Autopsy rates and diagnostic errors in a Swiss community hospital: a ten-year retrospective analysis

Abstract

Historically, autopsy has served as a procedure to determine the cause of death. With the surge in diagnostic techniques and a change in mindset, autopsy rates have decreased globally to less than 10%. Besides providing the cause of death and insights into disease mechanism, autopsy may assess clinical performance via discrepancy analysis. The objectives of our study were to determine autopsy rates in a Swiss community hospital and major discrepancies between clinical and autopsy reports according to the Goldman classification. Of 1711 registered in hospital deaths between 2012 and 2021, 123 patients were subjected to autopsy, resulting in an average autopsy rate of 7.2%. Younger patients had higher autopsy rates (17.0% for patients aged 50–59 years), and sex-adjusted male-to-female autopsy rate was 1.6. Major discrepancies were identified in 18 cases (14.6%); 10 (8.1%) class I, and 8 (6.5%) class II errors. Complete agreement was reached in 74 cases (60.2%), and 31 cases (25.2%) were classified as indeterminate category VI. The most frequent class I errors were infectious diseases (n = 5) with two cases of endocarditis, two cases of aspergillosis, and one case of perforated diverticulitis. The most frequent class II error was advanced malignant disease (n = 4). In conclusion, the autopsy rate at our hospital was at the higher end of other contemporary reports, and major discrepancy rates occurred at a comparably low frequency. Nevertheless, the study revealed that despite extensive diagnostics, cases prone to misdiagnosis still exist. In this context, autopsy represents an important tool for quality control.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00428-025-04060-2.

Introduction

Autopsy, one of the oldest methods for revealing pathological processes, has a long history, beginning over 2000 years ago. The word derives from Greek, meaning “to see for oneself.” Postmortem examination experienced a surge during the Renaissance when scholars systematically started investigating human anatomy and pathology. In recent decades, unprecedented medical progress has been made with great advances in diagnostic and therapeutic techniques that have substantially changed the face of medicine. At the same time, autopsy rates started to decline steadily, a process that is ongoing [1]. In Switzerland, a decrease of 72% occurred between 1993 and 2012 [2] with similar trends in other countries such as Germany and the USA [3, 4]. Current autopsy rates in Switzerland are estimated to be below 4% [5]. Such low autopsy rates question the performance of high-quality pathology examinations due to a sheer lack of sufficient training and constitute a loss of valuable case-specific and systematic insights as an opportunity for quality control and further medical advancement [6]. Goldman and colleagues compared autopsy results to clinical diagnoses and reported major discrepancies in more than 20% of cases over 20 years, irrespective of medical advancements during this period [7]. They introduced a classification system to grade discrepancies, which was later modified by Battle et al. [8], and thereby laid the foundation for comparable investigations.

This retrospective quality assessment examines autopsy practice at a Swiss community hospital over a 10-year period. We analyzed autopsy rates over the years, cause of death distribution, and the concordance between clinical and autopsy diagnoses using the Goldman classification. Additionally, we documented incidental findings and compared our results with similar studies to provide further insights into the value of autopsies in medical practice.

Material and methods

Patients

This retrospective analysis was conducted at a community hospital in Wetzikon Switzerland over the time period from 2012 to 2021. Numbers of inpatient treatments continuously increased to currently over 10,000 treatments annually. Overall, the mortality rate is approximately 2%, and autopsy is offered to the relatives of all deceased patients; consent is sought by the attending physician. Clinicopathological conferences to review cases in an interdisciplinary setting take place on a regular basis.

All patients who underwent clinical autopsy were included in our analysis, and their corresponding medical and postmortem reports were examined. Patients under 18 years of age as well as obstetric and neonatal deaths were excluded. Data on sex, age, and duration of hospital stay were recorded as well as whether a patient had been resuscitated or treated in the intensive care unit prior to death. Clinical records were retrieved from the hospital’s computer-based system. All included patients were subjected to a complete autopsy, including neuropathological and histological examinations, conducted at the Institute of Pathology at the cantonal hospital Winterthur. The final report included the principal findings, followed by a detailed description of changes in each organ.

