personal and clinical tobacco-related practices and attitudes of u.s. medical students

7
Personal and clinical tobacco-related practices and attitudes of U.S. medical students Erica Frank a,b, , Lisa Elon c , Elsa Spencer b a University of British Columbia, Department of Health Care and Epidemiology, and Department of Family Practice, 5804 Fairview Avenue, Vancouver, BC, Canada b Emory University School of Medicine, Department of Family and Preventive Medicine, 49 Jesse Hill Jr. Drive, Atlanta, GA 30303, USA c Emory University Rollins School of Public Health, Department of Biostatistics and Bioinformatics, 1518 Clifton Rd, Atlanta, GA 30322, USA abstract article info Available online 2 July 2009 Keywords: Tobacco Cigarettes Medical students Physician health Physician Objectives. Medical students' tobacco-related practices particularly matter because practicing physicians' smoking predicts their tobacco counseling; the objective of this study was to determine when this relationship between personal and clinical tobacco practices develops, and to determine predictors of medical students' personal smoking habits, and predictors of their patient tobacco counseling practices. Methods. We surveyed the Class of 2003 between 1999 and 2003, at freshman orientation (n = 1836), entrance to wards (n =1616), and senior year (n =1441) in a nationally representative sample of 16 U.S. medical schools (response rate=80.3%). Tobacco use questions came from Centers for Disease Control and Prevention datasets, and tobacco counseling questions from validated instruments. Results. 12% of female and 15% of male U.S. medical students report smoking, with no differences in usage over time. More tobacco counseling training and strongly believing in prevention signicantly predicted both more perceived counseling relevance and frequency. Additionally, intention to practice primary care predicted relevance (OR = 3.5, 95% CI: 2.54.9), and tobacco users were 77% (95% CI: 64%94%) as likely as non-users to report frequently counseling smokers. Conclusions. U.S. medical students are less likely to smoke than other young U.S. adults, but more likely than U.S. physicians, and showed no clear decrease during medical school. It is encouraging that medical students with more exposure appreciate tobacco counseling's importance more, and are more likely to counsel. Students' personal tobacco use was also associated with counseling frequency. These data should help educators seeking better methods to reduce tobacco use. © 2009 Elsevier Inc. All rights reserved. Introduction Tobacco use is the leading global cause of preventable disease and premature death (World Health Organization, 2005). Major health care organizations and authorities recommend that physicians provide their patients with regular tobacco interventions (World Medical Association, 2007). Despite these recommendations, most smokers are not advised to quit during their clinical encounters (Fiore, 2000; Frank et al., 1991). A seven-study meta-analysis by the U.S. Public Health Service (Fiore, 2000) found that even brief (3 min of) physician advice resulted in a small but statistically signicant increase in smoking quit rates (summary OR: 1.3; 95% CI: 1.11.6). We wished to examine medical students' tobacco-related practices for three main reasons. First, because their tobacco use may predict tobacco use in the general population. We hypothesized that U.S. medical students would, like U.S. physicians, have low cigarette smoking rates. U.S. men physicians have low cigarette use (b 4%) (Nelson et al., 1994), and U.S. women physicians are also less likely (3.7%) than other high socioeconomic status (SES) U.S. women (8%) and other U.S. women (25%) to smoke cigarettes, to consume fewer cigarettes/day (12 vs.16 for high SES or 18 for all other women), and to quit smoking at a younger age (Frank et al., 1998). Second, medical students' tobacco use rates are of interest because physicians' personal tobacco use has been shown to be a predictor of their patient counseling (Frank, 2004), but the point at which this relationship develops is unknown And third, because physicians have been encouraged to be role models (World Medical Association, 2007) on tobacco use, but North American medical students' personal tobacco use habits have not been extensively examined. Methods General methods All medical students in the Class of 2003 at 16 U.S. schools were eligible to complete three questionnaire administrations during their medical training: at freshman orientation, orientation to wards (typically between their second and third years), and in their senior year. Surveys were administered between 1999 and 2003. Our sample of schools was designed to be representative of all U.S. medical schools (Frank et al., 2008, 2004a). The Healthy Doc questionnaire protocol was approved by the Emory University IRB. Paper copies of the questionnaires were usually administered Preventive Medicine 49 (2009) 233239 Corresponding author. University of British Columbia, Department of Health Care and Epidemiology, and Department of Family Practice, 5804 Fairview Avenue, Vancouver, BC, Canada. Fax: +1404 616 6847. E-mail addresses: [email protected], [email protected] (E. Frank). 0091-7435/$ see front matter © 2009 Elsevier Inc. All rights reserved. doi:10.1016/j.ypmed.2009.06.020 Contents lists available at ScienceDirect Preventive Medicine journal homepage: www.elsevier.com/locate/ypmed

Upload: erica-frank

Post on 31-Oct-2016

215 views

Category:

Documents


2 download

TRANSCRIPT

Preventive Medicine 49 (2009) 233–239

Contents lists available at ScienceDirect

Preventive Medicine

j ourna l homepage: www.e lsev ie r.com/ locate /ypmed

Personal and clinical tobacco-related practices and attitudes of U.S. medical students

Erica Frank a,b,⁎, Lisa Elon c, Elsa Spencer b

a University of British Columbia, Department of Health Care and Epidemiology, and Department of Family Practice, 5804 Fairview Avenue, Vancouver, BC, Canadab Emory University School of Medicine, Department of Family and Preventive Medicine, 49 Jesse Hill Jr. Drive, Atlanta, GA 30303, USAc Emory University Rollins School of Public Health, Department of Biostatistics and Bioinformatics, 1518 Clifton Rd, Atlanta, GA 30322, USA

⁎ Corresponding author. University of British Columband Epidemiology, and Department of Family PracVancouver, BC, Canada. Fax: +1 404 616 6847.

