Trends: What Drove Private Health Insurance Spending on Mental Health and Substance Abuse Care, 1992-1999? The decline in spending is a change from trends in the 1980s, when rising costs alarmed policymakers and employers alike. Tami L. Mark, Ph.D., M.B.A. Rosanna M. Coffey, Ph.D. Tami Mark (tami.mark@thomson.com ) is an Associate Director, and Rosanna Coffey is a Vice President of th e Research and Pharmaceutical Division, of Medstat in Washington, DC The substance abuse and Mental Health Se rvices Administration (SAMHSA) has funded research that examines mental health a nd substance abuse (MH/ SA) spending nationally from 1986 to 1996 and from 1987 to 1997. 1 This research looks at aggregate spending by type of service and payer but does not explain the factors behind slower rates of spending for MH/SA services relative to those for all health care. Does this spending trend result from fewer people being treated, fewer servi ces per person, or lower costs per unit of service? How do these underlying spendi ng components contribute to spending on inpatient care, outpatient servic es, and prescription drugs? The SAMHSA studies have not analyzed thes e types of questions because the data did not consistently provide th e detail necessary to decompose spending into these underlying factors. A number of reports and papers funded by SAMHSA and others have analyzed various aspects of MH/SA servic es use and spending; space precludes a full description of these studies here. 2 This paper adds to the emerging literature by focusing on trends in employer-based private insuran ce spending, which might differ in important ways from total mental health care sp ending. We examine the underlying factors influencing the trends in MH/SA spending in the private sector, by decomposing changes in covered private health insurance spending during 19922261999 into changes in the probability of use, intensity of use, and cost per service used. Study Methods Data for this study came from the Medstat MarketScan 256 database, which compiles claims information from the private health insura nce plans of large employers. The database covers employees, their dependents, and early retirees of participating companies. Medstat collects and standard izes claims from more th an 100 different insurance companies, including Blue Cross and Blue Shie ld plans and third-part y administrators. In 1992 about seventy-five employers contributed to MarketScan 256 . In 1999 about forty employers did so. There were 5.0 million covered lives in 1992 and 3.5 million in 1999. Trends: What Drove Private Health Insurance Spending on Mental Health and Substance Abuse Care, 1992-1999? 1 All of the employers were self-insured or had Employee Retirement Income Security Act (ERISA) plans. Some employers in MarketScan 256 change from year to year. To account for changing population characteristics attributable to chan ging employers, we present results only for the sample of twenty-two employers that were in MarketScan 256 in 1992 and 1999, representing 1,802,382 people in 1992 and 1,664,676 people in 1999. Information about the firms is confidential; however, the distri bution of covered lives by industry in the twenty-two employers was as follows: o il and gas extraction mining (5 percent); manufacturing226durable goods (10 percen t); manufacturing226nondurable goods (33 percent);manufacturing226durable and nondurab le goods (5 percent); transportation226 communication226utilities (19 perc ent); services (19 percent); retail (5 percent); and combination of manufacturing, transporta tion226 communication226utilit ies, services, and finance226insurance226real estate (5 percent). Th e distribution of employees by plan type for these twenty-two employers in 1992 was 90 percent fee-for-servi ce (FFS), 8 percent preferred provider organization (PPO), 2 pe rcent point- of-service (POS) plan, and 1 percent exclusive provider organization. The distribution in 1999 was 25 percent FFS, 41 percent PPO, 31 percent POS, 3 percent hea lth maintenance organization (HMO), and 1 percent exclusive prov ider organization. Claims data include both the covered charge and the portion of the charge paid by the enrollee. The focus of this analysis was onl y on the covered charges, since out-of-pocket trends are only imperfectly captured by cl aims. Encounter records from HMOs in MarketScan 256 do not contain charges for inpatient and outpatient services. Charges for these plans were imputed using a regressi on model. The dependent variables were inpatient and