|Home | About | Journals | Submit | Contact Us | Français|
To compare adults with different insurance coverage in care for alcohol, drug abuse, and mental health (ADM) problems.
From a national telephone survey of 9,585 respondents.
Follow-up of adult participants in the Community Tracking Study.
Self-report survey of insurance plan (Medicare, Medicaid, unmanaged, fully, or partially managed private, or uninsured), ADM need, use of ADM services and treatments, and satisfaction with care in the last 12 months.
Logistic and linear regressions were used to compare persons by insurance type in ADM use.
The likelihood of ADM care was highest under Medicaid and lowest for the uninsured and those under Medicare. Perceived unmet need was highest for the uninsured and lowest under Medicare. Persons in fully rather than partially managed private plans tend to be more likely to have ADM care and ADM treatments given need. Satisfaction with care was high in public plans and low for the uninsured.
The uninsured have the most problems with access to and quality of ADM care, relative to the somewhat comparable Medicaid population. Persons in fully managed plans had better rather than worse access and quality compared to partially managed plans, but findings are exploratory. Despite low ADM use, those with Medicare tend to be satisfied. Across plans, unmet need for ADM care was high, suggesting changes are needed in policy and practice.
The recent Surgeon General's Report on Mental Health (U.S. Department of Health and Human Services 1999) emphasized the importance of systematically addressing unmet need for mental health care nationally; similar concerns have been raised about substance abuse (Institute of Medicine 1990a; 1990b). Policy formulation, however, would benefit from recent data on the distribution of unmet need for alcohol, drug, and mental health (ADM) care across diverse types of insurance coverage.
For example, persons with psychiatric disorders are more likely than others to become uninsured (Sturm and Wells 2000), but it is unclear whether the uninsured have greater unmet need for ADM care than insured groups. One study suggests that the uninsured may have similar ADM need but lower access to mental health care than persons with Medicaid (Norquist and Wells 1991). Further, despite public concerns about access or quality of managed care for ADM conditions, it is unclear whether persons in more intensively managed plans face greater unmet need (Hall et al. 1997; Mechanic and McAlpine 1999). Most prior studies comparing ADM care among private plan types were based on regional samples or specific plans and were conducted prior to recent developments in managed care. These studies suggest similar or greater access to mental health care in staff model Health Maintenance Organizations (HMOs) relative to fee-for-service; while studies of depressed patients suggest similar access but lower quality of care for some prepaid patients compared to fee-for-service patients (Manning et al. 1987; Diehr et al. 1984; Sturm et al. 1995; Wells et al. 1996). Young et al. (2001) found similarly low rates of appropriate care for depressive or anxiety disorder among adults in the U.S. population with no insurance or public or private insurance.
This paper provides profiles of access to, unmet need for, and quality of ADM care for a national sample of adults with telephones who have public, private, or no insurance. We hypothesized that there would be lower access to and higher unmet need for ADM care among the uninsured relative to Medicaid or the privately insured, and similar access to ADM across private plans. We were uncertain about expectations for quality of care given limited prior data.
We analyzed data from Healthcare for Communities (HCC), which reinterviewed a random sample drawn from 30,375 adult telephone respondents from the Community Tracking Study (Kemper, Blumenthal, Corrigan et al., 1996). For HCC, 14,985 respondents were selected to oversample persons with high psychological distress, specialty ADM use, and family income below the federal poverty level. We obtained 9,585 eligible responses (64 percent response) and weighted data for the sampling design and nonresponse to represent the noninstitutionalized U.S. population (Sturm et al. 1999).
We categorize main insurance plan as uninsured, Medicaid, Medicare, or unmanaged, partially managed, and fully managed private plans. Plans with gatekeeping and preauthorization for specialty care plus a closed provider panel were considered fully managed, those with some of these features were considered partially managed, and plans without these features were unmanaged.
We used survey items to assess age, family income, sex, marital status, ethnicity (white, black, Hispanic, other), and education.
