The results of this prospective cohort study show that a substantial proportion of women did not follow their physicians’ recommendation for antibiotic treatment and many of them used NSAIDs, although NSAIDs were rarely prescribed or recommended by their physicians. Dipstick results, urine culture results and symptom severity were not strongly associated with women’s decisions for or against the recommended treatment. Only the decision to take NSAIDs was significantly associated with the symptom severity at baseline. The UTI symptoms significantly declined across days, irrespective of whether women decided to take an antibiotic, NSAIDs, none or both. However, we detected a significant cross-level interaction between the change in symptoms and use of antibiotics, meaning that the change in symptom severity was somewhat more pronounced in women taking antibiotics than in the remainder.
Strengths and limitations
The study provides information about management of uncomplicated UTI in primary care beginning with the consultation, investigations and tests and proceeding to the physicians’ treatment decisions and the women’s actual management strategies at home, including interactions with and outcomes of the symptom course.
Although the number of participating women was rather small, the data suggest that the study population was representative compared with other UTI studies in Germany. Baseline data, such as the proportion of patients with a positive urine culture (approx. 75%) [15–18], the proportion of patients with E. coli infections (ca 75%) and the susceptibility data, were comparable with the results of other observational studies in Germany [19, 20]. Most importantly, the sample was large enough to investigate longitudinal changes in symptom severity across days with high statistical power, especially when using multilevel modelling on basis of 769 reports within person and 120 units between person. Finally, since we found a significant decline in symptom severity across days and, given the number of patients and data reports, we can conclude with high statistical power that women had a good chance of recovery irrespective of whether they decided to take antibiotics.
In contrast, an observational study is not adequate to compare the outcomes of the two treatment approaches. Therefore, we cannot conclude that a symptomatic approach is more or less equivalent to antibiotic treatment. In a previous trial we could demonstrate a better treatment success of fosfomycin in terms of symptom burden compared with ibuprofen (15).
Although the safety of the symptomatic treatment approach was not a main focus of the study, we should emphasize that only one case of pyelonephritis occurred in a woman who was treated with antibiotics first and that only 13 women had a second episode of a UTI in the following 3 weeks, nearly half of whom had been treated with antibiotics.
The observational study allowed us to investigate the daily symptoms of patients with uncomplicated UTIs. Growth curve modelling allows the estimation of different growth patterns and the estimation of inter-individual differences in intra-individual change over time, and it is more robust to violations of assumptions than, for example, repeated-measures ANOVA.
Comparison with existing literature
Previous research investigated both the physicians’ treatment approach [21–23] and the patients’ management of UTIs [3, 10, 24]. These studies provided data about tests used in local practices, the susceptibility and resistance of UTI bacteria and/or physicians’ guideline adherence [25–27]. In patient-focused studies, variations in symptom presentation or patient views of the reason for their infection have been investigated. Our study sheds light on whether the decisions made in the consultation were implemented by patients at home, illuminating the gap in the doctor-patient interaction.
Although not all women were prescribed antibiotics, the rate of antibiotic prescriptions was rather high (87%), but it is in accordance with data from other countries [25, 28], for example, an antibiotic prescription rate of 82% in a recent Hong Kong study in primary care, and data from a Spanish study with an even higher proportion of antibiotic treatment (96%).
In several randomized controlled trials [15–18], the symptom course of UTIs, usually assessed and documented by the women themselves, was compared between those immediately prescribed antibiotics and those prescribed symptomatic treatment. The symptoms mostly resolved in both groups, with a somewhat longer duration with symptomatic treatment. Similar to the results of these RCTs, we also found that the women taking antibiotics recoverd somewhat faster. No less remarkable is the fact that women who did not take antibiotics also reported a rapid decline of the UTI symptoms,, as also found by Little et al.  under standardized conditions.
We found only moderate, non-significant associations between dipstick results, which were immediately available at the consultation and the women’s decision whether to take antibiotics. There was no association between the results of the urine culture, available several days after the first consultation, and the initial decision to take antibiotics. On first view, this result may be surprising because one might have supposed that their decision intuitively follows the ‘real biochemical facts’. This is obviously too simplistic a view that reduces patients to their disease, as Di Paleo et al.  suggested it in their review of personalized medicine; rather, the women seem to balance the invasive character of an antibiotic drug against the severity of symptoms.
Implications for practice
This study is another plea for patient participation and shared decision making to form key parts of patient-centred care , this time in the case of uncomplicated UTIs. Women seem to know the best treatment approach to manage their UTI symptoms. Sensitive listening to patient preferences in the consultation may encourage physicians to recommend and prescribe symptomatic treatment more often than antibiotic medicines. However, GPs who prefer to delay antibiotic treatment are sometimes frustrated with patients who expect to get well quickly with antibiotics. They are faced with a complex diversity of factors influencing the culture of antibiotic prescribing, as described in an Irish study , and have to accept that the path to prudent prescribing is long and strenuous. In this respect, studies such as ours may provide physicians with arguments that can motivate and support more women in choosing a symptomatic treatment, at least initially.
Following the NICE  principles of medicines optimization, physicians could help women by discussing their preferences and what is important to them about managing their condition and their medicines and recognize and accept that the women’s values and preferences may be different from their own.
Doctors should understand that women’s disease management will be affected by individual preferences for particular treatment modalities, the avoidance of certain side effects and a personal benefit-harm trade-off analysis of the available interventions and may differ in the level of priority they give to health and symptom recovery compared to other problems . In the end, their decisions seem to have been wise because those who decided to take only NSAIDs fared nearly as well as those who took antibiotics.