Abstract
Importance
Although many factors influence the management of carotid artery stenosis, it is not well understood whether a preference toward procedural management exists when procedural volume and physician compensation are linked in the fee-for-service environment.
Objective
To explore evidence for provider-induced demand in the management of carotid artery stenosis.
Design, Setting, and Participants
The Department of Defense Military Health System Data Repository was queried for individuals diagnosed with carotid artery stenosis between October 1, 2006, and September 30, 2010. A hierarchical multivariable model evaluated the association of the treatment system (fee-for-service physicians in the private sector vs salary-based military physicians) with the odds of procedural intervention (carotid endarterectomy or carotid artery stenting) compared with medical management. Subanalysis was performed by symptom status at the time of presentation. The association of treatment system and of management strategy with clinical outcomes, including stroke and death, was also evaluated. Data analysis was conducted from August 15, 2015, to August 2, 2016.
Main Outcomes and Measures
The odds of procedural intervention based on treatment system was the primary outcome used to indicate the presence and effect of provider-induced demand.
Results
Of 10 579 individuals with a diagnosis of carotid artery stenosis (4615 women and 5964 men; mean [SD] age, 65.6 [11.4] years), 1307 (12.4%) underwent at least 1 procedure. After adjusting for demographic and clinical factors, the odds of undergoing procedural management were significantly higher for patients in the fee-for-service system compared with those in the salary-based setting (odds ratio, 1.629; 95% CI, 1.285-2.063; P < .001). This finding remained true when patients were stratified by symptom status at presentation (symptomatic: odds ratio, 2.074; 95% CI, 1.302-3.303; P = .002; and asymptomatic: odds ratio, 1.534; 95% CI, 1.186-1.984; P = .001).
Conclusions and Relevance
Individuals treated in a fee-for-service system were significantly more likely to undergo procedural management for carotid stenosis compared with those in the salary-based setting. These findings remained consistent for individuals with and without symptomatic disease.
This database study examines evidence of provider-induced demand in the management of carotid artery stenosis.
Key Points
Question
Is the variation in care for carotid stenosis seen between reimbursement systems consistent with the concept of provider-induced demand?
Findings
In this database study of 10 579 individuals with carotid artery stenosis, the adjusted odds of undergoing carotid endarterectomy or stenting were significantly higher in the fee-for-service setting than in the salary-based setting for both symptomatic and asymptomatic disease.
Meaning
Individuals with both symptomatic and asymptomatic carotid stenosis treated in a fee-for-service system were significantly more likely to undergo procedural management when compared with those in the salary-based system.
Introduction
Stroke is a leading cause of morbidity and mortality in the United States and other developed countries, with an estimated prevalence of 8% and an annual incidence of 2% to 8.8% among the elderly. A major preventable cause of ischemic stroke is occlusive disease of the extracranial carotid arteries due to atherosclerosis. The resulting carotid stenosis (CS) can be managed either through reduction of risk factors and medical management or interventions such as carotid endarterectomy or carotid artery stenting.
The choice of therapy for CS depends on several clinical factors, including the degree of stenosis and whether the patient is symptomatic or asymptomatic from the lesion. Several large randomized clinical trials have provided evidence of current professional societal guidelines for the treatment of CS. Medical management, consisting of antiplatelet therapy, use of a cholesterol-lowering agent, and control of hypertension, is recommended for all patients with CS. In addition, carotid endarterectomy is also recommended for symptomatic patients with 50% to 99% stenosis, as well as for patients with 60% to 99% stenosis who are asymptomatic but have a good risk profile. Based on the results of randomized clinical trials and meta-analyses, stenting is generally reserved for patients with high surgical risk and is performed by clinicians who have placed a high volume of carotid artery stents.
Recently, there has been greater support for medical management in all asymptomatic patients regardless of the degree of stenosis. Proponents of this strategy point to the small relative benefit of surgery compared with medical management in these patients in the age of statins. Complicating this debate is the fact that some of the stakeholders may be influenced by the potential financial gains or losses that accompany any broad change in treatment guidelines.
In the United States, most physicians work in a fee-for-service (FFS) compensation structure in which their income is associated directly with clinical volume. Procedures in particular are reimbursed at a higher relative rate than noninterventional management because of their intensity and complexity. Provider-induced demand (PID) is an economic term that refers to a greater demand for services than what would otherwise be expected in a perfect market. In this context, the term provider refers to physicians who care for patients in different reimbursement models. Provider-induced demand can develop when physicians or other health care professionals influence the consumption of health care through increased procedures, more complex procedures, or both. Provider-induced demand arises from information asymmetry between patients, who are almost inevitably less informed as to the details of their conditions and options for treatment, and clinicians, on whom patients then rely for treatment recommendations. Thus, a potential conflict of interest exists when a clinician serves as the patient’s agent in making medical decisions while also acting as the supplier of those same services being recommended. The conflict can be exacerbated in FFS systems in which clinicians can financially benefit from the recommendations they make on behalf of their patients.
