Comparison of Risk-Standardized Readmission Rates of Surgical Patients at Safety-Net and Non–Safety-Net Hospitals Using Agency for Healthcare Research and Quality and American Hospital Association Data | Health Disparities | JAMA Surgery | JAMA Network
[Skip to Navigation]
Table 1.  Hospital Characteristics
Hospital Characteristics
Table 2.  Mean RSRRs by State
Mean RSRRs by State
Table 3.  Multivariable Linear Regression Model to Predict RSRRs
Multivariable Linear Regression Model to Predict RSRRs
1.
Al-Amin  M.  Hospital characteristics and 30-day all-cause readmission rates.  J Hosp Med. 2016;11(10):682-687. doi:10.1002/jhm.2606PubMedGoogle ScholarCrossref
2.
Borza  T, Oreline  MK, Skolarus  TA,  et al.  Association of the hospital readmissions reduction program with surgical readmissions.  JAMA Surg. 2018;153(3):243-250. doi:10.1001/jamasurg.2017.4585PubMedGoogle ScholarCrossref
3.
Lewen  ME, Altman  S; Committee on the Changing Market, Managed Care, and the Future Viability of Safety-Net Providers.  America’s Health Care Safety-Net: Intact but Endangered. Washington, DC: National Academy Press; 2000.
4.
Ayanian  JZ, Weissman  JS, Schneider  EC, Ginsburg  JA, Zaslavsky  AM.  Unmet health needs of uninsured adults in the United States.  JAMA. 2000;284(16):2061-2069. doi:10.1001/jama.284.16.2061PubMedGoogle ScholarCrossref
5.
Eisert  SL, Durfee  MJ, Welsh  A, Moore  SL, Mehler  PS, Gabow  PA.  Changes in insurance status and access to care in an integrated safety net healthcare system.  J Community Health. 2009;34(2):122-128. doi:10.1007/s10900-008-9136-2PubMedGoogle ScholarCrossref
6.
Coleman  DL.  The impact of the lack of health insurance: how should academic medical centers and medical schools respond?  Acad Med. 2006;81(8):728-731. doi:10.1097/00001888-200608000-00009PubMedGoogle ScholarCrossref
7.
Arbaje  AI, Wolff  JL, Yu  Q, Powe  NR, Anderson  GF, Boult  C.  Postdischarge environmental and socioeconomic factors and the likelihood of early hospital readmission among community-dwelling Medicare beneficiaries.  Gerontologist. 2008;48(4):495-504. doi:10.1093/geront/48.4.495PubMedGoogle ScholarCrossref
8.
Bernheim  SM.  Measuring quality and enacting policy: readmission rates and socioeconomic factors.  Circ Cardiovasc Qual Outcomes. 2014;7(3):350-352. doi:10.1161/CIRCOUTCOMES.114.001037PubMedGoogle ScholarCrossref
9.
Eslami  MH, Rybin  D, Doros  G, Farber  A.  Care of patients undergoing vascular surgery at safety net public hospitals is associated with higher cost but similar mortality to nonsafety net hospitals.  J Vasc Surg. 2014;60(6):1627-1634. doi:10.1016/j.jvs.2014.08.055PubMedGoogle ScholarCrossref
10.
Hoehn  RS, Wima  K, Vestal  MA,  et al.  Effect of hospital safety-net burden on cost and outcomes after surgery.  JAMA Surg. 2016;151(2):120-128. doi:10.1001/jamasurg.2015.3209PubMedGoogle ScholarCrossref
11.
Horwitz  LI, Bernheim  SM, Ross  JS,  et al.  Hospital characteristics associated with risk-standardized readmission rates.  Med Care. 2017;55(5):528-534. doi:10.1097/MLR.0000000000000713PubMedGoogle ScholarCrossref
12.
Bazzoli  GJ, Chen  HF, Zhao  M, Lindrooth  RC.  Hospital financial condition and the quality of patient care.  Health Econ. 2008;17(8):977-995. doi:10.1002/hec.1311PubMedGoogle ScholarCrossref
13.
Joynt  KE, Jha  AK.  Who has higher readmission rates for heart failure, and why? implications for efforts to improve care using financial incentives.  Circ Cardiovasc Qual Outcomes. 2011;4(1):53-59. doi:10.1161/CIRCOUTCOMES.110.950964PubMedGoogle ScholarCrossref
14.
Healthcare Cost and Utilization Project. SID data elements. https://www.hcup-us.ahrq.gov/sidoverview.jsp. Modified July 24, 2018. Accessed May 7, 2018.
15.
American Hospital Association. Data collection methods. http://www.ahadata.com/. January 1, 2018. Accessed May 7, 2018.
16.
Horwitz  L, Partovian  C, Lin  Z,  et al. Centers for Medicare and Medicaid Services. Hospital-wide all-cause unplanned readmission measure: final technical report. https://www.cms.gov/Medicare/Quality-Initiatives-Patient-Assessment-Instruments/HospitalQualityInits/Downloads/Hospital-Wide-All-Cause-Readmission.zip. July 1, 2012. Accessed May 7, 2018.
17.
Hoyer  EH, Padula  WV, Brotman  DJ,  et al.  Patterns of hospital performance on the hospital-wide 30-day readmission metric: is the playing field level?  J Gen Intern Med. 2018;33(1):57-64. doi:10.1007/s11606-017-4193-9PubMedGoogle ScholarCrossref
18.
Rattan  R, Parreco  J, Lindenmaier  LB,  et al.  Underestimation of unplanned readmission after colorectal surgery: a national analysis.  J Am Coll Surg. 2018;226(4):382-390. doi:10.1016/j.jamcollsurg.2017.12.012PubMedGoogle ScholarCrossref
19.
Kane  RL, Shamliyan  TA, Mueller  C, Duval  S, Wilt  TJ.  The association of registered nurse staffing levels and patient outcomes: systematic review and meta-analysis.  Med Care. 2007;45(12):1195-1204. doi:10.1097/MLR.0b013e3181468ca3PubMedGoogle ScholarCrossref
20.
Blegen  MA, Goode  CJ, Spetz  J, Vaughn  T, Park  SH.  Nurse staffing effects on patient outcomes: safety-net and non–safety-net hospitals.  Med Care. 2011;49(4):406-414. doi:10.1097/MLR.0b013e318202e129PubMedGoogle ScholarCrossref
21.
Horwitz  LI, Wang  Y, Altaf  FK,  et al.  Hospital characteristics associated with postdischarge hospital readmission, observation, and emergency department utilization.  Med Care. 2018;56(4):281-289.PubMedGoogle Scholar
22.
Wakeam  E, Hevelone  ND, Maine  R,  et al.  Failure to rescue in safety-net hospitals: availability of hospital resources and differences in performance.  JAMA Surg. 2014;149(3):229-235. doi:10.1001/jamasurg.2013.3566PubMedGoogle ScholarCrossref
Limit 200 characters
Limit 25 characters
Conflicts of Interest Disclosure

