- Email: [email protected]
Evaluation of the “Angelina Jolie Effect” on Screening Mammography Utilization in an Academic Center
ORIGINAL ARTICLE
HEALTH SERVICES RESEARCH AND POLICY
Evaluation of the “Angelina Jolie Effect” on Screening Mammography Utilization in an Academic Center Marco D. Huesch, MBBS, PhD a , Susann Schetter, DO a, Joel Segel, PhD b, Alison Chetlen, DO a Abstract Purpose: The aim of this study was to understand the impact on screening mammography at our institution, comparing weekly utilization in the 2 years before and the 2 years after Ms Angelina Jolie disclosed in the New York Times on May 13, 2013, that she had had a prophylactic double mastectomy. Methods: All 48,110 consecutive screening mammograms conducted at our institution between May 16, 2011, and May 16, 2015, were selected from our electronic medical record system. We used interrupted time series statistical models and graphical methods on utilization data to understand utilization changes before and after Ms Jolie’s news. Results: The graphed trend of weekly screening mammogram utilization failed to show changes around the time of interest. Analytical models and statistical tests also failed to show a step change increase or acceleration of utilization around May 2013. However, graphical and time series analyses showed a flattening of utilization in the middle of 2014. Conclusions: In our well-powered analysis in a large regional breast imaging center, we found no support for the hypothesis that this celebrity news drove increased screening. Key Words: Screening, breast cancer, celebrity news, mammography, influence, time series analysis J Am Coll Radiol 2017;14:1020-1026. Copyright 2017 American College of Radiology
Patients can potentially be influenced by celebrity medical information in the media. Understanding whether such celebrity influence exists is important because this influence can lead to positive and negative public health effects. Consider an early positive example involving celebrity breast cancer. In 1973 and 1974, respectively, the wife of the president and the wife of the vice president publicly disclosed news of breast cancer [1]. Ford’s lesson for other women was said to be “straightforward: get a mammogram, which she had not done” [2]. The
a
Department of Radiology, Milton S. Hershey Medical Center, Hershey, Pennsylvania. b Department of Health Policy and Administration, Pennsylvania State University, University Park, Pennsylvania. Corresponding author and reprints: Marco D. Huesch, MBBS, PhD, Department of Radiology, Milton S. Hershey Medical Center, 500 University Drive, Mailcode H-066, Hershey, PA 17033-0850; e-mail: [email protected]. The authors have no conflicts of interest related to the material discussed in this article.
influence of these disclosures was strongly associated with contemporaneous increases in breast examination consultations [3] and to increases in the still not widely available mammograms [4], as well as more diagnoses [5] and more treatment of breast cancer [6]. Similarly positive public health impacts of celebrity news can be found in other settings. After news in the traditional media of President Reagan’s colon cancer, early detection with fecal occult blood tests increased [7]. In 2000 after the death of her husband from colon cancer, Katie Couric underwent a colonoscopy live on the Today Show, which led to a short-term spike in colonoscopy screening [8]. However, less benevolent effects are also possible. Misleading media information could drive suboptimal consumer decisions. One example is celebrity influence on vaccination rates based on autism fears [9]. Widespread media reports of increased risk with vaginal births after Cesarean sections were associated with a rapid drop in the rate of this procedure [10]. Another example lies in the context of hereditary breast cancers. ª 2017 American College of Radiology
1020
1546-1440/17/$36.00
n
http://dx.doi.org/10.1016/j.jacr.2017.03.016
Although increasing prevention and screening are desirable in appropriate risk strata, overuse of BRCA gene screening in patients with a low pretest probability would be undesirable [11]. More than one-third of US women with BRCA1 or BRCA2 mutations already choose prophylactic mastectomy, more than in many other countries [12]. Overuse of nonmedically indicated prophylactic surgery in women with lower risk would be another concern [13]. To understand the potential impact (positive, negative, or none) of celebrity news, the case of Angelina Jolie may be important. She revealed her prophylactic double mastectomy for BRCA1 gene marker status in the New York Times on May 13 (online) and May 14 (print) [14]. This highly publicized news had a global impact [15]. Risk monitoring with screening MRI and risk reduction with prophylactic mastectomy may be appropriate for women with the BRCA1 mutation and others of this risk stratum [16], but would be inappropriate for women in general choosing to follow Ms Jolie’s lead without knowledge of their level of lifetime risk. Indeed, Ms Jolie hoped that women “will be able to get gene tested, and that if they have a high risk they, too, will know that they have strong options” [14]. Although patient-centered care requires consideration of screening regimens and therapies that are respectful of patient preferences [17], fear of cancer may increase the trend toward aggressive risk-reducing procedures. For example, a blogger wrote on the same day as Ms Jolie’s article: “Angelina Jolie went from having an 87% chance of getting breast cancer to a 5% chance. . . . If I am accepted for BRCA testing, and my results show that I, too, am carrying the BRCA gene, I am certain my decision will be crystal clear as well” [18]. Some existing research investigates the “Angelina effect.” In our previous work, we showed that at the time of the New York Times article, social media had displayed dramatic but transient increases in mentions of breast cancer in the United States [19]. Others have also shown increases in Internet searches related to breast screening both in the United States and internationally [20,21]. Such effects would have clearly positive public health impacts, because increased screening for breast cancer would save lives [22-29]. Yet to our knowledge there has been no research into utilization of screening and little research into utilization of genetic testing or bilateral risk-reducing mastectomies. Two studies in the United Kingdom show dramatic and almost immediate increases in referral for genetic testing from clinics and genetics centers [15] and Journal of the American College of Radiology Health Services Research and Policy n Huesch et al
n
subsequently in the same population for bilateral riskreducing mastectomies [30]. Although genetic testing referrals were judged to be appropriate based on internal review [15], subsequent surgeries in the next 6 to 24 months increased among women regardless of mutation carrier status and was greater among those without mutations [30]. Finally, a recent national study among commercially insured US women also found significant increases in BRCA genetic testing after the Jolie editorial [31]. However, this study found no overall increase in mastectomies and some evidence that incremental genetic testing may have been among women with a lower pretest probability of a BRCA mutation [31]. We hypothesized that there would be an Angelina effect in screening mammography. Our rationale was 3-fold, based on patient salience, physician salience, and workup processes. We anticipated that underscreened or unscreened women may be influenced to start appropriate screening by the celebrity news leading to greater attention to, and awareness of, the prevalence of breast cancer and the risks of not screening. Similarly, appropriately screened women on an annual cycle or underscreened women on a 2-year screening cycle might bring future screening schedules forward. We also anticipated that referring physicians might be influenced to discuss or recommend mammography with their patients after reading this widely publicized news. Finally, as part of the workup among high-risk or potentially high-risk women, a first step would be a screening mammogram if this had not already been done. In this study, we investigate this hypothesis of increased utilization of screening mammography and report on our institutional experience with screening mammography in the 2 years before and after Ms Jolie’s public disclosure.
