EFFECTIVENESS OF SHARED DECISION-MAKING FOR ELDERLY DEPRESSED MINORITY PRIMARY CARE PATIENTS

Abstract

Objective:

We assessed the impact of a Shared Decision-Making intervention among elderly depressed minority primary care patients not currently receiving treatment.

Methods:

202 English and Spanish-speaking primary care participants aged 65+ who scored positive on the PHQ-9 (≥10) were randomized at the physician level to receive a brief Shared Decision-Making intervention or Usual Care. Primary analyses focused on patient adherence to either psychotherapy or antidepressant medication, and reduction in depression severity (HAM-D) over 12 weeks.

Results:

Patients randomized to physicians in the Shared Decision-Making condition were significantly more likely than patients of physicians randomized to Usual Care to receive a mental health evaluation or initiate some form of treatment (39% vs. 21%), and to adhere to psychotherapy visits over 12 weeks. There were no differences between groups in adherence to antidepressant medication or in reduction of depressive symptoms.

Conclusions:

Among untreated elderly depressed minority patients from an inner-city municipal hospital, a brief Shared Decision-Making intervention was associated with greater initiation and adherence to psychotherapy. However, low treatment adherence rates across both groups and the intervention’s lack of impact on clinical outcomes highlight the need to provide focused and accessible mental health services to patients choosing active treatments.

Major depression is a common but treatable mental health condition among older adults (1, 2). The majority of older adults who receive treatment are seen in primary care settings; however, treatment rates remain low particularly among racial and ethnic minorities (3–6). Involving patients more actively in shared decision-making (SDM) might promote uptake and adherence to depression treatment (7).

SDM has been proposed as a primary strategy to achieve good patient-centered care (8, 9). SDM involves a collaborative process whereby patients articulate personal values and preferences and clinicians provide information such that both participants may arrive at mutually-agreed upon treatment decisions (8,10,11). SDM interventions, typically one-session encounters, increase patient involvement in decision-making, knowledge of treatment options, and proportion of patients who make treatment decisions (12–14). Despite these immediate benefits, research is needed to determine the impact of SDM on treatment adherence and health outcomes.

We propose that SDM may be beneficial to elderly depressed minority individuals, as SDM enhances autonomy and empowerment in a manner that directly addresses cardinal features of depression such as helplessness and hopelessness (7). SDM strategies have been shown to benefit individuals with mental health conditions (15), and one study demonstrated that SDM can engage depressed patients in primary care settings (16). SDM interventions, however, have yet to be adequately tested, particularly among older depressed adults and among racial and ethnic minorities.

Several factors must be taken into account when designing SDM interventions for elderly depressed minority primary care patients. First, in contrast to younger patients, older adults report fewer experiences of SDM in their care (17). Ethnic minority and immigrant populations may experience cultural and linguistic barriers to SDM, and have reported receiving less information about their condition and less respect for their treatment preferences (18–20). Second, older adults may exhibit greater acceptance of the physician’s traditional dominant decision-making role, and greater medical burden limiting their ability to consider treatment alternatives (7). Third, stigmatic concerns, negative beliefs about depression, and tangible barriers such as transportation and travel distance may limit treatment adherence (7).

We therefore developed an SDM intervention for untreated elderly depressed, minority primary care patients and tested its effectiveness in improving both treatment adherence and clinical outcome. We hypothesized that depressed patients of physicians randomized to SDM would show: 1. greater initiation and adherence to antidepressant medication or psychotherapy; and 2. greater reductions in depression severity (Hamilton Depression Rating Scale; HAM-D) (21) over 12 weeks, in comparison to depressed patients of physicians randomized to Usual Care (UC).

METHODS

Overview:

The study was conducted at Lincoln Medical and Mental Health Center, part of NYC’s Health and Hospital Corporation. Lincoln is an acute care inner-city public benefit hospital in South Bronx, and provides services to a racially and ethnically diverse population. Outpatients in the general medical and geriatric subspecialty units were eligible for recruitment. The study was approved by the Weill Cornell Medical College and Lincoln IRBs. All participants provided written informed consent. Recruitment ran 4/2010–11/014.

