Health Affairs Blog: National vs. California Comparison: Detailed Data Help Explain The Risk Differences Which Drive Covered California’s Success

By Al Bingham, Michael Cohen and John Bertko

July 11, 2018

Since the beginning of enrollment under the Affordable Care Act (ACA) in 2014, California’s exchange, Covered California, and the state’s individual market have shown remarkable stability. In the exchange, in addition to significant enrollment, issuer participation has been robust, premium increases have consistently been well below national averages, and the risk mix of enrollees has been stable and fairly healthy. Notable aspects of this success include significant outreach and marketing combined with “active purchasing” strategies to foster a competitive environment. But, questions arise about which of Covered California’s activities, versus other characteristics of the California market, contribute to making California more successful than most other exchanges.

To identify meaningful differences in the much-better-than-average risk mix of the California exchange, Wakely Consulting Group evaluated average risk scores of enrollees in the California individual market over the past three years (2015, 2016 and 2017), both on- and off-exchange. We then compared California risk scores to other states’ and contrasted California’s market dynamics with the rest of the country, as represented by the 30-plus states in the Wakely database (consistently used in this post). Finally, we examined potential reasons for the difference between California and the rest of the country.

This post provides risk scores separated into on- and off-exchange categories – a level of detail not available in the results of the annual risk adjustment transfer notice that was published by the Centers for Medicare and Medicaid Services (CMS) on July 9. Accordingly, our analysis allows for a more nuanced understanding of the differences between California and other states.

Our Methods

We used data from the Wakely National Risk Adjustment Reporting Program(WNRAR). WNRAR provides participating insurers with interim information for the risk adjustment program under the ACA using the standard Department of Health and Human Services (HHS) hierarchical condition categories (HCC) risk adjustment methodology. This information is critical for health plan issuers for purposes of pricing health plans and estimating the ACA individual and small group risk adjustment transfer payments. The WNRAR project has reported estimated transfer payments and market averages to over 85 participating health insurance carriers for over 30 states since 2014.

Risk scores associated with the HHS risk adjustment model measure actuarial risk and are adjusted for metal level, enrollment in a Cost Sharing Reduction (CSR) variant, and an enrollee’s duration. To better measure health differences between California and other states, risk scores were adjusted to exclude certain non-health related factors (Note 1). To measure morbidity consistently over this period, we calculated average risk scores using the 2017 risk adjustment model for all three years.

California Versus Other States

As the largest state exchange, Covered California has had premium growth rates that have consistently been below national averages. For example, over the 4-year period 2015 to 2018, Covered California’s average premium increased only 7.2 percent after consumer shopping and any changes in metal tier enrollment, much lower than the national average. California’s on-exchange enrollment, as a percent of potential exchange enrollment, is one of the highest in the country. And the state-wide average risk scores, both on- and off-exchange, have consistently been in the lowest 10 percent of states each year (based on the authors’ calculations using CMS-reported data).

In examining combined on- and off-exchange individual market average risk scores, we find a marked difference between California and other states (Note 2). On average, California’s individual market risk score is about 20 percent lower than the other states’ average for each of the three years examined (2015, 2016, and 2017) (Exhibit 1). This is partly driven by the fact that fewer California individual market enrollees, relative to the other states, had a serious health condition, as measured by HCCs. As shown in Exhibit 1, 81 percent of California individual market enrollees did not have an HCC indicator in 2017, versus 78 percent in other states (a difference of nearly 15 percent fewer enrollees with an HCC). The difference in prevalence of HCC indicators may be because California exhibits a healthier population profile, or because California insurers enrolled a cohort of healthier people as a share of all potential enrollees (relative to other states), or some combination of the two.

Exhibit 1: Select Characteristics Of California’s And Other States’ Individual Markets (On- And Off- Exchange)

Source: Authors’ analysis of WNRAR data

These risk score relativities are not driven by differences in metal level enrollment, as California’s risk scores are lower than the national average even when controlling for metal level (Exhibit 2).

Exhibit 2: Individual Market (On- And Off-Exchange) Risk Scores Differences By Metal Level, California Versus Other States

Source: Authors’ analysis of WNRAR data

Another key difference between California and the rest of the country is the stability of the off-exchange market. As shown in Exhibit 3, while in the rest of the country off-exchange enrollment decreased nearly 30 percent from 2015 to 2017, in California the off-exchange market was essentially constant in the Wakely study companies (Note 3). This is especially significant as off-exchange enrollees are, on average, healthier, than on-exchange enrollees. From the Authors’ analysis of WNRAR data, off-exchange members in non-California Wakely study states have approximately 20 percent higher risk scores due to having more HCC conditions present. The average off-exchange risk score was 1.04 in California versus 1.26 in non-California states in 2015. The corresponding scores were 1.14 versus 1.37 in 2017.

Exhibit 3: Change In Off-Exchange Enrollment — 2015 To 2017 (Member Months), California Versus Other States

Source: Authors’ analysis of WNRAR data 

Health risk measures, not age differences, are driving the risk score differences between California and the rest of the country for the total (on- and off-exchange) individual market. Risk scores can be decomposed into two components – a portion that measures age/gender (demographic portion) and a portion that measures health risk (HCC portion). California’s demographic portion does not differ substantially from the rest of the country (i.e., the average of all three years is 0.309 compared to 0.313 for the rest of the nation) (Exhibit 4). However, the health risk portion shows a marked difference. The average HCC-only risk score over the three years was 0.824 for California, while the average for non-Californian states was 1.05.

