BRIEF REPORT
Use of Interactive Voice Response to Improve Colorectal Cancer Screening Hannah Cohen-Cline, MPH,*w Karen J. Wernli, PhD,*z Susan C. Bradford, MPH,z Mary Boles-Hall, BA,z and David C. Grossman, MD, MPH*zy
Introduction: Only 63% of the US population aged 50–75 years is adherent with recommended colorectal cancer (CRC) screening guidelines. Efforts are needed to increase screening for CRC. Methods: We evaluated a quality improvement intervention conducted in 2010 at Group Health in Washington State to assess the use of interactive voice response (IVR) systems to improve CRC screening. Eligible members were aged 50–81 years, received primary care from a contracted physician, and were due for CRC screening. A total of 13,279 members were identified and randomly assigned to either: (1) an intervention with IVR to encourage CRC screening (n = 10,000), or (2) usual care (n = 3279). The primary outcome was receipt of any recommended CRC screening test at 6 months postintervention or index date. We used Cox proportional hazards to model time until receipt of CRC screening at 6 months. Results: Adherence to CRC screening was 10.0% in the intervention arm and 7.8% in the usual care arm at 6 months. Randomization to the IVR intervention arm was associated with a 32% increased likelihood of receiving CRC screening (hazard ratio = 1.32; 95% confidence interval, 1.14, 1.52) compared with usual care. The difference in CRC screening uptake was apparent by 3 months and persisted for 12 months postintervention (log-rank test, P = 0.0012). Discussion: Our analysis provides “real-world” evidence that IVR is effective when delivered by a commercial health plan, and may be a useful tool for increasing adherence to screening guidelines among patients outside an integrated care practice. Key Words: colorectal cancer screening, adherence, interactive voice response systems (Med Care 2014;52: 496–499)
C
olorectal cancer (CRC) is a leading incident cancer and cancer mortality in the United States.1,2 Overall, only 63% of the US adults report receiving CRC screening,
From the *Group Health Research Institute; Departments of wEpidemiology; zHealth Services, University of Washington; and yDepartment of Clinical Improvement and Prevention, Group Health, Seattle, WA. D.C.G. is a shareholder in Group Health Physicians, which contracts exclusively with Group Health Cooperative to provide medical services. The remaining authors declare no conflict of interest. Reprints: Karen J. Wernli, PhD, Group Health Research Institute, 1730 Minor Ave, Suite 1600, Seattle, WA 98101. E-mail: [email protected]. Copyright r 2014 by Lippincott Williams & Wilkins ISSN: 0025-7079/14/5206-0496
496 | www.lww-medicalcare.com
which is the lowest adherence level compared with other cancers.3 To improve adherence to CRC screening, outreach programs are turning to multiple ways to encourage participation. One promising method is the use of interactive voice response (IVR) systems, or automated telephone calls. With IVR technology, participants interact directly with a recorded voice linked to a computer database that can deliver health-related messages and store data received from the call.4 IVR systems offer low-cost solutions for contacting large populations. In managed care settings, 2 studies have evaluated the effectiveness of IVR to increase CRC screening and show conflicting results with one improving adherence to CRC screening and the other suggesting no association.5,6 There are few published studies that assessed the effectiveness of IVR in a nonmanaged care setting. The goals of our evaluation were to understand whether IVR could be effective to engage individuals overdue for CRC screening in community practice settings and to determine if the effect would persist over time.
METHODS We evaluated a pilot quality improvement program for CRC screening adherence comparing the use of IVR to usual care. The targeted population was insured by Group Health in Washington State and had a primary care physician who was a member in the contracted network of providers. The study protocol was approved by the Group Health Research Institute Institutional Review Board. The Committee granted a waiver of consent and HIPAA Authorization to access Group Health records to identify potential subjects and collect data to conduct this analysis.
Participants Men and women aged 50–81 years who were not adherent to CRC screening were eligible. We used administrative data (ie, claims of receipt of services) to determine adherence to CRC screening and defined not adherent as lacking: (1) a colonoscopy within 10 years; (2) a flexible sigmoidoscopy or barium enema within 5 years; or (3) a fecal occult blood test (FOBT) or fecal immunochemical test (FIT) within 12 months of the intervention date.
