.

Average SAPS score vs. waiting time for appointment for each unit.

Fitting a linear regression model to model this relationship produced the results shown in Table 13. There is little evidence of an association between the SAPS score and waiting time for the first appointment (p = 0.203).

TABLE 13 - Coefficients from a hierarchical linear regression model for SAPS score vs. (log)-waiting time (n = 1576)

Variable	Coefficient (95% CI)	p-value
Waiting time (log-transformed)	0.187 (–0.101 to 0.474)	0.203

Modified Newcastle–Farnworth questionnaire score

We investigated the relationship between the modified Newcastle–Farnworth questionnaire score and consultant presence, unit volume and weekend opening hours [see modified Newcastle–Farnworth questionnaire; URL: www.journalslibrary.nihr.ac.uk/programmes/hsdr/140441/#/ (accessed 2 June 2020)]. The Newcastle–Farnworth questionnaire score was calculated as the average of the individual component scores.

The relationship between the Newcastle–Farnworth questionnaire score and consultant presence is shown in Figure 44. Figure 44 suggests that the Newcastle–Farnworth questionnaire score decreases as consultant presence increases.

FIGURE 44.

Newcastle–Farnworth questionnaire score vs. planned consultant presence for each unit.

The relationship between the Newcastle–Farnworth questionnaire score and unit volume is shown in Figure 45. Figure 45 suggests that the Newcastle–Farnworth questionnaire score decreases as unit volume increases.

FIGURE 45.

Newcastle–Farnworth questionnaire score vs. unit volume for each unit.

The relationship between the Newcastle–Farnworth questionnaire score and weekend opening hours is shown in Figure 46. Figure 46 suggests that the Newcastle–Farnworth questionnaire score is not associated with weekend opening hours.

FIGURE 46.

Newcastle–Farnworth questionnaire score vs. weekend opening for each unit.

Fitting multivariable linear regression models for the Newcastle–Farnworth questionnaire score produced the results shown in Table 14. There is evidence of negative associations with planned consultant presence (p = 0.027) and unit volume (p = 0.021).

TABLE 14 - Coefficients from hierarchical linear regression models for the Newcastle–Farnworth questionnaire score with confounder adjustment

E, ethnicity; G, gravidity; L, LMUP score; P, parity.

Variable	Adjustment	Coefficient (95% CI)	p-value
Consultant time (%)	FD + MA (n = 1597)	–0.003 (–0.005 to –0.0003)	0.027
Volume (per 100 visits)		–0.003 (–0.005 to –0.0004)	0.021
Weekend opening (hours)		0.003 (–0.001 to 0.007)	0.188
Consultant time (%)	FD + MA + DS + GAP (n = 1566)	–0.003 (–0.005 to 0.0001)	0.063
Volume (per 100 visits)		–0.002 (–0.005 to –0.0001)	0.041
Weekend opening (hours)		0.004 (–0.001 to 0.008)	0.092
Consultant time (%)	FD + MA + DS + GAP + E + P + G + L (n = 1517)	–0.001 (–0.004 to 0.001)	0.268
Volume (per 100 visits)		–0.002 (–0.004 to 0.0004)	0.114
Weekend opening (hours)		0.003 (–0.001 to 0.008)	0.113
Consultant presence (yes/no)	FD + MA (n = 1597)	–0.05 (–0.14 to 0.05)	0.310
Volume (≥ 2500 visits)		–0.02 (–0.12 to 0.08)	0.641
Weekend opening (yes/no)		0.02 (–0.08 to 0.11)	0.768
Consultant presence (yes/no)	FD + MA + DS + GAP (n = 1566)	–0.03 (–0.13 to 0.06)	0.507
Volume (≥ 2500 visits)		–0.02 (–0.13 to 0.08)	0.659
Weekend opening (yes/no)		0.02 (–0.08 to 0.13)	0.658
Consultant presence (yes/no)	FD + MA + DS + GAP + E + P + G + L (n = 1517)	–0.004 (–0.09 to 0.08)	0.925
Volume (≥ 2500 visits)		–0.01 (–0.10 to 0.08)	0.858
Weekend opening (yes/no)		0.02 (–0.07 to 0.11)	0.637

Staff satisfaction

We investigated the association between staff experience of providing care and consultant presence, unit volume and weekend opening.

The relationships between staff experience of providing care and consultant presence, unit volume and weekend opening are described in Table 15. There is a large observed difference of 17% in the percentage of staff who ‘witnessed potentially harmful errors, near-misses or incidents in the last month’ between the units with and units without consultant presence. In addition, the pressure felt by staff is 17% higher in units that are closed at weekends than in units open at weekends.

Secondary outcome analysis: key findings

• There was no association between consultant presence or weekend opening with the proportion of women attending for multiple follow-up visits.

• The proportion of women attending for multiple follow-up visits was significantly increased in high-volume units.

• Consultant presence and weekend opening had no effect on PUL rates.

• PUL rates were significantly higher in high-volume units.

• The rate of negative laparoscopies for suspected ectopic pregnancies was excessively high.

• Patient satisfaction tends to decrease with increasing unit volume and consultant presence.

• The proportion of staff reporting excessive work pressure is higher in units that are closed over weekends.

Short Assessment of Patient Satisfaction score

Short Assessment of Patient Satisfaction score ranged from 11 to 28 points in 38 participants (possible score range of 0–28 points). The median score was 25 points; this was lower than in the overall study as we oversampled adverse outcomes, which tended to have a lower score. Sixteen of the 38 women (42.1%) scored 27 or 28 points.

Question 6 [How much do you agree with the statement ‘the time you had with the (doctor/other health-care professional) was too short’?] was when most women dropped from a full score. Question 4 (How satisfied were you with the choices you had in decisions affecting your health care?) was the next hardest to endorse, with four women reporting that they were either dissatisfied or very dissatisfied. Conversely, question 5 [How much of the time did you feel respected by the (doctor/other health-care professional)?] was the easiest to endorse, with all but one woman saying that they felt respected all or most of the time. Overall, 37 of the 38 women reported being satisfied or very satisfied with the care they received (question 7).

Attendance at EPAUs

Women in this study each had different journeys through the EPAU service. Most reported their reason for attendance as bleeding or spotting and/or pain, although some used the service because of other pregnancy-related health complaints, such as vomiting, nausea, feeling unwell or lethargy. Some attended for early reassurance because they had prior history of pregnancy loss and, although not common, women were occasionally referred to clarify an inconclusive pregnancy test result.

Women in the sample either self-referred or were referred by a health-care professional (i.e. GPs, A&E staff or their midwives).

Within the EPAU, most women recalled having seen a sonographer, midwife or a nurse. Reception staff were next most commonly reported, followed by a doctor, an unidentified health-care professional, a health-care assistant or a consultant. One participant reported meeting a registrar and another reported meeting a student health-care professional. Women often struggled to distinguish between the different types of health-care professionals and their level of seniority. A ‘sonographer’ could potentially have been a midwife or a doctor.

Transabdominal and transvaginal ultrasound scans were the most common procedures women underwent at the EPAU, although blood tests, other physical health examinations and urinalyses were also reported (in descending frequency).

