[Skip to Content]
Sign In
Individual Sign In
Create an Account
Institutional Sign In
OpenAthens Shibboleth
[Skip to Content Landing]
Figure 1.
Medicare Beneficiaries Who Received Inappropriate Imaging Tests by HRR Quartile
Medicare Beneficiaries Who Received Inappropriate Imaging Tests by HRR Quartile

A and B, In both bar graphs, quartile 1 had the lowest inappropriate imaging rate; quartile 4, the highest. In all panels, CT indicates computed tomography; HRR, hospital referral region; PET, positron emission tomography.

Figure 2.
Regional-Level Inappropriate Prostate Cancer Imaging by Regional-Level Inappropriate Breast Cancer Imaging
Regional-Level Inappropriate Prostate Cancer Imaging by Regional-Level Inappropriate Breast Cancer Imaging

Size of bubble reflects the inverse variance of inappropriate breast cancer imaging in each hospital referral region; all rates are age standardized.

Table 1.  
Characteristics of Medicare Beneficiaries With Incident, Low-Risk Prostate or Breast Cancer
Characteristics of Medicare Beneficiaries With Incident, Low-Risk Prostate or Breast Cancer
Table 2.  
Risk of Inappropriate Prostate Cancer Imaging by Regional and Demographic Characteristics Determined Using Logistic Regression
Risk of Inappropriate Prostate Cancer Imaging by Regional and Demographic Characteristics Determined Using Logistic Regression
Table 3.  
Risk of Inappropriate Breast Cancer Imaging by Regional and Demographic Characteristics Determined Using Logistic Regression
Risk of Inappropriate Breast Cancer Imaging by Regional and Demographic Characteristics Determined Using Logistic Regression
1.
Wennberg  JE, Fisher  ES, Skinner  JS.  Geography and the debate over Medicare reform.  Health Aff (Millwood). 2002;(Suppl Web Exclusives):W96-114.PubMedGoogle Scholar
2.
American Board of Internal Medicine Foundation.  The Choosing Wisely campaign: five things physicians and patients should question.http://www.choosingwisely.org/wp-content/uploads/2013/01/about_choosingwisely_fivethings.pdf. Accessed May 5, 2012.
3.
Cassel  CK, Guest  JA.  Choosing Wisely: helping physicians and patients make smart decisions about their care.  JAMA. 2012;307(17):1801-1802.PubMedGoogle ScholarCrossref
4.
Brody  H.  Medicine’s ethical responsibility for health care reform: the Top Five list.  N Engl J Med. 2010;362(4):283-285.PubMedGoogle ScholarCrossref
5.
Schnipper  LE, Smith  TJ, Raghavan  D,  et al.  American Society of Clinical Oncology identifies five key opportunities to improve care and reduce costs: the Top Five list for oncology.  J Clin Oncol. 2012;30(14):1715-1724.PubMedGoogle ScholarCrossref
6.
Rao  VM, Levin  DC.  The overuse of diagnostic imaging and the Choosing Wisely initiative.  Ann Intern Med. 2012;157(8):574-576.PubMedGoogle ScholarCrossref
7.
Mariotto  AB, Yabroff  KR, Shao  Y, Feuer  EJ, Brown  ML.  Projections of the cost of cancer care in the United States: 2010-2020.  J Natl Cancer Inst. 2011;103(2):117-128.PubMedGoogle ScholarCrossref
8.
Medicare Payment Advisory Commission.  Medicare payment policy: report to the Congress.http://www.medpac.gov/documents/reports/march-2012-report-to-the-congress-medicare-payment-policy.pdf?sfvrsn=0. Accessed May 10, 2012.
9.
Dinan  MA, Curtis  LH, Hammill  BG,  et al.  Changes in the use and costs of diagnostic imaging among Medicare beneficiaries with cancer, 1999-2006.  JAMA. 2010;303(16):1625-1631.PubMedGoogle ScholarCrossref
10.
Iglehart  JK.  The new era of medical imaging: progress and pitfalls.  N Engl J Med. 2006;354(26):2822-2828.PubMedGoogle ScholarCrossref
11.
Iglehart  JK.  Health insurers and medical-imaging policy: a work in progress.  N Engl J Med. 2009;360(10):1030-1037.PubMedGoogle ScholarCrossref
12.
Mitchell  JM.  Utilization trends for advanced imaging procedures: evidence from individuals with private insurance coverage in California.  Med Care. 2008;46(5):460-466.PubMedGoogle ScholarCrossref
13.
Hillman  BJ, Goldsmith  JC.  The uncritical use of high-tech medical imaging.  N Engl J Med. 2010;363(1):4-6.PubMedGoogle ScholarCrossref
14.
Hara  T, Kosaka  N, Kishi  H.  PET imaging of prostate cancer using carbon-11-choline.  J Nucl Med. 1998;39(6):990-995.PubMedGoogle Scholar
15.
Souchon  R, Rouvière  O, Gelet  A,  et al.  Visualisation of HIFU lesions using elastography of the human prostate in vivo: preliminary results.  Ultrasound Med Biol. 2003;29(7):1007-1015.PubMedGoogle ScholarCrossref
16.
Ito  H, Kamoi  K, Yokoyama  K, Yamada  K, Nishimura  T.  Visualization of prostate cancer using dynamic contrast-enhanced MRI: comparison with transrectal power Doppler ultrasound.  Br J Radiol. 2003;76(909):617-624.PubMedGoogle ScholarCrossref
17.
Halpern  EJ, Rosenberg  M, Gomella  LG.  Prostate cancer: contrast-enhanced us for detection.  Radiology. 2001;219(1):219-225.PubMedGoogle ScholarCrossref
18.
Warner  E, Messersmith  H, Causer  P, Eisen  A, Shumak  R, Plewes  D.  Systematic review: using magnetic resonance imaging to screen women at high risk for breast cancer.  Ann Intern Med. 2008;148(9):671-679.PubMedGoogle ScholarCrossref
19.
 Americans speak on health reform: report on health care community discussions.http://web.archive.org/web/20121016185712/http://www.healthreform.gov/reports/hccd/concernsd.html. Accessed January 28, 2015.
20.
Keyhani  S, Falk  R, Bishop  T, Howell  E, Korenstein  D.  The relationship between geographic variations and overuse of healthcare services: a systematic review.  Med Care. 2012;50(3):257-261.PubMedGoogle ScholarCrossref
