Effect of Reminders of Personal Sacrifice and Suggested Rationalizations on Residents’ Self-Reported Willingness to Accept Gifts a Randomized Trial

Home , Sacrifice (2007), Sacrifice (2010)

ORIGINAL CONTRIBUTION

Effect of Reminders of Personal Sacrifice and Suggested Rationalizations on Residents’ Self-Reported Willingness to Accept Gifts A Randomized Trial

Sunita Sah, MBChB, BSc, MBA, MS Context Despite expanding research on the prevalence and consequences of con- George Loewenstein, PhD flicts of interest in medicine, little attention has been given to the psychological pro- cesses that enable physicians to rationalize the acceptance of gifts. HY DO MANY PHYSICIANS Objective To determine whether reminding resident physicians of the sacrifices made feel comfortable accept- to obtain training, as well as suggesting this as a potential rationalization, increases ing gifts from pharma- self-stated willingness to accept gifts from industry. ceutical and medical de- Design, Setting, and Participants Three hundred one US resident physicians from viceW manufacturers that raise ethical 2 sample populations (pediatrics and family medicine) who were recruited during March- concerns about conflicts of interest July 2009 participated in a survey presented as evaluating quality of life and values. (COIs)? Studies have examined the extent of physician-industry relation- Intervention Physicians were randomly assigned to receive 1 of 3 different online 1,2 surveys. The sacrifice reminders survey (n=120) asked questions about sacrifices made ships, identified strategies used in phar- in medical training, followed by questions regarding the acceptability of receiving gifts 3,4 maceutical sales and marketing, from industry. The suggested rationalization survey (n=121) presented the same sac- explored the potential effect of gifts on rifice questions, followed by a suggested possible rationalization (based on sacrifices physician prescribing behavior,1,5 and ad- made in medical training) for acceptance of gifts, before the questions regarding the vocated policies to reduce the influence acceptability of gifts. The control survey (n=60) asked about the acceptability of gifts of COIs.6 However, little work has ex- before asking questions about sacrifices or suggesting a rationalization. amined how physicians rationalize ac- Main Outcome Measures Physician self-stated acceptability of receiving gifts from ceptance of questionable ties to indus- industry. try. One of the few studies to do so found Results Reminding physicians of sacrifices made in obtaining their education re- that although physicians interviewed in sulted in gifts being evaluated as more acceptable: 21.7% (13/60) in the control group focus groups appreciated the hazards vs 47.5% (57/120) in the sacrifice reminders group (odds ratio, 1.81; 95% confi- posed by COIs, they used a variety of dence interval, 1.27-2.58; P=.001). Although most residents disagreed with the sug- strategies to rationalize placing them- gested rationalization, exposure to it further increased the perceived acceptability of gifts to 60.3% (73/121) in that group (odds ratio relative to sacrifice reminders group, selves in conflicted situations, includ- Ͻ ing not thinking about the COI, deny- 1.45; 95% confidence interval, 1.22-1.72; P .001). ing an effect on their prescribing Conclusions Providing resident physicians with reminders of sacrifices increased the behavior, rejecting responsibility for the perceived acceptability of industry-sponsored gifts. Including a rationalization state- problem, and using diverse techniques ment further increased gift acceptability. intended to resist or undo bias.7 JAMA. 2010;304(11):1204-1211 www.jama.com One specific rationalization uncov- ered in a study of third-year medical stu- dorsed the view that they were entitled rewards.9 The pharmaceutical industry dents justifies acceptance of gifts on the to gifts from industry because of hard- may recognize the value of treating phy- basis of the hardships associated with ship, described as “considerable debt and sicians well, and physicians in return may medical training and practice.8 Eighty minimal income.”8 Adams’ equity theory percent of respondents in that study en- postulates that individuals who believe Author Affiliations: Carnegie Mellon University, Pitts- they are underpaid will respond by low- burgh, Pennsylvania. ering their input (ie, their work contri- Corresponding Author: Sunita Sah, MBChB, BSc, MBA, For editorial comment see p 1233. MS, Carnegie Mellon University, 5000 Forbes Ave, butions) or by attempting to raise their Pittsburgh, PA 15213 ([email protected]).

