On March 31, 2016, the Centers for Medicare and Medicaid Services (CMS) will hold a day-long public meeting in Baltimore to discuss potential changes to its Affordable Care Act (ACA) risk-adjustment program. On March 24, CMS published a White Paper describing its current risk adjustment program and options that it is considering for modifying that program.
The Affordable Care Act included three premium stabilization programs that went into effect in 2014 together with its market reforms — the reinsurance, risk corridor, and risk adjustment programs. These “3Rs” work in tandem but have different purposes.
Background
The reinsurance program collects contributions from all insurers and third party administrators that offer health coverage in all markets during the first three years of the market reforms to reinsure individual plans—inside and outside of the marketplaces—for high-cost claims. It is based on the assumption that the individual market will disproportionately serve high-cost enrollees immediately following the banning of health status underwriting and pre-existing condition exclusions and that individual market insurers will need some help until the market stabilizes.
The risk corridor program is also in effect only for 2014, 2015, and 2016. It applies only to qualified health plans; it is based on the assumptions that insurers will find pricing these plans difficult during the first three years and that plans which take large losses should be compensated while those that make windfall profits can help out. A decision by Congress to limit appropriations for this program to the fees it collects from insurers caused a dramatic shortfall for 2014 (and likely for 2015 as well), dramatically limiting its effectiveness.
The third program, risk adjustment, covers all non-grandfathered comprehensive insurance plans in the individual and small group market. It is, unlike the other two “Rs,” a permanent program. It is intended to provide support for the ACA’s market and rating reforms.
Although the ACA outlaws risk selection, it is not that difficult for health insurers to avoid high-risk enrollees and attract low-risk enrollees through benefit design, networks and formularies, or marketing practices. The risk adjustment program is intended to transfer revenues from insurers that in fact cover low-risk enrollees to those that cover high-risk enrollees, providing a further disincentive for risk selection.
The problem is designing a program to achieve this. Risk adjustment programs have long been used in Medicare and Medicaid and in European countries with social insurance programs. The ACA risk adjustment model builds on the Medicare model, adapting it for the under-age-65 commercial market.
How The ACA Risk Adjustment Approach Works
The ACA risk adjustment model is described in the first three and in the final chapter of the White Paper. In fact there are separate models for infants, children, and adults. Risk scores are calculated for each enrollee of a plan who has been enrolled for at least one month in a fee-for-service health plan without any payments made on a capitated basis, and who has prescription drug and integrated mental health coverage.
The risk adjustment model is a concurrent model, which is to say it is based on information reported regarding actual plan enrollees during a program year rather than predicting experience from a prior year’s experience. Information is collected on enrollees through an external data gathering environment (EDGE) server. The Department of Health and Human Services (HHS) does not itself collect or maintain data on individual enrollees, but rather calculates risk scores using data in the possession of the insurers
Enrollees are given risk scores based on their age and gender, and on their HHS hierarchical condition categories (HCC) classification. HCCs are in turn based on ICD-10 codes. The CMS payment model currently uses 127 HCCs which reflect 3,439 diagnosis codes. A risk score is intended to predict plan expenditures based on the HCCs, age and sex categories, and, where applicable, disease interactions.
Dollar coefficients are estimated for these factors. These are calculated based on cost data drawn from the Truven MarketScan(r) Commercial Claims database, with costs trended forward to the current year. Plan liability expenditures, or risk scores, are calculated for each enrollee at each metal level using standardized benefit design parameters. Risk scores are adjusted for cost-sharing reduction payment eligibility, recognizing that increased actuarial value and reduced cost sharing is likely to result in increased utilization of health services. Expenditures are annualized for enrollees with less than 12 months experience.
Finally, the models calculate a relative plan liability risk score. This is based on the average of the risk scores of all individuals in the plan divided by the average of the mean plan liability at each metal level for all plans and issuers in a state, weighted for the distribution of enrollees across metal levels.
The relative plan liability risk score is a key component of the risk transfer formula. The risk transfer formula attempts to measure the difference between the revenues required by a plan given the health status expenditures of its actual enrollees and the revenues the plan can generate based on the allowable rating factors of its enrollees.
Required revenues are calculated as a product of the relative plan liability risk scores, an induced demand factor based on the plan’s metal level, and a geographic cost factor reflecting utilization and expenditure patterns in the geographic location of the plan’s enrollees. Revenues that a plan can generate are the product of the plan’s actuarial value, premiums it can charge given allowable age rating factors, induced demand associated with the plan’s metal level (to insure that plans are not paid simply for plan variations), and the geographic cost factor corresponding to the location of the plan’s enrollees.
The plan-required-revenue term is divided by the statewide average required revenue while the plan allowable-premium term is divided by the statewide average allowable premium. The difference between the two terms is multiplied times a statewide enrollment-weighted market average plan premium to determine the amount that insurers either pay into or collect from the program. The sum of all transfers must be budget neutral.
Changes to the risk adjustment formula that HHS made for 2015 and 2016, other than updates of program parameters, were mainly technical. For 2017, the risk adjustment model will include preventive services, which should increase the risk scores for plans with healthier enrollees who use preventive services. Recent regulatory developments have otherwise focused primarily on validation of risk adjustment reporting. For 2014, about 10 percent of premiums in the individual market and 6 percent in the small group market were transferred among insurers through the risk adjustment program.
