Variables used in predictive risk models
| Model 1 ‘Full’ model | Model 2 Primary care setting | Model 3 Hospital setting | Sensitivity analysis: Model 2 with GP use | |
| Sociodemographics/social support | ||||
| Age, gender, Aboriginal status, geographic remoteness, SES, marital status | X | X | X | X |
| Language spoken at home | X | X | – | X |
| Income, social isolation | X | – | – | |
| Health status and health behaviours | ||||
| Health condition count | X | X | X | X |
| Self-rated health, polypharmacy, anxiety/depression, BMI, smoking status, unsafe alcohol use | X | X | – | X |
| Functional status, falls | X | – | – | |
| Prior health service utilisation | ||||
| Previous hospitalisations, admission via ED, length of stay of previous hospitalisation, previous ED use | X | – | X | – |
| Primary care accessibility | ||||
| GP use | X | – | – | X |
| Total number of variables | 23 | 14 | 11 | 15 |
The survey variables ‘health condition count’ and ‘self-rated health’ were used to approximate the patient health history information that is commonly captured in GP practice management software.
BMI, body mass index; ED, emergency department; GP, general practice; SES, socioeconomic status.