Context Referral to high-volume hospitals has been recommended for operations with a demonstrated volume-outcome relationship. The characteristics of patients who receive care at low-volume hospitals may be different from those of patients who receive care at high-volume hospitals. These differences may limit their ability to access or receive care at a high-volume hospital.
Objective To identify patient characteristics associated with the use of high-volume hospitals, using California's Office of Statewide Health Planning and Development patient discharge database.
Design, Setting, and Participants Retrospective study of Californians receiving the following inpatient operations from 2000 through 2004: elective abdominal aortic aneurysm repair, coronary artery bypass grafting, carotid endarterectomy, esophageal cancer resection, hip fracture repair, lung cancer resection, cardiac valve replacement, coronary angioplasty, pancreatic cancer resection, and total knee replacement.
Main Outcome Measures Patient race/ethnicity and insurance status in high-volume (highest 20% of patients by mean annual volume) and in low-volume (lowest 20%) hospitals.
Results A total of 719 608 patients received 1 of the 10 operations. Overall, nonwhites, Medicaid patients, and uninsured patients were less likely to receive care at high-volume hospitals and more likely to receive care at low-volume hospitals when controlling for other patient-level characteristics. Blacks were significantly (P<.05) less likely than whites to receive care at high-volume hospitals for 6 of the 10 operations (relative risk [RR] range, 0.40-0.72), while Asians and Hispanics were significantly less likely to receive care at high-volume hospitals for 5 (RR range, 0.60-0.91) and 9 (RR range, 0.46-0.88), respectively. Medicaid patients were significantly less likely than Medicare patients to receive care at high-volume hospitals for 7 of the operations (RR range, 0.22-0.66), while uninsured patients were less likely to be treated at high-volume hospitals for 9 (RR range, 0.20-0.81).
Conclusions There are substantial disparities in the characteristics of patients receiving care at high-volume hospitals. The interest in selective referral to high-volume hospitals should include explicit efforts to identify the patient and system factors required to reduce current inequities regarding their use.
Efforts to improve the quality of surgical care in the United States have led many organizations to advocate the use of high-volume hospitals for certain procedures. For example, the Leapfrog Group,1 a consortium of more than 170 private and public organizations that insure nearly 35 million individuals, incorporated volume standards for 5 operations (abdominal aortic aneurysm [AAA] repair, coronary artery bypass grafting [CABG], esophagectomy, coronary angioplasty, and pancreatic resection) into their hospital referral criteria. These recommendations rely on the use of the number of procedures performed at a hospital, a structural characteristic, as a marker of quality.
This shift toward the use of hospital procedure volume as a surrogate metric of quality is fostered by numerous reports of a direct volume-outcome relationship for certain procedures, with patients at high-volume hospitals consistently having better outcomes.2-13 Projections based on such studies predict substantial numbers of lives saved as a result of regionalization to high-volume hospitals for certain high-profile operations.14,15
While referral to high-volume hospitals is generally believed to be a positive step toward improving the quality of care, there may be problems in its implementation. Studies have shown a paucity of such hospitals.16 In California, less than 15% of hospitals are high volume (by Leapfrog criteria) for AAA repair, CABG, and esophagectomy. It also appears that there are important differences in the racial3,17 and socioeconomic status of patients who receive care at high- and low-volume hospitals. It is unclear if selective referral to high-volume hospitals will account for such differences and whether its implementation will exacerbate or improve these disparities. While prior studies have limited their analysis to specific procedures, the purpose of this study was to determine whether the use of high-volume hospitals varies by race/ethnicity or insurance status in a broad segment of patients undergoing surgical care.
We examined patient characteristics and use of high-volume hospitals across 10 hospital-based procedures with known volume-outcome relationships among Californians during a 5-year period (2000-2004). Using patient-level data, we investigated whether certain patient characteristics—race/ethnicity and insurance status—were more or less likely to be associated with the receipt of surgical care at high- vs low-volume hospitals. We hypothesized that nonwhite minorities and uninsured individuals or those with safety-net insurance such as Medicaid would be less likely to receive care at high-volume hospitals and concomitantly more likely to receive care at low-volume hospitals.
