Healthy Options: study protocol and baseline characteristics for a cluster randomized controlled trial of group psychotherapy for perinatal women living with HIV and depression in Tanzania

This study is being conducted within three urban districts in Dar es Salaam, Tanzania: Ilala, Kinondoni, and Temeke. According to the 2015 Demographic and Health Survey (DHS), 96% of women aged 15–49 in Dar es Salaam had attended school, with 44% completing secondary school. Sixty-six percent of women were employed, with the most common employment being unskilled manual labor and domestic service [31]. The prevalence of HIV among pregnant women in Dar es Salaam has been estimated as 6.6% [32]. Study sites are government-run reproductive and child health (RCH) clinics in Dar es Salaam that provide a broad range of prevention of mother-to-child transmission of HIV (PMTCT)-related services including antiretroviral therapy (ART), counseling, as well as provision of nevirapine and HIV testing for infants at 6 weeks of age.

The study is a cluster randomized controlled trial with clusters randomized to intervention or control arms with a 1:1 allocation ratio. Clusters were pair-matched by geographic location within districts (Ilala, Kinondoni, and Temeke). Within each pair match we used a random number generator to determine which cluster would be the intervention cluster. Each cluster was defined as a primary health care facility with a RCH clinic that provided PMTCT services and its satellite facilities. Satellite facilities were those that referred patients to the primary health facility and were therefore administratively and clinically linked. Ten of the clusters had a single satellite facility and the remaining six did not have satellite facilities.

Healthcare facilities were eligible for inclusion in this study if they were government-managed RCH clinics in Dar es Salaam that were supported by a local non-governmental organization, Management and Development for Health (MDH). At the time of facility selection, MDH supported 203 of the 231 active RCH clinics in the region (88%). Facilities were eligible for inclusion in the study if they were dispensaries or health centers (i.e. hospitals, N = 3, were excluded), and if they saw at least 350 HIV-positive pregnant women in the year prior to the start of the study. Facility selection occurred in two rounds. In the first round of selection in March 2015 12 facilities were selected randomly, matched, and randomized. In November 2015 an additional four facilities were selected, matched, and randomized. A description of the study facilities can be found in Table 1.

Initial cluster selection and randomization occurred in March 2015 with 12 facilities. There were three changes in the selection of clusters after this initial randomization. First, 2 weeks after randomization one of the control sites was upgraded from a health center to a hospital, and was therefore not eligible to be a study site. The RCH clinic was moved to a nearby health center and this new health center thus replaced the original one as the control facility for this study. Second, in November 2015, there were fewer eligible women than expected identified in the study facilities. The study team thus identified, matched, and randomized an additional four facilities from the list of original list of eligible facilities. Third, in January 2016 one of the intervention facilities also was noted to have low enrollment due to fewer than expected women eligible at that site. Therefore, a new facility was randomly selected to replace the intervention facility from the two eligible facilities identified. This resulted in 16 study sites/clusters as compared to the original 12 sites that were planned.

Within the selected study clusters, women were eligible for participation if they were at least 18 years old, up to 30 weeks of gestation, HIV-positive and receiving ART, demonstrated symptoms consistent with MDD on the Patient Health Questionnaire-9 (PHQ-9) of a score of ‘9’ or above, planned to continue their post-partum care at the study facility, and without a current plan to harm herself, including the intent to commit suicide. A cut-off score of ‘9’ for the PHQ-9 was used based on the results from a prior validation study in Tanzania [33]. If the woman had suicidal thoughts a formal referral to the psychiatric nurse at the facility was made. Women were recruited from the study facilities from May 25, 2015 until the target sample size was reached on April 29, 2016. Research assistants conducted recruitment by approaching women attending the PMTCT clinics at these sites. All women who were pregnant and HIV-positive were approached and invited to participate after an interview to determine their eligibility. Women who had received their HIV-positive diagnosis on the day of recruitment were contacted 2 weeks later to determine eligibility. If eligible, women were invited to participate in the study.

