Surgeon Influence on Use of Needle Biopsy in Patients With Breast Cancer: A National Medicare Study

Abstract

Purpose

Use of needle biopsy is a proposed quality measure in the diagnosis and treatment of breast cancer, yet prior literature documents underuse. Nationally, little is known regarding the contribution of a patient's surgeon to needle biopsy use, and knowledge regarding downstream impact of needle biopsy on breast cancer care is incomplete.

Methods

Using 2003 to 2007 nationwide Medicare data from 89,712 patients with breast cancer and 12,405 surgeons, logistic regression evaluated the following three outcomes: surgeon consultation before versus after biopsy, use of needle biopsy (yes or no), and number of surgeries for cancer treatment. Multilevel analyses were adjusted for physician, patient, and structural covariates.

Results

Needle biopsy was used in 68.4% (n = 61,353) of all patients and only 53.7% of patients seen by a surgeon before biopsy (n = 32,953/61,312). Patient factors associated with surgeon consultation before biopsy included Medicaid coverage, rural residence, residence more than 8.1 miles from a radiologic facility performing needle biopsy, and no mammogram within 60 days before consultation. Among patients with surgeon consultation before biopsy, surgeon factors such as absence of board certification, training outside the United States, low case volume, earlier decade of medical school graduation, and lack of specialization in surgical oncology were negatively correlated with receipt of needle biopsy. Risk of multiple cancer surgeries was 33.7% for patients undergoing needle biopsy compared with 69.6% for those who did not (adjusted relative risk, 2.08; P < .001).

Conclusion

Needle biopsy is underused in the United States, resulting in a negative impact on breast cancer diagnosis and treatment. Surgeon-level interventions may improve needle biopsy rates and, accordingly, quality of care.

INTRODUCTION

There is a growing movement in oncology to measure quality of cancer care with the ultimate goal of increasing accountability and improving outcomes.^1–4 As outlined in a recent Institute of Medicine (IOM) report, fundamental to this movement is the need to elucidate the role that physicians play in promoting or impeding quality.^5,6 Yet even in breast cancer, one of the most commonly studied malignancies, the absence of national databases has been a key barrier preventing assessment of physician quality.^2,7

Recent creation of nationally comprehensive cohorts of patients with cancer permits, for the first time, detailed assessment of the role that treating physicians play in promoting cancer care quality on a national basis.^7,8 Use of needle biopsy to establish a diagnosis of breast cancer is an opportune quality measure in which to evaluate the role of the treating physician, because it was developed by the American College of Surgeons and endorsed by the National Quality Forum,^9–15 yet prior studies suggest ongoing underuse.^15,16

Accordingly, in a national cohort of Medicare beneficiaries age 66 years and older with incident breast cancer treated with breast-conserving surgery (BCS) and radiation, we sought to evaluate the influence of the patient's surgeon on use of needle biopsy. To validate the clinical benefit of needle biopsy, we also sought to determine whether receipt of needle biopsy was associated with a lower number of breast cancer surgeries before delivery of adjuvant radiation therapy.

METHODS

Cohort Creation

The study cohort was drawn from national Medicare data encompassing 100% of Medicare beneficiaries' final action institutional (inpatient and outpatient) and provider medical claims. A total of 89,712 incident breast cancers diagnosed from 2003 to 2007 in women age ≥ 66 years with fee-for-service, Part A and B coverage and treated with BCS and radiation were identified using a validated algorithm with more than 90% positive predictive value (Appendix Tables A1 and A2, online only).¹⁶ Limiting the study cohort to patients who underwent BCS followed by radiation ensured a population for whom needle biopsy is the preferred diagnostic option.^9–11,15

Outcomes

We evaluated three outcomes that follow the progression of most women through the continuum of breast cancer detection to treatment (Fig 1). The first outcome is timing of surgeon consultation, with surgeon consultation before biopsy defined as an encounter with a surgeon within 60 days before biopsy and surgeon consultation after biopsy defined as the first encounter with a surgeon taking place after needle biopsy. The second outcome is use of needle biopsy (yes or no), with needle biopsy defined as either core biopsy or fine-needle aspiration within 60 days before initial BCS. The third outcome is the number of breast cancer surgeries, with single breast cancer surgery defined as all BCS and nodal surgery (if indicated) performed on a single date before radiation and multiple breast cancer surgeries defined as any BCS and/or nodal surgery occurring on more than one date before start of radiation (Appendix Table A2). For the purposes of these outcomes, BCS included claims for both segmental mastectomy and excisional biopsy, because both procedures can functionally serve as BCS (Appendix Table A2).

Fig 1.

This flow diagram indicates the three key decisions confronted by patients and their providers during the evaluation and surgical treatment of breast cancer. The first outcome is timing of surgeon consultation (“Prebiopsy surgeon consultation?”) and corresponds to the multivariable model shown in Table 2. The second outcome is type of biopsy (“Type of biopsy determined”) and corresponds to the multivariable model shown in Table 3 in which patients are nested within their diagnosing surgeon (the surgeon who saw the patient before biopsy). The third outcome is number of breast cancer surgeries (“More cancer surgery performed?”) and corresponds to the multivariable model in Table 4, in which patients are nested within their treating surgeon (the surgeon who performed the actual breast-conserving surgery [BCS]). It should be noted that for patients who undergo prebiopsy surgeon consultation but do not receive needle biopsy, the BCS itself functions as a biopsy to establish the cancer diagnosis. BCS includes claims codes for both excisional biopsy and formal segmental mastectomy.