Data for all hospital discharges in California between January 1, 2000, and December 31, 2004, were obtained from California's Office of Statewide Health Planning and Development patient discharge database.18 The database includes discharge abstracts from all nonfederal hospitals in California and is compiled annually. Each discharge abstract includes codes for up to 20 inpatient procedures and 24 diagnoses per hospitalization. All procedures and diagnoses are coded using the International Classification of Disease, 9th Revision, Clinical Modification (ICD-9-CM). Also included are patient demographic information (race, Hispanic ethnicity, sex, expected source of payment, source of admission, age, and ZIP code of residence), outcomes (in-hospital mortality), and site of hospitalization (unique hospital identifier and ZIP code). Proxy income and urban vs rural residence location were assigned by patient ZIP code using data from the 2000 US Census.
Inclusion and Exclusion Criteria
Operations were selected based on established evidence from the literature for a volume-outcome relationship. In total, we selected 10 operations and procedures for evaluation: elective AAA repair, CABG, carotid endarterectomy, esophageal cancer resection, hip fracture repair, lung cancer resection, cardiac valve replacement, coronary angioplasty, pancreatic cancer resection, and total knee replacement.2-13 Each operation or procedure was studied independently. The specific ICD-9-CM codes used to identify the cohort sample are available from the authors on request. Patients with a non-California ZIP code were excluded, which ranged from 1.3% to 5.0% of the original sample size for each of the 10 operations. The unit of analysis was hospital discharge for each individual patient.
The dependent variable for our analyses was hospital volume category, divided among high-, medium-, and low-volume hospitals. Hospital volume for a given surgical procedure was defined as the mean annual procedural volume (ie, number of procedures performed divided by the number of years the hospital performed the operation) during the 5-year period (January 1, 2000-December 31, 2004). Mean volume was used because it represents an aggregate assessment of overall procedural experience during the period of analysis. The dependent variable was created by stratifying hospital volume into patient quintiles, with the highest 20% of patients by mean annual volume considered high volume, the middle 60% considered medium volume, and the lowest 20% considered low volume.
The independent variables included patient-level characteristics: race/ethnicity, sex, location (urban vs rural), insurance status, source of admission, age, comorbid conditions, income, and median distance to travel to the nearest high-, medium-, and low-volume hospital. Both race and ethnicity are reported by the hospital based on patient self-report. The race/ethnicity response options are defined by the California Office of Statewide Planning and Development and were reclassified into 5 mutually exclusive categories (non-Hispanic white [white], non-Hispanic black [black], Asian, Hispanic, and other). Patients reported as Hispanic were categorized as Hispanic, regardless of race. Insurance status was categorized based on the expected source of payment (Medicare, Medicaid, private insurance [any], other [eg, other government insurance], and uninsured [indigent, charity, no charge]). Source of admission was sorted into mutually exclusive groups (admitted from home, admitted through emergency department, transferred from another hospital, and other [non–emergency department]). Patients were grouped by age into 5 categories (<55, 55-64, 65-74, 75-84, and ≥85 years). Patient comorbidity was assessed using an adaptation of the Charlson Comorbidity Index.19,20 Each patient's index was calculated from the primary and 24 secondary diagnosis codes and was divided into 4 strata by score (0, 1, 2, and ≥3).
Patients were identified as living in a rural area based on their ZIP code and existing metropolitan statistical area definitions. Income was assigned based on the median income of the patient ZIP code reported in the 2000 US Census and then grouped by quintiles. Distance traveled was calculated by measuring the straight-line distance between the patient's residence and the hospital. The calculated distances were based on the longitude and latitude of each ZIP code's center point. Where a patient's ZIP code was missing or where values for income, urban vs rural residence, and distance were unknown or could not be calculated, values were assigned using hot-deck imputation21 with sampling observations restricted to patients treated at the same hospital.
The study protocol was reviewed and approved by the institutional review board at the University of California, Los Angeles. Due to the nature of these secondary data analyses, waiver of patient consent was granted.