outpatient charges for MH /SA treatment and other treatment. The independent variables for estimating inpatien t charges for non- MH/SA treatment were major diagnostic categories of the diagnosis -related groups (DRGs), region, whether or not the patient was in a metropolitan statistical area (MSA), and age of claimant. The independent variables for estimating inpa tient charges for MH/SA treatment were MH/SA diagnosis, region, MSA, and age. The independent variables for estimating outpatient charges for non-MH/SA treatment were Berenson-Eggers Type of Service (BETOS) Codes (a system for classifying out patient claims into service types),major diagnostic categories, region, MSA, and age. The independent vari ables for estimating outpatient charges for MH/SA treatment we re BETOS group, MH/SA diagnosis, region, MSA, and age. A negative exponential model was used to estimate the parameters, because the spending data were log-normally distributed. Claims for MH/SA services were identified based on a primary diagnosis or use of a specialty MH/SA provider. 3 The diagnosis codes selected match those used in the SAMHSA spending estimates study that tracks MH/SA spending. They include diagnoses such as depression and schizophr enia but exclude dementia and mental retardation. Medication claims also were iden tified as MH/SA if th eir therapeutic class indicated a psychotropic medication. The th erapeutic classes were assigned by the Red Book classification system and include the following: opiate antagoni sts, anticonvulsants, antidepressants, tranquilizers/ antipsychotics, stimulants, barbiturates, benzodiazepines, anxiolytics/sedatives/ hypnotic s, antimanic agents, and miscellaneous central nervous Trends: What Drove Private Health Insurance Spending on Mental Health and Substance Abuse Care, 1992-1999? 2 system (CNS) agents. The medications selected included those prescribed by internists as well as those that may have had a primary indication for psychiatric conditions but were prescribed for other nonpsychiatric c onditions. A total of 105,527 people in 1992 and 113,188 in 1999 were identified wi th an MH/SA claim. Study Findings Exhibit 1 shows the overall seven-year per enro llee growth rate of total health care and MH/SA treatment spending. Total health care spending per enroll ee grew by 16 percent from 1992 to 1999, or 2.1 percent annually. In contrast, MH/SA spending fell by 18 percent or 2.8 percent an nually. Whereas MH/SA accounted for 7.2 percent of total health care spending in 1992, it accounted for only 5.1 percent in 1999. Exhibit 1 also shows the distribution of MH/SA spending by inpatient, outpatient, and pharmaceutical expenses. The proportion of spending going to inpatient care declined dramatically between 1992 and 1999 from 48 percent to 18 percent, falling 15.6 percent annually. In contrast, psychotropic drugs made up 22 pe rcent of total MH/SA expenditures in 1992 and 48 percent in 1999, growing 8.9 percent annually. Exhibit 1 also decomposes the spending rate by age group: up to age 17, ages 1822654, and age 55 and older. People age seventeen and younger had the highest proportion of MH/SA spending in 1992 and the greatest drop in the spending proportion. In absolute terms, children and adolescents saw a 71 per cent decline in MH/SA expenditures, adults saw a 21 percent decline, and older adults a 20 percent increase. The distribution of expenditures by service also differed by age group (Exhibit 1). In 1999 spending for psychotropic medications accounted for 24 percent of spending for people age 17 and younger, 49 percent for people ages 1822654, and 68 percent for people age 55 and older. The dramatic shift aw ay from inpatient spending and toward pharmaceutical spending occurred for all three age groups. EXHIBIT 1. Mental Health And Su bstance Abuse (MH/SA) Spending For People With Employment-Based Private Insurance, By Type Of Service, 1992 And 1999 All Ages Birth through Age 17 Ages 18-54 Age 55 and older Type of Service 1992 1999 1992 1999 1992 1999 1992 1999 All health spending per covered life $1,592 $1,843 $734 $86 8 $1,575 $1,770 $3,443 $3,918 MH/SA spending per covered life $115 $95 $98 $57 $129 $107 $88 $110 MH/SA as percent of all health 7.2% 5.1% 13.4% 6.6% 8.2% 6.1% 2.5% 2.8% Distribution of MH/SA spending by type of treatment Inpatient 48% 18% 37% 27% 42% 17% 38% 12% Outpatient 31 34 26 49 34 35 20 20 Psychotropic Drugs 22 48 7 24 24 49 42 68 Trends: What Drove Private Health Insurance Spending on Mental Health and Substance Abuse Care, 1992-1999? 