The main findings of this study were: (1) an increase and slightly earlier initiation of specialist outpatient PC, (2) a constant frequency and ongoing late initiation of generalist outpatient PC, (3) a reduction in the number of new PEG insertions in the last 3 months of life for patients with dementia and (4) a lower number of inpatient treatment days though an unchanged number of hospitalisations. In the following, we will discuss these results in comparison with earlier results and particularly with the published findings of Radbruch et al., who investigated EoLC on the basis of similar quality indicators in Germany for the years 2010 to 2014 (supplementary Table S
1) [
7].
Curative overtreatment
Our results regarding curative overtreatment present a mixed picture. The lower number of dementia patients with a new PEG insertion in the last 3 months of life can be interpreted as a step in the right direction. Nonetheless, the relatively high proportion of cancer patients receiving chemotherapy in the last month of life and the high number of overnight hospital stays suggest an ongoing pattern of curative overtreatment.
The number of new PEG tube insertions in the last 3 months of life in 2016 and 2017 was considerably lower than that found by Radbruch et al. (2010 to 2014: 2.5%) [
7]. It has been demonstrated that tube feeding does not improve clinically important outcomes, and it should therefore not be used, especially for patients with dementia [
25]. For these patients, van der Steen et al. recommend intensified hand feeding, rather than permanent enteral tube nutrition [
26]. Furthermore, insertion of a PEG tube is often perceived as burdensome by the general public and some health care professionals [
7]. The decrease in new PEG insertions found in the present study may indicate an increase in the use of intensified hand feeding, as well as a higher awareness amongst health care professionals of the clinical limitations of PEG tubes at the end of life. The decrease may have also been affected by recent political initiatives and legal regulations in Germany, which may have improved PC awareness amongst health care professionals. Additionally, the new legislation to improve PC [
14] may have encouraged the realisation of advance care planning concepts [
13]. Therefore, undesirable overtreatments such as PEG tube insertions might continue to be reduced, especially within nursing homes, where they are often used for dementia patients at the end of life [
13].
Compared to the findings of Radbruch et al. [
7], the present results showed a slight increase in the number of cancer patients receiving chemotherapy in the last month of life (2010 to 2014: 9.6%). It would be incorrect to assume that all chemotherapy administered in the last month of life is inappropriate, as such treatment may be reasonable for patients with a fast disease progression or when aimed at improving quality of life [
7]. However, exceedingly aggressive treatments (e.g. chemotherapy) at the end of life are indicative of poor EoLC, and they may negatively impact on patients’ quality of life [
24,
27,
28]. There are many reasons why chemotherapy may still be administered in the last month of life. Clinicians may overestimate the prognosis, applying inappropriate treatment and delaying PC [
29,
30]. Decisions on treatment intensity at the end of life may also be influenced by patient preferences. However, most patients at an older age prefer palliative treatment over life-extension treatment [
31]. Early end-of-life conversations about patients’ preferences and the timely initiation of PC may reduce the administration of chemotherapy, thereby improving patients’ quality of life and care [
32,
33]. Further PC education amongst health care professionals may encourage the provision of PC and reduce curative overtreatment [
34,
35].
Compared to the results of Radbruch et al. [
7], the present findings showed a consistent mean number of hospitalisations in the last 6 months of life, but a slightly lower (by approximately 2 days) number of inpatient treatment days (2010 to 2014: 1.7/18.6). The hospitalisation of patients with PC needs can sometimes be useful. However, hospital admissions with no medical indication may be deemed aggressive and burdensome by patients with PC needs at the end of life [
7,
36,
37]. In Germany, the number of days spent in inpatient treatment has decreased over recent years, mainly due to changes in the health care system [
38,
39]. Therefore, the lower number of hospital treatment days found in the present study cannot necessarily be interpreted as an indication of a reduction in curative overtreatment. Indeed, hospital admissions and treatment days may be influenced by a variety of factors, including the tendency for patients to feel safe in a hospital and general patient characteristics (e.g. age, ethnicity) [
40]. Furthermore, it is often difficult for physicians to determine the clinical need for hospital admissions [
37], and this may be one reason for the overall high number of hospitalisations at the end of life. Training in caregiving for terminally ill patients might improve this situation. Also, changes in the health care system to expand outpatient care alternatives for critically ill patients may be useful [
37]. Further studies should investigate the effects of various approaches to reduce unnecessary end-of-life hospital admissions, such as PC training for ambulance staff [
41].
Palliative undertreatment
Compared to the results of Radbruch et al., the present findings showed a reduction in palliative undertreatment for specialist outpatient PC, but a consistent level of generalist outpatient PC, and therefore ongoing palliative undertreatment [
7].
This consistency (2014: 28.0%) is highly remarkable, given the introduction of billing codes for generalist outpatient PC in Germany in 2013, which was expected to significantly increase the provision of this service. In fact, recent legal changes in Germany appear to have failed to achieve their intended goals, for a variety of reasons. As recently described, generalist outpatient PC requires great effort, especially from general practitioners [
42]. Thus, there may be a need for further legislation around health care structures and financial models [
42,
43]. A reform of payment models and funding approaches may improve widely access to PC, ensure best practice and prevent inverted incentives [
43,
44]. Additionally, general practice has taken on greater importance in recent years and, in line with this, the requirements and qualifications for general practitioners have become increasingly complex [
45]. However, the increased demand for primary care services has not been accompanied by an equivalent growth in the workforce; thus, time restraints on general practitioners might reduce their quality of care and lower their job satisfaction [
46,
47]. Overall, time-consuming bureaucratic procedures, personal commitments and inadequate qualifications may prevent general practitioners from timely initiating PC, and this needs to be addressed [
35]. Nevertheless, the present results do not enable any conclusions to be drawn relating to the daily care routines of general practitioners, since only billed health care services were included in the analysis.