1204 JAMA, September 15, 2010—Vol 304, No. 11 (Reprinted) ©2010 American Medical Association. All rights reserved. RATIONALIZING ACCEPTANCE OF INDUSTRY GIFTS think that they are worthy of that treat- Figure 1. The 3 Main Conditions ment. These justifications could over- ride reservations about the inappropri- Sacrifice reminders condition ateness of accepting gifts. Whereas research has shown that Sacrifice Conflict of Suggested medical students endorse the idea that questionsa interest questionsb rationalizationc hardships can justify acceptance of gifts, it has not documented a causal connection (or even correlation) between Suggested rationalization condition the perception of hardship and attitudes toward the acceptance of gifts. To Sacrifice Suggested Conflict of a c b test for a causal connection, we con- questions rationalization interest questions ducted a randomized study to examine whether early-career physicians who are reminded of personal sacrifices and, Control condition hence, provided with implicit justifi- Conflict of Sacrifice Suggested cations for ethically questionable be- interest questionsb questionsa rationalizationc havior would evaluate that behavior as more acceptable. We also tested Each group was presented with the same 3 sets of questions but in different order, as shown. whether providing a suggested poten- a Questions on the number of hours worked, hours of sleep, salary, and education-related debt. tial rationalization (that inadequate b Questions on the acceptability of receiving gifts. c compensation and poor working con- “Some physicians believe that the stagnant salaries and rising debt levels prevalent in the medical profession justifies accepting gifts and other forms of compensation and incentives from the pharmaceutical industry. To ditions might justify accepting gifts) fur- what extent do you agree or disagree that this is a good justification?” ther increases the reported acceptability of gifts. 2009 to 30 Pennsylvania family medi- asked about the sacrifices they had METHODS cine residency directors (representing made to obtain their medical educa- Participants potential access to 600 family medi- tion (eFigure 1; available at http://www We recruited from 2 resident popula- cine residents) and the next 3 in July .jama.com). In the suggested rational- tions to achieve a minimum target of 300 2009 to approximately 420 family medi- ization group, physicians first answered residents, consistent with power calcu- cine residency directors in the rest of the same sacrifice questions, then were lations (presented herein). The first the United States. The e-mails re- asked whether they agreed or dis- sample was pediatrics residents from the quested that residency directors for- agreed (on a 5-point Likert scale) with Children’s Hospital of Pittsburgh, Pitts- ward the information to their resi- the following statement: “Some physi- burgh, Pennsylvania, collected during dents. For this sample, the incentive for cians believe that the stagnant salaries March 2009. All 100 residents received responding was a portable media player and rising debt levels prevalent in the 3 e-mail requests from the chief pediat- for every 1 in 100 respondents. Be- medical profession justifies accepting ric resident encouraging them to com- cause we had no direct contact with the gifts and other forms of compensation plete a 3-minute survey and offering each family medicine directors, we could not and incentives from the pharmaceuti- a $20 online shopping voucher if more establish how many residents ulti- cal industry.” Next, both groups were than 80% responded. mately received the solicitation. asked a series of questions, mainly The second sample consisted of fam- about the acceptability of receiving gifts ily medicine residents. Pediatricians are Study Design from industry (eFigure 2). In the con- less likely than family physicians to re- The introduction to the survey de- trol group, physicians were asked these ceive samples, reimbursements, and scribed its purpose as collecting infor- questions before the sacrifice ques- payments for professional services from mation on quality of life, expectations, tions. In the sacrifice reminders and industry10; therefore, inclusion of fam- and values and did not mention COIs. control groups, agreement with the ra- ily medicine residents increases the gen- The study protocol was approved by the tionalization was elicited at the end of eralizability of the results. Six e-mail re- institutional review board of Carnegie the questionnaire. Agreement with the quests containing a link to the surveys Mellon University, Pittsburgh, and in- rationalization was defined to include were sent to family medicine resi- cluded a waiver for written consent. “neither agree nor disagree,” “agree,” dency directors via the program direc- The 3 main conditions were sacri- and “strongly agree.” tor of the Forbes Family Medicine Resi- fice reminders, suggested rationaliza- Sacrifice Reminder Questions. Six dency, West Penn Allegheny Health tion, and control (FIGURE 1). In the sac- questions were intended to remind phy- System, Pittsburgh: the first 3 in April rifice reminders group, physicians were sicians of sacrifices they made to obtain

