Benefits and costs of a free community-based primary care clinic.
This study estimates the benefits and costs of a free clinic
providing primary care services. Using matched data from a free clinic
and its corresponding regional hospital on a sample of newly enrolled
clinic patients, patients' non-urgent emergency department (ED) and
inpatient hospital costs in the year prior to clinic enrollment were
compared to those in the year following enrollment to obtain financial
benefits. We compare these to annual estimates of the costs associated
with the delivery of primary care to these patients. For our sample
(n=207), the annual non-urgent ED and inpatient costs at the hospital
fell by $170 per patient after clinic enrollment. However, the cost
associated with delivering primary care in the first year after clinic
enrollment cost $505 per patient. The presence of a free primary care
clinic reduces hospital costs associated with non-urgent ED use and
inpatient care. These reductions in costs need to be sustained for at
least 3 years to offset the costs associated with the initially high
diagnostic and treatment costs involved in the delivery of primary care
to an uninsured
Community health services
Primary health care (Economic aspects)
Community hospitals (Services)
Community hospitals (Economic aspects)
Medical care, Cost of (Forecasts and trends)
Fertig, Angela R.
Corso, Phaedra S.
|Publication:||Name: Journal of Health and Human Services Administration Publisher: Southern Public Administration Education Foundation, Inc. Audience: Academic Format: Magazine/Journal Subject: Government; Health Copyright: COPYRIGHT 2012 Southern Public Administration Education Foundation, Inc. ISSN: 1079-3739|
|Issue:||Date: Spring, 2012 Source Volume: 34 Source Issue: 4|
|Topic:||Event Code: 360 Services information; 010 Forecasts, trends, outlooks Computer Subject: Market trend/market analysis|
|Product:||SIC Code: 8399 Social services, not elsewhere classified|
|Geographic:||Geographic Scope: United States Geographic Code: 1USA United States|
It has been estimated that the annual cost of uncompensated health
services provided to uninsured people in the United States was $56
billion in 2008 (Hadley, Holahan, Coughlin, & Miller, 2008). In
particular, the utilization and costs of emergency departments (ED) for
non-urgent care has been a considerable and growing problem over the
last several decades (Baker & Baker, 1994; Grumbach, Keane &
Bindman, 1993). More recently, there is evidence that potentially
preventable hospitalizations are on the rise as well (Russo, Jiang &
Barrett, 2007). In response to this inefficient and costly utilization
of hospital care across the country, many hospitals and communities have
sought different mechanisms for providing primary care for the uninsured
or under-insured. These mechanisms have ranged from free-standing
clinics, to mobile units, to expansion of primary care services provided
in already established settings, such as schools or churches. Funding
for these initiatives, which provide both free or nearly free healthcare
in many cases, comes from federal, state and local governments,
faith-based organizations, non-profit groups, businesses, hospitals, and
Most policy makers, healthcare industry leaders, funders and providers of primary healthcare believe that these primary care clinics are "cost-effective" because access to preventive care is believed to be a more efficient use of scarce health resources relative to the use of ED or inpatient services that might eventually be needed for untreated health conditions. However, there are surprisingly few studies that have rigorously assessed the economic impact of the provision of free primary care within a community.
Several studies have evaluated the utilization of ED services post implementation of primary care provision. A study conducted by Zahradnik (2008) found that the provision of free or low-cost primary care to uninsured patients in Michigan resulted in per person reductions in the use of hospital EDs from 0.33 to 0.031 visits per month. In Kansas, after a policy was applied to provide primary care to medically underserved people, researchers found that ED visits by the uninsured declined 39% over the subsequent two years (Smith-Campbell, 2000). In another study, researchers estimated that a program that provides indigent patients with free primary care decreased ED utilization from 1.89 to 0.83 visits per year, and decreased charges for ED care by $457 per person (Davidson, Giancola, Gast, Ho, & Waddell, 2003). In Georgia, a study found that rural counties without a community health center (CHC) primary care clinic site had 33% higher rates of uninsured ED visits per 10,000 uninsured people than rural counties with a CHC (Rust et al., 2009). Finally, Young, D'Angelo, and Davis (2001) found that operating an inschool health center to provide primary care resulted in a significant decrease (p<0.03) in non-urgent ED use for the student population, from 44 visits to 26 visits per year for a sample of 216 students.
There is also evidence indicating that preventable hospitalization rates reflect inadequate ambulatory care. Parchman and Culler (1994) analyze hospital discharge data in Pennsylvania and find that areas with higher per capita rates of general practice physicians have lower rates of ambulatory care sensitive hospitalization, that is, hospitalizations that can be prevented with effective primary care which either prevents the onset of an illness or controls acute episodes of the disease.