Study Participants:

Eligibility criteria included: age≥65 years; English or Spanish-speaking; scoring≥10 on medical staff or RA-administered PHQ-9 (22); not receiving antidepressant medication or psychotherapy within past month; and able to give consent. Exclusion criteria included: bipolar, psychotic, or current substance abuse disorders via Structured Clinical Interview for Axis I DSM Disorders (SCID) (23); or dementia according to medical records. We chose Mini-Mental State Exam (MMSE) (24) scores ≥20 to account for the populations’ low education level (25).

Randomization:

We randomized physicians to: (a) SDM, whereby their patients were provided access to nurse-administered SDM; or (b) Usual Care (UC), whereby physicians engaged patients in depression treatment decisions as part of routine care. Consent forms for both groups described the study’s purpose as investigating whether SDM about depression treatment leads to improved outcomes, and encouraged physicians to recommend some type of treatment to eligible patients. Physicians in both conditions were provided guidelines on pharmacological treatment of depression. Opportunities for mental health care included physician medication evaluation and referral to hospital or community-based mental health clinics or depression groups. RAs were aware of randomization status following baseline assessment but not of study hypotheses.

Intervention:

Four registered nurses employed by Lincoln provided the manualized SDM intervention (7), under regular supervision by the PI. SDM consisted of a 30 minute in-person meeting followed by 2 weekly 10–15 minute telephone calls.

Nurses first discussed the patient’s depressive symptoms and provided psychoeducation. They elicited patients’ treatment experiences, preferences, and concerns regarding various treatment approaches. Nurses used decision aid materials to further clarify patients’ values by discussing the effectiveness, speed of onset, side effects, and costs associated with both antidepressant medication and psychotherapy. Nurses provided psychoeducational handouts for patients and family members to review at home. Nurses assisted with appointment scheduling and addressed practical barriers to care such as transportation as needed, or referred patients to in-house social work.

During follow-up calls, if patients encountered difficulty because of poor motivation, stigma, poor access, high cost, or lack of service availability, nurses attempted to address unresolved treatment barriers and re-engaged patients in SDM processes.

The PI conducted training in SDM by manual review, demonstrations, and role plays over one-month. Nurses were required to receive global adherence scores of 3 (“adequate”) to final role plays, according to the SDM Adherence Form that uses a 0–5 point scale evaluating SDM tasks described above. Intervention fidelity over the study’s course was assessed by a trained RA on 32 audiotaped sessions: 88% of these sessions received “satisfactory” or higher global ratings. Interrater reliability was assessed using Intraclass Correlation Coefficient based on 23 sessions rated jointly by another RA (ICC=0.81). The PI provided corrective feedback and weekly supervision to nurses.

Assessment and Outcomes:

RAs conducted baseline patient interviews shortly after physician visits. Participants whose physicians were assigned to SDM were interviewed prior to receiving the SDM intervention.

RAs assessed depression symptoms using the SCID (23), with diagnoses assigned by the PI after review of symptoms. Psychotic or manic symptoms, suicidal ideation, and substance abuse were reported to the participant’s PCP. The 24-item HAM-D (21) assessed depression severity, and the World Health Organization Disability Assessment Schedule II-12 item (WHODAS) (26) yielded a composite disability score. The Camberwell Assessment of Needs for the Elderly (CANE) (27) assessed social service needs. Medical morbidity was estimated using the Chronic Disease Score from prescribed medications, updated to the 2005 American Hospital Formulary (28). History of antidepressant and psychotherapy use was determined by self-report.

One week following in-person SDM interventions, or interactions with physicians for UC patients, we asked: “What was your decision about how to be treated for your depression at this time?” Options included antidepressant medication, psychotherapy/counseling, both approaches, various complementary approaches (e.g., herbal remedies, exercise, religious/spiritual activities), or “do nothing at this time.” We administered the Satisfaction with Decision Scale (29), a 7-item Likert scale yielding a total score (example item: “I am satisfied with the decision I made about treatment for my depression.”).

