The Influence of Practice Structure on Urologists’ Treatment of Men with Low-Risk Prostate Cancer

Abstract

Background:

Vertical and horizontal integration among health care providers has transformed the practice arrangements under which many physicians work.

Objective:

To examine the influence of type of practice structure, and by implication the financial incentives associated with each structure, on treatment received among men newly diagnosed with low-risk prostate cancer.

Research Design:

We compiled a unique database from cancer registry records from four large states, Medicare enrollment and claims for the years 2005–2014 and SK&A physician surveys corroborated by extensive internet searches. We estimated a multinomial logit model to examine the influence of urologist practice structure on type of initial treatment received.

Results:

The probability of being monitored with active surveillance was 7.4 and 4.2 percentage points higher for men treated by health system and non-health system employed urologists (p<.01), respectively, in comparison to men treated by single specialty urology practices. Among multispecialty practices, the rate of active surveillance use was 3 percentage points higher compared to single specialty urology practices(p<.01). Use of intensity modulated radiation therapy (IMRT) among urologists with ownership in IMRT was 17.4 percentage points higher compared to urologists working in small single specialty practices.

Conclusions:

Physician practice structure attributes are significantly associated with type of treatment received but few studies control for such factors. Our findings—coupled with the observation that urologist practice structure shifted substantially over this time period due to mergers of small urology groups—provide one explanation for the limited uptake of active surveillance among men with low-risk disease in the US.

Introduction

Consolidation, through mergers and acquisitions among health care providers, has occurred at a rapid pace since 2005. Vertical integration whereby hospitals or health systems acquire physician practices means that an increasing share of the physician workforce have become employees. Although vertical integration has potential benefits including greater care coordination, quality improvements, innovation and cost saving through greater efficiencies, published research suggests that these benefits rarely materialize.^1–6 The same set of factors have fostered horizontal integration whereby small single specialty physician groups merge to form larger physician groups. An additional motivation for horizontal integration is to establish ownership in ancillary services and adjuvant therapies to augment practice revenues. An often-overlooked consequence of consolidation is the sharp decline of small single specialty community-based practices. This common practice model has been replaced by a heterogenous mix of practice structure arrangements. Physicians now are employees of a health system, belong to a large multispecialty practice or have joined a large for-profit single specialty group. However, the effects of this restructuring of physicians’ practice arrangements on service use, including highly reimbursed procedures and treatments, is unknown. This study addresses this significant knowledge gap by examining the influence of practice structure—and by implication the financial incentives associated with each structure—on treatment received among men newly diagnosed with low-risk prostate cancer.

Our study focuses on prostate cancer for several reasons. First, previous studies have established significantly higher use of intensity modulated radiation therapy (IMRT) by urologists with ownership in IMRT,^7–10 but did not evaluate other practice structure arrangements. Second, prostate cancer is the most common nondermatological cancer among men and it is costly to treat. In 2020, an estimated 191,930 American men were newly diagnosed with prostate cancer and more than 60% of incident cases occurred among men age 65 and older.¹¹ About 90% of them have clinically localized disease, which in most cases is indolent. The costs to Medicare for screening and treating prostate cancer in men aged 70 and older for three years after diagnosis was estimated to be $1.2 billion.¹² Thus, treatment decisions have significant financial ramifications for patients and insurers.

Third, there is no consensus regarding the optimal treatment approach for men with low-risk disease. Primary definitive (aggressive) treatments include prostatectomy, brachytherapy, and IMRT, a highly reimbursed treatment. Alternatively, a patient may receive less aggressive hormone therapy or monitoring (active surveillance). While there are significant cost differences between the alternative treatments, differences in survival between these clinically equivalent options are negligible.^13,14 Irrespective of treatment received, the relative 10-year survival rate among all men with prostate cancer is 98–99%.^11,14–16 Notably, IMRT and robotic surgery are both substantially more expensive than the older technologies they have replaced, but offer no additional benefit in terms of survival for low-risk disease.^14,17,18 The ongoing controversy regarding which men to treat has resulted in significant variation in practice patterns.¹⁹ Moreover, evidence shows that many men with low-risk prostate cancer are over-treated,^19,20 despite recognition that conservative monitoring approaches are more appropriate.^21,22 Thus, there is heightened opportunity in the case of prostate cancer for non-clinical factors, such as practice structure to influence urologists’ practice patterns.

