Impact of Continued Mailed Fecal Tests in the Patient-Centered Medical Home: Year 3 of the Systems of Support to Increase Colon Cancer Screening and Follow-Up Randomized Trial

Abstract

BACKGROUND

The current study was conducted to determine the effect of continuing a centralized fecal occult blood test (FOBT) mailed program on screening adherence.

METHODS

A patient-level randomized controlled trial was conducted in 21 patient-centered medical home primary care clinics between January 2010 and November 2012. A total of 2208 patients ranging in age from 52 to 75 years in a substudy of the Systems of Support to Increase Colon Cancer Screening and Follow-Up (SOS) trial were randomized at year 3 to continued automated interventions (Continued group), which included mailed information regarding colorectal cancer (CRC) screening choices, and were mailed stool kit tests or to a group in which interventions were stopped (Stopped group). The main outcomes and measures were the completion of CRC screening in year 3 and by subgroup characteristics, respectively.

RESULTS

Adherence to CRC screening in year 3 was found to be significantly higher in patients in the Continued group compared with those in the Stopped group (53.3% vs 37.3%; adjusted net difference, 15.6% [P<.001]). This difference was entirely due to greater completion of FOBT (adjusted net difference, 18.0% [P<.001]). Year 3 CRC screening rates were highest in patients in the Continued group completing FOBT in both years 1 and 2 (77.2%), followed by patients completing only 1 FOBT in 1 of the 2 years (44.6%), with low rates of CRC testing reported among patients not completing any FOBT within the first 2 years (18.1%).

CONCLUSIONS

A centralized mailed FOBT CRC screening program continued to be more effective than patient-centered medical home usual-care interventions, but only for those patients who had previously completed FOBT testing. Research is needed regarding how to engage patients not completing CRC testing after being mailed at least 2 rounds of FOBT tests.

INTRODUCTION

The lifetime risk of colorectal cancer (CRC) is >5%¹ in the United States, where, despite recent declines in incidence and mortality, CRC remains the second highest cause of cancer death.¹ Better treatments have improved survival rates, but morbidity and mortality could be more rapidly and cost-effectively reduced by achieving higher uptake of and long-term adherence to CRC screening.²

The US Preventive Services Task Force recommends CRC screening with colonoscopy every 10 years, flexible sigmoidoscopy every 5 years, or fecal testing (using a high-sensitivity fecal test) annually.³ Nearly all trials testing interventions to increase CRC screening rates have tested the effect of an intervention on only a single round of screening. To our knowledge, few trials to date have evaluated repeated interventions and their influence on long-term adherence.^4,5 We previously demonstrated that a low-cost, health information technology-facilitated mailed intervention with stepped increases in support was able to double CRC screening uptake and adherence rates over both years of the study (Systems of Support to Increase Colon Cancer Screening and Follow-Up [SOS] trial).^6,7 Although it appears obvious that continued mailed interventions might be required to maintain adherence to annual FOBT screening in the long term, to our knowledge this hypothesis has not been tested to date. Furthermore, it is unknown whether repeated mailed interventions are more effective for specific patient subgroups. We conducted a randomized controlled trial (RCT) to answer these questions.

MATERIALS AND METHODS

The study presented herein was conducted from November 2010 to December 2012 and was supported by a grant from the National Cancer Institute. Study procedures were approved by Group Health’s Institutional Review Board. The current study is registered at clinicaltrials.gov (NCT00697047).

