Videotape-Based Decision Aid for Colon Cancer Screening

Methods

Setting

We conducted a randomized, controlled trial in three community primary care practices in two small to moderate-sized cities in central North Carolina from May to November 1998. Two sites were primary care divisions of large multispecialty practices, and the third was a three-physician satellite clinic. The study was approved by the institutional review board before initiation.

Patients

Participants were adults 50 to 75 years of age who were scheduled to be seen for a new or ongoing health problem by one of the participating providers from the three practices.

Eligibility

Patients were excluded if they reported a personal or family history of colon cancer or if they had had home fecal occult blood testing (FOBT) in the past year or flexible sigmoidoscopy, colonoscopy, or barium enema in the past 5 years. Patients were also excluded if they were judged by the research assistant to be too ill to participate or if their appointments were only for laboratory blood work.

Group Assignment

Patients were contacted by telephone before their scheduled visits and asked to participate in a study of preventive care. Those who agreed to participate were asked to arrive 30 minutes before their scheduled appointments. The project's research assistants obtained written informed consent to participate and assessed eligibility. Eligible patients were randomly allocated by using a computerized random-number generator (16). Randomization was performed centrally and was not balanced among centers. Assignments were placed in sealed, opaque, sequentially numbered envelopes and were distributed to the three sites. The research assistants opened the envelope for each participant only after eligibility was confirmed.

Intervention

Intervention group participants were asked to watch an 11-minute educational video on colon cancer and then choose one of three color-coded, patient-directed brochures to indicate their interest in screening. A matching laminated card was then attached to the patient's chart.

The content of the video was based on the results of a previous study (6) examining the effect of different types of information on patient interest in screening. The video used simple language and concepts and was tested in two focus groups of adults older than 50 years of age before it was used in this study.

The video included information about susceptibility to colon cancer and availability of effective screening tests. It gave specific information about how home FOBT and flexible sigmoidoscopy are performed and explained the meaning of receiving a negative or positive test result. The video emphasized that screening had been shown to be effective and offered the patient a choice of FOBT, flexible sigmoidoscopy, or both. Finally, it presented brief vignettes featuring four patients who discussed their experiences with colon cancer screening. At the end of the video, the viewer was asked to choose the color-coded brochure that best reflected his or her current readiness to be screened for colon cancer. The video, available in an English-language version only, can be obtained through the Internet at http://www.med.unc.edu/medicine/edursrc/colon.htmor by contacting the authors.

The color-coded brochures were based on Prochaska and colleagues' stages-of-change model (17). The green brochure signified current readiness to be screened, yellow indicated some interest in screening but a need for additional discussion or information, and red indicated no interest in screening at that time. The content of the brochures reinforced the main messages of the video in a stage-appropriate manner. Practices were encouraged to add inserts with instructions specific to their testing procedures. A colored laminated card corresponding to the selected brochure was attached to each intervention group participant's chart before he or she went on to see the care provider. The providers and office staff were not blinded to intervention status.

Controls watched a video of similar length on car safety, seat belt use, and airbags and received a standard brochure on automobile safety. No cards were attached to their charts.

Outcome Measures

At baseline, before randomization, a self-administered questionnaire was used to collect data on participants' intent to ask their provider for colon cancer screening. Intent was measured by using a 4-point Likert scale (1 = not at all likely, 4 = very likely). After viewing a video, participants filled out a second questionnaire that again asked about intent to request screening. The research assistant recorded the participant's choice of brochure at that time. The participants then saw their providers. Afterwards, participants completed a third questionnaire that asked whether a conversation about colon cancer screening had taken place during the visit and whether any colon cancer test was ordered. Questions about automobile safety were included in each of the three questionnaires.

Three to 6 months after the clinic visit, a different research assistant who was blinded to participants' intervention status examined participants' clinic records in a standardized and validated manner to determine whether colon cancer screening tests were actually completed within 3 months of the index visit. Reviewers recorded all colon cancer tests that were completed, not just the tests ordered at the index visit. Dual reviews were conducted on a sample of the charts to ensure accuracy and reliability.