A variety of evidence supports the existence and extent of PID in the modern era. Groups have shown wide disparities in health care spending across geographic regions that is not accounted for by demographic or health factors. A direct association has also been found between the number of surgeons per capita and the rates of surgical procedures performed in a geographic area. Physicians also seem to respond to changes in reimbursement by adjusting the volume of patients cared for in predictable patterns, suggesting that physicians do not behave to maximize incomes but rather to achieve target incomes in a trade-off for other activities, such as leisure time. Those who argue against the existence of PID claim that physicians are bound by professional, ethical, and regulatory restrictions.
Prior work on PID has used data from studies in which health care use was assessed before and after a change in reimbursement or compensation structure. In this study, we sought to determine whether PID exists in the care of CS by analyzing a contemporary population of patients cared for by physicians with different compensation structures. Patients covered by the Military Health System’s TRICARE program can receive their care from both salaried military physicians and FFS physicians in the private sector. We hypothesized that, consistent with PID, patients in the TRICARE program would be more likely to receive interventions for CS from private-sector FFS physicians in the purchased care (PC) system than from their salaried counterparts in the direct care (DC) system. We also hypothesized that the differences in intervention rates in the DC and PC systems would be greater among patients with asymptomatic CS compared with patients with symptomatic CS, for which intervention is less debated, suggesting that the observed variation in rates depends on the level of discretion that the physician has in determining treatment options.
Methods
Patient Data and Cohort Selection
This project took place under the auspices of the Comparative Effectiveness and Provider Induced Demand Collaboration between Brigham and Women’s Hospital and the Uniformed Services University of the Health Sciences. Through this collaboration, we created a database consisting of patients residing within the continental United States enrolled in either TRICARE Prime or TRICARE Plus from October 1, 2006, to September 30, 2010. The source of these data is the Military Health System Data Repository. TRICARE is the health care program serving active members of the US Uniformed Services, retirees, and their families. TRICARE is a part of the US Department of Defense Military Health System and is administered by the Defense Health Agency. This study was approved by the Partners Institutional Review Board. As this was a retrospective analysis of a large database with no patient contact, the Partners Institutional Review Board waived the need for consent. Patient data were deidentified before analysis.
We restricted our analysis to patients enrolled in TRICARE Prime and TRICARE Plus, who are each assigned a primary care manager who oversees the patient’s care. Although TRICARE Prime enrollees could possibly receive care that is not performed within or reimbursed by the Military Health System, similar to health maintenance organizations, it seems more likely that the patient care will be consolidated within TRICARE since each enrollee has a primary care manager and, as such, their care is more completely documented. The same reasoning holds for TRICARE Plus enrollees, who are older than 65 years but subject to similar business processes (ie, are assigned a primary care manager). Second, patients in TRICARE Prime or Plus have low or no enrollment fees and copayments and no deductibles, which essentially eliminates the possibility that patient out-of-pocket costs might be a potential factor in treatment decisions. Medicare eligibility by age or disability is also tracked in the Military Health System Data Repository and was included as a covariate to account for dual enrollment. The Military Health System is distinct in both structure and mandate from the US Department of Veterans Affairs.
Statistical Analysis
Data analysis was conducted from August 15, 2015, to August 2, 2016. The longitudinal medical record of each patient was divided into encounters consisting of a CS diagnosis, a query for carotid intervention vs medical management only, and clinical end points of stroke, transient ischemic attack, acute myocardial infarction, and mortality. To minimize diagnostic variability, only individuals who received a diagnosis of CS by a cardiologist or vascular surgeon were included in the analysis. Clinical outcomes were then referenced to time periods of 30 days, 1 year, and 2 years after carotid intervention or diagnosis of CS. Patients were assigned to the DC or PC designation based the system in which they received their diagnosis.
Bivariate models were performed using 2-tailed t tests for continuous variables and χ2 tests for frequencies. Multivariable hierarchical logistic regression models (PROC MIXED; SAS Institute Inc) for the dependent outcome of carotid intervention (carotid endarterectomy or carotid artery stenting) were created for all encounters. Independent factors in the model included patient demographics and CS-associated clinical factors, with risk stratification by use of the hierarchical condition categories (HCCs). Analysis was clustered by catchment area as a proxy for geographical region. An HCC community score was also calculated to generate an index score of comorbidity for group comparisons. This calculation was repeated for clinical outcome variables including stroke, death, and a composite outcome variable including stroke, transient ischemic attack, and acute myocardial infarction. An α of .05, corresponding to P = .05 and 95% CIs, was used as the criterion for statistical significance. Data manipulation and statistical analysis were performed using SAS, version 9.4 (SAS Institute Inc).
