Short-Term Out-of-Pocket and Total Costs of Care After Ablation, Resection, or Transplant for Early-Stage Hepatocellular Carcinoma: A National SEER-Medicare Cost Comparison.

BACKGROUND. Differences in survival and morbidity among treatment options (ablation, surgical resection, and transplant) for early-stage hepatocellular carcinoma (HCC) have been well studied. Additional understanding of the costs of such care would help to identify drivers of high costs and potential barriers to care delivery. OBJECTIVE. The purpose of this article was to quantify total and patient out-of-pocket costs for ablation, surgical resection, and transplant in the management of early-stage HCC and to identify factors predictive of these costs. METHODS. This retrospective U.S. population-based study used the SEER-Medicare linked dataset to identify a sample of 1067 Medicare beneficiaries (mean age, 73 years; 674 men, 393 women) diagnosed with early-stage HCC (size ≤ 5 cm) treated with ablation (n = 623), resection (n = 201), or transplant (n = 243) between January 2009 and December 2016. Total costs and patient out-of-pocket costs for the index procedure as well as for any care within 30 and 90 days after the procedure were identified and stratified by treatment modality. Additional comparisons were performed among propensity score-matched subgroups of patients treated by ablation or resection (each n = 172). Multivariable linear regression models were used to identify factors predictive of total costs and out-of-pocket costs for index procedures as well as for 30- and 90-day post-procedure periods. RESULTS. For ablation, resection, and transplant, median index-procedure total cost was US$6689, US$25,614, and US$66,034; index-procedure out-of-pocket cost was US$1235, US$1650, and US$1317; 30-day total cost was US$9456, US$29,754, and US$69,856; 30-day out-of-pocket cost was US$1646, US$2208, and US$3198; 90-day total cost was US$14,572, US$34,984, and US$88,103; and 90-day out-of-pocket cost was US$2138, US$2462, and US$3876, respectively (all p < .001). In propensity score-matched subgroups, ablation and resection had median index-procedure, 30-day, and 90-day total costs of US$6690 and US$25,716, US$9995 and US$30,365, and US$15,851 and US$34,455, respectively. In multivariable analysis adjusting for socioeconomic factors, comorbidities, and liver-disease prognostic indicators, surgical treatment (resection or transplant) was predictive of significantly greater costs compared with ablation at all time points. CONCLUSION. Total and out-of-pocket costs for index procedures as well as for 30-day and 90-day postprocedure periods were lowest for ablation, followed by resection and then transplant. CLINICAL IMPACT. This comprehensive cost analysis could help inform future cost-effectiveness analyses.

Cost-Effectiveness of Screening Asymptomatic Carotid Stenosis by Atherosclerotic Cardiovascular Risk.

Importance: Extracranial internal carotid artery stenosis (50-99% arterial narrowing) is an important risk factor for ischemic stroke. Yet, the benefits and harms of targeted screening for asymptomatic carotid artery stenosis (ACAS) have not been assessed in population-based studies. Objective: To estimate the cost-effectiveness of one-time, targeted ACAS screening stratified by atherosclerotic cardiovascular disease (ASCVD) risk using the American Heart Association's Pooled Cohort Equations. Design Setting and Participants: We developed a lifetime microsimulation model of ACAS and stroke for a hypothetical cohort representative of US adults aged 50-80 years without stroke history. We used the Cardiovascular Health Study to estimate the probability and severity of ACAS based on individual characteristics (e.g., age, sex, smoking status, blood pressure, and cholesterol). Stroke risks were functions of these characteristics and ACAS severity. In the model, individuals testing positive for >70% stenosis with Duplex ultrasound and a confirmatory diagnostic test undergo revascularization, which may reduce the risk of stroke but also introduces complication risks. Diagnostic performance parameters, revascularization benefits and risks, utility weights, and costs were estimated from published sources. Cost-effectiveness was assessed from the health care sector perspective using a $100,000/quality-adjusted life year (QALY) threshold. Main Outcomes and Measures: Estimated stroke events prevented, lifetime costs, QALYs, and incremental cost-effectiveness ratios (ICERs) associated with ACAS screening. Costs (2023 USD) and QALYs were discounted at 3% annually. Results: We found that screening individuals with a 10-year ASCVD risk >30% was the most cost-effective strategy, with an ICER of $89,000/QALY. This strategy would make approximately 11.9% of the population eligible for screening, averting an estimated 24,084 strokes. Results were sensitive to variations in the efficacy and complication risk of revascularization. In probabilistic sensitivity analysis, screening those in lower ASCVD risk groups (0-20%) only had a 0.6% chance of being cost-effective. Conclusion and Relevance: A one-time screening may only be cost-effective for adults at a relatively high ASCVD risk. Our findings provide a framework that can be adapted as future clinical trial data continue to improve our understanding of the role of revascularization and intensive medical therapy in contemporary stroke prevention secondary to carotid disease.

