INTRODUCTION

Colorectal cancer (CRC) is the second leading cause of US cancer death.1 Failing to receive or stay current on CRC screening increases mortality risk over two-fold,2 yet nearly one-third of the US population aged 50 to 75 years old did not have a current CRC screening test in 2016.3 Direct mailing of a fecal immunochemical test (FIT) to patients has proven to increase screening4 but has not become widely adopted practice in part due to potential resource costs.5

As a behavioral health intervention, financial incentives can increase a variety of health behaviors6 and have been applied to FIT programs to increase efficiency and response. However, incentives have had limited success in increasing CRC screening with most trials indicating no effect7,8,9 and some finding modest effects10,11 or effects only in conjunction with other interventions.12 Targeting incentive programs according to patient risk or likely response is increasingly recognized as cost-effective practice, including for CRC screening.13,14,15,16 Little is known, however, about why and for whom financial incentives may succeed from the perspective of patients.17 We conducted a mixed methods study18 embedded within a pragmatic randomized controlled trial19 in order to explore why financial incentives often fail to increase rates of CRC screening and identify whether certain patient contexts and characteristics might improve their efficacy.

METHODS

From August to November 2018, we conducted an embedded mixed methods study18 with primary care patients at an urban, academic health system who were enrolled in a 4-arm randomized pragmatic clinical trial which evaluated and failed to detect significant effects of financial incentives on mailed FIT completion.19 The governing institutional review board approved all activities.

Participants

At trial enrollment (December 2015–July 2017), participants were aged 50–75 years and overdue for CRC screening (n = 897).19 Eligible patients received a mailed FIT kit and were randomized to one of four parallel arms to receive (1) no financial incentive; (2) an unconditional $10 incentive included with the mailing; (3) a $10 incentive conditional on FIT completion; or (4) participation in a lottery with a 1-in-10 chance of winning $100 conditional on FIT completion. Completion rates were not statistically superior among any of the incentive arms compared to the active control arm.19

Following trial completion, we invited a subset of eligible participants via mailed letter and follow-up telephone call to complete a one-time, semi-structured interview and questionnaire. We randomly identified and invited patients in batches, stratified by trial arm, until reaching data saturation. We chose to randomize and stratify sampling by trial arm to enhance variation of intervention exposure, primarily to understand how different financial incentives within the trial did or did not work.20 We contacted 369 patients and reached 128; of these, 71 agreed to participate, and 60 completed the interview, 15 per trial arm. To assess for representativeness, we compared the enrolled sample (n = 60) to each of the following groups: the overall trial sample (n = 897), the randomly selected recruitment sample (n = 369), those contacted for recruitment but not reached (n = 241), and those reached but not enrolled (n = 68). We found no significant differences by age or race/ethnicity (p < 0.05). For sex, males represented a significantly higher proportion of those contacted but not reached (46.5%) than of those who enrolled (31.7%; p = 0.038). There were no other significant differences by sex between groups.

Data Collection

As an embedded mixed methods study,21 domains of interest were identified a priori at the time of the pragmatic trial to understand why and how each of the four interventions, particularly the use of financial incentives, succeeded or failed to change behavior and improve FIT completion. We planned to quantify certain results, taking advantage of validated questionnaires where available, to speak to existing literature;22,23,24,25,26 where less relevant literature existed, we planned exploratory analysis and used open-ended questions. Utilizing convergent parallel design,21 we aligned domains across data collection tools (interview guide and questionnaire) to triangulate qualitative and quantitative findings. To evaluate how patient-level factors shaped mailed FIT completion, we developed a semi-structured interview guide using Andersen’s behavioral model27 including open- and closed-ended questions examining views on financial incentives and other facilitators or barriers to screening (see Online Supplement). Andersen’s behavioral model is a health services access and utilization framework which models individual care access and use as a function of the following three factors: predisposing factors, such as social structure and demographics; enabling factors, including personal, family, and community; and perceived and evaluated need factors.27 Participants also completed a brief questionnaire evaluating demographics,22 health,23 screening history,24 provider communication,25 and screening beliefs using previously validated measures.26 The questionnaire was verbally administered to participants at the time of the interview. Following verbal consent, trained research staff with no prior relationship to participants from the Mixed Methods Research Lab (MMRL) at the University of Pennsylvania conducted the interviews with eligible participants in-person or by phone (depending on participant preference). Interviews lasted 25–30 min on average, were audio-recorded, and were transcribed verbatim. All participants received $20.

