The Relationship Between Competition and Quality in Procedural Cardiac Care David B. Glick, MD, MBA,* Kristen Wroblewski, MS,† Sean Apfelbaum, BA,* Benjamin Dauber, BA,* Joyce Woo, BA,‡ and Avery Tung, MD* BACKGROUND: Anesthesiologists are frequently involved in efforts to meet perioperative quality metrics. The degree to which hospitals compete on publicly reported quality measures, however, is unclear. We hypothesized that hospitals in more competitive environments would be more likely to compete on quality and thus perform better on such measures. To test our hypothesis, we studied the relationship between competition and quality in hospitals providing procedural cardiac care and participating in a national quality database. METHODS: For hospitals performing heart valve surgery (HVS) and delivering acute myocardial infarction (AMI) care in the Hospital Compare database, we assessed the degree of intrahospital competition using both geographical radius and federally defined metropolitan statistical area (MSA) to determine the degree of intrahospital competition. For each hospital, we then correlated the degree of competition with quality measure performance, mortality, patient volume, and per-patient Medicare costs for both HVS and AMI. RESULTS: Six hundred fifty-three hospitals met inclusion criteria for HVS and 1898 hospitals for AMI care. We found that for both definitions of competition, hospitals facing greater competition did not demonstrate better quality measure performance for either HVS or AMI. For both diagnoses, competition by number of hospitals correlated positively with cost: partial correlation coefficients = 0.40 (0.42 for MSA) (P < 0.001) for HVS and 0.52 (0.47 for MSA) (P < 0.001) for AMI. CONCLUSIONS: An analysis of the Hospital Compare database found that competition among hospitals correlated overall with increased Medicare costs but did not predict better scores on publicly reported quality metrics. Our results suggest that hospitals do not compete meaningfully on publicly reported quality metrics or costs. (Anesth Analg 2015;120:220–9)
A
n increasing concern among health care consumers, insurers, and regulators is the quality of health care being delivered in the United States. Recent studies have suggested that American health care is plagued with errors,1 unexplained practice variability,2 guideline noncompliance,3 and other potential markers of substandard quality. Proposed strategies to improve quality have included increased regulation and reimbursement schemes based on achieving specific quality goals. Results of these strategies have been mixed, however, with reports of successes and failures.4–6 Another proposed approach to improving health care quality is to facilitate competition among hospitals.7 Economic market theories predict that hospitals competing to attract patients will seek to differentiate their product. From the Departments of *Anesthesia & Critical Care and †Health Studies, University of Chicago, Chicago, Illinois; and ‡University of Chicago Pritzker School of Medicine, Chicago, Illinois. Joyce Woo, BA, is currently affiliated with the Department of Pediatrics, University of Chicago, Chicago, Illinois. Accepted for publication July 14, 2014. Funding: All financial support for this work came from the University of Chicago Departments of Anesthesia & Critical Care and Health Studies. The authors declare no conflicts of interest. This report was previously presented, in part, at the American Society of Anesthesiologists annual meeting on October 16 (2011), October 16 and 20 (2010), October 20 (2009), and at the International Anesthesia Research Society annual meeting on May 22, 2011, which was the subject of an article in Anesthesiology News. Reprints will not be available from the authors. Address correspondence to David B. Glick, MD, MBA, Department of Anesthesia & Critical Care, University of Chicago, 5841 S Maryland Ave., MC 4028, Chicago, IL 60637. Address e-mail to [email protected]. Copyright © 2014 International Anesthesia Research Society DOI: 10.1213/ANE.0000000000000456
220 www.anesthesia-analgesia.org
