OVERLAPPING, NON-MUTUALLY EXCLUSIVE EVENTS AND INVERSE SURVIVAL ANALYSIS WITH APPLICATION IN MODELING OF STUDENTS’ CHRONIC DISEASES A CASE STUDY OF MOI AND U.O.E UNIVERSITIES

Abstract

The analysis of time to event(s) has taken roots in many fields including health studies, statistics, among others. Single and mutually exclusive events have been studied, and, methods, for handling such cases have been developed. However, events could be overlapping and nonmutually exclusive (ONME) in such a way that subjects experience multiple events simultaneously. The events could be taking long durations before their termination and the time could be long enough for consideration in analysis. In such cases, time to event analysis is important as well as accompanying it with time to non-event (inverse survival) analysis. No methods have been developed to handle cases of ONME events as well as combining survival and inverse survival analysis for extensiveness of analysis. The main objective of this research was to study events that are ONME and apply the concepts in modeling students’ chronic diseases. The specific objectives are to create illustrations for the various forms of ONME events, developing ways of handling data from such events, expounding on concepts of inverse survival analysis and to apply the ideas in conjunction with concepts of discrete-time, rightcensored survival and unbalanced three-stage hierarchical designs in modeling and analyzing chronic diseases among selected university students. The percentage of university students currently nursing chronic diseases is alarming, and for a long time, these students are ever thought to be healthy. However, research shows otherwise with up to 30% of the students nursing the diseases. For the first three objectives, AutoCAD software was used in developing the illustrations where hypothetical data was used. Mixed-study design was applied in sampling 739 students randomly and using a questionnaire for data collection in Moi University and University of Eldoret. Descriptive analysis and inferential analysis including t-tests, log-rank tests, MANOVA and regression were performed. Data management was carried out in R, Minitab, STATA and SPSS. All the tests were carried out at 5% level of significance. In objective 1, figures were used where the number of events, starting and ending points of each event and the study periods were found to be the factors affecting the number of illustrations. In objectives 2 and 3, five methods were developed in each case for handling ONME events. Each of the methods developed is appropriate based on the situation at hand, and the ONME events showed similar analysis approaches with the standard survival models. In objective 4, discrete-time models and ANOVA components were applied. Currently, there are approximately 14.6% of the students who are suffering from chronic diseases. At every level of study, the number of students who get sick is significant. The factors ‘family history’ ( 2  =29.03, p-value<0.0001), ‘involvement in drugs’ ( 2  =27.03, p-value<0.0001), ‘adopted lifestyles’ ( 2  =23.04, p-value<0.0001) and ‘extreme poverty’ ( 2  =5.14, p-value=0.0233) were found to be significantly associated with the chronic diseases. On the effects, the diseases were found to be negatively affecting the aspects of life of the infected students. Across gender and years of study, survival of the students was found to be the same, age was affected inverse survival negatively. It is concluded that, methods for handling ONME events and inverse survival as an emerging field in the near future adopt similar approaches to standard survival methods in analysis. Also, there is need for intervention among the university students as 14.6% is not a number to be ignored. It is recommended that the researchers and data managers to adopt the new ideas in this work in expanding the field of survival analysis, as well as stakeholders to come together and arrest the situation in universities before things slip out of hands