Cost-effectiveness of targeted and tailored interventions on colorectal cancer screening use
Colorectal cancer (CRC) screening is cost-effective but underused. The objective of this study was to determine the cost-effectiveness of targeted and tailored behavioral interventions to increase CRC screening use by conducting an economic analysis associated with a randomized trial among patients in a large, racially and ethnically diverse, urban family practice in Philadelphia.
The incremental costs per unit increase were measured in individuals who were screened during the 24 months after intervention. Percent increase in screening was adjusted for baseline differences in the study groups. Each intervention arm received a targeted screening invitation letter, stool blood test (SBT) cards, informational booklet, and reminder letter. Tailored interventions incrementally added tailored messages and reminder telephone calls.
Program costs of the targeted intervention were $42 per participant. Additional costs of adding tailored print materials and of delivering a reminder telephone call were $150 and $200 per participant, respectively. The cost per additional individual screened was $319 when comparing the no intervention group with the targeted intervention group.
The targeted intervention was more effective and less costly than the tailored intervention. Although tailoring plus reminder telephone call was the most effective strategy, it was very costly per additional individual screened. Mailed SBT cards significantly boosted CRC screening use. However, going beyond the targeted intervention to include tailoring or tailoring plus reminder calls in the manner used in this study did not appear to be an economically attractive strategy. Cancer 2008. © 2007 American Cancer Society.
There will be an estimated 153,760 new diagnoses of colorectal cancer (CRC) and 52,180 deaths from CRC in 2007.1 On the basis of evidence from randomized controlled trials and case–control studies, the U.S. Preventive Services Task Force and other organizations strongly support the periodic use of screening tests to reduce CRC morbidity and mortality among men and women aged ≥50 years.2–14 CRC screening offers the potential both for primary prevention through polyp removal and for secondary prevention through detection and treatment of early-stage disease.9, 11, 12
For average-risk adults aged ≥50 years, the American Cancer Society recommends annual stool blood testing (SBT) with fecal occult blood testing or fecal immunochemical tests and suggests the following intervals for other CRC screening tests: flexible sigmoidoscopy (FS) every 5 years, double-contrast barium enema X-ray every 5 years, or colonoscopy every 10 years. In addition, these guidelines include the combination of annual SBT and FS every 5 years.13 The Year 2010 Health Objectives for the Nation recognizes the importance of CRC early detection.5 The objectives for CRC screening for the population aged ≥50 years are 1) 50% should have had a screening SBT in the preceding 2 years, or 2) 50% should have ever received a screening FS.
The 2004 Behavioral Risk Factor Surveillance System findings indicate that 27% of Americans aged ≥50 years reported having an SBT within the past 2 years, and 54% reported ever having a sigmoidoscopy or colonoscopy.15 Reported exposure to endoscopy is likely to over estimate screening use, because both screening procedures and diagnostic procedures are reflected by this measure. Data from the National Health Interview Survey for 2005 indicate that only 35% of Americans report ever having had a CRC screening test.16 Economic evaluations of CRC screening reveal that it is highly cost-effective relative to many existing health-promotion and treatment interventions.3, 17–19 These findings highlight the need to increase CRC screening rates in the U.S.
We recently provided evidence of the effectiveness of targeted and tailored interventions to increase CRC screening in a primary care practice setting.20 Given the highly constrained resources for health promotion, decision makers require information on the cost-effectiveness of the interventions. Although there have been several studies of the economics of mammography screening promotion, few studies have evaluated the cost-effectiveness of CRC screening promotion.21, 22 Estimates for promotion strategies range from a cost of $11 to $978 per additional individual screened.23, 24 The former estimate was for a low cost intervention directed toward disadvantaged populations in primary care clinics, and the latter estimate was for a sophisticated tracking system in the Veterans Administration Health System. In this report, we present findings from a National Cancer Institute-funded, randomized controlled trial that was designed to test the impact and cost-effectiveness of targeted and tailored behavioral interventions on CRC screening use among patients in a large, racially and ethnically diverse, urban family practice in Philadelphia.