E-mail addresses: [email protected], erica.frank@ub

0091-7435/$ – see front matter © 2009 Elsevier Inc. Aldoi:10.1016/j.ypmed.2009.06.020

a b s t r a c t

a r t i c l e i n f o

Available online 2 July 2009

Keywords:TobaccoCigarettesMedical studentsPhysician healthPhysician

Objectives.Medical students' tobacco-related practices particularly matter because practicing physicians'smoking predicts their tobacco counseling; the objective of this study was to determine when thisrelationship between personal and clinical tobacco practices develops, and to determine predictors ofmedical students' personal smoking habits, and predictors of their patient tobacco counseling practices.

Methods. We surveyed the Class of 2003 between 1999 and 2003, at freshman orientation (n=1836),entrance to wards (n=1616), and senior year (n=1441) in a nationally representative sample of 16 U.S.medical schools (response rate=80.3%). Tobacco use questions came from Centers for Disease Control and

Prevention datasets, and tobacco counseling questions from validated instruments.

Results.12% of female and 15% of male U.S. medical students report smoking, with no differences in usageover time. More tobacco counseling training and strongly believing in prevention significantly predicted bothmore perceived counseling relevance and frequency. Additionally, intention to practice primary carepredicted relevance (OR=3.5, 95% CI: 2.5–4.9), and tobacco users were 77% (95% CI: 64%–94%) as likely asnon-users to report frequently counseling smokers.

Conclusions. U.S. medical students are less likely to smoke than other young U.S. adults, but more likelythan U.S. physicians, and showed no clear decrease during medical school. It is encouraging that medicalstudents with more exposure appreciate tobacco counseling's importance more, and are more likely tocounsel. Students' personal tobacco use was also associated with counseling frequency. These data shouldhelp educators seeking better methods to reduce tobacco use.

© 2009 Elsevier Inc. All rights reserved.

Introduction

Tobacco use is the leading global cause of preventable disease andpremature death (World Health Organization, 2005). Major healthcare organizations and authorities recommend that physiciansprovide their patients with regular tobacco interventions (WorldMedical Association, 2007). Despite these recommendations, mostsmokers are not advised to quit during their clinical encounters (Fiore,2000; Frank et al., 1991). A seven-study meta-analysis by the U.S.Public Health Service (Fiore, 2000) found that even brief (≤3 min of)physician advice resulted in a small but statistically significantincrease in smoking quit rates (summary OR: 1.3; 95% CI: 1.1–1.6).

Wewished to examinemedical students' tobacco-related practicesfor three main reasons. First, because their tobacco use may predicttobacco use in the general population. We hypothesized that U.S.medical students would, like U.S. physicians, have low cigarettesmoking rates. U.S. men physicians have low cigarette use (b4%)(Nelson et al., 1994), and U.S. women physicians are also less likely

ia, Department of Health Caretice, 5804 Fairview Avenue,

c.ca (E. Frank).

l rights reserved.

(3.7%) than other high socioeconomic status (SES) U.S. women (8%)and other U.S. women (25%) to smoke cigarettes, to consume fewercigarettes/day (12 vs. 16 for high SES or 18 for all other women), andto quit smoking at a younger age (Frank et al., 1998). Second, medicalstudents' tobacco use rates are of interest because physicians' personaltobacco use has been shown to be a predictor of their patientcounseling (Frank, 2004), but the point at which this relationshipdevelops is unknown And third, because physicians have beenencouraged to be role models (World Medical Association, 2007) ontobacco use, but North American medical students' personal tobaccouse habits have not been extensively examined.

Methods

General methods

All medical students in the Class of 2003 at 16 U.S. schools were eligible tocomplete three questionnaire administrations during their medical training:at freshman orientation, orientation to wards (typically between their secondand third years), and in their senior year. Surveys were administered between1999 and 2003. Our sample of schools was designed to be representative of allU.S. medical schools (Frank et al., 2008, 2004a).

The Healthy Doc questionnaire protocol was approved by the EmoryUniversity IRB. Paper copies of the questionnaires were usually administered

234 E. Frank et al. / Preventive Medicine 49 (2009) 233–239

after semi-mandatory activities (such as tests or class meetings); studentswere informed that questionnaires were anonymous and confidential, andparticipation was voluntary. Our response rate was 80.3%; a total of 2316individuals responded at some point to our survey.

Description of variables

Our primary professional outcomes were smoking cessation counselingvariables: 1) perceived relevance (“How relevant do you think talking topatients about smoking will be in your intended practice?”) and 2) frequency(“With a typical general medicine patient, how often do you actually talk topatients about tobacco?”). Relevance responses were queried at all timepoints, as “not at all”, “somewhat”, and “highly”; frequency responses werequeried only in senior year, and were “never–rarely”, “sometimes”, and“usually–always”.

Statistical analysis

There was a lack of independence between observations due to studentclustering within schools and students being measured multiple times; weadjusted variance estimates to account for the dependencies. For all analysesdescribed below, we used SUDAAN (Shah, 2006), software designed for theanalysis of clustered data, treating each school as a cluster and each student'smultiple responses as subclusters.

Bivariate associations between our three outcomes (tobacco use,counseling relevance, and counseling frequency) and independent variableswere tested using the Chi square test. Due to the number of associations beingtested, we limit our discussion of significant results to those with pb0.01.

We estimated model parameters with a SUDAAN procedure usingworking exchangeable generalized estimating equations with robust varianceestimation, and these are presented with their 95% confidence intervals. Weassessed model fit via standardized deviance residuals and the Hosmer–Lemeshow goodness of fit test (Hosmer and Lemeshow, 1989).

