The aim of this cross-sectional study was to carry out an epidemiological survey to estimate prevalence of MHDs among adults attending primary care facilities located in all five districts (Jaffna, Mannar, Mullaitivu, Vavuniya, Kilinochchi) of Northern Province, Sri Lanka.
Inclusion criteria were both sexes, 18 years and older, internally displaced due to conflict, attending public primary care facilities, either for the first time or as a follow-up visit. Children younger than 18 and people with severe mental illness or hearing or speech disability whose conditions prevented administration of the study questionnaire were excluded.
Sample size was calculated using a prevalence of anxiety or depression - the most common MHDs - of 50.0%, a conservative value chosen based on doubling previous study prevalence conducted in Jaffna, Mannar, and Kilinochchi districts of Northern Province of depression at 22.2% (95% CI, 18.2–26.5%) and anxiety at 32.6% (95% CI, 28.5–36.9%) [7]. Due to the conservative choice of 50% prevalence, this sample size should be adequate to detect higher prevalence of concurrent anxiety conditions as well with high precision. To achieve ±5% precision, considering a 95% confidence interval, a total population of 1,234,932 in the Northern Province, an estimated design effect (DEFF) of 2.2 based on previous literature, and an assumed response rate of 80.0%, the total required sample size was calculated to be 1025 [11].
Primary sampling units were public primary care facilities in Northern Province, which included divisional hospitals and primary medical care units. The most recent data (from 2013/14) details Northern Province had 54 reported divisional hospital, 35 reported primary medical care units, and the provincial number of outpatients in district hospitals and primary medical care units were 1,445,675 and 325,480 respectively [12].
A list of public primary care facilities in all Northern Province districts was compiled by type, and 25 facilities (clusters) were randomly selected. A flowchart of the study sample is presented in Fig. 1. Distribution of clusters was allocated proportionally to total number of IDPs in each district; districts with smaller number of IDPs were assigned fewer clusters. Allocation of clusters was to account for population displacement during the last stage of the conflict in 2009, and parts of the province, which experienced less displacement, but had larger populations. This strategy ensured an adequate representation of conflict severity and displacement typology and length. From each facility (cluster), 41 individuals were recruited by systematically selecting every third attendee from the facility registration desk.
Information was gathered using a structured interview covering demographic and socio-economic background, conflict and displacement experiences, mental disorder screening, social support/networks, and health service use. MHD screening tools utilized were: Hopkins Symptom Checklist-25 (HSCL-25) for depression and anxiety, Patient Health Questionnaire (PHQ-15) for expression of somatoform symptoms, Harvard Trauma Questionnaire (HTQ) for PTSD, Psychosis Screening Questionnaire (PSQ) for psychosis. The Lubben Social Network Scale (LSN) and Multidimensional Support Scale (MDSS) were used to assess social support and social networks. The threshold for each MHD was determined by the validated scoring process for each measurement utilised. This process is detailed below.
Most instruments had been previously used in a number of epidemiological studies in Sri Lanka, especially among conflict-affected populations including IDPs [7,8,9, 13]. These instruments are available in both Sinhalese and Tamil, the main languages spoken in the country. Tamil versions had been validated and previously used in the Tamil-speaking population in the Northern Province [7,8,9,10,11,12,13]. New instruments were adapted for use in the study setting through established procedures and field tested during a pilot study conducted at a separate primary care facility in Jaffna district [14].
The sociodemographic section of the questionnaire included 31 variables used in a previous epidemiological cross-sectional study among Tamil populations in Sri Lanka [8]. The conflict and displacement questionnaire was also previously used in Sri Lanka to explore displacement, conflict experience, services available throughout displacement, and experience of return [8]. Depression and anxiety were assessed using the Hopkins Symptom Checklist-25 (HSCL-25). The HSCL-25 has been translated into Tamil and used in a previous study in Northern Sri Lanka [7, 12A]. In concordance with previous research, a cutoff score of 1.75 for each anxiety and depression were used to identify positive cases [7]. Expression of somatoform disorders was assessed using the Patient Health Questionnaire (PHQ-15), where the Tamil language version of the PHQ has been used previously among IDPs and has adequate internal consistency [8, 9].
PTSD was measured using the first section of the Harvard Trauma Questionnaire (HTQ), which has been validated and used in a previous study in Jaffna district [7, 13]. Those that reported a three or a four to at least one question in the recurring symptoms sub-section, at least two questions on the arousal sub-section, and at least three questions in the avoidance sub-section were classified as screening positive for PTSD symptoms [7].
The Psychosis Screening Questionnaire (PSQ) was used to screen for symptoms of psychosis in respondents. The PSQ was previously used in a national prevalence study in Sri Lanka7 but not among post-conflict populations, although the PSQ has been used in conflict settings in other countries. Cutoff scores for psychosis were in agreement with previous research and sensitivity analyses were conducted excluding hypomania from this case definition [14, 15].
Lubben Social Network Scale (LSNS-6) and Multidimensional Support Scale (MDSS) were used to measure and assess social networks and support. Both instruments have been previously validated and used in Tamil language among IDP populations in the Northern Province [9, 16]. Using the LNS, participants with scores of 12 or greater were considered to have adequate social networks [17]. Social support availability and adequacy were assessed via the MDSS. As a cut-off point is not standard for this measure, total scores were sorted by tertiles to create availability and adequacy scores [17]. For the section on social support availability, summed scores greater than 32 out of 48 indicated high availability, scores of 17 to 32 indicated moderate availability, and scores of 16 or lower indicated low availability. For social support adequacy, scores of greater than 22 out of 33 indicated high adequacy, scores of 12 to 22 indicated moderate adequacy, and scores of 11 or less indicated low adequacy.
A modified, Sri Lankan version of the Client Service Receipt Inventory (CSRI) was used to record health service utilization aspects [18]. This instrument had been previously translated, back translated and adapted for local context [18]. The modified version used in the current study was piloted before use in the full study to ensure cultural appropriateness and validity.
Data were collected across 25 randomly selected primary care facilities between 20 June and 10 October 2016. All interviews were conducted on-site with data collection forms created using Kobo toolbox, version 1.4.8 [19] and data were hosted at a secure, encrypted server.
Data were downloaded from the secure server to SPSS version 20.0 and cleaned [20]. Data analysis was conducted using SAS version 9.3 [21]. To create a representative sample of all care-seeking individuals in the Northern Province, data were weighted according to total district population size to account for unequal probabilities of selection by district, and respondents from larger districts received larger weights than those from smaller districts. Demographics and prevalence of the five most prevalent MHDs – anxiety, depression, PTSD, expression of somatoform symptoms, and psychosis with hypomania – were obtained using survey means and frequency accounting for two-stage sampling design and sampling weights. Those presenting with any MHD were categorized as screening positive; comorbidities were defined as any combination of screening positive for two MHDs and the number of two-way comorbidities was calculated. Associations between MHDs and social, demographic, economic, and conflict and displacement-related factors were investigated via univariable logistic regression. Multivariable logistic regression considered each MHD individually and included demographics, conflict- and displacement-related factors, social support structures, clinic type and utilization, and participant’s concept of adequacy of care. The final model was selected using backwards selection with a removal threshold of 0.1. Multicollinearity between predictors was assessed using variance inflation factors (VIFs) and correlations.
Informed written consent was obtained and participants were free to withdraw at any time from the study. If participants were identified to have suicidal ideations or serious mental illness, research team members immediately referred them to specialized services following standard operating procedures. Maximum effort was taken to protect privacy during interviews and ensure confidentiality of data collected.