From prediction to prevention: enhancing hospitalisation risk algorithms through primary care data integration   — The Association Specialists
  1. Schedule
  2. Oral Presentations - Strengthening system responsiveness to health crises
  3. From prediction to prevention: enhancing hospitalisation risk algorithms through primary care data integration  

From prediction to prevention: enhancing hospitalisation risk algorithms through primary care data integration   (23918)

Cathy O'Callaghan 1 , Margo Barr 1 , Taksh Patel 2
  1. International Centre for Future Health Systems, Kensington, NSW, Australia
  2. Big Data, UNSW, Kensington, NSW, Australia

Background: Hospitalisation risk prediction algorithms are widely used to identify patients for integrated care programs. However, evidence from the Integrated Care for People with Chronic Conditions (ICPCC) program in New South Wales showed that while algorithms effectively identify patients with prior hospital use, they may miss individuals in the community with rising risk who have not yet presented to hospital. Clinician referral was therefore a critical complementary mechanism for identifying patients with complex health and social needs not captured by the algorithm. 

Aim: To examine how engagement with general practice influences hospitalisation risk, and whether incorporating primary care data improves risk prediction and supports earlier intervention.

Methods: This study uses linked data from the 45 and Up Study, a population-based cohort of 267,357 adults aged 45 years and over in New South Wales, in a retrospective cohort design. The analytic population is expected to include approximately 140,000–260,000 participants, depending on inclusion criteria. Participants will be followed using a 12-month look-back and 12-month follow-up framework based on individual survey dates. The primary outcome is any hospitalisation within 12 months. Two models will be developed and compared: (1) a model using demographic, clinical and prior hospital utilisation variables, and (2) an enhanced model incorporating primary care indicators, including general practice utilisation and preventive care measures derived from Medicare Benefits Schedule (MBS) data. Linked datasets include hospital admissions, emergency department, MBS, PBS, and mortality records.

Results: Prior ICPCC findings demonstrate that reliance on hospital-based datasets limits early identification of risk, highlighting the importance of integrating primary care data and clinician-informed selection processes.

Implications: This study provides implementation insights from integrated care programs by testing whether linking general practice and hospital data improves earlier identification of risk. Findings will inform more equitable, preventive and scalable approaches to risk stratification and integrated care.