Strengths and limitations
Although a randomized controlled trial might be considered the ideal design to assess the effect of diagnostic tests on treatment decisions, there is no consensus about the best study design to help understand the link between combination of diagnostic test, test interpretation and treatment decision [
6,
13,
14]. Since our study aimed at describing the use of diagnostic tests and the appropriateness of the treatment decision under daily practice conditions, a prospective observational design was a sensible way of capturing the whole decision-making process from consultation to final treatment decision [
15].
The observational study was carried out as part of the context assessment of a quality improvement program. It brings about some advantages and disadvantages.
On one hand, the external validity needs to be considered with caution. Only 39 practices out of 500 accepted to participate. Previous studies have shown that GPs that participate in research or quality improvement activities tend to prescribe fewer antibiotics and are possibly more interested in reducing inappropriate prescribing than their non-participating colleagues [
16,
17]. Therefore, the differences shown in this study may be conservative.
On the other hand, observation bias was minimized by the fact that GPs were interested in obtaining an accurate recording of their decisions under daily practice conditions. This type of medical audit has been extensively used in Denmark for more than 25 years [
18]. The cornerstone of the methodology is the bottom-up approach in which the GPs set their own improvement goals. Hence, there is no point to record data that do not reflect their everyday practice.
Finally, we cannot completely rule out residual confounding due to unmeasured variables. Nonetheless, the effect of clustering was considered in the assessment of predictors and differences between the diagnostic groups. The model to test the association between the diagnostic pathway groups and treatment decision was adjusted by the predictors for the combination of tests and differences in baseline characteristics between the diagnostic pathway groups.
Interpretation of the findings in relation to previous literature
This study shows the complexity of the decision-making process in a group perceived as “easy” patients in general practice [
19].
International guidelines recommend the inclusion of urine culture as part of the diagnostic process only in suspected complicated UTI to target the right antibiotic choice [
20,
21]. In our study, an uncomplicated versus complicated UTI did not predict the incorporation of urine culture within the diagnostic workup. It may be explained by the current debate on finding a new approach [
22] to define the type of UTI based on burden of symptoms, risk factors, and availability of appropriate antimicrobial therapy. For example, in the no culture group the median anamnestic score was lower than in the other two groups. It means GPs possibly considered these patients as “uncomplicated cases”. Furthermore, the higher the anamnestic score the higher the probability to send the urine culture to the hospital and the lower the probability to perform the urine culture in the practice.
Results from the tests performed previously did not predict incorporation of a test within the diagnostic work-up. It indicates a quality problem in the use of diagnostic tests. An obvious reason for the unsystematic use of diagnostic tests might be the payment scheme. Danish primary care system provides economic incentive for the GP to perform laboratory testing in practice, and this may have influenced the diagnostic workup.
Nonetheless, shortcomings of the dipstick and microscopy may lead to the incorporation of another test in order to gain certainty on the diagnosis. The use of dipsticks in primary care needs to take into consideration two important factors for the interpretation [
11]. Firstly, a minimum bladder incubation time of 4 h, what is difficult if we take into consideration that one of the main motives of consultation is frequency. Secondly, common uropathogens such as Enterococcus spp. and Staphylococcus spp. do not reduce nitrites. Microscopy requires expertise and the sensitivity is low with colony counts < 10
5 [
23]. Then, Danish GPs may rely more on the results of a urine culture.
Behavioural and medical reasons may explain the inclusion of urine culture in the diagnostic workup for most patients too.
Sackett et al. have long ago described four diagnostic approaches in clinical practice [
24]. A mixture of the hypothetico-deductive strategy where the clinician gathers information to either confirm or refute his/her hypothesis and the complete history strategy could explain our results.
In a country with a strong antibiotic stewardship program, GPs are aware of the societal consequences of unnecessary use of antibiotics. They need to find a way to counteract the systematic error called “confirmation bias” [
25]. This error arises from the tendency to seek and interpret evidence that predominantly confirms a pre-existing hypothesis. In our population, patients entered the study with a suspected UTI, thus for the GPs it may be difficult to withhold antibiotic prescribing. Not surprisingly, inappropriate use of antibiotics is mainly driven by overprescribing of antibiotics and only to a lesser extent by underprescribing.
Cardinal symptoms such as dysuria, frequency and urgency has a high negative predictive value [
26]. It means, they are useful to rule out a UTI when absent. However, in our study population, one of the inclusion criteria was dysuria and/or frequency so the negative predictive value could not be used to rule out a UTI.
Urine culture performed in a laboratory is currently regarded as the best available test to rule in or out an infection [
23]. Thus, GPs may see the use of urine culture as the best way to counteract “the confirmation bias”.
The clinical history plays a role when GPs decide whether to perform a urine culture in the practice or send it to the hospital. This has an effect as well on treatment decision.
In patients whose urine culture was performed in practice, the expectation of obtaining a result the following day combined with low anamnestic scores encouraged both GPs and patients to withhold antibiotics. The day of the index consultation, only 14% of patients were over-treated. However, incorrect interpretation of the culture resulted in 24% overtreatment for the final treatment decision.
For patients with hospital culture, a few days’ delay of culture result combined with a higher symptom score led to more antibiotic overprescribing at the day of the index consultation (27%). It is in line with a qualitative study on TDs in patients with suspected UTI, where GPs preferred to treat empirically due to the discomfort experienced by the patient [
19].
Perspectives
Implementation of effective interventions to reduce inappropriate use of antibiotics requires thorough knowledge of the context in which the decision to prescribe antibiotics is taken [
27]. For example, for the group of GPs that participated in this study, education on interpretation of urine culture in practice or improving communication with the microbiology department are interventions that might reduce the inappropriate prescribing of antibiotics.
This assessment should encourage the implementation of studies to assess the use of the available diagnostic tests to set-up improvement strategies that can reduce the unnecessary prescription of antibiotics within each context/country.
Finally, the results presented in this paper focus on the impact on treatment decision. Further research assessing the cost-effectiveness and impact on patients’ outcomes should contribute with information to evaluate the multi-dimensional value of these diagnostic strategies.