Family Involvement, Medication Adherence, and Depression Outcomes Among Patients in Veterans Affairs Primary Care

Low levels of social support and poor-quality relationships are strongly related to poor depression outcomes (1,2), particularly for older patients in primary care settings (3). However, there is little additional information about these associations to guide clinical care. Family members are a key health-related social support resource for many patients and are the support system most commonly accessed by clinicians. Nonetheless, many primary care providers report barriers to increasing family and friend involvement in patient care (4). More knowledge about the role of family involvement in recovery from depression can provide a basis for addressing social support issues in treatment of depression.

Many benefits of including family participation in evidence-based treatment of serious mental illness have been reported (5–9). In addition, research has indicated that mental health consumers report increased satisfaction with mental health services when their clinicians involve family in the consumers’ care (7), yet few studies have specifically addressed family involvement in the treatment of depression. The results of one randomized clinical trial suggested that brief cognitive-behavioral couples therapy significantly improved depression outcomes (10), but the study was small, was focused solely on marital relationships, and was limited to women participants with depression. Overall, the empirical evidence regarding family interventions for depression in primary care is sparse, and there is insufficient knowledge regarding patients’ preferences for family involvement in care.

Although there is evidence that social support improves well-being and recovery from psychiatric symptoms (6), the multidimensional nature of social support makes it difficult to understand its specific role as a risk or a protective factor for depression. For example, in older populations, received support does not necessarily buffer the effect of disability and depression, but perceived support may alleviate negative outcomes (11,12). Indeed, individuals who perceive a high level of support from family or friends often report higher levels of well-being than those with low support, indicating that perception of support may be more vital to outcomes than actual social support.

Social support, in general, and family relationships, in particular, have also been associated with better treatment adherence among patients with chronic conditions and severe mental illness (13–15). Less information is available that specifically relates social support and adherence among patients with depression, despite the evidence that nonadherence to depression treatment in primary care is common (16), with up to 68% of patients diagnosed as having depression discontinuing antidepressant treatment within three months (17). These findings further underscore the significance of family-involved interventions by suggesting their potential to improve both depression outcomes and adherence with treatment.

Clinician engagement of family in treatment is a complex activity. For example, low social support could signal a sicker, more treatment-resistant patient who requires greater vigilance or treatment intensity, without necessarily indicating that attempts to enhance social support would be beneficial. In addition, although family interventions for patients with serious mental illness have proven effective in some cases, a long tradition in mental health care warns of the pitfalls of breaching confidentiality or crossing treatment boundaries in cases where family relationships are problematic.

This study of depressed veterans in primary care analyzed their perceptions about social support, family involvement, and satisfaction with the extent of clinician involvement of family members in care. More specifically, satisfaction with the extent of clinician involvement of family in care was explored among patients with high versus low levels of family involvement. Additionally, the relationships between social support, family involvement, and patient outcomes were investigated over an 18-month period to determine whether specific types of social support or family involvement were associated with depression symptoms and medication adherence outcomes over time.

Methods

Data were derived from the Well-being Among Veterans Enhancement Study (WAVES). WAVES was a cluster randomized clinical trial testing the Translating Initiatives in Depression into Effective Solutions (TIDES) project, a Veterans Affairs (VA) depression quality improvement program. The overall TIDES and WAVES studies and methodologies have been described in detail (18,19).

Study sites and sample

Study sites.

WAVES included veterans from ten VA primary care clinics located in rural or semirural settings or small cities in five states.

Sample.

Between June 2003 and June 2004, trained interviewers from a survey firm under contract with the VA conducted computer-assisted telephone interviewing with VA patients who were enrolled in VA primary care and had an upcoming appointment to determine study eligibility and collect baseline data. To be eligible for WAVES, patients had to report major depressive symptomatology as measured by the Patient Health Questionnaire (PHQ-9), a validated depression screening measure (20–22). A total of 10,929 primary care patients were screened by the PHQ-9, and 1,313 patients scored ≥10, indicating major depression. Consenting veterans with PHQ-9 scores ≥10 were invited to enroll in WAVES and respond to additional survey items, including questions about social support, family involvement in care, health care utilization, perceptions about depression, sociodemographic characteristics, and other variables (N=761). Follow-up surveys were conducted seven and 18 months after enrollment. Approximately 72% (N=546) completed the seven-month survey, and approximately 52% (N=394) completed the 18-month survey. All results were adjusted for potential attrition bias at each time point (18). Institutional review boards at the University of Washington and at participating VA sites approved study procedures.

