Practice volume may affect the outcome of patients with breast carcinoma. Defining factors that influence the relation of volume and outcome for the diagnosis and treatment of breast carcinoma is important, because breast carcinoma is common, and care is decentralized.
Community-wide diagnosis and treatment of mammogram-detected breast carcinoma was examined using claims data from a single insurer representing 25% of the regional population. Among 1001 mammogram-directed breast biopsies, the rate of breast carcinoma diagnosed by stereotactic core needle biopsy (SCNB) or excisional biopsy with needle localization (EBNL) and the rate at which breast-conserving surgery (BCS) was used were analyzed. Outcome and practice volume were examined for surgeons, radiologists, and medical centers.
Two hundred twenty-four tumors were diagnosed by EBNL (604 diagnoses) and SCNB (397 tumors), for a 22.4% positive biopsy rate. The median number of procedures per physician was one. Positive biopsy rates for radiologists, surgeons, and medical centers did not correlate with practice volume. Positive biopsy rates for high-volume physician providers and medical centers ranged from 9% to 46%. The BCS rate was 45% and 64% for surgeons treating one or more than one claim, respectively. Tumor stage and surgeon case volume were the only independent predictors of BCS (P < 0.05).
The volume of patients treated by hospitals and individual physicians affects outcomes in many areas of medicine. A correlation between volume and outcome for treatment-related mortality is well established for cardiac surgery, cardiac angioplasty, and complex surgery for pancreatic, esophageal, liver, and pelvic malignancies.1–6 There is also evidence that case volume affects long-term survival after patients undergo surgery for breast carcinoma, with better survival for women who undergo such surgery at hospitals with a high case volume.7, 8
The reason for the correlation of volume and survival with breast surgery is unclear. Surgery for breast carcinoma does not carry a significant risk of perioperative mortality. The correlation of volume with outcome may be due to subtle differences in technique, variation in diagnostic and initial treatment decisions, or coordination and use of postoperative therapy. It is well documented that there is considerable variation in breast carcinoma treatment despite wide dissemination of screening and treatment guidelines.9–14 One solution to this problem is centralizing all care in high-volume centers. However, this is impractical and unproven for high-incidence conditions such as breast cancer and, itself, may be a barrier to patients accessing care. Another strategy is to evaluate the breast carcinoma care process to identify key steps in treatment where practice volume affects quality and to establish community-wide programmatic oversight and education to improve quality.
The outcome of care for patients with breast carcinoma is defined by decisions in the early phase of treatment, including the approach to diagnosis, surgical treatment, and application of adjuvant radiation and systemic therapy. One important aspect of care for patients with breast carcinoma for which provider case volume may affect outcome is the diagnosis and initial treatment of mammogram-detected carcinomas. These procedures are performed in virtually all acute care hospitals and in outpatient centers with widely variable volumes. An increasing fraction of breast carcinomas are diagnosed as an abnormality on mammogram prior to the appearance of a palpable mass.15–18 However, only a minority of mammogram-detected lesions for which biopsy is recommended are diagnosed as malignancies.19–24 Ideally, a minimum number of benign biopsies should be performed, and no malignancies should be missed. In most U.S. case series, the rate of malignancies among mammogram-directed biopsies ranges from 25% to 30%.23, 25–30 Lower rates of positive biopsy suggest that unnecessary breast biopsies are being performed and may serve as an indicator of the need for more comprehensive breast carcinoma program development and oversight. We hypothesized that the rate of positive biopsies for mammogram-detected breast carcinoma is a treatment outcome that can be measured on a community-wide basis from insurance claims data, that these factors are optimized in higher volume versus lower volume practice settings (surgeons, radiologists, and health care facilities), and that volume is related to the overall use of breast-conserving surgery (BCS) in this population.
MATERIALS AND METHODS
Identification of Patients
Insurance claims were used from Independent Health Association, Inc. (IHA), one of the three large managed-care organizations in the Western New York (WNY) region. IHA covered approximately 400,000 of the 1.6 million people in the region during the years 1995–1996. Confidentiality was assured by removing patient and provider identifiers from claims data to blind investigators. All medical records used in this study were collected and kept by IHA, blinded, and reviewed by study data managers at IHA headquarters to maintain confidentiality. The Institutional Review Board at Roswell Park Cancer Institute approved the study.
