Search

OBJECTIVEContinuous glucose monitoring (CGM) parameters may identify individuals at risk for progression to overt type 1 diabetes. We aimed to determine whether CGM metrics provide additional insights into progression to clinical stage 3 type 1 diabetes.RESEARCH DESIGN AND METHODSOne hundred five relatives of individuals in type 1 diabetes probands (median age 16.8 years; 89% non-Hispanic White; 43.8% female) from the TrialNet Pathway to Prevention study underwent 7-day CGM assessments and oral glucose tolerance tests (OGTTs) at 6-month intervals. The baseline data are reported here. Three groups were evaluated: individuals with 1) stage 2 type 1 diabetes (n = 42) with two or more diabetes-related autoantibodies and abnormal OGTT; 2) stage 1 type 1 diabetes (n = 53) with two or more diabetes-related autoantibodies and normal OGTT; and 3) negative test for all diabetes-related autoantibodies and normal OGTT (n = 10).RESULTSMultiple CGM metrics were associated with progression to stage 3 type 1 diabetes. Specifically, spending ≥5% time with glucose levels ≥140 mg/dL (P = 0.01), ≥8% time with glucose levels ≥140 mg/dL (P = 0.02), ≥5% time with glucose levels ≥160 mg/dL (P = 0.0001), and ≥8% time with glucose levels ≥160 mg/dL (P = 0.02) were all associated with progression to stage 3 disease. Stage 2 participants and those who progressed to stage 3 also exhibited higher mean daytime glucose values; spent more time with glucose values over 120, 140, and 160 mg/dL; and had greater variability.CONCLUSIONSCGM could aid in the identification of individuals, including those with a normal OGTT, who are likely to rapidly progress to stage 3 type 1 diabetes.

WAVES is a large, single-center dataset comprising 9 years of high-frequency physiological waveform data from patients in intensive and acute care units at a large academic, pediatric medical center. The data comprise approximately 10.6 million hours of 1 to 20 concurrent waveforms over approximately 50,364 distinct patient encounters. The data have been de-identified, cleaned, and organized to facilitate research. Initial analyses demonstrate the potential of the data for clinical applications such as non-invasive blood pressure monitoring and methodological applications such as waveform-agnostic data imputation. WAVES is the largest pediatric-focused and second largest physiological waveform dataset available for research.

Objective The purpose of this study is to evaluate the ability of three metrics to monitor for a reduction in performance of a chronic kidney disease (CKD) model deployed at a pediatric hospital. Methods The CKD risk model estimates a patient's risk of developing CKD 3 to 12 months following an inpatient admission. The model was developed on a retrospective dataset of 4,879 admissions from 2014 to 2018, then run silently on 1,270 admissions from April to October, 2019. Three metrics were used to monitor its performance during the silent phase: (1) standardized mean differences (SMDs); (2) performance of a “membership model”; and (3) response distribution analysis. Observed patient outcomes for the 1,270 admissions were used to calculate prospective model performance and the ability of the three metrics to detect performance changes. Results The deployed model had an area under the receiver-operator curve (AUROC) of 0.63 in the prospective evaluation, which was a significant decrease from an AUROC of 0.76 on retrospective data (p = 0.033). Among the three metrics, SMDs were significantly different for 66/75 (88%) of the model's input variables (p Conclusion This study suggests that the three metrics examined could provide early indication of performance deterioration in deployed models' performance.

Background: Informed by drivers of visit modality in specialty clinics, we sought to identify factors associated with visit modality selection (telemedicine vs. in-person) for new and return visits in primary care clinics at a large academic medical center (AMC). Methods: We used electronic health record data from 3/2020 - 5/2022 from 13 primary care clinics with 98 unique clinicians for new and 131 for return visits (21,031 new & 207,292 return visits), with about 55% telemedicine (telephone and video) use. We used hierarchical logistic regression and cross-validation methods to estimate the variation in visit modality associated with the patient, clinician, and visit factors (measured with Area Under the Curve). Results: For both new and return visits, there was significant variation in telemedicine use among clinicians (ranging from 0 - 100%) for specific clinical diagnoses (Figure 1). Most variation in telemedicine use was attributed to the clinician seen (new visits) and primary visit diagnosis (return visits). Other visit and patient characteristics were less predictive. For new visits, the clinician model had an AUC of 0.79, followed by clinic site 0.69, whereas for return visits, the primary diagnosis was 0.77, followed by the clinician seen at 0.65. Diagnoses most commonly seen via telemedicine include acute respiratory infection and suspected COVID-19 exposure. Diagnoses commonly seen in-person include annual physical and gynecological exams. Conclusions: Our findings show high variability in telemedicine use for the same diagnosis and differential drivers of visit modality between new and return visits in primary care clinics at a large AMC. Provider ID was the most predictive factor of visit modality for new patients, indicating that clinician preference and individual practice patterns influence the modality of care much more than patient preference. Primary care clinics may reduce the variability in visit modality through standardized processes that integrate clinical factors and patient preference.

OBJECTIVE Continuous glucose monitoring (CGM) parameters may identify subjects at risk of progressing to overt type 1 diabetes. We aimed to determine whether CGM metrics provides additional insights into progression to clinical Stage 3 type 1 diabetes. RESEARCH DESIGN AND METHODS One hundred and five relatives of type 1 diabetes probands (median age 16.8 years; 89% non-Hispanic White; 43.8% female) from the TrialNet Pathway to Prevention Study underwent 7-day CGM assessments and oral glucose tolerance tests (OGTTs) at 6-month intervals, the baseline data is reported here. Three groups were evaluated: individuals with 1) Stage 2 type 1 diabetes (n=42) with ≥2 diabetes-related autoantibodies and abnormal OGTT; 2) Stage 1 type 1 diabetes (n=53) with ≥2 diabetes-related autoantibodies and normal OGTT; and 3) negative test for all diabetes-related autoantibodies and normal OGTT (n=10). RESULTS Multiple CGM metrics were associated with progression to Stage 3 type 1 diabetes. Specifically, spending ≥5% time with glucose levels ≥140 mg/dL (p = 0.01), ≥8% time ≥140 mg/dL (p=0.02), ≥5% time ≥160 mg/dL (p = 0.0001) and ≥8% of the time spent at glucose levels ≥160 mg/dL (p=0.02) were all associated with progression to Stage 3 disease. Stage 2 participants and those who progressed to Stage 3 also exhibited higher mean day-glucose values, spent more time with glucose values over 120, 140 and 160 mg/dL, and had greater variability. CONCLUSIONS CGM could aid in the identification of subjects, including those with a normal OGTT, who are likely to rapidly progress to Stage 3 type 1 diabetes.

Algorithm-enabled patient prioritization and remote patient monitoring (RPM) have been used to improve clinical workflows at Stanford and have been associated with improved glucose time-in-range in newly diagnosed youth with type 1 diabetes (T1D). This novel algorithm-enabled care model currently integrates continuous glucose monitoring (CGM) data to prioritize patients for weekly reviews by the clinical diabetes team. The use of additional data may help clinical teams make more informed decisions around T1D management. Regular exercise and physical activity are essential to increasing cardiovascular fitness, increasing insulin sensitivity, and improving overall well-being of youth and adults with T1D. However, exercise can lead to fluctuations in glycemia during and after the activity. Future iterations of the care model will integrate physical activity metrics (e.g., heart rate and step count) and physical activity flags to help identify patients whose needs are not fully captured by CGM data. Our aim is to help healthcare professionals improve patient care with a better integration of CGM and physical activity data. We hypothesize that incorporating exercise data into the current CGM-based care model will produce specific, clinically relevant information such as identifying whether patients are meeting exercise guidelines. This work provides an overview of the essential steps of integrating exercise data into an RPM program and the most promising opportunities for the use of these data.

