BACKGROUND:Micro-randomization is a common method used to design and tailor AI-driven digital health interventions. However, applying it in real-world clinical settings can be challenging particularly when there are operational or resource constraints. We propose a novel design that integrates micro-randomization with treatment allocation policies to address such constraints, inspired by a pediatric type 1 diabetes (T1D) program. METHODS:We evaluated the design's properties through an extensive simulation study and developed a simulation-based power calculator, MRThreshold, to support such trial designs. RESULTS:Operational constraints that led to imbalance in treatment assignment affected efficiency. However, increasing resources had less impact relative to increases in study length (i.e., opportunities for micro-randomization). We observed a > 50% increase in power when lengthening a 16-week study to a 40-week study. Using our power calculator, we demonstrated that a 40-week study with 100 patients provides 84.0% power to detect a 2% change in time spent in glucose control, providing design considerations for our study. CONCLUSIONS:Careful consideration of study length, sample size, and operational capacity is essential for thoughtful design. Our novel design and tool balance micro-randomization and treatment allocation under operational constraints.

Objective: To evaluate large language model (LLM) performance on unprocessed electronic medical record (EMR) data for clinical registry abstraction. Methods: We evaluated LLM performance answering registry questions for the American College of Cardiology National Cardiovascular Data Registry (ACC NCDR). In a pilot study at an academic medical center, the model identified candidate data sources for each registry question and experienced abstractors used these results to define question-specific document sets. In a validation study at a second center with a second ACC NCDR registry, the LLM answered questions using the question-specific document sets. Before reviewing any output, two abstractors independently established the ground truth and assigned each question to one of six categories, ordered by the ambiguity and clinical reasoning required to resolve it: Medication/Event Flag, Binary Clinical Presence, Administrative, Quantitative Laboratory/Physiologic, Clinical Interpretation, and Event Timing. Results: The analytical sample comprised 9,430 abstractor answers reconciled to 4,715 consensus answers (501 pilot; 4,214 validation). In the pilot, candidate data sources per question averaged between 14.6 (SD 13.9) for demographics and 89.2 (SD 56.1) for history and risk factors. In validation, human inter-rater agreement was approximately 98% while 87% of LLM answers exactly matched consensus, 2% partially, and 9% did not. Mean question-level accuracy was 91.5% (SD 13.4%) across 157 questions with at least 20 answers, and declined as ambiguity increased, from 96% for Medication/Event Flag to 62% for Event Timing questions. Conclusions: LLMs answering clinical registry questions on unprocessed EMR data achieved far lower accuracy than human abstractors. LLM accuracy fell steadily as ambiguity and the level of required clinical reasoning increased.

BACKGROUND:The impact of telemedicine use on heart failure (HF) care is unknown. We assessed the association between telemedicine use and rates of diagnostic testing and prescribing for patients with HF. METHODS:We extracted electronic health record data for patients with HF at an academic cardiovascular center in Northern California. We assigned patients to the cardiologist providing >50% of their visits from March 2019 to May 2023. This interval was divided into 8 nonoverlapping 6-month periods, excluding March through May of 2020. All patients seen and orders placed in each period were included. Orders of interest included diagnostic tests (Holter monitors, electrocardiograms, echocardiograms, natriuretic peptide tests, chemistry panels) and new prescriptions (beta blockers, aldosterone antagonists, renin angiotensin inhibitors, hydralazine, nitrates, total guideline-directed medical therapy, diuretics). We assessed the association between clinician-period ordering rates and telemedicine use (proportion of visits via video/phone) using negative binomial regression with adjustment for the period, number of patients, and clinician random intercepts. We report incident rate ratios of per-patient ordering associated with a 50-percentage-point increase in telemedicine use. Subgroup analyses were conducted for patients with reduced/preserved ejection fraction. RESULTS:There were 7741 patients seen by 44 clinicians (1941 patients with HF with reduced ejection fraction seen by 28 clinicians). Mean age was 65 years (SD, 16.8) with 41% female. A 50-percentage-point increase in telemedicine use was associated with significantly reduced ordering of electrocardiograms (incident rate ratio, 0.30, [95% CI 0.25-0.35]), echocardiograms (0.70 [0.59-0.82]), natriuretic peptide tests (0.66 [0.49-0.88]), and chemistry panels (0.66 [0.56-0.76]). For the HF with reduced ejection fraction subgroup, this increase in telemedicine use was associated with reduced ordering of aldosterone antagonists (0.72 [0.52-0.99]) and total guideline-directed medical therapy (0.80 [0.69-0.94]). CONCLUSIONS:Greater clinician telemedicine use was associated with decreased diagnostic testing for patients with HF and reduced guideline-directed medical therapy initiation for patients with HF with reduced ejection fraction. Novel interventions are needed to ensure guideline-concordant care regardless of care modality.

