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

This decision analytical model estimates the projected 3-year association between public health interventions and opioid overdose-related outcomes among persons with opioid use disorder (OUD).


Introduction
Opioid-involved overdose deaths remain at high levels in the US, driven primarily by synthetic opioids such as illegally made fentanyl (IMF). 1,2In 2021, 80 816 opioid-involved overdose deaths occurred, 1 and an estimated 5.6 million persons aged 12 years and older had an opioid use disorder (OUD) in the US, based on the Diagnostic and Statistical Manual of Mental Disorders (Fifth Edition) criteria for heroin or prescription pain reliever use disorder in the National Survey on Drug Use and Health (NSDUH). 35][6][7] Expanding access to MOUD treatment is a cornerstone of the response to the opioid overdose crisis. 8,9ven the complexity of OUD and the epidemic of opioid-involved overdose deaths, systems models have utility as a tool to understand the potential impact of and opportunities for evidencebased interventions to prevent overdose. 102][13][14][15][16][17] While these models differ in scope, they highlight the need to combine interventions to achieve considerable reductions in opioidinvolved overdose deaths.However, only a few state-specific models exist that focus on the population with OUD, 12

Methods
We developed and calibrated MODIPHI, a national-level system dynamics simulation model 18 of the estimated US population aged 12 years and older with OUD using data from 2019 to 2020.The calibrated model was used to estimate the association between overdose prevention interventions simulated between 2021 and 2023, the prevalence of OUD and MOUD, and the number of nonfatal and fatal opioid-involved overdoses among persons with OUD.This study was reviewed by the US Centers for Disease Control and Prevention (CDC) and conducted consistent with applicable federal law and CDC policy in accordance with 45 CFR §46.Institutional review board approval was not sought and the need for informed consent did not apply because deidentified, retrospective, aggregate data were used in this study.This study followed the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) reporting guideline. 19

Disease Model of OUD
The disease model comprised 9 model states: (1) persons with OUD (not receiving MOUD treatment), (2) persons with OUD receiving MOUD treatment with durations of 1 month or less, (3) more than 1 month to 6 months or less, (4) more than 6 months to 12 months or less, and (5) more than 12 months; (6) persons in sustained remission from OUD (defined here as receiving MOUD or counseling-only treatment for at least 360 days and no reported opioid use in the past 90 days); (7)   nonfatal opioid-involved overdose; (8) fatal opioid-involved overdose; and (9) death from other causes (Figure 1).).We adjusted prevalence data reported in NSDUH since estimates of OUD from general population surveys have been shown to underestimate prevalence by up to 3-to 5-fold. 22,28ditional calibration targets included MOUD prevalence (eAppendix 9 in Supplement 1) with calibration results reported in eAppendix 10 in Supplement 1 (eFigure 1 and eTable 8 in Supplement 1).The projected number of nonfatal overdoses was validated with a recent estimate in the literature 29 (eAppendix 11 in Supplement 1).

Simulating Public Health Interventions
Evidence from previous models highlights the importance of combining interventions to achieve reductions in opioid overdose-related outcomes. 12,15As in previous studies, we simulated hypothetical improvements 13,17

JAMA Network Open | Substance Use and Addiction
Public Health Interventions and Overdose-Related Outcomes Among Persons With OUD past year OUD). 26,27These increases or decreases in parameter values were implemented stepwise between 2021 and 2023, representing gradual implementation of the intervention.For example, a 50% overall decrease in a parameter across the 3-year simulation time horizon was simulated by implementing a 16.7% stepwise decrease in the original parameter value at the beginning of each year to achieve a 50% decrease in the final year.

Data Analysis
We reported model estimates for outcomes of interest (nonfatal and fatal opioid overdoses, OUD prevalence, and MOUD prevalence) as a percentage change relative to projected outcomes in the baseline model scenario at the end of 2023.One-way sensitivity analyses were also conducted on model parameters as well as OUD and fatal overdose calibration targets within AnyLogic.