IHA claims from 1995 to 1996 were searched for women with CPT codes 19000-19260 (breast procedures), 38745 (axillary lymph node dissection), 88170 (fine-needle aspiration), 76095 (stereotactic core needle biopsy [SCNB]), 76096 (needle localization of surgical biopsy), and 76098 (specimen mammography). This search identified records from 2796 patients who underwent 1 of these index procedures. All medical claims for these patients were collected for a period of 6 months after the index breast procedure. For each patient, the biopsy result was classified as malignant (carcinoma) or benign based on the ICD-9 and CPT codes for the index procedure and for all subsequent claims for procedures and physician encounters for the 6 months after the index procedure. For all results classified by claims review as malignant, the operative reports and pathology reports from each breast procedure (index procedure and any subsequent procedures) or the hospital cancer registry report (for one hospital with an approved cancer registry) were reviewed to confirm the diagnosis of breast carcinoma and to define the disease stage (according to the 1997 American Joint Committee on Cancer criteria).31 To establish the validity of the methods used to identify malignant disease, pathology reports were reviewed for one-third of the biopsies that were classified as benign. These biopsies were confirmed as benign in 99.8% of the reports reviewed.
From the cohort of 2796 patients who underwent an index breast procedure during 1995–1996, 955 patients underwent an initial biopsy that included a CPT code indicating mammographic guidance (19125, 12126, 19290, 76095, 76096, and 76098). These mammogram-directed biopsies were classified as excisional biopsy with needle localization (EBNL) or SCNB. It was assumed that breast lesions biopsied with EBNL or SCNB procedures were mammogram-detected lesions that were not evident on physical examination. Records for patients who underwent more than one of these procedures were reviewed to identify the index biopsy procedure for each unique lesion based on the sequence and dates of procedures. There were a total of 1001 breast lesions that initially were biopsied by EBNL or SCNB among the 955 patients.
Claims information was used to determine the biopsy type, definitive surgery, radiologist provider, surgeon provider, and medical center for each of the 1001 mammogram-identified lesions. Provider specialty was defined by the first three digits of the provider code. The provider could not be determined in 3% of SCNB procedures and in 2–4% of EBNL procedures for the radiologist, surgeon, or medical center.
The positive rates for EBNL and SCNB were determined for surgeons, radiologists, and medical centers. Descriptive statistics, such as means and percentages, are reported. The chi-square test for independence was used to assess differences in positive biopsy rates by provider volume. The mean and variance in positive biopsy rates were calculated among providers with ≥ 10 patients and were used to estimate the true distribution of the positive biopsy rate. The lower 10th and 20th percentiles of these distributions were determined to identify providers and medical centers with positive biopsy rates below these limits. Factors that affected the use of BCS were evaluated in a multivariate model, which included type of initial biopsy, patient age, TNM stage, hospital volume, surgeon volume, and radiologist volume as predictor variables.
Biopsy Type and Disease Stage
Insurance claims records from a single health maintenance organization that covered 25% of the population in the WNY region identified 955 women who underwent mammogram-directed biopsies of 1001 lesions (397 SCNB procedures and 604 EBNL procedures) during 1995–1996, of which 224 lesions revealed malignancies. The positive biopsy rate, defined as the proportion of mammogram-directed biopsies that proved malignant, was 22.4%. For 6 of 1001 biopsies, the biopsy diagnosed the local recurrence of a prior malignancy. TNM staging could be defined for 217 of 218 incident malignancies. The stage distribution of the lesions is shown in Table 1. Fifty-seven lesions (26%) were diagnosed as ductal carcinoma in situ (55 lesions) or lobular carcinoma in situ (2 lesions). Among the 160 invasive tumors, 135 tumors (84%) measured ≤ 2 cm in greatest dimension (T1). Lymph node metastases were identified in 36 of the invasive tumors (23%). There were no differences in the TNM stage of tumors biopsied by SCNB or EBNL as an initial diagnostic procedure (P = 0.60).