Measures

Dependent variables.

The outcome variables of interest included depression severity and self-reported medication adherence. Depressive symptoms were defined by PHQ-9 score. Possible scores on the PHQ-9 range from 0 to 27, with higher scores indicating worse symptomatology. Scores ≥10 at baseline were an inclusion criterion for the original study.

Medication adherence was determined by three questions about the participants’ behavior in the prior six months. The first item inquired whether they had missed doses of medication prescribed by a doctor. The second item asked whether they had taken a medication differently than prescribed, including stopping a medication, changing the dose, or taking a medication for a longer period, without first getting a doctor’s agreement or had taken a prescription medication that was not prescribed for them. The third item asked if they had stopped taking medications because of side effects. These items were not specific to antidepressant medication adherence, although a majority of participants (>55%) reported receiving a prescription for emotional or mental problems within the prior six months.

Independent variables.

The key predictor variables of interest included family involvement in medical care and social support. Participants were asked to rate satisfaction with their clinicians’ involvement of family in their care during the past six months on a 5-point Likert scale ranging from “very satisfied” to “very dissatisfied.” To determine amount of involvement of family by providers, participants rated how much their VA health care providers involved family members in their care during the past six months on a 4-point Likert scale ranging from “a lot” to “not at all.” Finally, to determine likelihood of patient predilection to self-initiate family involvement, participants were asked to rate how likely they would be to discuss with family members a decision regarding treatment of a major medical condition on a 4-point Likert scale ranging from “very” to “not at all” likely.

Social support was assessed by eight items from the Medical Outcomes Study Social Support Survey (23). The same items were used in the Mental Health Awareness Project, a study funded by the National Institute of Mental Health comparing usual care with team-based quality improvement care for depression (24). Participants were asked to complete a sentence beginning “How much do you feel you can count on someone to” with indicators of either emotional or tangible social support. Items measuring emotional support included “show you love and affection,” “confide in or talk to about yourself and your problems,” “give you a hug when you want one,” “love and make you feel wanted,” and “turn to for advice about how to deal with personal problems.” Indicators of instrumental support included “help you if you were confined to bed,” “take you to the doctor if you needed it,” and “help with daily chores if you were sick.” Participants responded by using a 5-item Likert scale ranging from “completely” to “not at all.” Higher emotional and tangible support scores indicate more perceived support.

Additional sociodemographic and clinical characteristics that had the potential to influence depression or medication adherence were controlled in statistical analyses. For demographic characteristics, these included age, education level, current employment status, and type of treatment (TIDES or usual care) provided by the primary care clinic. Given the comorbidity of depression and other clinical characteristics, such as posttraumatic stress disorder (PTSD) and general medical conditions (25), both self-reported health and PTSD were also included as covariates. PTSD and poor health have also been linked to poorer adherence outcomes (26,27), further underscoring the importance of controlling for these factors in all statistical models. Self-reported health was based on the Health Status Questionnaire (28), in which participants are asked to rate their health on a 5-point scale, from 1, excellent, to 5, poor. Finally, PTSD was assessed by using the Primary Care PTSD Screen (PC-PTSD) (29,30), a measure of four PTSD symptoms (nightmares and intrusive thoughts, avoidance, hypervigilance, and emotional detachment) experienced in the past month. A cut-off score of three positive symptoms identifies PTSD with a sensitivity of .78, a specificity of .87, and diagnostic efficiency of .85 (30).

Data analysis

A chi square test examined differences between patient satisfaction and perception of providers’ involvement of family in the patients’ care. To account for the small cell sizes and ordinality of the variables, StatXact was used in the chi square analysis, and the exact p values were determined on the basis of a Monte Carlo estimate.

