Browsing by Author "Nadler, Adam"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    Admission Risk Score to Predict Inpatient Pediatric Mortality at Four Public Hospitals in Uganda
    (PLoS ONE, 2015) Mpimbaza, Arthur; Sears, David; Sserwanga, Asadu; Kigozi, Ruth; Rubahika, Denis; Nadler, Adam; Yeka, Adoke; Dorsey, Grant
    Mortality rates among hospitalized children in many government hospitals in sub-Saharan Africa are high. Pediatric emergency services in these hospitals are often sub-optimal. Timely recognition of critically ill children on arrival is key to improving service delivery. We present a simple risk score to predict inpatient mortality among hospitalized children. Between April 2010 and June 2011, the Uganda Malaria Surveillance Project (UMSP), in collaboration with the National Malaria Control Program (NMCP), set up an enhanced sentinel site malaria surveillance program for children hospitalized at four public hospitals in different districts: Tororo, Apac, Jinja and Mubende. Clinical data collected through March 2013, representing 50249 admissions were used to develop a mortality risk score (derivation data set). One year of data collected subsequently from the same hospitals, representing 20406 admissions, were used to prospectively validate the performance of the risk score (validation data set). Using a backward selection approach, 13 out of 25 clinical parameters recognizable on initial presentation, were selected for inclusion in a final logistic regression prediction model. The presence of individual parameters was awarded a score of either 1 or 2 based on regression coefficients. For each individual patient, a composite risk score was generated. The risk score was further categorized into three categories; low, medium, and high. Patient characteristics were comparable in both data sets. Measures of performance for the risk score included the receiver operating characteristics curves and the area under the curve (AUC), both demonstrating good and comparable ability to predict deathusing both the derivation (AUC =0.76) and validation dataset (AUC =0.74). Using the derivation and validation datasets, the mortality rates in each risk category were as follows: low risk (0.8% vs. 0.7%), moderate risk (3.5% vs. 3.2%), and high risk (16.5% vs. 12.6%), respectively. Our analysis resulted in development of a risk score that ably predicted mortality risk among hospitalized children. While validation studies are needed, this approach could be used to improve existing triage systems.