21.
Committee on Geographic Variation in Health Care, Spending and Promotion of High-Value Care Board on Health Care Services, Institute of Medicine of the National Academies.  Variation in health care spending: target decision making, not geography.http://books.nap.edu/openbook.php?record_id=18393. Accessed January 28, 2015.
22.
Abraham  N, Wan  F, Montagnet  C, Wong  YN, Armstrong  K.  Decrease in racial disparities in the staging evaluation for prostate cancer after publication of staging guidelines.  J Urol. 2007;178(1):82-87.PubMedGoogle ScholarCrossref
23.
Cooperberg  MR, Lubeck  DP, Grossfeld  GD, Mehta  SS, Carroll  PR.  Contemporary trends in imaging test utilization for prostate cancer staging: data from the cancer of the prostate strategic urologic research endeavor.  J Urol. 2002;168(2):491-495.PubMedGoogle ScholarCrossref
24.
Kindrick  AV, Grossfeld  GD, Stier  DM, Flanders  SC, Henning  JM, Carroll  PR.  Use of imaging tests for staging newly diagnosed prostate cancer: trends from the CaPSURE database.  J Urol. 1998;160(6, pt 1):2102-2106.PubMedGoogle ScholarCrossref
25.
Saigal  CS, Pashos  CL, Henning  JM, Litwin  MS.  Variations in use of imaging in a national sample of men with early-stage prostate cancer.  Urology. 2002;59(3):400-404.PubMedGoogle ScholarCrossref
26.
Makarov  DV, Desai  RA, Yu  JB,  et al.  The population level prevalence and correlates of appropriate and inappropriate imaging to stage incident prostate cancer in the Medicare population.  J Urol. 2012;187(1):97-102.PubMedGoogle ScholarCrossref
27.
Makarov  DV, Desai  R, Yu  JB,  et al.  Appropriate and inappropriate imaging rates for prostate cancer go hand in hand by region, as if set by thermostat.  Health Aff (Millwood). 2012;31(4):730-740.PubMedGoogle ScholarCrossref
28.
National Cancer Institute.  SEER-Medicare linked database.http://appliedresearch.cancer.gov/seermedicare/. Accessed September 15, 2009, 2009.
29.
Warren  JL, Klabunde  CN, Schrag  D, Bach  PB, Riley  GF.  Overview of the SEER-Medicare data: content, research applications, and generalizability to the United States elderly population.  Med Care. 2002;40(8)(suppl):IV-3-IV-18.PubMedGoogle Scholar
30.
Carlson  RW, Allred  DC, Anderson  BO,  et al; National Comprehensive Cancer Network.  Invasive breast cancer.  J Natl Compr Canc Netw. 2011;9(2):136-222.PubMedGoogle Scholar
31.
Wennberg  JE, Cooper  MM.  Appendix on the geography of health care in the United States, 1999.http://www.dartmouthatlas.org/downloads/methods/geogappdx.pdf. Accessed January 28, 2015.
32.
Wennberg  JE, Cooper  MM.  The Dartmouth Atlas of Health Care in the United States. Chicago, IL: American Hospital Publishing; 1996.
33.
Elixhauser  A, Steiner  C, Harris  DR, Coffey  RM.  Comorbidity measures for use with administrative data.  Med Care. 1998;36(1):8-27.PubMedGoogle ScholarCrossref
34.
Bynum  J, Song  Y, Fisher  E.  Variation in prostate-specific antigen screening in men aged 80 and older in fee-for-service Medicare.  J Am Geriatr Soc. 2010;58(4):674-680.PubMedGoogle ScholarCrossref
35.
Ma  X, Wang  R, Long  JB,  et al.  The cost implications of prostate cancer screening in the Medicare population.  Cancer. 2014;120(1):96-102.PubMedGoogle ScholarCrossref
36.
Ko  DT, Wang  Y, Alter  DA,  et al.  Regional variation in cardiac catheterization appropriateness and baseline risk after acute myocardial infarction.  J Am Coll Cardiol. 2008;51(7):716-723.PubMedGoogle ScholarCrossref
37.
Makarov  DV, Loeb  S, Ulmert  D, Drevin  L, Lambe  M, Stattin  P.  Prostate cancer imaging trends after a nationwide effort to discourage inappropriate prostate cancer imaging.  J Natl Cancer Inst. 2013;105(17):1306-1313.PubMedGoogle ScholarCrossref
Original Investigation
May 2015

Regional-Level Correlations in Inappropriate Imaging Rates for Prostate and Breast CancersPotential Implications for the Choosing Wisely Campaign

Author Affiliations
  • 1US Department of Veterans Affairs, Washington, DC
  • 2Department of Urology, New York University School of Medicine, New York
  • 3Department of Population Health, New York University School of Medicine, New York
  • 4New York University Cancer Institute, New York
  • 5Cancer Outcomes, Public Policy and Effectiveness Research (COPPER) Center, Yale University, New Haven, Connecticut
  • 6Department of Internal Medicine, Yale University School of Medicine, New Haven, Connecticut
  • 7Department of Medicine, New York University School of Medicine, New York
  • 8Department of Therapeutic Radiology, Yale University School of Medicine, New Haven, Connecticut
  • 9Center for Outcomes Research and Evaluation, Yale–New Haven Hospital, New Haven, Connecticut
  • 10Section of General Internal Medicine, Department of Medicine, Yale University School of Medicine, New Haven, Connecticut
  • 11Department of Health Policy and Management, Yale University School of Public Health, New Haven, Connecticut
  • 12Department of Epidemiology and Public Health, Yale University School of Medicine, New Haven, Connecticut
JAMA Oncol. 2015;1(2):185-194. doi:10.1001/jamaoncol.2015.37
Abstract

Importance  The association between regional norms of clinical practice and appropriateness of care is incompletely understood. Understanding regional patterns of care across diseases might optimize implementation of programs like Choosing Wisely, an ongoing campaign to decrease wasteful medical expenditures.

