Is there an unmet medical need for palliative care services in Germany? Incidence, prevalence, and 1-year all-cause mortality of palliative care sensitive conditions: real-world evidence based on German claims data

A retrospective real-world data (RWD) analysis was performed, based on statutory health insurance (SHI) claims data (cf. Swart et al. 2014). Source of data was the InGef (Institute for Applied Health Research Berlin) Research Database, which is an anonymized claims database comprising longitudinal data from approximately 6.7 million persons insured in one of the 70 German SHI companies covered by InGef. For the purpose of this analysis, the InGef Research Database was condensed to a sample of approximately 4 million subjects, which is considered representative of the German population in terms of age and sex, and shows high external validity regarding overall measures of morbidity, mortality, and drug use as described elsewhere (Andersohn and Walker 2016). In brief, the InGef Research Database contains socio-demographic information such as age, sex, and the region of residence. In addition, the database gives information on hospitalizations, outpatient physician visits, and outpatient drug prescriptions. The hospital data comprises information on the date of admission and discharge, the reason for discharge, and diagnostic and therapeutic procedures with the exact date, as well as diagnoses which can be distinguished in hospital main discharge diagnoses and secondary diagnoses. The outpatient data cover diagnostic and therapeutic information. Outpatient diagnoses can be distinguished into confirmed diagnoses, suspected diagnoses, status post diagnoses, and diagnoses ruled out. Inpatient and outpatient diagnoses are coded according to the German modification of the International Classification of Diseases 10th Revision (ICD-10-GM). Data on outpatient prescription of reimbursed drugs comprise information on the date of prescription and dispensation as well as the pharmaceutical reference numbers. Based on a pharmaceutical reference database information on the anatomical–therapeutic–chemical code (ATC code), the defined daily dose (DDD), the package size as well as the strength and formulation of the drug can be linked for each dispensed drug. Furthermore, information on the specialty of care-providing physicians can be obtained from the database (Andersohn and Walker 2016; Ohlmeier et al. 2018).

The study was conducted according to the Good Practice of Secondary Data Analysis (GPS) recommendations (Swart et al. 2015). To estimate the prevalence of palliative care conditions, a cross-sectional study was carried out. A retrospective cohort design was used to estimate the incidence and 1-year all-cause mortality of palliative care conditions. The analyses were based on the years 2014 to 2016, since information on relevant palliative care services were not comprised in the data prior to 2014, and since 2016 was the most recent data year available at the time the study was carried out.

Patients were included in the overall study population if they had continuous insurance coverage during 2014 and at least 1 day of insurance coverage in 2015. In the subpopulation of patients with an incident case with reimbursement claims for palliative care in 2015, the subsequent 1-year all-cause mortality was analyzed (identification see below).

The documentation of palliative care treatment is possible in both the outpatient and inpatient setting. Therefore, a palliative care condition was assumed, if one of the following criteria was met:

An incident palliative care condition was assumed in case of an absence of the abovementioned codes on palliative care treatment in 2014. The first documented palliative care treatment in 2015 was then defined as the start of the palliative care treatment.

Deaths in the 365-day observation period after the start of the palliative care treatment were identified in the dataset, if the reason for the end of the insurance period was “death”. The date of death was defined as the end date of the insurance period which indicated death. Validation studies of German claims data suggest an accurate and valid documentation of the status of death as well as the date of death (Ohlmeier et al. 2016; Ohlmeier et al. 2015).

Hospital main or secondary discharge diagnoses and confirmed outpatient diagnoses were considered in order to calculate the frequency of pre-specified diseases and to analyze the general comorbidity via Elixhauser comorbidity score (ECS) (Quan et al. 2005) in patients with palliative care conditions. One documented diagnosis in 2015 was sufficient to be identified as suffering from the specific comorbidity.

In total, n = 14,522 patients with an incident reimbursement code for a palliative care condition in 2015 were identified. Of those, 55.3% were female and 44.7% were male (Table 1). The mean age of patients with an incident reimbursement code for a palliative care condition was 76.9 [standard deviation (SD): 13.4] and was slightly higher in women than in men. The majority of the patients received AAPV as the initial palliative care treatment (76.1%), whereas the inpatient complex palliative treatment and the SAPV were rarely observed as an initial palliative care treatment (14.6% and 9.3% respectively). In men, the proportion of patients with inpatient complex palliative treatment or SAPV as the initial treatment was marginally higher compared to women.