The main outcome measures were differences in the proportions of patients in whom any colon cancer screening test (home FOBT, flexible sigmoidoscopy, or both) was ordered and completed. Office FOBT after rectal examination was recorded but was not counted as completed FOBT. Change in participants' intent to ask their provider for a screening test at the index visit, measured by using a 4-point Likert scale, was a secondary outcome measure.

Sample Size Calculation

We sought to detect a minimum difference between groups of 20% and estimated the proportion of participants in the control group who would have a screening test ordered to be 20% to 30%. For 80% power and a two-sided α value of 0.05, 88 participants in each group were required. We therefore aimed to enroll 100 participants in each group, with a minimum of 50 participants from each site.

Statistical Analysis

Data were analyzed by using Stata software, version 5.0 (16). Categorical data, including the proportions of patients who completed screening tests, were compared by using chi-square tests and are reported as differences in proportions with 95% CIs. Continuous data were compared by using the Wilcoxon rank-sum test. The main outcome of chart-verified screening test completion was measured on an intention-to-treat basis. The possibility of effect modification by practice or physician was examined by using stratified analyses and Mantel–Haenszel testing. Logistic regression was used to examine the effects of covariates on test ordering and completion. Analyses were adjusted for intraclass correlation due to clustering of the effect among physicians.

Role of the Funding Sources

Funding to develop and test the decision aid was provided by a grant from the National Cancer Institute. Fellowship support for Dr. Pignone was provided by the Robert Wood Johnson Clinical Scholars Program. The funding sources had no role in the collection, analysis, or interpretation of the data or in the decision to submit the study for publication.

Results

Enrollment

The process by which potential participants were recruited and entered into the study is shown in the Figure. Research assistants identified 1657 consecutive patients 50 to 75 years of age who had scheduled visits during the study period at the three practices. Of the 1657 potentially eligible patients, 651 were contacted, agreed to participate, and were enrolled. The most common reasons for nonparticipation (n = 1006) were inability to reach the patient by phone (23%), insufficient time to come in early on the patient's part (23%), and arrival of the patient too close to the appointment to be enrolled (20%). The research assistants judged 42 patients (4%) to be too ill to participate and excluded them. Of the 651 potential participants, 249 (38%) were found to be eligible and underwent random assignment. The final sample represents 15% of the 1657 potential participants initially identified.

The most common reasons for ineligibility (402 patients) were having had FOBT in the past year (44%), having had sigmoidoscopy in the past 5 years (22%), or having a history of colon cancer in a first-degree relative (18%). Ineligible participants were similar to eligible participants, except that they were slightly more likely to have Medicare as their main form of health insurance (56% vs. 52%) and were slightly less likely to be high school graduates (78% vs. 86%).

Because of an administrative error, 18 controls did not complete the second and third questionnaires. Four other participants also had missing data for the second and third questionnaires because of insufficient time to complete them. Chart review data were available for 247 of 249 participants (99%). The 249 eligible patients were randomly assigned to the intervention (n = 125) or control (n = 124) groups.

Nine physicians (three at each site) participated in the trial. Among the nine physicians, the number of patients enrolled in the trial varied from 6 to 55 (median, 23 patients). For 14 participants, provider data were missing.

Table 1 shows the participants' demographic characteristics. Overall, the mean age was 63 years; 87% self-reported their ethnicity as white and 13% as African-American. Sixty-one percent were women. All participants had some form of health insurance; only 4% had Medicaid. Twenty-one percent of participants had not completed high school or received their equivalency certificate. Intervention group participants were less likely than controls to have graduated from high school (73% vs. 84%) and were more likely to have Medicare as their main form of insurance (56% vs. 45%).

About one third of participants in each group reported that their visit was for a routine check-up rather than for a new problem or follow-up of an old problem. More than three quarters of participants (78%) reported their health to be excellent, very good, or good. One quarter of the participants had attended the practice for more than 5 years. The mean number of visits to the provider in the past year was 4.4 for intervention group participants and 4.2 for controls.

