Study design and context
This paper reports on a cohort study within a cluster randomised controlled trial (RCT), including cross-sectional baseline data and longitudinal data on changes from baseline to follow up. The aim of the RCT was to evaluate the effectiveness of the Primary Care 101 guideline training programme for primary health care providers [18, 19], and to assess whether the programme improved quality of care for specified chronic diseases. Primary Care 101 consists of three elements: a 101-page algorithmic guideline that covers common symptoms and conditions in adults; an educational outreach programme in which nurse trainers deliver interactive training sessions on-site to all staff at a facility, using the Primary Care 101 guideline and case scenarios; and additional prescribing provisions for nurses who successfully complete their training.
Thirty-eight clinics in the Eden and Overberg districts of the Western Cape, South Africa, were cluster randomised either to receive the Primary Care 101 training programme for health care providers, or to continue with usual care. Eligible patients, defined below, who provided consent were interviewed at baseline in 2011 and once more, 14 months later [19]. The analyses for this study included data from the whole RCT cohort at baseline and follow-up, combining the intervention and control arms.
Study population and sample
The study population comprised adults attending public sector primary care clinics in two districts of the Western Cape province of South Africa. The communities served by the public sector clinics in these two districts are characterised by high levels of unemployment and socio-economic deprivation. In 2011, unemployment rates were 22.5 and 17.0 % in the Eden and Overberg districts respectively [20], and the Eden district was rated as the poorest in the Western Cape province [21]. The study site is typical of many low resource rural and small urban settings in South Africa, in which the public sector primary health care clinics are nurse-led with some doctor support.
Thirty-eight of the largest primary care clinics in the Eden district and two Overberg sub-districts were selected. Each clinic services at least 10 000 attendances per year and they are staffed by nurse practitioners, doctors and community health workers. The study population was restricted to adults 18 years or older, planning to reside in the area for the next year, and capable of actively engaging in an interviewer-administered questionnaire at the time of recruitment.
Among patients who met these criteria, four groups representing patients with hypertension, diabetes, chronic respiratory disease and depression were identified. Patients were eligible for the hypertension and diabetes groups if they reported being on medication for hypertension or diabetes respectively. They were eligible for the respiratory group if they reported being on medication for chronic respiratory disease, or had symptoms of chronic respiratory disease and were not on treatment for tuberculosis. Patients were eligible for the depression group if they scored ten or more on the 10-item Centre for Epidemiologic Studies Depression Scale (CESD-10) [22]. Patients may have fulfilled inclusion criteria for more than one disease group. Participants were sampled consecutively within each clinic and invited to participate in the study, until the sample size required for each clinic was obtained. They were screened for eligibility by orally questioning them and, if they met the eligibility criteria, were then asked to provide informed consent to participate.
Data collection and coding
At baseline trained fieldworkers administered the electronic questionnaire and took clinical measurements after eligible participants provided informed consent. The baseline questionnaire included questions about demographic characteristics, comorbidities, and socio-economic factors. Participants were asked about the highest level of education they had achieved (no schooling, grade 1–7, grade 8–12 or tertiary/diploma), their employment status (employed, self-employed, student/learner or unemployed), and their employed and pension/grant income in the last month.
The presence of depression symptoms was assessed with the 10-item CES-D scale which was administered to all participants. The 20-item CES-D was originally developed by Radloff (1977) to measure symptoms of depression in the general population [23, 24]. A shortened 10-item version was created by Andresen et al. [22] The CESD-10 items are: “1. I was bothered by things that usually don't bother me. 2. I had trouble keeping my mind on what I was doing. 3. I felt depressed. 4. I felt that everything I did was an effort. 5. I felt hopeful about the future. 6. I felt fearful. 7. My sleep was restless. 8. I was happy. 9. I felt lonely. 10. I could not get going.” The individual items are scored from 0 (rarely or none of the time) to 3 (most of the time) and a score is assigned by totalling all item scores. The possible range of scores is 0–30 for the 10-item scale, with higher scores representing greater degrees of depressed mood [22]. Both the 10- and 20- item CES-D have been used and validated in a number of countries including among HIV infected individuals in South Africa [25, 26].
All participants were asked if they had received psychological counselling in the year leading up to their baseline interview. Counselling was defined as talking with someone in a way that helps to find solutions to problems, or receive emotional support, and not just receiving advice on how to take medication. Participants who reported receiving counselling from a mental health nurse, clinic counsellor, social worker, psychiatrist or psychologist were considered to have received counselling. Participants who reported receiving counselling from a mental health nurse, psychiatrist or psychologist were considered to have been referred to psychiatric services.
