In this pilot study, we tested the ability of MDS to elicit patients’ preferences for attributes relevant to CRC screening tests. This tool may be useful for improving shared decision making because it is relatively quick and fairly easy for patients to use, incorporates both patient education and patient preferences, and provides a framework for patients to discuss and choose a test with their physician. A graph like Figure generated from the MDS survey could be used to stimulate discussion between a primary care physician and a patient about which test would be best for them.
As expected, patients varied in how they prioritized CRC screening test characteristics. In particular, concern for invasiveness of the test and willingness to trade-off sensitivity were associated with age and health status. Overall, however, subjects rated sensitivity of the CRC screening test as the most important attribute. Sensitivity was also found to be important to patients in other studies17,25,26
and correlated with preferring colonoscopy as it did in ours17,25
. However, a rating-scale-based conjoint analysis in a racially diverse population found that patients were more concerned with test process then accuracy25
. This latter difference may have related to population differences as the patients in the rating-scale analysis were more racially diverse, had a higher percentage of women and less education. However, rating scale versus choice-based task differences may have contributed. We were surprised that “prep” was not important to most subjects, as in clinical practice many patients complain about this. Possibly, this attribute was not ranked high because it is required for all CRC screening tests except FOBT, which was not a preferred test in this population. A second unexpected finding was the population did not rank the risk of pain as important. This may be because most subjects had prior screening by colonoscopy (75%) and reported a positive or neutral experience with the test (91%). Two prior conjoint analysis studies also found that pain was considered less important than test accuracy and process25,26
. However another study found that pain was important, particularly to screening naïve patients, although less important than effectiveness and preparation24
Consistent with a large amount of literature describing patient preferences for different health care decisions, we found that subjects varied in their preferences for CRC screening tests33–36
. What is surprising and interesting in our population is that despite high levels of satisfaction with prior colonoscopy, a significant proportion of patients would choose colon capsule or CT colonography over colonoscopy. Most other groups have found that colonoscopy is not the most preferred study37–41
, although a survey study in a veteran population similarly found that colonoscopy was the preferred test42
. In the conjoint analysis survey by Marshall et al. in which screening rates were lower, 30% of patients preferred no screening; the preferred test for patients desiring screening was CT colonography26
. Hawley et al. did not find an association between prior screening with endoscopy and choice of screening test, and the preferred tests in their study were fecal immuonochemical testing and CT colonography. They did not find an association with gender25
. However, female gender has in the past been associated with lower screening rates by colonoscopy, so the preferred test in our study may not have been colonoscopy if our population included more females43,44
Some of the attributes in our study may have had complex relationships, i.e., the risk of pain and the need for sedation. Depending on perceived relationships, this may have impacted individual results. Additionally, our tool was not designed to address directionality or levels of attributes as has been done with conjoint analysis24–26
. This may have affected our results as ambiguity in the meaning of the attribute may affect selection of the attribute. However, inclusion of directionality for all test characteristics would have made the survey much more complex. Moreover, in contrast to conjoint analysis surveys, which require directionality to predict preference for specific options, our intention was to enable patients to prioritize characteristics (regardless of direction) in order to then be able to discuss their preferences with their physician.
Because the majority of patients had prior CRC screening by colonoscopy (75%), the reported preferences in our population may not be broadly applicable to populations with low screening rates or less exposure to colonoscopy. This is a significant limitation, as is the relative homogeneity of our population including mostly white, male patients. We did not consider cost in our study. This affects comparison of our results with other preference studies as this attribute is known to be important to patients15,16
. We chose not to address cost as the actual costs to individuals would have varied widely among non-veterans. In reality, cost and insurance coverage would limit the number of possible options available to patients, inflating the importance assigned to this attribute and limiting evaluation of other attributes. Additionally, we did not include frequency of screening as an attribute. Although frequency is important to patients, the interval for a given patient is difficult to predict as the results of the test impact the recommended screening interval.
In addition to “test sensitivity,” we have specifically evaluated patient’s value of attributes related to mode of the individual tests as has been done in recent conjoint analysis studies. We have additionally evaluated patient’s values for specific complications; these have not been included in most other decisional tools other than one study that evaluated risk of complications, but not specific complications24
. The latter appears to be important as older patients in our study were concerned about the need for sedation and the risk of a capsule getting stuck.
Our results add to literature underscoring the need to incorporate patient preferences into a shared decision-making process. This pilot study suggests that patients vary in how they prioritize colorectal cancer screening test attributes, this variation is associated with test preferences, and a MDS tool may help patients construct their preferences.