Association of Private Equity Firm Acquisition of Ophthalmology Practices with Medicare Spending and Utilization of Ophthalmology Services

Abstract

Purpose:

Private equity (PE) firms are increasingly acquiring ophthalmology practices; little is known of their influence on care utilization and spending. We studied changes in utilization and Medicare spending after PE acquisition.

Design:

Retrospective cohort study.

Participants:

762 clinicians in 123 practices acquired by PE between 2017 and 2018; 34,807 clinicians in 20,549 never-acquired practices.

Methods:

We analyzed Medicare fee-for-service claims (2012–2019) combined with a novel national database of PE acquisitions of ophthalmology practices using a difference-in-differences method within an event study framework to compare changes after a practice was acquired to changes in practices that were not acquired.

Main Outcome Measures:

Numbers of beneficiaries seen; intravitreal injections and medications used for injections; spending on ophthalmologist and optometrist services, ancillary services, and intravitreal injections.

Results:

Comparing changes over time for PE-acquired to non-acquired practices, there was a relative increase in beneficiaries seen per quarter per PE optometrist of 23.92% (4.20 beneficiaries, 95% CI, 1.73 to 6.67); no change for ophthalmologists; and a relative increase in spending per beneficiary of 5.06% ($9.66, 95% CI, −2.82 to 22.14). Spending on clinician services decreased 1.62% ($−2.37, 95% CI, −5.78 to 1.04), including a 5.46% ($17.70, 95% CI, −2.73 to 38.15) increase in per beneficiary per quarter spending on ophthalmologist services and a 4.60% ($−5.76, 95% CI, −9.17 to −2.34) decrease for optometrists. Ancillary services spending decreased 7.56% (-$2.19, 95% CI 4.19 to −0.22); intravitreal injections spending increased 25.0% ($20.02, 95% CI, −1.38 to 41.41). The number of intravitreal injections increased 5.10% (1.83, 95% CI, −0.1 to 3.80), including a 74.09% (8.38 injections, 95% CI, 0.01 to 16.74) in use of an expensive medication (ranibizumab) and a 12.91% decrease (−3.40 injections, 95% CI −6.86 to 0.07) for an inexpensive injection (bevacizumab). The event study showed consistent and often statistically significant increases in ranibizumab injections and decreases in bevacizumab injections after acquisition.

Conclusions:

Though not all results reached statistical significance, this study suggests that PE-acquired practices had little or no overall effect on utilization or total spending, but increased the number of unique patients seen per optometrist and the use of expensive intravitreal injections.

Précis

Compared to non-acquired practices, private equity-acquired practices had a 74% increase in intravitreal ranibizumab injections and a 13% decrease in intravitreal bevacizumab. The event study results show steady and significant changes over time in the usage of ranibizumab and bevacizumab injections. Other findings were mixed but suggested increased spending and utilization in PE-acquired practices.

INTRODUCTION

During the past decade, private equity firms have increasingly acquired physician practices,¹ especially in dermatology,^2,3 gastroenterology,^4,5 and ophthalmology.⁶ These specialties are particularly attractive because they are fragmented (mainly small and medium-sized practices), provide profitable procedural, ancillary, and self-pay services (e.g. cosmetic services), and are likely to see growth in demand for services as the U.S. population ages.⁷

From 2012–2019, private equity (PE) firms acquired 228 ophthalmology and optometry practices in the U.S., with most of the acquisitions occurring between 2017 and 2019.⁸ Typically private equity firms acquire majority ownership of a “platform practice” within a market – a practice that is well-managed and of at least moderate size – and then acquire additional practices to consolidate with this platform.

Practices are purchased via leveraged buyouts, in which most of the purchase price is provided by lenders rather than by the PE firm; the practices – not the firm – are responsible for paying back the loan. Private equity firms seek to generate average annual returns in the range of 20% for their investors. This is anticipated to come from management fees paid by practices to the PE firm, from the PE firm’s share of practices’ profits, which the firm seeks to increase, from the increased value of smaller practices when they are added to a platform practice, and from sale of practices (typically within 3–5 years) at prices anticipated to be much higher than their original purchase price. Physicians retain minority ownership of acquired practices, are paid large sums for the practice share acquired by private equity, have their annual income from the practice reduced (though additional incentive-related income is available), and receive a share of the purchase price when the practice is sold again.