Criticisms And Proposals For Change
The risk adjustment formula has come in for criticism. Some critics have charged that the formula disadvantages smaller and newer plans and plans with lower premiums. Others criticize the use of MarketScanâ data for estimating costs as not appropriate given the difference between the commercial population those data represent and marketplace enrollees. They also criticize the diagnoses included in the model and the failure of the model to consider prescription drug use, which might make it more accurate.
Partial Year Enrollments
The White Paper examines proposals for improving both the risk adjustment model and the risk adjustment transfer formula. One change it considers is taking into account partial year enrollments. Enrollees experiencing acute episodes may incur most of their expenses over a short period of time. If they enroll during a special enrollment period, have a high-cost episode of treatment, and then disenroll, their costs may not be reflected, or adequately reflected, in an HCC.
CMS analysis found that actuarial risk for adults with short enrollment periods, and in particular very short enrollment periods, was underpredicted by the risk adjustment model. The problem is how to address this. If partial year enrollment is simply added as another factor to the model, its impact is small. If separate models are developed based on specific enrollment periods (1-4 months, 5-8 months, and 9-12 months) prediction accuracy is increased, but so is complexity, and increased prediction accuracy may be illusory given small sample size. The issue of alleged “hit and run” enrollment is being aggressively pressed by insurers as they lobby to cut back on special enrollment periods, but since the risk adjustment program is a zero-sum game, adjusting risk adjustment is probably less attractive to insurers as a solution.
Adding Prescription Drug Utilization
The bulk of the chapter on risk adjustment changes is taken up by proposals regarding adding prescription drug utilization as a factor to the risk adjustment model. Adding drug utilization information could improve identification of health care costs in situations where individuals in fact have a chronic disease and are taking medication for it, but do not have a diagnosis code in their claims for a year because their provider failed to note it or for some reason (such as the stigma associated with the diagnosis) decided not to. The use of prescription drugs may also be a marker of the severity of a condition identified through a diagnosis. Drug prescribing data may be more timely and standardized than diagnosis data. Moreover, taking prescription drugs into account may reduce financial disincentives for prescribing expensive drugs.
Incorporating prescribing data also raises concerns, including the creation of perverse incentives for gaming or unnecessary prescribing. Many factors affect drug prescribing, including cost-sharing and drug utilization programs. Adding prescription drug data to the formula would increase administrative burden and complexity. Moreover, there are multiple indications for most drugs, thus it is often difficult to associate a particular drug with a particular condition.
The White Paper discusses criteria for incorporating prescribing into the risk adjustment model and a detailed framework for doing so. It examines the possibilities of designing and classifying prescription drug categories (RXCs) and then identifying drug-diagnosis pairs for incorporation into a hybrid model that also includes HCCs. It discusses alternative models in which the prescription of a drug would be used to impute an otherwise missing diagnosis or to indicate increased severity of conditions otherwise identified through diagnosis codes. The discussion concludes:
Based on the research performed so far, we believe that a hybrid model that includes prescription drug data in the HHS-HCC risk adjustment framework deserves consideration. This revision would need to be carefully designed and implemented, with an understanding that the potential gains in predictive power, accuracy, and fairness will come with costs of increased potential for gaming, incentives for greater prescription drug utilization, sensitivity of risk adjustment to variations in drug utilization unrelated to enrollee health status, and added administrative burden.
Separating Out High-Cost Cases, Moving To A Prospective Model, And Other Potential Changes
Third, the White Paper considers the creation of a supplemental risk adjustment model that would separately consider the costs of high-cost cases. High-cost cases above a threshold amount (say, $1 million) would be excluded from the current transfer model which would be applied without them. An additional transfer amount would be calculated to cover these high-cost cases. The paper gives an example of how this would work.
The paper next examines moving from a concurrent to a prospective model now that data are available as to the health status of marketplace enrollees. The paper decisively rejects this proposal. Finally, the paper considers moving away from its current distributed external data gathering environment (EDGE server) approach to actually obtaining access to deidentified enrollee data itself, and using these data to improve the risk adjustment program and increase transparency. CMS might be able to use these data, for example, to calibrate the risk adjustment model to account for socioeconomic status. CMS seeks comments on this approach.
The White Paper finally examines several proposals for changing the risk adjustment transfer formula. One would eliminate administrative costs from the statewide average premium used for the denominator. Critics argue that the current formula disadvantages plans with lower premiums and this change would help, but CMS is skeptical.
CMS also welcomes comments on other changes that would improve the accuracy of the transfer formula. It concludes, however, that for 2014,
- both the risk adjustment model and risk adjustment transfers were highly related to an issuer’s relative amount of paid claims,
- issuer size did not predetermine if they received a risk adjustment payment or charge,
- and although differences in insurers’ mix of enrollees across metal levels were related to the direction and magnitude of transfers, they were not the sole determining factor.
The effect of insurer size on transfers has been a particular concern of critics. CMS found that transfers varied much more for small insurers than for large, but at median small insurers received net payments in the individual market but paid charges in the small group market. As to metal level, insurers tended to receive transfers for gold and platinum plans and pay charges for bronze plans.
The paper will be discussed at the March 31 conference, which will also be available by webinar. It promises to be a fascinating seven and half hours for ACA policy wonks.
from Health Affairs Blog http://ift.tt/1qf2Y5e
No comments:
Post a Comment