A malignant neoplasm was observed in 56.4% of all patients with an incident reimbursement code for a palliative care condition; the prevalence of malignant neoplasms was higher in men (65.0%) than in women (49.5%) (Table 2). A metastatic state was present in 29.4% of the patients, again with a higher prevalence in men (34.3%) than in women (25.4%). The given percentages refer to the whole study population having cases with reimbursement claims for palliative care. The most frequent localizations in both sexes were malignant neoplasms of digestive organs, with a pre-valence of 16.8%, and malignant neoplasms of respiratory and intrathoracic organs, with a prevalence of 10.5%. According to expectations, malignant neoplasms of the breast were a frequently observed comorbidity in women (14.5%), whereas the prevalence of malignant neoplasms of genital organs was high in men (15.2%).

In 2015, the incidence of cases with reimbursement codes for palliative care conditions was 38.1 per 10,000 persons on average, and was higher in women than in men (41.3 and 34.9 per 10,000 persons respectively). The incidence increased with advancing age, and doubled from every age group to the next higher one. The incidence was highest in subjects aged 90 years or older (546.0 per 10,000 persons, Table 3). The age-stratified incidence of cases with reimbursement codes for palliative care conditions was higher in men than in women in most of the defined age categories. Taking into account the age structure of the German population, the extrapolation of the observed incidence estimates results in n = 153,985 women and n = 125,185 men with a first occurrence of a case with a reimbursement code for palliative care condition in 2015.

The prevalence of cases with reimbursement codes for palliative care conditions was 56.3 per 10,000 persons, and was also higher in women than in men (61.3 and 51.1 per 10,000 persons, respectively). Similarly to the incidence, the prevalence increased with advancing age, and doubled from every age group to the next higher one. The prevalence was highest in persons aged 90 years or older (762.6 per 10,000 persons, Table 4). The extrapolation of the age-stratified prevalence estimates to the German population resulted in more than 400,000 individuals (over 220,000 women and 180,000 men) with palliative care sensitive conditions in 2015.

As stated in the Methods section (subsection Study population) this study determines the 1-year all-cause mortality in patients with incident occurrence of reimbursement codes for palliative care conditions during the 1-year period subsequent to the index date. For instance, if someone was prescribed SAPV in November 2014 and died in October 2015, this case was included in the 1-year all-cause mortality. The 1-year all-cause mortality of the 14,522 patients with an incident occurrence of reimbursement codes for palliative care conditions was 67.5% and was slightly higher in men (69.1%) than in women (66.2%). Mortality was highest in the first month after initiation of palliative care treatment (35.0% 30-day mortality). The 1-year all-cause mortality was higher in patients with SAPV or inpatient complex palliative care (86.1% and 90.6% respectively) compared to patients with AAPV (60.8%) as initial treatment (Fig. 1). The logrank test indicates significant differences in the mortality between AAPV and other types of treatment as initial palliative care treatment (p <  0.0001).

These mortality figures, related to palliative care, contrast against a 1-year all-cause mortality of approximately 1% (9.8 per 1000 individuals) in the InGef Research Database of 4 million beneficiaries, which in turn is approximately 8% lower than the standardized mortality rate in Germany according to the Federal Statistical Office of Germany for the given year. Roughly 9800 of those 39,200 deceased in the InGef sample died with some sort of palliative care claims in the current or preceding year, which corresponds to approximately 25% of those dying.

Based on a representative German claims dataset, an incidence and prevalence of palliative care conditions of 38.1 and 56.3 per 10,000 persons, respectively, was observed. This corresponds to an annual number of 410.000 patients with a palliative care condition. Both incidence and prevalence increased sharply with advancing age. The 1-year all-cause mortality in patients with an incident palliative care condition was 67.5%, and was highest in the first thirty days after treatment initiation.