This study also addresses some of the limitations of prior research. First, we focus on low-risk prostate cancer—the cohort for whom active surveillance is most appropriate. Second, we examine a ten-year period during which there was substantial restructuring of the arrangements under which urologists work. Third, we examine a range of practice structure features, including single and multispecialty practices and employment arrangements. Previous research did not distinguish health system employment from multispecialty practices. Fourth, we examine a full spectrum of treatment options, whereas prior research focused only on IMRT^7,8 and/or active surveillance.²³

Conceptual Framework

Urologists, physician specialists who diagnose and sometimes treat prostate cancer, serve as the gatekeeper of the treatment team. Other specialists involved include diagnostic radiologists and radiation oncologists. Lacking clinical expertise, the patient must rely on his urologist to act as his agent in the prostate cancer treatment decision-making process. A urologist’s recommendation will consider multiple factors, including tumor attributes, potential side effects of treatment, the financial and time costs to obtain treatment, and patient health and demographic characteristics. Previous research suggests that the physician’s recommendation has considerable influence on the patient’s treatment choice, whereas patient preferences had no impact.^24–26

Urologists typically receive compensation for their services under different practice structure arrangements. First, some providers receive payment based on the number and types of services provided, with the level of payment determined—at least for Medicare beneficiaries—by the Medicare resource-based relative value scale (RBRVS) fee schedule. This is known as a fee-for-service (FFS) and is typical in small, single specialty urology practices. Second, urologists who have acquired ownership in IMRT services receive FFS payments for their own work and a share of the revenue generated when their patients undergo IMRT —an arrangement also known as “self-referral”.⁸ Most self-referring urology practices were formed through the merger of multiple small urology groups. Third, some urologists are members of multispecialty practices, which are also paid FFS but include primary care physicians and radiation oncologists. Although compensation paid to individual physicians within multispecialty practices varies, it is typically not tied directly to the number and types of services each physician provides. Fourth, some urologists are employed by a health care organization from which they receive a straight salary. Although these organizations bill Medicare on a FFS basis, they do not compensate their physician employees based on the number of services provided. Some employed urologists are paid a straight salary whereas others receive a salary plus bonus based on productivity and/or quality.

These practice structures create differing financial incentives around resource utilization. For urologists working under the FFS arrangement, incremental income is generated in proportion to the resources used for treatment. Urologists involved in self-referral practices face financial incentives to use more resources and to use treatments for which they receive a share of revenue. Urologists in large multispecialty practices face joint (cross-practice), if not individual, incentives related to resource utilization. However, urologists in large multispecialty practices are not likely to benefit directly from internal referrals for IMRT or other profitable treatments, so their financial incentives to recommend IMRT are diminished relative to urologists who work in practices with ownership in IMRT. Contrary to for-profit practices, financial incentives related to resource utilization are muted for urologists who are employed. Productivity-base bonuses on top of a salary may incentivize urologists to increase resource, but these effects tend to be minimal.

Data and Methods

We rely on data from four state cancer registries—California, Florida, New Jersey, and Texas—merged with enrollment and claims data for beneficiaries enrolled in traditional Medicare (TM) and with physician surveys from SK & A (OneKey). We selected these states because they are populous, racially/ethnically diverse and urologists work under a variety of practice structure arrangements. Importantly, cancer registry records enable us to identify incident cases of low-risk disease (stage I), include date of diagnosis and collect a rich set of clinical variables that are not available in Medicare claims data. The cancer registry data includes information on initial treatment, but no data on hormone therapy or chemotherapy. Receipt of these treatments can only be identified using Medicare claims. In addition, we use Medicare claims data to identify individual physicians, hospitals and cancer center providers. We use supplemental information from the SK & A physician surveys to help identify practice structure for each physician. We merged registry and Medicare data at the individual level using an encrypted Medicare beneficiary identifier and an encrypted patient state registry identifier. We constructed a physician-year using the UPIN and the NPI number associated with each practicing urologist. Next, we merged these core items obtained from individual claims with practice level characteristics of each urology group from the SK&A physician surveys for the years 2007, 2010 and 2013. Lastly, our analytic file includes zip-code level educational attainment and household income variables derived using the American Community Survey. Our merged data span calendar years 2005–2014 and includes men ages 65 and older who were newly diagnosed with low-risk prostate cancer during the time period January 1, 2005 through June 30, 2014 (to allow for six months of follow-up).