Exhibit 4: California Versus Other States — Demographic And Health Differences

Source: Authors’ analysis of WNRAR data 

Key Drivers

Why is California different from the rest of the country? Policy decisions and an active purchaser strategy likely drive much of the difference. States that expanded Medicaid have markedly different risk scores than states that did not. These differences are directly attributable to difference in the percent of enrollees with serious conditions. Since the federally-collected data used in the Wakely analysis do not include income, it is not possible to definitively identify causal mechanisms related to Medicaid eligibility. However, there is a strong correlation between the policy decision not to expand Medicaid and the prevalence of sicker enrollees in the on- and off-exchange individual market. Many analysts have found that lower-income enrollees (i.e., those between 100 and 138 percent of the Federal Poverty Level that are served on-exchange when there is no Medicaid expansion) are costlier to insure. California’s choice to expand Medicaid may explain part of its exceptionally low risk score.

Another key policy-related difference is that state-run exchanges (also called “state-based” exchanges, or SBE) have lower risk scores on average than exchanges operated by HHS on states’ behalf (Exhibit 5). This is true holding constant enrollment by metal level. The primary difference in risk scores is driven by health differences rather than age differences. While there is overlap among states that expanded Medicaid and those that have SBEs, states that did not expand Medicaid all have federally-facilitated exchanges (FFEs). Interestingly, the subset of states with SBEs have, on average, even healthier exchange enrollees than the partially overlapping subset of those that expanded Medicaid.

There is some evidence that SBEs, on average, tended to have higher advertising and outreach budgets and other strategies, which may partly explain their greater enrollment of individuals without serious health conditions.

Exhibit 5: Exchange Risk Scores By Type Of State

Source: Authors’ analysis of WNRAR data 

As noted earlier, Covered California is an “active purchaser,” offering standard plan designs to make consumers’ choices easier to understand. There has been ample research that reduction of “hassle costs” and consumer confusion increases take-up. This could have benefits to the risk pool, as healthier enrollees may be strongly deterred by hassle costs. In addition, active purchasing has served to help control premium increases by increasing the level of competition for the best marketing positions (i.e., issuers which offer products that are the lowest and second-lowest Silver plans). The lower premium growth would increase take-up among the unsubsidized.

We believe Covered California’s active purchaser strategy and other policy decisions (e.g., Medicaid expansion and prohibiting transition policies for Covered California insurers), along with much higher marketing and outreach spending and efforts – which we have found to be associated with better risk scores – have contributed to Covered California’s success in stabilizing the individual market, on- and off-exchange.

A Few Limitations

Our analysis has some limitations, primarily resulting from the available data. First, the data cover only 31 to 33 states (varying by year) and approximately 50 percent of national member months for each of the respective years. While we believe the data are representative, further analysis is needed, and it is possible some of the missing data could influence the results.

The federally-collected data used for this analysis are plan-level, and enrollee specific data are not identifiable. As such, disentangling covariates such as income level and identifying causal mechanisms, such as how increased advertising directly impacted propensity to take up coverage, are not feasible. However, we believe that the differences in risk are both real and meaningful, and that others will be able to use this work as a starting point for additional causal analyses.

Another limitation may be that state health factors, such as the prevalence of obesity in the population, is influencing the results. It may be that lower-than-national average obesity rates correlate with policy choices such as Medicaid expansion or operating a state-based exchange. It is the authors’ observation that many of the FFE and non-Medicaid expansion states are those with the highest prevalence of morbid obesity. This population characteristic, if replicated in the exchange population, could partially explain the higher occurrence of many correlated HCC conditions and, thus, higher risk scores. However, further studies are needed to disentangle the independent effects of states’ overall population health, the morbidity of the individual market, and policy choices.

Finally, the individual market environment is rapidly changing. In 2018, changes to exchange marketing and outreach by the new administration coupled with a much shorter open enrollment may have affected enrollment patterns. While SBE enrollment was flat between 2017 and 2018, states relying on HHS’ Healthcare.gov experienced noticeable enrollment decreases. Differences between states could become even larger in the future. State variation in terms of rules governing short-term duration plans, association health plans, state based reinsurance programs, and individual mandates may also influence the findings in this post and differences between states in future years.

Given that the ACA was designed with the goal of state control, it is not surprising that state policy choices and state differences have played a large role in the success and failure of the exchanges. Our research shows that Covered California’s success in enrollment has been accompanied by real differences in health between itself and the rest of the country. These real and pronounced differences have helped California have lower premium increases compared to the rest of the country. While it is difficult to identify causal relationships in health policy, this research suggests that several policy choices California made could help explain its success.

Author’s Note

The WNRAR project data used for this study included only state level summaries, not individual issuer results. The state level summaries were blinded and aggregated to protect the confidentiality of the participants’ results by state. For more information please visit the Wakely Consulting Group website.

Acknowledgement

This analysis would not have been possible without the Wakely National Risk Adjustment Reporting (WNRAR) project. WNRAR provides interim information for the risk adjustment program under the Affordable Care Act. Health plans participating in the project receive estimated risk adjustment payables and receivables, estimates of health plan specific risk scores by state and market, and detailed reports that show the key drivers behind risk scores. Wakely performed all analyses of the WNRAR data for this study. The authors thank the Wakely WNRAR team, and especially Tyler Steiner, for providing the risk score data used to produce the averages. We also thank Michael Gillespie for his significant contribution to developing the underlying analyses.

Note 1

All references in this post will be to these “adjusted risk scores” – a term that will be used interchangeably with the term “risk scores.”

Note 2

“Other states” refers to the states in which all other participating issuers in the WNRAR study do exchange business. Overall the WNRAR study includes slightly more than 50 percent of all member months in the exchanges across more than 30 states for each year of the study.

Note 3

Due to potential measurement errors for off- versus on-exchange enrollment in 2016, we have excluded this year.