Results
From October 1, 2006, to September 30, 2010, a total of 47 420 Prime and Plus enrollees had at least 1 diagnosis of carotid artery stenosis. Of those, 10 579 were diagnosed by either a cardiologist or vascular surgeon, and 1307 (12.4%) also underwent a carotid intervention (endarterectomy or stenting) rather than medical management alone. On average, patients using the PC system (n = 3589) were younger than those using the DC system (n = 6990) (mean [SD] age, 58.1 [7.6] vs 69.5 [11.0] years at first diagnosis; P < .001; 61.7 [6.4] vs 69.5 [9.3] years at procedure; P < .001), more likely to be symptomatic (641 [17.9%] vs 271 [3.9%]; P < .001), more likely to be female (1777 [49.5%] vs 2838 [40.6%]; P < .001), more likely to be married (3022 [84.2%] vs 5509 [78.8%]; P < .001), and less likely to be eligible for Medicare (346 [9.6%] vs 4722 [67.6%]; P < .001) (Table 1). Patients in the PC group were also less likely to report current or prior tobacco use (962 [26.8%] vs 3149 [45.1%]; P < .001) and had lower rates of diabetes (406 [11.3%] vs 1220 [17.5%], P < .001), congestive heart failure (153 [4.3%] vs 400 [5.7%]; P = .001), previous myocardial infarction (77 [2.1%] vs 255 [3.6%]; P < .001), peripheral vascular disease (198 [5.5%] vs 652 [9.3%]; P < .001), chronic obstructive pulmonary disease (194 [5.4%] vs 508 [7.3%]; P < .001), and renal failure (136 [3.8%] vs 533 [7.6%]; P < .001) (Table 1). In univariate analysis, with medical management, the percentage of patients who developed stroke was significantly higher among the PC group than those in the DC group at 30 days (PC, 271 of 2943 [9.2%]; DC, 145 of 6329 [2.3%]; P < .001), 1 year (PC, 353 [12.0%]; DC, 282 [4.5%]; P < .001), and 2 years (PC, 393 [13.4%]; DC, 361 [5.7%]; P < .001) (Table 2). The percentage of patients who developed transient ischemic attack was also higher in the PC group than the DC group at 30 days (PC, 315 [10.7%]; DC, 252 [4.0%]; P < .001), 1 year (PC, 457 [15.5%]; DC, 450 [7.1%]; P < .001), and 2 years (PC, 524 [17.8%]; DC, 565 [8.9%]; P < .001). Acute myocardial infarction at 30 days (PC, 56 [1.9%]; DC, 79 [1.2%]; P = .01) was also higher among the PC group. Death at 1 year (PC, 51 [1.7%]; DC, 165 [2.6%]; P = .009) and at 2 years (PC, 82 [2.8%]; DC, 327 [5.2%]; P < .001) was lower among the PC group.
Table 1. Demographics of Study Cohort.
| Characteristic | Valuea | P Value | ||
|---|---|---|---|---|
| Carotid Stenosis (n = 10 579) |
Direct Care (n = 6990) |
Purchased Care (n = 3589) |
||
| Any procedure | 1307 (12.4) | 661 (9.5) | 646 (18.0) | <.001 |
| Age, mean (SD), y | ||||
| At initial diagnosis | 65.6 (11.4) | 69.5 (11.0) | 58.1 (7.6) | <.001 |
| At procedure | 65.7 (8.9) | 69.5 (9.3) | 61.7 (6.4) | |
| Symptom status at presentation | ||||
| Symptomatic | 912 (8.6) | 271 (3.9) | 641 (17.9) | <.001 |
| Asymptomatic | 9667 (91.4) | 6719 (96.1) | 2948 (82.1) | |
| Race/ethnicity | ||||
| White | 4284 (40.5) | 2724 (39.0) | 1560 (43.5) | <.001 |
| Black | 719 (6.8) | 455 (6.5) | 264 (7.4) | |
| Other | 334 (3.2) | 270 (3.9) | 64 (1.8) | |
| Unknown | 5242 (49.6) | 3541 (50.7) | 1701 (47.4) | |
| Marital status | ||||
| Married | 8531 (80.6) | 5509 (78.8) | 3022 (84.2) | <.001 |
| Single | 2048 (19.4) | 1481 (21.2) | 567 (15.8) | |
| Sex | ||||
| Female | 4615 (43.6) | 2838 (40.6) | 1777 (49.5) | <.001 |
| Male | 5964 (56.4) | 4152 (59.4) | 1812 (50.5) | |
| History of smoking | 4111 (38.9) | 3149 (45.1) | 962 (26.8) | <.001 |
| Beneficiary category | ||||
| Dependent | 4338 (41.0) | 2695 (38.6) | 1643 (45.8) | <.001 |
| Sponsor | 6241 (59.0) | 4295 (61.4) | 1946 (54.2) | |
| Medicare eligibility | ||||
| Yes | 5068 (47.9) | 4722 (67.6) | 346 (9.6) | <.001 |
| No | 5511 (52.1) | 2268 (32.5) | 3243 (90.4) | |
| HCC, mean (SD) | 0.7 (0.6) | 0.8 (0.6) | 0.6 (0.5) | <.001 |
| Comorbidities | ||||
| Diabetes | 1626 (15.4) | 1220 (17.5) | 406 (11.3) | <.001 |
| CHF | 553 (5.2) | 400 (5.7) | 153 (4.3) | .001 |
| Acute MI | 154 (1.5) | 94 (1.3) | 60 (1.7) | .18 |
| Prior MI | 332 (3.1) | 255 (3.6) | 77 (2.1) | <.001 |
| Stroke | 259 (2.5) | 154 (2.2) | 105 (2.9) | .02 |
| COPD | 702 (6.6) | 508 (7.3) | 194 (5.4) | <.001 |
| Renal failure | 669 (6.3) | 533 (7.6) | 136 (3.8) | <.001 |
| Peripheral vascular disease | 850 (8.0) | 652 (9.3) | 198 (5.5) | <.001 |
| Peripheral vascular disease (complicated) | 137 (1.3) | 101 (1.4) | 36 (1.0) | .06 |
Abbreviations: CHF, congestive heart failure; COPD, chronic obstructive pulmonary disease; HCC, hierarchical condition category; MI, myocardial infarction.