Table 1. Stage Distribution of Mammogram-Detected Incident Breast Carcinoma, 1995–1996a
Claims for biopsies were submitted by 70 different radiologists in the WNY region. Eighteen radiologists submitted claims for SCNB procedures on 389 mammogram-identified lesions (Table 2). The number of SCNB procedures performed by each radiologist ranged from 1 to 239 procedures. Two radiologists who each submitted claims for > 100 SCNB procedures collectively performed 88% of all procedures. No other radiologist submitted claims for more than eight SCNB procedures. The positive biopsy rates among radiologists with claims for 1–2 SCNB procedures, 3–9 SCNB procedures, and ≥ 20 SCNB procedures ranged from 21% to 22% (P = 0.98).
Table 2. Rate of Positive Mammogram-Directed Breast Biopsies by Number of Claims Submitted, 1995–1996
Sixty-three radiologists performed the needle localization for EBNL on 589 lesions (Table 2). The total number of EBNL procedures performed by each radiologist over 2 years ranged from 1 to 86 procedures. Fifty-four percent of the radiologists who performed EBNL procedures submitted claims for ≤ 1 procedure per year. Twelve high-volume providers (10 or more claims) performed 445 of 589 procedures (76%). The positive biopsy rates among radiologists with claims for 1–2 EBNL procedures, 3–9 EBNL procedures, 10–19 EBNL procedures, and ≥ 20 EBNL procedures ranged from 15% to 24% (P = 0.93).
One hundred twenty-five surgeons submitted claims for EBNL procedures on 552 mammogram-identified lesions. The total number of EBNL procedures performed by each surgeon over 2 years ranged from 1 to 47 procedures. Table 2 shows the number of procedures and the number of malignancies identified by surgeons with claims for 1–2 EBNL procedures, 3–9 EBNL procedures, 10–19 EBNL procedures, and ≥ 20 EBNL procedures. Seventy surgeons (56%) submitted claims for ≤ 1 procedure per year. Thirteen high-volume providers (≥ 10 total claims submitted) performed 265 of 552 procedures (48%). The positive biopsy rates among the four groups ranged from 21% to 31% (P = 0.90).
Medical Center Characteristics
EBNL procedures were performed at 29 different medical centers. The number of EBNL procedures performed at these centers ranged from 1 to 121 procedures. Table 2 shows the number of procedures and the number of malignancies identified at medical centers with claims for 1–2 EBNL procedures, 3–9 EBNL procedures, 10–19 EBNL procedures, and ≥ 20 EBNL procedures. Ten centers (34%) performed ≤ 1 procedure per year. Ten high-volume centers (10 or more claims) performed 519 of 584 procedures (89%). The positive biopsy rates among medical centers with claims for 1–2 EBNL procedures, 3–9 EBNL procedures, 10–19 EBNL procedures, and ≥ 20 EBNL procedures ranged from 19% to 35% (P = 0.55). In contrast to EBNL claims, SCNB claims rarely included medical centers' identifiers (30 of 397 procedures). Therefore, an analysis of SCNB procedures by medical center was not performed.
Variation in Positive Biopsy Rates among Individual High-Volume Physicians and Centers
There was considerable variation among individual physicians and medical centers. For individual physicians and centers submitting claims on only a few procedures, the positive biopsy rates ranged predictably from 0% to 100%. However, the positive biopsy rates for high-volume physicians and medical centers (10 or more claims) varied by greater than 5-fold. The positive rate for individual providers by practice volume is summarized in Figure 1.
The positive biopsy rates ranged from 10% to 46% (median, 18%) for radiologists submitting 10–19 total claims and from 11% to 36% (median, 23%) for radiologists submitting ≥ 20 total claims. The positive biopsy rates ranged from 9% to 38% (median, 30%) for surgeons submitting claims for 10–19 EBNL procedures and from 20% to 48% (median, 30%) for surgeons submitting claims for ≥ 20 EBNL procedures. Medical centers identified in 10–19 EBNL claims had positive biopsy rates of 9–31% (median, 24%), and centers identified in ≥ 20 EBNL claims had positive biopsy rates of 10–28% (median, 23%).
The variance among individuals in each group was measured by using the mean and standard deviation for high-volume providers (individuals submitting more than 10 claims). These values were used to estimate the true distribution of the positive biopsy rate. The 10th and 20th percentiles at the lower extreme of this distribution were calculated for radiologists, surgeons, and medical centers. The positive biopsy rates of 3 radiologists were below the 20th percentile for the group. Among 3 surgeons with positive biopsy rates below the 20th percentile, 2 surgeons were below the 10th percentile. Two medical centers had positive biopsy rates below the 10th percentile for the group.
