Abstract
Purpose
The association between a high annual hospital-specific case volume and improved postoperative outcomes after oesophageal cancer resection (OCR) is well described. The Swiss government started centralisation of OCR in 2013. This study investigates this centralisation in terms of the annual hospital-specific case volume of OCR (hospital volume) and its effect on short-term postoperative outcomes.
Methods
National inpatient registry data of all hospitals providing OCR (ICD codes C15 and at least one of the CHOP codes starting with 424, 425, or 426) between 2013 and 2022 in Switzerland were analysed. The primary endpoint was in-hospital mortality. Secondary endpoints were postoperative complications and length of hospital stay (LOS).
Results
1’535 cases were identified. The annual number of hospitals providing OCR ranged between 30 in 2013 and 14 in 2021. Hospital volume varied between one and 40 patients per year. Overall mortality was 5.3% (n = 81) declining from 9.3% in 2013 to 3.8% in 2022 (t=-0.511, p = 0.040). Increasing hospital volume was associated with lower risk for mortality (odds ratio (OR) = 0.723, 95% confidence interval (CI) 0.53 to 0.98, p = 0.037) and postoperative complications (OR = 0.705, 95%CI: 0.62 to 0.83, p < 0.001). There was no association of hospital volume with LOS (β = 0.007, 95%CI -0.093 to 0.108, p = 0.889). However, there were significant differences across quantiles (joint p < 0.001).
Conclusions
Higher hospital volume is associated with lower mortality and fewer postoperative complications following OCR, whereas no association was observed with LOS.
Clinical trial registration
ClinicalTrials.gov, NCT07022652, https://clinicaltrials.gov/search?id=NCT07022652.
Supplementary Information
The online version contains supplementary material available at https://doi.org/10.1007/s00423-026-04125-z.
Keywords: Oesophageal cancer, Surgery, Hospital volume, Mortality, Postoperative complications
Introduction
The effect of centralisation on postoperative outcomes after oesophageal cancer resection (OCR) in Switzerland has not been highlighted yet. Oesophageal cancer is a rare malignancy in Switzerland. It accounts for only 2% of all cancers, with an annual incidence of 603 cases between 2017 and 2021 [1]. The prognosis of oesophageal cancer is poor. The 437 annual oesophageal cancer-related deaths between 2017 and 2021 resulted in a low 5-year overall survival of only 30% [2]. The therapeutic management of oesophageal cancer (neoadjuvant therapy, interventional therapy, surgical technique, adjuvant therapy) was driven by innovation and changed considerably in recent years, resulting in a complex therapy algorithm [3–7]. Together with the often polymorbid patients, the very demanding surgical technique, and the challenges in postoperative management, a specialised, high-volume, interdisciplinary treatment team is required [8, 9]. The centralisation of such complex surgical procedures in specialised centres aims to build up such treatment teams, resulting in improved quality of the treatment and increased patient safety [10]. The association between a high annual hospital-specific case load of OCR (hospital volume) and improved short-term postoperative outcomes as well as long-term survival has been well documented in current scientific literature [11, 12]. International benchmark centres suggest a threshold between 20 and 40 for high-volume centres [13, 14]. A threshold value of 12 was introduced for the first time in 2013 in Switzerland [15]. The threshold and other regulations resulted in only eight hospitals with a performance mandate for OCR [16]. Between 2001 and 2012, 69 different hospitals performed OCR in Switzerland, and the in-hospital mortality of 5.9% was significantly higher compared to international benchmark centres from ESODATA (2.0%) or DUCA (2.5% to 4.4%) [13, 14, 17–19].
This retrospective registry study provides Swiss national data on short-term postoperative outcomes after elective OCR in Switzerland from 2013 to 2022. It aims to examine the effect of the ongoing centralisation through the association between the hospital volume and in-hospital mortality, postoperative complications, and length of hospital stay (LOS).
Material and methods
Study design
This is a retrospective population-based analysis from the Swiss national inpatient registry between 2013 and 2022. The anonymised data was obtained from the Swiss Federal Statistics Office (FSO) via a data contract for the one-time delivery and use of individual data (without personal identification) in Switzerland. At the time of ordering the data in 2024, the data from 2023 was not yet available. The data contained detailed information from all Swiss hospitals’ inpatient cases between 2013 and 2022 as described in the supplementary material.
Case selection
The cases were selected according to their “International Statistical Classification of Diseases and Related Health Problems 10th Revision” (ICD) and “Swiss Classification of Operations” (CHOP) codes [20, 21]. A case was selected for OCR when at least one of the ICD codes started with C15 (malignant neoplasm of the oesophagus [oesophageal cancer] and at least one of the CHOP codes started with 42.4 (excision of the oesophagus), 42.5 (intrathoracic anastomosis on the oesophagus), or 42.6 (in front of the sternum oesophagostomy). An ICD based surrogate was used to construct the “Union Internationale Contre le Cancer” (UICC) stage: Stage I/II (all other), III (ICD code C77), IV (ICD code C78 or C79) [22].
Exclusion criteria included emergency admission, incomplete case documentation, age under 20 years, and hospitalisation in non-general or non-surgical hospitals. Incomplete documentation occurs due to the strictly full calendar year (from 1 st January to 31 st December) based classification of the FSO. Data from patients hospitalised overnight on the 31 st of December are therefore incomplete [17]. The inclusion of 18- and 19-year-olds was not possible due to the data structure with five-year age groups. Other hospitals than general and surgical hospitals were excluded because of the low levels of comorbidities and the relatively low risk of dying in the hospital [23].
Primary and secondary endpoints
Hospital volume was defined as the annual hospital-specific case load of OCR. Each case was assigned to the hospital volume of the hospital per corresponding calendar year. Like this, cases at high-volume hospitals contribute more (weighting).
