Abstract
BACKGROUND
Surgical repair or drainage is the standard treatment for benign esophageal perforation. The Food and Drug Administration has approved the use of esophageal stents for the management of malignant esophageal stricture and/or fistula. We hypothesize that increasing enthusiasm and experience with esophageal stents has led to greater use of stents for the management of benign esophageal perforation.
METHODS
We performed a retrospective cohort study (2007–2014) of patients with benign esophageal perforation using MarketScan®—a commercial claims database. Subjects had six months of follow-up. Regression was used for risk-adjustment.
RESULTS
A total of 659 patients (mean 49 years, 41% women) were treated for benign esophageal perforation (surgical repair: n=449, 69%; surgical drainage: n=110, 17%; stent: n=100, 15%). Stent use increased from 7% in 2007 to 30% in 2014 (p-trend<0.001). Over the same period, surgical repair decreased from 71% to 53% (p-trend=0.001), while surgical drainage did not change (p=0.24). After adjustment for other factors that could vary over time, stent use increased by 28% per year (incidence rate ratio=1.28, 95% confidence interval 1.17–1.39). There were no significant changes in risk-adjusted deaths, discharges home, readmissions, or costs over the same period (all p-trend>0.05).
CONCLUSIONS
The use of stents for the management of benign esophageal perforation has increased by over four fold in just eight years, but short-term outcomes have not changed over time for this population of patients. A national registry for off-label use of esophageal stents may clarify the indications for and risks and benefits of stenting benign esophageal perforations.
Benign perforation of the esophagus is a rare but highly morbid disease with a mortality of approximately 10–20%.1 Untreated contamination of the mediastinum and pleural spaces leads to sepsis, multi-organ failure, and death. Definitive surgical treatment includes primary repair of the esophageal defect and debridement, decortication and drainage of the mediastinum and pleural spaces. Some patients are ineligible for operative management because of severe underlying comorbid conditions, poor functional status, complications of perforation (e.g. septic shock, respiratory failure, etc.), or delay in presentation. In such cases, the goals of treatment are to control the esophageal leak, drain contaminated thoracic cavities, and administer antibiotic and nutritional support. Drainage alone is often appropriate for isolated cervical perforation and creation of a controlled esophagocutaneous fistula (e.g. via T-tube drainage) combined with wide drainage has been used to control an esophageal leak when primary repair is not possible.
In the mid-2000’s, the Food and Drug Administration (FDA) approved the use of esophageal stents for the management of malignant stricture and/or fistula. The use of stents for the management of benign esophageal perforation is a logical extension of the original FDA approved indication, and yet its use for this indication remains “off-label.” A growing number of observational studies suggest that stents may be as effective as primary surgical therapy with lower associated risks, length-of-stay (LOS), and costs.2–4 We hypothesize that off-label use of stents for management of benign esophageal perforation has increased during the past decade. A secondary goal of this study was to describe trends in overall outcomes and costs over time, and their association with varying approaches to treatment.
PATIENTS AND METHODS
We performed a retrospective cohort study of adults diagnosed and treated for esophageal perforation between January 15, 2007 and June 30, 2014 with six months of follow-up using the Thomson Reuters MarketScan® Research Databases. This database includes information from employer- and health plan-sourced enrollment and claims data with person-level healthcare utilization across inpatient and outpatient settings. These data are available for workers and their dependents covered by employer-sponsored private health insurance. Diagnoses are codified using the International Classification of Diseases, 9th Revision (ICD-9), and procedures and services are codified using the Current Procedural Terminology (CPT) and Healthcare Common Procedure Coding System (HCPCS) codes (see Appendix A). Patients were required to have continuous health plan enrollment two weeks prior to and six months after esophageal perforation (unless disenrollment was due to death within six months) to ensure complete capture of claims. The MarketScan® database only includes claims data for the supplemental insurance of those patients enrolled in Medicare. As such, patients enrolled in Medicare would have incomplete healthcare utilization data and so patients over age 64 were excluded. Other exclusions were a diagnosis of esophageal cancer in the year prior to and six months after perforation, and patients who underwent esophageal stenting for any reason in the year prior to perforation. This study was approved by the University of Washington Institutional Review Board.
