Physicians must incorporate patients’ preferences into their recommendations to practice patient-centered medicine effectively (Rittenhouse & Shortell, 2009). To do this effectively, physicians not only need to empathize with patients’ worries and understand their tastes (Levinson, Lesser, & Epstein, 2010) but also counteract their potential biases (Redelmeier, Rozin, & Kahneman, 1993). Consciously or not, physicians may use their own experience as healthcare consumers as a starting point for understanding their patients’ perspective, especially in the case of routine preventive procedures that the physicians have experienced. Conversely, asking what a physician would do for herself is a “reasonable question from the patient's perspective, as knowing the stocks your broker is buying” (Earl, 2008). Although physicians may be unwilling to share such personal information in person, anonymous surveys provide an alternative means for tallying such personal opinions. The current study takes this approach, examining the relationship between personal practices and patient recommendations in the context of breast cancer screening.
When making choices for others in situations involving potential losses, people tend to choose more cautiously than when choosing for themselves (Atanasov, 2013). This pattern is especially pronounced in the medical decision-making context (Ubel, Angott, & Zikmund-Fischer, 2011; Zikmund-Fisher, Sarr, Fagerlin, & Ubel, 2006). Physicians tend to recommend more risk-averse treatments for their patients than the patients would choose for themselves, and patients are often surprised by the high level of risk-aversion exhibited by physicians, as well as the self-other differences in physician decisions (Garcia-Retamero & Galesic, 2012). Given the high impact of physician recommendations on eventual treatment decisions, (Quill & Brody, 1996) it is important to examine whether the gap in risk preferences for self versus patients demonstrated in hypothetical choice studies are also reflected in the actual physician practices.
In the context of prevention, physicians are more likely to recommend lifestyle choices similar to their own. Physicians who engage in healthy behaviors, such as dieting, exercise, and cholesterol testing, are more likely to recommend these behaviors to their patients (Brotons et al., 2005; Frank, Rothenberg, Lewis, Brooke, & Belodoff, 2000). Most research on the relationship between personal choices and professional recommendations has focused on choices with proven benefits and minimal costs, other than patient effort. Less research has examined medical procedures with potential downsides, such as diagnostic tests with high false-positive rates or those with radiation exposure. For example, regular exercise and cholesterol screening is generally considered beneficial for most adults, but the costs and benefits of commencing annual mammography screening from age 40 are subject to debate.
The context of this study, breast cancer screening, is suitable for the examination of personal and professional risk preferences for at least two reasons. First, most female physicians are likely to have personally considered their screening options, and to remember their past choices. Second, cancer-screening guidelines differ among professional associations, leaving physicians with several defensible options. The diversity of professional opinions was underscored by the U.S. Preventive Services Task Force's (USPSTF) recent recommendations that regular mammography screening should generally commence at the age of 50, as well as the guideline that teaching patients to perform breast self-examination does not yield significant survival benefits (Editors, 2010; Nelson, Tyne, Naik, Bougatsos, Chan, & Humphrey, 2009).
An important assumption underlying the study hypotheses, and their relationship to the literature, is that undergoing diagnostic testing is the risk-averse choice from the physician and patients’ perspective. Screening involves a small but certain utility loss, linked to financial costs, time, pain, and anxiety. Additional costs may include false-positive diagnoses. The ultimate goal of screening is the reduction of long-term serious health risks. Whether the risk-averse choice is also the better one depends on the balance of evidence about costs and benefits of the procedure and is not beyond the scope of this study.
Our first hypothesis was that female physicians would be less likely to personally undergo cancer-screening procedures than to recommend these procedures to patients of their age and with a similar medical history. Results of hypothetical choices studies in the medical domain are consistent with the hypothesized pattern—higher risk aversion for patients than for self (Ubel et al., 2011; Zikmund-Fisher et al., 2006). We also hypothesized that physicians who have recently undergone a given procedure will be more likely to recommend it to patients like themselves, even after accounting for age and medical history. This hypothesis was motivated by the literature on personal lifestyle choice and patient recommendations.
The survey instrument was sent to a quasi-randomly selected sample of 500 obstetricians and gynecologists from the Collaborative Ambulatory Research Network (CARN), a group of practicing obstetricians and gynecologists belonging to the American College of Obstetricians and Gynecologists (ACOG), who have volunteered to complete two to five research studies each year.