Clinical need is defined as probable 12-month psychiatric disorder or substance abuse. Psychiatric disorder was assessed by major depressive, dysthymic, or generalized anxiety disorder on the Composite International Diagnostic Interview, Short Form (CIDI-SF); probable panic disorder by a positive CIDI item and a role limitation on the SF-12; and probable severe mental disorder by a positive CIDI lifetime mania item or report of an inpatient stay for psychosis or a doctor's diagnosis of schizophrenia (Kessler et al. 1998; Ware, Kosinski, and Keller 1995; World Health Organization 1995). Substance abuse was assessed by alcohol abuse on the AUDIT (World Health Organization 1995) or illicit substance use as reported on items adapted from the CIDI-SF (Kessler et al. 1998). Perceived need was based on responses to two items asking individuals if they “needed help for emotional or mental health problems, such as feeling sad, blue, anxious, or nervous” or “needed help for alcohol or drug problems.” Mental health-related quality of life was assessed by the MCS-12, the global mental scale of the SF-12 (Ware, Kosinski, and Keller 1995). Medical comorbidity was assessed by a count among 17 common chronic medical conditions.
Access to care in the last 12 months was measured by self-report of any outpatient mental health specialty visit and any visit to a general medical provider that included either counseling, suggestions to cut down on drinking or avoid recurrences, or a specialty referral. We also developed an indicator of having any ADM care, that is, inpatient, day-treatment, or residential care, or an emergency room or outpatient visit for ADM problems.
Persons with perceived ADM need, but no ADM care, were classified as having unmet need, while those reporting delays in care or receiving less care than needed are classified as having delayed care.
Among persons with clinical or perceived need, we assessed satisfaction with care available in the prior 12 months using single-item measures for overall care, care for emotional or mental health problems, and care for substance abuse problems. Many did not answer items about satisfaction with substance abuse care, and we analyze only complete responses.
To distinguish active treatment from visits with assessments only, we developed an indicator of having inpatient, day-treatment, or residential ADM care, use of prescribed psychotropic medication daily for a month or more, or “potentially therapeutic” outpatient care, that is, four or more ADM visits or use of counseling techniques (i.e., improving skills in relationships, coping with loss, ways to relax, taking responsibility for substance abuse problems, avoiding recurrences, or participating in enjoyable activities).
To describe plan groups, we regressed each health and demographic indicator on insurance status. For main analyses, we regressed each of the 10 ADM care indicators on insurance status. Analyses of unmet need were limited to respondents with perceived ADM need. Analyses of satisfaction and active treatment were limited to persons with clinical or perceived need. To determine whether plan differences were due to confounding with individual characteristics, we conducted parallel analyses controlling for covariates. We used logistic regression for dichotomous dependent variables and linear regression for continuous variables. We tested the overall effect of plan type using an F test for the set of plan coefficients. With 10 ADM care indicators, we focus on describing patterns of ADM care when the overall plan effect is significant at p < .005, a criterion met for all analyses. We describe results for Medicare as a unique population and focus on six pair-wise plan comparisons from the regression coefficients. We contrast fully managed with unmanaged and partially managed plans, using an exploratory criterion of p < .10 due to lack of empirical precedent. We compare the uninsured with Medicaid and each private plan type. For these comparisons, one-tailed tests are appropriate and results are in a consistent direction, so a formal Bonferroni correction of p < .004 is too conservative (Miller 1981). Thus, we focus on results significant at p < .01. To illustrate results, we generated means and percentages (either unadjusted or adjusted for covariates) and calculated standard errors using the parameters of the regression models. Some variables (especially income) had missing data, and we used a multiple imputation method (Little and Rubin 1987; Rubin 1987; Schafer 1997). All analyses are adjusted for the clustered sampling (SUDAAN Software 1997; Yates and Grundy 1953) and are nationally weighted. This appendix is available at http://www.blackwell-synergy.com or at http://www.hospitalconnect.com/hsr/Main.html.
Demographic factors differ by plan type, but private groups are similar (Table 1). Persons with Medicare or Medicaid are more likely to be female than persons in other plans. Minorities are overrepresented among the uninsured and Medicaid groups, and education level is lower in these plans. Prevalence of probable psychiatric disorder varies widely and is highest for those with Medicaid and lowest for those with Medicare. The uninsured have the highest probability of a substance abuse problem while persons with Medicare have the lowest.
Uninsured adults have significantly lower access to each type of ADM care than do persons with Medicaid (e.g., for any ADM care, T=6.78, p < .001), and significantly lower access to any ADM care and general medical ADM care than persons with fully or partially managed private plans (e.g., for any ADM care, T=4.84, p < .001 for fully managed plans). Among private plans, fully managed plans tend to have higher access to any ADM care and general medical ADM care compared to unmanaged plans (Ts are 3.16 and 3.12 respectively, each p < .002) and partially managed plans (Ts are 2.37, p = .02 and 2.13, p = .05, respectively). In addition, use of specialty ADM care tends to be higher for fully managed than unmanaged plans (t=1.69, p = .10) (Table 2). Use of specialty ADM care is highest for Medicaid and the lowest for Medicare.
