Forrest CB, Whelan E. Primary Care Safety-Net Delivery Sites in the United StatesA Comparison of Community Health Centers, Hospital Outpatient Departments, and Physicians' Offices. JAMA. 2000;284(16):2077-2083. doi:10.1001/jama.284.16.2077
Author Affiliations: The Primary Care Policy Center for the Underserved, Department of Health Policy and Management, Johns Hopkins School of Public Health, Baltimore, Md.
Context The US primary care safety net is composed of a loose network of community
health centers, hospital outpatient departments, and physicians' offices.
National data on how the mix of patients and services differ across sites
Objective To develop and contrast national profiles of patient and service mix
for primary care.
Design, Setting, and Patients Comparative analyses of 3 national surveys of primary care visits occurring
in 1994: for data on physician's office visits, the National Ambulatory Medical
Care Survey (NAMCS); for hospital outpatient department data, the National
Hospital Ambulatory Medical Care Survery (NHAMCS); and for data on community
health centers, the Bureau of Primary Health Care's 1994 Survey of Visits
to Community Health Centers. A time trend analysis also was conducted using
the 1998 NAMCS and NHAMCS.
Main Outcome Measures National estimates of primary care visit rates, types of patient presentation,
patient case-mix, disposition of patients, and management interventions in
1994, and compared with 1998 data.
Results The US population made 1.3 primary care visits per person in 1994, which
accounted for 43.5% of all ambulatory visits to physicians' offices, community
health centers, and hospital outpatient departments. Primary care visits per
person were 20% lower for Hispanics and 33% lower for black, non-Hispanic
persons compared with white, non-Hispanic persons. Visits to community health
centers were more likely to be made by ethnic minorities, patients with Medicaid
or no insurance, and rural dwellers than visits made to the other delivery
sites. Visits at hospital outpatient departments were made by sicker populations
and were characterized by less continuity than the other delivery sites. Controlling
for patient mix, visits made to hospital outpatient departments were more
commonly associated with imaging studies, minor surgery, and specialty referrals
than those made to physicians' offices. In 1998, the US population made an
estimated 3.4 visits per person, 45.6% of which were primary care visits.
National estimates of primary care visit rates and patient mix and practice
pattern comparisons between hospital outpatient departments and physicians'
offices were similar in 1998 and 1994.
Conclusions Expanding community health centers will likely improve access to primary
care for vulnerable US populations. However, enhancing access to of physicians'
offices is also needed to bolster the safety net. The greater service intensity
and poorer continuity for primary care visits in hospital outpatient departments
that we observed raises concern about the suitability of these clinics as
primary care delivery sites.
In 1994, more than 600 community health centers provided primary care
to 7.1 million individuals nationwide,1 increasing
to 8.1 million by 1998.2 This program is the
federal government's primary mechanism for attempting to ensure access to
primary care for populations residing in medically underserved areas. However,
the Bureau of Primary Health Care (BPHC), Health Resources and Services Administration,
Department of Health and Human Services, estimates that community health centers
provide services for just 1 in 6 persons who lack access to a primary care
practitioner.1 A patchwork of locally funded
hospital outpatient departments, physicians' offices, and community clinics
therefore supplies most of the safety net primary care. This local structure
has led to substantial variation across communities in the composition and
capacity of the primary care safety net.3,4
Understanding how the mix of patients and the content of services differ
by the primary care delivery site is important to decision makers who develop
policy aimed at improving access to health care. Because of limitations in
data sources, there have been no comprehensive national studies that compare
primary care delivery among community health centers, hospital outpatient
departments, and physicians' offices. The National Center for Health Statistics
conducts separate annual surveys of visits to office-based practitioners5 and hospital outpatient departments.6
In 1994, the BPHC conducted the only national Survey of Visits to Community
Health Centers.7 The objectives of this study
were to use these 3 surveys to develop and contrast national profiles of patient
and service mix for primary care visits made to community health centers,
hospital outpatient departments, and physicians' offices. Using data from
the 1998 National Center for Health Statistics surveys,8,9
we also assess how these profiles have changed over time.
Visit data from physicians' offices were obtained from the National
Ambulatory Medical Care Survey (NAMCS).5 This
survey's target universe included visits made to nonfederally employed, office-based
physicians in the United States; hospital outpatient departments were excluded.
Selected physicians completed questionnaires for a systematic sample of patient
visits made during 1 week.
