Personal and Clinical Skin Cancer Prevention Practices of US Women Physicians FREE

THE MOST frequently diagnosed cancer in the United States, skin cancer, is largely preventable. Even so, of the 3 most common types, basal cell and squamous cell cancer account for a million new cancers per year, while cutaneous malignant melanoma accounted for approximately 44,200 new cases in 1999. Melanoma is associated with the greatest mortality of all skin cancers, with 7300 estimated deaths in 1999.¹ Recently, primary care physicians have been encouraged to educate their patients to protect their skin from sun damage^2- 3 or to screen patients at risk for skin cancer.^4- 5 Little is known, however, about predictors of compliance with these strategies. Having more knowledge about the determinants of patient counseling and screening could guide interventions on physician characteristics that are amenable to change and potentially improve the promotion of physicians' prevention practices.⁶

Several studies of physicians' practices^6- 12 have identified characteristics of clinicians who tend to be more avid preventionists, but prevention practices specifically related to skin cancer have been examined^13- 18 in limited populations (eg, single specialties, small geographic regions). Furthermore, these studies have often lacked adequate control for confounding factors and have not extensively addressed issues surrounding both skin cancer prevention and screening. In addition, they have not examined physicians' personal practices relative to prevention and screening for skin cancer, even though these factors may be important predictors of counseling and screening practices.^6- 7,9- 10 In this study, we explored the effect of such factors on skin cancer counseling and screening variables among 4501 respondents to the Women Physicians' Health Study (WPHS), a questionnaire-based study of a representative sample of US women physicians.

The design and methods of the WPHS and the basic demographic, personal, and professional characteristics of respondents^19- 21 have been described elsewhere. In brief, the study used the American Medical Association's database of all US physicians with an MD degree to randomly select 2500 physicians from each of the past 4 decades' graduating classes (1950-1989), including active, part-time, professionally inactive, and retired physicians aged 30 to 70 years who were not in residency training programs. Between September 1993 and October 1994, 4 mailings were sent out (final n=4501). Based on our nonresponse findings, we weighted the data by decade of graduation (to adjust for our stratified sampling scheme) and by decade-specific response rate and board certification status (to adjust for our identified response bias), allowing us to make inferences to the entire population of women physicians graduating from medical school between 1950 and 1989.

Our primary outcome variables were responses to 3 questions (on counseling and screening, personal sunscreen use, and personal skin examination), each based on national recommendations at the time of the survey.²² Response options to the question "Considering your typical patients: How often do you usually discuss or perform screening for skin cancer/sunscreen use?" were as follows: "every visit," "every ≤1 year," "every >1-2 years," "every >2-3 years," "every >3-5 years," "only at the initial visit," "only if clinically indicated," or "never." We initially divided the responses into 3 categories: frequent ("every visit" or "≤1 year"), less frequent, and never. For the question "How often do you wear sunscreen when spending ≥1 hour outdoors?" the response options were "always," "nearly always," "sometimes," "seldom," "never," and "never outdoors for ≥1 hour." For the question "How long has it been since your most recent skin exam for cancer (by clinician)?" the response options included "≤1 year," ">1-2 years," ">2-3 years," ">3-5 years," ">5 years," "never done," and "don't know."

Analysis of the counseling or screening questions was limited to physicians working at least 5 clinical hours per week who were not radiologists or pathologists. In addition to examining practices by region (North, West, South, and East) and by melanoma mortality (high, medium, low), to capture ultraviolet (UV) exposure and latitude, we used the UV index, a composite measure of the expected risk of overexposure to the sun.²³ The UV index uses a scale of 0 (low risk) to 10+ (very high risk). We assigned to each physician the average UV index of a major city in her state to create a dichotomous variable (low, 1-3; high, 4-10+). Indices were obtained from the National Weather Service (C. Long, MS, National Weather Service, Silver Spring, Md, written communication, July 1998). Although there are limited data on consumer or physician awareness of the UV index, we chose to use this measure to present our data because it captures UV intensity better than region. We considered family medicine, general practice, general internal medicine, pediatrics, obstetrics-gynecology, and public health (all self-designated) to be primary care specialties.