Study Participants

SOS is an ongoing RCT of 4675 patients aged 50 to 73 years who are overdue for CRC screening (no colonoscopy within 9 years, no flexible sigmoidoscopy within 4 years, and no fecal test within 9 months) who were enrolled in the study between August 2008 and November 2009.⁸ Potentially eligible patients received an invitation letter, followed by a telephone call to confirm eligibility (no prior CRC, inflammatory bowel disease, or life-threatening illnesses such as renal failure) and willingness to participate. Those verbally agreeing to participate were mailed information regarding participation (no written informed consent was required) and were randomized to receive either 1) usual care or 1 of 3 stepped-care interventions: an electronic health record (EHR)-linked automated mailed program that included information regarding CRC screening choice, a number to call for endoscopy, and mailed FOBT for those not calling (Automated); this plus brief telephone assistance from a medical assistant to complete their choice of CRC screening test (Assisted); or this plus nurse navigator ongoing support for overcoming screening barriers (Navigated). Methods, recruitment, and results of the parent 2-year study have been published previously.^6–10 The study setting is Group Health Cooperative, an integrated health care system in Washington State and 21 Group Health-owned primary care medical centers.

Study participants were eligible for the study reported herein (the “year 3 study”) if they were originally randomized to 1 of the 3 active intervention arms (Automated, Assisted, or Navigated), had not died, were still enrolled in Group Health, were aged ≤75 years, and had not opted out of participation before year 3 study randomization (Fig. 1). Participants who in years 1 or 2 had a diagnosis of CRC, underwent colonoscopy, or had a positive FOBT with or without a follow-up colonoscopy also were considered ineligible because they no longer needed routine screening. Patients completing flexible sigmoidoscopy in years 1 or 2 were eligible for interventions in year 3 because, at the time, Group Health recommended combining screening sigmoidoscopy with FOBT testing.

Figure 1.

Study flow diagram. CRC indicates colorectal cancer; FOBT, fecal occult blood test; Y3, year 3.

Participants eligible for the year 3 study were randomized to either the Stopped or Continued intervention arms using concealed computer-generated random sequences and a permuted block design, stratified by prior randomization arm, clinic, and whether the participant had completed a FOBT in year 1 or 2. The study team was blinded to the randomization assignment, except for the programmer and research assistants managing the mailings.

Interventions

Patients in the Stopped arm received usual care in year 3. As part of usual care at Group Health, patients are assisted in completing preventive health screening tests through an annual birthday letter reminding them of screening tests that are due (including CRC screening). In addition, all Group Health primary care medical centers are certified patient-centered medical homes (PCMHs).¹¹ As part of the PCMH, before clinic visits, medical assistants or nurses (collectively referred to as medical assistants (MAs)) use the EHR to generate lists of patients’ unmet prevention needs (eg, flu shots, CRC screening). At the visit, the MA gives the form to the physician. Patients who are overdue for CRC screening receive a FOBT kit or discuss flexible sigmoidoscopy or colonoscopy with their physician (“inreach”). MAs also review the physician’s list of patients during the month of their birthday and contact patients by telephone or secure electronic message and mail FOBT kits to those overdue for CRC screening (“outreach”).

Participants in the Continued arm received usual care plus continued automated mailed interventions (the Automated intervention in years 1 and 2 of the SOS study). Automated interventions were generated by a study database linked to Group Health’s data warehouse and EHR. Participants due for screening (12 months from their last completed FOBT, or 11 months from the last mailed FOBT if not completed or they did not have a colonoscopy in the interim) received up to 3 mailings. The first mailing included a letter stating that they were due for CRC screening and that they would soon receive a FOBT kit, a pamphlet on CRC screening choices (fecal test, flexible sigmoidoscopy, or colonoscopy), and a number to call if they preferred colonoscopy or flexible sigmoidoscopy. Those calling were told that most insurance plans cover colonoscopy and flexible sigmoidoscopy, and were instructed to call their health care team directly to arrange this (in the parent trial Assisted and Navigated patients received additional support for arranging and completing endoscopy, but not in the year 3 study). The second mailing, sent 2 weeks later, included a letter with the date of their last FOBT test (or no FOBT in the last 2 years), the importance of completing FOBT annually, a FOBT kit with pictographic instructions, and a pre-stamped return envelope for mailing it to the laboratory. Study FOBTs were processed by the Group Health laboratory and documented in the EHR. Those not completing a study or clinic FOBT received a mailed reminder.

