Initial Cancer Treatment in Certified Versus Non-Certified Hospitals

Abstract

Background

According to the National Cancer Plan in Germany, all cancer patients should receive high-quality care in accordance with evidence-based treatment guidelines. Certification programs were established for this purpose but have not yet been comprehensively evaluated.

Methods

In the WiZen project, which was supported by the Innovation Fund (supported project number 01VSF17020), controlled cohort studies were performed to investigate whether initial treatment in hospitals with or without a certificate from the German Cancer Society was associated with a difference in overall survival (primary endpoint) in patients with cancer of the colon, rectum, lung, pancreas, breast, cervix, prostate, endometrium, and ovary, head and neck cancer, and neuro-oncological tumors. The studies were based on nationwide data from adult insurees of the AOK statutory health insurance carrier for the years 2009–2017.

Results

The majority of patients with all entities except breast cancer received their initial treatment in non-certified hospitals. Initial treatment in a certified hospital was found to be beneficial in terms of overall survival for all cancer entities, even after extensive adjustment for patient- and hospital-related confounders. The hazard ratio (HR) ranged from 0.97 (95% CI: [0.94; 1.00]) for lung cancer to 0.77 [0.74; 0.81] for breast cancer, corresponding to an absolute risk reduction (ARR) for overall survival of 0.62 months for lung cancer to 4.61 months for cervical cancer.

Conclusion

The WiZen study shows for the entities studied that initial cancer treatment in a certified center is associated with lower mortality. Despite the recommendations of the National Cancer Plan, however, more than 40% of all cancer patients still receive their initial treatment in a non-certified hospital. The preferential provision of initial care in certified hospitals would be likely to improve overall survival. Although the study design does not permit any conclusion with regard to causality, the findings seem robust considering that a control group was used, confounders were taken into account, and the study population was of large size.

Cancer is the second most common cause of death in Germany and one of the most common chronic diseases (1). In 2018, over 497 000 people were diagnosed with cancer nationwide. About 229 000 men and women died of cancer in the same year (2). Cancer represents one of the most significant burdens affecting patients, relatives, and the healthcare system measured in terms of quality of life loss, burden of disease, and treatment costs (3, 4). Despite intensified efforts at prevention, age-specific incidence rates for cancer remain stagnant. At the same time, therapeutic advances have been made for various cancer entities, so that cancer is increasingly being classified as a chronic disease with a continuously rising prevalence of cancer diseases (1).

In terms of health policy, cancer control has a very high priority in Germany. The National Cancer Plan in Germany, launched jointly by the Federal Ministry of Health and other participants in 2008, focuses not only on the further development of cancer screening and patient orientation, but also in particular on the optimization of oncological (care) structures (5). An infrastructure consisting of three central interconnected components has therefore been established:

• The German Guideline Program in Oncology develops evidence-based treatment guidelines with appropriate quality indicators.

• The certification of oncological treatment facilities focusses on the implementation of evidence-based treatment guidelines and verifies adherence to quality indicators (6).

• Clinical cancer registers (CCRs) record the clinical courses and treatments of all cancer patients, including guideline-based quality indicators (7).

Study evidence to date does not allow a conclusive assessment of the effects of treatment at certified hospitals/centers for individual cancer entities, but regional analyses indicate survival and economical advantages associated with treatment in certified hospitals (8–13). Patients and doctors believe that certification is commonly associated with good-quality care and a high treatment success rate (14–16). Given the major importance of certification as part of the German National Cancer Plan and the outstanding effort certification requires from the hospitals, a comprehensive controlled study of the effects of cancer treatment in certified centers is warranted.

In this light, we launched the study “Effectiveness of Care in Certified Cancer Centers (WiZen)” (Innovation Fund number 01VSF17020) involving 11 cancer entities to investigate whether initial treatment in hospitals with or without certification was associated with a difference in overall survival. The starting hypothesis was that initial treatment in certified hospitals is associated with better outcomes for patients.