Subjects were divided into three groups—those treated with surgical repair, esophageal stent without repair, or surgical drainage alone—based upon treatment codes during the first seven days of index hospitalization for esophageal perforation. If patients were stented and subsequently surgically drained, they were counted in the stenting group. Information was also collected on measured potential confounders. Age at the time of diagnosis, sex, and health plan type were obtained through enrollment data. The Charlson comorbidity index (CCI) was determined based on claims from inpatient and outpatient claims during the 30 days following diagnosis using methods by Quan.5 Because endoscopy can lead to esophageal perforation, we also measured use of both diagnostic and therapeutic endoscopy during the 14 days prior to perforation as a surrogate for iatrogenic perforation. If patients had no documented endoscopy during the 14 days prior, we considered the perforation to be spontaneous (i.e. Boerhaave’s).
Outcomes included discharges to home after index hospitalization and death during the index or readmissions within six months of diagnosis (MarketScan® does not ascertain outpatient deaths). Claims in the six months after diagnosis of esophageal perforation were used to describe healthcare utilization as measured by the LOS of the index hospitalization and readmissions. Total direct medical expenditures were measured in the six month period after perforation. Out-of-pocket costs (copayments, deductibles, and coinsurance) were also measured. This analysis takes the perspective of the insurer and the patient as total direct costs include both reimbursements to the insurer and out-of-pocket expenses for the patient. Costs were adjusted for inflation using the Consumer Price Index for Medical Care Services to 2014 United States (US) dollars.6
Differences between treatment groups were assessed by Analysis of Variance and Kruskal Wallis tests for normally and non-normally distributed continuous variables, respectively and by Chi-squared tests for categorical data. Unadjusted trends over the period of this study were assessed using a non-parametric test of trend (continuous variables) or a Mantel Haenszel test of trend (binary variables).7,8 Patients who are treated with esophageal stents and surgical drainage are often sicker than those who are treated by conventional surgical repair. Because claims data do not provide sufficiently granular information about the patient’s physiologic status, severity of comorbid conditions, and functional status, we did not perform comparative analyses of outcomes, utilization, or costs between management approaches. In order to estimate trends in treatments over time, however, we did adjust for potential changes in patient characteristics over time. Adjustment covariates included age, sex, Charlson comorbidity index, health plan type, and when applicable, presence of prior endoscopy. We assumed that stent use would occur in at least 10% of perforations. Accordingly, we used Poisson regression to estimate trends in the use of stents over time relative to surgical treatments because we assumed stent use to not be rare.9 To measure temporal trends in outcomes, we modeled differences in in-hospital death (logistic regression), hospital LOS, total six-month costs (generalized linear model [GLM] with an identity link and gamma distribution), risk of discharge to home (Poisson, among those who survived the index hospitalization), and hospital readmission (logistic, for those who survived the index hospitalization) over time. Each adjusted analysis was performed among the total population of patients as well as a planned subgroup analysis for patients with presumed Boerhaave’s perforation (i.e. excluding patients who underwent endoscopy two weeks prior to hospitalization for perforation). Covariate data was missing for only health plan type and in only 4% of cases. There were no differences in treatment type between those with and without missing covariate data (p>0.05), and the proportion of patients with missing covariate data did not change over time (p>0.05). All analyses were case complete, and all adjusted analyses used robust standard errors. Statistical significance was defined by p<0.05. Analyses were performed using commercially available software packages (Stata v. 14.2 IC, StataCorp LP College Station, TX).
RESULTS
Over this eight year study, 659 patients were hospitalized and treated for esophageal perforation (Table). Patients were predominantly middle-aged men. A majority had at least one comorbid condition and were enrolled in a low-deductible health plan. Just over a third underwent endoscopy two weeks prior to hospitalization for perforation, and a majority of these endoscopic procedures were therapeutic. Most patients were treated by surgical repair (69%) with the remainder of patients treated with either surgical drainage (17%) or via esophageal stenting (15%). Overall, typical index LOS was 13 days. While a majority of patients discharged home (77%), the rate of readmission was high (39%). Over the six month follow-up period, 6% of patients died during a hospitalization, and most deaths occurred during the index hospitalization. The median six month total cost of care after benign esophageal perforation for a given patient was $145,942, and the median six month out-of-pocket cost was $5,034. The total costs of care for the 659 patients in this study was $151.6M.