Objective:To assess the training and the future workforce needs of paediatric cardiac critical care faculty.Design:REDCap surveys were sent May−August 2019 to medical directors and faculty at the 120 US centres participating in the Society of Thoracic Surgeons Congenital Heart Surgery Database. Faculty and directors were asked about personal training pathway and planned employment changes. Directors were additionally asked for current faculty numbers, expected job openings, presence of training programmes, and numbers of trainees. Predictive modelling of the workforce was performed using respondents’ data. Patient volume was projected from US Census data and compared to projected provider availability.Measurements and main results:Sixty-six per cent (79/120) of directors and 62% (294/477) of contacted faculty responded. Most respondents had training that incorporated critical care medicine with the majority completing training beyond categorical fellowship. Younger respondents and those in dedicated cardiac ICUs were more significantly likely to have advanced training or dual fellowships in cardiology and critical care medicine. An estimated 49–63 faculty enter the workforce annually from various training pathways. Based on modelling, these faculty will likely fill current and projected open positions over the next 5 years.Conclusions:Paediatric cardiac critical care training has evolved, such that the majority of faculty now have dual fellowship or advanced training. The projected number of incoming faculty will likely fill open positions within the next 5 years. Institutions with existing or anticipated training programmes should be cognisant of these data and prepare graduates for an increasingly competitive market.

Despite a fingerstick hemoglobin requirement and 56-day minimum donation interval, repeat blood donation continues to cause and exacerbate iron deficiency.Using data from the REDS-II Donor Iron Status Evaluation study, we developed multiclass prediction models to estimate the competing risk of hemoglobin deferral and collecting blood from a donor with sufficient hemoglobin but low or absent underlying iron stores. We compared models developed with and without two biomarkers not routinely measured in most blood centers: ferritin and soluble transferrin receptor. We generated and analyzed "individual risk trajectories": estimates of how each donors' risk developed as a function of the time interval until their next donation attempt.With standard biomarkers, the top model had a multiclass area under the receiver operator characteristic curve (AUC) of 77.6% (95% CI [77.3%-77.8%]). With extra biomarkers, multiclass AUC increased to 82.8% (95% CI [82.5%-83.1%]). In the extra biomarkers model, ferritin was the single most important variable, followed by the donation interval. We identified three risk archetypes: "fast recoverers" (10% risk of any adverse outcome on post-donation day 56), "slow recoverers" (60% adverse outcome risk on day 56 that declines to35% by day 250), and "chronic high-risk" (85% risk of the adverse outcome on day 250).A longer donation interval reduced the estimated risk of iron-related adverse outcomesfor most donors, but risk remained high for some. Tailoring safeguards to individual risk estimates could reduce blood collections from donors with low or absent iron stores.

IntroductionPopulation-level algorithm-enabled remote patient monitoring (RPM) based on continuous glucose monitor (CGM) data review has been shown to improve clinical outcomes in diabetes patients, especially children. However, existing reimbursement models are geared towards the direct provision of clinic care, not population health management. We developed a financial model to assist pediatric type 1 diabetes (T1D) clinics design financially sustainable RPM programs based on algorithm-enabled review of CGM data.MethodsData were gathered from a weekly RPM program for 302 pediatric patients with T1D at Lucile Packard Children’s Hospital. We created a customizable financial model to calculate the yearly marginal costs and revenues of providing diabetes education. We consider a baseline or status quo scenario and compare it to two different care delivery scenarios, in which routine appointments are supplemented with algorithm-enabled, flexible, message-based contacts delivered according to patient need. We use the model to estimate the minimum reimbursement rate needed for telemedicine contacts to maintain revenue-neutrality and not suffer an adverse impact to the bottom line.ResultsThe financial model estimates that in both scenarios, an average reimbursement rate of roughly $10.00 USD per telehealth interaction would be sufficient to maintain revenue-neutrality. Algorithm-enabled RPM could potentially be billed for using existing RPM CPT codes and lead to margin expansion.ConclusionWe designed a model which evaluates the financial impact of adopting algorithm-enabled RPM in a pediatric endocrinology clinic serving T1D patients. This model establishes a clear threshold reimbursement value for maintaining revenue-neutrality, as well as an estimate of potential RPM reimbursement revenue which could be billed for. It may serve as a useful financial-planning tool for a pediatric T1D clinic seeking to leverage algorithm-enabled RPM to provide flexible, more timely interventions to its patients.

Background: Guidelines recommend observation for atrioventricular node recovery until postoperative days (POD) 7 to 10 before permanent pacemaker placement (PPM) in patients with heart block after congenital cardiac surgery. To aid in surgical decision-making for early PPM, we established criteria to identify patients at high risk of requiring PPM. Methods: We reviewed all cases of second degree and complete heart block (CHB) on POD 0 from August 2009 through December 2018. A decision tree model was trained to predict the need for PPM amongst patients with persistent CHB and prospectively validated from January 2019 through March 2021. Separate models were developed for all patients on POD 0 and those without recovery by POD 4. Results: Of the 139 patients with postoperative heart block, 68 required PPM. PPM was associated with older age (3.2 versus 1.0 years; P =0.018) and persistent CHB on POD 0 (versus intermittent CHB or second degree heart block; 87% versus 58%; P =0.001). Median days [IQR] to atrioventricular node recovery was 2 [0–5] and PPM was 9 [6–11]. Of the 100 cases of persistent CHB (21 in the validation cohort), 59 (59%) required PPM. A decision tree model identified 4 risk factors for PPM in patients with persistent CHB: (1) aortic valve replacement, subaortic stenosis repair, or Konno procedure; (2) ventricular L-looping; (3) atrioventricular valve replacement; (4) and absence of preoperative antiarrhythmic agent (in POD 0 model only). The POD 4 model specificity was 0.89 [0.67–0.99] and positive predictive value was 0.94 [95% CI 0.81–0.98], which was stable in prospective validation (positive predictive value 1.0). Conclusions: A data-driven analysis led to actionable criteria to identify patients requiring PPM. Patients with left ventricular outflow tract surgery, atrioventricular valve replacement, or ventricular L-Looping could be considered for PPM on POD 4 to reduce risks of temporary pacing and improve care efficiency.

Background COVID-19 spurred rapid adoption and expansion of telemedicine. We investigated the factors driving visit modality (telemedicine vs. in-person) for outpatient visits at a large cardiovascular center. Methods We used electronic health record data from March 2020 to February 2021 from four cardiology subspecialties (general cardiology, electrophysiology, heart failure, and interventional cardiology) at a large academic health system in Northern California. There were 21,912 new and return visits with 69% delivered by telemedicine. We used hierarchical logistic regression and cross-validation methods to estimate the variation in visit modality explained by patient, clinician, and visit factors as measured by the mean area under the curve. Results Across all subspecialties, the clinician seen was the strongest predictor of telemedicine usage, while primary visit diagnosis was the next most predictive. In general cardiology, the model based on clinician seen had a mean area under the curve of 0.83, the model based on the primary diagnosis had a mean area under the curve of 0.69, and the model based on all patient characteristics combined had a mean area under the curve of 0.56. There was significant variation in telemedicine use across clinicians within each subspecialty, even for visits with the same primary visit diagnosis. Conclusion Individual clinician practice patterns had the largest influence on visit modality across subspecialties in a large cardiovascular medicine practice, while primary diagnosis was less predictive, and patient characteristics even less so. Cardiovascular clinics should reduce variability in visit modality selection through standardized processes that integrate clinical factors and patient preference.

