A Competing-Risk Approach for Modeling Length of Stay in Severe Malaria Patients in South-East Asia and the Implications for Planning of Hospital Services

Abstract

© The Author(s) 2018. Published by Oxford University Press for the Infectious Diseases Society of America. Background Management of severe malaria with limited resources requires comprehensive planning. Expected length of stay (LOS) and the factors influencing it are useful in the planning and optimisation of service delivery. Methods A secondary, competing-risk approach to survival analysis was performed for 1217 adult severe malaria patients from the South-East Asia Quinine Artesunate Malaria Trial. Results Twenty percent of patients died; 95.4% within 7 days compared to 70.3% of those who were discharged. Median time to discharge was 6 days. Compared to quinine, artesunate increased discharge incidence (subdistribution-Hazard ratio, 1.24; [95% confidence interval 1.09-1.40]; P =.001) and decreased incidence of death (0.60; [0.46-0.80]; P <.001). Low Glasgow coma scale (discharge, 1.08 [1.06-1.11], P <.001; death, 0.85 [0.82-0.89], P <.001), high blood urea-nitrogen (discharge, 0.99 [0.99-0.995], P <.001; death, 1.00 [1.00-1.01], P =.012), acidotic base-excess (discharge, 1.05 [1.03-1.06], P <.001; death, 0.90 [0.88-0.93], P <.001), and development of shock (discharge, 0.25 [0.13-0.47], P <.001; death, 2.14 [1.46-3.12], P <.001), or coma (discharge, 0.46 [0.32-0.65], P <.001; death, 2.30 [1.58-3.36], P <.001) decreased cumulative incidence of discharge and increased incidence of death. Conventional Kaplan-Meier survival analysis overestimated cumulative incidence compared to competing-risk model. Conclusions Clinical factors on admission and during hospitalisation influence LOS in severe malaria, presenting targets to improve health and service efficiency. Artesunate has the potential to increase LOS, which should be accounted for when planning services. In-hospital death is a competing risk for discharge; an important consideration in LOS models to reduce overestimation of risk and misrepresentation of associations.