Despite this evidence that the provision of primary care may lower utilization and costs of ED and inpatient services, there is little evidence to establish whether there are financial returns for investing in primary care. Therefore, the purpose of this study is to estimate whether the financial benefits of reducing the use and costs of hospital care outweigh the costs of investing in primary care.
To our knowledge, there is only one study that has examined the returns on investment of the provision of primary care. Oriol et al. (2009) examined the returns on investment of a mobile healthcare unit providing primary care and found that the community saved $36 for every $1 invested in the mobile healthcare unit. Oriol et al. (2009) compute the savings by multiplying the state average per visit preventable ED cost with the number of preventable ED visits they assume were avoided. To do this calculation, they have to make the strong assumption that if the sample population did not have access to the mobile van, 80% of van visits would have resulted in utilization of ED care. The strategy of this study differs from that employed by Oriol et al. (2009) in that we are able to estimate the change in non-urgent ED and inpatient hospital costs directly by comparing actual one-year prior to clinic enrollment costs with one-year post-enrollment costs. Moreover, we have all hospital costs so we can also estimate the savings due to the prevention of hospitalizations, as well as more accurately estimate the costs of primary care delivery by including other hospital costs, which are a potentially important part of the cost associated with the delivery of primary care. As a result, this study will provide much more accurate and comprehensive estimates of the costs and benefits of providing primary care in a free clinic setting.
Data for this analysis come from a large free clinic in northern Georgia. The clinic offers free medical care to the indigent, homeless and low-income people of their community, who have no health insurance and who cannot afford medical care. Hospital data come from a large regional hospital serving the same community as the clinic. Clinic data collection. The new clinic patient sample was chosen based on clinic enrollment date. All persons attending the clinic as a first-time patient between January 1st, 2006 and June 29th, 2007, were considered for inclusion in this study. Of the 289 patients identified from the clinic population, 18 clinic patients could not be linked to the hospital database as ever having received services, and therefore were excluded from consideration in this study. An additional 5 patients were dropped from this study because their clinic files could not be located, and 56 patients were dropped because they did not receive any clinic services in the one-year post-clinic enrollment (and therefore, the benefits of clinic enrollment may not have been realized). An additional 3 patients were dropped because of inpatient hospital admission length-of-stays that were outside of the 95% confidence interval for the sample (20 days, 40 days, and 49 days, respectively), and thus potential outliers. Therefore, the final sample size included in this study is 207. Clinic costs come from the 2007 clinic financial report.
Hospital data collection. Clinic patients were identified in the hospital's electronic medical records system based on the following clinic population identifiers: name, date of birth, gender, clinic enrollment date, and social security number (when available). Based on clinic enrollment data, all hospital data available for the clinic population were abstracted from the electronic medical records system for the one-year pre-clinic enrollment date to one-year post-clinic enrollment date. Hospital data categories for each patient included the following: ED services and their associated costs subcategorized as urgent or non-urgent, the total cost for inpatient visits, and the total cost of all other hospital services. Other hospital services included: ancillary care, diabetes education, imaging center services, hospice care, lab pathology, mental health services, mobile mammogram, outpatient chemotherapy, radiation oncology, outpatient rehabilitation, workers compensation rehabilitation, short-stay surgery, and wound repair. The total cost of a hospital service is defined as the sum of the variable cost, direct fixed cost, and overhead cost. All cost values were converted to 2009 U.S. dollars using the hospital and related services component of the Consumer Price Index (Bureau of Labor Statistics, 2009).
The purpose of this study is to compare the costs associated with providing primary care to the population under consideration with the savings to the hospital that result from having a free clinic in the region. The time frame for assessing costs and benefits was one year. The savings we include are derived from the decreased utilization of non-urgent ED care and inpatient hospital care, because clinic visits may be substitutes for non-urgent ED visits and may prevent hospitalization as discussed in the introduction. The primary care costs we include are the costs for clinic services, not including the value of volunteered services, and the costs of diagnostic and other hospital services associated with the delivery of that primary care. We add other hospital services to the cost side of the equation because we hypothesize that initial access to primary care for an uninsured population will involve the need for a variety of tests and services for their backlog of previously undiagnosed and untreated conditions.
We estimated hospital savings from the presence of the free clinic using a two-part model (Duan, Manning, Morris, & Newhouse, 1983; Cameron & Trivedi, 2005), which estimates separately the use of a hospital category (e.g. inpatient care) and the resulting costs conditional on any use. Specifically, we model the patient's decision to have a specific type of hospital service using a probit model. Then, we estimate the costs of the service using ordinary least squares (OLS) for the sample with non-zero costs. In both models, the regressors include patient's age, gender, race and ethnicity, a list of five conditions the patients have at enrollment (hypertension, diabetes, arthritis, depression, and/or asthma), and a post-enrollment indicator. (1) The expected hospital expense for each individual is estimated by multiplying the predicted value of the probability of any expense from the first regression and the predicted value of the expense.