Data are presented as number (percentage) of patients unless otherwise indicated.
Table 2. Clinical Outcomes of Intervention or Medical Management.
| Characteristic | Group, No. (%) | P Value | |
|---|---|---|---|
| Direct Care | Purchased Care | ||
| Medical Managementa | |||
| Death | |||
| 30 d | 18 (0.3) | 10 (0.3) | .65 |
| 1 y | 165 (2.6) | 51 (1.7) | .009 |
| 2 y | 327 (5.2) | 82 (2.8) | <.001 |
| Stroke | |||
| 30 d | 145 (2.3) | 271 (9.2) | <.001 |
| 1 y | 282 (4.5) | 353 (12) | <.001 |
| 2 y | 361 (5.7) | 393 (13.4) | <.001 |
| Transient ischemic attack | |||
| 30 d | 252 (4) | 315 (10.7) | <.001 |
| 1 y | 450 (7.1) | 457 (15.5) | <.001 |
| 2 y | 565 (8.9) | 524 (17.8) | <.001 |
| Acute myocardial infarction | |||
| 30 d | 79 (1.2) | 56 (1.9) | .01 |
| 1 y | 229 (3.6) | 117 (4) | .40 |
| 2 y | 348 (5.5) | 147 (5) | .32 |
| Composite 1b | |||
| 30 d | 158 (2.5) | 277 (9.4) | <.001 |
| 1 y | 424 (6.7) | 393 (13.4) | <.001 |
| 2 y | 641 (10.1) | 457 (15.5) | <.001 |
| Composite 2c | |||
| 30 d | 442 (7) | 538 (18.3) | <.001 |
| 1 y | 937 (14.8) | 758 (25.8) | <.001 |
| 2 y | 1277 (20.2) | 861 (29.3) | <.001 |
| Procedured | |||
| Death | |||
| 30 d | 2 (0.3) | 4 (0.6) | .40 |
| 1 y | 16 (2.4) | 14 (2.2) | .76 |
| 2 y | 25 (3.8) | 19 (2.9) | .40 |
| Stroke | |||
| 30 d | 44 (6.7) | 72 (11.2) | .004 |
| 1 y | 68 (10.3) | 99 (15.3) | .006 |
| 2 y | 78 (11.8) | 112 (17.3) | .004 |
| Transient ischemic attack | |||
| 30 d | 52 (7.9) | 73 (11.3) | .03 |
| 1 y | 80 (12.1) | 102 (15.8) | .053 |
| 2 y | 88 (13.3) | 119 (18.4) | .01 |
| Acute myocardial infarction | |||
| 30 d | 17 (2.6) | 12 (1.9) | .38 |
| 1 y | 36 (5.4) | 23 (3.6) | .10 |
| 2 y | 50 (7.6) | 30 (4.6) | .03 |
| Composite 1b | |||
| 30 d | 45 (6.8) | 74 (11.5) | .003 |
| 1 y | 78 (11.8) | 107 (16.6) | .01 |
| 2 y | 96 (14.5) | 124 (19.2) | .02 |
| Composite 2c | |||
| 30 d | 94 (14.2) | 125 (19.4) | .01 |
| 1 y | 148 (22.4) | 181 (28.1) | .02 |
| 2 y | 176 (26.6) | 213 (33) | .01 |
Times reported as time from diagnosis. Direct care, n = 6329; purchased care, n = 2943.
Stroke or death.
Stroke, transient ischemic attack, myocardial infarction, or death.
Carotid endarterectomy or stenting; times reported as time from procedure. Direct care, n = 661; purchased care, n = 646.
