- Email: [email protected]
Understanding Failure to Rescue After Esophagectomy in the United States
Journal Pre-proof Understanding Failure-to-Rescue after Esophagectomy in the United States Zaid M. Abdelsattar, MD, MSc, Elizabeth Habermann, MPH, PhD, Bijan J. Borah, PhD, James P. Moriarty, MSc, Ricardo L. Rojas, BA, Shanda H. Blackmon, MD, MPH PII:
S0003-4975(19)31614-5
DOI:
https://doi.org/10.1016/j.athoracsur.2019.09.044
Reference:
ATS 33185
To appear in:
The Annals of Thoracic Surgery
Received Date: 29 January 2019 Revised Date:
26 August 2019
Accepted Date: 14 September 2019
Please cite this article as: Abdelsattar ZM, Habermann E, Borah BJ, Moriarty JP, Rojas RL, Blackmon SH, Understanding Failure-to-Rescue after Esophagectomy in the United States, The Annals of Thoracic Surgery (2019), doi: https://doi.org/10.1016/j.athoracsur.2019.09.044. This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. © 2019 by The Society of Thoracic Surgeons
Understanding Failure-to-Rescue after Esophagectomy in the United States Running head: Failure to rescue and esophagectomy
Zaid M. Abdelsattar, MD, MSc1; Elizabeth Habermann, MPH, PhD2,3 ; Bijan J. Borah, PhD2,3; James P. Moriarty, MSc2; Ricardo L. Rojas, BA2; Shanda H. Blackmon, MD, MPH1
1 Department of Surgery, Mayo Clinic, Rochester, MN Department of Health Sciences Research, Mayo Clinic, Rochester, MN 2 Kern Center for the Science of Health Care Delivery, Mayo Clinic, Rochester, MN
Correspondence/Reprint requests: Shanda Blackmon, M.D., M>P.H., FACS Professor of Surgery, Department of Surgery, Division of Thoracic Surgery Mayo Clinic; 200 1st Street, SW; MA-12W Rochester, MN, 55905 email: [email protected]
Word count: Abstract: 231; Manuscript: 2109; Pages: 23; Tables: 4; Figures: 2 Key words: Esophagectomy; Outcomes; Population based; Failure to resuce; mortality
ABSTRACT BACKGROUND: Data on failure to rescue (FTR) after esophagectomy are sparse. We sought to better understand the patient factors associated with FTR, and assessed whether FTR is associated with hospital volume. METHODS: We identified all patients undergoing esophagectomy between 2010 and 2014 from the AHRQ Nationwide Readmission Database. We defined FTR as mortality after a major complication. Multiple logistic regression was used to identify patient factors and hospitalvolume associations with FTR. RESULTS: Of 26,820 patients undergoing an esophagectomy, 7,130 (26.6%) experienced a major complication. Of those, 1,321 did not survive the index hospitalization (FTR rate 18.5%). Risk factors for FTR included: Increasing age (aOR=1.06, p<.001), congestive heart failure (aOR=2.07, p<0.001), bleeding disorders (aOR=2.9, p<.001), liver disease (aOR=2.37, p=0.001), and renal failure (aOR=2.37, p=0.002). At the hospital level, there was wide variation in FTR rates across hospital volume quintiles, with 21.2% of patients suffering a complication not surviving to discharge at low volume hospitals, compared to 13.4% at high volume hospitals (p<0.001). At low volume hospitals, the highest FTR rates were acute renal failure (35.3%), postoperative hemorrhage (31.9%), and pulmonary failure (28.1%). CONCLUSIONS: One in five esophagectomy patients suffering a complication at low volume hospitals do not survive to discharge. Several patient factors are associated with death after a major complication. Strategies to improve the recognition and management of complications in at-risk patients may be essential to improve outcomes at low-volume hospitals.
Despite substantial advances in surgical care, esophagectomy remains a high-risk operation with complication rates of up to 46%1 and national mortality rates up to 11%,2 with wide variation between hospitals. Studies have shown that differences in mortality after inpatient surgery are largely related to death after a complication, an outcome termed failure-to-rescue (FTR).3 Focusing on FTR has gained momentum as there is a growing body of evidence showing that the incidence of complications and mortality are not correlated at the hospital level.4 In other words, hospitals with higher complication rates do not necessarily have higher mortality rates. Rather, high-mortality hospitals are not as good at the recognition and management of complications once they occur, and thus have high FTR rates.5 Data on FTR after esophagectomy are sparse. Although hospital volume and certain patient risk factors have been associated with FTR,5,6 the underlying mechanisms are not fully elucidated. It is not known whether low volume hospitals have a different burden of at-risk patients, or are failing to rescue patients from complications due to structural limitation, process gaps, or non-adherence to evidence based guidelines. Elucidating the mechanisms underlying FTR is imperative to develop effective quality improvement strategies. In this context, we use national data from the Agency for Healthcare Research and Quality to better understand the patient- and hospital-level associations with FTR. Specifically, we aim to examine (1) whether low- and high-volume hospitals have different rates of FTR and (2) whether there are certain complications that lead to higher rates of FTR at low- and highvolume hospitals in the United States.
PATIENTS AND METHODS Data Source We used data from the Nationwide Readmissions Database (NRD).7 The NRD is part of a family of databases and software tools developed for the Healthcare Cost and Utilization Project (HCUP). The NRD is a unique and powerful database designed to support various types of national analyses for all payers and the uninsured. This database addresses a large gap in health care data: the lack of nationally representative information for all ages and all payers. Unweighted, the NRD contains data from approximately 17 million discharges each year. Weighted, it estimates roughly 36 million discharges. Developed through a Federal-StateIndustry partnership sponsored by the Agency for Healthcare Research and Quality, HCUP data inform decision-making at the national, state, and community levels. Variables in the database include patient demographics, disease-specific diagnostic codes, calculated severity measures, outcomes, and hospital-level characteristics. The NRD is a publicly available de-identified database. This study was exempt from review by the institutional review boards at the Mayo Clinic in Rochester MN.
Patient Population All patients having an esophagectomy between 2010 and 2014 were eligible for study inclusion. These patients were identified using ICD-9 procedure codes 42.4, 42.41, 42.42, 42.51, 42.52, 42.54, and 42.87 either in the principal or secondary procedure positions. Esophageal cancer was tagged as having ICD-9 diagnosis codes of 150-159.9. Patients less than 18 years of age were excluded.
