Factors associated with the duration of symptoms in adult women with suspected cystitis in primary care

Objective The aim of this study was to identify factors associated to the duration of symptoms of cystitis. Patients and methods We conducted a nested survival study using Druti study data. Druti was a cross-sectional survey conducted in adult women visiting a general practitioner in France, for a suspected urinary tract infection between January 2012 and February 2013. For this study, urine cultures were systematically performed for all women. The evolution of symptoms were monitored daily for two weeks. This nested study considered only women with suspected cystitis from Druti; women with pyelonephritis were excluded. To identify independent predictors for duration of symptoms, a Cox proportional hazards regression model was performed. Results In Druti, 449 patients had a suspected cystitis. Among them, 440 had a follow up at two weeks. Out of the 440 patients, 424 had a prescription of antibiotic treatment (96.4%). The urine culture was positive for 326 patients (74.1%). The median duration of symptoms after consultation was two days (interquartile 1–3). The absence of urinary frequency (median two days versus three days, p = 0.008), age over 55 years (median two days versus three days, p<0.001) and patient’s bet about the presence of a urinary tract infection (median two days, p = 0.021) were associated to a longer duration of symptoms. Positive culture (p = 0.99) and presence of a multi-drug resistant organism (p = 0.38) did not influence the duration of symptoms. Conclusion In a real-life study, factors influencing the duration of symptoms are clinical factors. The delay before re-evaluation in case of persistent symptoms after treatment could be adapted according to the initial clinical examination.


Results
In Druti, 449 patients had a suspected cystitis. Among them, 440 had a follow up at two weeks. Out of the 440 patients, 424 had a prescription of antibiotic treatment (96.4%). The urine culture was positive for 326 patients (74.1%). The median duration of symptoms after consultation was two days (interquartile 1-3). The absence of urinary frequency (median two days versus three days, p = 0.008), age over 55 years (median two days versus three days, p<0.001) and patient's bet about the presence of a urinary tract infection (median two days, p = 0.021) were associated to a longer duration of symptoms. Positive culture (p = 0.99) and presence of a multi-drug resistant organism (p = 0.38) did not influence the duration of symptoms. PLOS

Introduction
Urinary tract infections (UTIs) are one of the most common bacterial infections with an estimated global incidence of 150,000 million per year [1,2]. In the United States, 10.8% of adult women have had at least one physician-diagnosed suspected UTI over the past year [3]. In France, incidence rate of medical consultation for suspected UTI among women over 18 years old had been estimated at 3,200 per 100,000 women in 2012 [4]. According to French recommendations, the clinical diagnosis for cystitis is based on the presence of voiding symptoms associated or not with the nitrites or leukocytes on urinary dipstick test [5]. The utility of the urinary dipstick test is questioned in the literature, as the likelihood of UTI is over 90-95% when clinical symptoms such as dysuria, urinary frequency without vaginal discharge or irritation are present [6,7]. Likelihood of UTI thus varies according to the combination of clinical symptoms. In the French recommendations, two situations are defined for cystitis: uncomplicated cystitis and cystitis with complication risk [5].
The clinical evolution of symptoms is important to know in order to give better information to patients. The aim of this study was to identify factors influencing the duration of symptoms among women visiting a general practitioner (GP) for a suspected cystitis.

Design and study population
We conducted a nested survival study using Druti study data [4]. Druti was a national crosssectional prospective survey conducted in France between January 2012 and February 2013 by GPs of a French practice-based surveillance network, the "Sentinelles" network [22]. The aim of Druti was to describe the patterns of antibiotic resistance of Enterobacteriaceae involved in community-acquired UTIs.
Patients included were female aged 18 and older visiting their GP for suspected UTI (i.e. complaining of at least one clinical symptom of UTI-dysuria, urinary frequency or urinary urgency) for less than seven days. They had to agree to participate, to avoid any antibiotic during the seven previous days, and to be able to provide a midstream urine sample during the consultation. Population of the Druti study is described elsewhere [4].
For our nested study, a first population was formed, excluding patients with substantial arguments for pyelonephritis (temperature above 38.5˚C) [23] or who had not been followed- up to two weeks. From this population, a subgroup, also excluding patients with pelvic or back pain, was formed to rule out any patient who may have pyelonephritis.