Objective  To determine whether regional rates of inappropriate prostate and breast cancer imaging were associated.

Design, Setting, and Participants  Retrospective cohort study using the the Surveillance, Epidemiology, and End Results–Medicare linked database. We identified patients diagnosed from 2004 to 2007 with low-risk prostate (clinical stage T1c/T2a; Gleason score, ≤6; and prostate-specific antigen level, <10 ng/mL) or breast cancer (in situ, stage I, or stage II disease), based on Choosing Wisely definitions.

Main Outcomes and Measures  In a hospital referral region (HRR)-level analysis, our dependent variable was HRR-level imaging rate among patients with low-risk prostate cancer. Our independent variable was HRR-level imaging rate among patients with low-risk breast cancer. In a subsequent patient-level analysis we used multivariable logistic regression to model prostate cancer imaging as a function of regional breast cancer imaging and vice versa.

Results  We identified 9219 men with prostate cancer and 30 398 women with breast cancer residing in 84 HRRs. We found high rates of inappropriate imaging for both prostate cancer (44.4%) and breast cancer (41.8%). In the first, second, third, and fourth quartiles of breast cancer imaging, inappropriate prostate cancer imaging was 34.2%, 44.6%, 41.1%, and 56.4%, respectively. In the first, second, third, and fourth quartiles of prostate cancer imaging, inappropriate breast cancer imaging was 38.1%, 38.4%, 43.8%, and 45.7%, respectively. At the HRR level, inappropriate prostate cancer imaging rates were associated with inappropriate breast cancer imaging rates (ρ = 0.35; P < .01). At the patient level, a man with low-risk prostate cancer had odds ratios (95% CIs) of 1.72 (1.12-2.65), 1.19 (0.78-1.81), or 1.76 (1.15-2.70) for undergoing inappropriate prostate imaging if he lived in an HRR in the fourth, third, or second quartiles, respectively, of inappropriate breast cancer imaging, compared with the lowest quartile.

Conclusions and Relevance  At a regional level, there is an association between inappropriate prostate and breast cancer imaging rates. This finding suggests the existence of a regional-level propensity for inappropriate imaging utilization, which may be considered by policymakers seeking to improve quality of care and reduce health care spending in high-utilization areas.

Introduction

Researchers have estimated 30% of resources spent on health care in the United States do not improve the health of patients.1 In 2012, the American Board of Internal Medicine Foundation and Consumer Reports launched Choosing Wisely, a national effort to encourage appropriate use of health care resources by inspiring discussion among patients, physicians, and other stakeholders.2,3 As part of Choosing Wisely,3,4 the American Society of Clinical Oncology (ASCO) released a Top Five list of tests and procedures that could be used less often without compromising patient care.5 This list identified several opportunities for promoting high-value care, including reducing unnecessary diagnostic imaging.6 Specifically, 2 items on ASCO’s list recommend decreasing imaging to stage patients with low-risk prostate and breast cancers,5 clinical practices estimated to be inefficient yet stubborn to eradicate.6

Despite recommendations against their routine use, imaging modalities unlikely to inform medical decision making713 are frequently offered to patients with incident prostate1417 and breast18 cancers. Factors such as fear of malpractice (defensive medicine), physician and patient preferences, duplication of care secondary to fragmentation, and poor record keeping may facilitate inappropriate imaging and feature prominently in cancer care delivery.19 While extensive literature documents regional variation in the cost and utilization of health care, little is known about regional-level drivers of utilization.20 In spite of this knowledge gap, a report from the Institute of Medicine (IOM) suggested that regional factors were not as important as individual-level decision making in driving inappropriate utilization.21 Studying prostate and breast cancer imaging patterns in the pre–Choosing Wisely era may help determine the potential existence of a yet undiscovered regional-level driver of health care resource utilization and the extent to which it might affect cancer care.

Previous studies have documented wide regional imaging variation and its correlates within prostate cancer. Such variation is not entirely explained by regional variations in patient-level characteristics.2226 The variation appears to be driven by regional tendencies toward imaging utilization: regions with lower inappropriate imaging also have lower appropriate imaging, while regions with higher inappropriate imaging similarly have higher appropriate imaging.27 Regional imaging patterns for other cancers, such as breast, have not been studied; neither have regional-level imaging associations across cancer types. Because prostate and breast cancers affect different patient populations and are often treated by different specialists, there should not be an association between their imaging patterns. A correlation between regional rates of prostate and breast cancer imaging would suggest that regional imaging behaviors share common determinants. Such a finding might contradict current IOM recommendations against regionally targeted health care utilization policies and suggest that changing regional culture, or other environmental-level factors, in high-utilization regions might improve care across diseases.

