Antibiotic management of urinary tract infection in elderly patients in primary care and its association with bloodstream infections and all cause mortality: population based cohort study
BMJ 2019; 364 doi: https://doi.org/10.1136/bmj.l525 (Published 27 February 2019) Cite this as: BMJ 2019;364:l525Linked Editorial
Antibiotic prescribing in primary care
Linked News
Delaying antibiotics in over 65s with UTI may increase risk of sepsis
All rapid responses
Rapid responses are electronic comments to the editor. They enable our users to debate issues raised in articles published on bmj.com. A rapid response is first posted online. If you need the URL (web address) of an individual response, simply click on the response headline and copy the URL from the browser window. A proportion of responses will, after editing, be published online and in the print journal as letters, which are indexed in PubMed. Rapid responses are not indexed in PubMed and they are not journal articles. The BMJ reserves the right to remove responses which are being wilfully misrepresented as published articles or when it is brought to our attention that a response spreads misinformation.
From March 2022, the word limit for rapid responses will be 600 words not including references and author details. We will no longer post responses that exceed this limit.
The word limit for letters selected from posted responses remains 300 words.
1. Per the research, the data source is the CPRD database that depends on Read codes for data collection. Read Codes are based on a co-dependent hierarchy where each code is based on a Parent code. When making the diagnosis of UTI, we are unsure as to whether every time “UTI” or “Asymptomatic Bacteriuria” or any cause of mortality is identified, if data is over/under represented secondary to using only Parent codes in the CPRD database, doubling of the point cases if both Parent code and subset code is taken or any specific way that eliminates or clarifies this error.
2. Miscommunication gap - There has not been standardized way to diagnose the UTI across the said practices in the data. It is unclear what vague symptoms are considered significant for UTI by one GP and is considered non-significant by another GP. There is no standardization to show what symptoms were strong enough to be classified as UTI and rest were considered Asymptomatic Bacteriuria.
3. For non-verbal, acute encephalopathic patients admitted to the hospital, there were no symptoms that would be concerning for UTI. However, since the patient had index UTI and had bacteriuria, these cases were included in the study and would contribute to confoundation of the results.
4. The outcome was decided by bacteremia, sepsis and septicemia in patients who had index UTI. However, there is no results shown on what kind of bacteremia these patients had, was it similar to microbial culture of the urine and if patient had another source of bacteremia. An index UTI may not be the cause of Bacteremia in the next 60 days in patients who have significant comorbidities predisposing them to other infections.
5. Patients older than 85 years, living in a deprived area, with a high Charlson comorbidity index score, were mainly managed using either deferred antibiotics or a no antibiotics approach which seems to be highly contradictory to medical management as these patients would be candidates for getting antibiotics. Seeing their predisposition, they would be more susceptible to having bacteremia, an increased all cause mortality and hence is a factor that may cause inaccurate results. Similarly, increased length of stay would be in itself higher because of their comorbidities and hence may not be directly correlated with antibiotics use for the UTI.
6. Big data issue: Although a large sample size is effective, it has it’s own disadvantages. There is often a difficulty in compartmentalization of data across various EMRs, between different Practitioners and this may affect the results in being under or over reported. Also, upcoding or use of specific diagnostic codes is a common source of systematic error which will significantly affect any outcome.
7. Apply propensity score matching technique to balance the covariate distributions among the comparison groups: The study showed significant differences in patients’ baseline characteristics among the four comparison groups. The potential problem for any observational study is that the study groups were not comparable due to the imbalanced covariates among them. As a result, the treatment effects could be biased, as patients were not randomly allocated. The proposed propensity score matching is a method to reduce the bias in the estimation of treatment effect with observational datasets. Propensity scores provide a summary measure to control for multiple confounders simultaneously.
Competing interests: No competing interests
We advise caution in using findings from this study to guide antibiotic prescribing decisions for older people in routine clinical care (1). In addition to Prof Hay’s considerations (2) there are additional plausible explanations for the associations found in this retrospective study of electronic primary care records of people aged 65 and older that do not imply causation.
