Effect of Sequential or Active Choice for Colorectal Cancer Screening Outreach

Key Points Question Does the offering of choice between colonoscopy and fecal immunochemical testing (FIT) change participation in screening outreach? Findings In this 3-arm pragmatic randomized clinical trial of 438 patients, there was no statistically significant increase in screening when offering the choice of FIT compared with colonoscopy only, but fewer patients selected colonoscopy in the choice arms. Meaning Offering the choice of FIT in a colonoscopy outreach program did not substantially increase screening participation, but the framing of choice altered patient decision making.


Initial Protocol 23 24
Abstract 25 A 3-arm randomized trial assessing whether the rate of completion of colorectal cancer screening 26 is increased when patients receive a sequential choice in screening options (colonoscopy followed 27 by Fecal Immunochemical Testing (FIT)) or an active choice (FIT or colonoscopy offered 28 together) versus colonoscopy alone. The targeted population is patients within the University City 29 and Valley Forge Community Care Associates (CCA) practices at the University of Pennsylvania 30 Health System. 31 32 Study Instruments 33 The primary endpoint being evaluated is the rate of participation in colorectal cancer screening between 34 the two intervention arms (Sequential Choice, Active Choice) versus the control arm (colonoscopy only). 35 The FIT is a well validated tool for colorectal cancer screening and is one of the screening modalities 36 recommended by the USPSTF. The control arm will be sent a letter inviting them to schedule screening 37 colonoscopy directly through the Gastroenterology Call Center. If not scheduled within 4 weeks, subjects 38 will receive a mailed reminder to call to schedule. The Sequential Choice arm will be sent the same initial 39 outreach as the control arm. If not scheduled within 4 weeks, subjects will receive a mailed reminder 40 including the call center number as well as the option to complete a FIT kit mailed along with the 41 reminder. The Active Choice arm of the study will be sent a letter offering the choice of Colonoscopy or 42 completion of the FIT kit included with the initial mailing. If subjects in this group have not either 43 scheduled colonoscopy nor completed FIT within 4 weeks, they will receive a mailed reminder to call to 44 schedule colonoscopy or to complete the original mailed FIT. 45 46 A sub-sample of 90 subjects will be called to complete a questionnaire over the phone 4 months after 47 initial outreach was mailed. The subjects will be asked to confirm their eligibility (e.g. that they had not 48 had CRC screening within the USPSTF CRC screening guidelines) and provide additional demographic and 49 socioeconomic information so that we can better understand what populations, if any, may have 50 differential response rates. We will also ask them about their perception of the impact and design of 51 CRC screening outreach. Demographic and socioeconomic questions are modified from demographic 52 questions on the Behavioral Risk Factor Surveillance System survey, administered by the Centers for 53 Disease Control and Prevention. See attached sub-sample questionnaire. 54 55 Group Modifications 56 For subjects in the Control (Colonoscopy only) arm of the study, the post outreach phone questionnaires 57 will not include questions regarding mailed FIT. 58 59 60 Method for Assigning Subjects to Groups 61 Subjects will be randomly assigned Study ID numbers and then randomized to one of three study arms 62 stratified by the two practice locations (University City and Valley Forge) using a computer-generated 63 randomization algorithm. The research coordinator will record the randomization assignments on a 64 master list which will be maintained by the research coordinator on a password protected computer in a 65 locked office. The research coordinator and research assistants will assemble the mailings based on this 66 master list. 67 68 69 70 71 Administration of Surveys and/or Process 72 90 subjects will be randomly selected for the questionnaire. We anticipate the post outreach 73 questionnaires to take 10 minutes to complete over the phone. The research staff will make no more 74 than three attempts to speak directly with the subject. Based on a previous project where we reached 75 about 50% of patients via phone call, we anticipate reaching approximately 45 subjects (15 in each 76 arm) to complete this sub-sample questionnaire. 77 78 Administration of Surveys 79 All subjects will complete a baseline and post-intervention survey. The baseline survey will collect basic 80 demographics and baseline medication adherence and blood pressure monitoring frequency and be 81 conducted over the phone after the participant has been consented. The post-intervention survey will 82 collect similar adherence and monitoring information, as well as qualitative data regarding patient 83 perceptions about the interventions. The post-intervention survey will be completed at the in-person 4 84 month visit. These surveys should take no more than 15 minutes to complete. Demographic and socio-85 economic questions are modified from demographic questions on the Behavioral Risk Factor 86 Surveillance System survey, administered by the Centers for Disease Control and Prevention. 