Univariate analysis was performed using χ2 and t tests where appropriate. Graphs and tables were used as needed to examine the distribution of each variable. The crude distribution of race/ethnicity, type of insurance, and in-hospital mortality within each volume category was calculated for each operation. The neighborhoods of high- and low-volume hospitals were compared using the 2000 US Census linked to the hospital ZIP code; the variables of percentage of non-Hispanic whites older than 50 years and percentage of persons living below 200% of the poverty threshold were examined.
Generalized ordered logistic regression22 was used to estimate the impact of the primary predictors (race/ethnicity and insurance status) on the use of high- or low-volume hospitals, accounting for other patient characteristics (sex, rural/urban residence, source of admission, age, Charlson Comorbidity Index, income, and distances to closest high-, medium-, and low-volume hospitals). This model was selected instead of dichotomous logistic regression or standard ordered logistic regression because the outcome (hospital volume category) had 3 ordered categories and the model failed the proportional odds assumption of standard ordered logistic regression. The model predicted a patient's chance of receiving care in a particular hospital volume category (high, middle, or low) based on the patient's characteristics. Regression analysis did not take into account clustering of patients within hospitals, because choice of hospital was the outcome being modeled.
Relative risks (RRs) for receiving care at either high- or low-volume hospitals based on race/ethnicity and insurance status were calculated from the regression results. The results are expressed as RRs with respect to the reference groups white and Medicare. Ninety-five percent confidence intervals for the RRs were calculated using bootstrapping with 1000 repetitions and the bias-corrected technique.23 All other patient-level characteristics were included in the analysis to account for differences among patients. These results are available from the authors on request. All statistical analyses were performed using Intercooled Stata version 7 (StataCorp, College Station, Tex). P<.05 was used to establish statistical significance.
A total of 719 608 patients received 1 of the 10 operations. Table 1 lists demographic information for the patients receiving each operation or procedure. Seventy-five percent were white, 4% were black, 5% were Asian, 12% were Hispanic, and 4% were other race/ethnicity; 45% were women. Sixty percent had Medicare, 29% had private insurance, 5% had Medicaid, and 3% were uninsured. Sixty-seven percent of the patients were older than 65 years. Thirty-two percent had a Charlson Comorbidity Index of 0, while 35%, 19%, and 14% had an index of 1, 2, or 3 or greater, respectively. Fifty-three percent of patients were admitted from home, 31% were admitted through the emergency department, and 11% were transferred from another hospital. Six percent of patients lived in a rural area. The median distance traveled to a hospital was always greater to a high-volume hospital in comparison with medium- or low-volume hospitals. In general, the number of California hospitals meeting the Leapfrog Group criteria was low, ranging from 1% for AAA repair to 20% for coronary angioplasty.
For all 10 operations, in-hospital crude mortality rates decreased as hospital volume increased. For example, the in-hospital mortality for cardiac valve surgery was 5.4% at high-, 6.3% at medium-, and 7.4% at low-volume hospitals, while the in-hospital mortality for coronary angioplasty was 1.2% at high-, 1.8% at medium-, and 2.1% at low-volume hospitals. The largest differences between in-hospital mortality at high- vs low-volume hospitals were seen for pancreatic cancer resection (3% vs 13%) and esophageal cancer resection (7% vs 11%). The number of hospitals performing each operation ranged from 131 for CABG to 359 for hip fracture repair. Compared with low-volume hospitals, the high-volume hospitals were more often located in neighborhoods with higher percentages of non-Hispanic whites older than 50 years and lower percentages of persons living below 200% of the poverty threshold.
Table 2 describes patient characteristics in high- and low-volume hospitals for each of the 10 procedures, stratified by both race/ethnicity and insurance status. For example, across the 4 race/ethnicity categories for total knee replacement, the percentage of patients with private insurance was higher for patients going to a high-volume (28%-34%) vs a low-volume (14%-29%) hospital, while the percentage of patients without insurance was lower for patients going to a high-volume (1%-3%) vs a low-volume (2%-6%) hospital. In general, these unadjusted analyses demonstrate disparities for both race/ethnicity and insurance status.