Results
The decreasing the time to access and be linked to treatment, increased prescribing and availability of MOUD, and implementation of peer support programs and efforts to address stigma. 6,7,35Efforts to decrease both nonfatal and fatal opioid-involved overdoses can encompass overdose education and  harm reduction efforts including increased access to and distribution of naloxone and drug-checking services. 14,36,37

Limitations
This study has several limitations.First, the model does not simulate the impacts of specific intervention programs among persons with OUD due to limited data.We also cannot comment on the investments or intensity of efforts necessary to achieve the change in the magnitude of hypothetical interventions simulated here.Second, this model does not distinguish between type of opioid involved in overdoses.The validation of nonfatal overdoses utilized 1 estimate among people who inject drugs, which may not be representative of people with OUD.Accurately tracking nonfatal overdoses is limited by their timely identification and lack of data sources linking nonfatal and fatal overdoses at the individual level. 44,45Fourth, due to limitations in data quality, comparability, and availability, we used 2019 to 2020 data to calibrate the model and could not

Conclusions
This and previous models incorporate limited information on MOUD treatment duration.We developed a national-level simulation model of persons with OUD, Modeling OUD Dynamics Informing Public Health Interventions (MODIPHI), to estimate the relative change in opioid overdose-related outcomes that can be achieved by scaling public health interventions among persons with OUD in the short-term.This model additionally leverages new treatment and fatal overdose datasets to improve our understanding of population dynamics associated with recovery from OUD, including the risk of OUD recurrence and nonfatal and fatal overdose risk by explicitly modeling different durations of MOUD treatment.Findings from this model can provide insights about the potential outcomes associated with increasing population reach of or investments in public health interventions among people with OUD in the short term.
directly in the rates of transitions in the OUD disease model (eg, increase in initiation of treatment with MOUD, reduction in recurrence of OUD [ie, discontinuation of MOUD combined with past 90-day opioid use], and reduction of nonfatal and fatal opioid overdose rates) resulting from combinations of unspecified public health interventions.We chose this approach because limited data exist to precisely estimate the impact of evidence-based interventions when scaled at a national level.By focusing on the intended goals achieved through an unspecified intervention, we can better identify which transition pathways and intermediate outcomes are most critical to improving opioid overdose-related outcomes.In this study, we modeled 4 possible intervention scenarios aiming to yield the following: scenario A increased MOUD initiation and decreased OUD recurrence among persons receiving MOUD for 6 months or less (defined as early-stage) and persons receiving MOUD for more than 6 months (defined as late-stage); scenario B decreased fatal overdose rates and decreased recurrence of OUD; scenario C decreased nonfatal overdose rates and decreased recurrence of OUD; and scenario D increased MOUD initiation and decreased fatal overdose rates.To simulate the change in specific model transitions from combined interventions, the corresponding parameters were increased or decreased by a percentage of the baseline parameter value.Most parameters were decreased by 10% to 50% of the original parameter value, except for the MOUD initiation rate, which was increased by 50% to 200%.We chose this larger range for MOUD initiation given substantial variations and reduction in MOUD prevalence observed in NSDUH between 2019 and 2020 (18.1% in 2019 vs 11.2% in 2020 of persons aged 12 years or older with a

Figure 4 .
Figure 4. Percentage Change in Projected Model Outcomes Relative to the Baseline Scenario for Nonfatal Overdoses, Opioid Use Disorder (OUD) Prevalence Without Medications for Opioid Use Disorder (MOUD), and MOUD Prevalence Under Model Scenario C a Nonfatal overdoses Baseline scenario estimate = 5 102 289 nonfatal overdoses b A

