Referral Regions Research Articles

Estimates of intra-cluster correlation coefficients from 2018 USA Medicare data to inform the design of cluster randomized trials in Alzheimer’s and related dementias

BackgroundCluster randomized trials (CRTs) are increasingly important for evaluating interventions embedded in health care systems. An essential parameter in sample size calculation to detect both overall and heterogeneous treatment effects for CRTs is the intra-cluster correlation coefficient (ICC) of both outcome and covariates of interest. However, obtaining advance estimates for the ICC can be challenging. When trial outcomes will be obtained from routinely collected data sources, there is an opportunity to obtain reliable ICC estimates in advance of the trial. Using USA national Medicare data, we estimated ICCs for a range of outcomes to inform the design of CRTs for people living with Alzheimer’s and related dementias (ADRD).MethodData from 2018 Medicare Fee-for-Service beneficiaries, specifically, 1,898,812 individuals (≥ 65 years) with diagnosis of ADRD within 3436 hospital service areas (treated as clusters) and 306 hospital referral regions (treated as fixed strata), were used to calculate unadjusted and adjusted ICC estimates for three outcomes: death, any hospitalizations, and any emergency department (ED) visits and three covariates: age, race and sex. We present both overall and stratum-specific ICC estimates. We illustrate their use in sample size calculations for overall treatment effects as well as detecting treatment effect heterogeneity.ResultsThe unadjusted overall ICCs for death, hospitalizations, and ED visits were 0.001, 0.010, and 0.017 respectively. Stratum-specific ICCs varied widely across the 306 HRRs: median 0.001, 0.010 and 0.025 for death, hospitalizations, and ED visits respectively and 0.007, 0.001, and 0.080 for age, sex and race. An interactive R Shiny app is provided that allows users to retrieve estimates overlayed on a map of the USA.ConclusionsWe presented both adjusted and unadjusted ICCs for outcomes as well as unadjusted ICCs for covariates of potential interest from population-level data in the USA and demonstrated how the estimates may be used in sample size calculations for CRTs in ADRD.

Development and Evaluation of Pediatric Acute Care Hospital Referral Regions in Eight States

To develop a set of pediatric acute care hospital referral regions for use in studying pediatric acute care delivery, and test their utility relative to other regional systems. We used state-level administrative databases capturing all pediatric acute care in eight states to construct novel referral regions. We first constructed pediatric hospital service areas (PHSAs) based on 5,837,464 pediatric emergency department encounters. We then aggregated these PHSAs to pediatric hospital referral regions (PHRRs) based on 344,440 pediatric hospitalizations. Finally, we used three measures of spatial accuracy (localization index, market share index, and net patient flow) to compare this novel region system with the Dartmouth Atlas, designed originally to study adult specialty care, and the Pittsburgh Atlas, designed originally to study adult acute care. The development procedure resulted in 717 novel PHSAs, which were then aggregated to 55 PHRRs across the included states. Relative to HRRs in the Dartmouth and Pittsburgh Atlases, PHRRs were fewer in number and larger in area and population. PHRRs more accurately captured patterns of pediatric hospitalizations, (eg, mean localization index: 69.1 out of 100, compared with a mean of 58.1 for the Dartmouth Atlas and 62.4 for the Pittsburgh Atlas). The use of regional definitions designed specifically to study pediatric acute care better captures contemporary pediatric acute care delivery than the use of existing regional definitions. Future work should extend these definitions to all US states to enable national analyses of pediatric acute care delivery.

Regional variation in diagnostic intensity of dementia among older U.S. adults: An observational study.

Geographic variation in diagnosed cases of Alzheimer's disease and related dementias (ADRD) could be due to underlying population risk or differences in intensity of new case identification. Areas with low ADRD diagnostic intensity could be targeted for additional surveillance efforts. Medicare claims were used for a cohort of older adults across hospital referral regions (HRRs). ADRD-specific regional diagnosis intensity was measured as the ratio of expected new ADRD cases (estimated using population demographics, risk factors, and practice intensity) compared to observed ADRD-diagnosed cases. Crude new ADRD diagnosis rate ranged from 1.7 to 5.4 per 100 across HRRs. ADRD-specific diagnosis intensity ranged from 0.69 to 1.47 and varied most for Black, Hispanic, and the youngest (66-74) subgroups. Across all subgroups, ADRD diagnosis intensity was associated with 2-fold difference in receiving an ADRD diagnosis. Where one resides influences the likelihood of receiving an ADRD diagnosis, particularly among those 66-74 years of age and minoritized groups. Rate of new Alzheimer's disease and related dementias (ADRD) case identification varies geographically across the United States. Variation in case identification is greatest in Black, Hispanic, and young-old groups. Intensity of diagnosis (ie, case identification) unrelated to population risk differs across place. Likelihood of receiving an ADRD diagnosis varies 2-fold based on place of residence.

Evolution of overall survival (OS) after curative-intent lung cancer surgery in a population-based Mississippi Delta cohort.