Covariates and Analyses

Patient and structural covariates are listed in Table 1. Charlson comorbidity index was determined using claims generated from 12 months to 1 month preceding initial BCS.¹⁷ Using ArcGIS v10.1 (ESRI, The Redlands, CA), distance from the centroid of the patient's zip code to the centroid of the zip code of the nearest radiologic facility performing a breast needle biopsy in the year of diagnosis was determined for each patient in the cohort (such radiologic facilities were located using nationally comprehensive Medicare claims of all patients with breast cancer).

Table 1.

Patient Demographic and Clinical Characteristics for Entire Cohort and Use of Needle Biopsy

Characteristic	Entire Cohort		Underwent Needle Biopsy		P*
	No. of Patients	%	No. of Patients	%
Entire cohort	89,712	100.0	61,353	68.4
Age, years					< .001
66-69	18,101	20.2	12,572	69.5
70-79	52,101	58.1	35,926	69.0
≥ 80	19,510	21.7	12,855	65.9
Race/ethnicity					< .001
White	82,775	92.3	56,855	68.7
Black	4,888	5.4	3,134	64.1
Other	2,049	2.3	1,364	66.7
Medicaid coverage					< .001
Yes	5,446	6.1	3,298	60.6
No	84,266	93.9	58,055	68.9
Year of BCS					< .001
2003	20,080	22.4	12,205	60.8
2004	18,846	21.0	12,220	64.8
2005	18,132	20.2	12,559	69.3
2006	18,165	20.23	13,290	73.2
2007	14,489	16.2	11,079	76.5
Charlson comorbidity score					< .001
≤ 1	81,892	91.3	56,299	68.7
2	5,413	6.0	3,548	65.5
≥ 3	2,407	2.7	1,506	62.6
Receipt of chemotherapy†					.05
Yes	12,376	13.8	8,559	69.2
No	77,336	86.2	52,794	68.3
County of residence					< .001
Rural	17,719	19.8	10,706	60.4
Urban	71,993	80.3	50,647	70.3
Distance to nearest radiologic facility performing needle biopsy, miles					< .001
≤ 0.6	22,403	25.0	15,663	69.9
0.7-3.4	22,400	25.0	15,946	71.2
3.5-8.1	22,476	25.1	15,722	70.0
≥ 8.2	22,410	25.0	14,010	62.5
Unknown	23	0.0	12	52.2
Timing of surgeon consultation					‡
Before biopsy	61,312	68.4	32,953	53.7
After biopsy	28,400	31.6	28,400	100
Mammography before surgical consultation or needle biopsy§					< .001
Yes	77,808	87.3	53,968	88.0
No	11,904	12.7	7,385	12.0
Visit with PCP before surgical consultation or needle biopsy§					< .001
Yes	58,487	65.2	39,585	68.1
No	31,225	34.8	21,768	69.7
No. of breast cancer surgeries					< .001
Median	1		1
Range	1-8		1-7
Multiple breast surgeries					< .001
Yes	40,381	45.0	20,651	51.1
No	49,331	55.0	40,702	82.5

Abbreviations: BCS, breast-conserving surgery; PCP, primary care provider.

^*Pearson's χ² test, the Wilcoxon test, or the Cochran-Armitage test for trend was used to test for statistical significance at the P = .05 level.

^†Receipt of chemotherapy within 12 months of initial BCS served as a marker of cancer severity.

^‡This P value is not presented because, by definition, receipt of needle biopsy in patients with surgeon consultation after biopsy is 100%.

^§Defined as mammography or PCP encounter, respectively, occurring within 60 days prior to the first of the following two events: surgeon consultation or biopsy.

Each patient was assigned a treating surgeon, defined as the surgeon who performed the initial BCS, in accordance with prior methods.¹⁸ Each patient seen by any one of these surgeons before biopsy was also assigned a diagnosing surgeon, defined as the last surgeon to see the patient within 60 days before biopsy, which may or may not have been the same as the treating surgeon. Surgeon characteristics were determined via linkage to the American Medical Association Physician Master File.

Descriptive statistics characterized associations of patient-level covariates with needle biopsy receipt using Pearson's χ², Wilcoxon, and Cochran-Armitage tests as appropriate. We also mapped the distribution of women in our cohort by hospital referral region (HRR) and overlaid the locations of all radiologic facilities performing needle biopsy (zip code centroids) to examine whether facilities were generally available in areas with large numbers of patients. To evaluate regional variation in receipt of needle biopsy, the proportion of patients receiving needle biopsy at the HRR level was calculated and mapped. A multilevel, intercept-only model of patients nested within HRR evaluated the proportion of variation in needle biopsy receipt attributable to the HRR level.

For the outcome timing of surgeon consultation, logistic regression identified significant patient and structural factors, with goodness of fit evaluated using the Hosmer-Lemeshow test. For the outcome use of needle biopsy among patients with surgeon consultation before biopsy, data structure was hierarchical with patients nested within diagnosing surgeons. We thus applied a multilevel, random intercept, logistic regression model.^19–25 A random intercept was included to account for correlated outcomes for patients treated by the same diagnosing surgeon, whereas all predictors were modeled as fixed effects. Using the intercept-only model, the intraclass correlation coefficient was calculated to determine the proportion of total variance attributable to the surgeon level.^19,26 The ratio of the generalized χ² statistic to its df was evaluated for residual overdispersion. For the outcome of single versus multiple breast cancer surgeries, our data had a hierarchical structure of patients nested within treating surgeons, and thus, a similar multilevel model was used.

For fitting all multivariable models, covariates were included if significant in unadjusted logistic regression at P < .05. Models were iteratively refined to minimize colinearity, with backward selection to remove covariates without statistically significant effects in the final model. Patients and surgeons with unknown covariate values were excluded. Odds ratios were converted to risk ratios.²⁷

All analyses were conducted using SAS version 9.3 (SAS Institute, Cary, NC). This study was granted exempt status by The University of Texas MD Anderson Cancer Center Institutional Review Board.