At 4, 8, and 12 week follow-up points, RAs administered the HAM-D and Cornell Service Use Index (31), which documented frequency and type of mental health services and antidepressant use.

Data Analyses:

Descriptive statistics for demographic and baseline clinical variables by randomization are reported in Table 1. We included all participants who completed baseline assessments in intent-to-treat analyses.

Table 1:

Patient Demographic and Baseline Sample Characteristics

Variable	Total (N=202)	SDM (N=114)	UC (N=88)	P-value	Stddiff#
Demographics
Age, Years	72.1±5.5	72.2±5.4	71.9±5.6	0.67a	0.061
Education, Years*	7.8±3.9	7.8±3.5	7.7±4.5	0.96a	0.008
Sex, Female	164(81.2)	92(80.7)	72(81.8)	0.84c	−0.029
Ethnicity, Hispanic	184(91.1)	101(88.6)	83(94.3)	0.16c	−0.206
Race				0.25c	0.242
White	111(55.0)	58(50.9)	53(60.2)
Black	59(29.2)	34(29.8)	25(28.4)
Other+	32(15.8)	22(19.3)	10(11.4)
Baseline Assessments
HAM-D	18.8±6.2	18.8±6.0	18.9±6.5	0.92a	−0.014
Anxiety Disorder	25 (12.4)	10 (8.8)	15 (17)	0.08d	−1.771
WHODAS	29.2±8.5	29.4±8.6	28.9±8.4	0.68a	0.058
MMSE	24.9±2.6	24.8±2.7	25.0±2.6	0.63a	−0.069
Medical Burden (CDS)*	6.1±3.1	6.1±3.1	6.1±3.1	0.88a	−0.022
CANE	6.0[4.0,8.0]	6.0[4.0,8.0]	6.0[4.0,8.0]	0.60b	0.075

^*Data not available for all subjects. Missing values: Education, Years = 1, Medical Burden (CDS) = 6.

Statistics presented as Mean ± SD, Median [P25, P75], Median (min, max) or N (column %).

p-values:

^a=ANOVA F-test,

^b=Kruskal-Wallis χ ²-test,

^c=Pearson’s χ²-test,

^d=Fisher’s Exact test.

Statistics presented as Mean ± SD, Median [P25, P75], N (column %).

^#|StdDiff|>0.278 indicates imbalance between the two treatment groups.

⁺Other Race: N=1, Asian (SDM); N=1, Other (SDM); N=30, Pacific Islander (10 SDM, 20 UC)

StdDiff = standardized differences

SDM = Shared Decision-Making; UC = Usual Care

HAM-D = Hamilton Rating Scale for Depression; higher scores indicating greater depression severity

WHODAS = World Health Organizational Disability Assessment Scale; higher scores indicating lower functioning

MMSE=Mini-Mental State Exam; higher scores indicating greater cognitive functioning

CDS = Chronic Disease Scale; higher scores indicating more severe medical burden

CANE = Camberwell Assessment of Need for the Elderly; higher scores indicating greater social service need

Primary analysis (1): Adherence.

Our primary analysis evaluated the effectiveness of SDM on treatment adherence among 103 SDM and 78 UC participants. Patients were not assigned treatment, but those in either the SMD or UC conditions could have pursued antidepressant medication, psychotherapy, or both. Thus, adherence was first operationalized at a composite level by the combined proportion of possible antidepressant pills taken (0–84 divided by 84 as an upper limit) and possible psychotherapy sessions attended (0–12 divided by 12 sessions as an upper limit) over 12 weeks. Thus, a single outcome expressed as a number ranging from 0–2 was used for each subject. To account for non-normality of data and physician-level correlation, we used a marginal generalized estimating equations (GEE) model which provides robust inference to large data with skewed distributions. To assess the effect of SDM (vs. UC) on adherence level, we used an identity link and exchangeable covariance structure to account for physician-level correlation.

