Effectiveness of a decision aid for promoting colorectal cancer screening in Spain: a randomized trial

Colorectal cancer (CRC) is the third most common type of cancer in the world and one of the main causes of death by cancer, showing a wide geographical variation in incidence and mortality [1]. In western countries, these rates have declined from several decades ago due to the introduction of screening techniques (e.g., fecal occult blood test -FOBT-, colonoscopy, sigmoidoscopy), and the improvement in treatment [2‐4]. CRC screening is therefore recommended for people above 50-years old, and population programs have been implemented in several countries at a national or regional level, however screening rates are still low [5‐7]. Research has shown that screening uptake is significantly influenced by a wide range of sociodemographic, clinical, psychosocial, lifestyle and health system-related variables [8‐11].

Proposed interventions to increase and improve recruitment for CRC screening have included reminders by health professionals, informational material in different formats (printed, videos, interactive computer programs), mailed FOBT kits, group education, counseling or provider or practice-directed interventions [12‐14]. One of these practice-based interventions is the application of decision aids (DA). These tools include explanations of the available treatment or screening/diagnostic procedures, evidence-based quantitative information about their benefits and risks, and promote implicit or explicit clarification of patients’ preferences about testing and its potential consequences [15, 16].

DAs have shown to improve patients’ decisional process (e.g., knowledge of the disease and treatments, risk perceptions, decisional conflict, congruence between preferences and choices) in many physical and mental health conditions [17]. In the field of CRC, Volk et al. [18] recently published a systematic review and meta-analysis on the effectiveness of DAs. Pooled results showed that, compared to no-intervention, they significantly increased participants’ knowledge, and screening intention and uptake, but when control groups were given general CRC screening information, only the difference in knowledge was significant.

In Spain, public population-based CRC screening programs have been heterogeneously implemented across and within its 17 autonomous regions, since the first initiative launched in Catalonia in 2000 [19]. Programs follow the European Guidelines, are entirely free of charge and biennially invite men and women aged 50 to 69 years to a FOBT test; colonoscopy as a diagnostic test is offered to those FOBT positive. Although screening rates have progressively increased through the successive rounds of the program, reaching about 40% of eligible population, they still are below the optimal recommended value or standards [20].

In this context, the aim of this study is to assess the effectiveness of a DA on the basic decisional processes (knowledge, decisional conflict, intention) about CRC screening of Spanish primary care patients eligible for the procedure. Secondarily, since in our region (Canary Islands) the public screening program has been implemented only in some areas, we wanted to assess the effect of the DA in participants who previously had been invited (and refused) to be screened compared to those who had never been invited.

We performed a randomized controlled trial (RCT) to assess the effectiveness of a web-based DA developed to promote informed decisions about CRC screening. Inclusion criteria were: age 50–69 years-old, having no CRC history or current symptoms, no family history of CRC, and no previous CRC screening. Two primary care centers in Tenerife (Canary Islands, Spain) participated in the study, carried out during 2016. One of them (hereafter, center A) attends patients in an area where the public CRC screening program was not implemented at that moment, although it was expected to be accessible in the next years. The other (center B) is in an area where the screening program have been running since 2009, and therefore its patients between 50 and 69 years-old had received invitations and reminders in the past to undergo the screening (but not specific information about the procedure). Collaborating physicians in both centers assessed the eligibility of their consecutive patients, and invited those eligible to be contacted by telephone by a member of our team, who explained the study in more detail and established an appointment in their primary care center.

Computer-based simple randomization was performed by a statistician not involved in the study, and the researcher who recruited participants and established an appointment by phone was blinded to allocation (we used a centralized off-site computer allocation process). All participants were received in a room of their primary care center by other researcher, who had the allocation list; all participants were assigned to the group in which they had been randomized. Intervention participants signed informed consent and reviewed the DA in the computer; the researcher was sitting beside them and briefly explained them the functioning of the DA and gave support in navigation when necessary. After that, participants filled the questionnaires assessing the outcome variables, in the same web-based interface, with no mediation of the researcher. Control participants only signed informed consent and completed the questionnaires, in a different interface of the web site. Therefore, all measures were assessed only once (after reviewing the DA in the intervention group), except knowledge in the DA group, which was assessed both before and after the DA application.