Intervention We identified a total of 13,279 eligible individuals. We randomized 10,000 individuals to the intervention and 3279 Medical Care
Volume 52, Number 6, June 2014
Medical Care
Volume 52, Number 6, June 2014
individuals to usual care. Because the intervention was originally implemented as a pilot quality improvement initiative, the decision was made to maximize the number of individuals who could receive the IVR intervention with the available resources. Telephone scripts and protocols for the calls were jointly developed by Group Health and Eliza Inc. (Boston, MA). The IVR calls were made over 6 weeks in winter 2010. The intervention was a single IVR telephone call (average length = 5 min) to the primary telephone number listed in the participant’s records. The call included the following features: (1) assessment of prior CRC screening; (2) information about the benefits of screening and elicitation of the barriers to screening; and (3) offer of a FOBT kit mailed to the participant’s home. The IVR call mentioned both FOBT and colonoscopy as recommended screening tests. If the IVR system left a message, only 1 additional message was sent. When there was no answer or a busy signal at the telephone number, up to 6 total attempts were made to reach the participant. When participants requested a FOBT kit, Group Health mailed the kit within 1 week. Participants returned the completed FOBT kit in-person to their physician’s office. Routinely, defined as usual care, a personalized outreach letter is mailed annually to all Group Health members before their birthday, informing them of upcoming preventive service needs, including cancer screening. Participants in both the intervention and usual care could have received the outreach letter at any point during the 12-month follow-up period near their birthday.
Use of IVR to Improve CRC Screening
randomized to receive the IVR call, 821 participants were not called because of incomplete contact information or contact information that was not unique to r3 participants and was assumed to be work phone numbers. Of the 8005 participants remaining in the intervention arm, a total of 2711 (33.9%) participants initiated the IVR call; of those who began the call, 1606 participants (59.2%) completed the call (Fig. 1). We describe patient characteristics by receipt of the intervention and usual care. We calculated P-values of difference of using the w2 tests for categorical variables and the t tests for differences in mean. We used Cox proportional hazards to evaluate the effectiveness of IVR on CRC screening. Person-time accumulated from the date of the IVR call until disenrollment, date of first CRC test, or 1 year from the call date. Participants in the usual care arm were randomly assigned an index date from the window of time that the IVR calls were conducted in winter 2010. Our comparison groups were those formed by randomization. To address the concerns of residual confounding, we a priori included age, sex, and prior CRC screening as covariates in the model. We assessed whether the effect of the IVR call on completion of CRC screening at 6 months was modified by age group (below 65 y, 65 y and above), sex, or receipt of prior CRC screening. Interaction terms were retained in the model only if they were statistically significant (P < 0.05).
13,279 in potentially eligible study population
Ascertainment of CRC Screening We determined receipt of any form of CRC screening from administrative data. The main outcome variable was the receipt of any recommended CRC screening (ie, colonoscopy, FOBT or FIT, flexible sigmoidoscopy, virtual colonoscopy, and barium enema) at 6 months follow-up to the IVR call identified through ICD-9, HCPC, and CPT codes. We evaluated all tests received within 12 months of the call date. In instances where >1 test occurred during follow-up, we selected the first test received. To evaluate uptake over time, we assessed receipt of CRC screening at 3, 9, and 12 months postintervention. Information on age (continuous), sex, residence (urban, large rural, small rural, isolated), insurance type (commercial, individual/family, Medicaid, Medicare), and years of enrollment in Group Health (continuous) were obtained from membership files. Race and ethnicity were imputed using an algorithm that relies on surname and geocoded residential address. The overall distribution using this algorithm has been shown to match with self-report race and ethnicity from the US Census.7
3,279 randomized to Usual Care
10,000 randomized to Intervention
Excluded n=274 members:
Excluded N=1,995 members:
•
•
• •
Disenrolled before "call" date (n=19) Were determined to have received CRC screening prior to study start (n=102) Had previous diagnosis of CRC, Crohn's disease, or colitis diagnosis (n=153)
• •
•
Disenrolled before call date (n=389) Were determine to have received CRC screening prior to the study start (n=325) Had previous diagnosis of CRC, Crohn's disease, or colitis diagnosis (n=460) Did not receive the intervention (n=821)
Did not participate in the intervention (N=5,294)
Statistical Analysis We excluded participants who: (1) disenrolled from Group Health before the call date for the intervention group or the index date for usual care; (2) were determined to have received screening before study start; and (3) had a prior diagnosis with CRC, Crohn disease, or colitis. Of those r
2014 Lippincott Williams & Wilkins
998 requested FOBT kit
FIGURE 1. Inclusion and exclusion criteria in the evaluation of IVR as an intervention to improve colorectal cancer screening. IVR indicates interactive voice response. www.lww-medicalcare.com |
497
Medical Care
Cohen-Cline et al
Volume 52, Number 6, June 2014
TABLE 1. Characteristics of Group Health Members Randomized to Intervention With IVR Call Compared With Usual Care in 2010 n (%)
Male Age, mean (SD) (y) White* Geographic location Urban Large rural Small rural Isolated Other Western Washington Continuous enrollment pre-IVR call, mean (SD) (y) Previous CRC screening Insurance Medicaid Medicare
Intervention (N = 8005)
Usual Care (N = 3005)
3371 (42.1) 60.5 (7.3) 6934.4 (86.6)
1394 (46.4) < 0.001 60.6 (7.3) 0.44 2584.5 (86.0) 0.06 0.15 2148 (71.5) 612 (20.4) 110 (3.7) 135 (4.5) 0 1535 (51.7) 0.10 8.0 (4.4) 0.29
5889 1539 254 319 3 3963 8.1
(73.6) (19.2) (3.2) (4.0) (50.0) (4.5)
2057 (25.7)
778 (25.9)
32 (0.4) 907 (11.3)
9 (0.3) 331 (11.0)
P
0.84 0.67
*Calculated from an algorithm based on geocoded address and surname. CRC indicates colorectal cancer; IVR, interactive voice response.