The three comparison groups were:
1. UTI Read code+same-day immediate antibiotic (87%)
2. UTI Read code+ antibiotic on day 1-7 (6%) – so called ‘deferred’ antibiotic in the paper.
3. No prescription: UTI Read code and no antibiotic recorded (7%)
The authors found higher risk for hospital admission and death in groups 2 and 3.
The ‘immediate antibiotic group’ is likely to be over represented by well patients who may have had underlying asymptomatic bacteriuria (ASB) or lower urinary tract symptoms but without meeting Public Health England Guidelines threshold for UTI diagnosis (3). The authors attempted to exclude ASB, but they do not say how. ASB, although very common, is very rarely coded in routine clinical practice. Dysuria which is the most predictive symptom for diagnosis of UTI in this age group (3,4) was only coded in 1.2% of all patients and in 18% of those where symptoms were recorded in the 30 days preceding diagnosis.
True delayed/back-up prescribing, when an antibiotic prescription is given to patients at the time of consultation for the patient or carer to collect only if symptoms do not improve or get significantly worse, was not used in this study and is still a fairly uncommon practice, especially among frail, high risk patients. The apparently ‘deferred antibiotic’ group is likely to include mostly those with true UTIs – with symptoms sufficient to be diagnosed and coded initially, then confirmed on urine culture at which point they received antibiotics.
Higher risk patients are likely to be over represented in the ‘no prescription group’, including those seen out of hours or on home visits where a diagnosis has been coded and antibiotics prescribed, but, as is often the case, not captured in the electronic medical record. This group may have also included patients admitted to hospital soon after diagnosis and those with a UTI diagnosis recorded in the medical record following hospital admission but where there is a date discrepancy with Hospital Episodes Statistics admission data; or those near end of life where a pragmatic decision has been made to withdraw or not initiate potentially life prolonging treatment.
Patients included in the ‘deferred’ and ‘no prescription’ groups were older than those included in the ‘immediate’ group (mean age 79.1 and 79.3 versus 76.9); had twice the proportion of patients with an indwelling catheter; and considerably greater proportions of patients with antibiotic exposure or hospital admission in the 30 days prior to the index UTI, all of which are risk factors for E.coli bacteraemia. Poorer outcomes could plausibly be the result of antibiotic resistance from recent exposure or hospital acquired infection with an atypical or less easy-to-treat organism. Although the authors adjusted for covariates, these patients are likely to be less healthy, so residual confounding is probable.
To prove immediate antibiotic prescription for possible UTI in older people is safer than back-up/deferred would require a randomised control trial to ensure that the same types of patients are represented in each group. The fact remains that UTI is very difficult to diagnose in older people, especially if one relies on urine culture. Furthermore, many older people included in studies like this will have been diagnosed and treated for UTI but will not have a true UTI.
Other explanations for the observed rise in BSI in older patients are lower thresholds for admission of the increasing numbers of elderly frail patients that are now managed in the community and increasing rates of antibiotic resistant infections driven by society’s expectation that doctors should “do something”, our intolerance of uncertainty and overuse of antibiotics for soft symptoms of UTI.
References
(1) Gharbi M, Drysdale JH, Lishman H, et al. Antibiotic management of urinary tract infection in elderly patients in primary care and its association with bloodstream infections and all-cause mortality: population-based cohort study. BMJ 2019;364:l525
(2) Hay A. Antibiotic prescribing in primary care. BMJ 2019;364:I525
(3) Diagnosis of urinary tract infections: quick reference guide for primary care. Public Health England 2018. Accessed 13/03/2019 at https://www.gov.uk/government/publications/urinary-tract-infection-diagn...
