Predicting the use of outpatient mental health services: do modeling approaches make a difference?

HOW DO CHANGES IN INCOME, EMPLOYMENT AND HEALTH INSURANCE AFFECT FAMILY MENTAL HEALTH SPENDING?

Using eight two-year panels from the Medical Expenditure Panel Survey data for the period 2004 to 2012, we examine the effect of economic shocks on mental health spending by families with children. Estimating two-part expenditure models within the correlated random effects framework, we find that employment shocks have a greater impact on mental health spending than do income or health insurance shocks. Our estimates reveal that employment gains are associated with a lower likelihood of family mental health services utilization. By contrast employment losses are positively related to an increase in total family mental health. We do not detect a link between economic shocks and mental health spending on behalf of fathers.

The Impact of State Medical Malpractice Reform on Individual-Level Health Care Expenditures

BACKGROUND

Past studies of the impact of state-level medical malpractice reforms on health spending produced mixed findings. Particularly salient is the evidence gap concerning the effect of different types of malpractice reform. This study aims to fill the gap. It extends the literature by examining the general population, not a subgroup or a specific health condition, and controlling for individual-level sociodemographic and health status.

METHODS

We merged the Database of State Tort Law Reforms with the Medical Expenditure Panel Survey between 1996 and 2012. We took a difference-in-differences approach to specify a two-part model for analyzing individual-level health spending. We applied the recycled prediction method and the bootstrapping technique to examining the difference in health spending growth between states with and without a reform. All expenditures were converted to 2010 U.S. dollars.

RESULTS

Only two of the 10 major state-level malpractice reforms had significant impacts on the growth of individual-level health expenditures. The average annual expenditures in states with caps on attorney contingency fees increased less than that in states without the reform (p < .05). Compared with states with traditional contributory negligence rule, the average annual expenditures increased more in both states with a pure comparative fault reform (p < .05) and states with a comparative fault reform that barred recovery if the plaintiff's fault was equal to or greater than the defendant's (p < .05).

CONCLUSIONS

A few state-level malpractice reforms had significantly affected the growth of individual-level health spending, and the direction and magnitude of the effects differed by type of reform.

Incremental dementia-related expenditures in a medicaid population

OBJECTIVES

With the growing number of older adults, understanding expenditures associated with treating medical conditions that are more prevalent among older adults is increasingly important. The objectives of this research were to estimate incremental medical encounters and incremental Medicaid expenditures associated with dementia among Indiana Medicaid recipients 40 years or older in 2004.

METHODS

A retrospective cohort design analyzing Indiana Medicaid administrative claims files was used. Individuals at least 40 years of age with Indiana Medicaid eligibility during 2004 were included. Patients with dementia were identified via diagnosis codes in claims files between July 2001 and December 2004. Adjusted annual incremental medical encounters and expenditures associated with dementia in 2004 were estimated using negative binomial regression and zero-inflated negative binomial regression models.

RESULTS

A total of 18,950 individuals (13%) with dementia were identified from 145,684 who were 40 years or older. The unadjusted mean total annualized Medicaid expenditures for the cohort with dementia ($28,758) were significantly higher than the mean expenditures for the cohort without dementia ($14,609). After adjusting for covariates, Indiana Medicaid incurred annualized incremental expenditures of $9,829 per recipient with dementia. Much of the annual incremental expenditure associated with dementia was driven by the higher number of days in nursing homes and resulting nursing-home expenditures. Drug expenditures accounted for the second largest component of the incremental expenditures. On the basis of disease prevalence and per recipient annualized incremental expenditures, projected incremental annualized Indiana Medicaid spending associated with dementia for persons 40 or more years of age was $186 million.

CONCLUSIONS

Dementia is associated with significant expenditures among Medicaid recipients. Disease management initiatives designed to reduce nursing-home use among recipients with dementia may have much potential to decrease Medicaid expenditures associated with dementia.

ESTIMATING TREATMENT EFFECTS ON HEALTHCARE COSTS UNDER EXOGENEITY: IS THERE A 'MAGIC BULLET'?

Methods for estimating average treatment effects, under the assumption of no unmeasured confounders, include regression models; propensity score adjustments using stratification, weighting, or matching; and doubly robust estimators (a combination of both). Researchers continue to debate about the best estimator for outcomes such as health care cost data, as they are usually characterized by an asymmetric distribution and heterogeneous treatment effects,. Challenges in finding the right specifications for regression models are well documented in the literature. Propensity score estimators are proposed as alternatives to overcoming these challenges. Using simulations, we find that in moderate size samples (n= 5000), balancing on propensity scores that are estimated from saturated specifications can balance the covariate means across treatment arms but fails to balance higher-order moments and covariances amongst covariates. Therefore, unlike regression model, even if a formal model for outcomes is not required, propensity score estimators can be inefficient at best and biased at worst for health care cost data. Our simulation study, designed to take a 'proof by contradiction' approach, proves that no one estimator can be considered the best under all data generating processes for outcomes such as costs. The inverse-propensity weighted estimator is most likely to be unbiased under alternate data generating processes but is prone to bias under misspecification of the propensity score model and is inefficient compared to an unbiased regression estimator. Our results show that there are no 'magic bullets' when it comes to estimating treatment effects in health care costs. Care should be taken before naively applying any one estimator to estimate average treatment effects in these data. We illustrate the performance of alternative methods in a cost dataset on breast cancer treatment.

