Survival Analysis: Alternate Approach

Abstract

Existing models of survival analysis dealing with group data have advantages and limitations too. Assumptions of the models need to be verified. Problem arises when one or more assumptions of a model are not satisfied. Methods of survival analysis, inter alia assumes homogeneity of treatment and related factors during the follow-up periods. However, in practice, such assumptions do not hold. The proposed method (Geometric mean approach) is non-parametric, simple and satisfies desired properties from measurement theory angle. Focusing on individual patient, it helps in mathematical diagnosis of disease like cancer of a particular type, disease intensity in terms of the chosen measurable factors/variables, identification of bad prognosis factors of an individual and quantification of progress or deterioration of a patient over time (analogous to hazard function of an individual). The method can help the researchers and practitioners to make meaningful analysis and drawing meaningful conclusions including estimation of hazard function of sample patients without making any assumption. Empirical verifications of the proposed method along with its robustness and estimation of hazard function and clinical validations are proposed as future studies.