This paper introduces a special issue examining data infrastructure for patient-centered outcomes research that addresses health equity. Addressing the persistence and breadth of health inequalities in the United States requires ongoing expansion of the capacity to detect, monitor, and understand them. Data infrastructure is itself an equity policy matter: decisions about what data to collect, on whom, and in what form determine whose health inequities can be observed and, consequently, addressed. The 9 research papers in this issue are organized around 4 themes. The first examines new data sources and standards that extend what health systems can observe about the populations they serve. The second demonstrates how linking data across sectors enables analysis of social determinants that clinical records alone cannot capture. The third addresses strategies for disaggregating populations that are small, internally heterogeneous, or historically excluded from data collection. The fourth considers implementation—the organizational and institutional conditions under which new data systems are successfully adopted in practice. Together, the papers illustrate both the scientific possibilities created by deliberate investment in data capacity and the policy stakes of sustaining it.
Background: The development of clinical tools to combat extreme heat events (EHEs) is urgently needed. However, the collection, linkage, and application of data and technology required to address the health consequences of EHEs—through individualized decision-making, population-focused interventions, and health system planning—remain in its infancy despite the wealth of data infrastructure in health care systems. Methods: In this paper, we describe a use case for data-intensive system architecture that can enable best practices for addressing EHE-related health risks in older adults with cardiovascular disease (CVD), a population uniquely vulnerable to EHEs. Results: Descriptions of various data sources integrated into a modular approach are discussed that allows multilevel (ie, individual-level, population-level) evaluation of EHE-related risk. Individual data streams include batched data from personal digital health devices such as wearables, indoor temperature sensors, and electronic medical record data linked through unique identifiers. Data collection, processing, and analysis as well as related challenges (eg, data quality, processing requirements, and health care system attribution) are also discussed. How this data architecture can then address important preclinical, clinical, and related questions are then described, including: (1) which physiological signals (including cardiovascular and sleep measures) may best anticipate EHE-related health care utilization in older adults with CVD; (2) how do heat thresholds that increase EHE-related health care utilization differ by medication use and type, and comorbidities; and (3) how does indoor versus outdoor temperature measures influence these associations—all understudied aspects of EHE risk in older adults. Conclusions: With considered effort and expertise, a modular data architecture that allows the combination of different elements will enable the development of clinical tools to address EHE-related health risk among older adults with CVD at multilevel scales.
Background: Medicaid race and ethnicity data quality continues to pose challenges for patient-centered outcomes research. Collection practices and data systems vary; according to the Centers for Medicare & Medicaid Services Data Quality Atlas, many states have race and ethnicity data quality of concern (medium, high, or unusable), including Maryland. Objective: This study links data from 3 sources to reduce the percentage of Maryland Medicaid enrollees with an unknown race and ethnicity and improve the accuracy of population-level data. Research Design: People enrolled in Maryland Medicaid (MMIS2) at any point during calendar year 2023 ( N =1,898,041) were matched to data from Maryland’s state health insurance marketplace (MHBE) and designated health information exchange (CRISP). Enrollees were assigned a single race and ethnicity value from the data sources in the following order: MHBE; CRISP; historic MMIS2; current MMIS2. If a participant had an unknown race and ethnicity, the next source was used. Results: Most Medicaid enrollees (97.8%) were found in MHBE and/or CRISP. The study methodology allowed for greater disaggregation of the population by race and ethnicity and reduced the percentage with an unknown race and ethnicity from 23.0% to 1.0%. The distribution of enrollees by race and ethnicity after applying the study method was better aligned with American Community Survey benchmarks than the original data. Conclusions: These results will help stakeholders in Maryland better identify disparities in patient-centered outcomes and advance health equity goals. This methodology can assist researchers and policymakers in other states with similar data quality issues, as part of a larger effort to improve Medicaid race and ethnicity data.