(4) Loeb M, Brazil K, Lohfeld L, et al. Effect of a multifaceted intervention on number of antimicrobial prescriptions for suspected urinary tract infections in residents of nursing homes: a cluster randomised controlled trial. BMJ. 2005 Sep; 331(7518):669. Available from: www.ncbi.nlm.nih.gov/pmc/articles/PMC1226247/
Competing interests: Stephen Granier works as consultant to Public Health England (PHE) and contributed to the development of Urinary Tract Infection Guidelines and PHE antibiotic audits. Haroon Ahmed is funded by Health and Care Research Wales on a Fellowship using CPRD to investigate UTI-related outcomes in older people. Leah Jones works for Public Health England’s Primary Care Unit on the development of the TARGET Antibiotics Toolkit. Cliodna McNulty leads on development of PHE UTI diagnostic guidance, and is on the NICE Managing Common Infections Committee. Leah Jones works for Public Health England’s Primary Care Unit on the development of the TARGET Antibiotics Toolkit
This article reports a 5% 60-day mortality in untreated older patients with “lower UTI.” For patients who received immediate treatment, mortality was 1.6%. An ominous prognosis for "lower UTI" seems confirmed and antibiotic treatment seems life-saving. The authors accordingly recommend prompt antibiotic treatment. (1) This harmful recommendation rests on the perfectly ambiguous use of the term “uti.”
How was the term used here? The authors report that patients with asymptomatic bacteriuria were excluded, but only 7.5% of their subjects had any “specific or nonspecific signs or symptoms” of a “uti.” Only 1.3% of subjects had dysuria. We are left to wonder who these patients were. Over 90% meet criteria for asymptomatic bacteriuria, assuming they even have bacteriuria, and this is a largely harmless finding. (2)
The article thus provides no data about management of older adults with symptoms referable to the urinary tract.
Severity of illness seems far too great for either asymptomatic bacteriuria or conventional “lower uti.” Of patients who were immediately treated with antibiotics, 16% (27% of men) were hospitalized or had died within 60 days. One purely speculative explanation is that clinicians may have used “uti” for convenience when a complex patient had an ill-defined worsening and incidental bacteriuria.
Can the better outcomes with antibiotic treatment be explained? Patients who received delayed or no antibiotics were less healthy at baseline than those who received immediate antibiotics. They were older with more comorbidities and higher rates of recent antibiotic use and hospitalization (27% and 34% vs. 9%). Residual confounding seems likely despite adjustments. Speculatively, patients being managed with a more palliative approach might have been less likely to receive antibiotics and more likely to die.
Antibiotic overuse is a major public health problem and this article may increase the already great pressure for unnecessary treatment of asymptomatic bacteriuria (2). Some news reports have been quite alarmist. Antibiotic treatment is already a fairly common practice when "UTI" is diagnosed from "offensive urine." The goal of treatment seems to be improvement in the specimen's bouquet.
The article reinforces two important principles: that unmindful use of the term “uti” can be directly harmful (3) and that older patients with vague signs of systemic illness should be carefully attended and may benefit from antibiotic treatment.
1. Gharbi M, Drysdale JH, Lishman H, et al. Antibiotic management of urinary tract infection in elderly patients in primary care and its association with bloodstream infections and all cause mortality: population based cohort study. BMJ 2019; 364:l525. doi: 10.1136/bmj.l525
2. Lindsay E. Nicolle, Suzanne Bradley, Richard Colgan, James C. Rice, Anthony Schaeffer, Thomas M. Hooton; Infectious Diseases Society of America Guidelines for the Diagnosis and Treatment of Asymptomatic Bacteriuria in Adults, Clinical Infectious Diseases, Volume 40, Issue 5, 1 March 2005, Pages 643–654, https://doi.org/10.1086/427507
3. Finucane, T. E. (2017), “Urinary Tract Infection”—Requiem for a Heavyweight. J Am Geriatr Soc, 65: 1650-1655. doi:10.1111/jgs.14907
Competing interests: No competing interests
Gharbi et al. (1) reported that non-prescribed or deferred antibiotics significantly increased the bloodstream infection rate and mortality within 60 days compared with immediate antibiotics among patients with urinary tract infection aged 65 years or older. Based on binary endpoint, we used the data in Table 1 to re-estimate the number needed to harm (NNH) and the corresponding 95% confidence interval. Taking the immediate antibiotics group as the reference, the NNH for an occurrence of death in the no antibiotics group (26.3; 95% CI, 24.4 to 28.6) was lower than that in the deferred antibiotics group (83.3; 95% CI, 69.6 to 103.9).