Methods

In the WiZen project, controlled cohort studies were performed to investigate for 11 cancer entities (cancer of the colon, rectum, pancreas, breast, cervix, ovary, endometrium, lung, prostate, head and neck, and neuro-oncological tumors) whether initial treatment in hospitals with or without a certificate from the German Cancer Society (DKG) was associated with a difference in overall survival (primary outcome measure, other primary outcome measures 1-/5-year survival and 30-day mortality). The choice of cancer entities was based upon the presence of an implemented DKG certification program at the time of the study concept and the mappability of the addressed entities in the billing data of the statutory health insurance system (SHI) and Clinical Cancer Registers (CCR). The WiZen study was approved by the Ethics Committee of the TU Dresden (reference number: EK95022019), was registered at ClinicalTrials.gov (ID: NCT04334239), and conducted in accordance with the Declaration of Helsinki and the EU General Data Protection Regulation.

The studies were based on nationwide pseudonymized SHI data from all insurees of the AOK statutory health insurance carrier for the years 2006–2017, provided by the AOK Research Institute. For this purpose, insuree-related data from the service areas of insuree master data (Section 284 of the German Social Security Code [SGB] V), outpatient care (Section 295 SGB V), inpatient care (Section 301 SGB V), and medication prescriptions (Section 300 para. 1 SGB V) were combined across sectors. In accordance with good practice of secondary data analysis (17), a diagnosis-free phase (2006 to 2008) was used to determine incidental cancer, so the analysis period was in fact from 2009 to 2017.

Structural characteristics of the hospitals (number of beds, university hospital status, teaching hospital status, ownership) were obtained from the structured quality reports according to Section 136 SGB V. The DKG provided data on DKG certification of the hospitals (start, end, suspension of certificate validity).

The data were pseudonymized at the patient and hospital level and transmitted in encrypted form. Data analysis was conducted at the Center for Evidence-based Healthcare (ZEGV) of the TU Dresden.

Patients aged at least 18 years at diagnosis and with an initial diagnosis of the considered cancer entities according to ICD-10-GM within the time period 2009 to 2017 were included. The choice of ICD numbers was defined by panels of clinical experts. A detailed description of the codes used and other methodological details may be found in the final WiZen trial report (18). Patients whose initial diagnosis date and death date were identical and/or who had implausible data were excluded. Other exclusions were made for change of health insurance carrier, patients without an inpatient primary diagnosis of the observable entity, and initial treatment (“index treatment”) in a hospital during the year prior to the hospital receiving its DKG certification. This takes into account the fact that certification effects may not only present on conferring the certificate, but at least in part already in preparation for certification.

The primary endpoint of the SHI data analysis as presented here was overall survival time as from the start of the index treatment. Survival times of patients without a date of death or with a death date after 2017 (end of follow-up) were treated as right-censored.

Intervention was considered to be the initial treatment after diagnosis in a hospital certified by the DKG. Patients with initial treatment in a DKG-certified hospital formed the intervention group; patients with initial treatment in a non-certified hospital were the control group. Initial treatment was considered to be the time point, if documented, of excision of the primary diagnosis of the respective entity, otherwise the first stay in hospital. Since direct assignment was not possible where hospitals had multiple sites, all sites were assigned the status of a DKG-certified hospital if one of the hospitals held this status.

Age (18–59, 60–79, ≥80 years), sex, and disease severity (distant metastases, other oncological diseases, comorbidities) were included as influencing variables (confounders) at patient level for risk adjustment of the estimated certification effects (eTable 1). The year of initial treatment was also taken into consideration to protect against secular trends. The entity-specific selection of comorbidities was made with the involvement of clinical expertise and according to Elixhauser et al. prior to data analysis (19). At hospital level, we took into consideration the number of beds (1–299, 300–499, 500–999, ≥1000 beds), the function of the institutions as a university hospital and/or teaching hospital, and ownership (public/non-profit/private) as influencing variables (20).

eTable 1. Estimated effect of initial treatment in a certified center according to duration of certification, in each case in comparison with an initial treatment in a non-certified hospital (overall survival) for the observable entities*.