Chronic medication prescribed at the time of each participant’s interview for depression, hypertension, diabetes and respiratory disease was recorded. Fieldworkers photocopied all available prescription charts for the year preceding the interview. The trial manager (NF) analysed the prescription charts to identify medication for chronic conditions prescribed for each participant at the time of their interview.
It is common practice in the Eden and Overberg districts for amitriptyline or imipramine to be prescribed at a low dose (25 mg daily) for pain management and insomnia. We considered amitriptyline and imipramine at a dose less than 50 mg daily to be sub-therapeutic for depression [6]. Other antidepressants were not prescribed at sub-therapeutic doses [27]. We therefore defined being on an antidepressant at a therapeutic dose as prescription of amitriptyline or imipramine of 50 mg or more daily, or on any other antidepressant.
Disease-specific control indicators were measured at baseline and follow-up [19]. Systolic and diastolic blood pressure were measured in all participants. Ten year risk of cardiovascular deaths was calculated, based on age, sex, systolic blood pressure, smoking status, reported diabetes and body mass index [28]. The severity of respiratory disease was assessed with the Symptom and Activity domains of the St Georges Respiratory Questionnaire (SGRQ) [29] in participants enrolled in the respiratory disease group. Glycated haemoglobin (HbA1c) was measured in a sub-sample of 704 diabetic participants from 20 randomly selected clinics.
The following clinic characteristics were identified at baseline: availability of a pharmacist, availability of drug supply away from clinic, psychiatric nurse at clinic, doctor at clinic every day, clinic location, clinic patients per year, clinic patients per nurse per year, and intervention versus control clinic.
At follow-up the questionnaire, clinical measurements and prescription data were collected and recorded as for the baseline data. Baseline data collection began in March 2011 and ended in October 2011. Follow-up data collection started in May 2012 and ended in January 2013.
Statistical methods
The statistical analyses investigated associations between participants’ health and socioeconomic indicators, and their symptoms and treatment of depression. We also investigated associations between depression symptoms reported at baseline and subsequent changes in participants’ income and employment, ten year risk of death from cardiovascular disease and, in participants with hypertension, diabetes, or respiratory disease, in blood pressure control, glycaemic control and respiratory symptoms respectively. Analyses of treatments included the following clinic characteristics as potential explanatory variables: pharmacist in clinic, drug supply available away from clinic, psychiatric nurse at clinic, doctor at clinic every day, clinic location, clinic patients per year, clinic patients per nurse per year, and intervention versus control clinic. These clinic characteristics were investigated because they could potentially influence access to necessary treatment directly, or be indirect indicators of the quality of care.
In all analyses the study’s cluster sampling design was accounted for in regression models with robust adjustment for intra-clinic cluster correlation of outcomes, using Stata version 12.0 statistical software [30]. A p value 0.05 or less was considered statistically significant.
Intervention or control arm of the randomised controlled trial was accounted for in all longitudinal analyses. Variables independently associated with the outcome in each model were selected using backwards stepwise selection. At each step, explanatory variables with a p value of less than 0.10 were removed from each model. The purpose of stepwise selection of explanatory variables for each model was to estimate the effects of each socioeconomic indicator without confounding by other socioeconomic indicators or patient characteristics. Even though all of the socioeconomic indicators could theoretically have causally influenced depression and its care, it was not appropriate to keep all of them in every model because of the likelihood that overadjustment for collinear variables would obscure relevant associations.
The primary analyses of variables associated with depression symptoms were multiple linear regression models with CESD-10 score as the continuous outcome variable. Secondary analyses of depression symptoms used multiple logistic regression models with CESD-10 scores coded as high (greater than or equal to ten) or low (less than ten).
Analyses with treatments (antidepressant medication, counselling, or referral to psychiatric services) as outcomes were confined to participants with CESD-10 scores greater than or equal to ten at baseline and used multiple logistic regression models. Primary analyses of antidepressant medication coded treatment as present only if drug doses were defined as therapeutic. Secondary analyses coded antidepressant treatment as present at any dose.
Longitudinal data analysis was as follows. Changes between baseline and follow-up in depression symptoms, antidepressant medication, employment or income used analysis of covariance (ANCOVA) in the multiple regression models, that is, with the follow-up variable as outcome and with the baseline variable as a covariate. This was done to account for regression to the mean, that is, individuals with exceptionally high or low values at baseline would at follow-up tend to have values closer to the mean, due to chance alone [31]. Analyses of changes all included trial arm as a potential explanatory variable.
The trial is registered with Current Controlled Trials (ISRCTN20283604). Ethical approval for the trial was obtained from the University of Cape Town Human Research Ethics Committee and the Western Cape Provincial Department of Health. All participants provided informed consent to participate in the study.