Independent Variables Coexisting conditions were identified by the appropriate ICD-9-CM diagnosis codes and defined using the Elixhauser Method.8 To define postoperative surgical complications, we used methodology originally created by the Complication Screening Project.9–11 Hospital volume was calculated independently for each year of data. The inclusion of multiple years of data is commonplace in analyzing national level data from the Healthcare Cost and Utilization Database (HCUP). While it may be true that some hospitals may be represented more or less when using multiple years, this is in fact taken into account by design of the survey sampling methodology of the HCUP data and post-stratification.12 Total weighted number of esophagectomies by year was determined for each hospital, regardless of whether the individual hospitalizations were of the FTR population. We used volume quintiles for ease of presentation and understanding.13 Hospital volume was then categorized by the yearly quintiles of the esophagectomy frequency distributions. This was done due to the sampling methodology of the NRD. Hospital characteristics reported included hospital bedsize, hospital teaching status, and hospital ownership. Outcome Measures The primary outcome was FTR. For a patient to be eligible for this complication that patient must have had an eligible complication which included: GI bleeding, myocardial infarction, pneumonia, pulmonary failure, venous thromboembolism, acute renal failure, post-operative hemorrhage and surgical site infection. Secondary outcomes included overall in-hospital mortality, morbidity and 30-day readmission. For privacy reasons, NRD identifier for an individual varies by year, and hence it is not possible to track the same individual across multiple years. Therefore, for 30-day readmission the eligible population was restricted to those
discharged prior to December in a given year. This was done in order to ensure that patients had enough time to fully capture a possible readmission within 30 days. Statistical Analysis Per the instructions of HCUP documentation, analyses included appropriate statistical procedures to accommodate the weighting and sampling variables of the data. Analyses consisted of both univariate and multivariate analyses; univariate analyses were bases on ttests for continuous covariates, and chi-squared tests for categorical covariates. Univariate analyses of patient characteristics were presented by (i) surviving to discharge versus not surviving to discharge; and by (ii) esophagectomy volume quantiles. Multivariate analyses of the study outcome were based on logistic regression modeling. While hospital characteristics were compared descriptively they were not included in multivariate modeling. The reasoning for this was to only include hospital level covariates that were specifically relevant to our study population. While high volume esophagectomy hospitals would likely be from larger hospitals, it may not always be the case. The HCUP methods series suggest against using hierarchical modeling with analyses using the sampling weights, therefore we did not use hierarchical modeling.14 To ensure the reliability of the results; however, the analysis was repeated with volume as a continuous variable and using hierarchical modeling, the repeated analyses were qualitatively similar and are shown in Supplemental Table 1. Models were evaluated for discrimination using the c-statistic and for information bias using the Akaike information criterion (AIC) statistic. The c-statistic evaluates model discrimination and represents the area under the receiver operator characteristic curve. A value of 0.5 indicates that the model is equivalent to chance; a value of 1.0 indicates perfect discrimination. In other words, the c-statistic is the probability that a random patient who experienced the outcome will have a higher risk score than a random patient who did not. The AIC is an estimator of the
relative quality of statistical models for a given set of data, and balances over- and under-fitting. Model performance with and without hospital volume were as follows: C statistic with volume is 0.821, C statistic without volume is 0.8. The AIC with volume is 7380 and 7419 without volume. Both parameters suggest better discrimination and model fit with the inclusion of hospital volume. All analyses were performed in SAS software, Version 9.4, SAS Institute Inc., Cary, NC, USA. All statistical tests were two-sided using a 0.05 level of significance.
RESULTS In the 5-year study period between 2010 and 2014, there were 26,820 patients who underwent an esophagectomy, and 7,130 (26.6%) experienced a major complication. Of those 1,321 (18.5%) did not survive the index hospitalization (i.e. average FTR rate = 18.5%). Patients not surviving to discharge were older (mean + SD; 68.4 + 10.4 vs. 62.9 + 11.1, p=0.04), and were more likely to have 2 or more comorbidities (79.2% vs. 70.1%; p<0.001) as shown in Table 1. Specifically, patients not surviving to discharge were more likely to have congestive heart failure (13.1% vs. 4.1%, p<0.001), bleeding disorders (22.8% vs. 6.7%; p<0.001), liver disease (6.3% vs. 2.9%; p=0.005), and renal failure (10.7% vs. 4.1%; p<0.001). Median household income, sex, and esophageal cancer diagnosis were not associated with mortality. Patient characteristics stratified by hospital volume quintile are shown in Table 2. Mean age was comparable between quintiles. Patients at low-volume hospitals were less likely to have an esophageal cancer diagnosis compared to high-volume hospitals (34% vs. 47%; p<0.001). Medicare and/or Medicaid were collectively the most common payers in all quintiles. None of these case-mix differences were associated with failure to rescue in the multivariable model however (Table 3).. In the multivariable analysis, independent risk factors for FTR included: Increasing age (aOR=1.06, p<.001), congestive heart failure (aOR=2.07, p<0.001), coagulopathy (aOR=2.9, p<.001), chronic liver disease (aOR=2.37, p=0.001), and chronic renal failure (aOR=2.37, p=0.002) as shown in Table 3. When compared to hospitals in the highest volume quintile, hospitals in the lowest volume quintile were associated with worse outcomes. (p=0.007). Hospital characteristics are shown in Table 4. Hospitals in all the five volume quintiles tended to be of large bed-size (60% to 78%). Hospitals in the lowest volume quintile were more likely to have non-teaching status, and be private-for-profit hospitals, while those in the highest volume quintile were more likely to be teaching, and private non-profit. Government hospitals
accounted for 14.1% of all hospitals and were roughly equivalent between quintiles. Hospital characteristics were collinear with hospital volume. At the hospital level, there was wide variation FTR rates across hospital volume quintiles, with 21.2% of patients suffering a complication not surviving to discharge at lowvolume hospitals, compared to 13.4% at high-volume hospitals (p<0.001), as shown in Figure 1. Readmission rates were slightly higher at low-volume hospitals (32% vs 30.6%; p=0.004). Figure 2 shows the incidence and mortality rate of specific complications at low- and high-volume hospitals. Renal failure, pneumonia and surgical site infections were the 3 most common complications at both low- and high-volume hospitals. Although the incidence rates were not significantly different for postoperative hemorrhage, respiratory failure and venous thromboembolism, the mortality rates associated with these complications were very different (p<0.001) between low and high volume hospitals as shown in Figure 2. At low-volume hospitals, the highest FTR rates were with acute renal failure (35.3%), postoperative hemorrhage (31.9%) and pulmonary failure (28.1%). The highest FTR rates at high-volume hospitals were with MI (24.4%).
COMMENT In this nationally representative study on failure to rescue after esophagectomy, we found that: 1) There is wide variation in FTR rates across hospitals with a strong volumeoutcome relationship; 2) Several preoperative patient risk factors such as increasing age, bleeding diathesis and, liver or kidney failure are associated with increased odds of not surviving to discharge after a complication; and that 3) high-volume hospitals rescue patients from specific complications more effectively than low-volume hospitals. Failure to rescue has emerged as an important outcome to gauge hospital quality as it reflects the effectiveness of a given hospital’s response to the occurrence of a complication.15,16 On the other hand, the occurrence of a postoperative complication is more closely related to patient factors than the actual delivery of care. Consistent with previous studies, we found a strong relationship between hospital volume and FTR rates,3,4,17 with wide variation between hospitals. Nationally, the FTR rate of 18.5% is comparable to other published studies from Europe.18,19 Although we identified several patient risk factors associated with FTR, the hospitallevel variation was not attributable to major differences in patient case-mix. In one study, Liou and colleagues6 demonstrated that older age, markers of frailty, major infectious complications (e.g., pneumonia and sepsis) and complications of end organ dysfunction (e.g., acute renal failure) were patient level predictors of FTR. In this study, we identified similar patient risk factors. The importance of our results however, lies in the differences in FTR rates between the hospitals despite similar complication profiles. This again highlights the utility of FTR as an outcome to gauge the quality of healthcare delivery. Although certain patients are more likely at-risk, failed rescue efforts ultimately result from hospital deficiencies in managing complications that are occurring at similar incidence rates.
There are at least 2 aspects of healthcare delivery that could account for the differences in FTR rates between low- and high-volume hospitals. First, evidence based processes of care aimed at the timely and effective management of postoperative complications may not be adhered to at low-volume hospitals. One such example is the Surviving Sepsis Campaign.20 Second, hospital structural resources, such as presence of dedicated surgical ICU intensivists and multidisciplinary teams may also contribute to the ability of hospitals to effectively rescue patients from complications.17 In the present study, FTR rates at low-volume hospitals were highest after acute renal failure, postoperative hemorrhage and respiratory failure, supporting the aforementioned assumptions. The success of the high volume center in managing complications is likely multifaceted. Whether this success is a derivative of caring for a wider breadth of patients overall, or specific esophageal complications is unknown. Previous general surgical literature may suggest that a given hospital’s risk adjusted outcomes following non-cancer surgery may not predict its outcomes following cancer surgery.21 On the other hand, caring for a wide breadth of patients may make more specialists available. In this study, despite similar incidence rates of respiratory failure, for example, FTR rates were dramatically higher at low volume centers. One explanation to this might be in the absence of intensivists with expertise in thoracic surgical patients. Further mixed methods research is warranted to fully elucidate these differences.