278 Background: Surgical resection is the most important curative modality for patients with lung cancer, with long-term OS influenced by both surgical quality and cancer biology. We compared surgical care and outcomes over 4 eras in the population-based Mid-South Quality of Surgical Resection (MS-QSR) cohort, an ongoing regional surgical and pathologic quality improvement (QI) dissemination and implementation project. Methods: We implemented a surgical lymph node collection kit and a novel pathology gross dissection method across 12 hospitals encompassing >95% of lung cancer resections in 5 contiguous Dartmouth Hospital Referral Regions. We analyzed the MS-QSR cohort at 5-year intervals characterized by stepped-wedge implementation of specific QI interventions: Era 1 (2004-2008; baseline, no QI), Era 2 (2009-2013; quality feedback, surgical QI pilot), Era 3 (2014-2018; surgical QI implementation), and Era 4 (2019-2023; combined surgical QI and pathology QI implementation). We used standard statistical methods to compare surgical quality; Kaplan-Meier plot, log-rank test, and Cox proportional hazards model with Era 1 as reference to compare OS across eras. Results: Of 7240 resections, 14%, 31%, 29%, and 26% were performed in Eras 1-4, respectively (Table). Attainment of the stringent International Association for the Study of Lung Cancer (IASLC) definition of complete resection increased from 0% in Era 1 to 32% in Era 4; attainment of the American College of Surgeons Operative Standard 5.8 (ACS_OS 5.8), the current US definition of surgical quality, improved from 4% in Era 1 to 67% in Era 4; 120-day mortality rate decreased from 10% to 4% (p<0.0001 for all comparisons). With median follow-up of 4.7, 5.5, 5.6, and 2 years for Eras 1-4, 3-year OS rates were 60%, 64%, 70%, and 79%; 5-year OS rates were 48%, 52%, 58%, and 70%, respectively (p<0.0001). The Hazard Ratio (HR) for mortality decreased from 0.91 (95% CI: 0.8-0.99) in Era 2 to 0.49 (95% CI: 0.43-0.56) in Era 4. Conclusions: Surgical quality and OS sequentially improved in this diverse, high-risk population-based Mississippi Delta surgical resection cohort. We are evaluating the full range of factors associated with this striking OS improvement. Ongoing work is exploring biologic correlates to this improvement. Variable Era 1 N= 988 Era 2 N=2235 Era 3 N=2107 Era 4 N=1910 Median age (Q1, Q3) 68 (62, 74) 68 (62, 74) 68 (61, 74) 68 (61, 74) Male sex (%) 514 (52%) 1234 (55%) 1082 (51%) 977 (51%) Black race 221 (22%) 416 (19%) 450 (21%) 451 (24%) Attained ACS_OS 5.8 40 (4%) 533 (24%) 1050 (50%) 1272 (67%) IASLC complete resectionUncertain completeness 0 (0%)930 (94%) 194 (9%)1932 (86%) 435 (21%)1594 (76%) 606 (32%)1227 (64%) 120-day mortality 100 (10%) 197 (9%) 139 (7%) 72 (4%) 5-year OS 48 (45-51) 52 (50-55) 58 (56-60) 70 (66-74) HR (95% CI) Reference 0.91 (0.83-0.99) 0.75 (0.69-0.83) 0.49 (0.43-0.56)

Application of Community Detection Methods to Identify Emergency General Surgery–Specific Regional Networks

There is growing interest in developing coordinated regional systems for nontraumatic surgical emergencies; however, our understanding of existing emergency general surgery (EGS) care communities is limited. To apply network analysis methods to delineate EGS care regions and compare the performance of this method with the Dartmouth Health Referral Regions (HRRs). This cross-sectional study was conducted using the 2019 California and New York state emergency department and inpatient databases. Eligible participants included all adult patients with a nonelective admission for common EGS conditions. Interhospital transfers (IHTs) were identified by transfer indicators or temporally adjacent hospitalizations at 2 different facilities. Data analysis was conducted from January to May 2024. Admission for primary EGS diagnosis. Regional EGS networks (RENs) were delineated by modularity optimization (MO), a community detection method, and compared with the plurality-based Dartmouth HRRs. Geographic boundaries were compared through visualization of patient flows and associated health care regions. Spatial accuracy of the 2 methods was compared using 6 common network analysis measures: localization index (LI), market share index (MSI), net patient flow, connectivity, compactness, and modularity. A total of 1 244 868 participants (median [IQR] age, 55 [37-70 years]; 776 725 male [62.40%]) were admitted with a primary EGS diagnosis. In New York, there were 405 493 EGS encounters with 3212 IHTs (0.79%), and 9 RENs were detected using MO compared with 10 Dartmouth HRRs. In California, there were 839 375 encounters with 10 037 IHTs (1.20%), and 14 RENs were detected compared with 24 HRRs. The greatest discrepancy between REN and HRR boundaries was in rural regions where one REN often encompassed multiple HRRs. The MO method was significantly better than HRRs in identifying care networks that accurately captured patients living within the geographic region as indicated by the LI and MSI for New York (mean [SD] LI, 0.86 [1.00] for REN vs 0.74 [1.00] for HRR; mean [SD] MSI, 0.16 [0.13] for REN vs 0.32 [0.21] for HRR) and California (mean [SD] LI, 0.83 [1.00] for REN vs 0.74 [1.00] for HRR; mean [SD] MSI, 0.19 [0.14] for REN vs 0.39 [0.43] for HRR). Nearly 27% of New York hospitals (37 of 139 hospitals [26.62%]) and 15% of California hospitals (48 of 336 hospitals [14.29%]) were reclassified into a different community with the MO method. Development of optimal health delivery systems for EGS patients will require knowledge of care patterns specific to this population. The findings of this cross-sectional study suggest that network science methods, such as MO, offer opportunities to identify empirical EGS care regions that outperform HRRs and can be applied in the development of coordinated regional systems of care.

