Objectives: to examine the relationship between seven predictor variables (recorded on Day 3 of hospital admission) and discharge destination in non-elective medical patients aged 65+ years.
Design: prospective cohort.
Setting: eight centres in six European countries.
Predictor variables: age, gender, living alone, physical function (three categories based on Barthel Index), cognition (Katzman's orientation-memory-concentration test), main body system affected (based on International Classification of Diseases), number of geriatric giants (GGs) involved in the referral (a GG being a problem with falling, mobility, continence or cognition).
Main outcome measures: discharge destination (by Day 90) in three categories: 'HOMESAME' (return to previous residence), 'INSTIN90' (discharge to alternative residence or still in hospital at 90 days), 'DEADINHO' (death in hospital),
Results: in 1,626 patients, discharge destination was HOMESAME in 84.7%, DEADINHO in 8.9% and INSTIN90 in 6.4%. Mean duration of stay was 17.7 days, median 12. Univariate analyses showed a statistically significant relationship between all seven predictor variables and discharge destination. Physical function was the best single predictor with a seven-fold difference in adverse outcome rates between the best and worst categories. On multiple logistic regression, significant predictor variables were as follows. (i) For DEADINHO: physical function, cognition, gender; (ii) for INSTIN90: physical function, living alone, GGs, age, gender. Multiple linear regression identified physical function, GGs and living alone as predictors of loge length of stay.
Conclusion: case-mix systems to compare risk-adjusted hospital outcome in older medical patients need to incorporate information about physical function, cognition and presenting problems in addition to diagnosis.