|
MARKETWATCHThe Relationship Between Health Plan Advertising And Market Incentives: Evidence Of Risk-Selective Behavior
Medicare beneficiaries are now facing advertising from an unprecedented number of health plans that are offering prescription drug coverage. Previous Medicare managed care efforts have been undermined by risk selection, the practice of enrolling healthier and therefore less costly patients. In this study we explore how the content of health plan advertising is related to the competitiveness of the health plan market. We find that increased competition is associated with greater use of advertising that targets healthier patients.
WITH PASSAGE OF THE Medicare Prescription Drug, Improvement, and Modernization Act (MMA) and its new prescription drug benefit, Medicare beneficiaries are being confronted by advertising from an unprecedented number of Medicare managed care plans, preferred provider organizations (PPOs), and prescription drug plans.1 Previous Medicare managed care efforts were undermined by risk selection or "cream skimming," in which health plans selectively enrolled healthier patients and avoided sicker ones.2 Most of the time, health plans are paid a fixed premium for each enrollee, regardless of the enrollee’s health status. In such a situation, patients with chronic illnesses are a potential financial liability, because they are more likely to incur high costs. Compared with reducing costs or improving quality, risk selection is a relatively easier mechanism for health plans to increase profits.3 A substantial body of research demonstrates that Medicare managed care enrollees are, on average, healthier than Medicare fee-for-service (FFS) enrollees.4 Provisions in MMA attempt to deter risk selection, but there is concern that these provisions are insufficient and that health plans will still have an incentive to engage in risk selection.5 Although the occurrence of risk selection is well documented, there is little research on its mechanism. Some believe that health plans actively recruit healthier patients through advertising, but this has never been empirically demonstrated.6 Others believe that health plans deliberately structure their benefits so that sicker patients voluntarily leave the plan.7 However, since risk selection is seen even at the time of initial enrollment, pre-enrollment factors such as advertising probably contribute to it a great deal.8 Only one study has examined the ads and recruitment techniques of health plans. Patricia Neuman and colleagues found that Medicare managed care plans targeted healthy seniors by including images of active seniors, using a small typeface, and recruiting at events that were accessible only to more-active seniors.9 This study was limited, however, by having no comparison groups of plans or markets. Our study is the first to examine risk selection in a national sample of health plan print ads and how risk selection has changed in those ads over time. We focused on the relationship between frequency of risk-selective ads and health maintenance organization (HMO) market share. HMO market share has often been used as a marker of competition.10 Higher HMO market share is associated with lower premiums and profit margins.11 Therefore, it likely increases the incentive to use risk selection. We therefore hypothesized that there would be an increasing frequency of risk-selective advertising in markets with greater HMO penetration at a point in time (2000) and nationally as HMOs became more prevalent over time (1970–1999).
We created a methodology for coding risk-selective characteristics in ads and used an initial sample of ads to generate a set of ten risk-selective characteristics. We then tested our hypotheses in two new samples of ads. Ad acquisition and inclusion/exclusion criteria. We obtained health plans’ print ads from two databases. For the cross-sectional analysis, we obtained them from VMS VoiceTrak (Los Angeles), a database containing all health plan ads appearing in local newspapers in the 137 largest U.S. markets in 2000. For the longitudinal analysis, we obtained ads from the John W. Hartman Center for Sales, Advertising, and Marketing History at Duke University. We included all plan ads in the archive that appeared during 1970–1999 in a major urban newspaper or national magazine. During this period, the health insurance market became more competitive, and our goal was to test whether there was an associated change in risk-selective characteristics. For both analyses we looked at all types of health plan ads, including Medicare, HMO, and PPO products. If risk selection is being used by one type of plan, other plans must respond to remain competitive.12 We sampled among ad occurrences, rather than among unique ads, because this approach better reflects a plan’s advertising strategy. An ad used many times would be expected to have more impact on enrollment than an ad used only once.13 We dropped ads from the analysis if the health plan was not the ad’s focus. Coding ads. We performed content analysis, a technique used in advertising research to describe and analyze in a systematic and quantitative way the written and visual characteristics appearing in a set of ads.14 Detailed coding instructions were created for each of the risk-selective characteristics, and coders received training with practice ads. All ads were analyzed by one of three coders, who were blinded to the hypotheses. Seventy-five ads were coded by more than one coder. Intercoder reliability was assessed using the kappa statistic; we excluded from our analysis any characteristics for which the percentage agreement was less than 90 percent or the kappa was less than 0.4.15 Preliminary list of risk-selective characteristics. Because our method was novel and there was no prior literature, we