Save A Leg – Exploring Disparities in Community Characteristics Between Regions of High and Low Leg Amputation Rates

ObjectiveTo explore the disparities in community characteristics between the regions with the highest and lowest leg amputation rates. ApproachThis study used an exploratory method to describe the disparities between community characteristics of these two regions. The study team used several data sources: 1) We used the Dartmouth Atlas Healthcare Map to identify the highest and lowest leg amputation areas. This data was based on leg amputations per 1,000 Medicare enrollees by Hospital referral regions (HRRs) in 2015. 2) We used population estimates and income data from The Census Bureau. 3) We gathered climate, crime index, walk score, bike lanes, and park data from numerous websites. ResultsFlorence, SC has the highest leg amputation rate and Bridgeport, CT has the lowest across the nation. Florence also shows higher cardiovascular disease and diabetes discharge rates. Florence’s population has a slightly higher percentage of persons 65 years and over, a higher percentage of bachelor’s degrees, a higher per capita income, and a lower poverty rate. Florence also has fewer parks, is less walkable or bikeable, and has worse crime. It is worth noting that Bridgeport is close to Yale New Haven Hospital, which has a Limb Preservation Program that could contribute to the region’s extremely low leg amputation rate. ConclusionsThere are disparities in community characteristics between the regions with the highest and lowest leg amputation rates. ImplicationsFuture work needs to investigate the association between community characteristics and leg amputation rates and assess the impact of the Limb Preservation Program.

Exposure to acute ambient temperature extremes and neonatal intensive care unit admissions: A case-crossover study

BackgroundExtreme in utero temperatures have been associated with adverse birth outcomes, including preterm birth and low birthweight. However, there is limited evidence on associations with neonatal intensive care unit (NICU) admissions, which reflect a range of poor neonatal health outcomes. MethodsThis case-crossover study assesses the associations between ambient temperature changes during the week of delivery and risk of NICU admission. Data from the Consortium on Safe Labor (2002–2008) were linked to ambient temperature at hospital referral regions. Adjusted hazard ratios (HR) and 95 % confidence intervals (CI) estimated NICU admission risk with a 1 °C increase on each day of the week of delivery and of the average weekly temperature, adjusted for particulate matter ≤2.5 μm (PM2.5) and relative humidity. We also estimated associations with 1 °C increases and 1 °C decreases in temperatures during weeks of site-specific extreme heat (>90th and 95th percentiles) and cold (<5th and 10th percentiles), respectively. ResultsThere were 27,188 NICU admissions with median (25th, 75th) temperature of 16.4 °C (5.8, 23.0) during the week before delivery. A 1 °C increase in temperature during the week of delivery was not associated with risk of NICU admission. However, analyses of extreme temperatures found that a 1 °C decrease in weekly average temperatures below the 10th and 5th percentiles was associated with 30 % (aHR = 1.30, 95 % CI 1.28, 1.31) and 47 % (aHR = 1.47, 95 % CI 1.45, 1.50) increased risk of NICU admissions, while a 1 °C increase in weekly average temperatures above the 90th and 95th percentiles was associated with more than two- (aHR = 2.29, 95 % CI 2.17, 2.42) and four-fold (aHR = 4.30, 95 % CI 3.68, 5.03) higher risk of NICU admission, respectively. ConclusionsOur study found temperature extremes in the week before delivery increased NICU admission risk, particularly during extreme heat, which may translate to more adverse neonatal outcomes as extreme temperatures persist.

Prostate magnetic resonance imaging utilization and its relationship with advanced prostate cancer detection

