This is an 18-month cluster randomized controlled trial using a SMART design with two randomizations (the second using an embedded dynamic treatment regimen), with a target enrollment of 800 patients with one of six solid tumors (stage I-III breast, prostate, colorectal, endometrial, and head/neck cancer; stage I-II non-small cell lung cancer) who are being treated with curative intent. To engage the PCP early in the process, the investigators will enroll patients at one of their first visits with a cancer specialist (e.g., surgeon, radiation or medical oncologist). The investigators estimate that 80 unique community-based PCP clinics across North Carolina will be involved in the study. The investigators will determine the effectiveness of the multi-level intervention compared with usual care on (1) Healthcare Effectiveness Data and Information Set (HEDIS) quality measures of management of the three CVD comorbidities using laboratory testing (glycated hemoglobin \[A1c\], lipid profile) and blood pressure measurements; (2) medication adherence assessed pharmacy refill data using Proportion of Days Covered (PDC); and (3) patient-provider communication (Patient-Centered Communication in Cancer Care, PCC-Ca-36).
Key aspects of the intervention are to: reframe the message to patients, PCPs, oncologists, and the respective health care teams in these two settings to emphasize the importance of optimizing the management of comorbidities during and after cancer therapy; promote a change in the workflow in both PCP and oncology practices; enhance PCP-oncologist relationships, and utilize EHR technology. There are two phases of the intervention, both having patient- and PCP-level components. During the first phase of the intervention, occurring with the first randomization, the investigators will test the effectiveness of a self-guided, informational strategy (iGuide). For PCP clinics that do not achieve the HEDIS targets, a booster phase with tailored (patient-level) and targeted (PCP-level) strategies will be tested with a second randomization (iGuide2).
The intervention has been designed to leverage communication tools with in our electronic health record (EHR), Epic . For patients who agree to receive study communications through Epic's patient portal, MyChart, we created an automated messaging system using a dynamic, rules-based protocol which determines the proper language to send each subject based on their study arm and time since enrollment. We developed a similar rules-based automated messaging system for provider communications. This method sends letters to a provider's Epic inbox, called InBasket, for Duke providers and PCPs in the community who have limited access to the Duke EMR through a portal called MedLink. Epic will send letters to PCPs outside of Duke who do not have MedLink access by fax. These outside providers will also be given the opportunity to enroll in MedLink if they are interested.
All participants in the study will be given a survivorship care plan based on the American Society of Clinical Oncology (ASCO) template. Because the investigators will be recruiting participants at our cancer centers and community practices, there will be an inevitable contamination across cancer specialists. Thus, the investigators did not include an oncology-level intervention. However, cancer specialists are integral to the patient- and PCP-level interventions. At the end of the study, patients and PCPs will be mailed a newsletter with a summary of the study findings. Lastly, it is inevitable that some patients will change their PCP during the study. When notified of the change, the research team will send the new PCP the intervention materials.
Supplement:
A retrospective data analysis of Duke cancer registry data integrated with EHR data elements linked by the medical record number will done. The study cohort will consist of older adults ≥65 years who have ≥1 cardiovascular comorbidity (hypertension, type 2 diabetes, dyslipidemia) and underwent cancer surgery for solid tumors at DUHS from January 1st, 2017 to December 31st 2019. Cancers of interest are breast, prostate, colorectal, endometrial, gastric, esophageal, liver, pancreatic, renal cell, bladder, ovarian, head/neck, and non-small cell lung cancer. Cancer diagnoses will be grouped according to the International Classification of Diseases for Oncology, 3rd revision (ICD-O-3). Key variables to be abstracted include demographic, clinical (comorbidities \& pertinent history, medications, cancer treatment, labs), geriatric-specific (frailty, function, cognition, operative, post-operative, and PCP follow-up data.
The qualitative component will involve semi-structured interviews with PCPs of older adults with solid tumors who underwent cancer surgery within a 12-month period in the DUHS. Participating PCPs will complete an audio-recorded semi-structured interview that will last approximately 30 minutes. Qualitative interview: We will design a semi-structured interview guide assessing perspectives on (1) frequency, modes, and perceived quality of communication with surgical providers during transitions in care from surgery (2) perceived barriers to effective care coordination with surgical providers (3) how communication and care coordination can be improved. Open-ended prompts may include "Please describe your experience with care coordination with surgical providers of older adults undergoing cancer surgery?" "What challenges have you faced when communicating with surgical providers of your older adult patients?"; "What are your thoughts on ways to improve communication and care coordination after an older adult with solid tumor undergoes surgery?", "Please give examples of particularly good or poor care coordination with surgical providers?" The interview guide will be designed with assistance from the experienced research staff in the Duke Behavioral Health and Survey Research Core.
Power re-estimation:
Our original power calculations were completed using methods described in Eldridge et al (ref). At the time of project development, the number of practices (40 per arm) and number of patients per practice (10) were unknown, but we estimated the 10 when computing planned sample size for the study. Following enrollment patterns over time, it has been determined that the average number of patients per practice is much smaller, 1.81, presently. If we had this information at time of study planning, based on the design effect and our original effect size, we see that with smaller sample size than originally planned/proposed, we still retain high power, primarily due to the reduction in the design effect (DE). Using the revised calculations, we estimate a power of 0.85-0.9 for 250 projected randomized participants and a power of 0.9 for 275 projected randomized participants.