Figure 5 .a
Figure 5. Percentage Change in Projected Model Outcomes Relative to the Baseline Scenario for Fatal Overdoses, Opioid Use Disorder (OUD) Prevalence Without Medications for Opioid Use Disorder (MOUD), and MOUD Prevalence Under Model Scenario D a Fatal overdoses Baseline scenario estimate = 145 237 fatal overdoses b A decision analytical model study provides insight about the population dynamics of MOUD treatment and opioid overdose among persons with OUD and the association between hypothetical outcomes of combined public health interventions and opioid overdose-related outcomes nationally.Findings suggest that expansion of evidence-based interventions to reduce the risk of overdose fatality among persons with OUD, such as through harm reduction efforts (eg, overdose education or naloxone distribution) are critical to achieve maximal reductions in fatal opioid-involved overdoses in the short-term.These results also emphasize that efforts to increase MOUD initiation and retain persons in treatment (eg, linkage to care, low-barrier treatment, or behavioral interventions)engender marked improvement in MOUD and OUD prevalence but may have limited influence on fatal opioid-involved overdoses in the short-term.A multifaceted and multisector approach that includes collaborations across health systems, public health, public safety, and community-based organizations will be important to successfully implement and scale up this comprehensive suite of interventions required to both reduce opioid-involved overdose fatalities in the short-term and sustain improved outcomes over the long-term.

JAMA Network Open | Substance Use and Addiction Public
Persons in the OUD model state could initiate MOUD, enter remission without MOUD, or experience a nonfatal or fatal overdose or death from other causes.We assumed all persons who experienced a nonfatal overdose subsequently transitioned to the OUD model state Health Interventions and Overdose-Related Outcomes Among Persons With OUD Persons in any MOUD state or remission could experience nonfatal or fatal overdoses, a death from other causes, or recurrence of OUD (defined as discontinuation of MOUD combined with past 90-day opioid use).Annually, persons newly diagnosed with OUD were added to the OUD model state.eAppendix 1 in Supplement 1 provide details of model states and the study population.
JAMA Network Open.2024;7(4):e244617.doi:10.1001/jamanetworkopen.2024.4617(Reprinted)April 3, 2024 2/14 Downloaded from jamanetwork.combyguest on 04/06/2024 after a time delay of 1 day, regardless of their prior model state, but could once again initiate MOUD treatment.aboutlength of time in treatment, completion or cessation of treatment, past 30-day and 90-day opioid use, and past 90-day opioid overdose.We used MarketScan Commercial and Multi-State Medicaid Databases, accessed through the MarketScan Treatment Pathways platform, to obtain nonfatal overdose rates among persons diagnosed with OUD not receiving MOUD (eAppendix 4, eTable 2, and eTable 3 in Supplement 1).Nonfatal overdose rates among persons receiving MOUD for less than 1 month, and rates of fatal overdose and death from other causes among persons receiving MOUD and those in remission wereFigure 1.Schematic of Natural History Model of Opioid Use Disorder (OUD) and Medications for Opioid Use Disorder (MOUD) a Parameters were calibrated.b Transition is modeled with a time delay of 1 day.c Parameters derived from MarketScan Commercial and Multi-State Medicaid Databases.d Recurrence of OUD as defined by recent opioid use and discontinuation of MOUD.e Parameters derived from MOUD study 20 data.f Parameters derived from scientific literature.

Decreased Fatal Overdoses and Decreased OUD Recurrence
baseline model projected outcomes through 2023, assuming parameter values calibrated with 2019 to 2020 historical data remained constant between 2021 and 2023 (eTable 8 in Supplement 1).

