One of the key concerns in the implementation of insurance parity laws, which require equivalent coverage of behavioral and general medical care, has been the impact on treatment use, spending, and overall costs of insurance premiums. Several studies have indicated that parity laws have minimal financial impact (1–5) and modest effects on use of and perceived access to care (6–10). An additional aspect of insurance parity, the implications for the quality of mental health treatment, is less well studied (11–13). Evidence of increased treatment quality, combined with limited financial impact, would arguably strengthen the overall rationale for parity laws.
A limited number of studies have examined the impact of parity laws on treatment quality, and the results are mixed. Busch and others (12) compared the quality of care for depression among enrollees in the Federal Employees Health Benefits Program before and after passage of federal parity laws and found only modest improvement. Trivedi and others (13), however, found strong increases in the rate of follow-up care after psychiatric inpatient treatment among Medicare plans that implemented mental health parity. Notably, the focal measure of parity in Trivedi and others’ study was equivalence of cost-sharing between general medical and behavioral treatment. This raises the question—was the finding of increased psychiatric follow-up primarily a result of reduced cost-sharing, or did other aspects of parity, such as the elimination of treatment limits, also contribute to the increase in follow-up care?
Oregon’s 2007 parity law provides a unique opportunity to assess whether follow-up rates for psychiatric care change in the absence of changes in cost-sharing between psychiatric and general medical care. The law was one of the most comprehensive state parity laws at the time of its passage. It defined parity for addiction and mental health conditions in terms of both quantitative coverage limits (for example, parity in application of any numerical limits on treatment use) and nonquantitative coverage limits (for example, parity in application of utilization review or other management techniques that may limit treatment). The Oregon law’s comprehensive definition of parity has strong parallels with the 2008 federal Mental Health Parity and Addiction Equity Act (MHPAEA) (1).
McConnell and others (1) found that after the policy was implemented, there were small but statistically nonsignificant increases in expenditures for mental health and addiction services among continuously enrolled members of four representative Oregon-based health insurers. The parity law’s implementation resulted in changes in both quantitative and nonquantitative coverage limits in the four plans for mental health and substance abuse–related services. However, parity in cost-sharing—for example, copayments—between general medical and behavioral health coverage was maintained throughout the study period, with only one plan indicating a change (increase) in cost-sharing for general medical and behavioral treatment.
This study estimated the effect of Oregon’s parity law on rates of outpatient behavioral health care that was received within 30 days following a discharge from a psychiatric hospitalization. To assess the effect, we conducted two analyses. The first analysis compared rates of follow-up care associated with discharges of individuals who were enrolled in Oregon insurance plans that were or were not affected by the 2007 parity law. The second analysis was limited to discharges of individuals who were enrolled in plans affected by the parity law. It compared rates of follow-up care for discharges of individuals who had or had not met quantitative limits on coverage in a single coverage year. The results of these analyses provided evidence of the impact of Oregon’s parity law on rates of follow-up care after a psychiatric inpatient stay, independent of the general effects of cost-sharing, such as changes in copayments. Given that the Oregon parity law parallels many features of the federal MHPAEA, the study results provided some evidence of the potential effects—beyond cost-sharing effects—of national parity law on follow-up rates.
The study employed a difference-in-difference design, consistent with the natural experimental setting, to assess changes in rates of follow-up care received within 30 days of a psychiatric inpatient stay that were due to implementation of Oregon’s parity law. In the primary analysis, we compared rates of follow-up care provided before and after the parity law for discharges of individuals who were continuously enrolled in Oregon insurance plans subject to the parity law (intervention group) and of a comparison group consisting of individuals continuously enrolled in self-insured plans (and thus exempt from state parity laws) in Oregon, Washington, and Northern California (control group). The secondary analysis was restricted to discharges of individuals in the intervention group. In this analysis, we compared rates of follow-up care before and after the parity law for discharges of individuals who had or had not met quantitative mental health coverage limits.
We used data spanning two years before and two years after implementation of the Oregon parity law, on January 1, 2007. The data were collected for McConnell and others’ (1) general study of the law. The data included insurance claims from individuals who were continuously enrolled in plans from four representative Oregon insurers subject to the parity law and claims from Thompson Reuters’ MarketScan database for individuals who were continuously enrolled in commercial, self-insured plans from Washington, Northern California, and Oregon that were exempt from state parity laws. Study years for individuals in the intervention group were based on renewal dates, given that parity was enforced on January 1, 2007, or at the first point of annual renewal for that year. Study years for individuals in the control group were based on calendar year (2005–2008). Discharge observations were assigned a study year based on the date of discharge.