This study has several limitations. First and foremost, this is an observational study from administrative data.22 Administrative data lack clinical details about the severity of the presenting disease and resultant complication (e.g., varying degrees of sepsis) that may influence the outcome. Certainly, some patients are inherently of higher risk and not all
esophagectomies are technically equal. However, the wide variation in FTR rates across hospitals suggests inefficiency and room for improvement. On the other hand, several features of using this administrative data source are particularly useful for this study. The number of patients and hospitals in this data set exceeds, by far, the number of hospitals in any clinical registry. Furthermore, we were able to include all age groups and all payers. Second, we do not have data on the surgeon’s specialty, the composition of the care team, nor the timing of the occurrence of the complication(s). Previous research has shown that thoracic surgeons are more likely to rescue patients when compared to general surgeons after an esophagectomy.23 In addition, we could not capture the exact circumstances of the death and whether the patient and/or their family decided to withdraw care after the occurrence of major complications. However this is assumed to be randomly distributed between hospitals. Notwithstanding these limitations, the findings presented herein are relevant to patients, hospitals and payers, highlighting the potential for “lessons learned” form high performers. This study is the largest publication to date on FTR after esophagectomy. The present study also exposes several issues in the care of the patient planned for an esophagectomy. Whether the answer to minimize the variation in outcomes lies in referral to a high-volume hospital at time of diagnosis (a.k.a. regionalization) is unknown. While there are opponents and proponents to this approach, the fee-for-service healthcare model in the US makes regionalization challenging. Public reporting of outcomes, targeted feedback programs, and collaborative quality-improvement initiatives are more likely to succeed than volume-based referrals. An additional perspective is whether low-volume hospitals should transfer the patient to a high-volume hospital after the occurrence of a complication. This approach might have detrimental effects on the outcomes at high-volume hospitals, not to mention the additional incurred costs of taking care of complications. This approach might also foster surgeon non-
ownership of patients. Payers, policy-makers, and professional organizations should prioritize platforms that have the potential to reduce mortality in all circumstances. In conclusion, one in five patients who suffer a complication at low-volume hospitals do not survive to discharge after esophagectomy. Several patient factors are associated with death after a major complication. Strategies to improve the recognition and management of complications in at-risk patients may be essential to improve outcomes at low-volume hospitals. Further mixed-methods research is warranted to identify “lessons learned” from high-volume hospitals or hospitals with low FTR rates.
REFERENCES 1.
van Hagen P, Hulshof MCCM, van Lanschot JJB, et al. Preoperative
Chemoradiotherapy for Esophageal or Junctional Cancer. N Engl J Med. 2012;366:2074-2084. doi:10.1056/NEJMoa1112088 2.
Finks JF, Osborne NH, Birkmeyer JD. Trends in hospital volume and operative mortality
for high-risk surgery. N Engl J Med. 2011;364(22):2128-2137. doi:10.1056/NEJMsa1010705 3.
Ghaferi AA, Birkmeyer JD, Dimick JB. Variation in Hospital Mortality Associated with
Inpatient Surgery. N Engl J Med. 2009;361(14):1368-1375. doi:10.1056/NEJMsa0903048 4.
Ghaferi AA, Birkmeyer JD, Dimick JB. Complications, failure to rescue, and mortality
with major inpatient surgery in medicare patients. Ann Surg. 2009;250(6):1029-1033. doi:10.1097/SLA.0b013e3181bef697 5.
Ghaferi AA, Birkmeyer JD, Dimick JB. Hospital volume and failure to rescue with high-
risk surgery. Med Care. 2011;49(12):1076-1081. doi:10.1097/MLR.0b013e3182329b97 6.
Liou DZ, Serna-Gallegos D, Mirocha J, Bairamian V, Alban RF, Soukiasian HJ.
Predictors of Failure to Rescue After Esophagectomy. Ann Thorac Surg. 2018;105(3):871-878. doi:10.1016/j.athoracsur.2017.10.022 7.
Overview of the Nationwide Readmissions Database (NRD). https://www.hcup-
us.ahrq.gov/nrdoverview.jsp. 8.
Elixhauser A, Steiner C, Harris DR, Coffey RM. Comorbidity measures for use with
administrative data. Med Care. 1998;36(1):8-27. http://www.ncbi.nlm.nih.gov/pubmed/9431328. Accessed July 24, 2014. 9.
Iezzoni LI, Daley J, Heeren T, et al. Identifying complications of care using administrative
data. Med Care. 1994;32(7):700-715. http://www.ncbi.nlm.nih.gov/pubmed/8028405. Accessed July 16, 2015.
10.
Weingart SN, Iezzoni LI, Davis RB, et al. Use of administrative data to find substandard
care: validation of the complications screening program. Med Care. 2000;38(8):796-806. http://www.ncbi.nlm.nih.gov/pubmed/10929992. Accessed October 29, 2018. 11.
Gonzalez A a, Abdelsattar ZM, Dimick JB, Dev S, Birkmeyer JD, Ghaferi A a. Time-to-
Readmission and Mortality After High-Risk Surgery. Ann Surg. 2014;00(00):1-7. doi:10.1097/SLA.0000000000000912 12.
Introduction to the HCUP Nationwide Readmission Database. https://www.hcup-
us.ahrq.gov/db/nation/nrd/Introduction_NRD_2010-2016.pdf. 13.
Birkmeyer JD, Siewers AE, Finlayson EV a, et al. Hospital volume and surgical mortality
in the United States. N Engl J Med. 2002;346(15):1128-1137. doi:10.1056/NEJMsa012337 14.
Houchens R, Chu B, Steiner C. Hierarchical Modeling Using HCUP Data HCUP
Methods Series Report # 2007-01 Online.; 2007. https://www.hcupus.ahrq.gov/reports/methods/2007_01.pdf. 15.
Measuring performance. National Quality Forum.
https://www.qualityforum.org/Measuring_Performance/Measuring_Performance.aspx. 16.
Patient safety indicators: technical specifications. AHRQ Quality Indicators.
17.
Ghaferi A a, Osborne NH, Birkmeyer JD, Dimick JB. Hospital characteristics associated
with failure to rescue from complications after pancreatectomy. J Am Coll Surg. 2010;211(3):325-330. doi:10.1016/j.jamcollsurg.2010.04.025 18.
Busweiler LA, Henneman D, Dikken JL, et al. Failure-to-rescue in patients undergoing
surgery for esophageal or gastric cancer. Eur J Surg Oncol. 2017;43(10):1962-1969. doi:10.1016/j.ejso.2017.07.005 19.
Nimptsch U, Haist T, Krautz C, Grützmann R, Mansky T, Lorenz D. Hospital volume, in-
hospital mortality, and failure to rescue in esophageal surgery. Dtsch Aerzteblatt Online. 2018;115(47):793-800. doi:10.3238/arztebl.2018.0793
20.
Dellinger RP, Levy MM, Carlet JM, et al. Surviving Sepsis Campaign: International
guidelines for management of severe sepsis and septic shock: 2008. Crit Care Med. 2008;36(1):296-327. doi:10.1097/01.CCM.0000298158.12101.41 21.
Abdelsattar ZM, Krell RW, Campbell D a, Hendren S, Wong SL. Differences in Hospital
Performance for Noncancer vs Cancer Colorectal Surgery. J Am Coll Surg. 2014;219(3):450459. doi:10.1016/j.jamcollsurg.2014.02.034 22.
