Monday, October 21, 2013: 1:00 PM - 2:15 PM
Key Ballroom 8,11,12 (Hilton Baltimore)
Category Reference for Presentations
AHEApplied Health Economics DECDecision Psychology and Shared Decision Making
HSPHealth Services, and Policy Research METQuantitative Methods and Theoretical Developments

* Candidate for the Lee B. Lusted Student Prize Competition

Session Chairs:
Jessica Ancker, MPH, PhD and Pei-Jung Lin, Ph.D.
1:00 PM
Ann Partridge, MD, MPH1, Karen R. Sepucha, PhD2, Anne O'Neill, M.S.3, Kathy D. Miller, M.D.4, Emily L. Baker3, Chau T. Dang, M.D.5, Donald W. Northfelt, M.D.6, George W. Sledge Jr., M.D.4 and Bryan P. Schneider, M.D.4, (1)Dana-Farber Cancer Institute, Boston, MA, (2)Massachusetts General Hospital, Boston, MA, (3)Dana Farber Cancer Institute, Boston, MA, (4)Indiana University Cancer Center, Indianapolis, IN, (5)Memorial Sloan-Kettering Cancer Center, New York, NY, (6)Mayo Clinic, Scottsdale, AZ
Purpose: Biomarker information can risk stratify patients based on potential for benefit/toxicity from therapy. Ideally, a biomarker will identify those who benefit with limited or no toxicity. However, for some medicines, such as bevacizumab, early biomarker studies suggest that patients who may benefit also have increased toxicity.  The purpose of this study was to examine how biomarker information would impact patients' preferences for therapy in this situation.

Method:  We surveyed participants at the 18 month follow-up assessment in a large, international double blind randomized controlled trial, ECOG5103. For this trial, participants with breast cancer were randomized to receive adjuvant chemotherapy with either placebo or bevacizumab.  We asked patients for their preferred treatment (either chemotherapy A alone or chemotherapy A+B) in two hypothetical scenarios: 1) baseline without biomarker information; and 2) after learning that they tested positive for a “B-receptor” which increased both the benefit and toxicity of chemotherapy A+B. The risk information was given in both numerical (table) and graphical (100-person pictograph) format. We asked participants for the main reason for their choice. McNemar’s test was used to examine changes in treatment preferences.

Result: 439 patients completed both scenarios on 18-month survey. Table 1 shows the participants’ treatment preferences in each scenario. The increase in benefit and toxicity associated with the positive biomarker information in scenario 2 led 60/439 (14%) participants to switch their preference. Among participants who changed preference, those randomized to receive bevacizumab were more likely to switch to chemotherapy A in scenario 2. Among all participants, the main reason reported for their treatment preference in scenario 2 was greater benefits of chemotherapy A+B (64%), the lower risks with chemotherapy A (20%) and positive biomarker (10%). 

Table 1: Participants’ treatment preferences in scenario 1 and 2 (chemo=chemotherapy)



Scenario 2: With “positive B-receptor”




Preferred chemo A

Preferred chemo A+B


Scenario 1: Without

biomarker information

Preferred chemo A




Preferred chemo A+B









Conclusion:  Information about a positive biomarker indicating increased benefit and increased toxicity from additional chemotherapy did not change many participants’ preferred treatment. The majority (70%) preferred the most aggressive course of treatment in both scenarios. Whether patients not enrolled in the trial would be more sensitive to the increased toxicity information is unclear.

1:15 PM
Michael Pignone, MD, MPH1, Stephanie B. Wheeler, PhD, MPH1, Sarah T. Hawley, PhD, MPH2, Carmen Lewis, MD, MPH1, Trisha Crutchfield, MHA, MSIS1, Kristen Hassmiller Lich, PhD, MHSA1, Paul M. Brown, PhD, MS3, Ravi K. Goyal, MS1, Emily Gillen, MA1 and Jane Laping, MS, MPH1, (1)University of North Carolina at Chapel Hill, Chapel Hill, NC, (2)University of Michigan, Ann Arbor VA Health System, Ann Arbor, MI, (3)University of California, Merced, Merced, CA
Purpose: To use a discrete choice experiment (DCE) to learn about how vulnerable adults in North Carolina value different aspects of colorectal cancer (CRC) screening programs.

