Wee CC, McCarthy EP, Davis RB, Phillips RS. Physician Counseling About Exercise. JAMA. 1999;282(16):1583-1588. doi:10.1001/jama.282.16.1583
Author Affiliations: Division of General Medicine and Primary Care, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, Mass.
The Patient-Physician Relationship Section Editor:
Richard M. Glass, MD, Interim Coeditor, JAMA.
Context The increase in sedentary lifestyle may contribute to the rise in obesity
nationally. Although guidelines suggest that physicians counsel all patients
about exercise, physicians counsel only a minority of their patients. Whether
patient factors influence physician counseling is not well established.
Objectives To examine and to identify factors associated with exercise counseling
by US physicians.
Design and Setting National population-based supplemental (Year 2000) survey to the 1995
National Health Interview Survey.
Participants Of the 17,317 respondents to the Year 2000 supplemental survey, 9711
adults had seen a physician in the previous year, and 9299 responded when
asked about physician counseling on exercise.
Main Outcome Measure Physician counseling to begin or to continue to exercise.
Results Of 9299 respondents, 34% reported being counseled about exercise at
their last visit. After adjustment for other sociodemographic and clinical
factors, women were slightly more likely to be counseled, with an adjusted
odds ratio (AOR) of 1.15 (95% confidence interval [CI], 1.02-1.29). Physicians
counseled older patients (>30 years) more often than younger patients; those
aged 40 to 49 years were counseled most often (AOR, 1.71 [95% CI, 1.34-2.20]).
Patients with incomes above $50,000, those with higher levels of physical
activity, college graduates, and patients who were overweight to obese (body
mass index: 25 to ≥30 kg/m2) were more likely to be counseled,
as were patients with cardiac disease (AOR, 1.81 [95% CI, 1.52-2.14]) and
diabetes (AOR, 1.87 [95% CI, 1.46-2.38]). Counseling did not vary by physician
specialty or patient race.
Conclusion The rate of physician counseling about exercise is low nationally. Physicians
appear to counsel as secondary prevention and are less likely to counsel patients
at risk for obesity. The failure to counsel younger, disease-free adults and
those from lower socioeconomic groups may represent important missed opportunities
for primary prevention.
During the last decade, the prevalence of obesity has increased substantially.1- 3 Part of this rise has
been attributed to increasingly sedentary lifestyles. Recent data suggest
that most residents of the United States do not participate in regular physical
activity at recommended levels.4
In addition to reducing obesity, regular physical activity has been
shown to reduce morbidity and mortality associated with coronary artery disease,
hypertension, diabetes, and osteoporosis.5- 11
Because there is strong evidence that exercise is beneficial, several organizations,
including the US Preventive Services Task Force, recommend that physicians
advise all patients seen in a primary care setting to increase physical activity,
despite the limited available data on the effectiveness of counseling about
However, physician surveys suggest that physicians generally counsel only
a minority of patients.19- 22
Physicians frequently cite time limitations as a reason for not counseling.22 Furthermore, physician specialty, personal fitness,
and perceived success at advising patients are also correlated with physician
The influence of patient sociodemographic and clinical characteristics
on physician counseling practice is less clear. Failure to communicate effectively
about exercise, particularly with women, racial and ethnic minorities, and
members of lower socioeconomic groups who are at high risk for weight gain
and obesity, may be contributing to the increase in obesity in these groups.2,3,23
We used data from a nationally representative sample collected as part
of the 1995 National Health Interview Survey (NHIS) to examine physician counseling
about exercise and to identify factors associated with counseling.
The NHIS is a nationwide, in-person household survey conducted by the
US Census Bureau for the National Center for Health Statistics.24
The NHIS uses a multistage probability design to permit continuous sampling
of the US civilian noninstitutionalized population. The overall response rate
in 1995 was 94%. Approximately 102,000 persons (including children) from approximately
39,000 households responded to the core survey, which elicited information
on sociodemographic factors including occupation, insurance coverage, basic
health status, visits or contacts with health care providers, days hospitalized,
height, and weight. Respondents also were asked about functional limitations
and whether they had a usual place for health care, a usual health care practitioner,
and the clinician's specialty.
