Classification of Homelessness in Medicaid Claims and Homeless Services Administrative Records: Application of a...

People experiencing homelessness have high healthcare utilization, with multiple indicators used to identify them in medical data. The association between the claims-based ICD-10 Z code for homelessness (Z59.0) and documented homelessness service use history is unknown. This study examined (1) Z59.0 code prevalence, (2) patterns of Z59.0 coding compared with that of known homeless service use, and (3) factors associated with Z59.0 coding.

The authors used Medicaid claims linked with homelessness service use recorded in the Homeless Management Information System for 19 New Jersey counties. The outcome of interest was the prevalence of Z59.0 coding on Medicaid inpatient or emergency department claims from 2016. A concordance framework was applied to assess the alignment between Z-coded status and documented homelessness service use. Binary logistic regression models identified predictors of Z-code-based homelessness status.

Lower-bound estimates of sensitivity, positive predictive value, and Kappa statistic suggested weak agreement between homelessness classification within Medicaid claims and administrative records of homeless services. Homeless service use histories were significantly associated with higher odds of being Z59.0 coded. Male sex, being aged >60 years, and Medicaid expansion eligibility were all associated with higher odds of being Z59.0 coded. Substance use disorders and serious mental illnesses had the strongest associations with the presence of a Z59.0 code. Wide variation in the use of Z codes across hospitals was evident.

Estimates suggest limited agreement in the classification of homelessness between Z59.0-coded claims status and known homelessness service utilization. The lack of systematic use of Z codes for homelessness observed by this study limits their usefulness to document health outcomes for people experiencing homelessness.