Intent To Ask for Screening and Test Preferences

At baseline, the mean score for intent to ask for screening (measured on a 4-point Likert scale) was 2.2 in both groups. After the participants watched the video, the mean ± SD score for intent to ask for screening was significantly higher in the intervention group (3.1 ± 1.0) than the control group (2.5 ± 1.1) (P < 0.001, Wilcoxon rank-sum test). In the intervention group, 50.0% of participants moved from low (1 or 2) to high (3 or 4) intent to be screened after watching the video compared with 26.5% of controls.

Test preferences after watching the video differed between the two groups. In the intervention group, 66% of participants preferred FOBT alone, 17% preferred flexible preferred sigmoidoscopy alone, 12% preferred both tests, and 5% preferred no tests; the respective values for controls were 60%, 7%, 25%, and 9%.

Conversations about Screening

When asked after their visits whether they had had any conversation about colon cancer screening, 68.5% of intervention group participants and 43.4% of controls reported some conversation (difference, 25.1 percentage points [95% CI, 12.7 to 37.6 percentage points]).

Ordering Screening Tests

Table 2 shows the effect of the decision aid on patient self-reported ordering of tests. Data were available for 230 participants. After visits with their providers, 33.9% of participants in the intervention group and 19.8% of controls reported that home FOBT had been ordered, and 17.1% of intervention group participants and 8.5% of controls reported that flexible sigmoidoscopy had been scheduled. For the main outcome of having either or both tests ordered, 47.2% of intervention group participants and 26.4% of controls had a test ordered (difference, 20.8 percentage points [CI, 8.6 to 32.9 percentage points]).

Chart review was done 3 to 6 months after the visit to determine the proportion of participants for whom tests were ordered within 3 months of the index visit. The results shown in Table 2 include all 249 participants. Differences between testing as determined by chart review and participants' self-report of tests ordered on the day of the visit were small. The difference in the combined outcome of either or both tests ordered was slightly smaller (40.8% of intervention group participants and 23.4% of controls; difference, 17.4 percentage points [CI, 6 to 28.8 percentage points]).

Screening Test Completion

Table 3 shows the effect of the decision aid on the proportion of participants for whom screening tests were completed, as verified by chart review. Chart review data were available for 247 of the 249 participants; one participant in the intervention group died, and another one's chart could not be located. For purposes of analysis, these two participants were treated as not having completed a screening test.

In the intervention group, 28.5% of participants completed home FOBT compared with 20.2% of controls, and 17.6% of intervention group participants completed flexible sigmoidoscopy compared with 4.8% of controls. For the main outcome of any test completed, the absolute difference between the intervention (36.8%) and control (22.6%) groups was 14.2 percentage points (CI, 3.0 to 25.4 percentage points).

To examine whether features other than the intervention may have affected the outcomes of interest, we performed additional analyses. We first examined the magnitude of effect of the intervention, stratified by practice and by physician. For tests ordered, modest variation in the effect size was seen among practices and among physicians. Overall, the unadjusted relative risk for having a test ordered was 1.79 (CI, 1.23 to 2.58). The three practices had effect sizes of 1.39, 1.39, and 3.56. The P value calculated by using the Mantel–Haenszel test for heterogeneity was 0.14; this value suggests possible differences in the magnitude of effect by practice but did not reach statistical significance because of the relatively small sample. Among the nine physicians, most values for relative risk were between 1.2 and 2.1, but two physicians had outlier values of 0 and 9.2. The P value calculated by using the Mantel–Haenszel test for heterogeneity was 0.56, suggesting little heterogeneity.

Similar analyses were performed to examine variation in effect size for test completion. The crude estimate of relative risk was 1.66. Little variation was seen among practices: Relative risks were 1.26, 1.49, and 1.96 (P > 0.2). Variation among physicians was larger, with relative risks ranging from 0 to 12.8 (P > 0.2).