Aim: Identify, simulate, and evaluate the formal and informal patient-level and unit-level factors that nurse managers use to determine the number of nurses for each shift. Background: Nurse staffing schedules are commonly set based on metrics such as midnight census that do not account for seasonality or midday turnover, resulting in last-minute adjustments or inappropriate staffing levels. Methods: Staffing schedules at a pediatric intensive care unit (PICU) were simulated based on nurse-to-patient assignment rules from interviews with nursing management. Multivariate regression modeled the discrepancies between scheduled and historical staffing levels and constructed rules to reduce these discrepancies. The primary outcome was the median difference between simulated and historical staffing levels.Results: Nurse-to-patient ratios underestimated staffing by a median of 1.5 nurses per shift. Multivariate regression identified patient turnover as the primary factor accounting for this difference and subgroup analysis revealed that patient age and weight were also important. New rules reduced the difference to a median of 0.07 nurses per shift.Conclusion: Measurable, predictable indicators of patient acuity and historical trends may allow for schedules that better match demand.Implications for Nursing Management: Data-driven methods can quantify what drives unit demand and generate nurse schedules that require fewer last-minute adjustments.

ContextEarly initiation of continuous glucose monitoring (CGM) is advocated for youth with type 1 diabetes (T1D). Data to guide CGM use on time-in-range (TIR), hypoglycemia, and the role of partial clinical remission (PCR) are limited.ObjectiveOur aims were to assess whether 1) an association between increased TIR and hypoglycemia exists, and 2) how time in hypoglycemia varies by PCR status.MethodsWe analyzed 80 youth who were started on CGM shortly after T1D diagnosis and were followed for up to 1-year post diagnosis. TIR and hypoglycemia rates were determined by CGM data and retrospectively analyzed. PCR was defined as (visit glycated hemoglobin A1c) + (4*units/kg/day) less than 9.ResultsYouth were started on CGM 8.0 (interquartile range, 6.0-13.0) days post diagnosis. Time spent at less than 70 mg/dL remained low despite changes in TIR (highest TIR 74.6 ± 16.7%, 2.4 ± 2.4% hypoglycemia at 1 month post diagnosis; lowest TIR 61.3 ± 20.3%, 2.1 ± 2.7% hypoglycemia at 12 months post diagnosis). No events of severe hypoglycemia occurred. Hypoglycemia was rare and there was minimal difference for PCR vs non-PCR youth (54-70 mg/dL: 1.8% vs 1.2%, P = .04; ConclusionAs TIR gradually decreased over 12 months post diagnosis, hypoglycemia was limited with no episodes of severe hypoglycemia. Hypoglycemia rates did not vary in a clinically meaningful manner by PCR status. With CGM being started earlier, consideration needs to be given to modifying CGM hypoglycemia education, including alarm settings. These data support a trial in the year post diagnosis to determine alarm thresholds for youth who wear CGM.

Hispanic populations generally experience more adverse socioeconomic conditions yet demonstrate lower mortality compared with Non-Hispanic White (NHW) populations in the US. This finding of a mortality advantage is well-described as the “Hispanic paradox.” The Coronavirus Disease 2019 (COVID-19) pandemic has disproportionately affected Hispanic populations. To quantify these effects, we evaluated US national and county-level trends in Hispanic versus NHW mortality from 2011 through 2020. We found that a previously steady Hispanic mortality advantage significantly decreased in 2020, potentially driven by COVID-19-attributable Hispanic mortality. Nearly 16% of US counties experienced a reversal of their pre-pandemic Hispanic mortality advantage such that their Hispanic mortality exceeded NHW mortality in 2020. An additional 50% experienced a decrease in a pre-pandemic Hispanic mortality advantage. Our work provides a quantitative understanding of the disproportionate burden of the pandemic on Hispanic health and the Hispanic paradox and provides a renewed impetus to tackle the factors driving these concerning disparities.

We previously examined a type 1 diabetes care model based on the use of a remote patient monitoring (RPM) tool that analyzes CGM data. The care model was associated with improved glucose management and reduced provider effort. We compare the financial feasibility of this care model in value-based care (VBC) and fee-for-service (FFS) settings. We set parameters based on the current deployment of the care model. In the FFS setting, we estimated Medicare reimbursement rates and labor costs for in-person diabetes education. We compared these to costs in a VBC setting where providers make prospective payments for RPM services. We evaluated the financial impact of the care model across a variety of parameters (Fig) . In a FFS setting, we estimated that in-person care incurs net costs of $1per patient. Under a VBC contract, we estimated that prospective payments are offset by reduced labor costs, incurring net costs of $65 per patient (i.e., 49% savings) . These results are sensitive to input parameters (Fig) . We identified financially viable payment models for an algorithm-enabled telemedicine-based care model associated with improved T1D glucose management. The analysis may inform operational planning at the clinic level as well as reimbursement policy design. Disclosure R.Leonard pei: None. P.Dupenloup: None. P.Prahalad: None. R.Johari: None. A.Addala: None. D.P.Zaharieva: Research Support; Insulet Corporation, International Society for Pediatric and Adolescent Diabetes, Leona M. and Harry B. Helmsley Charitable Trust. D.M.Maahs: Advisory Panel; Abbott Diabetes, Eli Lilly and Company, Medtronic, Novo Nordisk, Sanofi, Consultant; Aditx Therapeutics, Inc., Biospex. D.Scheinker: None. Funding Helmsley Charitable Trust.ISPAD-JDRF Fellowship.R18DK122422.SDRC.LPCH Auxiliaries.The Stanford REDCap platform (http://redcap.stanford.edu) is developed and operated by Stanford Medicine Research IT team. The REDCap platform services at Stanford are subsidized by a) Stanford School of Medicine Research Office, and b) the National Center for Research Resources and the National Center for Advancing Translational Sciences, National Institutes of Health, through grant UL1 TR001085.Stanford Maternal and Child Health Research Institute.

We previously showed that the use of a tool for remote analysis of continuous glucose monitor data, Timely Interventions for Diabetes Excellence (TIDE) , was associated with improved glucose management for patients with type 1 diabetes. To facilitate the deployment of TIDE at two partner institutions, we developed an operational planning tool. We designed an interactive dashboard to match capacity and demand with a variety of operational, population, and workforce parameters. To improve the dashboard’s accessibility, simplicity, and interpretability, we performed iterative design based on feedback from users and two partner institutions. We set parameter values based on the first deployment of TIDE and conducted sensitivity analyses of the capacity-demand match. The dashboard (Figure 1) contains seven modifiable parameters; generates a table illustrating the match between capacity and demand; displays a timeseries plot of capacity and demand; and calculates three alternative capacity plans that satisfy annual patient demand. It is available online (www.bit.ly/surf-tide) . Small absolute reductions in per-patient review time produced the most significant changes in clinic capacity. We developed an interactive dashboard that facilitates data-driven operational planning for clinics seeking to deploy algorithm-enabled telemedicine-based diabetes care. Figure 1. Capacity Planning Dashboard Disclosure A.Chang: None. M.Z.Gao: None. J.Ferstad: None. D.M.Maahs: Advisory Panel; Abbott Diabetes, Eli Lilly and Company, Medtronic, Novo Nordisk, Sanofi, Consultant; Aditx Therapeutics, Inc., Biospex. P.Prahalad: None. R.Johari: None. D.Scheinker: None. Funding NIH R18DK122422

In the Pilot 4T study (n=135) , remote patient monitoring (RPM) was associated with improved time in range (TIR) and lower A1c. Measuring differences in how patients respond to messages from the care team is essential for the design of effective, personalized care models. We analyzed electronic health record (EHR) and continuous glucose monitor (CGM) data to estimate the week-over-week impact of RPM messages on patients’ TIR. We applied statistical clustering methods to divide patients who received messages into two groups based on their changes in TIR and compared the characteristics of these groups. Receiving a message was associated with a greater mean week-to-week improvement in TIR after a low-TIR week [4.9 percentage points (pp) after a message vs. 2.5pp without a message; p < 0.001]. A patient’s TIR improvement after receiving a message is greater on average if the same patient’s TIR improved following previous messages. We identified two groups of patients with significantly different responses to messages. The group with greater mean TIR improvement after messages contains a larger proportion of non-white, non-English speaking, and publicly insured patients. We found that messages were associated with different magnitudes of TIR improvement across two clusters of patients. Identifying patients who benefit more from RPM could facilitate the personalization of management strategies. Disclosure J.Ferstad: None. P.Prahalad: None. D.M.Maahs: Advisory Panel; Abbott Diabetes, Eli Lilly and Company, Medtronic, Novo Nordisk, Sanofi, Consultant; Aditx Therapeutics, Inc., Biospex. E.Fox: None. R.Johari: None. D.Scheinker: None. Funding Helmsley Charitable Trust, ISPAD-JDRF Fellowship,NIH R18DK122422,Stanford Diabetes Research Center,LPCH Auxiliaries,Stanford Maternal and Child Health Research Institute, The Stanford REDCap platform (http://redcap.stanford.edu) is developed and operated by Stanford Medicine Research IT team. The REDCap platform services at Stanford are subsidized by a) Stanford School of Medicine Research Office, and b) the National Center for Research Resources and the National Center for Advancing Translational Sciences, National Institutes of Health, through grant UL1 TR001085.

