Equitable access to mental healthcare integrated in primary care for people with severe mental disorders in rural Ethiopia: a community-based cross-sectional study

Cross-sectional, community-based study.

The study was conducted in Sodo district, one of the 15 districts of the Gurage zone, in the Southern Nations, Nationalities and Peoples’ Regional state of Ethiopia. Sodo is located about 100 km from Addis Ababa, the capital of Ethiopia. Sodo district combines the rural–urban and highland-lowland geography which is representative of the country [5]. The population of the district was estimated to be around 160,000 at the time of the study [17]. The settlement patterns are mostly rural, with only 10% of people living in urban centres [17]. The district has one asphalt road linking it to the neighbouring Oromia region in the North and another Gurage district in the South. The urban settlements largely follow the main road. The district has eight health centres, which serve an average of 40,000 people in urban sub-districts and 25,000 people in rural areas. The health centres are each linked to approximately five health posts run by health extension workers (HEWs; female community health workers with 1 year of training in preventive care after completing grade 10). Prior to the introduction of mental healthcare within all eight health centres through PRIME, the nearest specialist mental healthcare facility was a psychiatric nurse-led out-patient unit located around 30 to 50 km away, or in-patient services in Addis Ababa. Part of the PRIME intervention involved 9 days of training for non-specialist PHC staff (mainly public health officers and nurses) to equip them to deliver mhGAP evidence-based packages of care, including diagnosis, assessment and management of selected priority disorders (SMD, depression, epilepsy and alcohol use disorders) [5].

Sixty-five health extension workers and 26 key informants (mainly project data collectors who are also members of the community and community leaders) were trained to detect people with probable SMD. The training made use of case vignettes that illustrated common presentations of SMDs in this setting. Key informant detection of people with SMD has been shown to be sensitive, compared to house-to-house screening, in a previous study conducted in the neighbouring district [18]. Psychiatrists and supervised psychiatric nurses provided the half-day training. Those who received the training were requested to conduct house-to-house visits to identify people who had symptoms/behaviours that corresponded to the description of mental disorders provided during the training. People identified as having probable SMD were encouraged to attend their local health centre on a designated date. Out of people who attended the health centre, those diagnosed by mhGAP-trained PHC workers as having SMD (diagnoses of psychosis or bipolar disorder according to the mhGAP intervention guidelines) were then assessed by project psychiatric nurses using a standardized, semi-structured diagnostic interview schedule [19]. Where the PHC staff and the psychiatric nurses disagreed, diagnosis by the psychiatric nurses took precedence. People who were confirmed to have SMD, including primary psychotic disorders (schizophrenia, schizoaffective disorder) or affective disorders (bipolar disorder or severe depression/depression with psychotic features), were invited to participate in the study. Six months post-implementation of the new integrated primary mental healthcare service, lists of referrals made by the HEWs and key informants were collected from each sub-district (kebele). The referral lists were checked against the registry of those receiving treatment at the health centres. Follow-up visits were carried out to the houses of people who were identified as having probable SMD and referred to PHC but who had not accessed care by the end of 6 months. Both the PRIME project and the PHC registers were reviewed to identify cases who had not been identified originally by community informants, and who might have been eligible for care during the study period.

Access to care was measured by reviewing attendance of people referred to PHC with probable SMD at 6 months after the introduction of the service.

The five item, lay-interviewer administered Psychosis Screening Questionnaire (PSQ) [20], was used to determine presence of probable psychotic symptoms in those who were referred but who did not attend PHCs. Barriers that ever stopped, delayed or discouraged access to mental health care were measured using the Barriers to Access to Care Evaluation (BACE-3) which included a 12 item treatment stigma sub-scale [21]. The Short Explanatory Model Interview (SEMI) [22] was used to elicit causal attributions.

People who attended PHC-based care underwent a confirmatory diagnostic interview using the Operational Criteria Checklist (OPCRIT) [19] by psychiatric nurses.

The WHO 12 item Disability Assessment Schedule (WHODAS 2.0) [23] was used as a continuous measure of disability associated with mental disorders. The WHODAS has been validated for use in people with SMD in rural Ethiopia [24]. Social support was measured using the three item Oslo-3 scale, which measures the extent of the person’s support network, interest shown by others and ease of obtaining practical help [25]. The presence of alcohol use disorder was measured using the Alcohol Use Disorders Identification Test (AUDIT) [26]. A cut-off score of ≥ 8 was used to define presence of alcohol use disorder. Experience of discrimination due to mental illness was measured using the discrimination and stigma scale (DISC-12), using a sub-scale measuring the scope and content of experienced discrimination [27]. To assess physical and sensory impairment, the Brief Physical Impairment Rating Checklist (BPIRC) was used, which was constructed from items from the Washington Group’s disability measure and Work, Family and Well-being (WFW) scales [28, 29]. Constituent items of the scale evaluate physical disability in terms of mobility, ability to use hands, vision, hearing with or without aids and speaking.