MATERIALS AND METHODS
Identification of Study Participants
Potential study participants included adult patients at the Jefferson Family Medicine Associates (JFMA) practice, which is a large urban practice located at Thomas Jefferson University in Philadelphia. In 2001, we used the JFMA practice billing database to identify patients ages 50 years to 74 years of age who had no prior diagnosis of colorectal neoplasia or inflammatory bowel disease, had had at least 1 visit to JFMA within the previous 2 years, had complete contact information available (ie, address and telephone number), and did not have a recent CRC screening examination.
Control group participants received usual care during the course of the study. Participants in the standard intervention (SI) group initially received a mailed standard intervention (ie, a CRC screening invitation letter, SBT cards, informational booklet, and reminder letter). The screening invitation letter encouraged recipients to complete SBT screening and to schedule an FS screening examination. Participants in the tailored intervention (TI) group simultaneously were sent the standard intervention plus 2tailored “message pages.” These pages included brief messages that addressed barriers to SBT and FS screening that were identified through analyses of baseline survey data. Participants who received the tailored intervention plus a telephone call (the TIP group) were mailed the standard intervention and the tailored message pages and also were designated to receive 1 reminder telephone call for each round of the intervention. In the telephone call, a trained health educator reviewed the mailed materials and encouraged participants to consider screening. Before initiating the reminder calls, study health educators reviewed a self-study guide on study procedures. In addition, mock telephone interviews were conducted to assess interviewer manner and methods in conducting the reminder call. Corrective feedback was provided to increase caller completeness and effectiveness. During the study, call logs were obtained to track caller activity. These logs were reviewed during periodic staff meetings to identify and resolve issues related to delivery of the reminder calls.
Participants in each of the intervention groups (SI, TI, and TIP) received 2 “rounds” of contact. An initial round of intervention contacts was delivered soon after randomization, and a second round of intervention contacts was delivered approximately 1 year later. The same types of contacts that were delivered during the initial round were repeated for participants who did not have an abnormal CRC screening result or who did not have colonoscopy screening according to either self-report or billing records. We administered a midpoint survey to study participants during the study. Finally, we administered an endpoint survey and completed an endpoint chart audit approximately 24 months after participant randomization. The study was reviewed and approved by the Institutional Review Board of Thomas Jefferson University and was initiated in 2002.
To create tailored message pages, the research team initially conducted multivariate analyses using baseline survey data to identify Preventive health model (PHM) constructs (ie, perceived salience and coherence, susceptibility, worries and concerns, self-efficacy, response-efficacy, social support). Performance of these analyses served to identify PHM variables that were associated significantly (P < .05) with being in higher decision stages for SBT and FS screening, respectively. A similar approach to identifying test-specific perceptions was reported by Rawl et al.25
Next, we selected up to 3 PHM variables that had the strongest associations with outcome and developed “motivating” and “reinforcing” educational messages keyed to scores on these variables. Specifically, 2 messages were generated for each variable. A “motivating” message was keyed to a score that was associated negatively with decision stage (eg, low self-efficacy), and a “reinforcing” message was keyed to a score that was associated positively with decision stage (eg, high self-efficacy). Messages were deposited electronically onto template pages, 1 for SBT and 1 for FS. In accordance with JFMA policy, performance of both SBT and FS screening was promoted by sending pages for both screening tests.