Background Long-term opioid use has negative health care consequences. Patients who undergo surgery are at risk for prolonged opioid use after surgery (POUS). While risk factors have been previously identified, no methods currently exist to determine higher-risk patients. We assessed the ability of a variety of machine-learning algorithms to predict adolescents at risk of POUS and to identify factors associated with this risk. Methods A retrospective cohort study was conducted using a national insurance claims database of adolescents aged 12-21 years who underwent 1 of 1297 surgeries, with general anesthesia, from January 1, 2011 to December 30, 2017. Logistic regression with an L2 penalty and with a logistic regression with an L1 lasso (Lasso) penalty, random forests, gradient boosting machines, and extreme gradient boosted models were trained using patient and provider characteristics to predict POUS (≥1 opioid prescription fill within 90-180 days after surgery) risk. Predictive capabilities were assessed using the area under the receiver-operating characteristic curve (AUC)/C-statistic, mean average precision (MAP); individual decision thresholds were compared using sensitivity, specificity, Youden Index, F1 score, and number needed to evaluate. The variables most strongly associated with POUS risk were identified using permutation importance. Results Of 186,493 eligible patient surgical visits, 8410 (4.51%) had POUS. The top-performing algorithm achieved an overall AUC of 0.711 (95% confidence interval [CI], 0.699-0.723) and significantly higher AUCs for certain surgeries (eg, 0.823 for spinal fusion surgery and 0.812 for dental surgery). The variables with the strongest association with POUS were the days' supply of opioids and oral morphine milligram equivalents of opioids in the year before surgery. Conclusions Machine-learning models to predict POUS risk among adolescents show modest to strong results for different surgeries and reveal variables associated with higher risk. These results may inform health care system-specific identification of patients at higher risk for POUS and drive development of preventative measures.

OBJECTIVES: To quantify and describe patient-generated health data. METHODS: This is a retrospective, single-center study of patients hospitalized in the pediatric cardiovascular ICU between February 1, 2020, and February 15, 2020. The number of data points generated over a 24-hour period per patient was collected from the electronic health record. Data were analyzed by type, and frontline provider exposure to data was extrapolated on the basis of patient-to-provider ratios. RESULTS: Thirty patients were eligible for inclusion. Nineteen were hospitalized after cardiac surgery, whereas 11 were medical patients. Patients generated an average of 1460 (SD 509) new data points daily, resulting in frontline providers being presented with an average of 4380 data points during a day shift (7:00 am to 7:00 pm). Overnight, because of a higher patient-to-provider ratio, frontline providers were exposed to an average of 16 060 data points. There was no difference in data generation between medical and surgical patients. Structured data accounted for >80% of the new data generated. CONCLUSIONS: Health care providers face significant generation of new data daily through the contemporary electronic health record, likely contributing to cognitive burden and putting them at risk for cognitive overload. This study represents the first attempt to quantify this volume in the pediatric setting. Most data generated are structured and amenable to data-optimization systems to mitigate the potential for cognitive overload and its deleterious effects on patient safety and health care provider well-being.

Introduction: To evaluate the correlation between the presence of an independent EHR (compared to a shared EHR system within an adult hospital system) and an externally-derived third party ranking of children’s hospitals. Methods: Children’s hospitals that ranked in the top fifty of the 2019-2020 US News and World Report (USNWR) were included in the analysis. The mean and median ranking of children’s hospitals with independent versus a shared EHR was evaluated. The 2019-2020 USNWR rankings of the top twenty adult hospitals in the United States were then evaluated. For each children’s hospital with an associated adult hospital that was both ranked, it was noted as to whether the EHR for the children’s hospital was independent or shared and statistical differences in rankings compared. Results: Among the top 50 children’s hospitals included, the median USNWR ranking for hospitals was statistically different with an independent EHR than with a shared EHR (13 vs. 30.0) (p = 0.002). The 21 top ranked adult hospitals were associated with 17 children’s hospitals ranked in the top 50. The median ranking for those with an independent EHR was statistically different for those with independent EHR versus shared EHR (7 vs. 28) (p = 0.002). Conclusion: Children’s hospitals with an independent EHR are associated with higher scores on an independent external ranking of hospital quality compared to those which share an EHR with a partner adult hospital.

Current consensus definition for acute kidney injury (AKI) does not specify how baseline serum creatinine should be determined. We assessed how baseline determination impacted AKI incidence and association between AKI and clinical outcomes. We retrospectively applied empirical (measured serum creatinine) and imputed (age/height) baseline estimation methods to pediatric patients discharged between 2014 and 2019 from an academic hospital. Using each method, we estimated AKI incidence and assessed area under ROC curve (AUROC) for AKI as a predictor of three clinical outcomes: application of AKI billing code (proxy for more clinically overt disease), inpatient mortality, and post-hospitalization chronic kidney disease. Incidence was highly variable across baseline methods (12.2–26.7%). Incidence was highest when lowest pre-admission creatinine was used if available and Schwartz bedside equation was used to impute one otherwise. AKI was more predictive of application of an AKI billing code when baseline was imputed universally, regardless of pre-admission values (AUROC 80.7–84.9%) than with any empirical approach (AUROC 64.5–76.6%). AKI was predictive of post-hospitalization CKD when using universal imputation baseline methods (AUROC 67.0–74.6%); AKI was not strongly predictive of post-hospitalization CKD when using empirical baseline methods (AUROC 46.4–58.5%). Baseline determination method did not affect the association between AKI and inpatient mortality. Method of baseline determination influences AKI incidence and association between AKI and clinical outcomes, illustrating the need for standard criteria. Imputing baseline for all patients, even when preadmission creatinine is available, may identify a more clinically relevant subset of the disease.

Summary We consider the problem of optimal feedback control of an unknown, noisy, time‐varying, dynamic system that is initialized repeatedly. Examples include a robotic manipulator which must perf...

Hospital-acquired venous thromboembolisms (HA-VTEs) are increasingly common in pediatric inpatients and associated with significant morbidity and cost. The Braden QD Scale was created to predict the risk of hospital-acquired pressure injury (HAPI) and is used broadly in children's hospitals. This study evaluated the ability of the Braden QD Total score to predict risk of HA-VTE at a quaternary children's hospital.To analyze the predictive potential of the Braden QD Total score and subscores for HA-VTEs, the researchers performed univariate logistic regressions. The increase in a patient's odds of developing an HA-VTE for every 1-point increase in each Braden QD score was evaluated. Each model was evaluated using a 5-fold cross-validated area-under-the-curve of the corresponding receiver operating characteristic curve (AUROC).This study analyzed 27,689 pediatric inpatients. HA-VTE occurred in 135 patients. The odds of HA-VTE incidence increased by 29% (odds ratio 1.29, 95% confidence interval [CI] 1.25-1.34, p0.001) for every 1-point increase in a patient's Braden QD Total score. The AUROC was 0.81 (95% CI 0.77-0.85).The Braden QD Scale is a predictor for HA-VTE, outperforming its original intended use for predicting HAPI and performing similarly to other HA-VTE predictive models. As the Braden QD Total score is currently recorded in the electronic health records of many children's hospitals, it could be practically and easily implemented as a tool to predict which patients are at risk for HA-VTE.