Background: Common Data Models (CDMs) standardize health data from disparate observational sources, enabling the use of real-world data in more efficient, collaborative observational research studies. Until recently, the Observational Medical Outcomes Partnership Common Data Model (OMOP CDM) had limited infrastructure and ontology options to store a person’s race and ethnicity data. This made using these data challenging for patient-centered outcomes research (PCOR) and health equity studies. Objective: The primary goals are to outline the Observational Health Data Sciences and Informatics (OHDSI) community’s collaborative efforts to improve race and ethnicity representation in the OMOP CDM and to explain how these enhancements enable a more detailed and comprehensive representation of individuals’ racial and ethnic identities. These improvements ensure that research findings are both relevant to everyday clinical practice and applicable across diverse demographic groups. Methods: The OHDSI community collaboratively addressed limitations in race and ethnicity representation within the OMOP CDM through a structured, inclusive process. Key workgroups worked together, engaging diverse stakeholders. Patient input played a pivotal role in shaping enhancements like multivalue storage. The multistep enhancement process included community-wide feedback at every step. Results: The OHDSI community significantly enhanced the OMOP CDM’s ability to represent race and ethnicity data, improving its granularity, inclusivity, and flexibility. These updates expand the race and ethnicity value sets, address multiracial and multiethnic identities, and enable more accurate and granular PCOR and health equity studies. Conclusion: This enhancement to the OMOP CDM reduces the gap between data stored in source systems and data converted to the OMOP CDM. The enhanced data model enables more detailed, nuanced health equity research and eliminates biases previously associated with limited demographic representation.
Background: Out-of-home placement (OOHP) through child welfare, residential substance treatment, or extended inpatient care has life-altering impacts on child health and generates an outsized proportion of pediatric Medicaid costs. Early risk identification may prevent OOHP but is challenged by the difficulty of navigating state and federal data sharing regulations and multiple disconnected data systems. Identifying ways to streamline and simplify this process are critical to OOHP prevention. Objectives: To describe the development and implementation of an expert-derived data-driven algorithm to identify OOHP risk among Medicaid-enrolled children residing in 2 Ohio counties between 2022 and 2024 (n=27,000), discuss practical considerations for data sharing and linkage across multiple agencies, and identify lessons learned. Findings: A cross-sector team of government and academic partners developed a 3-category OOHP risk stratification algorithm incorporating data from Medicaid claims, child welfare information systems, area-level social determinants of health, and patient-reported health risk assessments as part of Ohio’s Integrated Care for Kids Model. Substantial administrative and legal processes were required to link data sources. Lessons learned include: (1) understanding legal, political, and security structures associated with use of administrative data is critical; (2) strong working relationships with data partners can ensure success; (3) as administrative data are dynamic and may be retroactively updated, appropriate lookback periods are necessary for accuracy; and (4) while executing data use agreements can be challenging, the relationships built can have lasting benefits. Conclusions: Data-driven risk stratification models have the potential to reduce cost, time, and redundancies in identifying families who would benefit from OOHP prevention supports. Balancing data-sharing challenges with the value gained from linking data is vital.
BACKGROUND:The COVID-19 pandemic led to economic and policy changes that increased Medicaid enrollment and reduced disenrollment. Tracking the impact of these changes on the composition of enrollees is crucial for resource allocation and program evaluation. However, Medicaid data currently lack the necessary demographic, social, and economic information about enrollees. OBJECTIVE:Enhance Medicaid enrollment data by linking it with nationally representative survey data to compare the composition of enrollees across various social determinants of health characteristics before and after the COVID-19 pandemic. RESEARCH DESIGN:We utilize individual-level Medicaid enrollment records (TAF, 2018-2021) before and during the pandemic, linked to restricted American Community Survey (ACS, 2021) microdata. RESULTS:Almost 95% of enrollees in our analytic sample received an anonymous identifier (Protected Identification Key), and nearly 1% were found in ACS data. By comparing Medicaid enrollees in different enrollment cohorts, we find that the pandemic caused significant compositional changes, particularly among the newly enrolled. Our findings indicate that those experiencing a major health or economic shock, either directly or through a family member, relied on Medicaid, likely as a temporary source of health insurance during the pandemic. CONCLUSIONS:Linking individual-level records between Medicaid and ACS data effectively addresses a crucial gap in current data capacity. The integrated data can be utilized, repurposed, and expanded by incorporating additional survey and administrative records to enhance the utility of Medicaid data for future research.
