Predictive risk modelling in health: options for New Zealand and Australia

Abstract

Predictive risk models (PRMs) are case-finding tools that enable health care systems to identify patients at risk of expensive and potentially avoidable events such as emergency hospitalisation. Examples include the PARR (Patients-at-Risk-of-Rehospitalisation) tool and Combined Predictive Model used by the National Health Service in England. When such models are coupled with an appropriate preventive intervention designed to avert the adverse event, they represent a useful strategy for improving the cost-effectiveness of preventive health care. This article reviews the current knowledge about PRMs and explores some of the issues surrounding the potential introduction of a PRM to a public health system. We make a particular case for New Zealand, but also consider issues that are relevant to Australia. What is known about the topic? PRMs are an alternative method to threshold modelling and clinical knowledge for determining a patient’s risk of a future event. PRMs are already in use in New Zealand and Australia to predict the occurrence of a disease. However, Kaiser Permanente in the US, and the UK’s National Health Service are using PRMs to predict health service usage (e.g. risk of future emergency hospitalisation) at the individual level. What does this paper add? This paper discusses issues including model parameters, data requirements and ethical considerations for using a PRM as a service planning tool in Australia and New Zealand. What are the implications for practitioners? PRMs could be used as the health service equivalent of disease risk assessments. New Zealand and Australia already have routinely collected data that could be used to predict various adverse, costly and potentially preventable health service events.