©2010 American Medical Association. All rights reserved. (Reprinted) JAMA, September 15, 2010—Vol 304, No. 11 1205 RATIONALIZING ACCEPTANCE OF INDUSTRY GIFTS their medical training, including hours The remaining survey questions con- ratios (ORs) of the main dependent vari- worked, hours of sleep, salary,and edu- sisted of characteristics such as post- able (acceptability of gifts) across con- cation-related debt (eFigure 1). Those graduate year and the hospital and state ditions. We used contrast-coding dum- in the sacrifice reminders and sug- the respondent practiced in (eFigure 3). mies for the main conditions, which gested rationalization groups were fur- The survey was pretested on a sample provide estimates of the effect of the sac- ther randomly assigned to “feel-rich” of 5 physicians. Protocol analysis was rifice reminders condition vs the con- and “feel-poor” subgroups designed to used to refine question wording and ex- trol condition and of the suggested ra- manipulate the degree of perceived sac- amine appropriateness of response cat- tionalization condition vs the sacrifice rifice. Because the mean annual salary egories. A pilot, without incentives for reminders condition. Although dummy for residents was estimated at about participation, was then conducted variables commonly compare both con- $42 000 and the mean hours of sleep among 73 internal medicine residents ditions to a baseline, we were inter- while on call at approximately 2.5 in 3 Pittsburgh hospitals (Shadyside, ested in whether the suggested rational- (based on the pilot described herein), Montefiore, and Veterans Affairs) to as- ization condition (which includes the response categories for salary and sess the appropriateness of the sacri- sacrifice reminder questions) gives an sleep were varied to give high- or low- fice manipulation and the reliability of additional effect over the sacrifice re- category mean responses. The differ- the scale. minders condition.11 Covariates for ing response options ensure that more Sample Size. On clicking the link to sample and postgraduate year were ini- physicians in the high categories an- participate, each physician was ran- tially included in every model. swer in lower response options than domly assigned by a computerized ran- The first model included indicator those in the low categories; eg, for the dom number generator (using a uni- variables for the sacrifice reminders con- feel-poor subgroups, the lowest cat- form distribution) to 1 of 5 conditions dition (relative to the control condi- egory for salary is $0 to $100 000 and (FIGURE 2). Using variance estimates tion) and for the suggested rationaliza- the highest category is $350 000 or from the pilot group, a target sample tion condition (relative to the sacrifice higher vs $0 to $20 000 and $50 000 size of 300 residents (n=60 in each of reminders condition). Analyses were or higher, respectively, for the feel- 5 conditions) provided 90% power to conducted on the complete sample as rich subgroups. To enable comparison detect a difference of 0.25 between well as separately for the 2 samples (pe- of responses between the groups, the population means, 1⁄20 of the range of diatric residents and family medicine resi- control group was also randomized to the 5-point scale. With 93 pediatric residents). The second model added fur- feel-rich and feel-poor subgroups, but dent responses from Children’s Hospi- ther variables to gain a more specific the related questions were asked after tal of Pittsburgh, we set a target of 210 picture of the factors that are associated the dependent variable (gift acceptabil- responses from family medicine resi- with gift acceptability; this model added ity) was collected, so this randomiza- dents and closed the family medicine a dummy variable for agreement with the tion did not constitute an experimen- survey after 230 responses. rationalization statement and interac- tal manipulation. tions between conditions (sacrifice re- The final question in this section, Statistical Analysis minders and suggested rationalization) “How do you feel about your working The main dependent measure was the ac- with this dummy variable. conditions?” served as a manipulation ceptability of receiving gifts, as derived To examine the effect of the rich-vs- check for the feel-rich and feel-poor from responses to the 10 COI ques- poor manipulation, the sample was re- subgroups. Having poor working con- tions. Analysis of the pilot data revealed stricted to only those in the sacrifice re- ditions was defined as those who re- high scale reliability (Cronbach ␣=0.85). minders and suggested rationalization sponded “okay,”“bad,” and “very bad.” Responses to the 10 questions were groups (since the rich-vs-poor ques- It was expected that those in the feel- summed, with the scale reversed so that tions came after measures of the depen- poor subgroups would have a more higher numbers correspond to greater ac- dent variable in the control group), and negative perception of their working ceptability. This resultant range of scores, the final model included explanatory conditions. from 10 to 50, was then divided by the dummy variables for the suggested ra- COI Questions. The scale eliciting maximum possible score (50), result- tionalization and feel-poor conditions. attitudes toward COIs consisted of 10 ing in a scale that could range from 0.2 In addition to ORs, we report com- items querying the acceptability of dif- to 1. We converted this scale, using the parative percentages and ␹2 statistics. ferent practices (eFigure 2), with re- mean, to a dichotomous 1 (acceptable) To explore differences between the sponses ranging from 1 (strongly agree) vs 0 (unacceptable) variable. Any score samples and conditions and to check to 5 (strongly disagree). To disguise the at or above the mean was given an ac- manipulations, we used analysis of vari- survey’s true focus, the COI questions ceptable rating and scores below the ance, ␹2 tests, and logistic regression. were interspersed with 4 questions ad- mean were converted to unacceptable. Before conducting analyses of vari- dressing quality of life and ethical is- The main analysis consisted of logis- ance, we tested for equality of vari- sues arising in medicine. tic regression models to estimate odds ances across groups using the Levine