The 1994 National Hospital Ambulatory Medical Care Survey (NHAMCS) was
used for visits to hospital outpatient departments.6
This survey's target universe included visits made to outpatient departments
of nonfederal, short-stay hospitals. Using patient records, hospital staff
completed questionnaires for a random sample of visits that occurred during
a randomly assigned 4-week reporting period. We excluded visits made to emergency
departments from the sample.
In 1994, the BPHC replicated the NHAMCS survey method (as described
above) for a probability sample of visits made to community health centers.7 The target universe included visits to community health
centers, including all of their delivery sites that received funds through
the BPHC grant program under Section 330 of the Public Health Service Act
(42 USC 2546). Grantee centers in operation fewer than 2 years were excluded
from the sampling frame.
Additional details of the surveys' methods are available elsewhere.5- 7 The response rates for
in-scope physicians or clinics were 95% for hospital outpatient departments,
100% for community health centers, and 70% for physicians' offices. For the
items in this study, the survey instruments from all 3 data sources used the
same questions, which were listed in the same order.
We also used the 1998 NAMCS8 and 1998
NHAMCS9 for a 1994-1998 time trend analysis.
Because physicians practicing in community health centers were in-scope for
the NAMCS, these 2 data sources provided similar national estimates as the
3 from 1994. The methods for the 1998 surveys were the same as for counterparts
done in 1994. Response rates for in-scope physicians or clinics in 1998 were
68% for physicians' offices and 98% for hospital outpatient departments.
Several exclusion criteria were applied to the data sets to obtain samples
of primary care visits. In the NAMCS data sets, we selected visits made to
general pediatricians, family physicians, general internists, and obstetricians/gynecologists;
in the NHAMCS data sets, visits to medical, obstetric, and pediatric clinics
were selected; and in the BPHC data set, we selected all visits because community
health centers are organized to deliver primary care.
We removed visits made to specialists or specialized clinics falsely
identified as generalists or general clinics because self-designated specialty
information can overestimate the number of generalist physicians.10 Physicians in the NAMCS data set who did not provide
any preventive care (defined by reasons for encounter and diagnosis codes)
were excluded; similarly, clinics in the NHAMCS data set that did not provide
preventive care during at least 10% of their visits were excluded. Physicians
and clinics that recorded at least 20 visits and gave diagnosis codes in the
same clinical area (cardiology, endocrinology, etc) for greater than 50% of
these visits also were excluded from the sample. Applying these exclusion
criteria yielded unweighted sample sizes for the 1994 data sets of 2878 (100%
of total) primary care visits made to 48 community health centers, 9853 (29.3%
of total) made to offices of 409 physicians, and 10,371 (35.6% of total) made
to 162 hospital outpatient departments. The exclusion criteria applied to
the 1998 data sets yielded 6913 visits (29.6% of total) made to the offices
of 263 physicians, and 11,600 (39.4% of total) made to 175 hospital outpatient
Diagnosis and reason for encounter codes were used to identify routine/preventive
care visits. Specific items from each survey allowed further classification
of patient presentation by new vs established patients; among new patients,
those referred by another physician; and among established patients, those
presenting with new health problems. (These items were not available in the
1998 surveys.) Higher percentages of established patients presenting with
new problems suggest better continuity of care.11
The case-mix index was based on the Johns Hopkins Adjusted Clinical
Groups (ACG) Case-Mix system.12 The ACG system
assigns all International Classification of Diseases, Ninth
Revision, Clinical Modification (ICD-9-CM)13 codes to 1 of 32 morbidity groups, or diagnostic
groups called ADGs. In all surveys, up to 3 ICD-9-CM codes were recorded for each visit; thus, each
visit could be assigned up to 3 unique ADGs. The case-mix index was the sum
of ADG-specific resource intensity weights. The weights were obtained from
ADG β coefficients derived from regression analyses done for 4 large
commercial health plans in which total patient charges were regressed on the
32 ADGs. (Details of this method, including the specific weights used, are
available from the authors.) To simplify interpretation of the case-mix index,
raw scores were standardized to a mean (SD) of 50 (10) for the combined sample.
Higher case-mix index scores indicate greater medical complexity, sicker patients,
and higher expected resource use.