For patients who underwent a procedure, there was no significant difference in death at 30 days (PC, 4 of 646 [0.6%]; DC, 2 of 661 [0.3%]; P = .40), 1 year (PC, 14 [2.2%]; DC, 16 [2.4%]; P = .76), and 2 years (PC, 19 [2.9%]; DC, 25 [3.8%]; P = .40). Stroke at 30 days (PC, 72 [11.2%]; DC, 44 [6.7%]; P = .004), 1 year (PC, 99 [15.3%]; DC, 68 [10.3%]; P = .006), and 2 years (PC, 112 [17.3%]; DC, 78 [11.8%]; P = .004) and transient ischemic attack at 30 days (PC, 73 [11.3%]; DC, 52 [7.9%]; P = .03) and 2 years (PC, 119 [18.4%]; DC, 88 [13.3%]; P = .01) were also each more common in the PC group; however, at 2 years, acute myocardial infarction was more common among the DC group (PC, 30 [4.6%]; DC, 50 [7.6%]; P = .03) (Table 2). Composite outcomes were more common among the PC group than the DC group at all time points (Table 2).
In multivariable analysis, patients in the PC system were more likely to receive an intervention for CS than were patients in the DC system (odds ratio [OR], 1.629; 95% CI, 1.285-2.063; P < .001) (Table 3). The odds of receiving an intervention were also higher among symptomatic patients than among asymptomatic patients (OR, 9.487; 95% CI, 7.611-11.824; P < .001). A higher HCC score was associated with lower odds of receiving an intervention (OR, 0.815; 95% CI, 0.730-0.910; P < .001). In subanalyses by symptomatic status, the effect of receiving care in the PC system on carotid intervention was higher in symptomatic patients (OR, 2.074; 95% CI, 1.302-3.303; P = .002) compared with asymptomatic patients (OR, 1.534; 95% CI, 1.186-1.984; P = .001) (Table 4).
Table 3. Multivariable Factors Associated With Receiving Carotid Procedure vs Medical Management.
| Factora | Odds Ratio (95% CI) | P Value |
|---|---|---|
| Purchased vs direct care | 1.629 (1.285-2.063) | <.001 |
| Age | 1.027 (1.015-1.038) | <.001 |
| Male vs female | 1.245 (1.120-1.384) | <.001 |
| Race/ethnicity | ||
| Black vs white | 0.478 (0.350-0.653) | <.001 |
| Other vs white | 0.853 (0.520-1.399) | .65 |
| Unknown vs white | 0.908 (0.785-1.050) | .14 |
| Married vs single | 1.042 (0.859-1.264) | .68 |
| Symptomatic vs asymptomatic | 9.487 (7.611-11.824) | <.001 |
| Eligible for Medicare vs not eligible for Medicare | 0.691 (0.550-0.868) | .002 |
| Smoking vs nonsmoking | 1.939 (1.677-2.242) | <.001 |
| Hierarchical condition category score | 0.815 (0.730-0.910) | <.001 |
C statistic, 0.732 (Hosmer-Lemeshow goodness-of-fit; P = .81).
Table 4. Medical Management Among Asymptomatic and Symptomatic Cohorts.
| Factor | Odds Ratio (95% CI) | P Value |
|---|---|---|
| Patients With Asymptomatic Carotid Artery Stenosis | ||
| Purchased vs direct care | 1.534 (1.186-1.984) | .001 |
| Age | 1.023 (1.011-1.036) | <.001 |
| Male vs female | 1.280 (1.108-1.478) | <.001 |
| Race/ethnicity | ||
| Black vs white | 0.501 (0.360-0.698) | <.001 |
| Other vs white | 0.896 (0.540-1.487) | .55 |
| Unknown vs white | 0.891 (0.752-1.057) | .34 |
| Married vs single | 1.055 (0.819-1.358) | .68 |
| Eligible for Medicare vs not eligible for Medicare | 0.741 (0.550-0.999) | .049 |
| Smoking vs nonsmoking | 2.017 (1.693-2.404) | <.001 |
| Hierarchical condition category score | 0.756 (0.673-0.849) | <.001 |
| Patients With Symptomatic Carotid Artery Stenosis | ||
| Purchased vs direct care | 2.074 (1.302-3.303) | .002 |
| Age | 1.042 (1.016-1.068) | .001 |
| Male vs female | 1.148 (0.874-1.508) | .32 |
| Race/ethnicity | ||
| Black vs white | 0.431 (0.217-0.855) | .04 |
| Other vs white | 0.726 (0.373-1.413) | .98 |
| Unknown vs white | 0.919 (0.715-1.181) | .16 |
| Married vs single | 0.985 (0.698-1.389) | .93 |
| Eligible for Medicare vs not eligible for Medicare | 0.551 (0.330-0.920) | .02 |
| Smoking vs nonsmoking | 1.731 (1.310-2.287) | <.001 |
| Hierarchical condition category score | 1.017 (0.800-1.295) | .89 |
Clinical outcomes, including stroke, death, and a composite outcome, were also examined in a multivariable model controlling for age, HCC score, Medicare eligibility, sex, race/ethnicity, marital status, smoking history, and symptomatic status. Outcomes were examined at 30 days, 1 year, and 2 years from the time of diagnosis for patients in the medical management group and from the time of procedure (carotid artery stenting or carotid endarterectomy) for those in the procedure group. For all patients, those in the PC group had a higher odds than those in the DC group of stroke at 30 days (OR, 1.981; 95% CI, 1.544-2.542; P < .001), 1 year (OR, 1.599; 95% CI, 1.329-1.923; P < .001), and 2 years (OR, 1.486; 95% CI, 1.235-1.788; P < .001) (Table 5). Odds of all-cause mortality were also higher in the PC group than the DC group for all patients at 30 days (OR, 1.976; 95% CI, 0.424-9.203; P = .39), 1 year (OR, 1.673; 95% CI, 1.165-2.402; P = .005), and 2 years (OR, 1.777; 95% CI, 1.285-2.458; P < .001) (Table 5). Odds of composite outcomes were also significantly higher within the PC setting compared with the DC setting at each time point (Table 5).
