Search

Purpose Metamodels are simplified approximations of more complex models that can be used as surrogates for the original models. Challenges in using metamodels for policy analysis arise when there are multiple correlated outputs of interest. We develop a framework for metamodeling with policy simulations to accommodate multivariate outcomes. Methods: We combine 2 algorithm adaptation methods—multitarget stacking and regression chain with maximum correlation—with different base learners including linear regression (LR), elastic net (EE) with second-order terms, Gaussian process regression (GPR), random forests (RFs), and neural networks. We optimize integrated models using variable selection and hyperparameter tuning. We compare the accuracy, efficiency, and interpretability of different approaches. As an example application, we develop metamodels to emulate a microsimulation model of testing and treatment strategies for hepatitis C in correctional settings. Results Output variables from the simulation model were correlated (average ρ = 0.58). Without multioutput algorithm adaptation methods, in-sample fit (measured by R2) ranged from 0.881 for LR to 0.987 for GPR. The multioutput algorithm adaptation method increased R2 by an average 0.002 across base learners. Variable selection and hyperparameter tuning increased R2 by 0.009. Simpler models such as LR, EE, and RF required minimal training and prediction time. LR and EE had advantages in model interpretability, and we considered methods for improving the interpretability of other models. Conclusions In our example application, the choice of base learner had the largest impact on R2; multioutput algorithm adaptation and variable selection and hyperparameter tuning had a modest impact. Although advantages and disadvantages of specific learning algorithms may vary across different modeling applications, our framework for metamodeling in policy analyses with multivariate outcomes has broad applicability to decision analysis in health and medicine.

Medications for opioid use disorder (MOUD) are shown to reduce opioid use and the risk of overdose. People with opioid use disorder (OUD) who exit inpatient medically managed withdrawal programs (detox) without initiating MOUD and linking to outpatient care have high rates of overdose. While detox encounters provide a theoretical opportunity for MOUD initiation, this is not ubiquitous in the United States. We used simulation modeling to estimate the population-level health effects and cost-effectiveness of a policy encouraging MOUD initiation during inpatient detox encounters.We employed a dynamic population state-transition model to evaluate the effectiveness and cost-effectiveness of using detox programs as venues for initiating MOUD in Massachusetts, United States. We compared standard of care, where no detox patients initiate MOUD or link to outpatient MOUD providers, to strategies of offering MOUD to detox patients and linking those patients to outpatient MOUD.Budgetary impact to the Massachusetts health-care sector, incremental cost-effectiveness ratios (ICER) and total counts and percentage differences of fatal overdoses prevented.Initiating MOUD in detox with perfect linkage to outpatient MOUD would reduce fatal overdoses by 4.5% [95% confidence interval (CI) = 2.3-5.9], at an ICER of $56 000 per quality-adjusted life-year (QALY) gained, compared with the standard of care. With moderate linkage, fatal overdoses would be reduced by 2.3% (95% CI= 1.2-3.1) with an ICER of $78 500 per QALY gained, compared with standard of care. Budgetary increase to Massachusetts health-care spending ranged from 0.5-1%.A simulation model indicates that initiation of medications for opioid use disorder and linkage policies among detox patients in Massachusetts, USA could prevent fatal opioid overdoses in the opioid use disorder population and would be cost-effective from a health-care sector perspective.

BACKGROUND: Federally qualified health centers (FQHCs) serve diverse communities in the United States (U.S.) and could function as important venues to diagnose and treat hepatitis C virus (HCV) infections. OBJECTIVE: To determine HCV testing proportion and factors associated with treatment initiation, and treatment outcomes in a large sample of FQHCs around the U.S. DESIGN: Retrospective cohort study using electronic health records of three hundred and forty-one FQHC clinical sites participating in the OCHIN network in 19 U.S. states. PARTICIPANTS: Adult patients (≥ 18 years of age) seen between January 01, 2012, and June 30, 2017. MAIN MEASURES: HCV testing proportion, stratified by diagnosis of opioid use disorder (OUD); treatment initiation rates; and sustained virologic response (SVR), defined as undetectable HCV RNA 6 months after treatment initiation. KEY RESULTS: Of the 1,508,525 patients meeting inclusion criteria, 88,384 (5.9%) were tested for HCV, and 8694 (9.8%) of individuals tested had reactive results. Of the 6357 with HCV RNA testing, 4092 (64.4%) had detectable RNA. Twelve percent of individuals with chronic HCV and evaluable data initiated treatment. Of those, 87% reached SVR. Having commercial insurance (aOR, 2.11; 95% CI, 1.46–3.05), older age (aOR, 1.07; 95% CI, 1.06–1.09), and being Hispanic/Latino (aOR, 1.87; 95% CI, 1.38–2.53) or Asian/Pacific Islander (aOR, 2.47; 95% CI, 1.46–4.19) were independently associated with higher odds of treatment initiation after multivariable adjustment. In contrast, women (aOR, 0.76; 95% CI, 0.60–0.97) and the uninsured (aOR, 0.15; 95% CI, 0.09–0.25) were less likely to initiate treatment. Only 8% of individuals with chronic HCV were tested for HIV, and 15% of individuals with identified OUD were tested for HCV. CONCLUSIONS: Fewer than 20% of individuals with identified OUD were tested for HCV. SVR was lower than findings in other real-world cohorts. Measures to improve outcomes should be considered with the expansion of HCV management into community clinics. ELECTRONIC SUPPLEMENTARY MATERIAL: The online version of this article (10.1007/s11606-020-05701-9) contains supplementary material, which is available to authorized users.

