Annals of Internal Medicine 2001 (Aug 21); 135 (4): 262–268 ~ FULL TEXT
You may also enjoy this Dynamic Chiropractic extended review of this article
Ronald C. Kessler, PhD; Roger B. Davis, ScD; David F. Foster, MD;
Maria I. Van Rompay, BA; Ellen E. Walters, MS; Sonja A. Wilkey, BA;
Ted J. Kaptchuk, OMD; and David M. Eisenberg, MD
BACKGROUND: Although recent research has shown that many people in the United States use complementary and alternative medical (CAM) therapies, little is known about time trends in use.
OBJECTIVE: To present data on time trends in CAM therapy use in the United States over the past half-century.
DESIGN: Nationally representative telephone survey of 2055 respondents that obtained information on current use, lifetime use, and age at first use for 20 CAM therapies.
SETTING: The 48 contiguous U.S. states.
PARTICIPANTS: Household residents 18 years of age and older.
MEASUREMENT: Retrospective self-reports of age at first use for each of 20 CAM therapies.
RESULTS: Previously reported analyses of these data showed that more than one third of the U.S. population was currently using CAM therapy in the year of the interview (1997). Subsequent analyses of lifetime use and age at onset showed that 67.6% of respondents had used at least one CAM therapy in their lifetime. Lifetime use steadily increased with age across three age cohorts: Approximately 3 of every 10 respondents in the pre–baby boom cohort, 5 of 10 in the baby boom cohort, and 7 of 10 in the post–baby boom cohort reported using some type of CAM therapy by age 33 years. Of respondents who ever used a CAM therapy, nearly half continued to use many years later. A wide range of individual CAM therapies increased in use over time, and the growth was similar across all major sociodemographic sectors of the study sample.
CONCLUSIONS: Use of CAM therapies by a large proportion of the study sample is the result of a secular trend that began at least a half century ago. This trend suggests a continuing demand for CAM therapies that will affect health care delivery for the foreseeable future.
From the FULL TEXT Article:
Community surveys done over the past decade have
documented that a substantial proportion of Americans
use complementary and alternative medical
(CAM) therapies (1– 4), which have been defined as “interventions
neither taught widely in medical schools nor
generally available in U.S. hospitals” (1). Many managed
care organizations have responded to this evidence
by providing insurance coverage for some CAM therapies
(5). Furthermore, most U.S. medical schools have
begun offering courses on CAM therapies (6).
These responses imply that CAM therapies are perceived
to be a force to be reckoned with for some time
to come. Yet, little is known about the likelihood that
this will be the case. The prevailing assumption is that
CAM therapies were used by a fairly narrow segment of
the population until the 1970s, at which time the ideology
associated with the youth counterculture led to a
rapid dissemination and use of CAM therapies that has
persisted through the present (7). However, lack of rigorous
trend data from epidemiologic surveys have precluded
evaluating this assumption rigorously or projecting
the future growth of CAM therapies on the basis of
evidence of past trends.
In the current report, we present nationally representative
trend data of this sort from a prevalence study.
The data came from retrospective self-reports of a nationally
representative sample of the U.S. general population
in a 1997–1998 telephone survey (4) about age at
first use of 20 representative CAM therapies. In our
analysis, we studied trends by examining betweencohort
differences in rates of initiation of CAM therapy
use (8). In the absence of prospective data, which do not
exist, our results represent, to our knowledge, the most
accurate information currently available on U.S. trends
in CAM therapy use over the past half-century.
The telephone survey was conducted between November
1997 and February 1998 in a nationally representative
household sample. Random-digit dialing was
used to select households, and a random-selection
method was used to select one respondent 18 years of
age or older for interview in each sample household.
Eligibility was limited to English speakers without cognitive
or physical impairment that would prevent interview
completion. The average administration time was
30 minutes. A $20 financial incentive for participation
was offered. The Beth Israel Deaconess Committee on
Clinical Investigations, Boston, Massachusetts, approved
the survey methods.