developed a preliminary list of twenty-four possible risk-selective characteristics. This list was generated based on discussion with experts in HMO risk selection, previous anecdotal reports from the literature, and our own hypotheses.16 For example, Neuman and colleagues noted that the seniors depicted in Medicare ads were often physically active, and few were in hospital beds or using wheelchairs. This was the impetus for three of our risk-selective characteristics: image of someone engaged in a physical activity ("attract healthy"), image of someone in a hospital ("attract sick"), and image of someone in a wheelchair (or using a cane) ("attract sick"—dropped because seen in less than 3 percent of ads). This preliminary list of characteristics was tested on a randomly selected sample of 300 ad occurrences from the cross-sectional data set. We tested the relationship between the mean number of these risk-selective characteristics in the ads and in HMO market share. HMO market share. We used the Inter-Study Competitive Edge database to categorize metropolitan statistical areas (MSAs) into three groups: low (<25 percent), moderate (25–50 percent), and high (>50 percent) HMO market share.17 Final risk-selective characteristics. From the list of twenty-four possible risk-selective characteristics, we dropped seven because they were found in less than 3 percent of ads, and four because of low intercoder reliability. We dropped three other characteristics because we discovered in the development sample that the relationship between HMO market share and the characteristic would be difficult to interpret, even if coded accurately. For instance, we hypothesized that mentioning specific chronic diseases would attract sicker enrollees, but the unblinded reviewers found that the fine print of Medicare HMO ads stated that patients with end-stage renal disease were not eligible. This made it impossible for us to use this characteristic. This process left us with ten risk-selective characteristics: (1) image of someone engaged in a physical activity; (2) image of someone at work; (3) ad discusses walking clubs, health club membership, stress management, health education, or wellness programs; (4) ad discusses nutrition, dietary consultation, or weight loss; (5) goal of ad is to promote an Internet-based service; (6) image of a patient talking to a physician; (7) ad discusses coverage for hospitalization; (8) image of a patient in a hospital; (9) ad discusses prescription drug coverage; and (10) ad discusses twenty-four-hour help line for health concerns. Of these ten characteristics, our a priori expectation was that the first five would be disproportionately attractive to healthy patients and the second five would be disproportionately attractive to chronically ill patients.18 Validating risk-selective characteristics. We validated these ten selected characteristics using two data sets: another random sample of 300 ads from the cross-sectional data set, and the entire longitudinal data set. For the cross-sectional validation, our method was identical to the development phase. For the longitudinal validation analysis, the ads were categorized into five-year intervals from 1970 to 1999. Statistical analysis. For each ad we created two scores by adding the number of risk-selective characteristics found in the ad. The "attract healthy" and "attract sick" scores could each range from 0 to 5. Scores between high and low HMO market share and between the earliest and latest time periods were compared using a Wilcoxon rank-sum test because of non-normality of the "attract healthy" and "attract sick" scores. Overall trend was tested using univariate regression, with the score as the outcome and the HMO market share or time period as predictor.
Development of risk-selective characteristics. We obtained 300 randomly selected ad occurrences from the cross-sectional database to develop the risk-selective characteristics. Of these, we dropped fifty-two because they did not meet our entry criteria; we coded the 248 remaining ad occurrences for the presence of risk-selective characteristics. Overall, one-quarter of the ads from markets with low HMO penetration had "attract healthy" characteristics, in contrast to the one-third of the ads from markets with high HMO penetration (Exhibit 1  ). The opposite trend was true for "attract sick" characteristics. The test of trend was significant for both "attract healthy" and "attract sick" characteristics.
Validation in cross-sectional analyses. Another 300 randomly selected ad occurrences were obtained from the cross-sectional database to validate the model. Of these, 236 met our eligibility criteria. Across markets, the use of "attract healthy" ads was more common in markets with high HMO penetration than in markets with low HMO penetration (Exhibit 2 ). There was no clear trend for the "attract sick" ads. The difference between the "attract healthy" scores in ads in markets with high and low HMO penetration was statistically significant, but the difference between the "attract sick" scores in the two types of markets was not.
Validation in longitudinal analyses. Among the 693 eligible ads in the longitudinal database, there was a trend toward increased use of "attract healthy" characteristics from 1970–74 to 1995–99 (Exhibit 3 ). The differences in the mean scores of ads from the two periods were significant, as were the tests for trend. In contrast, there was a low percentage of "attract sick" ads in 1970–74, followed by a rise in 1975–79 and then a uniform downward trend from 1975–79 through 1995–99. A similar trend was seen in the mean "attract sick" scores. The overall "attract sick" longitudinal trend was not significant, however, whether assessed by a comparison of 1995–99 to 1970–74 ads or by regression analysis.