Sarrazin MSV, Rosenthal GE. Finding pure and simple truths with administrative data.
JAMA. 2012;307(13):1433-1435. doi:10.1001/jama.2012.404 23.
Gopaldas RR, Bhamidipati CM, Dao TK, Markley JG. Impact of surgeon demographics
and technique on outcomes after esophageal resections: A nationwide study. Ann Thorac Surg. 2013;95(3):1064-1069. doi:10.1016/j.athoracsur.2012.10.038
FIGURE LEGENDS FIGURE 1: Outcomes at the lowest and highest volume quintiles are shown. Mortality rates (10.9% vs 3.9%), morbidity rates (38% vs 25%), FTR rates (21% vs 13.5%) and readmission (32% vs 30.6%) all p<0.05.
FIGURE 2: Incidence and mortality rates of specific complications stratified by hospital volume. Renal failure, pneumonia, GI bleeding and surgical site infection incidence rates are statistically different (<0.001) between low and high volume centers. However, mortality rates are only statistically different for postoperative hemorrhage, pulmonary failure, renal failure and venous thromboembolism (p<0.001) between low and high volume hospitals, despite seeming similar incidences (asterisks).
Table 1: Patient characteristics by surviving to discharge status. Survived to Discharge Patient Characteristic Yes (n=25499) No (n=1321)
Age, y, mean + SD Sex; Male Esophageal cancer diagnosis Insurance Medicare Medicaid Private Insurance Other Median household income First quartile Second quartile Third quartile Fourth quartile Comorbidity index 0 1 2+
62.7 + 11.1 19712 (77.3)
68.4 + 10.4 972 (73.5)
11919 (46.7)
575 (43.5)
p Value 0.04 0.097 0.232 <0.001
11511 (45.2)
897 (67.9)
2064 (8.1)
95 (7.1)
10927 (42.9) 965 (3.7)
279 (21.2) 50 (3.7)
5525 (22.0) 6573.5 (26.2)
336.1 (25.6) 369 (28.2)
6527.8 (26.0)
318.1 (24.3)
6422.1 (25.6)
285.4 (21.8)
2491 (9.7)
79 (6.0)
5136 (20.1) 17872 (70.1)
195 (14.7) 1047 (79.2)
0.149
<0.001
Table 2: Patient characteristics by hospital volume quintile. Hospital Volume Quintile Patient Lowest Second Middle Fourth Characteristic Volume Volume Volume Volume Quintile Quintile Quintile Quintile Age; y, mean + SD 64 + 12.5 62.5 + 11.9 63 + 11.8 63.1 + 11.4 Sex; Male 68.7% 69.6% 73.2% 76.3% Esophageal cancer 34.1% 38.4% 42.0% 49.2% diagnosis Insurance Medicare 51.7% 48.9% 47.0% 48.0% Medicaid 15.2% 12.8% 10.6% 9.3% Private Insurance 28.9% 32.3% 36.5% 37.9% Other 4.1% 6.0% 5.9% 4.8% Median household income First (%) 26.6% 26.5% 22.9% 24.2% Second (%) 26.1% 27.3% 23.0% 23.8% Third (%) 22.7% 25.4% 28.5% 26.7% Fourth (%) 24.6% 20.9% 25.6% 25.3% Comorbidity index 0 8.6% 8.3% 7.0% 7.7% 1 13.6% 16.6% 16.8% 17.2% 2+ 77.7% 75.0% 76.1% 75.0%
Highest Volume Quintile 62.9 + 10.8 78.1%
<0.001 <0.001
47.2%
<0.001
Pvalue
<0.001 45.7% 7.2% 43.7% 3.3% 0.127 21.5% 27.0% 25.8% 25.7% <0.001 10.1% 20.8% 68.9%
Table 3: Adjusted odds ratios from the multivariable model identifying associations with failure to rescue. Characteristic Adjusted OR 95% Confidence Interval p Value Age 1.06 1.039 - 1.077 <0.001 Sex; female 1.05 0.744 - 1.481 0.782 Insurance Private Reference Medicare 1.37 0.907 - 2.078 0.134 Medicaid 1.30 0.78 - 2.172 0.313 Self-pay 2.28 0.513 - 10.144 0.279 Other 1.49 0.58 - 3.804 0.410 Median household income 1st quartile Reference 2nd quartile 0.94 0.609 - 1.437 0.760 3rd quartile 0.91 0.618 - 1.334 0.621 4th quartile 0.86 0.581 - 1.275 0.454 Year of diagnosis 2014 Reference 2010 1.59 1.079 - 2.339 0.019 2011 1.37 0.849 - 2.205 0.198 2012 1.39 0.92 - 2.087 0.118 2013 1.02 0.687 - 1.518 0.918 Hospital volume quintile Highest volume quintile Reference Lowest volume quintile 1.90 1.194 - 3.035 0.007 2nd volume quintile 1.30 0.79 - 2.135 0.302 Middle volume quintile 1.64 1.093 - 2.465 0.017 4th volume quintile 1.63 1.152 - 2.312 0.006 Comorbid conditions Metastatic cancer 0.95 0.647 - 1.398 0.798 AIDS 1.28 0.163 - 10.046 0.814 Alcohol abuse 1.29 0.793 - 2.087 0.307 Deficiency anemias 0.80 0.563 - 1.142 0.220 Rheumatoid arthritis 0.65 0.234 - 1.818 0.414 Chronic blood loss 0.94 0.311 - 2.846 0.914 Congestive heart failure 2.07 1.431 - 2.994 <0.001 Chronic pulmonary disease 0.97 0.738 - 1.274 0.827 Coagulopathy 2.90 2.084 - 4.033 <0.001 Depression 0.62 0.39 - 0.997 0.049 Diabetes, uncomplicated 0.82 0.578 - 1.157 0.256 Diabetes with chronic complications 1.06 0.467 - 2.382 0.897 Drug abuse 0.55 0.173 - 1.719 0.300 Hypertension 0.46 0.351 - 0.604 <0.001 Hypothyroidism 0.68 0.349 - 1.316 0.251 Liver disease 2.37 1.397 - 4.002 0.001 Fluid and electrolyte disorders 2.72 2.093 - 3.522 <0.001 Other neurological disorders 1.13 0.558 - 2.299 0.731 Obesity 0.41 0.233 - 0.724 0.002 Paralysis 2.16 0.974 - 4.777 0.058 Peripheral vascular disorders 1.54 0.87 - 2.721 0.138 Psychoses 0.67 0.324 - 1.387 0.280 Pulmonary circulation disorders 2.34 1.493 - 3.666 <0.001 Renal failure 2.37 1.389 - 4.028 0.002 Solid tumor without metastasis 0.90 0.448 - 1.796 0.759 Peptic ulcer disease excluding bleeding 3.91 0.755 - 20.281 0.104 Valvular disease 0.68 0.323 - 1.447 0.320 Weight loss 1.38 1.034 - 1.852 0.029
Table 4: Hospital characteristics by hospital volume quintile.
