Background:

Disparities in disease prevalence, treatment and progression of chronic kidney disease (CKD) /Kidney Failure (KF) have been documented. There is, however, lack of published data on access to care measures in CKD/KF.

Methods:

Access to care in adults with CKD (N18)/KF (N19) was examined using pooled, multi-year (2016-2020) Medical Panel Expenditure Survey (MEPS) data. Access was defined as self-reported: coverage, including ability to pay medical bills; usual source of care; site of usual care, and ability to secure appointments. Multivariable logistic regressions estimated associations of CKD/KF with access measures, adjusting for socio-demographic factors.

Results:

The study identified 596 adults with CKD/KF(N­_Weighted=1.1M; 68% white; 49% female). The CKD/KF group was significantly more likely to report: public insurance (84% vs 33%); a usual source of care (91% vs 72%); hospital as usual source(28% vs 20%); ability to secure medical appointment (47% vs 33%), and difficulty in paying medical bills(9.7% vs 4.5%), vs general population (all p<0.05).

Multivariable analyses indicated: those with CKD(OR _CKDvsGen:3.94;95%CI:1.88-6.67) and KF(OR _KFvsGen:2.63;1.21-3.70) were significantly more likely to report a usual source of care (all p<.0.05); however, there was no difference in ability to schedule appointments across the comparison groups (OR­­ _CKDvsGen:1.32;0.26-6.64; OR _KFvsGen:1.20;0.36-3.93) [all p>0.05].

Conclusions:

While the majority of patients with CKD/KF reported a usual source of care, more patients in this group reported difficulties in paying medical bills versus general population. Given their high disease burden and unmet needs, understanding the drivers of healthcare utilization is key to addressing access barriers in the CKD/KF population.

Background: Complications following total hip arthroscopy (THA) and total knee arthroscopy (TKA) result in hospital readmissions and increased financial burden on the patient and the health system^1,2. The Centers for Medicare and Medicaid Services enforce fee penalties to hospitals with higher-than-expected readmissions, prompting Renown Health to use logistic regression and machine learning models to predict patients at higher risk of adverse surgical outcomes. Previous research show that electronic health record data can accurately predict hospital readmissions (AUC=0.85)³. Methods: Patient-level predictors include social determinants of health (SDoH) that are traditionally self-reported and encompass 5 domains (economic, educational, healthcare access, environment, and social) that explain 30-55% of health outcomes^4,5. This research proposes novel indicators of SDoH not previously studied in predicting surgical outcomes and are not subject to patients’ willingness to self-report. Patient no-show history, bad-debt history, and residential address change history quantitatively measure exposure to barriers within one or more of the SDoH domains and may add value to model performance when predicting outcomes following THA/TKA procedures. Results: Preliminary results show that THA/TKA patients with 5+ no-shows have 3.1 higher crude odds of readmission compared to those with 0 no-shows (OR:3.1, 95%CI: 2.61-3.69, p<.001). Additionally, THA/TKA patients with a no-show rate up to 5% have 2.24 higher crude odds of readmission than those with a 0% no-show rate (OR: 2.24, 95%CI: 1.92-2.61, p<.001). Discussion: Preliminary results generated interest to determine if new objective indicators of SDoH in logistic regression and machine learning methods improve model performance for predicting THA/TKA surgical outcomes.

Introduction:

This abstract underscores the significance of Household Pulse Survey data in evaluating COVID-19's impact on the U.S. population. The project aims to develop an R package, utilizing tools such as usethis, Rtools, and R Studio. Methodology involves acquiring CSV files from the Census website, optimizing storage efficiency via Parquet conversion, and sharing datasets on GitHub.

Methodology:

Leveraging usethis and Rtools, we acquire Census CSVs and convert them to Parquet for efficient storage without data loss. Parquet files are transformed into .rda datasets in R Studio, aligning with R’s native data format and ensuring reproducibility.

Results:

The meticulous approach yields an R package housing diverse Household Pulse Survey insights. The optimized pipeline enhances data retrieval for researchers. The finalized package, including 63 datasets, is shared on GitHub, fostering collaboration and transparency.

Discussion:

The workflow's broader implications emphasize the open-source contribution's impact on advancing data science. Efficiency gains from CSV to Parquet conversion optimize storage and data retrieval. Adoption of .rda datasets ensures compatibility and flexibility. The GitHub repository promotes collaboration and continuous improvement in varied data science contexts. Overall, the {hpsr} R Data Package facilitates accessible and comprehensive analysis of COVID-19's impact on the U.S. population.

Introduction: Individuals with diabetes often present with co-occurring chronic conditions (CC), or multimorbidity. We adopted a more expanded definition of multimorbidity that describes Complex multimorbidity (C-MM) as the occurrence/co-occurrence of CC, functional limitations (FL), and geriatric syndromes (GS). We aimed to describe C-MM in midlife and older adults with self-reported diabetes and identify the combinations of conditions constituting C-MM.

Methods: We conducted a cross-sectional study of respondents 50 years or older from the 2018 Health and Retirement Study. Conditions constituting C-MM were defined using self-reported CC, FL, and GS. We conducted Association Rule Mining to identify the most common combinations of conditions constituting C-MM among respondents with diabetes and by age and race.

Results: Of 15,636 respondents, 28.7% reported having diabetes. Nearly 66% of respondents with diabetes reported CC, FL, and GS, whereas only 1.7% had none, 9.8% had one condition, and 22.9% had 2 conditions constituting C-MM. More than half reported at least six conditions, including 22.1% with 10 or more conditions. The most common combinations of conditions included at least one CC, FL, and GS, regardless of stratification by age or race. However, there were no predominant combinations when stratifying by count of conditions constituting C-MM.

Discussion: Our findings indicate that FL and GS are highly prevalent in midlife and older adults with diabetes. This study highlighted the importance of characterizing MM in broader terms rather than limiting its definition to the co-occurrence of CC alone, and the heterogeneity of conditions for midlife and older adults with diabetes.