The survey was developed by the authors. The document was limited to one double-sided page to minimize the time burden associated with completing it. The first page consisted of demographic questions such as gender, age, race, and state of residence. Physicians were also asked whether they would recommend preventive procedures to patients of their “age, with a medical/family history similar to” theirs. The second page included questions about physicians’ personal medical practices. The response scales for both professional and personal practices were binary (yes/no). The survey also included questions about the number of breast self-examinations (BSEs) performed by the physician in past 2 years and the number of mammograms undergone in the past 5 years. Thus, personal prevention choices and patient recommendations were compared within subjects. Frequency of defensive medicine practices was estimated by asking physicians how often they tend to recommend more screening than medically necessary due to malpractice concerns1 (Anderson, Strunk, & Schulkin, 2011). The response scale varied from 0 (never) to 10 (always).
The first mailing was sent in February 2011, and was followed by a second mailing in April 2011. Enclosed in the mailing was the survey questionnaire, a cover letter explaining the purpose of the study, and a postage-paid envelope. The study protocol was reviewed and approved by the Institutional Review Board at ACOG.
All data were analyzed using Statistical Analysis System (SAS) version 9.3. The rates of personal use and recommendations for each procedure were compared using McNemar's matched samples test (McNemar, 1947). The Wilcoxon nonparametric test was used to compare the frequency of screening by physician age group, due to the nonnormality of the screening frequency distribution. Logistic regression models were constructed to examine the multivariate relationship between the outcome measure, patient recommendations, and predictors—personal practices, demographic characteristics, and self-reported defensive medicine practices.
The response rate was 54.8% (274 out of 500) overall, and 54.0% (135 out of 250) among women. The current study focused exclusively on data from the 135 female respondents, in order to directly compare physicians’ patient recommendations with their personal practices. Respondents’ average age was 49, 86% considered themselves generalists, and 85% provided breast care. The mean number of breast care patients per week was 31, which translated to about 2,800 breast care patients per week for the subsample sample of female physicians who responded to this question. About three in ten respondents (28%) reported breast cancer history in genetic relatives. More detailed demographic information is reported in Table 1.
Guidelines and Recommendations
Table 2 presents the guidelines published by major medical associations, along with the percentage of physicians who would recommend the procedure to a patient like themselves, and the percentage who reported having used the procedures at least once in the past 5 years. USPSTF guidelines are notably more conservative for both mammography screening and breast self-examination. ACOG and the American Cancer Society recommend that annual mammography screening starting should start age 40, and that physicians should decide whether to teach breast self-examination techniques based on specific circumstances. In the cases of disagreement about guidelines (e.g., mammography screening between ages 40 and 49, BSE), physicians tended to side with the professional guidelines favoring more screening, especially in their recommendations to patients.
Among female physicians younger than 40, 40% had undergone mammography screening at least once in the past 5 years, whereas 35% reported that they would recommend the procedure to patients of their age and with similar medical history, a nonsignificant difference (McNemar's test, S < 1.00, p > .20). Among all female physicians aged 40 and above (N = 108), the rate of mammography recommendations (95%) was significantly higher than the rate of personal use (89%, S = 5.44, p = .020). The main source of disparity was the subsample of women aged 40–49 (N = 45), 82% of whom had undergone the procedure, though significantly more (91%) reported that they would recommend it to a patient like themselves (McNemar's test, S = 4.00, p = .045). Among female physicians aged 50 and older (N = 57), virtually all (98.3%) reported recommending mammography screening, and most had undergone the procedure (93.0%). These rates did not differ significantly (S = 1.80, p = .18).
Frequency of screening provides additional detail about personal practices, distinguishing between those who practiced regular annual screening and those who report undergoing only one screening in the past 5 years. Despite ACOG and ACS recommendations for annual screening between ages 40 and 50, only 18% of physicians in our sample 40s reported undergoing mammography screening at least once per year for the past five. Among women in their 50s and 60s, 48% had performed screenings at least annually. Reported frequency of screening varied widely by age group (Figure 1). Most notably, women aged 45–49, who had been in the ACS and ACOG recommended screening age group during the past 5 years, had undergone screenings approximately every year and 10 months (N = 17, M = 0.54, SD = 0.27), whereas women in their 50s had undergone screening at a significantly higher rate (N = 43, M = 0.76, SD = 0.37), corresponding to one screening every year and 4 months (Cohen's d = 0.70, Wilcoxon test, z = 2.46, p = .014).
