Hospital Volume, In-Hospital Mortality, and Failure to Rescue in Esophageal Surgery
An analysis of German hospital discharge data
; ; ; ; ;
Background: In Germany, complex esophageal surgery is often performed in hospitals with low case numbers. For these procedures, an association exists between hospital case numbers and treatment outcomes, possibly because of differences in complication management. This aspect of the association between volume and outcome in esophageal surgery has not yet been studied in Germany.
Methods: On the basis of nationwide hospital discharge data (DRG statistics) from the years 2010 to 2015, the association between volume and outcome was analyzed in relation to in-hospital mortality, the frequency of complications, and the mortality of patients who had complications.
Results: 22 700 cases of complex esophageal surgery were identified. The probability of dying after esophageal surgery was much lower in hospitals with very high case numbers (median, 62 per year) than in those with very low case numbers (median, two per year), with an odds ratio (OR) of 0.50 (95% confidence interval, [0.42; 0.60]). At least one complication was documented for more than half of all patients; no association was found between the frequency of complications and the hospital case volume. The in-hospital mortality among patients who had complications was 12.3% [11.1; 13.7] in hospitals with very high case numbers and 20.0% [18.5; 21.6] in hospitals with very low case numbers. Of the 4032 procedures performed in 2015, 83% were for cancer of the esophagus.
Conclusion: These findings indicate that the quality of care for patients undergoing esophageal surgery in Germany could be improved if more patients were treated in hospitals with high case numbers. The observed association between case numbers and outcomes is tightly linked to failure to rescue.
Operations on the esophagus are highly complex interventions that place exacting demands both on the surgeon and on the whole team involved in pre-, peri-, and postoperative care. It has long been recognized that there is a strong association between the outcome of esophageal surgery and the volume of patients treated. Meta-analyses of studies from different countries have shown that the in-hospital mortality following esophageal interventions to treat malignant neoplasia is inversely related to the number of operations carried out at the institution concerned (1–3). A recently published German study confirmed this association and showed that in this country esophageal procedures often take place in hospitals with very low case volumes (4).
The findings of international research indicate that in-hospital mortality after esophageal surgery depends on how complications are managed, i.e., on the ability of the treatment team to react swiftly and appropriately to complications (5–6). For complex operations this effect seems to account for a large part of the volume–outcome relationship (7–8). To the best of our knowledge, this issue has not yet been investigated for esophageal interventions in Germany.
This retrospective observational study of the situation regarding esophageal surgery in Germany is based on nationwide hospital discharge data and analyzes the volume–outcome relationship in respect of in-hospital deaths, the incidence of complications, and the mortality among patients who develop complications (“failure to rescue”).The findings provide an empirical basis for the assessment of care and the planning of measures to improve care quality.
The diagnosis-related groups (DRG) statistics microdata were accessed remotely in a controlled fashion via the Research Data Center of the German Federal Statistical Office (Destatis) (9). The DRG statistics contain the data records of all inpatient treatments billed according to the DRG system. The information available for each of these treatment cases includes age, sex, diagnoses (coded according to the International Statistical Classification of Diseases and Related Health Problems, 10th revision, German modification, ICD-10-GM), procedures (coded according to the German Procedure Classification, OPS), length of hospital stay, and reason for discharge. The hospital where the treatment was performed can be identified by means of the anonymized institutional code number contained in every data record. Data from the period 2010 to 2015 were included.
The units of analysis were treatment cases in which adult patients had undergone complex surgery on the esophagus. The treatment cases were identified by means of the OPS codes. Total and partial esophageal resections, cases of gastrectomy involving subtotal esophageal resection, and complex reconstructions of the esophageal lumen were all included. This definition corresponds to the minimum volume regulations of the Federal Joint Committee (G-BA)—with the exception of the codes, added in 2012, for implantation or exchange of a magnetic antireflux system. On the basis of the documented principal and secondary diagnoses, the treatment cases were divided into indication groups according to underlying disease (eTable 1).
Volume of cases at treating hospital
In order to obtain the volume of complex esophageal interventions carried out at each hospital, the treatment cases were aggregated at the level of institutional code number and year. The treatment cases were then sorted according to the volume of their treating hospital and classified into five groups with similar numbers of cases (volume quintiles). The volume quintile categorized the relative volume of cases of the respective treating hospital in each treatment year (ranging from very low to very high). The absolute annual number of cases of the treating hospital was also regarded as a continuous variable in a separate analysis. In the course of manuscript revision, a post-hoc analysis was requested with the treatment cases divided into hospital-equivalent tertiles.