We combined the results from the randomized practice-based trial to increase CRC screening rates with an estimate of intervention costs to assess the cost-effectiveness. The intermediate outcome—screening rate—was determined directly from the randomized trial described above.20 An endpoint chart audit for each study participant was performed at approximately 24 months after initial randomization. Screening compliance (the effectiveness measure) was defined as having had 1 or more documented SBT of any type or a self-reported or documented FS, colonoscopy, or barium enema x-ray within 24 months after randomization. Data regarding screening use also were obtained from JFMA medical records, the JFMA electronic billing database, and an SBT laboratory database. To count an occurrence of SBT screening after randomization, we used documentation from the medical chart, the SBT laboratory, or the practice electronic billing database. The decision to require such documentation for SBT is consistent with reports in the literature that self-reported SBTs tends to be inaccurate.26 We accepted occurrences of endoscopy (ie, FS and colonoscopy) screening and barium enema X-ray screening if these procedures were documented in the medical chart, if they were present in an electronic billing database, or if they were reported on a survey questionnaire. In each of these instances, we required a performance date within the study observation period. The decision to accept self-reported endoscopy and barium enema X-ray procedures in the presence of performance data was consistent with the finding that self-reports of recent memorable procedures are highly reliable.26, 27
Activity cost estimation was used to determine the cost of each intervention. This estimate consisted of determining separate labor, supply, and overhead costs for each activity. Then, the total for each activity was divided among the intervention arms based on the number of patients in each group and the number of patients involved in each activity. Costs incurred for research purposes only, including the baseline and midpoint surveys for the usual care and SI groups, were excluded from the analysis. Intervention cost data were collected from 3 sources: questioning staff to approximate the amount of time required for specific tasks, purchase orders and invoices, and current market prices of supplies. Overhead cost was calculated as 30% of the direct cost to approximate overhead in medical care settings.28 Invoice data were collected between 2000 and 2003 and were adjusted to 2006 U.S. dollars by using the Consumer Price Index (CPI).29
All personnel costs were calculated by multiplying the hours spent on each activity by the adjusted salary per hour of the personnel.30 The base salary for each position (project manager, data manager, research assistant, primary investigator, and coinvestigator) was used to calculate the adjusted salary per hour for the project staff. The number of hours spent performing the specific tasks were estimated retrospectively. The overall time spent on each activity was used to compute a unit time cost per task.
Supplies included questionnaires, printing costs, envelopes, paper, postage, and all other resources that were used to produce and distribute the surveys and intervention materials. The costs of these materials were calculated by itemizing the materials needed for each activity that was completed directly by Thomas Jefferson University. These materials were priced according to current market rates. Other materials that were used by outside sources, such as ListMasters and Mathematica Policy Research Inc. (MPR), were accounted for by using invoices from project records. These costs were adjusted by using the CPI29 and were valued in 2006 U.S. dollars. Computer use and computer software were not included in the supply cost, because they were considered part of the overhead.
Indirect costs, such as administrative services, utilities, and office space, were calculated by multiplying the total direct cost (base) by a hypothetical indirect rate of 30%. This approach is consistent with other rates reported in the literature for healthcare activities.30 The indirect cost base is the summed personnel costs and supply costs. Total charges from MPR were divided into direct and indirect costs based on the assumed 30% overhead rate.
Incremental cost-effectiveness ratios (ICERs) were computed by dividing the added cost by the increased percentage of individuals being screened moving from the control group to the SI group and then to the TI and the TIP groups. Important, uncertain parameters included the effect sizes of the interventions and the overhead rate. The 95% confidence intervals on effect sizes were used to assess the sensitivity of ICERs to the effect size parameter. The overhead rate was varied from 25% to 35% to assess the sensitivity of the ICERs to the overhead cost assumption.
There were approximately 6300 patients aged 50 to 74 years who were using the family practice clinic according to the billing records. Of those patients, 2579 were eligible for the study and were contacted for the baseline survey. The estimated CRC screening rate in the practice before the intervention ranged from 20% to 25%. We completed a baseline telephone survey with 1546 individuals and randomized respondents to 1 of 4 study groups: the control group (n = 387), the SI group (n = 387), the TI group (n = 386), and the TIP group (n = 386). Table 1 shows the demographic and social characteristics of the study populations. Majority categories were women, nonwhite, age <60 years, and not married. Approximately 50% of the participants had more than a high school education.
