Population ageing in China has heightened concerns about health equality in later life. This study investigates how heterogeneous employment-retirement trajectories during the peri-retirement period shape later-life health disparities in China. Employment history and retirement arrangements are critical social determinants of health, while existing studies treat retirement as a discrete event rather than a dynamic process. From a life course perspective, we argue that health disparities in old age reflect the cumulative effects of distinct pathways out of the labour force. Using six waves (2011-2020) of the China Health and Retirement Longitudinal Study (CHARLS), we focus on the peri-retirement period (ages 45-63), which spans the statutory retirement ages in China. We apply sequence analysis and cluster analysis to identify typical employment-retirement trajectories, and estimate trajectory-specific life expectancy (LE) and healthy life expectancy (HLE) using Markov multistate models. To examine mechanisms, we use Shapley decomposition with counterfactual one-sided replacement to assess the contributions of health deterioration, recovery, and mortality to observed HLE disparities. We identify distinct trajectories by gender. Among men, four trajectories emerge: sustained agricultural work, formal non-agricultural employment with full retirement, long-term non-employment with full retirement, and sustained informal non-agricultural work. Among women, three trajectories emerge, with no equivalent sustained informal non-agricultural trajectory, reflecting gendered constraints in labour market participation. Substantial health disparities exist across these trajectory groups. Women have higher LE than men but slightly lower HLE, consistent with the gender health-survival paradox. The formal non-agricultural employment-full retirement group shows the most favourable LE and HLE, whereas the long-term non-employed group fares worst. As age rises, LE disparities across trajectories remain stable, while HLE disparities converge. Shapley decomposition shows that health deterioration and recovery capacity are the primary drivers of HLE disparities.

This study advances the life course research by reframing labor force exit as a dynamic process. Methodologically, it uses sequence analysis, multistate modelling, and counterfactual decomposition into a unified framework for trajectory identification, association estimation, and mechanism testing. The findings show that employment-retirement trajectories embody cumulative differences in income, social insurance coverage, and health resilience. In China's context, where social insurance schemes are stratified by employment type and sector, these trajectories reflect systematic differences in access to institutional protection. The results carry policy implications for China's ongoing gradual delay of statutory retirement ages. Targeted labor market support should be extended to peri-retirement workers, especially those in informal employment and long-term non-employment. Expanding social insurance coverage for disadvantaged groups and reducing institutional stratification in benefits could enhance health recovery capacity. Addressing gendered disadvantages requires recognising women's unpaid care work and ensuring more fair pension entitlements.