We used SUDAAN²⁴ to perform χ², F, and t tests to determine whether counseling or screening was related to certain personal and professional characteristics. To identify predictors of counseling among primary care physicians, we first used polychotomous logistic regression in SUDAAN to model the 3 levels of counseling (frequent, less frequent, and never) as a function of several personal and professional characteristics. Upon discovering similar predictors for frequent and less frequent counseling and considering that there is no consensus on appropriate counseling or screening intervals for skin cancer, we used logistic regression to compare the dichotomous outcome of whether respondents ever provide counseling for skin cancer/sunscreen use. For univariate tests, P<.05 was considered significant. To determine which variables should remain in the final logistic regression model, we used a criterion of P<.10 on the Wald F test. Variables were selected using a modified version of backward selection and a 20% "change-in-estimate" confounder selection strategy.²⁵ The Hosmer-Lemeshow goodness-of-fit test was used to assess goodness of fit.²⁶

Basic demographic and professional characteristics of women physicians have been outlined previously.^19,21 Of the 3032 nondermatologists who were active clinicians, 27% were frequent counselors for skin cancer screening and sunscreen use (Table 1). Almost half (49%) were less frequent counselors, and 24% never counseled. Physicians who reported good health were more likely to be either frequent or less frequent counselors than those who reported fair or poor personal health. Both higher frequency of sunscreen use and receiving a clinical skin examination in the previous 2 years were significantly associated with frequent counseling or screening. No significant differences in counseling or screening were found by age, ethnicity, UV index, family or spousal history of skin cancer, or personal history of skin cancer.

Primary care physicians were more likely than other nondermatologists to be frequent counselors (37% vs 14%). Among primary care physicians, obstetrician-gynecologists were least likely to be frequent counselors (16%) and most likely to never counsel (25%). Pediatricians reported counseling or screening frequently (43%), similar to family medicine practitioners (44%) and internists (36%), but pediatricians also reported a higher frequency of never counseling or screening (16%) than either of those 2 groups (3% and 6%, respectively). By practice type, hospital and medical school practice had the lowest percentages of frequent counselors, and approximately one third of physicians in both subgroups reported never counseling or screening. By site (location), urban practice had the highest percentage of never counseling or screening. High self-confidence in counseling or screening (reported by 38%), a perception that counseling or screening was highly relevant to one's practice (reported by 23%), and report of extensive training in counseling or screening (reported by 14%) were all significantly related to being a frequent counselor. No significant (P<.05) differences in counseling or screening were seen by UV index, melanoma mortality, work hours per week, work control, career satisfaction, board certification, or amount of continuing medical education (data not shown).

Nearly half of the physicians wore sunscreen always (19%) or nearly always (29%) when spending 1 hour or more outdoors (Table 2). More than one fourth reported using sunscreen sometimes (27%) and more than one fifth reported rarely or never using sunscreen (22%). A personal history of skin cancer, a family or spousal history of skin cancer, receiving a recent clinical skin examination, and age 30 to 49 years were associated with more frequent sunscreen use. The age association held when the analysis was restricted to those without personal or family histories of skin cancer and excluded those who never spent 1 hour or more outdoors (P<.001, data not shown).

More than half of white women reported always (22%) or nearly always (33%) using sunscreen; Hispanic women had the next highest combined rate (40%). Black women had the highest percentage reporting seldom or never using sunscreen (58%), and Asian women had the highest proportion reporting not being outdoors for more than 1 hour (12%). Living in a region with a high UV index was not significantly (P=.08) associated with increased sunscreen use. Nearly half of family medicine practitioners, pediatricians, obstetrician-gynecologists, and public health physicians but less than half of the general internists or general practitioners reported always or nearly always wearing sunscreen. Finally, high self-confidence, high relevance, and extensive training were associated (P<.05) with more frequent sunscreen use (data not shown).