Table 1. Sample Characteristics by 4 Study Groups: 24-Month Outcomes (n = 1546)
|Sex|| || || || || || || || || || ||.963|
| Men||513||33.2||131||33.9||129||33.3||124||32.1||129||33.4|| |
| Women||1033||66.8||256||66.1||258||66.7||262||67.9||257||66.6|| |
|Age, y|| || || || || || || || || || ||.648|
| <60||955||61.8||238||61.5||241||62.3||248||64.2||228||59.1|| |
| 60–69||460||29.8||120||31||113||29.2||102||26.4||125||32.4|| |
| ≥70||131||8.5||29||7.5||33||8.5||36||9.3||33||8.5|| |
|Race (n = 1530)|| || || || || || || || || || ||.511|
| Nonwhite||948||62||237||61.9||226||59||245||64.1||240||62.8|| |
| White||582||38||146||38.1||157||41||137||35.9||142||37.2|| |
|Marital status (n = 1535)|| || || || || || || || || || ||.564|
| Not married||883||57.5||215||55.7||215||55.8||228||59.8||225||58.7|| |
| Married/living as married||652||42.5||171||44.3||170||44.2||153||40.2||158||41.3|| |
|Education (n = 1539)|| || || || || || || || || || ||.160|
| <High school||206||13.4||49||12.7||48||12.4||50||13.1||59||15.3|| |
| High school||556||36.1||123||31.9||149||38.6||153||40.1||131||34|| |
| >High school||777||50.5||214||55.4||189||49||179||46.9||195||50.6|| |
Table 2 presents the cost per eligible patient of delivering the SI, TI, and TIP interventions. The SI cost $42 per individual in the target group. The average cost of the TI was $150, more than 3 times as costly as the SI. The TIP added another $50 per individual to the cost of the TI. The higher cost in the tailored interventions is because of the time and materials required for conducting and processing the questionnaires and telephone calls necessary for tailored interventions. Approximately 30% of the cost of the SI is for personnel compared with 19% of the cost of the TI (without the telephone call) and 35% of the cost of the TIP (with the telephone call).
Table 2. Cost of the Interventions
|Identify target population||1523||1519||1519|
|Training and planning meetings||3829||9572||21,058|
|Baseline and midpoint survey administration||—||24,931||24,931|
|Two rounds of interventions (screening invitation letter, SBT cards, information booklet, and reminder letter)||5668||5653||5653|
|Two rounds of producing tailored message pages||—||1355||1355|
|Two rounds of mailed reminders||1495||1492||1492|
|Two rounds of tailored reminder telephone calls||—||—||3492|
|Total direct cost||12,515||44,522||59,500|
|Overhead (30% of direct cost)||3755||13,357||17,850|
|Cost per individual†||42||150||200|
Table 3 presents estimates of the base case incremental cost-effectiveness analysis derived from the randomized trial of alternative interventions to increase CRC screening rates. Although the randomized trial provides evidence that the targeted intervention was effective and statistically significant at conventional levels, the between-intervention results were not statistically significant at the conventional α = .05 level. Approximately 33% of individuals were screened in the usual care group compared with 46% in the SI group, 44% in the TI group, and 48% in the TIP group.
Table 3. Incremental Cost-effectiveness (Intervention Cost per Additional Individual Screened)
|Usual care||0||—||32.56 (1.00)||—||—||—|
Because the purpose of the economic evaluation was estimation and not hypothesis testing, we used the point estimates as the “best” available estimates of program effects. This effect was augmented with sensitivity analyses of key variables to address the impact of uncertainty on the cost-effectiveness ratios.31 Odds ratios that compared each intervention arm with the referent control group category are displayed in the table. Also shown are the incremental costs, incremental effects, and incremental cost-effectiveness ratios of moving from a less resource-intensive SI to more resource-intensive TI and TIP interventions.
The ICER was $319 when moving from no intervention to the SI. The SI was more effective than the TI and dominated the TI, because the SI also was less costly. Comparing the SI with the TIP, the ICER was $5843. This relatively high cost was because the TIP was considerably more costly per individual than the SI and was only marginally more effective. Because of the finding that the SI dominated the TI, there was no comparison made between the TI and the TIP.
Table 4 shows how the ICERs were affected when effect size estimates were varied across the 95% confidence limits of the effects and when the overhead rate was varied from 25% of direct costs to 35%. Sensitivity was examined for the comparison between the usual care control condition and the SI and between the SI and the TIP. The base-case values for effect size and overhead rates were the midpoints of the range over which sensitivity was assessed. The alternative parameter estimates did not have a large impact on the ICERs: The TIP estimate ranged from $5550 to $6113, and the ICERs for the usual care versus the SI ranged between $300 and $339.