This perspective examines what is lost when race and ethnicity data are not adequately collected or analyzed. It highlights the importance of these data for identifying disparities, informing the design of effective interventions, and ensuring that clinical trials achieve meaningful and equitable impact. It also discusses how the absence of such data can limit the broader applicability and effectiveness of research findings. The conclusion is that when high-quality data are available and analyzed thoughtfully, they enable the identification of inequities, illuminate underlying mechanisms, guide targeted interventions, support accountability, and ultimately improve the health of all Americans.
Background: Uninteroperable health data impedes care and research—especially for people with multiple chronic conditions (MCC) or from disadvantaged communities. By enabling the sharing of person-centered data, standards-based care plans may improve health outcomes and reduce disparities. Objective: Develop and test interoperable electronic e-Care planning tools to collect, aggregate, and share person-centered data for MCC healthcare and research. Methods: Using participatory and agile design, we developed, implemented, and evaluated e-Care plan tools—including data standards and clinician- and patient/caregiver-facing e-Care plan applications (eCarePlanner and MyCarePlanner). The Consolidated Framework for Implementation Research Process Redesign and the Systems Engineering Initiative for Patient Safety model informed a mixed methods evaluation to gather feedback from patients, caregivers, and clinicians across formative, iterative, and summative stages. Results: The apps successfully connected with 4 electronic health records at 17 institutions. Evaluation participants included 57 patients/caregivers and 15 clinicians. Patients and caregivers were predominantly aged 65 years or older, White, and well-educated. Most participants were comfortable using the app (97%). Most felt app loading was timely (90%), and the app would support complex care coordination (63%). Usability scores were lowest for the ability to improve care team communication (48% agreed/strongly agreed) and for inconsistencies across app sections (37% agreed/strongly agreed). Key themes from interviews included challenges in moving health information across settings and the apps’ potential to overcome these challenges, thereby improving the frequency and quality of care planning. Conclusion: This project provides a proof-of-concept for standards-based tools to collect, aggregate, and share patient-centered health and social data across health care and research settings.
Background: Social adversity contributes to poor health outcomes for children after liver transplantation (LT), including greater morbidity and mortality. Although social risk screening is standard practice in pediatric primary care, its adoption in pediatric LT (pLT) remains limited, despite the significant prevalence of social risks among these families. Objectives: To evaluate barriers and facilitators to social risk screening implementation in pLT settings across a multicenter learning health network. Research Design: A mixed-methods study utilizing surveys and semistructured interviews. Subjects: Health care practitioners involved in pLT care across North American centers in the Starzl Network for Excellence in Pediatric Transplantation (SNEPT) Network. Participants included multidisciplinary team members, notably transplant physicians, social workers, and research coordinators. Measures: We surveyed center leads to identify center-level implementation challenges. We conducted 1-on-1 interviews with transplant team members to identify barriers and facilitators. We analyzed qualitative data using the Capability, Opportunity, Motivation-Behavior model, an implementation science model for developing targeted interventions. Results: We surveyed 10 centers, of which 40% of liver transplant clinics reported actively screening patients for Social Determinants of Health (SDoH). Most practitioners indicated reliance on social workers and cited limited resources as barriers to implementation. We also conducted interviews with 18 practitioners across 11 centers. Reported barriers included uncertainty about the tool’s added value, time and space constraints during patient encounters, and challenges with data entry and sharing across a multicenter network. Facilitators included institutional support, interdisciplinary collaboration, and integration into electronic health records. Conclusion: Efforts to increase adoption should focus on improving practitioners’ experiences with the tool as well as further assessing and disseminating its potential value in improving outcomes. Strategies for addressing logistical challenges to adapting workflows and simplifying network data management should be considered. This study establishes a foundation for improving screening rates and data capacity.