Cause of death

A synopsis of clinical and autopsy records was used as a basis to establish the cause of death. A general internal medicine specialist (CMH) analyzed each case and assigned COD to one of 11 predefined categories similar to Friberg et al. [9]: cardiopulmonary failure, myocardial infarct, aortic rupture, pulmonary embolism, cerebrovascular lesion, gastrointestinal hemorrhage, pneumonia, sepsis, neoplasm, other, and not identified. Cardiopulmonary failure was defined as terminal insufficiency of the heart or lungs as it occurs due to persistent myocardial injury, obstructive pulmonary disease, or other conditions leading to circulatory arrest. Sepsis was defined as multiorganic dysfunction due to excessive systemic inflammation caused by infection. Indicative pathological tissue alterations include severe inflammatory changes of the affected organ with detected microorganisms (e.g., pyelonephritis, suppurative peritonitis) as well as general accompanying processes such as hyperemic spleen, reactive hypercellular bone marrow changes, reactive lymphadenopathies, or septic spreading (e.g., pyemic foci in the myocardium). Whenever the lung was the source of infection without septic implications in autopsy, cases were referred to as pneumonia.

Goldman classification

Clinical and autopsy diagnoses were categorized into major or minor diagnoses. Major findings included either the principal underlying COD or related conditions demanding urgent attention. Minor findings included features unrelated to death and without immediate influence on treatment, such as concomitant and antecedent conditions. In cases where diagnostic and treatment efforts were withdrawn and comfort care was established, all major autopsy diagnoses that were deemed to have developed under end-of-life conditions were not considered in the discrepancy analysis. Clinical and postmortem findings were subsequently compared, and discrepancies were classified according to the system established by Goldman et al. and modified by Battle et al. (Tab. 1). The system retrospectively values the influence on case management, with a focus on the probable negative impact on survival. A general internal medicine specialist (CMH) conducted a discrepancy analysis, doubtful cases were reviewed and discussed with a pathology expert (PKB), and a consensus was reached for each case. Only major diagnoses were considered in the final discrepancy analysis such that cases were assigned to classes I, II, V, or VI. Concomitant diagnoses unrelated to COD were listed as incidental findings and not attributed to a minor discrepancy class.

Table 1.

Discrepancy classes according to Goldman et al., modified by Battle at al. Major errors are assigned to Goldman classes I and II, minor errors to classes III and IV. Class V represent cases where clinical and autopsy reports agree, class VI contains non-classifiable cases

Class	Degree	Type	Criteria
I	major	adverse impact	correct diagnosis would have affected management, potential survival impact (e.g. pulmonary embolism treated as pneumonia)
II	major	equivocal impact	correct diagnosis would not have influenced management (e.g. intracranial hemorrhage with undetected pneumonia)
III	minor	clinically accessible	no immediate treatment consequence but eventually relevant (e.g. dysplastic colorectal polyps)
IV	minor	clinically unaccessible	occult finding with only epidemiological or genetic significance (e.g. hepatic hemangioma)
V	agreement	non-discrepant
VI	non-classifiable	non-classifiable	immediate death, refusal of treatment, fruitless autopsy, cause of death unknown (e.g. malignant arrhythmia)

Ethics

An autopsy was performed only with the consent of the relatives. According to a position paper by the Swiss Ethics Committees on research involving humans [10] and based on the Swiss Federal Human Research Act (HRA, RS 810.30), approval by the ethics committee was not required for this study because it served as a quality control tool to improve already established medical practice.

Results

Study population

Between 2012 and 2021, 79,993 patients were hospitalized and 1711 deaths were reported (Table 2). One hundred twenty-three patients were subjected to autopsy and included in the study. The male-to-female autopsy ratio was 1.8 (male n = 79; 64%). The median age was 77 years (43–93 years), and the majority (n = 77; 63%) of the autopsied patients were aged 70 to 89 years. Of autopsies, 92.7% (n = 114) were requested by general internal medicine, the remaining by surgery, and none by other specialties. One-third was either reanimated prior to death (36/123) or treated in the intensive care unit (41/123). The median length of hospitalization was 6 days (1–95 days), and intermediate stays were most common (41/123 3–7 days, 33/123 8–14 days).

Table 2.

