Theme: Multilevel Modelling
14.00 – 14.50: Professor Fiona Steele, Co-Director, Centre for Multilevel Modelling, University of Bristol
Title: “Adjusting for Selection Bias in Longitudinal Analyses of the Relationship between Employment Transitions and Health Using Simultaneous Equations Modelling”
There is substantial interest in understanding the association between labour force participation and mental health, and in particular the impact of unemployment on wellbeing. While panel data allow detailed examination of the dynamics of the relationship between changes in labour force participation and mental health, selection bias remains a serious concern. We test for two types of selection effect: (i) direct selection (where prior health affects employment status), and (ii) indirect selection (due to unmeasured characteristics influencing both health and employment outcomes). We then examine the impact of adjusting for selection biases on estimates of the effect of employment transitions on mental health. We investigate the relationship between men’s employment transitions and mental health using data from the British Household Panel Survey, 1991-2009. We model the effect of a change in employment status between years t-1 and t on mental health at t, adjusting for mental health at t-1. Using a dynamic simultaneous equations model we allow explicitly for an effect of health at t-1 on employment transitions between t-1 and t to allow for direct selection. The health and employment equations include individual-specific random effects which are correlated across equations to allow for indirect selection due to shared unmeasured influences.
14.50 – 15.40: Professor Chris Roberts, Centre for Biostatistics, University of Manchester
Title: “Partial Nesting: Some Implications for Design and Analysis”
In some setting heterogeneity of clustering effects is confounded with treatment effect estimation. This can occur in trials evaluating group treatments and non-pharmacological therapies. One example of this partial nesting in which a clustering occurs in one group but not in the comparator leading to clusters of size one in a multilevel analysis. This talk will consider the statistical design and analysis implications for binary and continuous outcome measures, presenting the results of a series of simulation studies.
15.40 – 16.00: Tea/coffee break
16.00 – 16.50: Dr Wendy Harrison, Leeds Institute of Genetics, Health and Therapeutics (LIGHT), University of Leeds
Title: “Multilevel latent class casemix modelling: a novel approach to accommodate patient casemix”
Using routinely collected patient data we explore the utility of multilevel latent class (MLLC) models to adjust for patient casemix and rank NHS Trust performance. We contrast this with ranks derived from Trust standardised mortality ratios (SMRs). Our outcome is survival at 3 years after diagnosis and we adjust for patient age, sex, stage at diagnosis and socioeconomic status. The multilevel latent class analysis identified two patient classes and two Trust classes, and differences in the ranked Trust performance between the MLLC model and SMRs were all within estimated 95% confidence intervals. This approach to casemix adjustment allows us to rank Trust performance whilst facilitating the evaluation of factors associated with the patient journey (e.g. treatments) and factors associated with the processes of healthcare delivery (e.g. delays). Further research can demonstrate the value of modelling patient pathways and evaluating healthcare processes across provider institutions.