Table 5. Multivariable Odds Ratios of Clinical Outcomes for Purchased Care vs Direct Care.
| Time Frame | Odds Ratio (95% CI)a | P Value |
|---|---|---|
| All-cause mortality | ||
| 30 d | 1.976 (0.424-9.203) | .39 |
| 1 y | 1.673 (1.165-2.402) | .005 |
| 2 y | 1.777 (1.285-2.458) | <.001 |
| Stroke | ||
| 30 d | 1.981 (1.544-2.542) | <.001 |
| 1 y | 1.599 (1.329-1.923) | <.001 |
| 2 y | 1.486 (1.235-1.788) | <.001 |
| Composite outcome 1b | ||
| 30 d | 1.974 (1.530-2.548) | <.001 |
| 1 y | 1.601 (1.328-1.929) | <.001 |
| 2 y | 1.530 (1.256-1.863) | <.001 |
| Composite outcome 2c | ||
| 30 d | 1.725 (1.422-2.092) | <.001 |
| 1 y | 1.633 (1.384-1.925) | <.001 |
| 2 y | 1.577 (1.346-1.800) | <.001 |
Adjusted for symptomatic status, age, hierarchical condition category score, sex, race/ethnicity, marital status, smoking history, and Medicare eligibility. Odds ratio >1 favors the purchased care group.
Stroke or death.
Stroke, transient ischemic attack, myocardial infarction, or death.
Discussion
Carotid stenosis is a prevalent manifestation of atherosclerotic disease with significant morbidity and mortality. Although several prospective clinical trials have shown a potential benefit for carotid intervention in many patients, particularly those with symptomatic or high-grade stenosis, management of CS remains a topic of much debate. Advocates of medical management for CS assert that recent developments in medications and management goals have greatly improved outcomes of medical management compared with intervention. This clinical equipoise helps to establish an environment in which provider discretion and patient preference may influence treatment decisions. It also creates a situation in which nonclinical factors, including potential financial compensation, may influence care.
When controlling for demographic and clinical variables, our analysis found a significant increase in the odds of undergoing an intervention for CS in the PC setting of the Military Health System. What is unique about the population in question is that, rather than demonstrating variation in health care by geography or over time, the current study demonstrates variation in health care use by clinician reimbursement type in a synchronous national cohort. The finding that the odds of intervention are higher among patients in the PC group compared with those in the DC group suggests that such variation may be associated with clinician compensation structure.
Our findings are consistent with prior publications that demonstrate PID. Helmchen and Lo Sasso demonstrated increasing rates of encounters and procedures among primary care physicians when transitioning from a salaried reimbursement structure to an FFS model. Nguyen and Derrick also demonstrated an increase in clinician volume to recoup losses to decreasing Medicare reimbursement, with a 10% decrease in Medicare reimbursement triggering an estimated 4% to 8% increase in volume for the clinicians facing the largest price reductions. Falling birth rates have also been shown to trigger an increase in cesarean deliveries, a more lucrative procedure compared with vaginal delivery.
Although our results are consistent with these findings, based on these results alone it is difficult to claim whether the differences in intervention rates between patients in the DC setting and those in the PC setting represent overtreatment in the PC system or undertreatment in the DC system. We did attempt to address this issue, however, via subanalysis using symptomatic status in CS as a proxy for appropriateness of care. Most would agree that, for symptomatic patients, there is greater consensus about the benefits of intervention than there is for asymptomatic patients. Thus, one would expect a higher rate of appropriate (non-PID) intervention in symptomatic patients; in turn, the comparative rate of intervention for these patients would suggest that the system performed at what could be thought to be a more appropriate level of care. In this study, symptomatic patients, who by definition were at higher risk for complications after intervention, were also more likely to receive intervention in the PC system, suggesting a more aggressive clinical approach by clinicians in the PC setting. In other words, interventions are often indicated for many diseases, and higher rates of interventions alone are not detrimental. Our data showed that symptomatic patients within the DC system were less likely to undergo a procedure from which they may have benefited, which may represent undertreatment within the DC system rather than overtreatment in the PC system. Taken together, the evidence also suggests a component of clinical aggressiveness and risk-taking that explains some of the differences in health care use seen. This finding is consistent with prior work showing an association between clinicians in competitive markets and their use of riskier kidneys in renal transplantation for patients with fewer alternatives.