BackgroundThe rise in advanced prostate cancer has coincided with increased use of Magnetic Resonance Imaging (MRI), leading to the hypothesis that this increase in surveillance registries is an artifact of more sensitive imaging tools. We assessed the association between regional variation in prostate MRI and advanced prostate cancer diagnoses. MethodsWe utilized SEER-Medicare data (2004–2015), including men > 65 diagnosed with localized prostate cancer. The predictor variable was the utilization of prostate MRI in each hospital referral region (HRR, representing regional healthcare markets). We compared the proportion of disease recorded as locally advanced or of regional risk group (cT3, cT4, and cN1) which would plausibly have been detected by prostate MRI. We conducted adjusted multivariable analysis and performed correlation analysis with Spearman rank coefficient at the level of the HRR. Sensitivity analysis for years 2011 to 2015 was conducted. ResultsOf 98,921 men diagnosed, 4.01% had locally advanced or regional disease. The median prostate MRI utilization rate was 4.58% (IQR [3.03%, 8.12%]). Adjusted multivariable analysis revealed no statistically significant correlation between MRI utilization and proportion of advanced prostate cancer (aOR = 1.01, 95% CI, [0.99,1.03]) in each region. The correlation between MRI usage and advanced diagnosis was not significant (Spearman Ρ = 0.09, P = 0.4). Sensitivity analysis conducted between 2011 and 2015 showed similar results (aOR = 1.008, 95% CI, [0.989, 1.027]; Spearman Ρ = 0.16, P = 0.1). ConclusionsDuring our study period, HRR-level utilization of MRI was not associated with higher incidences of advanced prostate cancer. This suggests the rising advanced prostate cancer diagnoses observed in this period are unlikely an artifact of greater sensitivity of modern imaging tests, but potentially due to other factors such as changes in screening or risk factors. With increased utilization and evolving techniques in recent years, the association between MRI and advanced prostate cancer detection warrants continued monitoring.

Geographic variation in shortfalls of dementia specialists in the United States.

Dementia specialists-neurologists, geriatricians, and geriatric psychiatrists-serve a critical clinical function in diagnosing early-stage Alzheimer's disease and determining eligibility for treatment with disease-modifying therapies. However, the availability of dementia specialists is limited and varies across the United States. Using data from the Area Health Resources Files, we found that the median density of dementia specialists across hospital referral regions in United States is 28.8 per 100 000 population aged 65 years and older (interquartile range 19.3-43.6). We derived thresholds of 33-45 dementia specialists per 100 000 population aged 65 years and older as the provider density necessary to care for older adults with mild cognitive impairment and dementia. Based on these thresholds, we estimated that 34%-59% of the population aged 65 years and older resided in areas with potential dementia specialist shortfalls. The extent of potential shortfalls varied by state and rurality. A better understanding of potential gaps in the availability of dementia specialists will inform policies and practices to ensure access to services for people with cognitive impairment and dementia.

Association between tissue loss type and amputation risk among Medicare patients with concomitant diabetes and peripheral arterial disease

ObjectivePrior studies have described risk factors associated with amputation in patients with concomitant diabetes and peripheral arterial disease (DM/PAD). However, the association between the severity and extent of tissue loss type and amputation risk remains less well-described. We aimed to quantify the role of different tissue loss types in amputation risk among patients with DM/PAD, in the context of demographic, preventive, and socioeconomic factors. MethodsApplying International Classification of Diseases (ICD)-9 and ICD-10 codes to Medicare claims data (2007-2019), we identified all patients with continuous fee-for-service Medicare coverage diagnosed with DM/PAD. Eight tissue loss categories were established using ICD-9 and ICD-10 diagnosis codes, ranging from lymphadenitis (least severe) to gangrene (most severe). We created a Cox proportional hazards model to quantify associations between tissue loss type and 1- and 5-year amputation risk, adjusting for age, race/ethnicity, sex, rurality, income, comorbidities, and preventive factors. Regional variation in DM/PAD rates and risk-adjusted amputation rates was examined at the hospital referral region level. ResultsWe identified 12,257,174 patients with DM/PAD (48% male, 76% White, 10% prior myocardial infarction, 30% chronic kidney disease). Although 2.2 million patients (18%) had some form of tissue loss, 10.0 million patients (82%) did not. The 1-year crude amputation rate (major and minor) was 6.4% in patients with tissue loss, and 0.4% in patients without tissue loss. Among patients with tissue loss, the 1-year any amputation rate varied from 0.89% for patients with lymphadenitis to 26% for patients with gangrene. The 1-year amputation risk varied from two-fold for patients with lymphadenitis (adjusted hazard ratio, 1.96; 95% confidence interval, 1.43-2.69) to 29-fold for patients with gangrene (adjusted hazard ratio, 28.7; 95% confidence interval, 28.1-29.3), compared with patients without tissue loss. No other demographic variable including age, sex, race, or region incurred a hazard ratio for 1- or 5-year amputation risk higher than the least severe tissue loss category. Results were similar across minor and major amputation, and 1- and 5-year amputation outcomes. At a regional level, higher DM/PAD rates were inversely correlated with risk-adjusted 5-year amputation rates (R2 = 0.43). ConclusionsAmong 12 million patients with DM/PAD, the most significant predictor of amputation was the presence and extent of tissue loss, with an association greater in effect size than any other factor studied. Tissue loss could be used in awareness campaigns as a simple marker of high-risk patients. Patients with any type of tissue loss require expedited wound care, revascularization as appropriate, and infection management to avoid amputation. Establishing systems of care to provide these interventions in regions with high amputation rates may prove beneficial for these populations.

The oncology care model and initiation of systemic therapy for cancer.

