This is a cross-sectional study conducted in Kathmandu, the capital and most populous city of Nepal. We recruited caregivers of outpatients with recorded mental illness from Department of Psychiatry, Tribhuvan University Teaching Hospital (TUTH). TUTH is a site for seeking psychiatric health services at an affordable cost, and has a regular flow of patients from different parts of the country.

For the assessment of perceived stigma, we developed a questionnaire based on items from Internalized Stigma of Mental Illness (ISMI) and an earlier study by Ostman et al [4, 23], while that for attitude was adapted from Community Attitude towards Mental Illness (CAMI) tool [24]. Both the ISMI and CAMI were not designed specifically to assess caregiver’s attitude in a hospital-based setting, so we selected only the items from these tools that were relevant to our study objectives (for details, see questionnaire items- S1 File). The items on stigma measured the intensity of caregivers being stereotyped and the items on attitude reflected caregivers’ individual attitude towards PWMI. Response to each of these items was recorded in a scale of 1–5. A bilingual translator translated the questionnaire into Nepali language. We pre-tested the questionnaire among all the caregivers visiting the outpatient department on 23^rd April 2014 (n = 10) of TUTH. These participants were not included in the main study. Based on the pretesting, questionnaire sequence was organized, language was modified and few questions with repetitive responses were removed.

We recorded information on caregivers' sex (Male or Female), religion (Hindu, Buddhist or Others), residence (Rural or Urban), marital status (Married or Unmarried), educational status (Illiterate, Primary level, Secondary level or Higher), type of family (Nuclear or Extended), occupation (Agriculture, Business, Service, Foreign employment or Others) and relation with patient (Parents, Spouse, Children, Siblings, In-laws or Others). Caregivers belonging to a Village Development Committee (a relatively rural area with population below 20,000 and lacking basic infrastructures) were grouped as rural residents and those from municipality (a relatively urban area with population above 20,000 and having basic infrastructures) were grouped as urban residents. Family that constituted only a couple and their children was grouped as a nuclear family while other types of families were categorized as an extended family.

Patients’ information included sex (Male or Female) and age (15–24 years, 25–34 years, 35–44 years, 45–54 years or ≥55 years). Type of mental illness of the patient was categorized based on ICD-10 classification (Organic disorders, Schizophrenia, Mood disorders, Neurotic, stress-related and somatoform disorders, Behavioural syndromes, Mental retardation or Other diseases of nervous system) [25]. Duration of illness was recorded as the duration between interview and the date of diagnosis. It was categorized into four groups (<1 year, 1–5 years, 5–10 years or ≥10 years) for statistical analysis. Information regarding the treatment site used in first visit was recorded; response to this had two alternatives (Traditional or Modern). Information on the use of treatment modalities other than the allopathic was also recorded (Yes or No).

The data were collected from 170 caregivers visiting Outpatient Department of Department of Psychiatry, TUTH between May and July 2014. Caregivers below 18 years of age were excluded from our study due to ethical reasons as it was not feasible to take parental consent in the Outpatient Department.

Caregivers of patients who visited during the study period were approached for face-to-face interviews. In the process of recruitment, we first checked the Outpatient Card of patient for confirmation of a mental illness. Each card in the reception was checked and the patient was then followed. Caregivers of the patient were identified and approached when they accompanied the PWMI for outpatient consultation. Among those approached, 29 caregivers (14.57%) denied to participate in the study. Finally, 170 caregivers were then interviewed in the waiting room placing them at a distance from the patient. Authors themselves (DN, SD and ST) conducted all the face-to-face interviews.

Data were entered in EpiData 3.1 and was analyzed using SPSS version 17 (SPSS, Chicago, IL, USA). Negatively worded items were first reverse-coded so that for every item higher score reflected negative attitude or greater perceived stigma. Scores were then summed up to yield a composite attitude score and perceived stigma score. The practice of summing of the scores helps in better comprehension of the difference in the scores between subgroups and is common in studies assessing stigma or attitude [26, 27]. Possible score for attitude was 18–90, in which higher score meant a more negative attitude. Perceived stigma score could range from 12–60 where a higher score implied greater perceived stigma.