Most female physicians reported that they would recommend breast self-examination to patients, and a marginally higher proportion reported having performed the procedure on themselves in the past 2 years (N = 108, 86% vs. 92%, S = 2.67, p = .10). These small differences were consistent across age groups. The median physician respondent had performed 12 BSEs over the past year.
Predictors of Patient Recommendations
In addition to comparing how many physicians would undergo versus recommend screening, we examined who is more or less likely to recommend the procedures. The results of the logistic regression analyses predicting patient recommendations are shown in Table 3. Two main patterns were evident in both mammography screening and BSEs. First, there was a strong relationship between personal practices and patient recommendations (OR > 5.00, p < .01 for both procedures): physicians who had personally used a procedure were significantly more likely to recommend it. Second, physicians’ self-reported defensive medicine concerns did not significantly predict patient recommendations (p > .10 for all procedures).
In addition to age, personal use of procedure was the best predictor of mammography screening recommendations. The main source of variance in recommendations was the subsample of women below 50. Among respondents above 50, recommendations were almost uniformly in favor of the procedure. Breast cancer family history and defensive medicine practices were not significant predictors. Controlling for all these variables, physicians in private or community practice were more likely to recommend mammography screening and breast self-examination than those practicing at university hospitals and government facilities. Distance to the closest mammography screening facility, a proxy for convenience, did not predict frequency of personal screening or recommendations. Geographic region, physician's ethnicity, and subspecialty were not associated with patient recommendations. Sensitivity analyses with various combinations of demographic predictors (not shown) confirmed the basic results: strong association of recommendation with personal practices, and weak association with self-reported defensive medicine practices.
For breast self-examination, the number of BSEs performed by physicians within the past 2 years was the only significant predictor of recommending the procedure. Age was a marginally significant positive predictor, whereas breast cancer family history and defensive medicine practices were not.
The large majority of physicians recommended screening practices that they personally followed. However, in cases of disagreement, physicians tended to recommend practices they did not follow, rather than follow screening patterns they did not recommend. Screening mammography in women above 40, the starting age for reimbursement under the Patient Protection and Affordable Care Act (PPACA) (Sebelius, 2010) fits this pattern. The variance of patient recommendations across physicians was not explained by self-reported defensive medicine practices.
The current results show a distinct tendency to recommend aggressive breast cancer screening in line with American Cancer Society and the ACOG guidelines, rather than relatively conservative recommendations issued by the USPSTF. However, mammography recommendations were somewhat less aggressive compared to previous surveys, whereas personal practices more conservative still.
The proportion of ob-gyns recommending mammography screening to women of ages 40–49 in our sample (92%) was somewhat lower than recent survey estimates from a 2006–2007 survey, in which 99% of ob-gyns of both genders and across age groups recommended the procedure (Meissner, Klabunde, Han, Benard, & Breen, 2011). The largest and most important difference between personal decisions and professional opinions was in screening frequency. Only 18% of physicians in their 40s and 48% of those aged 50–69 had undergone mammography screening annually in the past 5 years. In comparison, Meissner et al. (2011) found that 80% recommended annual screening for women aged 40–49, and 98% recommended this frequency for women aged 50 or above. A recent survey provides a more direct comparison with the current results, because it used a sample (CARN) comparable to ours, and was administered after the publication of the USPSTF guidelines. In this sample, 50% of physicians recommended annual screening to women of ages 40–49 who chose to have such screening, and 97% recommended annual screening to women 50–69 years of age (Anderson, Pearlman, Griffin, & Schulkin, 2013).
This between-study comparison suggests that there are substantial differences between personal choices and professional opinions about screening frequency. In the current study, the within-subjects design may have resulted in a conservative estimate of these differences. The timing of our study may also have diminished self-other differences, because the questions regarding patient recommendations focused on the present, whereas personal health practices focused on the past 2–5 years. Given the recent tendency to favor more conservative mammography screening practices, (Anderson et al., 2013) it is possible that physicians’ personal preferences at present may be more conservative than their past behaviors.
