Chris Schilling - An Australian microsimulation model of osteoarthritis

  • Presenting author: Chris Schilling (University of Melbourne)

  • Authors: Chris Schilling, Yushy Zhou, Cade Schadbolt, Sid Rele, Michelle Dowsey, Peter Choong

  • Session: C02B - Health [3] - Wednesday 11:00-12:30 - Marietta-Blau Hall

Introduction

Simulation modelling is a powerful tool for assessing the impact of diseases and guiding healthcare policy decisions. This paper introduces AUS-OA, a dynamic microsimulation model of osteoarthritis and its treatment in Australia. AUS-OA builds on a range of datasets and expertise drawn over many decades, with the aim of providing policy makers, healthcare providers and patients with valuable insights into the burden and treatment of osteoarthritis across Australia.

Methodology

A synthetic population of individuals with age, sex, BMI, osteoarthritis and other attributes was derived from the Household, Income and Labour Dynamics in Australia (HILDA) survey, and weighted to the Australian population. BMI changes over time were modelled based on previous analysis that differentiates BMI progression by age, gender, socio-economic position and current BMI. Osteoarthritis was characterised using the Kellgren-Lawrence scale, with incidence and progression derived from analysis of HILDA and the Osteoarthritis Initiative (OAI) longitudinal studies. Joint replacement as a final treatment for osteoarthritis was modelled based on a proportional hazards competing risk model. Primary care and pharmaceutical utilisation and costs were derived from analysis of the SMART registry which links joint replacement data with the Medicare Benefits Schedule (MBS) and the Pharmaceutical Benefits Schedule (PBS). Both HILDA and the SMART registry were used to derive health-related quality of life outcomes based on an individual’s health conditions and treatments. AUS-OA was calibrated with observed data between 2013 and 2022. Base case and ‘what-if’ scenarios can then be simulated from 2023 until 2053, to evaluate the impact of socio-demographic trends, changes in costs and advances in treatment and models of care for osteoarthritis. For an initial application, AUS-OA was used in an economic evaluation of a decision support tool that improves patient selection into surgery.

Results

Comparisons of univariate and bivariate distributions across key population attributions (e.g. age, BMI, osteoarthritis, joint replacement) between the synthetic and actual populations highlight that the AUS-OA model accurately reflects the Australian population over the calibration period. The evaluation of the decision support tool highlighted both an increase in Quality-Adjusted Life Years and a reduction in health costs as a result of improved patient selection into surgery, relative to no use of the tool. A probabilistic sensitivity analysis showed that the tool is a dominant strategy for the majority of key epidemiological and economic parameter combinations.

Conclusion

AUS-OA combines research about the epidemiology and health economics of osteoarthritis and its treatments, into a consistent and repeatable analytical framework. This allows testing of policy change, improvements to models of care and development of innovative treatments.