Creating a parsimonious typology of physician financial incentives

Abstract

In order to create an empirically derived parsimonious typology of physician financial incentives that will be useful for future research, we used data from the nationally representative 2004–2005 Community Tracking Study Physician Survey (N = 6,628). Linear regression analyses informed by economic theory were used to identify the combinations of incentives associated with an overall financial incentive to expand services to individual patients. The approach was validated using two nonparametric methods (CART analysis and data mining techniques) and by examining the relationship between the resulting typology and other measures of physician behavior including hours worked, visit volume, and specialty-adjusted income. Of the 6,628 physicians surveyed, approximately 25% (1,605) reported an overall incentive to increase services and 75% (5,023) reported either neutral incentives or incentives to decrease services. Men, who were approximately 75% of respondents, were slightly more likely to report incentives to increase services (P < 0.05). There were no differences in reported incentives according to specialty. We created two typologies (one with eleven categories and the other with a collapsed set of six categories) based on combinations of variables measuring ownership, base compensation methods, and financial incentives. The percentage with an overall incentive to increase services ranges from 6% for employed physicians compensated via fixed salary to 36.7% for owners in low capitation environments with either individual or practice level productivity incentives. The criterion validity of the typology was established by examining the relationship with adjusted physician income, hours worked, and visit volume, which showed generally consistent relationships in the expected direction. A parsimonious typology consisting of six mutually exclusive groups reasonably captures the continuum of incentives to increase service delivery experienced by physicians.

1 Introduction

Per capita health care costs are higher in the United States than in any other developed country (Anderson et al. 2007). For example, federal spending on health care through the Medicare and Medicaid programs now consumes 4.6% of GDP and is projected to grow to 9% of GDP by 2035 (Congressional Budget Office 2007). Rapid growth in the provision of services to individual Medicare beneficiaries is a key element in the recent spurt in Medicare cost increases (Medicare Payment Advisory Commission (MedPAC) 2005). Similar trends beleaguer private sector payers. While some of this growth may be due to the introduction and diffusion of new technologies, physicians’ responses to financial incentives, both external from payment systems and internal through their practice, are also likely to be important. Moreover, financial incentives represent potentially powerful policy levers that CMS and other health care payers can use to influence the volume and mix of services provided. Although physician services account for less than 25% of overall health care spending, physician decision making drives 90% of health care service use and spending (Kaiser Family Foundation 2008).

One major challenge when defining and studying financial arrangements is the complex nature of physician practice in the US (Landon et al. 2005; Landon et al. 1998). Physician practices range from solo practices to large multidisciplinary groups consisting of hundreds of physicians. These practices are often nested within intermediary organizations (such as independent practice associations) and contract with multiple external payers that use different combinations of incentives. Theory suggests that each physician organization will come up with a single “best fit” method for compensating physicians that balances their values and professionalism with the goal of maximizing income across all external contracts (Glied and Zivin 2002). Nationally representative information about such arrangements, however, is not available, and what little information exists has rarely been collected in a systematic and reproducible way.

In this paper, we use detailed data from a large, nationally representative survey of physicians to construct and validate a parsimonious typology of physicians’ financial incentives. The goal of this typology is to reduce the variety and complexity of physicians’ financial incentives to a more limited and manageable set of categories that can be used to examine the influence of physician financial incentives on the costs and intensity of care.

Compensation arrangements are often complex, and require many variables to describe them. Moreover, individual compensation variables are often collinear and also require systematic exploration of interactions between them in order to understand fully their effects. Consequently, a simplified typology that accounts for these analytic issues will be useful for researchers trying to understand the influence of financial incentives on physician behavior.

2 Methods

2.1 Conceptual framework

We conceptualize the medical practice as the economic environment that is most proximate to the individual physician-patient interaction (Landon et al. 1998). Medical practices organize themselves and establish incentives and rules governing the delivery of care that allow the practice to operate best in its particular market. Empirical observations made by Glied and Zivin have specifically informed this conceptualization of the role played by financial incentives by indicating that medical practices facing a complex set of financial and administrative policies from multiple payers set up relatively simple sets of rules that spread financial risk (and reward) and organize care without necessarily maximizing income (Glied and Zivin 2002). The incentives faced by an individual physician are those most likely to influence clinical choices and depend on the methods used for rewarding and assigning risks within these varied organizational structures (Casalino 1992; Conrad et al. 1996; Kralewski et al. 1996; Murray et al. 1992; Ransom et al. 1996; Stearns et al. 1992).