At the conclusion of their visit the majority of women were asked to attend for further follow-ups or to continue with their routine antenatal care. Some women, however, were referred on to other health-care professionals for further medical care.

Pathway 1: rapid positive diagnosis, positive outcome (anxious presentation)

Women presented to the EPAU with anxiety about their pregnancy. After investigation, these women were discharged with the positive outcome of a viable pregnancy, relieving their anxiety (at least to some extent). Of the 16 women who followed this clinical care pathway, 14 went on to have a livebirth; however, two women later lost the pregnancy (one had a surgically managed miscarriage and the other a termination for medical reasons).

Pathway 2: delayed positive diagnosis, positive outcome (sustained anxiety because of diagnostic uncertainty)

Women were anxious when they presented to the EPAU, but the initial scan or scans were inconclusive (often because of the pregnancy being too early to visualise on the scan), meaning that their concerns were not allayed. Women in this clinical care pathway often visited the EPAU on multiple occasions. All women on this pathway were eventually diagnosed with a viable pregnancy; this helped to relieve their anxieties to some extent and they were discharged from the EPAU. Four women followed this pathway.

Pathway 3: rapid negative diagnosis, negative outcome (anxious/upset)

Women in this pathway presented to the EPAU with concerns about their pregnancy and it was rapidly confirmed that the pregnancy was lost. As a result, their main emotion then became ‘upset’. There were seven women in this pathway (six had a miscarriage and one had a PUL).

Pathway 4: delayed negative diagnosis, no intervention required, negative outcome (anxious/upset after diagnostic uncertainty)

Women whose initial scan was inconclusive and whose pregnancy was deemed non-viable at their follow-up scan were discharged. Women in this pathway experienced anxiety at presentation, uncertainty while they waited for a diagnosis and became upset on confirmation that the pregnancy was lost. Five women followed this pathway (four had a miscarriage and one had a PUL).

Pathway 5: rapid negative diagnosis, intervention required, negative outcome (anxious/upset with procedural uncertainty)

In this pathway women presented to the EPAU and were immediately confirmed to have a non-viable pregnancy or to be experiencing a pregnancy loss that required medical or surgical intervention. In this typology women experienced anxiety, followed by uncertainty surrounding the procedure and, finally, upset at the loss of their pregnancy. Five women followed this pathway (two had ectopic pregnancies, two had miscarriages and one initially had an ongoing pregnancy confirmed, but later re-presented to the service requiring surgical management of a miscarriage).

Pathway 6: delayed negative diagnosis, intervention required, negative outcome (anxious with sustained uncertainty)

Women in this pathway had to attend their EPAU on multiple occasions to establish whether or not the pregnancy was viable. When they finally received the negative diagnosis, it was coupled with the need for medical or surgical intervention. Having presented with anxiety, they then experienced uncertainty at both the diagnostic and procedural stages. Three women followed this pathway (two had miscarriages and one had a molar pregnancy).

Other emotional outcomes

Some women also reported experiencing emotions which were found not to be specific to either emotional typologies or clinical care pathways. These emotions were grouped as ‘blame, guilt, or shame’, ‘isolation and vulnerability’ and ‘reassurance’.

The first of these, ‘blame, guilt, or shame’, was exclusively identified in women who had lost their pregnancy and was in relation to women not knowing why they were losing the pregnancy. Only five women mentioned feelings of ‘blame, guilt, or shame’ in their interviews, but those who did were found to question whether or not they themselves were at fault in some way for the loss.

Ten women spoke about ‘isolation and vulnerability’ in their interviews. During these interviews, it was noted that these feelings were usually present when women were discussing waiting for results to be discussed (especially if the waiting area they were in was public), how they felt immediately after being told they were going to lose their pregnancy, if they felt unsupported by EPAU staff, or if they had attended the EPAU alone.

Finally, ‘reassurance’ was evident in more than half of the interviews and across women with a range of clinical outcomes. The main essence of this theme was one of ‘taking positive emotions away from negative events’, whereby women were often relieved and reassured that their pregnancy was fine and they could continue with their pregnancy, or that they had indeed lost their pregnancy and they could now put that event behind them and try for another baby:

Yes, I think a big difference, it was when my partner came over, because I was . . . I wasn’t by myself any more. Yes, by the end of the night I was feeling better. When I realised, OK, it wasn’t my fault, I didn’t do anything to create this, it just happens, it’s very common and that my partner was there, so yes, slowly, slowly, I started to feel better. Then I had a very hard week. That week was quite hard, but yes.

Participant 016 (miscarriage, pathway 5, configuration VcW)

These supplementary emotional outcomes identified through our analysis are important to illustrate how emotionally charged women’s journeys through the EPAU can be. In addition, they emphasise how important psychological care is in EPAU consultations, patient management and in the aftercare provided to women.

Theme 1: barriers

The theme of ‘barriers’ included data on the difficulties women encountered when accessing and attending the EPAU. This was analysed by unit configuration only. Barriers included factors such as lack of awareness of the service, inability to obtain appointments and accessibility (i.e. travelling and parking facilities at the EPAU/local hospital).

Women reported that obtaining appointments was the most commonly raised barrier to accessing EPAUs. Analysis of awareness and accessibility by unit configuration showed no distinctive pattern. Lack of awareness of the EPAU service was a service-wide issue, whereas accessibility of the service in terms of distance from home and parking (including associated charges) was a location-specific issue (e.g. participants who had attended EPAUs in Scotland reported no parking charges, but those who had attended EPAUs in England did).

Unit configuration had an effect on women’s ability to obtain appointments. Inconvenient opening times and lack of available appointments were the main issues raised by women, with weekend closures highlighted as a particularly important issue. Availability of only daytime appointments was raised by some women as a barrier, particularly for follow-up appointments. Women mentioned the need to take time off work or to arrange child care to attend these appointments. Although the ability to obtain appointments was raised across all unit configuration strata, there were some differences. For example, over half of the women who had attended EPAUs that were closed over the weekend (i.e. configurations vcw, vCw, Vcw and VCw) reported the ability to obtain appointments as a barrier; however, approximately one-third of women who had attended EPAUs that were open at weekends (i.e. configurations vcW, vCW, VcW and VCW) also reported the ability to obtain appointments as a barrier:

I’d seen my GP on the Friday and this was the Monday. I had to wait the weekend out basically . . . But I think I probably would have gone sooner if it hadn’t been the weekend.

Participant 013 (molar, pathway 6, configuration Vcw)

Theme 2: efficiency

This theme includes data from women on how efficiently the EPAU they attended was run, including delays and waiting times (potential examples of understaffing) and how timely the service was delivered. It was analysed by unit configuration only. There was no particular pattern by configuration, but configuration types Vcw and VcW were notable for higher proportions of women (both with 50% of women per unit type) reporting ‘good’ efficiency. Examples of efficiency centred on short waiting times and few delays, leading to a smooth, clear process through the EPAU, with clinicians doing what they said they would do:

Obviously the unit is quite understaffed, and I know that a lot of the time it closes which isn’t great. But yes, in general it’s really good. I can’t really fault anything other than levels of staffing . . . I felt a bit rushed.