Hospital Characteristic Hospital Bed Size Small Medium Large Hospital Teaching Status Metropolitan non-teaching Metropolitan teaching Non-metropolitan hospital Hospital Control/Ownership Government Private, non-profit Private, for-profit
Lowest Volume Quintile
Second Volume Quintile
Hospital Volume Quintile Middle Fourth Volume Volume Quintile Quintile
Highest Volume Quintile
P-value <.001
9.4% 29.7% 60.9%
17.2% 26.9% 55.9%
10.2% 23.8% 66.1%
8.4% 18.5% 73.1%
11.2% 11.2% 77.7% <.001
67.9% 31.9% 0.3%
40.2% 50.6% 9.1%
42.4% 52.4% 5.2%
24.4% 73.4% 2.2%
3.8% 96.0% 0.2% <.001
16.0% 54.5% 29.5%
16.3% 70.5% 13.2%
14.6% 71.5% 13.8%
11.1% 82.0% 6.9%
14.5% 79.0% 6.5%
S0003-4975(19)31614-5
DOI:
https://doi.org/10.1016/j.athoracsur.2019.09.044
Reference:
ATS 33185
To appear in:
The Annals of Thoracic Surgery
Received Date: 29 January 2019 Revised Date:
26 August 2019
Accepted Date: 14 September 2019
Please cite this article as: Abdelsattar ZM, Habermann E, Borah BJ, Moriarty JP, Rojas RL, Blackmon SH, Understanding Failure-to-Rescue after Esophagectomy in the United States, The Annals of Thoracic Surgery (2019), doi: https://doi.org/10.1016/j.athoracsur.2019.09.044. This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. © 2019 by The Society of Thoracic Surgeons
Understanding Failure-to-Rescue after Esophagectomy in the United States Running head: Failure to rescue and esophagectomy
Zaid M. Abdelsattar, MD, MSc1; Elizabeth Habermann, MPH, PhD2,3 ; Bijan J. Borah, PhD2,3; James P. Moriarty, MSc2; Ricardo L. Rojas, BA2; Shanda H. Blackmon, MD, MPH1
1 Department of Surgery, Mayo Clinic, Rochester, MN Department of Health Sciences Research, Mayo Clinic, Rochester, MN 2 Kern Center for the Science of Health Care Delivery, Mayo Clinic, Rochester, MN
Correspondence/Reprint requests: Shanda Blackmon, M.D., M>P.H., FACS Professor of Surgery, Department of Surgery, Division of Thoracic Surgery Mayo Clinic; 200 1st Street, SW; MA-12W Rochester, MN, 55905 email: [email protected]
Word count: Abstract: 231; Manuscript: 2109; Pages: 23; Tables: 4; Figures: 2 Key words: Esophagectomy; Outcomes; Population based; Failure to resuce; mortality
ABSTRACT BACKGROUND: Data on failure to rescue (FTR) after esophagectomy are sparse. We sought to better understand the patient factors associated with FTR, and assessed whether FTR is associated with hospital volume. METHODS: We identified all patients undergoing esophagectomy between 2010 and 2014 from the AHRQ Nationwide Readmission Database. We defined FTR as mortality after a major complication. Multiple logistic regression was used to identify patient factors and hospitalvolume associations with FTR. RESULTS: Of 26,820 patients undergoing an esophagectomy, 7,130 (26.6%) experienced a major complication. Of those, 1,321 did not survive the index hospitalization (FTR rate 18.5%). Risk factors for FTR included: Increasing age (aOR=1.06, p<.001), congestive heart failure (aOR=2.07, p<0.001), bleeding disorders (aOR=2.9, p<.001), liver disease (aOR=2.37, p=0.001), and renal failure (aOR=2.37, p=0.002). At the hospital level, there was wide variation in FTR rates across hospital volume quintiles, with 21.2% of patients suffering a complication not surviving to discharge at low volume hospitals, compared to 13.4% at high volume hospitals (p<0.001). At low volume hospitals, the highest FTR rates were acute renal failure (35.3%), postoperative hemorrhage (31.9%), and pulmonary failure (28.1%). CONCLUSIONS: One in five esophagectomy patients suffering a complication at low volume hospitals do not survive to discharge. Several patient factors are associated with death after a major complication. Strategies to improve the recognition and management of complications in at-risk patients may be essential to improve outcomes at low-volume hospitals.
Despite substantial advances in surgical care, esophagectomy remains a high-risk operation with complication rates of up to 46%1 and national mortality rates up to 11%,2 with wide variation between hospitals. Studies have shown that differences in mortality after inpatient surgery are largely related to death after a complication, an outcome termed failure-to-rescue (FTR).3 Focusing on FTR has gained momentum as there is a growing body of evidence showing that the incidence of complications and mortality are not correlated at the hospital level.4 In other words, hospitals with higher complication rates do not necessarily have higher mortality rates. Rather, high-mortality hospitals are not as good at the recognition and management of complications once they occur, and thus have high FTR rates.5 Data on FTR after esophagectomy are sparse. Although hospital volume and certain patient risk factors have been associated with FTR,5,6 the underlying mechanisms are not fully elucidated. It is not known whether low volume hospitals have a different burden of at-risk patients, or are failing to rescue patients from complications due to structural limitation, process gaps, or non-adherence to evidence based guidelines. Elucidating the mechanisms underlying FTR is imperative to develop effective quality improvement strategies. In this context, we use national data from the Agency for Healthcare Research and Quality to better understand the patient- and hospital-level associations with FTR. Specifically, we aim to examine (1) whether low- and high-volume hospitals have different rates of FTR and (2) whether there are certain complications that lead to higher rates of FTR at low- and highvolume hospitals in the United States.
PATIENTS AND METHODS Data Source We used data from the Nationwide Readmissions Database (NRD).7 The NRD is part of a family of databases and software tools developed for the Healthcare Cost and Utilization Project (HCUP). The NRD is a unique and powerful database designed to support various types of national analyses for all payers and the uninsured. This database addresses a large gap in health care data: the lack of nationally representative information for all ages and all payers. Unweighted, the NRD contains data from approximately 17 million discharges each year. Weighted, it estimates roughly 36 million discharges. Developed through a Federal-StateIndustry partnership sponsored by the Agency for Healthcare Research and Quality, HCUP data inform decision-making at the national, state, and community levels. Variables in the database include patient demographics, disease-specific diagnostic codes, calculated severity measures, outcomes, and hospital-level characteristics. The NRD is a publicly available de-identified database. This study was exempt from review by the institutional review boards at the Mayo Clinic in Rochester MN.
Patient Population All patients having an esophagectomy between 2010 and 2014 were eligible for study inclusion. These patients were identified using ICD-9 procedure codes 42.4, 42.41, 42.42, 42.51, 42.52, 42.54, and 42.87 either in the principal or secondary procedure positions. Esophageal cancer was tagged as having ICD-9 diagnosis codes of 150-159.9. Patients less than 18 years of age were excluded.
