Point-of-care tests for urinary tract infections to reduce antimicrobial resistance: a systematic review and conceptual economic model

Article history

The research reported in this issue of the journal was commissioned and funded by the Evidence Synthesis Programme on behalf of NICE as award number NIHR135710. The protocol was agreed in December 2022. The draft manuscript began editorial review in February 2023 and was accepted for publication in May 2024. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors’ manuscript and would like to thank the reviewers for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this article.

Copyright statement

Copyright © 2024 Tomlinson et al. This work was produced by Tomlinson et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. This is an Open Access publication distributed under the terms of the Creative Commons Attribution CC BY 4.0 licence, which permits unrestricted use, distribution, reproduction and adaptation in any medium and for any purpose provided that it is properly attributed. See: https://creativecommons.org/licenses/by/4.0/. For attribution the title, original author(s), the publication source – NIHR Journals Library, and the DOI of the publication must be cited.

2024 Tomlinson et al.

Epidemiology and burden of urinary tract infections

Urinary tract infection (UTI) is one of the most common infections worldwide and is the most commonly seen bacterial infection in general practice. 1 UTI is also the most common hospital-acquired infection in the UK, accounting for almost one in four of all infections, most of which are associated with catheter use. 2 UTI can affect the lower urinary tract when the infection is in the urethra (urethritis) or bladder (cystitis), or the upper urinary tract when the infection is in the kidney (pyelonephritis). The incidence of UTI generally increases with age and is higher in women than in men; a 2019 study reported that around 83% of UTIs in primary care between 2011 and 2015 in England were in women. 3 Lifetime incidence of UTI in women is estimated at approximately 50–60%. 3 Risk factors for recurrent uncomplicated UTIs include frequent intercourse, vulvovaginal atrophy, change of the local bacterial flora, history of UTI, diabetes mellitus and a non-secretor blood type. 1,4

There are several classifications of UTI, depending on the location and frequency of infection and whether the patient is symptomatic. Classifications of uncomplicated UTI are summarised in Table 1. A proportion of patients will suffer from chronic UTI. There is no accepted definition of this, and its prevalence is unclear, but it is generally accepted that these patients will suffer ongoing symptoms with no or little relief between attacks. 5 This is in contrast to recurrent UTI, where symptoms do resolve completely between attacks.

Complications including pyelonephritis, kidney failure and sepsis may arise as a consequence of UTI. Additionally, infections during pregnancy can cause pre-term delivery and low birth weight. Risk factors for complicated UTI include structural or neurological abnormalities, pregnancy, catheterisation, certain infecting organisms and comorbidities such as immunosuppression. 6

The most common cause of both uncomplicated and complicated UTIs is Escherichia coli. 3 A recent UK-based surveillance study found that E. coli was isolated from 67% (113/169) of positive urine samples. Other bacteria identified in positive samples included Klebsiella pneumoniae (9%), Citrobacter koseri (5%), Enterococcus spp. (5%) and Staphylococcus saprophyticus (3.5%). 7

Presentation of urinary tract infections

Clinical presentation of UTI varies according to patient group and can be non-specific, making it difficult to identify those who may have a UTI. Symptoms can include dysuria (discomfort/pain/burning with urination), increased daytime frequency, urgency, abdominal/suprapubic pain, haematuria, and changes in urine smell, appearance or consistency. 8,9 In those aged > 65 years, symptoms can be less specific and include delirium, lethargy, a reduced ability to carry out activities of daily living, and anorexia. 6

Diagnosis

The accurate and timely diagnosis of UTI is important to ensure appropriate treatment to help resolve symptoms and improve quality of life and also to reduce the risk of long-term complications such as pyelonephritis, kidney disease and sepsis. 10

Urinary tract infections are currently diagnosed using a combination of dipstick tests and laboratory-based urine culture, which usually includes antimicrobial sensitivity testing (AST). Dipstick tests involve dipping a specially treated paper or plastic strip into a urine sample to identify the presence of leukocyte esterase (LE), nitrites and blood. These tests can be used as initial screening for UTI as they can be performed by general practitioners (GPs) and give a result very quickly (within a few minutes), but their accuracy is limited, particularly in certain populations such as men, those aged > 65 years and those who are catheterised, in whom they are not recommended. 11 They are also unable to provide information on the pathogenic cause of the infection or on AST. Thus, even when these tests are used to help diagnose a UTI, follow-up laboratory testing using culture is often needed to confirm the infection and to determine AST. The European Committee on Antimicrobial Susceptibility Testing (EUCAST) provides guidance on AST that includes definitions of susceptibility testing categories with the aim of harmonising breakpoints in Europe. 12

Culture can take 24–72 hours depending on geographical location and local laboratory facilities, and in some cases, where there are delays in getting urine samples to a laboratory or in processing the test once samples arrive at a laboratory, results can take up to 1 week to be returned to the GP. Public Health England guidance recommends culture in the following groups to help diagnose a UTI:11

• men

• people aged > 65 years

• babies aged < 3 months

• children aged < 16 years who do not respond to treatment within 24–48 hours

• pregnant women

• those with suspected complicated UTI (pyelonephritis or sepsis)

• those with failed antibiotic treatment or persistent symptoms

• those who have recurrent UTI

• catheterised patients

• those testing dipstick negative for nitrites but positive LE

• those aged < 3 years, with positive dipstick for nitrite and LE

• those with the following risk factors for resistance:

  • abnormalities of genitourinary tract

  • renal impairment

  • care home resident

  • hospitalised for > 7 days in last 6 months

  • recent travel to country with increased resistance

  • previous resistant UTI.

However, there are also limitations associated with culture and exactly how a UTI should be defined. Culture can be negative even when a UTI is present, particularly in the case of antibiotic-resistant bacteria. Laboratory guidelines differ in how culture result should be interpreted to confirm the presence or absence of UTI13 and recommend different diagnostic criteria depending on age, symptoms and how urine was collected. Culture has additional limitations in populations such as frail older people in whom long-term colonisation can make diagnosis particularly difficult, and where culture cannot accurately identify those with a UTI.

Treatment of urinary tract infections

An acute uncomplicated UTI generally resolves within around 9 days without treatment,14 but most patients with UTI will be prescribed antibiotics. Treatment also involves giving advice on self-care such as analgesia and hydration. National Institute for Health and Care Excellence (NICE) guidelines on antimicrobial prescribing for UTI recommend that antibiotics are prescribed immediately in pregnant women, men, and children aged < 16 years. 15 In non-pregnant women, a backup antibiotic (to be taken only if symptoms persist for 48 hours or worsen) or an immediate antibiotic may be prescribed. While dipstick tests and culture are often used to inform the diagnosis and decision on whether to prescribe antibiotics, in some patients antibiotics will be prescribed based on symptoms and examination alone. A recent study of treatment of lower UTI in primary care in England found that the majority of patients (80%) were given empiric antibiotic treatment on the day of diagnosis and that for the majority (83%) no evidence of urine sample collection for laboratory investigation was in their electronic health records. 16 If urine is sent for culture and AST, then the antibiotic choice should be reviewed when the AST results are available. The NICE guidelines contain detailed recommendations of which antibiotic to prescribe as first choice or second choice (if the first choice is not effective or suitable) in different populations. First-choice antibiotics are based on empiric treatment (treatment given based on experience, without exact knowledge of the cause or nature of UTI), usually with nitrofurantoin or trimethoprim. Second-choice antibiotics include pivmecillinam (a penicillin) or fosfomycin in adults and amoxicillin or cefalexin in children. 15 Empiric antibiotics may have side effects, can be less effective than targeted antibiotics (antibiotics targeting the causative pathogen) and increase the risk of antibiotic resistance developing (see Antibiotic prescribing and resistance).

An acute recurrent UTI is managed in the same way as acute UTI. NICE guidelines on antimicrobial prescribing for recurrent UTI recommend giving advice on behavioural and personal hygiene measures and self-care treatment to reduce the risk of future UTI. Post-menopausal women with recurrent UTI may be recommended vaginal oestrogen if other measures are not effective. Antibiotic prophylaxis can be considered if none of the other measures is effective. An alternative to this that is being increasingly used is methenamine hippuirate (Hiprex), a non-antibiotic option. This should not be started until the acute UTI has been treated and resolved. Initial prophylaxis should include single-dose antibiotics; if this is not effective, then daily antibiotic prophylaxis can be trialled. This has associated risks of resistance and possible adverse effects. 15

There are currently no NICE guidelines on the treatment of chronic UTI. Patient organisations suggest that treatment may involve high-dose, extended-course (3–6 months) oral antibiotics or the instillation of antibiotics directly into the bladder. 17 Many patients will also seek relief from alternative therapies for which there is little evidence of effectiveness. 18

Antibiotic prescribing and resistance

Almost 75% of antibiotic prescribing occurs in primary care,19 with UTI contributing to a large proportion of this. Antimicrobial resistance, and in particular antibiotic resistance, is one of the greatest public health challenges faced today. The World Health Organization (WHO) highlights this as one of the current biggest threats to global health, food security and development. 20

The 2017 English Surveillance Programme for Antimicrobial Utilisation and Resistance (ESPAUR) report says that more than 1 million UTI samples were analysed in NHS laboratories across England in 2016, and that resistance was a ‘common’ observation. A recent surveillance study, published in June 2020, found that around 30% of E. coli, the most common cause of UTI, was resistant to trimethoprim, and around 1% was resistant to nitrofurantoin. 7 This is consistent with data from a study that evaluated the Flexicult test, which reported that around 20% of those with a microbiologically confirmed UTI had an infection that was resistant to any first-line antibiotic (nitrofurantoin, trimethoprim or fosfomycin). 7

Population

The population for this scope is people with suspected UTI who:

• would have an initial dipstick test in current practice (population 1)

• would not have an initial dipstick test in current practice (population 2).

People with suspected sepsis are not included in the scope. Subgroups of interest include:

• people with suspected acute UTI

• people with suspected recurrent UTI

• people with suspected chronic UTI

• women aged < 65 years

• women aged > 65 years

• men aged < 65 years

• men aged > 65 years

• adults with indwelling urinary catheters

• babies, children and young people aged < 16 years

• children aged < 3 months

• pregnant women

• people who are frail or have dementia

• people who are pre-, peri- or post-menopausal

• people on prophylactic antibiotics for treatment of UTI

• people of different ethnicities

• people with a higher risk of complicated UTI (e.g. people with neurogenic bladder, diabetes, polycystic kidney disease or people who are immunocompromised)

• people with suspected pyelonephritis.