87 88 Objectives 89 1.1 Objectives 90 The specific aim of this study is to assess the effectiveness of two different mailed outreach activities 91 (sequential choice of colonoscopy then FIT, active choice of colonoscopy and FIT) versus colonoscopy 92 only in increasing participation in CRC screening. 93 94 1.2 Primary Outcome Variable 95 The primary outcome is CRC screening completion (FIT or colonoscopy) within 4 months of initial 96 outreach 97 98 1.3 Secondary Outcome Variable(s) 99 The secondary outcome is the choice of screening test between FIT and colonoscopy. 100 101 Additional outcomes include demographic and socioeconomic characteristics of subjects who 102 participate, as well as exploratory qualitative data regarding experience with the different outreach 103 methodologies. Additional outcome is the percentage of FIT screening results that are positive, 104 percentage of those positive FIT tests that receive follow-up diagnostic colonoscopy, and percentage of 105 colonoscopies that find adenomas, advanced adenomas, and cancer. 106 107 Colonoscopy is the predominant form of screening in this country due to perceived effectiveness by 116 providers, but it entails a significant cost in time, resources, and perceived discomfort. The fecal 117 immunochemical test (FIT) is an attractive screening option as it is less invasive than traditional lower 118 endoscopy and can be mailed to patients to complete at home. A recent study has shown that offering 119 the choice of colonoscopy or stool-based testing in a clinic setting increases screening rates, but receipt 120 of colonoscopy has better durability since stool-based testing has to occur every year. 121 122 In this study, we will be using population-based outreach screening to evaluate the feasibility of a 123 proactive screening program to promote fecal immunochemical testing (FIT) and/or colonoscopy 124 through mailed outreach by leveraging principles of behavioral economics. We know that mailed FIT 125 outreach circumvents the need for an office visit and eliminates friction in the screening process, since 126  Approximately 900 potentially eligible subjects will be identified via a data abstraction by Penn Data 140 Store. Through other projects in primary care practices at Penn Medicine we found the accuracy rate of 141 the EMR algorithm to be approximately 75%. As such, we anticipate we will have enough patients to 142 enroll 423 subjects (and randomize 141 into each arm). We estimate a base return rate for the 143 colonoscopy only (control) arm to be 5%, and we will consider a meaningful increase in response rate to 144 be 10 percentage points for both the sequential choice and active choice arms as compared to control. 145 This will be sufficient sample size to detect a 10 percentage point increase in response rate using a two-146 tailed chi-square test with 80% power and a 5% level of significance. 147 148 1.2 Data analysis 149 The primary outcome is CRC screening completion (FIT or colonoscopy) within 4 months of initial 150 outreach. We will conduct a chi-square analysis using Stata to compare arms 2 and 3 to arm 1 separately 151 using intent-to-treat protocol. We will also compare arms 2 and 3 as a secondary analysis. We will 152 quantitatively analyze the choice of screening test and evaluate the survey results by study arm. As 153 exploratory analyses, we will evaluate response by practice location, age, gender, race/ethnicity, and 154 income at the level of zip code. 155 156 Analysis will be conducted by blinded members of the research team at least four months after the last 157 FIT is mailed. 158 159 Study Design 160 1.1 Design 161 Randomized: Subjects will be randomly assigned Study ID numbers and then randomized to one of 162 three study arms stratified by practice location using a computer-generated randomization algorithm. 163 The research coordinator will record the randomization assignments on a master list which will be 164 maintained on a password protected computer in a locked office. The research coordinator and research 165 assistants will assemble the mailings based on this master list. 166 167 Blinding: The investigators will be blinded to the randomization assignment. The research coordinator 168 and research assistants will be unblinded. Approximately 900 potentially eligible subjects will be identified via a data abstraction by Penn Data 231 Store. Through other projects in primary care practices at Penn Medicine we found the accuracy rate of 232 the EMR algorithm to be approximately 75%. As such, we anticipate we will have enough patients to 233 enroll 423 subjects (and randomize 141 into each arm). We estimate a base return rate for the 234 colonoscopy only (control) arm to be 5%, and we will consider a meaningful increase in response rate 235 to be 10 percentage points for both the sequential choice and active choice arms as compared to 236 control. 237 This will be sufficient sample size to detect a 10 percentage point increase in response rate using a two-238 tailed chi-square test with 80% power and a 5% level of significance. 900 potentially eligible subjects will be identified via a data abstraction by Penn Data Store using a data 285 query algorithm that identifies patients who meet the inclusion criteria. Chart review will be conducted 286 to source 423 eligible participants from this pool of subjects. 287 288

Subject Compensation 289
Participants will not be financially compensated for their participation. 290 291