Private equity firms argue that they provide professional management, capital for growth and for improving quality and efficiency, and more autonomy (and possibly a higher purchase price) than practices would have if acquired by a hospital.^9–16 Critics argue that PE firms – and therefore the acquired practices – are under such pressure to generate profits that quality of care may suffer, prices, utilization, and spending on care may increase nationally, and physicians in acquired practices may not be permitted to adequately advocate for their patients.^{9,13,14,17–21}

There is little empirical evidence on the impact of private equity acquisition of practices on the quality, cost, or utilization of care. In this paper, we provide evidence on changes in the utilization and cost of care after private equity acquisition of ophthalmology and optometry practices (henceforth called “ophthalmology practices”). Private equity firms have very strong incentives to generate high net revenues from practices that they acquire, so we hypothesized that, after PE acquisition, there would be an increase in the number of beneficiaries seen per clinician and in per beneficiary Medicare spending on ophthalmology services, ophthalmology-related ancillary services, and profitable injectable drugs, compared to practices that were not acquired.

METHODS

Study Design

We created a national database of PE acquisitions of ophthalmology practices, identified the clinicians (ophthalmologists and optometrists) in these practices, and merged this database with Medicare claims data. We used cross-sectional methods to compare PE-acquired practices to practices that were never acquired, before and after acquisition. We used difference-in-difference analysis to compare changes in spending and utilization over time in PE-acquired vs. never acquired practices, and event study methods to show changes over time.

Data Sources

Using similar methods to our previous research, we identified ophthalmology practices that were acquired by a PE firm between 2017 and 2018. We chose these years because relatively few ophthalmology practices were acquired prior to 2017 and because it is necessary to have at least one year post-acquisition to measure outcomes. We used the S&P Capital IQ and Irving Levin databases, which report acquisitions, including the announcement date, the name of the acquired practice, the platform practice that the PE firm used to acquire a practice, and the PE firm that owns the platform practice.

We used Web searches to identify the clinician owner(s) of the acquired practices and owners’ National Provider Identifiers (NPIs). The owners’ NPI was used to identify the practice’s Tax Identifier Number (TIN) and all other clinicians (ophthalmologist or optometrist) that billed under the TIN of an acquired practice using the Medicare Data on Provider Practice and Specialty (MD-PPAS) file.

We merged the PE-TIN dataset with a 20% random sample of Medicare fee-for-service claims from 2012 to 2019.

Study Population

The clinician sample was constructed using the following criteria (eTable 1, available at https://www.aaojournal.org): First, we included only practices that were primarily ophthalmology and/or optometry, excluding TINs in which less than 75% of clinicians were ophthalmologists or optometrists. Second, if a clinician billed under multiple practices in a Hospital Referral Region (HRR) during a quarter, that clinician was counted once during that quarter. If a PE clinician billed under a PE and non-PE TIN in a given quarter, they were attributed to the PE TIN, and non-PE TIN observations were excluded from the treatment (PE) and control (non-PE) groups in that quarter. No physician appeared in multiple HRRs or practices during a given quarter.

Third, once a physician was attributed to a PE practice, they were included in the PE (treatment) group in all subsequent quarters, even if they billed under a non-PE TIN in a subsequent quarter. Fourth, since acquired practices sometimes switch to the TIN of the PE-acquired platform practice, PE clinicians were assigned to their pre-acquisition TIN for all subsequent years to separate clinicians in acquired practices from clinicians in the platform practice that were not originally part of the acquisition. Fifth, clinicians who never billed a PE TIN and billed multiple non-PE TINs within a quarter were attributed to the TIN under which they submitted the most claims during a quarter. Sixth, if a practice was reacquired by a PE firm during our study period, we considered only the first acquisition in the treatment group. Practices reacquired in 2012 or later after being acquired prior to 2012 were excluded. Seventh, ophthalmologists that did not perform an intravitreal injection in a given quarter were excluded in the analysis of intravitreal injections.

STUDY VARIABLES, 2011–2019

Outcomes

Our outcomes included per clinician-quarter total ophthalmology-related per beneficiary spending, spending on clinician services, spending on ancillary services, spending on intravitreal injections, spending on ophthalmology services, spending on optometry services, number of unique beneficiaries seen by ophthalmologists and by optometrists, the number of intravitreal injections per 100 beneficiaries, and the medication used for these injections.