Independent Variables Coexisting conditions were identified by the appropriate ICD-9-CM diagnosis codes and defined using the Elixhauser Method.8 To define postoperative surgical complications, we used methodology originally created by the Complication Screening Project.9–11 Hospital volume was calculated independently for each year of data. The inclusion of multiple years of data is commonplace in analyzing national level data from the Healthcare Cost and Utilization Database (HCUP). While it may be true that some hospitals may be represented more or less when using multiple years, this is in fact taken into account by design of the survey sampling methodology of the HCUP data and post-stratification.12 Total weighted number of esophagectomies by year was determined for each hospital, regardless of whether the individual hospitalizations were of the FTR population. We used volume quintiles for ease of presentation and understanding.13 Hospital volume was then categorized by the yearly quintiles of the esophagectomy frequency distributions. This was done due to the sampling methodology of the NRD. Hospital characteristics reported included hospital bedsize, hospital teaching status, and hospital ownership. Outcome Measures The primary outcome was FTR. For a patient to be eligible for this complication that patient must have had an eligible complication which included: GI bleeding, myocardial infarction, pneumonia, pulmonary failure, venous thromboembolism, acute renal failure, post-operative hemorrhage and surgical site infection. Secondary outcomes included overall in-hospital mortality, morbidity and 30-day readmission. For privacy reasons, NRD identifier for an individual varies by year, and hence it is not possible to track the same individual across multiple years. Therefore, for 30-day readmission the eligible population was restricted to those
discharged prior to December in a given year. This was done in order to ensure that patients had enough time to fully capture a possible readmission within 30 days. Statistical Analysis Per the instructions of HCUP documentation, analyses included appropriate statistical procedures to accommodate the weighting and sampling variables of the data. Analyses consisted of both univariate and multivariate analyses; univariate analyses were bases on ttests for continuous covariates, and chi-squared tests for categorical covariates. Univariate analyses of patient characteristics were presented by (i) surviving to discharge versus not surviving to discharge; and by (ii) esophagectomy volume quantiles. Multivariate analyses of the study outcome were based on logistic regression modeling. While hospital characteristics were compared descriptively they were not included in multivariate modeling. The reasoning for this was to only include hospital level covariates that were specifically relevant to our study population. While high volume esophagectomy hospitals would likely be from larger hospitals, it may not always be the case. The HCUP methods series suggest against using hierarchical modeling with analyses using the sampling weights, therefore we did not use hierarchical modeling.14 To ensure the reliability of the results; however, the analysis was repeated with volume as a continuous variable and using hierarchical modeling, the repeated analyses were qualitatively similar and are shown in Supplemental Table 1. Models were evaluated for discrimination using the c-statistic and for information bias using the Akaike information criterion (AIC) statistic. The c-statistic evaluates model discrimination and represents the area under the receiver operator characteristic curve. A value of 0.5 indicates that the model is equivalent to chance; a value of 1.0 indicates perfect discrimination. In other words, the c-statistic is the probability that a random patient who experienced the outcome will have a higher risk score than a random patient who did not. The AIC is an estimator of the
relative quality of statistical models for a given set of data, and balances over- and under-fitting. Model performance with and without hospital volume were as follows: C statistic with volume is 0.821, C statistic without volume is 0.8. The AIC with volume is 7380 and 7419 without volume. Both parameters suggest better discrimination and model fit with the inclusion of hospital volume. All analyses were performed in SAS software, Version 9.4, SAS Institute Inc., Cary, NC, USA. All statistical tests were two-sided using a 0.05 level of significance.
RESULTS In the 5-year study period between 2010 and 2014, there were 26,820 patients who underwent an esophagectomy, and 7,130 (26.6%) experienced a major complication. Of those 1,321 (18.5%) did not survive the index hospitalization (i.e. average FTR rate = 18.5%). Patients not surviving to discharge were older (mean + SD; 68.4 + 10.4 vs. 62.9 + 11.1, p=0.04), and were more likely to have 2 or more comorbidities (79.2% vs. 70.1%; p<0.001) as shown in Table 1. Specifically, patients not surviving to discharge were more likely to have congestive heart failure (13.1% vs. 4.1%, p<0.001), bleeding disorders (22.8% vs. 6.7%; p<0.001), liver disease (6.3% vs. 2.9%; p=0.005), and renal failure (10.7% vs. 4.1%; p<0.001). Median household income, sex, and esophageal cancer diagnosis were not associated with mortality. Patient characteristics stratified by hospital volume quintile are shown in Table 2. Mean age was comparable between quintiles. Patients at low-volume hospitals were less likely to have an esophageal cancer diagnosis compared to high-volume hospitals (34% vs. 47%; p<0.001). Medicare and/or Medicaid were collectively the most common payers in all quintiles. None of these case-mix differences were associated with failure to rescue in the multivariable model however (Table 3).. In the multivariable analysis, independent risk factors for FTR included: Increasing age (aOR=1.06, p<.001), congestive heart failure (aOR=2.07, p<0.001), coagulopathy (aOR=2.9, p<.001), chronic liver disease (aOR=2.37, p=0.001), and chronic renal failure (aOR=2.37, p=0.002) as shown in Table 3. When compared to hospitals in the highest volume quintile, hospitals in the lowest volume quintile were associated with worse outcomes. (p=0.007). Hospital characteristics are shown in Table 4. Hospitals in all the five volume quintiles tended to be of large bed-size (60% to 78%). Hospitals in the lowest volume quintile were more likely to have non-teaching status, and be private-for-profit hospitals, while those in the highest volume quintile were more likely to be teaching, and private non-profit. Government hospitals
accounted for 14.1% of all hospitals and were roughly equivalent between quintiles. Hospital characteristics were collinear with hospital volume. At the hospital level, there was wide variation FTR rates across hospital volume quintiles, with 21.2% of patients suffering a complication not surviving to discharge at lowvolume hospitals, compared to 13.4% at high-volume hospitals (p<0.001), as shown in Figure 1. Readmission rates were slightly higher at low-volume hospitals (32% vs 30.6%; p=0.004). Figure 2 shows the incidence and mortality rate of specific complications at low- and high-volume hospitals. Renal failure, pneumonia and surgical site infections were the 3 most common complications at both low- and high-volume hospitals. Although the incidence rates were not significantly different for postoperative hemorrhage, respiratory failure and venous thromboembolism, the mortality rates associated with these complications were very different (p<0.001) between low and high volume hospitals as shown in Figure 2. At low-volume hospitals, the highest FTR rates were with acute renal failure (35.3%), postoperative hemorrhage (31.9%) and pulmonary failure (28.1%). The highest FTR rates at high-volume hospitals were with MI (24.4%).
COMMENT In this nationally representative study on failure to rescue after esophagectomy, we found that: 1) There is wide variation in FTR rates across hospitals with a strong volumeoutcome relationship; 2) Several preoperative patient risk factors such as increasing age, bleeding diathesis and, liver or kidney failure are associated with increased odds of not surviving to discharge after a complication; and that 3) high-volume hospitals rescue patients from specific complications more effectively than low-volume hospitals. Failure to rescue has emerged as an important outcome to gauge hospital quality as it reflects the effectiveness of a given hospital’s response to the occurrence of a complication.15,16 On the other hand, the occurrence of a postoperative complication is more closely related to patient factors than the actual delivery of care. Consistent with previous studies, we found a strong relationship between hospital volume and FTR rates,3,4,17 with wide variation between hospitals. Nationally, the FTR rate of 18.5% is comparable to other published studies from Europe.18,19 Although we identified several patient risk factors associated with FTR, the hospitallevel variation was not attributable to major differences in patient case-mix. In one study, Liou and colleagues6 demonstrated that older age, markers of frailty, major infectious complications (e.g., pneumonia and sepsis) and complications of end organ dysfunction (e.g., acute renal failure) were patient level predictors of FTR. In this study, we identified similar patient risk factors. The importance of our results however, lies in the differences in FTR rates between the hospitals despite similar complication profiles. This again highlights the utility of FTR as an outcome to gauge the quality of healthcare delivery. Although certain patients are more likely at-risk, failed rescue efforts ultimately result from hospital deficiencies in managing complications that are occurring at similar incidence rates.
There are at least 2 aspects of healthcare delivery that could account for the differences in FTR rates between low- and high-volume hospitals. First, evidence based processes of care aimed at the timely and effective management of postoperative complications may not be adhered to at low-volume hospitals. One such example is the Surviving Sepsis Campaign.20 Second, hospital structural resources, such as presence of dedicated surgical ICU intensivists and multidisciplinary teams may also contribute to the ability of hospitals to effectively rescue patients from complications.17 In the present study, FTR rates at low-volume hospitals were highest after acute renal failure, postoperative hemorrhage and respiratory failure, supporting the aforementioned assumptions. The success of the high volume center in managing complications is likely multifaceted. Whether this success is a derivative of caring for a wider breadth of patients overall, or specific esophageal complications is unknown. Previous general surgical literature may suggest that a given hospital’s risk adjusted outcomes following non-cancer surgery may not predict its outcomes following cancer surgery.21 On the other hand, caring for a wide breadth of patients may make more specialists available. In this study, despite similar incidence rates of respiratory failure, for example, FTR rates were dramatically higher at low volume centers. One explanation to this might be in the absence of intensivists with expertise in thoracic surgical patients. Further mixed methods research is warranted to fully elucidate these differences.
