Inpatient TIA and stroke care in adult patients in Germany - retrospective analysis of nationwide administrative data sets of 2011 to 2017

Background Comprehensive administrative data on TIA and stroke cases and treatment modalities are fundamental for improving structural conditions and adjusting future strategies of stroke care. Methods The nationwide administrative database (German federal statistical office) was used to extract all adult inpatient TIA and stroke cases and corresponding procedural codes for the period 2011–2017. Numbers were specified according to age, sex, stroke unit (SU) and critical care treatment (ICU), early transfer, and in-hospital mortality. Findings Inpatient adult TIA/stroke cases increased from annually 102,406 / 250,199 (2011) to 106,245 / 264,208 (2017). 84% of strokes were ischemic (AIS) also having the highest relative increase most likely due to more accurate coding within the time period, 68.2% of AIS were treated on SUs. 78% of hemorrhagic strokes were intracerebral hematomas (ICH; rather than subarachnoid hemorrhages [SAH]). Hemorrhagic strokes were increasingly treated on SUs (32.6% [2011], 37.8% [2017]). 68.8% of SAH were treated on ICUs (ICH:36.3%, AIS:10.3%). Early transfer in AIS increased (2.0 to 3.1%). Hemorrhagic strokes were associated with higher in-hospital mortality (SAH:19.6%, ICH:28.2%, AIS:7.3%). Interpretation The absolute increase of strokes presumably reflects the aging society and more awareness for cerebrovascular disease. The relative increase of AIS may be attributable to an increased neurological expertise. The increasing amount of early transfers in AIS reflects new specialized treatment options. Our findings reflect the need for structural adjustments in inpatient stroke care.


The numbers of treated adult TIA and strokes cases
in German hospitals has further increased in the observed time period 2011-2017. The proportion of hemorrhagic to ischemic stroke has not changed. 2. The rate of treatment on specialized wards like strokes units (SU) and intensive care units (ICU) has further increased for both ischemic and hemorrhagic stroke. 3. Hemorrhagic strokes are more frequently treated on ICUs, whereas cerebral infarctions and TIAs are more often treated on SUs. Hemorrhagic strokes are associated with higher in-hospital mortality. 4. For AIS, early transfer, e.g., to specialized neurovascular centers providing mechanical thrombectomy has increased in the time period. 5. The analysis of timely evolution of administrative data is important for future adjustments of infrastructure for inpatient stroke care.

Introduction
The Global Burden of Disease Study recently provided data on global, regional and country-specific epidemiological data of stroke [1][2][3][4]. Like in many other epidemiological analyses, insight in stroke epidemiology mostly relies on population-based, regionally limited, observational cohort or hospital-based registries, all of which bare specific constraints [5][6][7]. Due to the necessity to code both diagnoses and treatment procedures for reimbursement, the German DRG registry provides accurate and comprehensive data not only on all inpatient ischemic stroke / TIA cases and treatment modalities in German hospitals [8]. Analysis of all hospitalized stroke cases can provide new insights into evolving trends of both ischemic and hemorrhagic stroke subtypes in everyday practice. Therefore, data from the German federal statistical office of all adult stroke patients hospitalized from 2011 to 2017 were used to identify frequencies of all inpatient stroke subtypes as well as treatment modalities on specialized wards. In addition, age-and sex-related differences, early transfer rates and in-hospital mortality rates were evaluated.