In addition, a supplemental survey ("Year 2000 Supplement") was administered
to 1 randomly selected adult, aged 18 years or older, from one half of the
responding households (n=17,317 respondents). Respondents were queried about
6 common chronic medical conditions (diabetes; asthma, emphysema, chronic
bronchitis, or tuberculosis; chronic kidney disease; chronic liver disease
or cirrhosis; cancer; and chronic cardiac-related disease). They were also
asked about tobacco use, attempts to lose weight, and their participation
in a list of leisure or sport activities in the 2 weeks prior to the interview
and the frequency of participation. Respondents who reported having had a
medical check-up (any visit interpreted by the respondent to be a check-up)
within the previous year (n=9711) were also asked, "During your last (medical)
check-up, did the doctor recommend that you begin or continue to do any type
of exercise or physical activity?"
Respondents to the Year 2000 supplement who reported a medical check-up
within the previous year and who were asked about physician counseling about
exercise were eligible for this study sample. We considered those who reported
that their physician advised them either to start or to continue exercise
to have received counseling.
Using information available in NHIS, we defined factors hypothesized
to influence physician counseling about exercise. These included patient sociodemographic
factors such as age, sex, race, marital status, education, income, insurance
type, and region of the country. We also considered clinical factors, such
as comorbid illness, overall illness burden (self-reported health status,
number of hospital days in previous year, visits or contacts with a health
care provider, difficulty walking), tobacco use, and body mass index (BMI)
(defined as weight in kilograms divided by square of height in meters). We
considered the specialty of the patient's usual health care provider (general
internist or family practitioner; gynecologist; other specialist; or other
health care provider type) and whether the patient had a usual place for health
We performed bivariable analyses to compare unadjusted exercise counseling
rates across various factors. We fit logistic regression models using backward
elimination and the Wald χ2 test to identify sociodemographic,
clinical, and physician factors associated with physician counseling. Two-tailed P<.05 was considered statistically significant. The
final model included all statistically significant independent factors and
confounders. Confounders were factors that, when added to the final model,
altered the regression coefficients of any significant independent variable
by at least 10%.
We performed additional analyses to examine the stability of our findings.
First, we adjusted our primary multivariable model for whether patients were
attempting to lose weight at the time of survey administration to account
for patient-initiated counseling. We hypothesized that patients who were actively
attempting to lose weight would have been more likely to initiate counseling
with their physicians; this might explain any association between physician
counseling and factors correlated with patients' goals of losing weight, such
as BMI and health-seeking behaviors. Second, we adjusted for what patients'
physical activity levels were in the 2 weeks prior to the survey. We classified
respondents into low (sedentary), moderate, or high activity level based on
a validated method described previously.25
Those who participated in moderate activity (eg, walking, gardening, bowling)
more than 4 times a week or vigorous activity (eg, running, swimming, playing
tennis) more than 2 times a week were classified as having high physical activity
level. Respondents reporting 1 to 4 moderate-level activities or 1 to 2 vigorous
activities a week were classified as practicing moderate physical activity.
All others were considered sedentary. Third, we examined the effect of respondent
occupations on factors associated with physician counseling by performing
subgroup analyses for respondents who reported occupations involving high
levels of physical activity (eg, police or firefighters; cleaning or building
servicepeople; laborers) and in those who reported sedentary occupations (eg,
executives, administrators, or managers; scientists; health care providers;
teachers, librarians, or counselors; technicians). This classification was
based on NHIS-defined categories24; we did
not include categories that could not be readily classified. We were able
to classify 61% of those eligible who reported having an occupation (n=5321).
We also examined the effect of missing income data on our model, because approximately
14% (n=1260) of our study sample lacked information about income. To obtain
an annual estimate of household income for those with missing data, we multiplied
by 12 the monthly income imputed by the National Center for Health Statistics
based on age, sex, race, family status, and other economic and health characteristics.25 We applied our model to this larger sample (including
respondents with imputed data) to test the stability and generalizability
of our findings.
We used statistical analysis software for proper variance estimation
in all analyses.26,27 Results
were weighted to adjust for nonresponse to reflect US population estimates.
We used Taylor series linearization to estimate SEs.26,27
Of the 17,317 respondents to the Year 2000 supplement, 9711 reported
having seen a physician for a medical check-up in the preceding year, and
9299 responded when asked about physician counseling about exercise. Table 1 displays the sociodemographic characteristics
of the study sample (n=9299) and the unadjusted rates of counseling by these
characteristics. The overall rate of physician counseling about exercise was
34%. Counseling rates did not vary significantly by physician specialty (P=.69). Of adults who reported a general internist or family
practitioner as their usual provider, 35% received counseling to exercise.