We set out to determine the regional rates of ASCO-defined inappropriate imaging among Medicare beneficiaries with low-risk prostate or breast cancers. We hypothesized that there might be a positive regional-level association between inappropriate prostate and breast cancer imaging.27 To test this hypothesis, we identified patients with incident prostate and breast cancers from the Surveillance, Epidemiology, and End Results (SEER) Medicare-linked database, determined whether they underwent inappropriate imaging, and then used these results for regional characterization. Understanding whether prostate and breast cancer imaging are associated at a regional level might promote more nuanced, regionally tailored interventions, such as legislation, payment reform, and educational initiatives, to improve population health and provide higher-value care.

Box Section Ref ID

At a Glance

  • There are high rates of inappropriate imaging among patients with low-risk prostate cancer (44.4%) and low-risk breast cancer (41.8%).

  • At the level of the hospital referral region (HRR), inappropriate prostate cancer imaging rates were associated with inappropriate breast cancer imaging rates (ρ = 0.35; P < .01).

  • At the patient level, a patient with prostate cancer had higher odds of undergoing inappropriate imaging if he lived in a region with higher inappropriate breast cancer imaging.

  • Utilization of health care resources across diseases may be determined by regional-level factors, a novel finding.

Methods
Study Design and Data Source

We conducted a retrospective cohort study using the SEER-Medicare database. SEER is a program of the National Cancer Institute that distributes data from high-quality tumor registries across the US accounting for >25% of the population.28 Information on cancer diagnosis, location, stage, histology, and patient demographics is available.29 SEER-Medicare links tumor registry data with Medicare hospital, physician, and outpatient claims for 93% of patients ≥65 years old.28,29 The Yale Human Investigations Committee determined that this study did not constitute “human subjects research” and thus did not require institutional review board approval or participant written informed consent.

Study Sample

We constructed cohorts of patients with low-risk prostate and breast cancer based on clinical criteria specified in the ASCO Choosing Wisely guidelines.5 For prostate cancer, this included men diagnosed with clinical stage T1c/T2a prostate cancer with a Gleason score of 6 or lower and a prostate-specific antigen (PSA) level below 10 ng/mL; for breast cancer, this included women diagnosed with in situ, stage I, or stage II disease. For both samples, we applied the following inclusion criteria: year of diagnosis from 2004 to 2007; first or only cancer diagnosis; known month of diagnosis (to determine a timeframe from which comorbidity-defining claims could be examined); diagnosis not found on autopsy or death certificate; and ages 67 to 94 years. Additionally, all patients had to be continuously enrolled in Parts A and B fee-for-service Medicare and could not have been diagnosed with another primary cancer from 2 years prior to diagnosis through 1 year following diagnosis.

Construction of Variables

We again used the ASCO Choosing Wisely guidelines to define inappropriate imaging in our prostate and breast cancer cohorts.5 Our primary outcome (a dichotomous variable) was receipt of positron-emission tomography (PET), computed tomography (CT), or bone scan , determined by using Healthcare Common Procedure Coding System (Centers for Medicare & Medicaid Services) and International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) procedure codes (eTable 1 in the Supplement). For patients with prostate cancer, the timeframe was from 2 months before diagnosis until either 6 months after diagnosis or until initiation of treatment (prostatectomy, radiation therapy, orchiectomy, or androgen deprivation therapy), whichever came first. For patients with breast cancer, the timeframe included 2 months before through 6 months after diagnosis. We used different timeframes because of differences in the way each cancer is treated. The ASCO guidelines refer specifically to diagnostic imaging conducted for disease staging. For prostate cancer, once treatment is initiated, imaging may be performed in response to treatment complications or to monitor disease progression.26 For breast cancer, patients are often treated with surgery and adjuvant radiation necessitating postoperative imaging to measure progress or plan radiation therapy.30 Therefore, we did not count imaging performed on or after the date of treatment as inappropriate.

We performed the analysis at the HRR level.31 An HRR represents a predefined regional market for tertiary health care at a major regional referral hospital. We characterized imaging patterns across HRRs, which have been used previously to define health care markets.31 Patient zip code was used to assign each patient to an HRR.32

Covariates included patient-level factors such as age, race, income, marital status, comorbidity, and socioeconomic status, based on median household income at the census tract or zip code level. Comorbidity was assessed using a list of conditions suggested by Elixhauser.33 We searched inpatient, outpatient, and physician claims billed between 24 and 3 months prior to diagnosis for ICD-9-CM codes appearing on at least 1 inpatient claim or 2 or more outpatient or physician claims billed at least 30 days apart. A comorbidity score was equal to each patient’s number of conditions.

We also included several HRR-level variables. To address possible confounding from regional-level infrastructure, we included the HRR-level number of hospital beds and the number of radiologists, both per 100 000 HRR residents from the Dartmouth Atlas of Healthcare website.31 To address the possible effect of market-level competition, we calculated an HRR-level Herfindahl index (HI) based on hospital use of prostate or breast cancer surgery. HI is calculated as follows:Image description not available.where si is the market share of prostate or breast surgery for hospital i in the HRR, and N is the number of hospitals. An HI = 1 represents a market dominated by a single hospital monopoly, while an HI = 1/N represents perfect competition.

Statistical Analysis

For prostate cancer, HRR-level imaging rates were calculated for HRRs with 10 or more patients with prostate cancer. Because the age distribution within HRRs varied, the imaging rates were age standardized using direct age adjustment. This same method was used to calculate breast imaging rates. At the patient level, we used χ2 and rank-sum analyses to assess bivariate associations between covariates and receipt of imaging.

We performed analyses at the HRR level and at the patient level. At the HRR-level, we assessed correlation between HRR-level inappropriate prostate and breast imaging by calculating the Pearson coefficient between the square root–transformed rates of imaging, weighted by the inverse variance of the HRR level, square root–transformed inappropriate breast cancer imaging rate. We used this transformation so that the imaging rates would follow a normal distribution, and we used the inverse variance of the square root–transformed breast cancer imaging rate estimates to account for precision in our estimates, given the variation in populations among the HRRs.

