Health-Related Behaviors of Women Physicians vs Other Women in the United States FREE

THE NUMBER of women physicians in the United States has grown enormously in the last several decades. In the 1960s, less than 10% of US medical school graduates were women¹; there are now more than 100000 women physicians, and women constitute 40% of current medical school classes.² However, little is known about the personal health-related behaviors or other characteristics of this increasingly important sector of medical practitioners; the Women Physicians' Health Study (WPHS) seeks to address that shortcoming.

Knowledge of women physicians' health behaviors may be especially useful in setting realistic national objectives for disease prevention in women. For example, while we know what our ideal goals might be (eg, a 0% smoking rate), we do not know what goals may be achievable realistically. Socioeconomic status (SES; educational attainment, occupational prestige, and income) has been shown to be a consistent predictor of mortality and a wide variety of risk factors and morbidities.^3- 5 Because of women physicians' high SES as well as their health knowledge, it is worth exploring whether their personal adoption of health-related behaviors might represent a goal for currently attainable health practices. To determine whether women physicians meet these expectations, we used 3 samples to compare women physicians with women of high SES and other women in the United States. The 3 samples are the 4501 respondents to WPHS and the 1316 women of high SES and the 35361 other women who responded to the 1992 US Behavioral Risk Factor Surveillance System (BRFSS) survey.

The design of WPHS has been more fully described elsewhere.⁶ The WPHS surveyed a stratified random sample of women with doctor of medicine degrees in the United States; the sampling frame is based on the Physician Masterfile of the American Medical Association, Chicago, Ill, a database intended to record all people with doctor of medicine degrees residing in the United States and its possessions. Using a sampling scheme stratified by decade of graduation from medical school, we randomly selected 2500 women from each of the last 4 decades' graduating classes (1950-1989). We oversampled older women physicians, a population that would otherwise have been sparsely represented by proportional allocation because of the recent increase in numbers of women physicians. We included active, part-time, professionally inactive, and retired physicians, aged 30 to 70 years, who were not in residency training programs in September 1993, when the sampling frame was constructed. In that month, the first of 4 mailings was sent out; each mailing contained a cover letter and a self-administered 4-page questionnaire. Responses were accepted until October 1994 (final number of respondents, 4501).

Of the potential respondents, an estimated 23% were ineligible to participate because their addresses were wrong or they were men, deceased, living out of the country, or interns or residents. Our response rate was 59% of physicians eligible to participate. We compared respondents and nonrespondents in 3 ways: we used our telephone survey (comparing our telephone-surveyed random sample of 200 nonrespondents with all the respondents to the written survey), the Physician Masterfile of the American Medical Association (comparing all respondents with all nonrespondents), and an examination of survey mailing waves (all respondents, from wave 1 through 4) to compare respondents' and nonrespondents' outcomes for a large number of key variables. From these 3 investigations, we found that nonrespondents were less likely than respondents to be certified by a specialty board. However, respondents and nonrespondents did not consistently or substantively differ on other tested measures, including age, ethnicity, marital status, number of children, alcohol consumption, fat intake, exercise, smoking status, hours worked per week, frequency of being a primary care practitioner, personal income, or percentage actively practicing medicine.

Based on these findings, we weighted the data by decade of graduation (to adjust for our stratified sampling scheme) and decade-specific response rate and board-certification status (to adjust for our identified response bias). The analysis weights (within decade) for respondents who are board certified and those who are not board certified, respectively, are 3.4 and 5.5 (1950s), 9.3 and 17.7 (1960s), 17.9 and 36.5 (1970s), and 28.3 and 63.9 (1980s). Using these weights allows us to make inference to the entire population of women physicians who graduated from medical school between 1950 and 1989.

Unless otherwise noted, all WPHS questions used in this report were taken directly from the BRFSS questionnaire to maximize comparability; it was our a priori intention to compare items from the 2 questionnaires. The BRFSS telephone surveys are conducted by state health departments in collaboration with the Centers for Disease Control and Prevention, Atlanta, Ga; a detailed technical description of survey methods is available.⁷ Each year, participating states select independent probability samples of adult residents with telephones in the noninstitutionalized, civilian population. Data used in this report are those collected from all states for which analyzed variables were available in 1992, the last year for which information about postgraduate and professional degrees was obtained. The BRFSS response rate ranged from 42.2% to 94.8% for the surveyed states, with a median response rate of 82.9%.