Individual Adherence Components.

We assessed the effect of SDM on five components of adherence over 12 weeks: (1.1) initiation of any mental health care, including mental health evaluation, psychotherapy or antidepressant medication (any vs. none); (1.2) initiation of psychotherapy (any vs. none); (1.3) frequency of psychotherapy visits; (1.4) initiation of antidepressant medication (any vs. none); and (1.5) total number of pills taken. Components (1.3) and (1.5) are count variables and modeled using an overdispersed variant of negative binomial regression where the variance increases linearly with the mean. Since the majority of (1.3) and (1.5) were zeroes in both arms, we used a zero-inflated negative binomial regression model. A zero-inflated model is a two-part model which explicitly models two populations – one that is generated from the underlying distribution (negative binomial) which can have counts of 0, 1, 2, and so on, and an additional population of zeroes that adds to the zeroes of the first population. The second population reflects individuals who did not initiate mental health services or antidepressants. A zero-inflated model, therefore, explicitly models the source of extra zeroes due to lack of treatment initiation and provides the outcome count conditional on treatment initiation. A logistic mixed effects model was used with an overdispersion parameter and physician-level random intercept to account for physician-level clustering to estimate treatment effects (SDM vs. UC) for outcomes (1.1), (1.2) and (1.4).

We present odds ratios and frequency counts for (1.1), (1.2) and (1.4). For (1.3) and (1.5) the predicted probability of having additional zeroes (belonging to the second population above) determined the mean count conditional on belonging to the first population above (Table 3). We report the incidence ratio for the conditional model and the odds ratio for the zero-inflation model along with their CI.

Table 3:

Effectiveness of SDM vs. UC on adherence and depression symptom severity

	Treatment Group
	SDM (n=103)	UC (n=78)	Regression Estimate/ OR/ IRR (Adjusted CI)	Test-statistic	P-value
Primary Aim 1: Adherence
1. Adherence Level	0.18±0.03	0.10±0.03	0.08 (−0.02, 0.17)	z=1.77	0.154*
Sub aim analyses
1.1. Initiation of any mental health care (i.e. MH evaluation, psychotherapy or antidepressant medication). (Yes/No)	40 (38.8%)	16 (20.5%)	2.46 (0.92, 6.61)	t179=2.59	0.01
1.2 Initiation of psychotherapy (Yes/No)	26 (25.2%)	5 (6.4%)	4.93 (1.13, 21.48)	t179=3.08	0.002
1.3 Frequency of psychotherapy visits
Zero-Inflation probability:	0.57±0.07	0.87±0.05	0.20 (0.04, 0.97)	z=−2.86	0.0043
Conditional (on not being zero) mean:	1.30±0.24	1.20±0.50	1.08 (0.30, 3.88)	z=0.17	0.865
0	71 (68.9%)	71 (91.0%)
1	15 (14.6%)	3 (3.9%)
2	11 (10.7%)	3 (3.9%)
3	4 (3.9%)	1 (1.3%)
4	2 (1.9%)	0 (0%)
1.4 Initiation of antidepressant medication	24 (23.3)	12 (15.4)	1.98 (0.46, 8.44)	t157=1.7	0.194
1.5 Total number of pills taken
Zero-Inflation probability:	0.76±0.04	0.85±0.77	0.60 (0.20, 1.80)	z=−1.3	0.192
Conditional (on not being zero) mean:	48.39±6.5	39.64±8.0	1.25 (0.63, 2.35)	z=0.91	0.363
Primary Aim 2: Depression	SDM (n=114)	UC (n=88)
Treatment × T ime Interaction				F3,580=0.75	1.0*
Baseline	18.7±0.6	18.9±0.7	−0.12 (−2.81,2.57)
Week 4	12.9±0.8	13.6±0.9	−0.73 (−4.03, 2.57)
Week 8	12.4±0.8	11.7±1.0	0.70 (−2.74, 4.15)
Week 12	12.9±0.8	12.0±0.9	0.88 (−2.40, 4.17)
Treatment
Averaged over time	14.2±0.6	14.1 ±0. 7	0.2 (−1.8,2.1)	F1,580=0.03

SDM = Shared Decision-Making; UC = Usual Care;

^*=adjusted p-value for multiple comparisons.