Technologies of interest

Guidance from Public Health England, ‘Health matters: antimicrobial resistance’,19 published in 2015, highlights the need for rapid diagnostic tools to help GPs quickly (i.e. within minutes) identify the strain of bacterial infection present and the antibiotics to which the infection is resistant or susceptible. This is also highlighted in the 2021/2 ESPAUR. Tests that give a more accurate, rapid diagnosis of UTI than current dipstick testing, with or without identifying the bacteria or providing information on AST, would have the potential to substantially improve diagnosis of UTI in primary care. Such tests may reduce inappropriate antibiotic prescribing in general, as well as improve appropriate targeting of antibiotics prescribed (see Antibiotic prescribing and resistance). 21 They would be particularly useful in those groups in whom dipstick testing is not recommended. Given the high proportion of those presenting with symptoms of UTI who are subsequently found not to have a UTI, novel tests would also have the potential to rule out UTIs, reducing the need for samples to be sent for laboratory testing.

The technologies of interest in this appraisal are novel point-of-care tests (POCTs) that may detect the presence of a UTI and provide information on the strain of bacterial infection present and/or the antibiotics to which the bacteria are susceptible. POCTs are defined as technologies that a healthcare professional can carry out outside a conventional laboratory setting. 22 Table 2 gives an overview of POCTs for diagnosing UTIs within the scope of this appraisal. These tests were identified as part of the appraisal process and were specified in the NICE scope. These are grouped into rapid tests (those that provide results in < 40 minutes) and culture-based tests (which take up to 24 hours to give results). The aim of these tests is to provide more accurate, rapid diagnoses of UTIs and improve antibiotic prescribing. The extent to which these POCTs can improve antibiotic prescribing will depend on how quickly they are able to provide results, how accurate they are, whether they provide additional information on the specific pathogen present in the urine, and whether they provide information on AST.

Potential alternative technologies

A number of technologies are currently in development that will be able to rapidly indicate the presence of bacteria, identify the bacteria present and/or provide information on antimicrobial susceptibility, but these do not have a Conformité Européenne or UK Conformity Assessment (UKCA) mark, and are not expected to obtain this in the next 12 months, and so cannot yet be considered for recommendation by NICE.

Comparator

The comparator for this assessment is the current standard of care: (1) urine dipstick followed by confirmatory culture and AST (if necessary; population 1) or (2) urine culture and AST carried out in the laboratory (population 2). This varies according to population. Further details of the treatment pathway are provided in Current treatment pathway.

Current treatment pathway

The exact treatment pathway varies according to the population (age, sex and whether catheterised). Figure 1 provides a general overview of the treatment pathway. A person presents to their GP with symptoms suggestive of UTI. Depending on the patient population, the person may receive dipstick testing. If this test is positive for nitrite and LE, the person will be diagnosed with UTI; in some populations (e.g. women aged < 65 years) a diagnosis can also be made based on a positive nitrite alone or LE, if also positive for blood. A sample may be sent to the laboratory for susceptibility testing. Decisions about whether to prescribe antibiotics, and which antibiotic to prescribe, are often made before culture results are available, particularly if the person has presented with severe symptoms. This means that antibiotics may need to be changed if culture and AST suggest that the person is taking an antibiotic that is not likely to be effective against their infection, or stopped if no infection is detected on culture.

FIGURE 1.

Outline of treatment pathway. Reproduced with permission from NICE. © NICE [2023] Point-of-care tests for urinary tract infections to improve antimicrobial prescribing: early value assessment(TS insert superscript 23). Available from www.nice.org.uk/guidance/hte7 All rights reserved. Subject to Notice of rights. NICE guidance is prepared for the National Health Service in England. All NICE guidance is subject to regular review and may be updated or withdrawn. NICE accepts no responsibility for the use of its content in this product/publication.

Public Health England has separate pathways for infants/children aged < 16 years, women aged < 65 years, men aged < 65 years, adults who are catheterised and adults aged > 65 years. 11

The treatment pathways differ in terms of whether an initial dipstick test is done, whether a urine sample should be sent to a laboratory for culture testing and when or if to prescribe antibiotics. Table 3 provides an overview of recommendations from the treatment pathways for these different groups:

Place of the technology in the treatment pathway

A POCT for suspected UTI would be used as an initial test to diagnose UTI. If its performance is sufficient, then its place in the treatment pathway, as an initial test to diagnose UTI, will be the same in all populations and prespecified subgroups (see Population).

A POCT’s role in UTI diagnosis will depend on whether it provides additional information on the specific pathogen present in the urine, whether it provides information on AST, and the time taken to produce the result. This will also affect the potential impact of the tests. Table 4 provides an overview of the potential role and impact of a new POCT based on its features.

Inclusion and exclusion criteria

Studies that met the criteria summarised in Table 5 were eligible for inclusion:

Given the tight timelines for conducting an early value assessment (EVA), it was necessary to restrict the review so that it could be undertaken within the available time. The review was therefore restricted to studies reported (published or unpublished) after 2000. We consider it likely that clinical practice, the spectrum of bacteria causing UTI, and the technical performance of tests evaluated will have changed such that studies published before this date are unlikely to provide useful information to inform this appraisal. Animal studies were excluded.

Study identification

Studies were identified using bibliographic and non-bibliographic search methods following guidance in the NICE technology appraisal manual and recent guidance on searching. 28,29

Protocol changes

• We had originally specified that studies would be included for objective 3 only if they evaluated a test that had not been considered as part of objective 1 or 2. However, owing to the very small number of studies that we identified that fulfilled the inclusion criteria for objective 3, we removed this restriction and included studies of any of the technologies of interest.

• In addition to Flexicult Human, we identified a number of studies of ID Flexicult. This test was not specifically in the scope but is included in the review as we consider it possible that ID Flexicult identifies the same information as the control field of Flexicult Human; however, this has not been confirmed by the company.

Search results

The searches of bibliographic databases and trials registries identified 728 unique references after deduplication. After initial screening of titles and abstracts, 38 reports were considered potentially relevant and retrieved for full-paper screening.

In total, 16 studies in 28 reports were included in the review. Two studies in six reports were included for objective 1. Sixteen studies in 20 reports were included for objective 2. Two of these studies were also included in objective 1, and separate reports of DTA substudies provided data for objective 2. Five studies in five reports were included for objective 3. Four of these studies were also included for either objective 1 or objective 2. The final study was a report of a qualitative substudy from one of the studies included for objective 1.

The process of study identification and selection is summarised in Figure 2. Table 6 provides an overview of the number of studies assessing each test for each of our three clinical objectives, stratified by test. There were no data for any of the objectives for the following tests: Astrego PA-100 system, TriVerity, Diaslide, Chromostreak or Uricult Plus. The majority of studies evaluated culture-based tests, which take up to 24 hours to provide results. Uriscreen was the only rapid test to be evaluated in more than one study.

FIGURE 2.

Flow of studies through the review process. (a) After the main searches had been completed, as additional test (UTRiPLEX) was added to the scope of the review; (b) studies and study reports contributed to more than one objective.

Table 7 provides an overview of the populations defined in the scope and whether data were available for these populations. The majority of populations were not specifically considered in the included studies, although they might have been included in studies that enrolled mixed populations.

We excluded studies published before the year 2000, as outlined in Chapter 4. These were excluded after title and abstract screening. Appendix 2 provides a summary of the 62 studies excluded for this reason, showing which test and objective they potentially evaluated. As these were only screened at title and abstract stage, they were not reviewed at full-text screening stage, and so it is likely that not all of these studies would have been included in the review had the date restriction not been applied. All evaluated culture-based tests: the majority (n = 47) evaluated Uricult, two evaluated Uricult Trio, seven evaluated Uriscreen, one evaluated Diaslide, and it was not possible to tell which test was evaluated in the remaining five.

Objective 1: what is the impact on clinical outcomes of using point-of-care tests to diagnose urinary tract infection, with or without additional pathogen identification and antimicrobial sensitivity testing?

Two individually randomised RCTs evaluated the clinical impact of using Flexicult Human (often referred to in studies as the Flexicult SSI urinary kit): the POCT for urinary tract infection in primary care (POETIC) trial8 and a Danish trial. 35 Both trials were conducted in primary care and enrolled women aged > 18 years with symptoms suggestive of uncomplicated UTI. In both studies, all participants also had a urine sample sent for laboratory culture, which meant that a diagnostic accuracy substudy could be performed; the results of these two substudies are included for objective 2 (see Objective 2: what is the accuracy of the point-of-care test for urinary tract infection diagnosis, pathogen identification and antimicrobial sensitivity testing?). 36,37 Both studies were considered at low risk of bias (see Appendix 1). Neither study was funded directly by the test manufacturer, although the manufacturer provided the tests in the Danish study.

The POETIC trial was conducted across four countries: England, the Netherlands, Spain and Wales. It randomised 654 participants: 329 to testing with Flexicult Human and treating based on results (England, n = 117; Wales, n = 109) and 325 to standard care informed by national guidelines (England, n = 117; Wales, n = 110). One male participant was then excluded, resulting in a sample of 653 women. Flexicult plates specific to the antibiotics most commonly used in each of the four regions were developed. GPs were free to determine how best to use the test. Examples of how it could be used included:

• to determine whether, and what antibiotic class, to prescribe the following day

• to prescribe empirically and to aid in a next-day review of the initial prescribing decision

• to provide delayed antibiotics prescription and to guide the use of delayed prescription.

The Danish trial randomised 376 women to two different Flexicult-based strategies: Flexicult Human (which incorporates susceptibility testing) or ID Flexicult (which does not include susceptibility testing). In both arms, GPs were advised to treat based on test results.