RESULTS

Patient Characteristics and Needle Biopsy Utilization

We identified 89,712 patients, 11,279 diagnosing surgeons, and 12,405 treating surgeons. Median patient age was 74.0 years (interquartile range [IQR], 70 to 79 years), 7.7% of patients (n = 6,937) were nonwhite, 6.1% (n = 5,446) had Medicaid coverage, and 87.3% (n = 77,808) underwent mammography within 60 days before surgeon consultation or needle biopsy (Table 1). Fifty percent of patients lived within 3.4 miles of a radiologic facility performing needle biopsy, and 90% lived within 18.5 miles of such a facility. Locations of radiologic facilities performing needle biopsy relative to numbers of patients in the cohort at the HRR level are shown in Figure 2A.

Fig 2.

(A) Distribution of radiologic facilities performing breast needle biopsy relative to hospital referral region (HRR) –level distribution of patients in the cohort from 2003 to 2007 (N = 89,689). Gold dots show the locations (zip code centroids) of all radiologic facilities that performed breast needle biopsy on Medicare beneficiaries with a diagnosis code of breast cancer between 2003 and 2007. The number of patients in the cohort who resided in each HRR is shown by quintile to demonstrate the distribution of patients relative to radiologic facilities. Values in parentheses indicate the number of HRRs falling within each quintile. White areas on the map indicate regions with no HRR assignment. HRR of residence could not be determined for 23 patients in the cohort. (B) Percentage of female patients with breast cancer receiving needle biopsy before breast-conserving surgery by HRR from 2003 to 2007 (N = 89,689). HRR-specific needle biopsy rates by quintile are presented for the years 2003 to 2007. In the legend, the percentages refer to quintile cut points for percentage of patients undergoing needle biopsy in each HRR, and the numbers in parentheses indicate the number of HRRs falling within each quintile.

Between 2003 and 2007, 68.4% of patients (n = 61,353) underwent needle biopsy, of whom 92.8% (n = 56,962) underwent core biopsy. The proportion of patients undergoing needle biopsy increased significantly over time, from 60.8% (n = 12,205/20,080) in 2003 to 76.5% (n = 11,079/14,489) in 2007 (P < .001; Table 1). Median interval from needle biopsy to BCS was 21 days (IQR, 14 to 29 days), and median interval from BCS to start of radiation was 40 days (IQR, 26 to 62 days).

Geographic Variation in Needle Biopsy Utilization

HRR-level geographic variation was noted in needle biopsy rates, from a low of 24.1% (95% CI, 16.2% to 34.3%; n = 20/83) in Bismarck, ND, to a high of 97.2% (95% CI, 90.3% to 99.2%; n = 69/71) in Lynchburg, VA (Fig 2B). A map showing HRR-level change in needle biopsy rates over time is available in the Data Supplement. In an intercept-only model of patients nested within HRR, 9.1% of variation in needle biopsy receipt was attributable to the HRR level (P < .001).

Patient and Structural Predictors of the Timing of Surgeon Consultation Relative to Biopsy

In total, 68.4% of patients (n = 61,312) underwent surgeon consultation before biopsy and 31.6% (n = 28,400) underwent surgeon consultation after biopsy (Fig 1). Among patients with surgeon consultation before biopsy, 46.3% (n = 28,359) did not undergo needle biopsy, 38.4% (n = 23,528) underwent needle biopsy performed by their surgeon, and 15.4% (n = 9,425) underwent needle biopsy performed by a nonsurgeon (typically a radiologist), for an overall needle biopsy rate of 53.7% (n = 32,953). By definition, all patients with surgeon consultation after biopsy (n = 28,400) underwent needle biopsy.

In multivariable analysis (n = 89,712), patient characteristics associated with surgeon consultation before biopsy included older age (70.7% of patients age ≥ 80 years underwent prebiopsy surgeon consultation v 67.0% of patients age < 70 years; relative risk [RR], 1.06; P < .001), black race (71.1% for blacks v 68.3% for whites; RR, 1.05; P < .001), Medicaid coverage (73.7% for Medicaid beneficiaries v 68.0% for non–Medicaid beneficiaries; RR, 1.07; P < .001), comorbid illness (73.6% for comorbidity score ≥ 3 v 68.0% for comorbidity score of 0 or 1; RR, 1.07; P < .001), earlier year of diagnosis (73.9% for diagnosis year 2003 v 61.1% for diagnosis year 2007; RR, 1.14; P < .001), rural residence (76.9% for rural residence v 66.3% for urban residence; RR, 1.06; P < .001), longer distance to the nearest radiologic facility performing needle biopsy (74.0% for patients > 8.1 miles v 67.1% for patients < 0.7 miles; RR, 1.07; P < .001), no mammogram in the 60 days before consultation (74.7% for patients without a preconsultation mammogram v 67.4% for patients with a preconsultation mammogram; RR, 1.11; P < .001), and a visit with a primary care provider (PCP) in the 60 days preceding consultation (69.5% for patients with a PCP visit v 66.2% for patients without a PCP visit; RR, 1.05; P < .001; Table 2).

Table 2.