Surgical Treatment of Breast Carcinoma
The treatment schema for all patients with breast carcinoma is shown in Figure 2. For women with invasive carcinoma, the overall BCS rate was 59% (58% for women who underwent an initial EBNL procedure and 61% for women who underwent an initial SCNB procedure).
Sixty-six surgeons submitted claims for the treatment of 167 women with invasive breast carcinoma. Nineteen of these surgeons submitted claims on 2 or more patients, and 47 of these surgeons submitted claims on 1 patient each. Patients who were treated by surgeons who did more breast carcinoma surgery were more likely to undergo BCS. BCS was performed in 77 of 120 patients (64%) by surgeons who submitted claims on 2 or more patients and in 21 of 47 patients (45%) by surgeons who submitted claims on a single patient. Factors that affected the use of BCS were investigated in a multivariate logistic model (Table 3). Factors included in the model were type of initial biopsy, tumor size, lymph node status, patient age, and breast biopsy volume for the hospital, radiologist, and surgeon. Provider and hospital volumes were grouped by the number of biopsy procedures (1, 2–10, and ≥ 11 procedures). Case volume and tumor size were the only factors that affected the use of BCS. Compared with women whose surgeons submitted 1 claim, women with surgeons who submitted 2–10 claims (odds ratio, 3.07; P = 0.025) and ≥ 11 claims (odds ratio, 2.95; P = 0.039) were more likely to undergo BCS.
Table 3. Multivariate Analysis of the Use of Breast-Conserving Surgery for Mammogram-Detected Invasive Breast Carcinoma, 1995–1996
Defining correlations between cancer care outcomes and the number of patients treated by individuals and institutions has important implications for health policy planning. Breast carcinoma is a leading cause of cancer death in American women, it is treated in a variety of practice settings, and variations in treatment can affect outcome. Complete information on short-term outcomes for patients who are diagnosed and treated for breast carcinoma by individuals has been difficult to compile, because care is provided by so many unaffiliated physicians and hospitals, and because there is no required recording or centralized reporting of patients or practice outcomes.
Insurance claims proved to be a valuable source of community-wide data on breast practice. Because breast diagnostic care is administered throughout the community, there are no other sources of this data. Cancer registries are of no value for diagnostic information, because they do not collect data on benign biopsies. Furthermore, approved cancer registries are maintained by only one-third of hospitals in the United States. State cancer registries, although they are more comprehensive, lack detailed treatment information.32 Insurance data may provide a more representative picture of community-wide care than published single-institution or multi-institution case series. Such series may be biased toward academic settings or large-volume practitioners who are willing and able to collect data, and they may specifically exclude patients who are treated by low-volume providers. Insurance data include all patients who are treated by providers participating with that payer. Claims data analysis supplemented by selected medical records may be one of the only sources on the treatment of patients with malignant disease that includes all providers in the community.
Two distinct correlations were seen between provider volume and breast diagnostic and cancer care. Case volumes for surgeons, radiologists, and medical centers were not correlated with positive biopsy rates for mammogram-detected breast lesions. However, there was enormous variation among individual physicians and medical centers with the highest number of cases. Conversely, another important outcome of treatment, the use of BCS for patients with invasive carcinoma, was related to practice volume. Surgeons with higher practice volume were more likely to use BCS, and surgeon case volume was an independent predictor of BCS in a multivariate model that included patient age, tumor size, lymph node status, and hospital and radiologist case volume.
Another important finding was the very low case volume of individual providers. The large majority of surgeons and radiologists performed very few procedures, whereas a small minority performed many procedures. Overall, 50% of patients with breast carcinoma were treated by physicians who performed a low volume of breast carcinoma-related procedures. This study does not define the true number of procedures performed by each provider, because those data were from only one payer. However, because the IHA covers about 25% of the regional population, a crude extrapolation would suggest that the mean number of malignancies treated per surgeon per year in WNY is 3–4 malignancies.
Variation among individual providers has not been described previously on a regional basis. Overall, the positive biopsy rate of 22.4% was within the range reported by most U.S. case series.20, 21, 33, 34 Although an ideal rate is unknown, individual large centers report consistent positive rates in the 1990s of 24–32%.23, 25–30 Although case volume was not correlated with positive biopsy rates, there was wide variation in positive rates among high-volume radiologists, surgeons, and hospitals. For example, the hospital with the 5th largest EBNL experience in the region diagnosed 3 malignancies among 31 biopsies. Three radiologists at that institution had positive biopsy rates of 14% (3 of 22 biopsies), 0% (0 of 6 biopsies), and 0% (0 of 3 biopsies).