	Duration of certification
	<1 year	1– <2 years	2– <5 years	5 + years
Colon cancer
HR [95% CI]	0.96 [0.92; 1.01]	0.91 [0.88; 0.96]	0.89 [0.86; 0.93]	0.9 [0.86; 0.94]
p-value	0.102	<0.001	< 0.001	< 0.001
Rectal cancer
HR [95% CI]	0.93 [0.88; 0.99]	0.93 [0.88; 0.98]	0.89 [0.86; 0.93]	0.86 [0.82; 0.91]
p-value	0.023	0.008	< 0.001	< 0.001
Pancreatic cancer
HR [95% CI]	0.91 [0.85; 0.97]	0.88 [0.81; 0.94]	0.89 [0.84; 0.95]	0.77 [0.68; 0.87]
p-value	0.004	< 0.001	< 0.001	< 0.001
Breast cancer
HR [95% CI]	0.82 [0.75; 0.89]	0.77 [0.72; 0.84]	0.78 [0.74; 0.82]	0.74 [0.71; 0.78]
p-value	< 0.001	< 0.001	< 0.001	< 0.001
Cervical cancer
HR [95% CI]	0.89 [0.77; 1.03]	0.86 [0.73; 1.01]	0.81 [0.71; 0.93]	0.72 [0.6; 0.88]
p-value	0.112	0.067	0.002	0.001
Endometrial cancer
HR [95% CI]	0.91 [0.81; 1.03]	0.87 [0.77; 0.99]	0.95 [0.86; 1.05]	0.95 [0.82; 1.09]
p-value	0.131	0.038	0.332	0.471
Ovarian cancer
HR [95% CI]	0.86 [0.77; 0.95]	0.97 [0.87; 1.08]	0.89 [0.81; 0.97]	0.76 [0.67; 0.87]
p-value	0.004	0.600	0.008	< 0.001
Lung cancer
HR [95% CI]	0.98 [0.94; 1.02]	0.97 [0.93; 1.01]	0.96 [0.93; 1]	0.93 [0.88; 0.98]
p-value	0.398	0.180	0.069	0.006
Prostate cancer
HR [95% CI]	0.81 [0.74; 0.88]	0.83 [0.76; 0.9]	0.85 [0.79; 0.92]	0.8 [0.72; 0.88]
p-value	< 0.001	< 0.001	< 0.001	< 0.001
Neuro-oncological cancer
HR [95% CI]	0.97 [0.88; 1.07]	0.91 [0.81; 1.03]	0.88 [0.78; 0.98]	0.72 [0.1; 5.13]
p-value	0.485	0.127	0.025	0.743
Head and neck cancer
HR [95% CI]	0.98 [0.91; 1.05]	0.94 [0.86; 1.02]	0.9 [0.83; 0.97]	0.82 [0.61; 1.09]
p-value	0.562	0.125	0.009	0.169

* Adjustment variables: age, sex, distant metastasis, other oncological disease, Elixhauser comorbidities, number of hospital beds, teaching hospital, university hospital, hospital ownership, year of index treatment – dummy coded (based on SHI data)

The effect of treatment in a certified versus non-certified hospital on overall survival was modelled using multivariable Cox regression analysis, taking into account the above mentioned influencing variables/confounders, and hazard ratios (HR) were calculated with 95% confidence intervals [95% CI]. Entity-specific differences in disease progression were considered using baseline hazard functions. Introducing a random effect (shared frailty) at hospital level allows Cox models to take into account potential correlation of outcomes of patients within the hospitals (21).

As part of sensitivity analyses, stratified calculations were performed according to hospital size (number of beds), certificate duration (for breast cancer, as this is where certification has existed the longest), and for patients with and without distant metastases at the time of index treatment.

Results

Based on an overall population of around 22 million adult AOK insurees in the year 2017 (22) and after applying inclusion and exclusion criteria, cohorts of patients with incidental cancer for the examined entities ranging between 10 596 (cervical cancer) and 172 901 (lung cancer) individuals were included in the study. Across entities, there was no clear difference between certified and non-certified hospitals in terms of patient characteristics, but larger hospitals were more frequently DKG-certified and smaller hospitals less frequently so (Table 1). Despite a moderate increase over time in the proportion of patients treated in DKG-certified hospitals, the majority of patients were treated in non-DKG-certified hospitals for all cancer entities during the period under review, with the exception of breast cancer (Figure 1).