The SK & A data have been used in published research to address issues related to consolidation.^27,28 However, researchers have raised concerns about relying on the SK & A survey alone to examine the extent and effects of vertical integration. Other researchers have compared the physician populations in the National Provider and Plan Enumeration System (NPPES) with the AMA Masterfile and the SK&A physician file.²⁹ Compared to the NPPES (the gold standard), the SK & A had the highest rates of missing data (28% to 50%) and this varied by specialty. We adopted several strategies (see appendix A) to ensure the physician-year file included all urologists in each state who treat prostate cancer.

We codified treatments into one of six mutually exclusive and exhaustive categories: active surveillance; prostatectomy; IMRT; brachytherapy; hormone therapy alone and other surgical procedures (TURP, cryosurgery). See Table 1A of appendix B for descriptions. We estimate a multinomial logit model comparing receipt of each type of treatment to active surveillance. Because practice structure type varies over time and has multiple categories, we specify the following empirical model:

$$\text{Pr }\left({\text{Y}}_{ijt}=\text{M}\right)=\text{Pr}\left({\beta }_1+\mu {\text{ PRACI}}_{ijt}\times {\text{TIME}}_t+{\beta }_2{X}_{ijt}+{\beta }_3{R}_{jt}+\delta {\text{LOCALITY}}_{jt}+\alpha {\text{YEAR}}_{it}+u_{ijt}>0\right)$$

Where i = the patient with newly diagnose low-risk prostate cancer, j = the payment locality where the patient received treatment and t = the year of diagnosis. M is the treatment option, where M = 1 is active surveillance (the base category); M = 2 is prostatectomy; M = 3 is IMRT; M = 4 is brachytherapy; M = 5 is hormones only; and M = 6 is other surgical procedures (TURP and cryosurgery).

The key right-hand side variables of interest are the indicators designed to capture practice structure (Table 2). These variables include: 1) small urology group without IMRT ownership; 2) medium to large urology practices with ownership in IMRT; 3) FFS multispecialty group; 4) employed by a health system); and 5) employed but not by a health system. We selected small urology group with no ownership in IMRT as the reference category because this was the predominant practice structure in 2005. The interaction of the practice structure indicators with year (TIME) acknowledges that practice structure for each urologist can change over time. X is a vector of patient-level characteristics including age, race/ethnicity, marital status, clinical characteristics (tumor grade and PSA level), while R is a vector of local area characteristics including median income and education of residents in the patient’s census tract of residence. LOCALITY is a vector of dummy variables to identify the Medicare payment locality within each state where the patient received treatment, and YEAR a vector of dummy variables for year of diagnosis. These control variables are described in Table 2A of the appendix B.

Table 2.

Multinomial Logit Estimates of Treatment Choice among Men with Newly Diagnosed Low-Risk Prostate Cancer:

Effects of Urologists’ Practice Structure

Practice Structure	Prostatectomy	Radiation Therapy (IMRT)	Brachytherapy	Hormone Therapy Alone	Cryosurgery or TURP
Medium to Large Urology Group with IMRT Ownership	.020 (.036) OR=1.02 (0.95,1.10)	.709(.031) *** OR=2.02 (1.91, 2.16)	−.045 (.040) OR=0.96 (0.88, 1.03)	−.123(.048) *** OR=0.88 (0.80, 0.97)	.084 (.049) OR=1.09 (0.99, 1.18)
Multispecialty Group	−.052 (.040) OR= 0.95 (0.88, 1.03)	−.264(.037) *** OR=0.77 (0.71, 0.82)	−.416 (.044) *** OR=0.65 (0.60, 0.72)	.029(.049) OR=1.03 (0.93, 1.13)	.078 (.052) OR=1.08 (0.97,1.20)
Employed, Health System	.089(.039) ** OR=1.09 (1.01,1.18)	−.532(.038) *** OR = 0.59 (0.55,0.63)	−1.233(.054) *** OR = 0.29 (0.26, 0.32)	−.465(.058) *** OR = 0.63 (0.56, 0.70)	−.302 (.060) *** OR=0.74 (0.66,0.83)
Employed no Health System	.520 (.086) *** OR=1.68 (1.42, 1.99)	−.527 (.091) *** OR=0.59 (0.49,0.71)	−.814 (.115) *** OR=0.44 (0.35, 0.55)	−.551(.147) *** OR=0.58 (0.43, 0.77)	−.660 (.159) *** OR=0.52 (0.38, 0.71)

Source: Analysis of cancer registry merged with, Medicare enrollment, claims data & urologist practice structure.

