Bian Liu

Introduction: Home-based medical care (HBMC), longitudinal medical care provided in a patient's home led by physicians or advanced practice providers, addresses complex medical and social needs of seriously ill older adults, particularly those with dementia. We examined whether HBMC use is associated with hospice enrollment and length of stay and whether this varies by dementia status. Methods: We analyzed data from 100% fee-for-service Medicare decedents in 2019, who were eligible for HBMC prior to death. Models used inverse probability weighting, adjusting for clinical, demographic, and regional characteristics. Results: HBMC users had greater odds of hospice use (odds ratio: 1.34; 95% CI: 1.33-1.35) compared with nonusers, regardless of dementia status. For hospice enrollees, HBMC use minimized both ultra-short (180 days) stays, with a stronger reduction in short stays for patients with dementia. Conclusion: As more older adults age in the community and hospice services expand to support patients with complex illnesses such as dementia, HBMC may offer a valuable strategy for effectively navigating prognosis uncertainty and the complexities of hospice eligibility requirements and enrollment.

Current evidence on the impact of post-cancer-diagnosis cigarette smoking cessation on survival is largely based on small-scale studies with limited cancer types. This retrospective cohort study included 8,120 adult cancer survivors from the 2000-2018 National Health Interview Survey Public-use Linked Mortality Files with death information through December 31, 2019. We performed propensity score weighting (PSW) and survival analysis to assess the association of smoking cessation within 3 years after a cancer diagnosis with all-cause mortality among adult cancer survivors of all cancer types, cancers with high relative survival rates, and four major cancers (lung, breast, prostate, and colorectal). Approximately 19.6% of cancer survivors who were cigarette smokers quit smoking within 3 years after their cancer diagnosis. Compared to those who did not quit, cancer survivors who quit had an additional median survival time (MST) of 4 (95% CI 3.1-4.9) years and a relatively lower risk of death (marginal hazard ratio (MHR) 0.86; 95% CI 0.76-0.98). A more pronounced survival benefit was found for survivors of cancers with a high 5-year relative survival rate (> 74%), where the quit group had a longer MST of 7 (95% CI 6.7-7.3) years and an MHR of 0.79 (95% CI 0.66-0.94). Except for colorectal cancer, significant survival benefits were also found for survivors of lung, prostate, and breast cancers. By integrating PSW with linked nationally representative data spanning nearly two decades, we demonstrated that smoking cessation within 3 years after a cancer diagnosis was associated with a significant survival advantage.

Background: People living with dementia frequently use Medicare skilled home health care and have unique usage patterns as compared to people without dementia, but little is known about variation in measured quality of home health care received by this population. Methods: Using 2021 Medicare Fee‐for‐Service Claims data, we examined receipt of high‐quality home health (i.e., care from an agency with a star rating > 3.5) as determined by two publicly available measures: the Quality of Patient Care Star Rating (based on standardized clinical status measures) and the Patient Survey Star Rating (based on satisfaction with care reported by patients or caregivers). For each quality measure, we mapped the county‐level high‐quality‐home health agency utilization rate among people living with dementia and compared differences in utilization of high‐quality home health agencies by dementia status. Results: We found significant county‐level variability in utilization of high‐quality home health. When quality was operationalized based on clinical status measures (i.e., Quality of Patient Care Star Rating), dementia patients did not receive care from lower quality agencies. However, when quality was operationalized based on satisfaction with care (i.e., Patient Survey Star Ratings), people living with dementia were less likely than those without dementia to receive care from high‐quality home health agencies. Conclusions: These findings highlight variability in receipt of high‐quality home health care among people living with dementia nationally and suggest a need for further investigation as to what constitutes high‐quality home health care in this population. To ensure home health meets the unique care needs of people living with dementia, policy makers should work to ensure quality measures are better aligned with the needs of people living with dementia, incentivize access to high‐quality home health care where services are limited, and promote systems to improve family caregiver identification and engagement with home health care.