To assess visit intensity, we predicted the duration of each visit based
on the age and sex of the patient, type of patient presentation, case-mix
index, and diagnosis codes. Estimating visit intensity using expected visit
duration is a method used previously in the development of Ambulatory Visit
Groups,14,15 a case-mix measure
developed in the 1980s using NAMCS data. Visit duration weights were the β
coefficients obtained from regressing visit duration in minutes on the covariates
described above using generalist and obstetric patient visits from the 1989-1993
Patient management practices were compared across the delivery sites
on the disposition of the patient following visits and whether any of the
following were done during the visits: medication(s) prescribed, laboratory
tests, imaging studies, minor surgery, or a blood pressure check during a
routine, preventive care visit. Each 1994 data set included a disposition
variable that identified visits in which no follow-up encounters were scheduled,
specialty referrals were made, or patients were admitted to hospital. Disposition
was not available in the 1998 data sets. Information on laboratory tests,
imaging studies, and minor surgery was not collected in the community health
Sampling weights that accounted for the multistage sample design and
nonresponse of in-scope practitioners were used to obtain national estimates
of numbers of visits made to each primary care delivery site. Population rates
of visits per 100 persons in the United States were based on the US Bureau
of the Census estimates of the mid-year, civilian, noninstitutionalized US
population.16 Physicians working in community
health centers are theoretically in scope for the NAMCS. For the 1994 primary
care visit estimates, we subtracted the count of visits to community health
centers from the count of visits for physicians' offices.
Type of patient presentation, diagnostic case-mix of patients, visit
intensity, and patient management of primary care visits were compared across
the 3 delivery sites. Statistical analyses using unweighted data contrasted
proportions (χ2 analysis) and means (t
tests) of physicians' offices with the 2 traditional safety-net delivery sites—community
health centers and hospital outpatient departments. Because delivery of preventive
care depends primarily on the age-sex distribution of a population, the regression
analyses for routine/preventive care controlled for age-sex groups only. Other
logistic regression analyses controlled for age, sex, payer, race/ethnicity,
case-mix, and rural residence (nonstandard statistical metropolitan areas).
Multivariable linear regression was used to adjust case-mix index means
for differences in other patient characteristics.
In 1994, approximately 747.8 million outpatient visits were made to
community health centers, hospital outpatient departments, and physicians'
offices, a rate of 2.9 visits per person (Table 1). Of the 747.8 million visits, an estimated 325 million
(43.5%) were primary care visits, a rate of 1.3 primary care visits per person.
Obstetric services for pregnant women, defined by diagnosis codes, comprised
5.8% of primary care visits made to community health centers, 3.9% to physicians'
offices, and 3.9% to hospital outpatient departments.
Community health centers constituted 4.0%, hospital outpatient departments
6.7%, and physicians' offices 89.3% of all primary care visits (Table 1). The distribution of primary care visits for patients with
either Medicaid or no insurance was 10.6% to community health centers, 11.5%
to hospital outpatient departments, and 77.9% to physicians' offices. A similar
distribution was found for primary care visits made by ethnic minorities:
11.2% to community health centers, 11.9% to hospital outpatient departments,
and 76.9% to physicians' offices. However, a smaller percentage of primary
care visits for ethnic minorities with either Medicaid or no insurance were
made to physicians' offices: 19.6% to community health centers, 17.5% to hospital
outpatient departments, and 62.9% to physicians' offices.
The number, percent distribution, and annual rates of primary care visits
are shown by delivery site and patient characteristics in Table 2. Primary care visits made by either uninsured persons or
those with Medicaid financing accounted for 65.4% of those made to community
health centers, 43.0% to hospital outpatient departments, and only 18.5% to
physicians' offices. Rural dwellers comprised 49.6% of community health center
primary care visits, which contrasted with 19.7% in physicians' offices and
5.1% in hospital outpatient departments.
Compared with the white, non-Hispanic population, the overall primary
care visit rates to all 3 types of facilities for the Hispanic population
were 20% lower and 33% lower for the black, non-Hispanic population. The Asian/Pacific
Islander and white, non-Hispanic populations had similar primary care visit
For the Hispanic vs white, non-Hispanic populations, rates of primary
care visits made to community health centers were 700% higher, 79% higher
to hospital outpatient departments, but 35% lower to physicians' offices.
For the black, non-Hispanic vs white, non-Hispanic populations, rates of primary
care visits made to community health centers were 550% higher, 93% higher
to hospital outpatient departments, and 48% lower to physicians' offices.
Table 3 compares the type
of patient presentation, diagnostic case-mix of patients, and predicted visit
duration of primary care visits made to the 3 delivery sites. Among established
patients, the odds that primary care visits were for new health problems were
significantly higher in community health centers compared with physicians'
offices and hospital outpatient departments. A greater proportion of primary
care visits made to hospital outpatient departments were for new patients,
many of whom were physician-referred, than those made to physicians' offices.
