Short-Term Analysis to Support Mental Health and Substance Use Disorder Parity Implementation. Behavioral Health Service Setting Variables


There are a number of different indicators that can be used to identify a behavioral health claim occurring in one of the three intermediate care settings of interest, and no one indicator is consistently used by all the plans. We therefore applied rules across a multitude of indicators when we tried to identify residential treatment episodes (and length of stay), partial hospitalization and intensive outpatient visits (IOV) across plans.

Residential Treatment. Identification of individuals receiving treatment in a residential treatment setting involved a multi-step process. First, in the inpatient data we identified anyone receiving care in: (a) a residential substance abuse facility (STDPLAC = 55); (b) a psychiatric residential treatment center (STDPLAC = 56); or (c) general residential treatment center (STDPROV = 35). We removed from these claims those that also indicated that the service setting was an inpatient hospital setting (STDPLAC = 21 or 51 -- meaning general inpatient hospital or psychiatric inpatient hospital) unless the revenue code and procedure codes indicated that the care was non-hospital residential treatment.10Second, in the outpatient claims data we identified cases where additional outpatient type services were attached to an inpatient hospitalization, but these had not been flagged and aggregated with the inpatient claims by Thomson Reuters. These were identified in one of two ways: (a) the outpatient claim included an H-code indicating hospital or residential based treatment (H0017, H0018, or H0019); and (b) CPT codes indicated hospital based interactive psychotherapy (CPT codes in the range 90823-90829).

Applying these rules we identified approximately 2,050 residential treatment episodes in the data. While this represents a non-trivial number of residential treatment episodes, analytically what matters for our assessment of the effect on health plan costs is the distribution of these episodes across plans. Table 3 shows that fewer than 15% of the health plans in the full sample (n=52 out of 432) had any episodes involving residential treatment, and the mean number of episodes was very small (n=1). And while the proportion of plans with residential treatment claims is higher in our analytic sample (nearly 18%), this is due to our disproportionately dropping plans with zero claims. The mean number of residential treatment claims rises to just two in the analytic sample.

TABLE 3. Proportion of Plans Experiencing a Residential Treatment, Partial Hospitalization Visit or Intensive Outpatient Claim  

  N Mean Mean
  # of Claims  
PANEL A -- Full Sample
Proportion of Plans with Residential Treatment Claim     432     12.5%   1
Proportion of Plans with Partial Hospitalization Claim 432 39.1% 14
Proportion of Plans with Intensive Outpatient Claim 432 77.8% 2,911
PANEL B -- Analysis Sample
Proportion of Plans with Residential Treatment Claim 290 17.9% 2
Proportion of Plans with Partial Hospitalization Claim 290 56.9% 21
Proportion of Plans with Intensive Outpatient Claim 290 98.3% 4,333

Table 4 shows the distribution of claims across plans more explicitly. Thirteen of the 52 plans in our final analytic sample had just one residential treatment claim in 2008, and another 18 plans had five or fewer claims. Only 4.5% of all plans (n=13) had more than ten claims processed for residential treatment.

TABLE 4. Distribution of Plans by Number of Visits for Intermediate Services (n=290)

  Intermediate Service Claims   Number of Plans with Claims
0 1   2 - 5     6 - 10     11 - 20     21 - 50     51 - 75   75+
Residential Treatment   238     13   18 8 6 6 0 1
Partial Hospitalization 125 27 28 23 24 50 11 18
Intensive Outpatient Therapy   5 1 8 2 4 23 11   236  

As is clear from these tables, a claim for residential treatment is a rare event in MarketScan’s 2008 data. The relatively low number of claims coupled with the bunching of positive values at very low levels of visits across health plans will make identification of the effects of covering these services highly imprecise and possibly biased.

Partial Hospitalization Visits. There were relatively few cases of partial hospitalization in the inpatient claims data, but a few did exist and were easily identified through either a CPT code (90816-90822) or hospital revenue code (REVCODE = 912). Most of the claims involving partial hospitalization were in the outpatient data. These claims were identified again through procedure codes (CPT codes of 90816-90822) and H-codes (H0035).

Combined, we identified over 3,700 claims related to partial hospitalization in the inpatient and outpatient data. This is nearly twice the number of claims identified for residential treatment, and far more plans experienced at least one claim for partial hospitalization (as indicated in Table 3 and Table 4). Several plans experienced multiple claims for partial hospitalization and with longer episode length, increasing the variability in number of visits across plans.

Intensive Outpatient Visits (IOV). Identification of IOV was based solely on information provided in the outpatient claims data. Identification of these cases was based on procedure codes (CPT-codes in the range of 90804-90815 or an H-code of H0015). Nearly 178,000 intensive outpatient claims were identified in the MarketScandata for 2008, with over 85% of health plans in our analytic database experiencing at least one claim. As shown in Table 3 (by the mean number of claims) and Table 4 (in terms of the distribution of number of visits), there are a large number of health plans that experienced multiple claims for intensive outpatient treatment. This is a far more common service being utilized across the health plans represented in MarketScan’s2008 data.

10. As revenue and procedure codes are used for reimbursement purposes, we have more confidence in these measures for indication of the type of care received then in the variable identifying the setting. This only affected six claims so even if they are improperly identified, it would not affect our results.

View full report


"mhsud.pdf" (pdf, 669Kb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®