Treatment for opioid use disorder in the Florida medicaid population: using a cascade of care model to evaluate quality

Abstract

Background:

A cascade of care (CoC) model may improve understanding of gaps in addiction treatment availability and quality over current single measure methods. Despite increased funding, opioid overdose rates remain high. Therefore, it is critical to understand where the health-care system is failing in providing care for people with opioid use disorder diagnoses, and to assess disparities in treatment receipt.

Objective:

Using a CoC framework, assess treatment quality and outcomes for opioid use disorder in the Florida Medicaid population in 2017/2018 by demographics and primary vs. secondary diagnosis.

Methods:

Data from Florida Medicaid claims for 2017 and 2018 were used to calculate the number of enrollees who were diagnosed, began medication, were retained on medication for a minimum of 180 days, and who died.

Results:

Only 28% of those diagnosed with OUD began treatment with an FDA approved medication. Once on medication, 38% of newly diagnosed enrollees were retained in treatment for 180 days. Those who remained in treatment for 180 days had a hazard ratio of death of 0.226 (95% CI = 0.174 to 0.294) compared to those that did not initiate treatment, a reduction in mortality from 10% without care to 2% with care.

Conclusions:

Initiating medication after diagnosis offers the greatest opportunity for intervention to reduce overdose deaths, though efforts to increase retention are also warranted. Analyzing claims data with CoC identifies system functioning for specific populations, and suggests policies and clinical pathways to target for improvement.

Introduction

As opioid overdose deaths have increased in the United States (1–3), resources have been directed to increasing capacity, particularly increasing the number of providers who are capable of prescribing buprenorphine (4–6). Rulemaking (7) and proposed legislation (8) has been directed toward increasing the number of patients to whom each provider can prescribe. These efforts at increasing availability are essential but not sufficient to ensure more people receive necessary treatment. Eisenberg (9) described a series of “voltage drops” that occur between provision of health-care coverage and provision of quality care. Continuous coverage, adequate provider capacity, ease of access in terms of location and time, treatment that addresses patient needs, preferences, and abilities to pay for, obtain, and use medications on schedule, and ongoing efforts to support treatment adherence are all necessary to provide care that improves health.

A treatment cascade is one way of assessing where gaps or voltage drops in the provision of effective care exist. Several authors have proposed using a cascade of care as a core measure for monitoring of addiction treatment system functioning and providing actionable feedback (10–13). More recently cascades of care have been created for systems of care in at least two states (CO (14) and RI (15)) and one Canadian province (16). A cascade of care begins with the population prevalence, then calculates the number/percent of the population who are identified or diagnosed as having a condition, the number/percent who begin treatment, the number/percent who continue in or complete treatment and the number/percent who have positive treatment outcomes. Use of a cascade of care can help system planners to better understand where there are system failures and develop targets for improvement.

Administrative claims data are perhaps the most readily available source of information that can describe systems-level quality and can be used to describe the process of a cascade of care. Claims data are available at a system level and are useful in tracking process and some limited outcome measures of performance. Therefore, claims data can be used for all but the first – prevalence – and possibly the last step – outcome – of a cascade of care. Prevalence of those with opioid use disorder can generally be estimated from national or local survey data, though there have been criticisms of these methods and alternative methods proposed (17). Treatment outcome can be estimated from surveys (15) or may need to be inferred from expected links between process and treatment outcomes from the literature (15,18,19). Another alternative is to look at proxy measures such as reduced health spending, or to examine reduction in negative outcomes that can be measured in claims data such as reduced hospitalization, emergency department visits or death.

Our goal was to test an approach of data extraction and analysis based on previously defined terms that could be widely adopted and generalizable. The potential audience is states and payers wishing to assess the function of their system as it relates to the provision of treatment for opioid use disorder, and to use the methods to identify opportunities for improvement in system-level functioning. We did this by applying the cascade of care approach to evaluate the quality of care for individuals with opioid use disorder in a single state, Florida. We examined the location of diagnosis and the type of diagnosis (primary versus secondary) as a way of disaggregating opportunities for intervention. A primary diagnosis implies a patient seeking or being offered treatment for opioid use disorder, where a secondary diagnosis implies treatment for some other condition as primary, but OUD being a significant contributor or important enough to note in a medical claim file. We examined opportunities for improvement from both perspectives as well as whether there were differences in treatment receipt by age, sex, and race/ethnicity.

