Health Decision Aids To Facilitate Shared Decision Making in Office Practice

In the day-to-day practice of medicine, patients and their physicians face an increasing number of complex choices regarding prevention, diagnosis, and treatment. The options usually have different profiles of potential benefits and risks (not to mention costs). Deber and colleagues (1, 2) have thoughtfully distinguished problem solving tasks in medicine, which involve one right course of action and provide little room for patient involvement, and decision-making tasks, in which there are several reasonable courses of action and patient involvement is important (1, 2). The optimal choice for a decision-making task depends not only on the probabilities of various outcomes with each strategy but also on the patient's relative preferences for the possible outcome states (and their timing) and his or her attitudes toward risk (3, 4). Health Decision Aids Health decision aids are designed to facilitate shared decision making by helping patients and their physicians choose among reasonable clinical options (5, 6). Although these aids vary widely in content, common denominators are the presentation of more than one reasonable strategy for a clinical management question and a description of the possible outcomes of the various options. Other components may include some form of vicarious experience with possible outcomes (through written or videotaped testimonies from former patients) or exercises designed to help clarify patients' values as they pertain to the decision at hand. More complicated, interactive decision aids allow patients to take an active part in determining the amount and type of the information they receive about their health problem. In general, decision aids are meant to supplement, not replace, the traditional process of patient counseling by clinicians. Although decision aids generally present a menu of standard management options, they have also been used to present the additional option of a randomized clinical trial for informed patients who are at personal equipoise regarding two treatment options (7, 8). Health decision aids take many forms. The most common are combinations of written and oral information (including audiotapes); personal counseling, sometimes supplemented by decision boards; linear videotapes; and interactive, computer-driven multimedia programs. Randomized trials have examined the effectiveness of each of these types of decision aids, addressing a variety of medical decisions. Of interest, although simple written brochures are probably the most widely used decision aids in office practice, they are the least well studied. Measuring a Good Decision Later in this paper I discuss the evidence on the impact of health decision aids. First, however, it is important to be clear about how one would expect such programs to affect patients and the process and outcomes of medical decision making. Several recent reviews have addressed this topic (9, 10). In the big picture, use of health decision aids should result in better decisions. But how does one measure a good decision? Commonly measured outcomes in trials of decision aids include patients' knowledge about the condition of interest and its management options, their satisfaction with the process of decision making or the decision, their perceived level of participation in the decision-making process, their level of conflict about the decision, the treatment choice actually selected, and overall or disease-specific health status. If the choice of a screening or treatment strategy is to be truly informed, patient knowledge about the decision seems a straightforward outcome measure. Investigators studying the impact of decision aids have usually used short, problem-specific questionnaires to assess the impact of decision aids on patient knowledge. Satisfaction with the decision-making process and the decision itself can be measured by using validated instruments (11, 12). Participation in decision making is commonly assessed by using a single-item (13) or multi-item scale (14), the Autonomy Preference Index (14). The degree of conflict a patient feels about a decision is often measured by using the Decisional Conflict Scale (5). Choice of treatment is also decision specific, but changes in what patients elect to do alone do not indicate whether the therapeutic choices are better with a decision aid. Another logical end point for choice of strategy is whether decision aids result in selections that are more consistent with a patient's assessed outcome preferences, risk attitudes, and time preferences. Finally, improved decision making may lead to better health status in some situations, which can be measured by using a host of disease-specific or generic health status measurements. Of note, older research with coaching interventions used to encourage patient participation at their office visits has suggested that increasing patients' participation in their care appears to improve health status, independent of effects on their diseases (15, 16). More recently, we found similar results in a trial of a decision aid for men facing a decision about treatment for benign prostatic hyperplasia (12). There is no clear consensus on which combination of these measures best assesses the impact of a decision aid. Other attributes, such as level of patient anxiety and trust in the physician, may be important as well. I believe that measurements of knowledge, satisfaction with the decision-making process, choice of management strategy, and decisional conflict form a core set of outcome measures for trials of health decision aids. Ideally, trials should also assess patients' preferences for the important outcomes of the candidate strategies (and perhaps risk attitudes and time preferences as well) to determine whether exposure to the decision aid indeed better tailors management choices to patient preferences. The Thorny Issue of Balance When patients face medical decisions with multiple options, there may be many stakeholders. Manufacturers of screening tests, pharmaceutical agents, and medical devices may have a financial interest in promoting a particular choice, as may different specialty groups. Financial conflicts aside, different parties may strongly feel that one particular course of action is right. Recent debates about the value of screening mammography for women in their 40s (17-19) or the value of prostate-specific antigen (PSA) screening (20-23) demonstrate these passions in action. Obviously, decision aids should be developed to present the risks and benefits of candidate options in a nonbiased way; but bias, like beauty, is in the eye of the beholder. Avoiding real or apparent financial conflicts of interest in the development of decision aids and obtaining assessments of balance from viewers (patients and physicians without an obvious stake in the management decision) are two approaches for minimizing bias while developing decision aids. However, ratings of balance by viewers can be tricky. For example, my colleagues and I have shown that although men with benign prostatic hyperplasia rated a decision aid balanced overall, men leaning toward surgery rated the program as being somewhat more supportive of surgery whereas men leaning toward nonsurgical treatment rated the program as being somewhat more supportive of avoiding surgery (24). Effectiveness of Health Decision Aids Two recent systematic reviews have described the results of studies of health decision aids. O'Connor and colleagues (25) conducted a Cochrane systematic review of randomized trials of health decision aids, while Molenaar and colleagues (26) examined both noncontrolled and controlled studies (26). O'Connor and colleagues' review covered trials published through early 1998. Only randomized, controlled trials comparing decision aids to controls or alternative interventions were included. Patients had to be facing actual management decisions, and the decision aids being evaluated had to include, at a minimum, information on options and outcomes relevant to the user's health. The authors identified 17 trials addressing 11 screening or treatment decisions. One outcome measured in 8 trials was patient knowledge about the condition of interest and its management options (Table 1). Four trials comparing a decision aid against usual care showed a significant improvement in patients' knowledge; the weighted mean difference in scores was 19 points (95% CI, 13 to 25 points) on a scale of 0 (none correct) to 100 (all correct). Four trials comparing a more intensive with a less intensive decision aid found a more modest (but still significant) improvement with the more complex program, with a weighted mean difference of 3 points (CI, 0.7 to 5 points). The clinical significance of a difference of this magnitude is uncertain. Decision aids also positively affected decisional conflict in 3 of 4 trials measuring that outcome, with a weighted mean difference of 0.3 (CI, 0.1 to 0.4) on a 0 to 5 scale. Only 1 of 5 studies measuring some aspect of patient satisfaction showed significant improvement with a decision aid (on satisfaction with the decision-making process), and the overall measure of effect was not statistically significant. Table 1. Effect of Decision Aids on Patients' Knowledge of Options and Outcomes Fourteen trials examined the impact of a decision aid on the choice of a management strategy (Table 2). Here, the results depended on the type of decision. For major surgery, decision aids led to more conservative therapeutic choices, with 21% to 42% reductions in selection of the most invasive treatment option and an overall relative risk for choosing the most invasive option of 0.74 (CI, 0.6 to 0.9). No significant effect was seen with decision aids for circumcision, BRCA1 gene screening, amniocentesis, or hormone replacement therapy. One study found a significant increase in hepatitis B vaccination with exposure to a decision aid. Three included trials addressing PSA scree