Identify all potential conflicts of interest that might be relevant to your comment.

Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.

Err on the side of full disclosure.

If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.

Not all submitted comments are published. Please see our commenting policy for details.

Limit 140 characters
Limit 3600 characters or approximately 600 words
    Original Investigation
    January 16, 2019

    Comparison of Risk-Standardized Readmission Rates of Surgical Patients at Safety-Net and Non–Safety-Net Hospitals Using Agency for Healthcare Research and Quality and American Hospital Association Data

    Author Affiliations
    • 1Department of Surgery, Boston Medical Center, Boston, Massachusetts
    • 2Center for Healthcare Organization and Implementation Research, Veterans Affairs Boston Healthcare System, Boston, Massachusetts
    • 3Biostatistics and Epidemiology Data Analytics Center, Boston University School of Public Health, Boston, Massachusetts
    JAMA Surg. 2019;154(5):391-400. doi:10.1001/jamasurg.2018.5242
    Key Points

    Question  After adjusting for hospital characteristics, is there a difference in surgical readmission rates between safety-net and non–safety-net hospitals?

    Findings  This study of 1 252 505 patients with surgical admissions linked data from the Agency for Healthcare Research and Quality’s 2011-2014 State Inpatient Databases for 4 geographically varied states with data from the 2014 American Hospital Association annual survey; hospital risk-standardized readmission rates were calculated. After controlling for hospital characteristics and patient factors, safety-net hospitals had slightly higher readmission rates after surgery.

    Meaning  Surgical patients appear to have experienced higher readmission rates at safety-net hospitals.

    Abstract

    Importance  Medical patients discharged from safety-net hospitals (SNHs) experience higher readmission rates compared with those discharged from non-SNHs. However, little is known about whether this association persists for surgical patients.

    Objectives  To examine differences in readmission rates between SNHs and non-SNHs among surgical patients after discharge and determine whether hospital characteristics might account for some of the variation.

    Design, Setting, and Participants  This observational retrospective study linked the Healthcare Cost and Utilization Project State Inpatient Databases of the Agency for Healthcare Research and Quality from January 1, 2011, through December 31, 2014, for 4 states (New York, Florida, Iowa, and Washington) with data from the 2014 American Hospital Association annual survey. After identifying surgical discharges, SNHs were defined as those with the top quartile of inpatient stays paid by Medicaid or self-paid. Hospital-level risk-standardized readmission rates (RSRRs) for surgical discharges were calculated. The association between hospital RSRRs and hospital characteristics was evaluated with bivariate analyses. An estimated multivariable hierarchical linear regression model was used to examine variation in hospital RSRRs, adjusting for hospital characteristics, state, year, and SNH status. Data were analyzed from June 1, 2017, through March 1, 2018.

    Exposures  Surgical care at an SNH.

