Continuously updated forecasting of SARS-CoV-2 in a regional health system.

OBJECTIVES: To build a model of local hospital utilization resulting from SARS-CoV-2 and to continuously update it with new data. STUDY DESIGN: Retrospective analysis of real performance resulting from a model deployed in a major regional health system. METHODS: Using hospitalization data from the Kaiser Permanente Mid-Atlantic States integrated care system during the period from March 10, 2020, through December 31, 2020, and a custom-developed genetic particle filtering algorithm, we modeled the SARS-CoV-2 outbreak in the mid-Atlantic region. This model produced weekly forecasts of COVID-19-related hospital admissions, which we then compared with actual hospital admissions over the same period. RESULTS: We found that the model was able to accurately capture the data-generating process (weekly mean absolute percentage error, 10.0%-48.8%; Anderson-Darling P value of .97 when comparing percentiles of observed admissions with the uniform distribution) once the effects of social distancing could be accurately measured in mid-April. We also found that our estimates of key parameters, including the reproductive rate, were consistent with consensus literature estimates. CONCLUSIONS: The genetic particle filtering algorithm that we have proposed is effective at modeling hospitalizations due to SARS-CoV-2. The methods used by our model can be reproduced by any major health care system for the purposes of resource planning, staffing, and population care management to create an effective forecasting regimen at scale.

Geographic variation in predictors of ED admission rates in U.S. Medicare fee-for-service beneficiaries.

INTRODUCTION: We study community-level factors associated with emergency department (ED) admission rates and assessed how they vary across geography. METHODS: We conducted an ecological study using 2012 data from 100% of U.S. Medicare Fee-for-Service beneficiaries to calculate county-level ED admission rates, adjusted by Hierarchical Condition Categories to control for patient health. We tested community-level measures related to healthcare market concentration, healthcare delivery, and socioeconomic factors potentially associated with admission rates and assessed whether these factors predicted ED admission rates across counties using ordinary least squares (OLS) regression and whether they varied across geography using geographically weighted regression (GWR). RESULTS: In 3031 U.S. counties, the ED admission rate varied from 3.9% to 82.2%. The lowest ED admission rates were concentrated in counties in Kansas, Oregon, and Vermont and the highest ED admission rates were in counties throughout Washington, Wyoming, Texas, and Colorado. The OLS model found several community-level factors that negatively impacted admission rates, specifically hospital market concentration, the rate of hospital beds with urgent care, and the rate of hospital beds. The factors that had a positive impact on the admission rate include the rate of MDs and factors for disadvantage, affluence, and foreign born/Hispanic. However, GWR showed the relationship between the ED admission rate and predictors varied across U.S. counties CONCLUSIONS: The association between healthcare market concentration, healthcare delivery, and socioeconomic factors with ED admissions differed across communities in Medicare beneficiaries. This suggests that policy and interventions to reduce ED admissions need to be tailored to specific community contexts.

Evaluating the Effects of Pioneer Accountable Care Organizations on Medicare Part D Drug Spending and Utilization.

BACKGROUND: The improvement of medication use is a critical mechanism that accountable care organization (ACO) could use to save overall costs. Currently pharmaceutical spending is not part of the calculation for ACO-shared savings and risks. Thus, ACO providers may have strong incentives to prescribe more medications hoping to avoid expensive downstream medical costs. METHODS: We designed a quasinatural experiment study to evaluate the effects of Pioneer ACOs on Medicare Part D spending and utilization. Medicare fee-for-service beneficiaries with Part D drug coverage who were aligned to a Pioneer ACO were compared with a random 5% sample of non-ACO beneficiaries. Outcomes included changes in Part D spending, number of prescription fills, percent of brand medications, and total Part A and B medical spending. We utilized a generalized linear model with a difference-in-differences approach to estimate 2011-2012 changes in these outcomes among beneficiaries aligned with Pioneer ACOs, adjusting for all beneficiary-level demographics, income and insurance status, clinical characteristics, and regional fixed effects. RESULTS: Being in an ACO did not significantly affect Part D spending (-$23.52; P=0.19), total prescriptions filled (-0.12; P=0.27), and the percent of claims for brand-name drugs (0.06%; P=0.23). The ACO group was associated with savings in Parts A and B spending of $345 (P<0.0001) per person per year. CONCLUSIONS: We found that beneficiaries aligned to Pioneer ACOs were not associated with changes in pharmaceutical spending and use, but were associated with savings in Parts A and B spending in 2012.

County-Level Variation in Emergency Department Admission Rates Among US Medicare Beneficiaries.

STUDY OBJECTIVE: Hospital-based emergency departments (EDs) are the gateway to hospital admissions for many Americans. Approximately half of all US hospital admissions originate from EDs, and more than 3 in 4 are among Medicare beneficiaries. Recent literature has demonstrated nearly 2-fold variation in both physician- and hospital-level ED admission rates. We study geographic variation at the county level in ED admission rates among Medicare fee-for-service beneficiaries. METHODS: Using the 100% population data from the Centers for Medicare & Medicaid Services (CMS), we analyzed beneficiaries continuously enrolled in Medicare fee-for-service Parts A and B who resided in the 50 states and the District of Columbia in 2012. The ED admission rate was aggregated to the county level. ED admission rates were adjusted with the CMS Hierarchical Condition Categories (HCC) risk score. The resulting HCC adjusted ED admission rate was mapped to display the variation by county. RESULTS: The average county HCC adjusted ED admission rate was 30.8% in the Medicare population. Counties in the lowest quintile had an ED admission rate of 19.9% or lower. By comparison, counties in the highest quintile had an ED admission rate of 40.3% or higher. CONCLUSION: Among Medicare beneficiaries, county-level ED admission rates varied 2-fold between counties in the lowest and highest quintiles. Future work should focus on exploring causes for this variation, such as racial ethnic composition, socioeconomic status, and health care delivery system characteristics and the research of effectiveness of policies that affect ED admission decisions.