The two-part model is used because hospital costs have an extremely skewed distribution, which make estimating with a simpler model inappropriate especially with a small sample. In particular, the frequency of some types of hospital care is small leading to a large number of zeros in the distribution, and the costs of care vary substantially such that the maximum expenditures can be high. This method also allows us to estimate the costs controlling for a variety of individual characteristics, such that the results are slightly more generalizable. We use this model to estimate the one-year costs pre-enrollment and one-year costs post-enrollment for non-urgent ED care and inpatient care. We can test for statistical differences between the pre- and post-enrollment costs using a standard t-test.
We estimate costs for the care of the 207 patients from the clinic by taking the average cost per visit from the financial report and multiplying that figure by the total number of visits incurred by the sample patients. We add to this figure the difference between the costs pre-enrollment and the costs post-enrollment for other hospital services. We estimated these hospital costs using the two-part model as described above. (2)
Table 1 provides some basic summary statistics on the patient population. The majority of patients are female, non-elderly adults. The clinic sees few children and seniors because most in these age groups are eligible for Medicaid or Medicare making them ineligible for services at the clinic. There is a large Hispanic clinic population, which is reflective of the community demographics. We can also observe whether a patient has certain health conditions from the clinic enrollment records. Thirty-five percent of clinic patients have hypertension, nearly a quarter has diabetes, and 11% has depression.
From the matched hospital records, we estimate that the costs associated with non-urgent ED visits declined by 25% and inpatient care costs declined by 15% (see Table 2). Both differences are statistically significant (p<0.01). The savings totals $35,146 for the 207 patients or $170 per patient.
From the clinic's 2007 financial report, we estimate that the average cost per visit is $54.44. The 207 new enrollees had 1,224 clinic visits in their first year. Thus, the total clinic expense for these 207 patients is $66,635, or $322 per patient. In addition, the hospital costs associated with the delivery of primary care to this population increased from $33,109 in the pre-enrollment period to $71,066 after clinic enrollment and represents a statistically significant increase. This amounts to an additional $37,957 in costs, or $183 per patient.
We find that, for a free clinic providing primary care services, the costs of non-urgent ED care and inpatient care decreased significantly post-clinic enrollment, by an estimated $170 per patient. However, there are important costs associated with that reduction that cannot be ignored: they include the costs of the clinic themselves and the costs of other hospital services for testing/diagnostics associated with that primary care, which total $505 per patient. Assuming that the high number of clinic visits and the increase in other hospital services is a temporary cost that falls after diagnosis and the initiation of disease management, the hospital savings per year must be sustained for at least 3 years to offset these initially high costs after enrollment.
Consistent with the literature, we find that the free clinic reduces non-urgent ED costs but our estimate of the net benefit is lower than that found by others (Oriol et al., 2009). This is due in part to the frequency of clinic visits needed to offset a non-urgent ED care visit. On the other hand, we find that the majority of the costs savings associated with clinic enrollment is derived from a reduction in the length of inpatient hospital stays.
There are several study design and data limitations to note. First, our study only included the utilization of hospital and clinic services for one-year post clinic enrollment. Thus, we must hypothesize about the future savings and costs associated with access to primary care services. We argue that a majority of savings may come in future years beyond the one-year post-clinic enrollment since the first year involves costly diagnosis and initiation of disease management.
Second, our benefit estimates do not include the health benefits resulting in improvements in health-related quality of life for the individual patients or their potential gains in productivity due to improved health. Further, our cost estimates do not include the value of volunteered services, which were not included in the clinic expense figures given above. Similarly, because we collected hospital costs directly from the hospital, we assume that the only physician costs that are included are from those physicians that are employed by the hospital. The costs associated with physicians not employed by the hospital, e.g. specialty care, are likely not included in the total hospital costs. Therefore our cost estimates are an underestimate from the societal perspective (where all costs would be included regardless to whom they accrue), but an accurate estimate from the free clinic and the hospital perspectives.
Third, if one considers enrollment in the free clinic as "the intervention," a more advanced evaluation analysis would include a control group that does not receive the intervention for comparison purposes. We were not able to conduct that type of study because we were analyzing data from persons who had already received the intervention (i.e., they had already enrolled in the clinic). Therefore, we could not make any random assignments to intervention or control retrospectively. Although it may have been possible to match to a control group not enrolled in the clinic, e.g., by analyzing healthcare utilization data for a similar population matched on age, gender, income, etc., these data were not available to us in the community. Thus, the pre-post design was the best method available for this analysis.