BACKGROUND:Housing wealth can be leveraged to maintain economic security when expenses exceed income, allowing households flexibility when managing a new chronic disease diagnosis. We linked housing and electronic health record (EHR) data to examine how housing wealth and stability relate to type 2 diabetes (T2D) outcomes. METHODS:EHR data from patients diagnosed with T2D were linked to housing data (tenure, foreclosure, home equity, reverse mortgage, and loan-to-value ratio) using residential address. Multinomial logistic regression models were used to estimate associations between housing variables measured in the year before T2D diagnosis and glycemic control a year or more after T2D diagnosis. RESULTS:Among 5810 patients with T2D and complete housing data, patients with more home equity [relative risk ratio (rrr)=0.52, 95% CI: 0.323-0.838, P<0.001] were less likely to experience HbA1c >9% 1 year after diagnosis, and owners who were actively extracting home equity through a reverse mortgage (rrr=2.137, 95% CI: 0.008-1.511, P=0.048) were also more likely to have HbA1c >9% compared with patients with HbA1c<6.5%. Homeowners were less likely than renters to experience HbA1c >9% (rrr=0.798, 95% CI: 0.672-0.948, P=0.010). CONCLUSION:Patients diagnosed with T2D who had more housing wealth were less likely to experience challenges with glycemic control after diagnosis.
Background: Learning from promising maternal health interventions is limited by the lack of standardized outcome measures needed for their evaluation. We describe the development and utilization of Maryland’s severe maternal morbidity (SMM) surveillance by the Maryland Maternal Health Innovation (MDMOM) program. Methods: To ascertain the development, attributes, and data collected by SMM surveillance, we use 2020–2024 MDMOM program records and 2023 surveillance data. Hospital reports and MDMOM’s 2023 maternal health care provider survey data are used to document policy and practice changes attributable to participation in SMM surveillance. Data collected from MDMOM’s “Learning from Adverse Events in Maryland” trainees are analyzed to explore knowledge score changes (Wilcoxon signed-rank test) and training satisfaction. Results: SMM surveillance is a simple, flexible, stable, high-quality system for identifying areas of care in need of improvement. Before October 2024, when state legislation mandated all hospitals to participate, 24 of 32 hospitals (80% of the state’s annual deliveries) engaged in SMM surveillance. The data are used to assess SMM levels and contributors, disparities (eg, intensity of care practices and outcome differences between patient groups), and preventability. Knowledge scores improved after the training on SMM surveillance findings ( P <0.001). Data can be used for quality improvement and to generate PCOR hypotheses. Conclusions: Maryland established hospital-based SMM surveillance, demonstrated its value, and gained legislative support for statewide scale-up. This model surveillance system can be replicated in other states, aiming to better understand and address SMM and SMM disparities.
BACKGROUND:Timely cancer treatment is a major component of high-quality cancer care and is associated with better outcomes. Disruption caused by physician departures may adversely affect timely care. OBJECTIVE:To examine the effects of breast cancer surgeon departures on timely breast cancer surgery. METHODS:Using Medicare fee-for-service claims, we identified patients 66-99 years of age diagnosed with breast cancer between 2017 and 2019 who underwent a cancer-directed surgery. Surgical oncologists, plastic and reconstructive surgeons, and general surgeons performing these surgeries were included in our breast cancer surgeon cohort. Surgeon movement was tracked using changes in billing locations. The exposure variable was the number of surgeon departures in the prior 12 months from the hospital service area (HSA) where a patient's diagnostic biopsy was performed. Multivariable hierarchical logistic regression was used to estimate the associations between surgeon departures and surgical delays of more than 60 days. RESULTS:Our study cohort included 122,159 patients with breast cancer, of whom 108,239 (89%) were non-Hispanic white and 97,565 (80%) resided in a metropolitan area. We found that 52,708 (43.1%) received care within an HSA that experienced at least 1 surgeon departure. Each additional surgeon departure was associated with a 4% increase in odds of delay >60 days (95% CI: 1.02-1.05; P<0.001). CONCLUSIONS:We found a modest association between surgeon departures and the odds of breast cancer surgical delay. These results suggest that, on average, hospitals are often able to deliver timely care in the context of surgeon departures. Additional research is needed to determine whether this finding generalizes to other cancer types and to identify the strategies that enable health system resilience to surgeon departures.
