Setting

The mhBeF program in Nepal was implemented in Pyuthan district between 2013 and 2015. Pyuthan is a district located in the mid hills of the Rapti zone in the mid-western development region of Nepal. Nepal ranks 145 of 187 countries on the human development index (HDI) in 2015. The objective of mhBeF was to develop and implement a comprehensive community-based mental health care plan, in accordance with the mhGAP guidelines, for persons with severe mental disorders (psychosis and bipolar disorder) and epilepsy in Liberia, Uganda, and Nepal [7].

Intervention arms

The study evaluated patients that were offered a comprehensive care package, which included; (a) clinical assessment and psychotropic treatment by primary health care (PHC) workers following the mhGAP Intervention Guide (after receiving a 9-day training); (b) psychosocial support through individual or family counselling and patient support groups (PSGs) by community counsellors (after receiving a 21-day training, in addition to a base training in counselling); and (c) conducting stigma reduction activities and ensuring follow-up care through home-based care by female community health volunteers (after receiving a 3-day training). Supervision of PHC workers by a psychiatrist was an integral component of the program. In addition the program used a newly developed procedure (Community Informant Detection Tool—CIDT) for pro-active case finding by community members to increase demand for mental health services. The CIDT has been developed in Nepal, and evaluated for accuracy [8] and impact on increased help seeking [9]. One group of the study participants was offered this comprehensive package of care, while a comparison group was offered a more basic set of services consisting only psychotropic medicines (enhanced Treatment As Usual, eTAU) by mhGAP trained health workers, to allow the assessment of the added value of a comprehensive package of services over a minimal package of services. While the program planned to cater for people with psychosis, epilepsy and bipolar disorder, during the implementation phase only 6 people with bipolar were identified and treated. These people have been excluded from the study. Allocation to treatment or control arms was done by health facilities, which were geographically separated to minimize contamination. In the treatment arm we included the health facilities of 12 Village Development Committees (VDC; the smallest administrative unit in Nepal), and health facilities of 5 VDCs were included in the control arm. We first listed all eligible VDCs within the district (excluding those where referral and supervision were not going to be feasible). Next, we separated the health facilities (n = 5) that had received some form of prior mental health training (which was less than the training that was given as part of this study) from those that had never received such training. The former were allocated to the eTAU arm. Among the latter we randomly selected 12 health facilities—the number was set by the maximum capacity for project implementation.

Sample

Patients with a primary diagnosis of psychosis or epilepsy established by PHC workers trained in mhGAP were recruited into a controlled cohort study. Inclusion into the study followed a three-staged process. First potential respondents were identified as possibly having one of the target disorders (i.e. psychoses and epilepsy) using the CIDT—as described above. Second, all identified individuals were screened using a brief screening instrument developed for the purpose. Third, if a person is found positive on one or more of the target disorders based on the screening procedure, he or she was assessed by trained PHC workers following the mhGAP guidelines for on the target disorders.

Procedures

All research assistants were selected from Pyuthan district and received a 3-week training. The training included basic communication and interviewing skills, research concepts and ethics, as well supervised practice of administering the surveys. Written consent was sought from all respondents, after a full explanation of the study. All individuals that were diagnosed and provided informed consent, baseline measurement was administered by research assistants. Recruitment took place between June 2014 and May 2015. Ethical approval for this study was obtained from the Nepal Health Research Council (NHRC) (Ref: 874; Reg 173/2013). Adverse events were monitored and responded to following TPO Nepal’s Adverse Events Reporting Procedure.