Background Telemedicine use has increased, but its impact on access to initial cardiac care remains understudied. We assessed the association of new patient visit modality with geographic reach and wait times at an academic cardiovascular center. Methods We extracted electronic health record data for NPVs from January 2017 to November 2023. We defined the center's traditional catchment area using the 80th percentile of patient–clinic distance for in‐person NPVs before March 2020. For NPVs from July 2021 to November 2023 (study period), we used multivariable regression to assess the association of visit modality with the likelihood of living outside the catchment area and the time from scheduling to initial visit. Results There were 16 407 NPVs (45.4% telemedicine). Average age was 58.6, with 50.4% women, 2.8% Black individuals, 12.7% Hispanic individuals, and 10.3% patients on Medicaid. Patients receiving telemedicine NPVs were more likely to live outside the catchment area (adjusted odds ratio, 1.83 [95% CI, 1.55–2.16]). Subgroup analyses revealed a larger effect for patients aged 65 to 79 years versus 45 to 64 years (adjusted odds ratio, 2.05 versus 1.53; P=0.03). Telemedicine NPVs had shorter times to visit than in‐person NPVs (−8 days [95% CI, −3 to −13]). This effect was larger for patients on Medicaid versus private insurance (−13 versus −6 days; P=0.03). Conclusions New patients seen via telemedicine were more likely to reside outside the center's catchment area and had shorter wait times. Effects were more pronounced for older patients and those on Medicaid. Future studies should evaluate whether strategic implementation of telemedicine can broaden the geographic reach of cardiovascular centers and improve access for new patients.

Clinics continue to adopt remote patient monitoring for type 1 diabetes (T1D) and care models shaped by algorithmic CGM data analysis. No clinic-facing quantitative framework currently exists to track the impact of such algorithm-directed care on patient outcomes and clinical workload. The Teamwork, Targets, Technology, and Tight Control (4T) Study provides precision, whole-population care enabled by algorithms that use continuous glucose monitoring (CGM) data to direct clinician attention to patients with deteriorating glucose management. We used data from the 4T Pilot (n=133) and 4T Study 1 (n=135), in which algorithms use CGM data to identify youth with T1D meeting criteria for clinical review and potential clinician contact. Through iterative data analysis and interviews with diabetes educators and clinicians, we identified metrics for reviewing and revising clinical workloads, glucose management, and timeliness of care. For each metric, we developed an interactive dashboard to provide clinical and administrative leaders with an overview of the program. The metrics to track clinical workload were the total number of youths: (1) in the program, (2) in each study, and (3) cared for by each clinician. The metrics to track glucose management were the number of youths meeting each criterion for review: (4) in total, (5) for each clinician, and (6) for each study. The metric to track timeliness of care was (7) the number of days since meeting criteria for clinical review. When presented at weekly program leadership meetings, the metrics facilitated data-driven decision making about clinical and operational components of the program. We propose a novel quantitative framework for diabetes care teams to supervise and enhance algorithm-directed whole-population T1D care. As the role of algorithms grows in directing clinical effort and prioritizing patients for care, this framework may help clinics track clinical workload, patient outcomes, and the timeliness of care.

A panel of experts in the use of continuous glucose monitoring (CGM) data in the treatment of diabetes met in Burlingame, California on October 27, 2025 to discuss the utility of the glycemia risk index (GRI) for clinical care research and population health management. The GRI composite metric is a single number (on a 0-100 percentile scale-lower is better) based on an expert-determined weighting of the seven individual components in the existing ambulatory glucose profile (AGP). The GRI describes the quality of glycemia based on glucose values collected in a 14-day CGM tracing, thus providing additional insights into CGM profiles beyond the AGP. During the meeting, the mathematical derivation of the GRI metric was presented along with its use for adult and pediatric individuals with diabetes and cancer who require medications that can adversely affect the glucose concentration. Examples where the GRI provided useful insights into the quality of CGM tracings were also discussed by the expert panel. In addition, a new smartphone application, the GRI Calculator, was presented. This app calculates the GRI of a CGM tracing and provides visualization of sequential CGM tracings for a specific individual. The GRI provides a reference measurement for the accuracy of artificial intelligence (AI) models assigning levels of glycemic quality to CGM tracings intended to match the assessments of clinicians. The GRI is now part of the data visualization panel for the Integration of Connected Diabetes Device Data into the Electronic Health Record (iCoDE-2) project, which standardizes both CGM and insulin dosing data. Further exploration of the potential value of the GRI for non-insulin users needs to be undertaken. The panel unanimously recommended that CGM manufacturers and developers of data visualization software for CGMs add the GRI to their data platforms for insulin users.