We measured total spending as the amount allowed by Medicare to be paid for ophthalmology and optometry services, including spending for in-office services including supplies, medications, procedures, diagnostic tests, and clinician visits, using CPT codes billed under each clinician’s NPI (eTable 5). We used Medicare’s Geographic Variation Public Use File - by National, State & County to adjust spending for geographic differences in input costs, such as rent and labor.

STATISTICAL ANALYSIS

We conducted two primary analyses. First, using Student t-tests for continuous variables, we compared practice, clinician, and beneficiary characteristics and utilization and spending outcomes for PE and non-PE practices in 2012 (pre-acquisition) to determine whether practices targeted for acquisition were different from non-PE practices and in 2019 to determine whether differences persisted.

Second, we used a difference-in-differences within an event-study approach to compare changes, between 2012 and 2019, in utilization and spending, within an HRR that included at least one acquired practice, between clinicians in practices acquired in 2017 and 2018 to changes for clinicians in practices that were never acquired.

Regressions included controls for clinician fixed-effects and quarter-by-year fixed-effects, and an interaction between whether a practice was acquired and a post-acquisition time indicator, which represented the impact of PE acquisition on outcomes. Given concerns of bias in settings with multiple treatment periods that began at different times, we used the difference-in-difference method proposed by Callaway et al.²² To achieve improved balance in our sample, clinicians must have been in a practice in at least four quarters before, during, and after acquisition. Outcomes were Winsorized at the top and bottom 1% to remove outliers. All analyses included robust SEs clustered at the practice level and were conducted using Stata, version 17.0 (StataCorp LLC). P-values were 2-tailed, and statistical significance was defined as P < 0.05. P-values were adjusted for multiple comparisons using the Benjamini-Hochberg method.

A main assumption required for the difference-in-differences is that pre-PE acquisition differences between the treatment and control clinicians would remain constant in the absence of a PE acquisition. To test this assumption, and to show changes over time, we conducted an event-study analysis. The event study framework is like a traditional difference-in-differences model, as the main variable of interest is the interaction between the treatment variable (acquisition by PE) and an indicator for the post-acquisition period. We graphed the coefficients from the interaction terms for each quarter relative to the time of PE acquisition and an indicator whether a practice was acquired by PE to test for preexisting trends.

We assessed the sensitivity of the results to (1) restricting our sample to clinicians who billed only a single TIN during the sample period, (2) restricting our sample to clinicians who were present in all quarters of the sample period, and (3) using an interaction of state and a continuous measure of year to control for state-specific linear time trends.

The institutional Review Board at Weill Cornell Medical College approved this study.

RESULTS

Practice, Clinician, and Beneficiary Characteristics

In 2012, before acquisition, there were 539 clinicians in 115 practices that were later acquired by PE and 27,198 clinicians in 17,474 practices never acquired (Table 1). In 2019, there were 757 clinicians in 123 PE acquired practices, and 26,803 clinicians in 16,174 never acquired practices. During the entire study period (2012–2019), there was a total of 762 clinicians (464 ophthalmologists and 298 optometrists) in 123 practices (33 in 2017 and 90 in 2018) acquired by PE (the treatment group) and 34,807 clinicians (11,649 ophthalmologists and 23,158 optometrists) in 20,549 never-acquired practices (control). In 2012, before any acquisitions, PE practices had 342 ophthalmologists, 95 of whom were retinal specialists. In contrast, there were 10,238 ophthalmologists not associated with PE, including 1,812 retinal specialists. After acquisitions in 2019, the number of ophthalmologists in PE practices rose to 757, with 135 retinal specialists. Non-PE ophthalmologists totaled 9,051, including 1,942 retinal specialists.

Table 1.