All P-values were 2-sided with significance level set at 0.05. All statistical analyses were performed using STATA release 12 (StataCorp., College Station, TX).
RESULTS There were no important differences in participant characteristics between the intervention and usual care arm after eligibility exclusions (Table 1). The sample was mostly women with mean age of 60 years, and an average length of enrollment of 8 years. About one quarter of members had prior CRC screening but were nonadherent. At 6 months, 10.0% of members in the intervention arm had received CRC screening compared with only 7.8% of members in the usual care arm. Randomization to the IVR intervention arm was associated with a 32% increase in the receipt of CRC screening [hazard ratio (HR) = 1.32; 95% confidence interval (CI), 1.14, 1.52] (Table 2). The effect of the intervention was strongest within the first 3 months (HR = 1.40; 95% CI, 1.16, 1.70) and remained relatively stable and statistically significant over time (HR at 12 mo = 1.20; 95% CI, 1.08, 1.34). Kaplan-Meier curves demonstrate that
FIGURE 2. Cumulative hazard of completion of the screening test within 365 days post-IVR call for intervention arm or index date for usual care arm. IVR indicates interactive voice response.
the intervention effect began almost immediately after the call and reached a maximum between 3 and 6 months (Fig. 2), and persisted until the end of follow-up (Kaplan-Meir log-rank, P = 0.0012). Sex, age group, and receipt of previous CRC screening did not significantly modify the association between IVR and receipt of CRC screening (data not shown). The distribution of type of screening test differed between the 2 groups (Table 3). Of those who received CRC screening, participants in the intervention arm were more likely to receive FOBT screening (36.1% compared with 28.2% at 6 mo) compared with the usual care group. Of the 998 members in the intervention arm who requested an FOBT kit, 273 had completed a test (27.3% return rate) (data not shown).
DISCUSSION Our results suggest that individuals not adherent to CRC screening who receive an IVR call from their health insurance plan are 32% more likely to get screened within 6 months relative to usual care alone. The increase in CRC screening occurred among men and women at all ages. At 6 months, the overall participation in CRC screening was low at 10.0% and 7.8% in the intervention and usual care arm, respectively. However, given that the eligible population was nonadherent to CRC screening guidelines,
TABLE 2. The Proportion of Participants Obtaining Colorectal Cancer Screening and Multivariate Cox Proportional Hazards Modeling Comparing Intervention to Usual Care Received Colorectal Cancer Screening [n (%)] Months Post-IVR Call
Intervention
3 6 9 12
483 803 1081 1329
(6.0) (10.0) (13.5) (16.6)
Usual Care 131 234 337 423
(4.4) (7.8) (11.2) (14.1)
HR
95% CI
1.40 1.32 1.23 1.20
1.16, 1.14, 1.09, 1.08,
1.70 1.52 1.39 1.34
P 0.001 < 0.001 0.001 0.001
Models are adjusted for age (continuous), sex (men/women), and receipt of previous screening (yes/no). CI indicates confidence interval; HR, hazard ratio; IVR, interactive voice response.