We compared mean, median, and modal out-of-pocket expenses for general medical and behavioral outpatient care (indicated, respectively, by two common CPT codes, 99213 and 90806) and found consistent copay rates for visits for general medical and behavioral care in each study year by individuals in the intervention and the control group. The study was approved by the institutional review boards at Oregon Health and Science University and Portland State University.
Study participant selection
The primary unit of analysis was an inpatient stay with a primary mental health diagnosis (International Classification of Diseases codes 295.xx−302.xx and 306.xx−314.xx) for individuals ages six to 64 that did not involve a transfer to another inpatient facility, an inpatient readmission within 30 days, or a discharge in the last month of the study data. This yielded a total sample of 888 psychiatric inpatient discharges, representing 737 individuals, over the four-year study period. Within the 353 discharge observations from plans affected by the parity law, we identified 88 discharges, representing 77 individuals, where either the preparity mental health inpatient or outpatient coverage limits had been met by the discharge date within a single coverage year.
Before the parity law, mental health coverage limits were applied over two enrollment years within each of the four Oregon plans in the study. Inpatient treatment limits ranged from 14 to 19 days, and outpatient treatment limits ranged from 29 to 45 visits, varying by plan and adult or child status (1). The study data did not include indication of a “starting year” for imposing the biennial treatment limits. Therefore, we identified only discharges of individuals who met limits within a single study year. This was accomplished by counting the mental health inpatient days and outpatient visits within an individual’s study year through the date of discharge and comparing them with the preparity limits for each plan and age group. Mental health outpatient visits were defined according to the study criteria below.
The main outcome variable was a binary indicator of whether an individual received an outpatient behavioral health visit within 30 days of discharge from a mental health inpatient stay. An outpatient behavioral visit was defined as any nonemergency outpatient care with a primary behavioral health diagnosis (295.xx−314.xx) and a CPT or HCPCS code consistent with an office, home, or noninpatient residential visit (Evaluation and Management CPT codes 992xx−994xx, exclusive of codes for emergency services [99281–99288], and HCPCS codes beginning with “H”).
To implement the difference-in-difference empirical design, we included binary variables for discharge before or after the parity law, discharge of individuals in the intervention group or the control group, and the interaction of discharge after the parity law and intervention-group indicators. When estimated using linear methods—for example, ordinary least squares—the coefficient of the indicator variable representing the interaction of discharge after the parity law and intervention-group indicators provides the difference-in-difference estimate, or the (marginal) effect of the parity law independent of group membership (net policy effect). When using nonlinear methods, for example, logistic regression, the coefficient of the indicator variable representing the interaction of discharge after the parity law and intervention-group indicators alone does not provide a consistent estimate of the difference-in-difference, but an estimate can be derived from (nonlinear) transformations from the set of coefficients estimated within the difference-in-difference framework (14).
Additional binary control variables representing clinical and demographic characteristics were also included in the empirical model. These included discharge diagnosis category (295.xx−299.xx or psychotic or nonpsychotic disorders), gender, insurance policyholder status (employee, spouse, or dependent), and calendar year quarter of discharge. Insurance policy holder status subsumes (available) direct measures of age (or distinction between adult and child), given that all but two individuals in the dependent group were age 26 or younger, and a vast majority were under 18 years of age. Calendar quarter indicators were included to account for general seasonal variation (primarily during the holiday season) in follow-up rates. The reference group was discharges of male employees in the first calendar quarter of a study year with a discharge diagnosis of a nonpsychotic disorder.
Chi square tests were used in the two study analyses to assess any differences in characteristics associated with discharges of individuals in the intervention and control groups. Logistic regression was used for initial estimation. The margins command in Stata was then used to obtain consistent estimates of the marginal change in the probability of follow-up that was due to parity (the difference-in-difference) and to provide estimates, in percentages, of the effects related to each study measure on follow-up rates (14,15). The estimated average marginal effects are reported in percentages reflecting the average of predicted change in the follow-up rate derived from the logistic regression results for each discharge, when each binary study variable is changed from 0 to 1. Standard errors were adjusted for general heteroskedasticity and clustering of discharges across individuals (16,17).
Table 1 presents characteristics associated with discharges of individuals in the intervention and control groups. Some slight differences were evident. None of the demographic differences between the groups, however, were found to be statistically significant.