Results

Personal tobacco use

Table 1 reports personal use of tobacco products among U.S. femaleand male medical students. Men reported using significantly more ofeach type of tobacco product, especially more cigars. The associationof students' tobacco use with personal, professional, classmate, andschool characteristics is shown in Table 2. After separating (data notshown) the current non-smokers into past and never-smokers, wefound that those with a family history of smoking were more likely tohave previously smoked, as compared to those without a familyhistory (past, 7% vs. 4%; never, 70% vs. 77%; p=0.006).

Patient counseling regarding tobacco

Table 3 shows that significant (pb0.01) correlates of higherperceived relevance and more frequent counseling were reporting:being female, using tobacco less often in the past month, intended

Table 1Tobacco product use among U.S. medical students in the Class of 2003, by gender.

Product Tobacco product use atleast once in the past30 days, % (standard error)

Females Males p-value

Cigar 3 (0.6) 14 (1.6) b0.0001Chewing tobacco, dip, or snuff b1 (0.1) 6 (1.4) 0.0046Pipe 1 (0.1) 2 (0.3) 0.0006Cigarettes: any use 12 (1.4) 15 (1.7) 0.020Cigarettes: CDC definition of usea 6 (1.1) 8 (1.3) 0.0048Tobacco use of any type or frequencyb 14 (1.3) 27 (2.6) 0.0002

a The Centers for Disease Control (CDC) defines smokers as those reporting eversmoking ≥100 cigarettes and currently smoking “some days” or “every day”.

b May include use of multiple types of tobacco.

primary care specialty (at every time point), a belief in counselingpatients more effectively if one does not smoke, and a belief thatphysicians should promote prevention. Two additional predictors ofhigher perceived relevance were having classmates that discouragedsmoking, and believing that counseling helps promote patient health.Especially among non-primary care students, higher perceivedrelevance was also reported by those at least as interested inprevention as in treatment. We were interested to determine thepoint at which the relationship between personal and clinical tobacco-related practices began. We found that the perceived relevance oftobacco counseling among heavy, light, and past or non-smokers onlydiffered marginally in students' senior year (p=0.027), and not atprevious time points (p≥0.14). We also investigated whetherperceived relevance was associated with both point in the educationprocess and intended specialty; those intending to go into primarycare had a peak in their perception of tobacco relevance at wardorientation, while those intending to go into non-primary care hadreduced perception of tobacco relevance during senior year.

In models (Table 4), a student's school and classmate discourage-ments of smoking were both highly nonsignificant (pN0.95), andhence dropped, in relevance and frequency models. The personaltobacco use variable was collapsed in the frequency model (Table 5)because equivalent odds ratios (OR) were reported for heavy and lightusers, and for past and never-smokers.

Of the variables tested in full models, two were significant inpredicting both higher perceived relevance and reported frequency:reporting extensive training in smoking cessation and stronglyagreeing that physicians should promote prevention. Intention tosub-specialize in a primary care sub-specialty was only predictive offinding smoking counseling more relevant (OR=3.5, 95% confidenceinterval: 2.5, 4.9), and not of increased counseling frequency. Students'personal tobacco use was associated with lower frequency (but notrelevance) of counseling.

Discussion

Personal tobacco use

RatesWe found that 14% of female and 27% of male medical students

report using any form of tobacco in the past 30 days, with 12% and 15%reporting smoking cigarettes at all (and 6% and 8% smoking cigarettesaccording to the stricter CDC definition (Centers for Disease Controland Prevention, 1998). These rates are similar to 1987 findings of 10%of medical students (Hughes et al., 1991) and residents (Baldwin et al.,1991) reporting using cigarettes in the past month, and are far lowerthan the 25% of 25–44 year old U.S. women and the 29% of 25–44 yearold U.S. menwho smoked cigarettes in 1999–2001 (Schoenborn et al.,2004). We found that 6% of male and b1% of female medical studentsused dip, snuff, or chewing tobacco ≥1×/month, vs. 3% of U.S. menand b1% of women using these smokeless tobacco products N9 days/month, in 2001 (Mumford et al., 2006). While these tobacco data aregenerally favorable when compared with the U.S. adult population,they are far higher than the b4% cigarette smoking rates reported inthe 1990s for U.S. men and women physicians (Frank, 2004; Nelsonet al., 1994), and we observed no significant change in tobacco useduring medical school, and no substantive decrease in usage ratessince the 1987 survey of physicians-in-training. This combination ofobservations suggests that most of the few medical students whosmoke will stop during residency.

Potential improvements in ratesOne hopes that these unfavorable comparisons of medical students

with physicians do not reflect a cohort effect, but merely amaturational process that medical schools could try to accelerate.This is possible, as we demonstrated in a four-year health promotion

Table 2Association of personal, professional, school, and classmate characteristics with personal tobacco use of U.S. medical students in the Class of 2003.

Characteristic Nb Tobacco Usea, % (standard error) p-value

Frequent or heavy Infrequent or light No current use

Overall 4893 4 (0.7) 17 (1.4) 79 (1.9)

Personal variablesTime point 0.14Freshmen orientation 1836 3 (0.6) 20 (1.8) 78 (2.3)Introduction to wards 1616 4 (0.8) 16 (1.5) 80 (2.0)Senior year 1441 4 (1.0) 15 (1.6) 80 (2.0)

Gender 0.0009Female 2245 2 (0.6) 11 (1.0) 86 (1.3)Male 2642 5 (0.9) 22 (2.0) 73 (2.6)

Ethnicity 0.013Asian 927 3 (1.0) 14 (1.3) 82 (1.9)Black/African–American 386 1 (0.3) 6 (1.9) 93 (2.1)Hispanic 202 4 (2.2) 14 (2.4) 82 (3.5)White, non-Hispanic 3223 4 (0.9) 19 (1.5) 76 (2.2)Other 239 3 (1.4) 19 (2.9) 78 (2.6)