Methods

Data

Prevalence was measured using the National Survey on Drug Use and Health (NSDUH) latest available (2016–2017) restricted data set online query tool applied to Florida. The online query tool is available at https://rdas.samhsa.gov/#/survey/NSDUH-2016-2017-RD02YR.

All other data for this analysis was Medicaid administrative data. We used FL Medicaid institutional and office claims and encounters (including managed care and fee for service claims), and pharmacy claims for 2017 and 2018.

As Florida did not expand Medicaid under the Patient Protection and Affordable Care Act (20), the population of Florida Medicaid is made up of people who are categorically eligible. Eleven percent of Florida adults age 19–64 are covered by Medicaid compared to 14% nationally (21,22) but the percent of adults 65 or over who are dually covered by Medicaid and Medicare is similar to the national average of 20% (21,22).

Measures

Prevalence was computed as a cross tabulation in the Substance Abuse and Mental Health Data Archive restricted use data analysis system online analysis tool. The query was a crosstab of opioid abuse or dependence and Medicaid insurance with a control variable of state equal to Florida and using survey weights in the 2016–2017 2-year restricted data set.

Population is all Medicaid recipients with an opioid use disorder diagnosis in the calendar year 2017. For consistent data collection and result comparison, recipients were required to have at least 11 months of Medicaid coverage or full coverage until death in 2017. For inclusion in the sample, we required 6 months of continued coverage following an opioid use disorder treatment (if the recipient had at least one treatment) or opioid use disorder diagnosis (if the recipient did not have treatment). The only exception was when a recipient died during this six-month term; then, full Medicaid coverage until their death was used to meet the Medicaid coverage requirement and inclusion in the sample; 25,866 individuals met these criteria. The analyses herein are based on these 25,866 service recipients whose service record was measured for the period January 1, 2017 through December 31, 2018.

Diagnosis Codes: Opioid use disorder diagnosis is based on ICD-10-CM codes with all forms of F11.xxx, including F11.11 (in remission) and F11.90 (opioid use unspecified, uncomplicated) which are sometimes left out of analysis such as this, but which we believe warrant inclusion based on practitioners sometimes not differentiating between these codes. All diagnoses (primary or any secondary) in claims (institutional or office) were used to define opioid use disorder diagnoses. A recipient was considered as having opioid use disorder if the recipient had at least one claim (including all claims from institutional and office settings) with ICD-10-CM code F11.xxx in the calendar year 2017. Service recipients may have had diagnoses of opioid use disorder prior to 2017; however, receiving a diagnosis in 2017 and their subsequent course of treatment (if any) was the focus of this analysis. We also conducted some analysis using a wash out period of the final three months of 2016 to assess the outcomes for people with new diagnosis only. We refer to these individuals as newly diagnosed. We also coded individuals based on whether their opioid use disorder was listed as primary or secondary.

Treatment Codes: Treatment receipt is captured from service or pharmacy claims with methadone receipt measured from clinical service claims, buprenorphine from pharmacy claims and injectable naltrexone from pharmacy claims. In Florida, injectable naltrexone is most often purchased in a pharmacy and delivered to the provider who provides the injection. We did not count service claims separately as it would be duplicative and service claims are often coded simply as an office visit. Methadone treatment is identified by CPT codes H0020 in combination with the modifier ‘HG’ (indicates Opioid Addiction Treatment Program) and an ICD-10-CM code Fll.xxx from clinical claims. The restriction is necessary as the code H0020 could be used for treating other substance use disorders. We did not include service claims for counseling or other treatment that was not associated with an FDA approved medication to treat opioid use disorder as research does not support these services as effective treatment in the absence of medication (23,24).

For each enrollee, the number of claims for methadone, buprenorphine and injectable naltrexone are calculated from the initial diagnosis and treatment claim in 2017. Buprenorphine includes all types of buprenorphine products including combined buprenorphine-naloxone products and sublingual and all long-acting buprenorphine products but excludes products generally used only for the treatment of pain.

Treatment initiation is defined as the earliest medication treatment (buprenorphine, methadone, or injectable naltrexone) following diagnosis. The treatment initiation date is used as an index date to calculate the number of treatments and whether or not the person had continuous treatment. Following NQF definitions (25), if a person had at least 24 Methadone treatment encounters (billed weekly) in addition to the initial index encounter within 180 days of treatment initiation, then we considered the person as having continuous methadone treatment; if an individual had prescriptions for at least 173 out of 180 days, then the individual is considered as having had continuous buprenorphine treatment; if a person had at least four injectable naltrexone prescriptions following the initial index prescription, then the person is considered as having had continuous injectable naltrexone treatment. These measures allow for some very brief lapses in care during the period of study.