BACKGROUND:For-profit hospital ownership has been associated with worse patient outcomes, but the mechanisms remain unclear. Underinvestment in nurse staffing is a plausible pathway linking for-profit ownership to performance. OBJECTIVES:To examine whether differences in nurse staffing mediate associations between for-profit hospital ownership and patient, nurse, and hospital outcomes. DESIGN:Cross-sectional study using 2024 data on hospital ownership, registered nurse survey responses, Medicare claims, and Hospital Compare measures. SUBJECTS:A total of 143 for-profit and 798 nonprofit adult, nonfederal, acute care hospitals across 10 US states; >1.17 million Medicare fee-for-service admissions; and 17,368 direct-care bedside registered nurses. MEASURES:Patient-level 30-day mortality and readmission (Medicare Provider Analysis and Review); hospital-level hybrid risk-standardized mortality and readmission and HCAHPS overall hospital rating (CMS Hospital Compare); nurse burnout (Maslach Burnout Inventory emotional exhaustion subscale) and related outcomes (Penn Nurses All survey). The hypothesized mediator was hospital-level nurse staffing. RESULTS:Among surgical patients, for-profit ownership was associated with a 0.48 percentage-point higher 30-day mortality (95% CI: 0.24-0.73), 45% mediated through staffing. Among medical patients, 1.13 percentage-point higher 30-day mortality (95% CI: 0.52-1.73), 40% mediated. Readmission findings were similar. For-profit status was associated with worse HCAHPS ratings (-4.45 percentage points; 95% CI: -5.20 to -3.60), 31% mediated. Nurse burnout was 9 percentage points higher (95% CI: 5.8-12.4), 67% mediated. CONCLUSIONS:For-profit hospital ownership is associated with worse patient and nurse outcomes, and differences in nurse staffing account for a share of these disparities.
BACKGROUND:Availability of childbirth services is declining nationally, especially in rural communities, where risks of maternal and infant mortality are elevated. OBJECTIVE:The study aims to describe the availability of hospital-based childbirth services and higher-level neonatal care in rural and urban US counties from 2010 to 2023 and to examine trends in availability. METHODS:Survey, administrative data, and primary sources were used to identify childbirth services (obstetric and basic well-infant services) and higher-level neonatal care (intermediate or intensive care) at short-term acute care hospitals in rural (n=1958) and urban (n=1186) US counties. For each year (2010-2023), counties were categorized into 3 mutually exclusive service categories: having (1) no childbirth services, (2) only childbirth services, or (3) both childbirth services and higher-level neonatal care. We estimated predicted percentages of rural and urban births occurring to residents of counties by service category using multinomial logistic regression models. RESULTS:The predicted percent of urban births occurring to residents of counties with both childbirth services and higher-level neonatal care increased from 87.0% in 2010 to 88.6% in 2023 (P=0.007), while this percent remained statistically stable among rural counties at 17.0% in 2010 and 16.6% in 2023 (P=0.762). In 2023, 83.4% of rural births occurred to residents of counties without higher-level neonatal care, compared with 11.4% of urban births. CONCLUSIONS:In 2023, 88.6% of urban births occurred to residents of counties with both childbirth services and higher-level neonatal care, compared with 16.6% of rural births. From 2010 to 2023, rural-urban differences in the availability of specialized care for high-risk infants widened.
Background: Racial and ethnic disparities in patient-reported experiences exist in the Veterans Health Administration (VA) and in VA-paid health care from non-VA providers (“community care”), but how these disparities compare is unknown. Objectives: We compared racial and ethnic disparities in patient-reported experiences between VA primary care and community care, and calculated differences in VA versus community care by patient race and ethnicity. Methods: Using cross-sectional data from the 2020 to 2023 VA Survey of Healthcare Experiences of Patients—VA Primary Care and Community Care surveys, we used linear and logistic regressions to assess patient-reported access, care coordination, communication, and provider satisfaction for each racial and ethnic group by setting, and calculated differences-in-differences by race and ethnicity and setting. Results: Primary care provided in community care had larger racial and ethnic disparities in patient-reported experiences of care than VA. Compared with White Veterans, age and sex-adjusted community care disparities exceeded VA disparities for communication among Black Veterans by 13 percentage points; care coordination for American Indian/Alaska Native, Black, and multirace Veterans by 22, 12, and 16 percentage points, respectively; and provider satisfaction for American Indian/Alaska Native Veterans by 34 percentage points. Conclusion: Our findings highlight potential unintended consequences of VA’s efforts to improve timely access by increasing reliance on community care, which may inadvertently widen disparities and contribute to worse care experiences. We suggest several policy implications, including tailoring existing patient tools to navigate community care to minoritized groups experiencing disparities, and providing VA and CC metrics to Veterans to support informed decisions.