The primary endpoint of this study was postoperative mortality, defined as in-hospital mortality.
Secondary endpoints were postoperative complications and length of hospital stay (LOS). The analysed postoperative complications comply with ESODATA specifications and are shown in Table 3 [13].
Table 3.
Complications
| Variable | Label | Total n=1’535 | |
|---|---|---|---|
| Complications overall | Yes | 1063 (69.3%) | |
| Reoperation° | 00.99.10 | 136 (8.9%) | |
| Myocardial infarction* | I21, I22 | 5 (0.3%) | |
| Cardiac arrest* | I46 | 22 (1.4%) | |
| Arterial fibrillation* | I48 | 353 (23.0%) | |
| DVT and PE* | I26.0, I26.9, I80.2 | 58 (3.8%) | |
| Pneumonia* | J13-18, J85.1 | 243 (15.8%) | |
| Pleural effusion* | J90, J91 | 368 (24.0%) | |
| Pleural empyema* | J86.0, J86.9 | 112 (7.3%) | |
| Aspiration* | J69.0, J69.8, J95.4 | 124 (8.1%) | |
| Atelectasis* | J98.1 | 137 (8.9%) | |
| Mediastinitis* | J85.3, J98.5 | 66 (4.3%) | |
| Pneumothorax* | J93, J95.80, S27.0 | 166 (10.8%) | |
| Haemothorax* | J94.2, S27.1, S27.2 | 9 (0.6%) | |
| Pulmonal insufficiency, respiratory failure* | J95.1, J95.2, J95.3, J96 | 237 (15.4%) | |
| ARDS* | J80 | 37 (2.4%) | |
| Acute kidney injury* | N17, N19 | 116 (7.6%) | |
| Urinary tract infection* | N10, N30.0, N30.9, N39.0 | 52 (3.4%) | |
| Urinary retention* | R33 | 41 (2.7%) | |
| Anastomotic leakage * | 2013 and 2014 | No code available | 368 (24.0%) |
| 2015 to 2022 | Code available | 1’167 (76.0%) | |
| No leakage | 926 (79.3%)ꜝ | ||
| K91.83 | 241 (20.7%)ꜝ | ||
| Anastomotic stenosis* | K22.2 | 69 (4.5%) | |
| Gastrointestinal ulcer* | K25, K28 | 29 (1.9%) | |
| Peritonitis and intraabdominal abscess* | K65 | 20 (1.3%) | |
| Delayed gastric emptying* | K31.88 | 50 (3.3%) | |
| Ileus* | K56, K91.3 | 55 (3.6%) | |
| Bleeding* | T81.0, S36.81 | 88 (5.7%) | |
| Shock* | T81.1 | 5 (0.3%) | |
| Lymphatic fistula* | I89.8, J94.0 | 58 (3.8%) | |
| Wound dehiscence* | T81.3 | 64 (4.2%) | |
| SSI* | T81.4, T89.02, L08.9 | 108 (7.0%) | |
| Other vascular complication* | T81.7 | 1 (0.1%) | |
| Injury to the heart* | S26.0 | 2 (0.1%) | |
| Injury to the lungs or pleura* | S27.0, S21.3, S27.6 | 39 (2.5%) | |
| Tracheobronchial injury* | S27.4, S27.5 | 10 (0.7%) | |
| Injury to the thoracic duct* | S27.82 | 5 (0.3%) | |
| Injury to the diaphragm* | S27.81 | 2 (0.1%) | |
| Injury to other or multiple thoracal organs* | S27.7, S27.83, S27.84, S27.88, S27.9 | 3 (0.2%) | |
| Recurrent laryngeal nerve palsy* | J38, G52.2 | 61 (4.0%) | |
| Other intraoperative injuries* | T81.2 | 15 (1.0%) | |
| Intraoperative retained foreign body* | T81.5, T81.6 | 1 (0.1%) | |
| Other complications* | J95.8, J95.9, K91.88, K91.9, T81.8, T81.9 | 200 (13.0%) | |
°: Swiss Classification of Operations” (CHOP) codes
*: International Statistical Classification of Diseases and Related Health Problems 10th Revision” (ICD)
ꜝ: Reference operations between 2015 and 2022
SD: Standard deviation; DVT: Deep venous thrombosis; PE: Pulmonary embolism; ARDS: Acute respiratory distress syndrome; n/a: Not available; SSI: Surgical site infection
Additional objectives were sociodemographic trends (number of operations, number of hospitals providing the operation, mean hospital volume, age, Charlson comorbidity index (CCI), sex, intensive care unit (ICU) stay, nationality, insurance, robotic access (CHOP code 00.99.50)) between 2013 and 2022 [24]. The CCI and ICU stay were assumed as described in the supplementary material.
Statistical analysis
The R statistical software was used for statistical analysis (www.r-project.org). A two-sided p-value of less than 0.05 was considered statistically significant. Continuous data were expressed as the mean with standard deviation (SD), interquartile range (IQR) and the range. In descriptive analysis, proportions were compared with the chi-square statistics, and continuous variables were compared with t tests and Mann‒Whitney U tests, as appropriate.
The associations between the hospital volume and mortality and postoperative complications were assessed by generalized additive regression models (GAM) adjusting for year of operation with random effect B-splines. Hospital volume entered the regression models as a continuous variable per 10 cases. The confounding variable set included sex (male vs. female), age (continuous), CCI (continuous), nationality (foreign vs. Swiss), insurance (general vs. private (semi-private and private)), hospital type (university vs. tertiary vs. other (basic hospitals and specialised surgical hospitals)). Results were additionally presented graphically with an easy-to-interpret marginal effect analysis as described in the supplementary material [25, 26]. Regarding mortality, an additional failure-to-rescue analysis was carried out, excluding all patients without a complication. For the LOS, an ordinary least square regression was complemented by an in-depth analysis by quantile regression analysis to assess the associations between the hospital volume and the different lengths of the LOS, not disturbed by outliers or skewed data [27]. The Khmaladze test was used to evaluate whether the quantile regression coefficients are constant across quantiles (location shift hypothesis) [28]. The quantile regression analysis is described in detail in the supplementary analysis.