In December 2011, Group Health and the SOS study switched from the guaiac-based 3-card Hemoccult SENSA (Beckman Coulter Inc, Brea, Calif) to a single-sample quantitative fecal immunochemical test (FIT), OC FIT-CHEK (Polymedco Inc, Cortlandt Manor, NY). Patients in the Continued arm received mailed SENSA stool kits before the change and OC-FIT-CHEK after, with pictographic and other materials updated accordingly. Group Health also maintains a registry for tracking positive FOBTs and the completion of follow-up colonoscopy, and reminds providers if colonoscopy is not completed. The study also employs a nurse navigator who contacts providers if documentation of completed on noncompleted colonoscopies is incomplete and/or patients need assistance completing colonoscopy. We have previously reported very high follow-up rates after positive FOBTs.¹²

Measures

Primary and secondary outcomes were reported as binary measures of CRC testing completion, and were collected from automated EHR data. The primary outcome was the completion of any CRC testing during year 3 (FOBT, colonoscopy, or flexible sigmoidoscopy). Patients due for CRC screening were offered a choice of any of these three tests in the letter before mailing the fecal kit and as part of usual care. We also evaluated completion of FOBT or colonoscopy as secondary outcomes, but not flexible sigmoidoscopy because only 3 study participants chose this test. Age, sex, tobacco history, and completion of FOBT in years 1 or 2 of the parent study were obtained from EHR data. Patient self-reported data were obtained at the time of enrollment in the parent study and included race/ethnicity, education, and marital status. Race accounting for mixed race and ethnicity was categorized as follows: white and not Hispanic, Hispanic, black and not Hispanic, Asian and not Hispanic or black, or other. Patient self-reported health was measured using the first item of the 36-Item Short Form Health Survey (“In general, would you say your health is excellent, very good, good, fair, poor?”).¹³

Statistical Analysis

Analyses followed a modified intent-to-treat approach. All randomized patients were included in the analysis except for randomized participants who were subsequently found to have disenrolled from the health plan (152 patients) or who had a colonoscopy (25 patients) or positive FOBT (2 patients) before randomization (Fig. 1). These participants were randomized in error because these ineligibility events were not available in the EHR data until after randomization had occurred due to data system lags for disenrollment and procedure data. For binary screening outcomes (completion of CRC screening during year 3), we reported the number and percentage of patients screened by randomization group. Generalized linear models with a logit link and robust sandwich variance estimator were used to estimate predictive margins, which are estimated probabilities of the outcome (screening) adjusted across the covariate distribution in the analysis sample. We used the predictive margins to estimate the difference in screening rates (95% confidence interval [95% CI]) between the Stopped arm and the Continued arm. Models were adjusted for age, sex, race/ethnicity, and education. Secondary analyses explored whether the intervention effect differed by participant characteristics. We tested for effect modification by age, sex, race, marital status, education, smoking status, and self-reported health status at the time of entry to the parent study. We also explored effect modification by screening completion during the first 2 years of the parent trial; number of primary care visits in year 3 (face-to-face visit with a physician, nurse practitioner, or physician assistant); and whether 1 of these visits was a preventive, well-care visit. Interaction terms between these variables and the intervention groups were included in the multivariable models, and Wald tests were used to assess the overall significance of effect modification. Analyses were performed using Stata statistical software (version 12.0; StataCorp, College Station, Tex).

We estimated that 2500 patients from the main study would be due for screening and eligible for re-randomization at year 3. This sample size provided 80% power to detect a 6% difference in screening between the Continued and Stopped arms during year 3. Power analyses assume 10% loss to follow-up, 2-sided tests, and a 5% type I error rate.