There are several possible explanations for the differences between professional and personal practices. First, physicians who recommended mammography screening but did not undergo the procedure may believe that screening does not yield a favorable balance between costs and benefits in their own circumstances, but may find it difficult to explain this to patients, or may fear that patients interpret such recommendations as lack of concern for their health. This explanation would be consistent with the literature documenting higher levels of caution for patients than for self (Zikmund-Fisher et al., 2006). The higher rate of mammography recommendations by physicians in private and community practice, whose livelihood is highly dependent on maintaining positive relationships with patients, was also consistent with this account.
Second, physicians may believe they can selfdiagnose better and do not need frequent mammography screening. This explanation is consistent with the marginally higher personal rate of BSEs. However, respondents who performed more BSEs did not use mammography screening less often, so it is unlikely that one procedure was used as a substitute for the other. Third, physicians are human, and like most humans, they may plan to undergo annual mammography screenings, but may be unable to follow up on these plans, given the significant time pressure they face. This explanation is inconsistent with the evidence that convenience, using distance to the nearest mammography screening facility as a proxy, was unrelated to the likelihood or frequency of screening. However, busy schedules may limit screening frequency even if the screening facility is conveniently located. Despite the differences in levels between recommendations and personal practices, the two were strongly and positively correlated: physicians who had personally undergone a procedure were more likely to recommend it to patients like themselves. This pattern may reflect physicians’ underlying beliefs about the costs and benefits of mammography screening and BSE in their age groups. Alternatively, it may be driven by respondents’ belief that their patients’ preferences are similar to their own.
The pattern of responses for breast self-examinations was particularly interesting in light of the controversies surrounding the procedure. In December 2009, the USPSTF issued a D-grade recommendation “against teaching breast self-examination,” whereas ACOG has retained their recommendation that the procedure should be considered for some patients, and the National Cancer Institute continues to recommend the procedure. Proponents of BSE claim that between “nearly 70%” (National Breast Cancer Foundation website) of breast cancers are found through self-exams, although no peer-reviewed source supporting this statistic was referenced. Critics cite randomized controlled trials, which have demonstrated that teaching self-examination increases the rate of tumor detection and the number of biopsies performed, but has no significant impact on cancer or all-cause mortality (Hackshaw & Paul, 2003).
Clinical breast examination, which may be a closer approximation of physician-performed BSE, received an I grade by the USPSTF, denoting that the current evidence is insufficient to assess its benefits and harms. The high rates of personal practices and professional recommendations for BSEs represent a notable example of physicians’ reluctance to adjust their practices in the direction of less testing in response to new evidence from clinical trials regarding the limited benefits of procedures (Kösters & Gøtzsche, 2011). For procedures like breast self-examination, which are associated with little short-term cost or harm, evidence against the procedure's benefits may need to be especially strong in order to induce behavior change.
Several limitations should be acknowledged. First, the survey asked about past behaviors and may thus be subject to recall biases. For example, respondents may have wanted to appear consistent between personal practices and recommendations. This may have diminished the differences in relative levels between personal use and professional recommendations, and made the correlation between the two stronger. Second, 80% of the physicians were Caucasian, which may limit the generalizability of the study to other ethnicities. Lastly, the survey did not directly examine all possible reasons, rational or not, for differences between personal practices and recommendations. Future research should examine this question in further detail.
Obstetricians and gynecologists in our sample tended to favor aggressive cancer-screening guidelines for their patients. However, the personal choices they had made in the recent past showed a tendency toward less testing, especially in the case of mammography screening in the 40s. More generally, surveying physicians’ personal practices represents a useful tool to measure their attitudes toward medical procedures. It provides physicians with an opportunity to safely express their personal preferences, in a complex professional environment featuring conflicting guidelines and defensive medicine concerns. Future research should elucidate the stability of self-other differences in physician decisions and measure the impact of such discrepancies on their ability to deliver balanced and consistent messages about screening recommendations.
This project was funded by grant, UA6MC1901[ZERO WIDTH SPACE]0, through the U.S. Department of Health and Human Services, Health Resources and Services Administration, Maternal and Child Health Research Program. The funding agreement ensured the authors’ independence in designing the study, interpreting the data, writing, and publishing the report.
1 Additional questions were included in the survey, but the responses were not analyzed in this study. These questions addressed preventive practices for young women, such as human papillomavirus vaccination and pap smear testing. This was due to the study's focus on personal decisions, and the make-up of our sample, women aged 35 and above, respondents could only answer the questions in regard to their young female relatives.