3 SOURCE: Authors222 calculations from MarketScan 256 data. Exhibit 2 shows the decomposition of trends in MH/SA spending by the probability of use, the intensity of use, and the charge pe r service. These compone nts are shown by type of service (outpatient, drugs, and inpatient). The probability of outpatient use rose 17.5 percent; the number of outpatient visits per user, onl y 4.5 percent. Mean spending per visit declined by 3.6 percent. The probability of receiving a prescription for a psychotropic medication declined slightly between 1992 and 1999. The mean numb er of prescriptions per user over the period, however, rose 34.8 percent. The mean number of prescriptions reflects both new prescriptions and refills and does not control for the length of the prescription regimen. Mean spending per prescription also increase d by 49 percent, and the mean spending per user of an MH/SA prescr iption increased by 101 percent. Among the medications, antidepressants were the most widely used (35 percent of claims in 1992 and 57 percent in 1999) (data not shown). Benzodiazepines and anxiolytics were the next-highest- prevalence group of medications (28 percen t in 1992 and 19 percent in 1999). Inpatient use showed a dramatic decline in both the probability of use and the length-of stay. The probability of use dropped 39.6 per cent. The average number of admissions per user was unchanged, which indicates that readmissions did not increase. Average lengthof- stay dropped dramatically as well: 55.1 percent. Mean spending per day increased by 11.8 percent. EXHIBIT 2. Probability of Mental Health/Substance Abuse (MH/SA) Service Use, Intensity, and Spending Per Unit of Service, for People of All Ages with Employment-Based Private Insurance, 1992 and 1999 Type of Service 1992 1999 Absolute Change Percent Change Outpatient Percent with any use 5.8% 6.8% 1.0 17.5% Mean number of visits per user 5.9 6.2 0.3 4.5 Mean expenditure per visit $81 $72 -$3 -3.6 Psychotropic drugs Percent with any use 16.0% 14.5% -1.5 -9.4 Mean number of prescriptions per user 4.5 6.0 1.5 34.5 Mean expenditure per prescription $35 $52 $17 49.0 Mean expenditure per user $156 $314 $158 101.0 Inpatient Percent with any admission 0.4% 0.3% -0.1 -39.6 Mean number of admissions per user 1.3 1.3 0.0 0.0 Mean length of stay (day s) 16.9 7.6 -9.3 -55.1 Mean expenditure per day $592 $662 $70 11.8 SOURCE: Authors222 calculations from MarketScan 256 data. NOTE: All differences between 1992 and 1999 are statistically signif icant at p < .05. Trends: What Drove Private Health Insurance Spending on Mental Health and Substance Abuse Care, 1992-1999? 4 Exhibits 32265 decompose these trends by ag e group. Children (unde r age eighteen) had a lower probability of use of MH/SA services than the population as a whole. In 1999, 6 percent had an outpatient MH /SA visit, 6.5 percent used an MH/SA medication, and 0.2 percent had an inpatient MH/SA stay. Intens ity of outpatient and psychotropic drug use was also higher in 1999 than in 1992, alt hough the number of admissions per user was the same. Trends in the probability of use, intensity of use, and cost per service for children mirror trends in the population as a whole. The pe rcentage with any prescription remained constant, although the number of prescriptions per user increased by 37 percent and drug spending per user increa sed by 109 percent. People ages 1822654 had a slightly higher probability of outpatient care or psychotropic medication use than did the overall population; their inpatient usage and trends were comparable to those of the overall population. People age fifty-five and older had a lower probability of outpatient and inpatient care than the population as a whole, but they had a much higher probability of using a psychotro pic medication. In 1992, 23.6 percent in this age group used a psychotropic medication; in 1999, 23 percent did so. People in this age group, on average, spent $150 on psychotro pic medications in 1992 and $307 in 1999. EXHIBIT 3. Probability of Mental Health/Substance Abuse (MH/SA) Service Use, Intensity, and Spending Per Unit of Service, for People Through Age Seventeen with Employment-Based Private Insurance, 1992 and 1999 Type of Service 1992 1999 Absolute Change Percent Change Outpatient Percent with any use a 4.1% 6.0% 1.9 46.3% Mean number of visits per user a 5.0 5.8 0.8 14.6 Mean expenditure per visit a $98 $81 $-17 -17.6 Psychotropic drugs Percent with any use 6.5% 6.5% 0.0 0.0 Mean number of prescriptions per user a 4.0 5.4 1.4 36.9 Mean expenditure per prescription a $25.9 $40.0 $14.1 54.6 Mean expenditure per user a $103 $215 $112 109.4 Inpatient Percent with any admission a 0.3% 0.2% -0.1 -34.9 Mean number of admissions per user a 1.3 1.3 0.0 0.0 Mean length of stay (days) a 24.0 9.2 -14.8 -61.8 Mean expenditure per day a $597 $571 -$26 -4.4 SOURCE: Authors222 calculations from MarketScan 256 data. a Differences between 1992 and 1999 are st atistically significant at p < .05. Trends: What Drove Private Health Insurance Spending on Mental Health and Substance Abuse Care, 1992-1999? 