Chapter IV. Data to Support Work on Disparities (continued)
Evaluation of a Learning Collaborative's Process and Effectiveness to Reduce Health Care Disparities Among Minority Populations
D. Firm Efforts to Collect Race and Ethnicity Data
1. Direct Support from the Collaborative
With a few exceptions, the Collaborative did not directly support firms' interest in developing their own racial/ethnic data. The exceptions involved the Collaborative's support to help firms learn more about Aetna's Web Portal Approach and HealthPartners' activities in collecting information in their clinics. Almost universally, firms expressed respect for Aetna's initiative and its important influence on the industry. (HealthPartners' reliance on the affiliated medical staff meant fewer firms appear to have viewed it as applicable in their setting.) The Collaborative's response to firm requests to learn about the Web Portal took time, but at the June 30, 2005, meeting of firms' senior leaders, Aetna agreed to provide Collaborative participants with a demonstration of the portal; the Aetna demonstration (as well as the HealthPartners demonstration) took place via a Web call on October 20, 2005.
It is unclear why the Collaborative support organizations did not do more to address the interest among some firms in primary data collection on race and ethnicity. The support organizations' relative lack of emphasis on data collection stands in strong contrast to RAND's active interest in geocoding and surname analysis and CHCS/IHI's focus on pilot interventions. Of course, participating firms have their own IT platforms and unique concerns and may not have needed or wanted the support organizations to advise them on internal systems. However, the support organizations could have helped the Collaborative respond to some firms' expressed interest in help that might enhance support for external policy change (perhaps in the form of uniform standards and procedures) to face the internal challenges inherent in race/ethnicity data collection. For example, several firms mentioned that establishing a "safe harbor" could enable them to overcome any perceived legal barriers to collecting race or ethnic data. Another firm voiced strong interest in having the Collaborative build a broader coalition to promote standardized national approaches for such data collection that would help firms get this information via existing administrative systems. While support organizations did not directly work with firms to address these issues, at least one sponsor independently supported work directly relevant to firm concerns.13
As we discuss further in Chapter VIII, the second phase of the Collaborative will focus more heavily on supporting firms in their efforts to collect primary data on race/ethnicity. Firms have decided not to attempt to directly influence federal policy but instead, using their collective power via the Collaborative, to help shape the form in which race/ethnicity data are collected and to better support firms that seek to collect such data so that their efforts are synergistic.
While the Collaborative may not have directly pushed firms to improve their ability to collect primary race/ethnicity data, many participating firms expanded their data collection and plan to continue to do so. Firm involvement in the Collaborative appears to be at least one of many factors supporting such expansion decisions. At the Collaborative's outset, only a few firms had a policy of collecting data from all members; all but one now have such policies (Table IV.2). The exception is a firm that collects data selectively on race/ethnicity and is not yet convinced that its needs are sufficient to justify more universal collection, considering the challenges inherent in introducing such a change across the organization and its diverse IT platforms.
Given the scale and complexity of firms participating in the Collaborative, firms' success in fulfilling their goals for data collection remains to be seen. None expects to succeed in the near term, as racial/ethnicity data collection is a multiyear effort that requires decisions involving trade-offs and calculated risks. Firms such as Aetna that instituted member Web portals found that although the majority of portal users reported information, building up data on a large share of the membership posed a challenge. Further, firms face trade-offs in determining which racial/ethnic codes and categories to use. More extensive codes may be more accurate, but can also prove burdensome to both staff and members.
Provider-based firms appear to have an advantage in collecting racial/ethnic data because they are more closely linked with providers and thus have enhanced ability to capture race/ethnicity at the point of service. For example, the medical group affiliated with one firm collects information on language preference at point of scheduling so that it can arrange for an interpreter if needed; race/ethnicity, which is perceived as more sensitive, is collected during the visit. Staff are trained in and provided with scripts for use in eliciting racial/ethnic information as well as with materials for distribution to patients who want to know more about why the information is requested. Most health plans, however, are not provider-based systems. The other strategy that at least one firm has pursued involves capturing data on its own employees because this is something they can control and implement.