Data were collected with structured questionnaires administered by trained lay interviewers whose educational level ranged from 10th grade to a college degree. Supervisors, with educational levels ranging from a diploma to college degree, were also trained and supported by the study investigators throughout the process. Both the data collectors and the supervisors were trained for a week on administering interviews to people with SMD and their caregivers. Their training sessions also involved practice interviews followed by feedback from the group and the trainers. All of the data collectors are residents of the district and thus are familiar with the study setting.

All interview questionnaires were translated into Amharic and translated back to English to evaluate their semantic validity. A team of experts involving psychiatric nurses, psychiatrists, social workers, HEWs, data collectors and field supervisors reviewed the data collection instruments. The review involved attention to language issues and cultural appropriateness of the wording. Comments and concerns raised during the expert consensus meeting were addressed prior to piloting the instruments. For people who accessed care, data were collected at the PHCs after they had seen a healthcare provider. At that time they were given a modest financial reimbursement of 100 birr (approximately $4.5) for transport and time costs. During the referral, none of the people referred with probable SMD were notified about the availability of a reimbursement. For those who were referred but who did not access care, data collectors made home visits.

Double-data entry was carried out using EpiData version 3.1 and exported to STATA version 14 [30] for analysis.

Contact coverage of mental healthcare was calculated as the number of people with SMD in the district who accessed the integrated mental health service in PHC divided by those who were in need of the service [31].

Open-ended responses on the SEMI were coded and categorised using a coding sheet which had been contextualised for Ethiopia previously. The univariate association of access to care at baseline and the categorical primary exposure variables (gender, residence and physical or sensory impairment), was assessed using the Pearson Chi-square test. For normally distributed variables (poverty index, age, total WHODAS score), Student’s t-test was used. Kruskal–Wallis H test was employed for continuous variables that were not normally distributed (total DISC-12 score). Multiple logistic regression was used to examine factors associated with not accessing mental health care, including the following pre-specified variables: residence, poverty, sex, presence of physical or sensory impairment, age, marital status, educational level, alcohol use disorder, social support, disability related to mental disorders, and discrimination. Crude and adjusted odds ratios with their corresponding 95% confidence intervals were reported. To explore the potential role of geographical distance from the health facility in explaining the association with rural residence, the multivariable model was rerun replacing rural residence with distance from PHC.

The study was approved by the Institutional Review Board of the College of Health Sciences of Addis Ababa University. The Amharic version of the information sheet was read to all of the study participants. Eligible study participants were asked to give their written informed consent or a finger print if they were not literate. Guardian/caregiver permissions were obtained to interview people who lacked capacity to give an informed consent because of SMD. Participants in need of medical attention were referred to the nearest PHC.

A total of 300 people were confirmed to have SMD by the psychiatric nurses. See Fig. 1 for a flow chart of identification, referral, attendance and confirmation of SMD. Two people with possible SMD refused to participate in project assessments. Six people with established diagnoses of SMD preferred to continue receiving care from specialist mental health services located in the neighbouring district or Addis Ababa. After 6 months of the new service being available, 61 people referred with probable SMD had not accessed PHC-based care. Twelve months after recruitment was closed, an additional 28 people had received treatment for SMD at different PHCs in the district. These 28 people were confirmed to be incident cases that would not have been eligible for treatment when treatment was introduced. Therefore, the contact coverage for SMD was calculated to be 81.3% (300/369). Of those who accessed the service, 257 (85.7%) had a primary psychotic disorder and 43 (14.3%) had bipolar disorder or major depressive disorder with psychotic features.

More than three-quarters of non-attenders (76.2%; n = 45) had one probable psychotic symptom on the PSQ (see Table 1). The most frequently reported description of the nature of the condition was in terms of mental illness or symptoms of mental illness (n = 22; 36.0%), supernatural (n = 9; 14.8%) or another type of health problem (n = 9; 14.8%). The most frequently reported causal attributions were “unknown” (n = 12; 19.7%), “worry and anger” (n = 11; 18.0%) and supernatural (n = 9; 14.8%) (Additional file 1).

The most frequently endorsed barriers to attending PHC for mental healthcare were the belief that the condition would get better without intervention (44.6%), concerns about the cost of treatment (44.6%) and transport (37.0%), and believing no treatment was available (36.7%). Concerns about stigma were less frequently endorsed (see Table 2).