Stress in past 2 weeks 0.027Little/none 1847 3 (0.7) 18 (2.1) 79 (2.5)Moderate 1946 3 (0.6) 16 (1.5) 81 (2.0)A lot 1065 6 (1.4) 16 (1.1) 77 (1.8)

Stress in past 12 months 0.0004Little/None 925 2 (0.7) 22 (2.0) 76 (2.5)Moderate 2340 3 (0.7) 16 (1.6) 81 (1.9)A lot 1586 5 (1.0) 16 (1.5) 79 (2.0)

Mean days of bad mental health in the past month 4805 6 (0.6) 3 (0.3) 3 (0.2) 0.0047Perceived general health 0.0095Excellent 1676 2 (0.5) 16 (1.6) 82 (1.9)Very good 2197 3 (0.9) 17 (1.7) 80 (2.2)Good/Fair/Poor 999 7 (1.3) 20 (2.0) 73 (2.7)

Professional variableIntended specialty 0.026Primary care 1685 3 (0.7) 14 (1.6) 83 (1.8)Non-primary care 2468 4 (0.9) 19 (1.7) 77 (2.5)Undecided 694 3 (0.7) 18 (2.1) 79 (2.2)

Smoking-related variablesFamily history of smoking 0.0051Yes 2665 4 (0.7) 18 (1.2) 78 (1.7)No 2017 3 (0.7) 17 (1.9) 81 (2.4)

“I am at high risk for lung cancer” 0.0002Agree/strongly agree 214 26 (4.0) 29 (4.0) 45 (4.0)Neither agree nor disagree 753 6 (1.0) 22 (2.4) 72 (2.7)Disagree/strongly disagree 3747 2 (0.4) 16 (1.3) 83 (3.7)

“I am at high risk for heart disease” b0.0001Agree/strongly agree 1427 5 (0.8) 20 (1.8) 74 (2.2)Neither agree nor disagree 1071 4 (0.7) 18 (1.9) 79 (2.2)Disagree/strongly disagree 2228 2 (0.7) 15 (1.3) 83 (1.8)

“I will be able to provide more credible and effective counseling if I don't use tobacco”c 0.014Strongly agree 469 2 (0.6) 11 (1.2) 87 (1.3)Agree 713 4 (1.1) 16 (2.3) 80 (2.3)Neither agree nor disagree/disagree/strongly disagree 187 12 (2.8) 26 (3.4) 62 (4.8)

School discourages smokingc 0.28Agree/strongly agree 737 3 (1.0) 16 (2.2) 80 (2.4)Neither agree nor disagree 418 5 (1.1) 14 (1.9) 81 (2.8)Disagree/strongly disagree 220 5 (1.1) 16 (3.1) 79 (3.5)

Classmates discourage smokingc 0.016Agree/strongly agree 717 3 (0.9) 14 (2.1) 82 (2.3)Neither agree nor disagree 433 4 (1.0) 14 (1.5) 82 (2.3)Disagree/strongly disagree 226 6 (1.0) 23 (3.5) 71 (3.5)

a Frequent/heavy use: Either N10 cigarettes smoked/day and/or N19 days of tobacco use. Infrequent/light use: 1–10 cigarettes smoked/day and/or 1–19 days of tobacco use. Nocurrent use: past-smokers or never-smokers who do not currently use tobacco.

b Includes all three time points, with most individuals responding at more than one point, providing estimates for the period of medical school as a whole.c These variables were queried in senior year only.

235E. Frank et al. / Preventive Medicine 49 (2009) 233–239

intervention on the Class of 2003 at Emory University School ofMedicine (using the Class of 2002 as controls) (Frank et al., 2007b).We found that by senior year, control men reported twice the tobaccouse reported bymen in the intervention group (43% vs. 22%, p=0.02),although at entry to wards they reported nearly identical levels (31%vs. 29%, p=0.8). Women's tobacco use was much lower than men'sthroughout that study.

GenderWhilemen tend to usemore tobacco than dowomen in the general

population, the gender differential was especially strong among thesemedical students, especially for cigar use, chew, dip, and snuff.Nationally, while cigarette smoking rates for boys and girls differedmodestly in high school (26% vs. 21%) (Tomar, 2003), they werevirtually identical in college students (29% vs. 28%) (Rigotti et al.,

Table 3Crude relationships of gender, ethnicity, tobacco use, intended specialty, and opinions of U.S. medical students in the Class of 2003 with their perceived relevancea of smokingcounseling and reported frequencyb of counseling patients on smoking.

n Relevance % highly(standard error)

p-value n Frequency % usually/always(standard error)

p-value

Personal variablesGender b0.0001 0.0016Female 2154 74 (1.4) 653 63 (2.4)Male 2490 63 (1.7) 740 50 (2.3)

Ethnicity 0.052 0.47Asian 894 68 (2.3) 257 57 (2.9)Black/African–American 368 75 (2.7) 116 66 (5.4)Hispanic 194 71 (2.9) 57 56 (3.8)White, non-Hispanic 2964 66 (1.4) 888 55 (2.0)Other 217 68 (2.8) 73 52 (5.6)

Family history of smoking 0.89 0.22Yes 2550 68 (1.3) 707 54 (2.2)No 1930 68 (1.4) 528 58 (2.2)

Tobacco use in the past month 0.0088 0.0012No current use 3663 69 (1.4) 1116 58 (1.9)≤10 cigs/smoking day or ≤19 days of any tobacco use 808 63 (2.0) 213 50 (2.9)N10 cigs/smoking day or N19 days of any tobacco use 156 58 (3.3) 57 47 (3.8)