Introduction: Early-onset type 2 diabetes (EOD), defined as type 2 diabetes diagnosed before age 45, is a growing public health problem. Adolescence is a critical period for prevention interventions because related health behaviors and biological processes are established during this time. Early screening for type 2 diabetes risk including prediabetes (preDM) could identify youth most in need of intervention but unlike for adults, no questionnaire-based youth risk screeners exist. In this study, we assessed the ability of adapted American Diabetes Association (ADA) screening guidelines in adolescence to predict EOD and prediabetes (preDM/EOD) in adulthood. Methods: We used data from the National Longitudinal Study of Adolescent to Adult Health (Add Health). We applied ADA guidelines (modified to align with available data) in Waves I and II, when participants were 12-19, and assessed ability to predict preDM/EOD in Waves IV and V, when participants were 24-43. We then calculated screening performance measures with (N=13,530) and without (N=14,540) accounting for the survey design. Results: In weighted analyses, 40% of participants (5,383) had preDM/EOD based on biomarkers at Waves IV-V. Of these participants, 1,272 were considered at risk according to the ADA screening criteria in Waves I-II (sensitivity=23.6%). Of 8,147 participants without preDM/EOD, 7,218 were not considered at risk in Wave I-II (specificity=88.6%). The screening guidelines had a positive predictive value (PPV) of 57.8% and a negative predictive value (NPV) of 63.7%. The F+ and F- measures were 33.5% and 74.1%, respectively. Unweighted analyses produced similar results. Conclusions: Adapted ADA youth screening guidelines demonstrated poor ability to predict preDM/EOD before age 45. Although specificity was high, sensitivity was low: 76% of participants who developed preDM/EOD would not have been identified as at risk in adolescence using the adapted ADA guidelines. Given the high prevalence of preDM/EOD, there is a need for better screening tools to identify youth at risk and better target diabetes prevention interventions.

Identifying hospitals that disproportionately serve minority and publicly insured patients is important because patients at these hospitals often experience worse outcomes. Studies commonly identify disproportion by using the top decile of hospitals with the greatest proportion of Black discharges nationally. Our study aimed to identify a broader measure that accounts for disproportion by multiple characteristics. Using fee-for-service Medicare data, we classified hospitals as either serving disproportionately or not, examined overlaps in classification, and assessed differences in hospital quality. We found that using a combined measure for any hospitals in the top decile or above a threshold of twice their local healthcare market average of Black, Hispanic, minority, or dual-eligible discharges classified 28.1% (n = 680/2420) of hospitals as serving disproportionately, compared to only 10% (n = 242/2420) when using the top decile of a single characteristic. The combined measure detected moderate differences in hospital star quality ratings (mean difference of 0.57–0.87, all p-values < 0.001; standardized mean difference: 0.50–0.79, 95% CIs all above 0). The combined measure identified hospitals that were smaller, more rural, and served other minorities, namely, Asian and American Indian populations. Future work should consider using this combined measure to more comprehensively identify hospitals that disproportionately serve minority or publicly insured patients.

Purpose We aimed to identify geographical areas of vulnerability, namely co-occurring heightened cancer prevalence and exposure to natural hazards. Methods Cancer prevalence data for four major cancers (lung, breast, colorectal, and prostate) from the Centers for Medicare and Medicaid Services were used along with National Risk Index (NRI) for 18 hazards, including hurricane and wildfire, from the Federal Emergency Management Agency. We examined county-level spatial correlations between cancer prevalence and NRI values using Lee’s L statistics. Results NRI values and cancer prevalence were positively correlated with substantial variations (global Lee’s L = 0.24, p < 0.05; local Lee’s L IQR −0.07 to 0.38). Out of 3106 counties in the contiguous United States, 455 (14.6%) had statistically significant spatial correlations between NRI values and cancer prevalence, of which 194 (43%) were hotspot counties with high NRI values correlated with high cancer prevalence. These hotspot counties were located mostly along coastlines, specifically the Atlantic and Gulf of Mexico with some pockets in the Midwest, primarily in urban areas (151, 77.8%), and within the catchment areas of National Cancer Institute-designated cancer centers (173, 89.2%). They also differed in the types of natural hazard and cancer, as well as community resilience and social vulnerability. Conclusions We identified several geographical areas in the United States with increased need, which may serve as priority areas for future research around the impacts of environmental exposures on cancer continuum. On a public health level, they also may guide prioritization efforts for environmental hazard planning and preparation.