We analyzed 3 samples: (1) women from the BRFSS, aged 30 to 70 years and of high SES (n=1316), high SES defined as having both a household annual income higher than $50000 and a postgraduate or professional degree; (2) all other women from the BRFSS aged 30 to 70 years (n=35361); and (3) all women from WPHS (n=4501). All analyses are weighted, using the sampling weights from WPHS and BRFSS. With the exception of Table 1 (basic demographic description), all estimated prevalences and means for all comparison populations are age-adjusted to the population distribution of women physicians in the United States.

According to the standard protocol of the Centers for Disease Control and Prevention,⁸ to accommodate the sample design, we used a software package (SUDAAN, Research Triangle Institute, Research Triangle Park, NC) that permits correct analysis of sample survey data that are weighted, clustered, and stratified.⁹ To compare women physicians with the women of high and not high SES in the United States, we concatenated the 4 WPHS strata with the 62 BRFSS strata and then analyzed the resulting data set with the 3 subpopulations of interest. The BRFSS and WPHS data sets are both sample survey data (ie, based on probability samples). The data are weighted, which is typical in sample survey data. All analyses in this article are done to make inference to the 3 larger populations: the total population of women physicians graduating from medical school between 1950 and 1989, women of high SES, and other women in the United States. Therefore, we are not simply describing our sample but using the sample to make inference to the larger populations; these are standard analytic procedures for sample survey data. Due to multiple testing and large sample sizes, significance levels are shown only if P<.01.

Women physicians as a group were younger and included a higher percentage of Asian Americans than either BRFSS population (Table 1). Both physicians and women of high SES had smaller percentages of African Americans than women in the general BRFSS population. For the purpose of this report, women in the general BRFSS population refers to women who are not of a high SES in column 1 of Table 1, Table 2, and Table 3. Women physicians were less likely to be married than women in the high SES group, but they were more likely to be married than women in the general BRFSS population. Nearly all women physicians reported incomes in the highest bracket of household income of the BRFSS population (>$50000).

Table 2 shows that women physicians were more likely to have never smoked than other women. If physicians were ever smokers, they were likely to have started at a slightly, although not significantly, older age than other women. Physicians were more likely to have quit smoking than women in the other 2 groups. As calculations based on Table 2 reveal, the great majority (83%) of physicians who had ever smoked stopped smoking; a somewhat smaller proportion (78%) of women of high SES and less than half (47%) of women in the general BRFSS population who had ever smoked were ex-smokers. Physicians were also likely to have quit for more years than women in the general population. Physicians who were current smokers report consuming fewer cigarettes per day and are more likely to report having had at least 1 day in which they abstained from cigarette smoking in the last year. Based on the numbers shown in Table 2, women physicians were approximately one seventh as likely to smoke, half as likely to have ever been smokers, and nearly twice as likely to have quit smoking than women in the general population.

Regarding reported alcohol drinking, only a quarter of physicians had not consumed alcohol in the previous month, a significantly lower rate of abstinence than other women. Of those physicians who consumed alcohol in the past month, they reported drinking alcohol an average of twice a week, drank fewer drinks per episode of drinking than did other women, and almost never drank more than 4 drinks per episode.

Physicians ate somewhat more fruits and vegetables and less fat than did less-advantaged women and somewhat fewer fruits and vegetables and more fat than other women of high SES. Physicians are also highly likely to consistently wear seat belts in a car; nearly all the women of high SES in the BRFSS population also consistently wear seat belts.

As shown in Table 3, women physicians underwent screening or testing more recently than women in the general BRFSS population for all the reported screening and testing parameters studied (with the exception of blood pressure, for which the results were equivalent). Compared with other women of high SES, physicians underwent cholesterol measurements more recently; performed equivalently to them in blood stool testing, proctoscopic and sigmoidoscopic examination, breast examination by a clinician in those aged younger than 40 years, and mammography; and underwent blood pressure measurements, breast examination by a clinician in those aged 40 years or older, and Papanicolaou testing less recently. As one way of investigating whether some physicians might be making evidenced-based decisions about deferring screening, we compared the rates for Papanicolaou testing for physicians who are part of a couple or sexually inactive (lower risk) with the rates for physicians who were single and sexually active (higher risk); there was no significant difference in the recency of undergoing a Papanicolaou test (P=.5).