By mandating public reporting of quality-related metrics, health care regulators may induce hospitals to compete with each other on quality, leading to overall improvements in care.5 In this way, the quality of health care delivery would improve as higher quality providers are rewarded and lower quality providers either improve or are eliminated from the marketplace. Anesthesiologists are central figures in efforts to meet many perioperative quality metrics. Such metrics, however, may not matter to patients,8 may be only distantly related to relevant health care outcomes,6 and may thus be a low priority for hospitals seeking marketplace success. Third-party payers, patient preferences, government interventions such as limiting specialty care providers by requiring certificates of need, and ethical concerns may also distort relationships between competition and quality in the health care market. To attract patients and physicians, hospitals may choose to compete by offering physical enhancements unrelated to health care, such as a more appealing physical environment, better quality food, or easier parking. Finally, hospitals in less competitive environments may not perceive a strong need to compete for patients since they enjoy a near-monopoly market position. Previous work examining the relationship between a hospital’s competitive environment and the quality of care it delivers has had mixed results.9–11 Those data, however, were mostly compiled before the introduction of publicly reported quality measure performance. Because public reporting is relatively new, the effect on hospital behavior is not yet known. More recent research on relationships between quality and competition has found similarly mixed outcomes in the public reporting era. One 2013 study12 found increased adoption of endovascular technologies for aneurysm repair in more competitive markets January 2015 • Volume 120 • Number 1
but no change in mortality or complications. Another study of quality metrics for heart failure found little effect of competitive environment.13 In a third study,14 hospitals in competitive markets with high health maintenance organization (HMO) penetration reduced costs and improved quality, whereas competition in markets without HMO penetration resulted in increased costs and no change in outcomes. With respect to patient decision making, some evidence suggests that patients may prefer higher quality hospitals15 but that the effect is small.16 Taken together, these data suggest that competitive environments may not lead to improved care quality, even with public reporting. To better understand the relationship between competition and quality in an era of publicly reported quality measures, we used the U.S. Department of Health and Human Services Hospital Compare database to obtain data on quality measure performance, hospital outcomes, and costs for an elective surgical procedure (heart valve surgery [HVS]) and a related emergent medical diagnosis (acute myocardial infarction [AMI]) that, because of its acuity, might allow for less consumer choice.17 The Hospital Compare database is part of the Center for Medicare and Medicaid Services Hospital Quality Initiative and contains approximately 70% of all hospitals providing cardiac care in the United States.18 We correlated outcomes, quality measure performance, and cost data for each hospital with their competitive environment, assessed by the number of hospitals in the same geographical area or by federally defined metropolitan statistical area (MSA). We hypothesized that competition, particularly for an elective procedure where patients are most able to choose among providers, would result in better performance on publicly reported quality measures.
METHODS
This study was reviewed and exempted from formal institutional review by the University of Chicago IRB. We abstracted 2007 and 2008 cost, competition, and quality data for hospitals performing HVS and participating in the publicly available U.S. Department of Health and Human Services Hospital Compare database. We included number of admissions, mortality and morbidity data, location by state and zip code, and hospital performance on quality metrics. To measure the degree of competition faced by each hospital, we first searched the database for all hospitals within 20 miles of the index hospital and screened for those with >10 HVS admissions per year. Hospitals meeting these criteria were then tallied to quantify the degree of competition. As a comparison group, we repeated the above analyses for the related but nonprocedural diagnosis of AMI. To evaluate a hospital’s performance on quality metrics, we used mortality rates, complication scores, and Center for Medicare and Medicaid Services Hospital Quality measures for each of the diagnoses. For 2007 and 2008, the database used these metrics to group hospitals by performance into the top 25%, middle 50%, or lowest 25% for each diagnosis. To obtain cost data, we used the Medicare cost report for each hospital and Diagnosis Related Group. To partially adjust for the effect of geographic variability on Medicare payments unrelated to hospital costs, we used median household income for the zip code of each hospital obtained via the 2000 U.S. census.
