Absolute Cardiovascular Disease Risk and Shared Decision Making in Primary Care: A Randomized Controlled Trial

INTRODUCTION

Involving patients in decisions about their own health care is considered important from an ethical,^1,2 clinical,^3–5 and public health⁶ perspective. Nevertheless, it is still far from being standard practice. Evidence-based decision aids are meant to facilitate the shift from paternalism to styles of practice that involve patients and their families. According to the last update of a Cochrane review on decision aids,⁷ only 45 of the more than 200 decision aids developed so far have met the inclusion criteria and were tested in randomized controlled trials (C. Bennett, oral communication, May 31, 2007).⁸ The authors conclude that decision aids reduce decisional conflict and increase both knowledge and active involvement in care. No or only slight effects could be found on satisfaction with the decision, health effects, and the kind of option chosen. Observable effects suggest a tendency toward more-conservative choices, fewer operations, and less-intensive screening. Effects on decisional regret⁹ after treatment decisions have not been clarified because of the small number of trials.

Most decision aids are intended to be worked through by patients on their own. Even so, there is a need for testing transactional decision aids applied in the encounter between doctor and patient.^10,11 Interactive, transactional decision aids underline the idea of sharing decisions between patient and doctor. Decision aids used within the consultation might also change physicians’ behavior and help overcome barriers against implementing shared decision making. Such decision aids tend to have a different structure, relying more on structured communication and provision of options between patients and well-prepared physicians during consultation.^12–15

The case for shared decision making is particularly strong with regard to cardiovascular disease (CVD) prevention. Here patients have to consider behavior change, for example, to give up smoking, to start exercise, or to take long-term medication, such as aspirin or drugs that lower blood pressure or cholesterol levels. Especially for the latter, there is a plethora of high-quality prospective studies with relevant end points available to inform patients’ and physicians’ decisions.^16–20 Family doctors are not only expected to inform and involve their patients; in CVD prevention the physician’s focus is also shifting from individual risk factors, such as hypertension or hyperlipidemia, to global risk as a guide for preventive action.²¹ Currently, neither individual CVD risk calculation^15,22 nor engagement of patients in their own health care decisions, although regarded as effective,²³ can be considered as standard care.

We developed a simple, evidence-based decision aid (ARRIBA-Herz) to help physicians achieve the double paradigm shift toward shared decision making and global CVD risk. The development process was based on recommendations of the CREDIBLE criteria published in the Cochrane review on decision aids. Central to ARRIBA-Herz is a script of 6 steps that fits in with family practice consultation (Table 1⇓). The concomitant materials, comprising a booklet for family doctors, a risk calculator, and an individual summary sheet for the patient, follow this structure to allow easy adoption in primary care practice (available in German and English at http://www.arriba-herz.de; see also the online-only Supplemental Appendix, available at http://www.annfammed.org/cgi/content/full/6/3/218/DC1).

Before the trial described in this report started, ARRIBA-Herz had undergone a 3-year phase I/II development process with more than 200 family doctors.^24,25 In 2006, after our study began, the International Patient Decision Aid Standards Collaboration (IPDAS) published 11 criteria to judge the quality of decision aids.²⁶ Most of these criteria have been fulfilled; however, patients had not yet participated in the development of ARRIBA-Herz. In a current phase IV study applying ARRIBA-Herz in routine care, opinions and suggestions of patients are extensively investigated. We describe the results of the phase III trial in family practice that evaluates the effectiveness of ARRIBA-Herz as judged by patients.

METHODS

Participants

All CME groups in the state of Hessen supervised by a large German quality management institute (AQUA) were screened for eligibility (Figure 1⇓ displays a flowchart of participants). We excluded CME groups if several members had taken part in previous meetings on ARRIBA-Herz or if they routinely used other cardiovascular risk calculators. Thirty CME groups comprised 162 family doctors who were eligible and agreed to participate. Some small CME groups were merged to make up 9 to 13 participants in each CME group, resulting in 14 CME groups with 162 physicians. One hundred twenty-seven family doctors attended the first of 2 educational sessions in either the intervention or the control arm. After the participating physicians had completed the educational sessions, we asked them to recruit up to 15 of their adult patients. Study nurses visited each practice to familiarize practice teams with study procedures and materials. To make sure that similar patients were included in both study arms, physicians were asked to approach all consecutive patients who had their cholesterol levels measured during a period of 4 weeks. We audited practices via telephone 3 times during the recruitment period to confirm consecutive recruitment. The only exclusion criterion on patient level was insufficient knowledge of German language. As a result, our sample included patients with and without clinical vascular disease.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Study flowchart—CME groups, practices, and patients.

CME = continuing medical education.

Rolling recruitment of patients from each practice was spread evenly from May 2005 to March 2006. Six months after the index consultation, all participating patients were invited to a follow-up visit.

Ninety-one family doctors recruited at least 1 patient each. Our analysis is based on these family doctors and their study patients. Five hundred fifty patients were recruited by family doctors in the intervention arm, and 582 by family doctors in the control arm. Two hundred six patients were lost to follow-up, of which 82 were lost because their physician no longer participated in the study (Figure 1⇑).