Background Untreated opioid use disorder (OUD) affects the care of HIV/HCV co-infected people who inject opioids. Despite active injection opioid use, there is evidence of increasing engagement in HIV care and adherence to HIV medications among HIV/HCV co-infected persons. However, less than one-half of this population is offered HCV treatment onsite. Treatment for OUD is also rare and largely occurs offsite. Integrating buprenorphine-naloxone (BUP-NX) into onsite care for HIV/HCV co-infected persons may improve outcomes, but the clinical impact and costs are unknown. We evaluated the clinical impact, costs, and cost-effectiveness of integrating (BUP-NX) into onsite HIV/HCV treatment compared with the status quo of offsite referral for medications for OUD. Methods We used a Monte Carlo microsimulation of HCV to compare two strategies for people who inject opioids: 1) standard HIV care with onsite HCV treatment and referral to offsite OUD care (status quo) and 2) standard HIV care with onsite HCV and BUP-NX treatment (integrated care). Both strategies assume that all individuals are already in HIV care. Data from national databases, clinical trials, and cohorts informed model inputs. Outcomes included mortality, HCV reinfection, quality-adjusted life years (QALYs), costs (2017 US dollars), and incremental cost-effectiveness ratios. Results Integrated care reduced HCV reinfections by 7%, cases of cirrhosis by 1%, and liver-related deaths by 3%. Compared to the status quo, this strategy also resulted in an estimated 11/1,000 fewer non-liver attributable deaths at one year and 28/1,000 fewer of these deaths at five years, at a cost-effectiveness ratio of $57,100/QALY. Integrated care remained cost-effective in sensitivity analyses that varied the proportion of the population actively injecting opioids, availability of BUP-NX, and quality of life weights. Conclusions Integrating BUP-NX for OUD into treatment for HIV/HCV co-infected adults who inject opioids increases life expectancy and is cost-effective at a $100,000/QALY threshold.

To estimate the clinical effects and cost-effectiveness of universal prenatal hepatitis C screening, and to calculate potential life expectancy, quality of life, and health care costs associated with universal prenatal hepatitis C screening and linkage to treatment.Using a stochastic individual-level microsimulation model, we simulated the lifetimes of 250 million pregnant women matched at baseline with the U.S. childbearing population on age, injection drug use behaviors, and hepatitis C virus (HCV) infection status. Modeled outcomes included hepatitis C diagnosis, treatment and cure, lifetime health care costs, quality-adjusted life years (QALY) and incremental cost-effectiveness ratios comparing universal prenatal hepatitis C screening to current practice. We modeled whether neonates exposed to maternal HCV at birth were identified as such.Pregnant women with hepatitis C infection lived 1.21 years longer and had 16% lower HCV-attributable mortality with universal prenatal hepatitis C screening, which had an incremental cost-effectiveness ratio of $41,000 per QALY gained compared with current practice. Incremental cost-effectiveness ratios remained below $100,000 per QALY gained in most sensitivity analyses; notable exceptions included incremental cost-effectiveness ratios above $100,000 when assuming mean time to cirrhosis of 70 years, a cost greater than $500,000 per false positive diagnosis, or population HCV infection prevalence below 0.16%. Universal prenatal hepatitis C screening increased identification of neonates exposed to HCV at birth from 44% to 92%.In our model, universal prenatal hepatitis C screening improves health outcomes in women with HCV infection, improves identification of HCV exposure in neonates born at risk, and is cost-effective.

