Retirement, a major transition in the life course, may affect many aspects of retirees' well-being, including health and health care utilization. Leveraging differential statutory retirement age (SRA) by occupation for China's urban female workers, we provide some of the first evidence on the causal effect of retirement on hospitalizations attributable to mental illness and its heterogeneity. To address endogeneity in retirement decisions, we take advantage of exogeneity of the differing SRA cut-offs for blue-collar (age 50) and white-collar (age 55) female urban employees. We apply a Fuzzy Regression Discontinuity Design (RDD) around the SRA cut-offs using nationally representative hospital inpatient claims data that cover these workers. We show that blue-collar females incur more hospitalizations for mental illness after retirement, while no similar change is found for white-collar females. Conditional on blue-collar females being hospitalized, probabilities of overall and ER admissions due to mental illness increase by 2.3 and 1.2 percentage points upon retirement, respectively. The effects are primarily driven by patients within the categories of schizophrenia, schizotypal and delusional disorders; and neurotic, stress-related and somatoform disorders. Moreover, the 'Donut' RDD estimates suggest that pent-up demand at retirement unlikely dominates our findings for blue-collar females. Rather, our results lend support to their worsening mental health at retirement. These findings suggest that occupational differences in mental illness and related health care utilization at retirement should be considered when optimizing retirement policy schemes.