Decreased Nonfatal Overdoses and Decreased OUD Recurrence
9,7,30 2. Percentage Change in Projected Model Outcomes Relative to the Baseline Scenario for Fatal Overdoses, Nonfatal Overdoses, Opioid Use Disorder (OUD) Prevalence Without Medications for Opioid Use Disorder (MOUD), and MOUD Prevalence Under Model Scenario A afrom 50% to 200% substantially reduced OUD prevalence (from −5.6% to −20.4%) and increased MOUD prevalence (from 35.6% to 124.4%).beincreased by up to 137%, although this required tripling the MOUD initiation rate over 3 years (from 9.3% to 27.9% among persons with OUD).Interventions with these goals may have a greater influence on fatal and nonfatal overdoses over a longer time horizon once a higher proportion of persons with OUD could be initiated into MOUD treatment.This highlights the need for interventions that increase awareness of OUD symptoms, clinician training for appropriate OUD screening and diagnosis, and efforts to reduce stigma so that persons with OUD can be appropriately linked to treatment.6,7,30DecreasingOUDprevalenceand increasing access to MOUD treatment are 2 important objectives of Healthy People 2030 31 and are components of the US Department of Health and Human Services Overdose Prevention Strategy.9Ourfindings suggest that, in the short-term, expansion of evidence-based interventions aiming to reduce risk of nonfatal and fatal overdose (eg, overdose education, naloxone distribution) are critical to achieve maximum reductions of fatal opioid overdoses.In the long-term, interventions aimed at increasing MOUD initiation, retention of persons in treatment, and recovery support are critical to reducing OUD prevalence and can potentially achieve further reductions in opioid overdoses.CDC's strategic priorities in the Division of Overdose Prevention 32 support public health interventions designed to achieve these objectives through programs such as Overdose Data to Action 33 and the 2022 CDC Clinical Practice Guideline for Prescribing Opioids for Pain. 34Examples include supporting capacity building at state, local, and community levels, increased MOUD provision and access, and naloxone distribution.Increasing MOUD access may occur through a variety of linkage to care and retention mechanisms, such as [−36.6%]) (Figure5).Decreasing the fatal overdose rate by 50% among persons with OUD not receiving MOUD resulted in an additional 16.8% decrease in fatal overdoses compared with a similar intervention among persons receiving MOUD (from −13.1% to −29.9%).Decreasing fatal overdose rates had minimal association with OUD and MOUD prevalence but increasing MOUD initiation rate a Increased MOUD initiation and decreased early-and late-stage OUD recurrence.bModelestimates are cumulative over the time horizon of simulated public health interventions (January 1, 2021, to December 31, 2023).cModelestimatesindicateprevalence at the end of the simulation (December 31, 2023).JAMA Network Open | Substance Use and AddictionPublic Health Interventions and Overdose-Related Outcomes Among Persons With OUD JAMA Network Open.2024;7(4):e244617.doi:10.1001/jamanetworkopen.2024.4617(Reprinted)April3, 2024 6/14 Downloaded from jamanetwork.combyguest on 04/06/2024 with OUD not currently receiving treatment with MOUD demonstrated the largest reduction in both nonfatal and fatal opioid overdoses due to the high proportion of persons with OUD