A multivariate linear regression examined the associations between baseline perceptions of family involvement and social support and depression severity over time, controlling for sociodemographic and clinical characteristics. This analysis was used to test the relationship between the main predictors (family involvement and social support) and the dependent variable (depression). We ran one model that incorporated depression values for each participant at baseline and at seven- and 18-month follow-ups. By incorporating all time points into a single model and then clustering by participant, we minimized the potential for a type I error that could otherwise be caused by multiple testing in separate models at three separate outcome time points. The clustering by participant in each model also accounted for intrapatient correlation over time.

In subsequent analyses, the relationship between the main predictors (family involvement and social support) and the binary medication adherence outcomes were investigated. Three generalized linear model logistic regressions calculated odds of each medication adherence outcome over time by using baseline predictors of family involvement and social support. Again, each model incorporated all three outcome values for each participant (baseline, seven months, and 18 months) to maximize power and minimize error. Sample weights were used in all regression models to adjust for differential enrollment or sampling bias; results with and without weighting were similar. In addition, computer-assisted telephone interviewing procedures resulted in minimal missing data (<5%). All analyses were conducted by using Stata/MP 11 (31).

Results

Participant characteristics

Participants’ mean age was 60 years (range 24–93), and the PHQ-9 mean score was 15.8 (range 10–18), indicating an older participant sample with high depressive severity. As shown in Table 1, the participants were primarily white, married men with at least a high school education. Most (N=611, 81%) participants reported poor or fair health, and many (41%) were also on disability. In addition, a majority typically lived with others and rated social support as moderately high. For example, 56% (N=427) said they could completely or mostly count on someone to help if they were confined to bed. Most participants also indicated a predilection to self-initiate family involvement, given that 62% would be very likely to discuss treatment of a major medical condition with family members.

Participants’ satisfaction

A cross-tabulation between patients’ satisfaction with family involvement and degree of family involvement by VA providers is provided in Table 2. Overall, 531 (73%) participants reported very little family involvement by their clinicians, and 434 (60%) reported satisfaction with their clinicians’ involvement of family members (χ²=152.6, df=12, p<.001). This suggested that many patients appeared to be satisfied with little family involvement. In contrast, among the 197 patients who reported family involvement by their providers (“somewhat” or “a lot”), 181 (92%) also reported being somewhat to very satisfied with their VA providers. Finally, a majority (N=81) of the 88 patients who were “somewhat” or “very” dissatisfied with family involvement by their VA providers also reported little to no family involvement by their VA provider. Of the 728 participants, only two reported any level of dissatisfaction with providers who involved their family members “a lot.”

Family involvement, social support, and depression

After the analyses were adjusted for age, education, employment status, PTSD, self-reported health, and primary care site, participants who reported more satisfaction at baseline with their providers’ involvement of family in care also reported significantly fewer depressive symptoms over time (p=.006). Social support (both emotional and tangible), perceptions regarding amount of involvement of family by VA providers, and patient predilection for involving family in care at baseline were not significantly associated with depressive severity at follow-up (Table 3). Recall that this analysis incorporated both seven- and 18-month follow-ups into a single model clustering by participant in an effort to minimize error associated with multiple testing in separate models. Additional analyses (data not shown) in which models were run separately for seven-month and 18-month outcomes yielded the same results.

Family involvement, social support, and adherence

The results of the logistic regression analyses, which tested whether baseline family involvement or social support was related to medication adherence over time, are presented in Table 4. Using a method similar to the linear regression analysis, we included all outcome time points into a single model to maximize power and minimize error. After controlling for relevant sociodemographic and clinical characteristics, the analysis found that neither emotional nor tangible social support was significantly associated with medication adherence. In terms of family involvement, only patient satisfaction with VA providers’ involvement of family members in care was significantly related to medication adherence. Specifically, patients who reported higher satisfaction with family involvement at baseline had lower odds over time of missing doses of medication (odds ratio [OR]=.82, p=.006) and of taking medication differently than prescribed (OR=.80, p=.007). No family involvement variables were significantly associated with stopping medications. Separate models for seven- and 18-month outcomes (data not shown) yielded similar results.