Table 1Characteristics of discharges among the intervention and control groups, 2005–2008a
| Add to My POL
|Intervention (N=353)||Control (N=535)|
|One inpatient stay|
| Psychotic disorder||237||67||386||72|
| Nonpsychotic disorder||116||33||149||28|
|Calendar quarter of discharge|
| 1 (Jan.–March)||113||32||137||26|
| 2 (April–June)||92||26||151||28|
| 3 (July–Sept.)||73||21||126||24|
| 4 (Oct.–Dec.)||75||21||121||22|
Table 2 presents the average marginal effects of the study variables on rates of follow-up care before and after the parity law for discharges of individuals in the intervention and control groups. The postparity effect identifies the estimate of the difference-in-difference, or the net policy effect of parity on follow-up rates, in percentages. The estimate (.114, p=.042) indicated that the 2007 Oregon parity law was associated with an increase of 11% in the rate of follow-up within 30 days of discharge from a psychiatric inpatient stay.
Table 2Estimated average marginal effects of study variables on rates of follow-up care after the parity law
| Add to My POL
|Postparity effect (difference-in-difference)||.114||.056||.042|
|Postparity period (control)||–.033||.034|
|Preparity period (intervention)||–.057||.040|
|Psychotic disorder discharge diagnosis||–.076||.031||.015|
|Spouse of insurance policyholder||–.064||–.043|
|Dependent of insurance policyholder||–.147||.035||<.001|
|Calendar quarter of discharge|
Table 3 presents characteristics of discharges of individuals in the intervention group. Before the parity law, the discharges of individuals who had met preparity quantitative coverage limits in a single enrollment year were associated with a lower rate of follow-up care. Discharges of individuals who had met preparity quantitative coverage limits were also more likely to involve individuals with multiple inpatient discharges throughout the study period, dependent status, a psychotic disorder, or a discharge date in the second calendar quarter of a study year. Differences between the individuals who had or had not met preparity coverage limits were statistically significant (p<.05), except in the case of gender.
Table 3Characteristics of discharges of individuals in the intervention group who had or had not met preparity quantitative coverage limitsa
| Add to My POL
|Limits met (N=87)||Limits not met (N=266)|
|One inpatient stay||<.001|
| Psychotic disorder||54||62||65||24|
| Nonpsychotic disorder||33||38||201||76|
|Calendar quarter of discharge||.049|
| 1 (Jan.–March)||19||22||94||35|
| 2 (April–June)||31||35||61||23|
| 3 (July–Sept.)||19||22||54||20|
| 4 (Oct.–Dec.)||18||21||57||22|
Table 4 presents the average marginal effects of the study variables on rates of follow-up care for discharges of individuals in the intervention group who had or had not met preparity quantitative coverage limits in a single coverage year. The postparity effect identified the estimate of the difference-in-difference, or the net policy effect of parity on the follow-up rates for discharges of individuals who had met parity limits within a single enrollment year versus those who had not. The postparity effect (.203, p=.028) indicated that the 2007 Oregon parity law was associated with an increase of 20% in the rate of follow-up within 30 days of a psychiatric inpatient stay for discharges of individuals who had met the preparity coverage limits.
Table 4Estimated average marginal effects of study variables on rates of follow-up care for discharges of individuals who had met preparity quantitative coverage limitsa
| Add to My POL
|Postparity effect (limits met)||.203||.093||.028|
|Postparity period (limits not met)||.021||.050|
|Preparity period (limits met)||–.153||.070||.028|
|Psychotic disorder discharge diagnosis||.067||.051|
|Spouse of insurance policyholder||–.042||.064|
|Dependent of insurance policyholder||–.138||.054||.011|
|Calendar quarter of discharge|
The preparity-period estimate (–.153, p=.028) indicated that follow-up rates for discharges of individuals who had met preparity coverage limits were 15% lower than for discharges of individuals who had not met coverage limits. Given that after the parity law, follow-up rates for discharges of individuals who had not met preparity coverage limits increased only slightly (postparity-period estimate=.021, p=.678), these results indicate that a central effect of the parity law was to raise follow-up rates for discharges of individuals who had met coverage limits up to, and potentially beyond, those of individuals who had not met coverage limits.
