Background

Much research has been carried out with the aim of influencing the performance of clinicians. The results have varied [1, 2]. As with any human behaviour, clinical practice is difficult to change. Some strategies that have been evaluated, like passive dissemination of clinical practice guidelines, have had little or no effect on practice [3]. Others, like educational outreach visits ("academic detailing") and multifaceted interventions, may be more effective than passive interventions [1].

The reasons why clinical practice sometimes is not consistent with current best evidence varies across clinical problems and from one clinician to another. A logical consequence of this is to tailor quality improvement strategies to address specific barriers [4]. Several trials of tailored interventions have been conducted. The methods used for identifying barriers to change have varied and there is limited evidence of the relative usefulness of different approaches. However, the choice of method for identifying barriers has implications, particularly with regards to resources, since some methods are time consuming and demand the involvement of many individuals. This represents a practical and financial constraint. On the other hand, if such approaches lead to the identification of important barriers that otherwise would have been overlooked, they may be worth the effort.

In this article we describe a simple approach we have used to identify barriers to changing professional practice. This was done as the first step in a process of developing an intervention to improve the pharmacological management of hypertension and hypercholesterolaemia [5]. The intervention focused on three specific recommendations in clinical practice guidelines for hypertension and hypercholesterolaemia [6–8] based on evidence of a gap between the recommendations and current practice in Norway:

  • Contrary to recommendations, physicians seem to rarely estimate the risk of cardiovascular disease before initiating treatment [9]

  • Sales of thiazides are low, despite these drugs being recommended as first-line medication [10]

  • Relatively few patients reach recommended treatment goals [11, 12]

We also report the results of a post hoc exercise and a survey we carried out to evaluate our approach to identifying barriers and interventions.

Methods

We developed the intervention through a process of identifying barriers to implementation of recommendations and measures specifically addressing these barriers ("tailoring"). The methods we used were structured reflection, searching for other relevant trials targeted at improving the management of hypertension or hypercholesterolaemia, conducting a survey among general practitioners and discussion with physicians during pilot testing of the intervention.

Structured reflection

The three authors reflected over possible barriers based in part on our own experience as physicians working in primary care in Norway. We used a worksheet to structure our reflection (see Additional file: 1). The worksheet included factors that might act as barriers in the practice environment, the professional environment, and related to physicians' knowledge, skills and attitudes. One worksheet was completed for each targeted behaviour: increasing the use of cardiovascular risk assessment before initiating treatment for hypertension or hypercholesterolaemia, increasing the prescribing of thiazides for the treatment of uncomplicated hypertension, and increasing the proportion of patients on medication for hypertension and hypercholesterolaemia that reach recommended treatment goals. The worksheet was used to facilitate our group discussion of possible interventions to address the identified barriers.

Our research group had recently completed a trial of a strategy for guidelines implementation when we were planning this study [13]. In that study the multifaceted intervention consisted of several passive components. Information and materials were distributed by mail and to a large degree we relied on the physicians themselves to make an effort at changing their practice. The observed changes in practice were small. In another trial we had found that the use of active sick leave for back patients was significantly increased through a proactive intervention compared to a passive one [14]. Based on these experiences our research group decided to test an active strategy in this study. Therefore we decided to use outreach visits ("academic detailing") prior to considering specific barriers.

We considered systematic reviews of interventions to improve professional practice when we designed our strategy [1]. We searched the Cochrane Group of Effective Care and Organisation of Care http://www.epoc.uottawa.ca trial register for trials of interventions targeted specifically at the management of hypertension or elevated cholesterol in general practice.

Questionnaire to physicians

We surveyed general practitioners about some of the interventions about which we were uncertain after our structured reflection. The details of the survey have been described elsewhere [9]. Briefly, 265 physicians who had participated in an earlier trial conducted by our research group [13] were asked to complete a questionnaire as part of the study-evaluation. We used that opportunity to seek answers to the following questions:

  1. 1.

    Do physicians assess cardiovascular risk before prescribing antihypertensive or cholesterol-lowering drugs?

  2. 2.