No method is available to systematically study SARS-CoV-2 transmission dynamics using the data that rideshare companies share with government agencies. We developed a proof-of-concept method for the analysis of SARS-CoV-2 transmissions between rideshare passengers and drivers.To assess whether this method could enable hypothesis testing about SARS-CoV-2, we repeated ten 200-day agent-based simulations of SARS-CoV-2 propagation within the Los Angeles County rideshare network. Assuming data access for 25% of infections, we estimated an epidemiologist's ability to analyze the observable infection patterns to correctly identify a baseline viral variant A, as opposed to viral variant A with mask use (50% reduction in viral particle exchange), or a more infectious viral variant B (300% higher cumulative viral load).Simulations had an average of 190,387 potentially infectious rideshare interactions, resulting in 409 average diagnosed infections. Comparison of the number of observed and expected passenger-to-driver infections under each hypothesis demonstrated our method's ability to consistently discern large infectivity differences (viral variant A vs. viral variant B) given partial data from one large city, and to discern smaller infectivity differences (viral variant A vs. viral variant A with masks) given partial data aggregated across multiple cities.This novel statistical method suggests that, for the present and subsequent pandemics, government-facilitated analysis of rideshare data combined with diagnosis records may augment efforts to better understand viral transmission dynamics and to measure changes in infectivity associated with nonpharmaceutical interventions and emergent viral strains.

Background A shared decision-making model is preferred for engaging prostate cancer patients in treatment decisions. However, the process of assessing an individual’s preferences and values is challenging and not formalized. The purpose of this study is to develop an automated decision aid for patient-centric treatment decision-making using decision analysis, preference thresholds and value elicitations to maximize the compatibility between a patient’s treatment expectations and outcome. Methods A template for patient-centric medical decision-making was constructed. The inputs included prostate cancer risk group, pre-treatment health state, treatment alternatives (primarily focused on radiation in this model), side effects (erectile dysfunction, urinary incontinence, nocturia and bowel incontinence), and treatment success (5-year freedom from biochemical failure). A linear additive value function was used to combine the values for each attribute (side effects, success and the alternatives) into a value for all prospects. The patient-reported toxicity probabilities were derived from phase II and III trials. The probabilities are conditioned on the starting state for each of the side effects. Toxicity matrices for erectile dysfunction, urinary incontinence, nocturia and bowel incontinence were created for the treatment alternatives. Toxicity probability thresholds were obtained by identifying the patient’s maximum acceptable threshold for each of the side effects. Results are represented as a visual. R and Rstudio were used to perform analyses, and R Shiny for application creation. Results We developed a web-based decision aid. Based on preliminary use of the application, every treatment alternative could be the best choice for a decision maker with a particular set of preferences. This result implies that no treatment has determinist dominance over the remaining treatments and that a preference-based approach can help patients through their decision-making process, potentially affecting compliance with treatment, tolerance of side effects and satisfaction with the decision. Conclusions We present a unique patient-centric prostate cancer treatment decision aid that systematically assesses and incorporates a patient’s preferences and values to rank treatment options by likelihood of achieving the preferred outcome. This application enables the practice and study of personalized medicine. This model can be expanded to include additional inputs, such as genomics, as well as competing, concurrent or sequential therapies.

Hospital-acquired pressure injuries (HAPIs) cause patient harm and increase health care costs. We sought to evaluate the performance of the Braden QD Scale-associated changes in HAPI incidence.Using electronic health records data from a quaternary children's hospital, we evaluated the association between Braden QD scores and patient risk of HAPI. We analyzed how this relationship changed during a hospitalwide quality HAPI reduction initiative.Of 23,532 unique patients, 108 (0.46%, 95% confidence interval [CI] = 0.38%-0.55%) experienced a HAPI. Every 1-point increase in the Braden QD score was associated with a 41% increase in the patient's odds of developing a HAPI (odds ratio [OR] = 1.41, 95% CI = 1.36-1.46, p0.001). HAPI incidence declined significantly following implementation of a HAPI-reduction initiative (β = -0.09, 95% CI = -0.11 - -0.07, p0.001), as did Braden QD positive predictive value (β = -0.29, 95% CI = -0.44 - -0.14, p0.001) and specificity (β = -0.28, 95% CI = -0.43 - -0.14, p0.001), while sensitivity (β = 0.93, 95% CI = 0.30-1.75, p = 0.01) and the concordance statistic (β = 0.18, 95% CI = 0.15-0.21, p0.001) increased significantly.Decreases in HAPI incidence following a quality improvement initiative were associated with (1) significant deterioration in threshold-dependent performance measures such as specificity and precision and (2) significant improvements in threshold-independent performance measures such as the concordance statistic. The performance of the Braden QD Scale is more stable as a tool that continuously measures risk than as a prediction tool.

CGM use is associated with improvements in A1c. We hypothesized that initiation of CGM in the 1st month after T1D diagnosis would improve A1c. Youth with newly diagnosed T1D from July 2018 to May 2020 (pilot cohort, n=122) were offered CGM initiation in the 1st month after T1D diagnosis (119 started CGM). We compared A1c outcomes in the pilot cohort with those diagnosed from 2014-2016 (controls, n=272) who were not offered early CGM. We visualized A1c trajectories using locally estimated scatter plot smoothing (Fig). We and assessed for differences in A1c trajectory by cohort via interaction terms in a linear mixed model adjusted for baseline characteristics (age, sex, race and insurance type). The mean A1c at diagnosis was higher in the pilot cohort (12.2 vs. 10.7%). The median age of diagnosis was 9.5 years [6.8, 13.3], 64% male, 38% non-Hispanic White, and 76% with private insurance in the pilot cohort. In this pilot, 89% initiated CGM in the first 30 days after diagnosis compared to 2% in the control cohort. After adjusting for baseline characteristics, the mean A1c of the pilot cohort was lower at 6 months (-0.81, p = 0.019), 9 months (-1.43, p = 0.013), and 12 months (-2.05, p = 0.012) post-diagnosis compared to the historic cohort. Early initiation of CGM was associated with a lower A1c compared to those in a historic cohort. These data support early initiation of CGM as standard of care in youth with T1D. Disclosure P. Prahalad: None. J. Senaldi: None. A. Freeman: None. A. Addala: None. D. P. Zaharieva: None. K. K. Hood: Consultant; Self; Cecelia Health, Cercacor, LifeScan Diabetes Institute. D. Scheinker: Advisory Panel; Self; Carta Healthcare. M. Desai: None. D. M. Maahs: Advisory Panel; Self; Abbott Diabetes, Dompe, Eli Lilly and Company, Medtronic, Novo Nordisk, Consultant; Self; aditxt. V. Ding: None. B. Leverenz: None. J. Hooper: None. A. Chmielewski: None. B. P. Conrad: Advisory Panel; Self; Abbott Diabetes. J. Leverenz: None. A. Martinez-singh: None. P. Sagan: None. Funding National Institutes of Health (R18DK122422, P30DK116074)