OBJECTIVE:To estimate the extent to which Medicare Advantage (MA) plans underestimate risk, the additional Part A and Part B (A&B) revenue insurers receive as a result, and how this differs across insurers and over time. STUDY SETTING AND DESIGN:We compared projected risk in plan-level bids submitted by insurers to the actual plan-level risk as subsequently reported by the Centers for Medicare and Medicaid Services (CMS). We estimated the additional A&B revenue MA insurers received due to underestimates. DATA SOURCES AND ANALYTICAL SAMPLES:A total of 34,604 MA plans offered by 317 insurers from 2008 to 2020 from MA Bid Data as well as MA enrollment and payment data, all from publicly available CMS sources. PRINCIPAL FINDINGS:From 2008 to 2020, the average ratio of actual risk to projected risk was 1.01 (95% Confidence Interval, 1.009-1.011) and plans received additional A&B revenue of $14.6 billion due to risk underestimates. The fraction of plans with a ratio above 1.0 rose from 45% in the smallest enrollment quintile to 62% in the largest enrollment quintile. Among the 10 largest insurers, Blue Cross Blue Shield Highmark underestimated risk by the largest amount, an average of 3% from 2008 to 2020, generating $272 per member per year in additional A&B revenue, followed by Centene (2.1%, $202/member). CONCLUSIONS:From 2008 to 2020, MA plans systematically underestimated actual risk, reducing benefits for beneficiaries and generating additional A&B revenue of approximately 1% of premiums. The frequency and magnitude of the underestimates increased with plan size. This additional A&B revenue represents a significant fraction of reported MA profit margins, which averaged approximately 3.5% of premiums over the period. CMS should more closely scrutinize the projected risk scores in the annual bids. The Department of Justice should consider whether systematic risk underestimation is a violation of the False Claims Act.

Context: Youth with type 1 diabetes (T1D) struggle to meet and sustain hemoglobin A1c (HbA1c) targets. Youth enrolled in the Pilot 4T Study improved HbA1c by 0.5% at 1 year, compared to historical controls. Objective: To assess 3 years of glycemic outcomes in the Pilot 4T Study. Methods: The Pilot 4T Extension cohort was prospectively followed to determine changes in HbA1c and continuous glucose monitoring (CGM) metrics over 3 years at the Stanford Medicine Children's Health Diabetes Clinic. Youth with T1D in the Pilot 4T Study enrolled in the extension phase started CGM in the first month of diabetes diagnosis, received intensified education and remote patient monitoring (RPM) weekly for the first year of diabetes diagnosis, and monthly RPM in the extension phase. HbA1c and CGM metrics were evaluated over the first 3 years of diagnosis. Results: In the Pilot 4T cohort, 78.5% (n = 102) of participants enrolled in the study extension phase and were followed through 3 years. The adjusted difference in HbA1c at 3 years was 1.2% (95% CI 0.7%-1.7%) lower in the Pilot 4T cohort than in the Historical cohort. In the Pilot 4T cohort, 68% and 37% met the <7.5% and <7% HbA1c targets at 3 years, respectively, compared to 37% and 20% in the Historical cohort. Conclusion: Youth with T1D in the Pilot 4T extension phase sustained improvements in HbA1c over 3 years. Focusing resources on intensive management during the first year after T1D diagnosis may impact long-term glycemia.