January 2015 • Volume 120 • Number 1
STATISTICS
Zero-order and partial Spearman rank correlation coefficients were used to initially assess the association among number of patients (volume), competition, cost, and median household income. Hospitals were also separated into quartiles based on their level of competition (quartile #1 = fewest number of competitors) and number of patients treated with each diagnosis (quartile #1 = lowest number of patients). Quality data were represented as top 25%, middle 50%, and bottom 25% for each measure in the database and also as a dichotomous variable (top 25% versus not) for statistical analysis. The effect of competition, volume, cost, and median household income on quality measure performance was determined using logistic regression with competition, volume, and cost quartiles as the primary independent variables, household income as a continuous covariate, and position in the top 25% of performers for each quality measure as the dependent variable. Competition, volume, and cost quartiles were used because of the skewness of the data and for ease of interpretation. The effect of competition on mortality and cost was then evaluated using logistic regression with competition quartile as the independent variable and mortality or cost above the median as the dependent variable. The resulting odds ratios and 95% confidence intervals (CIs) were reported. To assess the effect of variability due to low-volume hospitals, we repeated the above analyses at volume thresholds of 50 and 100 cases per year (instead of 10/year). Because we chose an arbitrary 20-mile radius to assess competitive environment, we also repeated the above analyses using each hospital’s MSA instead of geographical radius. Correlation between MSA and geographical definitions of competition was strong (e.g., r = 0.80 for HVS and 0.73 for AMI, P < 0.001). For all comparisons, two-sided P values 10 HVS and 1898 hospitals caring for >10 cases of AMI. Of the 24 states included in the database, California had the highest percentage of hospitals (16.7% for HVS), and Vermont the lowest (0.15%). The distribution of hospitals by state and procedure volume is shown in Table 1. For hospitals performing HVS, competition did not correlate with quality measure performance. Performance on each of the 6 publicly reported measures listed in Table 2 was not significantly different across competitive quartiles (Fig. 1). Raising our inclusion threshold to hospitals performing >50 and >100 cases and repeating our analysis produced similar results (Appendix Tables 1 and 2). Top 25% performance in any single measure or combination of measures did not correlate significantly with mortality. Logistic regression analysis with competition and cost quartiles as independent variables and mortality (above/below the median) as the dependent variable also found no significant relationship between competition and mortality (2nd Q odds ratio [OR] = 0.78 [95% CI, 0.51–1.21], 3rd Q OR = 0.63 [95% CI, 0.41–0.98], 4th Q OR = 0.68 [95% CI, 0.43–1.07]). We found that increased competition, defined using geographical area, correlated with higher average Medicare costs (zero-order Spearman rank correlations are presented
www.anesthesia-analgesia.org
221
Competition and Quality in Procedural Health Care
Table 1. Distribution of Hospitals by State for Each Diagnosis State Arizona California Colorado Florida Iowa Illinois Massachusetts Maryland Maine North Carolina New Hampshire New Jersey New Mexico Nevada New York Oregon Pennsylvania Rhode Island Texas Utah Virginia Vermont Washington Wisconsin
Heart valve surgery N (% of total) 25 (3.8) 109 (16.7) 17 (2.6) 72 (11.0) 12 (1.8) 58 (8.9) 15 (2.3) 9 (1.4) 3 (0.5) 21 (3.2) 4 (0.6) 18 (2.8) 5 (0.8) 10 (1.5) 40 (6.1) 12 (1.8) 60 (9.2) 3 (0.5) 86 (13.2) 11 (1.7) 18 (2.8) 1 (0.2) 17 (2.6) 27 (4.1)
Acute myocardial infarction care N (% of total) 52 (2.7) 263 (13.9) 37 (1.9) 173 (9.1) 54 (2.9) 131 (6.9) 64 (3.4) 43 (2.3) 33 (1.7) 83 (4.4) 21 (1.1) 71 (3.7) 20 (1.0) 21 (1.1) 177 (9.3) 45 (2.4) 143 (7.5) 10 (0.5) 221 (11.6) 16 (0.8) 74 (3.9) 11 (0.6) 54 (2.8) 81 (4.3)
Table 2. Quality Metrics for Heart Valve Surgery Antibiotic given within 1 h of incision Right kind of antibiotic Antibiotic discontinued within 24 h of surgery Deep venous thrombosis prophylaxis ordered Deep venous thrombosis prophylaxis received Glucose control after cardiac surgery
