Decision-analytic Modeling Studies Research Articles

Community-Based Participatory Research and System Dynamics Modeling for Improving Retention in Hypertension Care

The high prevalence of hypertension calls for broad, multisector responses that foster prevention and care services, with the goal of leveraging high-quality treatment as a means of reducing hypertension incidence. Health care system improvements require stakeholder input from across the care continuum to identify gaps and inform interventions that improve hypertension care service, delivery, and retention; system dynamics modeling offers a participatory research approach through which stakeholders learn about system complexity and ways to model sustainable system-level improvements. To assess the association of simulated interventions with hypertension care retention rates in the Nigerian primary health care system using system dynamics modeling. This decision analytical model used a participatory research approach involving stakeholder workshops conducted in July and October 2022 to gather insights and inform the development of a system dynamics model designed to simulate the association of various interventions with retention in hypertension care. The study focused on the primary health care system in Nigeria, engaging stakeholders from various sectors involved in hypertension care, including patients, community health extension workers, nurses, pharmacists, researchers, administrators, policymakers, and physicians. Simulated intervention packages. Retention rate in hypertension care at 12, 24, and 36 months, modeled to estimate the effectiveness of the interventions. A total of 16 stakeholders participated in the workshops (mean [SD] age, 46.5 [8.6] years; 9 [56.3%] male). Training of health care workers was estimated to be the most effective single implementation strategy for improving retention in hypertension care in Nigeria, with estimated retention rates of 29.7% (95% CI, 27.8%-31.2%) at 12 months and 27.1% (95% CI, 26.0%-28.3%) at 24 months. Integrated intervention packages were associated with the greatest improvements in hypertension care retention overall, with modeled retention rates of 72.4% (95% CI, 68.4%-76.4%), 68.1% (95% CI, 64.5%-71.7%), and 67.1% (95% CI, 64.5%-71.1%) at 12, 24, and 36 months, respectively. This decision analytical model study showed that community-based participatory research could be used to estimate the potential effectiveness of interventions for improving retention in hypertension care. Integrated intervention packages may be the most promising strategies.

Public Health Interventions and Overdose-Related Outcomes Among Persons With Opioid Use Disorder

Given the high number of opioid overdose deaths in the US and the complex epidemiology of opioid use disorder (OUD), systems models can serve as a tool to identify opportunities for public health interventions. To estimate the projected 3-year association between public health interventions and opioid overdose-related outcomes among persons with OUD. This decision analytical model used a simulation model of the estimated US population aged 12 years and older with OUD that was developed and analyzed between January 2019 and December 2023. The model was parameterized and calibrated using 2019 to 2020 data and used to estimate the relative change in outcomes associated with simulated public health interventions implemented between 2021 and 2023. Projected OUD and medications for OUD (MOUD) prevalence in 2023 and number of nonfatal and fatal opioid-involved overdoses among persons with OUD between 2021 and 2023. In a baseline scenario assuming parameters calibrated using 2019 to 2020 data remained constant, the model projected more than 16 million persons with OUD not receiving MOUD treatment and nearly 1.7 million persons receiving MOUD treatment in 2023. Additionally, the model projected over 5 million nonfatal and over 145 000 fatal opioid-involved overdoses among persons with OUD between 2021 and 2023. When simulating combinations of interventions that involved reducing overdose rates by 50%, the model projected decreases of up to 35.2% in nonfatal and 36.6% in fatal opioid-involved overdoses among persons with OUD. Interventions specific to persons with OUD not currently receiving MOUD treatment demonstrated the greatest reduction in numbers of nonfatal and fatal overdoses. Combinations of interventions that increased MOUD initiation and decreased OUD recurrence were projected to reduce OUD prevalence by up to 23.4%, increase MOUD prevalence by up to 137.1%, and reduce nonfatal and fatal opioid-involved overdoses among persons with OUD by 6.7% and 3.5%, respectively. In this decision analytical model study of persons with OUD, findings suggested that expansion of evidence-based interventions that directly reduce the risk of overdose fatality among persons with OUD, such as through harm reduction efforts, could engender the highest reductions in fatal overdoses in the short-term. Interventions aimed at increasing MOUD initiation and retention of persons in treatment projected considerable improvement in MOUD and OUD prevalence but could require a longer time horizon for substantial reductions in opioid-involved overdoses.

Evaluation of venous thromboembolism risk assessment models for hospital inpatients: the VTEAM evidence synthesis.