Of the initial sample of 9750 telephone numbers,
26% did not work, 17% were not assigned to households,
and 9% were unavailable despite six attempted
follow-up contacts. Of the remaining households, 481
were ineligible because of language barrier or cognitive
or physical incapacity. Of the 4167 total eligible respondents,
1720 (41.3%) completed the interview on initial
request. Of a random subsample of 1066 persons who
initially declined and were offered an increased stipend
($50), 335 agreed to participate. In all, 2055 interviews
were completed. After we extrapolated the conversion
rate to all persons who had initially declined and
weighted the data for the undersampling of those who
participated after initially declining, the weighted overall
response rate among eligible respondents was 60%.
The data were weighted for three factors: 1) probability
of selection within household as well as geographic
variation in cooperation (by region of the country and
urbanicity [local population density]), 2) nonresponse,
and 3) post-stratification for aggregate discrepancies between
the sample distributions and Census population
distributions on a variety of sociodemographic variables
(9, 10). More details on the sample design have been
presented elsewhere (4). Age data were missing for 6
respondents; our analyses are limited to the remaining
The interview was described to respondents as a survey
by investigators from Harvard Medical School about
the health care practices of Americans. Interviewers
made no mention of CAM therapies. The first substantive
questions concerned perceived health, functional
impairment due to health problems, interactions with
physicians, and history of chronic medical conditions.
Interviewers then queried respondents about their lifetime
and recent use of 20 CAM therapies—acupuncture,
aromatherapy, biofeedback, chiropractic care, commercial
diet programs, energy healing (for example,
laying on of hands), folk remedy, herbal medicine, homeopathy,
hypnosis, imagery, lifestyle diet (such as vegetarianism
or macrobiotics), massage, megavitamin therapy,
naturopathy, osteopathy, relaxation techniques,
self-help group, spiritual healing by others, and yoga.
Users of each therapy were asked their age at first use as
well as details about the conditions for which the therapy
was initiated. The final set of questions dealt with
Cohorts were aggregated into three subsamples: prebaby
boom (respondents $54 years of age at interview,
born before 1945); baby boom (34 to 53 years of age at
interview, born 1945–1964); and post– baby boom (18
to 33 years of age at interview, born 1965–1979). For
sociodemographic variables, we used two categories for
sex (male or female), four for race/ethnicity (non-Hispanic
white, non-Hispanic black, Hispanic, or other),
four for education (less than high school, high school
graduate, some college, or college graduate), four for
U.S. region (northeast, midwest, south, or west), and
four for urbanicity (residence in a large city, small city,
All analyses were performed with weighted data by
using SAS statistical software (11). To assess differences
in trends among cohorts, the Kaplan–Meier (12)
method was used to generate a graphic representation of
the cumulative lifetime prevalences of CAM therapy use
according to cohort. The significance of historical
changes in lifetime use was estimated by using discretetime
survival analysis (13), a method of survival analysis
appropriate for data in which events are recorded only at
discrete time points (for example, in yearly increments).
Discrete-time survival analysis was operationalized as a
logistic regression with person-year as the unit of analysis
and first use of CAM therapy as the outcome variable.
The predictors of primary interest were a series of
dummy variables for decades of historical time, and covariates
included sociodemographic and dummy variables
adjusted for the baseline hazard rate of each year of
a person’s life. This model results in an intercept for
each time period, and the odds ratios (ORs) can be
interpreted as the relative risk for the annual risk for use
of alternative therapy. Subsample models were estimated
to study sociodemographic variation in trends. Disaggregated
models were estimated to study trends in the
use of particular CAM therapies.
To adjust for the design effects introduced by
weighting of the data, the method of jackknife repeated
replications (14) was used to estimate standard errors
(SEs). For this method, we used user-written macros in
SAS statistical software. For this process, 50 random
primary sampling units were created with two random
half-samples in each unit for a total of 100 random
replicates. Jackknife repeated replication is a method
that uses simulations of coefficient distributions in subsamples
to generate empirical estimates of SEs and significance
tests. The ratios of the coefficients to these
adjusted SEs are used to compute the 95% CIs of the
ORs. Tests for the significance of sets of predictors
taken together were computed by using the Wald chisquare
test from coefficient variance– covariance matrices
based on the jackknife repeated replications simulations.