Participant 004 (ongoing, pathway 1, configuration VcW)

Theme 3: communication and information

Within this theme examples from women of where ‘communication and information’ had been sufficient (and participants had been satisfied) and insufficient (and therefore participants were unsatisfied) are included and have been analysed solely by emotional typology. Examples of ‘communication and information’ which left women unsatisfied included the use of jargon and giving out irrelevant information, which often compelled women to conduct their own research. Women who were provided with easily understandable information, in both verbal and written forms, and who were given the opportunity to ask questions, were more likely to be satisfied.

By typology, women who experienced an ‘anxious presentation’ emotional pathway were least likely to be satisfied with the level of ‘communication and information’. This was unexpected as these women had the fastest route through the EPAU and left with a viable, ongoing pregnancy. However, this was generally because their pregnancy complications and symptoms had not been explained, either at all or in sufficient detail. Another pattern was that of longer clinical journeys involving interventions (‘anxious/upset with procedural uncertainty’ and ‘anxious with sustained uncertainty’); the requirement for good ‘communication and information’ in these cases was higher, and women in this group were at greater risk of not receiving adequate or sufficient information:

. . . at the end of that appointment that was when we were then told, right, these are your options. I think probably a little bit of frustration as well that we weren’t told that the week beforehand but again I kind of understand why, because at that point I guess there’s the potential that it could have been a different result, hence why you have to go back on the second time, but it might have been nice just to have walked away from that first appointment feeling, OK, I’m coming back in a week’s time but the option that I will then have if the result is X will look like this, this, this, and then we kind of did our own research between the two appointments. It might have just given us a bit more time to think through what the right options really were.

Participant 026 (miscarriage, pathway 6, configuration vcw)

Theme 4: involvement in care decisions

This theme explored if and to what degree women felt adequately involved in their care. It was analysed by both unit configuration and emotional typology. The level of ‘involvement in care decisions’ varied: most women either felt sufficiently involved or reported that there were no decisions to be made. A few women, however, felt that they had complete control of their care. Some participants spoke about EPAU staff involving their partners in care decisions. Most women in configuration types vCW, VcW and VCw (between 75% and 100%) reported ‘good’ involvement in their care. Configuration types vcW and Vcw had the fewest women reporting ‘good’ involvement in decision-making. The common trait between these two configuration types was the fact that hospitals in neither configuration offered consultant-led sessions.

Only two women reported poor ‘involvement in care decisions’, limiting any meaningful analysis by typology. Given that the emotional typologies of ‘anxious presentation’ and ‘anxious/upset’ followed clinical care pathways which resulted in immediate discharge from the EPAU, with either an ongoing pregnancy or a miscarriage, most women in these typologies felt sufficiently involved, had no decision to make or made no comment. Two women who reported low levels of involvement had different emotional typologies and so a typology-specific issue was ruled out. Both women had been informed of the available management options, but they were not permitted to make their own decisions. Both these women gave low scores on question 4 of the SAPS regarding satisfaction with choices relating to their care:

. . . the nurse was explaining to me that, ‘Look, you can leave it happen naturally or we can keep you in and give you medication’, but it ended up that I didn’t really have a choice in the end, they were saying, ‘Look, we are just going to leave you, leave it to happen naturally’. So, at that point I was sent home and when I began to bleed very heavy and I didn’t know what to do . . .

Participant 019 (miscarriage, pathway 3, configuration VCW)

Theme 5: staff attitudes or approach

This theme investigated the way in which staff interacted with their patients at the EPAUs and was analysed by both unit configuration and emotional typology. All women, regardless of typology or level of satisfaction, reported good interactions with staff, describing them as kind, compassionate, empathic, competent and taking time to explain things, meaning that experiences were overwhelmingly positive in this theme. This finding supports the outcome on question 6 of the SAPS, with 37 out of 38 women saying that they felt respected most or all of the time:

I’d say the quality of care from the health-care professionals when I actually saw them was great, was nothing more than I would have expected. They had sound medical advice, they were able to support me emotionally. So, I think from the health-care professional’s point of view, I have absolutely no complaints, they were empathetic, they were lovely, they were supportive, they understood the situation, maybe not from a personal level but they were able to say the right things.

Participant 021 (miscarriage, pathways 1 and 5, configuration VCw)

However, one-third of women did report a negative experience at some point during their care, such as staff being cold, clinical or impolite. Some women reported feeling like ‘being on a conveyer belt’ during ultrasound scans. Analysis by unit configuration showed no clear pattern, other than vcw hospitals performing significantly worse than all others, with 50% of women reporting issues with staff. By emotional typology, shorter pathways (‘anxious presentation’, ‘sustained anxiety due to diagnostic uncertainty’ and ‘anxious/upset’) were less likely to result in experiences of poor staff attitudes, whereas longer pathways (‘anxious/upset after diagnostic uncertainty’, ‘anxious/upset with procedural uncertainty’ and ‘anxious/upset with sustained uncertainty’) were more likely to result in suboptimal interactions, possibly because these women had a greater number of interactions with more staff. Interestingly, participants who reported feeling rushed by staff all came from the ‘anxious presentation’ emotional typology, that is were found to have a viable pregnancy and did not need any further care from the EPAU. The fact that all the women who felt rushed were all in the typologies in which pregnancy was rapidly confirmed as either ongoing or lost highlights the need to remember that the EPAU’s job is not done at diagnosis. Women felt that the lack of any discussion or explanation as to their symptoms, particularly if the pregnancy was ongoing, affected their EPAU experience:

From memory, I remember it being quite quick. I don’t really know what I was expecting, I’ve never had a scan before, but I did remember thinking ‘Oh OK, is that . . . that’s it then, that’s that done?’ But again, she did say we are just checking the heart beat. I guess at 6 weeks there’s not really much else they can see, it’s so tiny, but I do remember thinking ‘Ooh OK that’s quick, we go now?’

Participant 034 (ongoing, pathway 1, configuration VcW)

Most experiences of poor care reported in this theme originated from the interaction with the person who was undertaking the scan, whom women sometimes perceived to be ‘cold’ or ‘clinical’. This is most likely because the person doing the scan was focused on technical aspects of the procedure, whereas the woman was wondering whether or not her pregnancy was healthy, and this came across as ‘clinical’. Simply explaining to the woman that the examination will be carried out in silence, as one woman experienced during her EPAU visit, could prevent health-care professionals being perceived as removed or detached:

The only thing I would say with the experience of going there is I wish there was some signs on the wall to say we will be quiet . . . When I went to the other unit, they had signs, ‘We’ll be quiet when we are looking at the scan’. I fed this back . . . ‘Please bear with us, we’ll speak to you as soon as we get the chance to’, sort of thing . . . Because they have that on the wall . . . It kind of reassures you that actually they are silent for a reason here, but when you are lying in that bed with the silence and you think ‘what the hell is . . . please just tell me is everything all right’, that would be beneficial.