1206 JAMA, September 15, 2010—Vol 304, No. 11 (Reprinted) ©2010 American Medical Association. All rights reserved. RATIONALIZING ACCEPTANCE OF INDUSTRY GIFTS test for homogeneity and checked for RESULTS 95% were in the first 3 years of resi- nonnormality. PϽ.05 was considered The number of participants in each group dency. Similar to the pilot, reliability statistically significant. The absence of and sample is shown in Figure 2. Ex- analysis yielded a Cronbach ␣ of 0.90 for baseline measures precluded intention- cluding nonresidents and respondents the 10 questions. The scale for accept- to-treat analysis; however, only 10 par- with incomplete data, the final sample ability of receiving gifts ranged from 0.24 ticipants were lost and were equally dis- consisted of 301 residents, 90 pediatric to 1, with a mean and median of 0.64 be- tributed across groups. All tests were residents (93% response rate before ex- fore conversion to a dichotomous ac- 2-sided. Data were analyzed using SPSS clusions) and 211 family medicine resi- ceptable/unacceptable score. The bi- software, version 16.0 (SPSS Inc, Chi- dents. The family medicine resident nary cutoff for “acceptable” was the mean cago, Illinois). An alternative analysis sample included responses from 26 dif- of the scale, 0.6401; scores at or above using a ratio scale and linear regres- ferent states and 65 hospitals/medical this cutoff were converted to accept- sions produced similar results and is centers (with 1-12 residents per hospi- able and any score below it was con- shown in eTable 1. tal). Aggregated across both samples, verted to unacceptable.

Figure 2. Participant Flow

100 Pediatric residents in 1 450 Family medicine residency hospital sent survey directorsa forwarded survey via e-mail via e-mail to residents

323 Residentsb responded 93 Pediatric residents 230 Family medicine residents

0 Excluded

323 Residents randomized

69 Residents randomized to 127 Residents randomized to 127 Residents randomized to control surveyc sacrifice reminders survey suggested rationalization survey 21 Pediatric residents 36 Pediatric residents 36 Pediatric residents 20 Completed survey as 91 Family medicine residents 91 Family medicine residents assigned 1 Did not complete survey 48 Family medicine residents 46 Completed survey as 67 Randomized to “feel-rich” 60 Randomized to “feel-poor” 65 Randomized to “feel-rich” 62 Randomized to “feel-poor” assigned survey survey survey survey 2 Did not complete 18 Pediatric residents 18 Pediatric residents 18 Pediatric residents 18 Pediatric residents survey 18 Completed survey as 17 Completed survey as 17 Completed survey as 18 Completed survey as assigned assigned assigned assigned 49 Family medicine residents 1 Did not complete 1 Did not complete 44 Family medicine residents 48 Completed survey as survey survey 43 Completed survey as assigned 42 Family medicine residents 47 Family medicine residents assigned 1 Did not complete 40 Completed survey as 46 Completed survey as 1 Did not complete survey assigned assigned survey 2 Did not complete 1 Did not complete survey survey

60 Included in analysis 64 Included in analysis 56 Included in analysis 62 Included in analysis 59 Included in analysis 20 Pediatric residents 18 Pediatric residents 17 Pediatric residents 17 Pediatric residents 18 Pediatric residents 40 Family medicine residents 46 Family medicine residents 39 Family medicine residents 45 Family medicine residents 41 Family medicine residents 6 Family medicine nonresidents 2 Family medicine nonresidents 1 Family medicine nonresident 1 Family medicine nonresident 2 Family medicine nonresidents excluded from analysis excluded from analysis excluded from analysis excluded from analysis (fellow) excluded from analysis 4 Attending (attendings) (attending) 1 Attending 1 Fellow 1 Fellow 1 Title unknown

aThe number of residency directors receiving requests is approximate. bSome respondents were nonresidents and were excluded from the analysis as shown below. cTo enable comparison of responses between the groups, the control group was also randomized to feel-rich and feel-poor subgroups, but only after the dependent variable was measured, so this did not constitute an experimental manipulation. Within the control group, 27 were allocated to feel-rich and 33 to feel-poor subgroups.

