BACKGROUND: To determine whether the serum osmolarity, calculated at the time of an emergency medical presentation, alone or combined with other predictors, could identify patients at low and high risks of an inpatient death by day 30. METHODS: A retrospective analysis of all emergency medical patients admitted to St. James's Hospital (SJH), Dublin between the 1st of January 2002 and the 31st of December 2009, using the hospital in-patient enquiry (HIPE) system, linked to the patient administration system, and laboratory datasets. Hospital inpatient mortality (30 days) was obtained from a database of deaths over the same period. Multivariate logistic regression was used to derive the best predictive model, with goodness of fit and Area Under the Receiver Operator Curves (AUROC) assessing the predictive accuracy. RESULTS: Univariate analysis identified two quantiles of <10% and >75% of the osmolarity distribution as being at an increased mortality risk. Their respective mortality rates were 13.7% and 15.7% respectively, with unadjusted odds rate that were 1.66 (1.47, 1.88): p<0.0001 and 3.14 (2.87, 3.43): p<0.0001. After adjustment for other outcome predictors, a significant association with increased mortality remained, with OR=1.82 (1.61, 2.06), p<0.0001. Although the calculated osmolarity alone was not sufficiently predictive with AUROC=0.74 (95% CI: 0.73, 0.76), when combined with other predictors, the AUROC increased to 0.86 (95% CI: 0.84, 0.88). CONCLUSION: Admission osmolarity, a simple calculation, is associated with the risk of mortality in acutely ill medical patients; deviations outside the normal range are relevant. A useful clinical predictive algorithm requires the incorporation of additional predictors.
BACKGROUND: To determine whether the serum osmolarity, calculated at the time of an emergency medical presentation, alone or combined with other predictors, could identify patients at low and high risks of an inpatient death by day 30. METHODS: A retrospective analysis of all emergency medical patients admitted to St. James's Hospital (SJH), Dublin between the 1st of January 2002 and the 31st of December 2009, using the hospital in-patient enquiry (HIPE) system, linked to the patient administration system, and laboratory datasets. Hospital inpatient mortality (30 days) was obtained from a database of deaths over the same period. Multivariate logistic regression was used to derive the best predictive model, with goodness of fit and Area Under the Receiver Operator Curves (AUROC) assessing the predictive accuracy. RESULTS: Univariate analysis identified two quantiles of <10% and >75% of the osmolarity distribution as being at an increased mortality risk. Their respective mortality rates were 13.7% and 15.7% respectively, with unadjusted odds rate that were 1.66 (1.47, 1.88): p<0.0001 and 3.14 (2.87, 3.43): p<0.0001. After adjustment for other outcome predictors, a significant association with increased mortality remained, with OR=1.82 (1.61, 2.06), p<0.0001. Although the calculated osmolarity alone was not sufficiently predictive with AUROC=0.74 (95% CI: 0.73, 0.76), when combined with other predictors, the AUROC increased to 0.86 (95% CI: 0.84, 0.88). CONCLUSION: Admission osmolarity, a simple calculation, is associated with the risk of mortality in acutely ill medical patients; deviations outside the normal range are relevant. A useful clinical predictive algorithm requires the incorporation of additional predictors.