Characteristics of Practices, Clinicians, and Patients

	2012 [Pre-acquisition]a			2019 [Post-acquisition]			Difference in the change for PE and non-PE practices between the pre- and post-acquisition periodsb
	PE-acquired practices	Non-PE-acquired practices	Unadjusted difference [95% CI]	PE-acquired practices	Non-PE-acquired practices	Unadjusted Difference [95% CI]	Unadjusted difference [95% CI]	P Value
Practice characteristics c,d
Practices [TINs], No.	115	17,474	-	123	16,174	-	-	-
Clinicians in sample, No.e	539	27,198	-	757	26,803	-	-	-
Ophthalmologist in sample, No. (%)	342 (63.45)	10,238 (37.64)	-	462 (61.03)	9,051(33.77)	-	-	-
Optometrists in sample, No. (%)	197 (36.55)	16,960 (62.36)	-	295 (38.97)	17,752 (66.23)	-	-	-
Retinal Specialist in sample, No. (%)f	95 (27.78)	1,812 (17.70)	-	135 (29.22)	1,942 (21.46)	-	-	-
Mean practice size
Ophthalmologists, mean	3.04	0.60	2.44 [2.18 to 2.75]	3.99	0.60	3.39 [1.14 to 5.64]	0.95 [0.60 to 1.35]	<.001
Optometrists, mean	1.70	0.99	0.71 [0.50 to 0.92]	2.60	1.13	1.47 [1.18 to 1.76]	0.76 [0.28 to 1.54]	0.001
Retinal Specialist, mean	0.82	0.10	0.64 [0.52 to 0.76]	1.09	0.08	0.82 [0.70 to 0.94]	0.18 [0.04 to 0.32]	0.011
Optometrists and ophthalmologists, meang	4.69	1.56	3.13 [2.84 to 3.43]	6.15	1.70	4.45 [3.50 to 5.40]	1.32 [1.10 to 1.54]	<.001
Clinician characteristics
Clinician age, mean	47.40	51.68	−4.28 [−4.84 to 3.79]	51.07	53.14	−2.07 [−2.51 to - 1.62]	2.21 [1.56 to 2.94]	0.05
Male, %	80.35	75.81	4.54 [2.61 to 6.47]	75.40	68.73	6.67 [5.03 to 8.29]	2.13 [−0.40 to 4.65]	0.10
Patient characteristics
Dual-eligible, %	12.91	18.53	−5.62 [−9.34 to −1.89]	10.59	14.78	−4.20 [−7.74 to - 0.64]	1.43 [−3.71 to 6.56]	0.59
Female, %	61.85	60.52	1.33 [−1.01 to 3.66]	60.04	59.84	0.2 [−2.03 to 2.43]	−1.13 [−4.25 to 2.20]	0.49
Age, years	74.70	73.85	0.85 [0.64 to 1.06]	75.93	74.95	0.98 [0.80 to 1.10]	0.13 [−0.15 to 0.41]	0.36

^aPE-acquired practice include all practices acquired in 2017 or 2018. Never PE-acquired practices include all practices that were never acquired by PE as of 2019.

^bComparison of practice, clinician, and patient characteristics of pre-acquisition practices with post-acquisition practices used t-tests.

^cThere were 539 clinicians in 2011 and 757 clinicians in 2019 who were in practices acquired by PE. From 2011 to 2019 were a total of 762 unique clinicians who were in practices acquired by PE firms. There was a total of 144 PE ophthalmologist that were retinal specialist, of which 95 were included in 2012 and 135 in in 2019.

^dIn 2012, 115 practices were present that were later acquired by PE. By 2019, PE had acquired 123 practices, including 8 additional practices that were not present in the pre-acquisition period and were acquired and appeared in our sample from 2013 to 2018.

^eClinicians include ophthalmologist and optometrists.

^fOf the 342 ophthalmologist in PE-acquired practices in 2012, 95 were retinal specialist. Of the 462 ophthalmologists in PE practices in 2019, 135 were retinal specialists.

^gMean does not equal the sum of means of ophthalmologists, optometrists, and retinal specialists due to rounding.

In 2012, practices later acquired by PE were larger than never-acquired practices – 4.69 total clinicians (3.04 ophthalmologists and 1.70 optometrists) compared with 1.56 clinicians in non-PE practices. Practices that were later acquired had more retinal specialists per practice than never-acquired practices (0.82 and 0.10, respectively). In 2019, post-acquisition, the mean number of total clinicians (6.15), ophthalmologists (3.99), optometrists (2.60), and retinal specialists (1.09) per practice increased and remained higher in PE practices compared to non-PE practices (1.70, 0.60, 1.13, 0.08 respectively).

In 2012, prior to acquisition, PE clinicians saw a lower percentage of dual-eligible beneficiaries (patients that have Medicare and Medicaid insurance) (12.91% vs 18.53% compared to non-PE practices. In 2019, after acquisition, these percentages decreased to 10.59% and 14.78% respectively.

