NBER WORKING PAPER SERIESDO URGENT CARE CENTERS REDUCE MEDICARE SPENDING?Janet CurrieAnastasia KarpovaDan ZeltzerWorking Paper 29047http://www.nber.org/papers/w29047NATIONAL BUREAU OF ECONOMIC RESEARCH1050 Massachusetts AvenueCambridge, MA 02138July 2021This research was supported by National Institute on Aging grant numbers P01AG005842 andP30AG012810. Dan Zeltzer acknowledges support from Israeli Science Foundation grant number1461/20. The views expressed herein are those of the authors and do not necessarily reflect theviews of the National Bureau of Economic Research.NBER working papers are circulated for discussion and comment purposes. They have not beenpeer-reviewed or been subject to the review by the NBER Board of Directors that accompaniesofficial NBER publications. 2021 by Janet Currie, Anastasia Karpova, and Dan Zeltzer. All rights reserved. Short sectionsof text, not to exceed two paragraphs, may be quoted without explicit permission provided thatfull credit, including notice, is given to the source.
Do Urgent Care Centers Reduce Medicare Spending?Janet Currie, Anastasia Karpova, and Dan ZeltzerNBER Working Paper No. 29047July 2021JEL No. I1,I11ABSTRACTWe examine the impact of the opening of a new urgent care center (UCC) on health care costsand the utilization of care among nearby Medicare beneficiaries. We focus on 2006–2016, aperiod of rapid UCC expansion. We find that total Medicare spending rises when residents of azip code are first served by a UCC, relative to spending in yet-to-be-served zip codes, whilemortality remains flat. We explore mechanisms by looking at categories of spending and byexamining utilization. Increases in inpatient visits are the largest contributor to the overallincrease in spending, rising by 6.65 percent within six years after UCC entry. The number ofemergency room visits that result in a hospital admission also increases by 3.7 percent. Incontrast, there is no change in the number of ER visits that do not result in admission to hospital,in visits to physicians outside a UCC, or in imaging and tests. Overall, these results provide littleevidence that UCCs replace costly ER visits or that they crowd out visits to patients' regulardoctors. Instead, the evidence is consistent with the possibility that UCCs—which areincreasingly owned by or contract with hospital systems—induce greater spending on hospitalcare.Janet CurrieDepartment of EconomicsCenter for Health and Wellbeing 185AJulis Romo Rabinowitz BuildingPrinceton UniversityPrinceton, NJ 08544and [email protected] KarpovaPrinceton [email protected] ZeltzerThe Eitan Berglas School of EconomicsTel Aviv UniversityP.O.Box 39040Tel Aviv [email protected]
I.IntroductionUrgent Care Centers (UCCs) have frequently been viewed as a way to curb health care costs(Weinick, Burns and Mehrotra, 2010; Allen, Cummings and Hockenberry, 2019; Merritt,Naamon and Morris, 2000; Corwin, Parker and Brown, 2016). If patients can be divertedfrom costly hospital emergency rooms (ERs) and treated at lower cost in UCCs, the potentialper-visit savings could be substantial. However, large health care systems are increasinglyoperating UCCs, and in other cases UCCs contract with health systems (Kaissi, Shay andRoscoe, 2016). Some commentators have expressed concern that in these cases UCCs areused as a way to funnel patients into hospital systems (Landro, 2016). As one hospitalexecutive has remarked: “The benefit to patients is getting them out of the ER so they canget treated most effectively and most efficiently for their non-life-threatening problems. . .And on the flip side, it’s provided us with some more feet through the door” (Experity,2016). If hospitals use affiliated UCCs to increase their shares of a fixed market, UCCs willnot necessarily affect health care utilization or costs. But if UCCs offer a means to growthe market by increasing provider-initiated demand for hospital services, UCC entry couldincrease overall costs with uncertain effects on health outcomes.This study examines the impact of the opening of a new UCC on health care costs andutilization of care among nearby Medicare recipients. We focus on the period 2006 to 2016,when the percentage of zip codes with a UCC serving Medicare patients grew from 28.3to 90.7. There are several advantages to focusing on Medicare, the public health insuranceprogram that covers elderly Americans. First, Medicare recipients accounted for 21 percentof total U.S. health care spending in 2019 (Martin et al., 2021). Second, we can observeall health care spending for Medicare recipients, which enables us to examine substitutionbetween UCCs and ERs, as well as the effects of UCCs on the use of physician’s offices andhospitals. Third, Medicare allows us to identify UCCs through their use of specific codes forplace of service.2
