Article Text

Download PDFPDF

Antibiotics in urinary-tract infections. Sustained change in prescribing habits by practice test and self-reflection: a mixed methods before-after study
  1. T Kuehlein,
  2. K Goetz,
  3. G Laux,
  4. A Gutscher,
  5. J Szecsenyi,
  6. S Joos
  1. Department of General Practice and Health Services Research, University Hospital Heidelberg, Heidelberg, Germany
  1. Correspondence to Dr Thomas Kuehlein, Department of General Practice and Health Services Research, University Hospital Heidelberg, Vossstr 2, Geb 37, D-69115 Heidelberg, Germany; thomas.kuehlein{at}


Background The German guideline recommends trimethoprim (TMP) for the treatment of uncomplicated lower-urinary-tract infections (uLUTI) in primary care. In the authors' research network, the participating general practitioners (GPs) were asked why they prescribe mostly quinolones instead. The GPs stated the perception of a high rate of therapy failure of TMP and strongly rejected the guideline.

Objective To examine prescribing behaviour for uLUTI and whether a practice test of TMP might effect a change in prescribing habits.

Methods The study was conducted using observational and qualitative elements. A first focus-group (n=6) assessed reasons for current prescribing behaviour. In a 3-month practice test, patients with uLUTI were prescribed TMP (150 mg twice for 3 days). In a second focus group, the GPs (n=12) were presented with the results of the practice test.

Results The first focus group revealed that prescribing was mainly driven by former hospital training and what was perceived as common therapy. GPs felt no need to change a successful regimen. In the practice test, TMP had a success rate of 94% (84 episodes of uLUTI). The second focus group revealed that the practice test had strongly changed opinions in favour of TMP. Self-reflection and ownership of data acquisition were seen as major contributions for change in prescribing. After the test period, TMP remained the antibiotic most often prescribed.

Conclusion Internal evidence and peer-group opinion are strong determinants for clinical decisions. A self-conducted practice test, together with self-reflection in a peer group, strongly supports the process of change.

  • Guideline
  • health plan implementation
  • ownership
  • urinary tract infections
  • trimethoprim
  • clinical practice guidelines
  • decision-making
  • general practice
  • health professions education
  • practice-based research network

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Practice-based research networks can act as laboratories that help to bridge the gap between research and practice.1 The CONTinuous morbidity registration NeTwork (CONTENT) is a project to create a morbidity registry in the field of general practice. Electronic patient records (EPR) were modified to allow for data entry with the International Classification of Primary Care (ICPC-2) and the episode of care as the ordering principles.2 The project has been described in detail elsewhere.3 Peer-group meetings with the participants are held regularly. In one of them, we asked the GPs about what they prescribed in uncomplicated urinary-tract infections (uLUTI). The ‘Dysuria’ guideline from the German Society of General Practice at that time recommended trimethoprim (TMP) or nitrofurantoin for uLUTI.4 Both were rarely prescribed. Instead, quinolones were most frequently prescribed. This shift towards broad-spectrum antibiotics is well documented all over Europe.5 The disconnection of research and practice is a problem for both sides.6 Our hypothesis was that a self-conducted practice test can effect change concerning attitudes and adherence to guidelines, and help to bridge the gap between research and practice.


Study design

We conducted the study in a before–after design with a mixed three-step intervention, surrounded by baseline (t1), follow-up (t2) and extended follow-up (t3) data on antibiotic prescribing for uLUTI (see also the flow chart in figure 1).

Figure 1

Before–after study design.

When asked why they do not prescribe TMP, the GPs said that in their experience it had a rate of almost constant therapeutic failure. We proposed a self-conducted practice test of TMP for uLUTI for a period of 3 months. Therapy failure would be easily visible via the CONTENT-EPR. If TMP really had a high rate of therapy failure, this result should be reported to the authors of the guideline, to change their recommendations. The GPs agreed to the proposal.

Step 1 was a focus-group session structured by key questions to explore the attitudes of the GPs towards guidelines and the basis of their current prescribing decisions. Step 2 was the practice test. To be close to normal working conditions, the diagnosis of uLUTI was completely left to the GPs. For the same reason, there was no form of active follow-up. For a period of 3 months, the GPs were asked to open a new episode of care in their EPR for every incident case and prescribe TMP 150 mg twice daily for 3 days. Therapy failure was defined as the necessity to prescribe a second antibiotic or referral to a urologist in that same episode. The guideline defines the cut-off of 14 days to discriminate between relapse of infection and new infection. Therefore, data were followed until 14 days after the trial period. Step 3 were focus groups and interviews where we confronted participants with the results and evaluated the effect on their opinions.

To see whether the test really had sustainably changed practice, we chose the quarters III and IV 2008 (t2) and the quarters I and II 2009 (t3) for follow-up.


GPs in Germany are self-employed and mostly working in solo or small-group practices.