    Main Outcomes and Measures  Readmission after an index surgical admission.

    Results  A total of 1 252 505 patients across all 4 years and states were included in the analysis (51.7% women; mean [SD] age, 52.7 [18.1] years). Bivariate analyses found that SNHs had higher mean (SD) surgical RSRRs compared with non-SNHs; significant differences were found for New York (9.6 [0.1] vs 10.9 [0.1]; P < .001) and Florida (11.6 [0.1] vs 12.1 [0.1]; P = .001). The SNHs also had higher RSRRs in these 2 states when stratified by hospital funding (nonfederal government SNHs in New York, 11.9 [0.2]; for-profit, private SNHs in Florida, 13.1 [0.2]; P < .001 for both); however, bed size was a significant factor for higher mean (SD) RSRRs only for New York (200 to 399 beds, 12.0 [0.4]; P = .006). Similar results were found for multivariable linear regression models; RSRRs were 1.02% higher for SNHs compared with non-SNHs (95% CI, 0.75%-1.29%; P < .001). Increased RSRRs were observed for hospitals in New York and Florida, teaching hospitals, and investor-owned hospitals. Factors associated with reduced RSRRs included presence of an ambulatory surgery center, cardiac catheterization capabilities, and high surgical volume.

    Conclusions and Relevance  According to results of this study, surgical patients treated at SNHs experienced slightly higher RSRRs compared with those treated at non-SNHs. This association persisted after adjusting for year, state, and hospital factors, including teaching status, hospital bed size, and hospital volume.

    Introduction

    Hospital readmissions are associated with worse patient outcomes and higher health care costs, with estimates as high as $17 billion per year. In 2012, the Centers for Medicare & Medicaid Services (CMS) implemented the Hospital Readmission Reduction Program (HRRP) to decrease preventable readmissions for specific medical conditions.1,2 In 2015, the HRRP was expanded to include elective total knee and hip arthroplasty. In addition, in 2014, CMS developed a 30-day all-cause hospital-wide readmission (HWR) measure for the purpose of public reporting. Application of the HWR measure may result in greater consequences for hospitals beyond those with diagnoses targeted by the current policies.1

    Comparison of hospital performance using such measures is particularly problematic for safety-net hospitals (SNHs). Safety-net hospitals are characterized by a mission to offer patients access to services, regardless of their ability to pay, and largely represent Medicaid recipients and uninsured or underinsured patients.3 Patients who are uninsured are more likely to delay seeking care and forgo necessary care and screening, resulting in complications secondary to chronic illnesses and worse clinical outcomes.4,5 The care required to treat these patients is costly for hospitals and for Medicaid and Medicare.4,6

    Safety-net hospitals are likely to have higher readmission rates compared with non-SNHs and receive worse financial penalties given their larger proportion of poor, uninsured, and vulnerable patients,7-9 particularly for medical patients.2 Although SNHs account for 15% of US hospitals, estimates show that they may incur as much as 38% of financial penalties as a result of the HRRP.10 Safety-net hospitals could also face additional financial cuts owing to decreases in disproportionate-share hospital funding.10,11

    Most of the literature on readmissions focuses on medical patients; little is known as to whether differences in readmission rates between SNHs and non-SNHs persist among surgical patients.2,4,10-13 Although 1 study2 found a decrease in readmissions for the targeted joint arthroplasties after HRRP implementation, the investigators did not evaluate the role of SNH status in readmissions or assess other operations besides those related to orthopedic joint operations. Given the gap in the literature, we sought to examine differences between SNHs and non-SNHs among surgical discharges and determine whether hospital characteristics might account for some of the variation in readmission rates. We hypothesized that no differences in readmission rates between SNHs and non-SNHs would occur after adjusting for other hospital characteristics.

    Methods

    We conducted a retrospective observational study to examine the association between hospital characteristics and postsurgical all-cause readmission rates in 4 states (New York, Florida, Iowa, and Washington) from January 1, 2011, through December 31, 2014. Analyses were conducted within each state by year and then aggregated across years when linking was possible. These 4 states were chosen based on geographic variability and their ability to link admissions across years (eTables 1 and 2 in the Supplement). This study was approved by the institutional review board of Boston University, Boston, Massachusetts, which waived the need for informed consent for use of deidentified data.