Differences in Aggregate Use Trends among Cohorts
At the time of interview, 67.6% of all respondents
had used at least one CAM therapy at some time in their
lives. Figure 1 presents Kaplan–Meier age-of-onset
curves showing trends in each cohort in the cumulative
probabilities of use according to age. Of note are the
dramatic differences in use among cohorts. This is seen
most clearly by focusing on cumulative probabilities of
use for age 33 years, the oldest age represented in all
three cohorts. Approximately 3 of every 10 respondents
in the pre– baby boom cohort used some type of CAM
therapy by the age of 33 years compared with 5 of 10 in
the baby boom cohort and 7 of 10 in the post– baby
Weighted Kaplan–Meier estimates
of age of first use
Historical Trends in Aggregate Use
The aggregate data in the Figure are presented in a
different format in the bottom row of Table 1, where
the risk ratios are shown from a discrete-time survival
model that estimated the effects of historical time in
predicting age at first use of CAM therapy among respondents
after adjustment for person-year and sociodemographic
variables. The contrast category is first use
before 1960. Consistent with the pattern in the Figure,
the results of the model for the outcome of any therapy
show monotonically increasing risk ratios in each decade
from the 1960s through the 1990s.
Possible demographic subsample differences in time
trends were examined by estimating separate subsample
models that were identical to the discrete-time survival
model for any therapy and by evaluating the statistical
significance of differences in trends across subsamples.
No statistically significant (0.05 level in two-sided tests)
differences in trends were found for sex, race/ethnicity,
education level, region of the country, or urbanicity.
Trends in Relative Risk for First Use of 20 Specific CAM Therapies
Table 1 A
Table 1 B
Trends in the Use of Specific Therapies
Table 1 also shows the risk ratios to estimate first
use of each of the 20 CAM therapies assessed. All trends
are statistically significant except for those for folk remedy,
naturopathy, and osteopathy; thus, in the United
States over the past half-century, use of most kinds of
CAM therapy has grown.
The data in Table 1 can be transformed to yield
information on changes over historical periods of time
rather than across cohorts. Data resulting from this
transformation (results not shown) yield interesting insights
into variations in the timing of societal adoption
of the different therapies. Use of all but 4 of the therapies
increased in the 1960s compared with pre-1960,
and the increase was with a risk ratio of 2.0 or greater
for 11 of these therapies. In the 1960s, growth in the use
of 4 therapies increased markedly—commercial diet
programs, lifestyle diet therapy, megavitamin therapy,
and self-help groups. Increased use of all 4 of these therapies
is consistent with the largely increased national
interest that was seen at that time in fitness and health as
a result of a fitness campaign initiated by President
Kennedy in the early 1960s (15).
Our analysis also shows markedly increased growth
in the use of alternative therapies in the 1970s as well;
use of all the therapies increased in this decade compared
with use in the 1960s, and this increase was substantial
for 4 of the therapies (biofeedback, energy healing,
herbal medicine, and imagery). In addition, 8 of the
20 assessed therapies had their largest rate of growth
during the transition from the 1960s to the 1970s
(biofeedback, energy healing, folk remedy, herbal medicine,
homeopathy, hypnosis, imagery, and spiritual
healing by others).
The 1980s saw more modest growth in the use of
CAM therapies. Use of 14 of the therapies increased
compared with the 1970s, but only 2 (massage and naturopathy)
had risk ratios that were markedly greater
than in the 1970s. Only yoga became significantly less
popular over this time period. Overall, a modest level of
growth continued in the 1990s, with 16 therapies having
increased use compared with the 1980s, but only 5
of these had markedly greater risk ratios than in the
1980s (aromatherapy, energy healing, herbal medicine,
massage, and yoga). Aromatherapy had the most dramatic
growth during this period. Although the risk ratios
for use of some of the therapies during the 1990s
were lower than in the 1980s, none of these declines was
Next, we examined the hypothesis that observed differences
in trends among cohorts were primarily due to
increased use of selected CAM therapies considered
mainstream (as opposed to alternative). To test this hypothesis,
we repeated our cohort analyses by fitting the
Kaplan-Meier age-of-onset curves to two additional
definitions of alternative therapy that excluded possibly
mainstream therapies. The intermediate definition excluded
biofeedback, commercial diet programs, hypnosis,
lifestyle diet, massage, osteopathy, and relaxation
techniques from the definitions used in this report; the
narrow definition also excluded imagery and megavitamin
therapy. The differences in trends between cohorts
remained significant (P , 0.001) for all definitions.