Distribution of patient characteristincs according to age, sex, median hospital stay, resuscitation status and treatment in the ICU as well as total registered deaths of male and female patients in the analyzed time period

Number of autopsies (n) - by age	123 - <50: 4 (3.3%) - 50–59: 16 (13.0%) - 60–69: 19 (15.4%) - 70–79: 35 (28.5%) - 80–89: 42 (34.1%) - >89: 7 (5.7%)
Median age (years)	77 (43–93)
Ratio male/female	1.8 (79/44)
Median hospital stay (days)	6 (1–95)
Resuscitated (n)	36 (29.3%)
ICU (n)	41 (33.3%)
Total male deaths (n) - by age	900 - <50: 26 (2.9%) - 50–59: 53 (5.9%) - 60–69: 132 (14.7%) - 70–79: 254 (28.2%) - 80–89: 340 (37.8%) - >89: 95 (11.0%)
Total female deaths (n) - by age	811 - <50: 24 (3.0%) - 50–59: 41 (5.1%) - 60–69: 88 (10.9%) - 70–79: 167 (10.6%) - 80–89: 331 (40.9%) - >89: 160 (20.0%)

Autopsy rate

The autopsy rate declined from 12.6% in 2012 to 4.7% in 2021, and the average rate was 7.2% (Fig. 1a). Except in 2012 and 2021, there was a male predominance of autopsied patients. Most patients who died during hospitalization were between 80 and 89 years old (671/1711), with an autopsy rate of 6.3% (n = 42) in this age group. There were fewer deaths in the 50–59 year age group (94/1711), but the autopsy rate was highest at 17.0% (n = 16) (Fig. 1b). The male-to-female sex-adjusted autopsy rate was 1.6 (male n = 79/900, 8.8%; female n = 44/811, 5.4%).

Fig. 1.

Autopsy rates. a Annual autopsy rates: 2012 12.6%, 2013 6.2%, 2014 5.4%, 2015 9.2%, 2016 8.1%, 2017 7.0%, 2018 9.4%, 2019 3.2%, 2020 5.1%, 2021 4.7%; b autopsy rates in age categories: < 50 y 8.0%, 50–59 y 17.0%, 60–69 y 8.6%, 70–79 y 8.3%, 80–89 y 6.3%, > 89 < 2.8%

Cause of death and major findings

Actual cause of death (COD) and hence major diagnosis were identified by autopsy in 95% of cases (n = 117). The most common disease was pneumonia (n = 24; 19.5%), followed by cardiopulmonary failure (n = 23; 18.7%), myocardial infarction (n = 18; 14.6%), and sepsis (n = 16; 13.0%) (Fig. 2a). Considering neoplastic lesions in general, 33% (n = 41) of reports listed at least one neoplasia as a main finding, but only in 10.6% (n = 13) of cases was it the presumed COD. In six patients, the COD could not be identified, but the corresponding pathology reports included marked cardiac abnormalities (e.g., hypertrophy, fibrosis, coronary sclerosis), suggesting a cardiac COD.

Fig. 2.

Cause of death and incidental findings determined by autopsy between 2012 and 2021. a Cause of death attributed after autopsy. b Incidental findings according to the organ system, grouped into cardiovascular, pulmonary, gastrointestinal, urogenital, cerebral, endocrine, and musculoskeletal origin

Incidental findings

Minor diagnoses were regarded as incidental findings and grouped into several categories; the most frequent are shown in Fig. 2b. Vascular pathologies, especially atherosclerosis, were most often noted, mirroring the advanced age and Western lifestyle of the study population. Amyloid angiopathy and tissue amyloidosis represent separate entities with distinct underlying pathomechanisms but are rarely diagnosed clinically in general. Some connate architectural variations were also noted, such as heterotopies or duplicities. Signs of metabolic and endocrine affections appeared mainly as hepatic steatosis and cholecysto-/docholithiasis or as thyroid and adrenal gland abnormalities. Neoplastic findings occurred in almost all organs, usually benign (e.g., leiomyoma, lipoma, hemangioma, adenoma) but also certain malignant variants (e.g., prostate carcinoma). Most malignant neoplasia appeared as a major diagnosis.

Discrepancy analysis

Agreement between clinical and autopsy reports was reached in 60.2% (n = 74), and discrepancies were revealed in 14.6% (n = 18) of cases. Following the Goldman classification system, ten cases (8.1%) were assigned to class I and eight cases (6.5%) to class II (Fig. 3a). In 2017, classes I and VI occurred more frequently at the expense of class V (Fig. 3b). The median hospitalization duration in major discrepant cases was 7.5 days, the median age was 75 years, and the sex distribution was equal.

Fig. 3.