Participant 029 (ongoing, pathway 2, configuration vcw)

Theme 6: continuity of care

In this theme women discussed how well their care was integrated. There were two types of continuity: (1) external (i.e. referral into the EPAU from A&E and discharge from the EPAU into other clinical or routine care) and (2) internal (i.e. within the EPAU service). This theme was analysed in relation to both unit configuration and emotional typology. Only six women reported aspects of poor continuity of care (two in each of configurations vcW, vCw and VCw). The two women in the VCw configuration had attended the same EPAU, which suggests that there may be a site-specific issue, rather than a configuration issue. Lack of communication between staff in individual EPAUs, and poor linkages between EPAUs and other parts of the hospital, GPs and community midwives, were issues which transcended unit type:

I think the overall system and process was . . . well, added to anxiety from the point of having to repeat myself so many times about what had happened, sometimes I felt like they didn’t really know me or know my situation and that actually if I hadn’t have said the right thing I could have ended up with a very different treatment path.

Participant 021 (miscarriage, pathways 1 and 5, configuration VCw)

By emotional typology, the main finding was that all women felt that the community-based health-care professionals were informed of their visit to the EPAU. Women diagnosed with miscarriages were particularly keen that this information was passed to their GP and/or midwife so that their post-loss care could be managed sensitively:

I don’t understand why an e-mail can’t be sent at the point that it’s clear that you’ve had a miscarriage to the GP to let them know that. I don’t know why we’re still relying on snail mail to get that information to GPs and that’s both times that I’ve had the miscarriage, the information has not made it to the GP by the time that I go to see the GP to get a sick line for my work and that must happen to a lot of people. And certainly the GP said, ‘I wish that I had been informed’, because he was on the verge of saying congratulations when I went in to see him and had no idea, so again other people might experience that and the doctor does say congratulations to them and it’s even more distressing to then have to say, ‘Well actually I’m here because I’ve had a miscarriage and I need a sick note’. So, I think that that is something that they could improve on and I don’t see why it would be a difficult thing to do.

Participant 015 (miscarriage, pathway 4, configuration vcW)

Although not particular to analysis either by unit configuration or by emotional typology, many women did report that they would welcome having access to, or at least some information about, psychological support services. In addition, there was a reported desire to have access to a form of aftercare or further check-ups to ensure not only that their physical health was continuing to improve, but that good psychological health was maintained after the loss of a pregnancy:

I suppose in one way it’s a little odd that you can have quite a physically horrific experience and potentially have no other interaction beyond that event. I don’t know if that is an oversight. I mean, I’m proactive, happy to go and see the doctor and I suppose I felt I had those options available to me. I just wonder if other people would and for anybody who didn’t have that support or the confidence, perhaps, I wonder if they could fall through the net in a very emotional point in their life.

Participant 014 (miscarriage, pathway 4, configuration vcw)

Women suggested that this could be provided by different health-care professionals, such as trained EPAU staff, health visitors, professional bereavement counsellors or specialist advisors. Other options worth considering are advice lines, listening services or informal buddies (volunteers who have had similar experiences in the past). What each suggestion had in common was the ability for women, and their partners, to speak about their loss, validate their feelings of grief, obtain reassurance about future pregnancies, and have the time to process and organise their thoughts surrounding their pregnancy loss:

. . . I think the main thing for me is just having some form of counselling or psychotherapy in place and just almost checking the welfare of the mother, for example . . . when I gave birth to my little one, the health visitor comes out and she checks on you, ‘Have you got baby blues? Is the baby OK?’ But when you have a miscarriage, it’s like you are not given the same treatment I suppose and I think what I would like to see in place in the future that it is treated like a real pregnancy, yes you have had a miscarriage but a health visitor is checking on, ‘Do you have depression, are you grieving, what support can we give you?’

Participant 019 (miscarriage, pathway 3, configuration VCW)

Theme 7: sensitive patient management

The final theme, ‘sensitive patient management’, was analysed by both unit configuration and emotional typology. This theme raised issues and best practices around privacy and practical sensitivity issues, such as the location of an EPAU service in terms of whether it was a separate unit or located within another hospital department (e.g. maternity or gynaecology ward) and the issues that such co-location can bring. Informing the GP or midwife that a pregnancy had been lost and cancelling scans and other appointments so that the women did not have to were examples of ‘sensitive patient management’:

No, they did tell me they would inform the midwife and my GP, and they would cancel any upcoming appointments, which they did do, which was really nice . . . Yes, that was really great because to ring up and have to say what’s happened wouldn’t have been very nice. So yes, it was nice to not have to think about cancelling any appointments.

Participant 038 (miscarriage, pathway 3, configuration Vcw)

By unit configuration, we found that women who attended low-volume EPAUs (i.e. vcw and vCw) were more likely to have a poor or mixed experience of ‘sensitive patient management’. Women were particularly concerned when the EPAU waiting area was shared with women at more advanced stages of pregnancy. They were also worried about privacy issues when personal information was discussed in a confined space in which the EPAU was run. Women were more likely to report good experiences of ‘sensitive patient management’ in high-volume EPAUs (e.g. Vcw, VcW and VCW): > 26% of women who had reported good experiences and < 8% of women who had reported poor or mixed experiences had attended an EPAU with configuration Vcw, VcW, VCw or VCW. This compared with < 11% of all women reporting good experiences and > 26% of women reporting poor or mixed experiences in EPAUs with configuration vcw, vcW, vCw or vCW.

The desire for a separate EPAU waiting area or building to maintain privacy was the dominant finding in this theme. When analysed by emotional typology, the desire for a separate EPAU waiting area or building to maintain privacy was present across all typologies. This desire stemmed from the fact that women who are losing pregnancies may become upset when waiting with women who are not only excited and visibly happy about their ongoing pregnancies, but also undergoing practices that go alongside ongoing pregnancies, such as offering flu vaccinations to pregnant women while they are waiting for scans:

Women going through miscarriages need to be in a slightly more isolated area, they don’t need to be walking into a door where there’s a woman walking out with her newborn baby, and I think the early pregnancy units, although they are run by midwives, I think they need to be carefully placed, not directly in the line of a labour ward or an antenatal clinic. I think that is so important.

Participant 001 (PUL, pathway 3, configuration vCw)

. . . you feel very vulnerable and in public when you are feeling a bit like oh this is a horrible thing that’s maybe happening, I would just rather have been in a quiet room somewhere on my own.

Participant 013 (molar, pathway 6, configuration Vcw)

. . . I had a woman come up to me and try and force me to have the flu jab because she was trying to go round and make it . . . you know, when it became recommended for you to have it in pregnancy and I said quite quietly, ‘I don’t need to, I’m here because I’m having a miscarriage’, and she didn’t hear, so I ended up having to say it quite loudly and then everyone stared. Yes, it sort of added to what was already a nasty experience.

Participant 006 (ectopic, pathway 5, configuration vcw)

I think I’d have found being so close to the maternity hospital and seeing these women coming out with their new babies or heavily pregnant women going into labour, I’d have found that really hard.

Participant 034 (ongoing, pathway 1, configuration VcW)

Privacy within the unit so that women are not overheard in their consultations and sensitivity when the EPAU staff are interacting with their patients in person or over the telephone were key factors in improving women’s experiences. However, overwhelmingly, the desire to separate EPAUs from other services was mentioned by all women irrespective of unit type, emotional typology and clinical outcome, as combined or shared EPAU services can be insensitive to women’s emotional states.

The analysis of themes is summarised in Table 17.