Table. Participant Characteristics No. (%) of Participants

Control Sacrifice Reminders Suggested Rationalization P Value for Differences Characteristics (n = 60) (n = 120) (n = 121) Between Groupsa Postgraduate year 1 26 (43.3) 50 (41.7) 45 (37.2) 2 17 (28.3) 30 (25.0) 36 (29.8) .85 3 13 (21.7) 34 (28.3) 35 (28.8) Ն4 and chief residents 4 (6.7) 6 (5.0) 5 (4.1) Working h/wk Յ50 8 (13.3) 8 (6.7) 11 (9.1) 51-60 10 (16.7) 22 (18.3) 23 (19.0) 61-70 18 (30.0) 31 (25.8) 41 (33.9) .20 71-80 19 (31.7) 41 (34.2) 38 (31.4) Ն81 5 (8.3) 18 (15.0) 8 (6.6) Annual salary, $b Feel poor 0-100 000 33 (100) 56 (100) 57 (96.6) 100 000-150 000 0 0 2 (3.4) 150 000-200 000 0 0 0 200 000-250 000 0 0 0 .22 250 000-300 000 0 0 0 300 000-350 000 0 0 0 Ͼ350 000 0 0 0 Feel rich 0-20 000 1 (3.7) 1 (1.6) 0 20 000-25 000 0 1 (1.6) 0 25 000-30 000 0 1 (1.6) 0 30 000-35 000 2 (7.4) 2 (3.1) 3 (4.8) .33 35 000-40 000 3 (11.1) 6 (9.4) 1 (1.6) 40 000-50 000 18 (66.7) 43 (67.2) 53 (85.5) Ͼ50 000 3 (11.1) 10 (15.6) 5 (8.1) On-call sleep per night, hb Feel poor 0-5 31 (92.9) 55 (98.2) 54 (91.5) 5-6 2 (6.1) 1 (1.8) 5 (8.5) 6-7 0 0 0 .28 7-8 0 0 0 8-9 0 0 0 Ͼ9000 Feel rich 0-0.5 6 (22.2) 13 (20.3) 8 (12.9) 0.5-1 7 (25.9) 11 (17.2) 10 (16.1) 1-1.5 2 (7.4) 6 (9.4) 9 (14.5) .09 1.5-2 8 (29.6) 14 (21.9) 8 (12.9) 2-3 4 (14.8) 11 (17.2) 18 (29.0) Ͼ3 0 9 (14.1) 9 (14.5) Non–on-call sleep per night, h Feel poor 0-6 3 (9.1) 13 (23.2) 7 (11.9) 6-7 13 (39.4) 28 (50.0) 39 (66.1) 7-8 17 (51.5) 13 (23.2) 11 (18.6) .07 8-9 0 2 (3.6) 2 (3.4) 9-10 0 0 0 Ͼ10 0 0 0 (continued)

Table. Participant Characteristics (continued) No. (%) of Participants

Control Sacrifice Reminders Suggested Rationalization P Value for Differences Characteristics (n = 60) (n = 120) (n = 121) Between Groupsa Non–on-call sleep per night, h Feel rich 0-2 0 0 0 2-3 1 (3.7) 1 (1.6) 0 3-4 0 2 (3.1) 0 .13 4-5 2 (7.4) 1 (1.6) 4 (6.5) 5-6 12 (44.4) 19 (29.7) 16 (25.8) Ͼ6 12 (44.4) 41 (64.1) 42 (67.7) Perception of working conditionsc Յ2 5 (8.3) 8 (6.7) 6 (5.0) 3 25 (41.7) 43 (35.8) 44 (36.4) .96 4 16 (26.7) 48 (40.0) 55 (45.5) 5 14 (23.3) 21 (17.5) 16 (13.2) Rationalizationd 1 21 (35.0) 29 (24.4) 19 (15.7) 2 21 (35.0) 52 (43.7) 45 (37.2) .005 3 11 (18.3) 26 (21.8) 35 (28.9) Ն4 7 (11.7) 12 (10.1) 22 (18.1) Borrowed money for medical training 49 (81.7) 83 (69.2) 96 (79.3) .09e Receipt of gifts acceptablef 13 (21.7) 57 (47.5) 73 (60.3) Ͻ.001e a By analysis-of-variance test unless otherwise indicated. b In the control group, participants were also randomized into subgroups of “feel rich” and “feel poor” after the dependent variable (acceptability of gifts) had been measured to enable testing for comparability of groups resulting from randomization. c Working conditions were rated on a 5-point Likert scale (1=very bad; 5=very good). d Rationalization was rated on a 5-point Likert scale (1=strongly disagree; 5=strongly agree). e By ␹2 test. f Acceptability was dichotomized as described in the “Methods” section of the text.

