METHODS: We accessed data from the Michigan Care Improvement Registry (MCIR) and the Michigan Community Health Automated Medicaid Processing System (CHAMPS). The MCIR is expected to include all children who were vaccinated, as required by Michigan law. A large proportion of children in Flint are enrolled in CHAMPS. The datasets included records for children born on or after January 1, 2010 to present, who resided in ZIP codes served by the Flint Water System (FWS). We first de-duplicated and merged records using unique identifiers and probabilistic matching to identify children at risk. Then, we selected the most recent address for enrollees and geo-coded them to obtain highly resolved spatial locations (latitude and longitude coordinates) of children <6 years of age. We estimated the spatial extent of the FWS using geocoded addresses listed in the active water utility database plus locations in close proximity to water service lines. Using GIS, we enumerated children who resided in the FWS area. A lower population range was calculated by subtracting children in the CHAMPS database and FWS area but not in MCIR. An upper population range was calculated by adding children in CHAMPS without a current address in the FWS area. For quality control, our estimates were compared to U.S. Census estimates.
RESULTS: We identified 9,622 (range: 7,863 – 11,153) children <6 years of age who resided in the FWS area and are considered to be at risk for exposure to water lead contamination.
CONCLUSIONS: Enumerating populations that fall within a non-standard boundary like a water distribution system can require careful handling and merging of disparate databases and involve tasks such as managing inconsistent variable formats and misspellings. Other challenges include defining water service boundaries, population mobility, and temporal limitations. This approach required a collaborative mix of geospatial, programming, and epidemiologic expertise between MDHHS and CDC.