498 | www.lww-medicalcare.com
r
2014 Lippincott Williams & Wilkins
Medical Care
Volume 52, Number 6, June 2014
Use of IVR to Improve CRC Screening
TABLE 3. Description of the Types of Colorectal Cancer Screening Tests Completed by 6 Months [n (%)] Screening Test FOBT Colonoscopy Other Total (N)
Usual Care
Intervention
66 (28.2) 166 (70.9) 2 (< 1) 234
289 (36.1) 500 (62.4) 12 (1.5) 801
we would not expect their screening rates to be as high as those already participating in CRC screening. Although 2.2% is a small difference and may be statistically significant due to our large sample size, it still reflects an important difference at a population level in terms of a much larger absolute number of individuals who could be impacted by a call; for example, for health care organizations are trying to outreach to as many people using limited resources. Although results of studies of the effectiveness of IVR have been mixed, recent research suggests that IVR calls are useful in increasing adherence with preventative care.8,9 Mosen and colleagues evaluated IVR to provide information about CRC risk and screening and allowed participants to request a FOBT kit. The intervention increased screening adherence by 31% at 6 months post-IVR call compared with the control arm (HR = 1.31; 95% CI, 1.10, 1.56; P < 0.001), similar to our results. The study found a statistically significant interaction by age group, with participants aged 71–80 years showing a greater intervention effect than participants aged 51–60 years, a finding that we did not confirm.5 In contrast, Simon and colleagues evaluated an intervention that used a single IVR call to improve adherence by addressing attitudes and self-efficacy and providing information about CRC screening and risk. The intervention did not offer FOBT cards but stressed the importance of contacting a physician to discuss CRC screening. The study detected no difference in CRC screening rates at 12 months postintervention.6 In comparison with our results, the differences between these studies may indicate that IVR alone is not enough to increase CRC screening but should be paired with mailing a FOBT kit. Furthermore, our results may have been stronger if the participants mailed their FOBT kits directly for analysis instead of returning them in-person. Our overall return rate for FOBT kits requested by IVR was < 30%, but many of those individuals received screening with a colonoscopy when they did not return their kit. An additional finding is the importance of up-to-date contact information. We were unable to call 821 (8.2%) members because of missing contact information, and among those we tried to call, 1052 (13.1%) members were unreachable. Hence, some participants did not directly
r
2014 Lippincott Williams & Wilkins
experience the intervention and this identifies an area for improvement. Our evaluation had several strengths. By ascertaining CRC screening at several key time points, we were able to analyze the changes in the intervention effect over 1 year. In our setting, the IVR calls were made from the participant’s health insurance plan rather than their primary care provider, and our results provide evidence that IVR may be as effective when delivered in this setting. However, because we relied only on administrative claims data to confirm receipt of CRC screening, we did not have access to other health and socioeconomic information from the members’ electronic medical records. Moreover, the use of administrative codes does not allow for sufficient discrimination between screening versus diagnostic colonoscopy, hence some examinations could have been conducted due to symptomatic presentation. In addition, although we were able to use an algorithm to determine the probable distribution of race and ethnicity in the population, the race/ethnicity variable cannot be used in Cox regression, hence we were unable to include it in our final model.7 Finally, our results may not be generalizable to more racially diverse populations or individuals who are not native English speakers. IVR is one tool that could be used within a health care delivery system to increase adherence with preventive care guidelines. Further, it would be of interest to assess the impact of recurrent IVR calls to promote continued adherence to CRC screening guidelines. Such a study would be important for informing policy regarding implementing regular IVR calls in a health care delivery system or health plan. REFERENCES 1. American Cancer Society. Cancer Facts & Figures 2013. Atlanta, GA: American Cancer Society; 2013;64. 2. Schottenfeld D, Fraumeni JF. Cancer Epidemiology and Prevention. New York, NY: Oxford University Press; 2006. 3. Centers for Disease Control and Prevention. Vital signs: colorectal cancer screening among adults aged 50-75 years—United States, 2008. MMWR Morb Mortal Wkly Rep. 2010;59:808–812. 4. Piette JD. Interactive voice response systems in the diagnosis and management of chronic disease. Am J Manag Care. 2000;6:817–827. 5. Mosen DM, Feldstein AC, Perrin N, et al. Automated telephone calls improved completion of fecal occult blood testing. Med Care. 2010; 48:604–610. 6. Simon SR, Zhang F, Soumerai SB, et al. Failure of automated telephone outreach with speech recognition to improve colorectal cancer screening: a randomized controlled trial. Arch Intern Med. 2010;170:264–270. 7. Elliott MN, Fremont A, Morrison PA, et al. A new method for estimating race/ethnicity and associated disparities where administrative records lack self-reported race/ethnicity. Health Serv Res. 2008;43:1722–1736. 8. Oake N, Jennings A, van Walraven C, et al. Interactive voice response systems for improving delivery of ambulatory care. Am J Manag Care. 2009;15:383–391. 9. Griffin JM, Hulbert EM, Vernon SW, et al. Improving endoscopy completion: effectiveness of an interactive voice response system. Am J Manag Care. 2011;17:199–208.