Clinical variablesIntended specialty Interaction N/Ac No interaction 0.0056Primary care 1632 80 (1.3) 421 64 (2.3)Non-primary care 2311 58 (2.0) 936 53 (2.4)Undecided 664 71 (2.0) 14 71 (13.1)

Intended specialty (stratified by time)Primary care b0.0001 –

Freshman orientation 791 76 (1.6) –

Introduction to wards 433 89 (1.4) –

Senior year 408 78 (2.9) –

Non-primary care 0.0027 –

Freshman orientation 659 61 (2.5) –

Introduction to wards 767 64 (2.2) –

Senior year 884 51 (2.6) –

Undecided 0.058 –

Freshmen orientation 353 67 (1.7) –

Introduction to wards 298 76 (3.0) –

Senior year 13 77 (11.3) –

“My medical school discourages students from smoking” 0.04 0.016Strongly agree 183 72 (3.0) 190 66 (2.5)Agree 519 60 (2.9) 547 52 (2.8)Neither agree nor disagree 398 55 (2.9) 413 56 (1.9)Disagree 170 56 (3.2) 177 56 (5.5)Strongly disagree 36 72 (9.1) 39 59 (8.7)

“My classmates discourage each other from smoking” 0.0077 0.13Strongly agree 131 69 (3.3) 136 62 (3.4)Agree 554 59 (3.3) 583 55 (3.0)Neither agree nor disagree 411 58 (3.5) 430 55 (2.7)Disagree 174 59 (4.7) 179 58 (6.0)Strongly disagree 36 72 (6.5) 39 54 (5.0)

Opinions on prevention-related items“I am less interested in prevention than treatment” (stratified by Intended specialty)All specialties b0.0001 0.014

Strongly agree 172 55 (3.7) 82 41 (6.0)Agree 817 56 (2.4) 307 53 (3.6)Neither agree nor disagree 1017 61 (2.0) 303 50 (2.6)Disagree 1895 73 (1.2) 504 61 (2.6)Strongly disagree 667 82 (1.2) 159 69 (2.8)

Primary care 0.016 0.29Strongly agree 24 88 (7.8) 9 67 (16.3)Agree 161 73 (3.6) 56 66 (6.2)Neither agree nor disagree 292 76 (2.5) 72 61 (4.9)Disagree 786 81 (2.0) 194 60 (3.5)Strongly disagree 347 86 (1.2) 77 73 (4.8)

Non-primary cared 0.0028 0.016Strongly agree 132 51 (3.9) 72 39 (6.1)Agree 549 50 (2.3) 241 50 (4.4)Neither agree nor disagree 560 53 (2.3) 223 45 (3.2)Disagree 800 64 (2.5) 298 61 (3.2)Strongly disagree 226 76 (3.0) 77 65 (5.9)

236 E. Frank et al. / Preventive Medicine 49 (2009) 233–239

Table 3 (continued)

n Relevance % highly(standard error)

p-value n Frequency % usually/always(standard error)

p-value

Opinions on prevention-related items“I am less interested in prevention than treatment” (stratified by Intended specialty)Undecided 0.0034 N/A

Strongly agree 15 47 (11.4) 0 N/AAgree 98 65 (4.4) 3 N/ANeither agree nor disagree 159 62 (4.0) 5 N/ADisagree 292 76 (2.4) 5 N/AStrongly disagree 88 82 (3.7) 1 N/A

“I will be able to provide more credible & effective counseling if I don't use tobacco” 0.0004 0.0005Strongly agree 456 70 (2.4) 469 67 (2.6)Agree 680 57 (2.9) 707 54 (2.3)Neither agree nor disagree/disagree/strongly disagree 165 45 (4.0) 183 36 (4.0)

Patients are more likely to adopt a healthy lifestyle if counseled 0.0031 0.032Strongly agree 761 76 (2.2) 239 61 (4.9)Agree 2890 68 (1.5) 844 58 (2.2)Neither agree nor disagree 718 62 (2.4) 191 42 (4.7)Disagree/strongly disagree 196 60 (4.0) 80 55 (5.7)

“Physicians have a responsibility to promote prevention” b0.0001 0.0042Strongly agree 1438 79 (1.4) 353 69 (2.4)Agree 2766 65 (1.5) 858 55 (2.3)Neither agree nor disagree/disagree/strongly disagree 356 46 (3.6) 142 32 (5.9)

a Perceived relevance to student's intended specialty of talking to smoking general medicine patients about smoking cessation. Includes all three time points, withmost individualsresponding at more than one point, providing estimates for the period of medical school as a whole.

b Reported frequency of talking to smoking general medicine patients about smoking cessation. This item was queried only in students' senior year.c Estimates in these cells are unstable due to either an interaction or low numbers of respondents.d Among non-primary care students, there was a significant trend (pb0.01) of more frequent counseling with more interest in prevention.

237E. Frank et al. / Preventive Medicine 49 (2009) 233–239

2007). Among college students the gender differences in tobacco use(38% in men vs. 30% in women) were primarily due to men's greateruse of cigars (16% vs. 4%; pb0.001) and smokeless tobacco (9% vs.b1%; pb0.001) (Rigotti et al., 2007).

ConsequencesRegarding consequences of personal tobacco use, these smokers

reported worse perceived general health; this is unusual for youngsmokers (Prokhorov et al., 2003). Medical students using tobacco alsoreported more days of bad mental health in the past month, and morestress in the past year. This is consistent with prior literature showingthat smokers exhibit more depression and anxiety (Jorm et al., 1999).We found that these future physicians generally acknowledged their

Table 4Multivariate correlates of perceived relevance of smoking cessation counseling tointended specialty, among senior year U.S. medical students in the Class of 2003.