In multivariate analysis, race and ethnicity were strongly associated with the use of high- and low-volume hospitals, even after accounting for sex, rural location, insurance status, source of admission, age, Charlson Comorbidity Index, income, and the patients' proximity to high-, medium-, and low-volume hospitals (Table 3). Black, Asian, and Hispanic patients were less likely, compared with white patients, to receive care at high-volume hospitals and more likely to receive care at low-volume hospitals. For example, among patients undergoing elective AAA repair, blacks were only 71% (95% confidence interval, 53%-93%) as likely as whites to be treated at high-volume hospitals after adjusting for the other patient-level characteristics in the model. For all 10 operations, black patients were significantly (P<.05) less likely to receive care at high-volume hospitals in 6 of the operations (RR range, 0.40-0.72), Asians less likely in 5 (RR range, 0.60-0.91), and Hispanics less likely in 9 (RR range, 0.46-0.88). Moreover, blacks, Asians, and Hispanics were more likely to receive care at low-volume hospitals. Utilizing a low-volume hospital was significantly associated with being black for 7 of the operations (RR range, 1.19-1.88), Asian for 7 (RR range, 1.23-1.77), and Hispanic for 9 (RR range, 1.11-1.64).
Insurance type or status also had a significant impact on the utilization of high- and low-volume hospitals, even after accounting for other patient characteristics (Table 4). Compared with Medicare patients, those receiving Medicaid or those with no insurance (uninsured) were almost uniformly less likely to be treated at high-volume hospitals and more likely to be treated at low-volume hospitals. Medicaid insurance was a negative predictor of utilization of high-volume hospitals in 7 of the operations (RR range, 0.22-0.66) and a positive predictor of utilization of low-volume hospitals in 8 (RR range, 1.09-2.55) when compared with Medicare. Similarly, uninsured patients, relative to patients with Medicare, were less likely to be treated at high-volume hospitals in 9 of the operations (RR range, 0.20-0.81) and more likely to be treated at low-volume hospitals in 7 (RR range, 1.09-3.31). Conversely, private insurance was a positive predictor of utilization of high-volume hospitals and a negative predictor of utilization of low-volume hospitals. Patients with private insurance were significantly more likely to receive care at high-volume hospitals for 3 of the operations (RR range, 1.14-1.46). For 4 of the operations, patients with private insurance were less likely to receive care at low-volume hospitals (RR range, 0.75-0.91) compared with Medicare patients.
Although debate remains regarding the importance of the volume-outcome relationship, evidence of this relationship has led many organizations to accept procedural volume as a structural proxy for quality. In validating these relationships, prior volume-outcomes studies have alluded to the potential for racial/ethnic disparities in the utilization of high-volume hospitals but have not rigorously tested this hypothesis accounting for disease severity and case mix.
As such, the aim of the current study was to examine the effect of race/ethnicity and insurance status on receipt of complex surgical care at high- or low-volume hospitals. We demonstrate consistent and robust disparities in the use of high- and low-volume hospitals across 10 complex inpatient procedures with respect to race/ethnicity and insurance status, while controlling for clinical and demographic covariates. In general, blacks, Asians, Hispanics, patients with Medicaid, and uninsured patients were less likely to go to high-volume hospitals for complex surgical procedures but more likely to go to low-volume hospitals, when compared with whites and patients with Medicare. Furthermore, patients with private insurance were significantly more likely to go to high-volume hospitals for 3 of the surgical procedures.
The current study demonstrates a significant disparity in the distribution of patients at high- and low-volume hospitals with respect to race/ethnicity and insurance status. Without explicitly addressing this disparity, selective referral has the potential to sustain or even exacerbate this current level of disparities. Furthermore, while the Leapfrog Group1 aims to improve the care for insured individuals, selective referral for those privately insured may not address the problem of caring for those who are uninsured (now estimated at 46 million24).
