Commercial and Medicare Members
Our sample of Commercial members for 2017 included 347, 471 individuals compared to 43, 174 members for the Medicare sample. The number of members in the primary SUD sample meeting the criteria for a substance abuse problem was also larger (n = 3494 vs. n = 1152). During 2017, Medicare members showed a higher rate of visits with a substance abuse diagnosis (2.6%) and a lower rate of early follow-up visits (16.1%) compared to Commercial members (0.9% and 31.4%, respectively). Our SUD sample of 2017 Medicare members was 42.8% male and included 9.4% who were age 65 and under compared to 49.8% male and 94.1% age 65 and under for Commercial members. The proportions of urban and rural residents were about equal.
Pharmacy Claims
Using a sample of claims level data from the first quarter of 2017, the percentage of prescriptions for several classes of drugs among the SUD sample are shown below in Table 1. It is noteworthy that the frequency (percentage) of prescriptions for opioid drugs is higher for the Medicare group, while the percentage of prescriptions for medically assisted treatment (MAT) for opioid problems is lower for this group.
Comparing primary care providers to other providers, the pattern was also different for Commercial and Medicare members in the SUD sample for prescriptions during the first quarter of 2017. For Commercial members, 57% of all prescriptions were from primary care providers, including 41% of opioid prescriptions and 47% of prescriptions for opioid medically assisted treatment (MAT). About half of antidepressant prescriptions were from primary care providers. For Medicare members, more opioid prescriptions were from primary care providers (61%), while the percent of opioid MAT prescriptions was similar to Commercial members (44%). Medicare antidepressant prescriptions from primary care providers were also higher (68%) compared to Commercial antidepressant prescriptions (51%).
Comorbid Conditions
Comorbid conditions among members included in the SUD sample compared to members not included in this sample were different for Commercial and Medicare members. Among Commercial members, 60% of members included in the SUD sample had a mental health diagnosis compared to 28.4% not included in this sample. Other differences among several selected comorbid conditions included: cardiovascular- 45% vs. 31%, hepatobiliary - 14% vs. 6%, and nutrition/metabolic - 64% vs. 54%. Among Medicare members these differences were even more pronounced and included: mental health diagnosis - 72% vs. 37%, vascular - 70% vs. 46%, diabetes - 46% vs. 31%, lung - 78% vs. 48%, neurologic - 64% vs. 38%, and notably cognitive disorders - 38% vs. 17%.
Commercial Members
Since the sample of Commercial members was much larger and the number of members meeting the criteria for a substance abuse problem was also larger, we focused on the commercial sample for the remaining analyses.
Primary SUD Sample Diagnostic CodesUsing a sample of initial claims for 2017, for Commercial members included in the SUD sample, ICD –10 diagnostic codes included 59% alcohol abuse and dependence codes, 9.5% opioid codes, and 18% of codes for cannabis abuse and dependence. Thus, a majority of claims in this sample were for alcohol abuse and dependence problems.
Primary SUD Sample Age Groups and Substance Type Using age groups empirically derived from decision tree analysis, Figure 1 shows that among Commercial members during 2017, alcohol abuse steadily increases with age. Opioid abuse peaks with the 24 to 27 age group and then declines, while abuse of other drugs decreases with age.
Primary SUD Sample Age Groups and Gender Across all substances, Figure 2 shows that more males are included in the SUD sample, and that substance abuse problems peak with the 23 to 31 year old empirically derived age group. Urban and rural member addresses were comparable across these age groups.
Secondary Substance Use Condition (SUC) Sample The second sample of members with a substance abuse condition (SUC sample) obtained using a different, less restrictive selection criteria based only on diagnosis was considerably larger (n = 45,275) or 13% of the 347, 454 Commercial members for 2017. This percentage is reasonably similar to the 8% rate of substance use problems found in a previous study.4
Figure 3 shows the percentage of Commercial members by gender included in the second SUC sample. The age groupings are empirically derived from decision tree analysis. More males are included and the percentage peaks with the 23 to 62 year old age group.
Figure 4 shows the rate of mental health conditions using empirically derived age groupings from decision tree analysis with the second SUC sample for Commercial members. The percentage of substance abuse conditions with a comorbid mental health condition peaks with the 23 to 31 year old age group.
Analytic Results Stepwise logistic regression analysis was used to determine factors associated with a substance abuse problem using the larger second SUC sample among Commercial members and selecting members within a more homogeneous regional market (n = 32,341). Several variables were associated with being included in this sample. Table 2 presents the associations and odds ratios for each of these variables.
As can be seen in Table 2 from the values of the Wald statistics (larger is more strongly associated), a mental health condition is most strongly associated with being included in the second SUC sample, then age groups, then gender, and to a much lesser extent diabetes or urban-rural member address. Based on the odds ratios, the model shows that a member with a mental health condition is 2.3 times as likely to be included in this SUC sample as a member without a mental health condition. Members between the ages of 23 and 31 and 32 and 62 are 4.4 and 4.5 times, respectively, as likely to be included in this sample as members in other age groups. The odds ratios would indicate that males are 1.5 times as likely as females to be included in this larger secondary SUC sample of Commercial members.
Primary SUD Sample Initial Follow-up Rate Turning back again to the primary SUD sample of Commercial members that is based on the more restrictive criteria for the substance abuse measure, a member may or may not return for an initial follow-up visit within the first 14 days. Of the 3,494 Commercial members included in this measure, 1,098 returned for a follow-up visit and 2,396 did not (31%). As noted above, in this sample alcohol made up almost 60% of the overall substance abuse problems, and the follow-up rate was 30%. For opioid abuse the overall follow-up rate was 38%. Interestingly, the opioid follow-up rate for males (42%) was higher than for females (31%).
Provider Groups and Initial Follow-up In this primary SUD sample an empirical grouping of provider specialties for opioid abuse resulted in three groups: primary care, inpatient and program treatment, and outpatient behavioral health. The difference in follow-up percentage for specialty groups is striking, but not surprising. Inpatient or program-based treatment has a relatively higher initial follow-up rate (71.6% for opioid drugs and 59.3% for alcohol and other drugs). The follow-up percentage for primary care is very low at 6.9% and 4.6%, respectively. Although in primary care a 14 day follow-up may not be necessary to address many substance abuse problems, nonetheless programs such as SBIRT (screening, brief intervention, and referral for treatment) may result in more patients receiving follow-up intervention in a timely manner. Follow-up in outpatient behavioral health settings is intermediate at 27.9% and 20.8%, respectively, perhaps reflecting more weekly or biweekly visits.
Analytic Results Since the short term follow-up rate is so low for primary care providers, it would be difficult to use additional analyses to try and determine subgroups with whom to intervene. Primary care providers in general are an important group for intervention from a health insurer perspective. Among hospital and program providers, there is more of a balance between members who follow-up and those who don’t, and so additional analyses can be useful in identifying subgroups for intervention. Using stepwise logistic regression analysis with Commercial members in the primary SUD sample, several variables are associated with not following up. Table 3 presents the associations and odds ratios for each of these variables.
As can be seen in Table 3 from the values of the Wald statistics, age groups are most strongly associated with not following up, then urban-rural member address, and then substance type. Based on the Wald statistics, the model shows that ages 17 to 22 and 32 to 40 are significantly more likely to not follow up (p< 0.00 and p< 0.02, respectively). The odds ratios show that age 17 to 22 is about 2.2 times more likely to not follow up than other age groups. An urban resident is 1.8 times as likely as a rural resident to not follow up. A member of the SUD sample using alcohol is 1.3 times as likely to not follow up compared to the other substance categories.