Table 1. Analysis populations certified/non-certified, number (n), interquartile range, (Q1; Q3), and percentage (%) by characteristic group (age in years/size of hospital) for all observable entities.

	Patients certified				Hospitals certified
Characteristic	yes	no		Characteristic	yes	no
Colon cancer				Colon cancer
Total (n)	40 861	68 826		Total (n)	311	777
Age in years (Q1; Q3)	(67; 81)	(68; 82)		1–499 beds (%)	51.4	92.1
Sex female (%)	48.9	50.5		500+ beds (%)	48.6	7.9
Rectal cancer				Rectal cancer
Total (n)	22 086	29 370		Total (n)	310	741
Age in years (Q1; Q3)	(63; 79)	(65; 80)		1–499 beds (%)	51.3	92
Sex female (%)	39.1	40.8		500+ beds (%)	48.7	8.0
Pancreatic cancer				Pancreatic cancer
Total (n)	5426	39 892		Total (n)	96	955
Age in years (Q1; Q3)	(64; 79)	(67; 81)		1–499 beds (%)	20.8	86.2
Sex female (%)	50.8	52.3		500+ beds (%)	79.2	13.8
Breast cancer				Breast cancer
Total (n)	91 269	52 451		Total (n)	280	730
Age in years (Q1; Q3)	(56; 76)	(57; 78)		1–499 beds (%)	53.9	88.8
Sex female (%)	99.1	98.8		500+ beds (%)	46.1	11.2
Cervical cancer				Cervical cancer
Total (n)	2825	7771		Total (n)	128	679
Age in years (Q1; Q3)	(45; 68)	(47; 73)		1–499 beds (%)	27.3	83.7
				500+ beds (%)	72.7	16.3
Endometrial cancer				Endometrial cancer
Total (n)	5879	24 222		Total (n)	130	735
Age in years (Q1; Q3)	(60; 77)	(62; 79)		1–499 beds (%)	29.2	85.3
				500+ beds (%)	70.8	14.7
Ovarian cancer				Ovarian cancer
Total (n)	4763	16 031		Total (n)	129	865
Age in years (Q1; Q3)	(55; 76)	(59; 79)		1–499 beds (%)	28.7	86.7
				500+ beds (%)	71.3	13.3
Lung cancer				Lung cancer
Total (n)	34 884	139 115		Total (n)	61	1 091
Age in years (Q1; Q3)	(60; 75)	(62; 77)		1–499 beds (%)	55.7	82.5
Sex female (%)	34.3	33.0		500+ beds (%)	44.3	17.5
Prostate cancer				Prostate cancer
Total (n)	24 430	57 112		Total (n)	116	862
Age in years (Q1; Q3)	(63; 74)	(65; 76)		1–499 beds (%)	43.1	83.9
				500+ beds (%)	56.9	16.1
Neuro-oncological cancer				Neuro-oncological cancer
Total (n)	4703	58 032		Total (n)	29	1191
Age in years (Q1; Q3)	(52; 74)	(57; 78)		1–499 beds (%)	10.3	83.5
Sex female (%)	55.6	54.0		500+ beds (%)	89.7	16.5
Head and neck cancer				Head and neck cancer
Total (n)	8173	44 576		Total (n)	44	828
Age in years (Q1; Q3)	(57; 75)	(56; 74)		1–499 beds (%)	13.6	80.1
Sex female (%)	26.9	24.9		500 + beds (%)	86.4	19.9

*: Percentages correspond to column percentages

Figure 1.

Proportion of patients treated in a certified hospital according to entity over time, SHI data

For all entities, there was consistently a longer overall survival time for patients who had received initial treatment in a certified hospital. In the fully adjusted regression analyses, the overall survival advantages of patients in DKG-certified hospitals ranged between three percent (lung cancer, HR = 0.97; [95% CI: 0.94; 1.00]) and 23 percent (breast cancer, HR = 0.77; [0.74; 0.81]), with ARRs varying between 0.62 months (lung cancer) and 4.61 months (cervical cancer). The model’s goodness-of-fit using Harrell’s C index varied between 0.67 and 0.81 (Table 2, Figure 2). Survival benefits were also consistently evident for all cancer entities at specific follow-up times (30 days, 1 and 5 years) (eTable 2).