Application of a data continuity prediction algorithm to an electronic health record-based pharmacoepidemiology study.

BACKGROUND AND OBJECTIVES: Use of algorithms to identify patients with high data-continuity in electronic health records (EHRs) may increase study validity. Practical experience with this approach remains limited. METHODS: We developed and validated four algorithms to identify patients with high data continuity in an EHR-based data source. Selected algorithms were then applied to a pharmacoepidemiologic study comparing rates of COVID-19 hospitalization in patients exposed to insulin versus noninsulin antidiabetic drugs. RESULTS: A model using a short list of five EHR-derived variables performed as well as more complex models to distinguish high- from low-data continuity patients. Higher data continuity was associated with more accurate ascertainment of key variables. In the pharmacoepidemiologic study, patients with higher data continuity had higher observed rates of the COVID-19 outcome and a large unadjusted association between insulin use and the outcome, but no association after propensity score adjustment. DISCUSSION: We found that a simple, portable algorithm to predict data continuity gave comparable performance to more complex methods. Use of the algorithm significantly impacted the results of an empirical study, with evidence of more valid results at higher levels of data continuity.

Predictors for Early Liver Cancer Survival After Ablation and Surgical Resection: A Surveillance, Epidemiology, and End Results Program-Medicare Study.

OBJECTIVE: To identify independent predictors of all-cause and cancer-specific mortality after ablation or surgical resection (SR) for small hepatocellular carcinomas (HCCs), after adjusting for key confounders. METHODS: Using Surveillance, Epidemiology, and End Results Program-Medicare, HCCs less than 5 cm treated with ablation or SR in 2009 to 2016 (n = 956) were identified. Univariate and multivariable Cox regression models for all-cause and cancer-specific mortality were performed including demographics, clinical factors (tumor size, medical comorbidities, and liver disease factors), social determinants of health, and treatment characteristics. We also determined the most influential predictors of survival using a random forest analysis. RESULTS: Larger tumor size (3-5 cm) is predictive of all-cause (hazard ratio [HR] 1.31, P = .002) and cancer-specific mortality (HR 1.59, P < .001). Furthermore, chronic kidney disease is predictive of all-cause mortality (HR 1.43, P = .013), though it is not predictive of cancer-specific death. Multiple liver disease factors are predictive of all-cause and cancer-specific mortality including portal hypertension and esophageal varices (HRs > 1, P < .05). Though Asian race is protective in univariate models, in fully adjusted, multivariable models, Asian race is not a significant protective factor. Likewise, other social determinants of health are not significantly predictive of all-cause or cancer-specific mortality. Finally, treatment with SR, in later procedure years or at high-volume centers, is protective for all-cause and cancer-specific mortality. In machine learning models, year procedure was performed, ascites, portal hypertension, and treatment choice were the most influential factors. DISCUSSION: Treatment characteristics, liver disease factors, and tumor size are more important predictors of all-cause and cancer-specific death than social determinants of health for small HCCs.

Beyond Survival: Adverse Events and Care Delivery Outcomes after Early Liver Cancer Treatment in a Nationally Representative Cohort.