One fourth of the physicians had received a skin examination performed by a clinician in the previous 2 years (Table 3). One tenth had received a skin examination more than 2 years previously, but almost two thirds (65%) had never received one. Physicians aged 30 to 49 years (in general and when restricted to those without personal or family histories of skin cancer) were less likely to have been examined in the previous 2 years than were those who were 50 to 70 years old (P<.001 for both comparisons, data not shown). Women with a personal or family history of skin cancer were more likely (even when the analysis was stratified by age) to have been examined in the previous 2 years (data not shown). White physicians were more likely to have had a skin examination than nonwhite physicians; no differences were seen between primary care physicians and non–primary care physicians or among primary care specialties. Again, strong self-confidence, high relevance, and extensive training about skin cancer prevention were associated (P<.05) with more recent clinical skin examination (data not shown).

Dermatologists (n=95) were significantly more likely than other physicians to counsel or screen patients for skin cancer. (Please note that the following sample numbers vary according to the item response rate, and the percentages are weighted.) Of those who worked at least 5 hours per week (n=89), almost two thirds (51/82 [65%]) counseled at every visit (vs 4.8% of other physicians), and no dermatologists (vs 23.6% of other physicians) reported that they had never counseled (data not shown). Eighty-two (87%) of 94 dermatologists always or nearly always used sunscreen, and about two thirds (62/86 [68%]) reported receiving a clinical skin examination in the previous 2 years. Other specialists counseled significantly less than primary care physicians (14% vs 37%), but there was little difference between their personal skin cancer prevention practices.

In our multivariate model, significant predictors of ever having provided skin cancer counseling or screening among primary care practitioners included using a sunscreen sometimes (but not more frequently), being highly self-confident about skin cancer prevention, finding the counseling or screening highly relevant to one's practice, and working in solo or duo/group settings (Table 4). A negative predictor was being a pediatrician or an obstetrician-gynecologist. We could not examine personal history of skin cancer in this model because all primary care practitioners who had such a history reported counseling or screening their patients for skin cancer prevention.

As physicians of first contact and as gatekeepers to the medical care system, primary care physicians constitute the most logical group of clinicians to practice skin cancer counseling or screening.⁶ In this group, we found that sunscreen use "sometimes" was the only personal predictor of counseling or screening among primary care providers when we used multivariate analysis. Surprisingly, "always" or "nearly always" using sunscreen was not associated with providing counseling or screening. This may reflect the belief of many physicians that the level of sunscreen use and risk of skin cancer are not related. Indeed, until recently,²⁷ there has been a lack of evidence that using sunscreen decreases or increases the risk of skin cancer. Establishing a linkage is difficult, however; increased sunscreen use coupled with extended sun exposure, the often inappropriate application of sunscreen, and unreliable sun-exposure histories makes it difficult to sort out the effects of sunscreen alone.²⁸

Although we did not find as strong a tie between personal habits and counseling or screening practice among primary care physicians in the multivariate analysis, we found an association with sunscreen use among all physicians in bivariate analysis. This is consistent with prior analyses from the WPHS showing that women physicians were more likely to report counseling or screening patients regarding cholesterol, exercise, alcohol use, smoking cessation, provision of influenza vaccine, breast and cancer prevention and detection, and hormone replacement therapy use if they personally practiced what they recommended.²⁹ In addition, smaller studies have shown that physicians who are regular exercisers are more likely to counsel their patients on exercise habits,^30- 31 nonsmokers to counsel on smoking,^30,32- 33 seat belt users to counsel on seat belt use,³⁰ and consumers of low-fat food and those previously screened for cholesterol to counsel on cholesterol.³⁴

Not surprisingly, self-confidence about counseling or screening and perceived relevance to the physician's practice were strong independent predictors of counseling or screening among primary care physicians. Prior research has shown that many primary care physicians believe that skin cancer counseling or screening has a low priority, that primary care physicians have insufficient knowledge to appropriately identify a potentially malignant lesion, and that they are discouraged by the low yield of examinations.^{6,14,18,35- 36}