Table 2. Estimated effect of initial treatment in a certified center in comparison with an initial treatment in a non-certified hospital (overall survival) for the observable entities using adjusted Cox regression analysis*¹.

	p-value		Median survival (certified)
HR [95% CI]	without correction*2	with correction	yes	no	Harrell’s C*3	ARR*4 (in months)
Colon cancer
0.92 [0.89; 0.95]	<0.001	<0.001	4.82	4.28	0.74	2.53
Rectal cancer
0.90 [0.87; 0.94]	<0.001	<0.001	4.62	4.05	0.73	2.88
Pancreatic cancer
0.88 [0.84; 0.93]	<0.001	<0.001	0.44	0.37	0.67	2.00
Breast cancer
0.77 [0.74; 0.81]	<0.001	<0.001	–*5	–	0.80	3.54
Cervical cancer
0.83 [0.76; 0.92]	<0.001	0.001	–	7.42	0.76	4.61
Endometrial cancer
0.92 [0.86; 0.99]	0.036	0.143	–	–	0.76	1.54
Ovarian cancer
0.88 [0.82;0.95]	0.001	0.002	3.23	2.77	0.75	3.80
Lung cancer
0.97 [0.94; 1.00]	0.055	0.219	0.67	0.64	0.69	0.62
Prostate cancer
0.83 [0.78; 0.88]	<0.001	<0.001	–	–	0.81	3.01
Neuro-oncological cancer
0.92 [0.86; 0.99]	0.035	0.142	5.67	4.72	0.73	2.34
Head and neck cancer
0.94 [0.89; 1.00]	0.038	0.152	4.13	3.73	0.68	1.79

ARR, absolute risk reduction (difference in simulated survival for a follow-up period of 8 years from initial treatment); HR, hazard ratio; CI, confidence interval,

*¹ Adjustment variables: age, sex, distant metastasis, other oncological disease, Elixhauser comorbidities, number of hospital beds, teaching hospital, university hospital, hospital ownership, year of index treatment – dummy coded (based on: SHI data)

*² Bonferroni correction (multiplied by 4, because of 4 primary outcomes)

*³ Pairs are formed between all individuals considered in the regression, and their real and simulated survival times are compared. Pairs are excluded when comparison of their actual survival time is not possible due to censoring. The proportion of concordant pairs is determined among all remaining pairs (i.e., the actual, as well as the simulated, survival time is longer for the same individual). This proportion is a measure for the model’s goodness-of-fit – the greater the value, the better can survival time be predicted.

*⁴ Reading example: The absolute risk reduction refers to the difference between the simulated survival times as from the initial treatment in and outside of a certified hospital. For colon cancer, this results in a mean survival time of 2.53 months longer as from initial treatment in a certified hospital compared with initial treatment in a non-certified hospital.

*⁵ Calculation not possible because the adjusted survival rates in the follow-up period were not below 0.5

Figure 2.

Hazard ratios for overall survival in hospitals with, versus those without, certification for all observable entities. Hazard ratios <1 indicate an association between longer overall survival and treatment in a certified center.

eTable 2. Overall survival rates (in %, with [95% CI]) for specific time points after index treatment by center status per entity.