PURPOSE: To compare secondary outcomes after ablation (AB), surgical resection (SR), and liver transplant (LT) for small hepatocellular carcinomas (HCCs), including resource utilization and adverse event (AE) rates. MATERIALS AND METHODS: Using Surveillance, Epidemiology, and End Results Program (SEER)-Medicare, HCCs <5 cm that were treated with AB, SR, or LT in 2009-2016 (n = 1,067) were identified using Healthcare Common Procedure Coding System codes through Medicare claims. Index procedure length of stay, need for intensive care unit (ICU) level care, readmission rates, and AE rates at 30 and 90 days were compared using chi-square tests or Fisher exact tests. Examined AEs included hemorrhage, abscess formation, biliary injury, pneumonia, sepsis, liver disease-related AEs, liver failure, and anesthesia-related AEs, identified by International Classification of Diseases, Ninth/10th Revision, codes. RESULTS: The median length of stay for initial treatment was 1 day, 6 days, and 7 days for AB, SR, and LT, respectively (P < .001). During initial hospital stay, 5.0%, 40.8%, and 63.4% of AB, SR, and LT cohorts, respectively, received ICU-level care (P < .001). By 30 and 90 days, there were significant differences among the AB, SR, and LT cohorts in the rate of postprocedural hemorrhage, abscess formation, biliary injury, pneumonia, sepsis, liver disease-related AEs, and anesthesia-related AEs (P < .05). By 90 days, the readmission rates after AB, SR, and LT were 18.6%, 28.2%, and 40.6% (P < .001), respectively. CONCLUSIONS: AB results in significantly less healthcare utilization during the initial 90 days after procedure compared with that after SR and LT due to shorter length of stay, lower intensity care, fewer readmissions, and fewer AEs.

Prioritizing Quality Measures in Acute Stroke Care : A Cost-Effectiveness Analysis.

BACKGROUND: The American Heart Association and American Stroke Association (AHA/ASA) endorsed 15 process measures for acute ischemic stroke (AIS) to improve the quality of care. Identifying the highest-value measures could reduce the administrative burden of quality measure adoption while retaining much of the value of quality improvement. OBJECTIVE: To prioritize AHA/ASA-endorsed quality measures for AIS on the basis of health impact and cost-effectiveness. DESIGN: Individual-based stroke simulation model. DATA SOURCES: Published literature. TARGET POPULATION: U.S. patients with incident AIS. TIME HORIZON: Lifetime. PERSPECTIVE: Health care sector. INTERVENTION: Current versus complete (100%) implementation at the population level of quality measures endorsed by the AHA/ASA with sufficient clinical evidence (10 of 15). OUTCOME MEASURES: Life-years, quality-adjusted life-years (QALYs), incremental cost-effectiveness ratios, and incremental net health benefits. RESULTS OF BASE-CASE ANALYSIS: Discounted life-years gained from complete implementation would range from 472 (tobacco use counseling) to 34 688 (early carotid imaging) for an annual AIS patient cohort. All AIS quality measures were cost-saving or highly cost-effective by AHA standards (<$50 000 per QALY for high-value care). Early carotid imaging and intravenous tissue plasminogen activator contributed the largest fraction of the total potential value of quality improvement (measured as incremental net health benefit), accounting for 72% of the total value. The top 5 quality measures accounted for 92% of the total potential value. RESULTS OF SENSITIVITY ANALYSIS: A web-based user interface allows for context-specific sensitivity and scenario analyses. LIMITATION: Correlations between quality measures were not incorporated. CONCLUSION: Substantial variation exists in the potential net benefit of quality improvement across AIS quality measures. Benefits were highly concentrated among 5 of 10 measures assessed. Our results can help providers and payers set priorities for quality improvement efforts and value-based payments in AIS care. PRIMARY FUNDING SOURCE: National Institute of Neurological Disorders and Stroke.

Long-Term Survival After Surgery Versus Ablation for Early Liver Cancer in a Large, Nationally Representative Cohort.

