Data
Data came from the HSE, a routine cross sectional survey that draws a nationally representative sample of persons residing in private households in England, which is openly available. The sample and focus of the survey vary each year. Data from the 2008 survey was used in this study and included a sample of 9,191 households with 15,102 adults aged 16 or over, and a total child sample of 7,521. Households were sampled proportionately across the 9 Government Office regions of England. Sampling was based on a multi-stage stratified random sampling design that used the postcode address file as a sampling frame. To improve power of analysis, boost samples and sampling weights were employed appropriately [4]. The primary focus of HSE 2008 was PA and fitness. The method of data collection involved face-to-face interviews, self-completion, clinical and physical measurements (including objective measurements of PA via accelerometers). Tools for data collection were validated in a pilot study conducted in 2007 prior to the main survey (further details can be found in Craig et al. [4]). To compensate for seasonal variation in responses, the time period for interviews covered January-December 2008. This study draws on 5,537 observations that constitute 40–60 year olds among the adult sample.
HRQoL
HRQoL is measured in the HSE using the summary measure of health state utility value derived from the validated EuroQol-5 Dimensions (EQ-5D) [13]. These utility scores were generated using the descriptive system of the EQ-5D questionnaire (UK version), a standard HRQoL instrument with preference weights which are attached to combinations of responses. The EQ-5D descriptive system describes HRQoL in five dimensions (i.e. mobility, self-care, usual activities, pain/discomfort, and anxiety/depression) with each dimension including three levels: no problems, some/moderate problems, and severe/extreme problems. Different health states are created from the responses to the descriptive system of the EQ-5D by combining one level from each of the dimensions. A tariff is then applied to these health states to generate utility scores [14]. The utility scores usually range from ‘1’ (perfect health) to ‘0’ (death), with states perceived to be worse than death having a negative utility score.
Physical activity
PA was accessed via a composite indicator, reflecting a combination of types of PA (i.e. walking, housework, occupational activity, and sports and exercise) and captured through subjective, and objective measurements. In HSE 2008, the subjective measure of PA was assessed through a validated questionnaire and the objective measure the most widely used accelerometer, actigraph (model GT1M), which is relatively portable (due to its small size) and associated with low running costs [4]. Accelerometer is a favoured method of objective measurement of physical activity given that it has an increased capacity to capture varied movements [4]. A pilot study was conducted to examine the feasibility of using the actigraph was and it was found to perform well. In the main survey, the actigraph was worn (at the waist) by a randomly selected sub-sample of the HSE 2008 respondents (4,507 adults – 16 plus years). Respondents were told to wear the actigraph during waking hours for 7 consecutive days following an initial discussion with interviewers who explained the use of the device and gave respondents phone numbers to call for assistance when faced with difficulties using the device. A book was given to participants to log daily use of the actigraph (duration of use/non-use). Daily use was considered ‘valid’ if the actigraph was worn for at least 10 hours. Kine software (3.0.98) was used to analyse the raw accelerometry data to generate a standardised measures. Further details on the use of accelerometer in the HSE 2008 can be found in Craig et al. [4].
To test the robustness of the relationship between PA and HRQoL across specific types of PA, separate measures were included for specific types of PA (i.e. walking, and sports and exercise). Data for walking included self-reports on all walking (regardless of intensity) ranging from country walks, walking to and from work/school, and any other walks. Sports and exercise covered self-reports on activities such as swimming, cycling, aerobics, workout at a gym, running, team sports, and press/sit-ups. These two types of PA were chosen and analysed separately because they reflect the main types of exercise referral interventions [8].
Each PA measure was operationalised as a binary variable that takes the value of one if ‘physically active’ (minimum of 90 minutes of at least moderate intensive PA was undertaken per week) or zero otherwise (not physically active). This definition of being ‘physically active’ is consistent with the literature on exercise referral interventions and PA for health guidance [15]. Based on previous research [11, 12, 16], we hypothesise that being ‘physically active’ would be positively correlated with HRQoL.
Covariates
The covariates were socio-demographic, economic, health and other variables that in previous research had been shown to correlate with HRQoL: gender, age, income, educational qualification, employment status, ethnicity, marital status, house tenure, smoking status, drinking status, region of residence, urbanisation, general health status, BMI, morbidities (e.g. problems with heart, muscoskeletal, ear, vision, mental, hypertension, stroke, diabetes), psycho-social wellbeing (measured via General Health Questionnaire scores) [17–23].
Analysis
Means (standard deviation - SD) and proportions were calculated for continuous and categorical data respectively. Chi square and Fischer’s exact tests were used to check the association between HRQoL (dependent variable) and dummy variables representing item non-response for independent variables in order to examine the mechanisms under which the missing data occurred (i.e. missing completely at random or not) [24]. If the pattern of missing data did not occur completely at random (i.e. HRQoL is significantly associated with item non-response for independent variable(s), a regression-based imputation method was used to replace missing values of continuous variables and a dummy variable specifying item-non response added. For the categorical variables, item non-response was included in the omitted category and a dummy variable for item non-response created [25].
To account for censoring in the measurement of HRQoL [26], Tobit regressions with upper censoring at 1.0 and robust standard errors were employed to model the relationship between HRQoL and PA controlling for covariates. Separate models were fitted for each indicator of PA to avoid unstable estimates resulting from the collinearity among those indicators. Hence four different models were estimated: model 1 (walking); model 2 (sport and exercise); model 3 (objective measurement); model 4 (subjective measurement). Hereafter, the models will be referred to by these names. Each model had two versions: (a) model excluding missing observations; and (b) model including imputed missing observations. To allow comparability between subjective and objective measures of PA, models 3 and 4 were estimated using the same sample (i.e. those with data for both subjective and objective measures).
The models were estimated with sampling weights that were calculated as the inverse of the probability of being a respondent in a household multiplied by the household weight which accounts for non-responding households [4]. Specification errors and goodness of fit of models were examined using the linktest [27], and Akaike Information Criteria (AIC) and Bayesian Information Criteria (BIC) [28] respectively. Pseudo R2 was computed by calculating the R2 between the predicted and observed values [29] and multicollinearity among independent variables was assessed [30, 31]. Threshold for statistical significance was set at ≤ 0.10 and analysis was undertaken in 2010 using Stata version 10.