Study area and study participants
We collected data between April 15 and June 20, 2017 at the outpatient mental health clinic of Mbarara Regional Referral Hospital (MRRH) in Mbarara District MRRH is a public health facility in southwestern Uganda located 270km from Kampala, the capital city. We recruited patients with a diagnosis of bipolar disorder (as confirmed by the attending clinicians) who were in remission phase, 18 years and older, both male and female, and were attending the outpatient mental health clinic at MRRH during the study period.. We excluded participants who presented with active symptoms of mental illness that would interview with their cognitive ability to understand the contents of the questionnaire and the consent documents and those who were physically unwell to stand the length of the interview at the time of recruitment.
Sample size estimation
We used an online OpenEpi software , http://www.openepi.com,, based on Kelsey and colleagues [48] to calculate the sample size. Using this formula, we used a confidence level of 95 %, power of 80 % and ratio of the sample size of the healthy population to that of participants with bipolar disorder to be one. We assumed that the percentage of the healthy population with a poor quality of life to be 20 and the percentage of patients with bipolar disorder with poor quality to be 40. Substituting all the numbers we arrived at sample size of 169 participants. We recruited participants consecutively as the came to the clinic for review every week until we reached our desired sample size.
Measures
We used a locally generated questionnaire to collection social demographic and clinical information of the participants including age, sex, marital status, level of education and source of income. Information on clinical factors included age at onset of the illness, number of episodes in the past year, number of hospital admissions in the past year and presence of suicidal thoughts and psychotic symptoms in the last acute episode. We assessed health related quality of life using the medical outcomes health survey short form-36 (SF-36) scale [49].The SF-36 is a 36-item, self-administered measure of QoL that was developed to examine the impact of disease on perceived well-being. It consists of eight subscales that measure different areas of functioning including: (1) physical functioning; (2) role disruption because of physical difficulties (role disruption-physical); (3) role disruption caused by emotional difficulties (role disruption-emotional); (4) social functioning; (5) mental health; (6) vitality; (7) general health; and (8) bodily pain. Each item is scored on 6 point Likert-type scale. Four of these domains are considered to relate to mental health, namely: 1) role limitations due to personal or emotional problems (“role-emotional”); 2) emotional wellbeing, 3) social functioning and 4) energy and fatigue. Together, these four scores form a single mental component scale, while the other four sub-scales (physical functioning, role disruption because of physical difficulties, vitality and general health) aggregate into a physical component scale. Physical (PCS) and mental (MCS) component summary measures were calculated by weighting each SF-36 item using a norm-based scoring method given in the guidelines [50, 51]. Since Uganda does not have a population norm, the norm used was that of the Tanzania population while adapting the SF 36 [52]. Pre-coded numeric values from the questionnaire, on a Likert type scale, were recoded according to the scoring key given and items were then scored in such a way that the higher the score, the better the health state [51]. Each item was scored on a 0 to 100 range with 0 being the lowest and 100 the highest possible score. The summary statistics (i.e. Physical and Mental Summary Scores, PCS and MCS respectively) were calculated by averaging the different normalized scales. Finally, the summary scores were then standardized to give a mean of 50 and a standard deviation of 10 by multiplying the PCS and MCS scores by 10 and adding 50 to the product. . Higher scores indicate better health on each of the sub-scales [22, 51, 53]
The tool has been used in Uganda and other East Africa countries with good reliability measures [52, 54, 55]. The tool was translated into Runyankore, the local language in southwestern Uganda), using recommended guidelines (Maxwell, 1996; Peters and Passchier, 2006; Gudmundsson, 2009). We adjusted the SF-36 to fit the context of the study setting. Several phrases and actions in the physical activity section (i.e. climbing a flight of stairs, playing golf, walking one block, walking several blocks and pushing a vacuum cleaner) were replaced with appropriate activities that suit the local context. For example, climbing a flight of stairs was changed to climbing a hill, pushing a vacuum cleaner changed to lifting a 20-litre can of water. The adaptations were made following the Swahili version of the SF-36 that was adapted for use in Tanzania [52].
Ethics
The study was approved by the Research Ethics Committee (REC) of Mbarara University of Science and Technology (#19/11-16) and Uganda National Council for Science and Technology (# SS4309). All study participants provided written informed consent and received no facilitation for participation in the study. The data collected were confidential and anonymous with no information linking the study participants to the data. Participants who got severe distress during the interview were referred to a counsellor for appropriate care.
Data analysis
We summarized nominal variables as frequencies and percentages while numerical variables were summarized as means and standard deviation (SD). We carried out a bivariate analysis using logistic regression for all the predictor variables and the outcome variables. The outcome variables were the physical and mental component categories of the quality of life scale.
. Quality of life summary scores (i.e. physical (PCS) and mental (MCS) summary Scores) were calculated by averaging the different normalized scales from the relevant subscales. Values below 50 indicated poor health-related quality of life and those above 50 indicated a good quality of life.[50, 51]. We recorded unadjusted odds ratios, p-values and confidence intervals and summarized findings in a tables. We controlled for sex and age to determine adjusted odds, at 95% confidence intervals and statistical significance was determined at a p-value of <0.05. We then ran a multivariable logistic regression analysis with all the independent variables by including in the model all factors that had a p-value of ≤0.3 at bivariate analysis.