Finally, the present results indicated a considerable increase in specialist outpatient PC relative to Radbruch et al.’s findings (2010 to 2014: 5.3%) [
7]. It has been estimated that, in recent years, approximately 10% of the deceased required specialist outpatient PC prior to their death [
48], but were unable to access this service [
7,
49]. One important reason for the increase in specialist outpatient PC found in the present study might be the wider availability of specialist outpatient PC following its regional implementation in the community [
50,
51]. In fact, the present findings indicate that the capacity for specialist outpatient PC has increased and it can be assumed that the estimated population need for specialist PC is met. Presumably, the legal changes and initiatives to raise awareness for palliative needs have contributed to this increase since 2014. This shows that structural and legal changes can be an important driver for further development in health care systems. Existing structures need to be improved and expanded from top to down and cannot solely develop on regional level. Differences in regional structures and processes of the specialist outpatient PC teams might play a key role. There might also be certain regional disparities between the counties in Lower Saxony. While the potential population need might be met in some regions, it is potentially missed in others. However, our data cannot distinguish whether those patients with the greatest needs are actually the ones provided with specialist outpatient PC.
Delayed PC
The present findings underlined the ongoing late initiation of generalist outpatient PC. In contrast, specialist PC was initiated slightly earlier, relative to the findings of Radbruch et al. (2010 to 2014: median of 22.0 days) [
7]. While the slightly earlier initiation of specialist outpatient PC found in the present study may suggest a step in the right direction, the number of days between the onset of this treatment and death – especially with regards to generalist outpatient PC – indicates an unchanged focus on the last months of life.
It is well known that the early initiation of PC improves many important outcomes, such as quality of life and the burden of symptoms [
3,
52‐
54]. PC must not be reserved solely for patients whose life-prolonging treatment options have been exhausted; rather, it should be considered shortly after diagnosis [
55]. Many physicians find it difficult to determine the appropriate time to initiate PC in the disease progression [
56,
57]. Prognostic uncertainties form a major barrier for the early identification of patients with PC needs, and the estimation of disease progression is especially difficult for patients with non-oncologic diseases [
6,
58]. Internationally, there are several instruments that support the identification of patients with potential PC needs [
59,
60]. One such instrument is the Supportive and Palliative Care Indicators Tool (SPICT-DE), which is available for use in the German context [
61,
62]. Its application in primary care is currently being evaluated [
18]. Nonetheless, identification instruments such as the SPICT-DE are not implemented widely and consistently throughout Germany [
63]. For this reason, further PC training for physicians and other health care professionals might represent an important step in supporting the identification of patients with potential PC needs and promoting the early initiation of PC [
64‐
66].
Methodological strengths and limitations
AOK-N is the largest statutory health insurance provider in Lower Saxony [
19], and thus a reliable data source for the present analysis. The population of AOK-N members is comparable to the general population in Germany and Lower Saxony, regarding gender and age [
67]. However, differences exist with respect to education and occupation, which is why lower socioeconomic groups may have been overrepresented in the current study [
67]. To counteract this possible bias, the present study did not focus on socioeconomic differences between groups. Furthermore, the results were based on a large sample of AOK-N members who died in 2016 or 2017, enabling robust analyses to be conducted.
One difficulty with all secondary analyses of health insurance data pertains to billing purpose. In the present study, conclusions regarding PC timing may have been unreliable in some cases. Data on inpatient stays and outpatient services (e.g. generalist outpatient PC) were highly reliable, as they contained the dates of service provision. However, data on specialist outpatient PC only contained the date of prescription, while the actual treatment by a specialised PC team may have been delayed. Furthermore, specialist outpatient PC may have been initially prescribed by hospital doctors, and such prescriptions were not observable in the current dataset. Nonetheless, all follow-up prescriptions in the outpatient sector were observed. Finally, the use of routinely collected data involves low expenditure for data collection and can be highly beneficial to reflect the care situation [
68]. However, it has to be taken into account that the actual care situation cannot be completely represented by routinely collected data.
Content-related strengths and limitations
Although the data enabled us to evaluate the quality of EoLC on the basis of documented procedures of care, they did not allow us to analyse potential consequences of the analysed indicators, such as the effects on patients’ quality of life.
Further limitations pertain to diagnostic accuracy. Criteria for the validity of diagnoses cannot prove whether the diagnoses were correct and if patients were treated accurately [
20]. Particularly in the outpatient sector, ICD-10 codes are often used imprecisely, due to variations in coding methods [
69]. Data on diagnoses can be affected by an individual coder as well as by financial incentives in the German health care system. Additionally, statutory health insurance data does not record cause of death.
While the comparison with the results of Radbruch et al. [
7] was reasonable to contextualise our data, considerable differences existed between the study samples. In contrast to Radbruch et al., the present study predefined chronic diseases with potential PC needs. Nonetheless, the utilisation of criteria for the validity of diagnoses was an important strength of our study. Only data from patients with valid chronic diagnoses were included in the analysis. Furthermore, the ICD-10 code list was based on the current literature [
2,
22] and compiled by an interdisciplinary panel of experts.