METHODS We followed complex but well-established analytical methods in wide use to understand changes over time in a noisy series of data [32-34]. We first used graphical methods with smoothed trend lines to allow immediate visual inspection of possible change in utilization. Weekly and even daily volume is usually very variable over time as schedules and capacity changes and may be thought of as random noise. Over longer horizons, utilization may follow a so-called secular trend at some slowly increasing pace, reflecting slow changes in insurance, technology, preferences, and demographics.
Angelina Jolie Effect
1021
We are interested in a potential structural break, independent of that short-run noise, and adjusting for that long-term trend. Such a structural break is an increase, or decrease, or acceleration up or down that is associated with the time point of the hypothesized news item. Accordingly, we used two well-established complementary methods to understand any changes in screening mammography utilization after the New York Times article appeared: interrupted time series analysis and a closely related time series analysis method. The first explicitly models a potential step change and slope change in trend around a known date. The second considers the whole time series and seeks to identify a structural change in the data, such as an increase, or decrease, or acceleration in the trend, up or down.
Interrupted Time Series Analysis Three key points in this method are that we needed to (1) control for any secular trend in utilization over time, (2) measure any immediate and enduring step change in the average utilization after the time point of interest, and (3) measure any change in the slope of the trend in utilization in the new time period, compared to the prior one before the time point of interest. We built a regression model in which the dependent variable was weekly screening mammography and the independent variables were (1) the week of the service (to control for secular trends), (2) an indicator variable that took in each week the value 0 before and 1 after Ms Jolie’s news (to estimate immediate and enduring step changes), and (3) a variable that took the value 0 for every week before the change and took successive positive integers in the weeks after the changes (to estimate changes in the slope of the trend after the time point of interest). We also allowed an indicator for seasonality at the quarterly level, and in one sensitivity test also at the monthly level. We specified robust standard errors to account for likely heteroskedasticity in the data over time. We computed t tests for the coefficients of the two key variables in (2) and (3) above, to understand whether they are independently significant. Finally, we computed a Chow test for a structural break around the May 2013 time point in the data [32]. This tested whether the variables in (2) and (3) above are jointly zero. A significant value indicates that it would be unlikely for both variables to be zero and that there is a significant association between screening volume and the time point of the celebrity news item. 1022
Time Series Analysis In this complementary analysis, the model was not set up to, nor did it force the data to conform to, a particular structural break occurring at a particular time point. Rather, the model allowed the data to suggest a possible time point, examining all possible such time points. We tested the suggested break point using a Wald test. A significant test statistic suggests that the null hypothesis of no structural break can be rejected. We obtained deidentified data through our institution’s Millennium Cerner (Cerner Corporation, Kansas City, MO, United States) electronic medical record system’s billing system for all mammography in the 4-year period bisected by the key date of May 13, 2013. For our main analysis on screening mammography, we selected professional billing Current Procedural Terminology codes G020226 for “digital screening mammography.” For sensitivity analyses in which we wished to understand relative changes in screening volumes compared with a diagnostic procedure unlikely to be affected by celebrity news, we selected G020626 for “unilateral diagnostic mammography” and G020426 for “bilateral diagnostic mammography.” Subtotals per day and per week for each service were derived and yearly totals confirmed with separate breast center management information center reports on production volume. This study was HIPAA compliant and approved by the Pennsylvania State University College of Medicine Institutional Review Board with determination number 00005728. RESULTS In our institution, a total of 48,110 digital screening mammograms were performed between May 16, 2011, and May 16, 2015, over a total of 208 weeks. We aggregated highly variable daily volumes (see Appendix Fig. 1a) into weeks because no services were performed on Sunday and very few on Saturday. As Figure 1 shows, there was substantial noise even on a weekly basis. A lowess smoothed trend line suggested a gently sloping increasing trend in the 2 years before the New York Times article, and a slightly flatter slope afterward, tending to plateau around July 2014, and then declining slightly. In Appendix Figure 1b, we show similar results for the daily screening volume in the 2 weeks before and after the news release. These graphical findings are clearly not consistent with the hypothesized structural break, which we proceeded to formally test with a series of statistical models. Journal of the American College of Radiology Volume 14 n Number 8 n August 2017
Weekly fullfield digital screening mammogram patient totals 350
Smoothed trendline (Lowess smoother, bandwidth 0.8)
250 200 100
150
Weekly volume
300
May 13, 2013; publication of Ms Jolie's NYT article
All 208 consecutive service weeks from 5-16-2011 through 5-16-2015; Hershey Medical Center
Fig 1. Weekly full-field digital screening mammograms.
Interrupted Time Series Analysis We constructed a series of progressively saturated models to regress on weekly screening mammogram volume. Adding weekly time from the Jolie article, in the second column of Table 1, produced a reasonably wellpredicting model in which increased time was associated with increased volume in a highly significant manner. This linear trend was precisely estimated across all the models and represents the average trend toward higher utilization across the 4 years. However, adding the step change indicator did not meaningfully improve the
model’s fit, and the coefficient on this indicator was not statistically different from zero. This suggests no immediate and enduring step change in volumes accrued after the celebrity news publication. Even despite no immediate step change, it is possible that previous trends accelerated (leading to a steeper, more positive slope) or decelerated (leading to a flatter less positive slope) or even reversed (leading to a declining, negative slope). The addition of a slope indicator in the model captured these possible changes. Table 1 shows that the addition of the slope indicator was associated with a modest increase in explanatory
Table 1. Interrupted time series analysis Null Model Constant Weekly time Step change Indicator Slope change Indicator Quarter indicator Month indicator Observations R2 (%)
231.3 (<.001)
Add Weekly Time
Add Step Change Indicator
Add Slope Change Indicator
Add Quarter Seasonality
Add Month Seasonality
212.4 (<.001) 0.18 (<.001)
214.1 (<.001) 0.13 (.08) 6.5 (.44)
206.0 (<.001) 0.29 (.002) 6.7 (.43)
169.2 (<.001) 0.38 (<.001) 2.6 (.76)
165.6 (<.001) 0.37 (<.001) 1.8 (.83)
0.31 (.04)
0.31 (.02)
0.31 (.02)
12.9 (<.001)
24.5 (.002) 3.9 (.11)
28.3
29.1
208 weeks 0.0
9.8
9.9
11.8
Results are estimates (P values for the t test statistic).
Journal of the American College of Radiology Health Services Research and Policy n Huesch et al
n
Angelina Jolie Effect
1023
power (R2 improved from 9.9% to 11.8%) and the coefficient on this indicator was precisely estimated and significant at a conventional level of P < .05. The sign on the coefficient was negative, implying a decrease in slope in the post period, consistent with the graphical presentation. Finally, adding indicators for seasonality showed significantly higher volumes associated with the fourth quarter compared with earlier quarters in any year. These were driven by Breast Cancer Awareness month in October, as the four transient spikes in Figure 1 show. Beyond this quarterly seasonality, there was no significant effect of addition of a monthly seasonality indicator. The high R2 in the last two models suggests well-fitting models. We tested the joint hypothesis that the step-change indicator and the slope change indicator are jointly zero in a Chow test with a P value of .067. The insignificant result suggested that the null hypothesis of no structural break in May 2013 could not be rejected.
