|
MARKETWATCHHow Disclosing HMO Physician Incentives Affects Trust
Opinions are deeply divided over whether rewarding physicians for lowering costs decreases trust in physicians or insurers. To explore the effects of disclosing physician payment methods in HMOs, members of two similar HMO plans were randomized to intervention and control groups, and the experimental arm was told how the HMO paid their primary care physician. Separate disclosures were developed for each plan, one describing primarily capitation payment, and the other (mixed-incentive plan) describing fee-for-service payment with a bonus that rewards cost savings, satisfaction, and preventive services. The disclosures pointed out more of the positive than the negative features of these incentives. We found that the disclosures doubled the number of subjects with substantial knowledge of the physician incentives and halved the number with no knowledge. Nevertheless, the disclosures had no negative effects on trust of either physicians or insurers. The capitated plan disclosure had a small positive effect on trust of physicians. Disclosing the positive and negative features of incentives and increasing knowledge of these incentives does not, in the short term, reduce trust in physicians or insurers and may have a mild positive impact on physician trust, perhaps as a consequence of displaying candor and increasing understanding of positive features.
Using financial incentives to encourage physicians to reduce health care costs is the subject of increasing attention and controversy in medical ethics, law, and public policy.1 Various legal sources regulate health maintenance organizations’ (HMOs’) use and disclosure of physician incentives; financial incentives have been the subject of several pronouncements from professional and ethical authorities and are the center of major class-action lawsuits against leading U.S. insurers.2 Most recently, the U.S. Supreme Court ruled that the use of incentives does not violate ERISA (the Employee Retirement Income Security Act), but the Court reserved judgment on the legality of failing to disclose incentives.3 Throughout this debate, advocates have made critical assumptions about the relationship between financial incentives and patients’ ability to trust physicians and insurers. Opponents of incentives claim that the financial conflicts of interest created by cost containment incentives undermine trustworthiness.4 Proponents claim that some incentives are consistent with trusting relationships.5 There is general agreement that because the doctor-patient relationship is a "fiduciary" one grounded in trust, financial conflicts of interest must be disclosed, regardless of the impact on trust in physicians.6 Similar positions also exist with respect to trust in health insurers.7 To date there has been limited empirical investigation of these assumptions and assertions. Most were formulated before there was empirical evidence of any kind. Recently, data have been reported from two focus-group studies, two opinion surveys, and two studies comparing trust levels in different types of insurance plans.8 These studies indicate that giving physicians financial incentives to contain costs is potentially troubling to patients but does not necessarily undermine the basis for trust. However, a longitudinal examination of disclosure in a real-world setting is needed to determine the actual effects on trust. Prior studies may over- or underreport the impact of incentives on trust, either because subjects are not sufficiently informed of the actual incentives in use or because subjects are asked only hypothetical questions, which do not measure the real impact on their level of trust.9 This study advances understanding of physician incentives and patient trust using a stratified controlled intervention that assesses the impact on trust of disclosing to members of two different HMO plans how the HMO pays its primary care physicians.10 The study investigates the following questions: (1) Does disclosing primary care payment methods have positive or negative effects on patients’ trust in their physicians or HMO? (2) Does disclosure of capitation payment have a different effect on trust than disclosure of fee-for-service (FFS) payment with a mixed-incentive bonus? (3) Does disclosure of either payment method improve HMO members’ knowledge of physician incentives? (4) Does disclosure have any effects on trust apart from its impact on knowledge of incentives?