One encouraging note is that firms starting to collect data have reported less member opposition than they feared—perhaps in part because all reporting is voluntary. It may also be that efforts are "under the radar screen." More broad based efforts might generate more concerns by advocates. On the other hand, some firms noted that high levels of distrust exist in some communities and for some subgroups. One said that, in their view, provider support may be essential to capturing such data because enrollees would ask their providers about responding to any firm request. At least one other firm also noted the importance of having engaged minority community leaders in discussions about primary race and ethnicity data collection at the outset of its efforts. These comments raise the possibility that some firm strategies may be more effective if they can engage other stakeholders in their pursuit.
E. HEDIS Data Submission on Common Measures
Sponsors originally designed the Collaborative with the expectation that firms would submit and share with other participants the HEDIS indicators for their diabetic population at a minimum of two time points (with an intervention occurring between the two). However, from the start, firms expressed reservations about this. When the Collaborative did make a formal but limited request, only a few firms responded
The record is somewhat ambiguous regarding how firms viewed the agreements about sharing common measures at the beginning of the Collaborative. From the start, AHRQ pushed for a consensus document outlining expectations. An initial draft proposed at the July 2003 meeting ushered in the formation of the Collaborative (Chapter II). The draft was revised with feedback from the participating firms and reissued at the July 2004 meeting as a Memorandum of Participation. In it, firms agreed to commit to three meetings, to work with RAND to improve the race/ethnicity data available to them, and to use accepted measures (starting with HEDIS indicators for diabetes) to support their work. However, the agreement also stated that firms would have substantial flexibility in responding to market conditions and noted that shared data, pilot designs, and results would be balanced against requirements for maintaining privacy, confidentiality, and proprietary interests.
The issue of common reporting of HEDIS data did not arise again until mid 2005 at a leadership meeting of the Collaborative. While we did not attend the meeting, we understand that it was convened in response to some participants' concerns that the emphasis on geocoding and surname analysis was detracting from efforts that might lead to tangible and "scalable" pilot interventions to member plans. AHRQ staff told us, for example, that the support organizations felt that as long as the Collaborative kept analyzing data rather than intervening on disparities known to exist, there would be no actual impact on disparities. According to the minutes, participants agreed:
The Collaborative needs to produce tangible and "scalable" benefits to its member plans for it to warrant their continued participation. In order to achieve such benefits however, the members concurred that they each had to make meaningful progress, contribute findings/models to the common good, and report to the Learning Organization fully and on a timely basis.
The meeting reportedly led to a re-energized Collaborative with more complete reporting and sharing among member firms and an agenda that shifted (at the sponsors' request) from geocoding and surname analysis to more active efforts at intervention.
We are not aware of the reasons for the delay, but CHCS/IHI did not formally request participating firms to submit HEDIS measures until November 3, 2005. The request was for firms to submit three or four specific HEDIS measures on diabetes for their full plan membership stratified by commercial, Medicare, and Medicaid lines of business. The request proposed four racial/ethnic categories consistent with those used in the geocoding and surname exercise. It also proposed submitting HEDIS 2005 (CY2004) data on December 15, 2005, followed by HEDIS 2006 (CY 2005) data on October 15, 2006, and potentially HEDIS 2007 (CY 2006) data a year later. The memorandum proposed a new Disparity Index and a Quality Index to track performance over time.
As observers on the call when the request was under discussion, we heard firms ask about the scope of the request; staff could not provide answers immediately. Recognizing staff difficulties in responding to questions about scope, CHCS surveyed firms in the Collaborative to learn more about their concerns. In an April 2006 document, CHCS stated that all but two firms responded, and only one declined to provide the suggested measures (a decision that was modified after further discussion). However, by April 2006, only one firm had responded with data, with another said to have committed to provide data by the June 2006 meeting.
In round three, we had the opportunity to ask firms about the reasons for their (lack of) response to the request for HEDIS indicators. Most responses to this question were relatively brief. In many cases, it appeared that firms did not spend considerable time debating whether to respond to the request. The decision not to provide data did not appear to be central to the firms. Often they did not give it much attention, either because they felt they had already expressed their lack of interest in providing these data or because they examined the request and felt it would be too time- and resource-intensive. Firm feedback indicates that our earlier interpretations regarding firms' lack of support for collecting common measures may be on target.