Among those with perceived need, unmet need and delays in care are greatest for the uninsured and those on Medicaid and lowest for those on Medicare. The uninsured have significantly more unmet need than each private plan group (e.g., comparison to fully managed plan has T=4.81, p < .001). Differences in unmet need and delays in care between uninsured and Medicaid plans and among private plan types are not consistently statistically significant. Among those with clinical or perceived need, rates of active treatment are significantly higher for Medicaid and fully managed care plans relative to the uninsured (Ts are 5.88 and 4.85 respectively, each p < .001). Active treatment given ADM need is somewhat more likely in fully managed than partially managed plans (T=2.46, p = .02). Satisfaction with each type of care is lowest among the uninsured, who have significantly lower satisfaction of each type compared to Medicaid (each p < .001) and private plans. Results are qualitatively similar in adjusted models but a few individual comparisons fall in significance to p = 0.10.
Our findings indicate consistently lower access to and satisfaction with ADM care and lower rates of active treatment given need for the uninsured relative to Medicaid and privately insured groups. These findings reinforce the need for policies that expand insurance coverage to the uninsured, especially those with ADM need. Recent federal programs to expand insurance coverage mainly target children and are variably implemented within and across states (Edmunds and Coye 1998a; 1998b). Further, Medicaid recipients had high rates of perceived unmet need and only moderate rates of receiving active treatment given need; so coverage expansion without addressing supply or public education policy may be insufficient. Among private plans, we observed a trend in exploratory analyses for higher rather than lower access to ADM care for fully managed compared to partially or unmanaged plans and for higher rates of active treatment given ADM need for fully, compared to partially, managed plans. This pattern may be due to the more fully developed provider networks or lower copayments in fully managed plans; but these results counter public opinion about lower access or quality in more fully managed plans. Further research is needed on differences in ADM care as a function of management intensity. Descriptively, we observed a pattern among Medicare recipients with low use of ADM services and only moderate rates of active treatment given need, yet very low perceived unmet need and relatively high satisfaction with ADM care. This profile implies there may be some important quality gaps among the elderly but little leverage for change. This suggests that there is an important role for public education programs about ADM care for the elderly.
The findings are based on self-reports over a long recall period; this may particularly affect the validity of estimates of general medical ADM care that are based on recalling specific provider actions within visits. We found that this approach leads to higher estimates of ADM care than in studies that rely on respondents to directly designate primary care visits as due to mental health problems (Regier et al. 1993; Kessler et al. 1994). The visits we assessed do not necessarily reflect effective ADM care, a concern commonly voiced about general medical visits (Mechanic et al. 1999; Young et al. 2001). However, even brief counseling reduces alcohol use and may be considered ADM care (Fleming et al. 1997; Gentileilo et al. 1995). Our study was limited to adults with telephones, so we might underestimate the extent of access or quality problems among poorer populations. We did not have data on carve-out ADM specialty management, another coverage domain that might affect access or quality (Sturm and Sherbourne 2000). We cannot comment on the effects of managed care within the public sector or on the severely mentally ill who are uncommon in a household sample. The survey had moderate response and was a follow-up of a prior survey with moderate response, which could bias results despite weighting.
In summary, we found low access to ADM care for the uninsured; programs extending insurance to this population might improve ADM care. In addition, access and quality of care may be higher rather than lower in fully managed plans compared to those less managed, emphasizing the importance of monitoring access and quality nationally across plan types. Public plans have more counterbalancing patterns: high access and satisfaction but substantial unmet need among Medicaid recipients; and moderate rates of active treatment given need, but low perceived unmet need and high satisfaction among Medicare recipients. These findings suggest a need for supply-side interventions within Medicaid and enrollee education within Medicare. Yet across plans, many individuals had unmet need for ADM care, reinforcing the Surgeon General's conclusion that broad policy and practice changes are needed to improve ADM care access and quality.
We would like to thank Fuan Yue Kung, M.S., for his programming support, and Lingqi Tang, Ph.D., and Ruth Klap, Ph.D., for their statistical advice and database management support.
Funded by The Robert Wood Johnson Foundation, Healthcare for Communities Grant No. 031280, and The National Institute of Mental Health (NIMH), Research Center on Managed Care for Psychiatric Disorders Grant No. MHO 1170-04.