Figure 1 shows the participants’ flow chart and Table 1 their baseline characteristics. One hundred and twenty patients were informed about the study by the recruiter researcher, from whom 13 decline participation. Therefore, 107 participants were included, 83 from center A and 24 from center B, and 53 and 54 patients were randomized to intervention (DA) and control groups, respectively. Mean age was 58.1 years-old, 57.9% were female and a 45.7% had completed secondary education. Patients from center A were significantly older than those from center B (mean of 59 vs. 55 years, respectively, p = 0.004) and included fewer women (52% vs.79%, p = 0.017).

Table 2 shows the results of the different ANOVA performed for each of the continuous outcomes measures. The effect of center was significant only for the “informed” subscale, with patients from center B showing significantly less conflict (p < 0.001).

Regarding the effect of the intervention on decisional conflict, the interaction term was significant for the total score and three subscales (informed, values and uncertainty); post-hoc analyses showed that the DA only was effective in center A (p < 0.001). For the other two subscales there were not significant main or interaction effects, but when analyzed separately in center A, t-tests yielded p-values of 0.003 (effectiveness) and 0.05 (support) favoring the DA. Differences ranged from 7.2 (support) to 62 points (informed).

Knowledge in the control group was not significantly different from pre-intervention scores in the DA group (p = 0.714), which after the intervention improved 16.6 points on average (p < 0.001). ANOVA yielded a significant main effect of the DA on knowledge (mean difference: 17.8%, p < 0.001).

In the whole sample, most participants stated an intention to undergo FOBT (92.4%) or colonoscopy (89.5%). In center A, the differences favoring the DA (FOBT: 100% vs. 89.5%; Colonoscopy: 95.3% vs. 81.6%) were significant for both procedures in the uncorrected analysis, but not when Yates’ correction was applied (Table 3). In center B, none of the analyses yielded significant results.

Table 4 shows the results obtained for each of the goals/concern evaluated. Reducing both the probability of developing colorectal cancer (78% of patients scoring 5 in a 1–5 scale) and dying by its cause (91%) were rated by participants as the most important issues when deciding about CRC screening, followed by the risk of complications (70%). Univariate associations between the importance attributed to each goal/concern and the intention to get screened were assessed by means of logistic regression and no significant results were observed. Including center as covariate or using non-parametric analyses (Mann-Whitney’s U, Kolmogorov-Smirnov’s test) did not change the results; therefore, a measure of concordance between goals/concerns and intention to be screened could not be constructed.

This study aimed to assess the effectiveness of a DA for CRC screening on decisional conflict and knowledge of targeted Spanish participants who had never been screened, recruited in primary care services. We also evaluated the participants’ intention to undergo the screening procedure. In addition, we included a small subsample of participants (center B) who had previously been invited (and refused) to be screened in the public program.

In center A, results confirmed the beneficial effect of the DAs on the patients’ decisional process found in previous studies in CRC [18] and other diseases [17]. Intervention group showed significantly higher knowledge (pre-post difference of 17 points) and lower decisional conflict than control participants, with a strong effect in the DCS Informed subscale (between-group difference of 62 points). Control participants showed an acceptable level of knowledge (60% of correct responses), and nonetheless they expressed a high level of feeling uninformed. This result is in line with the relatively low correlation between objective and subjective knowledge usually observed in previous research [30‐32] (r = 0.35 in this study), and highlights the need of educating about health topics not only to empower citizens with poor knowledge about these issues, but also to improve the subjective perception of been informed in those who already have a correct level of knowledge.