Inflation of type I error due to multiple comparisons for two primary aims was handled using Bonferroni’s method, and overall α=0.025 used. Since sub-components of the composite adherence measure are dependent on the composite adherence measure, corrections to type I error assuming sub-components to be independent would result in highly conservative p-values. Therefore, significance of sub-components were not adjusted but evaluated at α=0.025. CIs of all estimates correspond to these α values.

Primary analysis (2): Depression.

Our second primary analysis evaluated effectiveness of SDM on depression symptom reduction among 114 SDM and 88 UC participants. Linear mixed modeling with physician-specific random intercepts, subject-level repeated measures (both modeled with unstructured covariance), and fixed effects for time, treatment, and time×treatment interaction assessed the effect of SDM on HAM-D scores measured at baseline and weeks 4, 8, and 12. If mean HAM-D trajectories did not differ between treatment groups (i.e., time × treatment interaction p-value > 0.025), interaction terms were removed from the model and a time-averaged treatment effect was tested.

Sample size analysis:

As we proposed two primary hypotheses, we applied a multiplicity-adjusted alpha level of 0.025 for each hypothesis to detect a clinical meaningful, standardized effect size of 0.5. Incorporating within-physician, within-subject outcome ICCs in power analyses, 100 subjects per group would provide adequate statistical power for both primary hypotheses. Planned total enrollment for the study was 210, and we completed recruitment of 202 participants.

RESULTS

Participant Flow:

The study consented 45 of 46 eligible PCPs, and randomized 22 to SDM and 23 to UC. The study screened 245 patients who met preliminary eligibility criteria: PHQ-9≥10; English or Spanish-speaking; and under the care of participating physicians (Fig 1). Of these, 202 (82%) met selection criteria. These 202 patients were under the care of 24 PCPs who consented and were randomized: 114 were under the care of PCPs assigned to the SDM condition and 88 were under the care of PCPs assigned to UC. Twelve-week follow-up data on treatment adherence was available for 103 SDM and 78 UC participants, and on HAM-D ratings for 84 SDM and 66 UC participants.

Figure 1:

Consort Chart for Cliuical Trial of Shared Decision-Miking

Participant Characteristics:

Participating physicians had a mean age of 46 (sd=12.6), with 50% being female. Twenty-one percent classified themselves as Hispanic. Racially, 17% classified themselves as Black or African-American, 33% as Asian, 12% as White, and 21% as other race, while 17% did not report race. The 24 PCPs who saw eligible study patients were significantly older (t=−4.9(38), p<0.000) and had more years of experience (t=−4.6(39), p<0.000) than the 21 PCPs who did not see eligible patients. The 4 participating female nurses had a mean age of 52 (sd=6.0). Three of these nurses classified themselves as bilingual Hispanic, and 1 as African-American.

Patient participants (N=202) were aged 72 (sd=5.5; range=65 to 88), and 164 (81%) were female (Table 1). 184 patients (91%) classified their ethnicity as Hispanic. Racially, 59 (29%) classified themselves as Black or African-American. Mean education was 7.8 years (sd=3.9), and most reported being on Medicaid (148/197; 75%). Almost all participants met criteria for major (144; 71%) or minor depression (55; 27%) as assessed by the SCID, and they experienced a range of depression severity (HAM-D mean=18.8, sd=6.2). Twenty-five (12.4%) met criteria for concurrent anxiety disorders. Participants reported multiple chronic medical conditions (CDS mean=6.1, sd=3.1), disability scores in the mild severity range (WHODAS mean=29.2, sd=8.5), and an average of 5.2 (sd=3.1) social service needs (CANE). By protocol design, no participant was receiving depression treatment at study admission; 79 (39%) reported a prior history of depression treatment. There were no significant differences (31) in demographic or clinical variables between the randomized groups (Table 1).