Introduction There is no widely accepted testing approach for hepatitis C virus infection in correctional settings, and many U.S. prisons do not provide routine testing. The aim of this study was to determine the most effective hepatitis C virus testing strategy in one U.S. state prison and describe the population with reactive testing. Methods A retrospective analysis was performed using individuals entering the Washington State prison system, which routinely offers hepatitis C virus testing, to compare routine opt-out with current recommendations for risk-based and one-time testing for individuals born between 1945 and 1965. Additionally, liver fibrosis stage was characterized using aspartate aminotransferase to platelet ratio index and Fibrosis-4 index. Analyses were conducted in 2017. Results Between 2012 and 2016, a total of 24,567 (83%) individuals were tested for the hepatitis C virus antibody and 4,921 (20%) were reactive (test was positive). There were 2,403 (49%) that had hepatitis C virus RNA testing, with 1,727 (72%) showing chronic infection. Reactive antibody was more prevalent in individuals born between 1945 and 1965 compared with other years (44% vs 17%); however, most cases (72%) were outside of this cohort. Up to 35% of positive reactive tests would be missed with testing targeted by birth cohort and risk behavior. Of chronically infected individuals, 23% had at least moderate liver fibrosis. Conclusions Targeted testing in the Washington State prison system missed a substantial proportion of hepatitis C virus cases; of those with reactive testing, a sizeable proportion of people had at least moderate liver disease, placing them at risk for complications. Routine testing at entry should be considered by U.S. state prisons.

Background The opioid epidemic has been associated with an increase in hepatitis C virus (HCV) infections. Federally qualified health centers (FQHCs) have a high burden of hepatitis C disease and could serve as venues to enhance testing and treatment. Methods We estimated clinical outcomes and the cost-effectiveness of hepatitis C testing and treatment at US FQHCs using individual-based simulation modeling. We used individual-level data from 57 FQHCs to model 9 strategies, including permutations of HCV antibody testing modality, person initiating testing, and testing approach. Outcomes included life expectancy, quality-adjusted life-years (QALY), hepatitis C cases identified, treated and cured; and incremental cost-effectiveness ratios. Results Compared with current practice (risk-based with laboratory-based testing), routine rapid point-of-care testing initiated and performed by a counselor identified 68% more cases after (nonreflex) RNA testing in the first month of the intervention and led to a 17% reduction in cirrhosis cases and a 22% reduction in liver deaths among those with cirrhosis over a lifetime. Routine rapid testing initiated by a counselor or a clinician provided better outcomes at either lower total cost or at lower cost per QALY gained, when compared with all other strategies. Findings were most influenced by the proportion of patients informed of their anti-HCV test results. Conclusions Routine anti-HCV testing followed by prompt RNA testing for positives is recommended at FQHCs to identify infections. If using dedicated staff or point-of-care testing is not feasible, then measures to improve immediate patient knowledge of antibody status should be considered.

Background The expansion of the US opioid epidemic has led to significant increases in infections, such as infective endocarditis (IE), which is tied to injection behaviors. We aimed to estimate the population-level IE mortality rate among people who inject opioids and compare the risk of IE death against the risks of death from other causes. Methods We developed a microsimulation model of the natural history of injection opioid use. We defined injection behavior profiles by both injection frequency and injection techniques. We accounted for competing risks of death and populated the model with primary and published data. We modeled cohorts of 1 million individuals with different injection behavior profiles until age 60 years. We combined model-generated estimates with published data to project the total expected number of IE deaths in the United States by 2030. Results The probabilities of death from IE by age 60 years for 20-, 30-, and 40-year-old men with high-frequency use with higher infection risk techniques compared to lower risk techniques for IE were 53.8% versus 3.7%, 51.4% versus 3.1%, and 44.5% versus 2.2%, respectively. The predicted population-level attributable fraction of 10-year mortality from IE among all risk groups was 20%. We estimated that approximately 257 800 people are expected to die from IE by 2030. Conclusions The expected burden of IE among people who inject opioids in the United States is large. Adopting a harm reduction approach, including through expansion of syringe service programs, to address injection behaviors could have a major impact on decreasing the mortality rate associated with the opioid epidemic.