If the usual provider was a gynecologist or a specialist, patients reported
counseling rates of 36% and 37%, respectively. Rates of counseling also did
not differ by patient sex or race; however, physicians were more likely to
counsel patients who were older than 30 years, who were married, and who were
of higher socioeconomic status.
Table 2 shows the clinical
characteristics of the study sample along with unadjusted rates of counseling.
Physicians counseled overweight (BMI, 25 to <30 kg/m2) and obese
respondents (BMI ≥30 kg/m2) at higher rates than those with
lower BMI (<25 kg/m2). However, physicians were less likely
to counsel those who were more sedentary and more likely to counsel respondents
with cardiac disease and diabetes. Patients who reported frequent physician
visits or contacts also were more likely to be counseled.
Table 3 presents factors
that were independently associated with exercise counseling after adjustment
(n=7410). Age remained a significant correlate, with adults aged 40 to 49
years being counseled most often to exercise. Women were more likely to be
counseled than men, as were patients with incomes above $50,000 and higher
education. Patients who were uninsured or who had Medicaid insurance were
counseled less often. Compared with normal-weight patients (BMI, 18.5 to <25
kg/m2), obese patients (BMI ≥30 kg/m2) were significantly
more likely to receive counseling about exercise. Patients with cardiac disease
and diabetes were more likely to report counseling as well. Patients were
less likely to be counseled if they were single, used tobacco, had difficulty
walking, and had infrequent physician visits or contacts in the previous year.
Physician specialty, previous hospitalization, perceived health status,
and comorbid illnesses other than cardiac disease and diabetes were not independently
associated with physician counseling after multivariable adjustment and did
not confound the relationship between independent variables in our model and
physician counseling about exercise.
When we adjusted our final model for whether respondents were attempting
to lose weight, differences by patient sex were no longer significant (P=.48); the odds ratio for women was 0.96 (95% confidence
interval [CI], 0.85-1.08). All other correlates of exercise counseling remained
statistically significant. In a subanalysis of patients with complete activity
data (n=6508), adjusting for physical activity level in the 2 weeks prior
to survey did not alter our primary results substantially, even though sedentary
respondents were less likely to be counseled. Compared with those reporting
high levels of physical activity, the adjusted odds ratios for exercise counseling
were 0.36 (95% CI, 0.31-0.41) for those reporting the lowest physical activity
and 0.61 (95% CI, 0.52-0.71) for those reporting moderate activity. Stratifying
by sedentary (n=2704) and nonsedentary occupations (n=526) also produced similar
results to those of our primary analysis. In particular, the association of
physician counseling with patient income, education, and insurance type was
not diminished. Also, to test the generalizability and stability of our major
findings, we included patients with missing income data using National Center
for Health Statistics–imputed income and applied our final model to
this larger sample (n=8488). These results were consistent with those obtained
from our primary sample.
We found the national rate of physician exercise counseling to be 34%
among patients who saw a physician in the previous year for a medical check-up.
Women and racial and ethnic minorities were counseled as often as men and
whites after adjustment for other sociodemographic and clinical factors. However,
physicians counseled sedentary patients and those with lower socioeconomic
status less often. Moreover, physicians appeared to counsel as a form of secondary
prevention, as evidenced by higher counseling rates in patients who were already
obese, who were older than 30 years, or who had comorbid conditions.
Although our findings are consistent with physician surveys and previous
studies on behavioral counseling, this study is the first, to our knowledge,
to confirm these findings relating to counseling about exercise using a nationally
Physicians consistently report that they counsel a minority of their patients
Although few studies identify patient characteristics associated with physical
activity counseling, data from behavioral counseling on smoking cessation
show that physicians tend to counsel patients whose health is already compromised.28,29 Smaller studies also show that patients
who have heart disease or cardiovascular risk factors are more likely to be
counseled about physical activity.28- 30
The low rate of physician counseling is especially troubling given the
increasing prevalence of obesity and sedentary lifestyle nationally. Patients
who are in poorer health and who are obese may be more motivated to make lifestyle
changes and should be aggressively counseled. Failure to counsel healthier
and disease-free adults, however, may represent important missed opportunities.