Finally, we found that these physicians exceeded national goals¹⁰ for the year 2000 for all behaviors and screening practices for which goals have been stated (not shown); neither BRFSS population did so (although the women of high SES met more goals than the general population of women from the BRFSS). The behaviors for which these goals were examinable were prevalences of current tobacco use, abstinence among smokers from smoking for at least 1 day in the past year, seat belt use, blood stool testing, proctoscopic and sigmoidoscopic examination, cholesterol measurement, blood pressure measurement, Papanicolaou testing, breast examination by a clinician in those younger than 40 years, and mammography, as well as mean age at onset of smoking (for those who were ever cigarette smokers), and mean alcohol use.

This is the first report comparing characteristics of a large sample of women physicians with other women in the United States. We found that women physicians generally reported healthy habits, exceeded national health behavior goals in all examinable cases, consistently had better health-related behaviors than women in the general population, and in many cases outperformed other women of high SES. This is noteworthy because high SES is usually a marker of healthy behaviors and good health status.^3- 5 Physicians' healthier habits were particularly pronounced in straightforward primary prevention activities, such as avoidance of cigarette smoking or heavy alcohol use.

Women physicians are more likely to be Asian American than women in either the general (not high SES) or the high SES samples. However, like other women of high SES, they are less likely to be African American. Women physicians were more likely to be married or single or never married and less likely to be separated or divorced than women in the general population. The American Medical Association has also found that physicians of both sexes are more likely to be married and less likely to be divorced than the US population as a whole.¹ We believe that the higher prevalence of being married in the high SES group is an artifact of our classification protocol, as higher household income is likely associated with being married. Some of these demographic characteristics may contribute to women physicians having health behaviors different from those of the general population; for example, the influences of ethnicity will be explored in subsequent analyses.

Reported cigarette smoking was the area in which women physicians exhibited the most exemplary behaviors. For example, women physicians were one seventh as likely to smoke, half as likely to have ever been smokers, and, if they ever smoked, nearly twice as likely to have quit smoking as women in the general population. Of the few women physicians who still smoked, they consumed fewer cigarettes per day and were more likely to have had at least 1 day in which they abstained from smoking in the last year than other women. Similar numbers have been reported for other recent studies^11- 16 of physicians' smoking habits, although all these studies but one¹⁶ had a small sample of physicians of either sex (N<400), and none reported the age of smoking onset, cigarettes consumed per day, or number of years elapsed since quitting.

ALTHOUGH ONLY one quarter of physicians reported complete abstinence in the past month, physicians who consumed alcohol reported drinking only twice a week on average, consuming less per episode of drinking than other women drinkers, and reported almost never drinking large amounts. The difference between physician and lay alcohol abstinence rates may reflect physicians' assessments of current scientific opinion on alcohol consumption (as well as their personal recreational preferences), since the contemporary medical literature^17- 18 generally supports the healthfulness of low-to-moderate alcohol consumption and the harmfulness of high consumption. A few small studies (N<300)^11,19- 20 and 1 large study¹⁶ of physicians have reported lower abstinence rates (<10% total alcohol abstinence) than we report. Hughes and colleagues¹⁶ found that only 6% of physicians abstained totally from alcohol, although 23% had not consumed alcohol in the past month. This suggests that the lower abstinence rates of our study, compared with those found for other surveys of physicians, may be a function of question structures, since neither BRFSS nor WPHS queried whether individuals ever drank alcohol, only whether they drank in the past month. Our rates could also reflect a temporal trend in physician drinking, since the other physician studies were conducted in the 1970s and 1980s. Alternatively, our rates may be a function of the sex of this physician population, since most studies of the general population, as well as the large study of drinking habits of physicians of both sexes, have found that women are more likely than men to abstain totally from alcohol use.¹⁶ Women physicians' diets were, in this limited examination, somewhat better than those of the general population of women and somewhat worse than those of other women of high SES. Subsequent analyses will examine these findings more thoroughly to determine in which dietary constituents women physicians fare particularly well or poorly, as well as the effects of such variables as ethnicity, marital status, specialty, and workload on diet. Both women physicians and other women of high SES consistently wear seat belts in a car.

For the screening and testing parameters examined, women physicians consistently had undergone screening and testing more recently than women in the general BRFSS population, although in general the results were only equivalent between women physicians and other women of high SES. It is interesting to speculate about why women physicians with presumably good access to clinical services, although performing well, performed worse than other women of high SES in the recency of undergoing Papanicolaou testing, blood pressure measurements, or breast examination by a clinician in those aged 40 years or younger. Reasons may include disagreement with screening recommendations (particularly if these individuals had previous results that were normal or were otherwise at low risk), perception of self-care as being sufficient, denial of vulnerability, embarrassment about using screening services, or scheduling difficulties. Few of these screening prevalences have been well-examined in other physician populations, although where they have been studied in small cohorts, results support our findings. A small study¹⁵ (N=119) examining physicians' personal habits for testing cholesterol levels found an 87% (compared with our 86%) testing prevalence in the previous 5 years. A study²¹ including 94 women physicians found that 84% (compared with our 86%) had undergone a Papanicolaou test within the last 3 years, 74% (compared with our 72%) of physicians aged 50 years or older (of both sexes) had ever undergone fecal occult blood screening, and 58% (compared with our 43%) had ever undergone sigmoidoscopy.