Risk-adjusted capitation rates for children: how useful are the survey-based measures?

OBJECTIVE

Despite the recognition by some experts that survey measures have the potential to improve capitation rates for those with chronic conditions, few studies have examined risk-adjustment models for children, and fewer still have focused on survey measures. This study evaluates the performance of risk-adjustment models for children and examines the potential of survey-based measures for improving capitation rates for children.

DATA SOURCES

The study sample includes 8,352 Medicaid children who were followed up for 2 years by the Medical Expenditure Panel Survey in 2000-2005.

STUDY METHODS

Children's information in 1 year was used to predict their expenditures in the next year. Five models were estimated, including one each that used demographic characteristics, subjectively rated health status, survey measures about children with special health care needs (CSHCN), prior year expenditures, and Hierarchical Condition Category (HCC), which is a diagnosis-based model. The models were tested at the individual level using multiple regression methods and at the group level using split-half validation to evaluate their impact on expenditure predictions for CSHCN.

PRINCIPAL FINDINGS

The CSHCN information explained higher proportion of the variance in annual expenditures than the subjectively rated health status, but less than HCC measures and prior expenditures. Adding the CSHCN information into demographic factors as adjusters would remarkably increase capitation rates for CSHCN.

CONCLUSIONS

Survey measures, such as the CSHCN information, can improve risk-adjustment models, and their inclusion into capitation adjustment may help provide appropriate payments to managed-care plans serving this vulnerable group of children.

A multi-worksite analysis of the relationships among body mass index, medical utilization, and worker productivity

BACKGROUND

The relationships between worker health and productivity are becoming clearer. However, few large scale studies have measured the direct and indirect cost burden of overweight and obesity among employees using actual biometric values. The objective of this study was to quantify the direct medical and indirect (absence and productivity) cost burden of overweight and obesity in workers.

MEASURES

A cross-sectional study of 10,026 employees in multiple professions and worksites across the United States was conducted. The main outcomes were five self-reported measures of workers' annual health care use and productivity: doctor visits, emergency department visits, hospitalizations, absenteeism (days absent from work), and presenteeism (percent on-the-job productivity losses). Multivariate count and continuous data models (Poisson, negative binomial, and zero-inflated Poisson) were estimated.

RESULTS

After adjusting for covariates, obese employees had 20% higher doctor visits than normal weight employees (confidence interval [CI] 16%, 24%, P < 0.01) and 26% higher emergency department visits (CI 11%, 42%, P < 0.01). Rates of doctor and emergency department visits for overweight employees were no different than those of normal weight employees. Compared to normal weight employees, presenteeism rates were 10% and 12% higher for overweight and obese employees, respectively (CI 5%, 15% and 5%, 19%, all P < 0.01). Taken together, compared to normal weight employees, obese and overweight workers were estimated to cost employers $644 and $201 more per employee per year, respectively.

CONCLUSIONS

This study provides evidence that employers face a financial burden imposed by obesity. Implementation of effective workplace programs for the prevention and management of excess weight will benefit employers and their workers.

Sociodemographic, perceived and objective need indicators of mental health treatment use and treatment-seeking intentions among primary care medical patients

We explored sociodemographic and illness/need associations with both recent mental healthcare utilization intensity and self-reported behavioral intentions to seek treatment. Data were examined from a community sample of 201 participants presenting for medical appointments at a Midwestern U.S. primary care clinic, in a cross-sectional survey study. Using non-linear regression analyses accounting for the excess of zero values in treatment visit counts, we found that both sociodemographic and illness/need models were significantly predictive of both recent treatment utilization intensity and intentions to seek treatment. Need models added substantial variance in prediction, above and beyond sociodemographic models. Variables with the greatest predictive role in explaining past treatment utilization intensity were greater depression severity, perceived need for treatment, older age, and lower income. Robust variables in predicting intentions to seek treatment were greater depression severity, perceived need for treatment, and more positive treatment attitudes. This study extends research findings on mental health treatment utilization, specifically addressing medical patients and using statistical methods appropriate to examining treatment visit counts, and demonstrates the importance of both objective and subjective illness/need variables in predicting recent service use intensity and intended future utilization.

Statistical procedures for analyzing mental health services data

In mental health services research, analyzing service utilization data often poses serious problems, given the presence of substantially skewed data distributions. This article presents a non-technical introduction to statistical methods specifically designed to handle the complexly distributed datasets that represent mental health service use, including Poisson, negative binomial, zero-inflated, and zero-truncated regression models. A flowchart is provided to assist the investigator in selecting the most appropriate method. Finally, a dataset of mental health service use reported by medical patients is described, and a comparison of results across several different statistical methods is presented. Implications of matching data analytic techniques appropriately with the often complexly distributed datasets of mental health services utilization variables are discussed.