Discussion

The results demonstrated that perceptions about family involvement were linked to both depression improvement and medication adherence among depressed primary care patients. More specifically, participants who were satisfied with their providers’ involvement of family members in the participants’ care reported lower depression severity and better medication adherence. Furthermore, a majority of participants who were satisfied with the current level of family involvement in their care also reported low family involvement. This finding underscored the importance of clinician responsiveness to patient wishes regarding family involvement in care. Furthermore, patients’ satisfaction with their providers’ involvement of family members was more strongly linked to depressive severity and medication adherence than social support, actual amount of family involvement, or patient predilection to self-initiate family involvement.

Although one must be careful to avoid drawing causal conclusions from these descriptive analyses, there does appear to be room for VA providers to better involve family members in patient care without compromising patient satisfaction. For example, very few participants reported any level of dissatisfaction with providers who involved their family members “a lot” (Table 2). We conclude, however, that providers should first inquire into their patients’ perceptions of their relationships with family members, point out that family involvement might be helpful, and discuss potential barriers to or problems associated with family involvement. Ultimately, the results suggested that providers ought to be guided foremost by patient preference for family involvement in care.

In this study, there were no significant associations between social support (tangible or emotional) and depression severity or medication adherence. This result suggested that family involvement and social support are distinct constructs and their effect on outcomes may have important differences. It is also possible that these findings were skewed because most participants reported relatively high levels of social support. Alternatively, this result may indicate that factors other than low social support are more closely tied to depressive symptoms or medication adherence.

There are several limitations to consider when evaluating the results of this study. First, the study relied solely on self-reported information from participants who screened positive for depression. Self-reported and actual medication adherence might differ slightly. In addition, participants reported adherence with all medications, which included but were not limited to medications for depression. Although many participants were prescribed medications for their emotional or mental problems, adherence patterns may differ by type of medication or by patient perceptions of various disease conditions.

In addition, measures of perceptions of family involvement and social support were available only at baseline; thus we were unable to investigate temporal sequence or changes in these variables over time and how these issues may also predict depression outcomes or medication adherence. Finally, this study focused on a sample of primarily older male veterans with multiple medical comorbidities. These results may not be generalizable to other patient populations, given that perceptions about family relationships and health care behaviors likely vary in groups by gender and ethnicity. At the same time, this sample is meaningful, given that studies on depression and family relationships in later life tend to underrepresent men, who are affected differently than women by efforts to improve depression care (32). Furthermore, older men with depression may be more resistant to treatment and are undertreated in comparison with women and younger groups (33,34). Given the prevalence of depression and the negative relationship between depression and overall health among older, white, men, an increased awareness of how social relationships can improve depression care in this population is useful, both within the VA system and in other health care settings.

It was beyond the scope of this study to interpret why most patients were satisfied with low family involvement, but it does merit further investigation. For example, the finding raised a number of questions, including whether limited family involvement over time has a negative impact on treatment adherence and recovery and whether patients are aware of how family relationships may benefit care. There may also be a variety of reasons why depressed patients choose not to reach out to family and friends for support. For instance, some patients experience conflicted or problematic family relationships, which would pose an obvious challenge to providers who seek to augment family involvement. It will be important for future research to determine the range of situations in which providers may improve treatment by increasing family involvement. For example, among patients who would like more family involvement in their care, providers may find that focusing on medication adherence can be a way to initiate conversations with family members and include them in care. Medication adherence may also be a useful topic for beginning conversations among families with conflicted ties. However, alternative ways to increase social support for patients with problematic family relationships would also be beneficial.

Conclusions

Although social support was not associated with clinical outcomes, the satisfaction of depressed veterans with family involvement by clinicians was an important factor linked to both depression severity and medication adherence over time. Patient satisfaction with clinicians’ involvement of family also appeared to be more important than the actual amount of family involvement in care. This research contributes to the empirical knowledge regarding patients’ preferences for family involvement in care among individuals with significant depression symptoms. Overall, a better understanding of family-involved depression treatment in primary care settings and methods for improving social support with and without family involvement are needed. Future studies should aim to identify successful methods of engaging patients and their families in treatment for depression in primary care settings as well as to continue to explore how to support providers in working with families to improve patient outcomes.