in Table 3). Multivariate analyses also demonstrated positive correlations between competition and cost (partial correlation = 0.40 adjusting for volume and income, P < 0.001). Furthermore, logistic regression analysis showed an incrementally positive effect of competition on the odds of having costs above the median (Table 4). When we repeated the above analyses using MSA instead of geographical radius, we found similar relationships between competition and cost (partial correlation = 0.42) and incrementally positive effects of competition on the odds of the costs being greater than the median cost. As a comparison group, we examined hospitals treating AMI, an emergent diagnosis related to HVS, but with fewer opportunities for consumer choice because of its acuity. Consistent with HVS, performance on the publicly reported quality measures for AMI listed in Table 5 did not differ depending on competitive environment as defined by geographical radius (Fig. 2). As with HVS, repeating the above analyses with volume thresholds of 50 or 100 cases per year produced similar results (Appendix Tables 3 and 4). When hospitals were grouped by top 25% versus bottom 75% quality measure performance, no effect of group on mortality was noted. For AMI, competition correlated inversely, albeit weakly, with mortality (partial correlation = −0.08, P = 0.002, N = 1628). As with HVS, multivariate analysis found a positive correlation between competition in AMI care and costs (partial correlation = 0.52 adjusting for volume and income, P
A
n increasing concern among health care consumers, insurers, and regulators is the quality of health care being delivered in the United States. Recent studies have suggested that American health care is plagued with errors,1 unexplained practice variability,2 guideline noncompliance,3 and other potential markers of substandard quality. Proposed strategies to improve quality have included increased regulation and reimbursement schemes based on achieving specific quality goals. Results of these strategies have been mixed, however, with reports of successes and failures.4–6 Another proposed approach to improving health care quality is to facilitate competition among hospitals.7 Economic market theories predict that hospitals competing to attract patients will seek to differentiate their product. From the Departments of *Anesthesia & Critical Care and †Health Studies, University of Chicago, Chicago, Illinois; and ‡University of Chicago Pritzker School of Medicine, Chicago, Illinois. Joyce Woo, BA, is currently affiliated with the Department of Pediatrics, University of Chicago, Chicago, Illinois. Accepted for publication July 14, 2014. Funding: All financial support for this work came from the University of Chicago Departments of Anesthesia & Critical Care and Health Studies. The authors declare no conflicts of interest. This report was previously presented, in part, at the American Society of Anesthesiologists annual meeting on October 16 (2011), October 16 and 20 (2010), October 20 (2009), and at the International Anesthesia Research Society annual meeting on May 22, 2011, which was the subject of an article in Anesthesiology News. Reprints will not be available from the authors. Address correspondence to David B. Glick, MD, MBA, Department of Anesthesia & Critical Care, University of Chicago, 5841 S Maryland Ave., MC 4028, Chicago, IL 60637. Address e-mail to [email protected]. Copyright © 2014 International Anesthesia Research Society DOI: 10.1213/ANE.0000000000000456
220 www.anesthesia-analgesia.org
By mandating public reporting of quality-related metrics, health care regulators may induce hospitals to compete with each other on quality, leading to overall improvements in care.5 In this way, the quality of health care delivery would improve as higher quality providers are rewarded and lower quality providers either improve or are eliminated from the marketplace. Anesthesiologists are central figures in efforts to meet many perioperative quality metrics. Such metrics, however, may not matter to patients,8 may be only distantly related to relevant health care outcomes,6 and may thus be a low priority for hospitals seeking marketplace success. Third-party payers, patient preferences, government interventions such as limiting specialty care providers by requiring certificates of need, and ethical concerns may also distort relationships between competition and quality in the health care market. To attract patients and physicians, hospitals may choose to compete by offering physical enhancements unrelated to health care, such as a more appealing physical environment, better quality food, or easier parking. Finally, hospitals in less competitive environments may not perceive a strong need to compete for patients since they enjoy a near-monopoly market position. Previous work examining the relationship between a hospital’s competitive environment and the quality of care it delivers has had mixed results.9–11 Those data, however, were mostly compiled before the introduction of publicly reported quality measure performance. Because public reporting is relatively new, the effect on hospital behavior is not yet known. More recent research on relationships between quality and competition has found similarly mixed outcomes in the public reporting era. One 2013 study12 found increased adoption of endovascular technologies for aneurysm repair in more competitive markets January 2015 • Volume 120 • Number 1