The results of the two trials are summarised in Table 8. The POETIC trial reported six different measures of antibiotic use. There was evidence that antibiotic prescribing at the initial consultation had reduced [odds ratio (OR) 0.56, 95% CI 0.35 to 0.88], but this did not impact on overall antibiotic prescription or on antibiotic use that was concordant with culture results (the primary outcome for the trial). Concordant antibiotic use is defined by Butler et al. 8 as ‘consumption of an antibiotic on day 3 (or day 1 or day 2 for Fosfomycin), for which a pathogen considered to be causing a UTI isolated in a laboratory was sensitive in vitro; or no antibiotic use by females who did not have a UTI on laboratory culture’. The Danish trial only reported on ‘appropriate antibiotic prescribing’; there was some evidence that appropriate prescribing was higher in the control arm than in the Flexicult Human arm (OR 1.11, 95% CI 1.03 to 1.99). Appropriate prescribing is defined by Holm et al. 35 as:

1. prescription of a first-line antibiotic to which the infecting pathogen is susceptible if the individual is found to have UTI in the reference

2. prescription of a second-line antibiotic if the individual has UTI but is allergic to the antibiotic or the pathogen is resistant to all first-line antibiotics

3. no antibiotic prescription if the individual is found to not have UTI in the reference.

Both trials also looked at improvement or duration of symptoms and microbiological cure. There was no evidence of any difference between groups for any of these outcomes. The POETIC trial looked at additional outcomes of enablement and resource use (re-consultation or hospital stay within 2 weeks) and found no differences between intervention groups. There were no data for the following outcomes prespecified in our protocol: mortality, health-related quality of life, recurrence, pyelonephritis, sepsis or adverse effects of antibiotics.

Objective 2: what is the accuracy of the point-of-care test for urinary tract infection diagnosis, pathogen identification and antimicrobial sensitivity testing?

Sixteen studies, reported in 20 publications, reported data on test accuracy and were included for this objective. 19,36,37,39–51 Studies were conducted in Denmark (n = 337,39,40), Wales (n = 219,51), Israel (n = 249,50), Hawaii (n = 141), Venezuela (n = 142), Belgium (n = 143), Mexico (n = 144), Philippines (n = 145), South Africa (n = 146), Republic of Korea (n = 147) and Argentina (n = 148), and one study was undertaken in Wales, England, Spain and the Netherlands (n = 136). Most studies were reported in English, with the exception of one in Korean47 and one in Spanish. 44 These were translated using Google Translate (Google Inc., Mountain View, CA, USA); the Spanish translation was checked by a member of the team whose native language is Spanish and was found to be accurate. One study was included from a manufacturer’s submission (submitted in response to a request for information) in the form of a draft manuscript that is academic in confidence. 51 All other studies were published as full reports. Table 9 provides an overview of the included studies’ key characteristics. Full details of each included study are reported in Appendix 3.

The majority of studies evaluated culture-based tests that take up to 24 hours to provide results. Four studies evaluated the Flexicult Human test (referred to in all studies as the Flexicult SSI Urinary Kit),19,36,37,39 three evaluated Uricult Trio,45–47 two evaluated ID Flexicult,37,40 two evaluated Dipstreak49,50 and one evaluated Uricult. 48 The only rapid test to be evaluated in multiple studies was the Uriscreen test, which was evaluated in four studies;41–44 UTRiPLEX43 and Lodestar DX51 were each evaluated in single studies. Two studies evaluated two tests of interest; one evaluated Flexicult Human and ID Flexicult and the other evaluated Uriscreen and UTRiPLEX. 37,43 The manufacturers’ submissions highlighted two ongoing studies that will provide data on the accuracy of the Astrego PA-100 AST test and the Lodestar DX, both rapid tests for UTI. (Confidential information has been removed). The Lodestar submission highlighted TOUCAN, a study evaluating the accuracy of three or four POCTs (details of these not yet available) in up to 800 women who consult their GP with symptoms of UTI. This study was due to complete in October 2023. 52

Four studies were laboratory-based; three tested fresh urine samples19,49,50 and one tested both fresh and stored urine samples. 51 The other 12 studies were conducted in primary or secondary care. Most of these studies performed the POCT in a near-patient setting but two performed the test in the laboratory. 43,48

Four studies recruited pregnant women,41,42,46,48 three studies recruited women with uncomplicated UTI,36,37,40 one study enrolled catheterised ICU patients,44 and three studies recruited children and/or infants aged under 18 years,43 16 years45 and 24 months. 47 Five studies analysed samples from mixed populations: two included people visiting outpatient clinics and hospitalised patients;49,50 one included symptomatic patients consulting the GP;39 and two tested samples submitted to the Public Health Wales microbiology laboratory. 19,51 No further information was provided on these mixed populations. Three studies specifically stated that those with recurrent UTI were excluded;37,44,45 information on whether those with recurrent or chronic UTI were eligible was not reported in the remaining studies.

Seven studies enrolled symptomatic patients36,37,39,40,43,45,47 and four enrolled asymptomatic patients. 41,42,44,48 One study comprised a mix of asymptomatic and symptomatic patients and stratified results accordingly. 46 The four laboratory-based studies did not specify whether urine samples came from symptomatic patients, but as they tested urine samples that had been referred to the laboratory it seems likely the sample comprised a mix of symptomatic and asymptomatic patients. 19,49–51

In the 10 studies that enrolled people and then took urine samples to test for UTI,36,37,40–43,45–48 the number of patients ranged from 117 to 2173 (mean 459 patients). Another study enrolled 57 patients and took multiple samples from each patient, giving a total of 108 samples. 44 In the five studies that tested urine samples rather than enrolling patients,19,39,49–51 the number of samples ranged from 121 to 955 (mean 578 patients).

One study was funded by the test manufacturer. 51 One study was funded by industry (not the test manufacturer) and non-industry. 45 Seven studies did not report funder details39,42,44,46,47,49,50 and all other studies were non-industry funded.

All included studies except for one51 assessed the accuracy of POCTs for detecting the presence of UTI. Three of these studies also reported data on antimicrobial sensitivity,19,36,39 one reported data on pathogenic cause,50 and one reported data on presence of E. coli. 47 Most studies used culture alone as the reference standard, with the exception of one study that used culture and microscopy, and culture, microscopy and spiral plating. 19 The threshold for culture varied between studies but was often reported as ≥ 10³ colony-forming unit (CFU), ≥ 10⁴ CFU or ≥ 10⁵ CFU (see Appendix 3).

Results

Figure 3 shows paired forest plots of estimates of sensitivity and specificity for the detection of presence of UTI together with 95% CIs, stratified by test. Summary estimates for tests evaluated in at least two studies are shown as diamonds on the plot. Results for each test are discussed below. Where evaluated, data are also presented for the detection of the pathogenic cause of the infection and for the accuracy of the test in detecting antimicrobial sensitivity. Table 11 provides a summary of whether data were available on diagnosis of UTI, pathogenic cause and antimicrobial sensitivity for each test. Full accuracy results are presented in Appendix 3.

TABLE 11 - Summary of whether data were available on diagnosis of UTI, pathogenic cause and antimicrobial sensitivity for each test

Test name	Presence of UTI	Pathogenic cause	Antimicrobial sensitivity
Rapid tests
Lodestar DX	✕	✓	✕
Uriscreen	✓	✕	✕
UTRiPLEX	✓	✕	✕
Culture-based tests
Dipstreak	✓	✓	✕
Flexicult Human	✓	✕	✓
ID Flexicult	✓	✕	✕
Uricult trio	✓	✓	✕
Uricult	✓	✕	✕

FIGURE 3.

Paired forest plots of individual study estimates and summary estimates of sensitivity and specificity for the detection of presence of UTI together with 95% CIs, stratified by test. FN, false negative; FP, false positive; TN, true negative; TP, true positive.

Lodestar DX

One study, funded by the test manufacturer, evaluated Lodestar DX. 51 The study was laboratory-based and evaluated the accuracy of Lodestar DX for detecting specific pathogens in fresh urine samples. It did not report the urine sampling method used and was judged as being at unclear risk of bias (see Table 9).

Pathogenic cause

Lodestar DX (n = 1) had good sensitivity 86% (95% CI 74% to 99%) and specificity 88% (95% CI 83% to 94%) for detecting E. coli in urine samples.

Uriscreen

Four studies evaluated Uriscreen. 41–44 One study analysed 156 children aged < 18 years in primary care in Belgium and conducted the POCT in the laboratory. 43 Three other studies conducted the POCT in a near-patient setting and analysed 378 pregnant women from antenatal clinics in Hawaii,41 150 pregnant women from antenatal clinics in Venezuela,42 and 108 samples from 57 catheterised ICU patients in Mexico. 44 Two studies used catheterised urine samples,42,44 one used mid-stream sampling41 and one used mid-stream or adhesive bags. 43 One study was judged as being at low risk of bias,43 two at unclear risk of bias41,42 and one at high risk of bias44 (see Table 9).

Presence of urinary tract infection

All four studies reported data on the accuracy of Uriscreen for detecting UTI, using the presence of foam to indicate the presence of UTI. Estimates of sensitivity ranged from 61% to 89% and specificity ranged from 43% to 89%. Summary sensitivity was 74% (95% CI 59% to 84%) and summary specificity was 64% (95% CI 41% to 82%). There were no clear reasons for the observed heterogeneity.

UTRiPLEX

One study evaluated UTRiPLEX. 43 The study analysed 292 children aged < 18 years in primary care in Belgium, although the test was conducted in the laboratory. The study collected urine samples using mid-stream sampling or adhesive bags, as per clinical practice. It was judged at low risk of bias (see Table 9).

Presence of urinary tract infection

Using the visualisation of ≥ 2 test lines after 6 minutes as the threshold, sensitivity was low (21%) but specificity was high (94%).

Flexicult Human

Four studies evaluated Flexicult Human. 19,36,37,39 This included test accuracy substudies from the two trials included for objective 1. 36,37 These two studies and one additional study were conducted in primary care settings in Denmark, Wales, and Wales, England, Spain and the Netherlands. The two test accuracy substudies from trials were restricted to women (aged > 18 years) with uncomplicated UTI; one of these analysed 183 women,37 and one analysed 289 women. 36 One study analysed 121 samples from a mixed population of symptomatic patients in Denmark,39 and one study was laboratory-based and used 200 fresh urine samples from a mixed population in Wales. 19 Mid-stream urine samples were collected in the two trial substudies. 36,37 The laboratory-based study collected samples using different methods, mid-stream sampling (n = 134) and catheter sampling (n = 7), and for 65 samples the method was unknown. One study did not report how urine samples were collected. 39 Three of the studies were judged to be at unclear risk of bias19,36,39 and one was judged to be at low risk of bias37 (see Table 9).