An alternative approach for quantifying treatment effect is based on survival endpoint, from which summary statistics such as survival rates or cumulative mortalities are often used to reflect the cumulative treatment effect up to a chosen time of clinical interest. The absolute risk reduction (ARR) at time t is defined as the difference in the survival rates between the treatment and control groups, and the corresponding number needed to treat (NNT) is the reciprocal of the ARR. The NNT can be interpreted as the number of patients needed to treat to prevent one death during the t-period follow-up. When the survival rate of the treatment group at time t is greater than that of the control group, a positive value of NNT is obtained, which is often termed as the number needed to benefit (NNB) because the treatment is more effective than the control; when the survival rate of the treatment group at time t is smaller than that of the control group, a negative value of NNT is obtained, which is often termed as NNH because the treatment is less effective than the control. Furthermore, the confidence interval of NNT can be obtained by inverting the lower and upper boundaries of the confidence interval of the ARR (2,3). For this study, taking immediate antibiotics as the reference, the estimated NNH based on the survival rates at 60 days (Figure 3A) was lower in the no antibiotics group (26.5; 95% CI, 24.6 to 27.8) than that in the deferred antibiotics group (83.2; 95% CI, 69.6 to 103.3).
There is no method to compare the two definitions of NNHs, i.e., the former is based on the estimated proportions using binary endpoints and the latter is based on survival rates at a particular time t (t = 60 days in this study) using the Kaplan-Meier survival curves. In general, the NNH based on survival endpoint incorporates patients’ exposure time, e.g., time to death or censoring, which thus provides more information than the NNH based on the binary endpoint. However, the estimated NNHs based on the binary or survival endpoints for the current study are practically identical because there is almost no censoring during the 60 days follow-up.
References:
1. Gharbi M, Drysdale JH, Lishman H, et al. Antibiotic management of urinary tract infection in elderly patients in primary care and its association with bloodstream infections and all cause mortality: population based cohort study. BMJ 2019;364:l525. doi: 10.1136/bmj.l525.
2. Altman DG. Confidence intervals for the number needed to treat. BMJ 1998;317(7168):1309-12.
3. Altman DG, Andersen PK. Calculating the number needed to treat for trials where the outcome is time to an event. BMJ 1999;319(7223):1492-5.
Competing interests: No competing interests
As with many useful publications in the media, the accompanying picture detracts from the message. It seems to show a GP using dipstick testing to diagnose urinary tract infection in a patient older than 65yrs, subliminally subverting current NICE guidance which emphasizes that factors other than the results of dipstick testing are taken into consideration.
Competing interests: No competing interests
In their observational study, Gharbi et al, suggested, ‘… patients with disease not severe enough to prompt antibiotic treatment are at risk of severe consequences”(1). Although sepsis and mortality rates were higher in the 7.2% of older adults with a UTI diagnostic coding but without an accompanying antibiotic treatment coding, there may be an alternative explanation: under-recording and data loss, rather than under-treatment.
About 5% of all GP prescribed antibiotics are prescribed by the out-of-hours (OOH) GP services (2). This prescribing is unlikely to be visible to researchers accessing primary care coding through CPRD records. GP coding from OOH services in most instances is a manual, not automated, process. The patient’s usual GP is likely to manually code the UTI diagnosis but much less likely to code the prescribed antibiotic treatment, resulting in diagnostic codes unaccompanied by treatment codes. If patients attending the OOH service are more acutely unwell, it would hardly be surprising if their subsequent sepsis rates were higher than in those treated by their own GP.
