Graph depicts the quartiles of safety-net burden. HBH indicates high-burden hospital; LBH, low-burden hospital; and MBH, medium-burden hospital.
Odds ratios (ORs) (diamond markers) for mortality and readmission in high-burden hospitals and risk ratios (RRs) (diamond markers) for cost after adjusting for patient factors and hospital volume are given with 95% CIs. CABG indicates coronary artery bypass graft.
eTable. Results From Multivariate Models for Cost, Mortality, and Readmissions After Adjusting for Patient Factors and Center Volume
Richard S. Hoehn, Koffi Wima, Matthew A. Vestal, Drew J. Weilage, Dennis J. Hanseman, Daniel E. Abbott, Shimul A. Shah. Effect of Hospital Safety-Net Burden on Cost and Outcomes After Surgery. JAMA Surg. 2016;151(2):120–128. doi:10.1001/jamasurg.2015.3209
Safety-net hospitals provide broad services for a vulnerable population of patients and are financially at risk owing to impending reimbursement penalties and policy changes.
To determine the effect of patient and hospital factors on surgical outcomes and cost at safety-net hospitals.
Design, Setting, and Participants
Hospitals in the University HealthSystem Consortium database from January 1, 2009, through December 31, 2012 (n = 231), were grouped according to their safety-net burden, defined as the proportion of Medicaid and uninsured patient charges for all hospitalizations during that time (n = 12 638 166). Nine cohorts, based on a variety of surgical procedures, were created and examined with regard to preoperative characteristics, postoperative outcomes, and resource utilization. Multiple logistic regression was performed to analyze the effect of patient and center factors on outcomes. Hospital Compare data from the Centers for Medicare & Medicaid Services were linked and used to characterize and compare the groups of hospitals.
Main Outcomes and Measures
Postoperative mortality, 30-day readmissions, and total direct cost.
For all 9 procedures examined in 231 hospitals comprising 12 638 166 patient encounters, patients at hospitals with high safety-net burden (HBHs) (vs hospitals with low and medium safety-net burdens) were most likely to be young, to be black, to be of the lowest socioeconomic status, and to have the highest severity of illness and the highest cost for surgical care (P < .01 for all). For 7 of 9 procedures, HBHs had the highest proportion of emergent cases and longest length of stay (P < .01 for all). After adjusting for patient characteristics and center volume, HBHs still had higher odds of mortality for 3 procedures (odds ratios [ORs], 1.81-2.08; P < .05), readmission for 2 procedures (ORs, 1.19-1.30; P < .05), and the highest cost of care associated with 7 of 9 procedures (risk ratios, 1.23-1.35; P < .05). Analysis of Hospital Compare data found that HBHs had inferior performance on Surgical Care Improvement Project measures, higher rates of surgical complications, and inferior markers of emergency department timeliness and efficiency (all P < .05).
Conclusions and Relevance
These data suggest that intrinsic qualities of safety-net hospitals lead to inferior surgical outcomes and increased cost across 9 elective surgical procedures. These outcomes are likely owing to hospital resources and not necessarily patient factors. In addition, impending changes to reimbursement may have a negative effect on the surgical care at these centers.
Quiz Ref IDSafety-net hospitals are defined by the Institute of Medicine as (1) having a legal mandate or an adopted mission to maintain an open-door policy for all patients, regardless of their ability to pay, or (2) having a substantial share of their patient mix consist of “uninsured, Medicaid, and other vulnerable patients.”1 A review of the quality of surgical care at safety-net hospitals found that they tended to have worse performance with regard to timeliness, patient centeredness, and equity of treatment.2 In some cases, these hospitals also had higher rates of mortality, hospital-acquired conditions, and other complications. The Centers for Medicare & Medicaid Services claims that variation in readmissions for patients of low socioeconomic status (SES) is owing to the quality of care they receive and that hospitals with more poor patients should not be held to lower standards.3 As such, hospital reimbursement penalties are being applied universally in an effort to curtail the costs and improve the quality of US health care.