Pharmacological prophylaxis during hospital admission can reduce the risk of acquired blood clots (venous thromboembolism) but may cause complications, such as bleeding. Using a risk assessment model to predict the risk of blood clots could facilitate selection of patients for prophylaxis and optimise the balance of benefits, risks and costs. We aimed to identify validated risk assessment models and estimate their prognostic accuracy, evaluate the cost-effectiveness of different strategies for selecting hospitalised patients for prophylaxis, assess the feasibility of using efficient research methods and estimate key parameters for future research. We undertook a systematic review, decision-analytic modelling and observational cohort study conducted in accordance with Enhancing the QUAlity and Transparency Of health Research (EQUATOR) guidelines. NHS hospitals, with primary data collection at four sites. Medical and surgical hospital inpatients, excluding paediatric, critical care and pregnancy-related admissions. Prophylaxis for all patients, none and according to selected risk assessment models. Model accuracy for predicting blood clots, lifetime costs and quality-adjusted life-years associated with alternative strategies, accuracy of efficient methods for identifying key outcomes and proportion of inpatients recommended prophylaxis using different models. We identified 24 validated risk assessment models, but low-quality heterogeneous data suggested weak accuracy for prediction of blood clots and generally high risk of bias in all studies. Decision-analytic modelling showed that pharmacological prophylaxis for all eligible is generally more cost-effective than model-based strategies for both medical and surgical inpatients, when valuing a quality-adjusted life-year at £20,000. The findings were more sensitive to uncertainties in the surgical population; strategies using risk assessment models were more cost-effective if the model was assumed to have a very high sensitivity, or the long-term risks of post-thrombotic complications were lower. Efficient methods using routine data did not accurately identify blood clots or bleeding events and several pre-specified feasibility criteria were not met. Theoretical prophylaxis rates across an inpatient cohort based on existing risk assessment models ranged from 13% to 91%. Existing studies may underestimate the accuracy of risk assessment models, leading to underestimation of their cost-effectiveness. The cost-effectiveness findings do not apply to patients with an increased risk of bleeding. Mechanical thromboprophylaxis options were excluded from the modelling. Primary data collection was predominately retrospective, risking case ascertainment bias. Thromboprophylaxis for all patients appears to be generally more cost-effective than using a risk assessment model, in hospitalised patients at low risk of bleeding. To be cost-effective, any risk assessment model would need to be highly sensitive. Current evidence on risk assessment models is at high risk of bias and our findings should be interpreted in this context. We were unable to demonstrate the feasibility of using efficient methods to accurately detect relevant outcomes for future research. Further research should evaluate routine prophylaxis strategies for all eligible hospitalised patients. Models that could accurately identify individuals at very low risk of blood clots (who could discontinue prophylaxis) warrant further evaluation. This study is registered as PROSPERO CRD42020165778 and Researchregistry5216. This award was funded by the National Institute for Health and Care Research (NIHR) Health Technology Assessment programme (NIHR award ref: NIHR127454) and will be published in full in Health Technology Assessment; Vol. 28, No. 20. See the NIHR Funding and Awards website for further award information.

Estimated US Cancer Deaths Prevented With Increased Use of Lung, Colorectal, Breast, and Cervical Cancer Screening

Increased use of recommended screening could help achieve the Cancer Moonshot goal of reducing US cancer deaths. To estimate the number of cancer deaths that could be prevented with a 10-percentage point increase in the use of US Preventive Services Task Force (USPSTF)-recommended screening. This decision analytical model study is an extension of previous studies conducted for the USPSTF from 2018 to 2023. This study simulated contemporary cohorts of US adults eligible for lung, colorectal, breast, and cervical cancer screening. Annual low-dose computed lung tomography among eligible adults aged 50 to 80 years; colonoscopy every 10 years among adults aged 45 to 75 years; biennial mammography among female adults aged 40 to 74 years; and triennial cervical cytology screening among female adults aged 21 to 29 years, followed by human papillomavirus testing every 5 years from ages 30 to 65 years. Estimated number of cancer deaths prevented with a 10-percentage point increase in screening use, assuming screening commences at the USPSTF-recommended starting age and continues throughout the lifetime. Outcomes were presented 2 ways: (1) per 100 000 and (2) among US adults in 2021; and they were expressed among the target population at the age of screening initiation. For lung cancer, estimates were among those who will also meet the smoking eligibility criteria during their lifetime. Harms from increased uptake were also reported. A 10-percentage point increase in screening use at the age that USPSTF recommended screening commences was estimated to prevent 226 lung cancer deaths (range across models within the cancer site, 133-332 deaths), 283 (range, 263-313) colorectal cancer deaths, 82 (range, 61-106) breast cancer deaths, and 81 (1 model; no range available) cervical cancer deaths over the lifetimes of 100 000 persons eligible for screening. These rates corresponded with an estimated 1010 (range, 590-1480) lung cancer deaths prevented, 11 070 (range, 10 280-12 250) colorectal cancer deaths prevented, 1790 (range, 1330-2310) breast cancer deaths prevented, and 1710 (no range available) cervical cancer deaths prevented over the lifetimes of eligible US residents at the recommended age to initiate screening in 2021. Increased uptake was also estimated to generate harms, including 100 000 (range, 45 000-159 000) false-positive lung scans, 6000 (range, 6000-7000) colonoscopy complications, 300 000 (range, 295 000-302 000) false-positive mammograms, and 348 000 (no range available) colposcopies over the lifetime. In this decision analytical model study, a 10-percentage point increase in uptake of USPSTF-recommended lung, colorectal, breast, and cervical cancer screening at the recommended starting age was estimated to yield important reductions in cancer deaths. Achieving these reductions is predicated on ensuring equitable access to screening.