Recommendation 1: general pregnancy and pre-pregnancy care and information

• Raise awareness of EPAUs among women of reproductive age.

• Provide women with the information that they need when they leave the EPAU, dependent on their outcome. For example, for women experiencing a spontaneous miscarriage this could be information on what to expect in terms of the amount and duration of bleeding, how long to wait before trying to conceive again and preconception advice, such as folic acid supplementation, to improve their chances of a subsequent ongoing pregnancy.

Recommendation 2: accessibility of EPAUs

• Provide suitable opening times.

  • Re-evaluate weekend opening and address bank holiday closures.

  • Provide alternative opening hours to accommodate working women and women in need of child care.

  • When this is not possible, ensure that there are alternatives for when the EPAU is closed (e.g. A&E or maternity assessment units) and that there are clear protocols in place to manage women with early pregnancy complications when the EPAU is closed.

• Increase availability of appointments.

  • Ensure that telephone lines are answered so that women do not have to leave a voicemail and wait to be contacted to be booked in.

  • Introduce a telephone advice line for out-of-EPAU opening hours.

  • Ensure that appointments are not subject to blocking on referral, apart from required checks to ensure appropriateness of attendance.

Recommendation 3: staffing

• Ensure that EPAUs are appropriately staffed.

  • When viable, prevent staff from having to take on multiple roles (such as EPAU receptionist, sonographer and consulting nurse).

• Maintain optimal staff attitudes and approaches to care.

  • Discuss each process and procedure with women, keeping them informed of where they are in the process and what to expect next.

  • Ensure that whoever is conducting the scans explains the scanning process and engages with the woman to avoid a ‘conveyer belt’ experience.

Recommendation 4: within-EPAU experience

• Increase EPAU efficiency.

  • Ensure that waiting times are kept short and delays are kept to a minimum.

  • Allocate appropriate amounts of time for appointments.

  • Highlight the reasons for certain procedures and delays.

• Offer women a smooth process through the EPAU.

  • Introduce posters into EPAUs that explain the processes women can expect to go through when they are attending the unit.

  • Make sure that women’s notes are passed on to the EPAU staff they will see next, to prevent having to re-explain history, symptoms, etc.

Recommendation 5: managing patients sensitively

• Provide a distinct but integrated EPAU service by physically separating EPAUs from other hospital services, when possible, and maintain good cross-health-care links for ongoing care.

  • Prevent co-location with other maternity services.

  • Waiting areas should be EPAU specific.

• Emphasise to staff the sensitive nature of EPAU visits and ensure that they act accordingly.

  • Ensure that staff confirm that they have the correct person when contacting women about appointments via telephone, before announcing it is the EPAU.

  • Ensure that reception staff understand that women may ask to cancel appointments because of pregnancy loss and allow them to do so without questioning.

  • Prevent staff from inappropriately undertaking routine processes with women attending the EPAU (i.e. recruiting for flu vaccinations).

• Ensure privacy.

  • Allow women to fill out a referral form on arrival, rather than making them discuss their issue verbally among other people waiting.

  • Ensure that consultations are not overheard by either providing background noise (e.g. television) or ensuring a private location.

Recommendation 6: communicating and decision-making

• Provide clear and accessible information that is specific to the condition a woman has and/or procedures they will have.

• Involve women in their care decisions.

  • Provide women with a full range of options that are appropriate to their condition.

  • Ensure that women’s choices are respected and followed (while still clinically safe).

• Involve women’s partners in the processes and procedures (should women want them to be).

• Plan for the journey a woman has ahead of them.

  • Provide (in written form) the information a woman will need to understand and use to make decisions about what happens when they are discharged from the EPAU; be that when they go home (whether or not the pregnancy is ongoing), or when they are heading to theatre for surgery, onto a ward or to another specialty.

Recommendation 7: continuity of care

• Review referral and discharge processes to ensure a smooth transition into and out of EPAUs.

  • Make sure that whoever is taking over care after discharge from the EPAU is fully informed of women’s notes from the EPAU.

• Ensure that there is efficient communication between the EPAU and the rest of maternity, ongoing and/or hospital care.

  • Send notes promptly (electronically) from the EPAU to health-care professionals who will next provide care for women who have visited the EPAU.

• Provide appropriate aftercare.

  • Develop new, or reinforce, links with psychological support services to work with women requiring ongoing support after using EPAU services.

  • Ensure that information on psychological support services is provided.

  • Explain to women the availability and location of follow-up (GP, community midwife, EPAU, etc.).

Resource use and costs

The analysis included costs associated with ultrasounds, blood tests, admissions and staff time for which data were available for 6531 patients. Total costs take into account repeated tests and admissions, as well as staff salary costs. The mean total cost per patient was £225 (SD £537). The main contributor to total costs was surgical admissions, followed by ultrasounds. The mean total cost per patient for each cost component is shown in Table 18. Costs were assessed for the within-study period only (up to 3 months post visit) and no discounting was applied.

Total cost varied from £1 to £4390. The distribution of total cost is shown in Figure 48. It is clear that the majority of patients had a total cost of < £200, with a minority of patients incurring higher costs because of admission to hospital.

FIGURE 48.

Distribution of total cost.

The mean total cost per patient masks large differences in mean total cost by FD. The mean total cost in patients with a normal pregnancy was £92, in patients with early embryonic demise it was £473 and in the 12 patients with molar pregnancy it was £1793 (Table 19).

We looked at the mean total cost depending on annual patient volume, consultant presence during opening hours and weekend opening. The mean total costs are shown in Table 20.

Lower-volume units and no consultant presence were associated with lower costs than their alternatives. Lack of weekend opening was also associated with a lower mean total cost. The differences became less pronounced when age and FD were adjusted/controlled for.

We adjusted total cost for FD, age, yearly volume, consultant presence and hours open at the weekend using a multilevel model, allowing for clustering at the unit level. The results of the regression are shown in Table 21. Age (b = 2.74) is statistically significant at the 0.05 level (p = 0.005). This means that, for every year increase in age, a £3 increase in total cost is predicted, holding all other variables constant. A FD of early embryonic demise, incomplete miscarriage, retained products of conception, complete miscarriage, ectopic pregnancy, inconclusive scan or molar pregnancy is also associated with higher costs of early pregnancy care. There was no evidence of a relationship between the other variables and total cost.

Quality of life

We measured patients’ quality of life at baseline using the EQ-5D-5L questionnaire to calculate an index score. Baseline quality of life was available for 3764 patients. The distribution of index scores is shown in Figure 49.

FIGURE 49.

Distribution of index scores of quality of life/EQ-5D-5L at baseline.

The mean score at baseline in each configuration type is shown in Table 25.

Cost per quality-adjusted life-year at 4 weeks (configuration level)

Table 31 shows the expected total cost and expected total QALYs at 4 weeks for each configuration, along with their 95% CIs, estimated from the probabilistic analysis. Configurations are ordered by increasing expected total cost, with configurations vCw and vCW having the lowest expected costs and configuration VCw having the highest expected costs. Configurations Vcw and vcW have the highest expected QALYs at 4 weeks. Configuration VcW has the lowest expected QALYs. As the configurations have no more than a 0.007 difference in expected QALYs between them, they could be assumed to be clinically equivalent and so a decision between them is effectively based on minimising total costs. Note that the CIs show that there is a high degree of uncertainty in these estimates.