Study sample. A medium-size (300,000-member) HMO was recruited, with membership located across one southeastern state, consisting almost entirely of employer groups. This HMO sells two similar plans side by side. One has a closed-panel independent practice association (IPA) structure, and the other is an open-panel point-of-service (POS) plan. The two plans have nearly identical physician networks statewide, and there are no major differences among the types of purchasers of each plan or in how the plans are marketed. For our purposes, the two plans differ primarily in that the first one (the "capitated plan") pays its primary care physicians based on capitation plus a mixed-incentive bonus, while the other (the "mixed-incentive plan") pays its primary care physicians FFS according to a discounted fee schedule plus the same mixed-incentive bonus. The bonus is paid according to a formula that includes meeting budgetary goals, maintaining patient satisfaction, and meeting certain performance measures that focus on preventive care. We used the following stratified random sampling design. Equal numbers of members from each plan were randomly selected, to compare disclosure effects from two different types of incentives. Participants were randomly assigned to intervention and control groups with a ratio of 5:3. Eligibility was restricted to one adult member per household who had been with the plan at least two years and had seen a primary care physician at least twice. Because this insurer has a longer-term membership than national norms, and because length of time with an insurer and with a physician are associated with trust, we attempted to recruit equal numbers of members who had been with the insurer two to four years and more than four years.11 The initial sample consisted of 53.2 percent longer-term subscribers because of a relative shortage of shorter-term eligible members available in the capitated plan (41.9 percent). Telephone surveys were conducted in two batches, separated by a year, beginning in September 1999 and September 2000. Attempts to reach 4,024 subscriber households resulted in 3,844 contacts. Of these, 598 (15.6 percent) were ineligible and 852 (22.2 percent) declined to be interviewed, resulting in initial interviews with 2,394 subjects, or 62.3 percent of potentially eligible contacts. At each subsequent round of interviews, either additional subjects became ineligible or attempts to reinterview them were unsuccessful or incomplete, so our sample consisted of 1,918 subjects by the conclusion of the study, 59.1 percent of those initially eligible. Disclosure intervention. In consultation with an expert review panel and with the HMO, an intervention was devised in which members of each plan were told how the plan pays their primary care (gatekeeping) physician. The disclosure was made initially in a letter mailed from the HMO’s medical director and then followed by a telephone call from an independent market research firm. Two written disclosures were devised, one for each plan, employing the following strategies, derived from a literature review and advice from an expert review panel.12 (1) The phrasing was simplified to an eighth-grade level, avoiding technical jargon. (2) The disclosure addressed both the basic structure of the incentive (fixed versus variable payment) as well as the direction and possible effects of the incentive (more versus less care). (3) Both the base payment method (capitation and FFS) and the supplemental bonus were disclosed. (4) Although the bonus aspect of each plan was the same, the capitation plan disclosure gave less emphasis to the bonus than did the mixed-incentive (FFS) disclosure. This was done to differentiate between the mix of incentives in the two payment methods and to emphasize that the capitation plan more strongly rewards cost savings. Anticipating that few members would read carefully or remember the written disclosure sent by mail, and needing for purposes of this study to ensure that the information was conveyed to the intervention group, the written disclosure was followed by a telephone call approximately one week later in which the caller read verbatim to each subject the core part of the disclosure and then asked a set of simple questions to test for basic comprehension. The caller told subjects whether or not they answered correctly. For subjects who answered incorrectly or were uncertain, the caller repeated the relevant portion of the disclosure statement and then asked the same comprehension question again. Surveys and analysis. Four to six weeks before this intervention, both intervention and control groups were surveyed by telephone about trust in their primary care physician; trust in the HMO; knowledge of incentives; relationship, characteristics, and past experiences with their primary care physician and with the HMO; health status and medical history; and demographics. The same survey, with appropriate modifications, was administered a second time, one month after the disclosure. Trust questions came from two independently validated scales (trust of physician and insurer) developed earlier by our team.13 Knowledge of incentives was assessed at multiple points using the same questions administered during the disclosure. Statistical methods. The effects of the disclosure intervention on physician and insurer trust scores was estimated using a mixed (generalized least-squares random effects) modeling technique using the natural logarithm for all dependent variables. The regressions adjusted for baseline differences between the groups that were significantly associated with trust scores as well as differences in time of survey (year and round). Dummy variables were created for both intervention groups, and the omitted control groups served as the comparison. We also conducted three subanalyses in which these models were estimated in the following subgroups: (1) subjects whose knowledge of the incentive increased in the follow-up interview by at least one additional question answered correctly out of four; (2) subjects with less or more than the twenty-fifth and seventy-fifth percentiles on the baseline physician and insurer trust scales; and (3) subjects enrolled with their plan more than and less than four years.
The baseline characteristics of the population across the four study groups are outlined in Exhibit 1  . Increases in the mean unadjusted physician and insurer trust scores were seen across all four groups (Exhibit 2). The percentage of subjects with substantial knowledge of the physician incentive (defined as correct responses to at least three of four knowledge questions) also increased in all groups except the capitated controls, for whom substantial knowledge decreased somewhat. Knowledge increases were much greater in both intervention groups (nearly 100 percent) than in either control group.
After we adjusted for baseline group differences as well as time effects, the disclosure was associated with a 1.4 percent increase in average physician trust in the capitated plan intervention group compared to the controls (p <.05) (Exhibit 3 ). Effects of the disclosure intervention on physician trust scores in the mixed-incentive plan intervention group, and on insurer trust scores in both of the intervention groups, were not significant (p <.05).
When only the subgroup of patients with increased knowledge in the postintervention period was observed, there was a doubling of the disclosure intervention effect on physician trust in the capitated plan group (3 percent increase in physician trust, p <.01).