Time Series Analysis In a separate analysis, we did not impose the constraints on the data that the interrupted time series analysis imposed (by insisting on one single break in the data at one particular time point). By allowing the time series to be tested for structural breaks at any of the 208 weekly points over the 4 years, we allowed other identified breaks to arise flexibly. We used the model in the last column of Table 1, leaving out the indicators for step and slope changes. After regression, we had a well-fitting model with an R2 of 27.3%. Based on all the 208 possible breaks, we then tested for the single most significant “unknown structural break,” if any. This analysis suggested a break in the first week of July 2014, with P values less than .02 across each of the supremum Wald, average Wald, and average likelihood ratio tests. It is important to point out that April or May 2013 were not identified as significant structural breaks. Examining the graph in Figure 1 shows that July 2014 is the center of the plateau before the volumes decrease through early 2015. Robustness Analyses Unlike screening mammography, diagnostic mammography utilization is less likely at the discretion of a patient compared with the decision of a physician, and so it is arguably less likely to be driven by celebrity news. However, there is likely to be similar short-run noise, and potentially similar long-run secular changes over time 1024
affect its utilization. This affords an opportunity to control directly for such noise in a model similar to a difference-in-difference design [35]. Graphically we show these different trends in Appendix Figure 2. In a separate analysis, we subtracted the diagnostic mammography volume each week from the screening mammography volume and repeated the interrupted time series analysis. In unreported results, there was little difference except a slightly better-fitting model (R2 rises to 36.7%) and a more precisely estimated negative sign on the slope indicator. Finally, we used a linear regression model for what is actually a count model [36]. The dependent variable can only take positive integer values depending on the week’s utilization. When we performed a similar analysis using an overdispersed negative binomial regression that more correctly modeled counts, we found qualitatively unchanged results (not reported).
DISCUSSION In our institution, we unexpectedly observed no Angelina Jolie effect in the utilization of screening mammography. We found no statistically significant increase in utilization after compared with before a prominent celebrity publicly disclosed BRCA gene status and bilateral risk-reducing mastectomies. Our findings in an adequately powered study fail to support our hypothesis that such a change was likely based on arguments of salience and workup processes. It is unlikely that women became more aware of breast cancer but preferred not to act on this awareness. For example, the risk might have been judged as not substantial enough. However, in related settings, even non-BRCA carriers have undertaken risk-reducing procedures for ovarian cancers, presumably to allay concerns of cancer risk [37]. In other related settings, addressing women’s legitimate fears and need for peace of mind has likely led to unnecessary contralateral prophylactic mastectomies [38], despite National Comprehensive Cancer Network guidelines that discourage such procedures except for BRCA mutation carriers [39]. Recent studies show that among women receiving a contralateral prophylactic mastectomy together with mastectomy of the affected breast, 70% did not have a clinical reason for having one done [40,41]. It is also unlikely that women became more aware of breast cancer risks but were unable to act on this because of capacity constraints in our institution. Our institution had sufficient capacity to flexibly accommodate shifts in demand as large as 50% in total volume from one month to the next. Journal of the American College of Radiology Volume 14 n Number 8 n August 2017
Finally, it is unlikely that other influences confounded the effects we focused on. For example, any confounding national policy effects such as the controversial United States Preventive Services Task Force recommendations in 2009 had an immediate and lasting effect nationally [42] and would not confound any possible Jolie effect in 2013. Instead, our findings show a slowing of screening mammography growth in our institution. Our overall experience of more than 48,000 screening mammograms over the 4 years is adequately powered for the tests we have carried out and did in fact precisely estimate a slowing of growth after the time point of interest. Our findings also fail to align with recent results from the United Kingdom, which found large increases in community referral for genetic testing [15] and similar increases in genetic testing found in the United States for the 1 to 2 months after that news [31]. Our results partly align with US results showing a stable mastectomy rate in the 2 months after the news of Ms Jolie’s mastectomy [31] but do not align with the significant and substantial increase in surgeries in the United Kingdom [30]. It remains a separate important point that physicians must ensure patients understand all their screening and treatment options and make informed choices [43]. If confirmed, results like ours offer a sobering reminder of the difficulties inherent in driving awareness of lifesaving breast cancer screening programs. They also cast some doubt on the utility of using celebrity endorsements to affect positive change. Improving financial and geographical access to screening programs and thus lowering the estimated proportion of one in three women failing to be appropriately screened remains the challenge of our time in breast imaging [44].
TAKE-HOME POINTS -
-
-
-
Celebrity medical news can increase patient demand for important health care services. In May 2013, Ms Angelina Jolie revealed her BRCA gene status and risk-reducing bilateral mastectomies. Subsequent studies have provided some evidence for increased gene testing and surgery after this news in the United Kingdom and in the United States, but no previous study has examined breast cancer screening utilization. In a study of over 48,000 consecutive screening mammograms, we found no changes after this news.
Journal of the American College of Radiology Health Services Research and Policy n Huesch et al
n
-
Spurring interest in potentially lifesaving breast cancer screening remains a difficult but important priority.
ADDITIONAL RESOURCES Additional resources can be found online at: http://dx. doi.org/10.1016/j.jacr.2017.03.016. REFERENCES 1. New York Times. New attitudes ushered in by Betty Ford. 1987. Available at: http://www.nytimes.com/1987/10/17/us/new-attitudesushered-in-by-betty-ford.html. Accessed May 2, 2017. 2. Lerner BH. Not so simple: the breast cancer stories of Betty Ford and Happy Rockefeller. Available at: http://www.huffingtonpost.com/barron-h-lerner/ not-so-simple-the-breast-_b_5733402.html. Accessed May 2, 2017. 3. Garfinkel L, Boring CC, Health CW. Changing trends: an overview of breast cancer incidence and mortality. Cancer 1994;74:222-7. 4. Kopans DB. Point: The New England Journal of Medicine article suggesting overdiagnosis from mammography screening is scientifically incorrect and should be withdrawn. J Am Coll Radiol 2013;10:317-9. 5. Bleyer A. Were our estimates of overdiagnosis with mammography screening in the United States “based on faulty science”? Oncologist 2014;19:113-26. http://dx.doi.org/10.1634/theoncologist.2013-0383. 6. Betsill WL Jr, Byrd B Jr, Hartmann WH. Breast cancer report. Cancer 1975;36:305-7. 7. Brown ML, Potosky AL. The presidential effect: the public health response to media coverage about Ronald Reagan’s colon cancer episode. Public Opin Q 1990;54:317-29. 8. Cram P, Fendrick AM, Inadomi J, et al. The impact of a celebrity promotional campaign on the use of colon cancer screening the Katie Couric effect. Arch Intern Med 2003;163:1601-5. http://dx.doi.org/ 10.1001/archinte.163.13.1601. 9. Hollywood Life. New report says Jenny McCarthy’s son may never have had autism after all. 2010. Available at: http://hollywoodlife.com/2010/02/26/ jenny-mccarthy-says-her-son-evan-never-had-autism/. Accessed May 2, 2013. 10. Price J, Simon K. Patient education and the impact of new medical research. J Health Econ 2009;28:1166-74. 11. Dawson S-J, Price MA, Jenkins MA, et al. Cancer risk management practices of noncarriers within BRCA1/2 mutation–positive families in the Kathleen Cunningham Foundation Consortium for Research Into Familial Breast Cancer. J Clin Oncol 2008;26:225-32. 12. Metcalfe KA, Birenbaum-Carmeli D, Lubinski J, et al, and the Hereditary Breast Cancer Clinical Study Group. International variation in rates of uptake of preventive options in BRCA1 and BRCA2 mutation carriers. Int J Cancer 2008;122:2017-22. 13. Morgan D, Sylvester H, Lucas FL, et al. Cancer prevention and screening practices among women at risk for hereditary breast and ovarian cancer after genetic counseling in the community setting. Fam Cancer 2009;8:277-87. 14. Jolie A. My medical choice. New York Times. 2013. Available at: http:// www.nytimes.com/2013/05/14/opinion/my-medical-choice.html?_r¼0. Accessed May 2, 2017. 15. Evans DG, Barwell J, Eccles DM, et al. The Angelina Jolie effect: how high celebrity profile can have a major impact on provision of cancer related services. Breast Cancer Res 2014;16:442. http://dx.doi.org/10. 1186/s13058-014-0442-6. 16. NCCN guidelines. Available at: https://www.nccn.org/professionals/ physician_gls/pdf/genetics_screening.pdf. Accessed December 19, 2016. 17. National Research Council. Crossing the quality chasm: a new health system for the 21st century. Washington, DC: National Academies Press; 2001. 18. Shubin A. BRCA test motivation. Weblog. 2013. Available at: http:// therichlifeinwinecountry.com/2013/05/brca-test-motivation.html. Accessed May 2, 2017.