TABLE 31 - Expected total costs, expected total utilities and expected net benefit at a £20,000 willingness-to-pay threshold with a 4-week time frame

NMB, net monetary benefit.

Configuration	Mean cost (£)	95% CI (£)	Mean utility gain	95% CI	Probabilistic NMB (£)	95% CI (£)
vCw	173	143 to 206	0.069	0.053 to 0.086	1203	882 to 1556
vCW	180	143 to 223	0.069	0.055 to 0.083	1191	915 to 1496
vcW	216	191 to 243	0.070	0.059 to 0.081	1176	956 to 1409
vcw	226	183 to 277	0.069	0.055 to 0.083	1146	873 to 1440
VcW	227	198 to 257	0.065	0.047 to 0.087	1079	705 to 1508
Vcw	240	180 to 308	0.072	0.059 to 0.085	1190	935 to 1463
VCW	258	224 to 296	0.068	0.055 to 0.081	1092	834 to 1373
VCw	280	232 to 332	0.067	0.055 to 0.081	1064	812 to 1345

The expected net benefit at a £20,000 willingness-to-pay threshold is highest for configuration vCw (£1203) and lowest for VCw (£1064).

We present the uncertainty surrounding the cost-effectiveness of the various configurations, using a CEAC (Figure 50). The CEACs plot the probability that each configuration type is the most cost-effective by computing the proportion of simulations for which that configuration had the highest net benefit for a given willingness to pay per unit increase in utility. None of the eight configurations compared had a probability of > 30% of being the most cost-effective at any willingness-to-pay value > £10,000, indicating a large degree of uncertainty in the optimal configuration type. The configuration with the highest probability of being cost-effective at a willingness-to-pay value of ≤ £40,000 is vCw. At higher willingness-to-pay values, configuration Vcw has the highest probability of being cost-effective (see Table 31). However, because of uncertainty, it is not possible from these data to conclusively recommend a particular EPAU configuration.

FIGURE 50.

Cost-effectiveness acceptability curve: plotted against different willingness-to-pay values per unit increase in utility (ceiling ratio) at 4 weeks.

Sensitivity analysis: cost per quality-adjusted life-year at 18 weeks by configuration (n = 316)

As a sensitivity analysis we repeated the analysis using QALY scores at 18 weeks rather than 4 weeks. The mean score at baseline for the patients was 0.85 (SD 0.14) and at an average follow-up of 130 days the mean score was 0.90 (SD 0.13). Table 33 shows baseline and follow-up index scores by configuration, along with the mean index score, the SD of the mean, the utility gain [mean index score × (18 weeks)] and the number of women who completed the questionnaire at both time points in each configuration type.

TABLE 33 - Quality-adjusted life-years at 18 weeks by configuration

Configuration	n	Baseline index score	Index score at 18 weeks	Mean index score	SD	Utility gain
vcw	68	0.865	0.893	0.879	0.104	0.304
vcW	33	0.843	0.900	0.872	0.092	0.302
vCw	37	0.884	0.917	0.900	0.099	0.312
vCW	12	0.864	0.898	0.881	0.125	0.305
Vcw	51	0.850	0.927	0.888	0.088	0.308
VcW	54	0.830	0.883	0.857	0.173	0.296
VCw	23	0.806	0.931	0.868	0.140	0.301
VCW	38	0.877	0.897	0.887	0.100	0.307

Table 34 shows the expected total cost and expected total QALYs at 18 weeks for each configuration type, averaged over the simulation sample, along with their 95% CIs. At this time point configuration vCw still has the highest expected QALYs and configuration VCw has the lowest. The expected net benefit at a £20,000 willingness-to-pay threshold is highest for configuration vCw (£6067), followed by configuration vCW (£5918) and lowest for configuration VcW (£5718).

TABLE 34 - Cost per QALY at 18 weeks

NMB, net monetary benefit.

Configuration	Mean cost (£)	95% CI (£)	Utility gain	95% CI	Probabilistic NMB (£)	95% CI (£)
vCw	173	144 to 205	0.312	0.247 to 0.379	6067	4768 to 7409
vCW	180	142 to 221	0.305	0.224 to 0.393	5918	4308 to 7681
vcW	216	191 to 244	0.302	0.242 to 0.367	5816	4624 to 7131
vcw	226	181 to 274	0.303	0.234 to 0.376	5840	4455 to 7306
VcW	227	199 to 256	0.297	0.185 to 0.422	5718	3488 to 8196
Vcw	240	180 to 308	0.307	0.25 to 0.369	5899	4741 to 7135
VCW	258	223 to 294	0.307	0.241 to 0.379	5888	4573 to 7305
VCw	279	233.5 to 332	0.301	0.212 to 0.399	5732	3970 to 7707

We again present the uncertainty surrounding the cost-effectiveness of the various configuration types using a CEAC (Figure 51). There is more uncertainty at this point, with none of the configurations having a probability of > 20% of being the most cost-effective at any willingness-to-pay value > £10,000. At a willingness-to-pay threshold of £20,000 per QALY, VcW has the highest probability of being cost-effective, followed by vCw and vCW. As at 4 weeks, no recommendations can be made about optimal cost-effective EPAU configurations based on these findings.

FIGURE 51.

Cost-effectiveness acceptability curve: plotted against different willingness-to-pay value per unit increase in utility (ceiling ratio) at 18 weeks.

Pre- and post-consultation anxiety

Patients were asked to report their anxiety pre and post consultation for every visit to the EPAU. A total of 3550 pre- and post-consultation anxiety scores were available. Figures 52 and 53 show the distribution of anxiety scores pre and post consultation. It is clear that pre consultation there is a relatively even distribution of anxiety scores, ranging from 0 to 100, but post consultation the majority of women report an anxiety score of < 20 points.

FIGURE 52.

Distribution of pre-consultation anxiety scores.

FIGURE 53.

Distribution of post-consultation anxiety scores.

Figure 54 shows that most women (78%) experienced a decrease in their anxiety following their EPAU appointment (difference in score < 0 points). Of the 195 women whose anxiety score did not change from pre to post consultation, 69 had a score of 0 points to begin with (i.e. not at all anxious) and 27 had a score of 100 points (i.e. the most anxious they could ever imagine) at both time points.

FIGURE 54.

Change in anxiety score pre and post consultation.

The mean anxiety score pre and post consultation and the mean change in anxiety for each configuration and each unit are shown in Tables 35 and 36. Configuration vCw had the largest mean decrease in anxiety, whereas configuration vcW had the smallest mean decrease in anxiety.

The mean change in anxiety score by diagnosis is shown in Table 37. Women diagnosed with a normal pregnancy during their visit had the biggest decrease in anxiety, as might be expected, whereas women with molar pregnancies experienced an increase in anxiety.

Table 38 shows that there is evidence of a relationship between change in anxiety score and all final diagnoses other than not pregnant and ‘other’. There is no evidence of a relationship between age or unit configuration and change in anxiety.

