Study Design
Who is in the Cohort?
The sampling unit was a family living in the same household, operationally defined as all individuals who live at the same address and identify each other as family members. The census defines household in terms of a group of persons who live together and need not be related, and make common provision for essentials for living.
All family members aged 10 years or older were eligible and invited to participate. Those aged 15 years or above completed the main questionnaire, while 10 to 14 year-olds completed a modified survey designed for adolescents. A household was enrolled in the cohort when all eligible members agreed to participate in the Wave 1 (Baseline) survey.
The cohort was compiled as a composite sample from several sources to cover about 20,000 households, or just under 1% of Hong Kong’s 2,368,796 households.
First, the “random core sample” was obtained using stratified random sampling, where residential addresses were randomly sampled in all 18 districts in Hong Kong with sample sizes proportionate to each of the district populations.
Second, once a household had been enrolled in the random core sample, the household head was invited to refer first degree relatives to also participate, in keeping with our study aim of assessing the extended (or non-household based) family unit.
Third and additionally, we over-sampled residents of three new towns (Tung Chung, Tin Shui Wai, and Tseung Kwan O) by randomly selecting an additional 2891 households. These towns are in relatively more remote locations and contain larger numbers of new mainland Chinese migrant families.
Fourth, we undertook purposive sampling of three special groups of interest who have undergone significant social readjustment in their recent life course with anticipated changes in family dynamics: couples married within the past six months, households with at least one member who has been hospitalised due to a serious illness (defined as congestive heart failure, coronary heart disease, cancer, hip fracture, or stroke) within the past six months, and households with a child entering primary school.
Fifth, we over-sampled six districts with the smallest number of residents so that they had sufficient sample size for district-specific reports.
Finally, we drew a sample of households where we randomly selected one member from each such household to complete the survey, as a comparator group in view of the full-household enrolment design.
From the six sources together (random core sample, first-degree relatives, new towns, randomly selected household members, over-sampled special groups and districts), the FAMILY Cohort included 20,279 households (46,001 participants).
How often have they been followed up?
Seven data collection time points have been completed from 2009 to 2014.
Wave 1 (Baseline) of household visits with interviewer-administered questionnaires was conducted from March 2009 to April 2011, and Wave 2 (Follow-up) was conducted from August 2011 to March 2014.
The overall follow-up rate for Wave 2 (Follow-up) was 69.6% (15 155 households, 32 016 participants), and the follow-up rates for random core, three new towns and randomly selected household members were 70.7%, 70.7% and 70.2% respectively. New eligible family members were also recruited in Wave 2 (Follow-up).
Five web-based and telephone follow-up questionnaires were separately administered at three, nine, and fifteen months following Wave 1 (Baseline), and four and eight months following Wave 2 (Follow-up). Out of the 42 489 age ≥ 15 participants who were eligible to be followed after Wave 1 (Baseline), 30 551 (71.9%), 32 110 (75.6%) and 30 738 (72.3%) completed the 3-month, 9-month and 15-month telephone/web-based survey follow-ups, respectively. Out of 32 418 participants who agreed to be followed after Wave 2 (Follow-up), 24 264 (74.8%) and 22 775 (70.2%) completed the 4-month and 8-month telephone/web-based survey, respectively.
What has been measured?
Wave 1 (Baseline) and Wave 2 (Follow-up) household visits consisted of a one-on-one home interview, not self-administered, with each eligible family member by trained interviewers who entered the data directly into tablet PCs. The interview lasted about one hour and covered lifestyle factors, a range of health indicators, happiness and family harmony, in addition to demographic and socioeconomic information.
Health indicators consisted of self-reported symptoms, medical history, and validated self-report scales of quality of life and depression.
Anthropometric measurements included weight, height, waist and hip circumference, body fat percentage, visceral fat level, heart rate, systolic and diastolic blood pressure.
Few psychological measurement scales were originally developed for use in Chinese populations, and imported instruments from individualistic cultures are unlikely to adequately capture the importance of family in the Chinese context. In particular, we proposed that family relationships are a central and specific axis for determining well-being, over measures of individual function that are more readily available and drive models that have been developed in more individualistic cultures. Our scale development efforts were guided by the lack of specific and indigenously developed of family harmony that are accessible to a population regardless of their education level, and are relatively brief.