Data collection
For each included patient, GP and patient filled in a questionnaire during the consultation. It included questions regarding the description of the current episode, the bet of the patient and the GP on diagnosis of UTI, the therapeutic management, clinical status (risk factors of complications, previous UTIs. . .) and patient's demographic characteristics. For the purposes of the Druti study, a urine sample was collected from all participants, and urine culture performed on all samples at the same laboratory (Ambroise Paré University Hospital, Paris). The GPs and the patients were blinded to the urine culture results.
To recover any missing data, a trained investigator phoned GPs and patients to verify the information, within two weeks of inclusion. A questionnaire was given to the patient at baseline to note the daily evolution of symptoms over the next two weeks. The question was whether patients were still disturbed by their cystitis with a daily report.
English translation of questionnaires used in this study have been published elsewhere [24]. The original version in French is available as Supporting information (S1-S3 Tables).

Selected variables
Selected variables to identify factors influencing the duration of symptoms were those described in the literature and those deemed relevant by the research team. Selected variables were age [11], clinical symptoms [12], duration of symptoms before consultation of more than three days [12], clinical status (previous UTIs [12], recurrent UTIs, UTI with complication risk), UTI with a suspected resistant germ isolate (i.e.: consultation and antibiotic used by the patient in the previous three months, hospitalisation within 12 months before the study), bet of the patient and the GP on the presence of a UTI [12], results of urine culture (positive urine culture [11,12,18], E. coli-positive urine culture, multi-resistant isolates [12,14,16,17,19]) and therapeutic management of the episode (antibiotic treatment [8,12,15,20], single-dose antibiotic treatment [21], antibiotic guidelines-compliant in terms of agent used, dose and duration, conferred to the French 2008 recommendations that were the current recommendations in 2012 [25]).
For this analysis, the French definition of cystitis with complication risk was used: cystitis occurring in a woman with urinary tract anomalies, or severe renal failure, or immunocompromising disease / treatment, or pregnancy, or aged 75 years and older or over 65 years old with criterion of fragility [5]. This definition differed from the European and the American definition of complicated cystitis, notably for diabetes mellitus that was no more a risk factor in France [26,27]. Because UTI were more frequent for post-menopausal women, age was included as a two classes variable inspired from literature: under 55 years and over 55 years [28,29]. Recurrent UTI was defined as more than three occurrences in the last year. Urine culture was defined as positive according to the recommendations of the REMIC (microbiology guidelines-bacteriology and mycology, 2010) of the French Society of Microbiology [30]. Multi-drug resistance (MDR) was defined as acquired resistance to at least three antibiotics according to the natural resistance of the isolate [31].

Statistical analysis
Characteristics of included patients were described as numbers and percentages and quantitative variables as mean (SD) or median (interquartile range, IQR). Survival outcome was defined as the time from consultation to cessation of symptoms within two weeks. Cessation of symptoms was defined as the time (number of days after the initial consultation) where patients were no more inconvenienced by the symptoms. Patients with symptoms at the fifteenth day were censored. Univariate and multivariate analysis was performed with a Cox proportional hazards regression to estimate unadjusted hazards ratios (HRs) and their 95% confidence intervals (95%CIs). Variables selected for multivariate analysis were those achieving a p-value <0.20 in univariate analysis, and those which seemed relevant by the research team and found as predictive in the literature [32].
To avoid introducing strongly correlated variables into multivariate Cox model, we assessed correlations using Cramer's V for categorical variables and the nonparametric Spearman's rank correlation for quantitative variables. In case of correlated variables, the most significant variable in univariate analysis or that deemed the most relevant by the research team was introduced in the multivariate analysis. The proportional hazards assumption was tested using Schoenfeld residual plots and tests [33]. We used stratified Cox models to deal with timedependent variables (i.e., inpatient/outpatient status) [34]. In case of stratified Cox models, interactions were assessed between time-dependent variables and others. Unlike the Druti study, we did not use a sampling design (stratification, stages and sampling weights). Analyses were performed using R software version 2.10.1 [35].
All analysis were performed for both our main population and the subgroup (patients without pelvic or abdominal pain).

Ethical approval
Before the inclusion, each patient completed a consent form. The data collected were anonymous. Each patient included received a 6-digit anonymity number: the first 4 digits corresponding to the doctor's identifier the last 2 digits corresponding to the patient number (01 for the first patient included). Thus, the third patient included by the doctor number 3760 was anonymized with the number 3760-03.
The study obtained research authorization from the French independent administrative authority protecting privacy and personal data (CNIL), number 911,485 (CCTIRS number 11.474), and from the local human investigation committee of Ile de France V.

Population
During the study, 87 GPs participated. Among 538 patients included in the Druti study, only 499 presented a suspected cystitis. Among them, 440 were followed for two weeks (Fig 1). There were 263 patients in the subgroup that were not suffering from pelvic or back pain.
The 440 patients followed and the 59 patients excluded for missing follow up were not different (Table 1).
Among the 440 included women, mean age was 46 years (min 18-max 97 years old). Concerning cystitis with complication risk, 59 (13%) women presented a single complication factor, three had two ( Table 2).
The average of period between the beginning of UTI symptoms and consultation was 1.9 days (95% CI [1.75;2.05]).
Antibiotic treatment had been received by 424 patients (96.4%). Out of them, 248 had a single dose (56.5%) and 347 received an antibiotic according to the recommendations ( Table 3).