RESULTS

Study Participants

In year 1 of the SOS trial, 3508 patients aged 50 to 75 years were randomized to receive 1 of 3 stepped interventions (Fig. 1). Of these, 2208 patients were still eligible for participation in the year 3 study, with 828 patients excluded because of colonoscopy before year 3, 318 patients excluded because they disenrolled, 70 patients excluded because they opted out of continued study participation, 28 excluded because they were aged >75 years, 17 patients had died, 5 were diagnosed with CRC, 16 had a positive FOBT in years 1 or 2 and had contraindications or declined follow-up colonoscopy, and 18 patients were excluded for other reasons (such as a diagnosis of ulcerative colitis). Of the 2208 eligible patients, 1106 were randomized to the Stopped arm and 1102 patients were randomized to the Continued arm. Patients in the Stopped and Continued arms were distributed equally by age, sex, race and ethnicity, health status, smoking status, and FOBT testing patterns in years 1 and 2 (Table 1).

TABLE 1.

Characteristics by Randomization Group

Characteristics	Stopped Arma	Continued Arma
	N=1106 No. (%)	N=1102 No. (%)
Age group at baseline, yb
50–64	962 (87.0)	936 (84.9)
65–73	144 (13.0)	166 (15.1)
Female sexb	595 (53.8)	592 (53.7)
Race/ethnicityb,c
Hispanic	32 (2.9)	34 (3.1)
Non-Hispanic
Black	64 (5.8)	58 (5.3)
Asian	64 (5.8)	56 (5.1)
White	874 (79.5)	873 (79.6)
Other	66 (6.0)	76 (6.9)
General healthb,c
Excellent/very good	688 (62.3)	716 (65.0)
Good	348 (31.5)	313 (28.4)
Fair/poor	68 (6.2)	72 (6.5)
Married or living with partnerb	822 (74.5)	822 (74.7)
Highest educational level attainedb,c
≤High school graduate or GED	159 (14.4)	158 (14.4)
Some college	333 (30.1)	344 (31.3)
≥Bachelor’s degree	614 (55.5)	598 (54.4)
Smoking statusb,c
Current	121 (11.2)	127 (11.7)
Former	315 (29.2)	278 (25.7)
Never	643 (59.6)	679 (62.6)
Preventive care visit in y 3
No	857 (77.5)	848 (77.0)
Yes	249 (22.5)	254 (23.1)
Primary care visits in y 3
No visits	320 (28.9)	335 (30.4)
1 visit	333 (30.1)	312 (28.3)
≥2 visits	453 (41.0)	455 (41.3)
FOBT screening in y 1 and 2 of parent study
None	299 (27.0)	293 (26.6)
1 y only	257 (23.2)	278 (25.2)
Both y	550 (49.7)	531 (48.2)
Parent study randomization group
Automated	381 (34.5)	377 (34.2)
Assisted	373 (33.7)	370 (33.6)
Navigated	352 (31.8)	355 (32.2)

Abbreviations: FOBT, fecal occult blood test; GED, General Educational Development.

^aAll patients received active interventions in years 1 and 2 of the parent study. In year 3, patients still eligible for colorectal cancer screening (eg, those with no prior colonoscopy or no colorectal cancer diagnosis) were randomized to Stopped or Continued interventions.

^bObtained at the time of enrollment in year 1 (parent study).

^cMissing data: race (11 patients), general health (3 patients), marital status (4 patients), education (2 patients), and smoking status (45 patients).

Primary Analyses

Patients in the Continued arm had significantly higher CRC testing completion in year 3 compared with those in the Stopped arm (53.3% vs 37.3%; adjusted net difference, 15.6% [P<.001]) (Table 2). This difference was entirely due to the significantly greater completion of FOBT (adjusted net difference, 18% [P<.001]). Patients in the Continued arm completed significantly fewer colonoscopies than those in the Stopped arm (7.1% in the Stopped arm and 4.8% in the Continued arm; adjusted net difference, −2.3% [P =.03]). No serious adverse events occurred as a result of study participation (hospitalization or deaths related to screening or follow-up diagnostic testing).

TABLE 2.