The Patient Protection and Affordable Care Act (PPACA) brought hope and uncertainty to safety-net hospitals and health care providers seeking to continue to serve the needs of their vulnerable populations. The law proposed to offset financial losses at these institutions by mandating health insurance coverage for all patients and expanding Medicaid coverage for patients with low SES across the country. However, the PPACA also calls for a $30 billion to $50 billion reduction in disproportionate-share hospital funding in the coming years.4,5 The Medicaid reimbursement is already lower than the cost of care5; moreover, half of the US states have resisted Medicaid expansion.4 As a result, safety-net hospitals that depend on disproportionate-share hospital funding may suffer financially in the coming years, exacerbating the current disparities in US health care.6
In light of these policy changes, we must understand the effect of patient factors and hospital payer mix on the cost of surgical care. Literature on this topic is sparse and the findings are conflicting.7,8 The aims of the present study were to analyze the effect of the safety-net burden on cost and short-term outcomes in a population undergoing surgery and to investigate the effect of patient and hospital factors on important quality-associated metrics. We hypothesized that safety-net hospitals would have worse outcomes and greater costs of care and that these differences would likely correlate with their patient populations having advanced illnesses and limited resources.
We used 2 data sources for our analyses. First, we assessed the University HealthSystem Consortium (UHC) Clinical Database/Resource Manager. The UHC is an alliance of 95% of the nation’s major not-for-profit academic medical centers, which include 118 academic medical centers and 298 of their affiliated hospitals. The Clinical Database/Resource Manager is an administrative data set that contains patient demographic, financial, procedural, and short-term outcome data and International Classification of Diseases, Ninth Revision (ICD-9) diagnostic codes provided by member medical centers. Second, we accessed the publicly available Hospital Compare data from the Centers for Medicare & Medicaid Services website, including Surgical Care Improvement Project (SCIP) measures, surgical complications, and emergency department timeliness and efficiency (or throughput). These data were linked to UHC hospitals for analysis of hospital groups. The institutional review board of the University of Cincinnati approved this study. Patient data were deidentified.
To study the effect of safety-net burden on surgical outcomes, a safety-net burden was assigned to each hospital (n = 231) as previously described.7,9 All patient encounters (n = 12 638 166) from January 1, 2009, through December 31, 2012, were included in the analysis. Data were collected from March through June 2014. The safety-net burden was defined as the proportion of all charges for patients with Medicaid coverage or no insurance. Hospitals were then stratified into groups of safety-net burden,9 as shown in Figure 1. Low-burden hospitals (LBHs) were in the lowest quartile (1.9%-14.2%); medium-burden hospitals (MBHs), in the middle 2 quartiles (14.5%-36.4%); and high-burden hospitals (HBHs), in the highest quartile (36.8%-100%).
A variety of orthopedic, thoracic, and abdominal surgical procedures were chosen for the analysis, representing a spectrum of disease sites and severity (Table 1). All cases during the study period were included, as identified by ICD-9 procedure codes. Case numbers per procedure ranged from 8046 for esophagectomy to 194 952 for hip replacement. The following patient encounter information was collected from the UHC: age, sex, race (white, black, Hispanic, Asian, or other), insurance type (private, government, or other), admission status (elective or emergent), overall length of stay (in days), length of stay in the intensive care unit (in days), in-hospital mortality rates, total direct cost, discharge disposition (home, rehabilitation, or other), and 30-day readmission rates. Patient SES quintiles were created based on US census data as previously described (1 indicates the lowest quintile; 5, the highest quintile).10,11 Patient severity of illness in the UHC (minor, moderate, major, or extreme) is derived from a proprietary formula that is based on all payer refined diagnosis related groups and has been validated in a nationwide data set that included 8.5 million discharges from more than 1000 hospitals.12
For this analysis, primary outcomes of interest were postoperative mortality, 30-day readmissions, and the cost of surgical care. Total direct costs are calculated from hospital charges using institution-specific Medicare cost-to-charge ratios and then adjusted with federally reported area wage indices to normalize regional variation in labor cost.13,14 We calculated the annual hospital procedure-specific volume and stratified hospitals into tertiles.15
Data were analyzed from March 2014 through February 2015. For univariate analysis, we used χ2 tests to compare categorical data and Wilcoxon rank sum tests to compare continuous variables. To assess the effect of patient and hospital factors on outcomes, several multivariate models were created. First, crude odds ratios (ORs) (for mortality and readmission) and risk ratios (for cost) were calculated for patients at HBHs, with LBHs constituting the reference group. Next, these analyses included adjustment for patient age, race, SES, and severity of illness followed by the inclusion of hospital volume. Final analysis adjusted for patient factors and center volume. We used a random-effects model to adjust for patient clustering within hospitals. Owing to data convergence issues, severity of illness was omitted from the mortality analysis for esophagectomy, kidney transplant, and pancreaticoduodenectomy.