Notes: The first row in each cell is the logit coefficient and standard error (in parentheses). The second row in each cell is the adjusted odds ratio and the third row in each cell is the 95% confidence interval surrounding the odds ratio. The reference category for type of practice structure is FFS single specialty practice with no ownership in IMRT. The comparison group for each procedure is active surveillance. Men who opt for active surveillance receive no treatment other than periodic tests to check for signs the cancer is growing.

The unit of observation in this analysis is the newly diagnosed prostate cancer patient who was treated in payment locality j and was diagnosed in year t. We do not estimate a fixed effects model where the unit of observation is the urologist-year for at least two reasons. First, we would not be able to evaluate the influence of practice structure on each type of treatment. Rather, using a fixed effects specification we can only evaluate the share of patients treated by each urologist in a given year who received treatment M (i.e., active surveillance). The estimation of a multinomial logit model enables us to examine the gamut of treatment options simultaneously. Second, we have an unbalanced panel as some older urologists retired during the time period and younger urologists joined specific practices. The exclusion of urologists who did not treat patients in every year during the time period 2005–2014 would result in a selected sample and yield biased estimates.

^***Significant at p<.01;

^**Significant at .01 ≤ p < .05.

Results

Our sample includes 87,922 episodes of men newly diagnosed with low-risk prostate cancer during the time period 2005 thru 2014. See consort diagram in appendix B for the rationale and counts of cases excluded. In 2005, about 83% of newly diagnosed low-risk prostate cancer cases were treated by urologists who belonged to small single specialty practices (Figure 1). By 2014, the practice structure accounted for only 39% of low-risk prostate cancer episodes. The reverse pattern characterizes urologists who joined larger practices with ownership in IMRT. In 2005, urology practices with IMRT ownership treated only 2.2% of low-risk prostate cancer episodes but this share increased to 32% by 2014. Multispecialty practices without IMRT ownership treated 7% of low-risk prostate cancer episodes in 2005 and this share increased to about 12.7%. Cases treated by health system employees more than doubled from 6.2% to almost 14% over the time period. Employed urologists not affiliated with a health system accounted for only 1–2% of episodes.

Figure 1.

Changes in Urologist Practice Structure Measured in Episodes of Low-Risk Prostate Cancer Cases Treated, 2005–2014

Note: Total Number of Episodes, by Year, are shown in brackets.

Figure 2 depicts changes in use of prostate cancer treatment over the time period 2005–2014. Use of active surveillance rose from 17.4% to 21%, an increase of less than 4 percentage points. Prostatectomy increased by more than 4 percentage points (16.8 to 21%). Use of IMRT jumped by 11 percentage points to 35.2% in 2014. In contrast, receipt of brachytherapy fell from almost 25% in 2005 to 8 percent in 2014. Use of hormone therapy alone declined slightly from 9.1% to 7.7%. Likewise, use of other surgical procedures fell by 1 percentage point.

Figure 2.

Changes in Prostate Cancer Treatments for Men with Low-Risk Prostate Cancer, 2005–2015

Note: Total Number of Episodes, by Year, are shown in brackets.

Table 1 reports patient characteristics for the overall sample and the sub-samples stratified by practice structure arrangement. Employed urologists treated higher proportions of men age 65–69 and lower percentage of men age 80 and older. Urologists who belonged to single specialty practices (with and without IMRT ownership) treated higher proportions of white Hispanic men relative to other practice structure types. Married men accounted for almost 67% of episodes treated by employed urologists who were not part of a health system but only 47% of men treated by single specialty practices with IMRT ownership. 84% of men treated by urologists with ownership in IMRT resided in Florida, New Jersey or Texas whereas about 50% of those treated by multispecialty practices or health system employed urologists resided in these three states.

Table 1.