This last advantage is important, because previously it has been difficult to develop eithera uniform definition or a comprehensive longitudinal data base of UCCs. Previous studiesof the impact of UCCs are based on case studies rather than on the universe of UCCs.There are a number of private organizations that offer UCC accreditation, but because notall UCCs pursue accreditation it is not possible to compile a comprehensive list of facilitiesfrom these sources. Moreover, accrediting organizations typically only have lists of currentmembers. Hence, the Medicare claims data offer the most complete available picture of U.S.UCC activity.To estimate the impact of UCC entry, we use a difference-in-differences (DID) framework,comparing zip codes around the time they are first served by a UCC to zip codes that hadnot yet experienced such an entry. We use the estimator proposed by de Chaisemartin andd’Haultfœuille (2020a), which is suitable for staggered entry with potentially heterogeneoustreatment effects.Our results show that when the residents of a zip code first begin to be served by aUCC, total Medicare spending rises while mortality remains flat. Inpatient spending is thelargest contributor to this increase, rising by 153 (2006 dollars) per patient, or 4.2 percentby six years after the entry. Part D spending on prescription drugs, home health spending,and Part B drug spending also rise by 108 per capita (9.55 percent), 43 per capita (9.83percent) and 19 per capita (4.63 percent), respectively.1 There are also increases in thenumber of ER visits that result in hospital admission (3.71 percent).The percentage increase in elective inpatient visits (5.62 percent) is larger than the percentage increase in non-elective visits (3.6 percent). By six years after entry, about a quarterof the increase in elective inpatient spending is occurring within 90 days after a UCC visitwhile about half of the increase occurs within six months of a UCC visit. In the case ofnon-elective visits, which are typically more urgent than elective visits, 14 percent of theincrease occurs within two days of a UCC visit.1Part B drugs are drugs that are usually administered by a health care provider but not during aninpatient visit.3
Several important categories of spending are not affected by UCC entry. In particular,there is no evidence of an effect on spending for hospital outpatient care, visits to physiciansoutside a UCC, imaging and tests, or the number of ER visits that do not result in admissionto hospital. Hence, there is little evidence that UCCs are substituting for these services.On the whole, there is little evidence that UCCs function as a low-cost alternative forcostly ER visits or that they crowd out visits to patients’ regular doctors. Instead, theevidence is consistent with the possibility that UCCs induce greater spending on hospitalcare, especially elective services, and for both prescription and Part B drugs.The rest of the paper proceeds as follows. Section II provides some background aboutUCCs and prior work. Section III provides an overview of the data, and Section IV discussesour methods. Results appear in Section V, followed by a discussion and conclusions inSection VI.II.BackgroundAccording to the American Academy of Urgent Care Medicine (2021), most UCCs employphysicians. They typically have X-ray and laboratory facilities and are able to treat wounds,injuries, fractures, asthma attacks and mild concussions. They do not offer surgery orinpatient care, and most do not have advanced imaging equipment (e.g., CT scans). If apatient requires more advanced care, the UCC will transfer the patient to an ER. From thepatient’s point of view, the most salient features of a UCC are likely that they have longeropening hours than typical doctor’s offices, and that they welcome walk-in patients.While the first UCCs were owned by physicians, UCCs are increasingly either owned byhospitals and health systems or by corporations that contract with health systems. According to the Urgent Care Association (2018), in 2008 54.1 percent of UCCs were owned byphysicians, 24.8 percent were owned by hospitals, and the rest were owned largely by corporations, often with private equity partners. By 2014, physician ownership had fallen to 404
