Cancer survivors and job tenure: transitions between employment and non- employment in a continuous-time Markov model

Presenter: Alain Paraponaris, INSERM, U912 (SE4S)

Abstract

Context: Advances in cancer treatments created opportunities for cancer survivors (CS) to take part in the labour force again. However CS still cope with strong difficulties to be employed after cancer diagnosis and treatment.

Objectives: The paper reconsiders the impact of cancer upon transitions from employment to non-employment in a case-control framework. It specifically targets the propensity to remain employed through disentangling obvious epidemiological arguments from workplace fundamentals linked to usual occupational transitions.

Methods: Individuals 58-year-old and less followed in a two-year period by two French national surveys (2004 CS Survey, Health Ministry and Employment Survey, National Institute for Statistics and Economic Studies) were matched with the help of the propensity score technique. Estimations based on continuous Markov transition processes were computed. There were three possible states: employed, not-employed (merging unemployed workers and inactive) and retired (absorbing state). Transitions in the labour market were compared for CS and their non-cancer (NC) counterparts. The estimations focused on the discrepancies in job tenure due to the socioeconomic status (SES), ‘low’ (farmers, artisans, factory workers, and drivers) or ‘high’ (top executives, intermediate professionals, and white-collars) and on the relative prognosis made at the diagnosis. Relative prognosis is a synthetic continuous index ranging from 0 (very bad) to 1 (excellent) depending on age, gender, localisation of cancer and five-year survival rate. Prognosis was considered as good (respectively bad) for values higher than, or equal to (less than) the median (0.674).

Results: Two years after diagnosis, CS were at smaller risk (77%) to remain employed than NC (90%) and at higher risk to move from employment to non-employment (20% versus 7%). They came more rarely from non-employment to employment (10% versus 27%) and stayed more often not-employed (73% versus 10%). Among low SES, the job tenure was 90% for NC versus only 68% for CS. Among high SES, percentages were respectively 89 and 81%. There were no obvious difference in the probability to remain not-employed for CS (75%) and NC (73%) with low SES, and a sharp discrepancy for CS with high SES (73% versus 50%). Among CS, good prognosis compared to bad prognosis lessened the detrimental impact of cancer upon job tenure for low SES (77% versus 62%) much more than for high SES (83% versus 78%). It reduced the transition probability from employment to non-employment more strongly for low (15% versus 34%) than for high SES (14% versus 20%). Good prognosis compared to bad prognosis dramatically boosted the probability to move from non-employment to employment for low SES (20% versus 3%) and lessened the probability to remain not-employed (73% versus 83%). For high SES, good prognosis stimulated in no way the return to employment (4% versus 17%) and reduced only slightly the probability to remain not-employed (71% versus 73%).

Conclusion: These results obviously question the policy-maker about which part of public policies to use in order to offset the detrimental and differentiated impacts of cancer upon job tenure. It then appears that both health protection and employment policies have to be called up.