Data were drawn from the household component of the 1998 and the 2007 MEPS, which are subsamples of the National Health Interview Survey (NHIS) (10, 11). Both surveys were sponsored by the Agency for Healthcare Research and Quality to provide national estimates of the use of, expenditures on, and financing of health care services. These surveys were conducted as national probability samples of the U.S. noninstitutionalized civilian population and were designed to provide nationally representative estimates to be compared over time. All data elements in these analyses except the physician provider specialty variable were the same in the two surveys.
A total of 22,953 participants provided data for the entire 1998 survey year from two separate overlapping panels, each of which included three rounds of interviews. The full-year response rate was 67.9% after factoring in the effects of nonresponse to NHIS, nonresponse to the first round of MEPS, and survey attrition from both panels (10). A total of 29,730 participants provided data for the entire 2007 survey, the most recent available data, for a full-year response rate of 56.9% (11). For both surveys, a designated informant was queried about all related persons who lived in the household.
The Agency for Healthcare Research and Quality devised weights to adjust for the complex survey designs and yield unbiased national estimates. The sampling weights also adjust for non-response and poststratification to population totals based on U.S. census data. More detailed discussions of the design, sampling, and adjustment methods have been presented elsewhere (10, 11).
The MEPS included a series of three in-person interviews during each study year. Respondents were asked to record medical events, as they occurred, in a calendar/diary that was reviewed in-person during each interview. Written permission was obtained from selected survey participants to contact the medical providers they mentioned during the survey to verify service use, charges, and sources and amounts of payments.
Mental health conditions.
The MEPS collected information on the diagnosis for each visit to hospital outpatient departments and office-based outpatient care. This information was coded in a manner that permitted classification according to ICD-9 categories by professional coders. The diagnoses associated with the psychotherapy visits were then grouped into the following mental health condition categories: schizophrenia and related disorders (codes 295, 297-299), depression and related mood disorders (codes 296.2, 296.3, 298.0, 300.4, 309.1, 311), anxiety disorders (codes 293.84, 300.0, 300.2, 300.3, 3008.3, 309.81), childhood disorders and mental retardation (codes 299, 312-315, 317-319, 307 [except 307.2 and 307.8]), adjustment disorders (codes 308 [except 308.3], 309.0, 309.1, 309.2, 309.4, 309.9), other mental disorders (codes 290-319 not specified above), and subsyndromal mental health-related conditions, including psychosocial circumstances (codes V40, V61, V62), sleep disturbance (780.5), malaise and fatigue (780.7), and nervousness (799.2). Separate variables classified emergency department visits and inpatient admissions with a diagnosis of a mental disorder (ICD-9-CM codes 290-319).
The MEPS asked respondents what type of care was provided during each outpatient visit and whether it was from a mental health specialist or other health care provider, using a set of response categories that included "mental health counseling or psychotherapy." Mental health counseling or psychotherapy is defined as "a treatment technique for certain forms of mental disorders relying principally on talk/conversation between the mental health professional and the patient." It specifically includes "individual, family, and/or group therapies" (12). Visits for psychotherapy or mental health counseling are considered psychotherapy visits.
The MEPS surveys asked respondents about medications bought or otherwise obtained by survey participants during the survey year. Psychotropic medications were grouped by therapeutic class as antidepressants, antipsychotic medications, anxiolytics/hypnotics, stimulants, and mood stabilizers (the latter included lithium, lamotrigine, carbamaze-pine, and valproate or valproic acid for respondents who were not treated for seizure disorders [ICD-9-CM code 345]).
The MEPS solicited information on the type of health care professionals providing treatment at each visit. We classified mental health providers of psychotherapy into three groups: social workers, psychologists, and psychiatrists. Information was not available in 1998 concerning psychotherapy visits provided by psychiatrists. A psychotherapy user was considered to have been treated by a given provider type if the user reported making one or more visits to that type during the survey year.
Expenditures and source of payment.
The MEPS data include sources of expenditures for each health care service. "Expenditures" refers to what is paid for the medical service and is defined as the sum of payments for each medical service that was obtained, including out-of-pocket payments and payments made by private insurance, Medicaid, Medicare, and other sources (10, 11). From these data, total expenditures were aggregated into outpatient medical care (outpatient visits and medications for all conditions), outpatient mental health care (outpatient visits and medications for mental health conditions), and outpatient psychotherapy. Summary variables were also constructed for six payment sources, including self-payment, private insurance, Medicaid, Medicare, other federal programs, and a residual group of other sources.
For each survey year, the percentage of persons using psychotherapy was computed overall and stratified by several sociodemographic characteristics. Trends were then examined by mental health condition group in the estimated national number of treated outpatients and their distribution with respect to treatment with psychotherapy but not psychotropic medication, psychotherapy and psychotropic medication, and psychotropic medication alone. The distributions of psychotherapy users were then examined by use of psychotropic medications, mental health provider groups, acute mental health service use, number of psychotherapy visits during the year, and self-assessed mental health status (excellent, very good, or good versus fair or poor). Among psychotherapy users, the mean number of psychotherapy visits in each survey year was compared overall and for psychotherapy users with and without psychotropic medication use as well as for those with good to excellent as compared with fair or poor self-rated mental health. In separate analyses, total national expenditures were estimated for all outpatient medical care, out-patient mental health care (outpatient visits for a mental disorder or condition), and psychotherapy. National psychotherapy expenditures were partitioned by payment source for the two survey years. The U.S. Consumer Price Index for medical care was used to adjust 1998 expenditures to 2007 dollars.
All statistical analyses were conducted using SAS, version 9.2 (SAS Institute, Cary, N.C.), using SURVEY procedures to accommodate the complex sample design and the weighting of observations. A series of logistic regressions, adjusted for age, sex, race/ethnicity, and insurance status, were performed to assess the strength of associations between year (with 1998 as the reference year) and psychotherapy use; results are presented as adjusted odds ratios with 95% confidence intervals. Similar analyses were conducted with psychotherapy and psychotropic medication use as well as psycho-tropic medication use only as dependent variables. Linear regression was used to assess change in the number of psychotherapy visits per year among respondents reporting psychotherapy use, and z tests were used to evaluate changes in total national expenditures. All tests were two-sided, and alpha was set at 0.05.