Presence of urinary tract infection

All studies provided data on the accuracy of the Flexicult Human test for diagnosing UTI. Three used culture alone as the reference standard. 36,37,39 One study used two reference standards: (1) culture and microscopy and (2) culture, microscopy and spiral plating. 19 Another study used three different reference standard definitions to define a UTI: ≥ 10⁴ CFU/ml pure culture of pathogen; ≥ 10⁵ CFU/ml mixed growth with one predominant pathogen; or ≥ 10³ CFU/ml of E. coli or S. saprophyticus (Public Health England/Health Protection Agency), ≥ 10⁵ CFU/ml pure culture of uropathogen or ≥ 10⁵ CFU/ml predominant culture a uropathogen with 3-log difference between the highest and next species (UK laboratory definition) and ≥ 10³ CFU of uropathogen (European definition).

The Flexicult Human thresholds for defining the presence of UTI varied. Two studies used ≥ 10³ CFU/ml,19,37 one used ≥ 10⁴ CFU/ml,39 and one used 10³ CFU/ml for pure culture of a pathogen and ≥ 10³ CFU/ml for predominant growth of a pathogen in mixture with normal flora. 36

Estimates of sensitivity ranged from 74% to 93% and of specificity ranged from 37% to 93%. Estimates were highest in the laboratory-based study of mixed urine samples (93% and 89%). 39 This study used a compound reference standard of culture, microscopy and spiral plating. Estimates were lower when the study used culture and microscopy as the reference standard (87% and 83%) and more similar to the reference standard used in the other studies. The summary estimates of sensitivity and specificity across all three studies in which the Flexicult Human test was conducted in primary care were 79% (95% CI 72% to 85%) and 67% (95% CI 30% to 90%).

Antimicrobial sensitivity

Three studies reported data for antimicrobial sensitivity. 19,36,39 Estimates of sensitivity ranged from 79% to 90% with a summary estimate of 87% (95% CI 83% to 90%). Estimates of specificity ranged from 72% to 94% with a summary estimate of 93% (95% CI 89% to 95%). 19,36,39

ID Flexicult

Two studies evaluated ID Flexicult. 37,40 Both studies conducted the ID Flexicult test in primary care in Denmark and recruited women with uncomplicated UTI and used mid-stream urine samples. One study analysed 158 people37 and the other analysed 117. One of these studies also evaluated Flexicult Human; this was the accuracy study nested within the trial that compared testing and treatment based on Flexicult Human with testing and treatment based on ID Flexicult. One study was judged as being at low risk of bias37 and one had unclear risk of bias40 (see Table 9).

Presence of urinary tract infection

The test had good sensitivity (90% and 88%), but estimates of specificity were lower, at 56% and 80%. Summary sensitivity was 89% (95% CI 84% to 93%) and summary specificity was 70% (95% CI 52% to 84%). The studies used thresholds of 10³ CFU/ml (primary pathogens) and 10⁴ CFU/ml (secondary pathogens) for the POCT.

Dipstreak

Two studies evaluated Dipstreak. 49,50 Both were conducted in Israel and were laboratory-based studies that tested fresh urine samples from mixed populations. One study analysed 795 mid-stream urine samples;50 the other analysed 818 samples (urine sampling method not reported). One study was judged at high risk of bias50 and one was judged at unclear risk of bias49 (see Table 9).

Presence of urinary tract infection

Both studies found Dipstreak to be highly accurate for detecting UTI. Sensitivity was estimated at 96% and 99%; both studies estimated specificity at 99%. Summary sensitivity was 95% (95% CI 94% to 99%) and summary specificity was 99% (95% CI 98% to 99%). One of these studies evaluated two Dipstreak thresholds (10⁴ and 10⁵ CFU/ml)49 and found similar results; the other did not report the Dipstreak threshold. 50

Pathogenic cause of urinary tract infection

Yagupsky et al. 50 reported that Dipstreak correctly identified the pathogenic cause of UTI in 211 out of 270 cases (the other 59 were not identified).

Uricult

One study evaluated Uricult. 48 It analysed mid-stream urine samples from 2173 pregnant women from antenatal clinics in Argentina, and performed the test in the laboratory. It was judged at high risk of bias (see Table 9).

Presence of urinary tract infection

The study reported very high estimates of sensitivity (98%) and specificity (100%) for Uricult for the detection of the presence of UTI, using a threshold of > 10⁵ CFU.

Uricult Trio

Three studies evaluated Uricult Trio. 45–47 Populations varied: one analysed 374 pregnant women in antenatal clinics in South Africa,46 one analysed 151 infants aged < 24 months from outpatient clinics in Republic of Korea47 and one analysed 200 children < 16 years from outpatient clinics in the Philippines. 45 The study in pregnant women stratified results according to whether women were symptomatic (n = 127) or asymptomatic (n = 247). All studies used mid-stream urine samples; one also used urine collection bags in infants and another used catheterisation where clean catch was difficult. One study was judged at high risk of bias,46 one was judged at unclear risk of bias,47 and one was judged at low risk of bias45 (see Table 9).

Presence of urinary tract infection

Estimates of sensitivity ranged from 59% to 78% and of specificity ranged from 49% to 85%. Summary sensitivity was 73% (95% CI 63% to 82%) and summary specificity was 70% (95% CI 52% to 84%).

Pathogenic cause

One study reported that for detecting the presence of E. coli infection, the sensitivity of Uricult Trio was 60% and the specificity was 96%.

Test comparisons

Two studies reported data on two POCTs included in the scope. 37,43 One evaluated both Flexicult Human and ID Flexicult. The other evaluated Uriscreen and UTRiPLEX. Both studies were set in general practice, assessed the accuracy of POCTs for the detection of UTI, and used culture as the reference standard. Both studies were judged to be at low risk of bias when assessed with QUADAS-C.

An accuracy study, nested within a trial, evaluated Flexicult Human and ID Flexicult. 37 The study recruited 341 women in Denmark who had uncomplicated UTI. Patients were randomised to be tested with Flexicult Human or with ID Flexicult. The study reported similar sensitivity and specificity with Flexicult Human (86% and 54%) and ID Flexicult (90% and 56%).

A prospective cross-sectional study evaluated the Uriscreen test and the UTRiPLEX test in children aged < 18 years in Belgium. 43 Three hundred samples were taken systematically and tested. However, far fewer results (156 vs. 292) were available for Uriscreen test than for the UTRiPLEX test because the former was introduced later in the trial, making a comparison of the tests difficult. Sensitivity and specificity were reported as 67% and 69% for Uriscreen and as 21% and 94% for UTRiPLEX.

We are unable to draw comparisons between the tests in other studies due to heterogeneity of population.

Comparison with standard urine dipstick tests

Six studies provided a direct comparison between the POCTs and standard urine dipstick testing for LE or nitrite. 40,41,43,47,48 Four of these defined a positive dipstick test as being positive for either LE or nitrite, one as being positive for both LE and nitrite, and one reported data separately for nitrite and LE dipstick tests. Three studies compared Uriscreen with standard dipstick testing and reported different findings, which may be related to how a positive dipstick test was defined (see Table 12). One study also evaluated UTRiPLEX which was found to be less sensitive but more specific than dipstick testing. Three studies compared culture-based POCTs with standard dipstick testing. All found that the POCTs were more sensitive and more specific than standard dipstick tests.

TABLE 12 - Estimates of sensitivity and specificity for standard dipstick tests and POCTs from studies that evaluated both tests

Study	Population	Test	Sensitivity (95% CI)	Specificity (95% CI)
Rapid tests
Boon (2022)43	Children aged < 18 years	UTRiPLEX	21 (8 to 40)	94 (91 to 97)
		Uriscreen	67 (38 to 88)	69 (60 to 76)
		Dipstick (either nitrite or LE positive considered positive)	32 (16 to 52)	86 (82 to 90)
Macias (2002)44	Catheterised ICU patients	Uriscreen	66.7	74.1
		Dipstick – nitrite only	66.7	45.2
		Dipstick – LE only	78.9	47.2
Millar (2000)41	Pregnant women (screening)	Uriscreen	70 (57 to 84)	45 (40 to 51)
		Dipstick (both nitrite and LE positive considered positive)	81 (69 to 93)	97 (95 to 99.2)
Culture-based tests
Pernille (2019)40	Women – uncomplicated UTI	ID Flexicult	88 (80 to 97)	80 (70 to 90)
		Dipstick (either nitrite or LE positive considered positive)	73 (59 to 84)	75 (63 to 85)
Mignini (2009)48	Pregnant women (screening)	Uricult	98 (96 to 99)	99.6 (99.3 to 99.8)
		Dipstick (either nitrite or LE positive considered positive)	53 (48 to 58)	92 (91 to 93)
Lee (2010)47	Children aged < 24 months	Uricult Trio	59%	85%
		Dipstick (either nitrite or LE positive considered positive)	50%	76.7%

Objective 3: what is the technical performance (other than accuracy) of point-of-care tests for urinary tract infection?

Five publications reported data on technical performance. Three reported data for Flexicult Human8,55 (two of these reported on the POETIC trial8,55) and two reported data for Uricult Trio. 45,46 Of these, one publication was also included for objective 18 and three were included for objective 2. 56 A further study56 appeared relevant to objective 3; however, it was excluded because it was only reported in a trial registry with no data and the trial author did not reply to a request for information. Results are provided in Appendix 3. There were no data for the following outcomes prespecified in our protocol: test failure rate; UTI-associated healthcare resources; health-related quality of life; and clinical outcomes, for example morbidity/mortality.

Conceptual modelling of costs, quality of life and cost-effectiveness

A decision-analytic model was conceptualised to estimate the incremental costs and quality-adjusted life-years (QALYs) for POCTs for UTI in comparison with culture with or without dipstick tests. The model described below is for all possible comparators and populations/subgroups described in Population. Separate models would be required for each population/subgroup.

In Review of evidence on cost-effectiveness, we review the clinical evidence identified in Chapter 5, and evidence identified by pragmatic searches, to narrow the focus on tests and populations where evidence and impact are greatest.

Conceptual model

Our conceptual model is illustrated in Figure 4. Arrows indicate the influence of components on the rest of the model.

FIGURE 4.

Conceptual model for POCTs in UTI. Boxes illustrate important elements to consider. Arrows illustrate influence. AS, antibiotic susceptibility; PHE, Public Health England.

