Wyatt P. Bensken

Objectives: To estimate the prevalence of polypharmacy among older adults with incident epilepsy and to describe the most common combinations of drug classes filled prior to epilepsy diagnosis. Polypharmacy—the concurrent use of multiple medications—is common in older adults with epilepsy, but little is known about its burden and specific patterns at the time of diagnosis. Methods: This retrospective longitudinal cohort study included 2 768 656 US Medicare beneficiaries aged 67 and older with Part D coverage. Beneficiaries were epilepsy‐free from 2016 to 2018; 13 819 (0.5%) were newly diagnosed with epilepsy in 2019. The prevalence of polypharmacy and combinations of therapeutic drug classes filled in 2018 were compared between those with and without incident epilepsy using frequent itemset mining and regression models to estimate adjusted rate ratios. Results: Beneficiaries with incident epilepsy filled prescriptions for a greater number of therapeutic drug classes in the year prior to diagnosis compared with those without epilepsy (median 9 [IQR 5–12] vs. 6 [IQR 3–9]) and were more likely to have filled prescriptions for five or more distinct drug classes (81.4% vs. 63.3%). Antidepressants, gastric medications, and loop diuretics were more prevalent among beneficiaries with incident epilepsy. The combination of calcium channel blockers and antihypertensives was most strongly associated with incident epilepsy after adjustment (adjusted rate ratio 1.54 [99% CI, 1.38–1.71]). Significance: Older adults with incident epilepsy experience a high burden of polypharmacy at the time of diagnosis, with nearly 90% prescribed five or more drug classes. Certain drug class combinations, such as antipsychotics and antidepressants, were more frequently observed among those with incident epilepsy. These findings underscore the importance of considering overall polypharmacy burden and specific drug class combinations when initiating epilepsy therapy in older adults. Plain Language Summary: Older adults who develop epilepsy often take many medications for other health conditions. In this study of US Medicare beneficiaries, people who were later diagnosed with epilepsy were taking about nine different types of drug classes in the year before diagnosis, compared with about six drug classes among those who did not develop epilepsy. Certain combinations of drug classes—especially those affecting the brain and nervous system—were more common among people who developed epilepsy. These findings show that many older adults have complex medication regimens when epilepsy is first diagnosed.

Objective: To demonstrate a large-scale EHR data transformation to the Observational Medical Outcomes Partnership (OMOP) Common Data model in the OCHIN network to support AI/ML analyses within the AIM-AHEAD national research consortium. We highlight customized workflows and infrastructure development processes designed to support OCHIN database use by AI/ML researchers of varied academic backgrounds and skillsets. Materials and methods: Automated and manual mappings were used to ingest and transform OCHIN’s i2b2-formatted database to the OMOP Common Data Model. Results: 360 million encounters across 10+ million OCHIN patients were mapped over the course of a single calendar year, with custom concepts created to represent social drivers of health. Discussion: The successful transformation of the OCHIN Research Data Warehouse benefitted from incorporating both customized and legacy data workflows. Conclusion: OCHIN’s efforts will facilitate parallel analyses across AIM-AHEAD datasets and support AI/ML model representativeness of affected populations.

Background: Timely cervical cancer screenings are critical for middle-aged to older Latinas. However, little is currently known about screening in this population. Latinas tend to receive disproportionate care in community health centers (CHC). Methods: A retrospective analysis of electronic health record data from a national network of US CHCs included Latinas and non-Hispanic Whites (NHW) aged >45 with known birth country and at least one in-person primary care visit between 2014 and 2022, across 527 community-based clinics in 15 states. We used generalized estimating equations logistic regression to examine covariate-adjusted receipt of cytology, overall and by country of origin. Results: Among 34,135 patients, Mexican Latinas had the highest unadjusted prevalence of receiving cytology or cotesting at 56%, whereas NHWs had the lowest prevalence at 37%. We found that Latina groups, except for US-born Latinas, individually had higher odds of receiving cytology or cotesting compared with NHWs. Latina groups born in Mexico [OR (95% confidence interval, CI), 1.44 (1.09–1.90)], Honduras [OR (95% CI), 1.39 (1.06–1.82)], El Salvador [OR (95% CI), 1.38 (1.09–1.73)], and Nicaragua [OR (95% CI), 1.36 (1.01–1.83)] had higher odds than NHWs. Conclusions: CHCs have important strengths in reaching diverse patient populations, with higher rates of cervical cancer screening in middle-aged to older Latinas than NHWs. Impact: This study adds novel information about screening rates among a heterogeneous sample outside of their reproductive years and in a large, multistate sample of CHCs. We add to a growing body of literature about the distinctive environment of CHCs.