Previous studies have shown that CGM use and frequent contact with the care team is associated with improved A1c. We hypothesized that youth with new onset T1D started on CGM with weekly CGM data review by a CDE would have improved outcomes compared to those on CGM alone. Youth with newly diagnosed T1D from 7/2018 to 3/2019 were started on CGM alone (control, n=46) while those from 3/2019 to 5/2020 were started on CGM with weekly data review by a CDE (GluVue, n=76). Dose adjustments were made via secure messaging. We visualized A1c trajectories using locally estimated scatter plot smoothing (Fig) and assessed for differences in A1c trajectory by GluVue status via cohort*time and cohort*time2 interaction terms in a linear mixed model regression adjusted for baseline characteristics (age, sex, race and insurance type). The mean (SD) A1c at diagnosis was 12.0% (1.8) in the control group and 12.3% (2.3) in the GluVue group. Baseline characteristics of control vs. GluVue: male 25% vs. 39%, Non-Hispanic White 41.3% vs. 35.5%, private insurance 80.4% vs. 73.7%. Mean A1c of the GluVue cohort was lower at 6 months (-0.45%), 9 months (-0.07%), and 12 months (-0.52%) post-diagnosis. A1c trajectories were significantly different between groups via regression analysis (p=0.004). Early initiation of CGM with remote monitoring was associated with lower A1c compared to CGM alone. These data support early CGM initiation plus care team contact in youth with T1D. Disclosure P. Prahalad: None. K. K. Hood: Consultant; Self; Cecelia Health, Cercacor, LifeScan Diabetes Institute. D. Scheinker: Advisory Panel; Self; Carta Healthcare. M. Desai: None. D. M. Maahs: Advisory Panel; Self; Abbott Diabetes, Dompe, Eli Lilly and Company, Medtronic, Novo Nordisk, Consultant; Self; aditxt. V. Ding: None. B. Leverenz: None. J. Hooper: None. J. Leverenz: None. P. Sagan: None. A. Martinez-singh: None. A. Freeman: None. D. P. Zaharieva: None. Funding National Institutes of Health (P30DK116074, R18DK122422)

To slow the spread of COVID-19, a shelter in place (SIP) order was imposed in California between 03/16/2020 - 05/31/2020. We assessed the impact of SIP on glycemic control in a pediatric population with T1D using a continuous glucose monitor (CGM). We hypothesized that glucose control would improve due to increased supervision at home. The retrospective study included 96 patients between the ages of 3-22 years who were diagnosed with T1D at least one year earlier. We analyzed CGM data during three time periods: baseline, SIP, and post SIP (06/01/2020 - 07/30/2020) and compared standard CGM metrics controlling for gender (56% male), race (55% White), ethnicity (70% non-Hispanic), age (mean 11 yrs ± 3.9), and insurance type (8.4% public). The mean time in range (70-180 mg/dL: TIR) increased across the three time periods: from 59.9% ± 15.1 at baseline to 62.8% ± 15.9 during SIP (p

The impact of non-clinical transfer delay (TD) from the ICU to a general care unit on the progress of the patient's care is unknown. We measured the association between TD and: (1) the patient's subsequent hospital length of stay (LOS); (2) the timing of care decisions that would advance patient care.This was a single center retrospective study in the United States of patients admitted to the surgical and neurosurgical ICUs during 2013 and 2015. The primary outcome was hospital LOS after transfer request. The secondary outcome was the timing of provider orders representing care decisions (milestones) that would advance the patient's care. Patient, surgery, and bed covariates were accounted for in a multivariate regression and propensity matching analysis.Out of the cohort of 4,926 patients, 1,717 met inclusion criteria. 670 (39%) experienced ≥12 hours of TD. For each day of TD, there was an average increase of 0.70 days in LOS (P 0.001). The last milestone occurred on average 0.35 days later (P 0.001). Propensity matching analyses were confirmatory (P 0.001, P 0.001).TD is associated with longer LOS and delays in milestone clinical decisions that progress care. Eliminating delays in milestones could mitigate TD's impact on LOS.

Introduction: In response to the COVID-19 pandemic, medical practices have expanded utilization of telehealth. Little is known about the operational impacts of transitioning from in-person to video visits in specialty clinics. In 2018, the Stanford South Asian Translational Heart Initiative (SSATHI), a preventive cardiology clinic focused on high-risk South Asian adults, introduced CardioClick, a program replacing in-person follow-up visits with video visits. Hypothesis: We hypothesized that implementation of video visits increased the efficiency of clinic operations. Methods: We extracted visit-level data from the EHR for 134 patients enrolled in CardioClick with video follow-up visits from June 14, 2018 to April 21, 2020 and a cohort of 276 patients enrolled in the in-person SSATHI prevention program with follow-up visits from September 11, 2014 to March 6, 2020. Results: Patients in CardioClick and the in-person cohort were similar in terms of age (mean 45 years), gender balance (23 vs 21% female), and cardiometabolic risk profiles. There were 181 video and 637 in-person follow-up visits. Video visits were shorter than in-person visits, both in terms of total clinic time [median 22 min (IQR 16, 29) vs 67 min (48, 100)] and provider time required [median 22 min (IQR 16, 29) vs 30 min (12, 58)]. Video visits were more likely to end on time (71 vs 11%, p Conclusions: In a preventive cardiology clinic, video follow-up visits required less clinic and provider time than in-person visits, were more likely to end on time, and were associated with increased same-day provider documentation completion. In conclusion, video visits offer benefits beyond their convenience and may increase the operational efficiency of specialty care practices focused on disease prevention, improving value in care delivery.

BackgroundThe emergence of novel, potentially vaccine-resistant strains of SARS-CoV-2 poses a serious risk to public health. The interactions between passengers and drivers facilitated by rideshare platforms such as Uber are, essentially, a series of partially standardized, random experiments of SARS-CoV-2 transmission. Rideshare companies share data with government health agencies, but no statistical method is available to aggregate these data for the systematic study of the transmission dynamics of COVID-19.MethodsWe develop a proof-of-concept model for the analysis of data from rideshare interactions merged with COVID-19 diagnosis records. Using simulated data with rideshare volumes, disease prevalence, and diagnosis rates based on a large US city, we use the model to test hypotheses about the emergence of viral strains and their transmission characteristics in the presence of non-pharmaceutical interventions and superspreaders.FindingsData from 10 simulated trials of SARS-CoV-2 propagation within the Los Angeles rideshare network resulted in an average of 190,387.1 potentially infectious rideshare interactions. Assuming access to data on 25% of the total estimated infections (Partial Reporting), these interactions resulted in an average of 409.0 diagnosed rideshare infections given our transmission model assumptions. For each of the 10 simulated trials, analysis given Partial Reporting could consistently differentiate between a baseline strain and an emergent, more infectious viral strain, enabling hypothesis testing about transmission characteristics.InterpretationSimulated evaluation of a novel statistical model suggests that rideshare data combined with COVID-19 diagnosis data have the potential to automate continued surveillance of emergent novel strains of SARS-CoV-2 and their transmission characteristics.

ObjectiveTo develop and scale algorithm-enabled patient prioritization to improve population-level management of type 1 diabetes (T1D) in a pediatric clinic with fixed resources, using telemedicine and remote monitoring of patients via continuous glucose monitor (CGM) data review.Research Design and MethodsWe adapted consensus glucose targets for T1D patients using CGM to identify interpretable clinical criteria to prioritize patients for weekly provider review. The criteria were constructed to manage the number of patients reviewed weekly and identify patients who most needed provider contact. We developed an interactive dashboard to display CGM data relevant for the patients prioritized for review.ResultsThe introduction of the new criteria and interactive dashboard was associated with a 60% reduction in the mean time spent by diabetes team members who remotely and asynchronously reviewed patient data and contacted patients, from 3.2±0.20 to 1.3±0.24 minutes per patient per week. Given fixed resources for review, this corresponded to an estimated 147% increase in weekly clinic capacity. Patients who qualified for and received remote review (n=58) have associated 8.8 percentage points (pp) (95% CI = 0.6–16.9pp) greater time-in-range (70-180 mg/dL) glucoses compared to 25 control patients who did not qualify at twelve months after T1D onset.ConclusionsAn algorithm-enabled prioritization of T1D patients with CGM for asynchronous remote review reduced provider time spent per patient and was associated with improved time-in-range.

