The construction of annualized PMPM total health care costs is based on all payments made by the plan (or a plan subcontractor in the case of behavioral health carve-outs) for general medical care, behavioral health services, and pharmaceutical claims incurred for enrollees. We calculated PMPM annual costs by summing up all health-related costs to the person level, then aggregating persons within the plan to generate a total cost per plan. Average costs are constructed by dividing the total cost per plan by the total number of member months observed in the data (as not all individuals are enrolled over the entire year), which generates a monthly estimate that can be annualized by multiplying by 12.
Table 5 shows some descriptive statistics on the average health care costs across plans, as indicated by average PMPM costs in total, and broken out for selective health categories (medical, behavioral health, pharmaceutical) for our 432 plans (Panel A) and then for the 290 plans in our final analytic sample (Panel B). Again, in looking at changes in mean and maximum values across Panel A and B it is easy to see to how the removal of artificially created “small” plans impacts PMPM costs. Interestingly, the removal of these “small” plans reduces our average PMPM cost for behavioral health services overall, and in the case of residential treatment and partial hospitalization, the reduction in average PMPM costs is fairly substantial. However, the average total PMPM cost, PMPM medical cost, and non-behavioral health prescription costs are all higher in the analytic sample.
TABLE 5. Descriptive Statistics and Sample Sizes for PMPM Cost Estimates
Focusing on values for the analytic sample (Panel B), the total average cost paid per enrollee across health plans represented in the data was $268.49, of which 4.6% ($12.22) was total PMPM for behavioral health services. The vast majority of the cost for behavioral health was for behavioral health prescriptions ($7.46) and not utilization of intermediate care services. Residential treatment, partial hospitalization and IOV combined represent only 19.6% ($2.40) of the total behavioral health costs to the health plan.11 Behavioral health prescription drugs represent the biggest share of total PMPM spending on behavioral health and are therefore likely to be a bigger driver of costs than intermediate services of any kind.
The last column of Table 5 provides some important insights regarding the number of plans for which we have information on utilization of intermediate behavioral health services. As indicated above, very few plans have claims for residential treatment and only about half of the plans have claims for partial hospitalization. Thus, even if these services were expensive, they represent a very small fraction of the average total plan cost. Residential treatment in particular represents less than one one-thousandths of a percent of total PMPM costs on average. Partial hospitalization represents only 0.2% of total PMPM cost on average. And although IOV are far more common across health plans, this category too represents less than 1% of total PMPM cost.
The fact that relatively few plans in our sample have claims reported for two of three intermediate services should not be surprising given that the utilization of these services is determined by events that are relatively rare in the general population and many plans do not provide coverage for these services. However, it does complicate our ability to model the impact of providing these services, as we are trying to model something that represents a tiny fraction of our dependent variable (total average plan medical costs). Although the MarketScansample included some health plans with generous behavioral health coverage, utilization of two of the three intermediate services even within these generous plans was relatively limited.
The highly skewed nature of the utilization data can be seen in Figure 1, which shows the distribution of the 75th percentile value for IOV (Figure 1a), residential treatment (Figure 1b) and partial hospitalization (Figure 1c). These figures represent the average length of stay or number of visits in a single episode for each plan rather than the number of claims. They show the distribution of these 75thpercentile values across plans (demonstrating on the y-axis the proportion of plans with the same value). Even when we look at the 75th percentile value across health plans we see that for two of the intermediate services, plan-utilization appears to be highly restrained. For IOV, the vast majority of health plans have claims involving episodes of 20 visits or fewer. In the case of residential treatment, the bulk of the health plans have zero episodes. The relatively few plans that do have claims, have 75thpercentile values for length of stay that are still generally quite low (although uniformly spread out between 1 and 40 days). Partial hospitalization is the only service where we see a fairly large spread in the 75th percentile value for episode length, but this seems to be driven basically by outliers, as the bulk of the plans have episode lengths well under 100 days.
FIGURE 1. Examination of the Value for Number of Visits/Days Covered at the 75th Percentile for Each Plan for Specific MH/SUD Services
FIGURE 1a. Intensive Outpatient Visits
FIGURE 1b. Residential Treatment
FIGURE 1c. Partial Hospitalization
11. Omitted from this table is the “other non-prescription MH/SUD spending,” which on average is $2.37 across plans.