Comparing Active Surveillance and Watchful Waiting With Radical Treatment Using Machine Learning Models Among Patients With Prostate Cancer.

In 2021, 59.6% of low-risk patients with prostate cancer were under active surveillance (AS) as their first course of treatment. However, few studies have investigated AS and watchful waiting (WW) separately. The objectives of this study were to develop and validate a population-level machine learning model for distinguishing AS and WW in the conservative treatment group, and to investigate initial cancer management trends from 2004 to 2017 and the risk of chronic diseases among patients with prostate cancer with different treatment modalities. In a cohort of 18,134 patients with prostate adenocarcinoma diagnosed between 2004 and 2017, 1,926 patients with available AS/WW information were analyzed using machine learning algorithms with 10-fold cross-validation. Models were evaluated using performance metrics and Brier score. Cox proportional hazard models were used to estimate hazard ratios for chronic disease risk. Logistic regression models achieved a test area under the receiver operating curve of 0.73, F-score of 0.79, accuracy of 0.71, and Brier score of 0.29, demonstrating good calibration, precision, and recall values. We noted a sharp increase in AS use between 2004 and 2016 among patients with low-risk prostate cancer and a moderate increase among intermediate-risk patients between 2008 and 2017. Compared with the AS group, radical treatment was associated with a lower risk of prostate cancer-specific mortality but higher risks of Alzheimer disease, anemia, glaucoma, hyperlipidemia, and hypertension. A machine learning approach accurately distinguished AS and WW groups in conservative treatment in this decision analytical model study. Our results provide insight into the necessity to separate AS and WW in population-based studies.

Estimated Reductions in Opioid Overdose Deaths With Sustainment of Public Health Interventions in 4 US States

In 2021, more than 80 000 US residents died from an opioid overdose. Public health intervention initiatives, such as the Helping to End Addiction Long-term (HEALing) Communities Study (HCS), are being launched with the goal of reducing opioid-related overdose deaths (OODs). To estimate the change in the projected number of OODs under different scenarios of the duration of sustainment of interventions, compared with the status quo. This decision analytical model simulated the opioid epidemic in the 4 states participating in the HCS (ie, Kentucky, Massachusetts, New York, and Ohio) from 2020 to 2026. Participants were a simulated population transitioning from opioid misuse to opioid use disorder (OUD), overdose, treatment, and relapse. The model was calibrated using 2015 to 2020 data from the National Survey on Drug Use and Health, the US Centers for Disease Control and Prevention, and other sources for each state. The model accounts for reduced initiation of medications for OUD (MOUDs) and increased OODs during the COVID-19 pandemic. Increasing MOUD initiation by 2- or 5-fold, improving MOUD retention to the rates achieved in clinical trial settings, increasing naloxone distribution efforts, and furthering safe opioid prescribing. An initial 2-year duration of interventions was simulated, with potential sustainment for up to 3 additional years. Projected reduction in number of OODs under different combinations and durations of sustainment of interventions. Compared with the status quo, the estimated annual reduction in OODs at the end of the second year of interventions was 13% to 17% in Kentucky, 17% to 27% in Massachusetts, 15% to 22% in New York, and 15% to 22% in Ohio. Sustaining all interventions for an additional 3 years was estimated to reduce the annual number of OODs at the end of the fifth year by 18% to 27% in Kentucky, 28% to 46% in Massachusetts, 22% to 34% in New York, and 25% to 41% in Ohio. The longer the interventions were sustained, the better the outcomes; however, these positive gains would be washed out if interventions were not sustained. In this decision analytical model study of the opioid epidemic in 4 US states, sustained implementation of interventions, including increased delivery of MOUDs and naloxone supply, was found to be needed to reduce OODs and prevent deaths from increasing again.

Projected COVID-19 Mortality Reduction From Paxlovid Rollout

This decision analytical model study assesses projections of simulated effects of Paxlovid rollout on hospitalizations and mortality using 10 models.

Comparing Projected Fatal Overdose Outcomes and Costs of Strategies to Expand Community-Based Distribution of Naloxone in Rhode Island