Stratifying by baseline trust scores showed no disclosure effects on physician trust among enrollees with low baseline physician trust (below twenty-fifth percentile scores), but for patients with high baseline physician trust scores (above seventy-fifth percentile scores), disclosure was associated with a 1.3 percent increase in physician trust scores in the intervention groups (Exhibit 4
There has been a great deal of controversy and speculation about the effect that financial conflicts of interest created by managed care have on patients’ ability to trust physicians and health plans. Many analysts believe that financial conflicts of interest undermine trust, others view them as necessary evils that make trust more tenuous, while still others maintain the possibility that incentives kept within proper bounds can be consistent with conditions of trust. From all sides of the debate, there is general agreement that such incentives should be disclosed in some fashion but considerable uncertainty about how effective such disclosures would be in conveying essential information and about how people will react to these disclosures. Much of this discussion is based on intuition and assertion. The little empirical information that exists is from questions posing hypothetical scenarios or from subjects who are largely uninformed about the issue.14 The findings reported here are based on a controlled study that observes the actual effects on trust of real-world incentive disclosures made to plan members who were subject to the incentives. Not only is this study’s description of incentives real rather than hypothetical, this study also measured the impact of disclosure on trust that was assessed independently of the disclosure, rather than simply asking subjects if their level of trust would change. However, our description of physician incentives emphasized more positive features (promoting health, eliminating unnecessary care) than did prior studies.
Knowledge of incentives.
Because our goal was to measure the effect of learning about incentives, we developed a disclosure method that went to unusual lengths to ensure that the essential information was conveyed. Information about incentives was disclosed by mail, followed by phone calls in which subjects’ understanding was tested and reinforced through repetition and simple quiz questions. This method was effective in greatly increasing knowledge of incentives (Exhibit 2
Contrary to the self-reported attitudes described above, we found that disclosing these incentives did not decrease trust in managed care physicians or health plans (Exhibit 3 Lack of negative effect. There are a variety of explanations for why disclosing these incentives did not lower trust, but none raises critical concerns about the robustness of this nonfinding. First, this health plan’s two physician payment methods contain components that lessen potential conflict of interest. Both methods included a bonus that was calculated, in part, on meeting preventive service goals and on patient satisfaction scores. However, the bonus also rewarded cost savings and was sizable—up to 25 percent, compared with the 10 percent levels reported elsewhere.15 Furthermore, these descriptions pointed out potentially negative consequences of these incentives, such as ordering fewer services. This goes beyond current practices and legal requirements for incentive disclosure, which merely describe the structure of the payment and not its behavioral or incentive consequences.16 Although many critics of these incentives would describe them in a more negative light than was done here, our descriptions were sufficiently balanced that in a separate regression analysis (not shown), higher knowledge of the incentives (apart from any disclosure) was associated with slightly lower physician trust in the study population as a whole (p < .05). Still, it is certainly possible that a much more negative description of incentives would produce a negative effect on trust. However, another plausible explanation is that once patients have formed opinions of their physicians based on actual experience with them, information about payment methods has little or no effect on their opinions, possibly because of the cognitive dissonance that otherwise would arise. This speculation is consistent with the finding by Audiey Kao and colleagues of no association between trust and patients’ perceptions of capitation versus FFS payment to their physicians.17 The lack of a negative effect also might be due to the short time interval (one month) between disclosure and the postintervention assessment of trust. The reality of the disclosure may not sink in until a patient’s next physician visit or encounter with the health plan. Conversely, the disclosure might have its greatest effect immediately and wear off over time. To explore these possibilities, we assessed trust in a subsample of the study population four months after the disclosure in the 1999 cohort (n = 985) and found no significant change in trust levels among the intervention groups, compared with controls. It is possible that making multiple disclosures might reinforce negative features of incentives, thereby eventually leading to a decline in trust, or that disclosures made by physicians at the point of treatment might have a much more negative impact on trust.18 However, consistent with current practice and with legal, ethical, and public policy discussions, we chose to assess the impact of only a one-time disclosure made by the health plan.19
Nonresults also could be attributable to insensitivity in our measures of trust. To explore this concern, we conducted a bivariate analysis (not shown) of the association between disclosure and subjects’ desire to change insurers or physicians. The disclosures produced a significant (p < .05) increase in the desire to change insurers, even though trust in insurers was not affected. However, disclosure made no difference in the desire to change physicians, and other studies indicate that trust and desire to change are well correlated for both physicians and insurers.20 Also, research published elsewhere shows that these trust measures are able to predict other relevant attitudes such as satisfaction, willingness to forgive mistakes, and willingness to recommend a physician or insurer to friends.21 Moreover, in this study these trust measures were related to other factors previously shown to be predictors of trust, such as self-reported prior disputes with the physician or insurer and adequate choice in selecting the physician or insurer (Exhibit 3
Finally, although most of the relevant characteristics of this population are similar to a random national sample on which these trust measures were validated, this population might be more trusting of managed care than is the case nationally or this health plan and its physicians may be more trustworthy.22 To explore these possibilities, we performed separate analyses in subpopulations of persons showing low trust levels and of shorter-term enrollees (two to four years) and found very similar (although nonsignificant) results for the intervention effects (Exhibit 4
Positive effects.