Procedures 292
Screening -Phase 1: We will submit a data request of patients from the CCA practices of University 293 City and Valley Forge from the Penn DataStore, based on an EMR algorithm that determines guideline-294 concordant colorectal cancer screening within PennChart. We estimate approximately 900 screen-295 eligible 296 patients will be identified through this EMR query, 423 of whom will be enrolled and randomized into 297 the three arms. We anticipate conducting chart reviews for two months. During these chart reviews, the 298 Research Assistants and the Research Coordinator will review the electronic medical record charts in 299 PennChart (EPIC) to review each of the eligible patients pulled from Penn Data Store 300 to confirm study eligibility. 301 302 Randomization -Phase 2: Subjects will be randomly assigned Study ID numbers and then randomized 303 with stratification to one of three arms using a computer-generated randomization algorithm. The 304 research coordinator will record the randomization assignments on a master list which will be 305 maintained by the research coordinator on a password protected computer in a locked office. The 306 research coordinator and research assistants will assemble the mailings based on this master list. 307 308 Outreach & Follow-up -Phase 3: Initial outreach letters will be sent to subjects. In the control arm the 309 outreach will include a phone number to the VIP colonoscopy scheduling center. For the sequential 310 choice arm, subjects will receive the same message as those in the control arm, indicating they are 311 overdue for screening and including a phone number to the VIP screening scheduling center. For the 312 active choice arm, subjects will receive both a mailed FIT kit and the option to call and schedule 313 colonoscopy at the same time. After four weeks, a reminder letter will be sent to all subjects in all three 314 arms who did not either schedule a colonoscopy appointment or complete and return a FIT kit. Subjects 315 in the Control arm will receive a reminder letter that includes the same phone number to schedule 316 colonoscopy. Subjects in the sequential choice arm will receive a second letter containing the phone 317 number to the VIP screening scheduling center as well as a FIT kit. The active choice arm will receive a 318 reminder that includes both the phone number to the VIP screening scheduling center and a reminder 319 that they may alternatively complete the FIT that was initially mailed. The Research Coordinator and 320 Research Assistants will be responsible for assembling the mailings. FIT kits will include a tube in which 321 to deposit the stool sample, directions on how to collect and mail the sample, a letter about CRC 322 screening, a lab requisition form, and a pre-paid return envelope. 323 324 Sub-sample Questionnaire -Phase 5: 325 A random subsample of 90 subjects will be selected to complete a questionnaire at least 4 months after 326 they received the mailing. Through this questionnaire, the subjects will be asked to confirm their 327 eligibility (e.g. that they had not had CRC screening within the USPSTF CRC screening guidelines) 328 and provide additional demographic and socioeconomic status information as well as qualitative 329 experience with the outreach materials and approach so that we can better understand what  330  populations,  331 if any, may be more likely impacted by different types of outreach modalities when engaging in CRC 332 screening. We anticipate these questionnaires to take 10 minutes to complete over the phone. The 333 research staff will make no more than three attempts to speak directly with the subject. 334 335 Analysis Plan 336 1.1 Power and Sample Size 337 Approximately 900 potentially eligible subjects will be identified via a data abstraction by Penn Data 338 Store. Through other projects in primary care practices at Penn Medicine we found the accuracy rate of 339 the EMR algorithm to be approximately 75%. As such, we anticipate we will have enough patients to 340 enroll 423 subjects (and randomize 141 into each arm). We estimate a base return rate for the 341 colonoscopy only (control) arm to be 5%, and we will consider a meaningful increase in response rate to 342 be 10 percentage points for both the sequential choice and active choice arms as compared to control. 343 This will be sufficient sample size to detect a 10 percentage point increase in response rate using a two-344 tailed chi-square test with 80% power and a 5% level of significance. 345 346 1.2 Data analysis 347 The primary outcome is CRC screening completion (FIT or colonoscopy) within 4 months of initial 348 outreach. We will conduct a chi-square analysis using Stata to compare arms 2 and 3 to arm 1 separately 349 using intent-to-treat protocol. We will also compare arms 2 and 3 as a secondary analysis. We will 350 quantitatively analyze the choice of screening test and evaluate the survey results by study arm. As 351 exploratory analyses, we will evaluate response by practice location, age, gender, race/ethnicity, and 352 income at the level of zip code. 353 354 Analysis will be conducted by blinded members of the research team at least four months after the last 355 FIT is mailed. 356