11031 Background: The CMS Oncology Care Model (OCM) was an episode payment model for patients with cancer receiving systemic therapy (chemotherapy, targeted therapy, immunotherapy, or hormonal therapy) that began in July 2016. Voluntarily participating practices received monthly care coordination payments of $160 per patient per month and were required to deliver enhanced services; they were eligible to share in savings if they achieved quality and spending targets (based on historic spending trended forward; all practices were in one-sided risk arrangements before 2019). Prior research identified savings of $499 per episode (excluding monthly payments) through OCM’s first 5 years, but no overall savings after including incentive payments. One concern about an episode payment model triggered by initiation of systemic therapy is that the financial incentives may prompt an increase in the total number of episodes. We assessed if OCM led to an increase in the likelihood of initiating systemic therapy. Methods: Using Medicare data, we studied care for beneficiaries enrolled in Parts A, B, and D of fee-for-service Medicare with index cancer diagnoses in 2010 through 2019. We assessed initiation of systemic therapy in the one year after the index diagnosis date. We studied two populations: (1) all patients with an index cancer diagnosis and no cancer diagnosis in the preceding two years, suggesting newly diagnosed or newly progressive cancers (incident cohort) and (2) patients with poor prognosis cancers. We used a difference-in-differences (DiD) analysis to assess systemic therapy initiation among patients with index cancer diagnoses who had office visits to medical oncology practices that were participating in OCM, compared with matched comparison practices (on number of patients attributed, number of physicians, number of medical oncologists, and academic affiliation, favoring matches within Hospital Referral Region), before and after OCM’s start in July 2016. Analyses adjusted for patient demographic and clinical variables. Results: Among 742,699 beneficiaries in the incident cancer cohort, 61.9% initiated systemic therapy within 1 year of their index diagnosis. Among 777,951 beneficiaries in the poor prognosis cohort, 58.2% initiated systemic therapy within 1 year of their index diagnosis date. OCM was not associated with the likelihood of initiating systemic therapy in the incident cohort (DID:-0.7 percentage point change, 95% CI:-1.9,0.4, P=0.19) or the poor prognosis cohort (DID:-1.0 percentage point change, 95% CI:-2.1,0.1, P=0.07). Conclusions: Despite financial incentives of episode payment models that may favor greater use of systemic therapy for patients with cancer, OCM did not increase the likelihood of initiating systemic therapy episodes among patients with newly diagnosed/newly progressive cancers or poor prognosis cancers who visited a medical oncology practice.

Acquisitions of safety-net hospitals from 2016-2021: a case series.

Safety-net hospitals have recently become targets of acquisition by health systems with the stated purpose of improving their financial solvency and preserving access to safety-net services. Whether acquisition achieves these goals is unknown. In this descriptive case series, we sought to determine the factors that contribute to safety-net hospital acquisition, and identify whether safety-net services are preserved after acquisition. We examined 22 acquisitions of safety-net hospitals from 2016 to 2021 and described characteristics of the acquired safety-net hospitals, their acquiring systems, and the operational fate of acquired hospitals. Relative to other hospitals in the same Hospital Referral Region in the year prior to acquisition, acquired safety-net hospitals tended to be smaller and have lower occupancy rates. Acquiring systems were geographically concentrated, with only 6 of 20 systems operating in more than 1 state. Safety-net hospitals frequently offered typical safety-net services prior to acquisition. However, after acquisition, 2 of the 22 acquired safety-net hospitals lost safety-net services, 3 hospitals ceased inpatient services, and 1 hospital closed entirely. These findings suggest that acquisition of safety-net hospitals may be associated with trade-offs related to the provision of safety-net services for the communities that stand to benefit from them most.

Geographic Variation of Antidementia and Antipsychotic Medication Use Among US Nursing Home Residents With Dementia.

Several antidementia medications have been approved for symptomatic treatment of cognitive and functional impairment due to Alzheimer disease. Antipsychotics are often prescribed off-label for behavioral symptoms. The aim of this study was to describe the basis for regional variation in antidementia and antipsychotic medication use. US nursing homes (n=9735), hospital referral regions (HRR; n=289). Long-stay residents with dementia (n=273,004). Using 2018 Minimum Data Set 3.0 linked to Medicare data, facility information, and Dartmouth Atlas files, we calculated prevalence of use and separate multilevel logistic models [outcomes: memantine, cholinesterase inhibitor (ChEI), antipsychotic use] estimated adjusted odds ratios (aOR) and 95% CIs for resident, facility, and HRR characteristics. We then fit a series of cross-classified multilevel logistic models to estimate the proportional change in cluster variance (PCV). Overall, 20.9% used antipsychotics, 16.1% used memantine, and 23.3% used ChEIs. For antipsychotics, facility factors [eg, use of physical restraints (aOR: 1.08; 95% CI: 1.05-1.11) or poor staffing ratings (aOR: 1.10; 95% CI: 1.06-1.14)] were associated with more antipsychotic use. Nursing homes in HRRs with the highest health care utilization had greater antidementia drug use (aOR memantine: 1.68; 95% CI: 1.44-1.96). Resident/facility factors accounted for much regional variation in antipsychotics (PCV STATE : 27.80%; PCV HRR : 39.54%). For antidementia medications, HRR-level factors accounted for most regional variation (memantine PCV STATE : 37.44%; ChEI PCV STATE : 39.02%). Regional variations exist in antipsychotic and antidementia medication use among nursing home residents with dementia suggesting the need for evidence-based protocols to guide the use of these medications.