percent and direct hospital ownership had increased to 37 percent. According to the UrgentCare Association (2018), big hospital chains such as Dignity Health, HCA, Aurora Health,Intermountain Health, and Carolinas Healthcare have all made significant investments inUCCs. Yee, Lechner and Boukus (2013) report that in some markets the vast majority ofUCCs are owned by hospitals.These trends in ownership suggest that arrangements with hospitals are likely to havebecome increasingly important over time. While there is little direct evidence that ownershipof UCCs matters to patient spending (and indeed, data about ownership of UCCs are notcomprehensively available), related research about the ownership of physician practices suggests that patients in organizations owned by hospitals have higher spending than patientstreated in similar physician-owned practices, although there is no consistent difference inquality (Ho et al., 2020). Similarly, Chernew et al. (2021) show that physicians affiliatedwith hospitals refer patients in need of lower-limb MRIs to hospitals, though the imagingcould easily be done in an outpatient setting.Much of the existing literature argues that UCCs have the potential to greatly lowerhealth care costs. For example, Weinick, Burns and Mehrotra (2010) examine the types ofconditions that bring people to the ER and estimate that as many as 27.1 percent of ERvisits could be treated at UCCs or retail clinics, for a savings of up to 4.4 billion annually.Similarly, Allen, Cummings and Hockenberry (2019) find that in areas with multiple UCCs,local non-emergent ER visits increase by 1.43 percent (over the adjusted mean rate of 70.58percent) after UCCs close each day. This finding suggests that UCC visits can substitutefor ER visits. Merritt, Naamon and Morris (2000) tracked patients before and after theirfirst visit to a UCC and report that these patients were subsequently less likely to use ERs.Corwin, Parker and Brown (2016) examine a cross section of Medicare beneficiary data in2012 and find that ER use is lower in areas with high UCC use, and vice versa.However, a few recent observers found zero or positive effects of UCC entry on ER use.Yakobi (2017) examines ER use in New York City and notes that although by 2015 there5
were over 100 UCCs operating in the city, there appeared to have been no impact on ERuse. Carlson et al. (2020) examine patients who were using the ERs at two academic medicalcenters for low severity conditions. They find that patients who lived within one mile of anopen UCC were less likely to utilize one of the ERs, but that proximity to a UCC had noapparent effect on the use of the other ER. Wang, Mehrotra and Friedman (2021) analyzecommercial insurance data and find that zip-code level increases in the rate of urgent carevisits were associated with small reductions in ER visits, but with an overall increase inspending on urgent care services. In related work, Xu and Ho (2020) also report that theentry of new, free-standing emergency departments had little impact on visits to hospitalERs but served merely to increase the overall use of emergency services. These conflictingresults of various studies suggest that it will be informative to look in the Medicare data atthe effects of UCC entry.III.DataOur main source of data is the Medicare Fee-for-Service population (CMS, 2006–2016). Weuse a 20-percent sample of the Medicare Master Beneficiary Summary Files (MBSF) and itssegments for 2006 to 2016. These files include information about the patient’s utilizationof care and spending, demographic characteristics, chronic conditions, dates of Medicareenrollment, date of death (if relevant), and zip code of residence. For identifying UCC andphysician visits and for measuring associated spending, we also use the 20-percent sampleof the Carrier Files, which record fee-for-service claims submitted by professional providersincluding physicians, physician assistants, and nurse practitioners. For analyzing inpatientspending, we use the 20-percent sample of the Inpatient Files, which contain claims submittedby hospital providers.In order to define UCC entry into a zip code, we take advantage of the fact that since2003 the Centers for Medicare and Medicaid Services (CMS), which oversees Medicare, has6
designated a specific place of service code for urgent care facilities.2 Using this code, we candetermine the number of Medicare beneficiaries residing in a given zip code who used a UCCin each year. If the share of patients residing in a zip code who use a UCC in a particularyear is at least one percent, we consider that zip code to be served by a UCC in that yearand in subsequent years. We sometimes refer to this for short as “entry,” but it should beclear that the UCC serving patients in a particular zip code need not necessarily be locatedin that zip code. These definitions are described in further detail in the appendix.A patient-year is included in the sample if the patient was covered by Medicare Parts Aand B (which cover inpatient and outpatient care, respectively) for the entire year, or if thepatient died sometime during the year. In addition, the patient must have been enrolled infee-for-service Medicare for all 12 months and be 65 or older. We remove zip codes that hadfewer than 100 fee-for-service beneficiaries in at least one year, leaving 14,562 zip codes.Appendix Table A1 shows the rapid penetration of UCCs over our sample period. Between 2006 and 2016, the fraction of Medicare beneficiaries with a UCC that began servingtheir zip code increased from 29.3 