To our knowledge, there is hardly a similar study on the utilization of palliative care on a national level which would take into account the whole spectrum of palliative care-sensitive conditions. Instead, there are recent studies that use pre-selected indications, such as various types of cancer, to investigate the utilization of palliative care in the terminal life span (Blaschke et al. 2019). Other studies use death registration data to estimate the possible demand for palliative care by combining recorded (disease-specific) deaths with diagnoses defined as potentially in need of palliative care at the end of life. Initially, Rosenwax et al. (2005) defined deaths from all types of benign and malign cancer and six non-cancer-related diseases as potentially being in need of palliative care. Murtagh et al. (2014) argue that the method applied by Rosenwax et al. could underestimate the need for palliative care, and develop an adapted method which partially adds or excludes diagnostic codes based on their presumed relevance (Murtagh et al. 2014). For instance, they consider chronic conditions of the heart, liver, kidneys, and the circulatory system, as well as neurodegenerative diseases and dementia, whereas deceased subjects with benign neoplasms are neglected (Murtagh et al. 2014).

Some other studies in different countries, e.g., in Ireland (Kane et al. 2015) and in 12 other countries around the world (Morin et al. 2016), used analogue methods to estimate the number of individuals who potentially need palliative care. For Germany, Scholten et al. (2016) used the Rosenwax method and the Murtagh method to determine the number of deceased individuals who in the year 2013 may have been in need for palliative care. Their estimation resulted in a total number of deaths of n = 363,689 (40.7% of all deaths) and n = 697,281 (78.0% of all deaths) respectively.

With an extrapolated number of over 400,000 individuals of the entire German population being eligible for palliative care services each year, the estimated prevalence in our study is located between the estimates of both methods. The fact that, using the Rosenwax method, the calculated number for potential need lies below the actual utilization figures for palliative care, would imply either that the Rosenwax method indeed underestimates the potential need, for instance by neglecting the need of patients with chronic diseases, or that the utilization of palliative care in Germany is already higher than the demand. This, however, is unlikely, because there is evidence for the lack in the provision of palliative care, especially in rural areas (Radbruch et al. 2015).

In contrast to Scholten et al. (2016), who determine the potential need for palliative care with various estimates, our study examines the present utilization of palliative care from a bottom-up perspective. Therefore, it is not surprising that Scholten’s estimate with the Murtagh method exceeds the prevalence found in our study. However, a gap between the use and the potential need seems to be noticeable, considering an estimated need that exceeds the actual utilization figures. Without having applied mathematical methods to prove this impression, the trends in the upward-sloping utilization figures over the various age groups appear to be rather similar: While the potential demand rises faster from 30 years onwards (Scholten et al. 2016), the utilization increases sharply in the age group of 70 years and above. This suggests an unmet need for palliative care, especially in the middle-aged patient group. A reason for the comparatively high unmet medical need for palliative care in middle-aged persons could lie in the predominance of curative treatment attempts for life-threatening diseases, although palliative care would already be indicated. With regard to this hypothesis, several studies suggest that earlier integration of palliative care, e.g., in parallel to curative treatment approaches, may be beneficial concerning quality of life (Temel et al. 2010; Zimmermann et al. 2014) and costs (May et al. 2015, 2018).

The main strength of this study is the large, unselected, and supraregional sample allowing robust estimations of the incidence, prevalence, and 1-year all-cause mortality of palliative care conditions. Further, recall and selection bias, both being a considerable methodological challenge to health services research in end-of-life studies, could be avoided since our study was based on routine claims data and did not depend on patients’ recall abilities or the willingness to participate.

The identification mechanism applied in this study is guided by the idea that a palliative care condition is given in cases where specific reimbursement claims are coded. This may occur in situations, (a) in which a care-giver sees the indication for a palliative care intervention, and/or (b) where an opportunity to bill such interventions is given. With this type of capturing algorithms, it cannot be excluded that actually existing, but not yet recognized situations with palliative care demands remain undetected. On the other hand, considering the level of specifity of the available reimbursement codes, it can be assumed that in the presence of an encoded claim for a palliative care intervention, a related palliative condition is factually existent. One can hardly imagine that palliative care services are provided without claiming the designated reimbursement code.

Although all available information can be linked on a person–individual level, the specific diagnoses triggering the utilization of care services in the sense of actual indications cannot be identified. Therefore, it was not possible to analyze, for instance, the distribution of diseases leading to the initiation of palliative care treatments. Furthermore, because of restrictions due to data safety requirements imposed by the SHI data provider, it was not feasible to perform in-depth analyses of the case development over time, to observe case histories on an individual level, or to assess a potential cross-over between general and specialized palliative care treatment.