Exploration of the use of decision analytic modelling in low back pain and sciatica

Abstract

Low back pain (LBP) is a global public health problem. Keele University developed the STarT Back tool to stratify LBP patients according to their risk of persistent disability, matching treatments to individual risk. A 12-month trial-based economic evaluation showed this stratified care model to be cost-effective. A recent trial, the SCOPiC trial, aimed to evaluate a modified stratified care model for sciatica patients consulting in primary care. However, the longer-term cost effectiveness of both care models is unknown.
To estimate the long-term cost-effectiveness of stratified care, two separate decision models were developed. The model conceptualisation process included expert consultations, and two systematic literature reviews assessing the use of decision analytic modelling in LBP and sciatica, and stratified care.
A de-novo state-transition cohort model was developed to estimate the cost-utility of stratified care for the management of LBP in primary care, from the NHS perspective, over a ten-year horizon. Model results provided support for the cost-effectiveness of the Keele stratified care model.
A de-novo individual-level simulation model was chosen to estimate the cost-utility of stratified care vs best usual care vs usual care for the management of those consulting with sciatica in primary care, from the NHS perspective, over a ten-year horizon. Model results suggest this model of stratified care is not cost effectiveness relative to best usual care.
Both cost-effectiveness results were robust to structural assumptions, however, sensitivity analyses highlighted how assumptions regarding health states, long-term patient prognosis and EQ-5D values could affect cost effectiveness results. Furthermore, the first Expected Value of Perfect Parameter Information (EVPPI) analyses in decision modelling for LBP and sciatica highlight the value of further research exploring transitons between health states.
The thesis concludes with recommendations for modelling in low back pain and sciatica, including the need to strengthen modelling methodologies and fully explore structural and parameter uncertainty.