In the peer-group meeting, 23 GPs participated in the CONTENT-project at the beginning of our study. Eleven of them attended the meeting.

  • Step 1: The GPs of the peer group meeting were invited to the first focus group. Six of them formed the focus-group in step 1. The others gave time constraints as the reason they could not attend.

  • Step 2: Ten GPs participated in the practice test. Nine had been in the peer-group meeting. One additional GP had not been in the peer-group meeting, but was willing to participate. Two of the GPs in the peer-group meeting declined to participate in the practice test. One stated that he could not bear the responsibility for such a test. The other was simply not interested.

  • Step 3: There were two focus groups, including nine (five and four) of the 10 GPs who had carried out the practice test. With the 10th GP, we conducted an individual interview. We also conducted single interviews with the two GPs who had denied participating in the practice test. Again, because of GPs' time constraints, the whole group was split up.

Data collection and analysis

Quantitative baseline and follow-up prescribing data were extracted from the CONTENT database. Contingency analyses were performed using χ2 tests.

The focus-group sessions and interviews were conducted by TK, who is leading the CONTENT project. TK is a GP with a lengthy practice experience. He is well known to the group as the peer-group moderator and as such might be described as somewhere in between an equal-ranking member, a facilitator and an opinion leader in the group. The focus-group sessions and interviews were recorded audio-visually and transcribed. A qualitative data analysis was carried out independently by TK and KG, a researcher with a background in medical sociology. We used Atlas.ti 5.2 for analysis.7 A categorising system according to Mayring was developed and consequently modified.8 Subcategories were added after a consensus process among the researchers. The results of the focus groups will be reported in more detail in a separate paper.


Baseline data

The baseline data displayed in figure 2 show the prescriptions of the GPs who later participated in the practice test. Before the intervention, quinolones were by far the antibiotics most often prescribed (56% of 516 prescriptions). Cotrimoxazole was prescribed in 27.7%. The recommended drugs TMP and nitrofurantoin were prescribed in 9.3% and 0.4% of cases, respectively.

Figure 2

Prescribing patterns of antibiotics before and after the trial.

Step 1 of the intervention

The radicalness of the rejection of the guideline recommendation was moderated in front of a recording camera. Asked how their current therapeutic decisions for uLUTIs might come into being, significant answers were: ‘To be honest—it is simply a habit’ (GP 0030) or ‘Some things you get used to, like a reflex’ (GP 0004). The discussion revealed that the current prescribing behaviour was mainly driven by what had been taught in hospital training and what was perceived as the usual therapy seen in hospital referral letters and letters from other specialists. The GPs felt no need to change a successful regimen.

Step 2 of the intervention

In the practice test, 83 episodes of care with ICPC-2 code U71 ‘cystitis/urinary tract infection, other’ and the prescription of TMP were recorded. In five (6%) of the cases, a second antibiotic or referral to urology was necessary (three with second antibiotic, four with referral to urology). Thus, the clinical success rate of TMP in uLUTI was 94%.

Step 3 of the intervention

The GPs univocally said that they could see no difference between the success of a quinolone or TMP. They spontaneously mentioned that they had at once taken over TMP into their standard repertory. Personal experience and peer-group opinion still were named as having the strongest influence on treatment decisions. The self-conducted acquisition and analysis of one's own data were seen as an important confirmation. Three factors, the initiation of the process in the protected time of the peer group meeting, ownership of the process and self-reflection in the focus groups, were seen as the most important facilitators of change. It was acknowledged that the guideline recommendation had even been unknown to some of the participants. Spontaneously, the similarity between the practice test and methods of clinical research trials was addressed. The GPs felt a narrowing of the formerly perceived research-practice gap: ‘… an interesting change in the structure of my thinking. To be able to see, (…) this is finally the way guidelines come into being’ (GP 0014). They stated that as a result of the whole process their confidence in research and guidelines had risen. However, one of the GPs said: ‘By means of this data acquisition we are now absolutely convinced of the guideline. But whether we can ever get the others into our boat, I disbelieve’ (GP 0034).


Figure 2 shows the use of antibiotics before (t1), in the quarters directly following the intervention (t2) and another 6 months later (t3).

After the intervention, prescribing of TMP rose from 9.3% (t1) to 37.8% (t2) (p<0.0001) and was stable 6 months later (37.5% (t3)). Quinolones dropped from 56.0% (t1) to 39.1% (t2) (p<0.0001) directly after the practice test. Six months later, there was a further slight reduction in the prescribing rate of quinolones to 34.4% (t3) (NS).


Principal findings

Given some limitations, our hypothesis that a self-conducted practice test can effect a change in attitudes and adherence to guidelines was supported by our study. Before the practice test, therapeutic decisions were mainly driven by what was perceived as prevailing practice. GPs felt no need to change a successful and common therapy. In the practice test, TMP had a success rate of 94%. The practice test changed the opinion towards TMP. Not only the practice test, but the whole process together with self-reflection in the focus groups was perceived as a fruitful way to deal with guideline recommendations. After the intervention, there was a sustained rise of TMP prescriptions.