    If not, would physicians be more likely to do so it they received a fee for this?

  3. 3.

    Do physicians comply with current regulations limiting the reimbursement of cholesterol-lowering drugs?

The last question was asked for two reasons. Firstly, we were considering making risk assessment a condition for reimbursement of the drugs. Secondly, the existing regulations were a possible barrier to adhering to our recommendations because they conflicted with these.

Pilot testing

During pilot testing of the intervention at two practices, which were selected for convenience, comments from physicians relevant to possible barriers were noted. We also informally evaluated each component of the intervention.

Post hoc focus groups and structured reflection exercise

After we had finished designing the intervention we had the opportunity of testing our method of structured reflection at a gathering of international researchers in the Research Based Education and Quality Improvement group (ReBEQI) http://www.rebeqi.org, December 2003. Each participant was asked to complete a worksheet to identify barriers and possible interventions related to the low use of thiazides among general practitioners. They were randomly allocated to four different groups where they collaborated on completing the worksheet. They were also asked to grade the importance of each barrier or intervention as minor, moderate or major. We disregarded those rated as minor. We compared the results from the four groups with the barriers and interventions we had identified.

Post hoc survey of physicians exposed to the intervention

While conducting the randomised trial to test the effectiveness of our multifaceted intervention we carried out telephone interviews with physicians allocated to the experimental group. They were asked if they adhered to our recommendations and, if not, why. The responses where noted down during the interviews.

Results

Barriers and interventions

Figure 1 illustrates the timeframe for the methods used to identify barriers and interventions. Tables 1, 2, 3 give an overview of the barriers and interventions that we identified for each clinical problem.

Figure 1
figure 1

Flow chart indicating time frame for methods used to identify barriers and interventions. * The trial period ended in December 2003, and data collection and analysis will be completed in September 2004.

Full size image
Table 1 Barriers to carrying out cardiovascular risk assessment, and possible interventions to address these
Full size table
Table 2 Barriers to prescribing thiazides for the treatment of hypertension, and possible interventions to address these
Full size table
Table 3 Barriers to reaching recommended treatment goals and possible interventions to address them
Full size table

Many of the barriers were related to a lack of knowledge and could be addressed through educational interventions. The use of educational outreach visits was logical since we had planned to use an active intervention, based on our previous experience, and since this type of intervention has consistently lead to improved professional behaviour in randomised trials [1]. Similarly, based on previous experience and the capabilities of the software we hade developed [13], we planned on using an electronic risk calculator, electronic prescriptions, patient information materials, and computerised reminders.

The search (July 2001) of the EPOC trial register for randomised trials with the word "hypertension" in any field yielded 58 references. Most were excluded after reading the abstracts, leaving eight, for which the full text was reviewed [15–22]. This did not lead to any changes in our intervention strategy. A search for randomised trials with the word "cholesterol" yielded 13 references. The full text was reviewed for only one of these [23]. This also provided little further guidance for designing our intervention. The nine trials that we reviewed are summarised in table 4.

Table 4 Studies targeting the management of hypertension and/or hypercholesterolaemia
Full size table

The survey results did not indicate that a fee for estimating cardiovascular risk before initiating drug therapy would affect practice [9]. The survey results also indicated that physicians are largely not affected by conditions for drug reimbursement [9]. Moreover, there were no mechanisms in place to enforce such regulations.

We did not identify additional barriers during pilot testing of the intervention with five physicians in two practices, but several of those already identified were confirmed, particularly barriers to prescribing thiazides.

Based on our findings and an assessment of the feasibility and evidence of effectiveness for various interventions, we designed a multifaceted intervention. The elements of the intervention are described in table 5.

Table 5 The final multifaceted intervention
Full size table

We also considered a number of interventions that we excluded. For example:

  • We considered placing computers in waiting rooms so that patients could assess their cardiovascular risk before seeing the physician, but concluded this would be costly and difficult to implement.

  • We considered providing pre-printed prescriptions, but found this would not to be relevant because most physicians use computerised systems for prescribing.