We then performed 2 patient-level analyses. First, we modeled the likelihood that an individual with prostate cancer would undergo inappropriate imaging as a function of the rate of HRR-level inappropriate breast cancer imaging rate in the HRR in which he was treated. We ranked each HRR by its rate of inappropriate breast cancer imaging, from highest to lowest, classifying each into quartiles. Next we determined using logistic regression the bivariate association between a patient with prostate cancer undergoing inappropriate imaging and the covariates described herein. Finally we tested the adjusted association between the likelihood of a patient undergoing inappropriate patient-level prostate cancer imaging and the HRR-level inappropriate breast cancer imaging quartile using hierarchical generalized linear regression with a logit link, accounting for clustering by HRR. We excluded covariates if their P values were greater than .05, leaving a parsimonious model including only the independent variable of interest (inappropriate breast imaging quartile) and statistically significant covariates. We then performed a second patient-level analysis using parallel methodology to regress patient-level likelihood of inappropriate breast cancer imaging on HRR-level inappropriate prostate cancer imaging quartile.

All analyses were performed using SAS software, version 9.2 (SAS Institute Inc), and SEER*Stat, version 7.0.9 (National Cancer Institute).

Results

We identified 9219 male Medicare beneficiaries with incident, low-risk prostate cancer, and 30 398 female beneficiaries with incident, low-risk breast cancer (Table 1). Men aged 80 to 84 years were most frequently imaged inappropriately for prostate cancer (49%), while women aged 67 to 69 years were most frequently imaged inappropriately for breast cancer (44%). In both groups, those with the most comorbidities were imaged most frequently (50% for both). For men with low-risk prostate cancer, frequency of inappropriate imaging was not associated with race, year of diagnosis, median household income, or marital status. However, rates of inappropriate imaging for women with low-risk breast cancer were associated with race (black women imaged most, 44%), year of diagnosis (imaging performed more frequently over time; 43% in 2007), median income (most frequent imaging for median income $50 000-$63 000, 43%), and marital status (married women imaged most, 42%). Patients with prostate cancer underwent more inappropriate imaging (44.4%) than those with breast cancer (41.8%), but those with breast cancer underwent PET scans more often (7.0% vs 0.3%) (Figure 1). The weighted Pearson coefficient for HRR-level age-standardized inappropriate prostate and breast cancer imaging rates was 0.35 (P < .01) (Figure 2). Additionally, we found that both HRR-level CT scan use and HRR-level bone scan use across prostate and breast cancers were positively correlated (Pearson coefficients of 0.27 [P = .01] and 0.19 [P = .09], respectively).

In the unadjusted model at the patient level, a man with low-risk prostate cancer had higher odds of undergoing inappropriate imaging if he lived in an HRR with higher inappropriate breast cancer imaging (Table 2; P = .01). In the multivariable model, the inappropriate imaging rates for these 2 cancers remained independently associated. A man with low-risk prostate cancer underwent inappropriate imaging 56.4%, 41.1%, 44.6%, and 34.2% of the time if he lived in the fourth, third, second, and first quartiles, respectively, of HRR-level inappropriate breast cancer imaging; this relationship between HRR-level rising percentages of prostate cancer imaging as a function of higher quartile of breast cancer imaging is depicted in Figure 1. Adjusted for age, comorbidity, and regional number of hospital beds, a man with low-risk prostate cancer had odds ratios (95% CIs) of 1.72 (1.12-2.65), 1.19 (0.78-1.81), and 1.76 (1.15-2.70) if he lived in the fourth, third, and second quartiles, respectively, of inappropriate breast cancer imaging, compared with the lowest quartile. Race, year of diagnosis, median household income, marital status, radiologist density, or HI were not significantly associated with inappropriate prostate cancer imaging. The association between patient-level inappropriate breast cancer imaging as a function of HRR-level inappropriate prostate cancer imaging remained positive but was not statistically significant (Table 3).

Discussion

We found that inappropriate imaging for both low-risk prostate and low-risk breast cancers was quite common in the pre–Choosing Wisely era. This suggests that Choosing Wisely was appropriately focused on these as important cancer care delivery problems. Furthermore, we found a positive association between regional-level inappropriate imaging for prostate and breast cancers. Consistent with this finding, both HRR-level CT and bone scan use were positively associated across cancer types. Patients with breast cancer were more likely to undergo resource-intensive imaging than those with prostate cancer. Regional-level breast cancer imaging rates more strongly influenced patient-level prostate cancer imaging (statistically significant) than vice versa (positive association, but not significant). This suggests that regions with higher rates of inappropriate breast cancer imaging may have infrastructure (ie, access to PET) or culture promoting imaging. Regions where prostate cancers were imaged frequently may not necessarily perform all imaging frequently, perhaps because the equipment for prostate cancer imaging is less specialized than that for breast cancer. Additionally, while regional-level imaging rates are associated, the patient-level prostate model demonstrates an inconsistent increase in prostate cancer imaging through the higher quartiles of breast cancer imaging. It is possible that the major prostate imaging division may come between the lowest quartile of breast cancer imaging and all of the other quartiles. This merits further exploration in subsequent research.

Our findings suggest that practice patterns may be a function of local propensities for health care utilization. This is a novel finding with great relevance to cancer policy. As patients with prostate cancer and breast cancer are a nonoverlapping cohort treated by nonoverlapping specialists, an association of inappropriate imaging between them suggests that regional culture and infrastructure contribute to health care utilization patterns across diseases. This finding is highly relevant for ASCO, since reduction of inappropriate prostate and breast cancer imaging are 2 of its Top Five priorities. Efforts to address inappropriate imaging must influence common local health system factors contributing to inappropriate imaging across provider types.