We defined ‘short term follow-up’ as having more than the initial treatment but not meeting the criteria for the definition of “continuous treatment” as defined for each medication above. The ‘initial-only’ group is comprised of those individuals who had a single treatment, but no further treatment recorded in the following 180 days. Individuals in the ‘no-treatment’ group were given a diagnosis of an opioid use disorder with no medication for opioid use disorder (MOUD) initiation following diagnosis.

Analysis

All claims in 2017 were used to identify opioid use disorder diagnoses and claims for 2017 and 2018 were used for service encounters or prescription claims. Demographics (date of birth, gender, race, dually eligibility, Medicaid eligible days) are from Medicaid enrollment and eligibility data. Initial and follow-up treatments are calculated on an individual level. After pulling all claims with a diagnosis of opioid use disorder, all measures were matched by enrollee unique ID (Medicaid ID) to create the analysis datasets. Inclusion criteria were applied to each measure for consistency. For each diagnosis, age-at-diagnosis is calculated and the individual is included if 18 years of age or older and enrolled for at least 11 months (or 330 days) in the time period. To calculate treatment utilization, Medicaid enrollment of at least 180 days from the initial diagnosis or treatment was required.

Chi-Square and Mantel-Haenszel (MH) Chi-Square were used to detect the association between measures and factors. We used a Cox proportional hazards model for survival analyses adjusting for age, race, ethnicity, and sex. We used SAS Version 9.4 for all analyses.

Results

Prevalence

Calculation of the expected prevalence of opioid use disorder for Florida Medicaid recipients from the NSDUH 2016/2017 Restricted Use Data Files online analysis tool is 1.83% (confidence interval 0.96–3.21%) or approximately 28,370 persons when the percentage is applied to the population based on our inclusion criteria.

Diagnosis

The number of adults (18 or older) who had at least 11 months of Medicaid coverage or full coverage until death in Florida in 2017 was 1,553,285. Of these Medicaid recipients, 27,568 (or 1.8%) had a diagnosis of opioid use disorder in their record in 2017. Of these 27,568 recipients, 25,866 met the inclusion criteria of having at least 6 months of coverage following their initial treatment (if the patient had opioid use disorder treatment) or initial diagnosis (if the patient did not have treatment).

Primary versus secondary diagnosis

Of the service recipients who received an opioid use disorder diagnosis in 2017, the type of diagnosis received, classified as Primary vs. Secondary/Other, varied predictably by setting. Those who received a primary diagnosis (44%) were found, most commonly, to have received it in community-based substance use disorder or mental health treatment settings (57%). In contrast, only 1% of individuals with a secondary opioid use disorder diagnosis received it from these settings; instead they commonly received their diagnoses in hospital or physician-office settings (90%).

All demographic factors, except rural/urban location, are significantly associated with whether service recipients received a primary or secondary diagnosis of opioid use disorder in 2017. Younger, female, and white, Asian, and American Indian individuals were more likely to have a primary diagnosis of opioid use disorder whereas Blacks and the Medicare eligible were more likely to have only a secondary diagnosis (see Table 1).

Table 1.

Characteristics of service recipients and service by diagnosis type and whether newly diagnosed in 2017 N = 25,866.

