A clinical score to predict survival from hyperglycemic crisis following general medical wards admission in a resource constrained setting

This study aims to develop a risk score model, based on available clinical data to assess absolute risk of dying among admissions for hyperglycemic crisis in Eastern Cape, one of South Africa’s most disadvantaged provinces. Data from 268 admissions for hyperglycemic crisis at Nelson Mandela Academic Hospital, Mthatha, for the 2 year periods of 2008 and 2009 were used to develop multivariate logistic regression and cox proportional hazards models for the time to onset of death and the area under curve (AUC) of the receiver-operating-characteristic curve (ROC). The risk score models included the following independent variables that were associated with mortality: systolic blood pressure (SBP) < 90 mm/Hg, Odd’s ratio (OR) = 13.3 (95%CI 2.1–83; p = 0.006) and Hazard ratio (HR) = 8.4 (95%CI 2–36; p < 0.0001), Leukocyte count >10 × 10⁶/L OR = 16.7 (95%CI 2.1–143; p = 0.0008) and HR = 5.7 (95%CI 1.8–7; p = 0.021) and Platelet count <150 × 10⁶/L OR = 11.6 (95%CI 13–43.5; p < 0.0001) and HR = 5.1 (95%CI 2.2–6.8; p < 0.001). The final models yielded good and significant area under the curve (AUC) for WBC >10 × 10⁶/L (AUC = 0.642, 95% CI 0.557–0.727; p < 0.001) and platelet count <150 × 10⁶/L (AUC = 0.651, 95% CI 0.567–0.741; p < 0.001) but lower and insignificantly discriminatory power of the models for SBP <90 mm/Hg (AUC = 0.573, 95%CI 0.456–0.691; p = 0.163). The proposed scoring system is 5 points where all three independent predictors are absent with additional 3 points for each independent predictor that is present. Mortality rates were 4.8% (N = 3/62), 17.1% (N = 19/111), 52.2% (N = 12/23) and 66.7% (N = 2/3) respectively for admissions with none (total score = 5points), any one (total score = 8points), any two (total score = 11points) and all three (total score = 14points) independent predictors. The present risk scores developed in the Eastern Cape Province of South Africa using easily obtained clinical parameters can help identify high risk hyperglycemic crisis, total score ≥8 who would benefit most from management in the high risk care unit. These tools need to be validated in other limited resource settings.