Our analysis accounts for the practice’s reliance on revenue from capitated contracts, as suggested by (Glied and Zivin 2002) and extends their work by focusing on the variety of direct compensation methods and other financial incentives practices use to pay physicians and influence the amount of care provided to patients. Our approach recognizes that physicians’ clinical decisions affect a practice’s revenues and costs, and that physicians, especially in larger practices, are likely to vary in the extent to which their own compensation is affected by their clinical decisions (Conrad et al. 1998); (Kralewski et al. 2000).

2.2 Data source

The data for this study are from the fourth round of the Community Tracking Study (CTS) Physician Survey, a nationally representative telephone survey of 6,628 direct patient-care physicians conducted between August, 2004 and August, 2005 for the Center for Studying Health System Change. The physician survey sample was designed to be representative of non-Federal physicians in the continental United States, as well as in selected communities, or sites. In the first stage of sample selection, 60 sites were randomly selected with probability in proportion to population (Center for Studying Health System Change 1996).

In the second stage, sample physicians were drawn from the American Medical Association’s and the American Osteopathic Association’s master files. The sample included active non-federal office and hospital-based physicians who spent at least 20 h per week in direct patient care, but excluded specialties such as radiology, pathology, and anesthesiology. Primary care physicians were over-sampled.

The overall survey response rate of 52.4% is comparable to response rates obtained for surveys of physicians in recent years. Characteristics of responders and non-responders to the survey differed only modestly (Reschovsky et al. 2001). Greater detail about the survey can be found elsewhere (Center for Studying Health System Change 2006; Inter-University Consortium for Political and Social Research). All analyses employ survey weights that account for the probability of sample selection and differential survey non-response.

2.3 Physician financial incentives variables

The key variables measuring physicians’ financial incentives were drawn from the 2004 to 2005 CTS survey (A full description of the wording is provided in Table 1). The survey first asks whether physicians are full owners, part owners, or employees because owners’ incomes are based primarily on practice profits.

Table 1.

Financial incentives questions from the community tracking study physician survey

Domain	Question
Practice level incentives	% of revenue from managed care
	% of revenue from largest managed care contract
	% of revenue from Medicare
	% of revenue from Medicaid
	% of revenue that is capitated
Base payment method and ownership	Are you a full or part owner of your practice?
	Are you a salaried physician?
	Are you paid in direct relation to the amount of time you work, such as by the shift or by the hour?
	Is your base salary a fixed amount that will not change until your salary is re-negotiated or is it adjusted up or down during the present contract period depending on your performance or that of the practice?
Eligibility for incentive payments	Are you currently eligible to earn income through any type of bonus or incentive plan?
	Are you eligible to receive end-of-year adjustments, returns on withholds, or any type of supplemental payments, either from this practice or from health plans?
	Are factors that reflect your own personal productivity used to determine your compensation?
	[If yes] Is productivity very important, moderately important, not very important, or not at all important in determining your compensation?
	Are results of satisfaction surveys completed by your own patients used to determine your compensation?
Determinants of incentives payments	[If yes] Are patient satisfaction survey results very important, moderately important, not very important, or not at all important in determining your compensation?
	Are results of practice profiling comparing your pattern of using medical resources to treat patients with that of other physicians used in determining your compensation?
	[If yes] Are profiling results very important, moderately important, not very important, or not at all important in determining your compensation?
	Are explicitly measures of quality used in determining your compensation?
	[If yes] Are quality measures very important, moderately important, not very important, or not at all important in determining your compensation?
	Is the overall financial performance of the practice used in determining your compensation?
	For each of the factors you mentioned, tell me whether it is very important, moderately important, not very important, or not at all important in determining your compensation?
Overall assessment of incentives	How would you describe your overall personal financial incentives in your practice? On balance, do these incentives favor reducing services to individual patients, favor expanding services to individual patients, or favor neither?