The primary treatment outcome was risk-adjusted in-hospital mortality, and the secondary treatment outcomes were the indicators for complications and the in-hospital mortality among patients in whom such a complication was documented (“failure to rescue”) (10). The complications were categorized as surgical (e.g., endoscopic intervention as a sign of anastomotic insufficiency) (6), septic (e.g., sepsis as secondary diagnosis), and cardiovascular (e.g., acute myocardial infarction as secondary diagnosis) (eTable 2). Since a single treatment case could involve more than one complication, various types were summarized, in that treatment cases with at least one type of complication were counted.
The characteristics of the treatment cases were first analyzed descriptively for each observation year, then the cumulative data were evaluated for the whole observation period.
In-hospital mortality was initially stratified according to volume quintile. We calculated not only the observed mortality but also the risk-adjusted mortality, by means of logistic regression procedures. To this end we used generalized estimation equations that took account of the distribution of patients over the hospitals. The regression models included age, sex, calendar year of treatment, and underlying disease. In addition, certain comorbidities were included that were very likely to have been present before the hospital stay (e.g., heart failure, chronic lung disease) and are possible risk factors for death in the hospital (6, 11) (eTable 3). For each treatment case, the expected risk of death was calculated and added up over the volume quintiles. The risk-adjusted mortality is the ratio of observed to expected mortality multiplied by the average mortality risk in the observation period.
To define more closely the association of the case volume at the treating hospital with in-hospital mortality, we calculated risk-adjusted odds ratios (OR) for the whole observation period. The OR estimated the risk of dying in the hospital depending on the case volume at the treating institution. To this end, the case volume of the treating hospital was added to the above-mentioned regression models, on one hand as volume quintile and on the other hand as a continuous variable. Based on a method described in detail elsewhere (4, 12), we calculated a threshold value for the annual volume per hospital from which the risk of death was below the national average.
The incidence of complications and the observed in-hospital mortality of patients with complications (“failure to rescue”) were stratified by volume quintile.
The statistical significance of the results was assessed on the basis of 95% confidence intervals (CI). All calculations were made using SAS Version 9.3.
Characteristics of treatment cases over time
The number of treatment cases involving a complex esophageal surgery rose from 3598 in 2010 to 4032 in 2015 (Table 1). The number of hospitals in which such surgeries were performed varied over time and was 406 in 2015. The median number of complex esophageal interventions per hospital and year rose from five to six during the observation period. In other words, in the year 2015 no more than six such operations took place in half of the hospitals where esophageal surgery was practiced.
Malignant neoplasia constituted the most frequently coded underlying illness. In 2015 (2010), malignant neoplasia was coded in 83% (81%) of treatment cases, and in a further 9% (11%) the coded diagnosis was malignant neoplasia of a neighboring organ or secondary malignant neoplasia.
The proportion of total esophageal resections varied between 48% and 52%. Within this group, the rate of total resections with abdominocervical access decreased from 13% in 2010 to 6% in 2015.
Characteristics of treatment cases by volume quintile
After classification of the treatment cases according to the volume of cases at the treating hospital in the year concerned, each volume quintile contained more than 4400 treatment cases for analysis (Table 2). In the quintile with very low numbers of cases, 260 hospitals carried out a median of only two complex operations on the esophagus annually. In the quintile with very high numbers of cases, 12 hospitals performed a median of 62 such operations each year. The results of the post-hoc analysis according to hospital-equivalent tertiles can be found in eTable 5.
The proportion of patients aged 65 years or over was higher in hospitals with a low volume (57% in the lowest volume quintile) than in those with a high volume (46% in the highest quintile).
The rate of malignant neoplasia of the esophagus as underlying disease was lower in the lowest volume quintile than in the highest quintile (74% versus 89%), while malignant neoplasia of neighboring organs was coded more frequently (16% versus 5%).
As for the type of surgery, the proportion of operations made up by total esophageal resection was lower in the lowest volume quintile (30%) than in the highest quintile (70%), and the proportion of partial resections was higher (59% versus 23%). Gastrectomy with subtotal esophageal resection was carried out in 10% of the patients in the bottom volume quintile and in 5% of those in the top quintile.
With regard to total esophageal resection, the proportion of surgeries with abdominocervical access was higher in hospitals with low volumes than in those with high volumes (bottom quintile 13% versus top quintile 6%).