Importance The United States is experiencing a crisis of opioid overdose. In response, the US Department of Health and Human Services has defined a goal to reduce overdose mortality by 40% by 2022. Objective To identify specific combinations of 3 interventions (initiating more people to medications for opioid use disorder [MOUD], increasing 6-month retention with MOUD, and increasing naloxone distribution) associated with at least a 40% reduction in opioid overdose in simulated populations. Design, Setting, and Participants This decision analytical model used a dynamic population-level state-transition model to project outcomes over a 2-year horizon. Each intervention scenario was compared with the counterfactual of no intervention in simulated urban and rural communities in Massachusetts. Simulation modeling was used to determine the associations of community-level interventions with opioid overdose rates. The 3 examined interventions were initiation of more people to MOUD, increasing individuals’ retention with MOUD, and increasing distribution of naloxone. Data were analyzed from July to November 2020. Main Outcomes and Measures Reduction in overdose mortality, medication treatment capacity needs, and naloxone needs. Results No single intervention was associated with a 40% reduction in overdose mortality in the simulated communities. Reaching this goal required use of MOUD and naloxone. Achieving a 40% reduction required that 10% to 15% of the estimated OUD population not already receiving MOUD initiate MOUD every month, with 45% to 60%% retention for at least 6 months, and increased naloxone distribution. In all feasible settings and scenarios, attaining a 40% reduction in overdose mortality required that in every month, at least 10% of the population with OUD who were not currently receiving treatment initiate an MOUD. Conclusions and Relevance In this modeling study, only communities with increased capacity for treating with MOUD and increased MOUD retention experienced a 40% decrease in overdose mortality. These findings could provide a framework for developing community-level interventions to reduce opioid overdose death.

Background Hepatitis C virus (HCV) testing and treatment uptake in prisons remains low. We aimed to estimate clinical outcomes, cost-effectiveness (CE), and budgetary impact (BI) of HCV testing and treatment in United States (US) prisons or linkage to care at release. Methods We used individual-based simulation modeling with healthcare and Department of Corrections (DOC) perspectives for CE and BI analyses, respectively. We simulated a US prison cohort at entry using published data and Washington State DOC individual-level data. We considered permutations of testing (risk factor based, routine at entry or at release, no testing), treatment (if liver fibrosis stage ≥F3, for all HCV infected or no treatment), and linkage to care (at release or no linkage). Outcomes included quality-adjusted life-years (QALY); cases identified, treated, and cured; cirrhosis cases avoided; incremental cost-effectiveness ratios; DOC costs (2016 US dollars); and BI (healthcare cost/prison entrant) to generalize to other states. Results Compared to “no testing, no treatment, and no linkage to care,” the “test all, treat all, and linkage to care at release” model increased the lifetime sustained virologic response by 23%, reduced cirrhosis cases by 54% at a DOC annual additional cost of $1440 per prison entrant, and would be cost-effective. At current drug prices, targeted testing and liver fibrosis–based treatment provided worse outcomes at higher cost or worse outcomes at higher cost per QALY gained. In sensitivity analysis, fibrosis-based treatment restrictions were cost-effective at previous higher drug costs. Conclusions Although costly, widespread testing and treatment in prisons is considered to be of good value at current drug prices.

Background The US Centers for Disease Control and Prevention and the U.S. Preventive Services Task Force recommend one-time hepatitis C virus (HCV) testing for persons born 1945-1965 and targeted testing for high-risk persons. This strategy targets HCV testing to a prevalent population at high risk for HCV morbidity and mortality, but does not include younger populations with high incidence. To address this gap and improve access to HCV testing, age-based strategies should be considered. Methods We used a simulation of HCV to estimate the effectiveness and cost-effectiveness of HCV testing strategies: 1) standard of care (SOC) - recommendation for one-time testing for all persons born 1945-1965, 2) recommendation for one-time testing for adults ≥40 years (≥40 strategy), 3) ≥30 years (≥30 strategy), and 4) ≥18 years (≥18 strategy). All strategies assumed targeted testing of high-risk persons. Inputs were derived from national databases, observational cohorts and clinical trials. Outcomes included quality-adjusted life expectancy, costs, and cost-effectiveness. Results Expanded age-based testing strategies increased US population lifetime case identification and cure rates. Greatest increases were observed in the ≥18 strategy. Compared to the SOC, this strategy resulted in an estimated 256,000 additional infected persons identified and 280,000 additional cures at the lowest cost per QALY gained (ICER = $28,000/QALY). Conclusions In addition to risk-based testing, one-time HCV testing of persons 18 and older appears to be cost-effective, leads to improved clinical outcomes and identifies more persons with HCV than the current birth cohort recommendations. These findings could be considered for future recommendation revisions.