Treatment Decisions:

The treatment modality chosen most frequently by both study groups, one week following the SDM intervention or UC clinical encounter, was psychotherapy (58% of SDM patients and 37% of UC patients; Table 2). We created a dichotomous variable indicating an active (medication or psychotherapy) or non-active treatment choice (complementary strategy or “do nothing”). Chi-square analysis indicated that SDM patients were more likely to choose an active approach (76%) following the clinical encounter than were UC patients (53%) (χ²(1)=9.35; p=0.002). Mean ratings on the Satisfaction with Decision Scale, administered 1 week following the SDM intervention or clinical encounter, did not differ across SDM (mean=3.06, sd=0.18) and UC patients (mean=3.1, sd=0.34, t=.09, df=166, p=0.93).

Table 2.

Treatment Choices of SDM and UC Patients^*

Treatment Choice	SDM (n=98)		UC (n=70)
	N	%	N	%
Psychotherapy/Counseling	57	58.2%	26	37.1%
Antidepressant Medication	7	7.1%	8	11.4%
Both Psychotherapy and Antidepressants	10	10.2%	3	4.3%
“Active” Treatment+	74	75.5%	37	52.9%
Complementary approach (herbal remedies, exercise, religious or spiritual activities)	11	11.2%	17	24.3%
“Do nothing at this time”	13	13.3%	16	22.9%
“Non-Active” Treatment	24	24.5%	33	47.1%

^*Data not available 16 SDM and 18 UC participants.

⁺χ²-test =9.35; df=1; p=0.002

SDM = Shared Decision-Making

UC = Usual Care

Primary analysis (1): Adherence Composite.

The majority of both SDM (61%) and UC (80%) conditions had a composite adherence measure of zero. The SDM arm had a significantly higher (χ²(1)=6.94, p=0.01) frequency of scores above the median compared to UC as determined by Mood’s median test. After accounting for physician-level correlation in a GEE model, the estimated composite adherence level was 0.18±0.03 among the SDM group and 0.10±0.03 for UC, which was not statistically different from 0 (Table 3).

Individual Adherence Components.

Patients in the SDM group had significantly higher odds of initiating any mental health care (MH evaluation, psychotherapy or antidepressant) within 12 weeks than patients receiving UC. Rates of treatment initiation were 39% vs. 21%, respectively. Examining specific treatments, the SDM group had significantly higher odds of initiating psychotherapy sessions than UC patients, but not of initiating antidepressant medication. SDM participants had significantly lower odds of belonging to the excess zeroes group in zero-inflated analyses, indicating that SDM participants were more likely to use psychotherapy services over time. Total number of pills taken over 12 weeks did not differ by study condition.

Primary analysis (2): Depression Status.

There was no difference in mean HAM-D score by treatment group over time (Table 3). Averaging treatment effect over time, we found no difference between SDM and UC conditions on depression status. Both SDM and UC patients experienced mild depressive severity at week 12 (Table 3).

CONCLUSIONS

Shared decision-making is a dynamic process through which patients and their providers jointly formulate a treatment strategy that aims to be both clinically effective and acceptable to patients. This RCT sought to determine SDM’s utility with a depressed elderly minority population for whom treatment is often inadequate or even absent. Could PCPs, nurses, and their older depressed patients discuss treatment options and formulate treatment plans that patients would subsequently implement?

The study’s principal finding was that an SDM intervention, in comparison to physicians’ Usual Care, almost doubled rates of mental health treatment initiation over a 12-week period (39% vs. 21%, respectively) (Table 3). While our composite measure of adherence to either psychotherapy or antidepressant medication failed to yield significant difference across treatment groups, participants in the SDM condition were more likely to initiate and adhere to psychotherapy visits than were UC patients. Our findings indicated that SDM patients reported modest treatment adherence over the 12-week course: 14.6% attended one individual or group psychotherapy session, and 16.5% attended between two and four sessions (Table 3). There were no differences between treatment groups in initiation or adherence to antidepressant medication. Finally, with regard to reducing depressive symptoms, both the SDM and UC groups experienced clinically significant reductions in depression severity (i.e., 6–7 points on the HAMD) at 12 weeks.