Certified	30 days	1 year	5 years
Colon cancer
yes	94.0 [93.7; 94.2]	75.6 [75.1; 76.0]	48.7 [48.1; 49.3]
no	92.3 [92.1; 92.5]	73.2 [72.9; 73.6]	46.7 [46.3; 47.1]
Rectal cancer
yes	95.9 [95.6; 96.2]	78.6 [78.1; 79.2]	49.2 [48.4; 50.0]
no	94.3 [94.1; 94.6]	73.9 [73.4; 74.4]	43.3 [42.7; 44.0]
Pancreatic cancer
yes	87.0 [86.1; 87.9]	37.6 [36.2; 39.0]	11.5 [10.2; 12.9]
no	81.3 [80.9; 81.7]	24.8 [24.4; 25.3]	6.5 [6.2; 6.8]
Breast cancer
yes	99.0 [99.0; 99.1]	94.2 [94.0; 94.3]	78.3 [77.9; 78.6]
no	98.2 [98.1; 98.4]	91.3 [91.1; 91.6]	71.9 [71.4; 72.3]
Cervical cancer
yes	97.8 [97.2; 98.3]	82.3 [80.9; 83.8]	59.1 [56.9; 61.4]
no	96.3 [95.9; 96.7]	76.8 [75.8; 77.8]	53.3 [52.1; 54.5]
Endometrial cancer
yes	98.2 [97.8; 98.5]	87.2 [86.3; 88.1]	66.7 [65.2; 68.3]
no	98.1 [97.9; 98.3]	86.7 [86.2; 87.1]	65.0 [64.3; 65.6]
Ovarian cancer
yes	94.7 [94.1; 95.3]	75.8 [74.5; 77.1]	42.6 [40.8; 44.5]
no	90.6 [90.1; 91.0]	65.1 [64.4; 65.9]	35.7 [34.9; 36.6]
Lung cancer
yes	94.1 [93.9; 94.4]	58.9 [58.3; 59.4]	28.0 [27.4; 28.6]
no	86.5 [86.4; 86.7]	41.8 [41.5; 42.1]	16.9 [16.7; 17.1]
Prostate cancer
yes	99.1 [99.0; 99.3]	94.0 [93.6; 94.3]	78.4 [77.8; 79.1]
no	98.4 [98.3; 98.5]	90.3 [90.0; 90.5]	71.2 [70.8; 71.7]
Neuro-oncological cancer
yes	97.1 [96.6; 97.6]	77.6 [76.3; 78.9]	61.0 [58.4; 63.7]
no	94.0 [93.8; 94.2]	66.2 [65.8; 66.5]	48.0 [47.5; 48.4]
Head and neck cancer
yes	98.0 [97.7; 98.3]	77.1 [76.1; 78.1]	47.0 [45.1; 49.1]
no	96.9 [96.8; 97.1]	74.6 [74.2; 75.0]	45.3 [44.8; 45.9]

All sensitivity analyses performed showed a very high degree of robustness of results and no apparent evidence for violation of the proportionality assumption with regard to center attributes (eFigure 1). Thus, in the stratified analysis, higher survival chances were evident in certified centers irrespective of the number of beds. For almost all entities, it was also shown that the association of treatment in a certified center with longer overall survival tended to be stronger the longer the certificate was held (eTable 1). The complete set of analysis results for all model specifications may be found in the final WiZen trial report (18).

eFigure 1 a–f.

Overall survival according to certification per entity (Kaplan-Meier estimate)

Discussion

Drawing on a large, comprehensive database, the WiZen project shows that, for the entities studied, initial treatment in a certified – as compared with a non-certified – hospital is associated with longer overall survival for patients with incidental cancer. The overall survival advantages ranged between three and 23 percent for the different entities and cohorts and after adjusting for different influencing variables at patient and hospital level.

Apart from the results of the SHI data analyses presented here, the WiZen study also evaluated pseudonymized data from the Clinical Cancer Registers (CCR) of Brandenburg/Berlin, Dresden, Erfurt, and Regensburg. This data is complementary to the SHI data and includes initial diagnoses for the period 2009 to 2017, together with demographic information and disease-specific data. The results of the CCR analysis confirm the association between treatment in a certified center and longer overall survival. In the cancer register cohorts, the survival benefit in certified hospitals was more pronounced in patients with localized and locally advanced stages (1–3) than in patients with advanced stage 4, where treatment is palliative and the primary treatment aim is often not to prolong overall survival. Furthermore, the CCR data also revealed longer recurrence-free survival for patients with R0 resected tumors treated in certified centers (18).

Several subgroup and sensitivity analyses indicate a high degree of robustness of treatment advantages in certified hospitals (18). The results can therefore be considered reliable and valid. Our analysis presents another important indicator for the importance of certification with respect to cancer treatment by showing, in most cases, greater overall survival advantages the longer certification has been in place (18, 23).