In 2021, the state of Rhode Island distributed 10 000 additional naloxone kits compared with the prior year through partnerships with community-based organizations. To compare various strategies to increase naloxone distribution through community-based programs in Rhode Island to identify one most effective and efficient strategy in preventing opioid overdose deaths (OODs). In this decision analytical model study conducted from January 2016 to December 2022, a spatial microsimulation model with an integrated decision tree was developed and calibrated to compare the outcomes of alternative strategies for distributing 10 000 additional naloxone kits annually among all individuals at risk for opioid overdose in Rhode Island. Distribution of 10 000 additional naloxone kits annually, focusing on people who inject drugs, people who use illicit opioids and stimulants, individuals at various levels of risk for opioid overdose, or people who misuse prescription opioids vs no additional kits (status quo). Two expanded distribution implementation approaches were considered: one consistent with the current spatial distribution patterns for each distribution program type (supply-based approach) and one consistent with the current spatial distribution of individuals in each of the risk groups, assuming that programs could direct the additional kits to new geographic areas if required (demand-based approach). Witnessed OODs, cost per OOD averted (efficiency), geospatial health inequality measured by the Theil index, and between-group variance for OOD rates. A total of 63 131 simulated individuals were estimated to be at risk for opioid overdose in Rhode Island based on current population data. With the supply-based approach, prioritizing additional naloxone kits to people who use illicit drugs averted more witnessed OODs by an estimated mean of 18.9% (95% simulation interval [SI], 13.1%-30.7%) annually. Expanded naloxone distribution using the demand-based approach and focusing on people who inject drugs had the best outcomes across all scenarios, averting an estimated mean of 25.3% (95% SI, 13.1%-37.6%) of witnessed OODs annually, at the lowest mean incremental cost of $27 312 per OOD averted. Other strategies were associated with fewer OODs averted at higher costs but showed similar patterns of improved outcomes and lower unit costs if kits could be reallocated to areas with greater need. The demand-based approach reduced geospatial inequality in OOD rates in all scenarios compared with the supply-based approach and status quo. In this decision analytical model study, variations in the effectiveness, efficiency, and health inequality of the different naloxone distribution expansion strategies and approaches were identified. Future efforts should be prioritized for people at highest risk for overdose (those who inject drugs or use illicit drugs) and redirected toward areas with the greatest need. These findings may inform future naloxone distribution priority settings.

Population-Level Implications of Sodium-Glucose Cotransporter-2 Inhibitors for Heart Failure With Preserved Ejection Fraction in the US

The expansion of sodium-glucose cotransporter-2 (SGLT-2) inhibitor use in patients with heart failure (HF) and left ventricular ejection fraction (LVEF) more than 40% following the EMPEROR-Preserved (Empagliflozin Outcome Trial in Patients with Chronic Heart Failure with Preserved Ejection Fraction) and the DELIVER (Dapagliflozin Evaluation to Improve the Lives of Patients with Preserved Ejection Fraction Heart Failure) trials have major implications in the US. To quantify the estimated US population-level impact of reducing worsening HF events with SGLT-2 inhibitors in individuals with LVEF more than 40%. This decision analytical model study used self-reported HF data from the National Health and Nutritional Examination Survey from 2015 to 2018, which was weighted across the entire US population and subsequently mapped onto newly eligible LVEF distributions from the Get With The Guidelines-Heart Failure registry. All patients older than 18 years with HF from the National Health and Nutritional Examination Survey were grouped into the following categories: all LVEF and LVEF more than 40%. Numbers needed to treat estimations over 3 years were obtained for outcome measures from the EMPEROR-Reduced (Empagliflozin Outcome Trial in Patients with Chronic Heart Failure with Reduced Ejection Fraction), EMPEROR-Preserved, DAPA-HF (Dapagliflozin and Prevention of Adverse Outcomes in Heart Failure), and DELIVER trials. Worsening HF events (unplanned HF hospitalizations, urgent HF visits requiring intravenous therapy, or cardiovascular death). A projected 4 794 524 (95% CI, 3 997 363-5 591 684) adults (57% male; 67% White; mean age, 66 years) with HF would be eligible for SGLT-2 inhibitors. Of this total population, 2 619 248 (95% CI, 2 183 759-3 054 737) would be estimated as newly eligible with LVEF more than 40%. Based on estimates from the EMPEROR-Reduced/EMPEROR-Preserved and DAPA-HF/DELIVER trials, a projected 624 247 (95% CI, 520 457-728 037) to 627 124 (95% CI, 522 855-731 392) worsening HF events could be prevented across the LVEF spectrum with SGLT-2 inhibitors over 3 years, of which 232 589 (95% CI, 193 918-271 260) to 282 879 (95% CI, 235 846-329 912) events could be prevented in individuals with LVEF more than 40%. Moreover, an estimated 468 904 (95% CI, 390 942-546 867) to 499 110 (95% CI, 416 125-582 094) total HF hospitalizations could be prevented across the LVEF spectrum, of which 172 870 (95% CI, 144 128-201 613) to 231 018 (95% CI, 192 608-269 428) could be prevented in individuals with LVEF more than 40%. In addition to the proven benefit in HF with LVEF of 40% and less, optimal implementation of SGLT-2 inhibitor therapy for HF with LVEF more than 40% can potentially prevent/postpone an additional approximately 250 000 worsening HF events over 3 years in the US.