OBJECTIVE: To compare survival outcomes (all-cause, cancer-specific, and disease-free) for small hepatocellular carcinomas (HCCs), less than or equal to 5 cm, after ablation (AB) and surgical resection (SR) after adjusting for key confounders. Secondarily, to understand differential survival outcomes of liver transplant (TR) compared with SR and AB. METHODS: Using Surveillance, Epidemiology, and End Results Program-Medicare, HCCs less than 5 cm that were treated with AB, SR, or TR in 2009 to 2016 (n = 1,215) were identified using Healthcare Common Procedure Coding System codes through Medicare claims. The TR group was subdivided into two groups: TR with prior treatment and TR without prior treatment. All-cause survival, cancer-specific survival, and disease-free survival were analyzed using Kaplan-Meier curves and compared between groups using log-rank tests and Cox regression analyses. Propensity score-matched comparison of AB and SR groups was performed, with groups matched on demographics, social determinants of health, medical comorbidities, and liver disease severity prognostic indicators. RESULTS: Median study follow-up time was 2.71 years (interquartile range 1.25-3.83). Unadjusted 1-, 3-, and 5-year cancer-specific survivals were 85.9%, 67.6%, and 56.3% for the AB group; 91.7%, 82.6%, and 81.7% for the SR group; 93.5%, 88.7%, and 79.4% for TR without prior treatment group; and 96.4%, 93.2%, and 93.2% for TR with prior treatment group (P < .0001). With SR as the reference group, the propensity-matched hazard ratios for AB were 2.04 (95% confidence interval: 1.51-2.77) for all-cause mortality, 2.44 (95% confidence interval: 1.56-3.80) for cancer-specific mortality, and 2.12 (95% confidence interval: 1.61-2.78) for disease recurrence. DISCUSSION: SR is superior to AB for small HCCs in a large, nationally representative, modern cohort, and in secondary analysis TR was superior to both.

Baseline haemoglobin A1c and the risk of COVID-19 hospitalization among patients with diabetes in the INSIGHT Clinical Research Network.

AIMS: To examine the association between baseline glucose control and risk of COVID-19 hospitalization and in-hospital death among patients with diabetes. METHODS: We performed a retrospective cohort study of adult patients in the INSIGHT Clinical Research Network with a diabetes diagnosis and haemoglobin A1c (HbA1c) measurement in the year prior to an index date of March 15, 2020. Patients were divided into four exposure groups based on their most recent HbA1c measurement (in mmol/mol): 39-46 (5.7%-6.4%), 48-57 (6.5%-7.4%), 58-85 (7.5%-9.9%), and ≥86 (10%). Time to COVID-19 hospitalization was compared in the four groups in a propensity score-weighted Cox proportional hazards model adjusting for potential confounders. Patients were followed until June 15, 2020. In-hospital death was examined as a secondary outcome. RESULTS: Of 168,803 patients who met inclusion criteria; 50,016 patients had baseline HbA1c 39-46 (5.7%-6.4%); 54,729 had HbA1c 48-57 (6.5-7.4%); 47,640 had HbA1c 58-85 (7.5^%-9.9%) and 16,418 had HbA1c ≥86 (10%). Compared with patients with HbA1c 48-57 (6.5%-7.4%), the risk of hospitalization was incrementally greater for those with HbA1c 58-85 (7.5%-9.9%) (adjusted hazard ratio [aHR] 1.19, 95% confidence interval [CI] 1.06-1.34) and HbA1c ≥86 (10%) (aHR 1.40, 95% CI 1.19-1.64). The risk of COVID-19 in-hospital death was increased only in patients with HbA1c 58-85 (7.5%-9.9%) (aHR 1.29, 95% CI 1.06, 1.61). CONCLUSIONS: Diabetes patients with high baseline HbA1c had a greater risk of COVID-19 hospitalization, although association between HbA1c and in-hospital death was less consistent. Preventive efforts for COVID-19 should be focused on diabetes patients with poor glucose control.

Impact of New York State's Health Home program on access to care among patients with diabetes.

INTRODUCTION: Access to care is essential for patients with diabetes to maintain health and prevent complications, and is important for health equity. New York State's Health Homes (HHs) provide care management services to Medicaid-insured patients with chronic conditions, including diabetes, and aim to improve quality of care and outcomes. There is inconsistent evidence on the impact of HHs, and care management programs more broadly, on access to care. RESEARCH DESIGN AND METHODS: Using a cohort of patients with diabetes derived from electronic health records from the INSIGHT Clinical Research Network, we analyzed Medicaid data for HH enrollees and a matched comparison group of HH non-enrollees. We estimated HH impacts on several access measures using natural experiment methods. RESULTS: We identified and matched 11 646 HH enrollees; patients were largely non-Hispanic Black (29.9%) and Hispanic (48.7%), and had high rates of dual eligibility (33.0%), Supplemental Security Income disability enrollment (49.1%), and multiple comorbidities. In the 12 months following HH enrollment, HH enrollees had one more month of Medicaid coverage (p<0.001) and 4.6 more outpatient visits than expected (p<0.001, evenly distributed between primary and specialty care). There were also positive impacts on the proportions of patients with follow-up visits within 7 days (4 percentage points (pp), p<0.001) and 30 days (6pp, p<0.001) after inpatient care, and on the proportion of patients with follow-up visits within 30 days after emergency department (ED) care (4pp, p<0.001). We did not find meaningful differences in continuity of care. We found small positive impacts on the proportion of patients with an inpatient visit and the proportion with an ED visit. CONCLUSIONS: New York State's HH program improved access to care for Medicaid recipients with diabetes. These findings have implications for New York State Medicaid as well as other providers and care management programs.