www.lww-medicalcare.com |
499
Use of Interactive Voice Response to Improve Colorectal Cancer Screening Hannah Cohen-Cline, MPH,*w Karen J. Wernli, PhD,*z Susan C. Bradford, MPH,z Mary Boles-Hall, BA,z and David C. Grossman, MD, MPH*zy
Introduction: Only 63% of the US population aged 50–75 years is adherent with recommended colorectal cancer (CRC) screening guidelines. Efforts are needed to increase screening for CRC. Methods: We evaluated a quality improvement intervention conducted in 2010 at Group Health in Washington State to assess the use of interactive voice response (IVR) systems to improve CRC screening. Eligible members were aged 50–81 years, received primary care from a contracted physician, and were due for CRC screening. A total of 13,279 members were identified and randomly assigned to either: (1) an intervention with IVR to encourage CRC screening (n = 10,000), or (2) usual care (n = 3279). The primary outcome was receipt of any recommended CRC screening test at 6 months postintervention or index date. We used Cox proportional hazards to model time until receipt of CRC screening at 6 months. Results: Adherence to CRC screening was 10.0% in the intervention arm and 7.8% in the usual care arm at 6 months. Randomization to the IVR intervention arm was associated with a 32% increased likelihood of receiving CRC screening (hazard ratio = 1.32; 95% confidence interval, 1.14, 1.52) compared with usual care. The difference in CRC screening uptake was apparent by 3 months and persisted for 12 months postintervention (log-rank test, P = 0.0012). Discussion: Our analysis provides “real-world” evidence that IVR is effective when delivered by a commercial health plan, and may be a useful tool for increasing adherence to screening guidelines among patients outside an integrated care practice. Key Words: colorectal cancer screening, adherence, interactive voice response systems (Med Care 2014;52: 496–499)
C
olorectal cancer (CRC) is a leading incident cancer and cancer mortality in the United States.1,2 Overall, only 63% of the US adults report receiving CRC screening,
From the *Group Health Research Institute; Departments of wEpidemiology; zHealth Services, University of Washington; and yDepartment of Clinical Improvement and Prevention, Group Health, Seattle, WA. D.C.G. is a shareholder in Group Health Physicians, which contracts exclusively with Group Health Cooperative to provide medical services. The remaining authors declare no conflict of interest. Reprints: Karen J. Wernli, PhD, Group Health Research Institute, 1730 Minor Ave, Suite 1600, Seattle, WA 98101. E-mail: [email protected]. Copyright r 2014 by Lippincott Williams & Wilkins ISSN: 0025-7079/14/5206-0496
496 | www.lww-medicalcare.com
which is the lowest adherence level compared with other cancers.3 To improve adherence to CRC screening, outreach programs are turning to multiple ways to encourage participation. One promising method is the use of interactive voice response (IVR) systems, or automated telephone calls. With IVR technology, participants interact directly with a recorded voice linked to a computer database that can deliver health-related messages and store data received from the call.4 IVR systems offer low-cost solutions for contacting large populations. In managed care settings, 2 studies have evaluated the effectiveness of IVR to increase CRC screening and show conflicting results with one improving adherence to CRC screening and the other suggesting no association.5,6 There are few published studies that assessed the effectiveness of IVR in a nonmanaged care setting. The goals of our evaluation were to understand whether IVR could be effective to engage individuals overdue for CRC screening in community practice settings and to determine if the effect would persist over time.
METHODS We evaluated a pilot quality improvement program for CRC screening adherence comparing the use of IVR to usual care. The targeted population was insured by Group Health in Washington State and had a primary care physician who was a member in the contracted network of providers. The study protocol was approved by the Group Health Research Institute Institutional Review Board. The Committee granted a waiver of consent and HIPAA Authorization to access Group Health records to identify potential subjects and collect data to conduct this analysis.
Participants Men and women aged 50–81 years who were not adherent to CRC screening were eligible. We used administrative data (ie, claims of receipt of services) to determine adherence to CRC screening and defined not adherent as lacking: (1) a colonoscopy within 10 years; (2) a flexible sigmoidoscopy or barium enema within 5 years; or (3) a fecal occult blood test (FOBT) or fecal immunochemical test (FIT) within 12 months of the intervention date.