  • We considered exposing conflicts of interest among clinical specialists who advocated using other first line drugs than thiazides, but elected not to do so.

  • We considered exposing techniques used in pharmaceutical advertisements, such as using relative risk reductions rather than absolute risk reductions [24], but concluded this would have at best a limited impact.

Post hoc focus groups and structured reflection exercise

Nineteen researchers were divided into four groups. All groups considered advocacy by drug companies to be a major barrier to change. Routines or habits were also included as an important barrier by all the groups, as well as lack of knowledge concerning the effectiveness of thiazides, their favourable adverse effects profile, and their low cost. All the groups also mentioned competing guidelines or diverging opinions as part of the problem. Three of the groups considered local or national opinion leaders as potential barriers to change. Patients' expectations or perceived expectations were also mentioned by three of the groups.

The interventions recommended by the groups to address the identified barriers are presented in table 6. All the groups suggested the use of computerised reminders to address physicians' lack of knowledge or their habits and routines. All the groups also suggested some form of interactive education, mainly as a counter force to promotional activities by the pharmaceutical industry, and patient information was suggested by three of the groups. Two suggested training physicians to address patient expectations. Two groups suggested developing clinical guidelines and two suggested audit and feedback, but one group considered this to be of minor importance.

Table 6 Interventions to address identified barriers (main results from post-hoc focus group and structured reflection exercise)
Full size table

Post hoc survey of physicians exposed to the intervention

Among the 195 physicians exposed to the intervention, 149 (76%) were contacted during the trial period and agreed to answer our questions. No major additional barriers were identified. However, some physicians questioned whether adhering to the recommendations would represent a good use of resources, specifically the recommended treatment goals.

Discussion

Addressing barriers to change with tailored interventions makes sense and there is some empirical support for this [1]. It is unclear, however, what methods are the most useful for identifying barriers and interventions.

Several qualitative methods can be used to identify barriers, such as interviews, focus groups and observation. These methods may be valuable, but they are relatively labour-intensive. We used a simpler approach to identifying barriers to change. Would the use of other methods have provided us with important additional information? Pilot testing and discussions with five physicians in two practices and interviews with 140 participating physicians did not indicate additional barriers. The post-hoc focus groups with international experts did not add much with regards to barriers and interventions. Several of these groups included "routines and habits" as a potential barrier, which was not explicitly mentioned among the barriers identified by the investigators. However, all interventions that were mentioned by more than one of the groups in the post-hoc focus group exercise were included in our multifaceted intervention. Our use of computerised reminders was based on the assumption that this would help to establish new routines, although we did not record routines and habits as a barrier when we developed the intervention.

There are inherent weaknesses in our approach. One is that the investigators undertaking the structured reflection were few and we were prejudiced by our own experiences. The lack of patient involvement is another limitation, which possibly lead to an under-emphasis of patient-mediated interventions. A weakness with the group of international researchers who participated in the post-hoc focus groups is their lack of familiarity with the Norwegian context.

A number of trials of tailored interventions have been conducted. The methods used to identify barriers to change have varied. Some investigators have simply hosted a meeting [25, 26], others have used questionnaires [23], conducted focus-groups [27–30] or interviews [31–33], or both [34]. Others have used a combination of several qualitative methods [35–37]. Some investigators have used identification of barriers as an intervention in itself [19, 38, 39]. The methods that were used have been poorly described in most of these studies.

Conclusions

Our simple approach to identifying barriers to improving practice appears to have been effective in identifying all of the important barriers, and it was efficient. However, we do not know for certain what barriers other methods would have identified or whether the intervention could have been more effective, if we had used other methods. Further work to address these questions is planned, including direct comparison of alternative methods and evaluations of theory-based approaches http://www.rebeqi.org.

The effectiveness of our multifaceted intervention is under evaluation in a randomised controlled trial.

Author contributions

All the authors participated in the process of structured reflection and in conducting the survey of physicians. AF was responsible for reviewing the results from previous research and pilot testing of the intervention. AF drafted the article while SF and ADO contributed to critical revisions of the manuscript. All authors read and approved the final manuscript.