Anderson B., Pearlman M., Griffin J., Schulkin J. Conflicting and changing breast cancer screening recommendations: Survey study of a national sample of ob-gyns after the release of the 2009 USPSTF guidelines. Journal of Healthcare Quality 2013;35(4):25–35.
Anderson B., Strunk A., Schulkin J. Study on defensive medicine practices among obstetricians and gynecologists who provide breast care. Journal of Healthcare Quality 2011;33:37–43.
Atanasov P. The losses of others: Meta-analysis of choices for self and others under uncertainty. 2013. Manuscript in preparation.
Brotons C., Björkelund C., Bulc M., Ciurana R., Godycki-Cwirko M., Jurgova E., et al.. Prevention and health promotion in clinical practice: The views of general practitioners in Europe. Preventive Medicine 2005;4:595–601.
Earl C. Do unto others. Journal of Clinical Oncology 2008;24:3331–3332.
Editors. When evidence collides with anecdote, politics, and emotion: Breast cancer screening. Annals of Internal Medicine 2010;152:531–532.
Frank E., Rothenberg R., Lewis C., Brooke F., Belodoff B. Correlates of physicians’ prevention-related practices, findings from the women physicians’ health study. Archives of Family Medicine 2000;9:359–367.
Garcia-Retamero R., Galesic M. Doc, what would you do if you were me? On self-other discrepancies in medical decision making
. Journal of Experimental Psychology-Applied 2012;18(1):38.
Hackshaw A., Paul E. Breast self-examination and death from breast cancer: A meta-analysis. British Journal of Cancer 2003;88:1047–1053.
Kösters J., Gøtzsche P. Regular self-examination or clinical examination for early detection of breast cancer. Cochrane Database of Systematic Reviews 2011;10.
Levinson W., Lesser C., Epstein R. Developing physician communication skills for patient-centered care. Health Affairs 2010;29:1310–1318.
McNemar Q. Note on the sampling error of the difference between correlated proportions or percentages. Psychometrika 1947;12:153–157.
Meissner H., Klabunde C., Han P., Benard V., Breen N. Breast cancer screening beliefs, recommendations, and practices. Cancer 2011;117:3101–3111.
Nelson H.D., Tyne K., Naik A., Bougatsos C., Chan B., Humphrey L. Screening for breast cancer: An update for the U.S. Preventive Services Task Force. Annals of Internal Medicine 2009;151:727–737.
Quill T.E., Brody H. Physician recommendations and patient autonomy: Finding a balance between physician power and patient choice. Annals of Internal Medicine 1996;125:763–769.
Redelmeier D., Rozin P., Kahneman D. Understanding patients’ decisions. Cognitive and emotional perspectives. Journal of the American Medical Association 1993;270:72–76.
Rittenhouse D., Shortell S. The patient-centered medical home. Will it stand the test of health reform? Journal of the American Medical Association 2009;301:2038–2040.
Ubel P., Angott A., Zikmund-Fischer B. Physicians recommend different treatments for patients than they would choose for themselves. Archicves of Internal Medicine 2011;7:630–634.
Zikmund-Fisher B., Sarr B., Fagerlin A., Ubel P. A matter of perspective: Choosing for others differs from choosing for yourself in making treatment decisions. Journal of General Internal Medicine 2006;21:618–622.
Pavel Atanasov, PhD, is currently a postdoctoral fellow at the University of Pennsylvania, where he completed his PhD in Psychology in 2012. His research focuses on decision making for self and others, behavioral economics, and crowd-based forecasting.
Britta L. Anderson, PhD, is Research Associate at the American College of Obstetricians and Gynecologists. She completed her PhD in Psychology in 2011 at American University. Her research focuses on medical decision making, numeracy, and women's health.
Joanna Cain, MD, is Professor of Obstetrics and Gynecology and Vice Chair for Faculty Development at the University of Massachusetts Medical School. She is a well-known leader in the field. Previously, Dr. Cain was Chair at Brown University.
Jay Schulkin, PhD, is Senior Director and Director of Research at the American College of Obstetricians and Gynecologists. He is also Research Professor in the department of obstetrics and gynecology at the University Washington School of Medicine.
Jason Dana, PhD, is currently a visiting Assistant Professor at Yale School of Management. He completed his PhD at Carnegie Mellon's department of social and decision sciences. His research focuses on judgment and decision making, quantitative psychology, and behavioral economics.