Characteristics	New diagnosis in 2017 (no treatment in Q4 2016)	Categories	At Least one Primary OUD Diagnosis		Secondary/Other OUD Diagnosis (No Primary) OUD		Association between Primary OUD and Factor
			N – SR	% of SR	N- SR	% of SR	MH Chi-square value (P-value)
Age group	No	18–25	181	93.3%	13	6.7%	25.9 (<.001)
		26–49	3,724	95.4%	180	4.6%
		50–64	712	89.7%	82	10.3%
		65 and up	48	90.6%	5	9.4%
	Yes	18–25	574	54.7%	476	45.3%	713.5 (<.001)
		26–49	5,287	51.9%	4,909	48.1%
		50–64	2,693	35.2%	4,961	64.8%
		65 and up	551	27.3%	1,470	72.7%
Dual Eligible	No	Dually Eligible	1,676	92.9%	128	7.1%	10.9 (=.001)
		Medicaid Only	2,989	95.2%	152	4.8%
	Yes	Dually Eligible	5,074	37.3%	8,535	62.7%	616.1 (<.001)
		Medicaid Only	4,031	55.1%	3,281	44.9%
Gender	No	F	3,492	94.6%	201	5.4%	1.32 (.25)
		M	1,173	93.7%	79	6.3%
	Yes	F	5,933	44.6%	7,368	55.4%	17.5 (<.001)
		M	3,172	41.6%	4,448	58.4%
Race	No	Asian	17	94.4%	1	5.6%	9.57 (.002)
		Black	143	88.8%	18	11.2%
		Hispanic	314	94.6%	18	5.4%
		American Indian or Native	38	97.4%	1	1.6%
		Not Determined	347	89.9%	39	10.1%
		Other	62	95.4%	3	4.6%
		White	3,744	94.9%	200	5.1%
	Yes	Asian	41	51.3%	39	48.7%	110.9 (<.001)
		Black	891	33.6%	1,758	66.4%
		Hispanic	872	44.1%	1,107	55.9%
		American Indian or Native	25	43.9%	32	56.1%
		Not Determined	1,401	40.6%	2,051	59.4%
		Other	145	43.8%	186	56.2%
		White	5,730	46.3%	6,643	53.7%
Rural/Urban	No	Rural	199	93.0%	15	7.0%	0.76 (.38)
		Urban	4,466	94.4%	265	5.6%
	Yes	Rural	534	48.0%	578	52.0%	9.68 (.002)
		Urban	8,571	43.3%	11,238	56.7%
Initial Setting	No	Community-based Care in Primary SUD or MH treatment setting	3,474	99.8%	6	0.2%	473.4 (<.001)
		Physician (Office-based)	521	79.1%	138	20.9%
		Hospital	135	57.2%	101	42.8%
		Other	535	93.9%	35	6.1%
	Yes	Community-based Care in Primary SUD or MH treatment setting	1,746	92.2%	147	7.8%	264.6 (<.001)
		Physician (Office-based)	3,009	37.2%	5,083	62.8%
		Hospital	2,257	28.9%	5,542	71.1%
		Other	2,092	66.7%	1,044	33.3%
Total	No		4,665	94.3%	280	5.7%	4,148.6 (<.001)
	Yes		9,105	43.5%	11,816	56.5%

^*: ‘Other’ includes diagnoses made in a range of settings including Federally Qualified Health Centers (FQHCs), Rural Health Clinic, Assistive Care, Case Management Agency. Among those collapsed in the ‘Other’ category, only ‘Independent Lab’ comprised greater than 5% of the total sample.

Treatment initiation, engagement and continuation

More than half of the people who were newly diagnosed during 2017 received only a secondary diagnosis (57%) and of these only 2% received any MOUD. Of those who received a primary diagnosis, and therefore presumably at least were offered treatment, only 24% began MOUD (Table 2). Most of the people in the continuous treatment group had begun treatment prior to the study period and continued their care into the study period. Of those individuals who received a new primary diagnosis and initiated MOUD, 42% remained in care for at least 180 days in the follow-up period.

Table 2.

Treatment initiation and retention by diagnosis type and whether newly diagnosed in 2017.

Treatment Type	New diagnosis in 2017 (no Treatment in Q4 2016)	Primary OUD		Secondary/Other OUD (N = 12,096)		Association among treatment type and OUD type: MH Chi-square value (P-value)
		N-Enrollees	%-Enrollees	N-Enrollees	%-Enrollees
Continuous Follow-up Treatment	No	2,813	98.8%	35	1.2%	654.6 (<.001)
Short-term Follow-up Treatment		1,637	93.8%	108	6.2%
Initial-only Treatment		85	59.9%	57	40.1%
No MOUD Treatment		130	61.9%	80	38.1%
Continuous Follow-up Treatment	Yes	923	97.8%	21	2.2%	2,423.4 (<.001)
Short-term Follow-up Treatment		1,030	90.2%	112	9.8%
Initial-only Treatment		241	62.0%	148	38.0%
No MOUD Treatment		6,911	37.5%	11,535	62.5%

In Florida, Medicaid-funded newly diagnosed individuals are about equally likely to receive methadone as buprenorphine, but the higher retention rate for methadone treatment leads to a 3:1 ratio of people in methadone treatment compared to buprenorphine for individuals in the continuous care group. Fifty-six percent of newly diagnosed individuals who began methadone treatment continued for 180 days, while 18.7% of those that began treatment with buprenorphine continued for 180 days. This ratio might change somewhat if we used less stringent standards for the definition of continuous treatment as prior research has indicated that results are sensitive to measure specification (26). Very few individuals (only 14) received injectable naloxone, and none received more than the initial treatment.