Differences Between Samples tween the feel-rich and feel-poor sacrifice questions were consistent with and Randomized Groups subgroups such that more responses the intended influence on what physi- Differences between the 2 samples are were in the lower categories (1 or 2) in cians thought about their working con- displayed in eTable 2. There were no the feel-poor than in the feel-rich sub- ditions. Poor working conditions were significant differences between the pe- groups. In regard to salary, 100% (148/ reported by 50.0% (74/148) of those in diatric and family medicine residents 148) of those in feel-poor subgroups re- the feel-poor subgroups compared with in working hours, salary, non–on-call sponded in the lower categories vs 2% 37.3% (57/153) in the feel-rich sub- sleep, whether they borrowed money (3/153) of those in feel-rich subgroups (χ2=4.97; OR, 1.68; 95% con- to fund their education, agreement with groups (χ2=290; PϽ.001), Similarly, fidence interval [CI], 1.06-2.67; P=.03). the rationalization, and acceptability of for on-call sleep, 100% (148/148) of gifts. However, relative to pediatrics those in feel-poor subgroups re- Rationalization Statement residents, family medicine residents sponded in lower categories vs 35.9% Of the residents, 37.5% (113/301) were more likely to be male (PϽ.001), (55/153) of those in feel-rich sub- agreed with the rationalization. More to be in their first year (PϽ.001), to re- groups (χ2=186; PϽ.001); for non– agreed with the rationalization state- port sleeping more hours when on call on-call sleep, 69.6% (103/148) re- ment in the suggested rationalization (PϽ.001), and to have a more posi- sponded in lower categories in feel- group (in which the statement came im- tive perception of their working con- poor subgroups vs 1.3% (2/153) in feel- mediately after the sacrifice questions ditions (P=.001). rich subgroups (χ2=275; PϽ.001). but before the COI questions) (47.1% There were no significant differ- Of the physicians surveyed, 94% [57/121]) compared with those who re- ences among the 3 randomized groups (282/301) thought that their working ceived it at the end of the survey in the in reported postgraduate year, work- conditions were okay, good, or very control or sacrifice reminders groups ing hours, salary, sleeping hours, edu- good on a 5-point scale ranging from (30.0% [18/60] and 31.7% [38/120], re- cation-related debt, and perception of very bad to very good. Only 6% (19/ spectively; χ2=7.94; P=.02) (compar- working conditions (TABLE). As in- 301) reported bad or very bad working the suggested rationalization group tended, there were significant differ- ing conditions. The results of the rich- with control and sacrifice reminder ences in the salary and sleep items be- vs-poor category manipulation on the groups combined, OR, 1.97; 95% CI,

ers group). Covariates for sample (when ined the effect of the rich-vs-poor ma- Figure 3. Acceptability of Receiving Gifts by Condition and Agreement With relevant) and postgraduate year were nipulation while controlling for the Rationalization initially included in every model but effect of the suggested rationalization. were in no case significant; thus, they It found that gift acceptability was Agree with rationalization were not included in these or subse- greater in the feel-poor subgroups than Disagree with rationalization 1.00 quent regressions. in the feel-rich subgroups (60.9% [70/ Results were similar when analyses 115] vs 47.6% [60/126]); χ2=4.25; OR,

0.75 were conducted in each resident sub- 1.71; 95% CI, 1.02-2.86; P=.04). group. Among pediatric residents, sacrifice reminders increased gift accept- 0.50 ability from 15.0% (3/20) in the control COMMENT group to 42.9% (15/35) (OR, 2.06; 95% Our results support the view that the 0.25 CI, 1.03-4.15; P=.04), and the ratio- perception of hardships may contrib- That Gifts Are Acceptable That Gifts Are nalization statement further increased ute to physician acceptance of gifts

Proportion of Respondents Reporting Proportion 0.00 gift acceptability to 48.6% (17/35) (OR, from the pharmaceutical industry. Control Sacrifice Suggested Reminders Rationalization 1.37; 95% CI, 1.00-1.90; P=.05 for dif- Even though few residents reported ference from sacrifice reminders group). that their working conditions were Those who agreed with the rationalization were more likely to accept gifts. Those who disagreed with the Among family medicine residents, sac- bad, reminding them about sacrifices rationalization were more likely to change their re- rifice reminders increased gift accept- to obtain their medical education sig- sponse about the acceptability of receiving gifts when they received sacrifice reminders or a suggested ra- ability from 25% (10/40) in the con- nificantly increased their readiness to tionalization. Error bars indicate 95% confidence trol group to 49.4% (42/85) (OR, 1.71; receive gifts. Providing a suggested intervals. 95% CI, 1.23-2.60; P=.01), and the ra- rationalization that low salaries and tionalization statement further in- education-related debt could poten- 1.23-3.18; P=.005). Respondents who creased gift acceptability to 65.1% (56/ tially justify accepting gifts increased reported poor working conditions were 86) (OR, 1.48; 95% CI, 1.21-1.82; the acceptability of industry- more likely to agree with the rational- PϽ.001 for difference from sacrifice re- sponsored gifts beyond the effect of ization (45.0% [59/131]) than those minders group). simple sacrifice reminders. who reported favorable working con- In a more detailed model that in- Furthermore, agreement with the ra- ditions (31.8% [54/170]; χ2=5.56; OR, cluded covariates for agreement with the tionalization statement was strongest 1.76; 95% CI, 1.10-2.82; P=.02). rationalization and interactions be- when it immediately followed the sac- tween agreement with the rationaliza- rifice reminders, indicating that feel- Acceptability of Receiving Gifts tion and the sacrifice reminders and sug- ings of hardship can increase justifica- All 10 of the COI items shifted in the pre- gested rationalization conditions, gift tions for ethically questionable dicted direction across groups; for all 10 acceptability was positively and signifi- behavior. Although those who agreed items, mean gift acceptability was high- cantly related to agreement with the ra- more with the rationalization were more est in the suggested rationalization group tionalization (OR, 10.61; 95% CI, 4.63- likely to view receiving gifts as accept- and lowest in the control group. 24.31; P Ͻ .001). The interactions able, those who disagreed with the ra- The first logistic regression model, between agreement and the sacrifice re- tionalization were most vulnerable to containing only variables for the main minders (OR, 0.21; 95% CI, 0.06-0.68; the influence of sacrifice reminders and conditions (suggested rationalization P=.009) and agreement and the sug- the suggested rationalization. This sug- and sacrifice reminders), found that re- gested rationalization (OR, 0.62; 95% CI, gests that “because you’re worth it” minding physicians of sacrifices made 0.39-0.996; P=.048) were also signifi- primes, such as those provided by sac- in obtaining their education resulted in cant. The pattern of main effects and the rifice reminders and suggested ratio- gifts being evaluated as more accept- interaction is depicted in FIGURE 3. For nalizations, are especially effective for able: 21.7% (13/60) in the control group those who accepted the rationaliza- those who would, in their absence, be vs 47.5% (57/120) in the sacrifice re- tion, gift acceptability was uniformly least likely to accept gifts. minders group (OR, 1.81; 95% CI, 1.27- high in all 3 experimental groups. In The justifications may not occur on 2.58; P=.001). Although most resi- contrast, for those who rejected the ra- a conscious level, since most respondents disagreed with the suggested tionalization, the experimental manipu- dents denied that their working con- rationalization, exposure to it further lations for sacrifice reminders and sug- ditions were poor and explicitly re- increased the perceived acceptability of gested rationalization had a substantial jected the suggested rationalization. gifts to 60.3% (73/121) of residents in effect on gift acceptability. Also, as demonstrated by the effect of the suggested rationalization group An additional model that included the feel-poor manipulation, the qual- (OR, 1.45; 95% CI, 1.22-1.72; PϽ.001 only the sacrifice reminders and sug- ity of working conditions is itself a sub- for difference from sacrifice remind- gested rationalization groups exam- jective judgment, one that can poten-