Limitations
This study has significant limitations. The source data are administrative and are subject to coding error and cases of inadequate specificity, as is most notable in the coding of the race/ethnicity value, for which 5242 patients (49.6%) had missing information, a previously described limitation of this data set. The diagnosis of CS is available for analysis, but the degree of stenosis is not available for inclusion. As such, we cannot completely control for the appropriateness of the decisions beyond patients having a diagnosis of CS and an assumption that clinicians act within the boundaries of acceptable medical practice. Information surrounding the compensation structure for specific physicians was not available, and it is likely that a subset of physicians within the PC system are employed by their institutions. However, while physicians may not be reimbursed in an immediate FFS manner, they and the system in which they operate remain subject to these pressures, reflected in productivity bonuses and Relative Value Unit goals that may incentivize physicians or physician groups toward the provision of procedures. Furthermore, while salaries are a major portion of clinician compensation, other incentives exist that may influence a clinician’s treatment decisions. Surgeons who perform a high volume of procedures may gain personal and social benefits of being recognized for their proficiency and expertise regardless of their compensation system. Interventionists also seek to maintain technical proficiency and thus would be motivated to maintain clinical volume. To the extent that these factors exist, they would likely bias the findings toward the null rather than toward our findings since they can affect clinicians in both the PC and DC systems. Finally, it is possible that these results reflect variation in patient mix or referral patterns incompletely captured by the data available. Although the HCC score was used to control for comorbidities, it is likely that clinical factors also affected management decisions at the level of individual clinicians. Despite this limitation, which is common to most studies that use administrative data, on balance, the overall trend in the provision of care does appear to be in line with PID as an economic model.
These limitations notwithstanding, our analysis represents a very large national cohort of patients who concurrently exist in 1 of 2 health care systems that differ by the compensation mechanism of its clinicians, a model somewhat unique to this data set. The assignment to each system greatly depends on the geographical location of the patient, a factor in itself not associated with the clinical presentation or outcome of treatment. The higher rates of intervention for CS in the PC setting are consistent with PID for this disease in this patient population. However, the treatment of symptomatic patients in this cohort suggests that the clinicians in the PC system are appropriately aggressive toward CS.
Conclusions
Although it is difficult to capture fully the factors that motivate patients and clinicians, with respect to the management of CS by surgeons and other interventionists, our results do appear to support the conclusion that PID may be at work. Given these findings, the health care community should focus on ways to detect the potential for PID and craft policies that will align the incentives of patients, clinicians, and society. Further analysis of the appropriateness of care and noncompensation incentives may improve our understanding of the role between incentives and health care use.
References
- 1.Centers for Disease Control and Prevention (CDC) Prevalence of stroke—United States, 2006-2010. MMWR Morb Mortal Wkly Rep. 2012;61(20):379-382. [PubMed] [Google Scholar]
- 2.Koton S, Schneider ALC, Rosamond WD, et al. Stroke incidence and mortality trends in US communities, 1987 to 2011. JAMA. 2014;312(3):259-268. [DOI] [PubMed] [Google Scholar]
- 3.Brott TG, Halperin JL, Abbara S, et al. 2011 ASA/ACCF/AHA/AANN/AANS/ACR/ASNR/CNS/SAIP/SCAI/SIR/SNIS/SVM/SVS guideline on the management of patients with extracranial carotid and vertebral artery disease: a report of the American College of Cardiology Foundation/American Heart Association Task Force on Practice Guidelines, and the American Stroke Association, American Association of Neuroscience Nurses, American Association of Neurological Surgeons, American College of Radiology, American Society of Neuroradiology, Congress of Neurological Surgeons, Society of Atherosclerosis Imaging and Prevention, Society for Cardiovascular Angiography and Interventions, Society of Interventional Radiology, Society of NeuroInterventional Surgery, Society for Vascular Medicine, and Society for Vascular Surgery. J Am Coll Cardiol. 2011;57(8):e16-e94. [DOI] [PubMed] [Google Scholar]
