Technical Appendix for Improving Treatment of Depression in a Low-Income
and Ethnically Diverse Population of Patients with Diabetes
Using the IMPACT Model and Project Dulce
This technical appendix provides additional information on the methods used in the study “Improving Treatment of Depression in a Low-Income and Ethnically Diverse
Population of Patients with Diabetes Using the IMPACT Model and Project Dulce.” Specifically, we provide an overview of the integration of the of Project Dulce and IMPACT programs, address study sample selection, and provide additional information regarding our analysis of diabetes self-care activities.
Overview of Project Dulce + IMPACT
Project Dulce employs a team consisting of a registered nurse/certified diabetes educator (RN/CDE), bilingual/bicultural medical assistant (MA), a bilingual/bicultural dietitian, and trained peer educators (or promotoras) to deliver diabetes care in the community clinic setting. The clinical goals are to meet the American Diabetes Association standards of care and to achieve improvements in A1c, blood pressure, and lipid parameters: A1c < 7%, BP < 130/80 mm Hg, and LDL < 100 mg/dL. Patients have an initial (50 minute) visit with a nurse and are asked to return for additional (50 or 25 minute, as required) visits with the nurse and for a (25 minute) visit with a dietitian. On average, participants have 5 visits with the nurse and 50% consult with a dietitian. In addition to these one-on-one visits, patients are encouraged to participate in an 8-week group self-management training program delivered by trained peer educators (or promotoras): approximately 50% participate in the class and average attendance is 4 classes. The peer educators are recruited from the patient population, have diabetes themselves, and are of the same cultural/ethnic group as the participants. They complete a 4-month, competency-based training and mentoring program. Classes are taught in the patient’s native language, and cover diabetes and its complications, the role of diet, exercise, and medication, and the importance of self-monitoring. The classes are collaborative, including interactive sessions in which the patients discuss their personal experiences and beliefs about diabetes. An emphasis is made to overcome misrepresented cultural beliefs and to encourage patients to take charge of managing their disease.
Project Dulce adapts easily to diverse cultures through the use of their peer educators, who are patients from the targeted ethnic group with diabetes who demonstrate leadership skills and have taken responsibility for their health. Project Dulce has been providing diabetes care and education in San Diego County since 1997. To date over 5000 patients have been served by the program, and the project has become a viable part of the community’s health care safety net. Research was conducted during the inception of Project Dulce (Philis-Tsmikias et al., Diabetes Care, 2004) to measure the impact of the program on culture-based beliefs, knowledge, and locus of control among the primarily Hispanic participants. The project demonstrated significant improvements in treatment satisfaction, diabetes knowledge, health locus of control (empowerment) and awareness of the efficacy of certain culture-based remedies, e.g., “eating nopales can cure diabetes.”
In this translational pilot study, we augmented the Project Dulce program with an IMPACT model of depression care management. Project Dulce participants were screened for depression using the PHQ-9. Those identified with clinically significant depressive symptoms (defined as a PHQ-9 score of 10 or greater) had a visit scheduled with the depression care manager. The depression care manager conducted a clinical and psychosocial history, reviewed educational materials with the patient, and discussed patient preferences for depression treatment with antidepressant medications and / or structured individual or group psychotherapy using Problem-Solving Treatment in Primary Care (PST-PC). PST-PC is a 6-8 session, brief, structured psychotherapy for depression. All patients received education about depression and behavioral activation. Problem solving therapy involves the patient identifying problems that may be contributing to their depression and developing an action plan to address one problem at a time. Behavioral activation involves encouraging the patient to participate in one enjoyable activity between sessions. Both approaches motivate patients to commit to taking concrete steps to control how they feel. The depression care manager consulted with the diabetes case manager and the patient’s primary care physician. New patients and patients needing treatment plan adjustments were discussed in weekly caseload review meetings with a consulting psychiatrist.
A number of specific issues were identified that needed to be addressed to make the IMPACT model culturally sensitive. A bilingual and bicultural social worker was employed who had extensive experience working with the Latino population in a counseling role, and who focused on cultural sensitivity in her graduate education. The depression program was integrated into Project Dulce, and benefited from the team approach and involvement of the peer educators (promotoras). All materials were translated into Spanish and adapted to a population with very low literacy. The delivery of care was made more flexible to respond to cultural norms as well as socio-economic barriers (for example, lack of transportation was a significant barrier to maintaining an “on-time” appointment schedule). The intervention addressed cultural beliefs regarding health care (for example, that doctor is the ultimate authority and not to be questioned). There was also a strong cultural stigma related to mental illness in this population. Basic education about depression, stress, and anxiety had a significant impact by helping patients to understand what contributes to depression, and how they could take control of how they feel through by action (empowerment) and self-responsibility.