Table.
Patient characteristics, outcomes, and costs by management of esophageal perforation
| Surgical Repair n = 449 |
Stent n = 100 |
Surgical Drainage n = 110 |
All patients n = 659 |
p-value | |
|---|---|---|---|---|---|
| Demographics | |||||
| age, mean (SD) | 49 (11) | 51 (11) | 48 (11) | 49 (12) | 0.17 |
| female, n (%) | 183 (41) | 43 (43) | 46 (42) | 272 (41) | 0.91 |
| Charlson Comorbidity Index, n (%) | 0.02 | ||||
| 0 | 205 (46) | 36 (36) | 40 (36) | 281 (43) | |
| 1 | 125 (28) | 22 (22) | 32 (29) | 179 (27) | |
| ≥2 | 119 (27) | 42 (42) | 38 (35) | 199 (30) | |
| Health Insurance Type, n (%) | 0.53 | ||||
| HMO or Capitated POS | 56 (13) | 12 (12) | 13 (12) | 81 (13) | |
| Low deductible insurance | 334 (78) | 74 (76) | 87 (83) | 495 (78) | |
| High deductible insurance | 40 (9) | 11 (11) | 5 (5) | 56 (9) | |
| EGD within 2 weeks prior, n (%) | |||||
| Therapeutic | 77 (17) | 32 (32) | 18 (16) | 127 (19) | 0.002 |
| Diagnostic | 77 (17) | 17 (17) | 24 (22) | 118 (18) | 0.87 |
| Any | 154 (34) | 49 (49) | 42 (38) | 245 (37) | 0.02 |
| 6-month Healthcare Utilization | |||||
| Length of Stay, median (IQR) | 12 (8–20) | 13 (5–27) | 16 (10–25) | 13 (8–22) | 0.03 |
| Readmission*, n (%) | 159 (36) | 52 (54) | 38 (38) | 249 (39) | 0.004 |
| 6-month total costs, $USD | |||||
| Median (IQR) | $126,573 (75,583–234,386) | $192,321 (93,922–319,313) | $156,894 (85,819–352,652) | $145,942 (78,182–264,105) | 0.005 |
| Mean (SD) | $225,445 (337,748) | $248,978 (216,302) | $231,845 (201,070) | $230,084 (302,445) | 0.78 |
| 6-month out of pocket costs, median (IQR) | $4,987 (2,539–8,418) | $5,906 (3,577–9,071) | $4,551 (2,573–8,009) | $5,034 (2,796–8,470) | 0.12 |
| mean (SD) | $6,864 (8,882) | $7,008 (6,043) | $5,897 (4,730) | $6,724 (7,941) | 0.48 |
| Outcomes | |||||
| Discharge to home*, n (%) | 346 (79) | 72 (72) | 74 (74) | 492 (77) | 0.73 |
| Death, n (%) | |||||
| Index hospitalization | 8 (2) | 4 (4) | 10 (9) | 22 (3) | 0.001 |
| Subsequent hospitalization | 5 (1) | 8 (8) | 2 (2) | 15 (2) | <0.001 |
| Any | 13 (3) | 12 (12) | 12 (11) | 37 (6) | <0.001 |
among those who survived their initial hospitalization
Figure 1 shows that the use of esophageal stents increased over four-fold from 7% in 2007 to 30% in 2014 (p-trend<0.001). The unadjusted rate of increase in stent use was 28% per year (incidence rate ratio [IRR] 1.28, 95% confidence interval [CI] 1.18–1.40). Over the same period, the frequency of surgical repair decreased by one quarter from 71% in 2007 to 53% in 2014 (p-trend=0.001). The unadjusted rate of decrease in surgical repair was 4% per year (IRR 0.96, 95% CI 0.93–0.98). Surgical drainage procedures did not change in frequency over the study period (p-trend=0.24). We explored whether any measured covariates changed significantly over time and found that enrollment in low-deductible insurance plans increased significantly (p=0.01) while enrollment in high-deductible plans decreased significantly (p<0.001) over the study period. Patient age, sex, comorbidity score, and frequency of prior EGDs did not change over the course of eight years (all p>0.05). Adjusting for changing patient characteristics over time revealed the same trends (i.e. rates) in treatment type.