Predictors of Timing of Surgeon Consultation Relative to Biopsy

Predictor	No. of Patients (N = 89,689)	No. of Patients With Prebiopsy Consultation	Unadjusted			Adjusted
			RR	95% CI	P*	RR	95% CI	P*
Age, years
< 70	18,098	12,131	1			1
70-79	52,091	35,371	1.01	1.00 to 1.02	.031	1.01	1.00 to 1.03	.012
≥ 80	19,500	13,792	1.05	1.04 to 1.07	< .001	1.06	1.05 to 1.07	< .001
Race/ethnicity
White	82,754	56,479	1			1
Black	4,886	3,472	1.04	1.02 to 1.06	< .001	1.05	1.03 to 1.07	< .001
Other	2,049	1,343	0.96	0.93 to 0.99	.009	0.96	0.93 to 0.99	.001
Medicaid coverage
Yes	5,442	4,012	1.08	1.07 to 1.10	< .001	1.07	1.05 to 1.09	< .001
No	84,247	57,282	1			1
Charlson comorbidity score
≤ 1	81,872	55,660	1			1
2	5,412	3,864	1.05	1.03 to 1.07	< .001	1.04	1.03 to 1.06	< .001
≥ 3	2,405	1,770	1.08	1.06 to 1.11	< .001	1.07	1.04 to 1.010	< .001
Year of BCS
2003	20,073	14,841	1.16	1.15 to 1.17	< .001	1.14	1.12 to 1.15	< .001
2004	18,841	13,342	1.17	1.16 to 1.19	< .001	1.18	1.16 to 1.19	< .001
2005	18,127	12,280	1.13	1.12 to 1.14	< .001	1.13	1.12 to 1.14	< .001
2006	18,160	11,972	1.09	1.07 to 1.10	< .001	1.09	1.07 to 1.10	< .001
2007	14,488	8,859	1			1
County of residence
Rural	17,713	13,612	1.06	1.05 to 1.07	< .001	1.06	1.05 to 1.08	< .001
Urban	71,976	47,682	1			1
Distance to nearest radiologic facility performing needle biopsy, miles
≤ 0.6	22,403	15,037	1			1
0.7-3.4	22,400	14,343	0.96	0.95 to 0.97	< .001	0.97	0.96 to 0.98	< .001
3.5-8.1	22,476	15,337	1.02	1.00 to 1.03	.011	1.03	1.02 to 1.04	< .001
≥ 8.2	22,410	16,577	1.10	1.09 to 1.12	< .001	1.07	1.05 to 1.08	< .001
Mammography before surgical consultation or needle biopsy
Yes	77,791	52,404	1			1
No	11,898	8,890	1.11	1.10 to 1.12	< .001	1.11	1.10 to 1.12	< .001
Visit with PCP before surgical consultation or needle biopsy
Yes	58,475	40,640	1.05	1.04 to 1.06	< .001	1.05	1.04 to 1.06	< .001
No	31,214	20,654	1			1

Abbreviations: BCS, breast-conserving surgery; PCP, primary care provider; RR, relative risk.

^*Statistical significance evaluated at the P = .05 level in both unadjusted and adjusted analyses. A RR > 1 indicates the factor was associated with a higher likelihood that the patient was evaluated by her surgeon before biopsy. A total of 23 patients were excluded as a result of unknown covariates. Model fit was acceptable (Hosmer-Lemeshow, P = .33).

Patient, Structural, and Surgeon Predictors of Biopsy Type

Among patients with surgeon consultation before biopsy, an intercept-only model (n = 11,279 surgeons and 61,312 patients [68.4% of the overall cohort]) indicated that 28.8% of variation in biopsy type was attributable to the surgeon level (P < .001). After adding fixed effects, patient and structural factors associated with needle biopsy were similar to those factors associated with prebiopsy surgeon consultation (Table 3). Surgeon-level factors associated with needle biopsy included lack of board certification (39.2% needle biopsy rate for patients of non–board-certified surgeons v 54.8% for patients of board-certified surgeons; RR, 1.32; P < .001), training outside the United States (44.1% needle biopsy rate for non–US-trained surgeons v 55.1% for US-trained surgeons; RR, 1.17; P < .001), earlier decade of medical school graduation (50.9% needle biopsy rate for graduation before 1980 v 64.1% for graduation in the 2000s; RR, 1.49; P < .001), low case volume (40.3% needle biopsy rate for one to five patients treated in cohort v 64.4% for ≥ 20 patients treated in cohort; RR, 1.66; P < .001), and specialization in general surgery (51.7% needle biopsy rate for general surgeons v 63.4% for surgical oncologists; RR, 1.11; P < .001).

Table 3.

Patient and Diagnosing Surgeon Factors Associated With Omission of Needle Biopsy Among Patients With Surgeon Consultation Before Biopsy (n = 10,981 diagnosing surgeons; n = 60,018 patients)