The time trends were assessed using the nonparametric Mann-Kendall trend test from the R trend library [29]. Its value is between − 1 (perfect descending trend) and + 1 (perfect ascending trend) with 0 representing no trend at all [30].
Ethics
This study was approved by the Ethics Committee of Eastern Switzerland (BASEC B2025-00702), registered under ClinicalTrials.gov (NCT07022652), and is compliant with the Strengthening the Reporting of Observation in Epidemiology (STROBE) guidelines [31]. It conforms to the provisions of the Declaration of Helsinki. There were no deviations from the published protocol.
Results
Case selection
The dataset provided by the FSO contained data of 14’438’938 cases. 22’837 cases had an oesophageal cancer diagnosis. 1’595 cases had an OCR. Fifty-nine cases were excluded due to emergency admission, and one case was excluded due to incomplete case documentation. 1’535 cases were included in the analysis (Fig. 1).
Fig. 1.
Case selection
Case baseline characteristics
The case baseline characteristics are shown in Table 1. The OCRs were predominantly performed at university and tertiary hospitals (39.4% and 57.4%) on men (78.4%) with a mean age of 64.7 years. Only 57.8% of the cases were treated at institutions with a hospital volume exceeding 12 cases (Fig. 2).
Table 1.
Case baseline characteristics and treatment
| Variable | Label | Total n=1’535 |
|---|---|---|
| Age (years) | Mean (SD) | 64.7 (9.7) |
| Range | 22.0-87.0 | |
| Sex | Male | 1203 (78.4%) |
| Female | 332 (21.6%) | |
| Nationality | Foreign | 177 (11.5%) |
| Swiss | 1358 (88.5%) | |
| Insurance | General | 1060 (69.1%) |
| Private | 475 (30.9%) | |
| CCI | Mean (SD) | 3.8 (2.0) |
| Range | 2.0-11.0 | |
| ICD based UICC stage surrogate | I/II | 1024 (66.7%) |
| III | 457 (29.8%) | |
| IV | 54 (3.5%) | |
| Type of hospital | University | 605 (39.4%) |
| Tertiary | 881 (57.4%) | |
| Other | 49 (3.2%) | |
| Hospital volume | Mean (SD) | 13.7 (9.2) |
| Range | 1.0-40.0 | |
| > 12.0 (50th percentile) | 648 (42.2%) | |
| > 18.0 (75th percentile) | 359 (23.4%) | |
| > 26.0 (90th percentile) | 141 (9.2%) | |
| Robotic access | Yes | 113 (7.4%) |
| ICU stay | Yes | 1350 (87.9%) |
| Hours on ICU | Mean (SD) | 145.8 (260.8) |
| Range | 3.0-5026.0 |
SD: Standard deviation; CCI: Charlson comorbidity index; ICD: International Statistical Classification of Diseases and Related Health Problems 10th Revision; UICC: Union Internationale Contre le Cancer; ICU: Intensive care unit
Fig. 2.
Histogram of frequency of oesophageal cancer resection according to the hospital volume. X-axis: year of operation. Y-axis: frequency of observations. (color should be used)
Mortality and failure-to-rescue
Overall mortality was 5.3% (n = 81). A higher hospital volume was associated with 27.7% lower odds for mortality for every ten additional cases (odds ratio (OR) = 0.723, 95% confidence interval (CI) 0.534 to 0.981, p = 0.037) (Fig. 3). According to the GAM logistic regression, higher age (OR = 1.028, 95%CI 1.002 to 1.055, p = 0.033) and higher CCI (OR = 1.218, 95%CI 1.095 to 1.355, p < 0.001) had higher odds for mortality. Year of operation was a significant random effect (p = 0.013) (Table 2). The failure-to-rescue analysis (n = 1’063) confirmed the association between mortality and hospital volume, with 19.0% lower odds for mortality for every ten additional cases; however, this did not reach significance (OR = 0.810, 95%CI 0.593 to 1.106, p = 0.184).
Fig. 3.
Marginal effect analysis of hospital volume and mortality. Red line: GAM logistic regression. Red shaded area: 95% confidence interval of the GAM logistic regression
Table 2.
GAM logistic regression for mortality
| Variable | Label | OR | 95%CI | p-value |
|---|---|---|---|---|
| Hospital volume | continuous | 0.723 | 0.534 to 0.981 | 0.037 |
| Age | continuous | 1.028 | 1.002 to 1.055 | 0.033 |
| Sex | Male | reference | - | - |
| Female | 0.973 | 0.557 to 1.701 | 0.924 | |
| CCI | continuous | 1.218 | 1.095 to 1.355 | <0.001 |
| Nationality | Foreign | reference | - | - |
| Swiss | 0.917 | 0.439 to 1.915 | 0.817 | |
| Insurance | General | reference | - | - |
| Private | 0.865 | 0.521 to 1.436 | 0.575 | |
| Type of hospital | University | reference | - | - |
| Tertiary | 0.833 | 0.487 to 1.425 | 0.504 | |
| Other | 0.752 | 0.196 to 2.890 | 0.678 | |
| Year of operation | Random effect | - | - | 0.013 |
Significant values are bold
OR: Odds ratio; CI: Confidence interval; CCI: Charlson comorbidity index
Postoperative complications
The complications are shown in Table 3. The overall complication rate was 69.3% (n = 1’063). A higher hospital volume was associated with 29.5% lower odds for overall complications for every ten additional cases (OR = 0.705, 95%CI 0.608 to 0.816, p < 0.001) (Fig. 4). According to the GAM logistic regression, higher age (OR = 1.022, 95%CI 1.010 to 1.034, p < 0.001) had higher odds while tertiary hospitals (OR = 0.749, 95%CI 0.572 to 0.980, p = 0.035) had lower odds for overall complications (Table 4).