Figure 1.
Trends in Management of Benign Esophageal Perforation
Figure 2 shows that the frequency of discharges home and six-month inpatient death rates did not change over time. Over the same period, the rate of readmission did not change significantly, nor did the median LOS or costs (all p>0.05). Adjusted analyses did not reveal any change in outcomes or costs over time (all p>0.05). While risk-adjusted LOS trended downward during the course of this study (p=0.04), this appeared consistent across treatment types.
Figure 2.
(a) Trends in Outcomes for Patients with Esophageal Perforation (b) trends in Utilization for Patients with Esophageal Perforation.
Patient characteristics and unadjusted outcomes, utilization, and costs varied by treatment (Table). As expected, patients treated with a stent or surgical drainage tended to have more comorbid conditions relative to those who underwent surgical repair. Therapeutic endoscopy prior to perforation was associated with esophageal stenting (p=0.02). Surgical repair was associated with the lowest unadjusted hospital mortality and readmission rate, shortest LOS, and least costs. Patterns of utilization were notably different between stenting and surgical drainage. LOS of was highest for surgical drainage (16 days) with comparable readmission rates relative to surgical repair. In contrast, compared with surgical repair, stenting was associated with a similar LOS but with the highest rate of readmission occurring in over half of patients. Six-month mortality rates were similar and high for patients treated with either esophageal stenting (12%) or surgical drainage (11%). However, among the surgical drainage group a majority of these deaths occurred during the index hospitalization (9%), whereas a majority of these deaths occurred during subsequent hospitalizations among those treated with esophageal stenting (8%). Esophageal stenting was associated with the highest unadjusted six-month costs of care relative to surgical interventions. A post-hoc exploratory analysis (controlling for number of subsequent hospitalizations) revealed that the greater costs seen among patients treated by esophageal stenting may be explained by the higher rate of readmission in this group.
A planned subgroup analysis was conducted among patients who had not undergone endoscopy in the two week period prior to hospitalization for perforation. Temporal trends and patterns of outcomes, utilization, and costs across varying management strategies were similar in this subgroup with presumed spontaneous (Boerhaave’s) esophageal perforation compared to the overall population of patients with esophageal perforation (Appendix B).
COMMENT
This study is the first to describe trends in the management of benign esophageal perforation in the US, as well as associated patient outcomes, healthcare utilization, and costs in the community-at-large. The key finding from this study is that the use of esophageal stents has increased by four-fold over eight years coincident with a proportional decrease in surgical repair and no evidence of a change in outcomes, utilization, or costs among the overall population.
The most likely explanation for the treatment trends observed in this study is that providers believe that esophageal stenting is frequently a reasonable alternative to surgical repair, and in some settings may be superior. It is logical to consider esophageal stenting for perforation management as a means to achieving source control. When coupled with adequate drainage of the mediastinum and pleural spaces, lung expansion, antibiotics, and nutrition, esophageal stenting as part of a comprehensive treatment plan might reasonably be expected to lead to similar outcomes as repair.10 Supporting this logic are several retrospective single-institution or single health-system observational studies that have reported excellent outcomes associated with esophageal stenting among both patients who were and were not candidates for surgical repair.2,3,10,11
The alternative explanation for the treatment trends in our study is that patients have become “sicker” over time. However, we found no evidence of a trend in comorbidity index over time. Additionally, there is no reason to believe that patients have more severe manifestations of esophageal perforation and/or delayed presentations over time. Symptoms of esophageal perforation are typically severe leading patients to seek care. Access to care has likely not changed dramatically in the US over the study period for a population of individuals with employer-provided health insurance.
Despite significant changes in the management of esophageal perforation over time, there were no trends in outcomes, utilization, or costs. One explanation for this finding is that esophageal stenting and surgical repair have equivalent outcomes when applied selectively. It is likely that there are patients better served by surgical repair, and others who have equivalent or even superior outcomes with stenting. In our study, however, we are not able to parse out the indications providers used for selecting one modality over the other. Another possibility is that esophageal stenting is inferior to surgical repair. If surgical outcomes are improving over time, then rapid adoption of esophageal stenting could “counter-balance” the improving surgical outcomes resulting in no overall change in this population’s outcomes over time. It is somewhat discouraging that the adoption of a new treatment option for patients with esophageal perforation would not be associated with improved outcomes over time, but perhaps this is due to uneven application of stenting for patients most likely to benefit from stenting when no guidelines exist to help providers apply alternative treatment algorithms optimally.