The Healthy Options intervention includes evidence-based components from problem-solving therapy (PST) and cognitive behavioral therapy (CBT) approaches [34, 35]. The PST component includes six psychosocial group therapy sessions plus one orientation session prior to delivery. Implementation strategies for the PST sessions are borrowed from NAMWEZA [36], a participatory and strengths-based program that was adapted for an urban Tanzanian context and demonstrated improved clinical outcomes in PLH [37]. This approach promotes awareness of one’s own strengths and assertive communication skills. The content of PST sessions includes acknowledging problems and identifying strategies for addressing them, focusing on positive thinking approaches, planning for the future and dreaming/fostering hope, as well as assertiveness and goal setting.

During the follow-up interview 6 weeks after delivery, women with elevated depressive symptoms consistent with a diagnosis of MDD (scoring ‘9’ or above on the PHQ-9) are invited to participate in eight CBT group sessions. The CBT sessions were adapted from the evidence-based “Thinking Healthy” intervention for psychosocial management of perinatal depression [38] for delivery in small groups of 8–10 women and include: discussing the link between one’s thoughts and feelings; sharing practical skills of caring for their infants and addressing daily challenges; discussing strategies for managing stress; fostering support from their social networks; and offering feedback on their CBT homework assignments with the goal of addressing depressed mood through cognitive restructuring.

Psychosocial support group sessions are facilitated in Swahili by lay community-based health care workers (CBHWs) who had prior experience with providing HIV-related peer counseling and support. A two-week in-depth training included didactic and experiential components whereby the CBHWs played the role of study participants in each of the sessions, with trainers serving as facilitators. The group sessions are approximately 2–3 h in duration and include 15–20 women, with smaller break-out groups for discussion and skills building, and are held in a variety of venues, such as health facilities, schools, and religious sites.

The control condition, referred to as “improved standard of care”, is a one-day training session offered to clinical staff including medical officers, clinical officers, nurses, medical social workers, and medical attendants at the RCH clinics where the study was recruiting participants. The training relies on evidence-based WHO mhGAP guidelines [39], focusing on: screening for depression; using an algorithm for assessment and management; attending to aspects of depression that can be addressed in a primary care setting, including exploration of current stressors, psychoeducation, reactivating social networks, offering structured physical activity, and follow-up; and identifying patients that need referral for specialized psychiatric care. This included access to counseling as well as anti-depressant treatment when necessary, as well as immediate care for individuals presenting with acute suicidal ideation. In the event of acute risk of suicide, study participants are escorted to a facility that offers services for further evaluation of suicidality by appropriately trained mental health professionals. This training session is also offered to the intervention sites and is designed to improve the capacity of all of the facilities to provide screening, treatment, and referral for women presenting with elevated levels of depressive symptoms.

The primary outcome of interest is a level of depressive symptoms consistent with MDD on the PHQ-9 [40] measured at baseline, 6 weeks and 9 months post-partum. A prior validation study was performed in a primary health care setting in Tanzania, which demonstrated that a lower cut-off score of ‘9’ maximized the sensitivity and specificity for identifying major depression in this setting. Therefore, a score of ‘9’ or above was used to reflect an elevated level of depressive symptoms consistent with major depression in this study [33]. The PHQ-9 has demonstrated good internal consistency (Cronbach’s alpha ranging from 0.86 to 0.89) and criterion validity (sensitivity and specificity of 88%). Secondary outcomes include intimate partner violence (IPV), social support, self-efficacy, hope, HIV-related stigma, and ART adherence. IPV is defined as report of either physical and/or sexual violence and items are derived from the DHS [41]. For social support the Duke University-UNC Functional Social Support Questionnaire is used; this scale has demonstrated positive correlations with other social support scales. A listing of ten items from this questionnaire has been used previously in Tanzania [8] and includes dimensions of instrumental and emotional support [42]. The General Self-Efficacy Scale is used to assess self-efficacy. This scale has demonstrated good internal consistency (Cronbach’s alpha ranging from 0.81 to 0.91) and construct validity has been shown through positive correlations with optimism and negative correlations with depression and anxiety [43]. Hope is assessed through a scale developed locally by Siril et al. [44] with individuals living with HIV in Tanzania. For HIV-related stigma, the HIV Stigma Scale developed by Berger et al. [45] is used. This measure was validated in a diverse sample in the U.S. and demonstrated construct validity through factor analysis and internal consistency (Cronbach’s alpha of 0.96). A sub-sample of these items also demonstrated construct validity of this scale in a resource-limited setting [46]. In addition, we assess self-reported ART non-adherence as the proportion of women who had missed at least 1 day of ART in the past 4 days [47, 48]. Finally, in order to assess women’s food security, an important indication of their current economic standing, we use a composite measure. Women are considered food insecure if they report that their household had any trouble satisfying their food needs in the past 6 months and/or they currently owe money at a shop where they purchased food.

