Trends in urine sampling rates of general practice patients with suspected lower urinary tract infections in England, 2015-2022: a population-based study

Data source and management

Data used for this work were obtained from the IQVIA Medical Research Database (IMRD) from January 2015 to December 2022, inclusive. The IMRD captures longitudinal non-identified patient electronic healthcare records collected from patients of all ages (>5.4 million) from a sample of GP practices using IQVIA-compatible practice management systems (this includes data from practices using EMIS Health and Vision) in England. The IMRD data have been shown to be representative of the general population in England in terms of geography, age and sex.17 18 IMRD is non-identified data, with reference IDs able to be linked across different data tables. SNOMED and Read codes were matched to IMRD-specific clinical codes to identify lower UTI diagnoses in order to create the study cohort with these codes also used to identify urine culture and dipstick tests. All drug code IDs for prescriptions with the generic drug name being one of the antibiotics included in the 2018 lower UTI NICE guidelines (nitrofurantoin, trimethoprim, cefalexin, amoxicillin, fosfomycin and pivmecillinam) were pulled from the prescribing file. These antibiotic prescriptions were linked to UTI diagnosis codes and termed UTI antibiotics, though some are not exclusively used for the treatment of UTIs in practice. Additional descriptive analyses looked at all antibiotic prescriptions on the same day as the initial UTI consultation. For this analysis, all drug code IDs for those antibiotics with an Anatomical Therapeutic Chemical Classification System code of J01 (antibacterials for systemic use) and select additional alimentary tract and nitroimidazole derivatives for protozoal disease were included. Any data for the same patient that occurred on the same day and with the same diagnosis or drug code ID were considered deduplicates.

UTI episodes were created to account for multiple UTI consultations within the same episode of infection. UTI episodes were defined as any UTI diagnosis code occurring within 14 days of each other based on treatment length recommendations (online supplemental figure S1). For each UTI diagnosis code, any urine culture or dipstick that occurred within 3 days was linked. Once testing codes were linked, UTI diagnoses occurring within 14 days of each other were merged to create the UTI episodes. Any codes identified via clinical review for a positive urine culture or dipstick found within 14 days of a UTI diagnosis code were identified to calculate positivity rates.

For the UTI antibiotics, once episodes with a UTI antibiotic prescription were identified, same-day prescriptions were pulled for those episodes where the prescription date was the same as the first UTI consultation.

Following the linkage of UTI episodes, demographic data were linked to each episode. Variables included year of birth (from which age at the time of episode was calculated), sex, Indices of Multiple Deprivation (IMD) decile and predefined region (London, South, North and East Midlands).

Statistical analysis

Once UTI episodes were fully linked, frequencies were run for a range of variables, including the number of episodes per month and the number of episodes per month with a urine culture, dipstick, UTI antibiotic and combinations of these variables. Trends were reviewed monthly, with annual frequencies calculated and tabled for comparison. These frequencies were further broken down by age group (<16, 16–44, 45–54, 55–64, 65–74, 75–84 and >85 years) and sex. To better contextualise these numbers, all frequencies were then calculated as percentages of all episodes, and for those broken down into age and sex, by the percentage of the population with an episode. Population counts were generated from the patient demography file provided in the IMRD data, with the number of patients registered with a GP calculated per year. All patient episodes were assigned to one of the four groupings in the NICE guidelines based on age, sex and pregnancy status. Based on the available data, pregnancy was identified using UTI pregnancy IMRD clinical codes. This is likely an underestimation of the prevalence of pregnancy in the cohort, and because of this, episodes including these pregnancy codes were removed from the final modelling. NICE groups were also used to identify episodes that received a first-line or second-line same-day UTI antibiotic or a combination of both.

In addition to the demographic characteristics, variables were included to characterise healthcare factors associated with UTI episodes. The type of episode was pulled from the consultation file and categorised as face to face, remote or other for those that did not fit either. GP practices were split into two sets of quartiles. The first based on the number of UTI episodes captured across the study period, with 1 being the lowest number and 4 being the highest. The second measure split practices into quartiles based on the number of patients registered with the practice. The number of patients registered at each practice was calculated using the registration numbers as of 2019, as it was the midpoint of the study period.

Following the trend analyses, univariate and multivariate logistic regression were run for variables selected a priori based on characteristics of interest available in the data. Variables included in the models were age group, sex, IMD decile, if the episode had a same-day UTI antibiotic, choice of same-day antibiotics (none, first line only, second line only, another UTI-specific antibiotic (as defined by the NICE guidelines), an alternative antibiotic or any combination of the above), the episode type, the year of the episode, the region of the practice, the number of episodes by practice quartile, the size of the practice by quartile. Following univariate analyses, multivariate models were run, with the outcome of interest being any urine test (culture and/or dipstick), along with additional models looking at urine tests for those with a same-day UTI antibiotic and without a same-day UTI antibiotic.

