Hospital use of common Z-codes for Medicare fee-for-service beneficiaries, 2017–2021

Abstract Recognizing the impact of the social determinants of health (SDOH) on health outcomes, in 2016, the Centers for Medicare and Medicaid Services recommended the use of International Classification of Diseases, 10th Revision (ICD-10), Z-codes to capture patients' health-related social needs. We examined changes in Z-code utilization to document health-related social needs for Medicare fee-for-service recipients among US hospitals between 2017 and 2021 across 5 common SDOH domains. We found that, while 56.9% of hospitals had at least 1 Z-code recorded in at least 1 patient per year, apart from those referring to housing needs, rates of Z-code adoption were low. Additionally, hospitals that were general medical, part of a teaching institution, affiliated with larger health systems, and of medium to large size had greater odds of utilizing Z-codes. Findings from this study highlight the need for continued efforts in promoting the consistent use of standardized SDOH capturing methods like Z-code documentation, such as provider training.


Introduction
The social determinants of health (SDOH), defined as the "conditions where individuals are born, grow, work, live, and age," play a pivotal role in shaping health outcomes. 1ecognizing this link, key regulatory and standard-setting agencies have introduced quality measures to foster health equity by gathering patient health-related social needs (HRSN) data. 2,3Notably, starting in 2024, the Centers for Medicare and Medicaid Services (CMS) will mandate HRSN screening in 5 common SDOH domains (housing, food, transportation, utilities, and interpersonal safety) during adult hospitalizations. 4In this context, International Classification of Diseases, 10th Revision (ICD-10), Z-codes that capture nonmedical determinants of health provide a standardized method for documenting and sharing SDOH across health care entities.However, their adoption has faced challenges, slowing widespread use.
We sought to examine change in the utilization of Z-codes in Medicare fee-for-service (FFS) hospitalizations in US hospitals from 2017 to 2021 by SDOH domain (Figure 1).6][7] However, less is known about facility and community characteristics associated with Z-code adoption.][10][11][12][13] We extend this work by examining facility and community characteristics associated with Z-code utilization across US hospitals across the 5 HRSN screening domains slated to be tracked by CMS, with a goal of providing insights into institutional practices, policies, and resources, which are pivotal for the systematic integration of SDOH data in hospitals that will be required in upcoming regulatory requirements surrounding HRSN screening and data collection. 4

Data and methods
Our analysis of the adoption of Z-codes in the United States combines data from the American Hospital Association (AHA) Annual Survey with total Medicare FFS claims from 42 million patients aggregated at the hospital level using the CareJourney analytic platform. 14The AHA survey annually captures data from over 6000 US hospitals and health care systems.Our analytic sample comprises 27 385 observations from 5685 hospitals over 5 years, encompassing all nonfederal, nonpediatric, acutecare hospitals processing at least 1 Medicare FFS claim annually.
We formed binary variables corresponding to each of the 5 HRSN categories (housing, food, transportation, utilities, and interpersonal safety) based on specific Z-code descriptions from the CMS (Table 1) and determined whether each hospital used at least 1 Z-code in any category, or in each specific category, for at least 1 patient annually (Table 2).Hospital features, such as teaching status, system membership, critical access designation (CAH), ownership, bed size, and region, were sourced from the AHA survey.County attributes were derived from the American Community Survey (percent uninsured, percent Black, percent Hispanic/Latinx, percent unemployed, percent high school education), the US Department of Agriculture (metropolitan status), and the University of Wisconsin's Area Deprivation Index.To connect claims measures to the AHA survey, we matched CMS Certification Number and Hospital Name identifiers to the AHA Hospital ID.County data were linked using Federal Information Processing Standards (FIPS) codes.
We calculated mean hospital and county characteristics across all observations (Table 3).We used generalized estimating equations for longitudinal data to generate unadjusted and adjusted odds ratios (ORs) (with 95% CIs) estimating relationships between hospital-level adoption of Z-codes, year, hospital characteristics, and county characteristics (Table 4).All analyses were conducted using STATA 16.0 SE (StataCorp LLC).This study of de-identified hospital-level data was exempt from the New York University Institutional Review Board.

