Short-Term Analysis to Support Mental Health and Substance Use Disorder Parity Implementation. Enrollee and Health Plan Characteristics

02/08/2012

The main demographic variables we could construct from enrollee information (aggregated up to the region level) were the following: percent male, percent children (i.e., <18 years of age versus 19-65 population), and the Charlson-DeyoIndex, which is a weighted index of 17 chronic illnesses (identified through ICD-9 codes) that are likely to generate inpatient hospitalization within the coming year (based on Deyo et al., 1992).8

The MarketScan database did not include public insurers or Medigap plans. Additional plan characteristics we could construct from the data included the size of the plan (measured by enrollees within the region), and the type of the plan (e.g., HMO, PPO, POS, consumer-directed health plan, etc.). Descriptive statistics for these variables for the full 432 plans and our final analytic sample of 290 plans are provided in Table 1.

The most noticeable consequence of moving from the full sample to the analysis sample is the sizeable decrease in the percent of small plans, from 38.4% to 8.3%. The reductions in small plans show up in other statistics as well. Differences in means and maximum values between the full sample (n=432) and the final analytic sample (n=290) shows that there is an important reduction in variance within our analytic sample in the Charlson-Deyo index. The maximum value falls from 0.667 from the full sample to just 0.166 in the analytic sample (n=290). By construction the maximum score possible for this index is 33, based on the weighting of the 17 diagnoses represented. In our encounter (individual level) data, we do observe some fairly high patient values. However, when these values are averaged over total plan enrollment, the typical value for the plan is much closer to the mean value observed across all individual encounters of care (0.023). One consequence is that the variance in this index across plans is extremely limited, and not likely to capture the plan heterogeneity in chronicity of patients that we had hoped it would. This reduced variance in the index across plans is an indication that we may not have adequately captured important differences in general health care utilization across plans.9

 TABLE 1. Enrollee and Plan Characteristics for the Full Sample (Panel A) and Final Analytic Sample (Panel B)

 

N

Mean

Min

Max

PANEL A -- Full Sample
Enrollee Characteristics
   % Male   432     47.9%   0 100%
   % Children (Age <18) 432 22.3% 0   62.5%  
   Charlson-Deyo Index 432 0.023 0 0.667
Plan Characteristics
   Small Plan (<100 full-year enrollees in region) 432 38.4% 0 100%
   Medium Plan (100 - 4,999 full-year enrollees in region) 432 38.3% 0 100%
   Large Plan (5,000 or more full-year enrollees in region)   432 23.4% 0 100%
   % HMO 432 14.8% 0 100%
   % PPO (capitated and non-capitated) 432 55.1% 0 100%
   % Exclusive Provider Org or Point of Service Plan 432 9.3% 0 100%
   % Consumer-Directed or Comprehensive Plans 432 16.4% 0 100%
PANEL B -- Analysis Sample
Enrollee Characteristics
   % Male 290 48.7%   19.7%   75.0%
   % Children (Age <18) 290 24.9% 0 39.2%
   Charlson-Deyo Index 290 0.024 0 0.166
Plan Characteristics
   Small Plan (<100 full-year enrollees in region) 290 8.3% 0 100%
   Medium Plan (100 - 4,999 full-year enrollees in region) 290 56.9% 0 100%
   Large Plan (5,000 or more full-year enrollees in region) 290 34.8% 0 100%
   % HMO 290 10.7% 0 100%
   % PPO (capitated and non-capitated) 290 61.0% 0 100%
   % Exclusive Provider Org or Point of Service Plan 290 7.6% 0 100%
   % Consumer-Directed or Comprehensive Plans 290 16.9% 0 100%

Plan benefit design can influence utilization of health care services by influencing the relative cost of the services to patients (through co-payments, deductibles and management techniques). Plan characteristics can proxy both the extent to which care is managed in order to control costs as well as the likely risk pool. In addition to the obvious types of measures (type of plan, region of operation, plan size), we were able to consider several benefit measures available through the benefit plan database. However, many of the potentially important benefit measures for parity were missing data or lacked clarity in terms of what benefit applied. Appendix Table A1 in Appendix 2 lists the plan benefits we had hoped to consider and the number of plans in our linked data set (out of 103) that actually contained this information. As the table highlights, very few of the actual benefits are systematically recorded for most plans, making the plan information far less useful than we originally anticipated. Thus, the only measures we were able to consider for analysis were the following: family deductible; medical co-insurance rate for outpatient visit; constructed measure of equality in inpatient co-insurance rates; constructed measure of equality in outpatient co-insurance rates; number of NQTLs; and behavioral health carve-out indicator. We obtained the first two plan measures directly from the benefit database. We constructed the remaining four measures using various reported plan benefit information, as described in Appendix 2.Table 2 provides descriptive statistics of these variables for the full sample and our reduced analytic sample.

TABLE 2. Descriptive Statistics for Plan Benefit Characteristics and Measures

 

N

Mean

Min

Max

PANEL A -- Full Sample
Average Family Deductible   432     $735.17   0   $4,000  
Co-insurance -- outpatient visit (amount paid by plan) 381 87.68%   70%   100%
Proportion of plans with equal co-insurance - inpatient 413 40.8% 0 100%
Proportion of plans with equal co-insurance - outpatient   413 70.0% 0 100%
Number of health plan NQTLs 432 2.75 0 4
Proportion of plans with behavioral health carve-out 432 66.4% 0 100%
PANEL B -- Analysis Sample
Average Family Deductible 290 $843.22 0 $4,000
Co-insurance -- outpatient visit (amount paid by plan) 261 86.6% 70% 100%
Proportion of plans with equal co-insurance - inpatient 279 82.8% 0 100%
Proportion of plans with equal co-insurance - outpatient 279 76.0% 0 100%
Number of health plan NQTLs 290 2.78 0 4
Proportion of plans with behavioral health carve-out 290 85.2% 0 100%
 

  1. Deyo R, Cherkin D, Ciol M (1992). “Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases.” Journal of Clinical Epidemiology, 45: 613-619.

  2. We also considered capturing variance in general health care utilization through indicators representing the fraction of plan enrollees who were either: (a) current or past smokers, or (b) obese. However, given the high correlation of these behaviors with behavioral health care utilization, measurement of these values in the same year as behavioral health care utilization would result in significant colinearity.

View full report

Preview
Download

"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®