    Data Sources

    We obtained the Agency for Healthcare Research and Quality (AHRQ) Healthcare Cost and Utilization Project State Inpatient Databases (SIDs) for New York, Florida, Iowa, and Washington. The SIDs capture 100% of nonfederal discharges in a state in a given year and contain unique patient identifiers that allow the linking of index admissions to readmissions. The SIDs provide patients’ sociodemographic characteristics, diagnoses, procedures, comorbidities, and clinical categories for diagnoses and procedures based on the AHRQ Clinical Classification Software (CCS).14 We obtained hospital characteristics from the 2014 American Hospital Association annual survey database.14

    Hospital Characteristics

    Similar to the AHRQ’s definition,15 we defined SNHs as those hospitals with the highest (top quartile) number of inpatient stays that were self-paid or paid by Medicaid. Hospital characteristics were classified into the following categories: teaching status (yes or no), ownership (nonfederal government, not-for-profit private, or for-profit private), metropolitan status (metropolitan, micropolitan, or rural), ambulatory surgery hospital (yes or no), and availability of cardiac procedures (cardiac surgery, cardiac catheterization, or neither).15 Bed size was grouped into the following 4 categories: less than 200, 200 to 299, 300 to 499, and 500 or more beds. Surgical volume was categorized into quartiles for descriptive analyses and as a continuous variable in multivariable models owing to its linear trend when used as quartiles. To calculate the nurse staffing ratio, we divided the nurse full-time equivalents by the total inpatient days at each hospital and categorized the ratios into quartiles.13,15

    30-Day All-Cause HWR After Surgery
    Study Population/Index Cohort

    We followed the CMS method of measuring HWR to define the surgical cohort.16 We included hospitalizations with an eligible AHRQ CCS surgical procedure category. We excluded in-hospital deaths, transfers to another acute-care hospital, discharges against medical advice, and admissions for primary psychiatric diagnoses, rehabilitation, or medical treatment of cancer. Unlike the CMS method, we excluded all patients who had obstetric procedures, because our focus was on operations applicable to all general hospital patients.

    Identification of 30-Day Unplanned Readmissions

    Following the CMS approach, we identified unplanned readmissions within 30 days after the index discharge, regardless of the cause of readmission. We excluded planned readmissions. If a patient had more than 1 unplanned admission within 30 days of discharge, only the first was considered a readmission. Owing to the limited linkage across years in the SID databases in Iowa (2012-2013) and Washington (2012-2013 and 2013-2014), the readmission outcome could only be partially linked to the index hospitalization if the index admission occurred in December of each year.

    Risk Adjustment

    The CMS approach adjusts for patients 65 years or older, comorbidities based on CMS condition categories, and the AHRQ CCS procedure categories. Based on the aims of our study and data availability in the SIDs, we modified this approach by adding sex and race/ethnicity into the risk-adjustment model11; using the AHRQ risk-adjustment software, which included 29 variables, to calculate comorbidity burden (eMethods 1 in the Supplement); and adjusting for the 88 principal diagnoses associated with the surgical operations relevant to the surgical population in this study (eTable 3 in the Supplement).

    Calculating Risk-Adjusted Readmission Rates

    Following the CMS approach, we estimated hierarchical generalized linear models for each year to calculate hospital-level 30-day risk-adjusted readmission rates (RSRRs) for our surgical cohort. This approach simultaneously modeled data at the patient and hospital levels to account for variance in patient outcomes within and between hospitals. At the patient level, we modeled the log-odds of 30-day readmission outcome after adjusting for patient risk factors. At the hospital level, we modeled hospital-specific effects, which accounted for the underlying risk of readmission at the hospital, after adjustment for patient risk factors. The hospital-specific effects also accounted for the clustering (nonindependence) of patients within the same hospital. This method generated the predicted and expected numbers of readmissions for each hospital. The hospital’s RSRR was calculated as a hospital’s ratio of predicted to expected readmissions multiplied by the overall observed readmission rate within each state.

    Statistical Analysis

    Data were analyzed from June 1, 2017, through March 1, 2018. We investigated variation in hospital characteristics by year and by state and conducted paired, 2-tailed t tests to compare RSRRs between SNHs and non-SNHs by state. We used analysis of variance tests to further explore these comparisons, including additional hospital characteristics. We aggregated data across the 4 states and 4 years and generated a multivariable linear regression model to examine variation across hospitals in the prediction of hospital RSRRs, adjusting for hospital characteristics, state, and year. We reported regression coefficients with 95% CIs and P values. Entries with missing data for all hospital characteristics were excluded from the model. Owing to some data missingness on ambulatory surgery hospital and availability of cardiac procedures, we recoded the missing values as an additional category for each of these 2 variables (eMethods 2 in the Supplement). Last, to assess whether specific surgical specialties had higher RSRRs than others, we examined differences in RSRRs between SNHs and non-SNHs using AHRQ’s CCS, a classification system based on the International Classification of Diseases, Ninth Revision, Clinical Modification, that categorizes approximately 3900 procedure codes into 231 clinically meaningful categories. We used P < .05 (2-sided) to determine statistical significance. All analyses were conducted using SAS software (version 9.4; SAS Institute, Inc).

    Results
    Patient and Hospital Characteristics

    Our surgical cohort included 1 252 505 patients across all 4 years and 4 states, with a mean (SD) age of 52.7 (18.1) years. Characteristics of the cohort are given in eTables 1 and 2 in the Supplement; in total, 48.2% were men and 51.7% were women; 66.0% were white; and 16.0% had Medicaid insurance.