Study sample selection
As discussed in the report, we screened participants in a diabetes management program for depression using the PHQ9. Sample selection is shown in Appendix Table 1. We screened 499 participants in Project Dulce for depression using the PHQ9: 464 (93%) were eligible to participate in Project Dulce + IMPACT; 154 (33%) scored as having symptoms of major depression using the PHQ9 (10+); and 99 (64%) completed the study. The only demographic characteristic was related to study participation: among those eligible to participate in the study, females were more likely to be identified as having symptoms of major depression than males (OR 1.9, P<.001). No other covariates were significant.
We conducted an additional analysis to address the effect of the intervention on those who dropped out, shown in Table 2. Depression symptoms were assessed using the PHQ9 at each visit. We modeled PHQ9 scores over the study period as a quadratic function of days in the study, controlling for baseline PHQ9 score, and including interaction terms with days in the study for those not completing the six-month follow-up. An F-test of the hypothesis that the parameter estimates on these two interaction terms was zero yielded an F-statistic (2,839) of 0.93, P=.395.
Table 1: Study Participant SelectionScreened / Not Depressed / Depressed / Completed Study
N / 499 / 310 / 154 / 99
Age Mean, SD / 52 / 12 / 52 / 12 / 52 / 11 / 53 / 9
Female N, %* / 341 / 68.3 / 192 / 61.9 / 123 / 79.9 / 83 / 83.8
Ethnicity N, %
Latino / 367 / 73.5 / 238 / 76.8 / 115 / 74.7 / 73 / 73.7
Non-Latino White / 82 / 16.4 / 40 / 12.9 / 26 / 16.9 / 18 / 18.2
African American / 20 / 4.0 / 8 / 2.6 / 9 / 5.8 / 6 / 6.1
Asian / 7 / 1.4 / 6 / 1.9 / 1 / 0.6 / 1 / 1.0
Other / 23 / 4.6 / 18 / 5.8 / 3 / 1.9 / 1 / 1.0
Primary Language N, %
Spanish / 343 / 68.7 / 226 / 72.9 / 107 / 69.5 / 70 / 70.7
English / 153 / 30.7 / 66 / 21.3 / 44 / 28.6 / 27 / 27.3
Other / 21 / 4.2 / 18 / 5.8 / 3 / 1.9 / 2 / 83.8
Note: Females were 1.9 times as likely to score as depressed: PHQ9 = 10+ (P<.001)
Table 2: Regression Predicting PHQ9 Scores
Source | SS df MS Number of obs = 845
------+------F( 5, 839) = 66.86
Model | 10339.7203 5 2067.94405 Prob > F = 0.0000
Residual | 25948.0478 839 30.9273514 R-squared = 0.2849
------+------Adj R-squared = 0.2807
Total | 36287.768 844 42.9949858 Root MSE = 5.5612
------
sumtotal | Coef. Std. Err. t P>|t| [95% Conf. Interval]
------+------
phqbase | .6082808 .0434136 14.01 0.000 .5230687 .6934928
days | -.0584357 .0107966 -5.41 0.000 -.0796272 -.0372443
days2 | .0001409 .0000587 2.40 0.017 .0000258 .000256
ns_days | -.0159083 .0253648 -0.63 0.531 -.0656943 .0338777
ns_days2 | .0001974 .0001979 1.00 0.319 -.000191 .0005858
_cons | 5.309063 .7456973 7.12 0.000 3.845412 6.772715
------
Note: days = days in study, and days2= days in study squared; ns_days and ns_days2 are interaction terms with days and days2 for participants who did not complete the six month follow-up. An F test of the two resulting parameter estimates yielded these results:
( 1) ns_days = 0
( 2) ns_days2 = 0
F( 2, 839) = 0.93
Prob > F = 0.3953
Diabetes self-care activities
We assessed diabetes self-care activities using Project Dulce’s Summary of Self Care Activities Survey (see Appendix 2). The survey includes five items related to nutrition, three items related to exercise, and four items related to glucose monitoring and medication adherence. We reduced the set of questions on glucose monitoring and medication adherence to three items by using the question for adherence to insulin users only for those reporting insulin use and the question for adherence to oral medications for those not reporting insulin use. We reversed the items as appropriate.
We calculated a Cronbach’s alpha statistic for each of these three sets of items. We also calculated composite scales and the Cohen’s d on the difference in composite scales between the baseline and follow-up. T-tests were used to assess the significance of these differences.
The composite scales for nutrition, exercise, and glucose monitoring / medication adherence behaviors had scale reliability coefficients of 0.58, 0.89, and 0.76, respectively. A statistically significant difference was found only in diabetes self care activities related to nutrition, which had a Cohen’s d statistic of 0.26 (P=.0187).
1