Distribution of Goldman classes 2012–2021. a Class I was attributed in 8.1%, II in 6.5%, V in 60.2% and VI in 25.2%. b Except in 2017, class V was assigned most frequently followed by class VI

The most frequent class I errors included infectious diseases; detailed findings are summarized in Table 3 Among class I discrepant cases, six out of ten were male, the median age was 72 years, the median hospitalization was 8.5 days, four patients were resuscitated, and five patients stayed in the ICU prior to death. In deaths after resuscitation and intensive care treatment, there was a slightly higher rate of class I (n = 4; 11% and n = 65; 12%) and class II (n = 4; 11% and n = 3; 7%) errors, whereas ICU patients had a greater rate of concordance with 66% (n = 29) class V and instead only 15% (n = 4) class VI cases (Online Resource 1a and b).

Table 3.

Major discrepant diangoses are incorporated into Goldman class I and II errors, class VI contains cases not attributable to an error class

	Cause	n
Class I (n=10)	Sepsis	4
	Myocardial infarct	2
	Pulmonary embolism	2
	Cerebral ischemia after MCAa thrombosis	1
	Pneumonia	1
Class II (n=8)	Neoplasm	4
	Sepsis	2
	Aortic rupture	1
	Subdural hematoma	1
Class VI (n=31)	Peracute death	14
	Comfort care	10
	No lethal autopsy finding	6
	Malcompliance	1

^aMCA middle cerebral artery

Class VI, in which either no morphological lethal finding was identified or in which diagnostic measures were not available (due to patient refusal/malcompliance, comfort care, immediate death after admission), was attributed to 31 patients (25.2%). Among these, 14 patients (45%) died within 1 or 2 days after admission without any diagnostic assessment. For ten patients (32%), further diagnostics were withheld and comfort care was established. One patient was malcompliant, one died unexpectedly from myocardial infarction in the course of hospitalization, and one died of cardiopulmonary failure with systemic amyloidosis and possibly malignant arrhythmia. Overall, COD remained undetermined in six cases after autopsy.

Discussion

Role and rates of autopsy

The clinical autopsy serves not only to determine COD but also to study underlying diseases, monitor emerging health threats, provide public health data, support education and research, and ensure medical quality control. Yet, its global decline since the 1960s continues due to advanced diagnostics, administrative and funding challenges, and reluctance to seek consent amid legal concerns. Blokker et al. performed a study in the Netherlands and assumed an annual decline in clinical autopsy rates of 0.7% [11]. In our study, we report a median autopsy rate of 7.2% at the community hospital in Wetzikon over a 10-year period from 2012 to 2021. This exceeds the average rate in Switzerland, which decreased from 13% in 1993 to less than 4% in 2012 [5].

Although the community hospital lacks its own pathology institute, requiring greater effort to arrange autopsies, the autopsy rate is just slightly below the rate of 9.75% reported in a recent study from the University Hospital of Zurich [12]. An evaluation conducted by the Swiss Federal Statistical Office between 2006 and 2011 revealed that of the 70,000 annual deaths in Switzerland, approximately 40% occurred in hospitals, 40% in nursing home facilities, and 20% at home [13]. As efforts are made to avoid hospitalization of very old patients and instead provide medical care in a community environment, disinclination toward autopsy might increase even further [14, 15].

Patient characteristics and cause of death

Determining COD is still the main incentive to request an autopsy, and the most likely case that triggers further investigation is consequently the sudden death of a young person. Hence, the 50–59 year age group had the highest autopsy rate of 17.0% in our study (5% of hospitalized patients). The rate for the oldest conversely was the lowest and is comparable to reported autopsy rates in the elderly [16]. Nonetheless, autopsies in the elderly may be very informative because medical care is often restricted with increasing age such that more diseases remain unidentified [17]. Indeed, many patients in our study, mostly those older than 80 years or with underlying neoplasia, expressed a wish for best supportive care during hospitalization, thereby foregoing further diagnostic procedures.

Interestingly, female deaths at the community hospital in Wetzikon were overall less likely followed up by autopsy than male (4.4 vs. 8.7%), an observation also reported in other autopsy studies [18]. A possible explanation is the longer life expectancy of women and the lower autopsy rate in older people as discussed above [19]. Other factors, such as varying occurrences of diseases and a different spectrum of comorbidities, might also play a role because complex cases rather trigger postmortem examination. Several investigations have also revealed that consent to autopsy is more readily given for male patients [20].