The association between regional variation and appropriateness of care is poorly understood. A recent systematic review found only 5 reviewable articles and concluded inappropriate utilization may not drive regional variation in the intensity or the cost of care.20 Other literature suggests that higher regional utilization of tests such as PSA, though not costly themselves, is associated with greater overall health care spending, greater care fragmentation, and more aggressive end-of-life care.34,35 Our study suggests that regions may have inherent characteristics, over and above market structure and competition, underlying health care decisions across diseases. Future work should further characterize determinates of these regional tendencies.

This study’s findings stand in contrast to the IOM’s position that individual-level decisions, rather than geography, drive regional health care spending variation. Rather, our findings are consistent with other recent publications suggesting that regional patterns do influence health care utilization.24,36 For instance, research recently describing the thermostat model27 (ie, regions with higher appropriate use will tend to have higher inappropriate use and vice versa) suggests that regions have inherent propensities to use health resources; but these regional effects have been thus far been explored only in single disease systems. Ko and colleagues36 observed that regions with higher cardiac catheterization rates performed more catheterizations among all patients, not only among those who needed catheterization.36 Our research group27 described a regional-level association between appropriate imaging in high-risk prostate cancer and inappropriate imaging in low-risk cancer, directly demonstrating the thermostat model. Other studies have explored the unintended consequences of policy efforts to decrease inappropriate utilization.22,37 As predicted by the thermostat model, decreases in inappropriate utilization are accompanied by decreases in appropriate utilization because the 2 are associated. The findings of the present study build on prior work by demonstrating that regional factors may contribute to imaging utilization even among different diseases, a novel observation. Understanding the common factors affecting regional cancer care—even factors working across distinct types of cancer—can help reveal the underpinnings of the thermostat model and may help guide the development of efficient policy solutions for decreasing inappropriate utilization.

Our study has important strengths. We used SEER-Medicare data to examine a large, population-based cohort of patients with prostate or breast cancer. This comprehensive resource permitted the novel, simultaneous study of 2 diseases affecting nonoverlapping cohorts, allowing us to demonstrate the similarity of imaging patterns in both. Rather than choosing arbitrary appropriateness criteria and facing the challenge of determining their compatibility across diseases, we defined appropriateness based on ASCO’s Choosing Wisely criteria. Framing inappropriate imaging around Choosing Wisely criteria ensures that our findings will be relevant for policy makers, clinicians, and patients seeking definitions of optimal care. However, ASCO’s Top Five list makes no mention of other potentially inappropriate imaging modalities, such as magnetic resonance imaging. Our analysis, therefore, almost certainly underestimates the degree of inappropriate imaging occurring in clinical practice.

Our study is limited because it uses retrospective, cross-sectional data. We were unable to determine long-term time trends in imaging because SEER only begins reporting PSA data in 2004, making it impossible to adequately determine prostate cancer risk in prior cohorts. Additionally, we did not include patients diagnosed after 2007. We acknowledge our data’s age and emphasize that our study focuses on the discovery of an association between inappropriate imaging in prostate cancer and breast cancer; this finding is not influenced by our data’s age. Our findings confirm the insights of the thought leaders who chose to highlight inappropriate cancer imaging through Choosing Wisely by demonstrating the extent of inappropriate imaging and showing how inappropriate imaging is concentrated among certain high-use regions, across diseases.

Our data limitations also restrict our ability to determine the causes of this regional association in inappropriate imaging. There was no comprehensive set of potentially confounding regional-level covariates and no qualitative data from physicians and patients that might explain the association. We were unable to determine whether any of the patients had symptoms attributable to advanced cancer, which might have prompted imaging.

Finally, the definitions for low-risk disease, though determined by a panel of expert clinicians and based on an abundance of clinical research, are potentially not known by or relevant to practicing clinicians and patients making the decisions in the community. This study should be repeated with data collected after the announcement of the ASCO Top Five list to determine whether inappropriate imaging rates have changed. It is unclear whether a Top Five list provides sufficient patient education or incentive to empower the patient to question the physician’s decision making, particularly during a time of high anxiety and stress.

Conclusions

We observed a regional-level association in the rates of inappropriate imaging of Medicare beneficiaries with low-risk prostate cancer and low-risk breast cancer. Our findings suggest that, contrary to the position of the IOM, regional-level factors may be important in determining utilization of health care resources. Therefore, researchers and policy makers should work to institute policy changes to improve quality of care and reduce health care spending in high-utilization areas in addition to focusing efforts on individual decision makers. Potential policy interventions might include formal educational programs to teach practitioners about specific guidelines, payment policies to reward appropriate care and penalize inappropriate care, and the design and implementation of clinical decision aids and support tools. Further research should be conducted to determine the causes of regional patterns of inappropriate imaging. Such research, including an evaluation of the clinicians and institutions performing these tests, might help optimize policy interventions aimed at improving the quality and lowering the cost of health care without decreasing access to care for those who need it.

Back to top
Article Information

Accepted for Publication: January 6, 2015.

Corresponding Author: Danil V. Makarov, MD, MHS, New York University School of Medicine, 550 First Ave, VZ30 Sixth Floor, Office 613, New York, NY 10016 (danil.makarov@nyumc.org).

Published Online: March 12, 2015. doi:10.1001/jamaoncol.2015.37.

Author Contributions: Dr Gross had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Makarov, Gross.

Acquisition, analysis, or interpretation of data: Makarov, Sen, Soulos, Gold, Yu, Ross.

Drafting of the manuscript: Makarov, Sen.

Critical revision of the manuscript for important intellectual content: Makarov, Soulos, Gold, Yu, Ross, Gross.

Statistical analysis: Makarov, Soulos, Gold.

Obtained funding: Gross.

