Design
A primary healthcare-based prospective cohort study of people with epilepsy.
2.2 Setting
The study was conducted in the Gurage zone in the Southern Nations, Nationalities, and Peoples’ Region (SNNPR) of south-central Ethiopia. The Gurage zone is predominantly rural, characterised by fertile semi-mountainous terrain. Welkite town is its administrative center. The study was conducted in four districts (Sodo, Eja, Wolikete, and Kebena), with a total estimated population of 450,000-500,000 people. The Ethiopian health care system is divided into three tiers of service delivery. The first level consists of primary healthcare units (health posts and primary health care (PHC) centres) and primary hospitals. PHC centres are generally staffed by nurses and health officers, serving a population of 25,000–40,000 people. Health posts are staffed by one or two community health extension workers, serving a population of 3000–5000 people. Secondary-level services are provided by general hospitals and serve as referral centres from the primary hospitals; and tertiary-level services include specialized hospitals.
This study was nested in the scale-up phase of the PRogramme for Improving Mental Health CarE (PRIME) project which was a UK Department for International Development (DfID-funded) research programme consortium across five LMIC(Ethiopia, South Africa, Uganda, India, and Nepal) (20). PRIME aimed to provide comprehensive evidence on how to integrate and scale up care for people with psychosis, depression, epilepsy, and/or alcohol use disorders. The focus was on integration in primary health care (PHC) settings using the World Health Organization’s mental health Gap Action Programme (mhGAP) intervention guide (21–23). The programme of care was first implemented in the Sodo district (8 PHC centres) and, from 2016 onwards, scaled up to the other 14 districts in the Gurage zone (one PHC centre per district). The four study districts for the current study were selected purposively because of their high commitment to integrating mental health care and logistical considerations.
Source population
The source population for this study was all people with a provisional diagnosis of convulsive epilepsy living in the four study districts of the Gurage zone.
Screening and recruitment of study participants
Case detection was carried out by community key informants and health extension workers (HEWs) who had been trained for two days to recognize people who may have active convulsive epilepsy, augmented by house-to-house screening by HEWs (21). Screen-positive individuals were referred to the nearby PHC centre and the diagnosis of epilepsy was confirmed by PHC workers who had been trained through PRIME and applied diagnostic algorithms outlined in the mhGAP intervention guide. This two stage screening method has been used previously (24) and was implemented in the PRIME study (21). The project psychiatric nurse then screened for eligibility, assessed for capacity to consent to participate in the study, and obtained informed consent before a person was recruited into this cohort study.
Inclusion criteria: (a) PHC worker diagnosis of active convulsive epilepsy: two or more unprovoked convulsions separated by greater than 24 hours, with one convulsion taking place within the preceding 12 months (25, 26); (b) Aged 18 years or above; (c) No plans to out-migrate in the next 12 months.
Exclusion criteria: (a) Communication difficulties due to cognitive or intellectual disability; (b) Unable to converse in Amharic, the official language of Ethiopia; (c) Lacking the capacity to consent after a psychiatric nurse assessment using the standardised approach used previously in this setting (27).
Sample size determination
Based on a large, prospective study, the mean quality of life score for people with epilepsy and depression was estimated to be 31.7 (SD = 13.06) compared to 19.3 (SD = 13.87) in those without depression (13). A total sample of 50 participants (25 with and 25 without co-morbid mental disorder) would be sufficient to detect this difference, with alpha 0.05 and power 0.8. To allow for the detection of a smaller difference in means (mean difference of 5.0), the required sample size was 88 in each group. To take account of clustering by district (n = 4), we assumed an intra-cluster correlation of 0.01 (28), resulting in a design effect of 1.21. Allowing for a 20% loss to follow-up, a total sample of 256 was required (128 per group).
Measurement:
Eligible people who gave informed consent to participate were interviewed at baseline (T0), and again after six months (T1) of follow-up. The hypothesised conceptual model is shown in Fig. 1.
Primary outcome (T0 and T1)
Quality of life was measured using the 10-item Quality of Life in Epilepsy questionnaire (QOLIE-10-p) (29). This questionnaire was derived from the original 89-item version QOLIE-89 with an additional eleventh item to give a weighted total score (30). The 10-item questionnaire has seven components: one item for each of five domains (seizure worry, overall quality of life, emotional well-being, energy, and cognitive functioning), two items on medication effects (physical effects, mental effects); and three items on social function (work, driving, social function). The total mean score ranges from 0-100 with a higher score indicating better quality of life. For this study, the instrument was adapted and construct validity was established (31). The English version of the QOLIE-10–p was initially designed to be self-administered. For this study the instrument was translated into Amharic, which is the local language, by the principal investigator and it was then back translated to English language by a non-mental health professional. The final Amharic version of the instrument was prepared after discussion of a group of psychiatrists with expertise in the area(31).
Secondary outcomes (T0 and T1)
Functional disability: was measured using the World Health Organization Disability Assessment Schedule version 2.0 (12 item WHODAS-2) (32). The WHODAS-2 is a generic instrument that measures health-related functional disability in six domains of life during the previous 30 days. Each item is scored on a Likert scale starting from “no difficulty” 1 to “mild” 2, “moderate” 3, “severe” 4, or “extreme” difficulty 5. The recommended polytomous scoring method was used for analysis. A higher total score indicates a higher functional disability. WHODAS-2 has been validated in people with chronic diseases, including epilepsy (33), and in Ethiopia (34). The 12 item even has shown superiority in understand ability and contextual relevance in a study done in people with severe mental disorders in the Ethiopian setting (34).