In sum, the lack of negative effect on trust appears to be a valid result. The positive effect on physician trust from disclosing the capitated incentive plan merits additional discussion. This positive effect may indicate that subjects chose to focus on the favorable attributes of the incentive, perhaps because of the cognitive dissonance that arises when one is told negative information about a physician that one trusts. This is suggested by the fact that the increase in correct responses to the comprehension questions asking about positive aspects was greater than to the questions asking about negative aspects (128 percent increase versus 92 percent). Also, the positive capitated effect doubles in magnitude among subjects who show some increase in incentive knowledge from the disclosure (Exhibit 3 The mere act of making a disclosure, regardless of any increased understanding, might display honesty and candor, which is a positive attribute of trust. This is consistent with the subgroup analyses just mentioned. However, this positive effect should have been revealed in trust in insurers, where it was not seen, rather than trust in physicians, where it was seen, since the disclosure clearly came only from the insurer. It is possible that a disclosure like this might be credited to both physician and insurer, but the nature of the two different types of trust, or our two different measures of trust, kept us from detecting an increase in trust in insurers. However, other factors that should perform similarly across the two types of trust, based on theory or on prior studies, did so.
Finally, the relatively small increase in one cell of the study might be a statistical artifact of the particular control variables we chose for this regression model. To the contrary, however, this result is robust across a variety of different regression models and analytical approaches that were used (some of which are not reported here). In particular, it is noteworthy that the magnitude of the positive capitated effect (~1.5 percent) is essentially the same in both the regression model (Exhibit 3
These findings have implications for several important legal, public policy, and operational issues. First, they show that the types of mixed-incentive payment methods disclosed here do not undermine trust and that their disclosure may actually increase trust. This suggests that reluctance to disclose physician incentives at the health plan level is not warranted based on a perceived threat to the doctor-patient relationship and that not all cost-minimizing physician incentives are ethically troubling to patients. Second, even the extensive and impractical methods used here to attempt to convey only limited knowledge of incentives fell well short of complete success. As a result, the aim of disclosure practices may need to be reconsidered for whether conveying full knowledge is the primary objective, or instead whether it is sufficient to provide members with the opportunity to learn about incentives if they so choose. These conclusions are tempered by several limitations of this study. The study population comes from a single insurer located in a southeastern state, only one type of disclosure method was tested, and subjects had little or no opportunity to discuss this information with their physicians. Although two different types of incentives were tested, their features and description were more positive than might be found or imagined elsewhere. Finally, we assessed trust effects only over a brief period of time, and we did not measure the impact of disclosure on trust-related behavior such as disenrollment, obtaining second opinions, questioning physicians’ recommendations, or appealing coverage decisions. For these reasons, the trust implications of physician payment methods merit continuing investigation.
Mark Hall is a professor of law and public health at the Wake Forest University School of Medicine in Winston-Salem, North Carolina. Elizabeth Dugan is a senior research scientist at the New England Research Institutes in Watertown, Massachusetts. Rajesh Balkrishnan is an assistant professor in the Department of Public Health Sciences, Wake Forest University School of Medicine. Donald Bradley is senior medical director, Quality Improvement and Medical Policy, at Blue Cross and Blue Shield of North Carolina in Durham. Hall occasionally consults with law firms that represent managed care organizations or physicians regarding litigation over financial incentives. This study was funded by the Robert Wood Johnson Foundation’s Initiative to Strengthen the Patient-Provider Relationship. Rajesh Balkrishnan conducted the statistical analyses. Survey Research Laboratory at the University of South Carolina conducted the surveys. Its director, Robert Oldendick, provided essential technical expertise. Analytical guidance and critical advice were provided by Mark Espeland and Sally Shumaker in the Department of Public Health Sciences at Wake Forest University. The authors are also grateful for the assistance and advice given by Beyaio Zheng, Douglas Levine, Aneil Mishra, Stephen Blackwelder, Wayne Steiner, Kristin Kidd, and Cynthia McQuellon.
This article has been cited by other articles:
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Top
Data And Methods
Study Results
