Understanding Society Scientific Conference 2019 - , University of Essex

Is Understanding Society attrition related to dynamic processes of substantive interest to survey users and how should we address this? Trajectories of long-term health conditions and survey attrition

My Programme   Add 
Day:Thu 4 Jul
  • Katie Saunders, University of Cambridge

From a methodological side we already know that attrition in Understanding Society is associated with self-rated health, with people who rate their health as “excellent” and “very poor” most likely to drop out over time. However from the policy side the aging UK population means that people are increasingly living with more than one long-term health condition. There is a need for evidence (particularly on the clustering and trajectories of long-term conditions, and the interrelationship between multimorbidity, disability, age, ethnicity, and other social factors) to inform health service planning and design for people with multimorbidity. Understanding Society provides a unique resource for addressing this evidential need; with rich social and health data; concerns about survey non-response however remain one of the key barriers to policymaker and clinician engagement with survey findings.  Characterising, and providing appropriate methodological guidance to address, the relationship between longitudinal health trajectories, trajectories of multimorbidity and survey attrition in Understanding Society is a priority. Within this Understanding Society Survey methods fellowship I am taking three approaches to address this; estimation of transition probabilities for long term conditions (and survey attrition) between survey waves, and modelling of within-person trajectories of multimorbidity using linear mixed models and survival models. An initial methodological issue arises from the fact that self-reported long term conditions change over time at a higher rate than would be seen in clinical records; predictors of inaccurate self-reporting of health conditions overlap with predictors of survey attrition. I will present findings from my initial work and practical methodological advice for survey users.