All data analyses were completed by using RStudio V.4.1.2 with data pulled from Microsoft SQL Server Management Studio V.18.

Patient and public involvement

None.

Principal findings

This large patient-level study demonstrates the changes seen in overall testing and prescribing for patients with a suspected lower UTI in GP, with marked changes seen following the start of the COVID-19 pandemic. The results presented in this work provide updated data around testing and prescribing for lower UTIs in GP in England, including data prior to and following the introduction of the 2018 NICE lower UTI guidelines and the COVID-19 pandemic. This work is also able to build directly on previous work looking at testing and antibiotic prescribing.

Prior to the start of the pandemic, the trends in testing for UTI patients were slowly but continuously increasing. This increase in testing, with the highest levels seen in 2019, coincides with NICE guidelines published in late 2018, with a mid-stream urine sample recommended for susceptibility testing in pregnant women and men, with urine culture for susceptibility recommended for under 16s.9 19 This period also coincides with the publication of the 2019–2024 UK NAP on AMR in 2019, which included the aim of reporting on the percentage of antibiotic prescriptions supported by a diagnostic test or decision-support tool by 2024.16 However, following the start of the pandemic in 2020, testing became less likely due to changes in how care was able to be delivered due to COVID-19 restrictions and national lockdowns. This is seen in the change in the percentages of remote and face-to-face consultations in 2020, with remote consultations making up 25.6% of UTI episodes in 2019, up to 64.6% of UTI episodes in 2020 and 68.7% in 2021. The increase in remote UTI consultants was largest in those under 65, who in 2019 made up 55.6% of remote consultations, increasing to 61.0% in 2020. As the rate of remote consultations increased, so did rates of coding for ‘suspected UTI’, which made up 76.0% of the first diagnosis code captured for UTI episodes in 2019, up to 89.3% in 2020. Overall, urine testing in 2021 and 2022, though higher than the percentage seen in 2020, did not recover to prepandemic levels seen in 2018 and 2019. The rates of testing were found to have rebounded earlier for males rather than females, likely due to the emphasis on testing for men versus only for a subpopulation of females. The change in testing during the pandemic, with little change seen in antibiotic prescribing, likely means that empiric prescribing without confirmation of the infectious agent was common, which might impact downstream rates of AMR.

Same-day antibiotic prescribing was consistent across the study period, even when testing rates decreased. When testing in UTI episodes decreased by 38.0% from 2019 to 2020, same-day antibiotic prescribing increased by 3.1%. This rise in same-day prescribing in 2020 was likely due to an abundance of caution early in the pandemic when patients could not be seen face to face. This aligns with other work, which found that though overall community antibiotic prescribing declined early in the pandemic when the decrease in GP encounters is taken into account, there is a noted increase in the proportion of encounters prescribed an antibiotic.20

Nitrofurantoin and trimethoprim were the most commonly prescribed UTI antibiotics, particularly for non-pregnant individuals over 16, which aligns with the NICE guidelines.9 When looking at same-day antibiotic prescribing as a factor influencing the odds of testing, a same-day prescription was associated with lower odds of testing. This might have been due to non-pregnant, non-catheterised women with two or more UTI symptoms being prescribed an antibiotic without requiring urinary testing, as per testing guidelines.8 9 When univariate regression is used to examine odds of testing by NICE group, women ≥16 years had the lowest odds of testing compared with men ≥16 years (OR 1.18) and children (OR 1.56). This population and the treatment guidance have likely driven the results seen in the multivariate regression analysis.9 Though from 2015 to 2019, females had a higher percentage of episodes with a same-day antibiotic and test than men, the larger number of episodes for females is a contributing factor.