Characteristics of Men with Newly Diagnosed Low-Risk Prostate Cancer, Overall and by Practice Structure Type

Type of Treatment	All Medicare Enrollees	Small Urology Group - no IMRT ownership	Medium to Large Urology Group - with IMRT Ownership	Multispecialty Group	Employed by a Health System	Employed not by a Health System
Number of cases	87,922	55,867	15,769	8,123	6,927	1,236
Patient Characteristics		PRAC1	PRAC2	PRAC3	PRA45	PRAC5
Age 65–69	33.1%	31.4%	35.7%	34.1%	37.9%	39.5%
Age 70–74	30.3%	30.1%	30.2%	30.9%	30.5%	31.1%
Age 75–79	21.9%	22.6%	21.0%	21.0%	19.9%	19.5%
Age 80–84	10.6%	11.4%	9.6%	9.4%	8.8%	7.8%
Age 85 plus	4.2%	4.5%	3.6%	4.6%	2.9%	2.0%
White Non-Hispanic	74.8%	74.2%	73.9%	78.6%	76.7%	77.5%
African American	6.6%	6.5%	7.2%	5.1%	7.8%	5.0%
White Hispanic	12.7%	12.9%	14.2%	10.8%	9.7%	9.9%
Asian	4.1%	4.3%	2.9%	3.9%	4.7%	6.2%
Race Other	1.8%	2.0%	1.7%	1.7%	1.0%	1.4%
Married	53.8%	55.9%	47.1%	50.9%	54.1%	67.2%
Divorced/Separated	4.1%	4.2%	3.3%	3.5%	4.6%	6.1%
Widowed	5.1%	5.7%	4.2%	4.2%	3.8%	3.8%
Never Married	6.1%	6.3%	5.4%	4.8%	7.6%	6.5%
Unknown M Stat	30.9%	27.9%	39.9%	36.6%	29.9%	16.4%
Dual Eligible	10.3%	10.9%	8.5%	9.5%	10.5%	10.8%
% Bachelor’s Degree	32.4% (17.9)	31.6% (17.5)	33.9% (17.6)	31.3% (18.6)	36.7% (20.2)	35.3% (17.6)
Median Income	$40,573 (14,877)	$40,224 (14,608)	$41,139 (15,079)	$39,695 (15,138)	$42,864 (16,038)	$42,096 (14,269)
PSA Positive/Elevated	70.5%	70.5%	68.9%	70.7%	73.6%	74.4%
PSA Negative/Normal	7.3%	6.7%	8.8%	7.4%	8.6%	9.1%
PSA Borderline	1.2%	1.5%	0.6%	0.8%	1.3%	1.1%
PSA Unknown	20.9%	21.3%	21.7%	21.2%	16.5%	15.5%
GRADE I	3.0%	2.4%	4.6%	3.3%	4.4%	4.2%
GRADE II	46.0%	46.7%	45.1%	45.1%	44.1%	44.3%
GRADE III	50.9%	50.9%	50.3%	51.5%	51.5%	51.5%
Employed by a Cancer Center	2.0%	XX	XX	XX	16.7%	45.4%
State-California	40.7%	44.6%	16.0%	50.4%	51.1%	60.6%
State-Florida	14.8%	13.2%	24.4%	11.0%	10.3%	13.6%
State-New Jersey	21.4%	22.1%	28.9%	8.4%	14.5%	14.1%
State-Texas	23.1%	20.1%	30.7%	30.3%	24.1%	11.7%

Table 2 reports logit coefficients and odds ratios from the estimation of multinomial logit regression model. The odds ratios are interpreted relative to the base category--urologists who belonged to small single specialty practices. Receipt of prostatectomy, as opposed to active surveillance, was significantly higher for men treated by employed urologists. The increased odds were 1.09 and 1.68 times as likely for men treated by health system and non-health system employed urologists respectively (p<.01). The increased odds of receiving IMRT, rather than active surveillance, was twice as likely for men treated by urologists with ownership in IMRT (p<.01). Conversely, the other practice structure types had lower use of IMRT compared to urologists who worked in small single specialty practices. The reduced odds were the lowest, 59% as likely, for employed urologists irrespective of health system affiliation (p<.01). Use of brachytherapy was likewise significantly lower for urologists in either multispecialty groups or employed practices. The lower odds were 29% as likely for men treated by urologists employed by a health system to 65% as likely for urologists who belonged to multispecialty groups (p<.01). Receipt of hormone therapy alone, as opposed to active surveillance, was 88% as likely for men treated by urologists with ownership in IMRT and 58% to 63% as likely for men seen by employed urologists (p<.01). Use of either TURP or cryosurgery 74% and 52% as likely for men treated by system and non-system affiliated urologists respectively (p<.01).