This study has some limitations. For example, reliability and validity have not been extensively tested in either questionnaire, and questions in the 2 comparison surveys were not asked in the same order. The questions were also of different types: WPHS used a confidential, self-administered questionnaire, and BRFSS used a telephone interview. This would tend to bias comparisons toward reporting healthier behaviors for the population surveyed by telephone (nonphysicians) because they would need to report undesirable behaviors directly to another person. Alternatively, despite the confidentiality of the written WPHS survey, women physicians may have felt more compelled than other women to provide what they perceived to be medically preferred responses. However, there are few practical alternatives to surveys for estimating behavior prevalence in populations. Also, BRFSS data were collected in 1992 (1992 data were used because this was the last year for which information on postgraduate and professional degrees was obtainable to define our high SES cohort) and WPHS data were collected during 1993 and 1994; there may have been some secular trends in our reported behaviors. There are also other limitations to BRFSS data, including variations in state participation and inclusion of individual items within survey years; these have been more extensively reported elsewhere.⁷ Another limitation is that for a few of these variables, such as the reported rates of blood stool testing, women physicians' higher rates may be an artifact of recognizing or remembering the performance of a test that other women may not fully comprehend or recall. Finally, although we learned much about our nonresponders, the WPHS response rate was 59% of eligible women, and the BRFSS response rate ranged from 42% to 95%, and there may be some response biases that were not accounted for by our weighting strategy.

Subsequent analyses of the WPHS data will examine other characteristics of women physicians, determine ways in which their characteristics and behaviors vary by such parameters as specialty and age, and examine more comprehensively the variables in which their behaviors did not surpass those of other women of high SES. Nonetheless, regardless of the results of subsequent analyses and stratifications, our results show that women physicians in the United States generally follow recommended health behavior guidelines, given the advantages of high income, substantial general and health-specific education, a prestigious occupation, and a social milieu that discourages egregiously unhealthy behaviors and despite any disadvantages that may accrue from their profession. While some of women physicians' more exemplary health habits may be attributable to their being of an even higher SES than the women of high SES from the BRFSS, some habits may also be due to their training and professional social environment. The limited studies addressing this question in men physicians have also demonstrated a high standard of physicians demonstrating healthy behaviors; whether they were compared with the general male population (N=1193; men physicians),²² or attorneys (N=289 physicians, 92% men),²⁰ men physicians in the United States have been shown to more frequently adopt healthy behaviors²⁰ and have lower mortality rates.²²

What are the implications of these findings? It is important to know physicians' health-related choices since physicians are social models and their personal health-related choices may influence their practices in counseling patients.^{15,21,23- 25} Additionally, women physicians may provide estimates of the achievability of goals for women in the general population. We found that, unlike other women (even those of high SES), women physicians exceeded the goals stated in Healthy People 2000: National Health Promotion and Disease Prevention Objectives¹⁰ in all examined behaviors. While there are some areas of health that we have not tested, these data suggest that women physicians may provide an imperfect but potentially useful set of standards for public health officials. Having a standard against which we can compare the behaviors of other populations could facilitate planning and interpretation of interventions to promote health for women. Finally, these data suggest that improving the SES (educational, income, and occupational) and/or the health-related knowledge of women in the general population may improve their health-related behaviors.

Accepted for publication June 3, 1997.

This study was funded by the American Medical Association's Education and Research Foundation, Chicago, Ill; the American Heart Association, Dallas, Tex; an institutional National Research Service Award (No. 5T32-HL-07034) from the National Institutes of Health (National Heart, Lung, and Blood Institute), Bethesda, Md; and the Emory Medical Care Foundation, Atlanta, Ga.

Corresponding author: Erica Frank, MD, MPH, The Women Physicians' Health Study, the Department of Family and Preventive Medicine, and the Department of Medicine, Emory University School of Medicine, 69 Butler St, Atlanta, GA 30303-3219 (e-mail: efrank@fpm.eushc.org).