First, firms are not necessarily able to report firm-wide HEDIS data by (proxy) race and ethnicity, at least universally. If firms capture such data, they typically do so for a subset of their population, such as those in particular disease management programs. To generate the measures requested by CHCS, many firms need IT support while all face competing obligations not under the control of firm staff involved in the Collaborative. The demands on resources associated with collecting common measures were cited as a large reason for the lack of support of these measures by at least four firms. We believe that the request for HEDIS data submission was later modified to give firms flexibility in defining their population of interest. Nonetheless, many firms were legitimately concerned about the potential utility of the indicators and remained sensitive about use of the indicators to compare firms or judge performance trends in the absence of their direct intervention. Our interviews clearly show that at least some firms remain relatively sensitive about the release of information on any aspect of their performance, particularly as related to measures distinguishing racial and ethnic subgroups. A couple of respondents also commented that too great a focus on comparative measures would have limited the opportunity for firms to be truly collaborative and to focus on understanding why disparities exist (rather than how much and at which firm they exist the most). However one participant from a support organization told us that, from personal experience, sharing data is the first step to true collaboration because it peaks the curiosity of participants in understanding what the data mean.
Second, the support that sponsors perceived from firms for publicly reporting HEDIS measures was based on tacit commitments made in Collaborative meetings, which, in hindsight, proved to be overstated. Many lead contacts viewed the request for HEDIS data as another of the many requests they got from sponsor and support organizations. Interviewees told us that they reside in complex organizations and should not be expected to make binding commitments for their firms in such meetings. One interviewee observed, for example, that even a CEO's commitment could be difficult to obtain on the spot since other staff in the organization will be operationalizing the policy response and need to be consulted. IT constraints, in particular, can be problematic, a fact that firms confirmed on the April 2005 call of the full Collaborative.
Third, the sponsors' original goal to generate HEDIS data over time for use in assessing improvements gained through interventions was unrealistic given the Collaborative's two-year time frame, the complexity of the participating firms, and their diverse interests. One firm had significant experience in public reporting on common measures with other firms in their own market, and noted that even that effort—which involved like firms in a common environment—was very difficult and took a long time to develop. A respondent from this firm said collecting common measures was a laudable goal but not a practical one for a group as diverse as the Collaborative, and within such a limited time frame. Furthermore, regardless of whether firms could submit HEDIS indicators for 2005 and 2006, they clearly did not conduct the types of interventions that would likely drive noticeable change in national indicators over only two years.
Fourth, despite the efforts CHCS indicated it made to involve some firm experts in formulating the solicitation on common measures, the solicitation could have more fully addressed the firms' likely questions and taken into account the diversity across firms and the particular challenges of firms without single integrated systems across locales and product lines. The memorandum requesting common measures did not indicate which product lines should be considered—HEDIS data are more available for managed care, especially HMOs, than for other products—and did not appear to anticipate the operational issues firms might face, especially if a company divided authority by payer (the Collaborative focused mostly on the commercial sector), used several IT platforms, or did not focus its geocoding work on its entire population as opposed to those in a particular state or region. CHCS's inability to address these concerns in advance or respond in the meeting may have inadvertently reinforced firms' existing reluctance to share data.
As researchers, we understand firms' concerns about interpreting the data they provide, although the sharing of data among participating firms might have helped strengthen the group process and reinforce the importance of addressing disparities. Firm data would likely have covered diverse populations—for example, those in disease management programs or all members—and time frames. While two years of data might have allowed firms to serve as their own control and show trends, the interpretation of such trends could be problematic in the absence of an intervention or in the case of one whose focus did not match the data. Hence, shared data seem useful in pointing out the existence (or lack) of disparities but not in comparing disparities across firms or identifying how effectively firms have addressed them.
The lesson we draw from the Collaborative's experience is that if a sponsor wants to convince firms reluctant to share data to do so, it must make a more compelling case for the value of shared data. It must also be certain that the request will be supported with assistance and is as operationally feasible as possible. Finally, the sponsor needs to be prepared to respond to the types of concerns firms may raise and make sure that firms express them openly in meetings and calls rather than ignoring the request.