    As shown in Table 1, the number of hospitals in each state remained relatively stable over time, with SNHs constituting almost 30% of all hospitals in each state for each year. New York (range, 155-162) and Florida (range, 193-196) had the greatest number of hospitals relative to Iowa (range, 93-99) and Washington (range, 77-79). Hospital characteristics varied based on geographic location; more teaching hospitals were found in New York (2011, 24 of 162 [14.8%]; 2014, 25 of 155 [16.1%]) and Florida (2011, 9 of 193 [4.7%]; 2014, 9 of 196 [4.6%]), relative to Iowa (2011, 1 of 97 [1.0%]; 2014, 2 of 99 [2.0%]) and Washington (2011, 2 of 79 [2.5%]; 2014, 2 of 78 [2.6%]). State topography was associated with hospital location; most Iowa hospitals were rural (2011, 56 of 97 [57.7%]; 2014, 53 of 99 [53.5%]), whereas metropolitan hospitals constituted the majority in New York (2011, 130 of 162 [80.2%]; 2014, 127 of 155 [81.9%]), Florida (2011, 178 of 193 [92.2%]; 2014, 179 of 196 [91.3%]), and Washington (2011, 58 of 79 [73.4%]; 2014, 58 of 78 [74.4%]). States with a greater number of hospitals were also more likely to have hospitals with higher numbers of beds, and very few hospitals had 500 or more beds in states with the lowest number of hospitals (Iowa and Washington). The presence of ambulatory surgery centers and cardiac intervention capabilities were more frequent in states with a greater number of hospitals as well; New York had a higher prevalence of hospital-associated ambulatory surgery centers (2011, 96 of 162 [59.3%]; 2014, 92 of 155 [59.4%]) compared with Florida (2011, 45 of 193 [23.3%]; 2014, 45 of 196 [23.0%]), Iowa (2011, 22 of 97 [22.7%]; 2014, 25 of 99 [25.2%]), and Washington (2011, 23 of 79 [29.1%]; 2014, 22 of 78 [28.2%]), and cardiac surgery–equipped facilities were more prevalent in Florida (2011, 59 of 193 [30.6%]; 2014, 59 of 196 [30.1%]) compared with New York (2011, 38 of 162 [23.4%]; 2014, 38 of 155 [24.5%]), Iowa (2011, 10 of 97 [10.3%]; 2014, 12 of 99 [12.1%]), and Washington (2011, 17 of 79 [21.5%]; 2014, 17 of 78 [21.8%]).

    Hospital RSRRs

    As presented in Table 2, RSRRs of SNHs were generally higher than those of non-SNHs, except in Washington. Significant differences were found for mean (SD) RSRRs of SNHs vs non-SNHs in New York (9.6 [0.1] vs 10.9 [0.1]; P < .001) and Florida (11.6 [0.1] vs 12.1 [0.1]; P = .001).

    Individual hospital characteristics were associated with mean RSRRs, primarily in New York and Florida. In both states, SNHs had higher RSRRs compared with non-SNHs, even after we examined differences by hospital ownership; mean (SD) RSRRs were highest for nonfederal government SNHs in New York (11.9 [0.2]; P < .001) and for-profit, private SNHs in Florida (13.1 [0.2]; P < .001). For bed size, higher RSRRs were observed in New York only for SNHs with 200 to 399 beds (mean [SD], 12.0 [0.4]; P = .006). Geographic location was significant in Iowa among SNHs in metropolitan areas (mean [SD] RSRR, 8.9 [0.1]; P = .04); these had the highest RSRRs.

    The association of hospital resources with RSRRs also varied across states. Safety-net hospitals without ambulatory surgery centers or cardiac intervention capabilities were generally associated with higher RSRRs. Significant mean (SD) differences for ambulatory surgery centers were found in Florida (11.2 [0.2]; P = .002) and Washington (8.4 [0.2]; P = .002) and for cardiac intervention capabilities in New York (10.4 [0.2]; P = .001) and Florida (11.8 [0.2]; P < .001). Last, we examined differences in RSRRs by nursing staffing and hospital volume (eTable 4 in the Supplement). Although SNHs had higher RSRRs relative to non-SNHs across all years, this association was only significant in Florida (nurse staffing, 11.5 [0.3] to 12.9 [0.2] [P = .02]; surgery volume, 11.6 [0.2] to 15.6 [1.1] [P < .001]).