357
Data Confidentiality 358 Paper-based records will be kept in a secure location and only be accessible to personnel involved in the 359 study. Computer-based files will only be made available to personnel involved in the study through the 360 use of access privileges and passwords. Prior to access to any study-related information, personnel will 361 be required to sign statements agreeing to protect the security and confidentiality of identifiable 362 information. Wherever feasible, identifiers will be removed from study-related information. Precautions 363 are in place to ensure the data is secure by using passwords and encryption, because the research 364 involves web-based surveys. 365 366 Subject Confidentiality 367 Information about study subjects will be kept confidential and managed according to the requirements 368 of the Health Insurance Portability and Accountability Act of 1996 (HIPAA). All PHI will be 369 maintained on UPHS servers. Source documents are maintained in PennChart. No source documents 370 will be printed or maintained in paper form at the study site. Data from PennChart will be recorded in 371 Penn Medicine's REDCap system. The investigator and study team (which includes the research 372 coordinator, and research assistants) will have access to PHI within PennChart and REDCap. We will 373 label all PHI within REDCap as identifiable information so that de-identified exports are possible. All 374 reports that include identifiable information will be stored on the Innovation Center secure drive, 375 maintained behind the UPHS firewall. Once data analysis and manuscripts have been published, the 376 databases will be removed from REDCap and the data will be de-identified on the secure drive. This 377 deidentified dataset will be stored for up to five years after analysis is complete and manuscripts have 378 been published. Once analysis is completed and any manuscripts are published, we will retain PHI no 379 longer than seven years in accordance with government regulations, applicable policies, and 380 institutional requirements. 381 382 Database Security/Protection Against Risk 383 To assure that patient, physician and other informant confidentiality is preserved, individual identifiers 384 (such as name and medical record number/physician billing identifier) are stored in a single password 385 protected system that is accessible only to study research, analysis and IT staff. This system is hosted on 386 site at The University of Pennsylvania (UPenn) and is protected by a secure firewall. Once a participant is 387 in this system, they will be given a unique study identification number (ID). Any datasets and computer 388 files that leave the firewall will be stripped of all identifiers and individuals will be referred to by their 389 study ID. The study ID will also be used on all analytical files. 390 The initial patient information collected for screening and recruitment will consist of name, address, 391 phone number, dates, medical records numbers and health plan account numbers. This information will 392 come from Electronic Chart reviews. 393 394

Sensitive Research Information 395
This Research does not involve collection of sensitive information about the subjects that should be 396 excluded from the electronic medical record. 397 398 Subject Privacy 399 We will only interact with the subsample of subjects with which we plan to call to conduct a follow-up 400 questionnaire. With these subjects, we will conduct phone calls in a private area. When we call subjects, 401 we will confirm the identify before administering the questionnaire. We will not be interacting with 402 subjects in person. 403 The risks associated with this study are no more than minimal. There is the potential risk of breach of 465 confidentiality. We will minimize this risk by using de-identified information whenever possible and by 466 maintaining all identifiable information on a secure drive and/or in a HIPAA-compliant system (e.g. 467

Data
REDCap). There is also the risk of psychological harm associated with being screened for cancer. We 468 will minimize this risk by communicating the results of the screening test to the subject in a timely 469 fashion and facilitating the scheduling of diagnostic testing if the screening test is positive (as is usual 470 practice for screening outreach programs The primary endpoint being evaluated is the rate of participation in colorectal cancer screening between 508 the two intervention arms (Sequential Choice, Active Choice) versus the control arm (colonoscopy only). 509 The FIT is a well validated tool for colorectal cancer screening and is one of the screening modalities 510 recommended by the USPSTF. The control arm will be sent a letter inviting them to schedule screening 511 colonoscopy directly through the Gastroenterology Call Center. If not scheduled within 4 weeks, subjects 512 will receive a mailed reminder to call to schedule. The Sequential Choice arm will be sent the same initial 513 outreach as the control arm. If not scheduled within 4 weeks, subjects will receive a mailed reminder 514 including the call center number as well as the option to complete a FIT kit mailed along with the 515 reminder. The Active Choice arm of the study will be sent a letter offering the choice of Colonoscopy or 516 completion of the FIT kit included with the initial mailing. If subjects in this group have not either 517 scheduled colonoscopy nor completed FIT within 4 weeks, they will receive a mailed reminder to call to 518 schedule colonoscopy or to complete the original mailed FIT. 519 520 A sub-sample of 90 subjects will be called to complete a questionnaire over the phone 4 beginning 521 approximately 6 months after initial outreach was mailed. The subjects will be asked to confirm their 522 eligibility (e.g. that they had not had CRC screening within the USPSTF CRC screening guidelines) and 523 provide additional demographic and socioeconomic information so that we can better understand what 524 populations, if any, may have differential response rates. We will also ask them The subjects will be 525 asked to explain their experience with CRC screening prior to our initial outreach (November 2017), as 526 well as about their perception of the impact and design of CRC screening outreach. Additional 527 questions will probe for potential barriers to screening participation and ways to improve 528 participation in the future. Demographic and socioeconomic questions are modified from demographic 529 questions on the Behavioral Risk Factor Surveillance System survey, administered by the Centers for 530 Disease Control and Prevention. See attached sub-sample questionnaire. 531 532 Group Modifications 533 For subjects in the Control (Colonoscopy only) arm of the study, the post outreach phone questionnaires 534 will not include questions regarding mailed FIT. 535 536 537 538 539

