Thirty-day mortality in critical care outreach patients with cancer: an investigative study of predictive factors related to outreach referral episodes.

AIM: To establish factors that predict outcome in critically ill, deteriorating cancer patients through critical care outreach referral episodes, characteristics and care reviews.
METHODS: A population-based prospective and retrospective study was undertaken with analysis exploring predictive factors regarding critically ill cancer patients referred to a critical care outreach team. Data collected included: diagnosis; presenting problem; early warning scores at referral and at deterioration; physiological and observation data; admission to critical care, length of stay; 30-day mortality; limitation of care including precipitating DNAR orders and documentation of not for CCU admission/intervention).
RESULTS: Data were collected on 407 episodes from 318 patients over a period of 8 months from 2006 to 2007. Outreach initiated decisions to limit care with medical teams in 32.2% (n=103/318) of all patients. Early warning scores were not predictive of outcome. A high heart rate at referral (HR), a high potassium, low SpO2 at time of deterioration were independently predictive of 30-day mortality. The logistic regression (LR) model, using these three variables correctly predicts the 30-day outcome of 71% of the patients, demonstrating a relatively high predictability in this patient population. The odds of mortality increase with a higher potassium, heart rate and as the oxygen saturation at deterioration (DSpO(2)) worsen. Management factors included limitation of care, which is highly associated with 30-day mortality. Cancer patients recently receiving chemotherapy may have an increased mortality once admitted to critical care. Being a haemato-oncology patient, or the timeliness of critical care outreach referral does not appear to affect 30-day mortality.
CONCLUSION: The LR model was able to predict 30-day outcome of 71% of the patients, demonstrating a reasonably high predictability in this cancer patient population. Critical care outreach initiated discussions on limiting treatment which had an effect on mortality.