Our finding that primary care physicians were less likely to counsel their patients on skin cancer and sunscreen use if those physicians were from an academic or hospital setting is consistent with a recent survey that found pediatricians in health maintenance organizations and private settings more likely to counsel about skin cancer than pediatricians in academic institutions or community health centers.¹⁵ Perhaps office and managed care settings are more likely to have guidelines addressing skin cancer prevention. Alternatively, hospital and academic center practice may involve dealing with more tertiary or acute medical problems; in such settings, primary or secondary skin cancer prevention may be limited by time and priorities. This is of concern, as physicians who work in hospitals and medical schools often serve as role models to medical students and residents, and perhaps they are in a better position to encourage counseling or screening practices.

In this survey, we found that 88% of primary care physicians reported counseling or screening for skin cancer for their "typical patient," a far greater proportion than the 30% of primary care physicians who reported bringing up or cautioning "most" patients about skin cancer in a 1989 American Cancer Society survey.³⁷ In that survey, there was little difference between the primary care specialties surveyed (rates were 32% for general and family medicine practitioners, 27% for internists, and 27% for obstetrician gynecologists). In contrast, in our survey both obstetrician-gynecologists and pediatricians had lower odds of ever counseling or screening for skin cancer than did internists. The American College of Obstetricians and Gynecologists has issued recommendations about skin cancer screening as part of an overall cancer evaluation and about counseling or screening both before and after the survey^38- 39; in addition, the American Academy of Pediatrics has emphasized sun-protective habits in its brochures and more recently in its policy statement.⁴⁰ There is little relevant literature on skin cancer prevention practices of obstetrician-gynecologists; a few studies have reported that many pediatricians counseled their patients on safe sun exposure practices despite lack of formal training on counseling.^15,17 Although our bivariate analysis found a high prevalence of frequent counseling among pediatricians, on par with that of internists and family medicine practitioners, perhaps the lower odds of ever counseling patients about skin cancer that we found among pediatricians in our multivariate analysis reflect the demands of prevention and treatment for children and adolescents, which are perceived as being more pressing than prevention of skin cancer, a disease with a long latency period.⁴¹ Nonetheless, most sun exposure occurs before the age of 18 years.⁴² During childhood and adolescence, sunburn, a signal of intense sun exposure, has been associated with an increased risk of melanoma,⁴³ and sun avoidance during childhood is thought to have more of an impact on melanoma risk than sun avoidance during adulthood.⁴⁴ Counseling children, adolescents, and parents (including expectant and new parents) may be the best opportunity for skin cancer prevention.

It was not surprising that being a dermatologist was powerfully related to skin cancer counseling or screening practices, given the interests, expertise, and risk of liability of dermatologists if they fail to detect a lesion. In a previous article by Frank et al²⁹ that examined 13 other prevention-related outcomes, dermatologists were generally less avid preventionists than were other physicians (even less avid than other non–primary care specialists).

That physician counseling or screening did not vary by the UV index, region, or melanoma mortality is disturbing. One would expect and hope that physicians who live in areas with a high UV index or areas with a high melanoma mortality rate would particularly encourage sunscreen use and other sun-safe behaviors, since this disease is strongly linked to UV exposure and latitude.⁴⁵ It was also surprising that family and spousal history did not predict counseling or screening among primary care providers, but personal history, which could not be examined in the logistic model, did significantly affect counseling or screening.