not receiving treatment with MOUD.While efforts to increase MOUD initiation and decrease OUD recurrence had limited association with fatal opioid overdoses over the short-term period simulated here (2021 and 2023), these interventions were associated with reducing OUD prevalence and increasing MOUD prevalence.Results of scenario A, which combined increasing MOUD initiation with decreasing OUD recurrence, estimated that OUD prevalence could be reduced by up to 23% and MOUD prevalenceFigure 3. Percentage Change in Projected Model Outcomes Relative to the Baseline Scenario for Fatal Overdoses, Opioid Use Disorder (OUD) Prevalence Without Medications for Opioid Use Disorder (MOUD), and MOUD Prevalence Under Model Scenario B a a Decreased fatal overdoses and decreased OUD recurrence.bModel estimates are cumulative over the time horizon of simulated public health interventions (January 1, 2021, to December 31, 2023).cModel estimates indicate prevalence at the end of the simulation (December 31, 2023).could 2,[38][39][40]Furthermore, we obtained fatal overdose rate estimates among persons receiving MOUD from a meta-analysis of studies 41 published before IMF driving increasing mortality rates.As a result, it is possible that our analysis underestimates the rate of fatal overdoses among persons in MOUD treatment and overestimates fatal overdose rates among persons with OUD not receiving MOUD treatment.Third, MODIPHI is an aggregate model and cannot track individual histories.As a result, the model does not differentiate probability of treatment initiation and receipt based on history of MOUD.Furthermore, the model does not account for the increased risk of fatal overdose among persons with past nonfatal overdoses.
22,28c Health Interventions and Overdose-Related Outcomes Among Persons With OUD Downloaded from jamanetwork.combyguest on 04/06/2024 project outcomes over a longer time horizon, which could reveal additional dynamics.Furthermore, the COVID-19 pandemic had impacts on data collection as well as health care access and utilization during 2020 and the projected 2021 to 2023 period, which may affect these results in unknown ways.However, we conducted sensitivity analyses on the 2020 calibration target of OUD prevalence reported in NSDUH and found low sensitivity, with identical results to those reported in our scenario analyses (eAppendix 14, eTable 12, and eFigure 9 in Supplement 1).Fifth, due to sample size constraints in the MOUD Study, the model cannot distinguish between specific MOUD treatments.Furthermore, responses in the MOUD Study were self-reported and subject to social desirability bias, excluded certain age groups in our study (ie, those aged 12-17 years), and excluded information about those who may have been lost to follow-up, dropped out of treatment, or died.Sixth, the study population represents persons with OUD, a patient population who are likely underdiagnosed or underestimated.22,28Additionally,proxyvariableswere used in SUDORS to identify possible OUD among opioid-involved overdose decedents.25Despitethis limitation, sensitivity analysis of the proportion of decedents with prior OUD in the SUDORS data showed our results to be robust (eAppendix 15 and eTable 13 in Supplement 1).Additionally, MODIPHI only represents persons with OUD and the model does not account for any future interventions that might reduce the incidence of OUD.