We used the margins command in Stata to calculate the marginal effect of each practice structure type on each treatment option assuming all other covariates were fixed at the sample means. Marginal effects must be evaluated in comparison to the baseline use rate. If the baseline use rate of a treatment is 25%, then a change of 20 percentage points represents an 80% increase in use. Marginal effects are reported in Table 3; use rates for each treatment for urologists who were members of small single specialty groups without IMRT ownership are reported at the bottom of Table 3.

Table 3.

Marginal Effects of Urologists’ Practice Structure Arrangement on Type of Treatment Received

Practice Structure	Active Surveillance	Prostatectomy	IMRT	Brachytherapy	Hormones Alone	TURP or Cryosurgery
Large Urology Group with IMRT Ownership	−.046***	−.041***	+.174***	−.049	−.026***	−.011***
Multispecialty Group no IMRT Ownership	+.030***	+.017***	−.038***	−.038***	+.014***	+.014***
Employed Health System	+.074***	+.088***	−.059***	−.100***	−.007	−.003
Employed no Health System	+.042***	+.173***	−.101***	−.073***	−.020***	−.020***

Notes: The reference category is urologists who were members of small single specialty groups with no ownership in IMRT. Type of treatment received for men seen by urologists who were members of small single specialty groups with no ownership in IMRT were as follows: active surveillance (15.9%), prostatectomy (18.3%), IMRT (30.4%), brachytherapy (19.6%), hormones alone (8.6%), and TURP/cryosurgery (7.1%). We used the margins command in Stata to calculate the marginal effect of each practice structure type on each treatment option where all other characteristics were held at their sample means. We then conducted a two-tailed t-test for difference between the means to ascertain if the mean predicted outcome for treatment M for men treated by urologists who were members of small single specialty groups with no ownership in IMRT was significantly different from each other practice structure type.

^***Significant at p<.01;

^**Significant at .01 ≤ p < .05

The probability of being monitored with active surveillance was significantly lower for men treated by urologists in FFS practices with ownership in IMRT; the marginal effect was −.046 which corresponds to a 29% reduction in use relative to FFS urologists working in small practices (p<.01). Multispecialty practices had greater use of active surveillance relative to urologists working in small FFS practices—3 percentage points higher (p<.01), which is almost a 20% increase. Use of active surveillance was even higher among employed urologists; 4.2 percentage points for non-system affiliated urologists and 7.4 percentage points for health system affiliated urologists (p<.01); these increases were 29% and 49% relative to urologists working in small practices (p<.01).

Receipt of prostatectomy was 4.1 percentage points lower (22% reduction in use) for men treated by large urology practices with IMRT ownership relative to those seen by urologists working in small practices (p<.01). Use of prostatectomy was 1.7 percentage points higher (9% increase) among men treated by urologists who were members of multispecialty practices (p<.01). For men treated by health system employed urologists, the odds of undergoing a prostatectomy was 8.8 percentage points higher, a 48% increase in use compared to urologists working in small FFS practices (p<.01). Urologists working at non-system affiliated cancer centers had prostatectomy use rates that were 17.3 percentage points (94%) higher than urologists who belonged to small FFS single practices (p<.01).

As expected, the probability of being treated with IMRT was 17.4 percentage points higher for men treated by urologists with ownership in IMRT, a 57% increase in use compared to men treated by urologists in small FFS practices (p<.01). Receipt of IMRT was 3.8 percentage points lower for men treated by urologists working in multispecialty practices, a 12.5% reduction in use compared to urologists in small FFS urology practices (p<.01). For men treated by employed health system affiliated urologists, receipt of IMRT was 5.9 percentage points lower, a 19% reduction in use (p<.01). Use of IMRT was 10 percentage points lower (a 33% reduction) for men treated by other employed urologists relative to urologists working in small practices with no IMRT ownership (p<.01).

