The American Journal of Managed Care January 2011
A System-Based Intervention to Improve Colorectal Cancer Screening Uptake
Using an electronic health record to identify and implement colorectal cancer screening in a population of eligible patients achieved higher uptake than a visit-based approach.
Objective: To determine whether mailing guaiac-based fecal occult blood tests (gFOBTs) directly to patients who are due for colorectal cancer screening would achieve higher screening uptake than using visit-based screening.
Study Design: Comparative effectiveness analysis.
Methods: We used an electronic medical record to identify 7053 New Mexico Veterans Affairs Health Care System patients aged 50 to 80 years who were due for screening in 2008. We invited 3869 randomly selected patients to participate in a randomized controlled trial comparing adherence with different fecal blood tests; 202 intervention patients were assigned to receive mailed gFOBTs. We identified the following 3 control groups who could receive only visit-based colorectal cancer screening: 3184 individuals who were not invited for the randomized controlled trial (control group 1), 2525 individuals who did not respond to invitations to participate in the randomized controlled trial (control group 2), and 255 individuals who could not be contacted (control group 3). We measured gFOBT screening within 3 months after enrollment in the intervention group, as well as gFOBT or colonoscopy screening within 6 months of identification as a control subject. We compared screening across groups using multivariate logistic regression analysis to adjust for sex, race/ethnicity, clinic site, previous gFOBT, and comorbidities.
Results: Colorectal screening occurred less often in each of the control groups (in 18.6% of control group 1, in 14.3% of control group 2, and in 18.8% of control group 3) than among patients mailed a gFOBT (48.5%). Adjusted odds ratios for screening among the control groups were all less than in the intervention group (adjusted odds ratios, 0.25, 0.19, and 0.23, respectively; all, P <.001).
Conclusion: Using an electronic medical record to identify screening-eligible patients and mailing them gFOBT cards achieved higher colorectal screening uptake than performing visit-based screening.
(Am J Manag Care. 2011;17(1):49-55)
Visit-based approaches to implementing cancer screening are inefficient.
- Visit-based approaches may not coincide with due dates for providing preventive services, and they overselect for sicker patients, who are more frequently seen in clinic but are less likely to benefit from preventive services.
- An electronic health record enables a healthcare system to identify the entire population of patients who are due for screening and facilitates screening efforts that directly reach these patients whether or not they are being routinely seen in clinic.
One system barrier arises from relying on routine physician office visits to implement screening. Interventions involving health system organizational changes, including the use of separate prevention clinics, direct patient reminders, and designation of nonphysician staff to perform prevention activities, are effective in increasing screening uptake.9 A “new model of primary care delivery” based on the chronic care model10 has been proposed to guide system changes to improve colorectal cancer screening rates.11 Important components of this model are an emphasis on using information systems to identify and remind patients who are due for screening, to track results, and to drive a team approach that allows for non–visit-based delivery of care.
The New Mexico Veterans Affairs (VA) Health Care System relies on patient visits to trigger screening alerts to implement screening, usually with gFOBT. Our group conducted a randomized controlled trial to compare testing adherence between fecal immunochemical tests (FITs) and gFOBTs; results showed greater adherence with FITs and are described elsewhere.12 To enroll patients for the intervention trial, we used the VA electronic health record system, the Veterans Health Information Systems and Technology Architecture (VistA), which allowed us to identify the entire cohort of New Mexico VA Health Care System patients who were eligible for colorectal cancer screening. We randomly selected a sample of 50.0% of these patients to be invited for the intervention study. However, this strategy also allowed us to identify several control groups who would require clinicvisits to initiate screening (Figure). Therefore, we performed a secondary analysis of our study data to evaluate whether using the electronic health record to identify eligible patients and then to directly mail gFOBT kits to enrolled patients would increase screening uptake compared with usual care screening that is based on clinic visits.
The study setting was the New Mexico VA Health Care System, including primary care clinics and laboratory at the Raymond G. Murphy VA Medical Center in Albuquerque, New Mexico, and community-based outpatient clinics (CBOCs) in Artesia, Gallup, Raton, Santa Fe, and Farmington, New Mexico. These primary care clinics were staffed by 36 healthcare providers, including 8 working in the CBOCs. The study was approved by the Raymond G. Murphy VA Medical Center Research and Development Committee and the University of New Mexico Human Research Review Committee, Albuquerque. The clinicaltrials.gov identifier is NCT00692211.
We used VistA files, including demographics, problem lists, procedure codes, pathology records, and screening reminders, to identify eligible patients. Inclusion criteria were active enrollment in a VA primary care clinic, age 50 to 80 years, and status of being due for colorectal cancer screening, for which gFOBT was appropriate. We used Current Procedural Terminology codes to exclude patients who had undergone colonoscopy in the past 10 years, flexible sigmoidoscopy in the past 5 years, or barium enema in the past 5 years. We used a laboratory results file to exclude patients who had undergone gFOBT in the past year and International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) and Systematized Nomenclature of Medicine codes to exclude patients with any history of adenomatous polyps, colorectal cancer, inflammatory bowel disease, or terminal illness.