OBJECTIVE:The use of continuous glucose monitoring (CGM) with remote patient monitoring (RPM) continues to grow. We evaluated the cost-effectiveness of CGM with RPM compared with self-monitoring of blood glucose (SMBG) and CGM alone. RESEARCH DESIGN AND METHODS:We simulated type 1 diabetes progression with a Markov model in 5-year-old patients over a 20-year, 50-year, and lifetime horizon. We tracked diabetic ketoacidosis (DKA), severe hypoglycemia (SH), and seven chronic complications: retinopathy, neuropathy, nephropathy, cardiovascular disease, end-stage renal disease, lower-extremity amputation, and blindness. We compared three interventions: SMBG, CGM, and CGM with RPM. Efficacy estimates were derived from meta-analyses of pediatric CGM studies and the results of the Teamwork, Targets, Technology, and Tight Glycemia Study (4T Study 1). We evaluated quality-adjusted life years (QALYs) and health care costs (2022 U.S. dollars) discounted at 3% annually. We performed extensive sensitivity analyses. RESULTS:Compared with SMBG, CGM increased QALYs by 0.09 and costs by $8,900 over 20 years; CGM with RPM increased QALYs by 0.37, and costs by $10,300. CGM with RPM yielded more QALYs at a lower incremental cost-effectiveness ratio compared with CGM ($27,400/QALY vs. $103,700/QALY, respectively). Results were robust across sensitivity analyses and time horizons. CGM with RPM remained cost-effective when achieving at least 30% of 4T's clinical efficacy. CONCLUSIONS:CGM with RPM delivers superior health outcomes compared with SMBG and CGM and is likely cost-effective for patients with newly diagnosed type 1 diabetes. Despite higher intervention costs, CGM with RPM can reduce complications costs and generate net health care savings.

Introduction and Objective: There is a growing shortage of pediatric endocrinologists, which affects the care available to people with type 1 diabetes’ (PwT1D). We benchmarked provider (physician, nurse practitioner (NP), or physician assistant (PA)) infrastructure for 14,324 PwT1D aged 1-21 years across 9 centers that provided staffing data. Methods: Centers reported provider staffing data to the T1D Exchange Collaborative while EHR data provided insurance, race/ethnicity, pump/CGM use, and HbA1c metrics. Means, percentages, and ratios were calculated to analyze EMR metrics and provider staffing across centers. Results: New diagnoses in 2022/2023 averaged 16% of the total PwT1D clinic population across the 9 centers. The clinic population was diverse (70% non-Hispanic White, 11% non-Hispanic Black, 11% Hispanic, 8% Other). Insurance coverage distribution varied widely between centers, with 28% to 58% of PwT1D publicly insured and 2% to 70% privately insured. Mean HbA1c ranged from 7.8% to 9.2% a year after diagnosis. Insulin pump use exceeded 50% in all centers and average CGM use was 78%. On average, there were 216 (99-332) PwT1D and 32 (15-60) new diagnoses per physician (FTE). Conclusion: These ratios highlight significant workload variation among providers, with some centers facing greater strain due to more cases/fewer providers. This underscores the need for adequate staffing to effectively manage T1D. S. Thapa: None. N. Rioles: None. D.M. Maahs: Advisory Panel; Abbott, Medtronic. Research Support; Dexcom, Inc. Consultant; Sanofi. S. Crossen: None. C. Demeterco-Berggren: None. K.K. Hood: Consultant; Sanofi. Advisory Panel; MannKind Corporation. Consultant; Havas Health. Research Support; embecta. L.M. Jacobsen: Advisory Panel; Sanofi. M.K. Kamboj: None. F. Malik: None. E.A. Mann: None. D. Scheinker: None. R.M. Wolf: Research Support; Novo Nordisk, Lilly Diabetes. Advisory Panel; Uneo. Research Support; Sanofi. O. Ebekozien: Research Support; Abbott. Advisory Panel; Sanofi. Research Support; Sanofi, Lilly Diabetes, Medtronic. The Leona M. and Harry B. Helmsley Charitable Trust

The COVID-19 pandemic disproportionately impacted the Hispanic population in the United States, leading to an unprecedented decline in the longstanding Hispanic mortality advantage (HMA) and highlighting the need to better understand the sociodemographic and structural factors driving these trends. To evaluate the association between county-level determinants (including demographic, socioeconomic, behavioral, healthcare, and structural factors) and declines in the HMA during COVID-19 pandemic. Data on non-Hispanic White (NHW) and Hispanic individuals were obtained from the Centers for Disease Control and Prevention Wide-Ranging Online Data for Epidemiologic Research and the Robert Wood Johnson Foundation County Health Rankings databases. County-level sociodemographic and structural factors. In this cross-sectional analysis of national county-level mortality data from 2019 to 2020, the primary outcome was the change in the HMA between 2019 and 2020. All-cause age-adjusted mortality rates (AAMRs) at the county level were used to calculate the HMA (NHW AAMR − Hispanic AAMR). Coronavirus disease 2019 (COVID-19)–specific AAMRs and county-level factors were identified. We analyzed 619 US counties with complete and reliable mortality and sociodemographic data from a total of 3193 counties. From 2019 to 2020, the mean AAMR per 100,000 persons increased by 12.5