Before PE acquisition, PE clinicians had higher total spending ($324.86 vs $187.43 per beneficiary per quarter), clinician services spending ($198.73 vs. $143.30), ancillary spending ($46.54 vs. $28.51), and intravitreal injection spending ($115.32 vs $30.85) compared to non-PE physicians (Table 2). They saw more beneficiaries per quarter (109.57 compared to 80.66 and 30.87 compared to 17.36 respectively) and performed a higher number of intravitreal injections per 100 beneficiaries per quarter (48.79 vs. 35.85).

Table 2.

Unadjusted Changes in Spending and Utilization After Private Equity Acquisition Compared to Practices Never Acquired

	2012 [Pre-Acquisition]			2019 [Post-Acquisition]			Unadjusted Differences Pre- and Post-Acquisition Periods
Outcomes	PE-acquired practices	Non-PE-acquired practices	Unadjusted difference [95% CI]	PE-acquired practices	Non-PE-acquired practices	Unadjusted difference [95% CI]	Unadjusted differencee [95% CI]	P Value
Total spending per beneficiary a	324.86	187.43	137.43 [127.61 to 147.25]	400.69	221.35	179.34 [171.01 to 187.65]	41.91 [29.39 to 54.42]	<.001
Spending on clinician services	198.73	143.30	55.43 [51.68 to 59.17]	188.39	146.79	41.60 [38.43 to 44.77]	−13.82 [−18.60 to −9.06]	<.001
Spending on ancillary services	46.54	28.51	18.03 [16.66 to 19.39]	37.93	26.03	11.90 [10.74 to 13.07]	−6.13 [−7.86 to - 4.37]	<.001
Spending on intravitreal injectionsb	115.32	30.85	84.47 [69.10 to 99.85]	273.26	126.66	146.60 [113.11 to 160.09]	62.13 [42.86 to 81.39]	<.001
Number of unique patients per clinician a
Number of unique patients for ophthalmologists	109.57	80.66	28.91 [24.74 to 33.10]	160.29	132.34	27.95 [24.24 to 31.66]	−0.96 [−6.21 to 4.29]	0.72
Number of unique patients for optometrists	30.87	17.36	13.51 [11.40 to 15.62]	61.63	30.36	31.27 [29.54 to 33.0]	17.76 [15.09 to 20.44]	<.001
Number of intravitreal injections c
Number of Injections Per 100 Beneficiaries	48.79	35.85	12.94 [9.24 to 16.63]	63.44	48.91	14.53 [11.68 to 17.78]	1.59 [−2.41 to 6.00]	0.403
Aflibercept Injectiond	17.58	11.67	5.91 [2.28 to 9.54]	55.11	40.83	14.28 [11.06 to 17.49]	8.37 [4.15 to 12.59]	<.001
Ranibizumab Injection	22.45	10.60	11.85 [8.41 to 15.27]	41.49	20.18	21.31 [18.48 to 24.14]	9.46 [5.56 to 13.37]	<.001
Bevacizumab Injection	31.64	25.99	5.65 [2.04 to 9.23]	26.37	28.50	−2.13 [−5.09 to 0.83]	−7.78 [−11.86 to - 3.68]	<.001

^aAll results are per clinician per quarter. The sample consists of 748,796 clinician-quarter (ophthalmologist and optometrists) observations, containing 762 PE clinicians and 32,807 non-PE clinicians. Spending measures are geographically adjusted using Medicare’s Geographic Variation Public Use File – by National, State, and County to adjust spending for geographic differences in input costs, such as rent and labor.

^bIncludes spending for the medication used and for the ophthalmologist service [that is, giving the injection]. Includes only ophthalmologists.

^cIncludes only ophthalmologists who gave at least one intravitreal injection [Afliberecpt, Ranibizumab, or Bevacizumab] in that quarter. Intravitreal injection sample consists of 58,749 ophthalmologist-quarter observations, containing 144 PE ophthalmologists and 2,345 non-PE ophthalmologists who provided at least one intravitreal injection in a given quarter.

^dAfliberecpt was approved for use on November 11, 2011. Therefore, we use the observations from the year 2013 to calculate the pre-acquisition period mean.

^eSample consists of clinicians who were present for at least 4 quarters before and after PE acquisition

Differential Changes in Clinician Outcomes Associated with PE Acquisitions

Table 3 presents adjusted difference-in-differences results, which compare changes over time in PE-acquired practices to those in non-PE practices. Results are expressed as the relative increase or decrease and as the percentage change of this relative increase or decrease compared to the pooled mean (across PE and non-PE practices) of the outcome variable in 2012.