Intervention We identified a total of 13,279 eligible individuals. We randomized 10,000 individuals to the intervention and 3279 Medical Care
Volume 52, Number 6, June 2014
Medical Care
Volume 52, Number 6, June 2014
individuals to usual care. Because the intervention was originally implemented as a pilot quality improvement initiative, the decision was made to maximize the number of individuals who could receive the IVR intervention with the available resources. Telephone scripts and protocols for the calls were jointly developed by Group Health and Eliza Inc. (Boston, MA). The IVR calls were made over 6 weeks in winter 2010. The intervention was a single IVR telephone call (average length = 5 min) to the primary telephone number listed in the participant’s records. The call included the following features: (1) assessment of prior CRC screening; (2) information about the benefits of screening and elicitation of the barriers to screening; and (3) offer of a FOBT kit mailed to the participant’s home. The IVR call mentioned both FOBT and colonoscopy as recommended screening tests. If the IVR system left a message, only 1 additional message was sent. When there was no answer or a busy signal at the telephone number, up to 6 total attempts were made to reach the participant. When participants requested a FOBT kit, Group Health mailed the kit within 1 week. Participants returned the completed FOBT kit in-person to their physician’s office. Routinely, defined as usual care, a personalized outreach letter is mailed annually to all Group Health members before their birthday, informing them of upcoming preventive service needs, including cancer screening. Participants in both the intervention and usual care could have received the outreach letter at any point during the 12-month follow-up period near their birthday.
Use of IVR to Improve CRC Screening
randomized to receive the IVR call, 821 participants were not called because of incomplete contact information or contact information that was not unique to r3 participants and was assumed to be work phone numbers. Of the 8005 participants remaining in the intervention arm, a total of 2711 (33.9%) participants initiated the IVR call; of those who began the call, 1606 participants (59.2%) completed the call (Fig. 1). We describe patient characteristics by receipt of the intervention and usual care. We calculated P-values of difference of using the w2 tests for categorical variables and the t tests for differences in mean. We used Cox proportional hazards to evaluate the effectiveness of IVR on CRC screening. Person-time accumulated from the date of the IVR call until disenrollment, date of first CRC test, or 1 year from the call date. Participants in the usual care arm were randomly assigned an index date from the window of time that the IVR calls were conducted in winter 2010. Our comparison groups were those formed by randomization. To address the concerns of residual confounding, we a priori included age, sex, and prior CRC screening as covariates in the model. We assessed whether the effect of the IVR call on completion of CRC screening at 6 months was modified by age group (below 65 y, 65 y and above), sex, or receipt of prior CRC screening. Interaction terms were retained in the model only if they were statistically significant (P < 0.05).
13,279 in potentially eligible study population
Ascertainment of CRC Screening We determined receipt of any form of CRC screening from administrative data. The main outcome variable was the receipt of any recommended CRC screening (ie, colonoscopy, FOBT or FIT, flexible sigmoidoscopy, virtual colonoscopy, and barium enema) at 6 months follow-up to the IVR call identified through ICD-9, HCPC, and CPT codes. We evaluated all tests received within 12 months of the call date. In instances where >1 test occurred during follow-up, we selected the first test received. To evaluate uptake over time, we assessed receipt of CRC screening at 3, 9, and 12 months postintervention. Information on age (continuous), sex, residence (urban, large rural, small rural, isolated), insurance type (commercial, individual/family, Medicaid, Medicare), and years of enrollment in Group Health (continuous) were obtained from membership files. Race and ethnicity were imputed using an algorithm that relies on surname and geocoded residential address. The overall distribution using this algorithm has been shown to match with self-report race and ethnicity from the US Census.7
3,279 randomized to Usual Care
10,000 randomized to Intervention
Excluded n=274 members:
Excluded N=1,995 members:
•
•
• •
Disenrolled before "call" date (n=19) Were determined to have received CRC screening prior to study start (n=102) Had previous diagnosis of CRC, Crohn's disease, or colitis diagnosis (n=153)
• •
•
Disenrolled before call date (n=389) Were determine to have received CRC screening prior to the study start (n=325) Had previous diagnosis of CRC, Crohn's disease, or colitis diagnosis (n=460) Did not receive the intervention (n=821)
Did not participate in the intervention (N=5,294)
Statistical Analysis We excluded participants who: (1) disenrolled from Group Health before the call date for the intervention group or the index date for usual care; (2) were determined to have received screening before study start; and (3) had a prior diagnosis with CRC, Crohn disease, or colitis. Of those r
2014 Lippincott Williams & Wilkins
998 requested FOBT kit
FIGURE 1. Inclusion and exclusion criteria in the evaluation of IVR as an intervention to improve colorectal cancer screening. IVR indicates interactive voice response. www.lww-medicalcare.com |
497
Medical Care
Cohen-Cline et al
Volume 52, Number 6, June 2014
TABLE 1. Characteristics of Group Health Members Randomized to Intervention With IVR Call Compared With Usual Care in 2010 n (%)
Male Age, mean (SD) (y) White* Geographic location Urban Large rural Small rural Isolated Other Western Washington Continuous enrollment pre-IVR call, mean (SD) (y) Previous CRC screening Insurance Medicaid Medicare
Intervention (N = 8005)
Usual Care (N = 3005)
3371 (42.1) 60.5 (7.3) 6934.4 (86.6)
1394 (46.4) < 0.001 60.6 (7.3) 0.44 2584.5 (86.0) 0.06 0.15 2148 (71.5) 612 (20.4) 110 (3.7) 135 (4.5) 0 1535 (51.7) 0.10 8.0 (4.4) 0.29
5889 1539 254 319 3 3963 8.1
(73.6) (19.2) (3.2) (4.0) (50.0) (4.5)
2057 (25.7)
778 (25.9)
32 (0.4) 907 (11.3)
9 (0.3) 331 (11.0)
P
0.84 0.67
*Calculated from an algorithm based on geocoded address and surname. CRC indicates colorectal cancer; IVR, interactive voice response.