5 EXHIBIT 4. Probability of Mental Health/Substance Abuse (MH/SA) Service Use, Intensity, and Spending Per Unit of Service, for People Ages 1822654 with Employment-Based Private Insurance, 1992 and 1999 Type of Service 1992 1999 Absolute Change Percent Change Outpatient Percent with any use a 6.8% 7.4% 0.6 8.8% Mean number of visits per user a 6.3 6.4 0.1 1.1 Mean expenditure per visit a $79 $78 -$1 -0.2 Psychotropic drugs Percent with any use a 18.3% 15.9% -2.4 -13.2 Mean number of prescriptions per user a 4.3 5.9 1.6 37.0 Mean expenditure per prescription a $38.5 $55.4 $17.0 44.1 Mean expenditure per user a $166 $326 $160 96.7 Inpatient Percent with any admission a 0.5% 0.3% -0.2 -41.6 Mean number of admissions per user 1.3 1.3 0.0 0.0 Mean length of stay (days) a 14.5 7.0 -7.5 -51.4 Mean expenditure per day a $597 $686 $89 14.9 SOURCE: Authors222 calculations from MarketScan 256 data. a Differences between 1992 and 1999 are st atistically significant at p < .05. EXHIBIT 5. Probability of Mental Health/Substance Abuse (MH/SA) Service Use, Intensity, and Spending Per Unit of Service, for People Age Fifty-Five and Older with Employment-Based Private Insurance, 1992 and 1999 Type of Service 1992 1999 Absolute Change Percent Change Outpatient Percent with any use 4.8% 5.4% 0.6 12.3% Mean number of visits per user a 5.0 5.6% 0.6 12.0 Mean expenditure per visit a $70 $73 $3 4.2 Psychotropic drugs Percent with any use 23.6% 23.0% -0.6 -2.8 Mean number of prescriptions per user a 5.0 6.5 1.5 28.7 Mean expenditure per prescription a $30 $49 $19 61.7 Mean expenditure per user a $150 $307 $157 104.7 Inpatient Percent with any admission 0.3% 0.2% -0.1 -40.0 Mean number of admissions per user 1.3 1.3 0.0 0.0 Mean length of stay (days) a 16.1 7.9 -8.2 -51.0 Mean expenditure per day a $546 $760 $214 39.1 SOURCE: Authors222 calculations from MarketScan 256 data. a Differences between 1992 and 1999 are st atistically significant at p < .05. Trends: What Drove Private Health Insurance Spending on Mental Health and Substance Abuse Care, 1992-1999? 6 Conclusions This analysis finds that private insurance spending on MH/SA treatment did not keep pace with total health care spending or with genera l price inflation. MH/SA as a proportion of total health care claims droppe d from 7.2 percent in 1992 to 5.1 percent in 1999. The decline in MH/SA spending is a chan ge from trends apparent in the 1980s, when the news media reported th at MH/SA cost growth had be en of considerable concern to employers. 4 A prior analysis of the MarketScan 256 data for 19862261988 by Richard Frank and colleagues found that charges for psychiatri c and substance abuse care rose at rates well above the rate for all health care : 20.1 percent and 32.4 pe rcent, respectively, compared with an overall ra te of increase of 13.0 percent. 5 Thus, while MH/SA expenditures were viewed as 223runaway224 in the late 1980s, data from the 1990s suggest that this was no longer the case. The decline in private insurance spending on MH/SA treatment was attributable to a dramatic decline in inpatient utilization227a decline that has been consistently found in other studies as well. One anal ysis indicated that inpatient dollars fell from 41.8 percent to 36.8 percent of total MH/SA spending. 6 The SAMHSA analysis did not decompose trends into inpatient and outpa tient treatment but did examin e trends by provider type and found a dramatic shift away from hospital- based care, particular ly care in specialty hospitals. 7 Our analysis indicates that the de cline in inpatient expenditures was attributable to declines in length-of-stay and in the per centage having any admission. Consistent with an analysis of other da ta, our analysis shows an increase in the probability of receiving outpatient treatment. 