Identifying Factors Associated with Loss to Follow-up Among Patients Reported to the New York State Cancer Registry.

Abstract

Background: State cancer registries in the United States are data sources for estimating population-based cancer survival. However, the completeness of patient follow-up can affect the accuracy of survival estimates. Like many registries, the New York State Cancer Registry (NYSCR) conducts patient follow-up largely through linkages with other data sources. Even after expending great effort on linkages, a small proportion of patients remain lost to follow-up (LTFU). In this study, we identified factors that are associated with the likelihood of LTFU in the NYSCR.

Methods: First primary cancers (sequence number, 00 or 01 and excluding death-certificate- and autopsy-only cases) diagnosed during 2000-2018 among New York State residents were selected for study. All patients were followed through December 31, 2018. Based on each patient's vital status and last contact date, follow-up status was categorized into 2 groups: patients LTFU and patients not LTFU. Patients LTFU were examined by demographic and tumor characteristics. Multivariate logistic regression analyses were performed to evaluate the association between demographic/tumor characteristics and likelihood of LTFU. For patients LTFU, the timing of LTFU (within 1 year, 1 to <5 years, 5 to <10 years, or >10 years) was further described. LTFU rates within 5 years after cancer diagnosis were also examined.

Results: Among 1,797,228 patients, 74,722 were LTFU prior to December 31, 2018, representing 4.2% of all patients and 7.6% of alive patients. About 60% of LTFU occurred within 1 year after cancer diagnosis. Compared to the reference group, logistic regression analyses indicated that patients LTFU were more likely to be female, Black, Asian/Pacific Islander (API), Hispanic, foreign born, insured by Medicaid, uninsured, aged <20 years, and living in New York City or metropolitan counties. Cases reported by laboratories only and physician offices also had a higher likelihood of LTFU. Similar patterns and effects were identified when evaluating 5-year LTFU.

Conclusion: Identifying factors associated with patient LTFU is important for cancer registries to improve follow-up data. We found that LTFU is not random; rather, certain patient groups have higher LTFU rates than others. For registries that conduct follow-up through linkages, it is critical to collect high-quality and complete demographic data, especially for females, children, the foreign born, and minority race/ethnicity groups.