All P-values were 2-sided with significance level set at 0.05. All statistical analyses were performed using STATA release 12 (StataCorp., College Station, TX).
RESULTS There were no important differences in participant characteristics between the intervention and usual care arm after eligibility exclusions (Table 1). The sample was mostly women with mean age of 60 years, and an average length of enrollment of 8 years. About one quarter of members had prior CRC screening but were nonadherent. At 6 months, 10.0% of members in the intervention arm had received CRC screening compared with only 7.8% of members in the usual care arm. Randomization to the IVR intervention arm was associated with a 32% increase in the receipt of CRC screening [hazard ratio (HR) = 1.32; 95% confidence interval (CI), 1.14, 1.52] (Table 2). The effect of the intervention was strongest within the first 3 months (HR = 1.40; 95% CI, 1.16, 1.70) and remained relatively stable and statistically significant over time (HR at 12 mo = 1.20; 95% CI, 1.08, 1.34). Kaplan-Meier curves demonstrate that
FIGURE 2. Cumulative hazard of completion of the screening test within 365 days post-IVR call for intervention arm or index date for usual care arm. IVR indicates interactive voice response.
the intervention effect began almost immediately after the call and reached a maximum between 3 and 6 months (Fig. 2), and persisted until the end of follow-up (Kaplan-Meir log-rank, P = 0.0012). Sex, age group, and receipt of previous CRC screening did not significantly modify the association between IVR and receipt of CRC screening (data not shown). The distribution of type of screening test differed between the 2 groups (Table 3). Of those who received CRC screening, participants in the intervention arm were more likely to receive FOBT screening (36.1% compared with 28.2% at 6 mo) compared with the usual care group. Of the 998 members in the intervention arm who requested an FOBT kit, 273 had completed a test (27.3% return rate) (data not shown).
DISCUSSION Our results suggest that individuals not adherent to CRC screening who receive an IVR call from their health insurance plan are 32% more likely to get screened within 6 months relative to usual care alone. The increase in CRC screening occurred among men and women at all ages. At 6 months, the overall participation in CRC screening was low at 10.0% and 7.8% in the intervention and usual care arm, respectively. However, given that the eligible population was nonadherent to CRC screening guidelines,
TABLE 2. The Proportion of Participants Obtaining Colorectal Cancer Screening and Multivariate Cox Proportional Hazards Modeling Comparing Intervention to Usual Care Received Colorectal Cancer Screening [n (%)] Months Post-IVR Call
Intervention
3 6 9 12
483 803 1081 1329
(6.0) (10.0) (13.5) (16.6)
Usual Care 131 234 337 423
(4.4) (7.8) (11.2) (14.1)
HR
95% CI
1.40 1.32 1.23 1.20
1.16, 1.14, 1.09, 1.08,
1.70 1.52 1.39 1.34
P 0.001 < 0.001 0.001 0.001
Models are adjusted for age (continuous), sex (men/women), and receipt of previous screening (yes/no). CI indicates confidence interval; HR, hazard ratio; IVR, interactive voice response.