8 Although the probability of such treatment increased, the intensity of outpatient care use remained relatively constant. This suggests that the decline in inpatient us e is not being replaced by more intensive outp atient usage, as one might expect. Our analysis also shows that inpatient MH /SA spending declines were only partially offset by growth in prescription drug spendi ng. Drug spending growth is a major cost containment concern among third-party payers . Psychotropic drugs are among the fastest growing in terms of drug spending. Neverthe less, while drug spending grew 8.9 percent a year, inpatient spending fell 15.6 percent a year , more than offsetting drug cost growth, so that overall MH/SA spending fell from 1992 to 1999. The percentage of the population using a prescription medication for MH/SA treatment was found to be very high (14.5 percent in 1999). This percentage is higher than that found in the Zuvekas study (5.6 pe rcent in 1996). The difference is probably the result of different definitions of 223psychotropic medication.224 Zuvekas used self-reported prescription data and only counted prescriptions where the household also reported on an associated MH/SA condition. 9 In our study many more people had a psychotropic drug prescription than an MH/SA diagnosis. Fo r example, in 1999, 6.8 percent of the population had an MH/SA diagnosis on an outpa tient claim, whereas 14.5 percent of the population used a psychotropic medication. In 1999, 85 percent of th e prescriptions in this study were classified as antidepressants, benzodiazepines, or anxiolytics. Some of these medications may be prescribed for c onditions that the general population would not Trends: What Drove Private Health Insurance Spending on Mental Health and Substance Abuse Care, 1992-1999? 7 recognize as MH/SA conditions. For example, a commonly prescribed medication in this study222s population is Ambien, which has a prim ary indication for insomnia. It was chosen for the study as a psychotropic medication be cause it fell into the therapeutic class labeled anxiolytic/sedative/ hypnotic. Nominal price increases for outpatient and i npatient services in the time period were generally modest. For outpatient services, th e price per vis it actually fell. It should be noted that prices are not quality-adjusted. For example, a day in the hospital may not include the same mix of serv ices in 1992 as in 1999. The decline in MH/SA spending as a percen tage of total spending was greatest for children (falling from 13.4 percent to 6.6 percent) and was al so sizable for adults ages 1822654 (falling from 8.2 percent to 6.1 percent). MH/SA spending as a percentage of total spending actually grew slightly for older a dults (age fifty-five and older), from 2.5 percent to 2.8 percent. Trends in the mix of services were consistent across age groups, with notable declines in inpatient use and growth in pharmaceutical usage. The main limitation of this study is that it is based on a convenience sample of twenty- two large, self-insured employers with prim arily FFS or PPO coverage, so it might not be representative of national tre nds. Nevertheless, the sample size is quite large, and the findings from the decomposition are similar to the findings of othe r studies using data representative of the non-in stitutionalized civilian populatio n for all payers, not just private health insurance. Overall, this analysis shows a profound change in the nature of MH/SA treatment being received by employees of large corporations. Part of this change may be the result of financing changes. Employer surveys indicate that the number of health insurance plans with no specific limitations on inpatient care for mental illness fell from 37 percent in 1990 to 13 percent in 1996. 10 At the same time, prescription drug benefits became more common. Managed care grew rapidly over the 1990s and has been shown to shift care away from inpatient treatment. Behavioral health care carve-outs, as well as 223carved-in224 prior authorization and utilization review programs, were increa singly used during the late 1980s and 1990s to control MH/SA spending. Although these programs are held responsible for inpatient and outpatient MH/S A costs, they typically do not manage drug costs. Thus, the incentive may be to shif t away from managed or 223budgeted224 items227 inpatient and outpatient care227to 223off- budget224 items227prescription drugs. Technologies also changed over the 1990s and may be driving the shif t in service types. For example, new medications such as sele ctive serotonin reuptake inhibitors (SSRIs) and atypical antipsychotic medications have reduced the side effects associated with psychotropic treatment of depression and schizophrenia, allowing for improved compliance and perhaps a reduced need for inpatient care. One question that this analysis cannot answ er is whether the mix of services being supplied is optimal to meet the needs of patie nts. Are people being admitted to hospitals Trends: What Drove Private Health Insurance Spending on Mental Health and Substance Abuse Care, 1992-1999? 