Open AccessValidation of an Automated Mortality Index using the Electronic Medical Record System in a Network of Acute Care Hospitals
Author(s) -
Deborah Morris,
Brynn E. Sheehan,
Rajan Lamichahane,
Kathie S. Zimbro,
Merri K. Morgan,
Parag Bharadwaj
Publication year - 2019
Publication title -
journal of hospital administration
Language(s) - English
Resource type - Journals
eISSN - 1927-7008
pISSN - 1927-6990
DOI - 10.5430/jha.v8n3p8
Subject(s) - medicine , receiver operating characteristic , logistic regression , medical record , emergency medicine , risk of mortality , electronic medical record , risk assessment , acute care , clinical practice , intensive care medicine , medical emergency , health care , physical therapy , computer security , computer science , economics , economic growth
Objective: Physicians struggle with prognostication for patients facing the final year of life. Practical tools which identify patients at the time of hospital admission who are at high risk of mortality would be helpful to provide timely access to supportive services, including palliative care and hospice. The PREDICT is a validated tool that predicts mortality risk but has not been implemented into electronic medical record (EMR) systems. The current study evaluated the validity of PREDICT within an EMR system and tracked patient mortality over 12 months.Methods: The study sample consisted of 3,488 adult patients admitted to a network of acute care hospitals. The PREDICT tool was evaluated for its ability to predict mortality within 6 and 12 months of hospitalization and was compared to the APR-DRG Mortality Risk Index (MRI).Results: A total of 299 patients (9%) were deceased within 12 months of hospital admission. Logistic regressions revealed that higher PREDICT scores were associated with greater risk of mortality within 6 and 12 months post-discharge. Receiver Operating Characteristic curve (ROC) analysis revealed that the overall PREDICT score significantly predicted mortality at 12 months (ROC = .767) and was a better predictor than the MRI.Conclusions: The PREDICT tool is a valid assessment of mortality risk and unlike the MRI, it can be readily automated in the EMR to help identify patients at greater risk of death. More research is needed to apply this tool in clinical practice and calibrate its performance across clinical settings.