Fourth, there were 18 clinic patients dropped from our study that could not be matched to the hospital's electronic medical record database as ever having received services. For these patients, it is possible that 1) they did receive hospital services but could not be matched because of documentation status or other administrative issues, 2) they died or moved and therefore did not have continuous service within the community, 3) they received services in another hospital setting outside the community, or 4) they legitimately did not receive any hospital services. If the latter is correct and they should have been included in the study, this would impact our total sample size and decrease our cost per case estimates (although it would not impact the pre/post changes). For any of the other scenarios, it was not possible to correct for the bias that their exclusion may have caused. This potential bias speaks to one of the many challenges in doing research that attempts to capture all of an individual's healthcare costs within a fragmented healthcare delivery system.
Finally, 56 patients were excluded from this study because they did not receive any clinic services post-enrollment. These cases were dropped because we thought the benefits of clinic care could not be realized in one visit only. While these limitations may affect the results, they are such that the estimates reported are as conservative as possible in terms of estimated benefits compared to costs.
Recent health care reform legislation will greatly increase the number of previously uninsured people seeking access to primary care services. There will also continue to be a large number of immigrants and others without access to insurance. As such, communities, hospitals, and other healthcare providers will need to continue exploring new mechanisms for providing primary care services to vulnerable populations. The results of this study suggest that access to primary care will reduce the inefficient use of ED resources for non-urgent care and will reduce costly hospitalizations even in the short-run. However, as many are predicting, there will be initial costs associated with access to primary care for an uninsured population with an accumulation of health problems that have been undiagnosed and untreated.
Baker, L.C., Baker, L.S. (1994). Excess cost of emergency department visits for nonurgent care. Health Affairs, 13, 162-167.
Bureau of Labor Statistics. (2009). Consumer Price Index Detailed Report. Retrieved from http://www.bls.gov/cpi/tables.htm.
Cameron, C., Trivedi, P.K. (2005). Microeconometrics: Methods and Applications. New York: Cambridge University Press.
Davidson, R.A., Giancola, A., Gast, A., Ho, J., Waddell, R. (2003). Evaluation of access, a primary care program for indigent patients: Inpatient and emergency room utilization. Journal of Community Health, 28, 59-64.
Duan, N., Manning, W.G., Morris, C.N., Newhouse, J.P. (1983). A Comparison of alternative models for the demand of medical care. Journal of Business & Economic Statistics, 1(2), 115-126.
Grumbach, K., Keane, D., Bindman, A. (1993). Primary care and public emergency department overcrowding. American Journal of Public Health, 83, 372-378.
Hadley, J., Holahan, J., Coughlin, T., Miller, D. (2008). Covering the uninsured in 2008: Current costs, sources of payment, and incremental costs. Health Affairs, 27(5), w399-w415.
Oriol, N.E., Cote, P.J., Vavasis, A.P., Bennet, J., Delorenzo, D., Blanc, P., Kohane, I. (2009). Calculating the return on investment of mobile healthcare. BMC Medicine, 7, 27.
Parchman, M., & Culler, S. (1994). Primary care physicians and avoidable hospitalizations. Journal of Family Practice, 39, 123-128.
Russo, C.A., Jiang, H.J., Barrett, M. (2007). Trends in potentially preventable hospitalizations among adults and children, 1997-2004. HCUP Statistical Brief #36. Agency for Healthcare Research and Quality, Rockville, MD. Retrieved from http://www.hcupus.ahrq.gov/reports/statbriefs/sb36.pdf.
Rust, G., Baltrus, P., Ye, J., Daniels, E., Quarshie, A., Boumbulian, P., Strothers, H. (2009). Presence of a community health center and uninsured emergency department visit rates in rural counties. The Journal of Rural Health, 25(1), 8-16.
Smith-Campbell, B. (2000). Access to health care: Effects of public funding on the uninsured. Journal of Nursing Scholarship, 32, 295-300.
Young, T.L., D'Angelo, S.L., Davis, J. (2001). Impact of a school-based health center on emergency department use by elementary school students. Journal of School Health, 71, 196-198.
Zahradnik, A.G. (2008). Does providing uninsured adults with free or low-cost primary care influence their use of hospital emergency departments? Journal of Health and Human Services Administration, 31, 240-258.
(1) We conducted a Chow test to determine whether we could estimate the pre-enrollment and post-enrollment data in the same regression to increase precision and found that both periods could be combined for both non-urgent ED care and inpatient care costs.
(2) A Chow test indicated that we could not combine the pre-enrollment and post-enrollment data in the same regression in the case of other hospital costs. Thus, the pre and post costs were estimated separately and the indicator for post-enrollment was not used in these regressions.
ANGELA R. FERTIG
PHAEDRA S. CORSO
University of Georgia
St. Joseph's University
Table 1 Sample Characteristics Clinic Population (N=207) Percent Female 59% Average Age 44 Percent Age<18 5% Percent 45
|Gale Copyright:||Copyright 2012 Gale, Cengage Learning. All rights reserved.|