Mortality as an outcome

The mortality rate of service recipients is a key measure in assessing the effectiveness and safety of treatment processes. Death dates were collected from the Medicaid enrollment database and matched to the study population. We included dates of death through 12/31/18.

Mortality is strongly associated with treatment duration, as mortality of the ‘no treatment’ group is more than 4 times the rate of those in continuous care. Individuals in continuous care had the longest survival rates. Though it is possible that individuals in the other treatment groups (short, initial, no treatment) may have died before receiving the dose of care required to be included in the continuous category, even after excluding individuals deceased with less than 180 days coverage, the mortality in the continuous follow-up group demonstrated the lowest mortality rate and the ‘no treatment’ group continued to demonstrate the highest mortality rate.

After controlling for demographics, the hazard ratio between the continuous follow-up treatment group and the no MOUD group is 0.226 (95% confidence interval 0.174 to 0.294) (Figure 1). There is no statistically significant difference between the hazard of death for the no-treatment group and the treatment initiation only group.

Figure 1.

Adjusted survival curves of patients with Opioid Use Disorders by treatment. The difference between Initial Only and No treatment is not significant. All other differences are statistically significant at p ≤ 0.05. Treatment type definition: Continuous Treatment – 173/180 days of MOUD; Short Term Follow up – more than one, but less than 173/180 days of MOUD; Initial-Only – one medication prescription or service following diagnosis, but no further MOUD; No-treatment – no medication initiated during 180 days following diagnosis.

The Florida medicaid opioid use disorder treatment cascade

Figure 2 displays an opioid use disorder treatment cascade for the state of Florida for Medicaid enrollees newly diagnosed in calendar year 2017, both overall and by whether the diagnosis was primary or secondary. ‘Prevalence’ reflects the number of Medicaid enrollees, 18 or older, expected to have an opioid use disorder based on NSDUH survey data analysis. ‘Identification’ represents the number of adult Medicaid enrollees who received a diagnosis in 2017. ‘Treatment initiation’ represents the number of those identified who received at least one medication service (such as a dosing visit to a methadone clinic) or prescription following the diagnosis. ‘Treatment continuation’ represents the number of enrollees who received medication for at least 180 continuous days as defined previously in 2017–2018.

Figure 2.

Opioid treatment cascade for Florida Medicaid population, newly diagnosed in 2017. Prevalence is from National Survey on Drug Use and Health online restricted data analysis system; all other data is from 2017 and 2018 Florida Medicaid claims. Prevalence is expected prevalence of active opioid use disorder in prior year; Identification, Treatment Initiation and Continuation are for those newly diagnosed in 2017 (no prior diagnosis or treatment in final quarter of 2016).

Discussion

Demographic differences

Older individuals and those who are black are less likely to receive a primary diagnosis and consequently are less likely to receive treatment for opioid use disorder. People who are dually eligible for Medicaid and Medicare are also less likely than people who are Medicaid eligible only to receive a primary diagnosis of opioid use disorder.

Importance of location and type of diagnosis

For Florida Medicaid recipients, where an individual initially receives a diagnosis and whether that diagnosis is labeled as primary or secondary to another injury or condition has significant life consequences. Service recipients who are diagnosed in a general medical or hospital setting and receive their diagnosis as secondary to another injury or condition are highly unlikely to receive MOUD. The consequences of not initiating care are dire, as the death rate is higher for each missed step of the treatment cascade. Given the high death rate, an opioid use disorder diagnosis at any level should be considered of enough concern to warrant initiation on a medication protocol regardless of whether the diagnosis is secondary to another condition. These results indicate a possible missed opportunity for treatment initiation that should be pursued and that results in disparities of care for people who are older, disabled, and black.

