Measuring the Impact of Providing Personalized Risk Information to Patients and Clinicians

Background Colorectal cancer (CRC) is the second-largest cancer-related cause of death in the United States, and low screening rates lead to thousands of preventable deaths. Leading organizations recommend several screening tests for people at average risk for CRC, including colonoscopy and stool testing (eg, fecal immunochemical testing or fecal immunochemical testing plus DNA testing [eg, Cologuard, Exact Sciences Corporation]). A validated prediction rule estimates the probability that a patient has an advanced colorectal neoplasm (ie, a CRC or advanced precancerous polyp). Informing patients and clinicians about the patient's chance of having an advanced colorectal neoplasm could motivate screening uptake, inform test choice, and improve decision-making. Objectives Our study had 2 objectives: Aim 1: To measure the impact on patient decision quality and screening uptake of providing patients and their clinicians with personalized messages about the patient's current risk of having an advanced colorectal neoplasm. Aim 2: To determine mediators and moderators of the interventions' effectiveness. Methods We conducted a 2 × 2 factorial design, randomized trial (ClinicalTrials.gov identifiers NCT04683731 and NCT04683744) involving clinicians and their patients who were due for CRC screening, were at average risk of CRC, and had an upcoming primary care appointment at either of 2 health care systems in Central Indiana. Patients were randomly assigned to view a decision aid about CRC screening either with a personalized message about advanced colorectal neoplasm risk (personalized decision aid) or without it (generic decision aid). Clinicians were randomly assigned to receive notifications that the patient was due for screening, either with a personalized message about the patient's advanced colorectal neoplasm risk (personalized notification) or without it (generic notification). The personalized message stated the patient's advanced colorectal neoplasm risk as high, moderate, low, or very low and recommended colonoscopy for patients with high advanced colorectal neoplasm risk and stool testing for those with low or very low risk. Patients completed questionnaires before (T0) and immediately after viewing the decision aid (T1), shortly after their primary care visit (T2), and 6 months after enrollment (T3) to assess secondary outcomes of perceived risk of CRC, knowledge about CRC screening, decisional conflict, intention to be screened, intended screening behavior (colonoscopy, stool test, or none), perceived shared decision-making, and satisfaction and usability of the decision aid. Screening uptake (co-primary outcome) and test choice (stool testing or colonoscopy, secondary outcomes) were extracted from the participants' electronic health records at 6 months. High-quality decision-making (co-primary outcome) occurred if the participant had adequate knowledge and received their intended screening test. Other secondary outcomes documented in the electronic health record included screening orders and the opening of clinician notifications. Logistic regression with random clinician effects and analysis of covariance models were used to analyze intervention effects. Potential mediators of change in screening intention, perceived CRC risk, and decisional conflict were analyzed by testing indirect effects within logistic regression models. Health care system as well as individuals' age, advanced colorectal neoplasm risk, adequate knowledge at baseline, and health literacy were examined as potential moderators of the intervention effects by including appropriate interaction terms in regression models. Results A total of 1084 patients and 214 clinicians participated. Patients had a mean (SD) age of 56.5 (6.2) years and were mostly female (59.7%). Almost all patients had completed high school (95.7%); 27.9% were Black, and 3.3% were Hispanic. Their mean (SD) advanced colorectal neoplasm risk score was low, at 3.0 (1.9). Primary hypotheses: There were no differences in the odds of screening uptake for either the decision aid or clinician intervention. Similarly, there were no differences in the odds of making a high-quality decision for either intervention. When patients who did not attend their clinic visit were excluded, though, the effect of the clinician notification on screening uptake was significant (odds ratio [OR], 1.36 [95% CI, 1.00‑1.84]; P = .048). Health system moderated the effect of the clinician notification on decision quality. At 1 health system, patients whose clinician was sent the personalized notification rather than the generic notification had significantly higher odds of having made a high-quality decision (OR, 1.39 [95% CI, 1.01-1.92]; P = .045). Secondary hypotheses: There were no differences in the odds of uptake of colonoscopy in the high-risk group for either intervention. Similarly, no difference was observed in the odds of uptake of stool tests in the low-risk group for either intervention. For uptake of stool tests, there was a 3-way interaction among health system, type of decision aid, and type of clinician notification. At 1 heath system, the odds that a patient would complete a stool test were higher when their clinician was sent a personalized notification instead of a generic notification if the patient saw the generic decision aid (OR, 3.24 [95% CI, 1.17-8.98]; P = .02), and the odds of the uptake of a stool test were also greater when the patient viewed the personalized decision aid instead of the generic decision aid if their clinician received the generic notification (OR, 2.91 [95% CI, 1.09-7.81]; P = .03). Other secondary outcomes: After viewing the decision aid, the personalized patient decision aid group had greater odds of intending to have any CRC screening (OR, 1.55 [95% CI, 1.03-2.33]; P = .03) and of having a stool test (OR, 2.17 [95% CI, 1.62-2.91]; P < .01), with decreased intention to have a colonoscopy (OR, 0.58 [95% CI, 0.43-0.79]; P < .01). At T1, the personalized decision aid group had higher intention to screen(P = .046), lower perceived CRC risk (P = .03), and lower decisional conflict (P < .01) than the generic decision aid group. At T2, those patients who viewed the personalized decision aid had higher knowledge scores (P = .02). There were no effects at T2 for the clinician notification. There were no differences by decision aid or clinician notification for the shared decision-making outcomes, satisfaction and usability scores, or screening orders. Mediation hypotheses were not supported. Conclusions Including a personalized message in a decision aid or a clinician notification about the patient's estimated chance of having an advanced colorectal neoplasm had no significant effects on screening uptake, decision quality, or test choice overall. The personalized clinician notification, however, increased the screening uptake for patients who attended their clinic visit as well as the odds of patients making a high-quality decision for all participating patients at 1 health system. At 1 health system, the personalized clinician notification and decision aid combined to increase the uptake of stool testing. Limitations Our study involved 2 health care systems in the Midwest with low representation of Hispanic and Asian patients and few patients with high-average advanced colorectal neoplasm risk, and it excluded non-English speakers. All patients viewed a general decision aid, which may have diluted the effect of the personalized clinician notification.