but no change in mortality or complications. Another study of quality metrics for heart failure found little effect of competitive environment.13 In a third study,14 hospitals in competitive markets with high health maintenance organization (HMO) penetration reduced costs and improved quality, whereas competition in markets without HMO penetration resulted in increased costs and no change in outcomes. With respect to patient decision making, some evidence suggests that patients may prefer higher quality hospitals15 but that the effect is small.16 Taken together, these data suggest that competitive environments may not lead to improved care quality, even with public reporting. To better understand the relationship between competition and quality in an era of publicly reported quality measures, we used the U.S. Department of Health and Human Services Hospital Compare database to obtain data on quality measure performance, hospital outcomes, and costs for an elective surgical procedure (heart valve surgery [HVS]) and a related emergent medical diagnosis (acute myocardial infarction [AMI]) that, because of its acuity, might allow for less consumer choice.17 The Hospital Compare database is part of the Center for Medicare and Medicaid Services Hospital Quality Initiative and contains approximately 70% of all hospitals providing cardiac care in the United States.18 We correlated outcomes, quality measure performance, and cost data for each hospital with their competitive environment, assessed by the number of hospitals in the same geographical area or by federally defined metropolitan statistical area (MSA). We hypothesized that competition, particularly for an elective procedure where patients are most able to choose among providers, would result in better performance on publicly reported quality measures.
METHODS
This study was reviewed and exempted from formal institutional review by the University of Chicago IRB. We abstracted 2007 and 2008 cost, competition, and quality data for hospitals performing HVS and participating in the publicly available U.S. Department of Health and Human Services Hospital Compare database. We included number of admissions, mortality and morbidity data, location by state and zip code, and hospital performance on quality metrics. To measure the degree of competition faced by each hospital, we first searched the database for all hospitals within 20 miles of the index hospital and screened for those with >10 HVS admissions per year. Hospitals meeting these criteria were then tallied to quantify the degree of competition. As a comparison group, we repeated the above analyses for the related but nonprocedural diagnosis of AMI. To evaluate a hospital’s performance on quality metrics, we used mortality rates, complication scores, and Center for Medicare and Medicaid Services Hospital Quality measures for each of the diagnoses. For 2007 and 2008, the database used these metrics to group hospitals by performance into the top 25%, middle 50%, or lowest 25% for each diagnosis. To obtain cost data, we used the Medicare cost report for each hospital and Diagnosis Related Group. To partially adjust for the effect of geographic variability on Medicare payments unrelated to hospital costs, we used median household income for the zip code of each hospital obtained via the 2000 U.S. census.
January 2015 • Volume 120 • Number 1
STATISTICS
Zero-order and partial Spearman rank correlation coefficients were used to initially assess the association among number of patients (volume), competition, cost, and median household income. Hospitals were also separated into quartiles based on their level of competition (quartile #1 = fewest number of competitors) and number of patients treated with each diagnosis (quartile #1 = lowest number of patients). Quality data were represented as top 25%, middle 50%, and bottom 25% for each measure in the database and also as a dichotomous variable (top 25% versus not) for statistical analysis. The effect of competition, volume, cost, and median household income on quality measure performance was determined using logistic regression with competition, volume, and cost quartiles as the primary independent variables, household income as a continuous covariate, and position in the top 25% of performers for each quality measure as the dependent variable. Competition, volume, and cost quartiles were used because of the skewness of the data and for ease of interpretation. The effect of competition on