Completion of CRC Screening Tests by Study Arm in Year 3

	Screened in Year 3		Adjusted Differences Between Groups, %a
	Stopped Arm N=1106	Continued Arm N=1102	Continued Minus Stopped
	No. (%)	No. (%)	Difference (95% CI)	P
Primary outcome
Any CRC screening testb	412 (37.3)	587 (53.3)	15.6 (11.9 to 19.3)	<.001
Secondary outcomes
FOBT	332 (30.0)	533 (48.4)	18.0 (15.4 to 20.6)	<.001
Colonoscopyc	78 (7.1)	53 (4.8)	−2.3 (−4.2 to −0.4)	0.02

Abbreviations: 95% CI, 95% confidence interval; CRC, colorectal cancer; FOBT, fecal occult blood test; NA, not applicable.

^aAdjusted for age, sex, race, and education.

^bFOBT, flexible sigmoidoscopy, or colonoscopy.

^cAs first test only.

Subgroup Analyses

The effectiveness of Continued interventions on screening in year 3 differed significantly by screening behavior in years 1 and 2 (test for interaction P =.001). The adjusted net difference in screening uptake was 26.7%, 11.0%, and 1.9%, respectively, in patients in the Continued versus Stopped arms who completed FOBT in both years 1 and 2, only 1 year, and neither year (Table 3). Patients in the Continued arm who completed FOBT testing in year 1 or 2 were more than twice as likely (adjusted relative risk [RR], 2.4; 95% CI, 1.7–3.0) and those completing FOBT in both years were 4 times as likely (adjusted RR, 4.0; 95% CI, 2.8–5.2) to complete CRC screening in year 3.

TABLE 3.

Completion of CRC Screening Among Patient Subgroups^a

	Screened in Year 3		Adjusted Differences Between Groupsb
	Stopped Arm N=1106	Continued Arm N=1102	Continued Minus Stopped
Outcome: Any Screen, Year 3	No. (%)	No. (%)	Difference (95% CI)	Within Subgroup Pc	Overall Pd
FOBT in y1 or y2					<.001
No FOBT in y1 or y2	N=299 48 (16.1)	N=293 53 (18.1)	1.9 (−5.8 to 9.6)	.64
Only 1 FOBT in y1–y2	N=257 87 (33.9)	N=278 124 (44.6)	11.0 (4.7 to 17.2)	.001
FOBT in both y1 and y2	N=550 277 (50.4)	N=531 410 (77.2)	26.7 (20.4 to 33.0)	<.001
Main study intervention group					.27
Automated	N=381 128 (33.6)	N=377 200 (53.1)	19.0 (12.4 to 25.7)	<.001
Assisted	N=373 138 (37.0)	N=370 195 (52.7)	15.1 (8.2 to 22.0)	<.001
Navigated	N=352 146 (41.5)	N=355 192 (54.1)	12.4 (6.3 to 18.4)	<.001
Age at baseline, y					.93
50 to <65	N=962 335 (34.8)	N=936 472 (50.4)	15.7 (11.2 to 20.1)	<.001
65–73	N=144 77 (53.5)	N=166 115 (69.3)	15.7 (5.8 to 25.6)	.002
Sex					.68
Male	N=511 181 (35.4)	N=510 258 (50.6)	14.7 (10.1 to 19.3)	<.001
Female	N=595 231 (38.8)	N=592 329 (55.6)	16.4 (10.9 to 21.8)	<.001
Education					.90
≤Some college	N=492 172 (35.0)	N=502 255 (50.8)	15.7 (10.8 to 20.6)	<.001
≥Bachelor’s degree	N=614 240 (39.1)	N=598 332 (55.5)	15.5 (10.6 to 20.4)	<.001
Marital status					.15
Married or living with partner	N=822 308 (37.5)	N=822 454 (55.2)	17.5 (13.9 to 21.0)	<.001
Other	N=281 103 (36.7)	N=279 133 (47.7)	10.2 (1.1 to 19.3)	.03
Race					.84
White, non-Hispanic	N=874 331 (37.9)	N=873 475 (54.4)	16.1 (11.6 to 20.5)	<.001
Black	N=64 22 (34.4)	N=58 27 (46.6)	13.7 (−1.8 to 29.2)	.08
Asian	N=64 25 (39.1)	N=56 34 (60.7)	20.4 (6.1 to 34.7)	.01
Hispanic	N=32 7 (21.9)	N=34 10 (29.4)	6.8 (−10.0 to 23.7)	.43
Other	N=66 25 (37.9)	N=76 37 (48.7)	11.5 (−1.7 to 24.7)	.09
Self-reported health					.003
Excellent/very good	N=688 290 (42.2)	N=716 404 (56.4)	13.9 (8.7 to 19.1)	<.001
Good	N=348 95 (27.3)	N=313 155 (49.5)	21.8 (17.44 to 26.2)	<.001
Fair/poor	N=68 26 (38.2)	N=72 27 (37.5)	−0.8 (−14.2 to 12.7)	.91
Smoking status at baseline					.33
Current	N=121 29 (24.0)	N=127 59 (46.5)	21.8 (8.4 to 35.2)	.001
Former	N=315 127 (40.3)	N=278 159 (57.2)	17.1 (10.0 to 24.1)	<.001
Never	N=643 253 (39.4)	N=679 361 (53.2)	13.5 (8.1 to 18.8)	<.001
Preventive care visit in y3					0.16
No	N=857 272 (31.7)	N=848 418 (49.3)	17.2 (12.5 to 21.9)	<.001
Yes	N=249 140 (56.2)	N=254 169 (66.5)	10.2 (3.3 to 17.1)	0.004
Primary care visits in y3					.06
No visits	N=320 66 (20.6)	N=335 145 (43.3)	21.4 (15.2 to 27.7)	<.001
1 visit	N=333 123 (36.9)	N=312 166 (53.2)	16.6 (7.3 to 25.9)	.001
≥2 visits	N=453 223 (49.2)	N=455 276 (60.7)	11.3 (6.0 to 16.5)	<.001