Table 4. Sensitivity Analysis
|Overhead = 25% of direct cost|| |
|Overhead = 30% of direct cost|
|Overhead = 35% of direct cost|
Our economic analysis demonstrates that the cost of the SI at $42 per individual in the study population and $319 per additional individual screened is well within the range of acceptable cost compared with other practice-based studies of CRC screening promotion (Table 5). Surprisingly, the SI dominated the TI, because it was less costly and more effective according to the point estimates of effect on screening behavior. Although the TIP was more effective than the SI, the TIP intervention was not attractive economically, because it cost an average of $200 per individual and almost $6000 per additional individual screened.
Table 5. Studies on the Cost-effectiveness of Colorectal Cancer Screening Promotion
|Wolf 200524||FOBT, flexible sigmoidoscopy, or colonoscopy||Veterans in a VA General Medical Clinic aged ≥50 y who were not screen-compliant||Provider workshop on CRC screening plus feedback on individual and group specific rates||Direct cost of personnel and estimated overhead||32.4||8.9||Payer, 978|
|Chirikos 200423||FOBT||Patients in primary care clinics serving disadvantaged populations of men and women ages 50–75 y||Clinician reminder of whether a number of cancer screening tests were due and division of office responsibilities to ensure tests were ordered and completed by patients||Materials, personnel and patient time, and overhead||11.9||28.18||Payer, 10.50; societal, 11.53|
|Church 200435||Although FOBT kits were mailed, the primary interest was in overall screening rate, regardless of method||All eligible individuals (aged ≥50 y) living in a nonurban county in the Minneapolis-St. Paul area||Direct mailing of FOBT kits plus letter without follow-up reminder and with a series of mail and telephone follow-up reminders||Materials, personnel, and mailing cost of purchasing and mailing FOBT kits and sending reminders||55.8||6.4*||Payer, 100.47†|
CRC screening use was substantially and significantly higher in all study intervention groups compared with the control group but did not differ significantly among intervention groups. In the intervention groups, participants were exposed to the TI (ie, a mailed CRC screening invitation letter, an informational booklet, SBTs, and a reminder). The screening rate in the control group was 33%, whereas CRC screening rates in the intervention groups ranged from 44% to 48%. We believe that this report is important in the CRC screening literature, because primary care practice-based research studies that assess cost-effectiveness are uncommon. However, it is noteworthy that our effectiveness results were consistent with the results from randomized trials that have used a mailed SBT strategy, message tailoring, and reminders in other settings with different study populations.
The findings reported are consistent with the growing literature on the impact of print materials and mailed SBTs related to CRC screening in primary care settings. Mant et al.32 reported that CRC screening was highest among study participants who were sent SBTs in conjunction with an invitation to a wellness office visit. Similarly, Goldberg et al.33 reported that mailing SBTs to patients with a scheduled office visit significantly increased CRC screening use. In another study, Dietrich et al.34 observed that distributing print information regarding CRC screening at a clinic visit and providing a follow-up telephone call to encourage screening use served to increase CRC screening use significantly. In a community study of CRC screening in Minnesota, Church et al.35 reported that mailing SBTs withand without reminders increased overall CRC screening use.