Figure 1 shows the observed and risk-adjusted in-hospital mortality stratified by volume quintile. The observed in-hospital mortality was 12.2% in the quintile of hospitals with very low volumes and 6.8% in the quintile with very high volumes. Risk adjustment narrowed this gap only slightly (to 11.4% versus 6.8%). Post-hoc analysis by hospital-equivalent tertile confirmed these results (eFigure).
Figure 2 shows the risk-adjusted OR. The risk of dying in hospital was around 18% lower (OR 0.82; 95% CI [0.70; 0.95]) in hospitals with low volumes than in hospitals with very low volumes. The reduction in hospitals with intermediate volumes did not attain statistical significance (OR 0.86 [0.74; 1.00]). In hospitals with high volumes the estimated reduction in mortality was 23% (OR 0.77 [0.65; 0.90]), and in those with very high volumes it was circa 50% (OR 0.50 [0.50 [0.42; 0.60]).
When the annual case volume was considered as a continuous variable, each additional case per hospital and year yielded a reduction in the likelihood of dying in hospital of circa 1% (OR 0.99 [0.99; 0.99]). With 10 additional cases per hospital and year, the risk of mortality went down by about 8% (OR 0.92 [0.90; 0.94]) (Figure 2).
The threshold value from which in-hospital mortality below the national average of 9.5% in the observation period could be expected was 26 treatment cases [21; 32]. In 2015, 26 of the 406 hospitals (6%) reached this threshold (data not shown).
Incidence of complications, in-hospital mortality among patients with complications
Overall, more than 50% of the patients who underwent complex esophageal surgery experienced at least one complication (Figure 3). There was no trend in incidence of complications along the volume quintiles. In contrast, the in-hospital mortality among patients with one or more complications was 20% [18.5; 21.6] in hospitals with very low volumes and decreased quintile by quintile with increasing volumes. In hospitals with very high volumes, 12. 3% [11.1; 13.7] of patients with complications died.
The incidence and associated mortality of the types of complications depicted in Figure 3 are listed separately in eTable 4.
In each volume quintile, at least one complication was documented in more than 90% of patients who died in hospital after a complex esophageal operation (eTable 4).
In the year 2015 there were still more than 200 hospitals in Germany with an annual volume of six or fewer complex operations on the esophagus. In agreement with earlier investigations (13), we observed only a weak trend towards centralization over the period 2010 to 2015.
Analysis of the in-hospital mortality in relation to the treating hospital showed a clear association between case volume and outcome. The greatest reduction of mortality was seen in the highest volume quintile; in the middle quintiles, the results did not vary significantly. In the 12 hospitals with a median of 62 cases per year, which treated one fifth of all patients in the study, mortality was half that in the 260 hospitals with a median of two cases annually, which also treated one fifth of the patients. A recently published study investigated the volume–outcome association in esophageal surgery to treat esophageal cancer (4). In contrast, our study included all types of esophageal surgery, while the risk adjustment took account of the different risks of death associated with different underlying diseases. The results of the two analyses are almost identical with regard to the association between volume and outcome. Esophageal cancer is the surgical indication for more than 80% of such operations.
Based on discharge data from all hospitals in Germany and including all indications, the study presented here estimated that the threshold above which mortality can be expected to be lower than the national average is 26 complex esophageal surgeries per year. In 2015, the most recent year analyzed, only 6% of hospitals attained this threshold. The statutory minimum volume is currently 10 cases per year, but a large proportion of hospitals do not achieve that figure (14, 15). In the Netherlands the minimum volume was raised to 20 in 2011 (16). The in-hospital mortality following complex esophageal surgery is higher in Germany than in countries such as the Netherlands, Great Britain, and Canada where these operations have been centralized (17–19).
Our analysis showed that the incidence of complications is not associated with the volume of cases. This finding has also been reported in studies from other countries (6, 7). Because the complexity of esophageal surgery means that they are associated with a host of perioperative risks, complications cannot always be avoided (20). However, the variation in survival of patients with complications shows that their management is an essential factor in explaining the volume–outcome association in complex esophageal surgery. Comparison with the mortality in the uppermost quintile shows that around 80% of the additional deaths in the lowest quintile can be attributed to “failure to rescue.” These findings correspond with international analyses describing the “failure to rescue” effect for various complex operations (5, 7–8, 21–23).