Angelina Jolie Effect
1025
19. Huesch MD. 2015. Celebrity influence? An observational study of consumer responses on social media and on an internet search engine to news of Angelina Jolie’s prophylactic mastectomies. CESR-Schaeffer Working Paper Series, Paper No. 2015-012. Available at: http://static. usc.edu/documents/WP_2015_012.pdf. Accessed May 2, 2017. 20. Noar SM, Althouse BM, Ayers JW, Francis DB, Ribisl KM, et al. Cancer information seeking in the digital age: effects of Angelina Jolie’s prophylactic mastectomy announcement. Med Decis Making 2015;35:16-21. 21. Juthe RH, Zaharchuk A, Wang C. Celebrity disclosures and information seeking: the case of Angelina Jolie. Genet Med 2015;17:545-53. 22. Otto SJ, Fracheboud J, Verbeek AL, et al. Mammography screening and risk of breast cancer death: a population-based case–control study. Cancer Epidemiol Biomarkers Prev 2012;21:66-73. 23. Coldman A, Phillips N, Warren L, Kan L. Breast cancer mortality after screening mammography in British Columbia women. Int J Cancer 2007;120:1076-80. 24. Tabár L, Vitak B, Chen TH-H, et al. Swedish two-county trial: impact of mammographic screening on breast cancer mortality during 3 decades. Radiology 2011;260:658-63. 25. Screening IUPoBC. The benefits and harms of breast cancer screening: an independent review. Lancet 2012;380:1778-86. 26. Webb ML, Cady B, Michaelson JS, et al. A failure analysis of invasive breast cancer: most deaths from disease occur in women not regularly screened. Cancer 2014;120:2839-46. 27. Chetlen A, Mack J, Chan J. Breast cancer screening controversies: who, when, why, and how? Clin Imaging 2016;40:279-82. 28. Tabar L, Gad A, Holmberg L, et al. Reduction in mortality from breast cancer after mass screening with mammography: randomised trial from the Breast Cancer Screening Working Group of the Swedish National Board of Health and Welfare. Lancet 1985;325:829-32. 29. Shapiro S, Venet W, Strax P, et al. Ten- to fourteen-year effect of screening on breast cancer mortality. J Natl Cancer Inst 1982;69:349-55. 30. Evans DG, Wisely J, Clancy T, et al. Longer term effects of the Angelina Jolie effect: increased risk-reducing mastectomy rates in BRCA carriers and other high-risk women. Breast Cancer Res 2015;17:143. http://dx.doi. org/10.1186/s13058-015-0650-8. 31. Desai S, Jena AB. Do celebrity endorsements matter? Observational study of BRCA gene testing and mastectomy rates after Angelina Jolie’s New York Times editorial. BMJ 355:i6357.
1026
32. Huesch MD, Ostbye T, Ong MK. Evaluating the impact of health policy interventions using aggregate population health data. Health Econ 2012;21:1234-49. 33. McDowall D, McCleary R, Meidinger EE, Hay RA. Interrupted time series analysis. In: Sage University papers series on quantitative applications in the social sciences. Thousand Oakes, CA: Sage; 1980. 34. Wooldridge JM. Econometric analysis of cross section and panel data. Cambridge, MA: MIT Press; 2002. 35. Card D, Krueger A. Minimum wages and employment: a case study of the fast food industry in New Jersey and Pennsylvania. The American Economic Review 1994;84:772-84. 36. Jones A. Applied econometrics for health economists, a practical guide. 2nd ed. Oxford, UK: Radcliffe; 2007. 37. Mannis GN, Fehniger JE, Creasman JS, et al. Risk-reducing salpingooophorectomy and ovarian cancer screening in 1077 women after BRCA testing. JAMA Intern Med 2013;173:96-103. 38. Katz SJ, Morrow M. Contralateral prophylactic mastectomy for breast cancer: addressing peace of mind. JAMA 2013;310:793-4. 39. Susan G. Komen Foundation. Why are rates of bilateral mastectomies rising? 2013. Available at: http://ww5.komen.org/Content.aspx? id¼6442452097. Accessed May 2, 2017. 40. Biolchini A. U-M study: fear may be driving women to over-treat breast cancer. 2012. Available at: http://www.annarbor.com/news/fear-worrymay-be-driving-women-to-over-treat-breast-cancer-university-of-michiganstudy-shows/. Accessed May 2, 2017. 41. Hawley ST, Jagsi R, Katz SJ. Is contralateral prophylactic mastectomy (CPM) overused? Results from a population-based study. J Clin Oncol 2012;30: abstr 26. Available at: http://meetinglibrary.asco.org/ content/104375-126. Accessed May 2, 2017. 42. Gray N, Picone G. The effect of the 2009 USPSTF breast cancer screening recommendations on mammography rates. Health Serv Res 2016;51:1533-45. 43. American Medical Association. Opinion 10.02: Patient Responsibilities. 1994. Available at: https://www.ama-assn.org/sites/default/files/mediabrowser/code-of-medical-ethics-chapter-2.pdf. Accessed May 2, 2017. 44. Centers for Disease Control and Prevention. Use of mammography among women aged 40 and over, by selected characteristics: United States, selected years 1987-2013. Table 70, page 246. Available at: http://www.cdc.gov/nchs/data/hus/hus15.pdf#070. Accessed May 2, 2017.