For descriptive statistics, frequency and percentage of each demographic characteristics were calculated. Median and inter-quartile range were also calculated for each item on attitude and perceived stigma scale. Shapiro Wilk test of the composite scores showed non-normal distribution so we used non-parametric tests for further analysis [28]. We then computed median score for each category, performed the Mann–Whitney U test or Kruskal Wallis H test to compare the scores between groups and Dunn-Bonferroni tests for post-hoc comparisons [29]. P-value less than 0.05 were considered statistically significant.

The study was approved by Institutional Review Board, Institute of Medicine, Tribhuvan University. We obtained prior permission from the hospital administration for the conduction of research. Participants were explained about study objectives and written consent was obtained. Participants received no incentive for participation into this study.

Significant difference in median attitude score by sex of caregiver (p<0.05) and sex of patient (p<0.05) was observed. Median attitude score when compared with the educational status of caregiver, a significant difference in scores between different groups was observed (p<0.001). Post hoc test showed significant difference in attitude scores between caregivers with higher education and those with primary education (p<0.001), and caregivers with higher education and illiterates (p<0.005). Similarly, median attitude score was significantly different between different types of mental illness (p<0.05). Post hoc test revealed difference in attitude score between the caregivers of patient having mood disorders and those having schizophrenia (p<0.05).

We observed a significant difference in median perceived stigma score by the sex of caregiver (p<0.05) and marital status of caregiver (p<0.001). Median perceived stigma score varied significantly by the educational status of caregiver (p<0.001). Post hoc test showed difference in perceived stigma scores between caregivers with higher education and those with primary education (p<0.001), and caregivers with higher education and those who were illiterate (p<0.001). Occupation of caregiver showed a significant difference in median perceived stigma score (p<0.05). On post-hoc analysis, difference in perceived stigma between caregivers in ‘Service’ category and those in ‘Others’ category (p<0.05), and caregivers in ‘Agriculture’ category and those in ‘Others’ category (p<0.05) was revealed. Also, difference in median perceived stigma score between different relation of caregivers with patient was observed (p<0.001). Post hoc test showed difference between caregivers who were children and those who were parents (p<0.001), and caregivers who were siblings and those who were parents (p<0.005). Similarly, difference in median perceived stigma score between those who used alternative treatment modalities and those who didn’t use alternative treatment modalities was revealed (p<0.05). (Table 1)

Overall median score and the inter-quartile range for attitude domain were 42 and 8 respectively. Upon analysis of individual items, the highest median attitude score (four) was found for three items: mentally ill people are violent, the mentally ill people have to be controlled like a young child, and person should be hospitalized soon after signs of a mental illness. Moreover, lowest median attitude score (one) was found for the statement that indicated that it is our responsibility to provide best care to PWMI. (Table 2)

Overall median score and inter-quartile range for perceived stigma among caregivers were 28 and 8 respectively. Item that asked if the caregiver has felt supported by people in carrying the burden of having a relative with mental illness received greatest median stigma score (four)—meaning higher stigma as measured by that particular item. All items other than this received median score of two—suggesting lower perceived stigma. (Table 3)

Before interpreting the findings of this study, few limitations need to be given due consideration. Caregivers of patients seeking no treatment and admitted in the in-patient clinics were not the participants in our study; therefore our study sample might have excluded PWMI in earlier stage and PWMI with a relatively severe illness. This study population might not be the true representative of background-population where the study-findings can be referred to, as caregivers of patients visiting hospital might have better attitude and lower perceived stigma in comparison to those in community setting. Among the approached caregivers, 14.57% denied to participate which might have affected the findings. In addition, we did not carry out a multivariate analysis so the confounding effect of other variables has not been addressed. Given the large number of tests conducted on relatively small sample size, probability of type-1 error is high. Finally, as the assessment was based on instrument developed by taking items from ISMI and CAMI with contextual relevancy into consideration, and has not been validated previously, rooms to question the construct validity of the instrument does exist.