BACKGROUND:Racial and ethnic disparities in patient-reported experiences exist in the Veterans Health Administration (VA) and in VA-paid health care from non-VA providers ("community care"), but how these disparities compare is unknown. OBJECTIVES:We compared racial and ethnic disparities in patient-reported experiences between VA primary care and community care, and calculated differences in VA versus community care by patient race and ethnicity. METHODS:Using cross-sectional data from the 2020 to 2023 VA Survey of Healthcare Experiences of Patients-VA Primary Care and Community Care surveys, we used linear and logistic regressions to assess patient-reported access, care coordination, communication, and provider satisfaction for each racial and ethnic group by setting, and calculated differences-in-differences by race and ethnicity and setting. RESULTS:Primary care provided in community care had larger racial and ethnic disparities in patient-reported experiences of care than VA. Compared with White Veterans, age and sex-adjusted community care disparities exceeded VA disparities for communication among Black Veterans by 13 percentage points; care coordination for American Indian/Alaska Native, Black, and multirace Veterans by 22, 12, and 16 percentage points, respectively; and provider satisfaction for American Indian/Alaska Native Veterans by 34 percentage points. CONCLUSION:Our findings highlight potential unintended consequences of VA's efforts to improve timely access by increasing reliance on community care, which may inadvertently widen disparities and contribute to worse care experiences. We suggest several policy implications, including tailoring existing patient tools to navigate community care to minoritized groups experiencing disparities, and providing VA and CC metrics to Veterans to support informed decisions.
BACKGROUND:Older injured patients are at risk for inappropriate pain management, including undertreatment of their pain and inappropriate medication use. The primary aim of this study was to examine the impact of geriatric comanagement on pain medication prescribing in hospitalized older adults. SETTING:Single-site urban academic hospital. PARTICIPANTS:Patients aged 65 years and older admitted to orthopedic trauma, general trauma, and neurosurgery services. METHODS:Pre-post intervention study from 2017 to 2020, using propensity score matching. We compared pain medication prescribing before and after implementation of a geriatric comanagement program. We tracked whether pain medications were received during hospitalization, and the total number of administrations per patient for each class of pain medications. RESULTS:A total of 2640 patients were included, and analytic sample sizes varied by outcome and subsample studied. Comanaged patients in the postintervention group were more likely to receive pain medications than the preintervention group (14.2%, 95% CI: 4.0-24.4, P=0.006), including opioids (20.1%, 95% CI: 2.5-37.6, P=0.025) and nonopioids (15.6%, 95% CI: 4.0-27.3, P=0.009). Comanaged patients had an increase in the administration of nonopioids (62.1%, P<0.001) without a significant increase in the administration of opioids. CONCLUSION:The introduction of geriatric comanagement was associated with changes in pain medication prescribing practices. While our study did not include pain outcome measures, the observed changes in pain medication prescribing add to the growing evidence supporting the benefits of geriatric comanagement models of care because appropriate prescribing facilitates pain management.