Three-month questionnaire

The subset of women (n = 364) contacted at a second time point to inform the longer-term health economic analysis were asked to complete a questionnaire asking about their health-care usage since their visit. Table 39 shows the number of women who said they had no health-care contacts of a particular type, the numbers who had one, and the numbers who had two or three or more contacts. The table shows that around half of women self-reported at least one GP consultation. This was similar for community midwife visits. Thirty-five per cent of women had at least one outpatient visit and 25% of women had at least one hospital midwife visit. Table 40 shows the number of women with normal pregnancies who made contact with health-care services following their visit to the EPAU. Similarly, Table 41 shows number of women with non-normal pregnancies who made contact with health-care services following their visit to the EPAU.

We calculated an average cost to the NHS for these women, taking into account costs associated with each health-care contact. We excluded any non-NHS costs, such as charities and private counselling. Women were asked to record the outpatient department they attended. Of those who answered, 30% said that they had attended an antenatal clinic, 18% said that they had attended an A&E department and 12% said that they had attended a maternity unit. The other 40% of women used a combination of terms such as ‘triage’, ‘out of hours’, and ‘Brook ward’. An outpatient visit was therefore costed as a routine ultrasound scan in an antenatal unit, as this was the most common response. The unit costs used in these calculations are shown in Table 42.

The mean cost per patient at a mean follow-up of 118 days, including EPAU and self-reported follow-up costs, was £407. The mean cost per configuration is shown in Table 43. Configuration VCW had the highest mean cost and configuration vcW had the lowest mean cost. Tables 44 and 45 show the mean NHS cost per configuration at 17 weeks in women with a normal and non-normal pregnancy, respectively.

Health economic evaluation: key findings

• Main contributor to total cost was surgical admission.

• Mean total cost per woman with incomplete miscarriage was seven times more and mean total cost per woman with an ectopic pregnancy was 15 times more than the cost of care for a woman with a normal pregnancy.

• Low-volume units with no consultant presence were associated with lower costs.

• There was a positive change in a woman’s overall health at 4 weeks, largely due to changes in the pain/discomfort and anxiety/depression dimensions of the EQ-5D-5L.

• Mean utility gain varied from 0.065 to 0.071 per patient by configuration type and from 0.058 to 0.072 per patient by unit. The mean utility gain was similar across the board and it was not linked to age, FD, consultant presence or number of patients seen.

• Three-quarters of women reported a reduction in anxiety scores following their visit to an EPAU. The largest decrease in anxiety was in women diagnosed with normal pregnancies, but anxiety scores increased in those diagnosed with molar pregnancy.

Staff categorisation and salary costs

Overall salary cost per 1000 patients

The salary cost of each EPAU’s workforce can be considered an efficiency factor for the overall workforce. On average, 150 patients attended each of the 44 EPAUs. However, the salary costs of the workforce in each unit is different because of the different types of staff utilised and the variable amount of time they spent with each patient across all visits. In view of this, the salary cost is determined by the workforce configuration and the time staff spend with patients.

There are differences between the EPAUs in the descriptions of staff types they employ; however, they broadly fit into eight strata:

1. administrative staff

2. health-care assistant

3. midwife (includes all types of midwives)

4. consultant

5. doctor (includes all doctors below consultant grade)

6. specialist nurse (includes all nurses above staff nurse)

7. nurse (includes all staff grade nurses)

8. sonographer.

Figure 55 shows the salary costs for each unit per 1000 patients. Although the costs are three times as high in the unit with the highest salary cost than the unit with the lowest salary cost, the overall variation is not statistically significant. The lowest salary cost (£7530) of all the units is coded JDG, whereas the highest salary cost (£23,310) is coded JPM. The average cost across all units is £13,500.

FIGURE 55.

Staff salary costs profile across all configurations per 1000 patients.

We analysed the data further by grouping units of similar key characteristics (volume, consultant presence and weekend opening) into eight different configuration types. Figure 56 shows the staff salary cost profile across all configurations. The widths of the bars indicate the overall salary cost for a particular staff type. The y-axis is arranged in order of individual staff salary [i.e. consultants earn the most (top of axis) and administrative staff earn the least (bottom of axis)].

FIGURE 56.

Staff salary costs profile across all configurations overall.

The largest overall cost is for sonographers. If the costs for all strata/bands of nurses are added together, then all nurses will be the largest cost, followed by sonographers and all strata/bands of doctors.

Table 50 shows the salary cost profile of each staff type for each of the eight configuration types. Configurations vcw, vcW, vCw, Vcw and VcW all have similarly below average overall salary costs, whereas configurations vCW, VCw and VCW all have similar above average overall salary costs. Configuration VCW has the highest (£16,505) overall salary costs across all configurations. There is a significant difference between the configuration types when grouped as consultant present compared with non-consultant present (p = 0.037); this is not surprising as consultants are the highest cost staff type.

TABLE 50 - Staff salary cost per 1000 patients by configuration and staff type

N/A, not applicable.

Staff type	Configuration cost (£)								Overall cost (£)
	Vcw	vcW	vCw	vCW	Vcw	VcW	VCw	VCW
All staff	12,453	12,170	12,374	15,043	12,755	12,539	15,575	16,505	13,557
Administrative staff	200	77	89	14	228	62	290	209	146
Health-care assistant	36	N/A	162	551	1598	558	39	284	300
Midwife	1378	45	1024	2465	73	583	N/A	5	724
Consultant	984	70	2443	1048	708	485	3682	1581	1300
Doctor	1251	1233	2171	2095	941	1894	3650	2960	2021
Specialist nurse	2554	824	1443	422	2312	2129	527	5011	2237
Nurse	2282	3570	1662	2140	2821	3118	3921	2360	2674
Sonographer	3768	6351	3380	6308	4074	3709	3466	4096	4155

Staff salary costs profiles by configuration and units

Figure 57a–h shows the overall staff salary costs per 1000 patients for each EPAU within each configuration type. The overall variation of salary costs within each configuration is not significant. Figure 58 shows the proportions of staff salary costs across all individual patients and all visits for all EPAUs.

FIGURE 57.

Staff salary costs per 1000 patients for each configuration type and individual EPAUs [mean salary cost for configuration (red line) ± SD (diagonally shaded area)]. (a) Configuration vcw; (b) configuration vcW; (c) configuration vCw; (d) configuration vCW; (e) configuration vCW; (f) configuration VcW; (g) configuration VCw; and (h) configuration VCW.

FIGURE 58.

Proportions of staff salary costs across all individual patients and all visits for all EPAUs.

We analysed the staff costs per each patient visit. We found that the staff salary costs for over half of all visits was < £10. In the stratum of staff salary costs > £30 there are a few patients whose visits have individual costs of between £100 and £380 (see Figure 58). In these cases, women typically stayed in the EPAU for several hours, having bloods taken, intravenous fluids administered, long consultations and/or counselling sessions, resulting in higher staff salary costs.

Figure 59 shows the distribution of staff salary cost in relation to the visits across all configuration types. Configuration VCW has the largest proportion of all visits that cost > £30, whereas configurations Vcw and VcW have the highest proportion of visits that cost < £10. Configuration vCW had the lowest proportion of visits that cost < £10. Configurations VCw and VCW utilise higher proportions of the more expensive staff types (i.e. consultants, doctors and specialist nurses), which resulted in a higher proportion of high-cost visits.