Our findings are consistent with the few prior reports on SDM interventions for depression. A German study of mid-life primary care patients found that SDM improved physician-reported patient participation in care, but did not reduce depression severity compared to usual care (16). Another RCT focused on mid-life primary care patients who were considering treatment with antidepressants found that SDM improved engagement in decision-making, but had no impact on medication adherence or depression severity (32). No prior studies have examined SDM for minority patients.

The modest effect SDM had regarding psychotherapy adherence, and its lack of effect on medication adherence or depression outcomes, may have been influenced by the substantial social service needs of participants in both groups. Prior research has documented that unresolved social stressors can be a barrier to adequate depression treatment (33) and that case management alone can improve depression symptoms in low-income older adults (34). While we did not assess case management services or resolution of social service needs, such services that participants in both groups may have received could have accounted for observed improvements in depressive symptoms. We suggest that future research examine these issues using appropriate follow-up assessments. Another factor that may have played a role in our observed reductions in depressive symptoms across groups was the possibility of a Hawthorne effect for participating physicians or patients. Consented physicians were encouraged to offer depression treatment to eligible patients. And patient participation in an RCT with careful attention to informed consent, in some ways a form of SDM, may have mitigated against differential effects.

Our modest findings also suggest that SDM interventions and outcomes are highly related to the capacity of primary care practices to provide patient-selected services. While most patients in the current study preferred psychotherapeutic services (Table 2), consistent with prior studies of depressed older patients (35–37), only 25% initiated this intervention (Table 3). We speculate that this modest participation rate may be due in part to significant efforts patients needed to take to obtain their preferred care. The hospital’s mental health clinic and local community mental health centers had waiting lists and prioritized services to the severely mentally ill. Considering the lack of motivation and energy experienced by many patients with depression, facilitating their access to community-based services is a vital aspect of the treatment process. This reality is particularly true for ethnically and racially-diverse elderly persons with little prior experience negotiating access to psychotherapy. Our findings in this regard raise the question of whether SDM should be undertaken in settings where it generates patient preferences for unavailable treatments.

Design limitations include a short-term 12-week follow up period, and inability to blind RAs to randomization status following baseline assessments. Our findings nonetheless suggest that PCPs and nurses can improve quality of patient-centered care for depressed persons by using a brief SDM intervention. These findings are consistent with the recent American College of Physicians recommendations for improving physician-patient partnerships (38). SDM compared to UC can increase the proportion of patients who select active treatments for depression, who initiate mental health treatment, and who adhere to it over time. SDM, however, had no impact on reducing the depressive symptom severity. We speculate that limited patient access to mental health services in our study’s community affected this outcome. Further research can determine whether SDM influences depression outcomes when such services are more readily available, such as in the context of integrated depression care management. Also of interest is our study’s finding that nurses working in an inner-city hospital and lacking prior mental health experience nevertheless can effectively participate in a depression-specific SDM intervention. This observation is fully consistent with earlier reports (39) demonstrating the vital role played by nurses when they collaborate with PCPs in caring for depressed patients.

Highlights.

The primary question addressed by the study was whether a Shared Decision-Making intervention for elderly depressed minority primary care patients would lead to improved adherence to mental health care and to reduction in depression severity.

The main finding of this study was that patients receiving Shared Decision-Making were significantly more likely than patients receiving Usual Care to initiate some form of treatment and to adhere to psychotherapy visits over 12 weeks. There were no differences between groups in adherence to antidepressant medication or in reduction of depressive symptoms. A brief Shared Decision-Making intervention can help untreated elderly depressed minority patients arrive at treatment decisions they are more likely to implement than those receiving a physician’s usual care.