- 4.Antoniou GA, Georgiadis GS, Georgakarakos EI, et al. Meta-analysis and meta-regression analysis of outcomes of carotid endarterectomy and stenting in the elderly. JAMA Surg. 2013;148(12):1140-1152. [DOI] [PubMed] [Google Scholar]
- 5.Mantese VA, Timaran CH, Chiu D, Begg RJ, Brott TG; CREST Investigators . The Carotid Revascularization Endarterectomy vs Stenting Trial (CREST): stenting vs carotid endarterectomy for carotid disease. Stroke. 2010;41(10)(suppl):S31-S34. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Raman G, Moorthy D, Hadar N, et al. Management strategies for asymptomatic carotid stenosis: a systematic review and meta-analysis. Ann Intern Med. 2013;158(9):676-685. [DOI] [PubMed] [Google Scholar]
- 7.Beck DE, Margolin DA. Physician coding and reimbursement. Ochsner J. 2007;7(1):8-15. [PMC free article] [PubMed] [Google Scholar]
- 8.Fisher ES, Wennberg DE, Stukel TA, Gottlieb DJ, Lucas FL, Pinder EL. The implications of regional variations in Medicare spending, part I: the content, quality, and accessibility of care. Ann Intern Med. 2003;138(4):273-287. [DOI] [PubMed] [Google Scholar]
- 9.Gruber J, Owings M. Physician financial incentives and cesarean section delivery. Rand J Econ. 1996;27(1):99-123. [PubMed] [Google Scholar]
- 10.Fuchs VR. The supply of surgeons and the demand for operations. J Hum Resour. 1978;13(suppl):35-56. [PubMed] [Google Scholar]
- 11.Christensen S. Volume responses to exogenous changes in Medicare’s payment policies. Health Serv Res. 1992;27(1):65-79. [PMC free article] [PubMed] [Google Scholar]
- 12.Krasnik A, Groenewegen PP, Pedersen PA, et al. Changing remuneration systems: effects on activity in general practice. BMJ. 1990;300(6741):1698-1701. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Das M. Is There Evidence Against the Induced Demand Hypothesis? Explaining the Large Reduction in Cesarean Rates. New York, NY: Columbia University; 2002. [Google Scholar]
- 14.Nguyen NX, Derrick FW. Physician behavioral response to a Medicare price reduction. Health Serv Res. 1997;32(3):283-298. [PMC free article] [PubMed] [Google Scholar]
- 15.Yip WC. Physician response to Medicare fee reductions: changes in the volume of coronary artery bypass graft (CABG) surgeries in the Medicare and private sectors. J Health Econ. 1998;17(6):675-699. [DOI] [PubMed] [Google Scholar]
- 16.Pope GC, Kautter J, Ellis RP, et al. Risk adjustment of Medicare capitation payments using the CMS-HCC model. Health Care Financ Rev. 2004;25(4):119-141. [PMC free article] [PubMed] [Google Scholar]
- 17.Horton NJ, Bebchuk JD, Jones CL, et al. Goodness-of-fit for GEE: an example with mental health service utilization. Stat Med. 1999;18(2):213-222. [DOI] [PubMed] [Google Scholar]
- 18.Halliday A, Harrison M, Hayter E, et al. ; Asymptomatic Carotid Surgery Trial (ACST) Collaborative Group . 10-Year stroke prevention after successful carotid endarterectomy for asymptomatic stenosis (ACST-1): a multicentre randomised trial. Lancet. 2010;376(9746):1074-1084. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Hobson RW II, Weiss DG, Fields WS, et al. ; The Veterans Affairs Cooperative Study Group . Efficacy of carotid endarterectomy for asymptomatic carotid stenosis. N Engl J Med. 1993;328(4):221-227. [DOI] [PubMed] [Google Scholar]
- 20.McPhee JT, Schanzer A, Messina LM, Eslami MH. Carotid artery stenting has increased rates of postprocedure stroke, death, and resource utilization than does carotid endarterectomy in the United States, 2005. J Vasc Surg. 2008;48(6):1442-1450, 1450.e1. [DOI] [PubMed] [Google Scholar]
- 21.Conrad MF, Boulom V, Mukhopadhyay S, Garg A, Patel VI, Cambria RP. Progression of asymptomatic carotid stenosis despite optimal medical therapy [published correction appears in J Vasc Surg. 2014;59(4):1179]. J Vasc Surg. 2013;58(1):128-135.e1. [DOI] [PubMed] [Google Scholar]
- 22.Sila CA, Higashida RT, Clagett GP. Clinical decisions: management of carotid stenosis. N Engl J Med. 2008;358(15):1617-1621. [DOI] [PubMed] [Google Scholar]
- 23.Helmchen LA, Lo Sasso AT. How sensitive is physician performance to alternative compensation schedules? evidence from a large network of primary care clinics. Health Econ. 2010;19(11):1300-1317. [DOI] [PubMed] [Google Scholar]
- 24.Hobson RW II, Mackey WC, Ascher E, et al. ; Society for Vascular Surgery . Management of atherosclerotic carotid artery disease: clinical practice guidelines of the Society for Vascular Surgery. J Vasc Surg. 2008;48(2):480-486. [DOI] [PubMed] [Google Scholar]
- 25.Adler JT, Sethi RKV, Yeh H, Markmann JF, Nguyen LL. Market competition influences renal transplantation risk and outcomes. Ann Surg. 2014;260(3):550-556. [DOI] [PubMed] [Google Scholar]
- 26.Zogg CK, Jiang W, Chaudhary MA, et al. Racial disparities in emergency general surgery: do differences in outcomes persist among universally insured military patients? J Trauma Acute Care Surg. 2016;80(5):764-775. [DOI] [PubMed] [Google Scholar]