    Model Including RSRRs

    As shown in Table 3, SNHs had 1.02% higher RSRRs relative to non-SNHs (95% CI, 0.75%-1.29%; P < .001), even after adjusting for hospital and patient characteristics. Relative to 2011 (reference category), significantly higher RSRRs were found in 2012 (0.35%; 95% CI, 0.10%-0.61%; P = .007) and 2013 (0.48%; 95% CI, 0.23%-0.74%; P < .001); RSRRs in 2014 were lower, albeit not significant (−0.06%; 95% CI, −0.32% to 0.20%; P = .64). The RSRRs in Florida were highest (2.64%; 95% CI, 2.10%-3.19%; P < .001), followed by New York (1.44%; 95% CI, 0.89%-1.99%; P < .001) and Washington (0.04%; 95% CI, −050% to 0.56%; P = .88); with Iowa as the reference category. Hospital teaching status (0.49%; 95% CI, 0.22%-0.76%; P < .001) and for-profit private hospital ownership (0.65%; 95% CI, 0.25%-1.05%; P = .002) were associated with increased RSRRs, compared with nonteaching hospitals and public hospitals, respectively. Presence of an ambulatory surgery center (−0.49%; 95% CI, −0.72% to −0.27%; P < .001) and cardiac catheterization capabilities (−0.59%; 95% CI, −0.90% to −0.28%; P < .001) were associated with reduced RSRRs, compared with hospital-based surgery programs and hospitals with cardiac surgery capacity, respectively. Surgical volume was linearly associated with reduced RSRRs (−0.48%; 95% CI, −0.65% to −0.30%; P < .001). Compared with the first quartile of nurse staffing ratio as reference category, the second (−0.42%; −0.71% to −0.14%; P = .004) and third (−0.41%; 95% CI, −0.70% to −0.12%; P = .006) quartiles of nurse staffing ratio were also significantly associated with reduced RSRRs; this association did not persist with the fourth quartile. Distributions of individual surgical specialties (as measured by CCS procedure categories) were similar between SNHs and non-SNHs (eFigure and eTables 5 and 6 in the Supplement).

    Discussion

    This study is the first, to our knowledge, to examine the association between risk-adjusted surgical readmission rates and SNH status. Similar to studies of medical readmissions, we found that SNHs had slightly higher RSRRs for surgical readmissions compared with non-SNHs. In addition, RSRRs varied across the 4 states, that is, RSRRs of New York and Florida were slightly higher compared with Iowa, albeit statistically nonsignificant. Furthermore, specific hospital characteristics were associated with variation in RSRRs. Teaching status and investor ownership were associated with increased RSRRs; in contrast, hospital resources (ie, ambulatory surgery center, cardiac intervention capabilities, and high surgical volume) were associated with decreased readmission rates. Regardless of surgical specialty, SNHs still had higher RSRRs compared with non-SNHs. Finally, contrary to what we hypothesized, despite controlling for hospital characteristics, SNHs still had 1.02% higher RSRRs relative to non-SNHs. Although a difference of 1.02% seems relatively small, under the HRRP, hospitals may be financially penalized for any excess readmission rates higher than the national mean. Thus, we believe the observed difference between SNHs and non-SNHs is clinically meaningful.

    Although our study focused specifically on surgical readmissions, our results are consistent with those of Horwitz et al,11 who found that factors such as SNH and teaching status were associated with higher RSRRs in surgical and gynecologic patients, as well as in medicine, neurology, cardiorespiratory, and cardiovascular patients. Higher RSRRs persisted for SNHs, even when controlling for high-risk patient characteristics,11,12 suggesting that regardless of whether the patient population is admitted for medical or surgical reasons, discrepancies exist between SNHs and non-SNHs that cannot necessarily be explained by hospital and/or patient characteristics.

    The effect of teaching status on readmission is likely multifactorial. Resident physicians at academic centers may be more inclined to readmit patients rather than send them to the emergency department when a postdischarge complication occurs.17 Teaching hospitals may also perform more complex operations owing to the availability of specialists and patient resources, a factor not addressed in our analysis.18

    The literature on nurse staffing ratios is inconsistent with respect to readmission rates. Some studies, such as that by Horwitz et al,11 have found that high nurse staffing ratios are associated with reduced hospital readmissions,13,19 whereas others fail to find a direct association with decreased patient morbidity.20 Our findings were also mixed; we found a slight reduction in readmission rates, but only for 1 quartile of nurse staffing ratio. The association between nurse staffing ratios and readmissions is complex and likely provides an example of a hospital resource that may affect outcomes differently.20

    Although we found that hospital surgical volume and selected hospital resources were associated with lower postsurgical RSRRs, these findings are not consistent in the literature. Similar to our findings, some studies11,21 have found that hospital procedural volume, the presence of an ambulatory surgery center, and cardiac intervention capabilities result in better outcomes. Other studies22 add more to the complexity of explaining differences between SNHs and non-SNHs. Wakeam et al22 found that access to hospital resources did not explain differences in outcomes between patients at SNHs vs non-SNHs. After adjusting for hospital resources, they still observed higher mortality rates among patients undergoing 8 major surgical operations in hospitals with a safety-net burden.22 The study by Wakeam et al22 and the present study were unable to explain why SNHs had higher morbidities or readmissions after controlling for hospital surgical volume and other procedural experience.