The strength of this study lies in its use of comprehensive DRG billing data from all German hospitals. This permits depiction of the true situation with regard to complex esophageal operations and their complications, including institutions with very low case numbers. It should be pointed out, however, that the DRG data for billing purposes are documented by the individual hospitals themselves, so differences in the incidence of the characteristics analyzed may be caused by variations in coding practice. Moreover, the use of the institutional code number to distinguish the hospitals may have led to a degree of inaccuracy in determining the case volumes. Since the institutional code number does not necessarily precisely identify the location of the hospital concerned, it is possible that two or more hospitals bill under the same code number (24). This could result in overestimation of the volume of cases in some hospitals, thus introducing distortion of the differences in in-hospital mortality towards the null hypothesis. It must be remembered that no causal relationship between volume and outcome can be proved by retrospective observational studies. Although it is probable that the case volume of a hospital is correlated with specific factors that determine the treatment quality (e.g., the surgeon’s experience, the density of nursing staff, or certain structural characteristics), these details are not contained in the data we used. The patients’ social status, which may yield pointers to possible inequalities of access to care, is also not included in the data and thus could not be analyzed.
The findings of our study indicate that the care quality of esophageal surgery in Germany could be improved if more patients were treated in hospitals with high case volumes. The observed volume–outcome association appears to depend essentially on the management of complications. It seems plausible that greater experience of highly complex interventions may lead to timely recognition of complications by physicians, nurses, and other members of the treatment team. Furthermore, hospitals with high volumes are generally equipped and structured in a way that enables the staff to react early and appropriately to complications, however serious.
The results of this study should be used to place the care provided in German hospitals on an evidential basis. Our findings should also be made known to patients and referring physicians, who will then be able to make better-informed decisions should esophageal surgery be planned.
Conflict of interest statement
Dr. Nimptsch and Prof. Mansky work in the Department of Structural Development and Quality Management in the Health System. This department is a chair endowed by Helios Kliniken GmbH.
The remaining authors declare that no conflict of interest exists.
Manuscript submitted on 20 March 2018, revised version accepted on
9 August 2018
Translated from the original German by David Roseveare
Dr. PH Ulrike Nimptsch
Fachgebiet Strukturentwicklung und Qualitätsmanagement
Technische Universität Berlin
10587 Berlin, Germany
For eReferences please refer to:
Department of General and Visceral Surgery, Sana Hospital Offenbach GmbH, Offenbach am Main: Dr. Thomas Haist
Department of Surgery, University Hospital Erlangen: Dr. Christian Krautz, Prof. Robert Grützmann, MBA
General, Visceral and Thoracic Surgery, Darmstadt Hospital GmbH, Darmstadt: Prof. Dietmar Lorenz
|1.||Markar SR, Karthikesalingam A, Thrumurthy S, Low DE: Volume–outcome relationship in surgery for esophageal malignancy: systematic review and meta-analysis 2000–2011. J Gastrointest Surg 2012; 16: 1055–63 CrossRef MEDLINE|
|2.||Wouters MW, Gooiker GA, van Sandick JW, Tollenaar RA: The volume–outcome relation in the surgical treatment of esophageal cancer: a systematic review and meta-analysis. Cancer 2012; 118: 1754–63 CrossRef MEDLINE|
|3.||Metzger R, Bollschweiler E, Vallböhmer D, Maish M, DeMeester TR, Hölscher AH: High volume centers for esophagectomy: what is the number needed to achieve low postoperative mortality? Dis Esophagus 2004; 17: 310–4 CrossRef MEDLINE|