Because the variation among physicians was greater than that among practices, we elected to perform logistic regression analyses for clustering by physician rather than by practice. Data on provider identity was available for 232 participants. After adjustment for clustering, the odds ratio for the effect of the intervention on test ordering was 2.54 (CI, 1.44 to 4.50), which was unchanged from the crude estimate. For test completion, the crude odds ratio was 2.14 (CI, 1.20 to 3.79); after adjustment for clustering, the confidence interval was slightly wider (0.94 to 4.84).

We next performed logistic regression analysis to examine the effect of several covariates on the main outcome of test completion. Adjustment for patient age, sex, previous completion of a screening test, and insurance status did not appreciably change the effect sizes for test ordering or completion.

Role of the Brochure

To test the validity of the stages-of-change model, we examined the relation between choice of brochure and rates of test ordering and completion. After viewing the video, 53% of intervention participants selected the green brochure, indicating an interest in being screened; 22% chose the yellow brochure, indicating that they were considering screening but needed more information; and 23% chose the red brochure, indicating they were uninterested in screening at that time.

Among intervention participants, 69% of those who selected the green brochure had a colon cancer screening test ordered, 41% of those who selected the yellow brochure had a test ordered, and 7% of those who selected the red brochure had a test ordered. In contrast, the proportion of controls for whom a screening test ordered was 26%. In the intervention group, 54% of participants who chose the green brochure completed tests, compared with 35% of those who chose yellow and 4% of those who chose red. The rate of test completion among controls was 21%.

Test Completion according to Level of Intent To Be Screened

We examined the effect of the decision aid within participants who had similar levels of intent to be screened after watching the video. Among the 136 participants with high intent (a score of 3 or 4), the rate of test completion was 46% in intervention group and 33% in the control group. Among the 89 participants with low levels of intent (scores of 1 or 2), the groups did not differ in the rate of test completion (20% vs. 19%).

Discussion

We found that a decision aid consisting of an educational videotape combined with a stages-of-change–based educational brochure and chart marker increased patient intent to request colon cancer screening, the proportion of patients who have screening tests ordered, and the proportion of patients who completed screening tests. We enrolled consecutive average-risk patients who had not been screened recently for colon cancer from community primary care practices. The absolute difference in the proportion of patients who completed screening was 14.2 percentage points (36.8% of the intervention group and 22.6% of controls). Approximately seven patients would need to use the decision aid to lead to one additional patient being screened.

The effect size that we found was similar to that of other interventions designed to improve screening. The fact that the providers and staff were not blinded and were aware of the nature of our study may have influenced the effect size for our intervention. Providers may have tried harder to screen participants in the intervention group, thereby inflating the effect size. On the other hand, awareness of the study may have increased efforts to screen all patients; this “Hawthorne effect” may have caused underestimation of the true difference. We minimized possible bias from repeated measurements by administering the same questionnaires to intervention and control participants.

Other studies have examined interventions to increase colon cancer screening in academic and community primary care settings (7, 8, 11-14, 18-24). Interventions directed at multiple steps in the process of care have generally been more effective in increasing performance of preventive care, including screening for colon cancer (19). A recent systematic review of physician prompting systems found that such systems increased FOBT screening by 14% (10).

Three studies examined the effect of a patient-held record of preventive care to increase colon cancer screening. They found increases in the proportion of patients completing FOBT of 1% (11), 4% (13), and 17% (21). Myers and colleagues (12) tested different combinations of patient interventions, including reminder telephone calls, instructional telephone calls, and an educational brochure, and increased screening by 10% to 21%. The addition of the instructional telephone call had a greater incremental effect than did the instruction booklet (12).

Videotape-based interventions have been used successfully to guide decision making about other preventive care issues, such as the decision to undergo prostate-specific antigen screening for prostate cancer (25). Barnas and associates (Personal communication; 1997) tested an 8-minute video in an internal medicine clinic of an urban public hospital. In a preliminary report from this trial, Meade and colleagues (20) found increased interest in and intent to be screened but no significant effect on the rates of tests ordered or performed.