For several reasons, our findings lead us to hypothesize that esophageal stenting is inferior to surgical repair in the community-at-large. Our analysis of outcomes after treatment reveal that surgical repair is associated with the best outcomes and surgical drainage is associated with the worst outcomes. This finding is expected given that some patients are “too sick” to undergo surgical repair and their outcomes are driven (i.e. confounded) by their underlying comorbid conditions or the severity or timing of the perforation. For this reason, we would also expect the outcomes of the esophageal stenting to be similar to surgical drainage, if stenting was used to treat the sickest patients. However, esophageal stenting appears to be “replacing” surgical repair rather than surgical drainage, and patterns of outcomes between stenting and surgical drainage were dissimilar. Patients treated with esophageal stenting had higher readmission rates. Our post-hoc analysis suggests that these readmissions appear to explain the higher six-month total costs of care associated with esophageal stenting. Finally, patients were more likely to die in a subsequent hospitalization compared to surgical drainage. This finding suggests that stented patients may experience delayed morbidity and mortality which could be explained by incomplete or absent source control (drainage of mediastinum and pleural spaces), suboptimal antibiotic administration, or failure to provide optimal nutrition.
Our study has several important limitations. The most important limitation is confounding from unmeasured variables that describe the severity of underlying comorbid conditions and functional status, complications of esophageal perforation, and time-to-presentation, degree of contamination, and physiologic impact on the patient. This limitation precludes knowing the indications for treatment. As a consequence, we cannot know for certain the reasons underlying temporal trends in management or compare outcomes across management strategies. Risk-adjustment strategies such as regression and propensity score analyses cannot overcome the problem of unmeasured confounding.1> Unfortunately, there are currently no clinical registries that ascertain variables relevant to the management of esophageal perforation, and all prior studies suffer from similar limitations as ours.13 In addition, by specifying a minimum of two weeks enrollment prior to perforation, we may miss diagnoses of esophageal cancer or stenting procedures. A sensitivity analysis of patients with one full year of pre-enrollment (n=421), however, revealed the same trends in choice of treatment and outcomes. Another limitation is generalizability. Through MarketScan® we cannot describe the specialty training or location of care received, both of which may impact patient outcomes. It is also likely that associated outcomes among all treatment strategies are better in our population compared to the uninsured, unemployed, and individuals older than 65 years of age. Finally, the MarketScan® database does not ascertain deaths that occur outside the inpatient setting. As a consequence, it is possible that we may have underestimated short-term mortality.
The lack of high level evidence about the efficacy and safety of stents for benign esophageal perforations, and the increase in off-label use of stents for this indication reported for the first time in this study, reveal an opportunity—creation of a national clinical registry for benign esophageal perforation. Although a randomized trial comparing treatment outcomes is preferred, it is unlikely to be feasible because benign esophageal perforations are rare,1> and a non-inferiority design typically requires a large sample size. In the absence of a trial, a clinical registry with prospectively defined data elements that ascertain relevant variables of interest (including indication, severity of underlying comorbid conditions and functional status, and severity and timing of perforation) could allow for a risk-adjusted comparison of treatment outcomes. Another benefit of such a clinical registry would be to monitor the quality of care and give providers feedback about their performance. Doing so may address the main hypothesis that arises from our current investigation—esophageal stenting is inferior to surgical repair in the community-at-large because of the lack of a comprehensive treatment plan that addresses mediastinal and pleural contamination, appropriate antibiotic use, and nutrition. Finally, administering such a clinical registry would allow surgeons and gastroenterologists to take a leadership role in addressing growing concerns about off-label use of medical devices. Examples such as the STS/ACC TVT Registry™ (https://www.ncdr.com/webncdr/tvt/publicpage/data-collection) provide a precedent for multi-disciplinary collaboration for building a disease-based registry for the purposes of quality improvement and scientific advancement with an eye towards achieving better patient outcomes.