The results reported here and in other studies support the use of targeted methods to increase CRC screening in primary care. Further research is needed to determine the degree to which telephone reminders can increase CRC screening use. Tailored CRC screening interventions have been tested outside of primary care settings with limited success.36, 37 Both targeted and tailored interventions have had significant effects, however, in population subgroups.38–40 In relation to intervention tailoring, it would be worthwhile to explore the impact of delivering messages related to factors that affect screening use.41 Exploring the impact of messages tailored to CRC screening preference is of particular interest.42–45
Other economic evaluations of CRC screening promotions in healthcare and community settings serve to provide a perspective for our results. Chirikos et al.23 reported results from a study of patients in primary care clinics that served disadvantaged populations of men and women ages 50 to 75 years. The primary intervention in their study included a clinician reminder of cancer screening tests due and the division of office responsibilities to ensure that tests were ordered and completed. Costs included personnel, materials, overhead, and patient time. The baseline screening rate was 12%. The ICER based on cost per additional individual screened was $10.50 from the payer perspective and $11.53 from the societal perspective. Adjustments were made in the estimates because of the joint promotion of several screening services. At the other end of the spectrum is a report by Wolf et al.24 on a screening-promotion program that was targeted to veterans in a Veterans Administration medical clinic aged > 50 years who were not compliant with screening guidelines. That intervention included a provider workshop on CRC screening plus feedback on individual and group-specific rates. Costs included those for personnel andoverhead. The baseline screening rate was 32%. The ICER based on cost per additional individual screened was $978 from a payer perspective. The authors estimated that the ICER could be reduced to approximately $200 by automating the information system.
Church et al.35 reported results from a study involving community residents in Minnesota who were targeted for a CRC screening educational campaign. In that study, residents who completed and returned a mailed baseline survey were assigned randomly to receive usual care only, mailed SBTs, or mailed SBTs with reminders. Data from a 1-year mailed endpoint survey was used to ascertain prospective CRC screening use. Baseline survey results indicated that 56% of study participants were adherent to CRC screening guidelines. At endpoint, self-reported screening in each of the study groups had increased from baseline as follows: Screening increased by 8% in the control group, by 13% in the mailed SBT group, and by 14% in the mailed SBT with reminders group. Costs included materials, personnel, and mailing and purchasing FOBT kits and sending reminders. The cost per additional individual screened with the mailed SBT plus reminders strategy was $100 from the payer perspective.
There are limitations to our study. First, the study was conducted with patients in only 1 primary care practice, limiting the extent to which study findings may be generalized. Second, actual delivery of telephone reminders was limited. We observed that only 72% of TIP group participants actually received 1 or more telephone reminders according to study protocol. Calls were not completed for the following reasons: participant not available for contact (37%), no answer (35%), disconnected telephone (16%), withdrawal from the study (7%), and no longer a JFMA patient (5%). Individuals who did not receive the telephone reminders may have differed from those who did receive the reminders. It also was the case that telephone reminders delivered to the TIP group were brief and focused on encouraging screening use rather than conveying tailored messages effectively. The potential impact of telephone contacts designed to amplify such messages is not known. Others may have more or less success with the interventions, depending on how well they can implement the strategies. The use of multiple information sources (ie, endpoint survey self-report, medical records, billing records) may have introduced some measure of bias to the outcomes assessment. We compared self-reported screening use across study groups and observed that the groups were comparable on this measure. Thus, it seems that differences in screening use across study groups are not likely to be attributable to self-report bias. The cost information is subject to the limitations of staff recall of time input, assumptions regarding the overhead cost rate, and inflation because of delivery of the interventions in the context of a research project and the higher costs associated with learning how to deliver a new program.
Although the accuracy of cost estimates would be improved with a fully allocated overhead cost model and prospective time input data collection, these were beyond the research budget for the current study, which was focused on the delivery of the intervention and prospective outcome data collection. All intervention strategies were subject to these possible biases; therefore, comparisons between the interventions, which were the basis of our efficiency analysis, should not be distorted. Average fixed costs for planning activities, development of materials, and programming for survey callers will decline as the target group is expanded in the short run (some costs are fixed). Target group size was selected to achieve statistical power for outcome assessment and not to achieve short-term or long-term efficiency (no costs are fixed). These issues are left for future research focused on production and cost function estimation and data from alternative size target populations and scale of operations. With due caution and adjustment to their own setting, decision makers may use these results for assessing potential effectiveness and relative cost-effectiveness of the interventions, although they may be able to deliver each of the interventions at a lower cost than presented here. Our bottom line is that mailed SBTs significantly boosted CRC screening use, and the cost of the SI for each additional individual screened was moderate compared with other screening-promotion interventions. Going beyond the TI to include tailoring or tailoring plus reminder telephone calls does not appear to be an economically attractive strategy.