Propensity score (PS) methods, including inverse probability of treatment weighting (IPTW) analysis, are increasingly applied to complex survey data in geriatric studies to infer causal effects. However, the comparative effectiveness of various PS estimation methods, particularly novel machine learning algorithms, has not been thoroughly explored when complex survey data are involved. We conducted a comprehensive simulation study to compare the following six PS estimation methods in IPTW analysis: Logistic Regression, Covariate Balancing Propensity Score, Generalized Boosted Model, Classification and Regression Tree, Random Forest (RF), and Super Learner. We considered 12 scenarios with varying treatment effects, degrees of non-linearity and non-additivity in the associations between covariates and the exposure, and levels of PS overlap. The performance of these six methods was assessed in terms of mean relative bias, root mean square error, and coverage probability. The results showed a similar performance across all methods when PS overlap was strong. However, RF consistently outperformed the other methods when PS overlap was not strong and under non-additive and non-linear scenarios. The results suggest RF to be a more effective approach for PS estimation than the other proposed methods when applying IPTW analysis to complex survey data for population average treatment effects. The methods were applied to data from the Medicare Beneficiary Current Survey for years 2002–2019 to estimate the impact of hospice use on end-of-life healthcare costs. Findings from the real-world example show that hospice use was significantly associated with reduced end-of-life healthcare costs of Medicare Beneficiaries.

Objective: To examine the extent of segregation between hospitals for Medicare beneficiaries by race, ethnicity, and dual‐eligible status over time.Data Sources and Study Setting: We used Medicare inpatient hospital provider data for fee‐for‐service (FFS) beneficiaries, and the Dartmouth Atlas of Health Care from 2013 to 2021 nationwide, for hospital referral regions (HRRs), and for and hospital service areas (HSAs).Study Design: We conducted time trend analysis with dissimilarity indices (DIs) for Black (DI‐Black), Hispanic (DI‐Hispanic), non‐White (including Black, Hispanic, and other non‐White) (DI‐non‐White), and dual‐eligible (DI‐Dual) beneficiaries. DIs between hospitals were contextualized and correlated with population compositions and residential DIs.Data Collection/Extraction Methods: We included 3177 hospitals with more than 250 Medicare FFS beneficiaries discharged per year. We cross‐linked data on hospital‐level patient race, ethnicity, and dual‐eligible status with geographic data and examined time trends using linear mixed models.Principal Findings: Nationwide DIs ranged from 0.23 to 0.53. HRRs and HSAs generally had low segregation (DI medians: 0.08–0.19, highest among Black, then non‐White, Hispanic, and dual‐eligible beneficiaries). However, some HRRs and HSAs had moderate or high segregation (DI‐Black >0.30 in 19.1% of HRRs and 5.8% of HSAs; DI‐non‐White >0.30 for two HRRs with high American Indian/Alaska Native populations). Time trends indicated small declines in segregation from 2013 to 2021 (0.15%–0.30% per year; all p < 0.001). DI‐Dual correlated moderately with non‐White populations.Conclusions: For Medicare FFS, we observe generally low and slightly declining levels of segregation across HRRs and HSAs, with notable exceptions. Improving race reporting and contextualizing select areas of higher segregation with their hospital and residential population compositions can help frame and understand health inequities. Interpretation of HRR‐level DI may require additional historical, demographic, and spatial context due to its potential to oversimplify, overstate, or obscure segregation. Future work should identify drivers and mitigators of segregation, including sorting patterns among health systems.

[This corrects the article DOI: 10.2196/53330.].