Modeling Optimal Clinical Thresholds for Elective Abdominal Hernia Repair in Patients With Cirrhosis

Patients with cirrhosis have increased risk of postoperative mortality. Several models have been developed to estimate this risk; however, current risk estimation scores cannot compare surgical risk with the risk of not operating. To identify clinical optimal thresholds to favor operative or nonoperative management for a common cirrhosis surgical scenario, the symptomatic abdominal hernia. This was a Markov cohort decision analytical modeling study evaluating elective surgery vs nonoperative management for a symptomatic abdominal hernia in a patient with cirrhosis. Transition probabilities and utilities were derived from the literature and from data using an established cirrhosis cohort in the Veterans Health Administration. Participants included patients who were referred to a surgery clinic for a symptomatic abdominal hernia. Data were obtained from patients diagnosed with cirrhosis between January 1, 2008 and December 31, 2018. Data were analyzed from January 1 to May 1, 2022. Expected quality-adjusted life-years (QALYs) were estimated for each pathway and iterated over baseline model for end-stage liver disease-sodium (MELD-Na) scores ranging from 6 to 25. Markov models were cycled over a 5-year time horizon. A total 2740 patients with cirrhosis (median [IQR] age, 62 [56-66] years; 2699 [98.5%] men) were referred to a surgery clinic for a symptomatic abdominal hernia; 1752 patients (63.9%) did not receive surgery. The median (IQR) follow-up was 42.1 (25.3-70.0) months. Using this cohort to estimate the mortality risk of operative and nonoperative pathways, an initial MELD-Na threshold of 21.3 points, below which surgery was associated with maximized QALYs was identified. Nonoperative management was associated with increased QALYs above this MELD-Na threshold. Although more patients experienced death with a surgical treatment decision across all initial MELD-Na values, this was counterbalanced by increased time spent in a resolved hernia state associated with increased utility. Model results were sensitive to the probability of hernia recurrence and hernia incarceration and utility decrement in the symptomatic hernia state. This decision analytical model study found that elective surgical treatment for a symptomatic abdominal hernia was favored even in the setting of relatively high MELD-Na scores. Patient symptoms, hernia-specific characteristics, and surgeon and center expertise may potentially impact the optimal strategy, emphasizing the importance of shared decision-making.

Handling Informative Premature Treatment or Study Discontinuation for Assessing Between-Group Differences in a Comparative Oncology Trial

This decision analytical model study examines premature treatment discontinuation in clinical trials for patients with advanced renal cell carcinoma.

Estimated COVID-19 Cases and Hospitalizations Averted by Case Investigation and Contact Tracing in the US

Evidence of the impact of COVID-19 case investigation and contact tracing (CICT) programs is lacking, but policy makers need this evidence to assess the value of such programs. To estimate COVID-19 cases and hospitalizations averted nationwide by US states' CICT programs. This decision analytical model study used combined data from US CICT programs (eg, proportion of cases interviewed, contacts notified or monitored, and days to case and contact notification) with incidence data to model outcomes of CICT over a 60-day period (November 25, 2020, to January 23, 2021). The study estimated a range of outcomes by varying assumed compliance with isolation and quarantine recommendations. Fifty-nine state and territorial health departments that received federal funding supporting COVID-19 pandemic response activities were eligible for inclusion. Data analysis was performed from July to September 2021. Public health case investigation and contact tracing. The primary outcomes were numbers of cases and hospitalizations averted and the percentage of cases averted among cases not prevented by vaccination and other nonpharmaceutical interventions. In total, 22 states and 1 territory reported all measures necessary for the analysis. These 23 jurisdictions covered 42.5% of the US population (approximately 140 million persons), spanned all 4 US Census regions, and reported data that reflected all 59 federally funded CICT programs. This study estimated that 1.11 million cases and 27 231 hospitalizations were averted by CICT programs under a scenario where 80% of interviewed cases and monitored contacts and 30% of notified contacts fully complied with isolation and quarantine guidance, eliminating their contributions to future transmission. As many as 1.36 million cases and 33 527 hospitalizations could have been prevented if all interviewed cases and monitored contacts had entered into and fully complied with isolation and quarantine guidelines upon being interviewed or notified. Across both scenarios and all jurisdictions, CICT averted an estimated median of 21.2% (range, 1.3%-65.8%) of the cases not prevented by vaccination and other nonpharmaceutical interventions. These findings suggest that CICT programs likely had a substantial role in curtailing the pandemic in most jurisdictions during the 2020 to 2021 winter peak. Differences in outcomes across jurisdictions indicate an opportunity to further improve CICT effectiveness. These estimates demonstrate the potential benefits from sustaining and improving these programs.

Assessment of Machine Learning–Based Medical Directives to Expedite Care in Pediatric Emergency Medicine

Increased wait times and long lengths of stay in emergency departments (EDs) are associated with poor patient outcomes. Systems to improve ED efficiency would be useful. Specifically, minimizing the time to diagnosis by developing novel workflows that expedite test ordering can help accelerate clinical decision-making. To explore the use of machine learning-based medical directives (MLMDs) to automate diagnostic testing at triage for patients with common pediatric ED diagnoses. Machine learning models trained on retrospective electronic health record data were evaluated in a decision analytical model study conducted at the ED of the Hospital for Sick Children Toronto, Canada. Data were collected on all patients aged 0 to 18 years presenting to the ED from July 1, 2018, to June 30, 2019 (77 219 total patient visits). Machine learning models were trained to predict the need for urinary dipstick testing, electrocardiogram, abdominal ultrasonography, testicular ultrasonography, bilirubin level testing, and forearm radiographs. Models were evaluated using area under the receiver operator curve, true-positive rate, false-positive rate, and positive predictive values. Model decision thresholds were determined to limit the total number of false-positive results and achieve high positive predictive values. The time difference between patient triage completion and test ordering was assessed for each use of MLMD. Error rates were analyzed to assess model bias. In addition, model explainability was determined using Shapley Additive Explanations values. There was a total of 42 238 boys (54.7%) included in model development; mean (SD) age of the children was 5.4 (4.8) years. Models obtained high area under the receiver operator curve (0.89-0.99) and positive predictive values (0.77-0.94) across each of the use cases. The proposed implementation of MLMDs would streamline care for 22.3% of all patient visits and make test results available earlier by 165 minutes (weighted mean) per affected patient. Model explainability for each MLMD demonstrated clinically relevant features having the most influence on model predictions. Models also performed with minimal to no sex bias. The findings of this study suggest the potential for clinical automation using MLMDs. When integrated into clinical workflows, MLMDs may have the potential to autonomously order common ED tests early in a patient's visit with explainability provided to patients and clinicians.