One methodological strength of the WiZen study is its large base population comprising around 22 million adult insurees of the AOK statutory health insurance carrier, which demonstrates a high degree of external validity. We consider the results originating from the period 2009 to 2017 to be fundamentally applicable to the present situation. Since caseload may have an impact on relevant results such as survival (24–26) and a minimum volume is required for DKG certification, some of the results could be due to volume effects. Since the SHI data were derived from one single health insurance company, it was not possible to quantify the total volume of patients in the respective attending hospitals. By using different data sources, observing a number of entities, and taking broad consideration of relevant patient, tumor, and hospital characteristics in terms of risk adjustment, the impact of certification was investigated and validated. It was not possible to exclude a potential bias due to the absence of randomization, but this was kept to a minimum by the comprehensive risk adjustment, allowing comparability of the certification effect across different cancer entities. Regardless of certification status, there was a positive association between hospital size and lower total mortality for the majority, albeit not all, of the examined entities. Hospital size (number of beds) served as a surrogate parameter for caseload. It cannot be excluded that some of the observed impact from certification was also induced by the caseload which is, after all, a criterion of certification. Since no details were included regarding patient socioeconomic status, a potential selection bias cannot be excluded, for instance, as a result of disproportionately more frequent initial treatment of more educated and higher-income patients in certified hospitals. Although no definitive causal conclusions can be drawn based on the study design, it seems unlikely to us that the association between initial treatment at a certified center and longer overall survival is not causal, given the known effect mechanism (guideline implementation, interdisciplinary treatment), control group design, statistical control for a large set of confounding variables, consistency of the presented association as shown by two complementary data sources, and the size of the study population.

DKG certification of hospitals ensures that evidence-based guidelines are implemented in line with the National Cancer Plan (5). According to the WiZen study, this is associated with relevant advantages for patients with respect to overall survival and recurrence-free survival (18). Even after applying a model maximally adjusted for confounders, the WiZen study reveals that there is a relatively strong impact on medicine from the certification of oncological centers for the vast majority of tumor entities. Certifying hospitals is therefore a highly effective complex intervention to reduce the disease burden of cancer in Germany. However, at the moment, cancer patients are not being specifically referred to certified hospitals: During the study period from 2009 to 2017 and across all entities with the exception of breast cancer, the majority of cancer patients were treated in hospitals not certified by the DKG. The systematic referral of cancer patients to certified centers—for instance, by applying appropriate specifications for service groups to hospitals within the framework of the intended hospital reform or also via corresponding regulations by the Federal Joint Committee (G-BA)—would have enormous potential: Extrapolated for the entire German population, around 33 000 life years could have been saved each year during the study period if all patients with incidental cancer of the examined entities had been referred to a certified center (total cancer mortality for selected cancer entities in Germany 2017: 147 662 patients [27]). The number of life years saved is derived from the ARR as calculated in the present study, extrapolated for all patients treated outside of certified centers. The preconditions for the validity of this extrapolation are, on the one hand, that the capacity and accessibility of the certified centers is suitable to meet the higher treatment volume and, on the other, that the observed effects are also applicable to those patients not previously treated in certified centers. Both conditions are met in our opinion: The characteristics of those treated in certified centers are essentially the same as those treated outside, confirming that the result are applicable to both groups. Certified centers are currently spread throughout the country in 435 hospitals (www.oncomap.de). This indicates that capacity is available to centralize initial inpatient treatment of cancer patients in certified centers. As things stand, however, exact calculations are still wanting. Usually, only stages of the oncological treatment pathway are conducted on an inpatient basis in hospital. Many treatment stages can also be undertaken near the patient’s home by qualified cooperation partners associated with the centers. In the present study, initial treatment was carried out in a DKG-certified center after the diagnosis had been established. Changing hospitals after this initial treatment is of little relevance for the present investigation, especially since these changes often occur between different treatment partners yet within the network of certified centers. According to the latest figures available for the year 2021, 41 percent of all incidental cancer cases in Germany are still being treated in non-certified institutions (PD Dr. Wesselmann, personal communication, DKG, 03/2023).