Hospital groups (HBHs and LBHs) were then compared with regard to Medicare Hospital Compare metrics.16 We chose the following 3 categories to assess safety and systems at the hospitals: SCIP measures because they are a focus of surgical quality improvement and value-based purchasing, surgical complications as a marker of safe and quality care, and emergency department throughput to assess hospital staffing and systems efficiency.
Statistical analysis was performed using SAS (version 9.4; SAS Institute Inc) and SigmaPlot (Systat) software. An α level of .05 was used for all significance tests.
Demographic information and postoperative outcomes for the patient cohorts are shown in Table 2. For all 9 procedures examined, patients at HBHs were the youngest and the most likely to be black, were in the lowest SES quintile, and had the highest severity of illness score and the highest cost of care (P < .001 for all). For 8 of the procedures, HBHs had the highest rates of postoperative mortality and 30-day readmissions (P < .05 for all). For 7 of the 9 procedures, HBHs had the highest proportion of emergent cases and longest length of stay (P < .01 for all).
Next, we investigated the influence of patient and hospital factors on adverse postoperative outcomes, specifically mortality, readmissions, and cost. On unadjusted analysis, HBHs had higher odds of mortality for 6 procedures, readmissions for 4 procedures, and cost for 7 procedures (eTable in the Supplement) (P < .05 for all). Quiz Ref IDAfter adjusting for patient factors (age, race, SES, and severity of illness), HBHs still had inferior perioperative outcomes. Postoperative mortality was worse for colectomy (OR, 1.72; 95% CI, 1.39-2.14), esophagectomy (OR, 2.15; 95% CI, 1.30-3.56), pancreaticoduodenectomy (OR, 2.91; 95% CI, 1.93-4.39), and ventral hernia repair (OR, 2.11; 95% CI, 1.61-2.75) (P < .05 for all). Odds of readmission were higher at HBHs for coronary artery bypass graft (OR, 1.18; 95% CI, 1.08-1.29), colectomy (OR, 1.73; 95% CI, 1.39-2.14), kidney transplant (OR, 1.28; 95% CI, 1.01-1.63), and ventral hernia repair (OR, 1.29; 95% CI, 1.16-1.43) (P < .05 for all). Costs were higher for 7 procedures, with increases ranging from 23% to 33% (P < .05 for all).
After adjusting for center procedure-specific volume, HBHs had higher rates of mortality and readmissions and higher costs for 5, 2, and 6 of the procedures examined (eTable in the Supplement) (P < .05 for all). Finally, when we adjusted for patient factors and center volume, HBHs still had a significantly increased likelihood of mortality for colectomy, pancreaticoduodenectomy, and ventral hernia repair (Figure 2A and B; P < .05) and increased readmissions for colectomy and ventral hernia repair (Figure 2B; P < .05). Most notably, HBHs still had greater than 20% increased cost for 7 of the 9 procedures examined (Figure 2C; P < .05). Cost and outcomes were never better at HBHs in any of the models.