Due to the observed effects of the COVID-19 pandemic, it is difficult to fully determine the impact of the 2018 NICE guidelines, as these were only published at the end of 2018.7 Prior to these guidelines, there were no national UK guidelines for the treatment and management of lower UTIs in adults, with NICE publishing a UTI Quality Standard in 2015.21 The Quality Statements as part of this publication with relevance to the NICE guidelines include adults ≥65 years recommended to have a full clinical assessment to diagnose UTI, men with symptoms of an upper UTI referred for urological investigation, urine culture recommended for adults with a UTI not responding to antibiotic treatment, and antibiotics not recommended to treat asymptomatic bacteriuria in non-pregnant women.21 22

The impact of demographic factors, such as age and sex, on urine testing, aligns with previous literature and much of the NICE guidance.1 7 9 23 24 Age was found to increase the odds of testing compared with those 16–44, except in those ≥85. This aligns with the emphasis placed on both testing children with UTI symptoms found in the guidance, as well as the decreased emphasis on testing in much older adults due to frequent asymptomatic bacteriuria1 9 24 with urine culture testing often only performed in older adults with recurrent UTI or when resistance is suspected.8 25–29 The 2015 NICE Quality Standards also emphasised that the diagnosis of UTI is particularly difficult in older people due to asymptomatic bacteriuria. The prevalence of bacteriuria may be so high that urine culture ceases to be a reliable diagnostic test, with older people, particularly in long-term care, often receiving antibiotic treatment for asymptomatic bacteriuria despite evidence of adverse effects with little to no compensating clinical benefit.22 Based on the NICE guidance, it would also be expected that urine testing would be completed for the majority of men with suspected UTIs.7 While males in this study were shown to have higher rates of testing in GP (53.8%), testing in females was found to be only 6.5% less. Though pregnant women were likely included where UTI pregnancy codes were not accurately captured, which may have increased the overall rates, this is still higher than would be expected based on the guidelines.7 This may represent overtesting in this population, but without additional data regarding patient symptoms and comorbidities, confirming the appropriateness of testing in this group is difficult.

Deprivation status was also found to impact the odds of testing, with those in the least deprived groups having greater odds of testing. This aligns with prior work around healthcare provided by IMD deciles, though lower testing may be a concern as work in similar populations has found that those in the more deprived IMD groups are more likely to have more resistant infections as well as being more likely to receive a greater number of antibiotics.30–33

Remote episodes were also associated with reduced odds of testing compared with face-to-face episodes. Furthermore, following the start of the COVID-19 pandemic in 2020, the odds of testing decreased compared with 2019, which coincided with a change in GP consultation location, whereby in 2020, the majority of consultations were remote. Additional factors, such as those practices with more episodes having greater odds of testing, are likely due to the greater opportunity for testing in these practices. Episodes occurring in London were also found to have the lowest odds of urine testing. This is likely due to many factors, including fewer UTI episodes in London compared with all three other regions and an established lower rate of overall and UTI-specific antibiotic prescribing in London, which would also impact testing rates.3 34 As established in other work,35 antibiotic use is higher in secondary care settings, and patients often have greater access to secondary care and other GP settings (walk-in and urgent care centres) in London compared with other regions with this data unavailable in the data set used here.

Comparison with other studies

Few recent studies have examined UTI testing and antibiotic prescribing, particularly following the 2018 NICE UTI guidelines. Many recent studies in this area have focused on UTIs in select populations rather than testing and prescribing across a broad sample of the population and often do not include the review of UTI testing and prescribing over time.10 36–39 The patterns in UTI episodes found in this study align with overall patterns in UTI burden found in previous work.1 37 The decreased number of UTIs seen after the start of the pandemic aligns with similar work completed in other populations.40 This decrease also aligns with similar reductions in appointments for other common indications in primary care in England, such as respiratory infections.18 41

The results of this study build on the work of a previous study examining testing and prescribing for UTIs from 2011 to 2015, inclusive.1 This earlier study reported that a urine sample for culture or microscopy was requested within 10 days for 25% of UTI episodes, lower than the 42.2% within 3 days found in this work.1 However, it should be noted that in this study, a urine culture within 3 days of any UTI diagnosis code within a UTI episode was included, likely increasing the rate. This prior study also found that for this 10-day testing window, the percentage of women aged 18–64 years with a urine test was 30% lower than men of the same age group (23.6% vs 30.6%).1 The difference in testing is similar to the result presented here, with the difference in urine testing between males and females found to range from approximately 0.2% (51.6% males, 51.7% females in February 2015) up to 36% (46.6% for males, 35.0% for females in July 2020). From 2011 to 2015, men aged 18–64 years were shown to have the highest testing level at 30.6% of episodes, a lower rate than seen in this study for males aged 16–64 years (57.3%).1 However, in the current study, a greater percentage of females were also found to have a urine test (48.2% for females 16–64) compared with the 2011–2015 study (23.6% for women 18–64). The prior study also found that same-day antibiotics were prescribed in 85.7% of episodes, higher than the 78.2% found in this study. However, this work found a higher rate of urine tests and same-day antibiotics (39.3%) than the rate of antibiotic prescribing and urine testing from 2011 to 2015 (17%).1

