Access to and Engagement in Substance Use Disorder Treatment Over Time

This study, conducted in 2018, used data from commercial and Medicaid managed care plans (“Medicaid plans”) on three measures reported to the NCQA: identification of alcohol and other drug services, initiation of alcohol and other drug abuse or dependence treatment, and engagement of alcohol and other drug abuse or dependence treatment (defined earlier) (10).

Approximately 400 commercial plans reported the measure rates (392 in 2010 and 404 in 2016), with a combined enrollment of 111 million individuals in 2016. Between 79 and 186 Medicaid plans reported the measures (79 in 2010 and 186 in 2016), with a combined enrollment of approximately 53 million individuals in 2016. To put the numbers in context, in 2016, these plans enrolled 71% and 80% of the U.S. population with commercial and Medicaid insurance coverage, respectively. All plans provided coverage for substance use disorder treatment. Independent organizations audited the data before they were reported to the NCQA. The data were de-identified and did not require institutional review board approval.

We calculated the average rate of each measure among all plans in each year. To estimate the change from 2010 to 2016, we used the generalized estimating equation (GEE) with Huber-White standard error (SE) estimates and an exchangeable working correlation structure (11). The GEE (also referred to as a “population-averaged” model) estimates the average response over the population when there are repeated correlated measurements (e.g., a plan’s performance in years 1 and 2 are correlated). These results are summarized by using slopes (β; average change between two consecutive years) and SEs.

We conducted a Pearson correlation analysis of the identification rate in 2015 and of the initiation and engagement rates in 2016 to test whether increasing identification rates were associated with falling initiation and engagement rates. Last, we calculated the difference in measure performance between low-performing (10th-percentile performance rate) and high-performing (90th-percentile) plans in 2010 and 2016. These analyses help identify realistic short-term opportunities for improvement (if some plans can reach a high performance rate, it is reasonable to conclude that other plans can as well). All analyses were conducted separately for commercial and Medicaid plans, and statistical significance was set at p<0.05.

Figure 1 shows average commercial and Medicaid plan performance on the measures in each year (see figure in the online supplement for details about plan performance). Between 2010 and 2016, commercial plan identification of individuals with substance use disorders increased from 1.0% to 1.6% (β=0.09, SE=0.00, p<0.001), the initiation rate declined from 41.9% to 33.7% (β=−1.47, SE=0.07, p<0.001), and the engagement rate declined from 15.8% to 12.1% (β=−0.56, SE=0.04, p<0.001). GEE results indicated that, on average, commercial plan identification increased by 0.09 percentage points per year and that initiation and engagement rates declined by 1.47 and 0.56 percentage points per year from 2010 to 2016, respectively.

For Medicaid plans, the rate of identification of individuals with substance use disorders increased from 3.3% to 6.7% (β=0.42, SE=0.05, p<0.001), the initiation rate was relatively stable at 42.9% in 2010 and 40.9% in 2016 (β=−0.25, SE=0.18, p=0.16), and the engagement rate was also relatively stable at 14.2% in 2010 and 12.7% in 2016 (β=−0.11, SE=0.14, p=0.46). GEE results showed that, on average, the Medicaid plan identification rate increased by 0.42 percentage points per year, and initiation and engagement declined slightly per year, but the declines were not statistically significant. In 2016, Medicaid plans had a higher rate of initiation than commercial plans (40.9% versus 33.7%, respectively), and both types of plans had a similar rate of engagement (12.7% versus 12.1%, respectively).

To inform whether increasing identification rates could have contributed to falling initiation and engagement rates, we tested whether there was a negative correlation between identification in one year and initiation and engagement in the subsequent year. In commercial plans, there was no statistically significant correlation between identification and initiation rates and a small positive correlation (r=0.14, p=0.01) between identification and engagement rates. Medicaid plans showed similar results: there was no statistically significant correlation between identification and initiation of treatment; there was a small but not statistically significant positive correlation between identification and engagement.

There was a significant gap in performance on the initiation and engagement measures between low-performing (tenth-percentile performance rate) and high-performing (90th-percentile) plans (see online supplement for performance rates). For example, the engagement rate in 2016 was 4.8% in the tenth-percentile and 21.3% in the 90th-percentile Medicaid plans. However, even the 90th-percentile plans did not exceed 50% in initiation or engagement rates in 2016. Between 2010 and 2016, the gap in performance on identification between low- and high-performing commercial plans showed no change but narrowed for initiation and engagement for commercial plans. For Medicaid plans, the gap between low- and high-performing plans broadened for identification and narrowed for both initiation and engagement. The narrowed performance gaps in initiation and engagement rates were mostly due to a decline in high-performing plans’ rates.

In 2016, identification of substance use disorders was 1.6% in commercial plans and 6.7% in Medicaid plans. Epidemiologic data from the NSDUH show that the prevalence rate of substance use disorders is 7.0% for commercial health plans and 8.7% for Medicaid plans (author analysis of 2015 NSDUH data). This suggests that the identification rate or the diagnosis or treatment rate of substance use disorders was lower than the self-reported prevalence for commercial plans and comparable for Medicaid plans.

Although the rates of identification of substance use disorders increased in both commercial and Medicaid plans, the rates of treatment initiation and engagement deteriorated for commercial plans and showed no significant change for Medicaid plans. We found a small, positive relationship between identification and engagement rates for commercial plans and found no relationship between identification, initiation, and engagement rates for Medicaid plans, which indicates that the lack of progress in initiation and engagement should not be attributable to identifying a larger number of people with substance use disorders in the denominator.

The decline in initiation and engagement in treatment for commercial plans and a lack of improvement for Medicaid plans is concerning, given the availability of effective treatment for substance use disorders and the growing evidence that initiation and engagement in treatment are associated with lower mortality (3–8). It is also concerning that the narrowing of the gap in performance in initiation and engagement rates between low- and high-performing plans was mostly due to a decline in performance among high-performing plans, which suggests that even high-performing plans struggled to improve substance use disorder treatment rates. It is noted that 80% of members who initiated substance use disorder treatment dropped out of treatment after the initial one or two visits. Consistent with the literature, this suggests that it is challenging to engage members with substance use disorders in treatment. Implementing elements of the chronic care model (e.g., clinical case management for patients with complex symptoms and mobilizing social support) may help increase engagement in treatment (12).

A limitation of this study was that most data were collected before national recognition of the opioid crisis. Future studies should examine how responses to this crisis might relate to changes in measure rates. Another limitation of this study was that we could not examine whether health plan and regional characteristics were associated with better performance. These factors should be explored in future research.

Despite the increase in identification of members with substance use disorder, there has been no improvement in rates of initiation and engagement—measures that have been shown to be related to reduced mortality. Concerted action and leadership across the federal government, states, third-party payers, educational institutions, professional associations, and substance use disorder treatment providers are needed to develop and implement protocols that increase substance use disorder treatment initiation and engagement.