Persistence of Use
Table 2 presents data on the percentage of therapies
used in the past 12 months by lifetime users, who were
classified according to the number of years since first
use. The unit of analysis for this table is persontherapies.
Of the 2049 survey respondents, 1386 reported
lifetime use of at least 1 CAM therapy, with an
average of 3.03 therapies used by users, for a total of
4199 person-therapies. Continued use of a specific
therapy in the 12 months before the interview was 47%
among therapies begun 5 to 10 years before the interview,
48.8% for therapies begun 11 to 20 years previously,
and 45.7% for therapies begun more than 20
years earlier. The lack of significant difference among
prevalence rates suggests that slightly fewer than half of
lifetime users continued to use CAM therapies in any
given year throughout their life course.
Persistence of Use among Lifetime
Users, according to Time since Onset
of Initial Use
The results reported here suggest that the lifetime
prevalence of CAM therapy use in the United States has
increased steadily since the 1950s. This increase appears
not to be concentrated in a single population sector or
decade. Furthermore, the trend is seen for several therapies,
although variation can be seen in the timing of the
introduction of different therapies between the 1960s
and 1990s. We also found powerful cohort effects that
are substantial and consistent across cohorts. The post–
baby boom respondents had a higher rate of lifetime use
by age 33 years than the pre– baby boom respondents
had by age 79 years.
These consistent and pervasive results should dispel
any suggestion (16) that use has increased for only singular
complementary or alternative modalities or that
the use of CAM therapies is a passing fad associated with
one particular generation or fringe segment of the population.
On the basis of the plausible assumption that
demand for CAM therapies, similar to that for conventional
health care, is sensitive to how much patients
must pay out of pocket (17), it seems likely that the
proportion of people using CAM therapies will increase
as insurance coverage for these treatments expands in
Although limitations in our data do not allow examination
of persistence of use over time, we found that
50% of all CAM therapy use that had been initiated at
least 5 years before the interview had persisted at the
time of the interview. This finding is consistent with the
finding of a previous report (18) that most CAM therapies
are used, at least in part, to prevent future illness or
to maintain health and vitality as part of lifestyle choices
linked to the perceived value of disease prevention and
health promotion. It seems reasonable to assume that
individuals’ interest in prevention will probably persist
throughout their lifetime, and this factor may help explain
the high rate of persistent use documented in
Our results are limited by the restriction of the sampling
frame to people who spoke English and who lived
in households with telephones, as well as by the relatively
low response rate (60%) and the use of a financial
incentive (4). Furthermore, we have no data on the accuracy
of self-reports on recollection of lifetime use or
age at first use of a therapy. Recall bias is a possibility,
and the degree of bias might be related to respondent
age. Although we could not correct for these limitations,
such adjustments might have made the estimated trends
A second potential confounding element of this
study is the constant flux in the labels for and the very
existence of particular CAM therapies. These may not
necessarily have remained stable throughout the examined
periods. For example, some unconventional medical
professions, such as “drugless practitioners” and
“sanipractors,” which were licensed in some states in the
1930s, have since disappeared (19).
Our finding of a pervasive increase in the use of 20
CAM therapies over the past half-century warrants reflection
on these trends in a historical context. History
does not work statistically, and the role of chance makes
prediction even more difficult. Specific historic events
(such as the invention of the polio vaccination or the
discovery of a link between acupuncture and endorphins)
and general secular trends (such as confidence in
medical progress) are all likely to influence the use or
abandonment of CAM therapies. In the future, episodes
of conventional or CAM practices associated with dramatic
positive developments or adverse events may tip
prevalence patterns in one direction or another. Also,
from a historical perspective, our data may not represent
a consistent trend of increased use of CAM therapies but
rather a distinct peak in a longer trend of constant fluctuation
in CAM therapy use. For example, survey data
from the 1920s and 1930s have revealed high rates of
unconventional therapy use (20), and government statistics
from 1900 document large numbers of registered
unconventional practitioners (21). Our data may, in
fact, be demonstrating a resurgence in CAM use after a
period of diminished use during the 1940s and 1950s.