Data Analysis

We analyzed the qualitative data using the constant comparative method, guided by modified grounded theory.18 We utilized a priori domains of interest based on Andersen’s behavioral model and inductively explored emergent themes within and across participants.18,27 We conducted a round of open coding on a subset of 4 interviews to identify initial themes. We then developed a coding dictionary guided by conceptual model (deductive codes) and themes identified during open coding (inductive codes) which included index, parent thematic, and child thematic codes with rules for each code type. All codes were applied at the question level for consistency. Two trained coders applied the refined codebook to the interview set, each coding 35 of the 60 interviews using NVivo, with 10 interviews independently double-coded by both coders. Overall inter-rater reliability was calculated on the double-coded interviews (kappa = 0.7), and we produced summary thematic reports. We then conducted targeted secondary analysis to quantify patient response by incentive impact. Two analysts independently coded the individual responses to the open incentives question and had 93.3% agreement; the discordant responses were resolved by a third reviewer. We conducted descriptive and bivariate analyses of quantitative data (Stata version 15.1, Stata Corp LP), using concurrent methods to triangulate quantitative patterns with qualitative data.18.

RESULTS

Patient Characteristics

The median age of participants was 60 years, and most were female (68%), non-Hispanic Black (68%), and without a college degree (53%). Over half (60%) preferred mailed FIT to colonoscopy or sigmoidoscopy (Tables 1 and 2).

Table 1 Participant Characteristics by Self-Reported Incentive Impact on CRC Screening: Predisposing Factors in Andersen’s Belief Model (ABM)a
Full size table
Table 2 Participant Characteristics by Self-Reported Incentive Impact on CRC Screening: Perceived and Evaluated Need and Enabling Factors (ABM)a
Full size table

Impact of Incentives

The majority of respondents (n = 49; 82%) reported incentives would not change their decision to complete a FIT (Table 1); 12% of these (n = 6) had never been screened for CRC using any modality. Those participants who reported incentives would impact their screening behaviors (n = 11) were significantly less likely to agree that CRC screening is beneficial (72.7% vs 95.9%; p < 0.05) or that CRC is curable if detected early (63.6% vs 98.0%; p < 0.05), and nearly half (n = 5; 46%) had never been screened (Table 3).

Table 3 Joint Display of Screening Beliefs by Impact of Financial Incentives on the Decision to Complete or Not Complete a FIT Kit
Full size table

Qualitative data supported the quantitative findings and clarified how financial incentives shape FIT decisions (Table 4).

Table 4 Joint Display on the Impact of Financial Incentives on Patient Decision to Complete or Not Complete a FIT Kit
Full size table

No Impact on Decision

Patients who stated incentives would not impact their decision-making largely reported they would complete the FIT regardless. These patients primarily cited their health as a driver of screening decisions, with financial incentives viewed as a bonus. Some, however, could never be financially motivated to complete the FIT, either because they preferred another screening method or would not participate in CRC screening regardless.

Impact on Decision

Among respondents who reported a financial incentive would impact their screening decision, several indicated they would return the FIT kit more quickly. Many noted, however, that any influence would depend on the amount, with suggestions varying from $10 to $500.

Motivators, Barriers, and Facilitators

To better understand how financial incentives impact decision-making, we assessed cross-cutting motivators, barriers, and facilitators to FIT completion (Table 5).

Table 5 Thematic Analysis Summary with Illustrative Participant Quotations on Cross-Cutting FIT Kit Motivators, Barriers, and Facilitators
Full size table

Motivators

Most commonly, respondents discussed personal beliefs, such as health preservation, mortality reduction, health scares, and known risk factors. Other motivators included the ease of mailed FIT and provider recommendation.