RESULTS

DISCUSSION

We were able to show that a decision aid based on absolute CVD risk can be successfully disseminated in the setting of family practice CME groups. Patients attending intervention practices were more involved and expressed higher satisfaction. Additionally, they showed less decisional regret after 6 months. Furthermore, shared decision making, contrary to what is sometimes suspected, did not lead to a worsening of the global CVD risk status. A longer follow-up period than a one-half year in regard to CVD risk and decisional regret to confirm these effects in the long run seems warranted but was not feasible in our study. Yet, a longer intervention effect of a single consultation on these outcome parameters also seems unlikely.

The literature is abundant with failed attempts to change professional behavior.^37,38 Several factors may have contributed to the intervention effect of ARRIBA-Herz. A development process using several feedback loops between family doctors and an academic department prepared the ground for the definitive phase III study reported here. We designed a multifaceted implementation strategy³⁹ consisting of educational seminars for existing CME groups, involvement of local opinion leaders (moderators of CME groups), printed materials, and aids for the consultation. We condensed insights from epidemiology, intervention studies, and risk communication in a single simple script to be used in family practice consultation. Successful implementation of innovative behavior among professionals, however, is often contingent upon external factors. In our case, pressure to contain prescribing costs and exaggerated claims of medication effect pose a dilemma to family doctors. ARRIBA-Herz seems to offer a practical solution to this dilemma.

Most decision aids on CVD prevention address one specific intervention only, such as antihypertensive drugs.⁴⁰ A Canadian decision aid covering a broad range of options has only been tested in a small trial and appears to be long and very complex.⁴¹ Another comprehensive decision support tool developed in the Netherlands on CVD prevention that focuses on lifestyle changes was administered to patients by their family doctor. Like most decision aids developed so far, this tool was worked through by patients on their own.⁴² This decision aid did not change relevant subjective or clinical outcomes, the latter defined as adherence to guidelines on cardiovascular risk by patient and physician. Most decision aids included in the Cochrane review⁷ comprehensively inform patients and invite them to reflect upon different options on their own before consulting their doctor. These decision aids improved knowledge but often did not lead to higher satisfaction after consultation, perhaps because the doctors’ behavior remained unchanged. Although the interactive decision aid tested in our study did not improve knowledge, it resulted in higher patient satisfaction and less decisional regret after 6 months. The small effect on knowledge might be because the decision aid was not designed as a consumer-oriented decision aid with comprehensive information material for patients, relying instead on the information of well-informed family doctors during consultation. On the other hand, we might have underestimated the effect by using only a 3-item sum scale.

The intervention induced a noticeable change in communication patterns as reported by patients (Tables 4⇑ and 5⇑). ARRIBA-Herz helped doctors and patients discuss individual concerns and pace information according to patients’ needs. The responses to most statements referring to the theoretically defined steps of the shared decision-making process (Table 5⇑) fit in with the study hypothesis.

For study outcomes we mainly defined soft variables, ie, patient participation, satisfaction, and decisional regret. Neither previous studies nor theoretical considerations suggest a substantial effect of decision aids on objective risk factors. We assessed CVD risk at 6 months’ follow-up primarily to exclude a deterioration of risk factors caused by the intervention. We observed a reduction of 10-year CVD risk in both groups. We interpret this finding as an effect of regression to the mean. Confidence intervals reached from less than to more than zero, but they are sufficiently narrow so that a deterioration of CVD risk through the intervention seems unlikely. Larger studies are still indicated. We strongly believe, however, that the case for shared decision making predominantly rests on the subjective argument. More involvement and less decisional regret of patients can be regarded as important goals of health care per se, not only measurable utilities.²

Given the pragmatic study design with cluster randomization, blinding of family doctors was not possible. Although we required participating practices to recruit identical patients, family doctors in the control arm recruited more diabetic and slightly more high-risk patients in secondary prevention, and their colleagues in the intervention arm recruited more patients with higher preferences toward more participation in the decision-making process. The former effect might also be due to cluster, ie, practice characteristics, since 8 practices of the control arm as opposed to only 3 in the intervention arm recruited relatively high numbers of diabetic patients. Six of the control practices with high numbers of diabetic patients had implemented specific programs for diabetic patients. Higher numbers of patients with a preference for participation in the intervention arm, as already mentioned in the result section, might also represent an intervention effect, as the questionnaire was applied after the index consultation. Some physicians extended the recruitment period to 6 weeks, so that consecutive recruitment was presumably not always performed. Thus, a selection bias in the intervention group toward discussing prevention with low-risk patients and with patients with a stronger preference toward active participation cannot be excluded, as was also suggested by one of our preliminary studies.²⁴ Because patient participation was the primary effectiveness measure, however, these imbalances do in our view not affect our main conclusions.

The setting of CME groups coordinated by the AQUA-Institute on behalf of the statutory ambulatory care organization allowed us to include family doctors who perhaps would not have taken part in a completely voluntary CME event. Furthermore, we excluded family doctors practicing certain features of the intervention already. We therefore assume that in our sample, the proportion of doctors adopting innovations at a later stage⁴³ is larger than in the population. Thus, we might have underestimated intervention effects.

In conclusion, a decision aid for the prevention of CVD has been effectively disseminated through family practice CME groups. It is based on absolute risk and has a script-like structure specifically adapted for the consultation between patient and family physician. We believe, however, that both consumer-oriented and consultation-based decision aids are important, might be combined, and deserve further study.