Evidence indicates that adults between the ages of 25 and 35 years, sedentary
individuals, and those with low socioeconomic status are at high risk for
obesity and weight gain.2,3,23,31
Our study reveals that these groups are particularly at risk for not being
counseled by physicians. Moreover, differences in leisure time or occupation-related
activity level did not explain the disparities by age and socioeconomic status.
These findings are consistent with results from 1 survey that showed that
physicians who estimated a higher proportion of their patients as nonsedentary
were more likely to counsel.22 The lower rates
of counseling in respondents with lower education and income levels, even
after adjusting for access to health care, illness burden, and activity level,
are particularly worrisome, because members of lower socioeconomic groups
have poorer health outcomes.23
Physicians' perceptions that counseling may be ineffective may explain
the low rate of physical activity counseling. Given competing time pressures
and limited resources, physicians may elect to pursue other, more proven preventive
measures such as cancer screening. Moreover, physicians may be targeting their
counseling efforts at those most likely to benefit from exercise. Despite
strong evidence suggesting that physical activity is beneficial, few data
are available on the long-term effectiveness of exercise counseling. Several
studies examining the efficacy of physician advice about exercise have shown
positive results.11- 15
The majority of these studies, however, were not randomized or well controlled;
many also lacked follow-up data beyond a few weeks and were limited in generalizability.12,15 The few available randomized trials
have shown conflicting results.11,13,14
The Johns Hopkins Medicare Preventive Services Demonstration Project, which
examined the effect of preventive health examinations in an elderly population,
found no significant increase in physical activity level in those randomized
to receive exercise counseling.11 However,
a second study demonstrated a statistically significant increase in exercise
activity after counseling. The OXCHECK trial randomized 2205 patients in England
to receive physical activity counseling and 5 follow-up visits by trained
nurses.13 After 3 years, a significantly higher
proportion of the intervention group (32.4%) compared with the control group
(29.1%) reported vigorous exercise more than once a month. Finally, perceptions
that counseling is less effective in certain patient populations may explain
the variable rates across certain patient subgroups. For example, physicians
may believe that counseling patients with lower socioeconomic status is less
effective. Data, however, demonstrate that low-income patients are actually
more likely to attempt behavioral change based on physician advice.32
Although we used data from a large, nationally generalizable sample,
our results are subject to recall bias. Some respondents may not have remembered
discussions with physicians about exercise even when they occurred. This bias
would underestimate actual counseling rates; nevertheless, patients' recall
may be an accurate reflection of the quality of physician discussions and
the influence of these discussions on the patient. Second, the rate of counseling
about exercise we report only represents counseling that took place during
the most recent medical check-up and does not necessarily reflect annual rates,
except in patients with only a single physician visit in the previous year.
In patients who had multiple physician visits in the previous year, we would
not have had access to data about counseling that occurred at an earlier visit.
However, those with more frequent health care provider visits and contacts
were more likely to report that counseling took place. Third, we were also
unable to differentiate between physician- and patient-initiated counseling.
The correlates of counseling we identified may represent correlates of health-seeking
behavior. Controlling for whether respondents were attempting to lose weight
did not alter our findings substantially, however. We attempted to adjust
for physical activity level, but the available data referred to activity during
the 2 weeks prior to survey and not necessarily to baseline activity at the
time of the physician visit. Patients who receive physician counseling about
physical activity may be more likely to initiate exercise, which would explain
our observation that persons who engaged in more physical activity were more
likely to be counseled. However, given the modest efficacy of physician counseling,
it is unlikely that this phenomenon completely explains the lower rates of
counseling in more sedentary patients. Finally, we had limited information
about clinician characteristics or practice beliefs other than specialty.
Our findings demonstrate that the rate of physician counseling about
exercise in the United States is low. Physicians often counsel patients to
exercise as a form of secondary prevention and undercounsel certain groups
at high risk for obesity, weight gain, and sedentary lifestyle. In particular,
the failure to counsel younger, disease-free, and sedentary adults and those
from lower socioeconomic groups might represent important missed opportunities
for primary prevention that could lead to adverse public health outcomes.
Given that the problems of obesity and sedentary lifestyle have reached epidemic
proportions in the United States, even modest benefits from relatively benign
interventions, such as counseling to increase physical activity, may have
a substantial public health impact. Additional studies are needed to evaluate
the effectiveness of physician counseling about exercise and the effectiveness
of interventions designed to improve physician counseling, especially in groups
at risk for not having discussions about exercise with their physicians.