    Strengths and Limitations

    A notable strength of this study is that it includes a large patient population using data from US hospitals in 4 geographically diverse states during a 4-year period. However, as with any study of administrative databases, there were some limitations. Missing linkage information for Iowa and Washington across years may have underestimated readmission rates. Second, we do not know whether these all-cause readmissions were related to the operations themselves or owing to reasons unrelated to the index admission. Third, we were unable to exclude patients who were discharged alive but who died within 30 days after index discharge because date of death after hospitalization was not available in the Healthcare Cost and Utilization Project data. Despite the potential for underestimating readmission rates in our study because of this, our methods were generally consistent with the CMS HWR measure. In addition, administrative databases do not include certain patient characteristics, such as primary language, level of education, and income, or variables pertinent to the postdischarge environment (eg, homelessness or level of social support), which have known associations with use of health care services.7 Finally, administrative data do not provide social determinants of health that may relate to patients’ disease burden and may affect 30-day readmission rates disproportionately in SNHs.

    Although readmissions are considered an important measure of hospital quality,1,2 evaluation of additional outcomes (eg, emergency department visits) may provide greater insight into the differences between SNHs and non-SNHs. Future research should examine other outcomes that may further elucidate these differences. Furthermore, multiple definitions of SNHs (eg, based on patient insurance status or patient socioeconomic status) exist; depending on which one is used, these definitions may affect whether a hospital is identified as an SNH or a non-SNH. Future studies should explore the effect on the various definitions of SNHs when comparing outcomes between SNHs and non-SNHs. Unlike medical readmissions, surgical readmissions are more likely to be a result of care within the hospital, rather than coordination of postdischarge care.2 Investigation of differences in readmission diagnoses through medical record review or some other methods might shed light into reasons for postoperative readmissions and provide opportunities for reducing potential preventable readmissions. Further analysis involving possible differences in emergency department visits between SNHs and non-SNHs could provide additional information regarding these 2 hospital populations, such as overall use of health care resources in the postdischarge environment.21

    Conclusions

    Even after accounting for hospital characteristics, surgical patients at SNHs had slightly higher readmission rates than their counterparts at non-SNHs, regardless of surgical specialty. Thus, SNHs will more likely to be subject to financial penalties through the HRRP, despite having traditionally fewer financial resources. Solutions for improving patient outcomes could include incentives for improvement rather than relative performance on risk-adjusted comparisons.8 Understanding the differences between SNHs and non-SNHs will likely be useful for improving quality of care as well as providing targets for SNHs to reduce their readmission rates.

    Back to top
    Article Information

    Accepted for Publication: September 23, 2018.

    Corresponding Author: Stephanie D. Talutis, MD, MPH, Department of Surgery, Boston Medical Center, 88 E Newton St, Ste C515, Boston, MA 02118 (stephanie.talutis@bmc.org).

    Published Online: January 16, 2019. doi:10.1001/jamasurg.2018.5242

    Author Contributions: Dr Talutis and Ms Wang had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

    Concept and design: Talutis, Chen, Rosen.

    Acquisition, analysis, or interpretation of data: Talutis, Chen, Wang.

    Drafting of the manuscript: Talutis, Chen, Rosen.

    Critical revision of the manuscript for important intellectual content: Talutis, Chen, Wang.

    Statistical analysis: Chen, Wang, Rosen.

    Administrative, technical, or material support: Talutis.

    Conflict of Interest Disclosures: None reported.