Financial incentives also incorporate the method of compensation. Specifically, did the physician receive a fixed or adjustable salary, was he/she paid on the basis of time worked (wage based), or some other form of variable compensation (such as share of practice revenues)? The survey also asked whether the physician received pay in the form of a bonus, withhold, or other performance based incentive; and whether the amount of compensation was affected by any of the following explicit factors: individual productivity, practice financial performance, results of patient satisfaction surveys, measures of quality, and comparative practice profiling. For each of these five factors, the physician then indicated how important it was (not very, moderately, or very) to determining their compensation. Owners of solo practices were not asked about factors affecting their compensation, for they were assumed to be remunerated solely on productivity. Finally, to separate “ambient” financial incentives from the external payment environment, the survey asked the percentage of practice revenue drawn from capitated contracts.

2.4 Constructing the typology

The typology was constructed by using linear regression analysis to identify the combinations of incentives associated with distinct responses to a survey question in which each physician was asked to rate on a 7-point Likert scale whether their overall financial incentives favor reducing services to individual patients, expanding services to individual patients, or neither. This previously validated survey question has been shown to be associated with payment arrangements and market level factors (e.g., HMO penetration, use of gatekeepers) that have been associated with decreased service intensity (Mitchell et al. 2000a). Based on previous analyses that demonstrated that the three-way classification of this variable (with neutral as the middle category) did not behave as an ordinal variable (Reschovsky et al. 2006), and given our interest in identifying incentives that increase costs by expanding services to patients, we converted the scale into a dichotomous dependent variable that classified physicians as reporting that their incentives favor increasing services to patients, as opposed to favoring reducing services or being neutral (fewer than 10% of physicians reported having a strong incentive to reduce services).

The independent variables were constructed as a set of mutually exclusive and exhaustive categories based on responses to the questions related to ownership, base compensation arrangements, and eligibility for incentive payments. Base compensation arrangements were defined as fixed or variable.

We estimated linear probability models without an intercept term against the full set of exhaustive and mutually exclusive independent variables that describe the various combinations of compensation and incentive structures. Omitting the intercept term allows the coefficients to be interpreted as the proportion of physicians within each category who reported an overall financial incentive to increase service delivery. We combined categories based on a combination of theory and empirical results (e.g., full and partial ownership were combined into a single category because their coefficients were not significantly different from each other).

Because the linear probability model (LPM) ignores the finite range and the relationship between the mean and variance of a binomial random variable, it may not yield maximally efficient inferences. However, when the dependent variable is binary and the predictor variables exactly partition the individuals (i.e., every individual is in exactly one category), the maximum likelihood estimate (MLE) of the parameters under linear regression and Bernoulli regression with any monotone link function (e.g., the identity, logit, or probit functions) are equivalent. Therefore, for a non-overlapping taxonomy, linear regression yields fully efficient inferences for effects of the components of the taxonomy. In addition, because the purpose of this analysis was (in conjunction with economic theory) to construct a typology for use as a predictor in subsequent analyses, the simplicity of the LPM and the fact that it evaluates interaction effects on the scale of the data outweigh concerns about estimation efficiency.

After reducing the number of ownership and compensation method categories, we next examined two sets of interaction terms. The first set interacted each category with a dichotomous variable indicating whether the practice received more than 35% of its revenue in the form of capitation payments (Preliminary analyses explored 30 or 40% as alternative cutoffs.). If the interaction had a statistically significant and substantively important effect, we created additional categories to distinguish between high and low capitation practices.

The second set of interactions was limited to physicians who reported that their compensation was significantly affected by five explicit incentives within their practice (individual productivity, practice financial performance, results of patient satisfaction surveys, measures of quality, and comparative practice profiling). We created dummy variables for each incentive rated as moderately or very important to determining compensation. We again interacted these terms and dropped interaction terms that did not have statistically significant and/or substantively important coefficients. These incentive interactions terms could have affected the results in either direction (i.e., increasing incentives to provide more services, or decreasing them).

As a sensitivity analysis, we retested the typology after stratifying by medical specialty (primary care, medical specialty, or surgical/procedural specialty). These analyses suggested that the typology was minimally sensitive to physician specialty, so we present a unified typology developed across all physicians.