Independent variables and effects Adj. oddsratioa

CIb lowerlimit

CI upperlimit

Relevance to tobacco counselingc (# highly relevant/n)Gender: females (378/574) 1.09 0.78 1.54Gender: males (354/648) 1.00 1.00 1.00Tobacco use: heavy/frequent or light/infrequent(122/237)

0.77 0.54 1.10

Tobacco use: never or past (610/985) 1.00 1.00 1.00Specialty: primary care (306/385) 3.53 2.54 4.89Specialty: non-primary care (426/837) 1.00 1.00 1.00Training: extensive (365/516) 2.15 1.60 2.87Training: some or none (367/706) 1.00 1.00 1.00Docs have responsibility to promote prevention:strongly agree (235/321)

2.61 1.38 4.93

Docs have responsibility to promote prevention:agree (448/336)

1.50 0.94 2.41

Docs have responsibility to promote prevention:neutral or disagree (49/117)

1.00 1.00 1.00

a Odds ratios are for perceiving counseling to be highly vs. somewhat or not at allrelevant, and for usually/always counseling vs. sometimes or never/rarely counseling.

b Confidence intervals (CI) reported are 95% confidence intervals.c Hosmer–Lemeshow goodness of fit test had a Chi-squared p-value of 0.43

(n=1222). Within-school clustering coefficient for perceived relevance of counselingwas 0.0072.

elevated risks of lung cancer and heart disease if they were smokers.The smokers who did not perceive themselves at high risk may haveeither believed that they would become ex-smokers before their risksbecame high, or understood that many smokers do not experiencelung cancer or heart disease.

Family historyThose students with a family history of smoking were more likely

to have tried smoking, and to have quit smoking. This is unlike priordata in non-medical populations, showing that children of smokersare more likely to be smokers themselves (den Exter Blokland et al.,2004; Otten et al., 2007). This may reflect other controlling andrescuing psychologies of those choosing medicine.

Table 5Multivariate correlates of reported frequency of smoking cessation counseling, amongsenior year medical students.

Independent variables and effects Adj. oddsratioa

CIb lowerlimit

CI upperlimit

Frequency of tobacco counselingc (# usually–always/n)Gender: females (363/521) 1.33 0.99 1.79Gender: males (349/724) 1.00 1.00 1.00Tobacco use: heavy/frequent or light/infrequent(120/243)

0.77 0.64 0.94

Tobacco use: never or past (592/1002) 1.00 1.00 1.00Specialty: primary care (249/390) 1.28 0.96 1.70Specialty: non-primary care (463/855) 1.00 1.00 1.00Training in smoking cessation: extensive (363/521) 2.33 1.73 3.13Training in smoking cessation: some or none(349/724)

1.00 1.00 1.00

Docs have responsibility to promote prevention:strongly agree (228/326)

3.22 1.45 7.15

Docs have responsibility to promote prevention:agree (441/794)

1.97 0.91 4.27

Docs have responsibility to promote prevention:neutral or disagree (43/125)

1.00 1.00 1.00

a Odds ratios are for perceiving counseling to be highly vs. somewhat or not at allrelevant, and for usually/always counseling vs. sometimes or never/rarely counseling.

b Confidence intervals (CI) reported are 95% confidence intervals.c Hosmer–Lemeshow goodness of fit test had a Chi-squared p-value of 0.38

(n=1245). Within-school clustering coefficient for reported counseling frequencywas 0.0052.

238 E. Frank et al. / Preventive Medicine 49 (2009) 233–239

Patient counseling regarding tobacco

Physician counseling on smoking cessation can have a powerfuleffect on patients' smoking habits (Duncan et al., 1992; Gilpin et al.,1993), (Agency for Healthcare Quality and Research &U.S. Departmentof Health and Human Services, 2003) so it was encouraging that wefound several malleable variables associated with students' tobaccocounseling attitudes and practices.

TrainingExtensive training in smoking cessation was strongly associated

with both perceived counseling relevance and counseling frequency.While we cannot determine causality, it is encouraging that studentswith more exposure to tobacco counseling education appreciate morethe importance of this activity for their future practices, and are morelikely to currently counsel their smoking patients. Prior literature(Cornuz et al., 1997; Strecher et al., 1991) also support that traininginterventions increase smoking cessation counseling.

AttitudesIndependent of intended specialty, students strongly agreeing that

it is a physicians' responsibility to promote prevention reportedsignificantly higher tobacco counseling relevance and frequency.Others have also found that prevention-related attitudes influenceperceived relevance and frequency of counseling (Foster et al., 2002;Rafferty and Frank, 1994).

Specialty choice and medical school milieuThe relationship with specialty choice likely operates through a

combination of attitude and exposure: medical students choosingprimary care have more prevention-oriented attitudes (Scott et al.,1992), and primary care rotations also provide more exposure toprevention, since practicing primary care physicians counsel patientsmore about prevention than do other specialists (Frank and Kunovich-Frieze, 1995; Frank et al., 2000). We found in bivariate analyses thatattending a medical school where the school or one's peersdiscouraged tobacco use was also modestly associated with greaterperceived tobacco counseling relevance and frequency; this is alsolikely an effect of attitudes and exposure.

Personal tobacco useStudents' personal tobacco use was associated with lower

counseling frequency (and with perceived frequency in senior year).We had hypothesized that this would be the case, since we and others(Frank et al., 2004a; Nelson et al., 1994) have shown that personalhealth practices are generally consistent and powerful correlates ofpatient counseling practices for physicians (and this relationship isspecifically true for smoking habits), and since we found this to be thecase for other behaviors in U.S. medical students (Frank et al., 2007a).Similarly, in a 2007 study of 661medical students in Bogota, Colombia,27% of current smokers (CS) vs. 52% of never/non-current smokers(NCS) “usually or always” counseled patients on smoking cessation(p=0.002). Likewise, 17% of CS vs. 5% of NCS disagreed with thestatement “I will be able to provide more credible and effectivecounseling if I don't use tobacco” (p=0.001).

