Consumer and Relationship Factors Associated With Shared Decision Making in Mental Health Consultations

This study took place in a psychiatric clinic at a VA medical center from January to September 2012. The clinic has 23 providers, serving approximately 5,200 veterans with diagnoses including posttraumatic stress disorder (PTSD), depression, and anxiety disorder. Three providers were approached and agreed to participate in this study. All providers were female. One was a psychiatrist; the others were advanced practice registered nurses. Consumers were eligible for participation if they were seeing one of the three consenting providers during the study period and were not experiencing symptom exacerbations that concerned the provider.

A research assistant explained the study to eligible consumers when they checked in with the clinic. Interested consumers completed an informed consent process. Consumers were told that the study was about communication but were not told of our interest in shared decision making, to help ensure that participation did not alter natural decision making during visits.

For the appointment, the research assistant placed an audio recorder in the exam room and waited outside. After the visit, consumers completed a series of questionnaires. Consumers were paid $10 for the clinic visit and $20 to complete the questionnaires. Providers were not paid. All procedures were approved by a local institutional review board and medical center review committee.

We collected information about demographic characteristics by consumer report and consulted medical records for information about mental health diagnoses. Medication adherence was measured with the ten-item Medication Adherence Rating Scale (MARS). The first four items (Morisky Scale) measure medication adherence behaviors; the last six items measure attitudes and beliefs about medications. The scale has adequate internal consistency, adequate test-retest reliability, and high positive correlations with the Medication Adherence Questionnaire (r=.79) and the Drug Attitude Inventory (r=.82) (8). Possible scores for the Morisky Scale range from 0 to 4, and lower values indicate higher self-reported adherence.

Illness self-management was measured with the Illness Management and Recovery Scale, client version, a questionnaire containing 15 items rated on a 5-point behaviorally anchored scale. Questions include items related to progress toward goals, knowledge about mental illness, involvement with significant others, symptom distress and coping, and alcohol and drug use. The scale is internally consistent, with good test-retest reliability and convergent validity (9).

The Mental Health Version of the Patient Activation Measure (PAM-MH) was used to assess patient activation, a term delineating the knowledge, skills, and confidence to manage illness (10). The 13-item PAM-MH has been used in several studies of persons with mental illness and shows strong reliability and validity (10). The length of the consumer’s relationship with the provider was measured through consumer self-report. The Working Alliance Inventory (WAI) was used to assess consumer-provider agreement on treatment goals (task subscale), collaboration to achieve these goals (goals subscale), and degree of emotional bond (liking and trust) between consumers and providers (11) (bond subscale). In its original development and validation, the WAI showed high reliability and demonstrated convergent, discriminant, concurrent, and predictive validity. We administered the short form (12 items) of the client version of the WAI and examined total scores and scores on the three subscales for task, bond, and goals.

Transcripts of the recorded sessions were divided among four trained coders, who independently coded each one for overall shared decision making, minimum shared decision making, and overall consumer-provider agreement. Overall shared decision making was evaluated by using the SDM Scale. The scale was adapted from Braddock’s Informed Decision Making Scale, which has demonstrated high reliability in several studies of decision making in primary care and surgery (1). The SDM Scale, which identifies nine elements of shared decision making, has been shown to reliably assess shared decision making in mental health consultations (4). [The elements and scoring of the scale are available online as a data supplement to this report.] Items are summed for an overall score ranging from 0 to 18.

Determination of the presence of a minimum level of shared decision making (SDM-Min) varied by the complexity of the decision. Decisions were classified as basic, intermediate, or complex depending on the level of medical consensus and the extent to which the treatment decision’s consequences could affect the consumer’s life. According to Braddock and colleagues (1), SDM-Min criteria for basic decisions are a discussion of the decision’s clinical nature (element 3) and either a discussion of the consumer’s desired role in decision making (element 1) or an exploration of the consumer’s preference (element 9). Intermediate decisions require those elements plus a discussion of alternatives (element 4), a discussion of pros and cons of the decision (element 5), and an assessment of consumer understanding (element 7). Complex decisions require all nine elements.

Agreement on decisions was coded as full agreement, passive agreement, or disagreement. The latter two categories were collapsed into one category, “lack of full agreement.”

To ensure consistency, one of every three transcripts was independently coded by all coders, who then met to compare coding and reach consensus. Each transcript was coded for overall score, SDM-Min score (0, indicating absence of minimum shared decision making; or 1, indicating minimum shared decision making), and overall agreement score (0, indicating lack of full agreement; or 1, indicating full agreement).

Correlation analyses (Pearson product-moment correlation for continuous measures and point-biserial correlation for dichotomous measures) were used to test the hypothesis that relationship factors (longer consumer-provider relationship and greater working alliance), and consumer self-management factors (greater medication adherence, illness management, and activation) would be related to greater shared decision making, likelihood of minimum shared decision making (SDM-Min), and consumer-provider agreement. We did not adjust for multiple comparisons, given that this adjustment can obscure potential findings in exploratory contexts (12). Results with p values of ≤.05 were considered statistically significant. SAS, version 9.3, was used for analyses.

A total of 102 consumers were approached, and 79 (78%) participated. The most common reasons for refusal were anxiety, lack of time, and lack of interest. Four consumers were unable to stay after their appointment to complete the questionnaires and were dropped from analysis. Results are presented only for visits that included a treatment decision, which was defined as a discussion and an explicit course of action (N=63).