8 appropriately and for an optimal length of time? Is pharmaceutical use too high, too low, or about right? How many outpati ent visits are adequate? For example, research suggests that optimal treatment for depression comprises medication and psychotherapy. 11 It would be useful to discern whether patients ar e now more or less likely to get medication treatment alone. Further, one might examine whether early hospital discharges are now more likely to be associated with adverse ev ents such as suicides or accidents. Clearly, future work needs to focus on these questions. The authors acknowledge the programming suppor t of Jack Li. This project was funded through a contract with the Center for Subst ance Abuse Treatment and Center for Mental Health Services, Substance Abuse and Me ntal Health Services Administration (SAMHSA). The content is solely the responsibility of the authors and does not necessarily reflect the officia l views of SAMHSA or the U .S. Department of Health and Human Services. Notes 1 R.M. Coffey et al., National Expenditures for Mental Health and Substance Abuse Treatment, 1997, Pub. no. SMA-00-3499 (Rockville, Md.: Center for Substance Abuse Treatment and Center for Mental Health Se rvices, Substance Abuse and Mental Health Services Administration, July 2000); T.L. Mark et al., 223Spending on Mental Health and Substance Abuse Treatment, 19872261997,224 Health Affairs (July/Aug 2000): 108226120; T.L. Mark et al., National Expenditures for Mental Health, Alcohol, and Other Drug Abuse Treatment, 1996, SMA-98-3255 (Rockville, Md.: SAMHSA, 1998); and D. McKusick et al., 223Spending for Mental Health and Substance Abuse Treatment, 1996,224 Health Affairs (Sep/Oct 1998): 147226157. 2 S.H. Zuvekas, 223Trends in Mental Health Services Use and Spending, 19872261996,224 Health Affairs (Mar/Apr 2001): 214226224; D.L. Leslie and R. Rosenheck, 223Changes in Inpatient Mental Health Utilization Costs in a Privately Insured Population, 1993 to 1995,224 Medical Care 37, no. 5 (1999): 457226468; Hay Group Management, Inc., 223Health Care Plan Design and Cost Trends2271988 through 1997,224 Report prepared for the National Association of Psychiatric Health Systems, Association of Behavioral Group Practices, and National Alliance for the Mentally Ill (Arlington, Va.: Hay Group, April 1999); and Coffey et al., National Expenditures for Mental Health and Substance Abuse Treatment, 1997. 3 Claims with an International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) code greater or equal to 290.0 and less than or equal to 315.0, or ICD-9-CM codes equal to 648.4, 797.xx, 648.3, V40.2, V40.3, V40.9, V61.0, V61.1, V66.3, V67.3, V70.1, V70.2, V71.0, V71.02, or V71.09. Trends: What Drove Private Health Insurance Spending on Mental Health and Substance Abuse Care, 1992-1999? 9 4 See, for example, A. Bruzzese, 223Double Digit Increase Per Worker Hits Health Costs,224 Employee Benefit News (March 1989); T.W. Ferguson, 223Any Wonder Medical Premiums Are Anything but Shrinking,224 Wall Street Journal, 22 May 1990; and L.D. Williams, 223Businesses Facing Rising Mental Health Care Costs Are Turning to Managed Care,224 Baltimore Sun, 5 August 1990. 5 R.G. Frank, D.S. Salkever, and S.S. Sharfstein, 223A New Look at Rising Mental Health Insurance Costs,224 Health Affairs (Summer 1991): 116226123. 6 Zuvekas, 223Trends in Mental Health Services Use and Spending.224 7 Coffey et al., National Expenditures for Mental Health and Substance Abuse Treatment, 1997. 8 Zuvekas, 223Trends in Mental Health Services Use and Spending.224 9 Ibid. 10 Hay Group, 223Health Care Plan Design and Cost Trends.224 11 Y. Burnand et al., 223Psychodynamic Psychotherapy and Clomipramine in the Treatment of Major Depression,224 Psychiatric Services 53, no. 5 (2002): 585226590; and C.F. Reynolds et al., 223Nortriptyline and Interpersonal Psychotherapy as Maintenance Therapies for Recurrent Major Depression: A Randomized Controlled Trial in Patients Older than Fifty-nine Years,224 Journal of the American Medical Association 281, no. 1 (1999): 3922645. Trends: What Drove Private Health Insurance Spending on Mental Health and Substance Abuse Care, 1992-1999? 10