Abbreviations: 95% CI, 95% confidence interval; CRC, colorectal cancer; FOBT, fecal occult blood test.

^aFOBT, flexible sigmoidoscopy, or colonoscopy.

^bAdjusted for age, sex, race, and education.

^cThe within-subgroup P value tests the significance of an intervention effect within the subgroup.

^dThe overall P value tests the significance of differences in the intervention effect across subgroups.

Patients with at least 1 primary care or preventive care visit had higher CRC testing rates in year 3 than those without visits, regardless of intervention group. The Continued intervention led to a significant increase in CRC testing completion in all primary care visit subgroups, but trended toward greater effectiveness among those with fewer visits (test for interaction P =.06).

The effectiveness of the Continued intervention did not appear to vary significantly by age, sex, education, marital status, or baseline smoking status. There were some differences observed by race, with smaller net increases in black individuals and especially Hispanic individuals, who also had lower screening rates compared with other groups, but these differences were not found to be statistically significant. Patients reporting fair or poor general health at baseline (year 1) in the Continued arm had CRC screening rates similar to those in the Stopped arm (37.5% vs 38.2%; P =.91). Patients in the Continued arm with excellent/very good or good self-reported health had significant incremental increases in CRC screening compared with those in the Stopped arm (P<.001).

DISCUSSION

The SOS trial demonstrated that continued interventions increased adherence to CRC screening. Specifically, it was found that among patients eligible for screening in year 3, those randomized to a continued centralized EHR-linked mailed CRC screening program were significantly more likely to be adherent to CRC screening than those randomized to the Stopped (usual care) interventions. In the current study, usual care included EHR reminders of overdue CRC screening, the provision of FOBT kits at the time of primary care visits, outreach calls, and/or mailed FOBT. Despite the intensity of usual care, a centralized FOBT kit mailing program led to a nearly 50% higher rate of CRC screening completion in year 3 (adjusted RR, 1.42; 95% CI, 1.28–1.55). These findings have important implications for health systems that want to longitudinally maintain high levels of screening adherence.

