Developing Tailored Health Content to Promote Colorectal Cancer Screenings Using Longitudinal Health Data

Background: In spite of physician and insurer led educational communication campaigns, colorectal cancer is the third-most common cancer in the U.S. as well as the second leading cause of cancer mortality. Colorectal cancer screenings decrease the mortality as well as the cost associated with the disease. Despite the proven benefits of a colorectal cancer screening (CRCS), only 58.6% of the HEDIS-eligible U.S. population completes a CRCS.

Program background: Using a variety of data sources including medical claims, HEDIS and third party demographic data, we developed tailored communication outreaches for six different subpopulations overdue for a colorectal cancer screening. The six tailored content messages differ in framing (positive vs. negative), referencing (self, self and others) and argument (one-sided vs. two-sided) tactics. The subpopulations emerged from a data modeling exercise and are ranked from least likely to most likely to complete a colorectal cancer screening and are separated by gender when the health practice varies by gender. The demographic profile and preventive care practiced by the subpopulations vary dramatically by subpopulation; for example the lowest ranked subpopulation is 1.4 times more likely to be an African American, 1.2 times more likely that their education is limited to high school and 3 times less likely to complete a breast cancer screening if eligible as defined by the HEDIS breast cancer screening measure.

Evaluation Methods and Results:  A total of 46,697 members overdue for a CRC screening were scheduled to receive an interactive voice response (IVR) communication call encouraging CRCS. 25% of each subpopulation was randomly assigned to the ‘general’ message group and 75% of each of the six subpopulations was randomly assigned to the ‘tailored’ message group. We measured the success of each tailored message by comparing the CRCS rates for the tailored group to the CRCS rates of the corresponding general message group. Screening success was linked to the communication if it occurred within 90 days of the received communication. Tailored, segment-specific content led to higher screening rates for four out of six subpopulations; with the least likely to complete subpopulation (disengaged) having significantly higher CRCS rates associated with the tailored content ( 4.21% tailored vs. 3.06% general, p= .09 Tailored Message n=3,516 and General Message n=1,176). 

Conclusions: The outcomes suggest the utility of tailored health content and indicate opportunity for further refinement of the tailored content for the documented intervention. Also, refining the content will require an iterative “test-and-learn” approach to evaluate the effects of changes in messaging tactics to changes in outcomes. The significant result for the lowest ranked subpopulation indicates this methodology may reduce CRC health disparities within this population.

Implications for research and/or practice: The results indicate that the specific health message content that is part of an IVR health screening reminder affect outcomes,  supporting the further development and experimentation with tailored IVR health messages for health behaviors. This technique may help reduce disparities in CRC incidence rates, CRC death rates and CRC screening rates through the application of analytical techniques that leverage large and disparate data sets.