Our analysis sought to provide a rigorous estimation of the opioid user population by grounding prevalence estimates in a range of datasets that are not often applied to this question. Although even this diversified approach may not capture the full scope of opioid use, it does allow construction of a more comprehensive picture of opioid users in New York City, including prevalence, service utilization, and demographic characteristics.
There is clearly a need for expanded drug services for opioid users in New York City. Even using our most restrictive estimations, there were over 69,000 opioid users residing in New York City in 2006. While over 21,000 individual opioid users did initiate treatment, this represents a small fraction of the identified opioid user population. Moreover, because close to half of treatment admissions are mandated each year in New York City, including one-third by criminal justice authorities (i.e. courts, probation, parole), one can assume a considerable portion of this fraction did not initiate treatment voluntarily.
The current drug treatment system alone cannot be relied upon to accommodate the needs of this population. Though maintenance treatment with an opioid agonist medication, (methadone or buprenorphine), is the best evidence-based treatment strategy for opioid dependence [29
], not all of the opioid users identified here would necessarily qualify for, or accept, medication assisted treatment. Furthermore, accommodating these individuals within the existing substance abuse treatment system would involve doubling the capacity of New York City opiate agonist treatment programs, which currently serve approximately 37,500 patients [14
]. Buprenorphine treatment in general medical settings provides an alternative means of treatment expansion, but slow uptake among physicians has thus far limited its reach.
While many opioid users would undoubtedly benefit from traditional drug treatment services, the availability of treatment slots is not the only barrier to engaging them in effective evidence-based care. Although these data sources largely capture the negative sequelae of opioid use (health problems, incarceration) that are more prominent among those with active substance use disorders, even non-dependent and early-stage users are vulnerable to drug-related problems, and may also have been identified here. Similarly, these data may include opioid users that were already enrolled in drug treatment, but nonetheless experienced problems related to drug use. Some opioid users could be served by less intensive treatment interventions, such as brief interventions or pharmacotherapy, provided in healthcare or community settings that are not traditionally considered part of the drug treatment system [30
]. Many would benefit from harm reduction approaches to prevent the negative consequences of opioid use. A greater diversity, as well as a greater number, of treatment providers would thus be needed to deliver care to the full range of opioid users identified here. Public health authorities could apply this data to assess and orient the structural configuration of existing treatment services, and to expand engagement, access points, service types and modalities, and venues to better reach the sizeable under-served opioid user population.
Our approach does have a number of limitations. Most notably, we were unable to match individual opioid users across data sources. Capture-recapture analysis, which relies on identifying unique individuals across multiple datasets, can potentially improve the accuracy of prevalence estimation, and has been applied in municipalities outside the US to estimate populations of illicit drug users [32
]. We chose not to use this approach because accessing fully identified individual-level data from multiple data sources was not feasible at the time of this analysis. Similarly, although it would be desirable to perform a validity check of the final estimates using an alternate method, this was beyond the scope of the present study. However, our findings do have face validity in light of prior estimates of the New York City opioid user population [1
Our estimation of the medical services population relied on hospital data from inpatient and emergency room presentations, and did not capture individuals who utilized only ambulatory care services. Analysis of insurance claims data could contribute that information, but was not accessible for this analysis. Our data also fail to capture services provided within the Veteran’s Administration (VA) system.
There are several limitations inherent to the datasets themselves. Substance use is often not identified in medical settings [36
] and substance use and other mental health diagnoses are poorly captured by medical administrative data [40
]. Additionally, none of the datasets that rely on service utilization (drug treatment, detoxification, or hospital inpatient services) capture those with less problematic opioid use, who have not experienced the health and social sequelae of addiction that often drive users to seek care. As a result, we relied on the NSDUH for the general population estimation of opioid use, despite its known limitations [16
]. Finally, our analysis is restricted to a single year-long period (2006), and thus does not capture trends in opioid use over time. This may be a particularly significant limitation with respect to prescription opioid use, which has been rising rapidly nationwide [9
Despite these limitations, our study contributes to the understanding of the prevalence of opioid use in New York City by rigorously defining the dimensions and characteristics of the opioid user population. More sophisticated epidemiologic analyses of these multiple datasets, including capture-recapture and multiplier methods, could provide a more accurate estimate of the hidden population of opioid users who do not appear in any existing data sources. Our goal here was to take a first step toward that end by defining the number and range of individuals who met our case definition of problematic opioid use in health-related datasets.