Recruitment. Recruitment was conducted in 2 phases. We initially identified 6405 patients who were eligible for screening on May 19, 2008, and did not have an upcoming primary care appointment in the next 2 months. This exclusion was to allow sufficient time to contact and enroll patients. We randomly selected 3221 patients for the intervention study and 3184 patients as control subjects using a random number generator.
We asked primary care providers to identify any patients for whom fecal blood test screening would be inappropriate. We then mailed cover letters to 3100 potential patients from their primary care providers describing the study and asking them to return a postcard if they were interested in participating. We called interested patients, confirmed their eligibility, and enrolled those wishing to participate.
Unexpected delays in enrolling patients prompted a second phase of recruitment in August 2008, when we identified another 5029 eligible patients. We randomly selected 1226 to be contacted. After obtaining approval from the primary care providers, we mailed invitation letters to 1088 patients. Overall, for the 2 phases we identified 7053 unique patients who were eligible for colorectal cancer screening and mailed 4188 letters to 3869 of these patients (Figure). We ended enrollment on December 15, 2008.
Intervention Patients. We used a Web-based random number generator (http://www.randomizer.org/) to randomly assign patients (n = 404) in groups of 8 to FIT kits (2 sample collectors) or to gFOBTs (3 cards). Assignments were based on consecutively returned postcards that contained only unique study identification numbers. We sent consent forms and Health Insurance Portability and Accountability Act of 1996 forms with the test kits along with an envelope addressed to the laboratory with prepaid postage for returning the test kits and regulatory forms. Herein, we evaluate screening uptake only in the gFOBT arm of the study because this was the only fecal blood test available to patients in the control groups. In addition, because screening adherence was higher among patients receiving FITs,12 we would bias results in favor of the intervention by including them in the analyses.
Control Subjects. We created 3 usual care control groups. Patients in these groups would be offered colorectal cancer screening when they were seen in clinic and the electronic health record indicated that they were due for screening. The screening reminder is based on the date of the most recent fecal blood test, lower endoscopic procedure, or barium enema; the screening reminder excludes patients who require surveillance testing based on previously diagnosed colorectal neoplasia. The first control group was composed of 3184 patients who were not randomly selected to be contacted for the intervention study. A second control group was composed of 2525 patients who were contacted for the intervention study but did not return a postcard regarding their interest in participating in the study (and did not live out of state [n = 29] or have their invitation letter returned as undeliverable [n = 113]). Overall, 1202 patients (31.1%) responded to an invitation letter by returning a postcard, including 291 who refused to participate. The third control group was composed of 255 patients who returned a postcard indicating interest in participating in the study but could not be enrolled because we could not reach them by telephone (n = 246) or their postcard was returned after enrollment ended (n = 9).
Baseline Data Collection. We used VistA to collect data for all intervention and control subjects, including age when identified as eligible for screening, sex, race/ethnicity (which is not routinely collected from patients), clinic site, gFOBTs during the previous 3 years, and the number of comorbidities based on ICD-9-CM problem list counts.
Stool Testing. Patients received gFOBT cards (Hemoccult II; Beckman Coulter, Fullerton, California). Patients were instructed to avoid aspirin (acetylsalicylic acid) and other nonsteroidal anti-inflammatory drugs, as well as rare meat, vitamin C, and foods containing peroxidase during the 3-stool collection period. All gFOBT cards were interpreted visually for positivity after adding developer.
Outcome Measurements. We used the VA electronic health record laboratory file to track gFOBT completion among patients in the gFOBT arm of the intervention trial and among control subjects. We also used Current Procedural Terminology codes to capture data on colonoscopies performed among control subjects. The New Mexico VA Health Care System does not perform flexible sigmoidoscopy or barium enemas for colorectal cancer screening. We collected testing data for control groups 1 and 2 for 6 months after identification as due for screening. For control group 3, we collected testing data for 6 months after receipt of the postcard indicating interest in participating in the study. We referred patients in the intervention study with positive gFOBT results for colonoscopy.
Baseline Statistics. We used descriptive statistics with 95% confidence intervals to characterize patients in the gFOBT arm of the intervention study versus the usual care control groups. Differences across groups were evaluated using analysis of variance for normally distributed continuous variables, Kruskal-Wallis test for skewed data, and c2 test for categorical variables.
Outcomes. We used multivariate logistic regression analysis to compare the proportion of intervention patients completing protocol gFOBTs within 3 months after enrollment in the intervention group (excluding 13 gFOBTs ordered by primary care providers) versus control subjects completing gFOBT or colonoscopy screening within 6 months of identification as a control subject. We adjusted for demographics (age, sex, and race/ethnicity), clinic site, previous gFOBT, and ICD-9-CM comorbidities. We modeled the square root of the comorbidity counts because of the skewed distribution. P <.05 was considered statistically significant.
Baseline characteristics of the cohort are given in Table 1. Most patients were middle-aged men with multiple comorbidities. Compared with intervention patients, control subjects were less likely to have had previous gFOBTs, more likely to receive care in a CBOC, and less likely to be identified as being of non-Hispanic white race/ethnicity.