From February 2017, information about the work of the MRC/CSO Social and Public Health Sciences Unit, University of Glasgow is available and updated on the University of Glasgow website.

Addressing non-response bias in health survey data

Reliable measures of alcohol intake are required to formulate and evaluate policies aimed at improving and maintaining population health and wellbeing. Alcohol retail data are limited to overall per capita consumption, but richer data can be obtained from national surveys such as the Scottish Health Surveys (SHeS). Survey data validity is dependent on their representativeness of the general population, but response levels are declining potentially making the surveys less representative. Currently, the sole attempt to address bias arising from such non-response is based on limited socio-demographic characteristics, and does not capture or adjust for health aspects, or allow for differences between responders and non-responders.

We aim to inform the monitoring and evaluation of the Scottish Goverment's Alcohol Strategy in the following ways:
  • To quantify non-response bias in SHeS by socio-demographic sub-groups, and assess the extent of representativeness by varying response levels;
  • To use advanced statistical methodology to address survey non-response induced imprecision of SHeS-based alcohol consumption estimates, with reference to general population estimates, and alcohol-related morbidity and mortality data;
  • To develop advanced generalised correction procedure to produce refined estimates of alcohol consumption.

Work completed so far has compared the alcohol-related harms observed in the follow-up of the SHeS participants with those seen in the general population, and uses insight on differentials to adjust survey-derived alcohol consumption estimates using advanced statistical techniques including multiple imputation. We found the odds of all cause mortality to be 11% lower among survey respondents compared to the general population and the odds of alcohol-related harm to be 31% lower. Our methodology led to increased estimates of consumption of between 6.5% and 9.6% among men (depending on the survey year) and between 0.0% and 2.2% among women, with this increase being most marked among males living in deprived areas.

Publications

2014

Gorman E, Leyland AH, McCartney G, White IR, Katikireddi SV, Rutherford L, Graham L, Gray L. Assessing the representativeness of population-sampled health surveys through linkage to administrative data on alcohol-related outcomes. American Journal of Epidemiology 2014;180 :941-948

open access  

2013

Gray L, White IR, McCartney G, Katikireddi SV, Rutherford L, Gorman E, Leyland AH. Use of record-linkage to handle non-response and improve alcohol consumption estimates in health survey data: a study protocol. BMJ Open 2013;3:e002647

open access