Methods: We used prior research, focus groups, and expert opinion to develop a DCE that examined four key attributes of potential CRC screening programs: 1) choice of screening tests offered (fecal occult blood testing (FOBT) alone, colonoscopy (COL) alone, choice of FOBT and COL, choice of FOBT, COL or radiological screening); 2) travel distance required to obtain screening (0, 15, 30, or 45 miles); 3) co-payment or reward for having screening ($1000 co-payment, $100 co-payment, $25 co-payment, $0 copayment, $10 reward, or $100 reward); 4) proportion of follow-up costs paid out of pocket (0%, 5%, 50%, 100%). We then used Sawtooth software to generate a 16-task DCE. After pilot testing to ensure comprehension, we enrolled a sample of English-speaking average-risk adults ages 50-74. Participants were recruited from rural NC communities with low rates of CRC screening, had either no insurance or only public insurance, and had low or fixed incomes. They received basic information about CRC screening and potential program features, then completed the DCE and survey questions. We analyzed DCE responses using Hierarchical Bayesian methods to produce group and individual-level part-worth utilities for the 4 attributes, and also calculated individual-level importance scores. Each individual’s highest importance score was considered the “DCE-calculated most important attribute.”

Results: We enrolled 150 adults. Mean age was 55.9 (range 50-74), 55% were women, 76% White and 19% African-American; 87% had annual household income under $30,000; and 51% were uninsured. From the survey, proportion of out of pocket follow-up costs was most frequently reported to be most important (42% of participants); testing options was next most frequent (32%).  From the DCE, follow-up cost was most frequently found to be the DCE-calculated most important attribute (49%), followed by screening test reward/co-payment (43%). Agreement between the survey and DCE-calculated most important attribute was modest (45%). On both the survey and DCE, participants valued having test choice and the opportunity to choose FOBT.

Conclusions: Screening test rewards /co-payments and follow-up costs are important program characteristics, particularly for vulnerable populations. Programs to encourage screening should take these factors into account to be most effective.

1:30 PM
Julia J. van Tol-Geerdink, PhD1, Jan Willem H. Leer, MD, PhD2, Carl Wijburg, MD, PhD3, Inge M. van Oort, MD, PhD4, Henk Vergunst, MD, PhD5, Emile J. van Lin, MD, PhD2, J. Alfred Witjes, MD, PhD4 and Peep F.M. Stalmeier, PhD6, (1)Radboud University Nijmegen Medical Centre, Nijmegen, Netherlands, (2)Dept. Radiation Oncology, Radboud Univ. Medical Centre, Nijmegen, Netherlands, (3)Dept. Urology, Rijnstate Hospital,, Arnhem, Netherlands, (4)Dept. Urology, Radboud Univ. Medical Centre, Nijmegen, Netherlands, (5)Dept. Urology, Canisius Wilhelmina Hospital,, Nijmegen, Netherlands, (6)Dept. Health Evidence, Radboud Univ. Medical Centre, Nijmegen, Netherlands

Implementation of decision aids in medical decision making is still low, partly because of fear that involving patients could have a negative impact. This study focuses on the effect of increasing patient involvement, by means of a decision aid, on regret in the context of the treatment choice for prostate cancer.


Between 2008 and 2011, patients with localized prostate cancer were individually randomized to 1) usual care (n=77) and 2) usual care plus a discussion on risks and benefits of different treatment options by means of a decision aid (N=163). The treatments options were radical prostatectomy, external beam radiotherapy and brachytherapy. This was a multicenter trial (3 sites) with imbalanced randomization (1:2). The primary outcome measure was regret, which was assessed before, and 6 and 12 months after treatment with the regret scale of Brehaut(1), and with three newly developed regret scales focusing on process regret, option regret and outcome regret. Additional information was gathered on patient characteristics, participation, knowledge, anxiety, healthy-related quality-of-life and decision evaluation.


The decision aid increased patient participation (P=0.002) and subjective knowledge (P=0.006). The effect of the decision aid was comparable on the three aspects of regret, but seemed to differ between patients with or without serious morbidity at 12 months. In patients with serious side effects, the use of a decision aid resulted in a trend to less option regret (P=0.075) and less Brehaut regret (P=0.061), with an effect size of 0.35 to 0.38, respectively.


Our data suggest that the decision aid had little effect on regret in patients without serious side effects, but it may lead to less regret in patients with serious side effects.