Methods
Analyses were based upon the latest German Diagnosis-Related Groups (G-DRG) data provided by the German federal statistical office (DRG-statistic, www.destatis.de) for the years 2011 to 2017. All in-patient stroke cases are encoded according to ICD-10-GM 1 and relevant operating and procedure keys (OPS-301 codes) issued by the German Institute of Medical Documentation and Information (DIMDI). Here, the following ICD main diagnosis codes were considered: G45.0-G45.99 (transient ischemic attack, TIA); I60.0-I60.9 (subarachnoid hemorrhage, SAH); I61.0-I61.9 (intracerebral hemorrhage, ICH); I63.0-I63.9 (cerebral infarction, AIS); I64 (stroke, not specified as hemorrhage or infarction). All case numbers were aggregated at the level of the 3-digit ICD codes. The agestandardized rates were calculated using the standard population of Germany based on the census of 2011 (Federal Statistical Office: Statistics on Natural Population Movement) [9]. In addition, the following OPS codes in combination with each considered main diagnosis were analyzed for all stroke subtypes 2 : 8-980 (basic intensivecare treatment); 8-98f (complex intensive-care treatment; from 2013 onwards); 8-981.0 (stroke unit treatment for more than 24 h and less than 72 h); 8-981.1 (stroke unit treatment for more than 72 h); 8-98b.*0/*1 (other acute stroke treatment without / with tele-consultation). For some analyses, OPS 8-980 and 8-98f were combined, as were the 3 subtypes of Stroke Unit care. Both first-ever and recurrent stroke cases were included, because the coding system cannot differentiate between them. Likewise, recurrent cerebrovascular events during hospital stay could not be analysed because these events are not coded consistently as a separate secondary diagnosis. Patients being transferred between hospitals during one treatment episode, (discharge key 06; transfer to another hospital), were censored accordingly in order to avoid any possible double/multiple coding. In addition, we assessed the number of acute stroke patients being transferred from one hospital to the other in the hyperacute phase for specific therapies such as mechanical thrombectomy, neurosurgical operations or intensive care treatment (so-called "hourly cases"). In-hospital mortality was assessed using discharge key 07 (death during hospital stay). For TIA and stroke subtypes, mean age with standard deviation, sex, and in-hospital mortality rate are provided. Only adult patients were considered. 3 The maps of the regional frequency of ICH and AIS in Germany are based on the 413 administrative districts and independent cities in 2017. The age standardized rates are calculated for each district / city. The pre-specified primary hypotheses were as follows: 1.  Table 1. Characteristic numbers comparing AIS (ICD I63) and ICH (ICD I61) cases are given in Table 2. The distribution of hospitalized patients with TIA and stroke subtypes is shown in Fig. 1. As most of epidemiological studies do not include patients with TIA, we present two illustrations with/without TIA cases. . Patients with ICH had a 4.9 times higher odds ratio (95%CI 4.4-5.5) to be treated on an ICU compared with AIS, which was more pronounced for younger age (see Additional file 1: Figure S1). SAH patients were more often admitted to ICUs than ICH patients (67% vs. 36% [2017]  10.3% of AIS were treated on ICUs in 2017 (see also Additional file 2: Figure S2). The proportion of AIS treated on "other stroke units" increased up to 2014 and stayed constant since then, accounting for 8.2% of all cases with AIS in 2017 (in 2011: without/with tele-consultation: 4.6/0.9%; in 2017: without/with tele-consultation: 5.2/3.0%). Early acute transfer to another hospital for further treatment increased only for AIS from 2.0 [2011] to 3.1% [2017], while for hemorrhagic stroke there was a decrease of transfer rates (24.8 to 19.9% for SAH, 13.5 to 12.6% in ICH). Inhospital mortality was highest for hemorrhagic stroke (SAH 19.6% / ICH 29.3% [2017]) compared to AIS (7.2%) and TIA (0.3%). The risk for in-hospital mortality was 6.4 fold (95%CI 5.98-9.92) higher for hemorrhagic stroke (see Additional file 3: Figure S3), especially in younger patients. In ICH, in-hospital mortality increased from 28.8% (2011) to 29.3% (2017), in AIS, in-hospital mortality decreased from 7.7% (2011) to 7.2% (2017).