2.5 Validation of the typology

We validated the typology created from the linear probability models in two ways. First, we replicated the analysis using two alternative analytic techniques: Classification and Regression Tree (CART) analysis (Breiman et al. 1984) and computerized “data mining” algorithms (Wojtusiak et al. 2006). Unlike our approach, which draws heavily from prior theory and research, these two methods are purely empirically based. Briefly, CART analysis creates a mutually exclusive and exhaustive “tree” that minimizes the variation within categories and maximizes the variation between categories. Computerized data mining follows a different approach in that it identifies possibly non-nested combinations of factors that are associated with the financial incentive to increase services. Unlike CART, these combinations of characteristics are not necessarily mutually exclusive (Kaufman and Michalski 2005).

After validating the method used to construct the typology, we evaluated its criterion validity by examining several additional measures of physician behavior that we hypothesized would also be related to basic payment incentives: patient visit measures (adjusted for time spent in clinical work), hours of direct patient care, the difference between actual income and “predicted” income (based on specialty, years of experience, board certification, time spent in clinical practice, and local cost of living), and a variable that asked physicians to rate their agreement with a question asking whether they could make clinical decisions without negatively affecting their incomes.

2.6 Factors associated with financial incentives

Finally, to better understand why practices adopt various combinations of financial incentives, we estimated a regression model that relates physician demographic, practice, and market characteristics to the typology-generated predicted probability of having an incentive to increase services. This analysis will indicate whether certain types of physicians, practice characteristics, and market characteristics are associated with combinations of financial incentives that encourage or discourage physicians to increase services to patients in ways predicted from theory.

3 Results

3.1 Sample characteristics and physician incentives to increase/decrease services

Of the 6,628 physicians surveyed, approximately 25% (1,605) reported an overall incentive to increase services, as compared with 75% (5,023) who reported either neutral incentives or incentives to decrease services. Approximately 75% of respondents were male, and males were slightly more likely to report incentives to increase services (P < 0.05). There were no differences in reported incentives according to specialty.

Aside from full ownership (31% of physicians), there were significant differences in basic compensation arrangements between physicians who reported incentives to increase services and those who did not. Physicians with an overall financial incentive to increase services were more likely to be partial owners (29 vs. 21%, P < 0.001), be paid with variable compensation related to productivity (85 vs. 64%, P < 0.001), be eligible for a bonus (61 vs. 51%, P < 0.001), and to report that productivity was an important factor in determining their compensation (58 vs. 48%, P < 0.001). In contrast, those in highly capitated practices (15 vs. 11%, P < 0.01) and those paid via a fixed salary (36 vs. 15%, P < 0.001) were less likely to report an incentive to increase services. Finally, those in small and medium-sized group practices were more likely to report incentives to increase services, while those in medical school and “other” practices such as community clinics were less so (Table 2).

Table 2.

Description of the sample stratified by incentives to increase services

Demographic characteristics	All physicians (n = 6628)	Incentives to increase services (n = 1605)	Incentives neutral or to decrease services (n = 5023)
Sex
Female	25.2	21.8	26.2*
Male	74.8	78.2	73.8*
Medical school training
US medical graduate	78.0	77.7	78.0
International medical graduate	22.0	22.3	22.0
Years in practice
2–10	30.6	34.0	29.5*
>10	67.7	63.8	69.0**
Specialty
Primary care	36.7	39.1	35.9
Specialists	63.3	60.9	64.1
Ownership and compensation
Full owner	31.2	31.3	31.2
Partial owner	23.1	29.9	21.0***
Not an owner	45.6	38.9	47.8***
Fixed compensation	31.8	15.2	36.0***
Variable compensation	68.2	84.8	64.0***
High capitation ( > 40%)	14.0	10.7	15.1**
Eligible for bonus	53.8	61.3	51.3***
Factors determining compensation
Productivity	50.1	58.2	47.6***
Profiling based on use of services	7.7	7.2	7.8
Results of patient satisfaction surveys	14.1	12.7	14.5
Specific quality measures	12.7	13.7	12.4
Practice setting
Solo/two physician practice	32.5	33.0	32.3
Small group ( < 10 physicians)	16.6	20.1	15.5**
Medium/large group	14.9	19.4	13.5***
Group/staff model HMO	4.5	3.5	4.8
Medical school	9.3	5.8	10.4***
Hospital-based	10.1	9.0	10.4
Other	12.0	9.0	13.0**
Patient population
Practice revenue from medicaid (%)	16.7	14.7	17.3***
Practice revenue from medicare (%)	31.6	32.0	31.5