    References
    1.
    Al-Amin  M.  Hospital characteristics and 30-day all-cause readmission rates.  J Hosp Med. 2016;11(10):682-687. doi:10.1002/jhm.2606PubMedGoogle ScholarCrossref
    2.
    Borza  T, Oreline  MK, Skolarus  TA,  et al.  Association of the hospital readmissions reduction program with surgical readmissions.  JAMA Surg. 2018;153(3):243-250. doi:10.1001/jamasurg.2017.4585PubMedGoogle ScholarCrossref
    3.
    Lewen  ME, Altman  S; Committee on the Changing Market, Managed Care, and the Future Viability of Safety-Net Providers.  America’s Health Care Safety-Net: Intact but Endangered. Washington, DC: National Academy Press; 2000.
    4.
    Ayanian  JZ, Weissman  JS, Schneider  EC, Ginsburg  JA, Zaslavsky  AM.  Unmet health needs of uninsured adults in the United States.  JAMA. 2000;284(16):2061-2069. doi:10.1001/jama.284.16.2061PubMedGoogle ScholarCrossref
    5.
    Eisert  SL, Durfee  MJ, Welsh  A, Moore  SL, Mehler  PS, Gabow  PA.  Changes in insurance status and access to care in an integrated safety net healthcare system.  J Community Health. 2009;34(2):122-128. doi:10.1007/s10900-008-9136-2PubMedGoogle ScholarCrossref
    6.
    Coleman  DL.  The impact of the lack of health insurance: how should academic medical centers and medical schools respond?  Acad Med. 2006;81(8):728-731. doi:10.1097/00001888-200608000-00009PubMedGoogle ScholarCrossref
    7.
    Arbaje  AI, Wolff  JL, Yu  Q, Powe  NR, Anderson  GF, Boult  C.  Postdischarge environmental and socioeconomic factors and the likelihood of early hospital readmission among community-dwelling Medicare beneficiaries.  Gerontologist. 2008;48(4):495-504. doi:10.1093/geront/48.4.495PubMedGoogle ScholarCrossref
    8.
    Bernheim  SM.  Measuring quality and enacting policy: readmission rates and socioeconomic factors.  Circ Cardiovasc Qual Outcomes. 2014;7(3):350-352. doi:10.1161/CIRCOUTCOMES.114.001037PubMedGoogle ScholarCrossref
    9.
    Eslami  MH, Rybin  D, Doros  G, Farber  A.  Care of patients undergoing vascular surgery at safety net public hospitals is associated with higher cost but similar mortality to nonsafety net hospitals.  J Vasc Surg. 2014;60(6):1627-1634. doi:10.1016/j.jvs.2014.08.055PubMedGoogle ScholarCrossref
    10.
    Hoehn  RS, Wima  K, Vestal  MA,  et al.  Effect of hospital safety-net burden on cost and outcomes after surgery.  JAMA Surg. 2016;151(2):120-128. doi:10.1001/jamasurg.2015.3209PubMedGoogle ScholarCrossref
    11.
    Horwitz  LI, Bernheim  SM, Ross  JS,  et al.  Hospital characteristics associated with risk-standardized readmission rates.  Med Care. 2017;55(5):528-534. doi:10.1097/MLR.0000000000000713PubMedGoogle ScholarCrossref
    12.
    Bazzoli  GJ, Chen  HF, Zhao  M, Lindrooth  RC.  Hospital financial condition and the quality of patient care.  Health Econ. 2008;17(8):977-995. doi:10.1002/hec.1311PubMedGoogle ScholarCrossref
    13.
    Joynt  KE, Jha  AK.  Who has higher readmission rates for heart failure, and why? implications for efforts to improve care using financial incentives.  Circ Cardiovasc Qual Outcomes. 2011;4(1):53-59. doi:10.1161/CIRCOUTCOMES.110.950964PubMedGoogle ScholarCrossref
    14.
    Healthcare Cost and Utilization Project. SID data elements. https://www.hcup-us.ahrq.gov/sidoverview.jsp. Modified July 24, 2018. Accessed May 7, 2018.
    15.
    American Hospital Association. Data collection methods. http://www.ahadata.com/. January 1, 2018. Accessed May 7, 2018.
    16.
    Horwitz  L, Partovian  C, Lin  Z,  et al. Centers for Medicare and Medicaid Services. Hospital-wide all-cause unplanned readmission measure: final technical report. https://www.cms.gov/Medicare/Quality-Initiatives-Patient-Assessment-Instruments/HospitalQualityInits/Downloads/Hospital-Wide-All-Cause-Readmission.zip. July 1, 2012. Accessed May 7, 2018.
    17.
    Hoyer  EH, Padula  WV, Brotman  DJ,  et al.  Patterns of hospital performance on the hospital-wide 30-day readmission metric: is the playing field level?  J Gen Intern Med. 2018;33(1):57-64. doi:10.1007/s11606-017-4193-9PubMedGoogle ScholarCrossref
    18.
    Rattan  R, Parreco  J, Lindenmaier  LB,  et al.  Underestimation of unplanned readmission after colorectal surgery: a national analysis.  J Am Coll Surg. 2018;226(4):382-390. doi:10.1016/j.jamcollsurg.2017.12.012PubMedGoogle ScholarCrossref
    19.
    Kane  RL, Shamliyan  TA, Mueller  C, Duval  S, Wilt  TJ.  The association of registered nurse staffing levels and patient outcomes: systematic review and meta-analysis.  Med Care. 2007;45(12):1195-1204. doi:10.1097/MLR.0b013e3181468ca3PubMedGoogle ScholarCrossref
    20.
    Blegen  MA, Goode  CJ, Spetz  J, Vaughn  T, Park  SH.  Nurse staffing effects on patient outcomes: safety-net and non–safety-net hospitals.  Med Care. 2011;49(4):406-414. doi:10.1097/MLR.0b013e318202e129PubMedGoogle ScholarCrossref
    21.
    Horwitz  LI, Wang  Y, Altaf  FK,  et al.  Hospital characteristics associated with postdischarge hospital readmission, observation, and emergency department utilization.  Med Care. 2018;56(4):281-289.PubMedGoogle Scholar
    22.
    Wakeam  E, Hevelone  ND, Maine  R,  et al.  Failure to rescue in safety-net hospitals: availability of hospital resources and differences in performance.  JAMA Surg. 2014;149(3):229-235. doi:10.1001/jamasurg.2013.3566PubMedGoogle ScholarCrossref
    ×