498 | www.lww-medicalcare.com
r
2014 Lippincott Williams & Wilkins
Medical Care
Volume 52, Number 6, June 2014
Use of IVR to Improve CRC Screening
TABLE 3. Description of the Types of Colorectal Cancer Screening Tests Completed by 6 Months [n (%)] Screening Test FOBT Colonoscopy Other Total (N)
Usual Care
Intervention
66 (28.2) 166 (70.9) 2 (< 1) 234
289 (36.1) 500 (62.4) 12 (1.5) 801
we would not expect their screening rates to be as high as those already participating in CRC screening. Although 2.2% is a small difference and may be statistically significant due to our large sample size, it still reflects an important difference at a population level in terms of a much larger absolute number of individuals who could be impacted by a call; for example, for health care organizations are trying to outreach to as many people using limited resources. Although results of studies of the effectiveness of IVR have been mixed, recent research suggests that IVR calls are useful in increasing adherence with preventative care.8,9 Mosen and colleagues evaluated IVR to provide information about CRC risk and screening and allowed participants to request a FOBT kit. The intervention increased screening adherence by 31% at 6 months post-IVR call compared with the control arm (HR = 1.31; 95% CI, 1.10, 1.56; P < 0.001), similar to our results. The study found a statistically significant interaction by age group, with participants aged 71–80 years showing a greater intervention effect than participants aged 51–60 years, a finding that we did not confirm.5 In contrast, Simon and colleagues evaluated an intervention that used a single IVR call to improve adherence by addressing attitudes and self-efficacy and providing information about CRC screening and risk. The intervention did not offer FOBT cards but stressed the importance of contacting a physician to discuss CRC screening. The study detected no difference in CRC screening rates at 12 months postintervention.6 In comparison with our results, the differences between these studies may indicate that IVR alone is not enough to increase CRC screening but should be paired with mailing a FOBT kit. Furthermore, our results may have been stronger if the participants mailed their FOBT kits directly for analysis instead of returning them in-person. Our overall return rate for FOBT kits requested by IVR was < 30%, but many of those individuals received screening with a colonoscopy when they did not return their kit. An additional finding is the importance of up-to-date contact information. We were unable to call 821 (8.2%) members because of missing contact information, and among those we tried to call, 1052 (13.1%) members were unreachable. Hence, some participants did not directly
r
2014 Lippincott Williams & Wilkins
experience the intervention and this identifies an area for improvement. Our evaluation had several strengths. By ascertaining CRC screening at several key time points, we were able to analyze the changes in the intervention effect over 1 year. In our setting, the IVR calls were made from the participant’s health insurance plan rather than their primary care provider, and our results provide evidence that IVR may be as effective when delivered in this setting. However, because we relied only on administrative claims data to confirm receipt of CRC screening, we did not have access to other health and socioeconomic information from the members’ electronic medical records. Moreover, the use of administrative codes does not allow for sufficient discrimination between screening versus diagnostic colonoscopy, hence some examinations could have been conducted due to symptomatic presentation. In addition, although we were able to use an algorithm to determine the probable distribution of race and ethnicity in the population, the race/ethnicity variable cannot be used in Cox regression, hence we were unable to include it in our final model.7 Finally, our results may not be generalizable to more racially diverse populations or individuals who are not native English speakers. IVR is one tool that could be used within a health care delivery system to increase adherence with preventive care guidelines. Further, it would be of interest to assess the impact of recurrent IVR calls to promote continued adherence to CRC screening guidelines. Such a study would be important for informing policy regarding implementing regular IVR calls in a health care delivery system or health plan. REFERENCES 1. American Cancer Society. Cancer Facts & Figures 2013. Atlanta, GA: American Cancer Society; 2013;64. 2. Schottenfeld D, Fraumeni JF. Cancer Epidemiology and Prevention. New York, NY: Oxford University Press; 2006. 3. Centers for Disease Control and Prevention. Vital signs: colorectal cancer screening among adults aged 50-75 years—United States, 2008. MMWR Morb Mortal Wkly Rep. 2010;59:808–812. 4. Piette JD. Interactive voice response systems in the diagnosis and management of chronic disease. Am J Manag Care. 2000;6:817–827. 5. Mosen DM, Feldstein AC, Perrin N, et al. Automated telephone calls improved completion of fecal occult blood testing. Med Care. 2010; 48:604–610. 6. Simon SR, Zhang F, Soumerai SB, et al. Failure of automated telephone outreach with speech recognition to improve colorectal cancer screening: a randomized controlled trial. Arch Intern Med. 2010;170:264–270. 7. Elliott MN, Fremont A, Morrison PA, et al. A new method for estimating race/ethnicity and associated disparities where administrative records lack self-reported race/ethnicity. Health Serv Res. 2008;43:1722–1736. 8. Oake N, Jennings A, van Walraven C, et al. Interactive voice response systems for improving delivery of ambulatory care. Am J Manag Care. 2009;15:383–391. 9. Griffin JM, Hulbert EM, Vernon SW, et al. Improving endoscopy completion: effectiveness of an interactive voice response system. Am J Manag Care. 2011;17:199–208.
www.lww-medicalcare.com |
499
