Understanding Homelessness ICD-10 Coding in Medicaid Claims: A Mixed-Methods Approach

Background & Purpose

People experiencing homelessness are often high utilizers of emergency departments and experience inpatient hospitalization more frequently. The ICD-10 introduced social determinant of health Z codes, which include an indication of homelessness (Z59.0); however, the association of this code with documented history of homelessness service use is unknown. Using an explanatory sequential mixed methods design, the overall aim of this project was to explain patterns in the use of the ICD-10 Z code for homelessness in Medicaid claims and elucidate providers’ explanations for the patterns and prevalence of Z codes in these claims.

Methods

The explanatory sequential mixed methods approach first analyzed the prevalence of Z codes in a linked administrative dataset consisting of HMIS data from 2014-2016 and Medicaid inpatient and emergency department clams from 2016. Quantitative analysis then informed qualitative pragmatic semi-structured interviews with inpatient and emergency department physicians (n = 18) focused on explanations for the patterns and prevalence these codes (QUANT --> qual). Interviews were analyzed via cross-case deductive and inductive content and cross-case thematic analysis. Results were integrated using a display table.

Results

Sensitivity, PPV, and kappa of Z-coded status’ reflection of homeless service use were poor, although homeless service use histories were significantly associated with higher odds of being Z-coded. Male sex, age over 60 years, and Medicaid expansion eligibility were all associated with higher odds of being Z-coded. Substance use disorders (SUD) and severe mental illnesses (SMI) had the strongest associations with the presence of a Z59.0 code. Wide variation in use of Z-codes across hospitals was evident.

Interviews revealed an idiosyncratic process by which providers screen for social determinants of health with no explicit role for Z coding in this process. Participants thought Z-coding prevalences were largely consistent with the homeless population, except for race/ethnicity and hospital-level Z coding practices. They suggested nuances within the homelessness designation not reflected in concordance analysis.

Conclusions & Implications

The lack of systematic use of Z codes for homelessness in this study is problematic because it limits the codes’ usefulness in planning and implementing strategies to address the health needs of people experiencing homelessness. Codes being consistent with provider expectations suggests further outreach could encourage Z code use, but incongruencies indicate potential issues with Z code uptake.

As internationally standardized tools, these codes represent an opportunity to consistently document multiple social determinants of health (including homelessness) across settings. Doing so would allow for assessment of local needs, monitoring of interventions to address them, and a broader view of population-level social determinants of health. Changes to hospital or state policy could improve uptake and consistency of use and changes to national billing protocol could provide an opportunity to increase reimbursement when considering these needs.