Table 3.

Adjusted Differential Changes in Spending and Utilization Associated with PE Acquisitions, 2011–2019

		Differences-in-differences adjusted estimatei
Outcome	Unadjusted mean, 2012j	Average treatment effect on treated [95% CI]k	Relative changel	P value	Adjusted P valuem
Total spending per beneficiarya	190.74	9.66 [−2.82 to 22.14]	5.06%	0.129	0.172
Spending on clinician services	145.71	−2.37 [−5.78 to 1.04]	−1.62%	0.174	0.209
Spending on ancillary servicesb	28.95	−2.19 [−4.19 to −0.22]	−7.56%	0.029	0.116
Spending on intravitreal injectionsc,d	80.07	20.02 [−1.38 to 41.41]	25.00%	0.067	0.116
Total spending per beneficiary for ophthalmologistse	316.88	17.70 [−2.73 to 38.15]	5.46%	0.088	0.209
Total spending per beneficiary for optometristsf,g	125.26	−5.76 [−9.17 to −2.34]	−4.60%	0.001	0.006
Number of unique patients per cliniciana
Number of unique patients per ophthalmologist	81.83	−1.07 [−6.19 to 4.04]	−0.11%	0.681	0.743
Number of unique patients per optometrist	17.56	4.20 [1.73 to 6.67]	23.92%	0.001	0.006
Number of intravitreal injectionsh
Number of Injections per 100 Beneficiaries	35.85	1.83 [−0.1 to 3.80]	5.10%	0.068	0.116
Number of Afliberecpt Injectionsd	12.51	0.69 [−4.07 to 5.45]	5.52%	0.854	0.854
Number of Ranibizumab Injections	11.31	8.38 [0.01 to 16.74]	74.09%	0.050	0.116
Number of Bevacizumab Injections	26.33	−3.40 [−6.86 to 0.07]	−12.91%	0.055	0.116

^aAll results are per clinician per quarter. Spending measures are geographically adjusted. The sample consists of 748,796 clinician-quarter (ophthalmologist and optometrists) observations, containing 762 PE clinicians and 32,807 non-PE clinicians.

^bAncillary services include services not directly provided by physicians, such as “scanning computerized ophthalmic diagnostic imaging, posterior segment”.

^cIncludes spending for the medication used and for the ophthalmologist service [that is, giving the injection]. Includes only ophthalmologist.

^dAfliberecpt was approved for use on November 11, 2011. Therefore, we use the observations from the year 2013 to calculate the pre-acquisition period mean.

^eIncludes all spending billed under an ophthalmologist national provider identification number [NPI]. See Appendix for spending subcategories for ophthalmologists and optometrists. Ophthalmologist sample consists of 316,482 ophthalmologist-quarter observations, containing 464 PE ophthalmologists and 11,649 non-PE ophthalmologists

^fIncludes all spending billed under an optometrist national provider identification number [NPI]. See Appendix for spending subcategories for ophthalmologists and optometrists. Optometrist sample consists of 432,314 optometrist-quarter observations, containing 298 PE optometrists and 21,158 non-PE optometrists

^gMean spending for ophthalmologists and optometrists does not add to mean total spending because there are different numbers of ophthalmologists and optometrists.

^hIncludes only ophthalmologists who gave at least one intravitreal injection in that quarter. Intravitreal injection sample consists of 58,749 ophthalmologist-quarter observations, containing 144 PE ophthalmologists and 2,345 non-PE ophthalmologists who provided at least one intravitreal injection in a given quarter.

ⁱCallaway and Sant’Anna difference-in-difference estimator used. All models included clinician and year fixed effects. Standard errors were adjusted for clustering at the practice [TIN] level.

^jUnadjusted mean is the pooled mean of both PE and non-PE clinicians in pre-acquisition period (2012).

^kSample consists of clinicians who were present for at least 4 quarters before and after PE acquisition.

^lRelative changes were derived from the sample by dividing the adjusted estimates for all outcomes by the unadjusted mean for the outcomes in the pre-acquisition period [2012].

^mBenjamini-Hochberg p-value adjustment was used for multiple comparisons.