Quiz Ref IDAnother way of comparing HBHs and LBHs used 3 categories from the Hospital Compare database (Table 3). The HBHs were significantly worse than the LBHs in 5 of the 10 SCIP measures, including receipt of prophylactic antibiotics 1 hour before surgery, appropriate venous thromboembolism prophylaxis, perioperative β-blocker use, discontinuation of prophylactic antibiotic therapy, and urinary catheter removal (P < .05 for all). The HBHs also had significantly higher rates in 4 of the 7 surgical complications recorded in the Hospital Compare database, including serious complications, deaths due to serious treatable complications after surgery, iatrogenic pneumothorax, and accidental cuts and tears (P ≤ .01 for all). Quiz Ref IDFinally, when assessing markers of emergency department throughput, HBHs were inferior to LBHs in all measures, including time from arrival to evaluation, admission decision time, times from arrival to departure for discharged and admitted patients, time for pain medicine administration to patients with long-bone fractures, and patients who left without being seen (P ≤ .002 for all). The HBHs were not better than LBHs in any of the Hospital Compare categories analyzed.
Health care payment and policy in the United States is undergoing significant reform, and how these changes affect the care of vulnerable populations at safety-net hospitals is of great concern. Quiz Ref IDOur analysis found, as expected, that patients at safety-net hospitals tend to have fewer resources and present with greater severity of illness. These hospitals have higher mortality and readmission rates and higher cost associated with surgical care. However, these inferior outcomes persisted after adjusting for patient characteristics and hospital procedure volume, suggesting that intrinsic qualities of safety-net hospitals lead to surgical care that is inferior and more expensive.
Much of the available literature suggests that patient characteristics, including SES and social support, influence surgical outcomes, such as readmission and mortality.11,17- 20 A study from an urban health system,21 where care presumably was standardized across patients, found that patients who were unmarried or of low SES were more likely to be readmitted. However, conflicting studies have shown hospital and community characteristics to be independently associated with decreased quality of care.22- 28 We found safety-net hospitals to have worse mortality and readmission rates, and these differences persisted after adjusting for patient age, race, SES, and severity of illness in addition to center volume. This finding contrasted with our hypothesis that patient factors drive outcomes at these hospitals and forced us to investigate center factors that may play a role.
Several studies29- 32 have found that poor hospital financial health may lead to increased mortality, hospital-acquired conditions, and other adverse events. We found that safety-net hospitals performed worse on most SCIP measures and had higher rates of most surgical complications recorded in the Hospital Compare database. When we analyzed measures of emergency department throughput, HBHs performed worse in all categories with adequate reporting. These latter measures primarily reflect the time a patient waits in the emergency department before being seen, treated, or discharged. As such, these measures are key markers of adequate staffing and systems efficiency. Hospitals with inadequate nurse or physician numbers to meet their patient volume or with inefficient systems will take longer to see patients and deliver care. Indeed, HBHs take longer to evaluate and treat patients in their emergency departments, which may very well correlate with systemic deficiencies in staffing or organization.
Safety-net hospitals are at a disadvantage in today’s evolving health care marketplace. Competitive forces increasingly necessitate that hospitals and providers become “focused factories” that specialize in certain services and jettison other programs in an effort to become more cost-effective.33,34 In times of financial stress, hospitals do not always reduce staffing but often cut back on less profitable services (ie, substance abuse programs) that are often important to the uninsured.35,36 Safety-net hospitals must care for all patients and are unable to streamline their services to maximize profit. As a result, these hospitals have a limited ability to compete in the modern market and are less able to adapt in times of financial stress without adversely affecting their patients.