Other existing studies examining the management of lower UTIs, including rates of testing and prescribing of antibiotics, have focused on older adults. One study looking at UTI incidence in older adults (>65 years) from 2004 to 2014 in UK primary care found an overall increase in UTIs across the study period in both men and women, with an increase in the use of UTI-specific antibiotics over the same period.37 This aligns with the increase seen in UTI episodes for males from 2015 to 2018 and in females from 2015 to 2017. While this study did not specifically look at antibiotic choice over time, the results did find that the percentage of UTI episodes with a same-day UTI-specific antibiotic prescription was lower (78.2%) compared with the 2004–2014 study (98%).37 Another study similarly looking at adults 65 and older with a community UTI diagnosis found that 87.3% of patients received an immediate antibiotic with men less likely to receive an immediate prescription.25 These results again align with the results found in this work, though overall same-day prescribing was slightly lower at 78.2%, with rates of same-day prescribing around 70.0% for males with females around 80.0%, with men possibly more likely to have delayed or withheld prescribing as found in the prior study.25

Strengths and limitations

The primary strength of this study is the size and established generalisability of the data set used. Population trends aligned well with earlier work, allowing this study to further build on these previous analyses.1 6 Regarding the limitations of this work, though there was a large amount of data accessible for this study, the data used were collected for clinical use and not for research purposes. Therefore, some variables may not be well completed, as they are not needed in day-to-day clinical use. This may limit the accuracy of the urine sampling variable and may not show the true prevalence of every laboratory sampling sent for analysis. However, as previously mentioned, rates were found to be in range and aligned with other studies, lending greater confidence to the results.1 An associated limitation is that patient medical data is provided via GP records, and therefore, all the results from a sample being cultured may not be recorded if no observation/consultation occurred from the result or if urinalysis was conducted at a care setting outside of a GP practice. As part of this work, other indications and comorbidities besides lower UTI diagnoses were not examined in relation to the antibiotics prescribed. This limits the ability of this work to determine the appropriateness of the antibiotic prescriptions. By examining antibiotics commonly used to treat UTIs as specified in the NICE guidelines in conjunction with a UTI diagnosis on the same day, this study has tried to limit the analyses to those prescriptions most likely written for UTI treatment.

In addition, data for other clinical characteristics may be incomplete as well. Due to the structure of the data, it was difficult to determine pregnancy dates within the cohort. Though diagnosis codes for UTI in pregnancy were available and used for this work, using these codes alone likely does not accurately capture all pregnancies in the cohort. The rate of episodes with a UTI in pregnancy code that was able to be identified in this cohort was around 0.4%, while data from 2011 to 2015 looking at UTI diagnosis and treatment found <1% of UTI episodes were for pregnant women.1 It is also likely that many UTIs in pregnancy would not be captured in GP records, as many would seek care and be tested in midwifery-led clinics or at other prenatal services.

When examining UTI episodes, and, in particular, those with positive results captured within 14 days of a urine culture or dipstick, it should be noted that many of these may be asymptomatic bacteriuria rather than a UTI. This is particularly true for older populations where asymptomatic bacteriuria is more common and often leads to unnecessary testing and antibiotic prescribing.26–29 42

Due to limitations in the diagnoses and symptom coding, we could not accurately capture if same-day prescriptions were designated as backup or delayed antibiotic prescriptions. Though these are part of the NICE guidance, with three READ/SNOMED codes designated by the Royal College of General Practitioners to capture data on these prescriptions,43 we found a very low rate of these codes in the data. In this study, there were no susceptibility results available for the urine cultures. This meant that appropriateness could not be fully determined for patients with multiple antibiotic prescriptions.

Regarding the methodology used, though binary logistic regression is a robust method for examining the odds of testing, an alternative approach using mixed effects models would have accounted for clustering by GP practice and patients with multiple episodes. This was unable to be completed within the project timeframe, however it is unlikely that the overall results would significantly change.

Implication for policy and practice

The proportion of urine cultures with a result reported within the medical records in a timely manner may suggest the need for updated policies around test results being recorded in the patient’s central record. This would allow for easier and more efficient tracking of patient conditions and may help decrease the need for future testing. However, it should be noted that the percentage of positive urine culture captured in this work does fall within rates found in other work, which have been found to range from between 20% and 50%.44–47 The coding used for delayed prescribing was difficult to assess in this population despite the inclusion of delayed and backup prescribing in the NICE guidelines. This, again, could support future policies around coding and capturing delayed prescribing at the first encounter in medical records.