Although our study indicated that 7 of 8 female primary care physicians counseled or screened a typical patient about skin cancer, the true proportion of all physicians who do so may be considerably less or more. First, these are self-reported data⁴⁶; we did not attempt to verify these data by reviewing medical records. Second, women physicians are usually more likely to offer preventive services than are men.⁶ Third, although this data set is from 1994, there is no evidence to suggest that a significant change in medical practice has occurred between then and the time this report was written. We should also consider that recent conflicting or negative recommendations (Table 5) regarding routine counseling on sun avoidance and sunscreen use, the utility of routine skin examinations, and both the expertise and type of provider needed to detect skin cancer early may have reduced the rate of current skin cancer preventive practices among many primary care physicians since our survey was undertaken. In 1989, the US Preventive Services Task Force (USPSTF) recommended routine skin cancer screening for persons at risk and counseling patients with increased outdoor exposure to use sunscreen and other sun-protective measures²²; in contrast, in 1996, the USPSTF cited insufficient evidence to recommend for or against routine screening for skin cancer by primary care providers and routinely counseling about sunscreen use. The USPSTF did recommend that primary care providers counsel adults and children at increased risk to avoid excess sun exposure, although the effectiveness of this counseling (ie, impact on actual patient behavior) has not been proven.³

In light of the mixed messages primary care physicians have received regarding the expertise needed for early skin cancer detection, a study that differentiated primary (counseling) vs secondary (screening/early detection) prevention in this group might be of special interest. By asking how often the physician discussed sunscreen use or performed screening, we obscured differences among physicians who only counseled, only screened, or both counseled and screened, and we could be reporting possibly inflated practices.

Other researchers might consider investigating the predictors of physician counseling for other primary prevention strategies that we did not analyze, including avoiding the midday sun, seeking shade, and wearing protective clothing. These approaches are frequently the key parts of a prevention strategy, but they are often seen as more difficult to change (and may be less frequently recommended). We also did not ask about characteristics of the physician's patient population to examine whether skin cancer counseling or screening varied by important patient characteristics (eg, race, skin type, work setting). From a previous study, we know that physicians are more likely to discuss skin cancer with patients who show signs of sun-damaged skin.³⁷

Our finding that 83% of white nondermatologists used sunscreen at least sometimes when outdoors is modestly encouraging. Hall and colleagues⁵⁵ reported that 66% of white women in the United States reported being "very" or "somewhat" likely to use sunscreen or sunblock when outdoors; this figure might be cautiously compared with our finding of 83%. We also found that white physicians were most likely and black physicians least likely to use sunscreen, which is comparable to the findings of previous studies showing that rates of sun-protective behaviors vary by the actual and perceived risk of skin cancer⁵⁶ and are lower among blacks.⁵⁷ Although we could not make a direct national comparison, the fact that 61% of white and at least 75% of nonwhite physicians had never had a skin examination is not surprising; the statistically significant difference between whites and nonwhites again reflects the lower perceived and actual risk among nonwhite persons. We did not obtain enough information from our respondents to determine the sun sensitivity of skin and thus used race/ethnicity as a proxy for risk (although increased UV exposure increases the risk of skin cancer among persons of all skin types).⁵⁸

In conclusion, by obtaining a nationally representative sample of specialties, we were able to comment on skin cancer prevention practices among dermatologists, primary care physicians, and other nondermatologists, all of whom were women. The results of this survey should help provide useful information on general trends and important predictors for skin cancer prevention among public and private health care providers nationally. Although there is not a consensus across medical groups on the efficacy of the early detection of skin cancer, most medical groups agree that counseling patients about limiting their UV exposure is important. Thus, education efforts should target primary care physicians who are less likely to ever counsel patients about skin cancer and sun exposure; other efforts should explore the reasons behind a low prevalence of counseling. Finally, further research needs to examine the effectiveness of physician counseling relative to other community or workplace interventions to limit UV exposure.

Accepted for publication September 20, 1999.

This study was supported in part by the Education and Research Foundation, American Medical Association, Chicago, Ill; by grant 1RO3CA7143401A2 from the National Cancer Institute, Bethesda, Md; by the Emory Medical Care Foundation, Atlanta, Ga; and by the Ulrich and Ruth Frank Foundation for International Health, Newton, Pa.

Corresponding author: Mona Saraiya, MD, MPH, Epidemiology and Health Services Research Branch, Division of Cancer Prevention and Control, Centers for Disease Control and Prevention, MS K-55, 4770 Buford Hwy, Atlanta, GA 30341 (e-mail: yzs2@cdc.gov).