Barriers

While not every patient reported barriers, the most commonly mentioned were personal factors, such as forgetting or losing the test or being too busy or ill. Some indicated they would delay or not complete the test because they felt healthy or feared the results. Structural issues, including cost and difficulty accessing a post office, were common. Respondents also mentioned test-related factors, particularly disgust, embarrassment, or reliability concerns; disgust was particularly common among those who indicated financial incentives would impact their decision-making.

Facilitators

The main facilitator mentioned was direct outreach, encompassing provider recommendation, education, and reminders. Preferred medium varied, including text message, email, telephone, mail, and in-person reminders. Many also discussed handoff issues, preferring to receive or complete the kit at their provider’s office.

DISCUSSION

While financial incentives have had limited to no success in improving at-home CRC screening rates to date, this intervention has proven an effective strategy to improve other health behaviors. As Moller and colleagues suggest,28 a better understanding of the factors motivating patients’ responses to financial incentives may identify patient contexts and characteristics which improve the efficacy of financial incentive programs. This study provides novel insight from the patient perspective into such factors, including indication that differing beliefs and motivations may require tailored intervention approaches to be most effective.

The majority of participants (49; 82%) indicated that financial incentives would not impact their decision to complete the FIT kit or not. These patients typically reported being motivated to complete the FIT regardless of financial incentives due to desires to stay healthy and follow doctor recommendations, which comported with higher perceived health benefits of screening and belief in the curability of CRC if detected early. While the mechanisms by which financial incentives could be effective vary by individual (e.g., cue to action), these findings suggest that financial incentives may not increase screening for most patients. Common barriers such as forgetfulness, busyness, mailing difficulties, and costs could be addressed through frequently mentioned facilitators, such as enhanced outreach, reminders, in-clinic hand-off of FIT kits, and reimbursement for any costs. Such interventions may not be effective, however, among the subset of participants who reported that financial incentives would not influence their decision but have never been screened (6/49; 12%). Previous research has examined persistent barriers to CRC screening, such as fear of results,29,30,31 but further investigation into such barriers among this predominately Black, urban population may be productive.

Participants who responded that incentives would influence decision-making (11; 18%) also reported lower perceived benefits of screening and ability to cure CRC if detected early. These beliefs may contribute to the group’s higher likelihood of never before being screened using any modality. Here, bolstering the effect of the cross-cutting facilitators mentioned above, financial incentives may influence patient decisions because these patients do not perceive many benefits to FIT completion and thus lack intrinsic motivation.28 Substantial financial incentives may be required, however, to sufficiently increase perceived benefit-to-effort ratio and to avoid the “peanuts” effect, where incentives are perceived as too small given a high-stakes context, such as one’s health, and subsequently undermine motivation.32 For example, of the subset of patients who reported never being screened but that incentives would impact their decision (5/11; 46%), none completed the FIT during the trial despite being randomized to receive a financial incentive; this may be because the incentive was too low.19

Limitations

This study has limitations, including use of self-reported data subject to recall or social desirability bias and sampling from a patient cohort at a single academic center. Participants differed significantly by sex from those who were contacted to participate but not reached; thus, participant beliefs and responses may differ in meaningful ways from those unreached and from those who decided not to participate in this study. Study participants were not significantly different, however, than the overall pragmatic trial participants in sex, age, or race/ethnicity, enabling contextual insight that can broadly help to explain why financial incentives did not succeed for most participants in the trial.

CONCLUSION

Findings indicate financial incentives in colorectal screening programs may impact patient decision-making to complete screening differently based on certain beliefs, with most patients indicating that incentives do not influence decisions. Future studies evaluating the impact of financial incentives should consider stratifying by baseline screening beliefs and history to further evaluate differential impact across patient screening beliefs. This may more accurately identify strategies to improve the targeting and cost-effectiveness of mailed FIT outreach programs, particularly financial incentive interventions, thus increasing the uptake of overall CRC screening.