Nurse Practitioner Care, Scope of Practice, and End-of-Life Outcomes for Nursing Home Residents With Dementia

Nursing home residents with Alzheimer disease and related dementias (ADRD) often receive burdensome care at the end of life. Nurse practitioners (NPs) provide an increasing share of primary care in nursing homes, but how NP care is associated with end-of-life outcomes for this population is unknown. To examine the association of NP care with end-of-life outcomes for nursing home residents with ADRD and assess whether these associations differ according to state-level NP scope of practice regulations. This cohort study using fee-for-service Medicare claims included 334 618 US nursing home residents with ADRD who died between January 1, 2016, and December 31, 2018. Data were analyzed from April 6, 2015, to December 31, 2018. Share of nursing home primary care visits by NPs, classified as minimal (<10% of visits), moderate (10%-50% of visits), and extensive (>50% of visits). State NP scope of practice regulations were classified as full vs restrictive in 2 domains: practice authority (authorization to practice and prescribe independently) and do-not-resuscitate (DNR) authority (authorization to sign DNR orders). Hospitalization within the last 30 days of life and death with hospice. Linear probability models with hospital referral region fixed effects controlling for resident characteristics, visit volume, and geographic factors were used to estimate whether the associations between NP care and outcomes varied across states with different scope of practice regulations. Among 334 618 nursing home decedents (mean [SD] age at death, 86.6 [8.2] years; 69.3% female), 40.5% received minimal NP care, 21.4% received moderate NP care, and 38.0% received extensive NP care. Adjusted hospitalization rates were lower for residents with extensive NP care (31.6% [95% CI, 31.4%-31.9%]) vs minimal NP care (32.3% [95% CI, 32.1%-32.6%]), whereas adjusted hospice rates were higher for residents with extensive (55.6% [95% CI, 55.3%-55.9%]) vs minimal (53.6% [95% CI, 53.3%-53.8%]) NP care. However, there was significant variation by state scope of practice. For example, in full practice authority states, adjusted hospice rates were 2.88 percentage points higher (95% CI, 1.99-3.77; P < .001) for residents with extensive vs minimal NP care, but the difference between these same groups was 1.77 percentage points (95% CI, 1.32-2.23; P < .001) in restricted practice states. Hospitalization rates were 1.76 percentage points lower (95% CI, -2.52 to -1.00; P < .001) for decedents with extensive vs minimal NP care in full practice authority states, but the difference between these same groups in restricted practice states was only 0.43 percentage points (95% CI, -0.84 to -0.01; P < .04). Similar patterns were observed in analyses focused on DNR authority. The findings of this cohort study suggest that NPs appear to be important care providers during the end-of-life period for many nursing home residents with ADRD and that regulations governing NP scope of practice may have implications for end-of-life hospitalizations and hospice use in this population.

Performance of health care service area definitions for capturing variation in inpatient care and social determinants of health.

To quantify the degree to which health care service area (HCSA) definitions captured hospitalizations and heterogeneity in social determinants of health (SDOH). Geospatial data from the Centers for Medicare and Medicaid Services, the Census Bureau, and the Dartmouth Institute. Drive-time isochrones from MapBox. Area Deprivation Index (ADI) data. 2017 inpatient discharge data from Arizona, Florida, Iowa, Maryland, Nebraska, New Jersey, New York, and Wisconsin, State Emergency Department Databases and State Inpatient Databases, Healthcare Cost and Utilization Project, Agency for Healthcare Research and Quality; and Fee-For-Service Medicare data in 48 states. Cross-sectional, descriptive analysis. The capture rate was the percentage of inpatient discharges occurring in the same HCSA as the hospital. We compared capture rates for each HCSA definition for different populations and by hospital type. We measured SDOH heterogeneity using the coefficient of variation of the ADI among ZIP codes within each HCSA. HCSA definitions captured a wide range of inpatient discharges, ranging from 20% to 50% for Public Use Microdata Areas (PUMAs) to 93%-97% for Metropolitan Statistical Areas (MSAs). Three-quarters of inpatient discharges were from facilities within the same county as the patient's residential ZIP code, while nearly two-thirds were within the same Hospital Service Area. From the hospital perspective, 74.7% of inpatient discharges originated from within a 30-min drive and 90.1% within a 60-min drive. Capture rates were the lowest for teaching hospitals. PUMAs and drive-time-based HCSAs encompassed more homogenous populations while MSAs, Commuting Zones, and Hospital Referral Regions captured the most variation. The proportion of hospital discharges captured by each HCSA varied, with MSAs capturing the highest proportion of discharges and PUMAs capturing the lowest. Additionally, researchers face a trade-off between capture rate and population homogeneity when deciding which HCSA to use.

Is Electronic Information Exchange Associated With Lower 30-Day Readmission Charges Among Medicare Beneficiaries?