Primary exposure (T0 only)
Common mental disorder (depression, anxiety, and somatic) symptoms: The Self Report Questionnaire (SRQ-20) was developed by World Health Organization (WHO) to screen for CMD symptoms in the past 30 days (35). The SRQ-20 items ask about depressive, anxiety, somatic symptoms, and suicidal ideation. The total score is calculated by summing up all positive symptoms, ranging from 0–20. The SRQ-20 was previously translated into Amharic and validated in perinatal women (36) and at primary healthcare level (36–38). A score of eight was the optimum cut-off point for detection of depression at PHC level (37). For people with epilepsy in the same setting the optimum cut-off score of SRQ-20 indicating common mental disorder was greater or equal to 9 (39).
Substance use: risky use of alcohol, khat, and tobacco was measured using the Alcohol, Smoking, and Substance Involvement Screening Test (ASSIST)(40). The ASSIST has eight questions, with questions one to seven asking about use and problems related to substance use, and the eighth question inquiring about the use of injectable drugs. The total score for specific substance involvement is calculated by summation of the assigned numerical numbers from questions number 2–7 for each substance class. Low risk is indicated by a score of 0–10 for alcohol and 0–3 for other substances, moderate risk is 11–26 for alcohol and 4–26 for other substances, and high risk is indicated by a score of 27 and above. The ASSIST has been contextually adapted in multiple countries including in Africa and Ethiopia (40–42). For this study, the ASSIST was modified to assess commonly used substances in the southern part of Ethiopia: alcohol and khat (43, 44).
Potential confounding variables (T0 only)
Socio-demographic characteristics: age, sex, education, marital status, income.
Epilepsy-related factors: duration of epilepsy and seizure frequency. At baseline, seizure frequency in the past one month was measured. At the follow-up time-point, the numbers of seizures in the last 6 months were recorded as follows: number per week (if < 1/day), number per month (if < 1/week), and number in the last 6 months (if < 1/month). The severity of seizures was grouped into three categories based on their frequency in the past 6 months: seizure:-None, low to moderate (1–2 seizures), and high seizure severity (greater and equal to 3 seizures). This categorization of seizure severity has been used in an African setting in a previous study (45).
Social support: This was measured using the Oslo-3 item Social Support scale (OSSS-3)(46). The OSSS-3 is a brief measurement of social functioning and has three items: The total score ranges from 3–14 with a higher score indicating better social support. OSSS-3 has been validated in an African setting (47) and has been used in several studies in Ethiopia (48).
Factor hypothesised to be on the causal pathway
Perceived stigma was measured using the stigma section of the Family Interview Schedule (FIS) questionnaire (49). This instrument has been translated into Amharic and has been used previously in rural Ethiopia to measure stigma in people with epilepsy and their caregivers and those with mental disorders (50, 51). Each item is rated in a four-point scale 0 “not at all”, 1 “sometimes”, 2 ”often”, and 3 “a lot” regarding the perceived stigma. A total score of one and above is considered as having the experience of perceived stigma.
Hypothesised effect modifier: epilepsy treatment engagement
Treatment engagement was operationalized as the number of times the person attended the PHC centre in the preceding 6 months. Self-reported attendance was recorded and augmented by a medical record review. Good treatment engagement was defined as attending ≥ 4 times during the follow-up period.
Data collection and management
All measures were carried out by experienced lay data collectors who have completed secondary school education. The lay data collectors were trained on the administration of the questionnaire for five days and practiced through role play before administering them to study participants. Immediately after the completion of data collection, the field supervisor checked the questionnaires for completeness. Data were double-entered using Epi-data version 3.1(52).
Data analysis
Data were analysed using STATA version 17 (53). For continuous variables, indicators of central tendency were calculated depending on the distribution (mean with standard deviation (sd) or median with Interquartile range (IQR)). Percentages and frequencies were used to summarize categorical variables. Simple descriptive analyses were used to summarise the socio-demographic and clinical characteristics at T0 and T1. Wilcoxon ranked sum test or Fisher’s exact test was used to examine the statistical significance of differences in baseline characteristics of those who were lost to follow-up and those who remained in the cohort. The dependent variables of change in quality of life and change in functional disability were calculated by subtracting the total scores at T1 from T0.
Univariate and multivariable linear regression models were fitted to evaluate whether the primary exposure (comorbid CMD symptoms) predicted a change in the outcome variables (QOL and functional disability) adjusting for baseline outcome data. The pre-defined potential confounding variables (measured at the baseline) were also entered into the multivariable model. The risk of alcohol use was entered into the model separately from the total SRQ-20 score (CMD symptoms). Effect modification by number of PHC centre visits (treatment engagement) was tested using interaction terms with a total SRQ-20 score. A likelihood ratio test was used to examine statistical significance.
Structural equation modelling (SEM) was then conducted using R version 4.3 (54) to examine direct and indirect pathways through seizure frequency linking co-morbid CMD symptoms with QoL or functional disability. The direct and indirect pathways linking to the outcome were drawn based on the pre-hypothesised conceptual model (Additional file 1). Separate SEM was fitted for QoL and functional disability as two separate outcomes.
Before fitting the full SEM, CFA was carried out for each of the latent constructs of CMD symptoms, stigma, quality of life, and functional disability to examine the fit of the measurement models. The goodness of fit of the models was checked for each latent construct using the Root Mean Square Error Approximation (RMSEA), Tucker-Lewis Index (TLI), and Comparative Fit Index (CFI). The significance of factor loadings of each item and the plausibility of the loadings were also examined. Weighted least square estimation was used for the complete data. The SEM was fitted again after multiple imputations of missing data using a chained Eq. (55).