Administrative, technical, or material support: Sen, Soulos, Gold, Gross.

Study supervision: Makarov, Gross.

Conflict of Interest Disclosures: Dr Makarov is a consultant for Castlight Health and for the US Food and Drug Administration. Dr Gross is a consultant for Johnson & Johnson, Medtronic Inc, and 21st Century Oncology. No other disclosures are reported. Dr Ross receives research support through Yale University from Medtronic Inc and Johnson & Johnson to develop methods of clinical trial data sharing.

Funding/Support: This research was supported by the Robert Wood Johnson Foundation, The Louis Feil Charitable Lead Trust, and the US Department of Veterans Affairs (VA), Veterans Health Administration, Health Services Research & Development Service (HSR&DS). Dr Makarov is a VA HSR&DS Career Development awardee at the Manhattan VA. The collection of the California cancer incidence data used in this study was supported by the California Department of Public Health as part of the statewide cancer reporting program mandated by California Health and Safety Code Section 103885; the National Cancer Institute (NCI) SEER Program under contract N01-PC-35136 awarded to the Northern California Cancer Center, contract N01-PC-35139 awarded to the University of Southern California, and contract N02-PC-15105 awarded to the Public Health Institute; and the Centers for Disease Control and Prevention (CDC) National Program of Cancer Registries, under agreement U55/CCR921930-02 awarded to the Public Health Institute. Dr Ross receives research support from the Centers for Medicare & Medicaid Services, to develop and maintain performance measures that are used for public reporting, and from the US Food and Drug Administration, to develop methods for postmarket surveillance of medical devices. Dr Ross is also supported by grant K08 AG032886 from the National Institute on Aging and by the American Federation for Aging Research through the Paul B. Beeson Career Development Award Program.

Role of the Funder/Sponsor: The funding institutions had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: The authors of this report are responsible for its content. The views expressed in this article are those of the authors and do not necessarily represent the views of the VA. The ideas and opinions expressed herein are those of the authors; endorsement by the State of California, Department of Public Health, the NCI, the CDC, or their contractors or subcontractors is not intended, nor should it be inferred.

Additional Contributions: We acknowledge the efforts of the NCI Applied Research Program; the Center for Medicare & Medicaid Services Office of Research, Development and Information; Information Management Services Inc; and the SEER Program tumor registries for the creation of the SEER-Medicare database. The interpretation and reporting of the SEER-Medicare data are the sole responsibility of the authors.