Predictor	No. of Patients	No. of Patients With Needle Biopsy	Unadjusted			Adjusted
			RR	95% CI	P*	RR	95% CI	P*
Patient factors
Age, years
< 70	11,836	6,436	1
70-79	34,629	18,748	1.01	0.98 to 1.03	.678			NS
≥ 80	13,553	6,994	1.07	1.04 to 1.11	< .001			NS
Race/ethnicity
White	55,317	29,869	1
Black	3,396	1,679	1.10	1.05 to 1.15	< .001			NS
Other	1,305	630	1.12	1.05 to 1.20	.002			NS
Medicaid coverage
Yes	3,935	1,825	1.14	1.09 to 1.18	< .001	1.06	1.02 to 1.11	.002
No	56,083	30,353	1			1
Year of BCS
2003	14,374	6,706	1.38	1.35 to 1.42	< .001	1.34	1.30 to 1.37	< .001
2004	13,021	6,527	1.28	1.24 to 1.31	< .001	1.24	1.21 to 1.28	< .001
2005	12,075	6,591	1.16	1.12 to 1.20	< .001	1.15	1.11 to 1.18	< .001
2006	11,814	6,998	1.02	0.99 to 1.06	.207	1.02	0.99 to 1.06	.187
2007	8,734	5,356	1			1
Charlson comorbidity score
≤ 1	54,491	29,366	1			1
2	3,789	1,958	1.06	1.01 to 1.10	.013	1.04	1.00 to 1.09	.039
≥ 3	1,738	854	1.09	1.03 to 1.16	.005	1.08	1.02 to 1.14	.014
Receipt of chemotherapy
Yes	8,622	4,879	1			1		< .001
No	51,396	27,299	1.14	1.11 to 1.18	< .001	1.15	1.12 to 1.18
County of residence
Rural	13,264	6,376	1.12	1.09 to 1.16	< .001	1.07	1.03 to 1.10	< .001
Urban	46,754	25,802	1			1
Distance to nearest radiologic facility performing needle biopsy, miles
≤ 0.6	14,739	8,112	1			1
0.7-3.4	14,059	7,729	1.00	0.97 to 1.03	.902	1.00	0.97 to 1.03	.980
3.5-8.1	15,021	8,396	1.00	0.97 to 1.03	.845	1.00	0.97 to 1.03	.861
≥ 8.2	16,199	7,941	1.09	1.06 to 1.12	< .001	1.04	1.01 to 1.07	.017
Mammography before surgical consultation or needle biopsy
Yes	51,529	28,109	1			1
No	8,489	4,069	1.26	1.23 to 1.28	< .001	1.26	1.23 to 1.29	< .001
Diagnosing surgeon factors
Board certified
Yes	56,709	30,960	1			1
No	3,309	1,218	1.51	1.45 to 1.57	< .001	1.32	1.25 to 1.38	< .001
US trained
Yes	52,480	28,852	1			1
No	7,538	3,326	1.38	1.32 to 1.43	< .001	1.17	1.11 to 1.22	< .001
Degree
MD	58,385	31,522	1
DO	1,633	656	1.40	1.30 to 1.50	< .001			NS
Decade of medical school graduation
Before 1980	25,905	13,064	1.53	1.26 to 1.75	< .001	1.49	1.25 to 1.69	< .001
1980-1989	22,544	12,568	1.31	1.05 to 1.54	.018	1.36	1.14 to 1.55	.001
1990-1999	11,413	6,446	1.25	0.98 to 1.49	.069	1.27	1.05 to 1.48	.018
2000+	156	100	1			1
Case volume, No. of patients†
≤ 5	13,496	5,407	1.81	1.77 to 1.85	< .001	1.66	1.62 to 1.70	< .001
> 5 and ≤ 10	12,069	5,968	1.49	1.43 to 1.54	< .001	1.41	1.37 to 1.46	< .001
> 10 and ≤ 20	14,554	8,009	1.31	1.26 to 1.36	< .001	1.27	1.22 to 1.32	< .001
> 20	19,899	12,794	1			1
Specialty
Surgical oncologist	7,944	5,055	1			1
General surgeon	50,714	26,187	1.42	1.34 to 1.51	< .001	1.11	1.04 to 1.19	.001
Other	1,360	936	1.04	0.91 to 1.18	.534	0.77	0.66 to 0.88	< .001
Sex
Male	46,545	23,938	1.40	1.34 to 1.46	< .001			NS
Female	13,473	8,240	1

Abbreviations: BCS, breast-conserving surgery; DO, doctor of osteopathy; MD, doctor of medicine; NS, not significant; RR, relative risk.

^*Statistical significance evaluated at the P = .05 level in both unadjusted and adjusted analyses. A RR > 1 indicates that the factor was associated with a lower likelihood of needle biopsy. In this model, 298 surgeons and 1,294 patients were excluded because of unknown covariates. The ratio of the generalized χ² statistic to its df was 0.85, indicating no residual overdispersion in the model.

^†Surgeon case volume was defined for each surgeon as the number of patients in the cohort who underwent BCS performed by that surgeon during the time interval of the study.

Association of Needle Biopsy With Number of Breast Cancer Surgeries

Overall, risk of multiple breast cancer surgeries was 33.7% (n = 20,651/61,353) for women who underwent needle biopsy versus 69.6% (n = 19,730/28,359) for women who did not (adjusted RR, 2.08; 95% CI, 2.07 to 2.11; P < .001). There was only a small difference in risk of multiple breast cancer surgeries based on the type of provider who performed the needle biopsy, with a 32.7% (n = 12,361/37,825) risk of multiple breast cancer surgeries if the needle biopsy was performed by a nonsurgeon compared with a 35.2% (n = 8,290/23,528) risk if the needle biopsy was performed by a surgeon (adjusted RR, 1.03; P = .005; Fig 1, Table 4).

Table 4.

Patient and Treating Surgeon Factors Associated With Number of Breast Surgeries (n = 12,013 surgeons; n = 88,272 patients)

