A preliminary mathematical model for head injury outcome prediction in a resource-limited settings

Background. Accurate prediction of outcomes following head injury is important for early clinical decision-making, long-term rehabilitation planning, counselling of relatives, and resource allocation. In resource-limited settings, prediction models based on readily available clinical, laboratory, and radiological parameters may provide practical support for early prognostication. Objective. To develop preliminary mathematical models for predicting head injury outcome at discharge and 6 months. Methods. This one-year prospective cohort study included 174 patients with complete data at the 68 Nigerian Army Reference Hospital, Yaba, Lagos. Clinical, laboratory, and radiological parameters were recorded. Outcomes were assessed using the Glasgow Outcome Scale–Extended (GOSE), an ordinal scale, at discharge and 6 months after injury. Linear regression was used as an approximation to derive preliminary prediction equations. Two predictor sets were analyzed: age, bilateral pupillary reactivity, and Glasgow Coma Scale (GCS) score; and age, total white blood cell count, neutrophil-to-lymphocyte ratio, monocyte-to-lymphocyte ratio, and cranial CT findings. Results. The clinical model showed strong apparent correlation with GOSE at discharge (R = 0.831) and at 6 months (R = 0.736), both p < 0.01. The laboratory-radiological model showed moderate apparent correlation with GOSE at discharge (R = 0.605) and at 6 months (R = 0.572), both p < 0.01. No statistically significant mean differences were observed between observed and predicted GOSE scores in the derivation cohort (p > 0.90). Conclusion. Preliminary linear regression models using routinely available parameters showed apparent association with short- and medium-term outcomes after head injury. The mathematical model based on age, pupillary reactivity, and GCS showed stronger apparent performance than the laboratory-radiological model. These findings should be interpreted cautiously because GOSE is ordinal, linear regression was used as an approximation, and no internal or external validation was performed.