Journal of the American College of Radiology Volume 14 n Number 8 n August 2017
HEALTH SERVICES RESEARCH AND POLICY
Evaluation of the “Angelina Jolie Effect” on Screening Mammography Utilization in an Academic Center Marco D. Huesch, MBBS, PhD a , Susann Schetter, DO a, Joel Segel, PhD b, Alison Chetlen, DO a Abstract Purpose: The aim of this study was to understand the impact on screening mammography at our institution, comparing weekly utilization in the 2 years before and the 2 years after Ms Angelina Jolie disclosed in the New York Times on May 13, 2013, that she had had a prophylactic double mastectomy. Methods: All 48,110 consecutive screening mammograms conducted at our institution between May 16, 2011, and May 16, 2015, were selected from our electronic medical record system. We used interrupted time series statistical models and graphical methods on utilization data to understand utilization changes before and after Ms Jolie’s news. Results: The graphed trend of weekly screening mammogram utilization failed to show changes around the time of interest. Analytical models and statistical tests also failed to show a step change increase or acceleration of utilization around May 2013. However, graphical and time series analyses showed a flattening of utilization in the middle of 2014. Conclusions: In our well-powered analysis in a large regional breast imaging center, we found no support for the hypothesis that this celebrity news drove increased screening. Key Words: Screening, breast cancer, celebrity news, mammography, influence, time series analysis J Am Coll Radiol 2017;14:1020-1026. Copyright 2017 American College of Radiology
Patients can potentially be influenced by celebrity medical information in the media. Understanding whether such celebrity influence exists is important because this influence can lead to positive and negative public health effects. Consider an early positive example involving celebrity breast cancer. In 1973 and 1974, respectively, the wife of the president and the wife of the vice president publicly disclosed news of breast cancer [1]. Ford’s lesson for other women was said to be “straightforward: get a mammogram, which she had not done” [2]. The
a
Department of Radiology, Milton S. Hershey Medical Center, Hershey, Pennsylvania. b Department of Health Policy and Administration, Pennsylvania State University, University Park, Pennsylvania. Corresponding author and reprints: Marco D. Huesch, MBBS, PhD, Department of Radiology, Milton S. Hershey Medical Center, 500 University Drive, Mailcode H-066, Hershey, PA 17033-0850; e-mail: [email protected]. The authors have no conflicts of interest related to the material discussed in this article.
influence of these disclosures was strongly associated with contemporaneous increases in breast examination consultations [3] and to increases in the still not widely available mammograms [4], as well as more diagnoses [5] and more treatment of breast cancer [6]. Similarly positive public health impacts of celebrity news can be found in other settings. After news in the traditional media of President Reagan’s colon cancer, early detection with fecal occult blood tests increased [7]. In 2000 after the death of her husband from colon cancer, Katie Couric underwent a colonoscopy live on the Today Show, which led to a short-term spike in colonoscopy screening [8]. However, less benevolent effects are also possible. Misleading media information could drive suboptimal consumer decisions. One example is celebrity influence on vaccination rates based on autism fears [9]. Widespread media reports of increased risk with vaginal births after Cesarean sections were associated with a rapid drop in the rate of this procedure [10]. Another example lies in the context of hereditary breast cancers. ª 2017 American College of Radiology
1020
1546-1440/17/$36.00
n
http://dx.doi.org/10.1016/j.jacr.2017.03.016
Although increasing prevention and screening are desirable in appropriate risk strata, overuse of BRCA gene screening in patients with a low pretest probability would be undesirable [11]. More than one-third of US women with BRCA1 or BRCA2 mutations already choose prophylactic mastectomy, more than in many other countries [12]. Overuse of nonmedically indicated prophylactic surgery in women with lower risk would be another concern [13]. To understand the potential impact (positive, negative, or none) of celebrity news, the case of Angelina Jolie may be important. She revealed her prophylactic double mastectomy for BRCA1 gene marker status in the New York Times on May 13 (online) and May 14 (print) [14]. This highly publicized news had a global impact [15]. Risk monitoring with screening MRI and risk reduction with prophylactic mastectomy may be appropriate for women with the BRCA1 mutation and others of this risk stratum [16], but would be inappropriate for women in general choosing to follow Ms Jolie’s lead without knowledge of their level of lifetime risk. Indeed, Ms Jolie hoped that women “will be able to get gene tested, and that if they have a high risk they, too, will know that they have strong options” [14]. Although patient-centered care requires consideration of screening regimens and therapies that are respectful of patient preferences [17], fear of cancer may increase the trend toward aggressive risk-reducing procedures. For example, a blogger wrote on the same day as Ms Jolie’s article: “Angelina Jolie went from having an 87% chance of getting breast cancer to a 5% chance. . . . If I am accepted for BRCA testing, and my results show that I, too, am carrying the BRCA gene, I am certain my decision will be crystal clear as well” [18]. Some existing research investigates the “Angelina effect.” In our previous work, we showed that at the time of the New York Times article, social media had displayed dramatic but transient increases in mentions of breast cancer in the United States [19]. Others have also shown increases in Internet searches related to breast screening both in the United States and internationally [20,21]. Such effects would have clearly positive public health impacts, because increased screening for breast cancer would save lives [22-29]. Yet to our knowledge there has been no research into utilization of screening and little research into utilization of genetic testing or bilateral risk-reducing mastectomies. Two studies in the United Kingdom show dramatic and almost immediate increases in referral for genetic testing from clinics and genetics centers [15] and Journal of the American College of Radiology Health Services Research and Policy n Huesch et al
n
subsequently in the same population for bilateral riskreducing mastectomies [30]. Although genetic testing referrals were judged to be appropriate based on internal review [15], subsequent surgeries in the next 6 to 24 months increased among women regardless of mutation carrier status and was greater among those without mutations [30]. Finally, a recent national study among commercially insured US women also found significant increases in BRCA genetic testing after the Jolie editorial [31]. However, this study found no overall increase in mastectomies and some evidence that incremental genetic testing may have been among women with a lower pretest probability of a BRCA mutation [31]. We hypothesized that there would be an Angelina effect in screening mammography. Our rationale was 3-fold, based on patient salience, physician salience, and workup processes. We anticipated that underscreened or unscreened women may be influenced to start appropriate screening by the celebrity news leading to greater attention to, and awareness of, the prevalence of breast cancer and the risks of not screening. Similarly, appropriately screened women on an annual cycle or underscreened women on a 2-year screening cycle might bring future screening schedules forward. We also anticipated that referring physicians might be influenced to discuss or recommend mammography with their patients after reading this widely publicized news. Finally, as part of the workup among high-risk or potentially high-risk women, a first step would be a screening mammogram if this had not already been done. In this study, we investigate this hypothesis of increased utilization of screening mammography and report on our institutional experience with screening mammography in the 2 years before and after Ms Jolie’s public disclosure.
METHODS We followed complex but well-established analytical methods in wide use to understand changes over time in a noisy series of data [32-34]. We first used graphical methods with smoothed trend lines to allow immediate visual inspection of possible change in utilization. Weekly and even daily volume is usually very variable over time as schedules and capacity changes and may be thought of as random noise. Over longer horizons, utilization may follow a so-called secular trend at some slowly increasing pace, reflecting slow changes in insurance, technology, preferences, and demographics.