References
1.
Wennberg  JE, Fisher  ES, Skinner  JS.  Geography and the debate over Medicare reform.  Health Aff (Millwood). 2002;(Suppl Web Exclusives):W96-114.PubMedGoogle Scholar
2.
American Board of Internal Medicine Foundation.  The Choosing Wisely campaign: five things physicians and patients should question.http://www.choosingwisely.org/wp-content/uploads/2013/01/about_choosingwisely_fivethings.pdf. Accessed May 5, 2012.
3.
Cassel  CK, Guest  JA.  Choosing Wisely: helping physicians and patients make smart decisions about their care.  JAMA. 2012;307(17):1801-1802.PubMedGoogle ScholarCrossref
4.
Brody  H.  Medicine’s ethical responsibility for health care reform: the Top Five list.  N Engl J Med. 2010;362(4):283-285.PubMedGoogle ScholarCrossref
5.
Schnipper  LE, Smith  TJ, Raghavan  D,  et al.  American Society of Clinical Oncology identifies five key opportunities to improve care and reduce costs: the Top Five list for oncology.  J Clin Oncol. 2012;30(14):1715-1724.PubMedGoogle ScholarCrossref
6.
Rao  VM, Levin  DC.  The overuse of diagnostic imaging and the Choosing Wisely initiative.  Ann Intern Med. 2012;157(8):574-576.PubMedGoogle ScholarCrossref
7.
Mariotto  AB, Yabroff  KR, Shao  Y, Feuer  EJ, Brown  ML.  Projections of the cost of cancer care in the United States: 2010-2020.  J Natl Cancer Inst. 2011;103(2):117-128.PubMedGoogle ScholarCrossref
8.
Medicare Payment Advisory Commission.  Medicare payment policy: report to the Congress.http://www.medpac.gov/documents/reports/march-2012-report-to-the-congress-medicare-payment-policy.pdf?sfvrsn=0. Accessed May 10, 2012.
9.
Dinan  MA, Curtis  LH, Hammill  BG,  et al.  Changes in the use and costs of diagnostic imaging among Medicare beneficiaries with cancer, 1999-2006.  JAMA. 2010;303(16):1625-1631.PubMedGoogle ScholarCrossref
10.
Iglehart  JK.  The new era of medical imaging: progress and pitfalls.  N Engl J Med. 2006;354(26):2822-2828.PubMedGoogle ScholarCrossref
11.
Iglehart  JK.  Health insurers and medical-imaging policy: a work in progress.  N Engl J Med. 2009;360(10):1030-1037.PubMedGoogle ScholarCrossref
12.
Mitchell  JM.  Utilization trends for advanced imaging procedures: evidence from individuals with private insurance coverage in California.  Med Care. 2008;46(5):460-466.PubMedGoogle ScholarCrossref
13.
Hillman  BJ, Goldsmith  JC.  The uncritical use of high-tech medical imaging.  N Engl J Med. 2010;363(1):4-6.PubMedGoogle ScholarCrossref
14.
Hara  T, Kosaka  N, Kishi  H.  PET imaging of prostate cancer using carbon-11-choline.  J Nucl Med. 1998;39(6):990-995.PubMedGoogle Scholar
15.
Souchon  R, Rouvière  O, Gelet  A,  et al.  Visualisation of HIFU lesions using elastography of the human prostate in vivo: preliminary results.  Ultrasound Med Biol. 2003;29(7):1007-1015.PubMedGoogle ScholarCrossref
16.
Ito  H, Kamoi  K, Yokoyama  K, Yamada  K, Nishimura  T.  Visualization of prostate cancer using dynamic contrast-enhanced MRI: comparison with transrectal power Doppler ultrasound.  Br J Radiol. 2003;76(909):617-624.PubMedGoogle ScholarCrossref
17.
Halpern  EJ, Rosenberg  M, Gomella  LG.  Prostate cancer: contrast-enhanced us for detection.  Radiology. 2001;219(1):219-225.PubMedGoogle ScholarCrossref
18.
Warner  E, Messersmith  H, Causer  P, Eisen  A, Shumak  R, Plewes  D.  Systematic review: using magnetic resonance imaging to screen women at high risk for breast cancer.  Ann Intern Med. 2008;148(9):671-679.PubMedGoogle ScholarCrossref
19.
 Americans speak on health reform: report on health care community discussions.http://web.archive.org/web/20121016185712/http://www.healthreform.gov/reports/hccd/concernsd.html. Accessed January 28, 2015.
20.
Keyhani  S, Falk  R, Bishop  T, Howell  E, Korenstein  D.  The relationship between geographic variations and overuse of healthcare services: a systematic review.  Med Care. 2012;50(3):257-261.PubMedGoogle ScholarCrossref
21.
Committee on Geographic Variation in Health Care, Spending and Promotion of High-Value Care Board on Health Care Services, Institute of Medicine of the National Academies.  Variation in health care spending: target decision making, not geography.http://books.nap.edu/openbook.php?record_id=18393. Accessed January 28, 2015.
22.
Abraham  N, Wan  F, Montagnet  C, Wong  YN, Armstrong  K.  Decrease in racial disparities in the staging evaluation for prostate cancer after publication of staging guidelines.  J Urol. 2007;178(1):82-87.PubMedGoogle ScholarCrossref
23.
Cooperberg  MR, Lubeck  DP, Grossfeld  GD, Mehta  SS, Carroll  PR.  Contemporary trends in imaging test utilization for prostate cancer staging: data from the cancer of the prostate strategic urologic research endeavor.  J Urol. 2002;168(2):491-495.PubMedGoogle ScholarCrossref
24.
Kindrick  AV, Grossfeld  GD, Stier  DM, Flanders  SC, Henning  JM, Carroll  PR.  Use of imaging tests for staging newly diagnosed prostate cancer: trends from the CaPSURE database.  J Urol. 1998;160(6, pt 1):2102-2106.PubMedGoogle ScholarCrossref
25.
Saigal  CS, Pashos  CL, Henning  JM, Litwin  MS.  Variations in use of imaging in a national sample of men with early-stage prostate cancer.  Urology. 2002;59(3):400-404.PubMedGoogle ScholarCrossref
26.
Makarov  DV, Desai  RA, Yu  JB,  et al.  The population level prevalence and correlates of appropriate and inappropriate imaging to stage incident prostate cancer in the Medicare population.  J Urol. 2012;187(1):97-102.PubMedGoogle ScholarCrossref
27.
Makarov  DV, Desai  R, Yu  JB,  et al.  Appropriate and inappropriate imaging rates for prostate cancer go hand in hand by region, as if set by thermostat.  Health Aff (Millwood). 2012;31(4):730-740.PubMedGoogle ScholarCrossref
28.
National Cancer Institute.  SEER-Medicare linked database.http://appliedresearch.cancer.gov/seermedicare/. Accessed September 15, 2009, 2009.
29.
Warren  JL, Klabunde  CN, Schrag  D, Bach  PB, Riley  GF.  Overview of the SEER-Medicare data: content, research applications, and generalizability to the United States elderly population.  Med Care. 2002;40(8)(suppl):IV-3-IV-18.PubMedGoogle Scholar
30.
Carlson  RW, Allred  DC, Anderson  BO,  et al; National Comprehensive Cancer Network.  Invasive breast cancer.  J Natl Compr Canc Netw. 2011;9(2):136-222.PubMedGoogle Scholar
31.
Wennberg  JE, Cooper  MM.  Appendix on the geography of health care in the United States, 1999.http://www.dartmouthatlas.org/downloads/methods/geogappdx.pdf. Accessed January 28, 2015.
32.
Wennberg  JE, Cooper  MM.  The Dartmouth Atlas of Health Care in the United States. Chicago, IL: American Hospital Publishing; 1996.
33.
Elixhauser  A, Steiner  C, Harris  DR, Coffey  RM.  Comorbidity measures for use with administrative data.  Med Care. 1998;36(1):8-27.PubMedGoogle ScholarCrossref
34.
Bynum  J, Song  Y, Fisher  E.  Variation in prostate-specific antigen screening in men aged 80 and older in fee-for-service Medicare.  J Am Geriatr Soc. 2010;58(4):674-680.PubMedGoogle ScholarCrossref
35.
Ma  X, Wang  R, Long  JB,  et al.  The cost implications of prostate cancer screening in the Medicare population.  Cancer. 2014;120(1):96-102.PubMedGoogle ScholarCrossref
36.
Ko  DT, Wang  Y, Alter  DA,  et al.  Regional variation in cardiac catheterization appropriateness and baseline risk after acute myocardial infarction.  J Am Coll Cardiol. 2008;51(7):716-723.PubMedGoogle ScholarCrossref
37.
Makarov  DV, Loeb  S, Ulmert  D, Drevin  L, Lambe  M, Stattin  P.  Prostate cancer imaging trends after a nationwide effort to discourage inappropriate prostate cancer imaging.  J Natl Cancer Inst. 2013;105(17):1306-1313.PubMedGoogle ScholarCrossref
×