After PE acquisition, the relative increase in number of beneficiaries seen per quarter by PE optometrists was 23.92% (4.20 beneficiaries, 95% CI, 1.73 to 6.67); there was no change for ophthalmologists.

The relative increase in spending per beneficiary was 5.06% ($9.66, 95% CI, −2.82 to 22.14). Spending on clinician services decreased 1.62% ($−2.37, 95% CI, −5.78 to 1.04), including a 5.46% ($17.70, 95% CI, −2.73 to 38.15) increase in spending on ophthalmologist services and a 4.60% ($−5.76, 95% CI, −9.17 to −2.34) decrease for optometrists. The relative decrease in ancillary services spending was 7.56% (−$2.19, 95% CI 4.19 to −0.22).

Spending on intravitreal injections increased 25.0% ($20.02, 95% CI, −1.38 to 41.41). The relative increase in the number of injections per 100 beneficiaries per physician-quarter-year was 5.10% (1.83, 95% CI, −0.1 to 3.80). The relative increase for the most expensive medications was 74.09% (8.38 injections, 95% CI, 0.01 to 16.74) for ranibizumab and 5.52% (0.69 injections, 95% CI, −4.07 to 5.45) for aflibercept; injections for the least expensive medication (bevacizumab) decreased by 12.91% (−3.40 injections, 95% CI −6.86 to 0.07).

The event studies show the changes in outcome variables for PE-acquired practices by quarter. All event study models indicated that the coefficients in the pre-acquisition period did not differ significantly between the treated and control clinicians, demonstrating that the parallel trends assumption fundamental to difference-in-differences analyses was not violated. The event study results are consistent with the differences-in-differences analyses (Figures 1 and 2; eFigures 1–3, available at https://www.aaojournal.org). In general, they suggest increasing differences between PE and non-PE practices over time, though confidence intervals are large in the later quarters, due to smaller sample size.

Figure 1.

Event Study: Changes in Total Spending for Ophthalmologists and Optometrists After PE Firm Acquisition Compared with Ophthalmologists and Optometrists in Practices Never Acquired, 2012–2019.

All results are per clinician per quarter. Spending measures are geographically adjusted. Includes all spending billed under an ophthalmologist national provider identification number [NPI]. Ophthalmologist sample consists of 316,482 ophthalmologist-quarter observations, containing 464 PE ophthalmologists and 11,649 non-PE ophthalmologists. Includes all spending billed under an optometrist national provider identification number [NPI]. Optometrist sample consists of 432,314 optometrist-quarter observations, containing 298 PE optometrists and 21,158 non-PE optometrists. All models included clinician and year fixed effects. Standard errors were adjusted for clustering at the practice (TIN) level. Sample consists of clinicians who were present for at least 4 quarters before and after PE acquisition.

Figure 2.

Differences of Private Equity Acquisition on the Number of Intravitrial Injections by Drug Type and Quarter, 2011–2019.

Includes only ophthalmologists who gave at least one intravitreal injection in that quarter. Intravitreal injection sample consists of 58,749 ophthalmologist-quarter observations, containing 144 PE ophthalmologists and 2,345 non-PE ophthalmologists who provided at least one intravitreal injection in a given quarter.

Subgroup, Secondary, and Sensitivity Analyses

Results were generally similar to the main analyses in secondary analyses that restricted the sample to clinicians who billed a single TIN (eTable 2), to clinicians who were present in all quarters of the study period (eTable 3), and in analyses that controlled for state specific time trends (eTable 4) (each available at https://www.aaojournal.org).

DISCUSSION

We identified 123 practices with 762 ophthalmologists and optometrists acquired by PE between 2012 and 2019. Our most striking finding is the 74% relative increase over time in use of ranibizumab, one of the most expensive intravitreal medications, in PE-acquired practices vs. non-PE practices, and a 13% decrease in use of a much less expensive medication (bevacizumab). It should be noted that these results did not reach statistical significance after adjustment for multiple comparisons, but the increasing use of ranibizumab beginning in the second quarter after acquisition, an increase that rose sharply and steadily from the sixth through eleventh quarter, is striking.

