Comparison of risk-adjustment systems for the medicaid-eligible disabled population

Exploring the predictive power of interaction terms in a sophisticated risk equalization model using regression trees

This study explores the predictive power of interaction terms between the risk adjusters in the Dutch risk equalization (RE) model of 2014. Due to the sophistication of this RE-model and the complexity of the associations in the dataset (N = ~16.7 million), there are theoretically more than a million interaction terms. We used regression tree modelling, which has been applied rarely within the field of RE, to identify interaction terms that statistically significantly explain variation in observed expenses that is not already explained by the risk adjusters in this RE-model. The interaction terms identified were used as additional risk adjusters in the RE-model. We found evidence that interaction terms can improve the prediction of expenses overall and for specific groups in the population. However, the prediction of expenses for some other selective groups may deteriorate. Thus, interactions can reduce financial incentives for risk selection for some groups but may increase them for others. Furthermore, because regression trees are not robust, additional criteria are needed to decide which interaction terms should be used in practice. These criteria could be the right incentive structure for risk selection and efficiency or the opinion of medical experts.

Is there one measure-of-fit that fits all? A taxonomy and review of measures-of-fit for risk-equalization models

This study provides a taxonomy of measures-of-fit that have been used for evaluating risk-equalization models since 2000 and discusses important properties of these measures, including variations in analytic method. It is important to consider the properties of measures-of-fit and variations in analytic method, because they influence the outcomes of evaluations that eventually serve as a basis for policymaking. Analysis of 81 eligible studies resulted in the identification of 71 unique measures that were divided into 3 categories based on treatment of the prediction error: measured based on squared errors, untransformed errors, and absolute errors. We conclude that no single measure-of-fit is best across situations. The choice of a measure depends on preferences about the treatment of the prediction error and the analytic method. If the objective is measuring financial incentives for risk selection, the only adequate evaluation method is to assess the predictive performance for non-random groups.

Florida's medicaid reform: informed consumer choice?

Florida is among the first states to implement Medicaid reform using a competitive consumer choice model. Using data from a 2006-07 Kaiser Family Foundation survey of Medicaid recipients newly enrolled in Florida's reform program, we examine how well they understood the changes taking place and their experiences in selecting a health plan. We find important gaps in people's understanding of major components of the reform: About 30 percent were not aware that they were enrolled in reform, and more than half had trouble understanding plan information. These problems were not particular to any group but instead were experienced broadly across the full Medicaid population.

A sampling bias in identifying children in foster care using Medicaid data

BACKGROUND

Prior research identified foster care children using Medicaid eligibility codes specific to foster care, but it is unknown whether these codes capture all foster care children.

OBJECTIVES

To describe the sampling bias in relying on Medicaid eligibility codes to identify foster care children.

METHODS

Using foster care administrative files linked to Medicaid data, we describe the proportion of children whose Medicaid eligibility was correctly encoded as foster child during a 1-year follow-up period following a new episode of foster care. Sampling bias is described by comparing claims in mental health, emergency department (ED), and other ambulatory settings among correctly and incorrectly classified foster care children.

RESULTS

Twenty-eight percent of the 5683 sampled children were incorrectly classified in Medicaid eligibility files. In a multivariate logistic regression model, correct classification was associated with duration of foster care (>9 vs <2 months, odds ratio [OR] 7.67, 95% confidence interval [CI] 7.17-7.97), number of placements (>3 vs 1 placement, OR 4.20, 95% CI 3.14-5.64), and placement in a group home among adjudicated dependent children (OR 1.87, 95% CI 1.33-2.63). Compared with incorrectly classified children, correctly classified foster care children were 3 times more likely to use any services, 2 times more likely to visit the ED, 3 times more likely to make ambulatory visits, and 4 times more likely to use mental health care services (P < .001 for all comparisons).

CONCLUSIONS

Identifying children in foster care using Medicaid eligibility files is prone to sampling bias that over-represents children in foster care who use more services.

Medicaid and managed care: a lasting relationship?

The Medicaid program made a major commitment to managed care during the past decade. Following turbulent early years, the marriage matured and stabilized because managed care models responded well to a number of the states' goals and Medicaid purchasers were willing to make key trade-offs on behalf of their beneficiaries that conformed to the designs of managed care products. The relative tranquility in Medicaid managed care contrasts sharply with turmoil in both the commercial and Medicare sectors. But continuing changes in the managed care marketplace and financial distress in state budgets present new challenges to the strength and durability of this relationship.

Disease burden profiles: an emerging tool for managing managed care

As health plans assume financial risk for providing health care services, effectively managing the health of a population remains one of the toughest challenges. This article shows how risk assessment methods can be used to measure disease burden in the full population and to discriminate levels of future health care needs within specific disease cohorts. We also examine and compare the predictive power of claims-based models within a diabetic cohort.

Managed care for the Medicaid disabled: effect on utilization and costs

The objective of this study was to describe the effect on health care utilization and costs of a program of managed care for the Medicaid disabled. The study was designed as a pre/post enrollment cohort comparison and was carried out in three Ohio counties. The subjects were disabled Medicaid-insured patients who voluntarily enrolled in a managed care program for at least 6 months between July 1, 1995 and December 31, 1997, and who had (1) at least one Medicaid claim in the 24-months pre-enrollment period and (2) overall satisfactory postenrollment encounter-level data. Ohio Medicaid provided claims and reimbursements (costs) for the pre-enrollment period and encounter-level data for the postenrollment period. Postenrollment costs were estimated by applying category-specific average pre-enrollment costs to postenrollment utilization data. We measured the following per patient-month: (1) trends in category-specific utilization and costs for up to 24 months before and after enrollment, (2) differences in overall and category-specific costs 1 year before and after enrollment, and (3) changes in the distribution of services 1 year before and after enrollment. Utilization categories included inpatient care, outpatient hospital (including emergency department) care, physician services, prescription medications, durable medical equipment and supplies, and home health care. We found that satisfactory encounter data were available in two of three counties. Of 1,179 enrollees, 592 met all inclusion criteria. Before enrollment, utilization and costs were increasing significantly in four of six categories and were unchanging in two. Postenrollment, decreasing utilization was observed for three categories, one remained unchanged, and two were increasing, but from a lower "baseline." Except for physician services and home health care, there were lower utilization and estimated costs in all categories in the year after enrollment. Estimated inpatient and total costs declined by $155/patient-month (44.9%) and $210/patient-month (37.1%), respectively. Findings were similar across sites. Inpatient care, outpatient hospital care, and prescription medications accounted for 97% of the reductions in estimated costs in the postenrollment period. Among patients voluntarily enrolled for at least 6 months, managed care for the Medicaid disabled was associated with striking decreases in health care utilization and estimated costs. The effect of managed care on these patients' satisfaction, access to specialized services, quality of care, and health outcomes are understood incompletely.