1210 JAMA, September 15, 2010—Vol 304, No. 11 (Reprinted) ©2010 American Medical Association. All rights reserved. RATIONALIZING ACCEPTANCE OF INDUSTRY GIFTS tially be used to support subconscious The limitations of this study were the as sex and other individual differ- rationalizations to accept gifts. Resi- self-reported nature of the data and the ences. dents’ rate of agreement to the ratio- possible nonrepresentativeness of one In summary, financial self-interest nalization in our study (37.6%) was of the samples. The 93% response rate may not fully explain physicians’ ac- much lower than that previously found in the Children’s Hospital of Pitts- ceptance of gifts. Rather, such accep- in medical students (80.3%).8 How- burgh sample ensures an adequate rep- tance may be facilitated by rationaliza- ever, the more deprived that physi- resentation of residents from 1 hospi- tions. Research has documented that cians felt (as determined by reporting tal. The family medicine sample was gifts are widespread10 and can influ- poor working conditions), the more added to increase power and enhance ence physician prescribing behav- likely they were to agree with the ra- the generalizability of the results by in- ior.1,17 This study helps explain how tionalization; and the more likely they cluding residents from another spe- well-intentioned physicians may use were to agree with the rationalization, cialty. However, we were unable to cal- subjective perceptions of hardships to the more likely they were to report gifts culate the response rate of family rationalize acceptance of such poten- as acceptable. medicine residents, and the mean re- tially biasing gifts.