Development and Validation of a Multivariable Prediction Model for Postoperative Intensive Care Unit Stay in a Broad Surgical Population

Despite limited capacity and expensive cost, there are minimal objective data to guide postoperative allocation of intensive care unit (ICU) beds. The Surgical Risk Preoperative Assessment System (SURPAS) uses 8 preoperative variables to predict many common postoperative complications, but it has not yet been evaluated in predicting postoperative ICU admission. To determine if the SURPAS model could accurately predict postoperative ICU admission in a broad surgical population. This decision analytical model was a retrospective, observational analysis of prospectively collected patient data from the 2012 to 2018 American College of Surgeons (ACS) National Surgical Quality Improvement Program (NSQIP) database, which were merged with individual patients' electronic health record data to capture postoperative ICU use. Multivariable logistic regression modeling was used to determine how the 8 preoperative variables of the SURPAS model predicted ICU use compared with a model inputting all 28 preoperatively available NSQIP variables. Data included in the analysis were collected for the ACS NSQIP at 5 hospitals (1 tertiary academic center, 4 academic affiliated hospitals) within the University of Colorado Health System between January 1, 2012, and December 31, 2018. Included patients were those undergoing surgery in 9 surgical specialties during the 2012 to 2018 period. Data were analyzed from May 29 to July 30, 2021. Surgery in 9 surgical specialties, including general, gynecology, orthopedic, otolaryngology, plastic, thoracic, urology, vascular, and neurosurgery. Use of ICU care up to 30 days after surgery. A total of 34 568 patients were included in the analytical data set: 32 032 (92.7%) in the cohort without postoperative ICU use and 2545 (7.4%) in the cohort with postoperative ICU use (no ICU use: mean [SD] age, 54.9 [16.6] years; 18 188 women [56.8%]; ICU use: mean [SD] age, 60.3 [15.3] years; 1333 men [52.4%]). For the internal chronologic validation of the 7-variable SURPAS model, data from 2012 to 2016 were used as the training data set (n = 24 250, 70.2% of the total sample size of 34 568) and data from 2017 to 2018 were used as the test data set (n = 10 318, 29.8% of the total sample size of 34 568). The C statistic improved in the test data set compared with the training data set (0.933; 95% CI, 0.924-0.941 vs 0.922; 95% CI, 0.917-0.928), whereas the Brier score was slightly worse in the test data set compared with the training data set (0.045; 95% CI, 0.042-0.048 vs 0.045; 95% CI, 0.043-0.047). The SURPAS model compared favorably with the model inputting all 28 NSQIP variables, with both having good calibration between observed and expected outcomes in the Hosmer-Lemeshow graphs and similar Brier scores (model inputting all variables, 0.044; 95% CI, 0.043-0.048; SURPAS model, 0.045; 95% CI, 0.042-0.046) and C statistics (model inputting all variables, 0.929; 95% CI, 0.925-0.934; SURPAS model, 0.925; 95% CI, 0.921-0.930). Results of this decision analytical model study revealed that the SURPAS prediction model accurately predicted postoperative ICU use across a diverse surgical population. These results suggest that the SURPAS prediction model can be used to help with preoperative planning and resource allocation of limited ICU beds.

Outcomes Associated With Social Distancing Policies in St Louis, Missouri, During the Early Phase of the COVID-19 Pandemic

