Utilization of predictive genetic testing remains an important issue for the translation of advances in genomics into improvements in disease prevention and chronic disease management. Concern exists about inequities in utilization across racial or socioeconomic groups as differential uptake of genetic testing may exacerbate existing health and health care disparities. Despite this concern, relatively little is known about what might be done to reduce disparities in utilization of genetic testing. In this study, self-reported likelihood of undergoing a hypothetical genetic test for cancer risk was lower in scenarios when the results are disclosed to the participant’s health insurance company, when the test must be obtained from a specialist and when the test marketing is specific to the participant’s racial group. However, the effect of these first two characteristics varied between people with different levels of health care system distrust, with individuals with high values distrust being particularly sensitive to the disclosure of results to insurance companies and individuals with high competence distrust being less sensitive to the need to obtain the test from the specialist. Although health care system distrust was slightly higher among African Americans than Whites (particularly values distrust), there were no differences in the effect of these characteristics between African Americans and Whites.
As hypothesized, certain characteristic of the delivery system reduce likelihood of undergoing genetic testing. Having to see a specialist to get the test increases the difficulty of obtaining the test and may raise concern about the cost of access. Disclosure of results to insurance companies is likely to raise concerns about insurance discrimination, which has been previously shown to be an important determinant of willingness to get tested.52
Race specific marketing also reduced likelihood of testing, an effect that was anticipated among African-Americans but was unexpected among Caucasians.
Targeting of genetic testing to any racial or ethnic group may create concerns about the use of the test results for racial or ethnic discrimination. Although the discussion of these issues has generally focused on groups with a history of discrimination (e.g. African Americans, Ashkenazi Jews), these results suggest racial targeting concerns may exist even among majority groups and that presenting tests as specific to one group may paradoxically reduce uptake across all groups.33,53
Consistent with previous work, we found a small, marginally significant difference in willingness to undergo testing between African Americans and Whites in unadjusted analysis.21
However, this difference was not significant after adjustment and the effect of the attributes on likelihood of testing did not differ by race. It is possible that racial differences in willingness to undergo testing have abated over time as knowledge about genetic testing has become more widespread and testing availability has increased. These results also suggest that policies about the delivery of testing by primary care providers, the disclosure of test results to insurers and race specific test marketing are not likely to create or exacerbate racial differences in uptake.
In contrast, although the overall effect of health care system distrust on likelihood of undergoing genetic testing was also relatively small and marginally significant, the strength of that association depends upon both whether the test is delivered by a primary care provider or a specialist and on whether the test results are disclosed to insurers. If test results are disclosed to health insurers, individuals with high values distrust will be particularly discouraged from undergoing testing. Values distrust encompasses concerns about the motives of the components of the health care system including insurers and it is not surprising that it heighten concerns about how insurers will use genetic test results.36
Given that disclosure to insurers lowers utilization overall, a policy that restricts disclosure of test results should both increase likelihood of test utilization and reduce disparities in utilization by health care system distrust. If the test is delivered by specialists, individuals with low competence distrust will be less likely to undergo testing whereas those with high competence distrust will be unaffected. One possible explanation for this finding is that while accessing specialists is inconvenient for all respondents, individuals who distrust the overall competence of the health care system place a greater utility on obtaining specialist input as specialists are often seen as having greater competence in their area of expertise. Thus, the overall effect of primary care provider delivery is to increase likelihood of testing, but mostly driven by those with low competence distrust. While it is accepted that disparities by race and socioeconomic status are problematic, the importance of differences in utilization by health care system distrust alone is uncertain. It is not clear that individuals with higher levels of health care system distrust represent a disadvantaged population or whether differences in utilization across distrust levels represent a socially unacceptable source of variation.
In addition, this study suggests that conjoint analysis may be a useful method for testing the potential effects of different decisions about health care delivery on utilization, including differential utilization among population subgroups. Conjoint analysis has been used extensively in marketing to predict future purchasing behavior, as well as to identify how preferences vary across subgroups- a process often termed market segmentation. Although this study does not allow us to test whether the preferences for the attributes of genetic testing would translate into actual behavior, the results have considerable face validity and the ability to assess conjoint scenarios for this relatively complex topic through a random digit dialing survey increases the feasibility of this method for other questions.
This study has several limitations. As noted above conjoint analysis measures preferences not actual behavior. Rates of genetic testing have been shown to be substantially lower than what was predicted based upon studies of preferences or intentions. However, the focus of the conjoint analysis is on the relative effect of different attributes rather than the absolute rate of testing. The response rate was relatively low and it is possible that non-responders differed from responders. We did not have data on non-responders to assess for the degree of response bias. Because conjoint analysis is a within-subject experimental design, the results are likely to be valid within the sample but may not generalize outside of the sample. We conducted the data collection by telephone because of concerns about completion rates using a phone-mail-phone strategy, particularly among minority participants. However, it is possible that the validity of the results could be affected by difficulty completing the conjoint scenarios without visual items. These data were collected in 2006 before the Genetic Information Non-discrimination Act of 2008. This legislation prevents discrimination by employers and health insurers on the basis of genetic information and may have reduced the adverse impact of insurance company disclosure on likelihood of testing.
In summary, willingness to undergo genetic testing is associated with several modifiable characteristics of how the tests are delivered. The effect of two of these characteristics (disclosure of test results to insurers and specialist vs primary care access) are modified in very specific patterns by values and competence distrust respectively), indicating that policy decisions about delivery of genetic testing may have differential effects across patients and may either widen or narrow existing disparities in testing utilization.