The trend of increased CAM therapy use across all
cohorts since 1950, coupled with the strong persistence
of use, suggests a continuing increased demand for
CAM therapies that will affect all facets of health care
delivery over the next 25 years. Evaluations of efficacy
and effectiveness by medical researchers and treating
physicians’ discussions with patients hold the promise of
minimizing adverse effects and maximizing the usefulness
of those CAM therapies that prove effective.
The authors thank the staff of DataStat, Inc., Ann
Arbor, Michigan, for assistance with telephone data collection.
In part by the National Institutes of Health (grants U24
AR43441 and K05 MH00507), the John E. Fetzer Institute, the American
Society of Actuaries, the Friends of Beth Israel Deaconess Medical
Center, the Kenneth J. Germeshausen Foundation, and the J.E. and Z.B.
Eisenberg DM, Kessler RC, Foster C, Norlock FE, Calkins DR, Delbanco
Unconventional Medicine in the United States:
Prevalence, Costs, and Patterns of Use
New England Journal of Medicine 1993 (Jan 28); 328 (4): 246–252
Use of alternative therapies: estimates from the 1994 Robert Wood Johnson Foundation National Access to Care Survey.
J Pain Symptom Manage. 1997;13:83-9
Druss BG, Rosenheck RA.
Association between use of unconventional therapies and conventional medical services.
Eisenberg DM, Davis RB, Ettner SL, Appel S, Wilkey S, Van Rompay M, et
Trends in Alternative Medicine Use in the United States,
from 1990 to 1997: Results of a Follow-up National Survey
JAMA 1998 (Nov 11); 280 (18): 1569–1575
Pelletier KR, Marie A, Krasner M, Haskell WL.
Current trends in the integration and reimbursement of complementary and alternative medicine by managed
care, insurance carriers, and hospital providers.
Am J Health Promot. 1997; 12:112-22
Wetzel MS, Eisenberg DM, Kaptchuk TJ.
Courses Involving Complementary and Alternative
Medicine at US Medical Schools
JAMA 1998 (Sep 2); 280 (9): 784-787
Berliner HS, Salmon JW.
The holistic alternative to scientific medicine: history and analysis.
Int J Health Serv. 1980;10:133-47
Shyrock HS, Siegel JS.
The Methods and Materials of Demography. vols. 1, 2.
Washington, DC: U.S. Bureau of the Census; 1971.
Holt D, Smith TM.
Journal of the Royal Statistical Society. 1979;142(Series A):33-46.
U.S. Bureau of the Census.
United States Population Estimates, by Age, Sex, Race, and Hispanic Origin, 1990 to 1997.
Accessed 10 June 2001.
SAS Users Guide, Release 6.12.
Cary, NC: SAS Institute; 1996.
Clayton D, Hills M.
Statistical Models in Epidemiology.
Oxford: Oxford Univ Pr; 1993:31-7.
Willett JB, Singer JD.
Investigating onset, cessation, relapse, and recovery: why you should, and how you can, use discrete-time survival analysis to examine event occurrence.
J Consult Clin Psychol. 1993;61:952-65
Lee ES, Forthofer RN, Lorimer RJ.
Analyzing complex survey data.
In: Lewis-Beck MS, ed.
Newbury Park, CA: Sage Publications; 1989.
The Health Movement: Promoting Fitness in America.
New York: Twayne Publishers; 1992.
A piece of my mind. Leeches, spiders, and astrology: predilections and predictions.
Shekelle PG, Rogers WH, Newhouse JP.
The effect of cost sharing on the use of chiropractic services.
Med Care. 1996;34:863-72
Why patients use alternative medicine: results of a national study.
Drugless healing in the 1920s: the therapeutic cult of sanipractic.
Pharmacy in History. 1986;28:14-25
The Healing Cults. A Study of Sectarian Medical Practice:
Its Extent, Causes and Control.
Chicago: Univ Chicago Pr; 1932.
American Physicians in the Nineteenth Century; From Sects to Science.
Baltimore: Johns Hopkins Univ Pr; 1985.
Return to the ALT-MED/CAM ABSTRACTS Page