According to the Swiss COD statistics, accidents dominate in the 25–44 year age group, whereas people between 45 and 84 years mostly die of cancer. Cardiovascular diseases are the main COD in the oldest age group of 85 and above [21]. The leading COD in our study traces back to infectious (pneumonia, sepsis) and circulatory diseases, which corresponds to results from other inquiries [4, 9, 22]. Neoplasia was a concomitant disease in one-third of our patients, but rarely declared the immediate COD. The diverging distribution of COD to national statistics might be explained by the selection of a specific subgroup of patients who underwent autopsy or by the fact that COD statistics are based on death certificates and not autopsies [23]. Moreover, victims of accidents are legally examined in Switzerland and therefore do not appear in our study at all. The incongruence of the distribution of autopsy-confirmed COD and official registers is widely known. As such, autopsy results might also draw attention to blind spots in health care and provide insight into population-specific differences on a more profound basis.

Discrepancy classification of unexpected findings

Most of our cases were assigned to Goldman category V, where autopsy confirmed clinical diagnosis. Goldman et al. reported approximately 10% class I errors in 1983, comparable to our study (n = 10; 8.1%) and other more recent studies [24]. The extensive meta-analysis by Shojania et al., with an estimated major error rate of 8.4 to 24.4% (class I, 4.1 to 6.7%) in U.S. institutions [25], mirrors the wide range of reported discrepancies of less than 5% to over 50% from other studies in different countries, institutions, and clinical settings [26, 27]. The University Hospital of Zurich currently reports 2.3% class I errors [12]. Evidence of higher discrepancy rates in community versus university hospitals exists [8, 28, 29]; however, there are no data available for Switzerland.

Discrepancies have decreased over time: A study from Berlin reported class I major errors in 25.8% of cases in 1988 and a reduction to 10.7% in 2008 [29]. A study from India states a decrease in discrepancies over time to a rate of 9.3% in 2010 [30]. Schwanda-Burger et al. reported a decrease from 30 to 7% in major errors between 1972 and 2002 at University Hospital of Zurich [2]. The rate of major diagnostic errors was higher in our subgroup of ICU patients (19% vs. 14.6%), an observation that has been made elsewhere and explained by the complex multimorbidity encountered in the ICU [31]. A shorter duration of stay and unexpected death are further thought to account for greater diagnostic uncertainty [27]. Indeed, more patients with shorter hospital stays were autopsied in 2015 and 2018, resulting in above-average autopsy rates (9.2% and 9.4% respectively). However, major errors did not accumulate during these years. Interestingly, the median hospital stay for class I error cases was 8.5 days, which is even longer than the median overall stay of 6 days in our cohort. Moreover, advanced age and female sex have been identified as risk factors for major discrepancies [8, 22, 29, 32]. In our 18 major error cases, there is no apparent tendency to confirm this.

Unidentified or incorrectly assigned infectious diseases most frequently accounted for class I errors. Aspergillosis, which is notoriously challenging to detect [33], was discovered in two cases. In two cases of autopsy-confirmed endocarditis, diagnostic procedures were performed but failed to reach the correct conclusion, thus leading to inappropriate treatment. However, it has to be emphasized that autopsy is not an appropriate tool to diagnose infections, and sepsis especially lacks distinctive pathological characteristics [34]. Then again, an autopsy might refute a clinical diagnosis of sepsis and reveal an alternative diagnosis [35]. Several studies have concluded that besides pneumonia and other infections, cardiovascular diseases and pulmonary embolism lead the ranks of major discrepancies [7, 29, 36]. In our two cases when acute myocardial infarction was missed, cardiopulmonary failure was suggested as COD on the basis of a history of cardiac disease with corresponding risk factors. Two unrecognized pulmonary embolisms occurred in patients with classical symptoms and a history of cancer as a predisposing factor. The most frequent class II discrepancy in our study was neoplasia.

Clinicians apparently cannot reliably predict discrepancies, and discrepancy rates are not affected by preliminary high clinical confidence [22, 36]. Moreover, our results confirm that extensive diagnostic methods do not fully eliminate discrepancies [37, 38]. Consequently, the autopsy remains a valuable tool for the proper detection of system-related errors and encourages differentiated thinking, especially when relevant conditions are at risk of being overlooked in time-sensitive situations [31, 39]. Profiles for patients prone to misdiagnosis could support a thorough investigation and help eliminate missed conditions in a prospective manner, thereby promoting adequate and cost-effective medical care.