BACKGROUND:Federally Qualified Health Centers (FQHCs) are essential safety-net providers, but traditional payment mechanisms may focus on volume over quality-of-care. Alternative payment models (APMs) have been introduced to enhance value-based purchasing. It is hypothesized that APMs would improve flexibility and incentivize quality of care, but evidence remains limited, especially for maternal and child health and cardiovascular conditions. OBJECTIVES:We aim to evaluate the impact of APMs on quality-of-care in FQHCs in the United States. METHODS:We used a 2-stage difference-in-differences design to compare FQHCs in states adopting APMs with those that did not, using data from 2014 to 2024 in 695 FQHCs. Quality-of-care was measured using 6 indicators, including 2 associated with maternal and child services (early prenatal care and dental sealants), 2 with preventive care (depression screening and tobacco use screening), and 2 with cardiovascular disease management (cardiovascular and ischemic vascular disease treatment). To examine heterogeneity of the effects, the analysis stratified FQHCs into 3 groups based on patient volume. RESULTS:APM adoption was significantly associated with improvements in 4 quality measures: a 3.82-percentage-point increase in early prenatal care, a 9.12 percentage point increase in dental sealants, a 5.08 percentage point increase in depression screening, and a 3.50 percentage point increase in ischemic vascular disease treatment. Stratified analyses showed that large FQHCs in APM-adopting states experienced statistically significant improvements in 3 indicators. CONCLUSIONS:Overall, APMs were associated with positive but modest improvements in quality-of-care at FQHCs.
BACKGROUND:Identifying clinically intended medication discontinuations at scale may help generate evidence to inform deprescribing. We developed an algorithmic approach to identifying such discontinuations, combining text strings applied to clinical documentation with medication order data for 5 different drug groups: oral hypoglycemics, statins, antihypertensives, bladder antimuscarinics, and antithrombotics. DESIGN:The study population (n=1588) comprised individuals aged older than 65 with ≥90-day gaps in dispensing classified through manual review as having a clinically intended medication discontinuation or not. This gold-standard cohort was randomly divided into development and validation subsamples for each drug group. We developed text strings from clinical documentation reflecting clinical intent to discontinue a medication (or not). Text strings were usually simple [eg, "stop (medication)"], but required tailoring to drug groups [eg, temporarily "hold (antithrombotic)"]. Text strings queried clinical documentation and were supplemented with order discontinuation data, if available. We calculated sensitivity and specificity for identifying intended discontinuations for text alone, discrete data alone, and the full algorithm across validation samples. RESULTS:Sensitivity and specificity for the full algorithm were 80% and 85% for oral hypoglycemics (n=467), 75% and 95% for statins (n=282), 82%, 75% for antihypertensives (n=599), 77% and 80% for bladder antimuscarinics (n=80), and 74% and 78% for antithrombotics (n=160). The full algorithm had higher sensitivity than either text or order data alone. CONCLUSIONS:Text-based approaches supplemented by medication order data can identify clinically intended medication discontinuations at scale with moderate specificity and sensitivity. This may reduce misclassification relative to using claims data to generate evidence about deprescribing.
BACKGROUND:While payers and health systems are increasingly interested in addressing health-related social needs (HRSNs) to reduce health care use and align with value-based care incentives and regulatory requirements, limited evidence enables actionable strategies to manage and address social needs across population groups. OBJECTIVES:To understand the relationship between HRSNs and health care access and use for adults with Medicare, Medicaid, and private health insurance coverage. RESEARCH DESIGN:Survey data from adult participants from the All of Us Research Program (2017-2023; n=126,490) and logistic regression were used to examine the association between food insecurity and housing instability and having a usual source of care, seeing a health care provider, and forgoing care due to cost-stratified by payer type. RESULTS:The prevalence of HRSNs varied substantially by insurance: food insecurity affected 49.20% of Medicaid beneficiaries 18-64 years of age, 11.89% of privately insured adults 18-64 years of age, and 5.20% of Medicare beneficiaries 65 years of age and older; housing instability affected 54.61%, 25.18%, and 14.99%, respectively. Food insecurity was associated with lower odds of having a usual source of care for Medicaid and privately insured adults, lower odds of use for privately insured adults, and higher odds of forgone care for all 3 payer types. Housing instability was associated with lower odds of having a usual source of care for all 3 papers and with higher odds of forgone care for all groups. CONCLUSIONS:HRSNs like food insecurity and housing instability vary substantially by payer type and influence health care access and utilization in different ways. This can inform the design of payer-specific health management strategies that incorporate HRSNs.
