Design
Secondary descriptive data analysis was conducted on baseline assessments from a diabetes prevention trial for Cambodian Americans with depression, clinicaltrials.gov identifier NCT02502929. Details of the methods have been reported previously [9]. Baseline data were collected from March 2016 to March 2019 on a rolling basis as enrollment in the trial continued.
Participants and sampling
All participants with baseline data were included in these analyses. The n = 186 resulted from a power analysis of the primary outcome of the trial (depression and HbA1c) [9]. Inclusion criteria were: (1) Cambodian or Cambodian-American; (2) Khmer speaking; (3) aged 35–75; (4) currently living in Connecticut, Rhode Island or Massachusetts (northeastern U.S.); (5) lived in Cambodia during 1975–1979 (the Pol Pot regime); (6) ambulatory; (7) take meals by mouth; and (8) elevated risk for diabetes per the American Diabetes Association Risk Test. Participants were also required to meet criteria for depression by (a) current antidepressant medication, and/or, (b) elevated depressive symptoms on the Khmer language Hopkins Symptom Checklist [10]. Exclusion criteria were: extant type 2 diabetes; seeing or hearing problems that would interfere with group sessions; major medical problems requiring intensive treatment; pregnancy or planning pregnancy; serious thinking or memory problems (e.g., schizophrenia or dementia); and 3 or more days in a psychiatric hospital or self-harm in the past 2 years.
Procedures
The study was conducted in accord with the principles of the Declaration of Helsinki and approved by the UConn Heaith institutional review board. Participants signed informed consent forms in their preferred language (Khmer or English). Bilingual, bicultural community health workers (CHWs) conducted surveys. Participants were paid $10 each in gift cards to a local pharmacy for completing the surveys and HbA1c assessment.
Measures
Demographic and clinical characteristics
Participants reported their age and sex. They also reported their ability to speak and read English, each on a 4-point scale from 0 = “not at all” to 4 = “very well”. HbA1c is the gold standard measure of glycemia. HbA1c was assayed at Quest laboratory and values are reported in percent units according to the National Glycohemoglobin Standardization Program (NGSP).
Social determinants
Healthcare system factors
Participants were asked how often in the past year they experienced difficulty communicating with their healthcare provider because of a language difference using a 5-point scale from “never” to “always” with higher scores indicating greater difficulty. They were asked a similar question referencing their pharmacist. Participants reported insurance status (Medicare, Medicaid vs private).
Educational attainment
Participants were asked, “How many years of education do you have?”.
Early adverse experiences
Participants were asked to report the number of years they had lived in the Khmer Rouge regime, and how many years they had lived in a refugee camp.
Income
Given the generally low household income of the population, income was categorized as < $20,000, $20,000-$30,000, $31,000-$40,000 or > $40,000.
Work
Participants were asked to choose their employment status: full time, part time, retired, homemaker, disabled, unemployed looking for work, or other.
Food
Food insecurity was assessed using the 6-item U.S. Household Food Security Survey (HFSS) module [11] using a 3-month time reference. The sum of affirmative responses produces a scale score (0–6). Higher scores indicate greater food insecurity.
Transportation
We asked two questions including, “can you drive?” and “do you have access to a car?” Responses options were 0 = ”no” and 1 = ”yes”; they were summed for a transportation score from 0–2.
Stress
We measured symptoms of depression with the 15-item depression subscale of the Khmer language version of the Hopkins Symptom Checklist [10]. We assessed symptoms of Post-Traumatic Stress Disorder (PTSD) with the 16-item symptom subscale of the Khmer language version of the Harvard Trauma Questionnaire [10].
Social support
Participants responded to four items from the Patient-Reported Outcomes Measurement Information System (PROMIS) test bank [12] and one item from the Enriched Social Support Instrument [13]. Response options were on a 5-point scale with higher scores indicating greater social support.
Social isolation
One item from PROMIS test bank [12] asks about “feeling isolated from other people”. Response options were on a 5-point scale with higher scores indicating greater isolation.
Addiction/substance use
Participants who endorsed drinking alcohol in the past year [14] went on to answer the four-item substance use subscale of the COPE [15] which assesses drinking alcohol as strategy to cope with distress. Response options are on a four-point scale from “I usually don’t do this at all” to “I usually do this a lot”.
Patient reported outcome variable
Pain. We measured past week pain occurrence and interference. Participants were asked, “Did you experience pain in the past 7 days?” Response options were “yes = 1” or “no = 0”. Participants who responded “yes” were asked, “Did pain make it difficult for you to do your day-to-day activities?” Participants who responded “yes” were asked, “How difficult did pain make it to complete your day-to-day activities?” Response options were “a little bit difficult = 1”, “moderately difficult = 2” or “very difficult = 3”. Two questions from the Pittsburg Sleep Quality Index (PSQI) [16] asked: “During the past month, have you had trouble sleeping because you have physical pain?” (yes = 1 or no = 0); affirmative responses were followed up with “How often?” with response options ranging from rarely = 1 to very frequently = 3. All responses were summed and pain scores could range from 0 (no pain) to 8 (pain that makes daily activities very difficult and with frequent trouble sleeping).
Statistical analysis
The pain score ranged from 0 to 8 with excess zeros (37%) and so a generalized linear model with negative binomial distribution and logit link was used to model predictors of pain. Bivariate analyses were performed for each predictor separately followed by a backward multivariate model that started with all significant bivariate predictors and removed non-significant predictors one at a time until only significant predictors remained. The analyses were conducted in SPSS v27.