METHODS: A deterministic match was performed between all NYC PWHA as of 12/31/2013 and patient-level SPARCS data for inpatient hospitalizations during 2013, using 16 match keys comprised of combinations of partial first and last names, date of birth, partial social security number, sex, and residential ZIP code. Match scores were generated based on the number of matching key elements, with a lower score indicating a lower likelihood of being a true match; low scoring matches were manually reviewed. SPARCS provides an HIV-positive indicator for each hospitalization event based on the International Classification of Diseases, Ninth Revision, Clinical Modification codes recorded during the hospitalization. Any records flagged with this HIV-positive indicator that did not initially match were also manually reviewed and probable matches were assigned accordingly. Once matches were determined, patient-level data were linked back to all hospitalization events during the year.
RESULTS: The final matched dataset included 16,648 unique PWHA, who had a total of 27,049 hospitalization events during 2013 in NYC hospitals. The SPARCS HIV indicator was useful for evaluating and identifying matches; of 1,948 records that were flagged for HIV and did not match programmatically, 690 (35%) were determined to be matches after review. Overall, HIV was indicated as a diagnosis in SPARCS for 12,987 (78%) of 16,648 matched patients. Most (69%) of matched PWHA were hospitalized only once during the year, and the maximum number of hospitalizations for a single patient was 28.
CONCLUSIONS: Matching the HIV surveillance registry with SPARCS hospitalization records provides a more complete picture of clinical care patterns among NYC PWHA. The matched data can be analyzed at the individual patient level as well as the hospitalization event level. Methods used to link these two data registries can be applied to additional years of hospitalization records and to future matches with other data sources.