percent to 92.2 percent. The fraction of beneficiariescovered is slightly higher than the fraction of treated zip codes over our sample period, indicating that zip codes with UCC entry are slightly larger than those without. AppendixTable A1 provides some support for our decision to ignore exits in our main results. In2016, the fraction of beneficiaries in a zip code currently served by a UCC is 87.1 percent.Comparing this figure to the percentage with an “entry” suggests that only 5.1 percent ofbeneficiaries lived in a zip code that experienced an exit. We show below that the resultsare quite similar whether we drop the zip codes with exits entirely, or treat an exit as a“negative” entrance. For these additional analyses, we define a zip code that experienced anexit as one in which the share of patients treated by a UCC dropped below one percent forat least two years in a row and did not rise again until the end of our sample period.Characteristics of patients, utilization of care and spending are shown in Table 1 for e of Service Code Set, accessed July 10, 2021.7
with and without UCCs in 2011, about halfway through our sample period, and in 2016, theend of this period. Spending is in 2006 US dollars throughout. The first panel of Table 1shows that the probability of UCC use, the number of UCC visits, and UCC spending are allan order of magnitude bigger in “treated” zip codes with UCC entry compared to “control”zip codes without such entry, as one would expect given our definition of entry.Panel B of Table 1 shows that total per capita spending was lower in treated zip codesin 2011, but somewhat higher in 2016. Treatment zip codes had lower mortality rates inboth periods. These differences could reflect differences in the areas where UCCs enteredover time. Previous research suggests that UCCs first entered wealthier urban areas andareas with higher rates of private insurance (Le and Hsia, 2016). One might well expect theeffect of UCC entry to be different in the richer areas where UCCs initially entered comparedto the areas where they entered later. In what follows, we use methods that are robust toheterogeneous treatment effects.Panel C of Table 1 shows the main categories of Medicare spending that we consider. Onecan see that the largest components of spending are (in order): inpatient visits; outpatienthospital visits; Medicare Part D drug benefits; and spending on skilled nursing facilities(SNF) and hospice. In each of these categories, per capita 2011 spending was lower intreated areas; however, by 2016, per capita spending on inpatient visits and Part D washigher in treated areas. As discussed above, one possible explanation for this crossover isthat poorer areas with less healthy populations were being added to the treatment groupover time. Spending on Part B drugs, imaging and testing, and visits to non-UCC physicianswere always higher in treatment areas despite their lower mortality rates, underscoring theconcern that there may be some unobserved differences between areas.Since inpatient spending is the largest category, Panels D–F of Table 1 break it downin several ways. Inpatient spending in acute care facilities (a category that includes mostgeneral-purpose hospitals) follows the pattern described above. Spending on non-electivevisits (i.e., visits that were relatively urgent) was lower in treatment areas in 2011 but8
became higher by 2016. Spending on elective visits was roughly even in treatment andcontrol areas.Because one concern about UCCs is that they may steer patients to hospitals, we breakdown inpatient spending on elective and non-elective visits further by considering only patients who visited a UCC in the past year or the previous year. Panel A shows that 6.3percent of patients in zip codes that were treated in 2016 visited a UCC during the year. Inthe two-year period 2015–2016, 10.2 percent of patients visited a UCC. Patients who visiteda UCC in either the current or the previous year spent on average 2313 on non-electiveinpatient admissions, nearly the same as the average beneficiary living in the same zip code( 2305). Of this spending, 8 percent ( 184) occurred within two days of a UCC visit and 11percent ( 259) occurred within seven days of a UCC visit. We see a similar pattern for elective visits: Patients who visited a UCC had spending of 1044 compared to an overall meanof 942. Because elective visits may be scheduled in a more leisurely way than non-electivevisits, we consider a longer time window and show that 23 percent ( 241) of the spendingoccurred within 90 days of a UCC visit and 39.6 percent ( 413) occurred within 180 days.Although they are only descriptive, these figures suggest that UCCs may be steering patientsto hospitals and that this possibility merits further investigation.Panel G of Table 1 shows comparable figures for ER visits. An ER visit can end in anadmission