Our conceptualisation was divided into short-term and long-term components. In the short term, the important elements to consider were the symptoms of complicated and uncomplicated UTI, characteristics and consequences of antibiotics, expected efficacy of antibiotics, and any response to ineffectiveness of antibiotics. In the long term, the model links to a generic model for UTI and covers the key complications of sepsis, pyelonephritis and kidney failure. Furthermore, the development or continuation of chronic or recurrent UTI was considered, and it was recognised that this would be particularly common in patients with risk factors such as catheters.

Costs were assumed to be from an NHS and PSS perspective and include all elements from the short-term or long-term components. The tests to compare are those detailed in Table 2, as described in Testing strategies.

Our conceptual model reflects the influence on the costs, health outcomes and model structures of the choice of populations and subgroups. UTIs themselves are categorised in acute, recurrent and chronic. Furthermore, UTIs divide into those that are uncomplicated and complicated at GP presentation, while our model reflects that patients with either uncomplicated or complicated UTI can still suffer complicated UTI at the end of testing and treatment.

Rates of complicated UTI, and the costs and health outcomes of the model, also depend on the subgroup under investigation. We conceptualised these to be broad and include the subgroups identified in Population.

Review of evidence on cost-effectiveness

In this section, we review the relevant evidence on cost-effectiveness identified by the clinical effectiveness review and separate pragmatic literature searches. We use this as a basis for narrowing the tests and subpopulations to only those that are feasible for modelling.

Evaluating costs, quality of life and cost-effectiveness

Using the conceptual model in Conceptual modelling of costs, quality of life and cost-effectiveness and evidence sources summarised in Review of evidence on cost-effectiveness, we developed a structure and identified the necessary evidence to evaluate the costs, quality of life and cost-effectiveness of POCTs for UTI. Our model also assesses the reduction in use of empiric/broad-spectrum antibiotics, and therefore antibiotic use overall, as POCTs with AST can yield targeted treatment and POCTs without AST can indicate when no UTI is present. An NHS and PSS perspective was taken with a lifetime horizon where costs and QALYs were discounted at an annual rate of 3.5%.

Our conceptual model could be extended to a full model with systematic literature reviews and other evidence-gathering exercises; the analyses described below are therefore what should be done if a full timescale for this work were ever to be made available, rather than the truncated timing of an EVA. However, a simple coded model for the tests and subgroups identified in Implications for cost-effectiveness modelling has been implemented in the R programming language. The results are not presented from this model as the evidence identified is too limited for the results to be meaningful, even for the subset of tests and populations evaluated.

Model structure

Our model structure comprises a decision tree over which the costs and consequences of testing for UTI would play out. Decision tree was the only type of model we identified as having been used previously in the UTI literature. 59,60,65,70,72 The key model assumptions are presented in Table 18.

TABLE 18 - Key structural and parameter assumptions of the cost-effectiveness model

Assumptions of the cost-effectiveness model
(i)	The underlying probability of UTI (p_uti) is the same regardless of the test used, but varies according to patient subgroup
(ii)	Test accuracy does not vary by subgroup. A notable exception is that manufacturers’ submissions note that Astrego can be used only in women
(iii)	Probability of antibiotic cure and side effects varies by population
(iv)	The probability of ‘healthy’ on targeted treatment (p_healthy_targ) is the same for each targeted antibiotic
(v)	As some tests can identify pathogenic cause or type of infection, despite not performing AST, the probability of ‘healthy’ on empiric treatment (p_healthy_emp) depends on the type of test used but not on which empiric antibiotic was prescribed
(vi)	Costs and health impacts of pyelonephritis, sepsis and kidney failure can be modelled as once-off costs and disutilities
(vii)	The probability of requiring more than one course of antibiotics is higher if we prescribe empiric antibiotics as there is a higher probability of the first course not targeting the correct bacteria
(viii)	Not modelling long-term impact of unnecessary antibiotic prescription. Instead modelling extent of empiric antibiotic treatment used for suspected UTI
(ix)	AST for patients without UTI will not detect specific antibiotic sensitivity, so these patients can only be falsely given broad-spectrum/empiric antibiotics
(x)	The UTI is eventually cured by targeted or empiric courses of antibiotics, although patients may suffer complications and remain at risk of recurrent/chronic UTI
(xi)	Antibiotic treatment may be given while awaiting culture-based testing results. All patients with a detected UTI will eventually be treated with antibiotics, but some may be treated only after the culture-based testing results have been received
(xii)	Patients started on antibiotics while awaiting culture-based testing will complete their course of antibiotics, even if the culture-based test eventually comes back negative
(xiii)	Patients without UTI may benefit from POCT or culture-based testing as underlying cause of symptoms may have specific antibiotic sensitivity
(xiv)	Patients suspected of UTI but with (true or false) negative test results may be given no further treatment or non-specific empiric/broad-spectrum antibiotics
(xv)	Costs and QALYs of complications do not vary by subgroups

Pyelonephritis, kidney failure and sepsis can be modelled as a once-off cost and quality-of-life decrement. We furthermore did not need to model a later recurrence of UTI. Such a repeat UTI would already be modelled by the decision tree model, as the tree does not distinguish between first and repeat UTI. We therefore did not adopt a long-term model, such as a cohort Markov model, for the long-term outcomes in Figure 4.

Our decision tree is illustrated in Figure 5. This structure is for rapid POCTs that perform AST or identify pathogenic cause (e.g. Astrego PA-100, Lodestar DX), POCTs with culture-based testing (e.g. Flexicult Human, ID Flexicult), and laboratory culture-based testing (with or without dipstick). The model could be extended to include no testing, as is often the strategy for women with uncomplicated UTI and typical symptoms. 78 Patients are assumed to have either a true UTI or no underlying UTI. Our conceptualisation is that the POCTs with AST or pathogenic cause identification would identify a patient as having UTI and a specific antibiotic to which the patient is susceptible, identify a patient as having UTI but not identify a specific antibiotic to which the patient is susceptible, or identify a patient as not having UTI. It is assumed that the POCTs with AST may not always detect the antibiotic to which the UTI is susceptible as they do not detect all possible bacteria. Laboratory culture-based testing can initially assign patients to broad-spectrum/empiric antibiotics before targeted treatment is enabled by the results of the test. Under all strategies, if no UTI is detected, the patient is assumed to be assigned to no further treatment. False positives (i.e. patients without UTI but diagnosed with UTI) are assumed to always receive broad-spectrum/empiric antibiotics.

FIGURE 5.

Decision tree structure for short-term modelling. ‘False positives’ do not incur further costs or consequences. We assume that these branches only incur the cost/disutility of treatment, and that they have no benefit from the POCT.

The probabilities of detecting UTI and, when with AST, detecting antibiotic susceptibility would differ between POCTs, as per the analyses in Results.

Treatment with broad-spectrum/specific antibiotics is modelled to include multiple courses of antibiotics. It also includes switching from one antibiotic to a targeted antibiotic in response to the results of a POCT or laboratory culture-based testing.

The decision tree then assumes that antibiotics would be assigned accordingly (e.g. targeted if specific susceptibility is known, empiric/broad-spectrum if unknown, and not given if known not to be a UTI). Empiric antibiotics are assumed to be potentially followed by targeted treatment if the initial antibiotic is unsuccessful and the results of culture-based tests become available. Treatment can be successful and leave a patient healthy without complications, or unsuccessful with complications from UTI. Our model assumes that all patients are eventually cured of UTI but may suffer complications, in line with Wang and LaSala,60 Sadler et al. 65 and all previous economic models we identified. Patients who recover without complications are ‘healthy’ but at risk of recurrent or chronic UTI. Dipstick with laboratory culture-based testing, or culture-based testing alone, is assumed to initially lead to broad/empiric antibiotic treatment as specific susceptibility is unknown.

The ‘UTI but not specific, empiric/broad treatment’ arms include additional costs and QALY losses from further testing being required to identify and prescribe an effective antibiotic. Recurrence of UTI takes place after the decision tree and may include chronic UTI.

Summary of evaluation of cost-effectiveness

In Conceptual modelling of costs, quality of life and cost-effectiveness, we developed a conceptual model that could be used for a future full economic evaluation of POCTs for UTI and the role of these tests in reducing antibiotic resistance. Our evaluation of the identified evidence (see Review of evidence on cost-effectiveness) and attempt to inform a decision tree implementation of the conceptual model (see Evaluating costs, quality of life and cost-effectiveness) reveal that the evidence is too limited for results to be meaningful. This is despite the restriction to a narrow set of tests and subgroups in Implications for cost-effectiveness modelling. We summarise below the areas where our evidence is most limited. However, these do not constitute formal gaps in the evidence. The clinical effectiveness systematic review of Chapters 4 and 5 was restricted to addressing objectives 1–3 of Chapter 3, which relate to the clinical efficacy, accuracy and technical performance of POCTs. Systematic literature reviews were not conducted on, for example, quality of life with UTI, efficacy of antibiotics for treating UTI, or costs related to complications of UTI. Our pragmatic search of Additional pragmatic searches for cost-effectiveness evidence identified eight previous economic models in UTI, but it was not a systematic search as no formal inclusion criteria was specified, the search terms were potentially insensitive, and screening was performed by only one analyst.

Evidence on test accuracy that could be used in our cost-effectiveness model is summarised in Table 19. The sensitivity and specificity of detecting UTI estimates for Flexicult Human and ID Flexicult were identified by the clinical effectiveness systematic review, but no reliable data were identified on the accuracy of detecting specific antibiotic sensitivity. For Lodestar DX, the sensitivity and specificity were identified for detecting E. coli estimates but not for detecting UTI overall.

There were more substantial evidence limitations in the other model parameters summarised in Table 15. No evidence was identified on the probabilities of sepsis and kidney failure resulting from UTI on targeted antibiotics, empiric antibiotics or no treatment. The probability of pyelonephritis on treatment was identified using NICE guideline NG109, but this guideline did not distinguish between targeted and empiric treatment and related to pregnant women. We would need to assume that this applies to non-pregnant women and the mixed population, which is questionable.

Full costing was possible for single courses of antibiotics to treat UTI (see Table 16). However, no evidence was identified on the probability of needing more than one course of antibiotics. There was also no evidence on the proportion of patients given antibiotics if their initial test did not detect UTI. Cost data on POCTs themselves were limited. The total cost per person of the Flexicult test was estimated in Butler et al. ,8 which included administration and interpretation costs, but similar estimates were not available for Lodestar DX or ID Flexicult. The manufacturer of Lodestar DX provided only the price of the test, plus an estimate of the distribution cost. The price per test of ID Flexicult was not provided by the manufacturer.