A combination of inadequate prescribing data capture coupled with case-selection bias among OOH attenders may account for the reported findings.
1. BMJ 2019; 364:1525. doi: https://doi.org/10.1136/bmj.l525
2. Edelstein M, Agbebiyi A, Ashiru-Oredope D, Hopkins S. Trends and patterns in antibiotic prescribing among out-of-hours primary care providers in England, 2010–14. J Antimicrobial Chemoth 2017;72:3490–3495. https://doi.org/10.1093/jac/dkx323 https://doi.org/10.1093/jac/dkx323 https://doi.org/10.1093/jac/dkx323
Competing interests: No competing interests
We have read with great interest the article by Gharbi et al in which the authors evaluate the risk of bloodstream infection, hospital admission and mortality within 60 days after an index case of urinary tract infection (UTI) diagnosed in primary care1. The results of the study show that elderly patients with a UTI in which antibiotics were not prescribed or were deferred had an increased risk of bloodstream infection, required more oftenly hospital admission and had an increased mortality rate when compared to those in which immediate antimicrobial was started. The authors conclude that early initiation of antibiotics in older adults with a UTI should be recommended to avoid serious complications. We belive that this conclusion without further clarification could lead to an unjustified increase in antimicrobial consumption.
Urinary tract infection is an important clinical problem affecting older populations and the main source of community acquired bacteraemia2,3. Despite its high frequency, UTI are frequently overdiagnosed in older populations leading to an incorrect diagnostic and antimicrobial treatment in the presence of nonspecific urinary symptoms2. In a study by Sundvall et al the authors observed that many nonspecific symptoms such as increased falls, fatigue or loss of appetite, which are frequently attributed to UTI and treated with antibiotics, were unlikely to be caused by bacteria in the urine4. According to the study by Gharby et al only 7.5% of the UTI episodes presented at least one of a range of specific urinary or non specific signs and symptoms. It would have been interesting that the authors had provided information on which urinary non specific signs and symptoms were considered to suspect a UTI. In addition, no information is given regarding the diagnostic criteria for the dignosis of UTI applied for the remaining 92.5% patients included in the study1. Finally no data on the type of bacteria isolated in the bacteremic episodes and reasons for hospital admission is provided which is of particular interest as the authors considered the later outcome regardless of the reason for hospita admission if it happened 60 days after the index UTI episode.
The authors recognise the uncertainties around the UTI diagnosis in elderly patients included in the study and that they cannot exclude a non-urinary source for the bacteremic episodes. Therefore, at the end it comes down to having clear criteria for the diagnoses of UTI in the older population. Recently a consensus decission tool for empiric antimicrobial treatment of suspected UTI in frail older adults, with and without indwelling urinary catheter, has been developed5. The applicability and impact of this decision tool in primary care and in the the acute care setting is yet to be established.
Lastly we believe that deferring antimicrobial in the elderly population with a suspected UTI is reasonable options, as it has been pointed by others, if signs or symptoms are not clearly indicative of an UTI or even in patients with low urinary symptoms awaiting the results of urine cultures although as pointed by the editorial that accompanying the article further research is needed to identify those in whom delaying treatment (while awaiting investigation) is safe2,6.
REFERENCES
1. Gharbi M, Drysdale JH, Lishman H, et al. Antibiotic management of urinary tract infection in elderly patients in primary care and its association with bloodstream infections and all cause mortality: population based cohort study. BMJ 2019; 364:l525. doi: 10.1136/bmj.l525
2. Nicolle LE. Urinary tract infections in the older adult. Clin Geriatr Med 2016; 32: 523-38.
3. Yahav D, Eliakim-Raz N, Leibovici L, Paul M. Bloodstream infection in older patients. Virulence 2016; 7: 341-52.
4. Sundvall PD, Elm M, Ulleryd P, et al. Interleukin-6 concentrations in the urine and dipstick analyses were related to bacteriuria but not symptoms in the elderly: a cross sectional study of 421 nursing home residents. BMC Geriatr 2014; 14: 88. doi: 10.1186/1471-2318-14-88.
