Enhancing the performance of predictive models for Hospital mortality by adding nursing data. (May 2019)

• Record Type: Journal Article
• Authors: Sanson, Gianfranco; Welton, John; Vellone, Ercole; Cocchieri, Antonello; Maurici, Massimo; Zega, Maurizio; Alvaro, Rosaria; D'Agostino, Fabio
• Is Part Of: International journal of medical informatics. Volume 125(2019)
• Page Start: 79
• Page End: 85
• Publication Date: 2019-05
• Subjects: Diagnosis-related groups -- Hospital mortality -- Nursing dependency index -- Nursing diagnosis -- Predictive models -- Risk adjustment; Medical informatics -- Periodicals; Information science -- Periodicals; Computers -- Periodicals; Medical technology -- Periodicals; Medical Informatics -- Periodicals; Technology, Medical -- Periodicals; Computers; Information science; Medical informatics; Medical technology; Electronic journals; Periodicals; Electronic journals; 610.285
• DOI: 10.1016/j.ijmedinf.2019.02.009
• Languages: English
• ISSNs: 1386-5056

Abstract

Abstract: Background: Mortality is the most considered outcome for assessing the quality of hospital care. However, hospital mortality depends on diverse patient characteristics; thus, complete risk stratification is crucial to correctly estimate a patient's prognosis. Electronic health records include standard medical data; however, standard nursing data, such as nursing diagnoses (which were considered essential for a complete picture of the patient condition) are seldom included. Objective: To explore the independent predictive power of nursing diagnoses on patient hospital mortality and to investigate whether the inclusion of this variable in addition to medical diagnostic data can enhance the performance of risk adjustment tools. Methods: Prospective observational study in one Italian university hospital. Data were collected for six months from a clinical nursing information system and the hospital discharge register. The number of nursing diagnoses identified by nurses within 24 h after admission was used to express the nursing dependency index (NDI). Eight logistic regression models were tested to predict patient mortality, by adding to a first basic model considering patient's age, sex, and modality of hospital admission, the level of comorbidity (CCI), and the nursing and medical condition as expressed by the NDI and the All Patient Refined-Diagnosis Related Group weight (APR-DRGw), respectively. Results: Overall, 2301 patients were included. The addition of the NDI to the model increased the explained variance by 20%. The explained variance increased by 56% when the APR-DRGw, CCI, and NDI were included. Thus, the latter model was nearly highly accurate ( c = 0.89, 95% confidence interval: 0.87–0.92). Conclusion: Nursing diagnoses have an independent power in predicting hospital mortality. The explained variance in the predictive models improved when nursing data were included in addition to medical data. These findings strengthen the need to include standardized nursing data in electronic health records.