Consumers’ ages ranged from 23 to 71 years (mean±SD=53±10). Nine (14%) were women, and all were veterans. Forty-seven (75%) were white, 13 (16%) were African American, two (3%) were Hispanic, and one (2%) was American Indian. Primary diagnoses were anxiety disorder, including PTSD (N=35, 56%); mood disorder (N=25, 40%); schizophrenia spectrum disorder (N=2, 3%); and other (N=1, 1%). Demographic characteristics were not significantly related to overall shared decision making score, SDM-Min, or agreement.

Consumer-provider relationship duration ranged from two months to ten years (mean=29.8±25.8 months). Shared–decision-making scores ranged from 3 to 13 (mean=9.4±2.4). About half of consumers (N=28, 44%) met SDM-Min criteria. Fifty-five (87%) consumers experienced full agreement, and eight (13%) experienced passive agreement. The decisions were considered basic for 18 (29%) consumers, intermediate for 43 (68%) consumers, and complex for two (3%) consumers.

The measures had satisfactory internal reliability (α≥.73), with the exception of MARS total score (α=.58) and the goals subscale of the WAI (α=.53), which were dropped from the analysis. Cronbach’s alpha for the Morisky Scale was .62.

Overall scores for shared decision making were not significantly associated with any measure (Table 1). SDM-Min score was associated with the bond subscale of the WAI (r=.29, p<.05). Medication adherence (Morisky Scale) was significantly associated with SDM-Min (r=–.32, p<.05), indicating that participants who met SDM-Min criteria were more likely to report taking medications as prescribed compared with participants without SDM-Min (mean scores of .8±1.0 and 1.5±1.2, respectively).

Overall agreement was associated with a shorter consumer-provider relationship (r=–.31, p<.05). Consumers with full agreement had had a longer relationship with the provider compared with consumers with a lack of full agreement (50.3±36.2 and 26.8±22.8 months, respectively). Overall agreement was related to medication adherence in the expected direction (r=–.28, p<.05), but it was not related to working alliance or other variables.

This study examined associations between shared decision making and factors related to illness management and the consumer-provider relationship. Contrary to our hypothesis, overall shared decision making was not related to measured variables. However, minimum levels of shared decision making (SDM-Min) were associated with the consumer-provider bond subscale of the WAI. Bond refers to feelings of liking and trust between consumers and providers. Given that SDM-Min requires either discussion of the consumer’s role in decision making or exploration of preferences, these positive feelings likely encourage such discussions. Conversely, adopting an approach of shared decision making could lead to a greater consumer-provider bond.

Medication adherence was also significantly associated with SDM-Min, indicating a positive relationship between shared decision making and taking medications as prescribed. This finding is consistent with other work showing shared decision making to be associated with adherence (13). Notably, the positive relationship between adherence and agreement on a decision is unsurprising, given that agreeing on treatment decisions should lead to greater treatment adherence.

Minimum levels of shared decision making may be more relevant than overall scores for shared decision making in terms of management of mental disorders and, more generally, chronic illness. Because individuals with chronic conditions typically visit providers regularly, shared decision making often occurs over numerous visits. Therefore, it may be unnecessary to revisit each element at each appointment (7). However, when treatment decisions are made, certain elements should still be discussed, suggesting that the SDM-Min score captured whether these essential conversations took place.

Overall agreement was associated with shorter consumer-provider relationships. Initially, this finding appears counterintuitive—consumers and providers who become more familiar over time with one another’s preferences and needs might be expected to have greater agreement. However, longer relationships might lead participants to feel more comfortable expressing disagreement rather than keeping silent.

It is surprising that patient activation was not related to shared decision making. Consumers with higher levels of activation take greater “ownership” of their health care, which should translate to greater involvement in treatment decisions. However, a previous study also found no relationship between patient activation and observer-rated activation in mental health visits (14). Activation might not manifest itself in communication—specifically, decision making—especially if consumers and providers have had a long relationship.

This study was limited in that we included only three providers and did not examine provider factors—such as years of experience, attitudes, and gender—that might influence shared decision making. Because all providers were female and most consumers were male, gender might have affected findings, particularly given that research indicates female patients are more involved in treatment and ask more questions (15). Second, insufficient statistical power might have obscured some relationships. However, many correlations were so low that greater power would have been unlikely to change results. Third, the cross-sectional study design precludes any conclusions related to causality. Fourth, excluding consumers experiencing symptom exacerbations might have limited our ability to examine a wider range of decision making, including the potential for greater provider paternalism. Fifth, the small magnitude of some correlations may call into question the clinical significance of some associations among variables. More research is needed to better understand these relationships. Finally, this study was conducted at a single VA medical center. There might be variations in shared decision making among different mental health clinics and between veterans and nonveterans.

This study adds to our knowledge of shared decision making and suggests important directions for future research. Specifically, further investigation of the role of minimum levels of shared decision making in chronic care and the influence of the consumer-provider relationship and of patient activation in shared decision making is merited. Future research should also explore the role of these factors over multiple provider visits.

This project was supported by grant CDA 10-034 from the Health Services Research and Development Service, Veterans Health Administration, U.S. Department of Veterans Affairs (VA). The views expressed in this report are those of the authors and do not necessarily represent the views of the VA or the United States government.

The authors report no competing interests.