Survival Analysis Modeling on Level of Exposure to the Environment and Health Hazards

Abstract

In this study the theoretical basis, model fitting, variable selection, and hypothesis testing of survival analysis models were explored and applied to real data to demonstrate how the model could be used to ascertain linkages between exposure and health hazards. Malaria cases from admission to destination state (death or censored), due to causal factors (level of exposure to malaria parasites) were collected over a one-year period from four hospitals within a district in Ghana and analyzed using survival analysis modeling techniques. The hazard ratio shows that exposed group was at a higher risk of malaria-death than less exposed group. The Kaplan-Meier curves reveals that dwellers who subscribed to insecticide treated nets (ITNs) as means of reducing exposure to the malaria parasite had better and longer survival prognosis than non-subscribers. The p-value (0.0035 is less than 0.0500) of the likelihood ratio test revealed that using ITN as means of reducing exposure to the parasite is significant. The graphical test conducted on the biomedical data for the two levels of exposure was parallel; this parallelism of the log-log graph is a litmus test for assessing the proportional hazards (PH) assumption. The results from the log-rank test (p-value = 0.0020 less than 0.0500) implies that if we had included all population elements into the study, and examined association between level of exposure to malaria parasite and health hazards for the two groups, the survival curves for less-exposed and exposed would have been different with less-exposed group having longer time of surviving health hazards than exposed group.