It is interesting to note that patients in the Stopped arm were found to be modestly but significantly more likely to complete a colonoscopy than those in the Continued arm (adjusted net difference for Stopped vs Continued, −2.3% [P =.03]). Physicians would be more likely to have discussions about CRC screening with patients who were overdue for screening, with more patients in the Stopped arm being overdue. Prior studies have reported that physicians are more likely to recommend screening colonoscopy.^14,15 Thus, offering a centralized mailed FOBT program only might have an unintended consequence of decreasing colonoscopy screening. In years 1 and 2 of the SOS study, intervention patients who received mailings plus additional telephone assistance had colonoscopy rates similar to usual care.⁷ Larger studies with longer follow-up are needed to determine whether reduced colonoscopy rates in the intervention group lead to negative clinical outcomes even when the mailed program increased CRC screening uptake overall.

We know of no prior RCT aimed at increasing CRC screening adherence in health care settings that has provided 3 rounds of mailed FOBT. The Minnesota Colon Cancer Control Study (1976–1992), one of the RCTs that established the efficacy of FOBT in decreasing CRC mortality and incidence,^16,17 reported adherence rates of >70% to 11 mailed FOBT screening rounds.¹⁸ However, the study was implemented in a research setting and non-compliant patients received repetitive calls. Large population-based screening programs provide additional information regarding long-term adherence to multiple rounds of mailed FOBT. In Scotland, among 251,578 eligible adults, adherence rates to biennial guaiac-based FOBT of 55% in year 1, 45% in year 2, and 48% in year 3 were reported, suggesting a drift in adherence from the initial year.¹⁹ In the Netherlands, adherence rates among 7501 eligible participants ranged from 63% to 68% for 3 rounds of biennial FIT.²⁰ Although these studies provide information regarding the level of adherence over time that may be achievable by a mailed FOBT program, they do not provide information concerning program comparative effectiveness compared with clinic-based CRC screening activities.

In the current study, although patients in the Continued arm had higher CRC screening rates than those in the Stopped arm, only 53% of participants completed CRC screening in year 3. These CRC screening rates are lower than rates reported by health plans as part of the Health Effectiveness Data and Information Set (HEDIS).²¹ However, it is important to note that the year 3 patients in the current study represent a subset of SOS participants. In year 1 of the SOS study, 68.3% of the active intervention participants completed CRC testing. However, by year 3, many patients had completed colonoscopy (828 of 3508 patients of the original active intervention cohort; 23.6%) and were no longer eligible for continued interventions. In contrast, patients who had never undergone screening continued to be eligible. Therefore, year 3 participants may represent a population that is more resistant to screening. In addition, in year 3 of the SOS study, Group Health switched from the 3-sample, guaiac-based SENSA to a 1-sample FIT. We previously demonstrated in another study performed at Group Health that adherence to FIT is greater than adherence to SENSA.²² Thus, screening rates might have been higher in both arms if FIT had been available throughout year 3.

Patients adhering to FOBT screening in years 1 and 2 were significantly more likely to screen in year 3, with no intervention effect noted among those not previously responding to mailings. In England, Lo et al²³ reported adherence rates of 66% in the third biennial round. Participation in the third round was found to be highest among individuals who had screened in both years 1 and 2 versus those not screening in either year (94.5% vs 14.6%). In the United States, Baker et al⁵ randomized community clinic, primarily Hispanic, Spanish-speaking patients who in the prior year completed mailed FIT screening to receive either a mailed FIT kit program that included free kits, low literacy instructions, paid postage, automated text and telephone reminders, and navigation or to usual care. Greater than 82% of the active intervention participants in the study by Baker et al completed FIT testing compared with 53% in the current study. Several factors may account for this difference. In the study by Baker et al,⁵ kits were mailed only to individuals completing FIT within the prior year, whereas we mailed kits to patients who had both completed and not completed FIT testing within the previous 2 years. In the current study, patients in the active intervention arm who completed FOBT testing in both year 1 and 2 had screening rates similar to those reported in the study by Baker et al,⁵ with >77% completing testing in year 3. In addition, the intervention in the study by Baker et al⁵ focused only on fecal testing, whereas the current study included screening choice. Many patients from the parent study chose colonoscopy in either year 1 or 2 and were no longer due for CRC screening and therefore not eligible for the year 3 study. Those remaining eligible may have been less willing to complete any type of CRC screening test.