Predictor	No. of Patients	No. of Patients With Multiple Breast Cancer Surgeries	Unadjusted			Adjusted
			RR	95% CI	P*	RR	95% CI	P*
Patient factors
Type of biopsy
Needle biopsy by nonsurgeon	37,272	12,232	1			1
Needle biopsy by surgeon	23,189	8,191	1.02	0.99 to 1.04	.070	1.03	1.01 to 1.05	.005
No needle biopsy	27,811	19,332	2.11	2.09 to 2.14	< .001	2.10	2.07 to 2.12	< .001
Age, years
< 70	17,755	8,532	1.25	1.22 to 1.28	< .001	1.26	1.24 to 1.29	< .001
70-79	51,279	23,666	1.21	1.19 to 1.23	< .001	1.23	1.21 to 1.25	< .001
≥ 80	19,238	7,557	1			1
Race/ethnicity
White	81,471	36,556	1
Black	4,798	2,307	1.03	0.99 to 1.07	.172			NS
Other	2,003	892	1.01	0.96 to 1.07	.619			NS
Medicaid coverage
Yes	5,363	2,379	1			1
No	82,909	37,376	1.07	1.03 to 1.11	< .001	1.11	1.08 to 1.15	< .001
Year of BCS
2003	19,630	7,917	1			1
2004	18,502	8,514	1.16	1.13 to 1.18	< .001	1.20	1.17 to 1.23	< .001
2005	17,879	8,543	1.21	1.19 to 1.24	< .001	1.29	1.26 to 1.32	< .001
2006	17,959	8,516	1.20	1.17 to 1.23	< .001	1.32	1.29 to 1.34	< .001
2007	14,302	6,265	1.11	1.08 to 1.14	< .001	1.25	1.22 to 1.28	< .001
Charlson comorbidity score
≤ 1	80,563	36,364	1.08	1.03 to 1.13	.004	1.10	1.04 to 1.15	< .001
2	5,340	2,370	1.04	0.98 to 1.10	.183	1.05	0.99 to 1.11	.106
≥ 3	2,369	1,021	1			1
Receipt of chemotherapy
Yes	12,137	5,949	1.10	1.07 to 1.12	< .001	1.08	1.05 to 1.10	< .001
No	76,135	33,806	1			1
County of residence
Rural	17,397	8,144	1		< .001	1
Urban	70,875	31,611	0.95	0.93 to 0.98		1.04	1.01 to 1.06	.004
Treating surgeon factors
Board certified
Yes	84,321	37,684	1
No	3,951	2,071	1.21	1.15 to 1.27	< .001			NS
US trained
Yes	78,462	34,917	1
No	9,810	4,838	1.16	1.11 to 1.20	< .001			NS
Degree
MD	86,316	38,749	1
DO	1,956	1,006	1.15	1.06 to 1.23	< .001			NS
Decade of medical school graduation
Before 1980	37,056	16,626	1.15	0.97 to 1.34	.107			NS
1980-1989	33,685	15,373	1.16	0.97 to 1.34	.100			NS
1990-1999	17,239	7,637	1.11	0.93 to 1.30	.243			NS
2000+	292	119	1
Case volume, No. of patients†
≤ 5	17,469	9,191	1.33	1.28 to 1.37	< .001	1.08	1.04 to 1.13	< .001
> 5 and ≤ 10	16,578	7,856	1.17	1.12 to 1.22	< .001	1.04	0.99 to 1.08	.063
> 10 and ≤ 20	20,914	9,449	1.12	1.07 to 1.17	< .001	1.03	0.99 to 1.07	.170
> 20	33,311	13,259	1			1
Surgical specialty‡
Surgical oncologist	14,356	5,489	1			1
General surgeon	72,990	33,578	1.32	1.27 to 1.38	< .001	1.14	1.09 to 1.19	< .001
Other	926	688	1.98	1.92 to 2.02	< .001	1.80	1.72 to 1.87	< .001
Sex
Male	66,227	30,587	1.14	1.10 to 1.18	< .001			NS
Female	22,045	9,168	1

Abbreviations: BCS, breast-conserving surgery; DO, doctor of osteopathy; MD, doctor of medicine; NS, not significant; RR, relative risk.

^*Statistical significance evaluated at the P = .05 level in both unadjusted and adjusted analyses. A RR > 1 indicates that the factor was associated with a higher likelihood of multiple (two or more) breast cancer surgeries on separate dates before starting adjuvant radiation therapy. In this model, 392 surgeons and 1,440 patients were excluded as a result of unknown covariates. The ratio of the generalized χ² statistic to its df was 0.91.

^†Surgeon case volume was defined for each surgeon as the number of patients in the cohort who underwent BCS performed by that surgeon during the time interval of the study.

^‡Medical specialty was determined using the coded provider specialty in both the Carrier file and the American Medical Association Physician Masterfile, with either source considered sufficient to document a specialty of surgical oncology.

DISCUSSION

In this nationally comprehensive study of women undergoing BCS and radiation for incident breast cancer, 68.4% were seen by a surgeon before diagnostic biopsy. Compared with patients with surgeon consultation after biopsy, patients with surgeon consultation before biopsy were more likely to be in their 80s, black, ill, and poor; reside in rural areas; live more than 8.1 miles from a radiologic facility performing needle biopsy; and not undergo mammography before surgeon consultation. In total, only 53.7% of patients with prebiopsy surgeon consultation underwent needle biopsy. The use of needle biopsy was heavily influenced by the patient's surgeon, with patients substantially less likely to undergo needle biopsy if their surgeon was not board certified, not trained in the United States, not a surgical oncologist, not recently trained, or cared for a lower number of patients in the cohort. The importance of the surgeon relative to other factors such as proximity to radiologic facilities is illustrated by the observation that even among patients who lived within just 0.7 miles of a radiologic facility and were seen by a surgeon before biopsy, only 55.2% underwent needle biopsy. Unfortunately, for patients who did not undergo needle biopsy, only 30.4% were able to complete all of their breast surgery in a single procedure, indicating that the majority of such patients underwent one open surgical procedure to diagnose their breast cancer followed by at least one, if not more, additional open surgical procedures to treat it.