Safety and effectiveness of metformin in patients with reduced renal function: A systematic review.

AIM: To examine clinical and safety outcomes associated with metformin use in patients with impaired renal function. MATERIALS AND METHODS: We searched PubMed and Embase databases from inception to August 2020, supplementing our search with a review of investigator files and reference lists of included studies. Any study reporting original data on metformin and patient-centred outcomes in patients with impaired renal function, defined as an estimated glomerular filtration rate (eGFR) of less than 60 mL/min/1.73m2 , was included. Post hoc meta-analysis was performed for the outcomes of mortality, cardiovascular events and acidosis. RESULTS: Nine small prospective studies enrolling patients with significantly impaired renal function identified only one case of clinically apparent lactic acidosis. Among 13 larger retrospective studies, seven examined the risk of mortality across patient subgroups; meta-analysis showed reductions in overall mortality at an eGFR of 45 mL/min/1.73m2 or higher but not at an eGFR of less than 45 mL/min/1.73m2 . Eight retrospective studies evaluated acidosis as an outcome; meta-analysis showed no increase in risk of acidosis except at an eGFR of less than 30 mL/min/1.73m2 , in which group the HR was 1.97 (95% CI 1.03-3.77). CONCLUSIONS: The literature shows metformin to be associated with reduced mortality and no increased risk of acidosis at an eGFR of 45 mL/min/1.73m2 or higher. Metformin appears to be associated with fewer benefits and possible increases in the risk of acidosis at an eGFR of less than 30 mL/min/1.73m2 . Consistent with US Food and Drug Administration guidelines, metformin should not be used at an eGFR less than 30 mL/min/1.73m2 , and further research on its risk-benefit profile at eGFR values approaching 30 mL/min/1.73m2 is warranted.

Cost-Effectiveness Study of Initial Imaging Selection in Acute Ischemic Stroke Care.

PURPOSE: National guidelines recommend prompt identification of candidates for acute ischemic stroke (AIS) treatment, requiring timely neuroimaging with CT and/or MRI. CT is often preferred because of its widespread availability and rapid acquisition. Despite higher diagnostic accuracy of MRI, it commonly involves complex workflows that could potentially cause treatment time delays. The purpose of this study was to analyze the impact on outcomes of imaging utilization before treatment decisions at comprehensive stroke centers for patients presenting with suspected AIS in the anterior circulation with last-known-well-to-arrival time 0 to 24 hours. METHODS: A decision simulation model based on the American Heart Association's recommendations for AIS care pathways was developed from a health care perspective to compare initial imaging strategies: (1) stepwise-CT: noncontrast CT (NCCT) at the time of presentation, with CT angiography (CTA) ± CT perfusion (CTP) only in select patients (initial imaging to exclude hemorrhage and extensive ischemia) for mechanical thrombectomy (MT) evaluation; (2) stepwise-hybrid: NCCT at the time of presentation, with MR angiography (MRA) ± MR perfusion (MRP) only for MT evaluation; (3) stepwise-advanced: NCCT + CTA at presentation, with MR diffusion-weighted imaging (MR DWI) + MRP only for MT evaluation; (4) comprehensive-CT: NCCT + CTA + CTP at the time of presentation; and (5) comprehensive-MR: MR DWI + MRA + MRP at the time of presentation. Model parameters were defined using evidence-based data. Cost-effectiveness and sensitivity analyses were performed. RESULTS: The cost-effectiveness analyses revealed that comprehensive-CT and comprehensive-MR yield the highest lifetime quality-adjusted life-years (QALYs) (4.81 and 4.82, respectively). However, the incremental cost-effectiveness ratio of comprehensive-MR is $233,000/QALY compared with comprehensive-CT. Stepwise-CT, stepwise-hybrid, and stepwise-advanced strategies are dominated, yielding lower QALYs and higher costs compared with comprehensive-CT. CONCLUSIONS: Performing comprehensive-CT at presentation is the most cost-effective initial imaging strategy at comprehensive stroke centers.