It is not surprising that although we studied disparity of race/ethnicity and insurance status separately, there was substantial overlap in these effects. For example, the crude rates of high- or low-volume hospital use stratified by race/ethnicity demonstrate that minorities receiving complex surgical care at low-volume hospitals generally had lower rates of private insurance coverage and higher rates of being uninsured compared with non-Hispanic whites. These results are consistent with data for health insurance status in the low-income, nonelderly population. The major source of coverage for nonwhites was Medicaid, while the largest source of coverage for non-Hispanic whites was employer based. In addition, non-Hispanic whites had lower rates of being uninsured, regardless of whether income was above or below 200% of the federal poverty threshold.25 These findings suggest that there is some important degree of collinearity that must be recognized between the variables of patient income, race/ethnicity, and insurance status. While collinearity is likely to understate the statistical significance of each variable because the effect is consistently in the same direction for each measure (ie, lower use for more vulnerable populations), it may be advantageous from the policy perspective because the solution for any single group will likely provide overlapping benefit to many.
In addition to the access issues related to payer status, there are other important patient- and physician-related factors that likely need to be addressed to lessen observed disparities. Patient-related factors, for example, range from a lack of knowledge about the volume-outcome association to inadequate basic transportation. Personal preference and individual cultural health behaviors are likewise important reasons why certain patients may choose not to go to a higher-quality health care facility even if they have access to complex surgical care at a high-volume hospital.26
Physician-related factors are also important. Bach et al27 showed that primary care physicians faced greater difficulties in obtaining access for black patients to high-quality physicians and hospitals. In addition, physicians caring for black patients were less likely to be board certified. Also implicit in the discussion of patient race/ethnicity is the issue of the physician's cultural competence. It has been repeatedly shown that culturally “competent” physicians improve the quality and effectiveness of care when the patient and physician are of concordant race/ethnicity.28 While there has been a push toward increasing the numbers of minorities in surgical training (and therefore the future surgical workforce), there remains an identified shortage at present. Although we were unable to distinguish between these patient- and physician-related factors in the current study, together they represent important areas that warrant further investigation.
Despite the presence of these patient- and physician-related factors, it is important to highlight the main finding of the current study: that the patient populations at high- and low-volume hospitals are different. This finding raises a question of whether the patient mix at high-volume centers is what influences the outcomes. On the surface, it may appear that the outcomes at high-volume hospitals should be similar for patients who usually seek care at low-volume hospitals; however, it remains unclear if high-volume hospitals would still have better outcomes if patients at low-volume hospitals were redirected there. Patients who usually seek care at low-volume hospitals may have a higher level of disease severity and burden of comorbid conditions.29 For example, uninsured patients have less physician or hospital continuity in the delivery of care, less utilization of preventive services, and an overall lower level of health compared with insured patients.30-32 It may be that by the time these patients present for surgical therapy their disease and comorbid conditions are more advanced than in patients with better preventive and primary care services.
This notion has been challenged by a recent study demonstrating that all races, not just minority patients, have higher mortality rates at low-volume hospitals and that differences in the use of such hospitals largely explained the higher mortality rates among blacks undergoing complex operations.33 That study suggested that the presumption of “worsened” case mix at low-volume hospitals may not be the dominant factor underlying differences in outcome at such hospitals. Another consideration is that the increased racial/ethnic minority population currently seen at low-volume hospitals may not be an ideal match with the high-volume hospitals based on patient-physician concordance.34,35 Addressing this issue may be an important factor when considering redirection of patients to high-volume hospitals.
This study has a number of limitations. Administrative data may lack important clinical information. The study may not be generalizable to other geographic regions. Disparities demonstrated in the 10 complex inpatient surgical operations may not be applicable to other procedures. Observational data cannot capture potential sources of bias arising from patient selection factors. The race/ethnicity variable may be subject to some inaccuracy due to incorrect self-reporting or the problems associated with coding mixed race.
Despite these limitations, our study makes several unique contributions to the literature, including consistent findings across 10 complex procedures and the ability to look at multiple ethnic and racial groups. Furthermore, we identified significant disparities secondary to race/ethnicity and insurance status, using multivariate analysis to adjust for multiple patient-level covariates. Finally, our study demonstrates robust findings in a large (12% of the US population), ethnically diverse population that includes all patients undergoing the selected procedures without restrictions based on demographics, insurance, or sampling. While there is significant interest among health care policy experts in improving quality by directing patients to high-volume hospitals, policy development should include explicit efforts to identify the patient and system factors required to reduce current inequities in the receipt of care at such hospitals.