Research on self-serving bias sug- sponses to the COI items of the family Author Contributions: Dr Sah had full access to all of gests that individuals’ views of what is medicine residents who chose to par- the data in the study and takes responsibility for fair is often biased in the direction of ticipate may not be representative of the the integrity of the data and the accuracy of the data 12 analysis. what benefits them personally. Ap- overall population. Given that they re- Study concept and design: Sah, Loewenstein. plied to COIs, the self-serving bias is sponded to the chance of obtaining a Acquisition of data: Sah. Analysis and interpretation of data: Sah, Loewenstein. likely to increase feelings of entitle- portable media player, it is possible that Drafting of the manuscript: Sah, Loewenstein. ment and disarm reservations that family medicine respondents include a Critical revision of the manuscript for important in- tellectual content: Sah, Loewenstein. might otherwise arise about accep- disproportion of physicians who are at- Statistical analysis: Sah, Loewenstein. tance of gifts. Moreover, people are gen- tracted to moderate-sized gifts. There Obtained funding: Loewenstein. erally not aware that they are subject are, however, several reasons to be- Administrative, technical, or material support: Loewenstein. to a self-serving bias, which means that lieve that this is not a major problem. Study supervision: Loewenstein. physicians are unlikely to recognize that First, since we were testing a causal Financial Disclosures: None reported. 13 Funding/Support: This study was supported by Car- they have been influenced by gifts. mechanism via experimental design, the negie Mellon’s internal departmental research Furthermore, individuals often en- use of blind randomization should pro- funds. Role of the Sponsors: The funder had no role in de- gage in minor forms of unethical be- duce comparability between groups, sign and conduct of the study; collection, manage- havior, falling short of behaviors that thus reducing the effect of response bias ment, analysis, and interpretation of the data; or prepa- ration, review, or approval of the manuscript. are sufficiently extreme to negatively even if the sample is not perfectly rep- Online-Only Content: eFigures 1 through 3 and affect their self-concept of being an hon- resentative of the larger population. Sec- eTables 1 and 2 are available online at http://www est person,14 and research on reciproc- ond, separate analysis of the samples re- .jama.com. Additional Contributions: Katherine Neely, MD, West ity suggests that favors given are likely vealed similar results. It is also possible Penn Allegheny Health System, assisted by sending to be paid back.15,16 There appear to be that the cultural backgrounds of the e-mails to enable residents to participate in the study. Robert Schoen, MD, University of Pittsburgh, pro- important psychological factors oper- residents could affect attitudes toward vided helpful comments and suggestions in the prepa- ating to weaken physicians’ compunc- accepting gifts as well as perceptions of ration of this article, and Howard Seltman, MD, PhD, Carnegie Mellon University, provided helpful statis- tions about accepting, or altering their personal sacrifice. Future studies could tical advice. None of these contributors received any behavior in response to, gifts. look at associations of culture as well compensation.

REFERENCES 1. Wazana A. Physicians and the pharmaceutical indus- 7. Chimonas S, Brennan TA, Rothman DJ. Physicians and tive on gifts to physicians from industry. JAMA. 2003; try: is a gift ever just a gift? JAMA. 2000;283(3): drug representatives: exploring the dynamics of the rela- 290(2):252-255. 373-380. tionship. J Gen Intern Med. 2007;22(2):184-190. 14. Mazar N, Amir O, Ariely D. The dishonesty of hon- 2. Campbell EG. Doctors and drug companies— 8. Sierles FS, Brodkey AC, Cleary LM, et al. Medical est people: a theory of self-concept maintenance. scrutinizing influential relationships. N Engl J Med. students’ exposure to and attitudes about drug com- J Mark Res. 2008;45(6):633-644. 2007;357(18):1796-1797. pany interactions: a national survey. JAMA. 2005; 15. Montague R. The perspective from neuroscience. 3. Fugh-Berman A, Ahari S. Following the script: how 294(9):1034-1042. In: The Scientific Basis of Influence and Reciprocity: drug reps make friends and influence doctors. PLoS 9. Adams JS. Inequity in social exchange. Adv Exp Soc A Symposium. June 12, 2007. https://services.aamc Med. 2007;4(4):e150. Psychol. 1965;2:267-299. .org/publications/index.cfm?fuseaction=Product 4. Steinman MA, Bero LA, Chren MM, Landefeld CS. 10. Campbell EG, Gruen RL, Mountford J, Miller LG, .displayForm&prd_id=215&cfid=1&cftoken Narrative review: the promotion of gabapentin: an analy- Cleary PD, Blumenthal D. A national survey of phy- =8BD66981-BC84-D32B-566C0DAEC4642307. Ac- sis of internal industry documents. Ann Intern Med. 2006; sician-industry relationships. N Engl J Med. 2007; cessed August 19, 2010. 145(4):284-293. 356(17):1742-1750. 16. King-Casas B, Tomlin D, Anen C, Camerer CF, 5. Steinman MA, Harper GM, Chren MM, Landefeld 11. Cohen J, Cohen P, West SG, Aiken LS. Applied Quartz SR, Montague PR. Getting to know you: repu- CS, Bero LA. Characteristics and impact of drug detail- Multiple Regression/Correlation for the Behavioral tation and trust in a two-person economic exchange. ing for gabapentin. PLoS Med. 2007;4(4):e134. Sciences. Hillsdale, NJ: Erlbaum; 1983. Science. 2005;308(5718):78-83. 6. Brennan TA, Rothman DJ, Blank L, et al. Health in- 12. Babcock L, Loewenstein G, Issacharoff S, Camerer 17. Zipkin DA, Steinman MA. Interactions between dustry practices that create conflicts of interest: a policy C. Biased judgments of fairness in bargaining. Am Econ pharmaceutical representatives and doctors in train- proposal for academic medical centers. JAMA. 2006; Rev. 1995;85:1337-1343. ing: a thematic review. J Gen Intern Med. 2005; 295(4):429-433. 13. Dana J, Loewenstein G. A social science perspec- 20(8):777-786.