Can Pay-for Performance Incentive Levels be Determined Using a Cost-Effectiveness Framework?

BACKGROUND: Healthcare payers in the United States are increasingly tying provider payments to quality and value using pay-for-performance policies. Cost-effectiveness analysis quantifies value in healthcare but is not currently used to design or prioritize pay-for-performance strategies or metrics. Acute ischemic stroke care provides a useful application to demonstrate how simulation modeling can be used to determine cost-effective levels of financial incentives used in pay-for-performance policies and associated challenges with this approach. METHODS AND RESULTS: Our framework requires a simulation model that can estimate quality-adjusted life years and costs resulting from improvements in a quality metric. A monetary level of incentives can then be back-calculated using the lifetime discounted quality-adjusted life year (which includes effectiveness of quality improvement) and cost (which includes incentive payments and cost offsets from quality improvements) outputs from the model. We applied this framework to an acute ischemic stroke microsimulation model to calculate the difference in population-level net monetary benefit (willingness-to-pay of $50 000 to $150 000/quality-adjusted life year) accrued under current Medicare policy (stroke payment not adjusted for performance) compared with various hypothetical pay-for-performance policies. Performance measurement was based on time-to-thrombolytic treatment with tPA (tissue-type plasminogen activator). Compared with current payment, equivalent population-level net monetary benefit was achieved in pay-for-performance policies with 10-minute door-to-needle time reductions (5057 more acute ischemic stroke cases/y in the 0-3-hour window) incentivized by increasing tPA payment by as much as 18% to 44% depending on willingness-to-pay for health. CONCLUSIONS: Cost-effectiveness modeling can be used to determine the upper bound of financial incentives used in pay-for-performance policies, although currently, this approach is limited due to data requirements and modeling assumptions. For tPA payments in acute ischemic stroke, our model-based results suggest financial incentives leading to a 10-minute decrease in door-to-needle time should be implemented but not exceed 18% to 44% of current tPA payment. In general, the optimal level of financial incentives will depend on willingness-to-pay for health and other modeling assumptions around parameter uncertainty and the relationship between quality improvements and long-run quality-adjusted life expectancy and costs.

Overcoming barriers to the use of metformin: patient and provider perspectives.

BACKGROUND: Metformin is the first-line treatment for type 2 diabetes mellitus. Observational studies show lower rates of use than would be expected from clinical guidelines. OBJECTIVE: We undertook a qualitative study of barriers to metformin use from the patient and provider perspective. DESIGN: Patient focus group, patient individual interviews, provider interviews, and chart review. Purposive sampling of patients and providers in New York State. PARTICIPANTS: Seven one-on-one patient interviews, one focus group with 13 patients, 10 provider interviews, and manual review of 1259 charts. APPROACH: Grounded theory. RESULTS: Both providers and patients cited potential health benefits as strong reasons to take metformin and describe it as the first-line drug for diabetes. Patients and providers both cited gastrointestinal side effects as the primary barrier to metformin use. Patients described adapting to these side effects and taking metformin at a time that minimizes them. In contrast, providers emphasized dose titration and the use of extended-release formulation as tools for minimizing side effects but were uncertain about the effectiveness of these strategies. CONCLUSION: Metformin is positively viewed by patients and providers, but gastrointestinal side effects are a barrier to its use. There is some consensus on basic measures to improve its tolerability, but these measures are not consistently applied and lack evidence establishing their effectiveness. Pragmatic clinical trial research on optimal dose, formulation, and counseling for new metformin users should be considered.

Cost Effectiveness of Assessing Ultrasound Plaque Characteristics to Risk Stratify Asymptomatic Patients With Carotid Stenosis.