Method for Assigning Subjects to Groups 540
Subjects will be randomly assigned Study ID numbers and then randomized to one of three study arms 541 stratified by the two practice locations (University City and Valley Forge) using a computer-generated 542 randomization algorithm. The research coordinator will record the randomization assignments on a 543 master list which will be maintained by the research coordinator on a password protected computer in a 544 locked office. The research coordinator and research assistants will assemble the mailings based on this 545 master list. 546 547 548 Administration of Surveys and/or Process 549 90 subjects will be randomly selected for the questionnaire. We anticipate the post outreach 550 questionnaires to take 10 minutes to complete over the phone. The research staff will make no more 551 than three attempts to speak directly with the subject. Based on a previous project where we reached 552 about 50% of patients via phone call, we anticipate reaching approximately 45 subjects (15 in each 553 arm) to complete this sub-sample questionnaire. 554 555 Administration of Surveys 556 All subjects will complete a baseline and post-intervention survey. The baseline survey will collect basic 557 demographics and baseline medication adherence and blood pressure monitoring frequency and be 558 conducted over the phone after the participant has been consented. The post-intervention survey will 559 collect similar adherence and monitoring information, as well as qualitative data regarding patient 560 perceptions about the interventions. The post-intervention survey will be completed at the in-person 4 561 month visit. These surveys should take no more than 15 minutes to complete. Demographic and socio-562 economic questions are modified from demographic questions on the Behavioral Risk Factor 563 Surveillance System survey, administered by the Centers for Disease Control and Prevention. 564 565 Objectives 566 1.4 Objectives 567 The specific aim of this study is to assess the effectiveness of two different mailed outreach activities 568 ( Colonoscopy is the predominant form of screening in this country due to perceived effectiveness by 593 providers, but it entails a significant cost in time, resources, and perceived discomfort. The fecal 594 immunochemical test (FIT) is an attractive screening option as it is less invasive than traditional lower 595 endoscopy and can be mailed to patients to complete at home. A recent study has shown that offering 596 the choice of colonoscopy or stool-based testing in a clinic setting increases screening rates, but receipt 597 of colonoscopy has better durability since stool-based testing has to occur every year. 598 599 In this study, we will be using population-based outreach screening to evaluate the feasibility of a 600 proactive screening program to promote fecal immunochemical testing (FIT) and/or colonoscopy 601 through mailed outreach by leveraging principles of behavioral economics. We know that mailed FIT 602 outreach circumvents the need for an office visit and eliminates friction in the screening process, since 603 patients can perform testing at home in minutes, but it is not clear how patients may respond to 604 different choice architecture about FIT versus colonoscopy as they have historically been seen as 605 competing, rather than complementary strategies. Behavioral economics suggests that choice 606 architecture may also impact response based on how the intervention is designed. For example, offering 607 the choice of colonoscopy with the mailed FIT kit may enhance participation (as compared to offering 608 colonoscopy alone) since it makes the decision an active choice, where the patient is choosing between 609 two options as opposed to the traditional opt-in approach. Offering mailed FIT after colonoscopy, as is 610 currently the standard during in-office visits, may also increase participation. By evaluating the 611 effectiveness of these alternative choice approaches, we will enhance the public health capacity and 612 efficiency to increase CRC screening uptake and reduce preventable death from this disease. 