Strengths and weaknesses of the study

The number of participating GPs at the time of the peer-group meeting was too small to allow for a control group. We therefore had to choose a before–after design. The selection of participating GPs probably favours a positive effect of our intervention. GPs take part deliberately in the CONTENT-project. They know each other and the moderator well. The overall intervention was based on a very personal relationship of mutual respect. For the follow-up data, a Hawthorne effect cannot be excluded. Unintentionally, the focus groups probably had a major effect on the change achieved. It might thus be difficult to repeat this change in prescribing behaviour on a larger scale. Research active practices are likely to be not fully representative for the wider primary care community.9 However, CONTENT practices participating in this study were clearly not more likely to adhere to this guideline than those in the wider community.

A strength of this study can be seen in creating a feeling of ownership in the group. We believe that this is one of the main forces to promote practice change. The GPs were involved in the study design from the beginning. An open discussion with the support of a facilitator instead of the teaching of scientific facts is not restricted to a focus group. The personal style of our intervention can thus be seen as weakness and strength at the same time.

Discussion of findings

Before the practice test, the GPs perceived guidelines as an alien power threatening their autonomy. They saw them mainly as an instrument to control costs. This finding corresponds well to the literature where guidelines are often seen as instruments in a power struggle between society and individual doctor. They are perceived as a norm that is to be implemented.10 Prescribing can be seen as the core element of medical professionalism. It has been described as ‘the battleground on which the cause of clinical autonomy is defended.’11 The peer-group meeting revealed accordingly the defensive position of the GPs against any form of influence on their prescribing from outside.

The focus group in step 1 revealed that the prescribing process in case of uLUTIs was more reflex-like than a cognitive decision. This phenomenon is a well-known barrier to change.12 13 Even the industry manages to have only little influence on these habits.14 15 In a study from the UK, only 5.4% of all prescriptions in a 12-month period were newly adopted to the repertory of the GPs.16 As seen by our participating GPs as well, hospital doctors had the strongest influence on these new prescriptions, followed by drug representatives.16 It should be mentioned that our practice test somewhat resembles the seeding trials of the pharmaceutical industry that serve as a means to entice doctors to prescribe new drugs.17 According to bacterial resistance rates in vitro, a far higher rate of therapy failure of TMP would have been expected.18 19 Spontaneous resolution of symptoms in many cases, independent of the antibiotic given, is highly probable.20 After the practice test, some participants acknowledged that they had not known the guideline recommendation they had rejected before. The reason for this might be the general distrust in guidelines and evidence-based medicine in Germany.21

Two major elements of managing change are self-reflection and ownership of the change process.22 23 The focus groups were not only a scientific means to evaluate the attitudes of the GPs concerning a guideline but also a means for the GPs themselves to reflect on their behaviour. This reflective process was perceived as very important by the GPs themselves. The outside trained facilitator is a key element in enabling the reflective process.24 Ownership has long been recognised as an important motivator for change,25 26 and so it was seen by our participating GPs. Research has shown that elements of our intervention such as peer-group meetings, ‘reflective practice,’ opinion leaders and active education strategies have effects on professional practice.27–29 What is new is the self-conducted practice test. We cannot say which elements of the intervention had the greatest impact. That only complex interventions can effect change is well known in the literature.30

Reflecting on actions and the reasons behind them may be an effective way of reshaping practical approaches in the future.31 A prerequisite for a practice test would be a suitable EHR, such as the CONTENT software, for data acquisition and analysis without adding much extra workload to the GPs. Further research on the effectiveness of a practice test for managing change is needed.

Meaning of the study

Some medical decisions seem to become reflex-like habits that are difficult to change when necessary. A self-conducted practice test, ownership of data acquisition and analysis together with self reflection strongly support the process of change. It has been proposed that one should speak of concordance, instead of compliance, when it comes to patient adherence to treatment recommendations by physicians.32 Analogously, we should seek to achieve concordance of doctors when it comes to the recommendations of guidelines. We hypothesise that, in the end, it is concordance that forms the key prerequisite to effect change.


We thank our colleagues, M Klinkman, from the University of Michigan Medical School, Ann Arbor (USA), and Rupa Patel, from the University of Washington, Seattle (USA), for language correction and advice.



  • Funding The CONTENT-project is funded by the German Ministry of Education and Research (BMBF)—grant number BMBF 01GK0601.

  • Competing interests None.

  • Ethics approval Ethics approval was provided by the Ethics Committee of the University of Heidelberg; the study protocol of the CONTENT-project was approved by the Institutional Review Board of the Medical School of the University of Heidelberg (approval number 442/2005).

  • Provenance and peer review Not commissioned; externally peer reviewed.