The results of the primary study analysis indicated that rates of follow-up within 30 days of a psychiatric inpatient stay increased in Oregon after the implementation of the 2007 Oregon parity law. In contrast to the results of a study by Trivedi and others (13), this increase occurred in an environment in which general cost-sharing for behavioral and general medical care was consistently equivalent. This suggests that the effect of parity laws on follow-up rates extends beyond a potential influence through copayments (and subsequent effects on utilization), working through removal of other quantitative or nonquantitative coverage limits as well.
The secondary study analysis indicated that among discharges of individuals enrolled in plans affected by parity, much of the change in follow-up rates that was due to parity came from improvements in follow-up rates for discharges of individuals who had met preparity coverage limits. One interpretation of this finding is that the quantitative coverage limits imposed for mental health treatment before parity were a direct impediment to successfully receiving postdischarge follow-up treatment. This analysis, however, does not provide clarity in regard to the mechanism or mechanisms by which parity led to improved follow-up.
As noted above, a vast majority of discharges of individuals who had met coverage limits did so only through the inpatient coverage limit. Thus if by hitting these limits, individuals were kept from completing follow-up within 30 days of discharge, one would have to surmise that some indirect mechanism was at play, given that a lack of existing coverage for outpatient care generally was not evident. One possibility is that these individuals did not distinguish between inpatient and outpatient coverage limits. That is, upon meeting their inpatient limits, they presumed that the limits extended to outpatient care and thus were less likely to follow through with postdischarge care.
A plausible alternative conjecture is that the improved follow-up rates reflected a supply side effect (provider), as opposed to a demand side effect (consumer). Notably, the group identified as having met preparity treatment limits consisted of individuals who were much more likely to have had multiple inpatient stays over the course of the study and who, by definition, had generally longer-than-average stays. These discharges were also much more likely to involve individuals with discharge diagnoses indicative of more severe and persistent mental illness and to involve children or young adults. Overall, the characteristics of the group that had met preparity treatment limits were indicative of higher-than-average expected treatment expense. Given that a key purpose of successful follow-up is to reduce inpatient recidivism, the increased rate of follow-up after parity may reflect efforts by insurers to reduce perceived expenditure risks given removal of quantitative treatment coverage limits (18). Notably, this would suggest that coverage limits may act as a substitute for providing treatment quality that could reduce treatment costs.
There were several potential limitations to this study. A primary concern is that the study results reflect some unique aspects of Oregon’s system of mental health care, the particular insurers sampled for the study, or some other unobserved aspects of the Oregon law or its implementation that cannot be generalized to other settings. The original study data included only individuals who were continuously enrolled throughout the four-year study period, which may have resulted in selection bias. Similar studies that compared results for samples of continuously enrolled and all enrolled individuals noted only modest differences (13).
The study’s measure of follow-up after a psychiatric inpatient stay is similar to, but less restrictive than, the Healthcare Effectiveness Data and Information Set (HEDIS) measure of follow-up after hospitalization for mental illness, which is commonly used to report psychiatric inpatient follow-up rates. Use of the exact HEDIS measure criteria or others may have yielded different results. The choice of follow-up measurement reflected concerns for preserving sample size and representativeness, as well as limitations to the study data necessary to accomplish exact HEDIS specifications. The measurement of discharges of individuals who had met preparity treatment limits did not capture discharges of individuals who had met limits over two coverage years. Nor can we verify that our accounting of accrued treatment reflected that of the actual insurers. Assessment of measurement bias that is due to limitations in identifying whether treatment limits are met would require a better understanding of how parity influences follow-up rates.
The study results did not directly identify the mechanism by which parity influences follow-up rates. Quantitative and nonquantitative coverage limits may directly inhibit psychiatric inpatient follow-up. Alternatively, the removal of these restrictions may create greater incentives to provide care in a more cost-effective manner. Further research on the relationship of parity to the provision of follow-up care and other measures of treatment process quality is clearly warranted.
The study results indicate that the 2007 Oregon parity law resulted in an increase in the rate of follow-up after psychiatric inpatient care in the two years after the law’s implementation. These results extend the findings of the positive effect of parity on follow-up rates beyond equity in cost-sharing to removal of quantitative (and nonquantitative) coverage limits in general. Given the similarities between the federal MHPAEA and the 2007 Oregon parity law, the increased follow-up rates found in Oregon may be a harbinger of a national effect when and if the MHPAEA is fully implemented.
The study was supported by grant 1R01DA020832-01A1 from the National Institute on Drug Abuse. Dr. McConnell is the principal investigator.
The authors report no competing interests.