In center B, the DA also significantly improved participants’ knowledge, but had no effect on decisional conflict. Although the statistical power of the analysis was very low due to the small sample size, differences were minimal for the DCS total score and its subscales. A possible floor effect could be present, since the control group obtained relatively low scores (29 points in the total score). The effect of center was significant only for the subscale informed, with less conflict in center B. This could be reflecting the effect of the public program’s invitations and reminders provided to these participants in the previous years, although this must be interpreted cautiously due the small sample size and the differences observed between centers in age and gender composition.

The percentage of participants who stated an intention to undergo screening was very high in the whole sample. Control group obtained a rate around 85%, a value markedly greater than the observed in previous trials with control groups without intervention [33, 34], or even in those in which control participants were given general information about CRC [35‐37]. Although intention significantly predicts screening uptake [26, 27], this seems to be an “inflated” result since screening rates in our region are around 40% [20]. A possible explanation is that we used a yes/no response format (whether the patient had intention to be screened or not) but not the “intensity” of intention (usually assessed with a Likert scale [37‐39]), nor did we make explicit a time frame in which the decision had to be made. This could make it to seem more “hypothetical” for participants (and maybe more prone to desirability bias), and lead those in a more contemplative stage of decision making, in terms of the Transtheoretical Model of behavior change [40], to answer affirmatively. Performing the study physically in a health context (primary care centers) could also increase this desirability bias. In center A, the effect of the intervention was significant only before applying correction for continuity; nonetheless, the study was underpowered to detect the between-group difference estimated in the meta-analysis of Volk et al. [18]: 66% in the DA groups versus 45% in control groups with no intervention (168 participants would have been needed, with a 95% confidence level, to obtain a statistical power of 80%), and therefore we cannot draw definitive conclusions about the effect of the DA on this measure. Finally, we did not find any significant associations between goals/concerns and intention, a result that could be influenced by the low variability observed for both measures.

There are several limitations in this study, beyond the absence of screening uptake assessment. We did not assess decisional conflict and intention at baseline, although randomization is expected to prevent biases in this sense. The absence of intervention in the control group may introduce a “novelty effect” in favor of the DA; to minimize as much as possible this potential bias we tried to offer the same experience (except for the application of the DA) to intervention and control participants: appointment in their primary care center, standardized instructions provided by the researcher, who did not mediated in participants’ responses to questionnaires, which were delivered in the same way. Regarding external validity, a selection bias could be present since we recruited participants in primary care centers, who might not be completely representative of the population targeted for CRC screening. With these limitations in mind, the results confirmed the improvements of participants’ knowledge observed in previous studies, and add evidence about the DAs effect on decisional conflict (a measure little used previously in the field of CRC screening), supporting their effectiveness to promote informed decisions about CRC screening. We also have provided preliminary evidence on the null effect of the DA on decisional conflict in participants who have been invited (and refused) to participate in a public screening program.

Beyond the positive effects of DAs effects on cognitive outcomes, however, their effect on intention and screening uptake are far from the desired rates from a public health perspective. The meta-analysis of Volk et al. [18] shows that the pooled rate in any DA group for screening intention/interest was 66% (95% CI: 54–77%) and only 33% (95% CI: 22–46%) for screening uptake. These results highlight the necessity to include a motivational component in interventions aimed to improved screening rates. On the other side, DAs outperform general information materials at improving knowledge, but not at increasing the rate of CRC screening, which raises questions about the quantity and specificity of information necessary or sufficient to decide to undergo the test.

Decision aids represent a useful resource to improve the quality of the decisional process about colorectal cancer screening, in terms of higher knowledge and less decisional conflict. However, preliminary evidence suggests that the effect on decisional conflict is not significant in people who have been previously invited to participate in a public screening program (although not specifically informed about the procedure). Future research should identify what subgroups could benefit more from DAs, as could occur with vulnerable populations [41, 42], and explore how values and motivational factors interact with knowledge and cognitive processes when making the decision about screening.