Supplementary Material
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. 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.
References
- 1.Biancari F, D’Andrea V, Paone R, et al. Current treatment and outcome of esophageal perforations in adults: Systematic review and meta-analysis of 75 studies. World J Surg. 2013;37(5):1051–1059. doi: 10.1007/s00268-013-1951-7. [DOI] [PubMed] [Google Scholar]
- 2.Freeman RK, Van Woerkom JM, Ascioti AJ. Esophageal Stent Placement for the Treatment of Iatrogenic Intrathoracic Esophageal Perforation. Ann Thorac Surg. 2007;83(6):2003–2008. doi: 10.1016/j.athoracsur.2007.02.025. [DOI] [PubMed] [Google Scholar]
- 3.Fischer A, Thomusch O, Benz S, Von Dobschuetz E, Baier P, Hopt UT. Nonoperative treatment of 15 benign esophageal perforations with self-expandable covered metal stents. Ann Thorac Surg. 2006;81(2):467–472. doi: 10.1016/j.athoracsur.2005.08.047. [DOI] [PubMed] [Google Scholar]
- 4.Schmidt SC, Strauch S, Rösch T, et al. Management of esophageal perforations. Surg Endosc Other Interv Tech. 2010;24(11):2809–2813. doi: 10.1007/s00464-010-1054-6. [DOI] [PubMed] [Google Scholar]
- 5.Quan H, Li B, Couris CM, et al. Updating and validating the charlson comorbidity index and score for risk adjustment in hospital discharge abstracts using data from 6 countries. Am J Epidemiol. 2011;173(6):676–682. doi: 10.1093/aje/kwq433. [DOI] [PubMed] [Google Scholar]
- 6.U.S. Bureau of Labor Statistics, Division of Consumer Prices and Price Indexes. [Accessed November 20, 2016];Consumer Price Index. http://www.bls.gov/cpi/
- 7.Cuzick J. A Wilcoxon-type test for trend. Stat Med. 4(1):87–90. doi: 10.1002/sim.4780040112. http://www.ncbi.nlm.nih.gov/pubmed/3992076. [DOI] [PubMed] [Google Scholar]
- 8.Clayton DG, MH . Statistical Models in Epidemiology. Oxford: Oxford University Press; 1993. [Google Scholar]
- 9.McNutt LA, Wu C, Xue X, Hafner JP. Estimating the relative risk in cohort studies and clinical trials of common outcomes. Am J Epidemiol. 2003;157(10):940–943. doi: 10.1093/aje/kwg074. [DOI] [PubMed] [Google Scholar]
- 10.Ben-David K, Behrns K, Hochwald S, et al. Esophageal perforation management using a multidisciplinary minimally invasive treatment algorithm. J Am Coll Surg. 2014;218(4):768–774. doi: 10.1016/j.jamcollsurg.2013.12.033. [DOI] [PubMed] [Google Scholar]
- 11.Freeman RK, Herrera A, Ascioti AJ, Dake M, Mahidhara RS. A propensity-matched comparison of cost and outcomes after esophageal stent placement or primary surgical repair for iatrogenic esophageal perforation. J Thorac Cardiovasc Surg. 2015;149(6):1550–1555. doi: 10.1016/j.jtcvs.2015.01.066. [DOI] [PubMed] [Google Scholar]
- 12.McMurry TL, Hu Y, Blackstone EH, Kozower BD. Propensity scores: Methods, considerations, and applications in the Journal of Thoracic and Cardiovascular Surgery. J Thorac Cardiovasc Surg. 2015;150(1):14–19. doi: 10.1016/j.jtcvs.2015.03.057. [DOI] [PubMed] [Google Scholar]
- 13.Biancari F. On the need of standards for reporting on esophageal perforation. G Chir. 33(8–9):254–258. http://www.ncbi.nlm.nih.gov/pubmed/23017283. [PubMed] [Google Scholar]
- 14.Vidarsdottir H, Blondal S, Alfredsson H, Geirsson A, Gudbjartsson T. Oesophageal perforations in Iceland: A whole population study on incidence, aetiology and surgical outcome. Thorac Cardiovasc Surg. 2010;58(8):476–480. doi: 10.1055/s-0030-1250347. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.