Our results are consistent with a recent IOM report entitled, “Variation in Health Care Spending: Target Decision Making, Not Geography,” which concluded that significant geographic variation exists in spending and utilization of general medical care and that such variation is largely unexplained. However, variation remains as units of analysis become progressively smaller, such that even within practices and across individual physicians meaningful variation exists.⁶ The findings of our study support and strengthen the argument of this IOM report, principally that the proportion of variance in needle biopsy use explained by the treating physician, 28.8%, is substantially greater than the proportion of variance explained by geographic region, 9.1%. Accordingly, our findings support the conclusion of this IOM report, which states that interventions to improve quality and value of care must be focused primarily on influencing provider behavior rather than modifying reimbursement globally in certain high-cost or low-value geographic regions.

Our results also identify multiple avenues whereby use of needle biopsy may be increased. Primarily, patients and PCPs should be empowered to partner directly with diagnostic radiologists to evaluate abnormal mammograms or clinical breast findings and to proceed to needle biopsy where appropriate. In addition, efforts to increase access to skilled radiologic facilities may also be beneficial, because patients residing more than 8.1 miles from a radiologic facility were more likely to undergo surgeon consultation before biopsy and less likely to undergo needle biopsy.

Our findings also indicate that surgeons often perform needle biopsy as an alternative to radiologic needle biopsy. Although encouraging surgeons to perform diagnostic biopsy may further increase use of needle biopsy to diagnose breast cancer, particularly in rural areas with suboptimal access to radiologists, prior literature indicates that needle biopsy performed by inexperienced physicians is frequently misleading and even harmful, with up to 25% of cancers missed by untrained physicians,²⁸ thus underscoring the need for adequate access and appropriate referral to skilled radiologic facilities.

Needle biopsy has been accepted as a diagnostic strategy to evaluate abnormal breast findings since the 1990s.¹¹ Consensus statements published in 2001,¹⁰ 2005,¹⁵ and 2009⁹ concluded that needle biopsy should be preferred over excisional biopsy and suggested that physician benchmarks for compliance with needle biopsy should exceed 90%. Needle biopsy achieves diagnostic accuracy similar to open surgical biopsy, but with a lower risk of postprocedure complications.²⁹ Furthermore, charges for needle biopsy are approximately 50% less than charges for open surgical biopsy; thus, increasing use of needle biopsy could reap substantial cost savings.³⁰ Needle biopsy also enhances accuracy of lymph node staging, with a false-negative sentinel lymph node rate of 15.3% in women diagnosed with excisional biopsy compared with 8.1% in women diagnosed with needle biopsy.³¹ Needle biopsy also serves to establish extent of disease preoperatively and can thus guide surgical decision making regarding whether to perform BCS or mastectomy. A final advantage of needle biopsy is that it provides a woman with information about her diagnosis before initiating surgical therapy and thus empowers her to be more involved in important decisions regarding neoadjuvant therapy, breast conservation versus mastectomy, radiation therapy, and breast reconstruction.

A growing literature has evaluated compliance with needle biopsy in the initial diagnosis of breast cancer. In the 1990s, only 24% of patients with breast cancer underwent needle biopsy.³² In the 2000s, a population-based study from Florida concluded that 70% of breast biopsies were needle biopsies.³⁰ A study from the National Cancer Database concluded that use of needle biopsy increased to 87% by 2008.³³ This study was limited, however, in that the definition of needle biopsy included open incisional biopsy, which could have resulted in inaccurately high needle biopsy rates. Furthermore, the National Cancer Database likely overestimates compliance with needle biopsy in the general population, because it includes only institutions participating in the American College of Surgeons Commission on Cancer program.

Results of this study are most applicable to older women with fee-for-service Medicare and may not readily extend to other patients and settings. To select a defined population with incident breast cancer for whom needle biopsy is indicated, we limited our cohort to women treated with BCS and radiation. It is possible that needle biopsy rates may vary in other patient groups with and without breast cancer. Furthermore, some patients in this cohort likely did not undergo needle biopsy as a result of legitimate reasons such as anxiety, transportation difficulties, lack of an imaging correlate for a palpable breast abnormality, or technical contraindications to biopsy, although all of these issues are relatively rare. Finally, although it is likely that needle biopsy rates have continued to increase since 2007, needle biopsy was considered the preferred standard of care during the entirety of our study interval, and thus our findings regarding the importance of timing of surgeon consultation relative to biopsy and the relevance of surgeon factors to use of needle biopsy are expected to remain relevant to contemporary practice. It should also be recognized that, to be consistent with the National Quality Forum–endorsed quality measure and prior literature, this study included both fine-needle aspiration and core biopsy in the definition of needle biopsy.^12,13 However, large-bore core needle biopsy is generally the preferred biopsy technique and carries the lowest false-negative rate.^34,35

Needle biopsy continues to be omitted in many patients for whom it is the preferred diagnostic approach, particularly those already known to be vulnerable to disparities in breast cancer care. This gap in quality doubles the likelihood of multiple breast cancer surgeries and has other harmful effects such as increased cost and decreased accuracy of sentinel lymph node biopsy. Surgeon-level interventions are needed to improve use of needle biopsy and, accordingly, quality of care.

Appendix

Table A1.

Inclusion and Exclusion Criteria for Cohort Selection

Step	Criteria	No. of Observations	No. of Patients Excluded
	All female Medicare beneficiaries (excluding entitlement solely as a result of ESRD or disability) age 67 years and older with BCS between 2003 and 2007 and radiotherapy to the breast within 12 months of BCS*	120,700
1	Exclude if the diagnosis code for breast cancer appeared only once or if there were two or more codes for distant metastasis from 12 months before to 12 months after the initial BCS	115,406	5,294
2	Exclude prevalent cases (defined as claim indicating history of breast cancer within 2 years prior to initial BCS, or with breast cancer diagnosis and breast cancer lymph node–directed surgery incurred between 12 and 24 months before initial BCS)	101,807	13,599
3	Exclude if patient was not continuously enrolled with Part A and B from 12 months before to 12 months after the initial BCS	99,193	2,614
4	Exclude if patient had any HMO coverage from 12 months before to 12 months after the initial BCS	94,903	4,290
5	Exclude if patient received chemotherapy or radiotherapy within 12 months before the initial BCS	91,184	3,719
6	Exclude if patient had a surgeon who performed the BCS, but his or her UPIN is unknown or unlinkable	90,153	1,031
7	Exclude if the patient did not live within one of the 50 US states	89,712	441

Abbreviations: BCS, breast-conserving surgery; ESRD, end-stage renal disease; HMO, health maintenance organization; UPIN, unique physician identification number.