Fig. 4.
Marginal effect analysis of hospital volume and postoperative complications. Red line: GAM logistic regression. Red shaded area: 95% confidence interval of the GAM logistic regression
Table 4.
GAM logistic regression for overall complication
| Variable | Label | OR | 95%CI | p-value |
|---|---|---|---|---|
| Hospital volume | continuous | 0.705 | 0.608 to 0.816 | <0.001 |
| Age | continuous | 1.022 | 1.010 to 1.034 | <0.001 |
| Sex | Male | Ref | - | - |
| Female | 1.283 | 0.972 to 1.692 | 0.078 | |
| CCI | continuous | 1.029 | 0.973 to 1.087 | 0.318 |
| Nationality | Foreign | Ref | - | - |
| Swiss | 0.728 | 0.506 to 1.047 | 0.087 | |
| Insurance | General | Ref | - | - |
| Private | 1.164 | 0.905 to 1.499 | 0.237 | |
| Type of hospital | University | Ref | - | - |
| Tertiary | 0.749 | 0.572 to 0.980 | 0.035 | |
| Other | 0.676 | 0.321 to 1.421 | 0.301 | |
| Year of operation | Random effect | - | - | 0.107 |
Significant values are bold
OR: Odds ratio; CI: Confidence interval; CCI: Charlson comorbidity index
The overall anastomotic leakage rate between 2015 and 2022 was 20.7% (n = 241). A higher hospital volume was not associated with lower odds for anastomotic leakage (OR = 0.861, 95%CI 0.682 to 1.086, p = 0.207). However, the occurrence of anastomotic leakage was strongly associated with mortality (OR 3.758, 95%CI 1.614 to 8.753, p = 0.002). The associations between all other complications and hospital volume and mortality are shown in the supplementary material table SM1.
Length of hospital stay
The mean LOS was 24.8 days (SD 18.5). In ordinary regression, no association between LOS and hospital volume was observed (β = 0.007, 95%CI -0.093 to 0.108, p = 0.889) (Fig. 5). The effect of hospital volume on LOS varied significantly across quantiles, with increasingly negative but individually non-significant coefficients at higher quantiles (joint p < 0.001) (Table 5).
Fig. 5.
Quantile regression for hospital volume and length of hospital stay (LOS). X-axis: 0.1 to 0.9 quantile of LOS. Y-axis: quantile regression coefficient of hospital volume. Black dots: regression coefficients. Red dashed line: β of the ordinary regression
Table 5.
Quantile regression for length of hospital stay
| Quantile | LOS (days) | Coefficient of hospital volume (95% CI) |
|---|---|---|
| 0.1 | 12 | 0.057 (-0.120 to 0.094) |
| 0.2 | 14 | 0.000 (0.000 to 0.107) |
| 0.3 | 15.2 | 0.056 (-0.036 to 0.093) |
| 0.4 | 17 | 0.000 (0.000 to 0.082) |
| 0.5 | 19 | 0.000 (0.000 to 0.083) |
| 0.6 | 22 | -0.028 (-0.088 to 0.160) |
| 0.7 | 25 | -0.000 (-0.123 to 0.074) |
| 0.8 | 31 | -0.065 (-0.177 to 0.081) |
| 0.9 | 43 | -0.182 (-0.451 to 0.100) |
LOS: Length of hospital stay; CI: Confidence interval
Trend analysis between 2013 and 2022
Trend analysis is shown in Table 6. The annual number of hospitals providing OCR decrease from 30 in 2013 to 18 in 2022 (tau=-0.750, p = 0.003) with its lowest number of 14 hospitals in 2021. The number of annual OCR also decreased from 183 in 2013 to 133 in 2022 (tau − 0.733, p = 0.003) with its lowest number of 121 in 2020. This resulted in a steady mean hospital volume between 14.4 in 2013 and 12.3 in 2022 (t=-0.270, p = 0.281). Further significant trends over time were evident for ICU stay (tau=-0.822, p < 0.001), private insurance (tau=-0.511, p = 0.040), and robotic access (tau = + 0.809, p = 0.001) (Table 6).
Table 6.
Trend analysis
| Variable | Sum / average | 2013 | 2022 | tau trend | p value* |
|---|---|---|---|---|---|
| Mortality (%) | 5.1 | 9.3 | 3.8 | -0.511 | 0.040 |
| Overall complications (%) | 69.2 | 63.9 | 72.9 | 0.289 | 0.245 |
| Number of operations | 1’535 | 183 | 133 | -0.733 | 0.003 |
| Number of hospitals | 19.9 | 30 | 18 | -0.750 | 0.003 |
| Mean hospital volume | 12.3 | 14.4 | 12.3 | -0.270 | 0.281 |
| Mean number per hospital | 7.9 | 6.1 | 7.4 | 0.244 | 0.325 |
| Mean age (years) | 64.7 | 64.7 | 66.3 | 0.156 | 0.531 |
| Mean CCI | 3.8 | 3.5 | 3.9 | 0.422 | 0.089 |
| Women (%) | 21.6 | 23.0 | 17.3 | -0.156 | 0.531 |
| ICU stay (%) | 87.2 | 96.2 | 76.7 | -0.822 | <0.001 |
| Swiss (%) | 88.4 | 85.9 | 90.2 | 0.045 | 0.857 |
| Private insurance (%) | 30.3 | 31.7 | 26.3 | -0.511 | 0.040 |
| Robotic access (%) | 8.07 | 0.0 | 23.3 | 0.809 | 0.001 |
| Mean LOS (days) | 24.8 | 26.7 | 25.5 | -0.200 | 0.421 |
*Significant values are bold
CCI: Charlson comorbidity index; ICU: Intensive care unit; LOS: Length of hospital stay
Discussion
The present retrospective Swiss national registry study is the first to examine the effect of centralisation of OCR on hospital volume and short-term postoperative outcomes in Switzerland. Cases operated at institutions with higher hospital volume were associated with lower mortality and lower postoperative complications. Regarding LOS, only a trend towards a negative association at higher quantiles was observed not reaching significance at any individual quantile.