Objective Excess administrative costs in the US health care system are routinely referenced as a justification for comprehensive reform. While there is agreement that these costs are too high, there is little understanding of what generates administrative costs and what policy options might mitigate them. Data sources Literature review and national utilization and expenditure data. Study design We developed a simulation model of physician billing and insurance-related (BIR) costs to estimate how certain policy reforms would generate savings. Our model is based on structural elements of the payment process in the United States and considers each provider's number of health plan contracts, the number of features in each health plan, the clinical and nonclinical processes required to submit a bill for payment, and the compliance costs associated with medical billing. Data extraction For several types of visits, we estimated fixed and variable costs of the billing process. We used the model to estimate the BIR costs at a national level under a variety of policy scenarios, including variations of a single payer "Medicare-for-All" model that extends fee-for-service Medicare to the entire population and policy efforts to reduce administrative costs in a multi-payer model. We conducted sensitivity analyses of a wide variety of model parameters. Principal findings Our model estimates that national BIR costs are reduced between 33% and 53% in Medicare-for-All style single-payer models and between 27% and 63% in various multi-payer models. Under a wide range of assumptions and sensitivity analyses, standardizing contracts generates larger savings with less variance than savings from single-payer strategies. Conclusion Although moving toward a single-payer system will reduce BIR costs, certain reforms to payer-provider contracts could generate at least as many administrative cost savings without radically reforming the entire health system. BIR costs can be meaningfully reduced without abandoning a multi-payer system.

Background The use of continuous glucose monitors (CGMs) is recommended as the standard of care by the American Diabetes Association for individuals with type 1 diabetes (T1D). Few hardware-agnostic, open-source, whole-population tools are available to facilitate the use of CGM data by clinicians such as physicians and certified diabetes educators. Objective This study aimed to develop a tool that identifies patients appropriate for contact using an asynchronous message through electronic medical records while minimizing the number of patients reviewed by a certified diabetes educator or physician using the tool. Methods We used consensus guidelines to develop timely interventions for diabetes excellence (TIDE), an open-source hardware-agnostic tool to analyze CGM data to identify patients with deteriorating glucose control by generating generic flags (eg, mean glucose [MG] >170 mg/dL) and personalized flags (eg, MG increased by >10 mg/dL). In a prospective 7-week study in a pediatric T1D clinic, we measured the sensitivity of TIDE in identifying patients appropriate for contact and the number of patients reviewed. We simulated measures of the workload generated by TIDE, including the average number of time in range (TIR) flags per patient per review period, on a convenience sample of eight external data sets, 6 from clinical trials and 2 donated by research foundations. Results Over the 7 weeks of evaluation, the clinical population increased from 56 to 64 patients. The mean sensitivity was 99% (242/245; SD 2.5%), and the mean reduction in the number of patients reviewed was 42.6% (182/427; SD 10.9%). The 8 external data sets contained 1365 patients with 30,017 weeks of data collected by 7 types of CGMs. The rates of generic and personalized TIR flags per patient per review period were, respectively, 0.15 and 0.12 in the data set with the lowest average MG (141 mg/dL) and 0.95 and 0.22 in the data set with the highest average MG (207 mg/dL). Conclusions TIDE is an open-source hardware-agnostic tool for personalized analysis of CGM data at the clinical population scale. In a pediatric T1D clinic, TIDE identified 99% of patients appropriate for contact using an asynchronous message through electronic medical records while reducing the number of patients reviewed by certified diabetes care and education specialists by 43%. For each of the 8 external data sets, simulation of the use of TIDE produced fewer than 0.25 personalized TIR flags per patient per review period. The use of TIDE to support telemedicine-based T1D care may facilitate sensitive and efficient guideline-based population health management.

BACKGROUND The use of continuous glucose monitors (CGMs) is recommended as the standard of care by the American Diabetes Association for individuals with type 1 diabetes (T1D). Few hardware-agnostic, open-source, whole-population tools are available to facilitate the use of CGM data by clinicians such as physicians and certified diabetes educators. OBJECTIVE This study aimed to develop a tool that identifies patients appropriate for contact using an asynchronous message through electronic medical records while minimizing the number of patients reviewed by a certified diabetes educator or physician using the tool. METHODS We used consensus guidelines to develop timely interventions for diabetes excellence (TIDE), an open-source hardware-agnostic tool to analyze CGM data to identify patients with deteriorating glucose control by generating generic flags (eg, mean glucose [MG] >170 mg/dL) and personalized flags (eg, MG increased by >10 mg/dL). In a prospective 7-week study in a pediatric T1D clinic, we measured the sensitivity of TIDE in identifying patients appropriate for contact and the number of patients reviewed. We simulated measures of the workload generated by TIDE, including the average number of time in range (TIR) flags per patient per review period, on a convenience sample of eight external data sets, 6 from clinical trials and 2 donated by research foundations. RESULTS Over the 7 weeks of evaluation, the clinical population increased from 56 to 64 patients. The mean sensitivity was 99% (242/245; SD 2.5%), and the mean reduction in the number of patients reviewed was 42.6% (182/427; SD 10.9%). The 8 external data sets contained 1365 patients with 30,017 weeks of data collected by 7 types of CGMs. The rates of generic and personalized TIR flags per patient per review period were, respectively, 0.15 and 0.12 in the data set with the lowest average MG (141 mg/dL) and 0.95 and 0.22 in the data set with the highest average MG (207 mg/dL). CONCLUSIONS TIDE is an open-source hardware-agnostic tool for personalized analysis of CGM data at the clinical population scale. In a pediatric T1D clinic, TIDE identified 99% of patients appropriate for contact using an asynchronous message through electronic medical records while reducing the number of patients reviewed by certified diabetes care and education specialists by 43%. For each of the 8 external data sets, simulation of the use of TIDE produced fewer than 0.25 personalized TIR flags per patient per review period. The use of TIDE to support telemedicine-based T1D care may facilitate sensitive and efficient guideline-based population health management.

Background: The emergence of novel, partially vaccine-resistant strains of SARS-CoV-2 poses a serious risk to public health. The interactions between passengers and drivers facilitated by rideshare platforms, such as Uber and Lyft, are essentially a series of partially standardized, pseudo-random experiments of SARS-CoV-2 transmission. Rideshare companies share data with government health agencies, but no statistical method is available to aggregate these data for the systematic study of the transmission dynamics of COVID-19. Methods: We develop a proof-of-concept method for the analysis of data from rideshare interactions merged with COVID-19 diagnosis records. Using simulated data with rideshare volumes, disease prevalence, and diagnosis rates based on a large US city, we test hypotheses about the emergence of viral strains and their transmission characteristics in the presence of non-pharmaceutical interventions and superspreaders. Findings: Data from ten simulated trials of SARS-CoV-2 propagation within the Los Angeles rideshare network resulted in an average of 190,387·1 potentially infectious rideshare interactions. Assuming access to data on 25% of the total estimated infections (Partial Reporting), these interactions resulted in an average of 409·0 diagnosed rideshare infections given our transmission model assumptions. For each of the ten simulated trials, analysis given Partial Reporting could consistently differentiate between a baseline strain and an emergent, more infectious viral strain, enabling hypothesis testing about transmission characteristics. Interpretation: Simulated evaluation of a novel statistical method suggests that rideshare data combined with COVID-19 diagnosis data have the potential to automate continued surveillance of emergent novel strains of SARS-CoV-2 and their transmission characteristics. Funding Information: Stanford University, Stanford. Declaration of Interests: The authors have no relationships or disclosures relevant to the contents of this paper. All authors have no conflicts of interest, financial or otherwise.