The variation in biopsy positive rate raises quality concerns. Rates below 15–20% suggest over-reading of mammograms and unnecessary biopsies. Low positive biopsy rates have major repercussions for health care costs, resource utilization, and quality-of-life issues for patients with benign breast biopsies. The impact of surgery and emotional stress associated with the possibility of a diagnosis of breast carcinoma may undermine support for mammography screening.35 Assuming that 40% of breast carcinomas in the United States are diagnosed by mammography, positive biopsy rates of 15% versus 25% nationally would increase the annual number of benign breast biopsies by 187,000 (from 280,000 benign breast biopsies to 467,000 benign breast biopsies).10, 36–38 There are few efforts to address this quality issue. Although practice self-auditing is a requirement for mammography center certification under the Mammography Quality Standards Act, these data are not reported centrally, and it is not clear that these audits have contributed to overall improvement in biopsy use.39, 40
Compared with the rate of positive breast biopsies among surgeons, there was a significant correlation between practice volume and the use of BCS. Nontumor-related factors that have been correlated previously with a higher rate of BCS include lower patient age, higher patient income, private hospital funding, urban hospital location, geographic region, recent year of surgery, large hospital size, and lower surgeon age.41–46 Due to the blinding of the data, institutional and provider demographic factors could not be included in a multivariate model. However, surgeon case volume was a predictor of the use of BCS independent of radiologist or medical center case volume, patient age, tumor size, and lymph node status.
The limitations of this study and the generalizability of the methodology used include the degree to which the data from one payer reflect care in an entire community. This will vary between insurance markets; however, at the very least, the data from one payer are useful to evaluate the care for clients of that payer. In the WNY region, there are only three major payers that cover about 90% of the insured population. Virtually all providers participate with all payers. It is unlikely that providers tailor care for patients who are diagnosed with breast carcinoma by payer; therefore, the results from patients with one payer are likely to be representative of practice patterns for the entire community. This study did not include the uninsured and only included the minority of older patients who were covered by managed Medicare contracts, which was reflected by the younger average age of patients with breast carcinoma in this study (55.1 years) compared with the national average age (63.0 years).47 This limitation was offset by the inclusion of all providers and hospitals.
Another limitation of this methodology is that the outcome of the procedure (carcinoma vs. not carcinoma) was assigned to the physician (surgeon or radiologist) who performed the procedure. In the case of the radiologist, a different radiologist may have recommended the biopsy. The radiologist performing the biopsy may find it difficult to reverse a biopsy recommendation made by a colleague. However, it is reasonable to assign the outcome to the radiologist who performed the procedure for community-wide analysis based on the assumption that the data will not be used directly for punitive or restrictive purposes.
There is mounting evidence that long-term survival after treatment for malignant disease is compromised if care is administered in a low-volume setting.1, 7, 8, 48, 49 One solution is to mandate centralization of all care to high-volume settings. This has been recommended for high-risk, low-volume surgical procedures, such as pancreatectomy.31 However, centralization of care for patients with breast carcinoma poses significant political and logistic problems and, in itself, may pose barriers to high-quality care, for example, in rural areas with long distances to specialty centers. Furthermore, these data underscore that the correlation between practice volume and outcome is complex and that volume is not synonymous with quality.
Quality-improvement programs based on community-wide practice analysis that identify the causes of volume-related outcome problems may be a practical alternative to mandated centralized care. Interventions to address identified quality issues need to be proactive but do not necessarily require restricting hospitals or physicians from providing these services. Identification of an institution or provider as an outlier should prompt an in-depth practice review in collaboration with the affected institution. Providers and centers should use these data in collaboration with regional experts to determine the causes of low positive biopsy rates, to correct the problem, and to evaluate overall management and referral practices for patients with breast carcinoma in their community.
Administrative claims data can be used by any payer/provider collaborative group to assess breast care. Ongoing community-wide quality review, including practice audits with feedback to providers and the public, coupled with proactive quality interventions are required to affect meaningful improvements in patient outcome.