A higher hospital volume was associated with lower mortality of nearly 28% for every ten additional cases. This finding aligns with other studies investigating the association between hospital volume and postoperative outcomes in OCR. US-American, UK, German, Japanese, and Korean national registry studies have described a similar association between hospital volume and postoperative mortality [32–37]. Their study populations ranged from 4’868 to 28’931 cases, and the periods examined spanned from three to 15 years between 1998 and 2017. Most analyses did not compare hospital volume as a marginal effect. Hospitals were divided into three groups: low-volume (threshold between 5 and 12 cases), mid-, and high-volume (threshold between 19 and 48 cases). These findings were confirmed in a recent meta-analysis, which showed a significantly reduced relative risk of postoperative mortality by 53% in high-volume hospitals compared to low-volume hospitals (threshold 45 cases, OR = 0.47, 95%CI 0.42 to 0.53) [11].
Overall, the mortality of 5.3% was higher compared to the international benchmark centres from ESODATA (2.0%) or DUCA (2.5% to 4.4) [13, 14]. After starting centralisation in 2013, the annual mortality rate was significantly reduced (tau=-0.511, p = 0.040) between 2013 (9.3%) and 2022 (3.8%). The lowest annual mortality rate of 2.0% in 2018 can keep up with these international benchmark centres.
A higher hospital volume was associated with lower postoperative complications of nearly 30% for every ten additional cases. Studies investigating the association between hospital volume and overall complications as a primary outcome are sparse. Most studies focus on mortality. The complete assessment of postoperative complications from large registries can be challenging. A nationwide analysis of 29’116 patients between 2016 and 2020 from the United States focusing on minimal invasive oesophagectomy described lower odds for major complications (OR = 0.83, 95%CI 0.70 to 0.96, p = 0.05) in hospital with a hospital volume > 16 [38]. DUCA showed less technical complications (OR = 0.82, 95%CI 0.70 to 0.96), less anastomotic leakage (OR = 0.80, 95%CI 0.66 to 0.97), and more textbook outcome (OR = 1.25, 95%Cl 1.07 to 1.46) in hospitals with hospital volume > 39 [39]. A recent meta-analysis of 177’566 patients also showed a lower leak rate in hospitals with hospital volume > 36 (9.6% vs. 8.3%, p = 0.040). Meta-regression confirmed an association between annual oesophagectomies per centre and anastomotic leakage (β=-0.0008, SD 0.0003, p = 0.0183) [40]. One possible explanation for the difference to our results regarding anastomotic leakage could be the generally higher hospital volume and threshold in these studies. The overall complication rate (69.3%) and the rate of anastomotic leakage (20.7%) were just slightly higher compared to ESODATA (60.5% and 12.5%) and DUCA (56.8% to 66.4% and 18.2% to 19.3%) [13, 14]. The extent to which the high-resource setting of the highly developed Swiss health care system can compensate for this remains unclear in this study.
The current Swiss hospital volume threshold of 12 was set in 2013 and confirmed in a first revision in 2019 at national level by the Swiss Conference of Cantonal Health Directors (GDK) based on the Intercantonal Agreement on Highly Specialised Medicine (HSM) [15, 16]. It is significantly below the threshold for high-volume centres proposed by ESODATA (n = 40 to 50), DUCA (n = 20), the German S3 oesophageal cancer guideline (n = 20), or the legally binding German national threshold (n = 26) [3, 13, 14, 41, 42]. In 2013, eight centres were granted a permanent mandate, whilst eight others were granted only a provisional one. After the first HSM revision in 2019, only eight centres with a permanent mandate remained. The big difference between the number of hospitals with a performance mandate and the actual number of hospitals providing oesophageal cancer resection as well as the fact that 57.8% of OCR were performed in hospitals below the stated threshold must be explained. Following the decisions taken by the GDK and HSM at national level, the cantonal health authorities are responsible for their final implementation and enforcement. The most powerful lever is funding. The procedures are only covered by health insurance in hospitals that have a performance mandate. However, Switzerland is a highly federalised country. The interest of individual cantons in having their own centres, appeals with suspensive effect against GDK decisions and HSM mandates, and the estimated delay in centralisation of 10 to 15 years between regulation and implementation are possible explanation for these numbers. However, the observation that the mean hospital volume remained stable despite a reduction in the number of hospitals and number of OCR performed per year does not fully support an effect of centralisation.