The case-mix index of patient visits was 2.8% higher, suggesting a sicker
patient population, in hospital outpatient departments than physicians' offices
and was similar between community health centers and physicians' offices (Table 3). The case-mix of primary care
visits made by the elderly population to the 3 types of delivery sites was
similar, whereas the case-mix index for primary care visits made by children
and adolescents was 1.7% higher in community health centers and 4.7% higher
in hospital outpatient departments than physicians' offices.
Visit intensity is a measure of the predicted duration of the visit
based on clinical characteristics of the patient and the type of patient presentation.
The overall visit intensity of hospital outpatient departments was 4.5% higher
than physicians' offices (Table 3).
The higher visit intensity in hospital outpatient departments translates into
visits expected to be about 40 seconds longer than those in physicians' offices;
visits for children and adolescents were predicted to be more than 2 minutes
(132 seconds) longer in hospital outpatient departments than physicians' offices.
Table 4 contrasts patient
management by primary care delivery site. The adjusted odds ratio of blood
pressure checks during routine check-ups was highest in hospital outpatient
departments. Controlling for differences in patient case-mix and other characteristics,
the service intensity of hospital outpatient department visits was higher
than those made to physicians' offices. Specifically, the adjusted odds ratios
of having an imaging study, minor surgery, or specialty referral were all
higher among primary care visits in hospital outpatient departments vs those
in physicians' offices.
In 1998, the US population made an estimated 3.4 visits per person,
45.6% of which were for primary care. Compared with 1994, similar proportions
of visits were made to hospital outpatient departments in 1998 (6.7% and 6.9%,
respectively) and made by ethnic minorities (20.9% vs 26.2%). The payer mix
of visits to physicians' offices and hospital outpatient departments in 1994
and 1998 was similar. In 1998, the case-mix of patient visits to hospital
clinics was 3.9% higher than physicians' offices; visit intensity was 5.6%
higher. For those management variables available in data sets from both years,
the trends in practice patterns between hospital outpatient departments and
physicians' offices found in 1994 held true in 1998.
This study used 3 nationally representative surveys to contrast primary
care visits made to community health centers, physicians' offices, and hospital
outpatient departments. The US population made an estimated 1.3 to 1.5 primary
care visits per person per year, which constituted 43.5% to 45.6% of all ambulatory
visits made to these delivery sites. Most primary care visits occurred in
physicians' offices, even for vulnerable subpopulations identified by a single
personal characteristic, such as race/ethnicity. However, about 40% of primary
care visits for ethnic minorities with either no insurance or Medicaid were
made to traditional safety-net delivery sites, such as community health centers
and hospital outpatient departments. This finding demonstrates the importance
of examining the impact of multiple vulnerable characteristics in studies
on access and service use.
Several caveats should be considered when interpreting this study's
data on primary care visit rates. Some safety-net delivery sites were not
included. The Veterans Affairs and Indian Health Service network of primary
care clinics were not part of our combined database. Exclusion of these federal
delivery sites is likely to have a small impact on visit rates estimates:
it has been estimated that Veterans Affairs clinics provide services to about
1.7 million low-income veterans.17 Although
locally funded health departments were not explicitly evaluated in this study,
physicians in these delivery sites were in the sampling frame of the NAMCS,
and visits to them were categorized in the physicians' offices group. Second,
the sampling frame of the NAMCS survey of physicians' offices excluded nurse
practitioners and physician assistants, who are becoming increasingly important
in the provision of primary care.18 Third,
we used algorithms to identify primary care providers in the 3 surveys and
then classified all visits to these physicians and clinics as primary care.
Our methods may have overestimated the number of primary care visits to the
extent that generalist physicians and obstetrician/gynecologists deliver some
specialty care services. On the other hand, the method may have underestimated
primary care visits to the extent that specialists provide primary care. To
examine how these potential biases may influence our estimates, we examined
the 1998 NAMCS survey,8 which contained an
item that asked physicians to report whether they were the patient's primary
care practitioner. Analysis revealed that 85.4% of visits to generalist physicians,
24.6% to obstetricians/gynecologists, and just 8.8% to specialists were for
patients for whom physicians reported they were the primary care practitioner.
Fourth, differences in survey administration between the NAMCS (physicians
completed the questionnaires) vs the other 2 surveys (medical record abstraction)
could affect the diagnosis coding and thus the morbidity assessments. Analysis
of the ICD codes recorded for each visit revealed
no significant differences in the mean numbers or the percentages with more
than 1 code among the 3 delivery sites.
Continuity of care, one of the hallmark features of primary care,11 was better in community health centers than other
delivery sites. Established patients at community health centers were more
likely to present with new health problems compared with the other delivery
sites. Better continuity of care allows practitioners to use more watchful
waiting and employ health care resources more judiciously, an association
that helps explain the lower likelihood of prescribed medications during primary
care visits to community health centers.