We know of no prior study that has evaluated the interaction between primary care visits and a centralized mailed FOBT program. Individuals with a primary care or preventive visit in year 3 had higher CRC screening rates than those without a visit. Previously in the same health care organization Fenton et al reported that receipt of a preventive care visit was found to be strongly associated with repeating FOBT annually over 2 years.²⁴ At a community health center network that served predominantly Spanish-speaking, uninsured patients, Liss et al found that having ≤3 primary care visits was strongly associated with repeat FOBT 9 to 18 months after the index FOBT.²⁵ In the current study, among patients with primary care visits and in the subgroup with a preventive visit, continued interventions still led to a significant incremental increase in CRC screening. However, the intervention effect magnitude was greater in the group of patients with fewer or no primary care visits.

Other patient characteristics including age, sex, and education were found to have little or no impact on intervention effectiveness, with the exception of self-reported health, for which the intervention was not effective among individuals self-reporting fair to poor health. In Australia, Duncan et al²⁶ reported that younger adults were significantly less likely to participate in any of the 3 rounds of a mailed FIT screening program. The results of the current study provide evidence that mailed FOBT programs can be just as effective in younger patients as in older patients, but younger patients simply undergo screening less often. Screening rates also were found to be lower among certain minority racial and ethnic groups, but small numbers limited the interpretation of subgroup differences. Duncan et al also examined psychosocial constructs associated with long-term CRC screening adherence.²⁶ Individuals with lower CRC screening self-efficacy (less confidence in their ability to complete screening) also were more likely to refuse screening. We previously reported psychosocial predictors of repeat screening in year 2 of our study.¹⁰ We found that higher self-efficacy scores had a modest but significant effect on the completion of FOBT testing in both years. However, enabling patients to complete CRC screening by offering them repeated mailed FOBT interventions (the study intervention) was found to be the strongest predictor of completion of FOBT testing in year 2.

The current study has limitations. The patients reported herein all had health insurance, were mostly white and non-Hispanic, and had higher levels of education than the surrounding community and the US population and therefore our findings may be less generalizable to other populations. Participants also provided verbal consent in year 1, and received another letter in year 3 with an option to call and opt out of further participation.⁹ Therefore, the participants in the current study were “volunteers” who may be more responsive to interventions. Our estimate of screening rates may be artificially high because some FOBTs and colonoscopies were performed for indications other than screening. Although this may affect our absolute estimates of screening, it would be unlikely to have an effect on cross-group comparisons. Conversely, the magnitude of increase in CRC uptake might have been higher in patients in the Continued compared with the Stopped interventions in a health care setting that had not implemented the PCMH. The strengths of the current study include the ability to capture both claims and EHR data, with prior validations indicating that we are able to capture almost all data correctly,⁸ and the ability to compare the effectiveness of clinic-based screening efforts with this plus a centralized mailed FOBT program. We previously reported that our mailed FOBT program was inexpensive and cost-saving compared with clinic-based efforts to increase CRC screening.^6,7

To our knowledge, the current study is the first RCT to test the effectiveness of an ongoing program to increase CRC screening uptake over 3 years. We found that a mailed FOBT program was more effective than usual care, even in a setting in which patients all had insurance and received care in a health care system that offers patients many other opportunities to get screened. The benefit was confined to those patients who previously completed a screening test. Different types of interventions are likely to be necessary for those individuals consistently refusing a mailed FOBT program and should be the focus of future research.