The reduction of surgeries between 2013 and 2022 was somewhat astonishing. Normally, as a database becomes well established, the number of surgeries increases [43]. The current national registry went in the opposite direction. There are various possible explanations. The incidence of oesophageal cancer in Switzerland decreased in the study period [44]. A more profound prevention and screening may result in an increase in detection of early-stage cases with possible endoscopic therapy. Definitive radiochemotherapy has established itself as an alternative for squamous cell carcinomas and multimorbid patients [3]. Centralisation of highly complex surgery with high perioperative risk can result in stricter case selection [45]. At least, the COVID-19 pandemic could also have contributed to a reduced number of OCR [46]. Despite the decrease in number of hospitals performing OCR, a significant proportion of the data reported in this study came from low- to mid-volume, high-resource centres, as defined by international standards. As the perioperative management of OCR is very challenging, focusing on high-volume institutions alone is not advisable. A highly competent, dedicated, and resilient surgeon must rely on an interdisciplinary team of intensive care physicians (84.6% of cases involved an ICU stay), gastroenterologists, interventional radiologists, nutrition professionals, nursing professionals, and other relevant disciplines [47–51]. Optimal outcomes rely on the synergy of a well-coordinated interdisciplinary team and a skilled surgeon with excellent intraoperative decision-making. A high hospital volume strengthens this interdisciplinary teamwork and increases the multidisciplinary clinical expertise. Failure-to-rescue is a capacity directly related to institutional experience and interdisciplinary infrastructure [42]. This becomes particularly evident in critical situations (seriously ill patient, complex tumour case, severe complications). It may explain the trend of a negative association between hospital volume and LOS at higher quantiles (fewer prolonged LOS) as described in this study. However, the failure-to-rescue analysis regarding mortality was insignificant. The reduced statistical power due to its smaller sample size could therefore be an explanation.
The data from this study, the studies mentioned above, and the international benchmark centres should be considered in the next allocation review for OCR in Switzerland, which started in July 2025.
Strengths and limitation
The strengths of this study are the high case number, the high external validity through presentation of real-world data, the national population-based coverage minimalizing selection bias, and the minimal costs per case using government-collected data.
There are some limitations next to the low- to mid-volume, high-resource setting mentioned above. Non-randomized studies only allow associations to be made, not causal statements. They carry a relevant risk of confounding by indication, time-related biases (changes in surgical, medical, and intensive care treatment) and bias from unobserved variables. The data used in this analysis were not primarily collected for research purposes and do not contain detailed clinical variables and depth. Due to the absence of the American Society of Anesthesiologists Physical Status classification, multimodal treatment strategy, surgical approach, operative details, histology, and the Clavien-Dindo classification of postoperative complications, several important patient characteristics are missing, as they are not reported in the FSO data. The transformation of clinical data into registry codes and back into clinical data makes such registry studies prone to reporting bias due to miscoding, underreporting, and incomplete documentation [52]. For example, regarding UICC staging, the ICD based surrogate likely underestimated stage III as lymph node metastases are often not coded as secondary diagnoses. This fact and the three additional degrees of freedom necessary for its implementation into the multivariate analyses were the reasons why the ICD-based UICC stage surrogate was not included in it.
The effect of centralisation, whereby sicker patients are more likely to be operated on at hospitals with higher hospital volume and university status, can lead to distortions [37, 53]. Together with a stricter case selection as mentioned above, this could be an explanation for the lower odds for complications in tertiary compared to university hospitals. If that is true, high-volume centres would operate sicker patients and the observed protective effect of hospital volume on mortality would be underestimated. These considerations are crucial in the clinical interpretation of the results.
Conclusions
Higher hospital volume is associated with lower mortality and fewer postoperative complications following OCR. These findings may support Swiss institutions in improving clinical outcomes, enhancing patient safety, and facilitating national comparisons with international benchmark centres.
Supplementary Information
Below is the link to the electronic supplementary material.
Acknowledgements
Acknowledgments to medical coding expert Dr. Therese Kim for helping with the data processing.
Author contributions
Yanic Ammann Conceptualisation, data curation, investigation, methodology, visualisation, writing - original draftRene Warschkow Conceptualisation, data curation, formal analysis, software, validation, visualisation, writing – review & editingMarie Klein Writing – review & editingKristjan Ukegjini Writing – review & editingIgnazio Tarantino Writing – review & editingThomas Steffen Conceptualisation, resources, project administration, supervision, writing – original draft, review & editing.
Funding
Thomas Steffen holds a grant from the “Stiftung Chirurgie”, St.Gallen, Switzerland, to financially support young researchers in translation, manuscript editing, and publication costs.
Data availability
The anonymised case data supporting this study’s findings can be obtained from the Swiss Federal Statistics Office (FSO).
Declarations
Ethics approval
This study was approved by the Ethics Committee of Eastern Switzerland (BASEC B2025-00702), registered under ClinicalTrials.gov (NCT07022652), and is compliant with the Strengthening the Reporting of Observation in Epidemiology (STROBE) guidelines. It conforms to the provisions of the Declaration of Helsinki.
Patient consent
The data was obtained anonymously from the Swiss Federal Statistics Office (FSO) via a data contract for the one-time delivery and use of individual data (without personal identification) in Switzerland.
Consent for publication
This submission is to Langenbeck’s Archives of Surgery. The material is original research, has not been previously published, and has not been submitted for publication elsewhere while under consideration.
Competing interests
The authors declare no competing interests.
Footnotes
What does this paper add to the literature?