613 614 Statistical Considerations 615 1.1 Power and sample size 616 Approximately 900 potentially eligible subjects will be identified via a data abstraction by Penn Data 617 Store. Through other projects in primary care practices at Penn Medicine we found the accuracy rate of 618 the EMR algorithm to be approximately 75%. As such, we anticipate we will have enough patients to 619 enroll 423 subjects (and randomize 141 into each arm). We estimate a base return rate for the 620 colonoscopy only (control) arm to be 5%, and we will consider a meaningful increase in response rate to 621 be 10 11 percentage points for both the sequential choice and active choice arms as compared to 622 control. This will be sufficient sample size to detect a 10 11 percentage point increase in response rate 623 using a two-tailed chi-square test with 80% power and a 5% level of significance. Type I error rate of 624 .025, accounting for two pairwise comparisons with Bonferroni correction (.05/2 = .025). 625 626 1.2 Data analysis 627 The primary outcome is CRC screening completion (FIT or colonoscopy) within 4 months of initial 628 outreach. We will conduct a chi-square analysis using Stata to compare arms 2 and 3 to arm 1 separately 629 using intent-to-treat protocol. We will also compare arms 2 and 3 as a secondary analysis. We will 630 quantitatively analyze the choice of screening test and evaluate the survey results by study arm. As 631 exploratory analyses, we will evaluate response by practice location, age, gender, race/ethnicity, and 632 income at the level of zip code. 633 634 Analysis will be conducted by blinded members of the research team at least four months after the last 635 FIT is mailed. 636 637 Study Design 638 1.1 Design 639 Randomized: Subjects will be randomly assigned Study ID numbers and then randomized to one of 640 three study arms stratified by practice location using a computer-generated randomization algorithm. 641 The research coordinator will record the randomization assignments on a master list which will be 642 maintained on a password protected computer in a locked office. The research coordinator and research 643 assistants will assemble the mailings based on this master list. 644 645 Blinding: The investigators will be blinded to the randomization assignment. The research coordinator 646 and research assistants will be unblinded. Subjects rights and welfare will not be adversely affected by the waiver of authorization and consent. 659 All subjects will have the opportunity to voluntarily participate in CRC screening. Arm 1 (Colonoscopy 660 only) will receive screening by colonoscopy if they elect. Arm 2 (Sequential Choice) will receive CRC 661 screening either by Colonoscopy or FIT if they elect. Arm 3 (Active Choice) will receive CRC 662 screening either by Colonoscopy or FIT if they elect. Each arm has the opportunity to engage in CRC 663 screening through routine care as well. 664 665 We believe that we would not be able to practically conduct the research without waiver of consent. If 666 we had to obtain either written or verbal consent ahead of time, it would substantially limit our study 667 population and it may alter their participation in the intervention. Thus, we would only learn about the 668 response rate for patients who we were able to speak to for consent. This, would limit the 669 generalizability to practice. Obtaining waiver of consent would allow us to avoid the potential 670 selection/volunteer bias for inclusion of patients particularly interested in screening that can occur when 671 consent is required. Since our main objective is to understand the potential influence varying outreach 672 strategies on subject behavior, we believe that obtaining consent would compromise our primary 673 objective. Additionally, we have received waiver of consent for similar studies related to colorectal 674 cancer screening outreach. 675 676 Verbal consent will be obtained from the subsample with whom we plan to conduct post intervention 677 interviews (see script Approximately 900 potentially eligible subjects will be identified via a data abstraction by Penn Data 709 Store. Through other projects in primary care practices at Penn Medicine we found the accuracy rate of 710 the EMR algorithm to be approximately 75%. As such, we anticipate we will have enough patients to 711 enroll 423 subjects (and randomize 141 into each arm). We estimate a base return rate for the 712 colonoscopy only (control) arm to be 5%, and we will consider a meaningful increase in response rate 713 to be 10 percentage points for both the sequential choice and active choice arms as compared to 714 control. 715 This will be sufficient sample size to detect a 10 percentage point increase in response rate using a two-716 tailed chi-square test with 80% power and a 5% level of significance. 717 718 719 720 721 722 to source 423 eligible participants from this pool of subjects. 765 766