Background Imaging may play an important role in identifying high-risk plaques in patients who have carotid disease and who could benefit from surgical revascularization. We sought to evaluate the cost effectiveness of a decision-making rule based on the ultrasound imaging assessment of plaque echolucency in patients with asymptomatic carotid stenosis. Methods and Results We used a decision-analytic model to project lifetime quality-adjusted life years and costs for 5 stroke prevention strategies: (1) medical therapy only; (2) revascularization if both plaque echolucency and stenosis progression to >90% are present; (3) revascularization only if plaque echolucency is present; (4) revascularization only if stenosis progression >90% is present; or (5) either plaque echolucency or stenosis progression is present. Risks of clinical events, costs, and quality-of-life values were estimated based on published sources and the analysis was conducted from a healthcare system perspective for asymptomatic patients with 70% to 89% carotid stenosis at presentation. Patients who did not undergo revascularization had the highest stroke events (17.6%) and lowest life-years (8.45), while those who underwent revascularization on the basis of either presence of plaque echolucency on ultrasound or progression of carotid stenosis had the lowest stroke events (12.0%) and longest life-years (14.41). The either plaque echolucency or progression-based revascularization group had an incremental cost-effectiveness ratio of $110 000/quality-adjusted life years compared with the plaque echolucency-based strategy, which had an incremental cost-effectiveness ratio of $29 000/quality-adjusted life years compared with the joint echolucency and progression-based strategy. Conclusions Plaque echolucency on ultrasound can be a cost-effective tool to identify patients with asymptomatic carotid artery stenosis most likely to benefit from carotid endarterectomy.

Development and Performance of the Pulmonary Embolism Result Forecast Model (PERFORM) for Computed Tomography Clinical Decision Support.

Importance: Pulmonary embolism (PE) is a life-threatening clinical problem, and computed tomographic imaging is the standard for diagnosis. Clinical decision support rules based on PE risk-scoring models have been developed to compute pretest probability but are underused and tend to underperform in practice, leading to persistent overuse of CT imaging for PE. Objective: To develop a machine learning model to generate a patient-specific risk score for PE by analyzing longitudinal clinical data as clinical decision support for patients referred for CT imaging for PE. Design, Setting, and Participants: In this diagnostic study, the proposed workflow for the machine learning model, the Pulmonary Embolism Result Forecast Model (PERFORM), transforms raw electronic medical record (EMR) data into temporal feature vectors and develops a decision analytical model targeted toward adult patients referred for CT imaging for PE. The model was tested on holdout patient EMR data from 2 large, academic medical practices. A total of 3397 annotated CT imaging examinations for PE from 3214 unique patients seen at Stanford University hospitals and clinics were used for training and validation. The models were externally validated on 240 unique patients seen at Duke University Medical Center. The comparison with clinical scoring systems was done on randomly selected 100 outpatient samples from Stanford University hospitals and clinics and 101 outpatient samples from Duke University Medical Center. Main Outcomes and Measures: Prediction performance of diagnosing acute PE was evaluated using ElasticNet, artificial neural networks, and other machine learning approaches on holdout data sets from both institutions, and performance of models was measured by area under the receiver operating characteristic curve (AUROC). Results: Of the 3214 patients included in the study, 1704 (53.0%) were women from Stanford University hospitals and clinics; mean (SD) age was 60.53 (19.43) years. The 240 patients from Duke University Medical Center used for validation included 132 women (55.0%); mean (SD) age was 70.2 (14.2) years. In the samples for clinical scoring system comparisons, the 100 outpatients from Stanford University hospitals and clinics included 67 women (67.0%); mean (SD) age was 57.74 (19.87) years, and the 101 patients from Duke University Medical Center included 59 women (58.4%); mean (SD) age was 73.06 (15.3) years. The best-performing model achieved an AUROC performance of predicting a positive PE study of 0.90 (95% CI, 0.87-0.91) on intrainstitutional holdout data with an AUROC of 0.71 (95% CI, 0.69-0.72) on an external data set from Duke University Medical Center; superior AUROC performance and cross-institutional generalization of the model of 0.81 (95% CI, 0.77-0.87) and 0.81 (95% CI, 0.73-0.82), respectively, were noted on holdout outpatient populations from both intrainstitutional and extrainstitutional data. Conclusions and Relevance: The machine learning model, PERFORM, may consider multitudes of applicable patient-specific risk factors and dependencies to arrive at a PE risk prediction that generalizes to new population distributions. This approach might be used as an automated clinical decision-support tool for patients referred for CT PE imaging to improve CT use.