Discussion
Our analyses refer to all hospitalized adult TIA and stroke cases in Germany from 2011 to 2017, in total representing 2.55 million cases. Due to the specific characteristics of the German coding system, our data represent a robust image of the inpatient incidence and treatment reality of TIA/stroke in Germany. Our results confirm previously published data and trends [18], but also cast some genuine insights on structural conditions. The age-adjusted inpatient rate for SAH of 9.0 per 100,  000 inhabitants in 2017, with more women being affected, is in line with previous data [11]. Since 10-15% of SAH die before hospital admission [17], a slightly higher incidence has to be assumed. The age-adjusted rates for hospitalized ICH and AIS were 31 and 256 per 100,000 inhabitants in 2017. Assuming a 10 year recurrence rate of up to 39% [12], the rate of first-ever ischemic or hemorrhagic stroke has to be reduced accordingly in our cohort. The Global Burden of Stroke study reported a worldwide age-adjusted annual firstever stroke incidence rate of 258 in 2010 [1], varying between 217 and 281 depending mainly on income status of the respective country. However, treatment data such as ours is difficult to compare with epidemiological data. In addition, the epidemiological term "stroke" used by the Global Burden of Disease Study Group excluded TIAs but included ICH and SAH [10]. Considering coded TIA cases is controversial due to numerous differential diagnoses. Béjot et al. reported on a varying ageadjusted incidence rate for stroke and TIA in Europe between of 95 and 290/100,000 and 28 and 59/100,000, respectively, with higher incidence in eastern countries and lower incidence in southern countries for the early twenty-first century [5]. We also found a regional heterogeneity in the inpatient rates within Germany and have illustrated this in Fig.2 by graphic demonstration for ICH and AIS referring to all 413 administrative districts and cities of the country [8]. Our data cannot reveal, however, whether these differences are real differences in incidence, e.g., due to socio-economic  127  119  134  9  7  11  32  34  31  259  261  257  9  7  10   2014  128  122  134  9  7  11  32  33  30  256  263  250  5  4  5   2017  121  115  126  9  7  11  31  32  29  256  268  244  2  2  3 ICD, international classification of diseases; TIA, transient ischemic attack; SAH, subarachnoid hemorrhage; ICH, intracranial hemorrhage; AIS, acute ischemic stroke; t, total; m, male; f, female variety or if this is solely bias. In the mentioned review of population-based studies it is also reported that ageadjusted incidence rates in men are 1.2 to 2 times higher than in women, possibly attributable to a higher prevalence of traditional vascular risk factors in men [5]. In our analysis we also found this difference for AIS and ICH, which even increased between 2011 and 2017. As for the proportions of different stroke subtypes, reviews describe wide variations for AIS from 55 to 90% of all cases, ICH from 10 to 25% and SAH from 0.5 to 5% [5].
In-hospital mortality slightly decreased in AIS patients during the observed time period, which can be explained by improved treatment strategies for AIS (e.g. higher use of thrombolysis, mechanical thrombectomy and hemicraniectomy). In contrast, in-hospital mortality slightly increased in ICH patients from 2011 to 2017, which might be attributed to an increasing use of oral anticoagulation in the general population [19].
Most of the review data refer to population-based, regionally limited, observational cohorts or hospital-based registries. Limitations of these study types have been discussed as focusing on urban areas, covering relatively small populations thus not reflecting the "true" composition of the population. Furthermore, the above mentioned studies have not analyzed the dispersion of treatment modalities such as specialized stroke unit (SU) or intensive care (ICU), which is a beneficial factor for patients´outcome. Given high hospitalization rates in Germany, the German DRG statistic has proved as a useful tool to generate valid data on both (diagnoses-related) inpatient rates as well as the distribution of (OPS code-related) treatment modalities such as i.v. thrombolysis, mechanical thrombectomy, or ICU-and SU-care [8,18], because correct and complete coding of both DRG-and OPS-codes is a prerequisite for reimbursement. While comprehensiveness of data is facilitated hereby, false economic incentives can be triggered. The German DRG-system is closely guarded, though, by a specialized medical service of the public health insurance system to avoid false coding. We therefore believe that most of the limitations discussed for administrative coding data (16) do not apply to the German system. Another inherent limitation, however, is the limitation to cases coded as such. Considering early transfers between hospitals in the first few hours, cases are considered as a transfer once they are coded as inpatient cases in the primary hospital (so called "hourly cases"). If they are not coded as an "inpatient" but only as an "outpatient" case, they will not appear as a transfer in the dataset, but as a regularly inpatient case of the secondary hospital. Regulatory rules provide the first mentioned proceeding even though it is unknown if all hospitals claim this proceeding. Therefore, transfer rates might be underestimated. However, we believe that due to reimbursement reasons every hospital has an inherent motivation to code these patients as "inpatient" cases. It can further be speculated if decreasing rates of early transfer rates as displayed may reflect improved allocation strategies of regional alliances between hospitals and emergency services, resulting in fewer necessary secondary transfers. Furthermore, our analysis is restricted to hospitalized cases only and cannot differentiate between first-ever and recurrent stroke. However, it is known that administrative data may result in an underestimation of disease incidence [13]. Further limitations include, e.g., the lack of data on stroke severity, vascular risk factors, medications, or functional outcome. Also, ethnicity, neuroimaging status, and other individual items cannot be accounted for due to the strict anonymization of the data set.
Our analyses provide genuine insights in the treatment and care reality in German hospitals. This is due to the fact that treatments on specialized wards like SU and ICU are separately reimbursed once coded in the system. Even though special requirements are to be fulfilled, it can be assumed that reimbursement is an effective incentive to provide any specialized treatment modality. We have formerly reported on treatment rates of systemic thrombolysis and mechanical thrombectomy in ischemic stroke and were able to show, that both rates have continuously risen in the past years to rates of 15.9 and 5.8% in 2017 [8,14]. Treatment rates on specialized units also increased over the observed time period. The demonstrated numbers illustrate that SU treatment is not limited to ischemic strokes alone and that the care reality reflects the according recommendations of the professional societies. In Germany, currently there are more than 325 SUs certified by the German stroke society with a constant rise over the last 20 years [15]. One reason was the nationwide implementation of systemic thrombolysis and mechanical thrombectomy during this time period [8,14]. AIS patients therefore are increasingly transferred to another hospital in the acute phase of treatment, mostly for interventional treatment. In order to guarantee for a comprehensive ability for interventional therapy over the country it is fundamental to provide a close net of SUs also in rural areas that can select patients qualifying for interventional therapy. This is partly met by networks encompassing "other stroke units" (i.e., not run by a neurological department), where neurological and neuro-radiological expertise are established by tele-medicine. We therefore hypothesized, that the amount of treatments on "other stroke units" should have increased in the observed time period. This has only been the case until 2014. We believe that this putative stagnation illustrates, that hospitals with formerly "other stroke units" have established a neurological department in the meantime.