^*<0.05;

^**<0.01;

^***<0.001

3.2 Constructing the typology: linear probability model

Table 3 presents two typologies constructed from the results of the linear probability models: one with eleven categories and a more parsimonious six-category version that combines several pairs of adjoining cells with similar values in the eleven-category typology. Each cell of the typology describes a particular mix of ownership, compensation, and financial incentives ordered by the proportion of physicians in the cell who reported an overall financial incentive to increase services to patients. The percentage with an overall incentive to increase services ranges from 6% for employed physicians compensated via fixed salary (Group 1a) to 36.7% for owners in low capitation environments with either individual or practice level productivity incentives (Group 6b). The range for the more parsimonious typology is from 8.3 to 32.2%. Owners without explicit additional incentives (17.1%) and full owners of solo practices are in the middle of the range (18.8%).

Table 3.

Expanded and reduced form typologies of physician financial incentives

Expanded typology	Proportion with incentives to increase services		Reduced form typology
Group 1a	0.061	0.083	Group 1
Employee			Employees
Fixed compensation			Fixed compensation
No explicit incentives
Group 1b	0.109
Employee
Fixed compensation
One or more explicit incentives
Group 2a	0.148	0.160	Group 2
Employee			Employees
Variable compensation			Variable compensation
No explicit incentives			No incentives-or-one or more explicit incentives and high capitation environment
Any capitation
Group 2b	0.168
Employee
Variable compensation
One or more explicit incentives
High capitation environment
Group 3a	0.171	0.186	Group 3
Owner (part or full)			Owners
No explicit incentives			No explicit incentives
Group 3b	0.188
Full owner of solo practice
Group 4a	0.199	0.204	Group 4
Owner (part or full)			Owners
One or more explicit incentives			High capitation and one or more explicit incentives-or-low capitation and incentive combinations other than productivity alone or productivity and practice financial performance
High capitation environment
Group 4b	0.206
Owners
One or more explicit incentives other than productivity alone or productivity and practice financial performance
Low capitation environment
Group 5	0.253	0.253	Group 5
Employee			Employee
Variable compensation			Variable compensation
One or more explicit incentives			One or more explicit incentives
Low capitation environment			Low capitation environment
Group 6a	0.303	0.323	Group 6
Owners			Owners w/low capitation and productivity only or productivity and practice financial performance incentives
Productivity and practice financial performance incentives only
Low capitation
Group 6b	0.367
Owners
Productivity incentive only
Low capitation

3.3 Validation using alternative methods

Using CART and the computerized data mining techniques described above, we developed alternative schemes for combining the ownership, compensation method, and incentive variables into typologies. The three methods produced similar qualitative typologies, in that having a fixed salary with no variable compensation appears to be the single most important factor associated with the absence of an overall financial incentive to increase services to patients. All three also identified high capitation and having productivity incentives as distinct elements of the typology. Predicted probabilities of having an overall financial incentive to increase services to patients constructed from the three approaches were highly correlated (ranging from 0.90 to 0.95).

3.4 Assessing the typology’s criterion validity

Table 4 shows how other relevant indicators of an incentive to increase service delivery to patients vary across the more parsimonious 6-category typology. While these results generally are consistent with those seen for incentives to increase services, they underscore the distinction between physician owners and non-owners. For instance, there is a linear increase in number of PCP visits in the past week, which ranges from 89 visits for employee physicians receiving fixed compensation to 128 visits for owners in a low capitation environment with additional productivity incentives. The one exception to this monotonic relationship is Group 5: employed physicians who receive variable compensation plus additional incentives in a low capitation environment had 104 visits. A similar pattern is observed for the other measures. For instance, Group 1 physicians’ earnings were just over $15,000 less than predicted, whereas Group 6 physicians earned over $20,000 more than predicted.