Measures

We selected a clinically-appropriate measure to assess symptom reduction for each disorder [1] Psychosis symptoms: people with psychosis completed the positive and negative symptoms scale (PANSS) (10), which has previously been validated in South Asia [11]. For this study we did not use the PANSS general scale, but only the positive and negative symptoms sub-scales (and for analyses combined positive and negative symptoms sub-scales) [2]. Depression symptoms: all participants also completed the Hamilton Depression Scale (HAM-D; [12]), which we included due to the increased risk of depression among people with epilepsy and psychosis [3]. Seizures: For people with epilepsy, a 9-item instrument to measure the number epileptic seizures in the previous 3 months (i.e., recent seizure) was used [13], which was scored dichotomously [4]. Disability: we measured functional disability for all patients through the 12-item version of the WHO Disability Assessment Schedule 2.0 (WHODAS). The WHODAS has been used extensively and validated in many low- and middle-income countries [14], including in Nepal [15]. We used the complex scoring method when calculating patients’ scores, which incorporates item weights [5]. Burden associated with mental illness, an important impact factor for caregivers and families of those with severe disorders, was measured through the burden assessment schedule (BAS; [16]) and two sections of the family interview schedule (FIS; [17]): symptoms and social behavior (FIS-SSB) and impact on caregiver (FIS-IC). The disability and burden measures were administered through family interview conducted by trained research assistants.

Translation of these instruments followed a standardized five-step procedure for translation of instruments for use in transcultural research [18]. The PANSS, HAM-D, and Epilepsy-9 were administered by trained clinicians. The other instruments were administered by trained research assistants.

The primary outcomes for the study were the two disorder-specific clinical outcomes (i.e., PANSS score for psychosis and recent seizure for epilepsy) and associated functional impairment (i.e., WHODAS score). All other domains are considered secondary outcomes. Each measure was administered at baseline and at 12 months follow-up.

Analysis

We compared baseline characteristics of the study arms using χ² tests for categorical data and an independent samples t-test for the single sample characteristic with continuous data, age of patient. To evaluate each clinical and functional outcome for the full cohort, we conducted a paired T-test for scaled outcomes, which compared baseline and endpoint scores on each measure for all patients. We evaluated the disorder-specific clinical outcome for epilepsy (i.e. recent seizures), using McNemar’s exact χ² for binomial paired data, which compared the proportion of patients with a recent seizure at baseline to that at endpoint. Analyses were carried out on a per protocol basis.

Differences in clinical and functional outcomes across treatment arms were assessed using an analysis of covariance for each outcome. We set the treatment group as a fixed-effect exposure and baseline score as a covariate, with the patient’s score at endpoint as the outcome. For measures that did not meet Levene’s test for equality of variance, we used a logarithmic transformation to account for heteroscedasticity [19]. We compared the reduction in risk of the dichotomous outcome for epilepsy seizure across treatment arms through logistic regression. Again, treatment arm was set as a fixed-effect exposure and recent seizure at baseline as covariate, with recent seizure at endpoint as the outcome.

Upon reviewing the data for the cohort, there were outliers in the distribution of change scores for multiple scaled measures, indicating the possibility of an underlying non-normal distribution of outcomes. We determined post hoc to use the nonparametric equivalent of the paired T-test, the Wilcoxon Signed-Rank Test, when comparing baseline and end point scores for each measure. It was not necessary to select a nonparametric method for analysis between treatment arms despite the presence of outliers since ANCOVA has demonstrated to be robust against violations of non-normality [20].

Limitations

The primary limitation of our analyses is a small sample size. The final sample size after attrition and exclusion of patients with missing data was n = 179 for measures administered through clinician interview, and fewer for those administered through family interviews. This limitation is common to studies in low-resource settings; moreover, a relatively small sample size is a hallmark of cohort studies. The findings for primary objectives are not substantially affected by the sample size. It is in the comparisons of secondary objectives where the small sample size affects our ability to conduct subgroup analyses and will further limit findings should additional inferential analyses be conducted. Future research may consider testing specific hypotheses in addition to those presented here. For example, how might treatment adherence affect disorder-specific outcomes, or are beliefs regarding the cause of psychosis associated with improvements in familial burden and functional impairment? Finally, as a result of the absence of a control group that received no treatment at all (deemed unethical in the context of this study), we cannot empirically attribute change to the treatment. However, given the chronicity of the epileptic and psychotic symptoms, and the strong effects found, it is highly plausible that treatment has contributed to such change.