Outpatient medical and mental healthcare utilization models among military veterans: results from the 2001 National Survey of Veterans

Using Andersen's (1995) [Andersen RM. Revisiting the behavioral model and access to medical care: does it matter? Journal of Health and Social Behavior 1995;36:1-10] behavioral model of healthcare use as our theoretical framework, we examined predisposing (i.e., sociodemographic), enabling (i.e., access resources), and need (i.e., illness) models of outpatient medical and mental healthcare utilization among a national sample of US veterans. Participants were 20,048 nationally representative participants completing the 2001 National Survey of Veterans. Outcomes were healthcare use variables for the past year, including the number of Veterans Affairs (VA) and non-VA outpatient healthcare visits, and whether VA and non-VA mental health treatment was used. Univariate results demonstrated that numerous predisposing, enabling and need variables predicted both VA and non-VA healthcare use intensity and mental healthcare use. In multivariate analyses, predisposing, enabling and need variables demonstrated significant associations with both types of healthcare use, but accounted for more variance in mental healthcare use. Need variables provided an additive effect over predisposing and enabling variables in accounting for medical and mental healthcare use, and accounted for some of the strongest effects. The results demonstrate that need remains an important factor that drives healthcare use among veterans and does not seem to be overshadowed by socioeconomic factors that may create unfair disparities in treatment access.

The Effects of State Parity Laws on the Use of Mental Health Care

OBJECTIVE

We used a quasiexperimental research design to measure the effect of state parity laws on the use of mental health care in the past year.

METHODS

We pooled cross-sectional data from the 2001, 2002, and 2003 National Surveys on Drug Use and Health. Our sample included 83,531 adults 18 years of age or over with private health insurance stratified by the level of mental and emotional distress experienced in the worst month of the past year. We used a state and year-fixed effects approach to measure the effect of parity. Similar to a difference-in-difference analysis, the effect of parity was measured by comparing pre-/postchanges in mental health service use within states that switched active parity status to changes in service use within states that did not change parity status in the same calendar year. For each subgroup, we report predictions of the percentage point change in any mental health care use, prescription drug use, and outpatient care use resulting from parity laws.

RESULTS

Depending on the time window used to define active parity status, we found that parity increased the probability of using any mental health care in the past year by as much as 1.2 percentage points (P<0.01) for the lower distress group and by as much as 1.8 percentage points (P<0.05) in the middle distress group. We found no statistically significant changes in service use for the upper distress group. Whether measured differences were attributable to changes in the use of prescription drug or outpatient care also depended on the definition of active parity status.

CONCLUSIONS

Overall, the results of this study suggest that state parity laws succeeded in expanding access to mental health care for those with relatively mild mental health problems.

Recent treatment history vs clinical characteristics in the prediction of use of outpatient psychiatric services

BACKGROUND

The use of outpatient psychiatric services has been shown to be a complex function of sociodemographic, clinical, and pathway variables. The relative contribution of each variable or groups of variables in explaining the variability in the use of outpatient psychiatric services, however, remains poorly documented.

METHODS

The subjects (N=382) were all patients admitted to an outpatient psychiatric clinic serving mostly a minority and low-income population. The charts of the patients were reviewed for sociodemographic, clinical, and pathway variables and the number of outpatient visits. The pathway variables studied were source of referral and most recent psychiatric treatment service used. Both bivariate and multivariate statistics were used to analyze the data.

RESULTS

Pathway variables were better predictors of the number of outpatient visits than clinical variables after controlling for sociodemographic variables.

CONCLUSION

Patients recently hospitalized may be sicker or have fewer social supports and therefore require more outpatient visits. Recent treatment history stands out as an important variable in the prediction of the number of outpatient mental health visits. More research is needed to examine the influence of pathway variables on treatment decisions.

The effects of state mental health parity legislation on perceived quality of insurance coverage, perceived access to care, and use of mental health specialty care

OBJECTIVE

To assess the impacts of recent state mental health parity legislation on perceived quality of health insurance coverage, perceived access to needed health care, and use of mental health specialty services by individuals with likely need for mental health care.

DATA SOURCES

The study sample came from two waves of a national household survey first fielded in 1997-1998 and then in 2000-2001. The analysis used a subset of the sample.

STUDY DESIGN

The study took the Difference-in-Difference-in-Difference approach to investigate changes in self-perceived quality of health insurance coverage and access to needed health care, and use of mental health specialty care by the group with mental disorders (relative to those without) in states with parity legislation of different comprehensiveness (relative to the nonparity states) in the years after the law (relative to before the law).

PRINCIPAL FINDINGS

Overall, there were no significant or consistent effects of the parity legislation. Descriptive statistics showed significant changes in some (but not all) outcome variables, but these results disappeared in detailed statistical analyses by controlling for important covariates.

CONCLUSIONS

The null findings of the effects of state mental health parity mandates suggest that under ERISA (Employee Retirement Income Security Act), the scope of state parity legislation may have been restricted because of large proportion of self-insured employers. Furthermore, comprehensiveness of state legislation appears to be related to the traditional level of use of mental health specialty care, which becomes another confounder for the potential policy effects.