mortality and cost was then evaluated using logistic regression with competition quartile as the independent variable and mortality or cost above the median as the dependent variable. The resulting odds ratios and 95% confidence intervals (CIs) were reported. To assess the effect of variability due to low-volume hospitals, we repeated the above analyses at volume thresholds of 50 and 100 cases per year (instead of 10/year). Because we chose an arbitrary 20-mile radius to assess competitive environment, we also repeated the above analyses using each hospital’s MSA instead of geographical radius. Correlation between MSA and geographical definitions of competition was strong (e.g., r = 0.80 for HVS and 0.73 for AMI, P < 0.001). For all comparisons, two-sided P values 10 HVS and 1898 hospitals caring for >10 cases of AMI. Of the 24 states included in the database, California had the highest percentage of hospitals (16.7% for HVS), and Vermont the lowest (0.15%). The distribution of hospitals by state and procedure volume is shown in Table 1. For hospitals performing HVS, competition did not correlate with quality measure performance. Performance on each of the 6 publicly reported measures listed in Table 2 was not significantly different across competitive quartiles (Fig. 1). Raising our inclusion threshold to hospitals performing >50 and >100 cases and repeating our analysis produced similar results (Appendix Tables 1 and 2). Top 25% performance in any single measure or combination of measures did not correlate significantly with mortality. Logistic regression analysis with competition and cost quartiles as independent variables and mortality (above/below the median) as the dependent variable also found no significant relationship between competition and mortality (2nd Q odds ratio [OR] = 0.78 [95% CI, 0.51–1.21], 3rd Q OR = 0.63 [95% CI, 0.41–0.98], 4th Q OR = 0.68 [95% CI, 0.43–1.07]). We found that increased competition, defined using geographical area, correlated with higher average Medicare costs (zero-order Spearman rank correlations are presented
www.anesthesia-analgesia.org
221
Competition and Quality in Procedural Health Care
Table 1. Distribution of Hospitals by State for Each Diagnosis State Arizona California Colorado Florida Iowa Illinois Massachusetts Maryland Maine North Carolina New Hampshire New Jersey New Mexico Nevada New York Oregon Pennsylvania Rhode Island Texas Utah Virginia Vermont Washington Wisconsin
Heart valve surgery N (% of total) 25 (3.8) 109 (16.7) 17 (2.6) 72 (11.0) 12 (1.8) 58 (8.9) 15 (2.3) 9 (1.4) 3 (0.5) 21 (3.2) 4 (0.6) 18 (2.8) 5 (0.8) 10 (1.5) 40 (6.1) 12 (1.8) 60 (9.2) 3 (0.5) 86 (13.2) 11 (1.7) 18 (2.8) 1 (0.2) 17 (2.6) 27 (4.1)
Acute myocardial infarction care N (% of total) 52 (2.7) 263 (13.9) 37 (1.9) 173 (9.1) 54 (2.9) 131 (6.9) 64 (3.4) 43 (2.3) 33 (1.7) 83 (4.4) 21 (1.1) 71 (3.7) 20 (1.0) 21 (1.1) 177 (9.3) 45 (2.4) 143 (7.5) 10 (0.5) 221 (11.6) 16 (0.8) 74 (3.9) 11 (0.6) 54 (2.8) 81 (4.3)
Table 2. Quality Metrics for Heart Valve Surgery Antibiotic given within 1 h of incision Right kind of antibiotic Antibiotic discontinued within 24 h of surgery Deep venous thrombosis prophylaxis ordered Deep venous thrombosis prophylaxis received Glucose control after cardiac surgery
in Table 3). Multivariate analyses also demonstrated positive correlations between competition and cost (partial correlation = 0.40 adjusting for volume and income, P < 0.001). Furthermore, logistic regression analysis showed an incrementally positive effect of competition on the odds of having costs above the median (Table 4). When we repeated the above analyses using MSA instead of geographical radius, we found similar relationships between competition and cost (partial correlation = 0.42) and incrementally positive effects of competition on the odds of the costs being greater than the median cost. As a comparison group, we examined hospitals treating AMI, an emergent diagnosis related to HVS, but with fewer opportunities for consumer choice because of its acuity. Consistent with HVS, performance on the publicly reported quality measures for AMI listed in Table 5 did not differ depending on competitive environment as defined by geographical radius (Fig. 2). As with HVS, repeating the above analyses with volume thresholds of 50 or 100 cases per year produced similar results (Appendix Tables 3 and 4). When hospitals were grouped by top 25% versus bottom 75% quality measure performance, no effect of group on mortality was noted. For AMI, competition correlated inversely, albeit weakly, with mortality (partial correlation = −0.08, P = 0.002, N = 1628). As with HVS, multivariate analysis found a positive correlation between competition in AMI care and costs (partial correlation = 0.52 adjusting for volume and income, P