Table 4.

Association of the reduced form typology with other related outcomes

Compensation arrangement	Incentives to increase services (from Table 3)	Visits (PCPs only)	Difference between actual and predicted income ($)	Direct patient care hours	Percent who agree that they can make clinical decisions without reducing income
Group 1	0.083	89.2	−15,177	40.0	0.91
Employees
Fixed compensation
Group 2	0.160	107.0	−8,236	42.1	0.88
Employees
Variable compensation
No incentives-or-incentives and high capitation environment
Group 3	0.186	125.4	−6,569	45.8	0.76
Owners
No explicit incentives
Group 4	0.204	125.3	22,041	47.5	0.79
Owners
Incentives [describe better]
Either high capitation or low capitation incentives combs
Other than productivity only or prod +PFP only
Group 5	0.253	104.3	−8,745	43.5	0.86
Employee
Variable compensation
Incentives
Low capitation
Environment
Group 6	0.323	127.5	20,674	49.7	0.81
Owners w/low capitation and productivity only or productivity + PFP only incentives

3.5 Regression model to assess factors associated with the typology

We used the six-category typology to construct a quasi-continuous dependent variable by assigning each physician the value (from Table 3) corresponding to the proportion of physicians in his/her typology cell who reported an incentive to increase services. More revenue from Medicaid, female sex, and being a foreign medical school graduate are all associated with a lower likelihood of reporting incentives to increase services. Conversely, compared to physicians in small group settings, all other practice settings except larger groups are associated with an approximately 5% lower proportion of reporting financial incentives to increase services. Perceptions of a more competitive market environment and a higher (more generous) private insurance physician payment index are both associated with a higher likelihood of reporting incentives to increase service volume (Table 5).

Table 5.

Regression results examining the likelihood of reporting incentives to increase services⁺

Variables	Beta (6 cat)	P-value	Beta (11 cat)	P-value
% Practice revenue from Medicaid
0–5	–	–	–	–
6–20	−0.009	0.002	−0.008	0.007
21+	−0.013	0.002	−0.014	0.001
% Practice revenue from Medicare
0–20	–	–	–	–
21–40	0.002	0.56	0.004	0.23
41+	−0.0002	−0.05	0.0001	0.97
Age
<35	–	–	–	–
35–39	0.005	0.40	0.005	0.38
40–44	0.011	0.04	0.012	0.03
45–49	0.009	0.08	0.009	0.08
50–54	0.015	0.004	0.015	0.005
55–59	0.008	0.13	0.009	0.06
60–64	0.015	0.02	0.014	0.02
65–69	−0.003	0.65	−0.002	0.80
70+	0.010	0.13	0.011	0.08
Sex
Female	−0.009	0.008	−0.009	0.003
Medical school training
Foreign medical school graduate	−0.011	0.005	−0.011	0.007
Specialty
Medical and surgical specialty	0.005	0.03	0.005	0.07
Practice setting
Solo/two physician practice	−0.050	< 0.001	−0.047	< 0.001
Small group	–	–	–	–
Medium/large group	−0.007	0.13	−0.003	0.49
Group/staff model HMO	−0.060	< 0.001	−0.057	< 0.001
Medical school	−0.058	< 0.001	−0.055	< 0.001
Hospital-based	−0.051	< 0.001	−0.049	< 0.001
Other	−0.078	< 0.001	−0.077	< 0.001
Perception of competition
Not at all competitive	–	–	–	–
Somewhat competitive	0.013	< 0.001	0.014	< 0.001
Very competitive	0.012	< 0.001	0.013	< 0.001
Number of PCPs per 10,000	0.0005	0.62	0.0003	0.79
Number of specialists per 10,000	−0.0001	0.79	−0.0000	0.92
Physician payment index	0.065	< 0.001	0.067	< 0.001
Percent uninsured	−0.0008	0.06	−0.0008	0.09

⁺For the dependent variable, physicians were assigned the average predicted probability among physicians in their group of reporting incentives to increase services. Beta coefficients are scaled to represent a 10% point change in the dependent variable