Fragmented readmissions, when admission and readmission occur at different hospitals, are associated with increased charges compared with nonfragmented readmissions. We assessed if hospital participation in health information exchange (HIE) was associated with differences in total charges in fragmented readmissions. Medicare Fee-for-Service Data, 2018. We used generalized linear models with hospital referral region and readmission month fixed effects to assess relationships between information sharing (same HIE, different HIEs, and no HIE available) and total charges of 30-day readmissions among fragmented readmissions; analyses were adjusted for patient-level clinical/demographic characteristics and hospital-level characteristics. We included beneficiaries with a hospitalization for acute myocardial infarction, congestive heart failure, chronic obstructive pulmonary disease, syncope, urinary tract infection, dehydration, or behavioral issues with a 30-day readmission for any reason. In all, 279,729 admission-readmission pairs were included, 27% of which were fragmented (n=75,438); average charges of fragmented readmissions were $64,897-$71,606. Compared with fragmented readmissions where no HIE was available, the average marginal effects of same-HIE and different-HIE admission-readmission pairs were -$2329.55 (95% CI: -7333.73, 2674.62) and -$3905.20 (95% CI: -7592.85, -307.54), respectively. While the average marginal effects of different-HIE pairs were lower than those for no-HIE fragmented readmissions, the average marginal effects of same-HIE and different-HIE pairs were not significantly different from each other. There were no statistical differences in charges between fragmented readmissions to hospitals that share an HIE or that do not share an HIE compared with hospitals with no HIE available.

Measuring Local-Area Racial Segregation for Medicare Hospital Admissions

Considerable racial segregation exists in US hospitals that cannot be explained by where patients live. Approaches to measuring such segregation are limited. To measure how and where sorting of older Black patients to different hospitals occurs within the same health care market. This retrospective cross-sectional study used 2019 Medicare claims data linked to geographic data. Hospital zip code markets were based on driving time. The local hospital segregation (LHS) index was defined as the difference between the racial composition of a hospital's admissions and the racial composition of the hospital's market. Assessed admissions were among US Medicare fee-for-service enrollees aged 65 or older living in the 48 contiguous states with at least 1 hospitalization in 2019 at a hospital with at least 200 hospitalizations. Data were analyzed from November 2022 to January 2024. Degree of residential segregation, ownership status, region, teaching hospital designation, and disproportionate share hospital status. The LHS index by hospital and a regional LHS index by hospital referral region. In the sample of 1991 acute care hospitals, 4 870 252 patients (mean [SD] age, 77.7 [8.3] years; 2 822 006 [56.0%] female) were treated, including 11 435 American Indian or Alaska Native patients (0.2%), 129 376 Asian patients (2.6%), 597 564 Black patients (11.9%), 395 397 Hispanic patients (7.8), and 3 818 371 White patients (75.8%). In the sample, half of hospitalizations among Black patients occurred at 235 hospitals (11.8% of all hospitals); 878 hospitals (34.4%) exhibited a negative LHS score (ie, admitted fewer Black patients relative to their market area) while 1113 hospitals (45.0%) exhibited a positive LHS (ie, admitted more Black patients relative to their market area); of all hospitals, 79.4% exhibited racial admission patterns significantly different from their market. Hospital-level LHS was positively associated with government hospital status (coefficient, 0.24; 95% CI, 0.10 to 0.38), while New York, New York; Chicago, Illinois; and Detroit, Michigan, hospital referral regions exhibited the highest regional LHS measures, with hospital referral region LHS scores of 0.12, 0.16, and 0.21, respectively. In this cross-sectional study, a novel measure of LHS was developed to quantify the extent to which hospitals were admitting a representative proportion of Black patients relative to their market areas. A better understanding of hospital choice within neighborhoods would help to reduce racial inequities in health outcomes.

Abstract 4846: Geographic and physician-level variation in the use of hypofractionated radiotherapy for breast cancer in the U.S.: A cross-classified multilevel analysis

