Going beyond ICD codes: identifying carceral status in California statewide billing and electronic health record data, 2010–2019

Carceral health researchers often struggle to collect health data in prisons and jails due to data sharing restrictions and a lack of willingness to collaborate from departments of corrections. Electronic health record data offer a promising alternative data source if carceral status in these records can be ascertained. Therefore, the goals of this paper are to (1) describe an algorithm to improve identification of incarceration status at time of hospitalization in patient discharge data (PDD) and (2) assess the concordance of this method with data from the department of corrections (DOC) by comparing PDD to hospitalizations reported by the California State DOC from 2011–2019. Study Setting and Design: Nonpublic PDD for all non-military hospitalizations in California-licensed hospitals were obtained from California Healthcare Access and Information. Three overlapping subsets of populations who may be incarcerated were constructed using the ICD-9 and ICD-10 codes for incarceration, Zip-code of residence and source of admission in PDD data. Additionally, data on hospitalizations among people incarcerated in state prisons were obtained through a public records request. Descriptive statistics on the demographic distributions and diagnosis categories were compared across populations from the incarcerated subsets of EHR and to hospitalizations reported by California Department of Corrections to estimate the concordance of the PDD populations with the DOC hospitalized population. In our sample of almost 29 million hospitalizations among people in California, three cohorts were constructed based on varying levels of certainty of the carceral status. Expanding upon ICD codes alone resulted in a population size that better approximates the true number of hospitalizations among incarcerated people as reported by data on hospitalizations in California state prisons. Cohort sizes ranged from 43,548 hospitalizations using ICD codes alone to 201,718 using the most inclusive definition, and this expanded cohort fell within the estimated range of 183,708–275,562 hospitalizations among incarcerated people statewide, compared to the ICD-code-only cohort, which substantially underestimated this total. Discrepancies in hospitalization cause were observed between the PDD cohort and hospitalizations reported by California DOC indicating that further investigation into how electronic health records are coded for people who are hospitalized during incarceration is needed. With this novel methodology to identify carceral status in electronic health records or PDD, a wide range of critical epidemiologic studies may be conducted on this subpopulation that is difficult to study.