Duration of symptoms and factors associated
The median duration of symptoms since the inclusion consultation was two days (IQR 1-3), 106 still had symptoms at the fourth day (24%) and 34 at two weeks (6.8%).
After univariate analysis, only clinical variables presented a p-value under 0.20 (Table 4). Positive urine cultures and multi-drug resistance of E. Coli were forced into the multivariate model.
As age was one factor used for the definition of cystitis with complication risk, the correlation risk between these both variables was very high. Therefore keeping both of them in the multivariate analysis was not recommended and only age was kept. Significant correlations were also found between GP's bet and patient's bet (p = 0.004), consultation in the previous three months and age (p<0.001), consultation in the previous three months and cystitis with complication risk (p = 0,008). Therefore, we kept age and patient's bet for the multivariate analysis.
The variable single dose antibiotic treatment did not meet the proportional hazards assumption (p < 0.001). So the model was stratified on this variable. No interaction between this variable and other variables in the model were found.  The final model is presented in Table 5. Presence of urinary frequency significantly reduced the duration of symptoms (median duration: two days vs three days, p = 0.008). Age ! 55 years old (median duration: three days vs two days, p < 0.001) and patient's bet on a positive urine culture (median duration: two days for both, p = 0.021) increased the duration of symptoms (Table 5).
For the subgroup of patients without pelvic or back pain, no predictor of healing time has been identified: results were not significant (data not shown).

Discussion
This is the first study which investigated factors predicting duration of symptoms for women with suspected cystitis including cystitis with complication risk. In these patients attending in primary care, the median duration of symptoms was two days since the inclusion consultation. Presence of a urinary frequency significantly reduced the duration of symptoms, an age 55 years and older and patient's bet on a positive urine culture increased this duration.
The duration of symptoms measured here was shorter than expected. In the literature, the mean and median duration of symptoms of treated uncomplicated cystitis were 3.5 days and four days respectively [11][12][13]. As in our study, the start date was the consultation date, corresponding to the beginning of antibiotic therapy. In two of these three studies, fewer patients received antibiotics than in our study [12,13]. Prescription of antibiotic reduces the duration of symptoms, while there was no significant differences between the different antibiotic management strategies [11][12][13]. Similary, contrary to our study, the presence of urinary frequency increased duration of symptoms in the literature [12]. Urinary frequency is probably the most troublesome symptom, impacting directly on the organization of everyday life. During the monitoring carried out, the question was "Are you still inconvenienced?" without specifically targeting each symptom. Thus, some patients with this disabling symptom reduction could answer "no" without being completely healed.
Regarding age, a previous Belgian study found no difference in the duration of symptoms between patients over 55 years and patients under 55 years (p = 0.984) [11] but this study excluded cystitis with complication risk. Among our patients aged 55 years and older, 30% have a UTI with complication risk while only 6% of patients under 55 years had one. The presence of cystitis with complication risk, which significantly increased the duration of symptoms in our univariate analysis, may explain this difference. This variable could not have been included in multivariate analyse because of its correlation with age. Further studies are needed to conclude on the respective share of age and the presence of complication risk.
The duration of symptoms was significantly increased when patient thought that they had a positive urine culture. Patients, who self-diagnose and try self-medication, consulted only when symptoms persist, worsen, cause confusion in everyday life, or when there was fear of complications [36,37]. Perhaps, these patients have initially more severe symptoms and the duration of symptoms had been shown to be longer in this case [11,12]. In our study, positive urine culture does not influence the duration of symptoms. This result is similar to a Norwegian study [18] but contradicts other studies [11,12] for which the duration was shorter with a positive culture. The presence of MDR was not associated with the duration of symptoms, unlike what is found in the literature [14,16,17,19]. However, even if the germ was resistant to the prescribed antibiotic, the absence of a relationship between Table 3. Description of the antibiotic prescribed.  clinical and bacteriological results was found in other studies, where the eradication rate after treatment was not related to the clinical cure rate [17,19]. Because antibiotic treatment did not meet the proportional hazards assumption, we cannot conclude about its effect on duration of symptoms with the multivariate analysis.