This retrospective Swiss national registry study between 2013 and 2022 confirms the association between an increasing annual-hospital specific case volume of oesophageal cancer resection and improved short-term postoperative outcomes. Furthermore, the necessity of further centralisation of this highly complex procedure in Switzerland is demonstrated in order to achieve the outcome of international benchmark centres.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- 1.Krebsregistrierungsstelle BfSN (2024) 13.05.2025] Krebs, Neuerkrankungen und Sterbefälle: Anzahl, Raten, Medianalter und Risiko pro Krebsart. Available from: https://www.bfs.admin.ch/bfs/de/home/statistiken/gesundheit/gesundheitszustand/krankheiten/krebs/daten.assetdetail.33250863.html
- 2.Krebsliga-Schweiz (2024) Krebs in der Schweiz: wichtige Zahlen
- 3.AWMF. S3-Leitlinie Diagnostik und Therapie der Plattenepithelkarzinome und Adenokarzinome des Ösophagus (2023) 13.05.2025] ; Available from: https://register.awmf.org/de/leitlinien/detail/021-023OL [DOI] [PubMed]
- 4.Kato H, Nakajima M (2013) Treatments for esophageal cancer: a review. Gen Thorac Cardiovasc Surg 61(6):330–335 [DOI] [PubMed] [Google Scholar]
- 5.Lewis S, Lukovic J (2022) Neoadjuvant Therapy in Esophageal Cancer. Thorac Surg Clin 32(4):447–456 [DOI] [PubMed] [Google Scholar]
- 6.Waters JK, Reznik SI (2022) Update on Management of Squamous Cell Esophageal Cancer. Curr Oncol Rep 24(3):375–385 [DOI] [PubMed] [Google Scholar]
- 7.Yang W, Niu Y, Sun Y (2023) Current neoadjuvant therapy for operable locally advanced esophageal cancer. Med Oncol 40(9):252 [DOI] [PubMed] [Google Scholar]
- 8.Chevallay M et al (2020) Esophageal cancer surgery: review of complications and their management. Ann N Y Acad Sci 1482(1):146–162 [DOI] [PubMed] [Google Scholar]
- 9.Visser MR, van Berge Henegouwen MI, van Hillegersberg R (2024) Centralization and quality control in esophageal cancer surgery: a way forward in Europe. Dis Esophagus. 10.1093/dote/doae035 [DOI] [PubMed]
- 10.Meguid RA et al (2009) The effect of volume on esophageal cancer resections: what constitutes acceptable resection volumes for centers of excellence? J Thorac Cardiovasc Surg 137(1):23–29 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Di J et al (2024) Hospital volume-mortality association after esophagectomy for cancer: a systematic review and meta-analysis. Int J Surg 110(5):3021–3029 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Wang Q et al (2023) Association of hospital volume and long-term survival after esophagectomy: A systematic review and meta-analysis. Front Surg 10 [DOI] [PMC free article] [PubMed]
- 13.Kuppusamy MK, Low DE (2022) Evaluation of International Contemporary Operative Outcomes and Management Trends Associated With Esophagectomy: A 4-Year Study of > 6000 Patients Using ECCG Definitions and the Online Esodata Database. Ann Surg 275(3):515–525 [DOI] [PubMed] [Google Scholar]
- 14.Voeten DM et al (2021) Outcomes of Esophagogastric Cancer Surgery During Eight Years of Surgical Auditing by the Dutch Upper Gastrointestinal Cancer Audit (DUCA). Ann Surg 274(5):866–873 [DOI] [PubMed] [Google Scholar]
- 15.HSM-Beschlussorgan. Entscheid zur Planung der hochspezialisierten Medizin (HSM) im Bereich der grossen seltenen viszeralchirurgischen Eingriffe: Oesophagusresektion (2013) ; Available from: https://www.fedlex.admin.ch/eli/fga/2013/1271/de
- 16.HSM-Beschlussorgan. Beschluss über die Zuteilung der Leistungsaufträge im Bereich der hochspezialisierten Medizin (HSM): Komplexe hochspezialisierte Viszeralchirurgie – Oesophagusresektion bei Erwachsenen (2019) ; Available from: https://www.gdk-cds.ch/fileadmin/docs/public/gdk/themen/hsm/hsm_spitalliste/BB_Oesophagusresektion.pdf
- 17.Güller U et al (2017) Lower hospital volume is associated with higher mortality after oesophageal, gastric, pancreatic and rectal cancer resection. Swiss Med Wkly 147:w14473 10.4414/smw.2017.14473 [DOI] [PubMed] [Google Scholar]
- 18.van der Werf LR et al (2020) Reporting national outcomes after esophagectomy and gastrectomy according to the Esophageal Complications Consensus Group (ECCG). Ann Surg 271(6):1095–1101 [DOI] [PubMed] [Google Scholar]
- 19.Kaech M et al (2021) Influence of the introduction of caseload requirements on indication for visceral cancer surgery in Switzerland. Eur J Surg Oncol 47(6):1324–1331 [DOI] [PubMed] [Google Scholar]
- 20.ICD-10-GM Version (2021) Systematisches Verzeichnis, Internationale statistische Klassifikation der Krankheiten und verwandter Gesundheitsprobleme, 10. Revision, Stand: 18. September 2020. 2020 19.04.2025]; Available from: www.bfarm.de – Kodiersysteme – Services - Downloads – ICD-10-GM – Version 2021
- 21.Schweizerische Operationsklassifikation (CHOP), Systematisches Verzeichnis - Version 2025. (2024) 19.04.2025] ; Available from: https://www.bfs.admin.ch/bfs/de/home/statistiken/kataloge-datenbanken.assetdetail.32128591.html
- 22.Rice TW et al (2017) Cancer of the Esophagus and Esophagogastric Junction: An Eighth Edition Staging Primer. J Thorac Oncol 12(1):36–42 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.van Walraven C et al (2009) A modification of the Elixhauser comorbidity measures into a point system for hospital death using administrative data. Med Care 47(6):626–633 [DOI] [PubMed] [Google Scholar]
- 24.Charlson ME et al (1987) A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 40(5):373–383 [DOI] [PubMed] [Google Scholar]
- 25.Li Z, Wood SN (2020) Faster model matrix crossproducts for large generalized linear models with discretized covariates. Stat Comput 30(1):1–7 [Google Scholar]