Medicare pays high prices for aflibercept and ranibizumab compared to bevacizumab) (2019 average sale price per dose of $1877, $1717, and $70 respectively)²³ and adds 6% (4.3% during sequestration) payment to practices for these medications. This provides a strong incentive ($267,000 annually for an average retinal specialist)²⁴ to use more expensive medications.²⁴ Additionally, and perhaps more importantly, as private equity consolidates practices, they are able to attain larger volume-based rebates from manufacturers, providing another incentive to use more expensive drugs. It is possible that ranibizumab and aflibercept have somewhat better clinical results than bevacizumab for certain conditions, but the differences are small enough and the controversy over which medication to use is strong enough that bevacizumab is widely used.^23,25,26 It is not clear why practices increased the use of ranibizumab so much more than similarly expensive aflibercept; possibly larger confidential rebates and discounts were offered for ranibizumab than for aflibercept.²⁴

By 2019, after PE acquisition, the number of optometrists per practice increased by nearly one optometrist (89%) per practice and the number of Medicare beneficiaries seen per optometrist increased by 23.9% compared to non-PE practices, while the number of ophthalmologists increased by 52% with no change in the number of beneficiaries per ophthalmologist. The absolute numbers are small, but these findings are consistent with the commonly stated hypothesis and empirical evidence that PE firms increase their use of non-physician clinicians, who are paid at lower rates.²⁷ The impact of substituting optometrists for ophthalmologists on the quality and overall cost of care is controversial.^28–31

Contrary to our hypotheses, the relative increase in overall spending on ophthalmology services was small (5%) and not statistically significant, and ophthalmology-related ancillary spending decreased by 8%. Also contrary to our hypotheses, though PE practices saw lower percentages of dual-eligible patients both pre and post-acquisition, they did not reduce the percentages of dual-eligibles over time compared to reductions by non-PE practices.

We are aware of three studies that focused on the number of PE acquisitions of ophthalmology and optometry practices. Zhu et al. identified 11 acquisitions between 2013 and 2016.³² Chen et al. identified 228 acquisitions with 2,146 clinicians between 2012 and 2019, including 127 acquisitions in 2017–2018, a number very close to the 123 practices that we identified for our analyses. Singh et al. identified 107 acquisitions in 2017–2018.⁴ The median numbers of ophthalmologists (5) and optometrists (2) in the 127 practices they found are comparable to the number of 4.0 and 2.6 that we identified.

To our knowledge, there is one other quantitative study of utilization and spending in PE-acquired ophthalmology practices. Singh et al. found higher allowed amounts, unique patients seen, and encounters in PE-acquired practices compared to non-PE practices.⁴ A study focused on dermatology practices, using commercial insurance claims, found 4–17% increases in number of patients seen and 3–5 % increases in prices (allowed amounts) after PE acquisition, compared to non-PE practices, with no significant changes in dermatology spending, use of biopsies, lesion destruction, or Mohs surgery.³ A study focused on hospital-based anesthesiologist services provided by physician management companies (PMCs) and private equity firms found a 13% increase in prices (allowed amounts) for PMC-employed anesthesiologists and a 26% increase for anesthesiologists employed by PE-backed PMCs.³³

This study has several limitations. First, there is no generally accepted claims-based measure of quality in ophthalmology care, so we were unable to measure quality. Second, though the event study figures suggest relative increases over time in PE-acquired compared to non-PE practices in total spending, spending on intravitreal injections, the number of ranibizumab injections, and beneficiaries seen by optometrists, wide confidence intervals in later quarters – due to smaller sample size – mean that these results should be interpreted with caution. Third, we were unable to assess the clinical appropriateness of the increase in ranibizumab injections and concomitant decrease in bevacizumab injections. Fourth, since our analyses are based on Medicare claims, we could not assess whether PE-acquired practices raise prices paid by commercial insurers, nor could we include services that not covered by Medicare for which patients pay directly. The potential to profit from such services is one factor that drives private equity interest in ophthalmology practices. Fifth, it is possible that outcomes would have changed if more years post PE acquisition were available. For example, use of optometrists as lower cost substitutes for ophthalmologists might increase over time, and particularly after a second sale of practices, in which a practices’ clinicians may have little or no say in choosing the company that buys them.

CONCLUSION

Though not all results reached statistical significance, this study suggests that PE-acquired practices had little or no overall effect on utilization or total spending, but increased the number of unique patients seen per optometrist and the use of expensive intravitreal injections. These results are relevant for policymakers, patients, ophthalmologists, and optometrists. Further research should include information on prices, quality (if adequate claims-based measures can be created), sales of acquired practices by private equity firms, and changes over additional years post-acquisition.