SUPPLEMENT 2. Data Sharing Statement
Dowell D, Ragan KR, Jones CM, Baldwin GT, Chou R. CDC clinical practice guideline for prescribing opioids for pain-United States, 2022.MMWR Recomm Rep. 2022;71(3):1-95.doi:10.15585/mmwr.rr7103a135.Worthington N, Gilliam T, Mital S, Caslin S. First responder assertive linkage programs: a scoping review of interventions to improve linkage to care for people who use drugs.J Public Health Manag Pract.2022;28(suppl 6): S302-S310.doi:10.1097/PHH.000000000000161136.Quinn K, Kumar S, Hunter CT, O'Donnell J, Davis NL.Naloxone administration among opioid-involved overdose deaths in 38 United States jurisdictions in the State Unintentional Drug Overdose Reporting System, 2019.Drug Alcohol Depend.2022;235:109467.doi:10.1016/j.drugalcdep.2022.109467The White House Briefing Room.Biden-Harris administration announces strengthened approach to crack down on illicit fentanyl supply chains.2023.Accessed February 21, 2024.https://www.whitehouse.gov/briefingroom/statements-releases/2023/04/11/fact-sheet-biden-harris-administration-announces-strengthenedapproach-to-crack-down-on-illicit-fentanyl-supply-chains/41.Sordo L, Barrio G, Bravo MJ, et al.Mortality risk during and after opioid substitution treatment: systematic review and meta-analysis of cohort studies.BMJ.2017;357:j1550.Estimation of Proportion of Decedents who Died From an Opioid-Involved Overdose With Prior Opioid use Disorder Using Data From the State Unintentional Drug Overdose Reporting System (SUDORS) in 26 Jurisdictions, 2019 to 2020 eTable 7. Estimation of Number of National Overdoses Among Individuals With Opioid use Disorder (OUD) Using Data From the State Unintentional Drug Overdose Reporting System (SUDORS) and National Vital Statistics System (NVSS) Mortality Data, 2019 to 2020 JAMA Network Open | Substance Use and Addiction Public Health Interventions and Overdose-Related Outcomes Among Persons With OUD Calibrating to Overall OUD Prevalence eAppendix 9. Calibrating to MOUD Prevalence eAppendix 10.Calibration Results eTable 8. Calibrated Parameter Ranges and Calibrated Values eFigure 1. Calibration Results Showing Pre-and Postcalibrated Model Outcomes in Relation to Targets eAppendix 11.Validating Nonfatal Overdoses and Reporting Overdoses eAppendix 12. Multiple Intervention Scenarios eTable 9. Summary of Annual Model Outcomes Under the Baseline Scenario eTable 10.Model Outcomes for Baseline and Intervention Scenarios When Interventions are set to the Maximum Effect Size Modeled.Outcomes Reported at end of Model Simulation (end of 2023) eTable 11.Summary of Relative Impact on Model Outcomes Across all Multiple Intervention Scenarios eFigure 2. Comparison of Model Outcomes From Scenarios A to D Over Time Relative to Baseline Scenario eAppendix 13.One-Way Parameter Sensitivity Analysis eFigure 3. One-Way Sensitivity Analysis Showing Association Between Parameter Perturbations Ranging From −100% to 200% and Fatal Overdoses Relative to Baseline Scenario eFigure 4. One-Way Sensitivity Analysis Showing Association Between Parameter Perturbations Ranging From −100% to 200% and Nonfatal Overdoses Relative to Baseline Scenario eFigure 5. One-Way Sensitivity Analysis Showing Association Between Parameter Perturbations Ranging From −100% to 200% and Number of Individuals With OUD not Receiving MOUD (OUD Prevalence) Relative to Baseline Scenario eFigure 6. One-Way Sensitivity Analysis Showing Association Between Parameter Perturbations Ranging From −100% to 200% and Number of Individuals Receiving MOUD (MOUD Prevalence) Relative to Baseline Scenario eFigure 7. One-way Sensitivity Analysis Showing Association Between Parameter Perturbations Ranging From −100% to 200% and Number of Individuals in Remission Relative to Baseline Scenario eFigure 8. One-way Sensitivity Analysis Showing Association Between Parameter Perturbations Ranging From −100% to 200% and Deaths From Other Causes Relative to Baseline Scenario eAppendix 14.Sensitivity Analysis of 2020 OUD Prevalence Data eTable 12. Summary of Annual Model Outcomes Under the Baseline Scenario Given Sensitivity Analysis of 2020 OUD Prevalence as a Calibration Target eFigure 9. Percentage Change in Projected Model Outcomes Relative to the Baseline Scenario for Fatal Overdoses, Nonfatal Overdoses, OUD Prevalence (Without MOUD) and MOUD Prevalence at the end of 2023 eAppendix 15.Sensitivity Analysis of SUDORS Data eTable 13.Summary of Annual Model Outcomes Under the Baseline Scenario Given Sensitivity Analysis of Increases in SUDORS Proportion of Decedents With Opioid-Involved Overdoses and Prior OUD eReferences 34. doi:10.1136/bmj.j155042.Stoové MA, Dietze PM, Jolley D. Overdose deaths following previous non-fatal heroin overdose: record linkage of ambulance attendance and death registry data.Drug Alcohol Rev. 2009;28(4):347-352.doi:10.1111/j.1465-3362.2009.00057.x43.Caudarella A, Dong H, Milloy MJ, Kerr T, Wood E, Hayashi K. Non-fatal overdose as a risk factor for subsequent fatal overdose among people who inject drugs.Drug Alcohol Depend.2016;162:51-55.doi:10.1016/j.drugalcdep.JAMA Network Open.2024;7(4):e244617.doi:10.1001/jamanetworkopen.2024.4617(Reprinted) April 3, 2024 13/14 Downloaded from jamanetwork.comby guest on 04/06/2024 eAppendix 8.