Angelina Jolie Effect
1021
We are interested in a potential structural break, independent of that short-run noise, and adjusting for that long-term trend. Such a structural break is an increase, or decrease, or acceleration up or down that is associated with the time point of the hypothesized news item. Accordingly, we used two well-established complementary methods to understand any changes in screening mammography utilization after the New York Times article appeared: interrupted time series analysis and a closely related time series analysis method. The first explicitly models a potential step change and slope change in trend around a known date. The second considers the whole time series and seeks to identify a structural change in the data, such as an increase, or decrease, or acceleration in the trend, up or down.
Interrupted Time Series Analysis Three key points in this method are that we needed to (1) control for any secular trend in utilization over time, (2) measure any immediate and enduring step change in the average utilization after the time point of interest, and (3) measure any change in the slope of the trend in utilization in the new time period, compared to the prior one before the time point of interest. We built a regression model in which the dependent variable was weekly screening mammography and the independent variables were (1) the week of the service (to control for secular trends), (2) an indicator variable that took in each week the value 0 before and 1 after Ms Jolie’s news (to estimate immediate and enduring step changes), and (3) a variable that took the value 0 for every week before the change and took successive positive integers in the weeks after the changes (to estimate changes in the slope of the trend after the time point of interest). We also allowed an indicator for seasonality at the quarterly level, and in one sensitivity test also at the monthly level. We specified robust standard errors to account for likely heteroskedasticity in the data over time. We computed t tests for the coefficients of the two key variables in (2) and (3) above, to understand whether they are independently significant. Finally, we computed a Chow test for a structural break around the May 2013 time point in the data [32]. This tested whether the variables in (2) and (3) above are jointly zero. A significant value indicates that it would be unlikely for both variables to be zero and that there is a significant association between screening volume and the time point of the celebrity news item. 1022
Time Series Analysis In this complementary analysis, the model was not set up to, nor did it force the data to conform to, a particular structural break occurring at a particular time point. Rather, the model allowed the data to suggest a possible time point, examining all possible such time points. We tested the suggested break point using a Wald test. A significant test statistic suggests that the null hypothesis of no structural break can be rejected. We obtained deidentified data through our institution’s Millennium Cerner (Cerner Corporation, Kansas City, MO, United States) electronic medical record system’s billing system for all mammography in the 4-year period bisected by the key date of May 13, 2013. For our main analysis on screening mammography, we selected professional billing Current Procedural Terminology codes G020226 for “digital screening mammography.” For sensitivity analyses in which we wished to understand relative changes in screening volumes compared with a diagnostic procedure unlikely to be affected by celebrity news, we selected G020626 for “unilateral diagnostic mammography” and G020426 for “bilateral diagnostic mammography.” Subtotals per day and per week for each service were derived and yearly totals confirmed with separate breast center management information center reports on production volume. This study was HIPAA compliant and approved by the Pennsylvania State University College of Medicine Institutional Review Board with determination number 00005728. RESULTS In our institution, a total of 48,110 digital screening mammograms were performed between May 16, 2011, and May 16, 2015, over a total of 208 weeks. We aggregated highly variable daily volumes (see Appendix Fig. 1a) into weeks because no services were performed on Sunday and very few on Saturday. As Figure 1 shows, there was substantial noise even on a weekly basis. A lowess smoothed trend line suggested a gently sloping increasing trend in the 2 years before the New York Times article, and a slightly flatter slope afterward, tending to plateau around July 2014, and then declining slightly. In Appendix Figure 1b, we show similar results for the daily screening volume in the 2 weeks before and after the news release. These graphical findings are clearly not consistent with the hypothesized structural break, which we proceeded to formally test with a series of statistical models. Journal of the American College of Radiology Volume 14 n Number 8 n August 2017
Weekly fullfield digital screening mammogram patient totals 350
Smoothed trendline (Lowess smoother, bandwidth 0.8)
250 200 100
150
Weekly volume
300
May 13, 2013; publication of Ms Jolie's NYT article
All 208 consecutive service weeks from 5-16-2011 through 5-16-2015; Hershey Medical Center
Fig 1. Weekly full-field digital screening mammograms.
Interrupted Time Series Analysis We constructed a series of progressively saturated models to regress on weekly screening mammogram volume. Adding weekly time from the Jolie article, in the second column of Table 1, produced a reasonably wellpredicting model in which increased time was associated with increased volume in a highly significant manner. This linear trend was precisely estimated across all the models and represents the average trend toward higher utilization across the 4 years. However, adding the step change indicator did not meaningfully improve the
model’s fit, and the coefficient on this indicator was not statistically different from zero. This suggests no immediate and enduring step change in volumes accrued after the celebrity news publication. Even despite no immediate step change, it is possible that previous trends accelerated (leading to a steeper, more positive slope) or decelerated (leading to a flatter less positive slope) or even reversed (leading to a declining, negative slope). The addition of a slope indicator in the model captured these possible changes. Table 1 shows that the addition of the slope indicator was associated with a modest increase in explanatory
Table 1. Interrupted time series analysis Null Model Constant Weekly time Step change Indicator Slope change Indicator Quarter indicator Month indicator Observations R2 (%)
231.3 (<.001)
Add Weekly Time
Add Step Change Indicator
Add Slope Change Indicator
Add Quarter Seasonality
Add Month Seasonality
212.4 (<.001) 0.18 (<.001)
214.1 (<.001) 0.13 (.08) 6.5 (.44)
206.0 (<.001) 0.29 (.002) 6.7 (.43)
169.2 (<.001) 0.38 (<.001) 2.6 (.76)
165.6 (<.001) 0.37 (<.001) 1.8 (.83)
0.31 (.04)
0.31 (.02)
0.31 (.02)
12.9 (<.001)
24.5 (.002) 3.9 (.11)
28.3
29.1
208 weeks 0.0
9.8
9.9
11.8
Results are estimates (P values for the t test statistic).
Journal of the American College of Radiology Health Services Research and Policy n Huesch et al
n
Angelina Jolie Effect
1023
power (R2 improved from 9.9% to 11.8%) and the coefficient on this indicator was precisely estimated and significant at a conventional level of P < .05. The sign on the coefficient was negative, implying a decrease in slope in the post period, consistent with the graphical presentation. Finally, adding indicators for seasonality showed significantly higher volumes associated with the fourth quarter compared with earlier quarters in any year. These were driven by Breast Cancer Awareness month in October, as the four transient spikes in Figure 1 show. Beyond this quarterly seasonality, there was no significant effect of addition of a monthly seasonality indicator. The high R2 in the last two models suggests well-fitting models. We tested the joint hypothesis that the step-change indicator and the slope change indicator are jointly zero in a Chow test with a P value of .067. The insignificant result suggested that the null hypothesis of no structural break in May 2013 could not be rejected.