Despite the well-established importance of health-related social needs in shaping patient outcomes, gaps remain in the literature examining these relationships at the individual level among patients undergoing cardiac surgery. This retrospective study used data from the 2016–2018 Nationwide Readmission Database to evaluate postoperative complications and readmissions in patients undergoing cardiac surgery (coronary artery bypass grafting, aortic surgery, valve surgery, or a combination) using individual-level social vulnerability clinically acknowledged using ICD-10 Z-codes. Six domains of ICD-10 Z-codes (employment, family, housing, psychosocial needs, socioeconomic status, dependence) were considered social vulnerabilities. Data were analyzed using stratification by social vulnerability status and multivariable logistic regression. Among the 846,837 included patients, dependence-related needs were the most documented domain. Patients with social vulnerability at any point were younger, had a longer length of stay, and had a higher prevalence of comorbid conditions, readmissions, and complications. For patients with social vulnerability, the odds ratio of complications was 1.12 (1.03–1.22), and the odds ratio of 90-day readmissions was 1.15 (1.03–1.27). Clinically acknowledged social vulnerability at any point was associated with higher odds of complications or readmissions after cardiac surgery. Z-codes may be useful for identifying nonmedical factors that can affect patient outcomes, but further standardization and assessment are needed.

Background: The Accelerating Data Value Across a National Community Health Center Network (ADVANCE) collaborative is 1 of 8 clinical research networks participating in PCORnet ^® , the National Patient-Centered Clinical Research Network. Providing data, engagement, and regulatory infrastructure, ADVANCE facilitates the inclusion of people who receive primary care in community-based health centers (CHC) across over 90 patient-centered research studies. Centrally coordinated by OCHIN, in partnership with Fenway Health, Health Choice Network, Oregon Health & Science University, and the University of Washington, ADVANCE comprises the nation’s most comprehensive network for health care and outcomes research in community-based primary care and hospital-based settings. Objective: This manuscript discusses ADVANCE’s unique past, present, and future approaches to strengthen infrastructure and capacity to enhance research in community-based primary care settings. Main Arguments: As community-based primary care organizations, CHCs have relationships, trust, and expertise with the communities they serve. ADVANCE provides critical data and engagement resources for including CHCs and their patients in research to improve patient-centered care and outcomes. Despite past investment, there is a recognized need for additional engagement, investment, capacity building, and research infrastructure to realize the full potential of CHC partnerships in clinical research and PCORnet ^® studies. By partnering with CHCs and tertiary hospitals, ADVANCE and the PCORnet infrastructure can provide enhanced access to clinical research opportunities that further support patient-centered, high-quality care delivery, and positive health outcomes while meeting CHCs’ priorities and goals, progressing their own research capacity and infrastructure, and contributing to scalable research readiness models.

Supplemental Table S1 describes the intermediate results and calculations of variable importance measures of the random forest analysis. Supplemental Table S2 provides a correlation matrix of variables that appeared in the results of CART and top 10 predictors from the random forest variable importance plot. Supplemental Figure S1 shows the trend of percentage of LSBC in Massachusetts from 2009 to 2017 to explain that missing a single year (2018) of Massachusetts data is unlikely to substantially impact the study results. Supplemental Figure S2 shows the distribution of percentage of late-stage breast cancer by state or region among SEER counties. Supplemental Figure S3 shows Classification and regression tree analysis in predicting percentage of LSBC. Supplemental Figure S4 shows the percentage of counties in states or regions by phenotypes.

Background:Disparities in the stage at diagnosis for breast cancer have been independently associated with various contextual characteristics. Understanding which combinations of these characteristics indicate highest risk, and where they are located, is critical to targeting interventions and improving outcomes for patients with breast cancer.Methods:The study included women diagnosed with invasive breast cancer between 2009 and 2018 from 680 U.S. counties participating in the Surveillance, Epidemiology, and End Results program. We used a machine learning approach called Classification and Regression Tree (CART) to identify county “phenotypes,” combinations of characteristics that predict the percentage of patients with breast cancer presenting with late-stage disease. We then mapped the phenotypes and compared their geographic distributions. These findings were further validated using an alternate machine learning approach called random forest.Results:We discovered seven phenotypes of late-stage breast cancer. Common to most phenotypes associated with high risk of late-stage diagnosis were high uninsured rate, low mammography use, high area deprivation, rurality, and high poverty. Geographically, these phenotypes were most prevalent in southern and western states, while phenotypes associated with lower percentages of late-stage diagnosis were most prevalent in the northeastern states and select metropolitan areas.Conclusions:The use of machine learning methods of CART and random forest together with geographic methods offers a promising avenue for future disparities research.Impact:Local interventions to reduce late-stage breast cancer diagnosis, such as community education and outreach programs, can use machine learning and geographic modeling approaches to tailor strategies for early detection and resource allocation.