Evidence on the costs and QALY impacts of sepsis and kidney failure in UTI was not identified.

These substantial weaknesses in our evidence base limit the utility of our model results for decision-making. Further systematic reviews and expert elicitation would be required to fully inform the model and use it in a full economic evaluation.

Statement of principal findings

There were limited data on the clinical effectiveness of POCTs for UTI. The majority of the included studies evaluated culture-based tests that take up to 24 hours to give a result: Flexicult Human (four studies), Uricult Trio (three studies), Dipstreak (two studies) and ID Flexicult (two studies). The rapid test Uriscreen was evaluated in four studies, with Lodestar DX and UTRiPLEX evaluated in single studies. We did not identify any relevant data on the rapid tests Astrego PA-100 system or TriVerity. The Astrego PA-100 system has the potential to be the most useful of the tests included in the scope for this appraisal as it is able to determine AST within 40 minutes. There were also no data on Chromostreak or Diaslide, but these are linked to the Dipstreak test or Uricult Plus, the latter of which is linked to the Uricult and Uricult Trio tests. These limited clinical effectiveness data also limited the feasibility of an economic evaluation.

Included studies assessed only the following specific populations defined in the scope: women (four studies, not stratified on age), pregnant women (four studies), children (four studies) and those with catheters (one study). There were no data on any of the other prespecified populations of interest. This further limited the scope of economic evaluation to these populations. However, those studies that enrolled a mixed population will most likely have included patients from these populations, but they did not report data separately for the different included populations.

There was very little evidence on the impact of using POCTs for UTI on clinical outcomes. We identified only two trials, both of which evaluated Flexicult Human: one compared with standard care and the other compared with testing with ID Flexicult (which can tell only if UTI is present and does not give information on antibiotic sensitivity). Both trials were judged at low risk of bias. Neither trial reported evidence of a difference in the primary outcome (concordant antibiotic use and appropriate antibiotic prescribing) between intervention groups. Although the study that compared Flexicult Human with standard care found that antibiotic prescribing was reduced at the initial consultation, it did not find a difference between groups for any other outcome related to antibiotic use. Neither study reported a difference between intervention groups for other outcomes: duration of symptoms/infection, patient enablement and resource use. There were no data on mortality or health-related quality of life. The lack of evidence on the impact of antibiotics prescribing also limited the feasibility of economic evaluation.

There were also limited data on the accuracy of POCTs for diagnosing UTI, detecting the pathogenic cause of the infection or detecting antimicrobial sensitivity. Although 16 studies were included for this objective, individual POCTs were each assessed in a maximum of four studies. Where there were data from multiple studies for a single test, studies were heterogeneous in terms of setting, population and where the POCT was performed (near patient or in a laboratory). The limited data suggested that performing the POCT in the laboratory may overestimate accuracy compared with performing the test in a near-patient setting, particularly for culture-based tests. Using stored rather than fresh urine samples was also found to overestimate accuracy in one study that used both types of sample. Some studies were judged to be at risk of bias, and so results should be interpreted with caution. Five were judged at high risk of bias because they had a large proportion of missing data (three studies), included multiple samples from the same patients (one study) and had selected enrolment of patients (one study). Blinding of the person interpreting the reference standard (usually culture) was often not reported and so this may have introduced bias in these studies. There were only two studies that reported direct comparisons between tests. Extreme caution should therefore be applied to the summary estimates and what these mean for the relative accuracy of the tests.

The Lodestar DX tests showed the greatest clinical value potential of the three rapid tests for which data were available. The study of Lodestar DX showed promising results for detecting the presence of E. coli, with 86% sensitivity (95% CI 74% to 99%) and 88% specificity (95% CI 83% to 94%). However, this study was conducted in a laboratory setting using fresh urine samples. It did not provide information on the accuracy of the test for detecting other pathogens using fresh urine samples (the test runs six panels each for different pathogens). Despite the importance of Lodestar DX for economic evaluation, this reliance on laboratory-based testing indicated that economic results could be biased in favour of Lodestar DX. Further data from a clinical setting are required to confirm the accuracy of this test.

Uriscreen was the most commonly evaluated rapid test. This simple test, which involves adding a test reagent powder that enables catalase detection followed by hydrogen peroxide to the urine sample, shaking the collection tube and then observing whether a foam ring has formed, is able to tell whether a UTI is present in a few minutes. However, it does not provide any information on antimicrobial sensitivity or on the pathogenic cause of the infection. The results suggested that both sensitivity and specificity were modest, with summary estimates of 74% (95% CI 59% to 84%) for sensitivity and 70% (95% CI 52% to 84%) for specificity. A single study of UTRiPLEX in children in primary care found very poor sensitivity (21%) but very good specificity (94%). This test, which uses a dipstick to detect inflammatory markers and provides data only on whether a UTI is present, is less likely to be of value given the poor sensitivity suggested by this study.

There were more data on culture-based POCTs, but these tests are less likely to be of value in a primary care setting because of the time they take to provide results (up to 24 hours), although they do provide results more quickly than standard laboratory-based culture. As demonstrated by the POETIC trial, the delay in providing results means that clinicians often start antibiotic treatment during the 24-hour wait for the result (reducing the tests’ value in avoiding unnecessary antibiotics). The limited data suggested that Dipstreak (two studies) and Uricult (one study) were highly accurate tests, but studies were at high or unclear risk of bias. Both studies of Dipstreak were performed in the laboratory and assessed urine samples from mixed populations (outpatient clinics and hospitalised patients), not all of whom would have presented with symptoms of a UTI. Further studies in a primary care setting are therefore needed to confirm these findings. Uricult was assessed by one study at high risk of bias using samples from secondary care and tested in the laboratory and reported very high sensitivity and specificity of 98% and 100%, respectively. However, studies of Uricult Trio, an extension of Uricult that provides additional information on whether Gram-negative, β-glucuronidase-producing organisms (e.g. E. coli) are present, reported more modest accuracy, with summary sensitivity and specificity estimated at 73% (95% CI 63% to 82%) and 70% (95% CI 52% to 84%), respectively. These studies were conducted in near-patient settings and so were likely to produce more reliable estimates for the use of this test in practice. Flexicult Human (four studies) and ID Flexicult (two studies) were found to be modestly accurate in the detection of UTI, with summary sensitivity of 79% (95% CI 72% to 85%) and 89% (95% CI 84% to 93%) and summary specificity of 67% (95% CI 30% to 90%) and 70% (95% CI 52% to 84%), although these data should be interpreted with caution because of the substantial variation across studies. All studies included in the meta-analysis were conducted and interpreted in primary care; one laboratory-based study of Flexicult Human reported higher estimates of sensitivity and specificity (this study was not included in the meta-analysis for this reason). Flexicult Human was shown to have good accuracy for AST, with summary sensitivity of 87% (95% CI 83% to 90%) and summary specificity of 93% (95% CI 89% to 95%). Two studies of culture-based tests provided information on the tests’ accuracy in correctly identifying the pathogenic cause. One study of Dipstreak reported sensitivity of 78% with no bacteria incorrectly identified (i.e. where bacteria were detected, all were correctly identified). A study of Uricult Trio looked only at the detection of the E. coli infection and reported sensitivity of 60% and specificity of 96%.

There were also very few data on the technical performance of the tests. We did not find any studies that reported only data on technical performance; all data for this objective came from five studies included for either objective 1 or objective 2 and relate to culture-based tests. Three studies evaluated Flexicult Human and two evaluated Uricult Trio. Technical performance data suggested that POCTs are easier to use and interpret than laboratory tests and produce results more quickly. The study of Uricult Trio reported fewer lost specimens using this POCT than with laboratory tests that need to be transported. The POETIC study included for objective 1 provided additional data on outcomes in the Flexicult Human arm only. These showed that it was quick to perform the test and obtain and record results and to discuss these with patients, although data on time between taking the sample and obtaining a test were not reported. A qualitative substudy of the POETIC trial suggested that around half of clinicians considered that Flexicult had increased their awareness about antibiotic prescribing and had positively impacted their prescribing habits. However, there were barriers to implementation, including limits on when the test can be used, difficulties in test result interpretation, limited resources, concerns about prolonging patient discomfort while awaiting test results, and the expense of maintaining a regular stock of tests. Only one study reported data on cost; Flexicult Human was reported to cost £48. (Confidential information has been removed). There were no other data on costs, and no data on test failure rate or health-related quality of life.

New POCTs would need to be more accurate and cheaper than standard dipstick tests or provide additional information to inform treatment. Although these tests give results within a few minutes, they are able only to suggest whether or not a UTI is present; they do not provide any information on pathogenic cause or on antimicrobial sensitivity. Six studies provided a direct comparison of POCTs with standard dipstick tests. These showed that culture-based tests were both more sensitive and more specific than standard dipstick tests. Results were more variable for the studies that compared rapid tests with standard dipstick tests.

We developed a conceptual model that could be used for a future full economic evaluation of POCTs for UTI and their role in reducing antibiotic resistance. This model identified pathways for benefit of POCTs, namely that they could reduce the use of empiric antibiotics and, by reducing the incidence of UTI complications and improving cure rates, reduce the healthcare costs and quality-of-life impacts arising from UTI.

The above limitations of the clinical evidence were compounded by limited findings of our further pragmatic searches for economic models. We found only eight previous economic models in UTI management, which provided limited evidence on rates of complications, treatment effects, quality of life and costs. We further explored NICE guidelines on antibiotics for UTI treatment, but these also yielded estimates of efficacy in a small range of subgroups and in broad ‘treated’ or ‘untreated’ groups. This made it impossible to show benefit of targeted versus empiric antibiotic treatment. Given the limitations in the clinical evidence, we restricted our potential implementation of the economic model to a mixed population (Lodestar DX vs. Flexicult Human) and in women with uncomplicated UTI (Lodestar DX vs. Flexicult Human vs. ID Flexicult). Even with this narrow comparison, it was decided that the results of our economic model would not be meaningful, and our findings are limited to the conceptual level.

Strengths and limitations of the assessment

Our systematic review followed published guidance on conducting systematic reviews of DTA studies26 and is reported in accordance with PRISMA 2020 guidance27 and PRISMA-DTA guidance, making our review processes transparent and robust. The protocol was pre-registered in the PROSPERO database (CRD42022383889). The only changes that we made to the protocol were to broaden our inclusion criteria such that objective 3 was not restricted to studies of tests that had not been evaluated for objective 1 or 2 and to include studies of ID Flexicult in addition to those of Flexicult Human.

