32 Derivation of a novel inpatient mortality prediction model for emergency department patients in singapore. (16th April 2018)

• Record Type: Journal Article
• Authors: Wu, Stella Xinzi; Liu, Nan; Hock Ong, Marcus Eng
• Is Part Of: BMJ open. Volume 8:Supplement 1(2018)
• Page Start: A12
• Page End: A12
• Publication Date: 2018-04-16
• Subjects: Medicine -- Research -- Periodicals; 610.72
• DOI: 10.1136/bmjopen-2018-EMS.32
• Languages: English
• ISSNs: 2044-6055

Abstract

Abstract : Aim: Inpatient mortality is an indicator of hospital performance and patient care. In this retrospective cohort study, we aimed to develop and validate an inpatient mortality model for use during ED consultation to efficiently risk stratify patients for better care and resource allocation. Method: Data was extracted from the Electronic Health Records (unified patient care data) of Singapore General Hospital in the year of 2014. Patients admitted through the ED were included and patients<21 were excluded. Variables, such as demographics, comorbidities, socioeconomic status and laboratory tests, were selected through literature review and clinician judgement before analysis with univariable and multivariable logistic regression. The model was assessed with receiver operating characteristic area under the curve. Results: Among the 35 699 patients admitted from ED in 2014, 690 died in-hospital. Univariate-analysis showed males, lower socioeconomic status, multiple comorbidities and increased acuity of illness as significant variables contributing to inpatient mortality. The final model included gender, use of medifund (financial aid), Charlson Comorbidity Index (CCI), albumin, creatinine, white blood cell counts, and number of ED visits within the past 1 year. This model (AUC 0.840) performed the best compared to other scores, including just using CCI, age, gender and principal diagnosis (AUC 0.723). Conclusion: A novel model for predicting inpatient mortality could effectively risk stratify patients early in the ED. This model may have future applications to improve management and disposition. Conflict of interest: None Funding: None