Study design and setting
We conducted a population based cross sectional study in all 52 Union Councils of District Abbottabad, Khyber Pakhtunkhwa province, Pakistan from March 2015 to August 2015. We recruited 2063 participants for our study. Abbottabad is the main district of Khyber Pakhtunkhwa province of Pakistan having more than 1.2 million population living in 52 union councils. The primary language spoken here is Hindko (used by 94% of the rural population and 75% of urban residents) followed by Urdu which is also spoken and understood in rural & urban areas [17].
Sample size
We used WHO software for Sample Size Determination in Health Studies. A difference of 2.5 in quality of life score was considered as significant between two groups. Pooled standard deviation of scores of two groups was taken as 12 from the pilot study. To detect this difference at 95% confidence level with 90% power with design effect of 1.8 and 10% inflation for missing information, the final calculated sample size was 968 in each group, which makes the total sample size of 1936 participants.
Sampling plan
Participants of this study were selected from all union councils (UCs) of District Abbottabad. Multi – stage cluster sampling technique was employed in this study. Each union council was further divided into several blocks called Mohallah. We did proportionate sampling according to the 1998 population census [18] of UCs for the selection of Mohallah & on the next stage households. In the first stage we randomly selected these blocks (Mohallah) in each of the UC from a list by using simple random sampling technique. In the next stage we selected households in that selected block by using a random sampling technique again. The total number of houses selected in each block was also proportional to the population size of respective block. For the selection of family type, from the list of household of each block, we made a list of joint & nuclear family system households and enrolled equal number of houses from both family types. A simple random sampling technique was used for the selection of person (≥18 years) from each house. Simple random sampling was done by applying the lottery method for selecting the ≥18 year’s participant for the study. The inclusion criteria used for selection of individual were age greater than 18 years and permanent resident of union council for at least 5 years. Guests and temporary residents were excluded from the study.
Study tool
We used, WHOQOL-BREF, a 26-item, self-administrated generic questionnaire that covers four domains of QOL (psychological 6 items, physical 7 items, social relationships 3 items and environmental 8 items). Each question scored on a scale from one to five, with high score indicating good QOL with the exception of three questions, which include pain and discomfort, need for medical treatment and negative feelings [19]. The seven items included in the physical health domain were mobility, daily life activities, pain, sleep, functional capacity and energy. The psychological domain measured negative thinking, self-image, positive approach, self-esteem, mindset, ability to learn, memory, consolidation, religion and the psychic conditions. Questions such as social support, sex-life and personal relationship come under the social relationship domain. The environmental health domain contains questions on financial assets, security, health and social services, living in natural environment, opportunities for advance learning experience, relaxation, and natural environment (air, noise, pollution and transportation) [20]. The total raw score for these five dimensions were transformed into 0 to 100 scale according to the standard procedure defined in WHO QOL user manual [19], and then analysis of this reconstruct score was done. Psychometric properties and validation of this WHOQOL-BREF questionnaire was done in National language “Urdu”. Cronbach’s alpha for each of five domains were 0.78, 0.71, 0.73 and 0.56 respectively [21]. To assess the feasibility and clarity of the items, a pilot study was conducted on 30 individuals conveniently selected from the study area. We also developed a structured demographic questionnaire which included variables such as age, gender, marital status, type of family (joint and nuclear), residence type (urban and rural), house ownership (owner, not owner), respondent education (no education, madrassa, can read/write, primary- up to grade 5, secondary education-up to grade 12 and tertiary-up to grade 16 or above), working status (employed, unemployed and retired).
The socio-economic characteristics were assessed by taking household conditions, sources of drinking water, sanitation facilities, availability of electricity, housing facilitates, possession of durable goods, mean of transport, inventory of house hold and personal items such as chairs, clocks, buckets, radios, television sets, fans, stoves or cookers, cars, and telephones. This list was composed of 21 such items used in Pakistan demographic and health survey in 2013 [22]. Wealth index was measured by an index constructed from principle component analysis (PCA) [23] of items indicating ownership of household durables and dwelling characteristics.
World Bank’s Social Capital Integrated Questionnaire (SC-IQ) was also used to study social capital among families. SC-IQ consists of 27 questions in six domains [24]. With the help of subject experts, we extracted five questions from the core questionnaire with having Cronbach’s alpha 0.64. These five selected questions were overall trust, trust in local government, trust in central government, community cooperation and safety at home.
Data collection
One day training session was conducted for administering the questionnaires prior to data collection for lady health workers of all UCs by principal investigator. The questionnaire was administered through face to face interviews in the households by trained lady health workers of that union council. To ensure privacy and confidentiality, interviews were conducted in an independent room or area separate from other members of the family.
Data analysis and statistical methods
Data was analyzed using Stata version 13.0 (Stata Corp, College Station, TX, USA). Descriptive statistics such as frequencies and proportions and means with standard deviation were calculated. We carried out univariate and multivariate linear regression analyses with domain scores as dependent and other variables as independent variables. In the multivariate analysis, we included all the variables and then used stepwise backward approach to eliminate variables with a p value > 0.05. Muti-level analysis was performed with two –level continuous random intercept model with individuals nested within clusters was applied to explore the variability explained by individuals and cluster level variables taking the correlated nature of data into account. P-value of <0.05 was considered as significant.