Introduction: The Virtual Diabetes Specialty Clinic (VDiSC) study demonstrated the feasibility of providing comprehensive diabetes care entirely virtually by combining virtual visits with continuous glucose monitoring support and remote patient monitoring (RPM). However, the financial sustainability of this model remains uncertain.Methods: We developed a financial model to estimate the variable costs and revenues of virtual diabetes care, using visit data from the 234 VDiSC participants with type 1 or type 2 diabetes. Data included virtual visits with certified diabetes care and education specialists (CDCES), endocrinologists, and behavioral health services (BHS). The model estimated care utilization, variable costs, reimbursement revenue, gross profit, and gross profit margin per member, per month (PMPM) for privately insured, publicly insured, and overall clinic populations (75% privately insured). We performed two-way sensitivity analyses on key parameters.Results: Gross profit and gross profit margin PMPM (95% confidence interval) were estimated at $-4 ($-14.00 to $5.68) and -4% (-3% to -6%) for publicly insured patients; $267.26 ($256.59-$277.93) and 73% (58%-88%) for privately insured patients; and $199.41 ($58.43-$340.39) and 67% (32%-102%) for the overall clinic. Profits were primarily driven by CDCES visits and RPM. Results were sensitive to insurance mix, cost-to-charge ratio, and commercial-to-Medicare price ratio.Conclusions: Virtual diabetes care can be financially viable, although profitability relies on privately insured patients. The analysis excluded fixed costs of clinic infrastructure, and securing reimbursement may be challenging in practice. The financial model is adaptable to various care settings and can serve as a planning tool for virtual diabetes clinics.

Background:Telemedicine use has increased significantly in cardiology clinics, but the impact of initial telemedicine evaluation on total visit usage is unknown. Objective:This study aimed to determine the effect of initial telemedicine evaluation on the number of follow-up visits within 6 months for common cardiovascular conditions at an academic health system. Methods:Electronic health records data were extracted for general cardiology visits. New patient visits (NPVs) were included occurring from June 1, 2020, to May 31, 2023, for 10 common cardiovascular conditions-atrial fibrillation or flutter, chest pain, coronary artery disease, dyslipidemia, dyspnea, heart failure, hypertension, palpitations, preoperative evaluation, and syncope or dizziness. The effect of initial telemedicine versus in-person evaluation on follow-up visits within 6 months was assessed using a 2-stage least squares instrumental variable model with the proportion of clinician telemedicine use as the instrument and adjustment for patient and visit characteristics. Results:There were 5528 NPVs conducted by 40 general cardiology clinicians during the study period. The average patient age was 56 (SD 17.5) years, 54.2% (2998/5528) were female, 43.2% (2389/5528) were non-Hispanic White, 24.7% (1368/5528) were Asian, 13.8% (761/5528) were Hispanic, 34.4% (1904/5528) were on Medicare, and 13.2% (729/5528) were on Medicaid. Of the NPVs, 53.5% (2959/5528) were conducted via telemedicine (2814/5528, 50.9% via video and 145/5528, 2.6% via phone). Telemedicine use for NPVs ranged from 0% to 100% (N=40) across individual clinicians. The average number of follow-up visits was 57 visits per 100 patients within 6 months across all diagnosis groups. Patients receiving telemedicine NPVs were more likely to have telemedicine follow-up visits than those receiving in-person NPVs (1354/1619, 83.6% vs 680/1533, 44.4%). In the instrumental variable analysis, the impact of initial telemedicine evaluation differed by presenting condition. There was an increase in follow-up visits for patients with syncope or dizziness (29.8 visits/100 patients, 95% CI 6.4-53.1), palpitations (34.9 visits/100 patients, 95% CI 18.6-51.1), chest pain (36.9 visits/100 patients, 95% CI 18.5-55.2), and dyspnea (37.0 visits/100 patients, 95% CI 11.8-62.0). There was a decrease in follow-up visits for patients with coronary artery disease (-29.5 visits/100 patients, 95% CI -50.3 to -8.6) and dyslipidemia (-24.5 visits/100 patients, 95% CI -40.2 to -8.8). There was no significant effect for patients presenting for atrial fibrillation or flutter, heart failure, hypertension, and preoperative evaluation. Conclusions:The effect of initial telemedicine evaluation on follow-up visits varied significantly by presenting condition in this cardiology practice. Telemedicine use resulted in increased follow-up visits for patients presenting with symptomatic complaints, while for those presenting with chronic conditions, there was no significant effect or a decrease in visits. Future studies should assess strategies to target initial care modalities to appropriate patients in cardiology clinics with early in-person evaluation for symptomatic patients.