FIGURE 59.

Distribution of staff salary costs across all individual patients visits by configuration type.

Average time spent with a patient

Figure 60 shows the average time spent with each patient across all staff types. The lowest average time spent with patients is 28 minutes (unit codes NSK and JDG), with the largest amount of time spent with patients being 71 minutes (unit code MJL). The average time spent across all units is 45 (SD 12) minutes. The observed overall variation is not statistically significant. A more detailed look at the specific units at either end of the average time distribution reveals that the longer times spent with patients are explained by longer consultations and counselling, whereas some patients required medical treatment (e.g. intravenous infusions). In units with shorter average times, there are no such patient interactions, which suggests that women requiring more complex care are referred to a nearby ward and not managed in the EPAU.

FIGURE 60.

Total time spent by all staff with each patient for each EPAU, showing overall mean and ± 1 SD.

Figure 61 shows the relationship between the time spent with patients and staff salary cost. The green markers are the individual unit salary costs and the red square is the average salary cost across all EPAUs. This figure shows the salary costs with the overall amount of time spent with the patient. The R² value of 0.6953 implies positive correlation between increasing time with patients and salary costs. Units with higher average salary costs are marked in green (i.e. OVA, YUS, SHS, VXL and JPM). When these five units are more closely examined, we find that the increased salary costs are explained by the higher use of consultants (i.e. OVA, YUS, SHS), doctors (i.e. OVA and VXL) and specialist nurses (i.e. JPM).

FIGURE 61.

Total salary costs per 1000 patients vs. the average total time spent by all staff with each patient for each EPAU, showing the mean and best linear fit.

Investigating the individual staff types’ time spent with a patient across all units reveals that there are some units which have individual patients who require a disparate amount of time compared with the vast majority of patients. The median of all the time spent with each patient for each staff type for each unit was computed and is shown in Figure 62.

FIGURE 62.

Median time each clinician type spends with patients across all units.

Figure 62 shows that the median times for consultant, nurse, sonographer and specialist nurse are all very similar. We found that the doctor and midwife staff types spend over 5 minutes more with patients than each of the other four staff types. The doctor staff type is made up of specialist registrars (88%) and specialist doctors, and has the second highest salary among all staff types. Table 51 shows that the doctor staff type sees more patients who have more complicated needs than any other staff type, as expected. Doctors are also typically asked to prepare women for admission, obtain consent for operative treatment and take blood samples. In view of this, the increasing length of time that doctors spend with patients is more likely to reflect the complexity of their tasks, rather than their lower efficiency. Similar considerations apply to midwives, who are typically expected to undertake more complex tasks than other nursing staff.

TABLE 51 - The proportion of interactions that each staff stratum spends with patients by FD

FD	Staff type (%)
	Consultant	Doctor	Midwife	Nurse	Sonographer	Specialist nurse
Normal/live intrauterine pregnancy	46	40	55	48	52	55
Early intrauterine pregnancy	11	11	11	14	13	9
Not pregnant	0	0	0	0	0	0
Complete miscarriage	13	12	9	10	11	13
Inconclusive scan (PUL)	2	4	2	2	3	2
Early embryonic demise	13	16	14	15	12	14
Incomplete miscarriage	4	8	6	6	5	4
Molar pregnancy	0	0	0	0	0	0
Retained products of conception	3	5	2	2	2	1
Ectopic pregnancy	6	4	1	2	2	2
Twin pregnancy	0	0	0	0	0	0
Other	0	0	0	0	0	0
Total	100	100	100	100	100	100

Workforce costs by diagnoses: salary costs

Figure 63 shows the staff salary cost per patient for all diagnoses. Figure 64 shows the total number of women in each configuration type across all diagnoses. In total, there were 6314 women with recorded diagnoses in the data set. The variations across configuration type are not significant. The average salary cost for staff attending to a woman in an EPAU across the UK is < £14 (range £10–17). Configuration VcW has the lowest staff salary costs and configuration VCW has the highest staff salary costs.

FIGURE 63.

Staff salary costs per patient for each configuration for all diagnoses.

FIGURE 64.

Total number of women in each configuration for all diagnoses. Note, there are 6531 women with timing data and 6409 valid diagnose in total, but only 6343 women with both timing data and diagnosis data.

Figure 65a–j shows the staff cost per patient by diagnosis stratum. The graphs are arranged in ascending order of staff cost based on the overall average cost per patient, as shown by the blue bar on each graph. Diagnoses of normal/live intrauterine pregnancy tend to have the lowest staff cost incurred for the time the patient interacts with, and is cared for, by the EPAU staff, with a cost of around £10 per patient across all configurations. Diagnoses of ectopic pregnancies are the most costly with an average of approximately £35, but reaching £46 for configuration VCW. The diagnoses across each configuration type have similar costs, except molar pregnancy. However, there are only 12 diagnoses of molar pregnancy in the data set.

FIGURE 65.

Staff salary cost per patient for the top 10 diagnosis strata by configuration type. Staff cost per patient with (a) normal/live intrauterine pregnancy; (b) early intrauterine pregnancy; (c) not pregnant; (d) complete miscarriage; (e) inconclusive scan (PUL); (f) early embryonic demise; (g) incomplete miscarriage; (h) molar pregnancy; (i) retained production of conception; and (j) ectopic pregnancy.

The differences observed were mainly due to the length of time spent by staff with women presenting with normal and abnormal pregnancies. Staff spend on average < 40 minutes with women presenting with normal pregnancies, but this increases to 60–75 minutes with women diagnosed with miscarriages and to > 100 minutes with women diagnosed with ectopic pregnancies. The spread in time spent by staff per patient shows differences between different configuration types. As a result, efficiency could be improved by shortening the length of interaction between staff and patients (Table 52).

TABLE 52 - Number of women seen by staff by diagnosis and strata (FD)

FD	Configuration (n)									Average	SD	SD/average (%)
	Vcw	vcW	vCw	vCW	Vcw	VcW	VCw	VCW	All
Normal/live intrauterine pregnancy	817	244	442	196	164	704	310	467	3344	418.00	238.96	57
Early intrauterine pregnancy	168	130	52	76	25	131	67	163	812	101.50	53.49	53
Not pregnant	3	1	1	1	0	1	2	3	12	1.50	1.07	71
Complete miscarriage	167	85	79	37	24	134	64	95	685	85.63	47.39	55
Inconclusive scan (PUL)	39	13	24	8	10	21	10	24	149	18.63	10.50	56
Early embryonic demise	134	62	85	49	49	192	86	143	800	100.00	51.46	51
Incomplete miscarriage	72	33	37	17	8	49	17	60	293	36.63	22.53	62
Molar pregnancy	3	0	1	1	0	2	2	3	12	1.50	1.20	80
Retained products of conception	40	8	7	5	7	23	8	25	123	15.38	12.61	82
Ectopic pregnancy	20	7	11	6	3	21	21	19	108	13.50	7.56	56
Twin pregnancy	0	1	0	0	0	0	0	1	2
Other	0	1	0	0	0	0	2	0	3
All	1463	585	739	396	290	1278	589	1003	6343	792.88	418.61	53