We conducted extensive literature searches designed to maximise the retrieval of relevant studies and did not apply any language or date restrictions to these searches. However, the review was restricted to studies published after the year 2000 so that it could be completed within the tight timescales of an EVA. We documented those studies considered potentially eligible but excluded due to their publication date; 62 studies were excluded for this reason. All evaluated culture-based POCTs; the majority evaluated Uricult/Uricult Trio, with a small number evaluating Uriscreen and Diaslide. We did not exclude studies based on the language of the report and included two non-English studies: one in Spanish and one in Korean. We used Google Translate to enable us to include these studies. The Spanish translation was checked by a member of the team whose first language is Spanish and this was found to be accurate; we were unable to verify the accuracy of the Korean translation. We conducted a formal assessment of the risk of bias of included studies using the RoB 2 tool for RCTs30 and the QUADAS-2 tool for DTA studies32 and its extension QUADAS-C79 to assess the two comparative accuracy studies included in the review. We modified QUADAS-2 to exclude the assessment of applicability. This is because our review question was broad with multiple populations and tests of interest. Instead of a formal assessment of applicability, we extracted information that could result in variation across studies and considered this in our synthesis of results. These data included population, setting, location of test performance, POCT and culture threshold, and reference standard. However, due to the small number of eligible studies that evaluated each individual test, it was not possible to draw strong conclusions regarding the impact of these features on test performance. Our synthesis included a meta-analysis where more than one study evaluated the same test. Included studies only were conducted in women, pregnant women, children and those with catheters. We calculated summary estimates of sensitivity and specificity across patient subgroups. This assumes that accuracy would not vary by subgroup, but this may not be the case; there were insufficient data to investigate whether accuracy varied across different populations. Estimates from these should be interpreted with caution due to clinical and statistical heterogeneity across studies. Estimates should not be applied to populations beyond those from which the estimates were drawn.

We did not include a formal assessment of publication bias due to the small number of included studies and the difficulties in assessing publication bias for DTA studies for which there is no clear threshold for ‘significance’. 26

We prespecified clearly defined, objective inclusion criteria. These specified that studies should be conducted in a population with suspected UTI. We interpreted this broadly such that studies in which pregnant women were screened for UTI and those in which mixed samples sent to the laboratory for testing were also included. However, we excluded studies that only assessed the technical validity of the tests, where control samples with known pathogens were tested using the POCT. These studies do not reflect how the test will perform in practice; they are an initial stage evaluation to determine whether the test can, in principle, be used to process patient urine. Such studies are likely to overestimate test performance. The submission from Astrego highlighted two technical performance studies of the Astrego PA-100 system, a test for which we did not identify any studies that fulfilled the inclusion criteria. 80,81 These studies showed that the test can, in principle, detect the presence of UTI and correctly identify antimicrobial sensitivity. This is potentially a very promising test as it can provide information on the presence of UTI and on antibiotic resistance within 10–15 minutes, but further data on the accuracy and clinical impact are needed. (Confidential information has been removed).

A potential limitation of the evidence base is exactly how UTI should be defined. The gold-standard test for UTI is culture, and the concept of significant bacteriuria, usually defined as > 10⁵ CFU/ml, was established in the 1960s by Kass from a study of 415 women attending a prenatal clinic who were screened for bacteriuria, of whom only 35 were culture positive. 82 However, there are limitations to culture as a reference standard. Culture can be negative even when a UTI is present, particularly in the case of antibiotic-resistant bacteria. Laboratory guidelines differ in how culture results should be interpreted to confirm the presence or absence of UTI13 and recommend different diagnostic criteria depending on age, symptoms and how urine was collected. All but one of the studies included in our review used culture alone as the reference standard, with thresholds to define the presence of UTI ranging from ≥ 10³ CFU to ≥ 10⁵ CFU. In some studies, this was based only on the presence of a single organism, while others had different thresholds for mixed growth, for example ‘Single organism 10⁴ CFU or two organisms when colony count of one > 10⁵ CFU’. One study used a compound reference standard consisting of culture, microscopy and spiral plating, which is likely to have given a more accurate determination of whether or not a UTI was present.

A further problem is potential contamination of urine samples or asymptomatic bacteriuria. 83 Culture does not distinguish between pathogenic and non-pathogenic bacteria, so bacteria growing on culture will not necessarily indicate the presence of a UTI, particularly in asymptomatic patients. The accuracy of all tests for UTI will depend on how the urine sample was collected and the potential risk of contamination. Where method of collection was reported, most studies included in this review used mid-stream urine samples or urine collection bags for children. Although such methods of urine collection do have a greater risk of contamination than other methods such as suprapubic aspiration or catheterisation, these methods reflect how urine is likely to be collected in practice and so were appropriate.

The available accuracy evidence drove our selection of the tests and subgroups to include in the economic evaluation. We took a pragmatic approach to prioritising the modelling of tests with the greatest potential for impact. This led us to focus on modelling rapid tests over culture-based tests and to prioritise tests that performed AST over those that only identified pathogenic cause, and both over those that only detected UTI. The only rapid tests with accuracy data were Lodestar DX, Uriscreen and UTRiPLEX; none of these can perform AST, and only Lodestar DX can detect pathogenic cause. The only culture-based tests with accuracy data that performed AST were Flexicult Human and ID Flexicult. We therefore aimed to model only Lodestar DX, Flexicult Human and ID Flexicult.

The limited evidence on accuracy further drove our selection of populations to include in the economic evaluation. Lodestar was only evaluated in a mixed population, while Flexicult Human and ID Flexicult were only evaluated in mixed and/or women with uncomplicated UTI. We therefore restricted modelling to a mixed population (Lodestar DX vs. Flexicult Human) and in women with uncomplicated UTI (Lodestar DX vs. Flexicult Human vs. ID Flexicult).

Sensitivity and specificity of detecting UTI estimates for Flexicult Human and ID Flexicult were identified by the clinical effectiveness systematic review, but no reliable data were identified on the accuracy of detecting specific antibiotic sensitivity. Sensitivity and specificity of Lodestar DX were identified for detecting E. coli estimates but not for detecting UTI overall.

We utilised cost-effectiveness evidence identified by the clinical systematic review, but this was limited to only two studies. We took a pragmatic approach to searching for additional cost-effectiveness evidence, with searches of Ovid MEDLINE, EMBASE and EconLit. We did not restrict to models and, by not specifying a PICOS, were able to flexibly include any study with potentially useful evidence. However, we found only eight studies, none of which modelled POCTs and none of which provided all of the evidence needed to inform our economic evaluation.

We used a broad conceptual model to reflect the influence of the choice of populations and subgroups on the costs, health outcomes and model structures. This covered all costs, outcomes, tests and populations specified in the scope. We furthermore designed a decision tree to reflect the short-term aspects of our conceptual model. Despite our prioritisation of tests and subgroups, broad approach to modelling and pragmatic approach to searching for evidence, we found that the evidence informing our economic model was too weak for results to be meaningful.

Uncertainties

Given the limited data available for this appraisal, a number of uncertainties remain. These include the accuracy of rapid tests for diagnosing UTI in primary care settings, the comparative accuracy of tests, whether accuracy varies according to population, how test interpretation varies between the laboratory and near-patient settings, the impact of recurrent or chronic UTI on test performance, and economic modelling.

We identified only a small body of evidence, with evidence particularly lacking for the more novel rapid POCT. There were insufficient data to investigate whether test performance differed across the different populations defined in the scope, or to consider how having recurrent or chronic UTI could impact on test performance.

Although the POCTs were designed to be carried out in a near-patient setting, nine studies performed a POCT in a laboratory setting. Six of these used samples sent to the laboratory, and the others collected the samples in antenatal clinics or primary care and then sent the samples to the laboratory for testing. Studies in which tests were performed in laboratories tended to overestimate accuracy compared with those carried out in near-patient settings. The only primary care settings in which studies were conducted were GP practices and antenatal clinics. There were no data on pharmacy settings. Further data are needed on how these tests perform in a near-patient setting.

The limitations of the clinical effectiveness evidence also limited the scope of the economic evaluation. Despite prioritising those tests and subgroups for which evidence and potential for impact were greatest, it was still decided that results of the economic model would not be meaningful for decision-making.

Although limited sensitivity and specificity data were identified for our prioritised tests (Lodestar DX, Flexicult Human and ID Flexicult), few reliable data were identified on the probability of identifying antibiotic susceptibility or pathogenic cause to direct targeted treatment. Sensitivity and specificity of Lodestar DX were identified for detecting E. coli estimates but not for detecting UTI overall.

There were more substantial evidence limitations of the other model parameters summarised in Table 15. No evidence was identified on probabilities of sepsis and kidney failure resulting from UTI on targeted antibiotics, empiric antibiotics or no treatment. The probability of pyelonephritis on treatment was identified using NICE guideline NG109, but the guideline did not distinguish between targeted and empiric treatment and related to pregnant women. No evidence was identified on the probability of needing more than one course of antibiotics. There was no evidence on the proportion of patients given antibiotics if their initial test did not detect UTI. We also need better ways of determining the long-term impact at both individual and societal levels of using point of care diagnostics to better target antibiotics.

Cost data on POCTs themselves were limited. The total cost per person of the Flexicult test was estimated in Butler et al. ,8 which included administration and interpretation costs, but similar estimates were not available for Lodestar DX or ID Flexicult. The manufacturer of Lodestar DX provided only the price of the test, plus an estimate of the distribution cost. The price per test of ID Flexicult was not provided by the manufacturer. Evidence on costs and QALY impacts of sepsis and kidney failure in UTI was not identified.

In addition to this evidence weakness, the structure of the model was subject to limitations. All assumptions in Table 18 could be questioned. In particular, the assumption that accuracy does not vary by subgroup could be challenged by Figure 3, for example pregnant women versus catheterised people for Uriscreen, specificity of Flexicult Human in mixed population versus women, or pregnant women versus children for Uricult Trio. As further evidence that test accuracy can vary by population, the manufacturers’ submissions note that Astrego can only be used in women.