The previously mentioned Emory study (Frank et al., 2007b) is theonly publication about which we know of the effect of discouragingmedical students' personal tobacco use on their tobacco counseling.That small (ncontrols=100 and ntreatment=106) study showed signifi-cant improvements inmalemedical students' personal tobaccouse, butonly very modest improvements in tobacco counseling (OR 1.1,p=0.6). However,we did demonstratemore significant improvementsin students' validated (Spencer et al., 2005) diet (OR=1.5, p=0.04)and exercise (OR=1.6, p=0.03) counseling practices from ourinterventions to improve their personal diet and exercise habits(Frank et al., 2007b). We therefore remain hopeful that improved

techniques to improvemedical students' tobaccohabits could still be anefficient mechanism to improve doctors' and patients' tobacco habits.

GenderWhile gender is not an especially malleable variable, policies that

have made medicine more hospitable for women may also producemore tobacco counselors. Women students were significantly morelikely to think that tobacco counseling would be highly relevant totheir future practices (both bivariately, and after controlling forintended specialty), and were somewhat more likely to currentlycounsel patients about tobacco. Prior literature has also shown thatwomen physicians generally counsel more than do men (Harvey,1996), though the data are inconsistent for tobacco counseling(Goldstein et al., 1998; O'Loughlin et al., 2007).

Strengths and limitationsOur overall and item response rates are excellent, and our sample

reflects a good and representative mix of U.S. schools regardinggeographic distribution, age, school size, NIH research ranking,private/public school balance, under-represented minorities, andgender. However, our institutions were not randomly selected, andtherefore our findings may not necessarily be fully generalizable to allU.S. medical schools, and the data are only from one country. A secondpotential limitation is that all outcome variables are self-reported.However, the surveys were anonymous and confidential, and penciland paper-based (vs. having to divulge socially undesirable practicesin-person or via phone survey). Further, substantial studies examiningthese same students' dietary (Spencer et al., 2005), school environ-ment (Frank and Kunovich-Frieze, 1995), and counseling (Frank et al.,2006) data have demonstrated both survey validity and reproduci-bility. To validate our counseling question (Spencer et al., 2005), wecompared 88 senior medical students' questionnaire responses aboutcounseling frequency (for diet, exercise, alcohol, and for cigarettesmoking) with their clinical assessments of four Standardized Patient(SP) cases with these same risk factors. For every risk factor, theproportion of SPs actually counseled was higher for those studentswho self-reported discussing that risk factor more frequently withtheir patients. Additionally, the odds of counseling an SP for any riskfactor were significantly higher (OR=1.8 to 2.8, pb0.05) whenstudents reported more frequent counseling. Further validation of ourfindings comes from the strong correlation (r=0.9, p=0.0002)between these medical students' assessments of their schools' healthpromotion environment and the assessments of their deans regardingthe school's environment (Frank et al., 2004b).

Conclusion

Attention by medical educators and health policy makers to thevariables we found to be significant could help reduce smoking in thegeneral population — by increasing the medical students' exposure totobacco counseling training, cultivating personal and clinical preven-tion-related attitudes in medical schools, encouraging primary carespecialty choices, discouraging medical students' personal tobaccouse, and promoting a medical climate that encourages women. Thesedata should give hope to those seeking new, efficient methods toreduce the prevalence of tobacco use.

Conflict of interest statementAll authors declare that there are no conflicts of interest.

References

Agency for Healthcare Quality and Research, U.S. Department of Health and HumanServices, 2003. Counseling to Prevent Tobacco Use. http://www.ahrq.gov/clinic/uspstf/uspstbac.htm

Baldwin, D.C., Hughes, P.H., Conard, S.E., Storr, C.L., Sheehan, D.V., 1991. Substance useamong senior medical students: a survey of 23 medical schools. JAMA 265,2074–2078.

239E. Frank et al. / Preventive Medicine 49 (2009) 233–239

Centers for Disease Control and Prevention, 1998. Behavioral Risk Factor SurveillanceSystem Survey Questionnaire. Atlanta, Georgia.

Cornuz, J., Zellweger, J.P., Mounoud, C., Decrey, H., Pécoud, A., Burnand, B., 1997.Smoking cessation counseling by residents in an outpatient clinic. Prev. Med. 26,292–296.

den Exter Blokland, E.A., Engels, R.C., Hale III, W.W., Meeus, W., Willemsen, M.C., 2004.Lifetime parental smoking history and cessation and early adolescent smokingbehavior. Prev. Med. 38, 359–368.

Duncan, C., Cummings, S.R., Hudes, E.S., Zahnd, E., Coates, T.J., 1992. Quitting smoking:reasons for quitting and predictors of cessation among medical patients. J. Gen.Intern. Med. 7, 398–404.

Fiore, M.C., 2000. Treating tobacco use and dependence: an introduction to the USPublic Health Service Clinical Practice Guideline. Respir. Care 45, 1196–1199.

Foster, K., Diehl, N., Shaw, D., Rogers, R., Egan, B., et al., 2002. Medical students' readinessto provide lifestyle counseling for overweight patients. Eat. Behav. 3, 1–13.

Frank, E., 2004. Physician health and patient care. JAMA 291, 637.Frank, E., Kunovich-Frieze, T., 1995. Physicians' prevention, counseling behaviors:

current status and future directions. Prev. Med. 24, 543–545.Frank, E., Winkleby, M.A., Altman, D.G., Rockhill, B., Fortmann, S.P., 1991. Predictors of

physician's smoking cessation advice. JAMA 266, 3139–3144.Frank, E., Brogan, D., Mokdad, A.H., Simoes, E., Kahn, H.S., Greenberg, R.S., 1998. Health-

related behaviors of women physicians vs other women in the United States. Arch.Intern. Med. 158, 342–348.