In the absence of a national strategy in response to the COVID-19 pandemic, many public health decisions fell to local elected officials and agencies. Outcomes of such policies depend on a complex combination of local epidemic conditions and demographic features as well as the intensity and timing of such policies and are therefore unclear. To use a decision analytical model of the COVID-19 epidemic to investigate potential outcomes if actual policies enacted in March 2020 (during the first wave of the epidemic) in the St Louis region of Missouri had been delayed. A previously developed, publicly available, open-source modeling platform (Local Epidemic Modeling for Management & Action, version 2.1) designed to enable localized COVID-19 epidemic projections was used. The compartmental epidemic model is programmed in R and Stan, uses bayesian inference, and accepts user-supplied demographic, epidemiologic, and policy inputs. Hospital census data for 1.3 million people from St Louis City and County from March 14, 2020, through July 15, 2020, were used to calibrate the model. Hypothetical delays in actual social distancing policies (which began on March 13, 2020) by 1, 2, or 4 weeks. Sensitivity analyses were conducted that explored plausible spontaneous behavior change in the absence of social distancing policies. Hospitalizations and deaths. A model of 1.3 million residents of the greater St Louis, Missouri, area found an initial reproductive number (indicating transmissibility of an infectious agent) of 3.9 (95% credible interval [CrI], 3.1-4.5) in the St Louis region before March 15, 2020, which fell to 0.93 (95% CrI, 0.88-0.98) after social distancing policies were implemented between March 15 and March 21, 2020. By June 15, a 1-week delay in policies would have increased cumulative hospitalizations from an observed actual number of 2246 hospitalizations to 8005 hospitalizations (75% CrI: 3973-15 236 hospitalizations) and increased deaths from an observed actual number of 482 deaths to a projected 1304 deaths (75% CrI, 656-2428 deaths). By June 15, a 2-week delay would have yielded 3292 deaths (75% CrI, 2104-4905 deaths)-an additional 2810 deaths or a 583% increase beyond what was actually observed. Sensitivity analyses incorporating a range of spontaneous behavior changes did not avert severe epidemic projections. The results of this decision analytical model study suggest that, in the St Louis region, timely social distancing policies were associated with improved population health outcomes, and small delays may likely have led to a COVID-19 epidemic similar to the most heavily affected areas in the US. These findings indicate that an open-source modeling platform designed to accept user-supplied local and regional data may provide projections tailored to, and more relevant for, local settings.

Development and Validation of a Machine Learning Model Using Administrative Health Data to Predict Onset of Type 2 Diabetes

Systems-level barriers to diabetes care could be improved with population health planning tools that accurately discriminate between high- and low-risk groups to guide investments and targeted interventions. To develop and validate a population-level machine learning model for predicting type 2 diabetes 5 years before diabetes onset using administrative health data. This decision analytical model study used linked administrative health data from the diverse, single-payer health system in Ontario, Canada, between January 1, 2006, and December 31, 2016. A gradient boosting decision tree model was trained on data from 1 657 395 patients, validated on 243 442 patients, and tested on 236 506 patients. Costs associated with each patient were estimated using a validated costing algorithm. Data were analyzed from January 1, 2006, to December 31, 2016. A random sample of 2 137 343 residents of Ontario without type 2 diabetes was obtained at study start time. More than 300 features from data sets capturing demographic information, laboratory measurements, drug benefits, health care system interactions, social determinants of health, and ambulatory care and hospitalization records were compiled over 2-year patient medical histories to generate quarterly predictions. Discrimination was assessed using the area under the receiver operating characteristic curve statistic, and calibration was assessed visually using calibration plots. Feature contribution was assessed with Shapley values. Costs were estimated in 2020 US dollars. This study trained a gradient boosting decision tree model on data from 1 657 395 patients (12 900 257 instances; 6 666 662 women [51.7%]). The developed model achieved a test area under the curve of 80.26 (range, 80.21-80.29), demonstrated good calibration, and was robust to sex, immigration status, area-level marginalization with regard to material deprivation and race/ethnicity, and low contact with the health care system. The top 5% of patients predicted as high risk by the model represented 26% of the total annual diabetes cost in Ontario. In this decision analytical model study, a machine learning model approach accurately predicted the incidence of diabetes in the population using routinely collected health administrative data. These results suggest that the model could be used to inform decision-making for population health planning and diabetes prevention.

Assessment of Incidence of and Surveillance Burden for Hepatocellular Carcinoma Among Patients With Hepatitis C in the Era of Direct-Acting Antiviral Agents

In the US, hepatocellular carcinoma (HCC), primarily associated with hepatitis C virus (HCV) infection, is the fastest rising cause of cancer-related death. Wider use of highly effective direct-acting antiviral agents (DAAs) substantially reduces the burden of chronic HCV infection, but the subsequent impacts with HCV-associated HCC remain unknown. To assess projected changes in the incidence rate of and surveillance burden for HCC in the era of DAA treatment for HCV. This decision analytical model study was performed from January 2019 to February 2020, using an individual-level state-transition simulation model to simulate disease progression, screening, and different waves of antiviral treatments for HCV in the US from 2012 to 2040. Current clinical management for chronic HCV infection. Model outcomes were projected temporal trends and age distribution of incident HCC cases and candidates for HCC surveillance among patients with viremia and patients with virologically cured HCV. The simulation model projected that the annual incidence of HCC among patients with viremia and patients with virologically cured HCV will continue increasing to 24 000 (95% uncertainty interval [UI], 18 000-31 000) cases until 2021. In patients with virologically cured HCV, incident HCC cases are projected to increase from 1000 (95% UI, 500-2100) in 2012 to the peak of 7000 (95% UI, 5000-9600) in 2031 with a subsequent decrease to 6000 (95% UI, 4300-8300) by 2040. The proportion of incident HCC cases that occur in individuals with virologically cured HCV is estimated to increase from 5.3% in 2012 to 45.8% in 2040. The number of candidates for HCC surveillance in the population with virologically cured HCV is projected to increase from 106 000 (95% UI, 70 000-178 000) in 2012 to the peak of 649 000 (95% UI, 512 000-824 000) in 2030 and decrease to 539 000 (95% UI, 421 000-687 000) by 2040, while the proportion of all candidates for surveillance who are virologically cured is estimated to increase from 8.5% to 64.6% during the same period. The average age of HCC incidence and surveillance candidates is estimated to increase from 55 in 2012 to 72 and 71, respectively, by 2040. The results of this study suggest that the burden of HCC will shift from patients with viremia to patients with virologically cured HCV, and to older populations. Appropriate management may be warranted for early detection of HCC in patients who may no longer be receiving specialty care for liver conditions.