Abstract Hypofractionated whole-breast irradiation (HF-WBI), a radiotherapy with a larger dose per fraction and a shorter schedule for early-stage breast cancer patients post breast-conserving surgery, shows equivalent safety and efficacy compared to conventionally fractionated whole-breast irradiation (CF-WBI) regarding local recurrence and disease-free survival. While several countries have set HF-WBI as the preferred standard of care for these patients, its uptake in the U.S. has lagged behind expectations. We used the Health Care Cost Institute (HCCI) database 2008-2017, covering a third of the U.S. employer-sponsored insurance population, to identify early-stage breast cancer patients. Patients were clustered at different geographic levels (Census region, state, core-based statistical area (CBSA), hospital referral region of Dartmouth Atlas, and zip code) and the radiation oncologist level. These two levels are cross-nested. We used a Bayesian cross-classified multilevel logistic model to simultaneously model the geographic units variability and physician-level variability in HF-WBI use across the U.S. Variation metrics included intracluster correlation coefficient (ICC) and median odds ratios (MOR). ICC with a value of 0 is equivalent to MOR of 1, indicating identical outcomes across all levels. After accounting for these variations in the cross-classified multilevel model framework, we evaluated the association between HF-WBI use and clinical and demographic factors.Our study included 79,747 women (74.0%) who underwent CF-WBI and 27,999 (26.0%) who received HF-WBI. HF-WBI adoption increased over time (2008-2017). Variability across radiation oncologists (ICC = 0.272, MOR = 2.88) was notably larger than geographic areas variability. Variability across CBSA (ICC = 0.097, MOR = 1.76) was the strongest among all the geographic levels. The variation in HF-WBI use was mainly attributed to physician-level variability (ICC=0.222, MOR=2.59) based on the cross-classified multilevel model. No substantial differences were found between young (aged 64 or younger with commercial health plans) and older patients (aged 65 or older with Medicare Advantage plans) in terms of both geographic and physician-level variations. After accounting for variability in radiation oncologists and CBSAs, older age, non-reception of chemotherapy, and several community-level factors, including longer distance from home to facility and higher community education level, were associated with a higher likelihood of HF-WBI use. These results may lend insights into potential facilitators/barriers to HF-WBI uptake. Future studies can further investigate the impact of the 2018 American Society for Radiation Oncology (ASTRO) guideline and the Covid-19 pandemic on HF-WBI use, which would illuminate potential improvements of healthcare practices. Citation Format: Yijia Sun, Loren Saulsberry, Chuanhong Liao, Donald Hedeker, Dezheng Huo. Geographic and physician-level variation in the use of hypofractionated radiotherapy for breast cancer in the U.S.: A cross-classified multilevel analysis [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 4846.

Disparities in Access to Multidisciplinary Cancer Consultations and Treatment for Patients With Early-Stage Non-Small Cell Lung Cancer: A SEER-Medicare Analysis

PurposeDisparities in access to multidisciplinary cancer consultations (MDCc) persist, and the role of physician relationships remains understudied. This study examined the extent to which multilevel factors, including patient characteristics and patient-sharing network measures reflecting the structure of physician relationships, are associated with MDCc and receipt of stereotactic body radiation therapy (SBRT) versus surgery among early-stage non-small cell lung cancer (NSCLC) patients. Materials and MethodsIn this cross-sectional study, we analyzed Surveillance, Epidemiology, and End-Result (SEER)-Medicare data for patients diagnosed with stage I-IIA NSCLC in 2016-2017. We assembled patient-sharing networks and identified cancer specialists who were locally unique for their specialty, herein referred to as “linchpins”. The proportion of linchpin cancer specialists for each hospital referral region (HRR) was calculated as a network-based measure of specialist scarcity. We used multilevel multinomial logistic regression to estimate associations between study variables and the receipt of MDCc and multilevel logistic regression to examine the relationship between MDCc and patient's first treatment. ResultsOur study included 6,120 patients with stage I-IIA NSCLC, of which 751 (12.3%) received MDCc, 1,729 (28.3%) only consulted a radiation oncologist, 2,010 (32.8%) only consulted a surgeon, and 1,630 (26.6%) had no consultations with either specialist within two months following diagnosis. Compared with patients residing in an HRR with a low proportion of linchpin surgeons, those in an HRR with a high proportion of linchpin surgeons had a 2.99 (95% CI: 1.87-4.78) greater relative risk of exclusively consulting a radiation oncologist (vs. MDCc) and a 2.70 (95% CI: 1.68-4.35) greater relative risk consulting neither specialist (vs. MDCc). Patients who received MDCc were 5.32 (95% CI: 4.27-6.63) times more likely to receive SBRT (vs. surgery). ConclusionsPhysician networks are associated with receipt of MDCc and treatment, underscoring the potential for leveraging patient-sharing network analysis to improve access to lung cancer care.

Obstetric imaging practice characteristics associated with prenatal detection of critical congenital heart disease in a rural US region over 20years.

To identify clinical practice characteristics associated with the frequency of prenatal critical congenital heart disease (CCHD) detection (i.e., the number of liveborn infants with postnatally confirmed CCHD identified on prenatal sonography) over 20years in a rural setting comprised of 11 primarily low-volume obstetric hospitals and the single tertiary academic hospital to which they refer. This was a retrospective cohort study of all patients in the referral region with an initial prenatal and/or postnatal diagnosis of CCHD from 01/01/2002 to 12/31/2021. The frequency of prenatal CCHD detection at the time of an obstetric ultrasound was reported, as was the change in detection over time. Critical congenital heart disease detection was assessed as a function of cardiac lesion type, practice setting, and practice characteristics. There were 271 cases with a confirmed postnatal CCHD diagnosis, of which 49% were identified prenatally. The majority of community practices each averaged <10 CCHD cases in total over the study period. Prenatal detection at the tertiary academic hospital's obstetric ultrasound unit was 64%, compared to 22% at the combined referring community practices (p<0.001), though CCHD detection improved over time in both settings. Professional accreditation by the American Institute of Ultrasound in Medicine, image interpretation by radiology or Maternal Fetal Medicine, and use of video clips of ventricular outflow tracts were associated with improved prenatal CCHD detection. Our data demonstrate the infrequency of CCHD cases at small-volume, rural hospitals and the substantial variation in prenatal CCHD detection across practice settings. Our methods allowed for the identification of practice characteristics associated with prenatal CCHD detection.