^*<0.05;

^**<0.01;

^***<0.001

4 Discussion

Physician financial incentives are potent levers available to policy makers and purchasers to influence the delivery of health care services. Understanding the combinations of payment strategies and organizational arrangements that influence physician behavior is important for understanding the potential impact of different policy initiatives on trends in health care spending. Our study has several notable findings. First, a parsimonious typology consisting of six mutually exclusive groups reasonably captures the continuum of complex financial incentives that encourage physicians to increase service volume. Second, we found few differences between primary care and specialist physicians, or between procedural-based and cognitive specialties, suggesting that physicians experience payment arrangements in a similar way, even if the magnitude of their total compensation differs markedly. Finally, full or partial ownership, particularly for those in small to medium-sized groups, appears to be an overriding factor in physician perceptions of their financial incentives.

We developed our typology based on responses to a previously validated item asking physicians to report whether their overall financial incentives were to increase services to patients, decrease services to patients, or were neutral. This question has been shown to be associated with physicians’ contractual arrangements with health plans as well as other aspects of their organizational arrangements. (Mitchell et al. 2000b; Reschovsky et al. 2006) Although our typology should eventually be linked to actual physician behavior, the magnitude of the differences we observed between groups on this summary item appears to be large. We further explored the validity of the typology through quantitative means and evaluated its criterion validity by examining other variables arguably related to incentives to increase services to patients, including overall number of visits per week, hours worked, and income.

Although there are many nuanced arrangements that are used to compensate physicians, the dominant factor influencing perceived incentives appears to be ownership (Mitchell 2005, 2008; Reschovsky et al. 2006). Physicians in small and medium sized groups mostly are owners of small businesses. As such, regardless of how the practice chooses to compensate member physicians, their livelihood depends on maximizing the profit from their medical practice. Given that the dominant form of physician payment continues to be some form of fee-for-service, owners clearly reflect the inherent incentives in FFS to increase services, as long as fees cover costs. This finding is clearly shown in analyses that examine end points such as visits and predicted income, which suggest that employed physicians who experience incentives to increase services (Group 5 in the typology) don’t necessarily respond like owners, perhaps because they do not experience the full financial benefit of actions they might take to increase service delivery. Although additional incentive programs including pay-for-performance have become more prevalent in recent years (Rosenthal et al. 2004; Rosenthal et al. 2006), our results suggest that the magnitude of these arrangements in 2004–2005 was not sufficient to match the incentives inherent in ownership and basic payment mechanisms for physicians who are not owners in spite of the fact that quality-based pay-for-performance measures largely address treatment patterns where services are underprovided, rather than treatment patterns where services are overprovided (Pham et al. 2009).

Our data are subject to several limitations. First, although our sample is the largest nationally representative study of physicians that we are aware of, the response rate leaves open the potential for bias. Prior analyses show few differences between responders and non-responders, and the CTS’ stringent weighting procedures should also dampen potential response bias (The American Association for Public Opinion Research (AAPOR) 2008).

Second, we developed our typology based on a single item regarding physicians’ perceived incentives. Although this item has been previously validated (Hadley and Mitchell 2002; Mitchell et al. 2000b), our typology may not apply to the other physician behaviors, including reducing services provided to patients, delivering high quality care, or increasing patient throughput. In addition, as shown in Table 4, we used several additional dependent variables to validate our approach. The relationships of the typology with these additional variables were all in the expected directions. Third, we did not validate our typology against actual physician behavior. This activity is planned in the future. Although the questions we used are among the most detailed we are aware of in a survey of physicians, we could not ask additional important questions that, for instance, would have defined the magnitude of incentives. Such detailed information is likely infeasible to collect in the context of this type of survey. Finally, we recognize that physician self selection into practices with different compensation structures may limit the typology’s use for making casual inferences.

In summary, we used nationally representative data from over 6,500 physicians on both their payment arrangements and practice characteristics to create a parsimonious typology of physician payment arrangements that can be used for future study of physician incentive arrangements. Our findings will be of use to policy makers and payers designing financial incentives and to future research seeking to quantify the influence of various compensation arrangements on the provision of care.