Using Metamodeling to Identify the Optimal Strategy for Colorectal Cancer Screening

ObjectivesMetamodeling can address computational challenges within decision-analytic modeling studies evaluating many strategies. This article illustrates the value of metamodeling for evaluating colorectal cancer screening strategies while accounting for colonoscopy capacity constraints. MethodsIn a traditional approach, the best screening strategy was identified from a limited subset of strategies evaluated with the validated Adenoma and Serrated pathway to Colorectal CAncer model. In a metamodeling approach, metamodels were fitted to this limited subset to evaluate all potentially plausible strategies and determine the best overall screening strategy. Approaches were compared based on the best screening strategy in life-years gained compared with no screening. Metamodel runtime and accuracy was assessed. ResultsThe metamodeling approach evaluated >40 000 strategies in <1 minute with high accuracy after 1 adaptive sampling step (mean absolute error: 0.0002 life-years) using 300 samples in total (generation time: 8 days). Findings indicated that health outcomes could be improved without requiring additional colonoscopy capacity. Obtaining similar insights using the traditional approach could require at least 1000 samples (generation time: 28 days). Suggested benefits from screening at ages <40 years require adequate validation of the underlying Adenoma and Serrated pathway to Colorectal CAncer model before making policy recommendations. ConclusionsMetamodeling allows rapid assessment of a vast set of strategies, which may lead to identification of more favorable strategies compared to a traditional approach. Nevertheless, metamodel validation and identifying extrapolation beyond the support of the original decision-analytic model are critical to the interpretation of results. The screening strategies identified with metamodeling support ongoing discussions on decreasing the starting age of colorectal cancer screening.

Simulation of Pool Testing to Identify Patients With Coronavirus Disease 2019 Under Conditions of Limited Test Availability

This decision analytical model study examines the feasibility of using pool testing to identify patients with coronavirus disease 2019 (COVID-19) in a setting with limited testing availability.

Decision-Analytic Modeling Studies in Prevention and Treatment of Iodine Deficiency and Thyroid Disorders: A Systematic Overview.

Background: Prevention and treatment of iodine deficiency-related diseases remain an important public health challenge. Iodine deficiency can have severe health consequences, such as cretinism, goiter, or other thyroid disorders, and it has economic implications. Our aim was to give an overview of studies applying decision-analytic modeling to evaluate the effectiveness and/or cost-effectiveness of iodine deficiency-related prevention strategies or treatments related to thyroid disorders. Methods: We performed a systematic literature search in PubMed/MEDLINE (Medical Literature Analysis and Retrieval System Online), EMBASE (Excerpta Medica Database), Tuft's Cost-Effectiveness Analysis Registry, and National Health System Economic Evaluation Database (NHS EED) to identify studies published between 1985 and 2018 comparing different prevention or treatment strategies for iodine deficiency and thyroid disorders by applying a mathematical decision-analytic model. Studies were required to evaluate patient-relevant health outcomes (e.g., remaining life years, quality-adjusted life years [QALYs]). Results: Overall, we found 3950 studies. After removal of duplicates, abstract/title, and full-text screening, 17 studies were included. Eleven studies evaluated screening programs (mainly newborns and pregnant women), five studies focused on treatment approaches (Graves' disease, toxic thyroid adenoma), and one study was about primary prevention (consequences of iodine supplementation on offspring). Most of the studies were conducted within the U.S. health care context (n = 7). Seven studies were based on a Markov state-transition model, nine studies on a decision tree model, and in one study, an initial decision tree and a long-term Markov state-transition model were combined. The analytic time horizon ranged from 1 year to lifetime. QALYs were evaluated as health outcome measure in 15 of the included studies. In all studies, a cost-effectiveness analysis was performed. None of the models reported a formal model validation. In most cases, the authors of the modeling studies concluded that screening is potentially cost-effective or even cost-saving. The recommendations for treatment approaches were rather heterogeneous and depending on the specific research question, population, and setting. Conclusions: Overall, we predominantly identified decision-analytic modeling studies evaluating specific screening programs or treatment approaches; however, there was no model evaluating primary prevention programs on a population basis. Conclusions deriving from these studies, for example, that prevention is cost-saving, need to be carefully interpreted as they rely on many assumptions.