Subject Compensation 767
Participants will not be financially compensated for their participation. 768 769

770
Screening -Phase 1: We will submit a data request of patients from the CCA practices of University 771 City and Valley Forge from the Penn DataStore, based on an EMR algorithm that determines guideline-772 concordant colorectal cancer screening within PennChart. We estimate approximately 900 screen-773 eligible 774 patients will be identified through this EMR query, 423 of whom will be enrolled and randomized into 775 the three arms. We anticipate conducting chart reviews for two months. During these chart reviews, the 776 Research Assistants and the Research Coordinator will review the electronic medical record charts in 777 PennChart (EPIC) to review each of the eligible patients pulled from Penn Data Store 778 to confirm study eligibility. 779 780 Randomization -Phase 2: Subjects will be randomly assigned Study ID numbers and then randomized 781 with stratification to one of three arms using a computer-generated randomization algorithm. The 782 research coordinator will record the randomization assignments on a master list which will be 783 maintained by the research coordinator on a password protected computer in a locked office. The 784 research coordinator and research assistants will assemble the mailings based on this master list. 785 786 Outreach & Follow-up -Phase 3: Initial outreach letters will be sent to subjects. In the control arm the 787 outreach will include a phone number to the VIP colonoscopy scheduling center. For the sequential 788 choice arm, subjects will receive the same message as those in the control arm, indicating they are 789 overdue for screening and including a phone number to the VIP screening scheduling center. For the 790 active choice arm, subjects will receive both a mailed FIT kit and the option to call and schedule 791 colonoscopy at the same time. After four weeks, a reminder letter will be sent to all subjects in all three 792 arms who did not either schedule a colonoscopy appointment or complete and return a FIT kit. Subjects 793 in the Control arm will receive a reminder letter that includes the same phone number to schedule 794 colonoscopy. Subjects in the sequential choice arm will receive a second letter containing the phone 795 number to the VIP screening scheduling center as well as a FIT kit. The active choice arm will receive a 796 reminder that includes both the phone number to the VIP screening scheduling center and a reminder 797 that they may alternatively complete the FIT that was initially mailed. The Research Coordinator and 798 Research Assistants will be responsible for assembling the mailings. FIT kits will include a tube in which 799 to deposit the stool sample, directions on how to collect and mail the sample, a letter about CRC 800 screening, a lab requisition form, and a pre-paid return envelope. 801 802 Sub-sample Questionnaire -Phase 5: 803 A random subsample of 90 subjects will be selected to complete a questionnaire at least 4 804 approximately 6 months after they received the mailing. Through this questionnaire, the subjects will 805 be asked to explain their experience with CRC screening outreach. Additional questions will probe for 806 potential barriers to screening participation, and ways to improve participation in the future. confirm 807 their eligibility (e.g. that they had not had CRC screening within the USPSTF CRC screening guidelines) 808 and provide additional demographic and socioeconomic status information as well as qualitative 809 experience with the outreach materials and approach so that we can better understand what 810 populations, if any, may be more likely impacted by different types of outreach modalities when 811 engaging in CRC screening. We anticipate these questionnaires to take 10 minutes to complete over the 812 phone. The research staff will make no more than three attempts to speak directly with the subject. 813 814 815 Analysis Plan 816 1.1 Power and Sample Size 817 Approximately 900 potentially eligible subjects will be identified via a data abstraction by Penn Data 818 Store. Through other projects in primary care practices at Penn Medicine we found the accuracy rate of 819 the EMR algorithm to be approximately 75%. As such, we anticipate we will have enough patients to 820 enroll 423 subjects (and randomize 141 into each arm). We estimate a base return rate for the 821 colonoscopy only (control) arm to be 5%, and we will consider a meaningful increase in response rate to 822 be 10 11 percentage points for both the sequential choice and active choice arms as compared to 823 control. This will be sufficient sample size to detect a 10 11 percentage point increase in response rate 824 using a two-tailed chi-square test with 80% power and a Type 1 error rate of .025, accounting for two 825 pairwise comparisons with Bonferroni correction (.05/2 = .025). 5% level of significance. 826 827 1.2 Data analysis 828 The primary outcome is CRC screening completion (FIT or colonoscopy) within 4 months of initial 829 outreach. We will conduct a chi-square analysis using Stata to compare arms 2 and 3 to arm 1 separately 830 using intent-to-treat protocol. We will also compare arms 2 and 3 as a secondary analysis. We will 831 quantitatively analyze the choice of screening test and evaluate the survey results by study arm. As 832 exploratory analyses, we will evaluate response by practice location, age, gender, race/ethnicity, and 833 income at the level of zip code. 834 835 Analysis will be conducted by blinded members of the research team at least four months after the last 836 FIT is mailed. 837

838
Data Confidentiality 839 Paper-based records will be kept in a secure location and only be accessible to personnel involved in the 840 study. Computer-based files will only be made available to personnel involved in the study through the 841 use of access privileges and passwords. Prior to access to any study-related information, personnel will 842 be required to sign statements agreeing to protect the security and confidentiality of identifiable 843 information. Wherever feasible, identifiers will be removed from study-related information. Precautions 844 are in place to ensure the data is secure by using passwords and encryption, because the research 845 involves web-based surveys. 846 847 Subject Confidentiality 848 Information about study subjects will be kept confidential and managed according to the requirements 849 of the Health Insurance Portability and Accountability Act of 1996 (HIPAA). All PHI will be 850 maintained on UPHS servers. Source documents are maintained in PennChart. No source documents 851 will be printed or maintained in paper form at the study site. Data from PennChart will be recorded in 852 Penn Medicine's REDCap system. The investigator and study team (which includes the research 853 coordinator, and research assistants) will have access to PHI within PennChart and REDCap. We will 854 label all PHI within REDCap as identifiable information so that de-identified exports are possible. All 855 reports that include identifiable information will be stored on the Innovation Center secure drive, 856 maintained behind the UPHS firewall. Once data analysis and manuscripts have been published, the 857 databases will be removed from REDCap and the data will be de-identified on the secure drive. This 858 deidentified dataset will be stored for up to five years after analysis is complete and manuscripts have 859 been published. Once analysis is completed and any manuscripts are published, we will retain PHI no 860 longer than seven years in accordance with government regulations, applicable policies, and 861 institutional requirements. 862 863 Database Security/Protection Against Risk 864 To assure that patient, physician and other informant confidentiality is preserved, individual identifiers 865 (such as name and medical record number/physician billing identifier) are stored in a single password 866 protected system that is accessible only to study research, analysis and IT staff. This system is hosted on 867 site at The University of Pennsylvania (UPenn) and is protected by a secure firewall. Once a participant is 868 in this system, they will be given a unique study identification number (ID). Any datasets and computer 869 files that leave the firewall will be stripped of all identifiers and individuals will be referred to by their 870 study ID. The study ID will also be used on all analytical files. 871 The initial patient information collected for screening and recruitment will consist of name, address, 872 phone number, dates, medical records numbers and health plan account numbers. This information will 873 come from Electronic Chart reviews. 874 875