1:45 PM
Heather L. Morris, BA, MS and Jennifer Elston Lafata, PhD, Virginia Commonwealth University, Richmond, VA
Purpose: The US Preventive Services Task Force, among others, has endorsed goal setting as a central component of health promotion. We describe patient-physician goal setting discussions during period health examinations (PHEs) as well as compare the mean length of visits with and without an explicit goal setting discussion.  

Method: Observational study of 485 PHEs to 64 primary care physicians practicing in a large, southeast Michigan Health System between 2007-2009.  Previously collected data from office visit audio recordings and direct observation were combined with pre-visit patient surveys and administrative records for patient and physician characteristics, respectively.  Office visit transcripts were queried in Microsoft Word to identify any use of the word “goal.”  Among visits containing the word “goal,” the first author used a structured worksheet to code goal-related discussions for context (health vs. other) as well as specific patient-centered and collaborative communication techniques, including those developed by Street et al. (2001) and Heisler et al (2003).  


Among the 485 visits, 98 goal discussions were identified, 59 of which were health-related.  These health-related goals occurred in 49 (10%) visits to 30 different physicians.  Visits with Caucasian patients were more likely to contain a health-related goal discussion (51%/n=30 followed by African Americans 37%/n=22) as were visits to female physicians (71%/n=42 vs. 29%/n=17, p=0.03), but no differences in the likelihood of health-related goal discussion was detected by patient gender, age, body mass index or health status, nor physician age, race or specialty.  The most commonly occurring health-related goals were those related to weight loss (30%), blood pressure (14%), and physical activity (10%).  Over half of goals (56%) were physician set, with the remainder evenly split between patient- and collaboratively-set.  Most (72%) included a discussion of goal-attainment strategies, but relatively few (<37%) contained discussion of the benefits of goal-attainment, patients’ beliefs, or physician partnership building.  When a health-related goal was discussed, the mean time with the physician significantly increased (30.2 vs. 26.6 minutes, p=0.01). 

Conclusion: Despite patients and physicians advocating the utility of PHEs for goal setting, we found only a minority of visits included an explicitly labeled goal.  Furthermore, barely a quarter of those reflected collaboratively set goals. How to foster effective, collaborative goal setting during busy primary care office visits remains an important topic for study.

2:00 PM
Joseph A. Ladapo, MD, PhD, NYU School of Medicine, NY, NY

Purpose: Physicians routinely perform cardiac stress testing to diagnose and risk-stratify patients suspected of having coronary heart disease (CHD), but its use has come under intense scrutiny recently because of concerns about explosive growth, the contribution of this growth to high and rising healthcare costs, and potential patient harms related to radiation exposure from radionuclide imaging. However, it is unknown whether trends in its utilization may be attributable to changing population demographics and clinical risk factors.

Method: We analyzed nationally representative cross-sections of adult ambulatory patient visits in the United States, using a sample of 872,498 visits from the National Ambulatory Medical Care Survey and National Hospital Ambulatory Medical Care Survey from 1993 to 2010. Patients were excluded if they had prior history of CHD. The main outcomes were survey-weighted measures of referrals for or performance of cardiac stress testing, with or without imaging. Multivariable logistic regression was used to test time trends, with year modeled as a linear predictor, and we included clinical risk factors (smoking, diabetes, dyslipidemia, and hypertension), an indicator for whether the patient's reason for visiting the physician was chest pain, physician specialty, and sociodemographic characteristics, including age, gender, race/ethnicity, geographic location, and insurance status.

Result: The annual number of adult ambulatory visits that resulted in a cardiac stress test being ordered or performed increased steadily from 1.86 million in 1993-1997 (31 in 10,000 visits) to 3.89 million in 2006-2010 (46 in 10,000 visits). After adjusting for clinical and sociodemographic characteristics, there was no evidence of a time trend in stress testing (P=0.48 for trend). However, the portion of cardiac stress tests performed with imaging increased from 64% (95% CI 48%-79%) in 1993-1997 to 87% (95% CI 79%-96%) in 2006-2010 (P<0.001 for trend).

Conclusion: The growth in physicians' use of cardiac stress testing does not appear to represent dynamic changes in over-testing or overuse. However, the increasing use of cardiac stress testing with imaging, much of which exposes patients to radiation, is unlikely to be related to changing patient demographics or clinic risk factors. It therefore supports concerns voiced by professional societies and insurers about inappropriate utilization of cardiac imaging.

Description: Macintosh HD:Users:jladapo:Dropbox:berger:smdm-stress testing:figure-volume of stress tests.png