We found that the use of ads that are attractive to healthy patients increased nationally from the 1970s through the 1990s as HMOs became more common and gained market share. Furthermore, in 2000, the use of such ads was more common in markets with higher HMO market share than in those with lower market share. There was no consistent relationship between "attract sick" ads and HMO market share in either the longitudinal or the cross-sectional analysis. These correlations suggest that as competition increases, health plans attempt to risk-select through advertising. Two random samples of plan ads from 2000 show that in markets with high HMO market share, the mean number of "attract healthy" characteristics in ads was 0.6 or higher, while in markets with low HMO market share, it was below 0.3. Study limitations. Print ads only. There are several important limitations to our study. Our analyses were limited to print ads, which are only one part of health plans’ marketing strategies. It is possible that radio ads, for instance, would show an opposite, mitigating trend. However, there is no database from which to obtain such ads (nor any a priori reason to expect to find this opposite trend). Validity of risk-selective characteristics. Although our risk-selective characteristics have face validity, future research will ideally look at whether these ten characteristics are truly attractive to different types of patient populations. Such research might include obtaining the subjective impressions of people with chronic diseases about whether an ad suggests to them that the health plan is suited for them. In this study we were limited to risk-selective characteristics for which there was sufficient intercoder reliability, and therefore more subtle and likely more prevalent risk-selective techniques could not be captured. Intent of ads. Because we looked at correlations, we could not address causality and whether the differences in ads’ content were deliberate attempts by health plans to risk-select. For example, the longitudinal changes might instead reflect secular trends and evolving consumer interests. However, the percentage of ads that mention prescription drug coverage—an area of rising consumer interest over the past three decades—declined from 1975 through 1999 (data not shown). If the changes in content of health plan ads primarily reflect trends in consumer interest, then the mention of drug coverage should have increased. In addition, we found a relationship between HMO market share and risk-selective advertising in our cross-sectional analyses, which suggests that at least part of the phenomenon reflects factors other than temporal trends. Failure of "attract sick" validation. It is not clear why the association between "attract sick" characteristics and HMO market share was not confirmed in the validation analyses. Although this could simply be a reflection of our sample size (since there was a trend in the longitudinal analysis), plans might believe that ads targeting healthier patients are sufficient to signal to the chronically ill that they should choose another plan. Alternatively, the structure of benefits, such as preauthorization for specialist referrals, might encourage sicker patients not to enroll or to disenroll. Policy implications. Because this is the first study of its kind, these results should be confirmed in further analyses. If confirmed, our results have important policy implications. Based on VoiceTrak’s estimates on cost for print space, health plans spent more than $70 million on newspaper advertising alone in 2000. The total advertising budget is much higher if all media types are included. Our findings imply that health plans are using that advertising to attract healthier patients. From a societal and clinical perspective, these resources are being misused. More importantly, if plans can be profitable through risk selection, they might not undertake the harder work of becoming profitable by improving quality and efficiency. MMA attempts to curb risk selection by paying health plans more for sicker patients and less for healthier patients.19 Whether or not the adjustment of payments, especially for the sickest patients, is sufficient remains to be seen. The advantages of risk selection might be stronger for stand-alone prescription drug benefit plans than for Medicare Advantage (MA) plans.20 Further research is needed to determine whether or not this advantage leads to differential response by plans in marketing for these two products. Our results suggest that analyzing the content of health plan ads could be a novel new method for assessing the impact of MMA’s adjusted payments and other new policies. A major advantage of studying changes in advertising is that plans can change advertising immediately and the data can be collected quickly—well before data on the utilization of enrollees can be acquired. In addition, health plans have full control over advertising, while outcomes or utilization could be subject to many influences beyond a plan’s control. WE FOUND preliminary evidence that health plans change the content of their advertising in response to market incentives. As initiatives such as Medicare’s prescription drug benefit create new situations in which beneficiaries have choices among health plans, the opportunities to risk-select through advertising increase, as does the need to study ads and consider policy options about advertising.
Ateev Mehrotra is a fellow in general internal medicine in the Department of Health Care Policy, Harvard Medical School, and in the Division of General Internal Medicine, Brigham and Women’s Hospital, both in Boston, Massachusetts. Sonya Grier is a Robert Wood Johnson Health and Society Scholar at the University of Pennsylvania in Philadelphia. Adams Dudley (adudley{at}itsa.ucsf.edu) is an associate professor of medicine and health policy in the Division of Pulmonary and Critical Care and the Institute for Health Policy Studies, University of California, San Francisco. The authors thank the ad coders for their time and effort. The primary funding source for this research was the Agency for Healthcare Research and Quality (Grant no. P01 HS10771-01). Ateev Mehrotra was supported by an institutional National Research Service Award (Grant no. 5 T32 HP11001-15).
This article has been cited by other articles:
| |||||||||||||||||||||||||||||||||||||||||||||||||||
Top
Study Methods
Study Results