- 26.Long JS, Freese J (2014) Regression Models for Categorical Dependent Variables Using Stata, 3rd ed. Stata Press, College Station, TX [Google Scholar]
- 27.Kazemi M et al (2021) Prediction of Hospitalization Length. Quantile Regression Predicts Hospitalization Length and its Related Factors better than Available Methods. Ann Ig 33(2):177–188 [PubMed] [Google Scholar]
- 28.Khmaladze EV (1981) Martingale Approach in the Theory of Goodness-of-Fit Tests. Theory Probab Its Appl 26(1):216–217 [Google Scholar]
- 29.Mann HB (1945) Nonparametric Tests Against Trend. Econometrica 13(3):245–259 [Google Scholar]
- 30.Kendall MG (1938) A new measure of rank correltation. Biometrika 30(1–2):81–93 [Google Scholar]
- 31.von Elm E et al (2007) The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. Lancet 370(9596):1453–1457 [DOI] [PubMed] [Google Scholar]
- 32.Kim BR et al (2021) The association between hospital case-volume and postoperative outcomes after esophageal cancer surgery: A population-based retrospective cohort study. Thorac Cancer 12(18):2487–2493 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Schlottmann F et al (2018) Esophageal Cancer Surgery: Spontaneous Centralization in the US Contributed to Reduce Mortality Without Causing Health Disparities. Ann Surg Oncol 25(6):1580–1587 [DOI] [PubMed] [Google Scholar]
- 34.Fuchs HF et al (2017) Mortality after esophagectomy is heavily impacted by center volume: retrospective analysis of the Nationwide Inpatient Sample. Surg Endosc 31(6):2491–2497 [DOI] [PubMed] [Google Scholar]
- 35.Nishigori T et al (2016) Impact of hospital volume on risk-adjusted mortality following oesophagectomy in Japan. Br J Surg 103(13):1880–1886 [DOI] [PubMed] [Google Scholar]
- 36.Nimptsch U, Peschke D, Mansky T (2017) Minimum Caseload Requirements and In-hospital Mortality: Observational Study using Nationwide Hospital Discharge Data from 2006 to 2013. Gesundheitswesen 79(10):823–834 [DOI] [PubMed] [Google Scholar]
- 37.Fischer C et al (2017) Volume-outcome revisited: the effect of hospital and surgeon volumes on multiple outcome measures in oesophago-gastric cancer surgery. PLoS One 12(10):e0183955 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Mallick S et al (2024) Association of hospital volume and operative approach with clinical and financial outcomes of elective esophagectomy in the United States. PLoS One 19(6):e0303586 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Voeten DM et al (2021) Overall Volume Trends in Esophageal Cancer Surgery Results From the Dutch Upper Gastrointestinal Cancer Audit. Ann Surg 274(3):449–458 [DOI] [PubMed] [Google Scholar]
- 40.Rahouma M et al (2024) Volume outcome relationship in postesophagectomy leak: a systematic review and meta-analysis. Int J Surg 110(4):2349–2354 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Bundesausschuss G (2020) Beschluss des Gemeinsamen Bundesausschusses über eine Änderung der Mindestmengenregelungen: Änderung der Nr. 3 der Anlage. [cited 2026 14.04.2026]; Available from: https://www.g-ba.de/downloads/39-261-4622/2020-12-17_Mm-R_Oesophagus_BAnz.pdf
- 42.Nimptsch U et al (2018) Hospital Volume, In-Hospital Mortality, and Failure to Rescue in Esophageal Surgery. Dtsch Arztebl Int 115(47):793–800 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Alluri RK, Leland H, Heckmann N (2016) Surgical research using national databases. Ann Transl Med 4(20):393 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Bundesamt-für-Statistik. Krebsmonitoring Schweiz, Speiseröhrenkrebs ICD-10 C15 (2025) 17.12.2025] ; Available from: https://krebs-monitoring.bfs.admin.ch/de/detail/C15
- 45.Ramsay AI et al (2025) Centralisation of specialised healthcare services: a scoping review of definitions, types, and impact on outcomes. Health Soc Care Deliv Res. 10.3310/REMD6648 [DOI] [PubMed] [Google Scholar]
- 46.Milito P et al (2022) Minimally invasive esophagectomy for cancer in COVID hospitals and oncological hubs: are the outcomes different? Eur Surg 54(2):98–103 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.Ashok A et al (2020) The enhanced recovery after surgery (ERAS) protocol to promote recovery following esophageal cancer resection. Surg Today 50(4):323–334 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48.Hua F et al (2023) Update on therapeutic strategy for esophageal anastomotic leak: A systematic literature review. Thorac Cancer 14(4):339–347 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49.Caccialanza R et al (2025) Nutritional support via feeding jejunostomy in esophago-gastric cancers: proposal of a common working strategy based on the available evidence. Updates Surg 77(1):153–164 [DOI] [PubMed] [Google Scholar]
- 50.Cao Y et al (2022) Effects of preoperative nutrition on postoperative outcomes in esophageal cancer: a systematic review and meta-analysis. Dis Esophagus. 10.1093/dote/doab028 [DOI] [PubMed] [Google Scholar]
- 51.Chen H, Ou Y, Mo L (2021) High-quality nursing can reduce the incidence of adverse events in esophageal cancer patients after operation in the intensive care unit and improve postoperative rehabilitation. Am J Transl Res 13(10):11851–11859 [PMC free article] [PubMed] [Google Scholar]
- 52.Lucyk K, Tang K, Quan H (2017) Barriers to data quality resulting from the process of coding health information to administrative data: a qualitative study. BMC Health Serv Res 17(1):766 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 53.Munir MM et al (2024) Association of Hospital Market Competition with Outcomes of Complex Cancer Surgery. Ann Surg Oncol 31(7):4371–4380 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The anonymised case data supporting this study’s findings can be obtained from the Swiss Federal Statistics Office (FSO).