Time Series Analysis In a separate analysis, we did not impose the constraints on the data that the interrupted time series analysis imposed (by insisting on one single break in the data at one particular time point). By allowing the time series to be tested for structural breaks at any of the 208 weekly points over the 4 years, we allowed other identified breaks to arise flexibly. We used the model in the last column of Table 1, leaving out the indicators for step and slope changes. After regression, we had a well-fitting model with an R2 of 27.3%. Based on all the 208 possible breaks, we then tested for the single most significant “unknown structural break,” if any. This analysis suggested a break in the first week of July 2014, with P values less than .02 across each of the supremum Wald, average Wald, and average likelihood ratio tests. It is important to point out that April or May 2013 were not identified as significant structural breaks. Examining the graph in Figure 1 shows that July 2014 is the center of the plateau before the volumes decrease through early 2015. Robustness Analyses Unlike screening mammography, diagnostic mammography utilization is less likely at the discretion of a patient compared with the decision of a physician, and so it is arguably less likely to be driven by celebrity news. However, there is likely to be similar short-run noise, and potentially similar long-run secular changes over time 1024
affect its utilization. This affords an opportunity to control directly for such noise in a model similar to a difference-in-difference design [35]. Graphically we show these different trends in Appendix Figure 2. In a separate analysis, we subtracted the diagnostic mammography volume each week from the screening mammography volume and repeated the interrupted time series analysis. In unreported results, there was little difference except a slightly better-fitting model (R2 rises to 36.7%) and a more precisely estimated negative sign on the slope indicator. Finally, we used a linear regression model for what is actually a count model [36]. The dependent variable can only take positive integer values depending on the week’s utilization. When we performed a similar analysis using an overdispersed negative binomial regression that more correctly modeled counts, we found qualitatively unchanged results (not reported).
DISCUSSION In our institution, we unexpectedly observed no Angelina Jolie effect in the utilization of screening mammography. We found no statistically significant increase in utilization after compared with before a prominent celebrity publicly disclosed BRCA gene status and bilateral risk-reducing mastectomies. Our findings in an adequately powered study fail to support our hypothesis that such a change was likely based on arguments of salience and workup processes. It is unlikely that women became more aware of breast cancer but preferred not to act on this awareness. For example, the risk might have been judged as not substantial enough. However, in related settings, even non-BRCA carriers have undertaken risk-reducing procedures for ovarian cancers, presumably to allay concerns of cancer risk [37]. In other related settings, addressing women’s legitimate fears and need for peace of mind has likely led to unnecessary contralateral prophylactic mastectomies [38], despite National Comprehensive Cancer Network guidelines that discourage such procedures except for BRCA mutation carriers [39]. Recent studies show that among women receiving a contralateral prophylactic mastectomy together with mastectomy of the affected breast, 70% did not have a clinical reason for having one done [40,41]. It is also unlikely that women became more aware of breast cancer risks but were unable to act on this because of capacity constraints in our institution. Our institution had sufficient capacity to flexibly accommodate shifts in demand as large as 50% in total volume from one month to the next. Journal of the American College of Radiology Volume 14 n Number 8 n August 2017
Finally, it is unlikely that other influences confounded the effects we focused on. For example, any confounding national policy effects such as the controversial United States Preventive Services Task Force recommendations in 2009 had an immediate and lasting effect nationally [42] and would not confound any possible Jolie effect in 2013. Instead, our findings show a slowing of screening mammography growth in our institution. Our overall experience of more than 48,000 screening mammograms over the 4 years is adequately powered for the tests we have carried out and did in fact precisely estimate a slowing of growth after the time point of interest. Our findings also fail to align with recent results from the United Kingdom, which found large increases in community referral for genetic testing [15] and similar increases in genetic testing found in the United States for the 1 to 2 months after that news [31]. Our results partly align with US results showing a stable mastectomy rate in the 2 months after the news of Ms Jolie’s mastectomy [31] but do not align with the significant and substantial increase in surgeries in the United Kingdom [30]. It remains a separate important point that physicians must ensure patients understand all their screening and treatment options and make informed choices [43]. If confirmed, results like ours offer a sobering reminder of the difficulties inherent in driving awareness of lifesaving breast cancer screening programs. They also cast some doubt on the utility of using celebrity endorsements to affect positive change. Improving financial and geographical access to screening programs and thus lowering the estimated proportion of one in three women failing to be appropriately screened remains the challenge of our time in breast imaging [44].
TAKE-HOME POINTS -
-
-
-
Celebrity medical news can increase patient demand for important health care services. In May 2013, Ms Angelina Jolie revealed her BRCA gene status and risk-reducing bilateral mastectomies. Subsequent studies have provided some evidence for increased gene testing and surgery after this news in the United Kingdom and in the United States, but no previous study has examined breast cancer screening utilization. In a study of over 48,000 consecutive screening mammograms, we found no changes after this news.
Journal of the American College of Radiology Health Services Research and Policy n Huesch et al
n
-
Spurring interest in potentially lifesaving breast cancer screening remains a difficult but important priority.
ADDITIONAL RESOURCES Additional resources can be found online at: http://dx. doi.org/10.1016/j.jacr.2017.03.016. REFERENCES 1. New York Times. New attitudes ushered in by Betty Ford. 1987. Available at: http://www.nytimes.com/1987/10/17/us/new-attitudesushered-in-by-betty-ford.html. Accessed May 2, 2017. 2. Lerner BH. Not so simple: the breast cancer stories of Betty Ford and Happy Rockefeller. Available at: http://www.huffingtonpost.com/barron-h-lerner/ not-so-simple-the-breast-_b_5733402.html. Accessed May 2, 2017. 3. Garfinkel L, Boring CC, Health CW. Changing trends: an overview of breast cancer incidence and mortality. Cancer 1994;74:222-7. 4. Kopans DB. Point: The New England Journal of Medicine article suggesting overdiagnosis from mammography screening is scientifically incorrect and should be withdrawn. J Am Coll Radiol 2013;10:317-9. 5. Bleyer A. Were our estimates of overdiagnosis with mammography screening in the United States “based on faulty science”? Oncologist 2014;19:113-26. http://dx.doi.org/10.1634/theoncologist.2013-0383. 6. Betsill WL Jr, Byrd B Jr, Hartmann WH. Breast cancer report. Cancer 1975;36:305-7. 7. Brown ML, Potosky AL. The presidential effect: the public health response to media coverage about Ronald Reagan’s colon cancer episode. Public Opin Q 1990;54:317-29. 8. Cram P, Fendrick AM, Inadomi J, et al. The impact of a celebrity promotional campaign on the use of colon cancer screening the Katie Couric effect. Arch Intern Med 2003;163:1601-5. http://dx.doi.org/ 10.1001/archinte.163.13.1601. 9. Hollywood Life. New report says Jenny McCarthy’s son may never have had autism after all. 2010. Available at: http://hollywoodlife.com/2010/02/26/ jenny-mccarthy-says-her-son-evan-never-had-autism/. Accessed May 2, 2013. 10. Price J, Simon K. Patient education and the impact of new medical research. J Health Econ 2009;28:1166-74. 11. Dawson S-J, Price MA, Jenkins MA, et al. Cancer risk management practices of noncarriers within BRCA1/2 mutation–positive families in the Kathleen Cunningham Foundation Consortium for Research Into Familial Breast Cancer. J Clin Oncol 2008;26:225-32. 12. Metcalfe KA, Birenbaum-Carmeli D, Lubinski J, et al, and the Hereditary Breast Cancer Clinical Study Group. International variation in rates of uptake of preventive options in BRCA1 and BRCA2 mutation carriers. Int J Cancer 2008;122:2017-22. 13. Morgan D, Sylvester H, Lucas FL, et al. Cancer prevention and screening practices among women at risk for hereditary breast and ovarian cancer after genetic counseling in the community setting. Fam Cancer 2009;8:277-87. 14. Jolie A. My medical choice. New York Times. 2013. Available at: http:// www.nytimes.com/2013/05/14/opinion/my-medical-choice.html?_r¼0. Accessed May 2, 2017. 15. Evans DG, Barwell J, Eccles DM, et al. The Angelina Jolie effect: how high celebrity profile can have a major impact on provision of cancer related services. Breast Cancer Res 2014;16:442. http://dx.doi.org/10. 1186/s13058-014-0442-6. 16. NCCN guidelines. Available at: https://www.nccn.org/professionals/ physician_gls/pdf/genetics_screening.pdf. Accessed December 19, 2016. 17. National Research Council. Crossing the quality chasm: a new health system for the 21st century. Washington, DC: National Academies Press; 2001. 18. Shubin A. BRCA test motivation. Weblog. 2013. Available at: http:// therichlifeinwinecountry.com/2013/05/brca-test-motivation.html. Accessed May 2, 2017.