Our choice of a decision tree (see Figure 5) to represent the conceptual model (see Figure 4) is a substantial structural uncertainty. All economic models in UTI that we identified used decision trees, but these were largely restricted to modelling pyelonephritis as a complication of UTI. Kidney failure, sepsis, recurrent UTI and chronic UTI are all potential long-term consequences of poor management of UTI. A Markov model, as illustrated in Figure 6, could be used to model the long-term consequences of complication branches of our decision tree.

FIGURE 6.

Possible long-term Markov model from decision tree. Hospitalisation is a factor for each of the complication states. Death is possible from any state.

Equality, diversity and inclusion

Our research was based on existing literature and so we had no control over the participants enrolled. We were broad in our inclusion criteria such that studies from any country and in any language of publication were eligible. We had intended to investigate how the accuracy of included tests varied across different populations, but there were insufficient data to allow us to do this.

Our team included researchers with a broad range of experience and expertise. The lead authors are junior researchers within Bristol TAG, who were given the opportunity to lead on the writing of this report to help develop their research skills and portfolio. They were supported by the two senior authors, who provided advice and mentorship to the junior researchers leading on the reviews and health economic modelling. The team included those with expertise in systematic reviews, health economics and medical statistics.

Patient and public involvement

We involved two patient representatives in this project who have lived experience of UTI. They attended meetings with the clinical effectiveness team (one at the beginning of the project and one closer to the end of the project), gave feedback on the plain language summary for the protocol and main report and wrote the following section about the difference that POCTs may make to patients with UTI. The involvement of patients had a positive impact on this project, particularly in highlighting the importance of not having to wait for test results. Discussions around this topic led us to stratify our results section into rapid tests and culture-based tests.

Impact on patients

The first and most important impact for patients is that a test that can be given immediately in the GP’s surgery, particularly if it suggests the appropriate antibiotic for treatment, can relieve symptoms much more quickly and effectively with less impact on antimicrobial resistance.

Urinary tract infections can be extremely painful and uncomfortable. They make leaving the house and being away from a toilet very difficult and they therefore impact on the ability of people to manage their everyday lives. For this reason, anything that can make treatment quicker and more effective is immensely valuable to patients. It also means that patients are less likely to attend accident and emergency departments, relieving pressure on those services and reducing a patient’s likelihood of coming into contact with other communicable diseases or spending long and painful hours waiting for treatment.

The benefit of being able to be diagnosed in a person’s local GP surgery in one visit would have a major impact on those with busy lives and would make life much better for those who find it difficult to get to the surgery. It would also reduce the number of appointments being booked, freeing up appointments for others to use.

In fact, these tests could be carried out at community pharmacies. It has been shown that during the COVID-19 pandemic more people sought advice and accessed pharmacies and trusted the advice provided. The fact that pharmacies are in the community and accessible, with longer opening hours, including at weekends, benefits patients.

If those GP-based tests can also suggest the most appropriate antibiotic or show immediately that the patient is unlikely to have a UTI, this will lead to less use of antibiotics overall, which must help to reduce antimicrobial resistance. This is positive for the future treatment of infections. This is also likely to cost less in antibiotic prescribing, which would be positive for the NHS.

Implications for practice

There is a clear need for a rapid test that can accurately diagnose UTI within a short time period in primary care settings, including GP surgeries and pharmacies. Ideally, such tests would also provide information on antimicrobial sensitivity, which would allow appropriate targeted antibiotic use, meaning that patients would be treated appropriately more quickly and the total burden of antibiotic prescriptions would be reduced. The only test within scope that meets these criteria is the Astrego PA-100 system. However, there are currently no data available on this test. Tests such as Lodestar DX that are able to rapidly identify pathogenic cause would also be of value as while these would provide direct information about which antibiotic the causative organism is susceptible to, they would help guide treatment as different pathogens are known to respond differently to certain antibiotics.

Flexicult Human, like the Astrego PA-100 system, is able to provide information on whether a patient has a UTI and on antimicrobial sensitivity. However, this test takes up to 24 hours to produce a result, and this is likely to be longer for samples that are taken on a Friday as the result would not be available until Monday. This makes it more difficult to implement in a primary care setting. Evidence from two trials suggested that using Flexicult had little impact on antibiotic prescribing or on other outcomes such as symptom duration or resource use. Accuracy of the test was found to be modest. Other culture-based tests had similar accuracy when conducted in near-patient settings.

Our conceptual model for economic evaluation found potential pathways to benefit of POCTs. They could reduce costs, improve quality of life and reduce antibiotic resistance by better targeting antibiotic use and reducing complications from UTI. However, we did not have sufficient evidence on test accuracy, targeted versus empiric antibiotic efficacy, or costs and quality-of-life impacts of UTI complications for our model to perform a meaningful comparison. A full evaluation is needed before any recommendation can be made regarding the cost-effectiveness of POCTs or their ability to impact antibiotic resistance.

Strong evidence that POCTs (1) reduce unnecessary antibiotic use, (2) improve symptoms or (3) are cost-effective is needed before such tests are introduced into the NHS.

Suggested research priorities

Given the paucity of data on POCTs for diagnosing UTI, further studies are needed to determine whether POCTs for people with suspected UTI have the potential to be clinically effective and cost-effective to the NHS. Future studies should prioritise those tests with the greatest potential to improve patient outcomes and reduce inappropriate antibiotic prescribing. The most promising tests, of those in scope, are the rapid POCT Astrego PA-100 system (which provides information on antibiotic susceptibility) and Lodestar DX (which provides information on pathogenic cause). Studies should also investigate the feasibility of introducing testing within a pharmacy setting, which could take pressure off GP practices and ensure quicker access to appropriate treatments in the current climate where it can be difficult to access GP appointments. Future research should also encourage the continued development of new diagnostic technologies.

The ideal study would use a similar design to the POETIC study: it would be conducted in primary care (GP surgery and/or pharmacy) and would randomise GP practices/pharmacies to either ‘test and treat appropriately’ or standard practice. Outcomes such as those from a recently published core outcome set84 (to improve standardisation and comparability between trials), for example symptom duration and AEs, would then be compared between intervention arms. Ideally, studies would also include a nested diagnostic accuracy study to provide additional information on the accuracy of the test. Either studies should enrol patients across multiple patient groups of interest (e.g. men, women, pregnant women, children), with results stratified according to patient subgroup, or separate studies should be carried out to determine whether results differ according to subgroups. Before such studies are conducted, it may be appropriate to conduct efficacy studies to demonstrate that the technology can work under ideal conditions, in which patient recovery is closely monitored, which cannot be done in a pragmatic RCT as described above.

In addition to further studies on the clinical effectiveness of POCTs, further research on potential cost-effectiveness and impact on antibiotic resistance is needed. This research could build on our conceptual economic model using systematic literature reviews to identify evidence. Such reviews should focus on the efficacy of empiric versus targeted antibiotic treatment of UTI, efficacy in preventing UTI complications, and both the cost and quality-of-life impacts of these complications.

Contributions of authors

Eve Tomlinson (https://orcid.org/0000-0002-0969-602X) (Research Associate, Evidence Synthesis, Systematic Reviews) contributed to and checked data extraction and quality assessment for objective 2, extracted data for objective 3, drafted the results sections for objectives 2 and 3 and assisted in drafting clinical effectiveness sections of the discussion.

Mary Ward (https://orcid.org/0000-0002-2321-2546) (Senior Research Associate, Health Economic Modelling, Health Economics) helped design the cost-effectiveness model, gathered input parameters, implemented the model in R and ran analyses.

Chris Cooper (https://orcid.org/0000-0003-0864-5607) (Research Fellow, Health Technology Assessment and Information Science, Systematic Reviews) designed and undertook the literature searches, contributed to the reporting of the systematic review, reviewed the company submissions and worked on the review of cost-effectiveness.

Rachel James (https://orcid.org/0000-0001-9299-9461) (Research Associate, Evidence Synthesis, Systematic Reviews) screened and extracted studies and reviewed the company submissions.

Christina Stokes (Patient Representative) provided a patient perspective on the project, edited the plain language summary and wrote the section of the report on impact on patients.

Samina Begum (Patient Representative) provided a patient perspective on the project, edited the plain language summary and wrote the section of the report on impact on patients.

Jessica Watson (https://orcid.org/0000-0002-8177-6438) (NIHR Clinical Lecturer, General Practice, Primary Care and Test Evaluation) provided clinical advice for the project.

Alastair D Hay (https://orcid.org/0000-0003-3012-375X) (Professor, Primary Care and Test Evaluation) provided clinical advice for the project.

Hayley E Jones (https://orcid.org/0000-0002-4265-2854) (Associate Professor, Medical Statistics and Test Evaluation) provided statistical advice and carried out the meta-analyses of diagnostic accuracy data.

Howard Thom (https://orcid.org/0000-0001-8576-5552) (Senior Lecturer, Health Economics) drafted the cost-effectiveness section of the report, designed the cost-effectiveness model and led the cost-effectiveness assessment.

Penny Whiting (https://orcid.org/0000-0003-1138-5682) (Professor, Clinical Epidemiology, Systematic Reviews and Test Evaluation) drafted the clinical effectiveness sections of the protocol and led the clinical systematic review.

All authors were involved in commenting on the final report. Penny Whiting is the senior author and guarantor.

Acknowledgements

We would like to thank Gus Hamilton for microbiological advice relating to this project, and Charlene Wisdom-Trew, Bristol TAG, for providing administrative support.

Data-sharing statement

All data extracted for the systematic review and the results of the risk of bias assessments are provided in full in the appendices to this report. The R code for the cost-effectiveness model is provided as a publicly accessible repository on GitHub.

Ethics statement

The research included in this report is secondary research and as such did not require ethical approval.

Information governance statement

There were no personal data involved in the production of this report.

Disclosure of interests

Full disclosure of interests: Completed ICMJE forms for all authors, including all related interests, are available in the toolkit on the NIHR Journals Library report publication page at https://doi.org/10.3310/PTMV8524.

Primary conflicts of interest: Alastair D Hay has been a member of the National Institute for Health and Care Research Efficacy and Mechanism Evaluation Funding Committee since 2021 and National Institute for Health and Care Excellence common infections guidelines since 2020.

Disclaimers

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

Pre-2000 studies

The table below provides an overview of the studies identified as potentially relevant during title and abstract screening that were excluded because they were published before the year 2000.

Studies excluded after full-text assessment

Studies included in manufacturer’s submission that did not meet inclusion criteria

Objective 1

Objective 2