Introduction: Patients with atherosclerotic cardiovascular disease (ASCVD) have high risk for recurrent ASCVD events despite statin use. Pooled cohort equations (PCE) are used for ASCVD risk prediction in primary prevention but there are no validated models for recurrent risk prediction in secondary prevention. Machine learning (ML) demonstrates promise in developing novel risk prediction models using electronic health record (EHR) data. Methods: We included adults with prior ASCVD from EHR data from an outpatient Northern California system between January 1, 2009 and December 31, 2018 with at least 2 visits at least 1 year apart and 5 years of follow up. The outcome was a recurrent ASCVD event defined as the first myocardial infarction, stroke, or fatal coronary artery disease in the 5 year follow-up period. We trained ML models to predict recurrent ASCVD risk: random forests (RF), gradient boosted machines (GBM), extreme gradient boosted models (XGBoost), and logistic regression with a standard L 2 penalty (LR) and an L 1 penalty (Lasso). We evaluated performance of ML models and the PCE on a 20% held-out test cohort using the areas under the receiver operating characteristic curves (AUCs). Results: Our cohort consisted of 32,192 patients with ASCVD (Mean age 70 years, 46% women, 12% Asian and 6% Hispanic). Less than half (49%) were on guideline directed statins. XGBoost and GBM were the best performing models for recurrent ASCVD risk prediction, while the PCE performed poorly (Figure). The top 20 predictive variables for recurrent ASCVD risk included prior events (ischemic stroke, myocardial infarction), traditional risk factors (age, blood pressure, lipid levels) and socioeconomic factors (income, education). Conclusions: EHR-trained machine learning models facilitated recurrent ASCVD risk prediction in real-world secondary prevention patients. Machine learning models developed from large datasets may help bridge contemporary gaps in ASCVD risk prediction.

Introduction: Optimal statin treatment decisions for primary prevention of atherosclerotic cardiovascular disease (ASCVD) rely on shared decision-making between patient and provider. We sought to develop a machine learning-based algorithm to personalize cholesterol treatment decisions using electronic medical record (EMR) data. Methods: We included EMR data for adults aged 40 to 79 with no prior ASCVD or statin therapy from an outpatient Northern California system between January 1, 2009 and December 31, 2018 with at least two visits at least 1 year apart and at least two low density lipoprotein cholesterol (LDL-C) values. The outcome was the LDL-C measured closest to one year after a patient’s second visit. We modeled four different treatment decisions: no statin use, low-intensity statin use, moderate-intensity statin use, and high-intensity statin use. We trained weighted-K-nearest-neighbor (wKNN) regression models to identify similar patients using each line of therapy to a candidate patient. The algorithm compared outcomes of these similar patients and recommended the treatment which predicted the lowest LDL-C after one year. Results: Our study cohort consisted of 50,911 patients (age 54.6 ± 9.84 years, baseline LDL-C 122 ± 34.2 mg/dL, follow-up LDL-C 121 ± 35.9 mg/dL) including 54% female, 47% Non-Hispanic White, 32% Asian, and 7.5% Hispanic patients. Among 8,551 test patients visiting in 2015 or later, 96.9%, 3.08%, and 0.05% were recommended to begin high-intensity, moderate-intensity, and low-intensity statins, respectively. With these recommendations, the LDL-C values at 1-year follow-up were predicted to be 21.5 ± 43.5 mg/dL (17.6%) lower per patient, on average (Figure). Conclusions: EMR-trained wKNN models are able to determine patient LDL-C trajectories under different lines of statin therapy. Machine learning models leveraging real-world datasets may provide useful statin therapy treatment recommendations for primary ASCVD prevention.

Introduction: Racial/ethnic diversity in clinical trials is essential to ensure that our evidence base reflects the population. We assessed the extent of reporting and representation of race/ethnicity in heart failure (HF) clinical trials referenced in the contemporary ACC/AHA HF guidelines. Methods: All randomized trials referenced in the 2013 ACC/AHA Heart Failure Guidelines and the 2017 Focused Update were included. The prevalence of reporting of race/ethnicity, the proportions of racial/ethnic subgroups enrolled, and subgroup analysis based on intervention type - pharmacologic, device, and other - were evaluated. Results: We identified 256 trials (545 233 subjects) published between 1950 and 2018. Among these, only 95 reported any race/ethnicity (37.1%), 94 reported white race (36.7%), 58 reported black race (22.7%), 16 reported Hispanic ethnicity (6.3%), and 23 reported Asian race (9.0%). In trials reporting white, black, Hispanic, and Asian race/ethnicity respectively, 76.4% (n = 299 153 of 299872) of patients were white, 11.7% (n = 25 274 of 215 905) of patients were black, 11.2% (n = of 8863 of 79 097) of patients were Hispanic, and 10.5% (n = 14925 of 141 504) of patients were Asian. Comparison of trial population proportions with US Census population demonstrates over-representation of white subjects, and under-representation of Hispanic and black subjects (Figure). Stratification by intervention type demonstrated that no device trials referenced in the guidelines report black or Asian race, and just one reported Hispanic race. Conclusions: Trials that dictate clinical care of patients with HF through informing contemporary ACC/AHA HF guidelines under-represent black and Hispanic populations. Additionally, 2/3rds of trials fail to report any race/ ethnicity at all. There is a need for guideline and practice-informing clinical trials to adequately represent all populations, and to provide clinicians the data they need to assess generalizability.

The pooled cohort equations (PCE) predict atherosclerotic cardiovascular disease (ASCVD) risk in patients with characteristics within prespecified ranges and has uncertain performance among Asians or Hispanics. It is unknown if machine learning (ML) models can improve ASCVD risk prediction across broader diverse, real-world populations. We developed ML models for ASCVD risk prediction for multi-ethnic patients using an electronic health record (EHR) database from Northern California. Our cohort included patients aged 18 years or older with no prior CVD and not on statins at baseline (n = 262,923), stratified by PCE-eligible (n = 131,721) or PCE-ineligible patients based on missing or out-of-range variables. We trained ML models [logistic regression with L2 penalty and L1 lasso penalty, random forest, gradient boosting machine (GBM), extreme gradient boosting] and determined 5-year ASCVD risk prediction, including with and without incorporation of additional EHR variables, and in Asian and Hispanic subgroups. A total of 4309 patients had ASCVD events, with 2077 in PCE-ineligible patients. GBM performance in the full cohort, including PCE-ineligible patients (area under receiver-operating characteristic curve (AUC) 0.835, 95% confidence interval (CI): 0.825–0.846), was significantly better than that of the PCE in the PCE-eligible cohort (AUC 0.775, 95% CI: 0.755–0.794). Among patients aged 40–79, GBM performed similarly before (AUC 0.784, 95% CI: 0.759–0.808) and after (AUC 0.790, 95% CI: 0.765–0.814) incorporating additional EHR data. Overall, ML models achieved comparable or improved performance compared to the PCE while allowing risk discrimination in a larger group of patients including PCE-ineligible patients. EHR-trained ML models may help bridge important gaps in ASCVD risk prediction.

Background We report hospitalization patterns from 2000 to 2016 for young children (ages 0-5 years old) in California who underwent 1 of the 20 most common inpatient procedures that required general anesthesia and evaluate the estimated probability of treatment at a tertiary care children's hospital (CH) by year. Methods We hypothesized that children ≤5 years old increasingly undergo care at tertiary care CHs for common inpatient surgeries or other procedures that require general anesthesia. Data from the California Office of Statewide Health Planning and Development dataset were used to determine procedure, patient age, year of procedure, and hospital name. Hospitals were designated as either tertiary care CHs, children's units within general hospitals (CUGHs), or general hospitals (GHs) based on the California Children's Services Provider List. A tertiary care CH was defined using the California Children's Services definition as a referral hospital that provides comprehensive, multidisciplinary, regionalized pediatric care to children from birth up to 21 years of age with a full range of medical and surgical care for severely ill children. We report the unadjusted percentage of patients treated at each hospital type and, after controlling for patient covariates and comorbidities, the estimated probability of undergoing care at a tertiary care CH from 2000 to 2016. Results There were 172,318 treatment episodes from 2000 to 2016. The estimated probability of undergoing care at a tertiary care CH increased from 63.4% (95% confidence interval [CI], 62.4%-64.4%) in 2000 to 78.3% (95% CI, 77.3%-79.4%) in 2016. Conclusions Children ≤5 years old undergoing common inpatient procedures that require general anesthesia increasingly receive care at tertiary care CHs in California.