Notable milestones in the cascade of care

In 2017, Florida Medicaid enrollees who had an opioid use disorder had a high probability of having been diagnosed when the prevalence is estimated using the National Survey on Drug Use and Health. However, only about 28% of those who received a diagnosis went on to initiate an FDA approved medication to treat their opioid use disorder. Care continuity is more common than not once treatment has begun, with 53% of those who were in care in 2017 having remained in care for at least 180 days. Retention was highest for recipients of methadone treatment with 66% of individuals on methadone during the study period retained in treatment for at least 180 days. The numbers of people receiving injectable naltrexone was not high enough to draw conclusions regarding retention rates.

Retention in care is inversely related to probability of death

There is a clear directional relationship between the length of retention in care and the probability of death. Ten percent of Florida Medicaid recipients with a diagnosis of opioid use disorder in 2017 that did not receive treatment died during the period of study. Patients in the continuous treatment group had a 2% death rate and a hazard ratio of death of .226 compared to those who did not receive treatment whose death rate was five times higher. The nature of the Florida Medicaid population with its extreme poverty and high percentage of enrollees with disabling conditions make it by definition a population with poor health and high risk, but 10% of the population dying in a two-year period is unprecedented and vastly higher than previous reports (27,28).

Limitations

There are a number of limitations impacting interpretations made in this study. The primary limitation is that the prevalence figure from the national household survey data is probably an underestimate as it leaves out the population of people who are institutionalized and those who are homeless and unsheltered. The absence of a group of people with high probability of having opioid use disorder and being potentially in our population of study probably creates an underestimate of prevalence and consequently an overestimate of the percent of the population that is diagnosed. Other methods have estimated the undercount of opioid use disorder diagnosis by the NSDUH to be as much as 400%.

The Florida Medicaid population is not representative of the total population of people with opioid use disorder as it is extremely poor and more likely to have other disabling conditions.

We limited our analysis to Medicaid enrollees who had at least 11 months of coverage in the period and 180 days of coverage following diagnosis. This was necessary to avoid assuming people who merely changed payers had dropped out of care. However, the churn in Medicaid is high and we lost over a third of the population of Medicaid enrollees for the time period of study due to the decision to limit the sample by coverage criteria. Examining the effect of loss of or change in coverage on treatment continuity would also be of value as it impacts a large number of people.

Since individual cause of death is not available for this analysis, death data represents all causes, which may or may not be directly related to opioid use disorder. Nevertheless, after adjusting for age and other demographic factors, the hazard ratio for death decreases the longer a service recipient remains in treatment. Cause of death by various treatment groups should be researched further.

Using the medicaid cascade of care for improving treatment delivery

We believe the cascade of care methodology used here is valuable for policy makers, medical, and social service practitioners. For policy makers, it can be used to prioritize limited outreach and treatment resources, recognizing that those at highest risk of dying do not have a primary opioid use disorder diagnosis, do not receive any treatment, and in Florida are more likely to be Black and older. The cascade of care could also be used as a basis upon which sophisticated simulation models of the opioid epidemic could be built to improve decision making for policy makers accounting for local context while using standard measures and data elements. The cascade of care framework has been used effectively in addressing gaps in HIV prevention and treatment through model development. Similar agent-based modeling approaches now being used in decision-making for ending the HIV epidemic (29) can provide guidance on key decisions to reduce mortality, improve recovery, and prevent opioid use disorder.

Clinicians who provide service recipients a secondary diagnosis of opioid use disorder should be aware of their high risk of mortality and, if unable to provide treatment themselves, should ensure linkage to appropriate care. Our study demonstrates that a large proportion of those who enter treatment do continue in ongoing care. This information can counter the commonly held view that most service recipients do not respond to treatment. Such a positive message around recovery may encourage individuals to enter and continue treatment.

Conclusion

Using a cascade of care approach to evaluating claims data for opioid use disorder was feasible and informative. The findings yield critical insights for systems leaders and policymakers. We were able to identify differences in treatment utilization based on where a diagnosis was made and to link retention in treatment to a primary outcome – death. Policy makers can use a cascade of care as a dashboard for systems planning, resource allocation and quality improvement activities.

The issues raised in our analysis are important for planning purposes in Florida and should be both replicated and examined in other states. We identified subpopulations who are less likely to receive treatment including people over age 50, people who are Black, and those who are dually Medicaid and Medicare eligible. Our study points to a need for better care coordination between general medical (physician offices, hospitals, ED) and specialty care providers. An opioid use disorder diagnosis, regardless of whether it is primary or secondary or who makes the diagnosis requires treatment. To leave these patients untreated can be a death sentence.