Angelina Jolie Effect
1025
19. Huesch MD. 2015. Celebrity influence? An observational study of consumer responses on social media and on an internet search engine to news of Angelina Jolie’s prophylactic mastectomies. CESR-Schaeffer Working Paper Series, Paper No. 2015-012. Available at: http://static. usc.edu/documents/WP_2015_012.pdf. Accessed May 2, 2017. 20. Noar SM, Althouse BM, Ayers JW, Francis DB, Ribisl KM, et al. Cancer information seeking in the digital age: effects of Angelina Jolie’s prophylactic mastectomy announcement. Med Decis Making 2015;35:16-21. 21. Juthe RH, Zaharchuk A, Wang C. Celebrity disclosures and information seeking: the case of Angelina Jolie. Genet Med 2015;17:545-53. 22. Otto SJ, Fracheboud J, Verbeek AL, et al. Mammography screening and risk of breast cancer death: a population-based case–control study. Cancer Epidemiol Biomarkers Prev 2012;21:66-73. 23. Coldman A, Phillips N, Warren L, Kan L. Breast cancer mortality after screening mammography in British Columbia women. Int J Cancer 2007;120:1076-80. 24. Tabár L, Vitak B, Chen TH-H, et al. Swedish two-county trial: impact of mammographic screening on breast cancer mortality during 3 decades. Radiology 2011;260:658-63. 25. Screening IUPoBC. The benefits and harms of breast cancer screening: an independent review. Lancet 2012;380:1778-86. 26. Webb ML, Cady B, Michaelson JS, et al. A failure analysis of invasive breast cancer: most deaths from disease occur in women not regularly screened. Cancer 2014;120:2839-46. 27. Chetlen A, Mack J, Chan J. Breast cancer screening controversies: who, when, why, and how? Clin Imaging 2016;40:279-82. 28. Tabar L, Gad A, Holmberg L, et al. Reduction in mortality from breast cancer after mass screening with mammography: randomised trial from the Breast Cancer Screening Working Group of the Swedish National Board of Health and Welfare. Lancet 1985;325:829-32. 29. Shapiro S, Venet W, Strax P, et al. Ten- to fourteen-year effect of screening on breast cancer mortality. J Natl Cancer Inst 1982;69:349-55. 30. Evans DG, Wisely J, Clancy T, et al. Longer term effects of the Angelina Jolie effect: increased risk-reducing mastectomy rates in BRCA carriers and other high-risk women. Breast Cancer Res 2015;17:143. http://dx.doi. org/10.1186/s13058-015-0650-8. 31. Desai S, Jena AB. Do celebrity endorsements matter? Observational study of BRCA gene testing and mastectomy rates after Angelina Jolie’s New York Times editorial. BMJ 355:i6357.
1026
32. Huesch MD, Ostbye T, Ong MK. Evaluating the impact of health policy interventions using aggregate population health data. Health Econ 2012;21:1234-49. 33. McDowall D, McCleary R, Meidinger EE, Hay RA. Interrupted time series analysis. In: Sage University papers series on quantitative applications in the social sciences. Thousand Oakes, CA: Sage; 1980. 34. Wooldridge JM. Econometric analysis of cross section and panel data. Cambridge, MA: MIT Press; 2002. 35. Card D, Krueger A. Minimum wages and employment: a case study of the fast food industry in New Jersey and Pennsylvania. The American Economic Review 1994;84:772-84. 36. Jones A. Applied econometrics for health economists, a practical guide. 2nd ed. Oxford, UK: Radcliffe; 2007. 37. Mannis GN, Fehniger JE, Creasman JS, et al. Risk-reducing salpingooophorectomy and ovarian cancer screening in 1077 women after BRCA testing. JAMA Intern Med 2013;173:96-103. 38. Katz SJ, Morrow M. Contralateral prophylactic mastectomy for breast cancer: addressing peace of mind. JAMA 2013;310:793-4. 39. Susan G. Komen Foundation. Why are rates of bilateral mastectomies rising? 2013. Available at: http://ww5.komen.org/Content.aspx? id¼6442452097. Accessed May 2, 2017. 40. Biolchini A. U-M study: fear may be driving women to over-treat breast cancer. 2012. Available at: http://www.annarbor.com/news/fear-worrymay-be-driving-women-to-over-treat-breast-cancer-university-of-michiganstudy-shows/. Accessed May 2, 2017. 41. Hawley ST, Jagsi R, Katz SJ. Is contralateral prophylactic mastectomy (CPM) overused? Results from a population-based study. J Clin Oncol 2012;30: abstr 26. Available at: http://meetinglibrary.asco.org/ content/104375-126. Accessed May 2, 2017. 42. Gray N, Picone G. The effect of the 2009 USPSTF breast cancer screening recommendations on mammography rates. Health Serv Res 2016;51:1533-45. 43. American Medical Association. Opinion 10.02: Patient Responsibilities. 1994. Available at: https://www.ama-assn.org/sites/default/files/mediabrowser/code-of-medical-ethics-chapter-2.pdf. Accessed May 2, 2017. 44. Centers for Disease Control and Prevention. Use of mammography among women aged 40 and over, by selected characteristics: United States, selected years 1987-2013. Table 70, page 246. Available at: http://www.cdc.gov/nchs/data/hus/hus15.pdf#070. Accessed May 2, 2017.
Journal of the American College of Radiology Volume 14 n Number 8 n August 2017