Spiten, Matthew J.

Abstract Background/Objective: Decentralized research has many advantages; however, little is known about the representativeness of a source population in decentralized studies. We recruited participants aged 18-64 years from four states from June to December 2022 for a prospective cohort study to assess viral epidemiology. Our aim was to determine the association between age, gender, race/ethnicity, rurality, and socioeconomic status (SES) on study participation in a decentralized prospective cohort study. Methods: We consented 9,286 participants from 231,099 (4.0%) adults with the mean age of 45.6 years (±12.0). We used an electronic decentralized approach for recruitment. Consented participants were more likely to be non-Hispanic White, female, older, urban residents, have more health conditions, and possessed higher socioeconomic status (SES) compared to those non-consented. Results: We observed an interaction between SES and race-ethnicity on the odds of consent (P = 0.006). Specifically, SES did not affect non-Hispanic white participation rates(OR 1.24 95% CI 1.16 – 1.32] for the highest SES quartile compared to those with the lowest SES quartile) as much as it did participants combined across the other races (OR 1.73; 95% CI 1.45 – 2.98]) Conclusion: The relationship between SES and consent rates might be disproportionately greater in historically disadvantaged groups, compared to non-Hispanic White. It suggests that instead of focusing on enrollment of specific minority groups in research, there is value in future research exploring and addressing the diversity of barriers to trials within minority groups. Our study highlights that decentralized studies need to address social determinants of health, especially in under-resourced populations.

Objectives/Goals: This study aims to evaluate the performance of a third-party artificial intelligence (AI) product in predicting diagnosis-related groups (DRGs) in a community healthcare system. We highlight a use case illustrating how clinical practice leverages AI-predicted information in unexpected yet advantageous ways and assess the AI predictions accuracy and practical application. Methods/Study Population: DRGs are crucial for hospital reimbursement under the prospective payment model. The Mayo Clinic Health System (MCHS), a network of clinics and hospitals serving a substantial rural population in Minnesota and Wisconsin, has recently adopted an AI algorithm developed by Xsolis (an AI-focused healthcare solution provider). This algorithm, a 1D convolutional neural network, predicts DRGs based on clinical documentation. To assess the accuracy of AI-generated DRG predictions for inpatient discharges, we analyzed data from 930 patients hospitalized at MCHS Mankato between March 2 and May 13, 2024. The Xsolis platform provided the top three DRG predictions for the first 48 hours of each inpatient stay. The accuracy of these predictions was then compared against the final billed DRG codes from the hospital’s records. Results/Anticipated Results: In our validation set, Xsolis achieved a top-3 DRG prediction accuracy of 71% at 24 hours and 81% at 48 hours, which is lower than the originally reported accuracy of 81.1% and 83.3%, respectively. Interestingly, discussions with clinical practice leaders revealed that the most valuable information derived from the AI predictions was the expected geometric mean length of stay (GMLOS), which Xsolis was perceived to predict accurately. In the Medicare system, each DRG is associated with an expected GMLOS, a critical factor for efficient hospital flow planning. A subsequent analysis comparing predicted GMLOS with the actual length of stay showed variances of -0.10 days on day 1 and 0.14 days on day 2, indicating a high degree of accuracy and aligning with clinical practice perceptions. Discussion/Significance of Impact: Our research underscores that clinical practice can leverage AI predictions in unexpected yet beneficial ways. While initially focused on DRG prediction, the associated GMLOS emerged as more significant. This suggests that AI algorithm validation should be tailored to specific clinical needs rather than relying solely on generalized benchmarks.

Abstract Background The incidence of respiratory syncytial virus (RSV)–acute respiratory infection (ARI) in community-dwelling adults after the Omicron variant of the COVID-19 pandemic is unknown. Our aim was to assess the incidence of RSV-ARI in adults aged 18 to 64 years over 2 consecutive RSV seasons (October–April 2022–2024) in 4 US states. Methods This community-based prospective cohort study comprised 7501 participants in Minnesota, Wisconsin, Florida, and Arizona. We calculated RSV-ARI and RSV–lower respiratory tract disease (LRTD) incidence and attack rates. We reported unadjusted incidence by age group, gender, race and ethnicity, Charlson Comorbidity Index, socioeconomic status, residential state, and rural/urban setting. Results Seasons 1 and 2 had 2250 and 2377 ARI episodes, respectively, with an RSV-ARI positivity rate of 5.5% for season 1 and 5.8% for season 2 among those tested. In season 1, the overall incidence of RSV-ARI was 27.71 (95% CI, 22.82–33.34) per 1000 person-years (1.49% attack rate). Almost half (49.0%) had RSV-LRTD, with an incidence of 13.53 (95% CI, 10.19–17.61) per 1000 person-years (0.73% attack rate). In season 2, the RSV-ARI and RSV-LRTD incidence rates were 26.39 (95% CI, 21.73–31.75) per 1000 person-years (1.51% attack rate) and 12.43 (95% CI, 9.31–16.26) per 1000 person-years (0.72% attack rate). RSV-ARI incidence peaked in November 2022 and December 2023. Conclusions Our observations suggest that RSV-ARI incidence and seasonal pattern are shifting to prepandemic RSV epidemiology.

Introduction: Using a digital process that leverages electronic health records (EHRs) can ease many of the challenges presented by the traditional enrollment process for clinical trials. We tested if automated batch enrollment using a technology-enabled subject recruitment system (TESRS) enhances recruitment while preserving representation of research subjects for the study population in our study setting. Methods: An ongoing community-based prospective adult cohort study was used to randomize 600 subjects who were eligible by age and residential address to TESRS (n = 300) and standard mailing method (n = 300), respectively, for 3 months. Then, TESRS was initiated and included automatic identification of patients’ preference for being contacted (online patient portal vs postal mail) from EHRs and automatic sending out of invitation letters followed by completion of a short online survey for checking eligibility and the digital consent process if eligible. We compared (1) median time to consent from invitation sent out per subject and total subjects recruited after a 3-month recruitment period, (2) the estimated study staff’s time, and (3) representation of sociodemographic characteristics (e.g., age, sex, race, SES measured by HOUSES index, and rural residence) between subjects recruited via TESRS and those via traditional mailing methods. Results: Median age of randomized subjects (n = 600) was 63 years with 52.0% female and 89.2% non-Hispanic White. Over a 3-month period, results showed consent rate via TESRS was 13% (39/297) similar to 11% (31/295) via standard mailing. However, recruitment was significantly faster with the TESRS approach (median 7 vs 26 days) given the study staff’s effort. Study staff’s time saved by using TESRS compared to standard mailing approach was estimated at 40 min per subject (equivalent to 200 h for 300 subjects). No significant differences in characteristics of research subjects from the study population were found. Conclusion: Our study demonstrated the utility of TESRS as a subject recruitment digital technology which significantly enhanced the recruitment effort while reducing the study staff burden of recruitment while maintaining the consistency of characteristics of recruited subjects. The strategy and support for implementing and testing TESRS in other study settings should be considered.