^*To ensure that patients received radiation therapy to the breast, we required that radiation be administered for a coded diagnosis of breast cancer and that patients did not undergo mastectomy at any time between their initial BCS and the initiation of radiation therapy.

Table A2.

Claims Codes Used to Identify Breast Cancer Cases and Their Treatments

Variable	Codes
Any needle biopsy	ICD-9 procedure codes: 85.11 CPT codes: 10021, 10022, 19000, 19001, 19100, 19102, 19103
Core needle biopsy	CPT codes: 19100, 19102, 19103
Breast cancer surgery*	ICD-9 procedure codes: 85.2, 85.20-85.23, 85.25, 40.3, 40.23, 40.51 CPT codes: 19120, 19301, 19302, 19110, 19125, 19126, 19160, 19162, 38740, 38745, 38525, 19162
Diagnosis of breast cancer	ICD-9 diagnosis code: 174.X
History of breast cancer	ICD-9 diagnosis code: V10.3
Distant metastasis	ICD-9 diagnosis code: 196.XX-199.XX (excluding 196.0, 196.3, 198.2, 198.81)
BCS†	ICD-9 procedure codes: 85.2, 85.20-85.25 CPT codes: 19110, 19120, 19125, 19126, 19160, 19162, 19301, 19302
Mastectomy	ICD-9 procedure code: 85.4, 85.41-85.48 CPT codes: 19180, 19182, 19200, 19220, 19240, 19303, 19304, 19305, 19306, 19307
Breast cancer lymph node–directed surgery	ICD-9 procedure code: 40.3, 40.23, 40.51, 85.43, 85.47 CPT codes: 38740, 38745, 38525, 19162, 19200, 19220, 19240
Radiation therapy for a coded diagnosis of breast cancer	ICD-9 procedure codes: 92.2, 92.20-92.27, 92.29, 92.3, 92.30-92.39, 92.4, 92.41 ICD-9 diagnosis codes: V58.0, V66.1, V67.1 CPT codes: 77371-77373, 77401-77525, 77761-77799, G0174, G0251, G0339, G0340 Revenue center codes: 0330, 0333
Chemotherapy receipt	ICD-9 diagnosis codes: V58.1, V66.2, V67.2 ICD-9 procedure code: 99.25 CPT codes: Q0083-Q0085, J8520, J8521, J8530, J8540, J8560, J8597, J8610, J8999, J9000-J9999 (exclude J9003, J9165, J9175, J9202, J9209, J9212-J9226, J9240, J9295, J9381, J9395), 96400-96549 Revenue center codes: 0331, 0332, 0335
Mammography (inclusive or diagnostic and screening mammograms)	ICD-9 diagnosis codes: V76.11, V76.12 ICD-9 procedure code: 87.37 CPT codes: 76082, 76083, 76085, 76090, 76091, 76092, 77051, 77052, 77055, 77056, 77057, G0202, G0203, G0204, G0205, G0206, G0207, G0236

Abbreviations: BCS, breast-conserving surgery; CPT, Current Procedural Terminology; ICD-9, International Classification of Diseases, Ninth Revision.

^*Breast cancer surgery includes codes for both BCS and axillary surgery. To determine the number of breast cancer surgeries, we summed the number of unique appearances of any code for breast cancer surgery occurring between the date of initial BCS and the date of initiation of radiation therapy (inclusive of both dates).

^†BCS included claims for both formal segmental mastectomy and excisional biopsy.

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

Although all authors completed the disclosure declaration, the following author(s) and/or an author's immediate family member(s) indicated a financial or other interest that is relevant to the subject matter under consideration in this article. Certain relationships marked with a “U” are those for which no compensation was received; those relationships marked with a “C” were compensated. For a detailed description of the disclosure categories, or for more information about ASCO's conflict of interest policy, please refer to the Author Disclosure Declaration and the Disclosures of Potential Conflicts of Interest section in Information for Contributors.

Employment or Leadership Position: None Consultant or Advisory Role: Wei Yang, Toshiba (C), Seno Medical (C) Stock Ownership: None Honoraria: Wei Yang, Amirsys Research Funding: Wei Yang, VuComp; Benjamin D. Smith, Varian Medical Systems Expert Testimony: None Patents, Royalties, and Licenses: None Other Remuneration: None

AUTHOR CONTRIBUTIONS

Conception and design: Jan M. Eberth, Ying Xu, Grace L. Smith, Yu Shen, Thomas A. Buchholz, Kelly K. Hunt, Sharon H. Giordano, Wei Yang, Chan Shen, Linda Elting, Benjamin D. Smith

Collection and assembly of data: Ying Xu, Grace L. Smith, Thomas A. Buchholz, Gary J. Whitman, Wei Yang, Chan Shen, Linda Elting, Benjamin D. Smith

Data analysis and interpretation: Jan M. Eberth, Ying Xu, Yu Shen, Jing Jiang, Dalliah M. Black, Gary J. Whitman, Wei Yang, Benjamin D. Smith

Manuscript writing: All authors

Final approval of manuscript: All authors