Continuous-Time Markov Model for Transitions Between Employment and Non-Employment: The Impact of a Cancer Diagnosis

This article investigates whether a cancer diagnosis can cause a permanent loss in employability. In this regard, we evaluate the impact of cancer on labor market conditions by constructing transition matrices to compare the transitions between occupational states. We obtain a set of statistics based on our estimations by using continuous-time Markov transition processes to study and compare the labor market dynamics in two populations: 1) individuals diagnosed with cancer and 2) individuals free of cancer in the general population. The consequences of cancer diagnosis were measured by the significant deviation in the transition matrix for cancer survivors in comparison to the prior matrix standardized according to the general population. We accounted for the probability that some individuals in the control group (i.e., the general population) could be diagnosed with cancer which is a key-issue in case-control studies. The absence of detailed information about the health statuses of the individuals in the control group required the implementation of the EM algorithm for maximizing the adapted likelihood function. We jointly estimated the probability of being diagnosed with cancer in the control group and the parameters of our model. Given that individuals are exposed differently to cancer depending on their activities, we stratified the dataset by socioeconomic status (SES) for two reasons: 1) to clearly distinguish between the cancer-specific effects and 2) to account for the other stigmatizing factors in the labor market that are inherent to the examined subpopulations (i.e., low- and high-SES groups). We also considered the systematic differences in the subjects' socioeconomic statuses and their abilities to return to work. We determined whether these differences are related to illness (e.g., cancer sites or prognosis) or occupation (e.g., physical demands).