Rural residence, but not gender, socio-economic status or physical/sensory impairment, was significantly associated with non-attendance for care (see Tables 3 and 4). In the multivariable model, people with probable SMD who did not attend the PHC-based mental healthcare were more likely to reside rurally and to have less mental disorder-related disability than those who accessed services. When the multivariable model was rerun with ‘rural residence’ removed, living more than 180 min walking distance away from PHC was associated significantly with not accessing care.

A thorough case ascertainment process was used to identify people with probable SMD. The involvement of community-based health extension workers and community key-informants facilitated the tracing and referral of people with probable SMD in the district. We did not identify any cases missed during the initial ascertainment who engaged subsequently. None of the 61 non-engagers were treated for SMD in the following year. All of those newly identified as contacting services (28 people) were reported to have developed the illness after the initial period of case ascertainment. This supports the previous finding that key informants have high sensitivity in detecting people with probable SMD in the context of rural Ethiopia [32].

Epidemiological estimates from neighbouring districts in Ethiopia indicate that schizophrenia and bipolar disorder have a combined prevalence of 0.97% of the adult population [33, 34]. Out of the 160,000 total population of the Sodo district, an estimated 77,586) are adults [17]. If we take a conservative estimate that 50% of people with SMD need ongoing care, this would amount to 423 people needing services in Sodo district (compared to the 369 we identified). The variation of our study findings from this epidemiological estimate could be explained by the under-representation of people with affective psychosis in our sample, with only 14.3% of people with SMD accessing services having an affective psychosis. This may be due to the episodic course of bipolar disorder and better functional recovery between episodes [35]. Therefore, these people are less likely to be identified by our methods of ascertainment.

Rural residence was associated with non-attendance for PHC-based care. Our analyses indicate that geographical distance from the health centre is an important contributor to the effect of rural residence: residing more than 180 min away from PHC was associated significantly with not accessing care. Geographical inaccessibility was also reported by field workers to be a barrier to attending for care. Some sub-districts in Sodo are hilly and do not have all-weather paths or roads. In our qualitative exploration of reasons for not engaging with care, respondents emphasised the difficulty conveying a person with SMD who does not agree to treatment when public transport is not available [36].

Although other studies suggested poorer households could have lesser access to health services [37] and this was anticipated to be a barrier in the formative qualitative work [7], we found that poverty was not a barrier for initial contact with PHC-based mental healthcare. Nevertheless, in those who did not attend, affordability of accessing care was an important concern. A possible explanation is the relative homogeneity of socio-economic status in the study site, and the challenge of distinguishing between levels of poverty. In our follow up qualitative work in this setting, affordability of treatment was reported to be a challenge to ongoing access to care for all participants [36]. Even those who had managed to initiate access to PHC-based mental healthcare doubted their ability to pay for ongoing access. Those who did managed to mobilise funds to attend at least once may have been motivated by the greater severity of disorder, as indicated by the higher level of mental disorder-related disability.

In our study, only 9 (14.8%) people who did not access PHC-based mental healthcare framed their problem in terms of supernatural aetiology. The most frequently provided explanation of the problem in those who did not engage was either mental illness or symptoms of mental illness (n = 22; 36.0%). This could be the effect of social desirability bias as there is an educational level difference between the respondents and the data collectors, even though the data collectors were recruited from the same community. We were not able to compare differences in causal attributions of mental illness between those who did not engage and those who did engage. This could be an important area of exploration for future studies.

This study focused on contact coverage with integrated mental health services, but this only reflects one-off contact with services and may not translate into mental health benefits (i.e. effectiveness coverage). There is need for prospective evaluation of the equity of access to long-term care in people with SMD in this setting. As discussed, from our qualitative work [36, 38], we anticipate that poverty will be an important factor in sustained engagement with mental health care.

Addressing the multiple challenges faced by rural residents is key to achieving equitable access to PHC-based mental healthcare in this setting. Interventions to expand access to affordable means of transportation may facilitate engagement by rural residents and people who have to travel long distances to get to the nearest health centre. Moreover, flexible service configurations such as service outreach into the community may be needed to achieve better coverage for rural residents.

Protection of economically vulnerable families against catastrophic healthcare expenditure through a universal health insurance coverage could be helpful. Families often experience unbearable economic distress and may fail to access healthcare due to a system that is structured around out-of-pocket financing [39]. Augmenting the current low-cost healthcare service delivery with other economic interventions such as livelihood diversification could facilitate access by the economically vulnerable. Interventions such as revolving credit services via involvement in service user groups could enable destitute households to engage with care. Strengthening rural literacy programs and introducing mental health awareness raising initiatives could enable people who did not access care due to the perception that mental illness is not treatable.