The medical complexity and patient management of primary care visits
in community health centers compared favorably with physicians' offices. None
of the measures in this study, however, assessed the social complexity of
patients, which should be greater in community health centers because of their
higher burden of low-income patients. The impact of patients' social risk
on the overall complexity and resultant resource intensity of primary care
visits is likely to have a positive effect.
Physicians working in community health centers were theoretically in-scope
in the NAMCS. To calculate primary care visit rates, we subtracted the count
of visits to community health centers from the count of visits for physicians'
offices. However, for the comparative analyses of patient mix and service
use, no such adjustment was made. The bias that may result from this overlap
in sampling frames is likely to be negligible because the number of community
health center visits possibly included in the physicians' office sample was
The medical complexity of patient visits in hospital outpatient departments
was greater than those made to other delivery sites. Using a measure of visit
intensity that predicted the visit duration, we estimated that the patient
population visiting hospital outpatient departments was 4% to 6% sicker than
the population visiting physicians' offices. This estimate is consistent with
the study by Lion and Altman14 from the 1970s,
which used a similar measure and found 5% to 15% higher visit intensity at
hospital outpatient departments vs physicians' offices. An important clinical
implication of the greater visit intensity is that primary care visits to
hospital clinics may last longer than those to physicians' offices—as
much as 2 minutes per visit for children and adolescents.
Even after differences in case-mix and other patient characteristics
were controlled for, the service intensity among primary care visits in hospital
outpatient departments exceeded that of visits made to physicians' offices.
Hospital outpatient department visits were more likely than visits to physicians'
offices to be associated with a specialty referral, minor surgery, and imaging
studies. These findings are consistent with studies using Medicaid claims
data that showed patients who use hospital outpatient departments incurred
higher risk-adjusted expenditures than those who used physicians' office.19,20 Greater resource use in hospital
outpatient departments could be a consequence of better availability of ancillary
and specialty services. Furthermore, the poorer continuity of care among visits
to hospital outpatient departments may contribute to a more aggressive practice
style, because of less familiarity with patients' health histories. Greater
service intensity and poorer continuity of care in hospital outpatient clinics
compared with other delivery sites also raise the concern over the suitability
of these clinics as primary care delivery sites. However, until studies are
done that compare health outcomes across delivery sites, we can make no conclusions
about which service use rate is best.
The main findings from this study are from data collected in 1994. To
assess the applicability of these data for current policy decisions, we conducted
sensitivity analyses using data from 1998. Our findings showed remarkable
similarity between the 1994 and 1998 results, suggesting that our primary
care visit analyses are applicable to current health care policy. Overall
primary care visit rates and patient mix were similar, and higher service
intensity in hospital outpatient departments compared with physicians' offices
was found in both years. This sensitivity analysis was limited by not including
specific community health center estimates and not explicitly evaluating the
impact of managed care. Some safety-net delivery sites have been responding
to the pressures of health care system change by forming integrated delivery
systems and expanding their contracts with managed care plans to attract paying
patients.21 Despite these new organizational
arrangements, the patient and payer mix at community health centers has remained
relatively constant over this time period, according to BPHC estimates (Jerrilynn
Regan, MSN, MPA, BPHC, Health Resources and Services Administration, oral
communication, August 2000).
Our estimates indicate that 1 in 25 primary care visits in the nation
occurred in community health centers. The patient population making visits
to community health centers was characterized by larger shares of ethnic minorities,
individuals with Medicaid or no insurance, rural dwellers, and women receiving
obstetric services compared with the other 2 delivery sites. An important
implication of these findings is that federal policy that affects the financing
or organization of community health centers will have its greatest impact
on access for these vulnerable subpopulations.
This study documented large inequalities between ethnic minorities and
whites in primary care visit rates. Compared with the white population, the
Hispanic population made 20% fewer primary care visits per person and the
black, non-Hispanic population made 33% fewer primary care visits per person.
Can expansions in the capacity of community health centers—the main
federal policy lever for improving nonfinancial access to care for vulnerable
populations—eliminate these disparities? If the number of community
health centers were doubled and the composition of patient populations did
not change for the delivery sites, the white-Hispanic primary care visit disparity
would be reduced by 50%, but the white-black disparity would be reduced by
just 24%. These are significant reductions in access disparities and move
our nation closer to a more equitable primary care system. However, eliminating
race/ethnicity inequalities in use of primary care will require policy directed
at improving vulnerable populations' access to mainstream physicians' offices
as well as traditional safety-net delivery site.