We report findings from testing a lung cancer screening decision aid in 50 primary care patients. We found that decision aid viewing was associated with greater knowledge of the benefits and harms of screening. At baseline, we found that participants tended to greatly overestimate the chances of benefitting from screening. After viewing, participants tended to have a more realistic understanding of the chances of benefitting. We also found that decision aid viewing led to improved understanding of two important but conceptually complex screening-related harms: false positives and overdiagnosis. To our knowledge, this study is the first to demonstrate that viewing a decision aid can help attenuate biased perceptions about the benefits and harms of lung cancer screening.
Our study complements and extends a limited body of empiric evidence about lung cancer screening decision aids. Previous studies were conducted in populations recruited from the community [19], or referral settings such as a tobacco cessation clinic [18] or a dedicated, sub-specialty screening program [20]. Our study was conducted in a primary care setting and allowed providers to exclude patients they believed to be poor screening candidates based on co-morbidity concerns. Thus, despite being relatively small, our study mimicked a systematic, practice-based approach to screening. Furthermore, two of three prior studies tested lung cancer screening decision aids in mixed populations that included relatively small numbers of screening eligible patients (n = 14 and n = 11, respectively) [18, 19]. All participants in our study were eligible for screening based on current USPSTF guidelines.
We found that baseline (pre-decision aid) screening preferences were heterogeneous, with roughly half of participants falling into each of our two preference groups (preferring screening vs. not). Although decision aid viewing was not associated with net changes in the proportions in each of these preference groups in the overall study sample, about one-quarter of participants in each group actually changed their screening preference after decision aid viewing (in opposite directions). Our findings suggest that we cannot assume screening-eligible primary care patients will be uniformly inclined toward or away from screening. Moreover, our results also suggest that a non-trivial proportion of primary care patients may change their preferences as they become more informed about screening.
Our results are consistent with the findings of Kinsinger et al.’s VHA pragmatic screening demonstration pilot [29] in that we observed heterogeneity in screening behavior. Among participants who preferred screening and saw a provider within 3 months of decision aid viewing, about half discussed LDCT scanning with their provider and about a third completed an LDCT. We also found that some participants who did not prefer screening after decision aid viewing ended up completing an LDCT. Kinsinger et al. similarly found that 50% of eligible patients completed screening [29]. While the VHA demonstration program developed and utilized paper-based decision support materials (from which we adapted our video decision aid), no data are available regarding the manner and extent to which they were used.
Our findings that 50% of patients preferred screening post-decision aid and 28% completed screening (among patients with post-decision aid primary care encounters), contrasts with findings from what is, to our knowledge, the only other study to assess screening completion in individuals receiving a lung cancer screening decision aid. That study, conducted by Mazzone et al. in patients attending a tertiary screening program, found that 95% of participants completed screening after undergoing a shared decision making visit [20]. We hypothesize that the overwhelming majority of patients who attend a dedicated lung cancer screening program assume that the purpose of referral to such a program is to complete screening, rather than to decide about screening. When considering Mazzone’s findings alongside our results, it suggests that the time frame during which screening-eligible patients actually make decisions about screening is before they attend a dedicated screening program. Tertiary lung cancer screening programs are now being implemented in the US, and many claim to be able to conduct shared decision making. Our findings add to the discourse regarding the appropriate implementation context for medical decision making about lung cancer screening.
We found that the more knowledge participants had about benefits and harms of screening, the less likely they were to prefer screening. The fact that we did not observe differences in screening behavior suggests the need for additional research aimed at understanding how we can help ensure that patients receive care that is consistent with their preferences. This will likely require interventions beyond providing patient decision aids.
Our study examined decision aid effects on knowledge about two important harms of screening: overdiagnosis and false positives. We chose these outcomes because: 1) overdiagnosis and false positive tests can cause substantial harms in screened populations; 2) CMS explicitly requires that information about overdiagnosis and false positives be included in lung cancer screening decision aids; and 3) these harms may not be recognized or understood by patients making medical decisions about cancer screening [9, 15, 25]. Other knowledge domains are also probably relevant to decisions about lung cancer screening [8]. Further research is needed to assess the validity of measures of decision-relevant knowledge and to better understand knowledge thresholds at which patients may be considered adequately informed to make medical decisions about cancer screening.
We found relatively low concordance (58%) between (post-decision aid) preferences and LDCT ordering during subsequent provider visits. It appears that a major driver of discordance was that many patients who preferred screening and did not have an LDCT ordered did not actually have an opportunity to discuss screening with their provider. This finding points to an important problem in implementation of patient decision support in health care. Decision aids are intended not to replace but to inform discussions between patients and providers about medical decisions [30, 31]. However, competing demands and lack of adequate provider time to deliver preventive services are important barriers to effective communication and decision-making [16]. Further, lung cancer screening is especially complex given both the risk for lung cancer in this population and the chances of serious complications as a consequence of screening. Thus, further research is needed to understand how best to structure lung cancer screening decision support interventions in the primary care medical home to ensure that there is adequate time for patients and providers to discuss and deliberate. Moreover, more research is needed to understand the role that domains other than knowledge and stated preferences play in the complex picture of lung cancer screening behavior.
Limitations
Our study has limitations. First, because of the one-group, pre-post study design we were unable to compare the effects of the decision aid with usual care. Second, this was a single site study and many patients declined initial eligibility assessment. Additionally, patients received an incentive to participate and were required to attend a separate study visit, both of which can affect participant behavior and sample representativeness. Thus, the degree to which the findings are generalizable to other screening-eligible populations is unclear. Nevertheless, our participants were similar to NLST participants in terms of average age (63 in our study, 62 in NLST), pack-years smoked (52 vs. 56, respectively), percent current smokers (46% vs. 48%, respectively), and percent female (48% vs 41%, respectively) [32]. Our sample reflects a more disadvantaged population than was studied in the NLST in that they were less likely to be white (58% vs 91%, respectively) and less likely to have received education beyond high school (50% vs 70%, respectively) [33]. Third, knowledge was assessed immediately following completion of the decision aid and may not reflect long-term retention of lung cancer screening information. Finally, our behavior assessments were relatively crude in that we did not differentiate between chest CT scans discussed and ordered for diagnostic reasons (i.e. to evaluate symptoms) vs. for true screening.
Another consideration is that the decision aid we tested was not targeted to lung cancer risk. Ideally, information about benefits and harms of screening would be tailored to the patient’s individual lung cancer risk, as occurs in the decision aid produced at the University of Michigan [33]. However, brief video decision aids such as ours or the one developed by Volk and colleagues [18] do not require high-level reading capability, entry of patient-specific data, or other interaction by the patient. Such a format offers potential advantages in terms of implementation, and may be better suited for low-literacy populations. Studies are needed to examine tradeoffs associated with using simpler versus more complex, tailored decision aids for lung cancer screening, particularly given the prevalence of low education and low literacy in the US screening-eligible population.