|4.||Nimptsch U, Mansky T: Hospital volume and mortality for 25 types of inpatient treatment in German hospitals: observational study using complete national data from 2009 to 2014. BMJ Open 2017; 7: e016184 CrossRef MEDLINE PubMed Central|
|5.||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: 1962–9 CrossRef MEDLINE|
|6.||Weledji EP, Verla V: Failure to rescue patients from early critical complications of oesophagogastric cancer surgery. Ann Med Surg (Lond) 2016; 7: 34–41 CrossRef MEDLINE PubMed Central|
|7.||Ghaferi AA, Birkmeyer JD, Dimick JB: Hospital volume and failure to rescue with high-risk surgery. Med Care 2011; 49: 1076–81 CrossRef MEDLINE|
|8.||Krautz C, Nimptsch U, Weber GF, Mansky T, Grützmann R: Effect of hospital volume on in-hospital morbidity and mortality following pancreatic surgery in Germany. Ann Surg 2018; 267: 411–7 CrossRef MEDLINE|
|9.||Forschungsdatenzentren der statistischen Ämter des Bundes und der Länder: Datenangebot. Fallpauschalenbezogene Krankenhausstatistik (DRG-Statistik). Wiesbaden: Forschungsdatenzentren der statistischen Ämter des Bundes und der Länder. www.forschungsdatenzentren.de/bestand/drg/index.asp (last accessed on 15 December 2017).|
|10.||Silber JH, Romano PS, Rosen AK, Wang Y, Even-Shoshan O, Volpp KG: Failure-to-rescue: comparing definitions to measure quality of care. Med Care 2007; 45: 918–25 CrossRef MEDLINE|
|11.||Steyerberg EW, Neville BA, Koppert LB, et al.: Surgical mortality in patients with esophageal cancer: development and validation of a simple risk score. J Clin Oncol 2006; 24: 4277–84 CrossRef MEDLINE|
|12.||Bender R: Quantitative risk assessment in epidemiological studies. Investigating threshold effects. Biometr J 1999; 41: 305–19 CrossRef|
|13.||de Cruppé W, Malik M, Geraedts M: Minimum volume standards in German hospitals: do they get along with procedure centralization? A retrospective longitudinal data analysis. BMC Health Serv Res 2015; 15: 279 CrossRef MEDLINE PubMed Central|
|14.||Nimptsch U, Peschke D, Mansky T: [Minimum caseload requirements and in-hospital mortality: observational study using nationwide hospital discharge data from 2006 to 2013]. Gesundheitswesen 2017; 79: 823–34 CrossRef MEDLINE|
|15.||de Cruppé W, Geraedts M: Mindestmengen unterschreiten, Ausnahmetatbestände und ihre Konsequenzen ab 2018. Komplexe Eingriffe am Ösophagus und Pankreas in deutschen Krankenhäusern im Zeitverlauf von 2006 bis 2014. Zentralbl Chir 2018; 143: 250–8 CrossRef MEDLINE|
|16.||Busweiler LA, Wijnhoven BP, van Berge Henegouwen MI, et al.: Early outcomes from the Dutch Upper Gastrointestinal Cancer Audit. Br J Surg 2016; 103: 1855–63 CrossRef MEDLINE|
|17.||Wouters MW, Karim-Kos HE, le Cessie S, et al.: Centralization of esophageal cancer surgery: does it improve clinical outcome? Ann Surg Oncol 2009; 16: 1789–98 CrossRef MEDLINE PubMed Central|
|18.||Finley CJ, Jacks L, Keshavjee S, Darling G: The effect of regionalization on outcome in esophagectomy: a Canadian national study. Ann Thorac Surg 2011; 92: 485–90 CrossRef MEDLINE|
|19.||Munasinghe A, Markar SR, Mamidanna R, et al.: Is It time to centralize high-risk cancer care in the United States? Comparison of outcomes of esophagectomy between England and the United States. Ann Surg 2015; 262: 79–85 CrossRef MEDLINE|
|20.||Gockel I, Ahlbrand CJ, Arras M, Schreiber EM, Lang H: Quality management and key performance indicators in oncologic esophageal surgery. Dig Dis Sci 2015; 60: 3536–44 CrossRef MEDLINE|
|21.||Nieman CL, Stewart CM, Eisele DW, Pronovost PJ, Gourin CG: Frailty, hospital volume, and failure to rescue after head and neck cancer surgery. Laryngoscope 2017; 128: 1365–70 CrossRef MEDLINE|
|22.||Spolverato G, Ejaz A, Hyder O, Kim Y, Pawlik TM: Failure to rescue as a source of variation in hospital mortality after hepatic surgery. Br J Surg 2014; 101: 836–46 CrossRef MEDLINE|
|23.||Johnston MJ, Arora S, King D, et al.: A systematic review to identify the factors that affect failure to rescue and escalation of care in surgery. Surgery 2015; 157: 752–63 CrossRef MEDLINE|
|24.||Nimptsch U, Wengler A, Mansky T: Kontinuität der Institutionskennzeichen in Krankenhausabrechnungsdaten – Analyse der bundesweiten DRG-Statistik von 2005 bis 2013. Z Evid Fortbild Qual Gesundhwes 2016; 117: 38–44 CrossRef MEDLINE|