Antibiotic Dose per tablet Dosage (number of tablet per day) Duration of treatment (number of days) Number of patients treated
The population of the study were women visiting their GP for suspected UTI. Women with suspected UTI who did not consult a GP (e.g., visit to another specialist, self-medication or spontaneous healing) were not taken into account. Indeed, a US study estimated that only 50% of UTIs had a medical visit [38]. The number of UTI (499) in our study could seem low given the number of GPs (87). However, it corresponded to patients that agreed to be included, that were followed two weeks after inclusion and that did not have fever. Another study from Druti has shown that the incidence of UTI in this population was consistent with the literature [39].
Our study did not take into account the severity of symptoms and the general symptoms (unwell, feverish. . .), which are usually present in up to two thirds of patients [11]. The presence of general symptoms would not have changed the median duration of symptoms because these signs were the first to disappear (median two days) [11]. In the literature, initial severe symptoms were correlated with a longer duration of symptoms [11,12]. In practice, there was no standardized assessment of the severity of symptoms in 2012 [40]. Each study about efficacy of antibiotic on UTI used its own questionnaire, not always described [41,42]. The initial assessment and re-evaluation of patients with suspected cystitis is not included in the French guidelines [5]. In this study, the question for the follow-up was "Are you still inconvenienced?", estimating an overall healing time. The duration of each voiding symptom could not be determined. However, the most important information is the overall healing sensation of patients rather than the evolution of each symptom. Contrary to the European guidelines, the French guidelines did not considered the previous treatment failure as a complicated UTI [5,26]. This criteria, which appears during the follow-up, was not considered in our study, as in the other [11][12][13]. At last, the complication factor "women aged between 65 and 75 years with criterion of fragility" could not have been taken into account for lack of data. Due to its correlation with age, the risk of complication was not used in the multivariate model. The lack of this data did not had impact on our results. Data collection had not taken into account the difference between back pain and pelvic pain. This could lead to the inclusion of patients suffering from pyelonephritis in our main population. This is why we performed a sub-group analysis without patients suffering from pelvic or back pain. Results of this analysis were not significant, probably due to the lack of power. However, incidence of pyelonephritis is considerably lower than incidence of cystitis: for women with recurrent infection, the ratio of episodes of acute pyelonephritis to episodes of acute cystitis has been reported to be 1:18 and 1:29 [43][44][45]. Although back pain is suggestive of pyelonephritis [46], near 50% of cystitis had lower back pain in a Belgian study [11]. Furthermore, patients with fever were excluded. Although lack of fever is common in the elderly, only 32 patients were 75 or older in our study [47]. Thus, despite maintaining the variable "pelvic or back pain" in our main population, there should be only a few patients with pyelonephritis included. This must be taken into account when analysing our results. Another symptoms like flank or abdominal pain, nausea or vomiting might occur during a pyelonephritis but the presence of these symptoms were not available in the database, preventing the distinction of pyelonephritis from cystitis in this way. Therefore in our study, the definition of pyelonephritis is limited to the presence of a temperature above 38.5˚C. The presence of vaginal discharge or irritation decreases the likelihood of UTI in some studies [6,7]. However, a more recent meta-analysis concluded that vaginal discharge was a weak predictor of absence of infection [48]. This is why we preferred not to exclude these patients. This study has several strengths. To our knowledge, this study is one of the first to systematically collect urine samples with centralized analysis from all women presenting with symptoms of a UTI in general practice. It is also the first to include cystitis with complication risk. GP did not use a standardized treatment, but continued to apply their usual patient-centered practice. This permitted to have data from the actual daily practice of GPs. Another strength was in the standardized collection of data, which limited information bias. Lastly, our study population is similar to the literature for the average age, the frequency of dysuria and urinary frequency [8,49,50]. The urinary urgency and pelvic pain were less frequent, the hematuria more frequent [8,11,14,51]. The urine culture was positive in 74.1% of patients, with 84% to E. coli, which is comparable to studies including women with suspected uncomplicated cystitis (respectively 64.2% to 74.6% and 75 to 95%) [12,51,52].

Conclusion
The predictive factors identified were clinical. There was no relationship between the evolution of symptoms and bacteriological results. According to the European and American guidelines concerning uncomplicated cystitis, a urine culture should be performed if symptoms do not resolve by the end of the treatment (ranging from one to five days) [26,27]. The French guidelines require informing patients that symptoms may persist for two to three days after treatment and that they may benefit from a urinary culture if their symptoms persist more than three days, whatever the initial clinical presentation is. In our study, one quarter of the patients remain symptomatic on the fourth day which means that one patient out of four should have a urine test according to the French guidelines. As predictive factors identified were clinical, this result could be taken into account for guidelines to perform a urine sample if symptoms persist after considering the initial clinical examination. This approach may be useful to the clinicians to adjust their speech to patient about the evolution of symptoms and the time before collecting a urine sample: be more patient with patients aged 55 and over and less patient for patients with urinary frequency.
Supporting information S1