We chose to compare our decision aid system with usual care rather than another effective approach such as a reminder system or audit and feedback. Thus, we tested whether a system initially directed to patients but reinforced with a chart marker could be an effective option for a practice initiating a process to improve rates of colorectal cancer screening. Our results suggest that performance of colon cancer screening can be increased by an intervention that informs patients, allows them to participate in the decision about how to be screened, and helps providers and staff act on the decision. We included the brochure and chart marker to encourage conversation and indicate the patient's level of interest in screening to the provider and office staff because previous research by Barnas and associates (Personal communication) indicated that using a patient video alone was insufficient to increase screening even though it improved knowledge.

Our analysis suggests that the video and chart marker worked synergistically to increase screening. The video increased patients' intent to ask their providers about screening. The chart marker appeared to increase screening among patients with high intent to be screened after watching the video but seemed to have no effect among patients with low intent.

We found an increase in the level of screening similar to the mean increase reported in studies that used reminder systems (10). The question of whether our system can produce additional benefit when added to a reminder or audit and feedback system will require a larger sample and is an important topic for future investigation.

We chose to include FOBT alone, flexible sigmoidoscopy alone, or both tests as screening options because these tests have been shown to be effective for screening, are supported by national guidelines (1, 2), and were the methods commonly used in the study practices. Barium enema and colonoscopy were not included because they were not used commonly as screening tests in these practices and have not yet been studied as thoroughly for screening. Future versions of the video will be modified to include these options for providers who choose to offer them.

By choosing to assess outcomes at 3 months, we attempted to balance the ability to measure the specific effect of the intervention visit against the possibility that we may have missed sigmoidoscopies that were slow to be scheduled and performed. However, since the number of sigmoidoscopies performed is about equal to the number ordered, we believe that our measure is accurate.

The final sample represented 15% of all age-eligible patients seen in the practices during the study period. Many of the patients not studied were those we were unable to contact, but we cannot generalize our results to patients who chose not to participate. Forty-one percent of potential participants self-reported recent screening with FOBT or flexible sigmoidoscopy, a proportion slightly higher than that in previous population-based surveys or chart audits from similar practices. The participants who were randomly allocated had not had recent screening (despite the relatively high level of screening in the practices) and thus may be more difficult to reach.

It should be noted that even in our intervention group, more than 60% of eligible patients were not screened. Among participants who expressed readiness to be screened by choosing the green brochure, the rate of test ordering was 69% and the rate of test completion was 54%. It is not clear what prevented the other 31% of interested participants from having tests ordered or why participants failed to complete ordered tests. Future efforts to increase screening could be directed toward ensuring that interested patients are offered and receive testing, perhaps through standing orders to nursing staff or a more formal physician and staff educational intervention. In contrast, a relatively large proportion of patients (23%) expressed that they did not want any screening at the index visit. Further qualitative and quantitative research is required to understand the various reasons for this lack of interest.

The full decision aid requires about 15 minutes of patient time: 11 minutes to view the video and 3 to 5 minutes to select and read the brochure and to have the research assistant attach the corresponding chart marker. Because much of the information about screening is provided in the video, little initial provider time is required. We did not formally study whether the video reduced total provider time, although anecdotal reports from participating providers suggested that this is the case. To further improve feasibility and efficiency, the video could be mailed in advance to eligible patients for home viewing.

In conclusion, we found that a majority of patients who are informed about colon cancer screening through a decision aid consisting of a video, brochure, and chart marker want to be screened. Informed patients also appear to be more likely than controls to have screening tests ordered and to complete them. Future research is needed to test whether informed patient choice, as facilitated by this decision aid, is superior or inferior to a process in which the provider selects a single colon cancer screening strategy for all patients. In addition, longitudinal research should be conducted to examine whether the decision aid continues to affect screening at future visits, since screening, especially with FOBT, requires ongoing adherence to be effective.