This combination of factors is of concern for safety-net hospitals in light of changing reimbursement policy. Safety-net hospitals care for vulnerable patients and receive decreased reimbursement compared with hospitals with a lower safety-net burden.5 Subsequently, they are more financially limited, which has been shown to have an adverse effect on patient outcomes and national hospital quality metrics.2 One study37 found that safety-net hospitals had smaller gains in quality performance over time and were more likely to incur financial penalties. Safety-net hospitals, which account for 15% of US hospitals, have been estimated to incur 38% of financial penalties as a result of the Hospital Readmissions Reduction Program.5 Coupled with the billions of dollars in disproportionate-share hospital funding cuts,4 safety-net hospitals could find themselves in a downward spiral of reduced funding leading to worse outcomes, followed by reimbursement penalties, that will further negatively affect patient care.6
If the objective of US health care reform is increased quality and decreased cost of care, universally applied reimbursement penalties may not be the ideal technique. We have shown that safety-net hospitals provide more expensive surgical care, potentially as a result of inefficient systems and staffing. Existing literature suggests that limited resources prevent hospitals from investing in care-improvement initiatives38,39 and that financial penalties for these providers may not induce the desired effect of improving outcomes and streamlining care but will likely do the opposite.27 These penalties may also exacerbate existing racial disparities in care.40 Perhaps a better solution would be to invest in these hospitals to improve quality and efficiency.41 Increases in nursing staff number and skill mix have been shown to lead to improved quality and reduced length of stay with a minimal effect on cost of care.42 Other options may include risk-adjusted policies or incentives not only for overall quality but also for improvement over time.43
This study has several limitations. First, it is retrospective and observational and cannot determine causation. Second, our data set is limited only to academic hospitals. Although a large population of hospitals was missed, we believe that the study findings were not limited. A large proportion of safety-net hospitals are urban academic centers; thus, our conclusions might be extrapolated to safety-net hospitals nationwide.44 In addition, academic medical centers face economic challenges similar to those of safety-net hospitals because they provide a broad range of expensive care rather than select profitable services. Finally, we were limited in evaluating only the patient and hospital characteristics available to us. Although we were able to analyze validated measures of SES and severity of illness,10,11,19 we lacked more granular patient-level details, such as marital status and medical comorbidities. Also, we could not compare hospital characteristics other than safety-net burden, procedure-specific volume, and information from the Hospital Compare database. We used a random-effects model to adjust for patient clustering, but some variables that affect the exposure-outcome relationship may remain unrecognized.45
We have found that the safety-net burden among academic hospitals is associated with surgical care that is more expensive yet has inferior outcomes. This relationship persisted after adjusting for patient characteristics and hospital volume. Among these differences, the increased cost of surgical care at safety-net hospitals was the most striking. Hospital Compare data from the Centers for Medicare & Medicaid Services website suggest that safety-net hospitals provide less efficient care that may be a result of inferior staffing or disorganized systems. Unadjusted reimbursement penalties based on these performance measures may exacerbate disparities in care. Whether the goal is to reduce health care expenditures or improve quality of care, special attention needs to be devoted to safety-net hospitals that are in a unique financial position and care for a vulnerable patient population.
Corresponding Author: Shimul A. Shah, MD, MHCM, Division of Transplant Surgery, Department of Surgery, University of Cincinnati School of Medicine, 231 Albert Sabin Way, ML 0558, Room MSB2006C, Cincinnati, OH 45267-0558 (email@example.com).
Accepted for Publication: June 24, 2015.
Published Online: October 14, 2015. doi:10.1001/jamasurg.2015.3209.
Author Contributions: Dr Shah had full access to all 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: Hoehn, Wima, Weilage, Abbott, Shah.
Acquisition, analysis, or interpretation of data: Hoehn, Wima, Vestal, Hanseman, Shah.
Drafting of the manuscript: Hoehn, Wima, Vestal, Weilage, Shah.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Hoehn, Wima, Hanseman.
Obtained funding: Shah.
Administrative, technical, or material support: Vestal, Weilage, Abbott, Shah.
Study supervision: Abbott, Shah.
Conflict of Interest Disclosures: None reported.
Funding/Support: This study was supported by the Department of Surgery, University of Cincinnati College of Medicine.
Role of the Funder/Sponsor: The funding source provided infrastructure and financial support but 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.