The likelihood of receiving brachytherapy was likewise significantly lower for employed and multispecialty practices relative to urologists working in small urology practices. The reduction in use of brachytherapy was greatest for health system employed urologists—10 percentage points—a 52% reduction in use (p<.01). For non-system affiliated urologists, the reduction in use of brachytherapy was 7.3 percentage points, a 37% decline (p<.01). Multispecialty practices had brachytherapy use rates that were 3.8 percentage points lower than FFS urologist in small practices, a 20% reduction in use (p<.01). Use of hormone therapy alone as the primary treatment was low regardless of practice structure arrangement (7.2% to 9.1%). Large urology practices with IMRT ownership had use rates of hormones alone that were 2.6 percentage points (30%) lower than urologists working in small FFS practices (p<.01). Use of hormone therapy alone was 2 percentage points (23%) lower for employed urologists not affiliated with a health system compared to urologists working in small practices (p<.01). In contrast, multispecialty practices had 16% higher use rates (p<.01). The influence of practice structure on receipt of other surgical treatments (TURP or cryosurgery) was similar to that for hormone therapy. Use of these other surgical treatments was 1.1 percentage points (15%) lower for men seen by urologists with ownership in IMRT (p<.01) and 2 percentage points (28%) lower for men treated by employed non-system affiliated urologists (p<.01). In contrast, multispecialty practices had use rates of other surgical treatments that were 1.4 percentage points or nearly 20% higher compared to urologists in small FFS practices (p<.01).

Discussion

Data limitations have been a key reason for the lack of empirical evidence to date regarding the influence of physician practice structure on type of treatment received. We addressed this gap in knowledge by constructing a unique database to analyze the impact of urologist practice structure on treatment received by men newly diagnosed with low-risk prostate. Our focus on low-risk prostate cancer is compelling because this disease is highly prevalent, costly to treat and there exist multiple treatment options that vary substantially with respect to cost but are clinically equivalent in terms of survival.

We find that practice structure significantly affects the type of treatment received and there are clear distinctions between employed urologists and those who work in FFS single specialty urology practices. The probability of being monitored with active surveillance was significantly lower for men treated by small and large single specialty practice urology practices where revenues are directly tied to the type and number of services received. Even multispecialty FFS practices had higher use rates of surveillance compared to single specialty urology practices. One possible explanation is that urologists who work in FFS multispecialty practices do not directly benefit financially from providing more costly treatments.

Use of prostatectomy was significantly lower among single specialty urology practices compared to employed and multispecialty practices whereas the reverse was the case for receipt of IMRT. As expected, the increase in use of IMRT among urologists with ownership in IMRT was 17.4 percentage points higher, a 57% increase relative to urologists working in small single specialty practices. The increased use is concerning because both the monetary and patient time costs to receive a course of IMRT are substantially higher than other aggressive treatments and active surveillance. Among the other practice structure types, receipt of IMRT was significantly lower compared to urologists working in small FFS practices but the difference was greatest for employed non-health system urologists (33% lower). Receipt of brachytherapy was significantly lower for all practice structure types (except urologists with ownership in IMRT) in comparison to urologists in small practices.

We recognize our study has limitations. First, we cannot distinguish if employed urologists are paid a straight salary or salary plus productivity-based bonus. Second, we lack data on the specific compensation approach used by large multispecialty practices and how their incentives differ from single specialty practices. Third, our analysis focuses on receipt of the primary treatment and did not consider intensity of treatment or receipt of adjuvant therapy (IMRT boost for surgery or brachytherapy). Fourth, although all men in our sample were diagnosed with stage I prostate cancer, we lack detailed data on the biopsy Gleason score which was only available in the NJ registry.

With mounting evidence regarding its safety and potential benefits, active surveillance has become the preferred treatment option for men with low-risk disease.^30,31 Surveillance avoids morbidity and spending in cases where treatment is unnecessary but also identifies men who are higher risk. This recognition prompted changes in practice guidelines that support the use of active surveillance for low-risk cases.³² That said, urologists in the US have been slow to adopt active surveillance for men with low-risk disease at the same rate as other developed countries,^33–36 and the factors that contribute to the slow uptake of active surveillance are poorly understood.

Our findings indicate there is considerable practice-level variation in use of active surveillance. The differences in practice patterns between urologists who work in settings where financial incentives and physician entrepreneurialism are strong versus employment arrangements where such incentives are weak are striking. Our findings have implications for Medicare reimbursement policy. Because treatment options for prostate cancer are clinically equivalent in terms of survival, Medicare could consider establishing reimbursement policies for prostate cancer that are based on NCCN practice guidelines.³⁰ In conclusion, an important implication of our findings is that practice structure attributes are significant predictors of type of treatment received for low-risk prostate cancer. Notably, few studies have recognized the importance of practice structure as a factor that influences physicians’ treatment decisions.

Data Availability:

The data that support the findings of this study are available from state cancer registries and the Center for Medicare and Medicaid Services. Restrictions apply to the availability of these data which were acquired and analyzed under data use agreements for this specific project.