A scoring system to predict in-hospital death in oldest-old patients with infections in Taiwan

BackgroundThe elderly population has the highest number of in-patient claims accounting for over 40% of in-patient medical expenses in Taiwan, and infectious diseases are one of the most common causes of hospitalization in elderly patients. The aims of the study were to identify independent predictors of mortality, and to develop a simple scoring system for use in a busy clinical medical setting.MethodsAll patients aged 75 years and older presenting to our emergency department or outpatient clinics who were subsequently admitted due to infections between March and July 2009 were enrolled.ResultsA total of 981 patients were included, and the mortality rate was 9.2%. Logistic regression revealed seven independent predictors of mortality [odds ratio (95% confidence interval)]: albumin <3 g/dL [3.72 (1.97–7.00)], creatinine >1.5 mg/dL [2.94 (1.69–5.09)], total dependence on admission [1.64 (1.11–2.43)], systolic blood pressure <100 mmHg [4.46 (2.02–9.84)], white blood cell count ≥10 or ≤4 × 109/L [2.23 (1.23–4.04)], total bilirubin >1.2 mg/dL [2.37 (1.31–4.31)], and malignancy history [2.73 (1.4–5.30)]. These were given weighting by discriminant analysis and used to create a receiver operating curve with an area under the curve of 0.820. The model was simplified and each variable was assigned scores of 3 or 4 to form an index predicting in-hospital mortality. Mortality increased with score: 0.9% for score <5; 6.9% for 6 to 11; 21.4% for 12 to15, 53.7% for 16 to 19, and all three subjects whose score was >19 died during hospitalization.ConclusionOur scoring system derived from seven readily available variables including activity level, malignancy history, systolic blood pressure, and a small number of laboratory parameters predicted in-hospital mortality in oldest-old patients with infections. This system may allow for the identification of high-risk patients soon after admission.