Autopsy furthermore enables the characterization of rare or underreported underlying diseases, which fall into Goldman classes III and IV. Maccio et al. reported undiagnosed conditions in 98% of autopsied patients [40]. They identified coronary artery disease in 24% and sub-/acute myocardial infarction in 15%. Our cohort exhibited any type of atherosclerosis in 90.2% (mostly with coronary affection of various degrees) and 73.2% cardiac necrosis or fibrosis. This difference could be attributed to the older age of our patients and the selection of a more distinctive collective at a university hospital. The rates of prostate cancer (2.2% vs. 3.3%) and cardiac amyloidosis (8% vs. 9.8%) were similar in our study. Cardiac amyloidosis, which is present in approximately 25% of people over 85 years [41], is a perfect example of an underrecognized disease with novel treatment options. Not only general internal medicine but also surgical patients benefit from autopsy through the promotion of procedural or technical improvement. Rastan et al. reported a high (77.8%) rate of clinically unrecognized postoperative complications in cardiac surgery, with 18.2% unrecognized surgery-associated pathologic findings, COD discrepancies in 23.1%, and unknown concomitant diseases with possible relevance in 19.2% of cases [42].

Cancer patients often have a long history of different therapies with phases of remission and relapse. Autopsy provides a last chance to examine and correlate the progress of the disease with clinical complaints and to evaluate the response and adverse effects of treatment [43]. One exemplary case in our study population was a patient with assumed successfully controlled metastatic lung cancer; however, autopsy revealed vital cancer cells with newly diagnosed cerebral metastasis. Results of other studies indicate that autopsy uncovers previously undiagnosed malignant neoplasia in 5 to 11% [44, 45]. Even if not immediate COD, malignancies might explain certain consequent diseases (e.g., thromboembolism), and their early diagnosis could facilitate treatment and prevent later complications.

Limitations and improvements

Our analysis has similar limitations to comparable studies. The retrospective design and unavailability of the treating physician’s judgment widened the range of interpretation. We sought to minimize interpretation bias by reviewing ambiguous cases. Clinical reports for our collective commonly formulate a hypothesis or even state a major diagnosis for COD even in ambiguous cases so that the autopsy actually provides a proper reference tool. Selection bias by autopsy of clinically complex cases may distort COD statistics for the general population and lead to overestimation of major discrepancies. Systematic analysis of large numbers of patients and increased autopsy rates overall would ultimately reduce selection bias. This goal could be achieved by enhancing communication skills to obtain autopsy consent from relatives as well as a renewed awareness of the value of postmortem examinations throughout the medical community and general population. In fact, the number of autopsies significantly increased over a short period of time when structured conversations between doctors and families took place [46, 47]. In a Swiss population, structured training in motivational interviewing and a standardized questionnaire brought an increase in autopsy rates from 12.8 to 26.9% [48]. Clinicians’ awareness of autopsies should be invigorated during medical education at universities [49], and ongoing collaboration between clinicians and pathologists should be emphasized, for example at clinicopathological conferences [50]. At the community hospital in Wetzikon, such conferences take place three times a year, setting the stage for case-related discussion and teaching with visual insights into diseases, thus strengthening the value of autopsies.

Conclusion

The autopsy rate of 7.2% at the community hospital in Wetzikon (2012–2021) reflects the global decline in postmortem examinations. However, our findings underscore the continued value of autopsies in quality assurance and medical education. A full concordance between clinical and autopsy diagnoses (category V) was observed in 60.2% of cases, while major diagnostic discrepancies (categories I and II) occurred in 14.6%, less frequently than in other studies. This emphasizes the high quality of medical care provided at the community hospital Wetzikon. Rather than being viewed as criticism, these discrepancies highlight opportunities for learning and improvement, ultimately benefiting future patients and advancing medical standards.

Supplementary information

Below is the link to the electronic supplementary material.

Author contribution

Cristina Manuela Hagen and Peter Karl Bode designed the study, interpreted the data, and drafted and revised the manuscript. Cristina Hagen acquired and analyzed the data. Urs Eriksson, Katrin Schmid, and Renata Flury critically reviewed the manuscript. All authors gave final approval for publication and jointly took responsibility for the work as a whole.

Declarations

Ethics approval

This research was conducted in a retrospective manner with anonymized patient data. It is in compliance with applicable ethical standards and does not require informed consent from the included subjects according to Swiss practice.

Conflict of interest

The authors declare no competing interests.