AIM AND METHODS There is no consensus definition on what constitutes a long stay in the intensive care unit, and little published information on the demographic characteristics, resource usage or outcomes of long-stay patients. We used data from the Australian and New Zealand Intensive Care Society Adult Patient Database to identify patients who had spent > 21 days in the ICU. We examined their resource usage, hospital type, diagnoses and outcomes, and trends in these characteristics over 5 years (2000-2004). RESULTS 6,565 patients (2.3% of all ICU patients) had one or more admissions > 21 days and accounted for 23% of total ICU bed-hour usage. Long-stay patients had a mean (SD) age of 60.3 (15.3) years and an APACHE III-J risk of death of 32.7% (21.3%). Metropolitan and tertiary hospitals had the highest proportions of long-stay patients. The three diagnoses most strongly associated with long ICU stay were neuromuscular disease (odds ratio [OR], 13.3; 95% CI, 10.2-17.4; P < 0.001), burns (OR, 6.0; 95% CI, 4.9-7.3; P < 0.001) and cervical spine injury (OR, 5.1; 95% CI, 3.4-7.5; P < 0.001), while the most common diagnosis was pneumonia (12.7% of total). During the period 2000- 2004, there was no significant change in the proportion, age, resource usage or outcomes of these patients. Overall observed mortality was 28% (predicted, 32.7%; 95% CI, 31.4%-34.5%). Of those aged >or= 80 years, 37% were discharged home, and 39% died. CONCLUSIONS Patients who spend > 21 days in the ICU use significant resources but appear to have worthwhile outcomes in all age brackets.