Frank, E., Rothenberg, R., Lewis, C., Fielding, B., 2000. Correlates of physicians'prevention-related practices: findings from the Women Physicians' Health Study.Arch. Fam. Med. 9, 359–367.

Frank, E., Galuska, D., Elon, L., Wright, E., 2004a. Personal and clinical exercise-relatedattitudes and behaviors of freshmen U.S. medical students. Res. Q. Exerc. Sport 75,112–121.

Frank, E., Hedgecock, J., Elon, L., 2004b. Personal health promotion at US medicalschools: a quantitative study and qualitative description of deans' and studentsperceptions. BMC Med. Educ. 4.

Frank, E., McLendon, L., Elon, L.K., Denniston, M., Fitzmaurice, D., Hertzberg, V.M., 2006.Medical students' self-reported typical counseling practices are similar to thoseassessed using standardized patients. MedGenMed 7.

Frank, E., Carrera, J.S., Elon, L., Hertzberg, V.S., 2007a. Predictors of US medical students'prevention counseling practices. Prev. Med. 44, 76–81.

Frank, E., Elon, L., Hertzberg, V., 2007b. A quantitative assessment of a 4-yearintervention that improved patient counseling through improving medical studenthealth. MedGenMed 9, 58.

Frank, E., Elon, L., Naimi, T., Brewer, R., 2008. Alcohol consumption and alcoholcounseling behaviors among a cohort of U.S. medical students. BMJ 337.

Gilpin, E., Pierce, J.P., Johnson, M., Bal, D., 1993. Physician advice to quit smoking: resultsfrom the 1990 California Tobacco Survey. J. Gen. Intern. Med. 8, 549–553.

Goldstein, M.G., DePue, J.D., Monroe, A.D., Lessne, C.W., Rakowski, W., et al., 1998.

A population-based survey of physician smoking cessation counseling practices.Prev. Med. 27, 720–729.

Harvey, L., 1996. Prevention advice rates of women and men physicians in primary careand other disciplines. Arch. Fam. Med. 5, 215–219.

Hosmer, D.W., Lemeshow, S., 1989. Applied Logistic Regression. John Wiley and Sons,New York, NY.

Hughes, P.H., Conard, S.E., Baldwin, D.C., Storr, C.L., Sheehan, D.V., 1991. Residentphysician substance use in the United States. JAMA 265, 2069–2073.

Jorm, A., Rodgers, B., Jacomb, P.A., Christensen, H., Henderson, S., Korten, A.E., 1999.Smoking and mental health: results from a community survey. Med. J. Aust. 170,74–77.

Mumford, E., Levy, D.T., Gitchell, J.G., Blackman, K.O., 2006. Smokeless tobacco use1992–2002: trends and measurement in the current population survey — tobaccouse supplements. Tob. Control 15, 166–171.

Nelson, D.E., Giovino, G.A., Emont, S.L., Brackbill, R., Cameron, L.L., et al., 1994. Trends incigarette smoking among US physicians and nurses. JAMA 271, 1273–1275.

O'Loughlin, J., Makni, H., Tremblay, M., Karp, I., 2007. Gender differences among generalpractitioners in smoking cessation counseling practices. Prev. Med. 45, 208–214.

Otten, R., Engels, R.C., van de Ven, M.O., Bricker, J.B., 2007. Parental smoking andadolescent smoking stages: the role of parents' current and former smoking, andfamily structure. J. Behav. Med. 30, 143–154.

Prokhorov, A., Warneke, C., de Moor, C., Emmons, K.M.M.J., Rosenblum, C., et al., 2003.Self-reported health status, health vulnerability, and smoking behavior in collegestudents: implications for intervention. Nicotine Tob. Res. 5, 545–552.

Rafferty, M., Frank, E., 1994. Office-based prevention: how can we make it happen?West. J. Med. 161, 190–191.

Rigotti, N., Lee, J.E., Wechsler, H., 2007. U.S. college students' use of tobacco products.J. Am. Med. Assoc. 284, 699–705.

Schoenborn, C.A., Adams, P.F., Barnes, P.M., Vickerie, J.L., Schiller, J.S., 2004. HealthBehaviors of Adults: United States, 1999–2001. National Center for Health Statistics,Hyattsville, MD.

Scott, C., Neighbor, W.E., Brock, D., 1992. Physicians' attitudes toward preventive careservices: a seven-year prospective cohort study. Am. J. Prev. Med. 8, 241–248.

Shah, B.V., 2006. SUDAAN. ed. Research Triangle Institute. Research Triangle Park, NCSpencer, E., Elon, L.K., Hertzberg, V.S., Stein, A.D., Frank, E., 2005. Validation of a brief

diet survey instrument among medical students. J. Am. Diet. Assoc. 105, 802–806.Strecher, V., O'Malley, M.S., Villagra, V.G., Campbell, E.E., Gonzalez, J.J., et al., 1991. Can

residents be trained to counsel patients about quitting smoking? Results from arandomized trial. J. Gen. Intern. Med. 6, 9–17.

Tomar, S.L., 2003. Trends and patterns of tobacco use in the United States. Am. J. Med.Sci. 326, 248–254.

World Health Organization, 2005. Fact Sheets: Smoking Statistics. http://www.wpro.who.int/media_centre/fact_sheets/fs_20020528.htm

World Medical Association, 2007. World Medical Association Statement on HealthHazards of Tobacco Products. http://www.wma.net/e/policy/h4.htm