5. van Buul LW, Vreeken HL, Bradley SF, et al. The Development of a Decision Tool for the Empiric Treatment of Suspected Urinary Tract Infection in Frail Older Adults: A Delphi Consensus Procedure. J Am Med Dir Assoc 2018;19 757-764.
6. Hay AD. Antibiotic prescribing in primary care. BMJ 2019. Feb 27;364:l780. doi: 10.1136/bmj.l780.
Competing interests: No competing interests
Re: Antibiotic management of urinary tract infection in elderly patients in primary care and its association with bloodstream infections and all cause mortality: population based cohort study
Dear Editor,
We read the above article with interest. Like other respondents before us, we were surprised by the large 7–8-fold increase in the odds of bloodstream infection when delaying or withholding antibiotics for suspected UTI in adults aged > 65 years in primary care [1]. We were also concerned by notable differences in patient characteristics between the treatment groups. As our research group has been undertaking similar analyses on the same CPRD-HES linked dataset, we attempted to replicate Gharbi et al.’s findings over the same time period.
Unlike Gharbi et al., we found no good evidence for an association between delayed or withheld antibiotic prescribing and increased odds of bloodstream infection in this patient group (OR 1.13, 95% CI 0.97-1.23, p-value = 0.105, NNH 1882) [2]. Patients who were not treated with antibiotics immediately (i.e. on the date of initial consultation) had increased odds of all-cause mortality (OR 1.17, 95% CI 1.09–1.26, p-value < 0.001), but they were also more likely to be admitted to hospital for reasons unrelated to UTI or bloodstream infection, and late treatment with an antibiotic was subject to likely confounding by indication / severity. In our analysis, only half of included bloodstream infection cases could be linked to UTI, and a further 25% of cases were directly linked to a non-urinary source of infection. This suggests that for some individuals, the decision to delay or withhold antibiotics may not have been related to UTI, but was instead a marker of diagnostic uncertainty or other factors that are not recorded in electronic health records.
Our analysis differed from that performed by Gharbi et al. in several ways. First, we modified the study design to ensure that we only included cases of community-onset lower UTI. Second, we considered patients with a delayed prescription or no prescription as a single group, since delayed prescribing is poorly recorded in general practice electronic health records. Third, we considered additional covariates such as whether the index consultation was a home visit, and fourth, we used diagnostic codes recorded in primary and secondary care to investigate the likely source of bloodstream infection. Finally, we performed thorough sensitivity analyses including propensity score analysis to assess the likely impact of confounding.
Our findings highlight uncertainty around the safety of delayed antibiotic prescribing strategies for UTI in adults > 65 years, and the challenges associated with analysing and interpreting results from studies based on electronic health records. There is an opportunity for improved diagnostic and risk prediction strategies to inform the use of delayed prescribing for uncomplicated lower UTI in this age group, but further work is required to understand whether this approach is acceptable to patients and clinicians.
Yours sincerely,
Mr Patrick Rockenschaub, on behalf of:
Dr Laura Shallcross
Dr Ruth Blackburn
Professor Nick Freemantle
Professor Irwin Nazareth
Professor Andrew Hayward
References:
[1] Gharbi M, Drysdale JH, Lishman H, et al. Antibiotic management of urinary tract infection in elderly patients in primary care and its association with bloodstream infections and all-cause mortality: population-based cohort study. BMJ 2019;364:l525
[2] Shallcross L, Rockenschaub P, Blackburn R, et al. Antibiotic prescribing for lower UTI in elderly patients in primary care and risk of bloodstream infection: A cohort study using electronic health records in England. PLoS Med 2020; 17(9): e1003336. https://doi.org/10.1371/journal.pmed.1003336
Competing interests: No competing interests