Referring cancer patients specifically to certified hospitals therefore has a very high potential and should be implemented as healthcare policy. In this respect, the positive evaluation of the WiZen study by the Innovation Committee of the Federal Joint Committee (G-BA) is to be welcomed. The resolution called on the G-BA’s Quality Assurance Subcommittee, among other things, to take the results of the WiZen study into account when defining minimum requirements for quality of structure, process and outcome, and for the development of data-based quality assurance procedures (28). The current recommendations of the Government Commission for a Modern and Needs-Based Hospital Care also offer opportunities to consider the WiZen results. For example, certification based on DKG criteria could be defined as a structural requirement for service groups treating cancer patients and thus become a precondition for the billing of corresponding healthcare services (29). Initial cancer treatment in certified hospitals appears to be wise, also from a health economic point of view: Thus, for bowel cancer, for example, a cost-effectiveness analysis by the German Cancer Research Center showed a longer survival time with lower treatment costs in certified versus non-certified hospitals (13).

eBox 1. Members of the WiZen study group.

• Prof. Dr. Jochen Schmitt

• PD Dr. Olaf Schoffer

• Dr. Veronika Bierbaum

• Christoph Bobeth

• Dr. Martin Rößler (ZEGV)

• Prof. Dr. Monika Klinkhammer-Schalke

• Kees Kleihues-van Tol, Bianca Franke (ADT)

• Dr. Michael Gerken,

• Jan Kurz

• Dr. Patricia Lindberg-Scharf

• Dr. Brunhilde Steinger (TZR)

• Christian Günster

• Patrik Dröge

• Thomas Ruhnke, Andreas Klöss (WIdO)

• Dr. Jürgen Malzahn (AOK BV)

• Dr. Torsten Blum (Berlin)

• PD Dr. Johannes Bründl (Regensburg)

• Prof. Dr. Marius Distler (Dresden)

• PD Dr. Marie-Therese Forster (Frankfurt/M.)

• Prof. Dr. Alois Fürst (Regensburg)

• Prof. Dr. Frank Griesinger (Oldenburg)

• Prof. Dr. Viktor Grünwald (Essen)

• Prof. Dr. Peter Hau (Regensburg)

• Dr. Bernd Hoschke (Cottbus)

• PD Dr. Elisabeth C. Inwald (Regensburg)

• PD Dr. Karin Kast (Dresden/Cologne)

• Prof. Dr. Rainer Keerl (Straubing)

• Dr. Max Kemper (Dresden)

• Prof. Dr. Oliver Kölbl (Regensburg)

• Prof. Dr. Peter Kummer (Regensburg)

• PD Dr. Thomas Papathemelis (Amberg)

• Prof. Dr. Pompiliu Piso (Regensburg)

• Prof. Dr. Bettina Rau (Neumarkt)

• Prof. Dr. Dr. Torsten E. Reichert (Regensburg)

• Prof. Dr. Christoph Reissfelder (Mannheim)

• Prof. Dr. Anton Scharl (Amberg)

• Prof. Dr. Christian Thomas (Dresden)

• Prof. Dr. Corinna Seliger (Heidelberg)

• Prof. Dr. Pauline Wimberger (Dresden)

eBox 2. Influencing variables used in the data analysis.

• Patient level

  • age at index treatment (in years, categorical: 18–59, 60–79, 80+)

  • sex (female/male)

  • distant metastasis (ICD-10-GM: C78–C79) before/at time of initial diagnosis of the observable entity (no/yes)

  • other oncological disease before/at time of initial diagnosis of the observable entity (no/yes)

  • Elixhauser comorbidities as selected by clinical experts (no/yes)

• Hospital level

  • no. of beds (1–299, 300–499, 500–999, 1000 +)

  • university hospital (no/yes)

  • teaching hospital (no/yes)

  • ownership (public/non-profit/private)

• Other information

  • year of initial treatment

eFigure 1 g–k.

Overall survival according to certification per entity (Kaplan-Meier estimate)