to the hospital or a discharge, so these two types of visits are further broken out.Zip codes with UCC entry had fewer ER visits on average than control zip codes (0.60 versus0.69 per beneficiary per year). This difference is concentrated in visits that did not result inan admission (0.40 vs 0.49 per beneficiary per year).Finally, the last panel shows the controls that are available. In addition to age and sex(which are constant over time and similar in treatment and control areas), we construct anapproximation to the Charlson Comorbidity Index (Charlson et al., 1987), which is frequentlyused to measure the burden of chronic disease in elderly people.3 The index is a weighted3Our approximation is based on the smaller set of chronic conditions that is available in MBSF File. Seethe appendix for the list of included conditions and their weights.9
sum of the number of chronic conditions where more severe conditions get higher weights.Table 1 shows that people are sicker in control areas than in treatment areas. The factthat the index rises over time in both treatment and control areas is consistent with UCCsmoving into areas where people are sicker: Moving the least sick patient from the controlto the treatment group would increase the mean level of sickness in both groups. In modelsexamining spending on drugs, we also control for the number of months of Part D coverage,since elderly people may move in and out of such coverage over time. See the appendix forfurther discussion of the definitions of the variables included in Table 1.IV.MethodsOur main analysis relies on a series of DID event-study analyses, examining the impactof UCC entry into a zip code on a range of outcomes, relative to yet-to-be-treated zipcodes. Several authors have shown that when treatment is staggered, standard DID estimatescan be biased in the presence of heterogeneous treatment effects (Goodman-Bacon, 2021;de Chaisemartin and d’Haultfœuille, 2020a,b). Sun and Abraham (2020) show that theestimated coefficients from two-way fixed effect regressions are not robust when there areheterogeneous treatment effects, while Borusyak and Jaravel (2017) show that in this caseit is also difficult to identify pre-trends. Since it is quite likely that UCCs will have differentimpacts in different areas, depending for example on how well served they are by other typesof providers, we use the estimator proposed by de Chaisemartin and d’Haultfœuille (2020a)which is unbiased in the presence of heterogeneous treatment effects.We compute dynamic treatment effects for each event-time l 0, where l 0 denotescontemporaneous treatment effects and l 0 denotes dynamic average treatment effects.This estimator corresponds to the average effect of the treatment on the treated patients(ATT), given the staggered entry design. Note that in keeping with the huge growth ofUCCs over our sample period and as discussed above, in our main results we ignore exits10
and assume that once a zip code is treated, it stays treated. Specifically, we estimate:TXβl "#XNz,tNz,t(Yz,t Yz,t l 1 ) (Yz,t Yz,t l 1 ) ,treatnon-treatNNt,lt,lz:τ t t lXωt,lt l 2z:τz(1)zwhere Yz,t is the average outcome for beneficiaries in zip code z and year t; τz is the year oftreatUCC entry in zip code z; Nz,t is the number of beneficiaries in zip code z and year t; Nl,tisnon-treatthe number of beneficiaries in zip codes treated for the first time in year t l and Nl,tis the number of beneficiaries in zip codes that had not been treated by year t. The weightsωt,l capture the relative size of the group of zip codes that had UCC entry in each year t fora fixed event-time l relative to all treated zip codes observed for event-time l.4 The term inbrackets is the DID estimator comparing the evolution of the outcome from period t l 1(the last period before the treatment) to t in groups treated for the first time in t l and ingroups which are yet to be treated at period t.We present plots with dynamic treatment effects for the six years following a UCC entry aswell as placebo pre-treatment effects for the six years preceding entry (with period 1 as thebaseline). We compute the placebo pre-period coefficients as proposed by de Chaisemartinand d’Haultfœuille (2020a) in order to be able to assess the parallel trends assumption.These coefficients are computed using the following equation:βlpre TXt l 2"preωt,lNz,t(Yz,t 2l 2 Yz,t l 1 ) treatNt,lz:τ t lz:τzNz,tnon-treatNt,lz tXX#(Yz,t 2l 2 Yz,t l 1 ) .(2)To incorporate the effects of control variables, we estimate a generalization of equation(2) in which (Yz,t and Yz,t l 1 ) are replaced by residuals from regressions of Y on X(the control variables) and year fixed effects. The control variables include: beneficiaryage, gender, and the approximated Charlson Comorbidity Index (see Section III for details).4treatThe weights are defined as follows: ωt,l Nt,l. PTt l 211treatNt,l.