Trained research assistants conducted structured eligibility and baseline surveys from May 2015 through April 2016. The eligibility form consisted of items indicated in the inclusion and exclusion criteria, including age, number of weeks of gestation, information regarding whether the woman planned to receive post-natal care at the same facility for 2 years after birth, and the depression screening questionnaire (PHQ-9). Prior to the interview all women provided written informed consent. Surveys were conducted in person and lasted approximately 30 min. Blinded data collection was not feasible given the clinic-based cluster design. Responses were recorded on paper and entered into a Microsoft Access database. Data were reviewed weekly for errors and inconsistencies and corrections were made as necessary. Follow-up data are collected at 6 weeks and 9 months after delivery. The follow-up data are collected electronically on smart phones. All data are exported to Stata version 14.2 for analysis (StataCorp LP, College Station, USA) [49].

Power calculations were conducted assuming randomization of 12 clusters, stratified by district, and pair-matched. We planned to screen 60 patients in each cluster. Given experience with other trials we assumed that 80% of patients would be enrolled from each cluster and a conservative 30% would drop out throughout the study, we anticipated having 34 patients in each cluster at the end of the study. Given this sample size (720 women screened), an anticipated depression prevalence of 73% in the control arm and 60% in the intervention arm 6 months after completion of the intervention [34], as well as a coefficient of variation of 0.05, the minimum detectable risk ratio for 80% power and 5% two-sided type I error was 0.79 [50]. With the addition of clusters after initial randomization, our power increased.

In this initial paper we present the baseline characteristics of participants in the intervention and control clusters. We include sociodemographic characteristics of participants as well as the baseline values for the primary and secondary outcomes of interest for the trial. The primary outcome will be depressive symptoms consistent with MDD. Secondary outcomes will include general self-efficacy, social support, HIV-related stigma, intimate partner violence, hope, and ART adherence. We present proportions for all categorical variables and the mean and standard deviation for all continuous variables and scales. Future analyses will determine the effect of the intervention on these primary and secondary outcomes.

For the outcomes analysis, an intent-to-treat strategy will be employed, consistent with what is typically performed for randomized controlled trials, with the primary outcome being depressive symptoms at follow-up. Secondary outcomes will include intimate partner violence, social support, self-efficacy, hope, HIV-related stigma, and ART adherence. Generalized estimating equations (GEE) will be used to compare outcomes for intervention and control groups at first and second follow-up assessments at 6 weeks and 9 months post-delivery, respectively. In order to account for the clustering of data we will utilize robust standard errors for GEE models. A sensitivity analysis will be performed to adjust for variables that have demonstrated some imbalance at baseline, such as education and marital status. Additional sensitivity analyses will be performed, including a pair-matched analysis as well as a non-parametric analysis to account for variables that are not normally distributed. Effect modification for the main outcome variable of depression will also be explored and will include variables such as baseline depression score, intimate partner violence, HIV status disclosure to partner, social support, HIV-related stigma, and hope. A parallel analytic strategy will be performed for all secondary outcomes.