A Model to Predict In-hospital Mortality in Elderly Patients with Community-acquired Pneumonia: a Retrospective Study

Abstract

ObjectivesTo develop and validate a predictive model for evaluating in-hospital mortality risk in elderly patients with community-acquired pneumonia.SettingTwo tertiary care hospitals with 2000 beds and 1000 beds respectively in Beijing, China.ParticipantsElderly patients (age ≥ 65 years) with community-acquired pneumonia admitted to the Department of Internal Medicine of the two hospitals from January 2010 to December 2019 or from January 2019 to December 2019 respectively.DesignIt was a retrospective study. After dividing the data set into training and validation sets, we created a mortality model that included patient demographics, hospitalization time, hospital outcome, and various clinical conditions associated with hospitalization. We then applied the model to the validation set.Main outcome measuresIn-hospital mortality.ResultsThe training cohort included 2,466 patients admitted to the one hospital, while the validation cohort included 424 patients at the other hospital. The overall in-hospital mortality rate was 15.6%. In the multivariable model, age, respiratory failure, heart failure, and malignant tumors were associated with mortality. The model had excellent discrimination with AUC values of 0.877 and 0.930 in the training and validation cohorts, respectively.ConclusionsThe predictive model to evaluate in-hospital mortality in elderly patients with community-acquired pneumonia, which was established based on administrative data, provides a simple tool for physicians to assess the prognosis of elderly patients with community-acquired pneumonia in Beijing.Strengths and limitations of this studyThe clinical data collected were obtained from a large retrospective cohort over a 10-year period, resulting in good reliability.The model based on administrative data can help healthcare workers determine the prognosis and outcomes of elderly patients of CAP, especially in resource limited regions in China.Our main outcome was in-hospital mortality, not 30-day mortality or longer.All included cases were collected from inpatient, outpatient and emergency patients were not included.The model was only verified in two hospitals in Beijing and further verification should be conducted in other areas and different levels of hospitals.