Estimates of the effect of UCC entry on drug spending and utilization as well as on totalspending also control for months of Part D enrollment. The reason for controlling for Part Denrollment is that although sample participants must have been in Medicare Parts A and Bfor 12 months, many of them have fewer than 12 months of enrollment in Part D, the partof Medicare that covers prescriptions drugs.In addition to the DID event-study graphs, we summarize our findings in tables thatreport weighted averages of the dynamic estimates for the post-treatment effects from equation (1). The weights correspond to the size of the group of zip codes that had UCC entryand are observed for event-time l relative to all treated zip codes.5 Standard errors areclustered at the zip code level and are computed using 200 bootstrap replications.V.ResultsFigure 1 shows the “first stage” results: The entry of a UCC in a zip code increases theuse of UCCs by Medicare beneficiaries living in that zip code. This is of course a necessarycondition for UCC entry to affect other types of utilization of medical care and spendingand is implied by our definition of UCC entry. Figure 1a shows that the probability that abeneficiary uses a UCC is essentially zero prior to entry and then jumps to about 1.5 percentin the year of entry, rising smoothly to over 4 percent by six years after entry.6 Similarly,Figure 1b shows that the average number of UCC visits jumps to about 0.02 visits perperson after UCC entry, rising to 0.065 visits per enrollee after six years. Figure 1c shows acorresponding increase in Medicare spending on UCC services, rising from zero dollars priorto UCC entry to around 6 per enrollee by six years after the year of entry. These figuresdemonstrate that UCC entry into a zip code clearly increased utilization of UCCs among5They are defined as follows:βpost LXωl βl ,where ωl l 06TXt l 2treatNt,lL XT.XtreatNt,l.l 0 t l 2Pre-entry UCC use may reflect occasional use by zip code residents of UCCs located in other areas.12(3)
Medicare beneficiaries living in that zip code, by all measures.Figure 2 shows DID estimates for the impact of UCC entry on total per capita Medicarespending and mortality. Figure 2a indicates that total spending per recipient began to riseafter UCC entry, increasing by almost 300 per person by six years after entry. Figure 2bshows that there is no corresponding change in mortality rates—death rates are unchangedfollowing UCC entry. Since the increase in expenditure is much higher than the amountspent directly on UCCs and has no effect on mortality, it begs the question of what categoriesaccount for the additional, apparently non-productive, spending?This question is addressed in Figure 3, which shows the four categories with the greatestincreases in spending. Figure 3a shows that inpatient spending accounts for the largest shareof the increase, rising by 153 per person by six years after entry (this increase amounts to4.23 percent of its baseline 2006 average). Figure 3b shows that there is also an increasein spending on Medicare Part D, the prescription drug benefit, of 108 (9.55 percent) aftersix years, suggesting that patients are receiving more prescriptions, or prescriptions for moreexpensive drugs, after a UCC opens. This observation is consistent with concerns about UCCover-prescription of drugs such as antibiotics (Laude et al., 2020; Urgent Care Association,2019; Incze, Redberg and Katz, 2018). Figure 3c shows that UCC entry also led to modestincreases in home health spending of approximately 43 per patient after six years. Althoughthere does not seem to be much research on this subject, it is possible that the availability ofadditional options for receiving both regular doctor visits and urgent care encourages someseniors to remain in their homes and therefore to make greater use of home health services.Finally, Figure 3d shows that there was also an increase in spending on Part B drugs ofabout 19 per person after six years. These are drugs that a patient wouldn’t usually givethemselves (such as injectable and infused drugs that are usually given in an outpatientsetting) but which could be administered in a setting such as a UCC.Figure 4 shows DID graphs for several other important categories of services that onemight expect to be impacted by UCC entry. Figure 4a shows that hospital outpatient13
spending was on a declining trend prior to UCC entry. Upon entry, spending flattens for a fewyears but then continues a downward trend. Figure 4b shows that spending on imaging andlab tests declined fairly smoothly through UCC entry (though perhaps with some flatteningright around entry). This may be in part because most UCCs do not have the most expensiveimaging equipment such as MRIs and CT scans. Figure 4c suggests that there was a smalldecline in spending on skilled nursing facilities four to six years after UCC entry, perhapsconsistent with the increase in spending on home health care noted above given that thesemay be substitutes.Figure 4d shows that visits to non-UCC physicians continued on a smooth, slightly upward trend through UCC entry, suggesting that there is little evidence of substitution ofUCC visits for non-UCC outp
While the rst UCCs were owned by physicians, UCCs are increasingly either owned by hospitals and health systems or by corporations that contract with health systems. Accord-ing to the Urgent Care Association (2018), in 2008 54.1 percent of UCCs were owned by physicians, 24.8 percent were owned by hospitals, and the rest were owned largely by corpo-File Size: 403KBAuthor: Janet Currie, Anastasia Karpova, Dan ZeltzerPage Count: 40