This study has several limitations. First and foremost, this is an observational study from administrative data.22 Administrative data lack clinical details about the severity of the presenting disease and resultant complication (e.g., varying degrees of sepsis) that may influence the outcome. Certainly, some patients are inherently of higher risk and not all
esophagectomies are technically equal. However, the wide variation in FTR rates across hospitals suggests inefficiency and room for improvement. On the other hand, several features of using this administrative data source are particularly useful for this study. The number of patients and hospitals in this data set exceeds, by far, the number of hospitals in any clinical registry. Furthermore, we were able to include all age groups and all payers. Second, we do not have data on the surgeon’s specialty, the composition of the care team, nor the timing of the occurrence of the complication(s). Previous research has shown that thoracic surgeons are more likely to rescue patients when compared to general surgeons after an esophagectomy.23 In addition, we could not capture the exact circumstances of the death and whether the patient and/or their family decided to withdraw care after the occurrence of major complications. However this is assumed to be randomly distributed between hospitals. Notwithstanding these limitations, the findings presented herein are relevant to patients, hospitals and payers, highlighting the potential for “lessons learned” form high performers. This study is the largest publication to date on FTR after esophagectomy. The present study also exposes several issues in the care of the patient planned for an esophagectomy. Whether the answer to minimize the variation in outcomes lies in referral to a high-volume hospital at time of diagnosis (a.k.a. regionalization) is unknown. While there are opponents and proponents to this approach, the fee-for-service healthcare model in the US makes regionalization challenging. Public reporting of outcomes, targeted feedback programs, and collaborative quality-improvement initiatives are more likely to succeed than volume-based referrals. An additional perspective is whether low-volume hospitals should transfer the patient to a high-volume hospital after the occurrence of a complication. This approach might have detrimental effects on the outcomes at high-volume hospitals, not to mention the additional incurred costs of taking care of complications. This approach might also foster surgeon non-
ownership of patients. Payers, policy-makers, and professional organizations should prioritize platforms that have the potential to reduce mortality in all circumstances. In conclusion, one in five patients who suffer a complication at low-volume hospitals do not survive to discharge after esophagectomy. Several patient factors are associated with death after a major complication. Strategies to improve the recognition and management of complications in at-risk patients may be essential to improve outcomes at low-volume hospitals. Further mixed-methods research is warranted to identify “lessons learned” from high-volume hospitals or hospitals with low FTR rates.
REFERENCES 1.
van Hagen P, Hulshof MCCM, van Lanschot JJB, et al. Preoperative
Chemoradiotherapy for Esophageal or Junctional Cancer. N Engl J Med. 2012;366:2074-2084. doi:10.1056/NEJMoa1112088 2.
Finks JF, Osborne NH, Birkmeyer JD. Trends in hospital volume and operative mortality
for high-risk surgery. N Engl J Med. 2011;364(22):2128-2137. doi:10.1056/NEJMsa1010705 3.
Ghaferi AA, Birkmeyer JD, Dimick JB. Variation in Hospital Mortality Associated with
Inpatient Surgery. N Engl J Med. 2009;361(14):1368-1375. doi:10.1056/NEJMsa0903048 4.
Ghaferi AA, Birkmeyer JD, Dimick JB. Complications, failure to rescue, and mortality
with major inpatient surgery in medicare patients. Ann Surg. 2009;250(6):1029-1033. doi:10.1097/SLA.0b013e3181bef697 5.
Ghaferi AA, Birkmeyer JD, Dimick JB. Hospital volume and failure to rescue with high-
risk surgery. Med Care. 2011;49(12):1076-1081. doi:10.1097/MLR.0b013e3182329b97 6.
Liou DZ, Serna-Gallegos D, Mirocha J, Bairamian V, Alban RF, Soukiasian HJ.
Predictors of Failure to Rescue After Esophagectomy. Ann Thorac Surg. 2018;105(3):871-878. doi:10.1016/j.athoracsur.2017.10.022 7.
Overview of the Nationwide Readmissions Database (NRD). https://www.hcup-
us.ahrq.gov/nrdoverview.jsp. 8.
Elixhauser A, Steiner C, Harris DR, Coffey RM. Comorbidity measures for use with
administrative data. Med Care. 1998;36(1):8-27. http://www.ncbi.nlm.nih.gov/pubmed/9431328. Accessed July 24, 2014. 9.
Iezzoni LI, Daley J, Heeren T, et al. Identifying complications of care using administrative
data. Med Care. 1994;32(7):700-715. http://www.ncbi.nlm.nih.gov/pubmed/8028405. Accessed July 16, 2015.
10.
Weingart SN, Iezzoni LI, Davis RB, et al. Use of administrative data to find substandard
care: validation of the complications screening program. Med Care. 2000;38(8):796-806. http://www.ncbi.nlm.nih.gov/pubmed/10929992. Accessed October 29, 2018. 11.
Gonzalez A a, Abdelsattar ZM, Dimick JB, Dev S, Birkmeyer JD, Ghaferi A a. Time-to-
Readmission and Mortality After High-Risk Surgery. Ann Surg. 2014;00(00):1-7. doi:10.1097/SLA.0000000000000912 12.
Introduction to the HCUP Nationwide Readmission Database. https://www.hcup-
us.ahrq.gov/db/nation/nrd/Introduction_NRD_2010-2016.pdf. 13.
Birkmeyer JD, Siewers AE, Finlayson EV a, et al. Hospital volume and surgical mortality
in the United States. N Engl J Med. 2002;346(15):1128-1137. doi:10.1056/NEJMsa012337 14.
Houchens R, Chu B, Steiner C. Hierarchical Modeling Using HCUP Data HCUP
Methods Series Report # 2007-01 Online.; 2007. https://www.hcupus.ahrq.gov/reports/methods/2007_01.pdf. 15.
Measuring performance. National Quality Forum.
https://www.qualityforum.org/Measuring_Performance/Measuring_Performance.aspx. 16.
Patient safety indicators: technical specifications. AHRQ Quality Indicators.
17.
Ghaferi A a, Osborne NH, Birkmeyer JD, Dimick JB. Hospital characteristics associated
with failure to rescue from complications after pancreatectomy. J Am Coll Surg. 2010;211(3):325-330. doi:10.1016/j.jamcollsurg.2010.04.025 18.
Busweiler LA, Henneman D, Dikken JL, et al. Failure-to-rescue in patients undergoing
surgery for esophageal or gastric cancer. Eur J Surg Oncol. 2017;43(10):1962-1969. doi:10.1016/j.ejso.2017.07.005 19.
Nimptsch U, Haist T, Krautz C, Grützmann R, Mansky T, Lorenz D. Hospital volume, in-
hospital mortality, and failure to rescue in esophageal surgery. Dtsch Aerzteblatt Online. 2018;115(47):793-800. doi:10.3238/arztebl.2018.0793
20.
Dellinger RP, Levy MM, Carlet JM, et al. Surviving Sepsis Campaign: International
guidelines for management of severe sepsis and septic shock: 2008. Crit Care Med. 2008;36(1):296-327. doi:10.1097/01.CCM.0000298158.12101.41 21.
Abdelsattar ZM, Krell RW, Campbell D a, Hendren S, Wong SL. Differences in Hospital
Performance for Noncancer vs Cancer Colorectal Surgery. J Am Coll Surg. 2014;219(3):450459. doi:10.1016/j.jamcollsurg.2014.02.034 22.
Sarrazin MSV, Rosenthal GE. Finding pure and simple truths with administrative data.
JAMA. 2012;307(13):1433-1435. doi:10.1001/jama.2012.404 23.