Results
Our analysis of hospital-based Z-code data found that, on average, 56.9% of hospitals had at least 1 Z-code recorded each year between 2017 and 2021.However, with the exception of housing, rates of Z-code documentation were low.While more than half of the hospitals in our sample recorded Z-codes relating to housing (56.0%), less than 10% documented a single Z-code on a single patient relating to food insecurity (2.6%), utility needs (5.0%), interpersonal safety (5.7%), or transportation (1.6%).For most Z-code categories, documentation increased over time, with the largest increase occurring for food insecurity (from 0.8% of hospitals in 2017 to 8.8% of hospitals in 2021).Only Z-code documentation for interpersonal safety decreased (from 5.8% of hospitals in 2017 to 5.5% of hospitals in 2021) (Table 2).
Most hospitals in our sample were general medical (77.3%).Approximately two-fifths were part of teaching hospitals (39.6%) and two-thirds were part of a larger health care system (66.7%).Approximately one-quarter of hospitals in our sample were CAHs (23.9%), and approximately half were nonprofit hospitals (53.1%).Most hospitals were small (36.5%) or medium-sized (38.4%) (Table 3).

Discussion
Z-code documentation is being encouraged by CMS for risk-adjustment and facilitating consistent comparisons and analyses of SDOH across health care settings.Despite this encouragement, we found a low rate of Z-code adoption across most domains, with the relatively higher documentation of housing Z-codes that we found being consistent with prior literature. 8,15Importantly, over the span of our study, there was merely a marginal uptick in Z-code adoption.This trend underscores a persistently low inclination towards Z-code documentation among health care providers, despite its evident significance.Given the unfolding of the COVID-19 pandemic during our study period, one might hypothesize an accelerated recognition of the importance of SDOH, since SDOH were related to COVID-19 outcomes.However, our findings suggest that, even in the face of such a significant global health crisis, the adoption rate of Z-codes remained stubbornly low.This raises questions on the preparedness and emphasis that health care settings place on holistic understanding of SDOH.To bridge this gap, hospitals may benefit from comprehensive training programs that emphasize the importance of all SDOH domains and for establishing robust systems and partnerships that can address the myriad of nonmedical needs patients may present with beyond housing.
Our study sheds light on specific facility characteristics influencing the utilization of Z-codes to document HRSN for Medicare FFS patients.Hospitals that were general medical, part of a teaching institution, affiliated with larger health systems, and of medium to large size had greater odds of utilizing Z-codes.It is plausible that these hospitals, given their larger infrastructure and resources, might be better positioned to integrate new workflows required to screen for HRSN and document need using Z-codes. 11Conversely, CAH and public hospitals have decreased odds of such utilization; given the pivotal role these hospitals play in serving more vulnerable populations, it is important to address the barriers they face in regularly documenting Z-codes.
There are several limitations to our study.First, our study is restricted to Medicare FFS claims and excludes other payers.Considering the influence of other value-based care incentive programs in Medicaid and managed Medicare on the documentation of social needs, our dataset might not fully capture the extent of Z-code utilization across hospitals.This is particularly true if these codes are documented more frequently among patients not covered by Medicare FFS.However, our metric for Z-code adoption-represented by hospitals with at least 1 Medicare FFS claim noting a Z-code-is a relatively conservative measure, setting a minimal threshold for documentation, the analysis of which might generate different results.Second, hospitals may be documenting HRSN in their electronic health records systems but not fully converting these needs into Z-codes for claims for billing purposes. 16Nonetheless, to the extent that CMS uses billing coding to calculate risk scores, those hospitals may miss reimbursement opportunities.Second, while we have captured patterns of Z-code capture at the county level, our hospital-based analysis does not incorporate individual patient characteristics. 8Third, our findings are constrained by the data available up to 2021, potentially missing recent shifts in practices or policies following COVID-19.Finally, our measure of Z-code documentation-at least 1 Z-code documented for at least 1 patient per year-presented a very low bar for "documentation."For Z-codes to be effectively used, a substantial proportion of patients will need to have documentation.Health Affairs Scholar, 2024, 2(1), 1-5

Figure 1 .
Figure 1.Percentage of hospitals with at least 1 Z-code recorded, overall and in each category, per year (2017-2021).Medicare fee-for-service claims data for 42 million patients were aggregated at the hospital level.

Table 1 .
Social determinants of health domains and Z-code cross-walk.

Table 2 .
Percentage of hospitals with at least 1 Z-code recorded per year (2017-2021).fee-for-service claims data for 42 million patients aggregated at the hospital level.a Describes change from 2017 to 2021. Medicare
n = 27 385 observations across 5685 hospitals.Hospital characteristics were derived from the American Hospital Association (AHA) survey.