Sensitive Research Information 876
This Research does not involve collection of sensitive information about the subjects that should be 877 excluded from the electronic medical record. 878

879
Subject Privacy 880 We will only interact with the subsample of subjects with which we plan to call to conduct a follow-up 881 questionnaire. With these subjects, we will conduct phone calls in a private area. When we call subjects, 882 we will confirm the identify before administering the questionnaire. We will not be interacting with 883 subjects in person. 884 The risks associated with this study are no more than minimal. There is the potential risk of breach of 948 confidentiality. We will minimize this risk by using de-identified information whenever possible and by 949 maintaining all identifiable information on a secure drive and/or in a HIPAA-compliant system (e.g. 950

Data
REDCap). There is also the risk of psychological harm associated with being screened for cancer. We 951 will minimize this risk by communicating the results of the screening test to the subject in a timely 952 fashion and facilitating the scheduling of diagnostic testing if the screening test is positive (as is usual 953 practice for screening outreach programs Approximately 900 potentially eligible subjects will be identified via a data abstraction by Penn Data Store. 988 Through other projects in primary care practices at Penn Medicine we found the accuracy rate of the EMR 989 algorithm to be approximately 75%. As such, we anticipate we will have enough patients to enroll 423 subjects 990 (and randomize 141 into each arm). We estimate a base return rate for the colonoscopy only (control) arm to be 991 5%, and we will consider a meaningful increase in response rate to be 10 percentage points for both the 992 sequential choice and active choice arms as compared to control. This will be sufficient sample size to detect a 993 10 percentage point increase in response rate using a two-tailed chi-square test with 80% power and a 5% level 994 of significance. 995 996

Data Analysis 997
The primary outcome is CRC screening completion (FIT or colonoscopy) within 4 months of initial outreach. We 998 will conduct a chi-square analysis using Stata to compare arms 2 and 3 to arm 1 separately using intent-to-treat 999 protocol. We will also compare arms 2 and 3 as a secondary analysis. We will quantitatively analyze the choice of 000 screening test and evaluate the survey results by study arm. As exploratory analyses, we will evaluate response 001 by practice location, age, gender, race/ethnicity, and income at the level of zip code. 002 003 Analysis will be conducted by blinded members of the research team at least four months after the last FIT is 004 mailed. Approximately 900 potentially eligible subjects will be identified via a data abstraction by Penn Data Store. 015 Through other projects in primary care practices at Penn Medicine we found the accuracy rate of the EMR 016 algorithm to be approximately 75%. As such, we anticipate we will have enough patients to enroll 423 subjects 017 (and randomize 141 into each arm). We estimate a base return rate for the colonoscopy only (control) arm to be 018 5%, and we will consider a meaningful increase in response rate to be 10 11 percentage points for both the 019 sequential choice and active choice arms as compared to control. This will be sufficient sample size to detect a 020 10 11 percentage point increase in response rate using a two-tailed chi-square test with 80% power and a Type 1 021 error rate of .025, accounting for two pairwise comparisons with Bonferroni correction (.05/2 = .025). 5% level 022 of significance. 023 024

Data Analysis 025
The primary outcome is CRC screening completion (FIT or colonoscopy) within 4 months of initial outreach. We 026 will conduct a chi-square analysis using Stata to compare arms 2 and 3 to arm 1 separately using intent-to-treat 027 protocol. We will also compare arms 2 and 3 as a secondary analysis. We will quantitatively analyze the choice of 028 screening test and evaluate the survey results by study arm. As exploratory analyses, we will evaluate response 029 by practice location, age, gender, race/ethnicity, and income at the level of zip code. 030 031 Analysis will be conducted by blinded members of the research team at least four months after the last FIT is 032 mailed. 033 Before analysis, we felt it necessary to revisit outcomes and provide a more detailed/thorough analysis plan to 045 ensure both were as complete as possible. The analysis plan was updated to be more descriptive of the specific 046 comparisons that would be made, to include communication modality and demographics. 047

5)
The FIT test is an at-home screening test for colon cancer where you send a swab of stool to the lab to 113 test for blood. Most doctors want their patients who are 50-75 years old to either do this test once per 114 year, or get a colonoscopy once every ten years. In the future, would you prefer to have a colonoscopy 115 every 10 years or take a FIT every year?