Corresponding Author: Clifford Y. Ko, MD, MS, MSHS, Department of Surgery, David Geffen School of Medicine, 10833 Le Conte Ave, 72-215 CHS, Box 956904, Los Angeles, CA 90095-6904 (cko@mednet.ucla.edu).
Author Contributions: Dr Liu 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: Liu, Zingmond, Ettner, Brook, Ko.
Acquisition of data: Zingmond, Ko.
Analysis and interpretation of data: Liu, Zingmond, McGory, SooHoo, Ettner, Ko.
Drafting of the manuscript: Liu, Zingmond.
Critical revision of the manuscript for important intellectual content: Liu, Zingmond, McGory, SooHoo, Ettner, Brook, Ko.
Statistical analysis: Liu, Zingmond, McGory, SooHoo, Ettner.
Obtained funding: Zingmond.
Study supervision: Zingmond, Brook, Ko.
Financial Disclosures: None reported.
Funding/Support: This study was supported by a Veterans Administration/UCLA Multicampus Ambulatory Healthcare Fellowship (Dr Liu), by Career Development Award NIA 1K08AG023024 (Dr Zingmond), and by the UCLA Robert Wood Johnson Clinical Scholars Program (Dr McGory).
Role of the Sponsors: The funding organizations had no role in the design and conduct of the study; in the collection, analysis, and interpretation of the data; or in the preparation, review, or approval of the manuscript.
2.Begg CB, Cramer LD, Hoskins WJ, Brennan MF. Impact of hospital volume on operative mortality for major cancer surgery.
JAMA. 1998;280:1747-17519842949
Google ScholarCrossref 3.Birkmeyer JD, Siewers AE, Finlayson EV.
et al. Hospital volume and surgical mortality in the United States.
N Engl J Med. 2002;346:1128-113711948273
Google ScholarCrossref 4.Halm EA, Lee C, Chassin MR. Is volume related to outcome in health care? a systematic review and methodologic critique of the literature.
Ann Intern Med. 2002;137:511-52012230353
Google ScholarCrossref 5.Hughes RG, Garnick DW, Luft HS, McPhee SJ, Hunt SS. Hospital volume and patient outcomes: the case of hip fracture patients.
Med Care. 1988;26:1057-10673185017
Google ScholarCrossref 6.Luft HS, Parker JD. Volume and mortality in coronary artery bypass grafting.
BMJ. 1995;311:1304-13057496262
Google ScholarCrossref 7.Patti MG, Corvera CU, Glasgow RE, Way LW. A hospital's annual rate of esophagectomy influences the operative mortality rate.
J Gastrointest Surg. 1998;2:186-1929834415
Google ScholarCrossref 8.Peck C, Peck J, Peck A. Comparison of carotid endarterectomy at high- and low-volume hospitals.
Am J Surg. 2001;181:450-45311448441
Google ScholarCrossref 9.Phillips KA, Luft HS, Ritchie JL. The association of hospital volumes of percutaneous transluminal coronary angioplasty with adverse outcomes, length of stay, and charges in California.
Med Care. 1995;33:502-5147739274
Google ScholarCrossref 10.Showstack JA, Rosenfeld KE, Garnick DW, Luft HS, Schaffarzick RW, Fowles J. Association of volume with outcome of coronary artery bypass graft surgery: scheduled vs nonscheduled operations.
JAMA. 1987;257:785-7893492614
Google ScholarCrossref 11.van Lanschot JJ, Hulscher JB, Buskens CJ, Tilanus HW, ten Kate FJ, Obertop H. Hospital volume and hospital mortality for esophagectomy.
Cancer. 2001;91:1574-157811301408
Google ScholarCrossref 12.Katz JN, Barrett J, Mahomed NN, Baron JA, Wright RJ, Losina E. Association between hospital and surgeon procedure volume and the outcomes of total knee replacement.
J Bone Joint Surg Am. 2004;86-A:1909-191615342752
Google Scholar 13.Kansagra SM, Curtis LH, Schulman KA. Regionalization of percutaneous transluminal coronary angioplasty and implications for patient travel distance.