Gopaldas RR, Bhamidipati CM, Dao TK, Markley JG. Impact of surgeon demographics
and technique on outcomes after esophageal resections: A nationwide study. Ann Thorac Surg. 2013;95(3):1064-1069. doi:10.1016/j.athoracsur.2012.10.038
FIGURE LEGENDS FIGURE 1: Outcomes at the lowest and highest volume quintiles are shown. Mortality rates (10.9% vs 3.9%), morbidity rates (38% vs 25%), FTR rates (21% vs 13.5%) and readmission (32% vs 30.6%) all p<0.05.
FIGURE 2: Incidence and mortality rates of specific complications stratified by hospital volume. Renal failure, pneumonia, GI bleeding and surgical site infection incidence rates are statistically different (<0.001) between low and high volume centers. However, mortality rates are only statistically different for postoperative hemorrhage, pulmonary failure, renal failure and venous thromboembolism (p<0.001) between low and high volume hospitals, despite seeming similar incidences (asterisks).
Table 1: Patient characteristics by surviving to discharge status. Survived to Discharge Patient Characteristic Yes (n=25499) No (n=1321)
Age, y, mean + SD Sex; Male Esophageal cancer diagnosis Insurance Medicare Medicaid Private Insurance Other Median household income First quartile Second quartile Third quartile Fourth quartile Comorbidity index 0 1 2+
62.7 + 11.1 19712 (77.3)
68.4 + 10.4 972 (73.5)
11919 (46.7)
575 (43.5)
p Value 0.04 0.097 0.232 <0.001
11511 (45.2)
897 (67.9)
2064 (8.1)
95 (7.1)
10927 (42.9) 965 (3.7)
279 (21.2) 50 (3.7)
5525 (22.0) 6573.5 (26.2)
336.1 (25.6) 369 (28.2)
6527.8 (26.0)
318.1 (24.3)
6422.1 (25.6)
285.4 (21.8)
2491 (9.7)
79 (6.0)
5136 (20.1) 17872 (70.1)
195 (14.7) 1047 (79.2)
0.149
<0.001
Table 2: Patient characteristics by hospital volume quintile. Hospital Volume Quintile Patient Lowest Second Middle Fourth Characteristic Volume Volume Volume Volume Quintile Quintile Quintile Quintile Age; y, mean + SD 64 + 12.5 62.5 + 11.9 63 + 11.8 63.1 + 11.4 Sex; Male 68.7% 69.6% 73.2% 76.3% Esophageal cancer 34.1% 38.4% 42.0% 49.2% diagnosis Insurance Medicare 51.7% 48.9% 47.0% 48.0% Medicaid 15.2% 12.8% 10.6% 9.3% Private Insurance 28.9% 32.3% 36.5% 37.9% Other 4.1% 6.0% 5.9% 4.8% Median household income First (%) 26.6% 26.5% 22.9% 24.2% Second (%) 26.1% 27.3% 23.0% 23.8% Third (%) 22.7% 25.4% 28.5% 26.7% Fourth (%) 24.6% 20.9% 25.6% 25.3% Comorbidity index 0 8.6% 8.3% 7.0% 7.7% 1 13.6% 16.6% 16.8% 17.2% 2+ 77.7% 75.0% 76.1% 75.0%
Highest Volume Quintile 62.9 + 10.8 78.1%
<0.001 <0.001
47.2%
<0.001
Pvalue
<0.001 45.7% 7.2% 43.7% 3.3% 0.127 21.5% 27.0% 25.8% 25.7% <0.001 10.1% 20.8% 68.9%
Table 3: Adjusted odds ratios from the multivariable model identifying associations with failure to rescue. Characteristic Adjusted OR 95% Confidence Interval p Value Age 1.06 1.039 - 1.077 <0.001 Sex; female 1.05 0.744 - 1.481 0.782 Insurance Private Reference Medicare 1.37 0.907 - 2.078 0.134 Medicaid 1.30 0.78 - 2.172 0.313 Self-pay 2.28 0.513 - 10.144 0.279 Other 1.49 0.58 - 3.804 0.410 Median household income 1st quartile Reference 2nd quartile 0.94 0.609 - 1.437 0.760 3rd quartile 0.91 0.618 - 1.334 0.621 4th quartile 0.86 0.581 - 1.275 0.454 Year of diagnosis 2014 Reference 2010 1.59 1.079 - 2.339 0.019 2011 1.37 0.849 - 2.205 0.198 2012 1.39 0.92 - 2.087 0.118 2013 1.02 0.687 - 1.518 0.918 Hospital volume quintile Highest volume quintile Reference Lowest volume quintile 1.90 1.194 - 3.035 0.007 2nd volume quintile 1.30 0.79 - 2.135 0.302 Middle volume quintile 1.64 1.093 - 2.465 0.017 4th volume quintile 1.63 1.152 - 2.312 0.006 Comorbid conditions Metastatic cancer 0.95 0.647 - 1.398 0.798 AIDS 1.28 0.163 - 10.046 0.814 Alcohol abuse 1.29 0.793 - 2.087 0.307 Deficiency anemias 0.80 0.563 - 1.142 0.220 Rheumatoid arthritis 0.65 0.234 - 1.818 0.414 Chronic blood loss 0.94 0.311 - 2.846 0.914 Congestive heart failure 2.07 1.431 - 2.994 <0.001 Chronic pulmonary disease 0.97 0.738 - 1.274 0.827 Coagulopathy 2.90 2.084 - 4.033 <0.001 Depression 0.62 0.39 - 0.997 0.049 Diabetes, uncomplicated 0.82 0.578 - 1.157 0.256 Diabetes with chronic complications 1.06 0.467 - 2.382 0.897 Drug abuse 0.55 0.173 - 1.719 0.300 Hypertension 0.46 0.351 - 0.604 <0.001 Hypothyroidism 0.68 0.349 - 1.316 0.251 Liver disease 2.37 1.397 - 4.002 0.001 Fluid and electrolyte disorders 2.72 2.093 - 3.522 <0.001 Other neurological disorders 1.13 0.558 - 2.299 0.731 Obesity 0.41 0.233 - 0.724 0.002 Paralysis 2.16 0.974 - 4.777 0.058 Peripheral vascular disorders 1.54 0.87 - 2.721 0.138 Psychoses 0.67 0.324 - 1.387 0.280 Pulmonary circulation disorders 2.34 1.493 - 3.666 <0.001 Renal failure 2.37 1.389 - 4.028 0.002 Solid tumor without metastasis 0.90 0.448 - 1.796 0.759 Peptic ulcer disease excluding bleeding 3.91 0.755 - 20.281 0.104 Valvular disease 0.68 0.323 - 1.447 0.320 Weight loss 1.38 1.034 - 1.852 0.029
Table 4: Hospital characteristics by hospital volume quintile.
Hospital Characteristic Hospital Bed Size Small Medium Large Hospital Teaching Status Metropolitan non-teaching Metropolitan teaching Non-metropolitan hospital Hospital Control/Ownership Government Private, non-profit Private, for-profit
Lowest Volume Quintile
Second Volume Quintile
Hospital Volume Quintile Middle Fourth Volume Volume Quintile Quintile
Highest Volume Quintile
P-value <.001
9.4% 29.7% 60.9%
17.2% 26.9% 55.9%
10.2% 23.8% 66.1%
8.4% 18.5% 73.1%
11.2% 11.2% 77.7% <.001
67.9% 31.9% 0.3%
40.2% 50.6% 9.1%
42.4% 52.4% 5.2%
24.4% 73.4% 2.2%
3.8% 96.0% 0.2% <.001
16.0% 54.5% 29.5%
16.3% 70.5% 13.2%
14.6% 71.5% 13.8%
11.1% 82.0% 6.9%
14.5% 79.0% 6.5%