JAMA. 2004;292:1717-172315479937
Google ScholarCrossref 14.Birkmeyer JD, Finlayson EV, Birkmeyer CM. Volume standards for high-risk surgical procedures: potential benefits of the Leapfrog initiative.
Surgery. 2001;130:415-42211562662
Google ScholarCrossref 15.Dudley RA, Johansen KL, Brand R, Rennie DJ, Milstein A. Selective referral to high-volume hospitals: estimating potentially avoidable deaths.
JAMA. 2000;283:1159-116610703778
Google ScholarCrossref 16.Liu JH, Etzioni DA, O’Connell JB, Maggard MA, Ko CY. Using volume criteria: do California hospitals measure up?
J Surg Res. 2003;113:96-10112943816
Google ScholarCrossref 17.Birkmeyer JD, Stukel TA, Siewers AE, Goodney PP, Wennberg DE, Lucas FL. Surgeon volume and operative mortality in the United States.
N Engl J Med. 2003;349:2117-212714645640
Google ScholarCrossref 19.Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation.
J Chronic Dis. 1987;40:373-3833558716
Google ScholarCrossref 20.Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with
ICD-9-CM administrative databases.
J Clin Epidemiol. 1992;45:613-6191607900
Google ScholarCrossref 21.Little RJA, Rubin DB. Statistical Analysis With Missing Data. New York, NY: Wiley-Interscience; 1987
22.Long JS. Ordinal outcomes—ordered logit and ordered probit analysis. In: Regression Models for Categorical and Limited Dependent Variables. Thousand Oaks, Calif: SAGE Publications Inc; 1997
23.Efron B, Tibshirani R. Bootstrap measures for standard errors, confidence intervals, and other measures of statistical accuracy.
Stat Sci. 1986;1:54-77
Google ScholarCrossref 26.Epstein AM, Weissman JS, Schneider EC, Gatsonis C, Leape LL, Piana RN. Race and gender disparities in rates of cardiac revascularization: do they reflect appropriate use of procedures or problems in quality of care?
Med Care. 2003;41:1240-125514583687
Google ScholarCrossref 27.Bach PB, Pham HH, Schrag D, Tate RC, Hargraves JL. Primary care physicians who treat blacks and whites.
N Engl J Med. 2004;351:575-58415295050
Google ScholarCrossref 28.Schrag D, Cramer LD, Bach PB, Cohen AM, Warren JL, Begg CB. Influence of hospital procedure volume on outcomes following surgery for colon cancer.
JAMA. 2000;284:3028-303511122590
Google ScholarCrossref 29.Glance LG, Dick AW, Mukamel DB, Osler TM. Is the hospital volume-mortality relationship in coronary artery bypass surgery the same for low-risk versus high-risk patients?
Ann Thorac Surg. 2003;76:1155-116214530004
Google ScholarCrossref 30.Taylor AK, Cohen JW, Machlin SR. Being uninsured in 1996 compared to 1987: how has the experience of the uninsured changed over time?
Health Serv Res. 2001;36:16-3116148958
Google Scholar 31.Okoro CA, Young SL, Strine TW, Balluz LS, Mokdad AH. Uninsured adults aged 65 years and older: is their health at risk?
J Health Care Poor Underserved. 2005;16:453-46316086008
Google ScholarCrossref 32.Ayanian JZ, Weissman JS, Schneider EC, Ginsburg JA, Zaslavsky AM. Unmet health needs of uninsured adults in the United States.
JAMA. 2000;284:2061-206911042754
Google ScholarCrossref 33.Lucas FL, Stukel TA, Morris AM, Siewers AE, Birkmeyer JD. Race and surgical mortality in the United States.
Ann Surg. 2006;243:281-28616432363
Google ScholarCrossref 34.Johnson RL, Roter D, Powe NR, Cooper LA. Patient race/ethnicity and quality of patient-physician communication during medical visits.
Am J Public Health. 2004;94:2084-209015569958
Google ScholarCrossref 35.Cooper-Patrick L, Gallo JJ, Gonzales JJ.
et al. Race, gender, and partnership in the patient-physician relationship.
JAMA. 1999;282:583-58910450723
Google ScholarCrossref