Factors related to and economic implications of inhospital death in German lung cancer patients - results of a Nationwide health insurance claims data based study

In Germany, around 51% of all-cause and tumour-related deaths occur in a hospital setting, whereupon males (♂ 57% vs. ♀ 45%) and patients with tumours in respiratory organs (60% vs. 50%) are affected above average [1]. Previous research has found that irrespective of a cancer diagnosis, there is a preference for dying at home or in a hospice; dying in a hospital setting is favoured only by few (Gomes 2012, Gomes 2013, Pinzon 2011 and Higginson 2013) [2‐5]. A palliative care setting may help patients express preferences about their place of death and their preferred treatment based on extensive information. Thus, patients in palliative care settings may more likely achieve their preferred end of life choices which supposed to improve their Quality of life (QoL) at the end of life. Although the number of inhospital deaths has declined in recent years, there is still an obvious discrepancy between the preferred and the actual place of death, especially in patients with respiratory tumours. Furthermore, this patient group is of high public health relevance, as respiratory tumours were the fourth leading cause of death in Germany in 2015 [6]. Two studies from the US and the UK found that, patients’ QoL at the end of life tends to be worse when they die in a hospital compared to when they die elsewhere, because they are more likely to experience physical and emotional distress and feel less at peace [7, 8]. Owing to different framework conditions, generalizing findings from one health care system to another is a sensitive issue. Of course, the results of the US-based study might not be fully replicable in the German setting, but since both systems are strongly “curatively” oriented and pay subordinate attention to palliative care, similar results can be expected for German populations. Additionally, inpatient care at the end of life generates higher expenditures than outpatient care as shown by Gaertner et al. (2013) [9] and Schwarzkopf et al. (2015) [10]. Thus, reducing the share of inhospital death in lung cancer patients is in the interest of both the patients concerned and health care service payers.

Costa (2014) published a systematic review on determinants influencing the place of death of terminally ill patients [11]. Interprofessional home palliative care and early referral to palliative services, inter alia, increased the number of home deaths. Based on these results, the German palliative care system has been improved, as Cremer-Schaeffer and Radbruch (2012) reported [12]. Since 2007, every German citizen has a legal claim to palliative home care. Furthermore, palliative care was implemented in educational programs for health professionals and the number of hospices increased overall. Consequently, Dasch et al. (2016) showed that the number of inhospital deaths in German lung cancer patients decreased between 2001 and 2011 (from 68,1% to 60,3% in men and from 70,7% to 49,6% in women) [1]. While trying to determine further factors influencing the place of death, Escobar Pinzón et al. (2011) found that nonworking relatives and a high care level are associated with home deaths in the German general population [4]. A study by Leak et al. (2013) describes characteristics of 104 lung cancer patients dying in emergency departments in North Carolina (USA) [13]. 71% of those patients died on their first visit and 65% of them were male. The most common chief complaint was respiratory distress. Considerably more patients (n = 143.627) were included in a UK nationwide study of O’Dowd et al. (2016) [14]. They identified sex, increasing age and social deprivation as factors associated with inpatient death.

To our knowledge, there is no representative nationwide study for a German population investigating multiple factors related to inhospital death as well as related to healthcare expenditures in German lung cancer patients as of yet.

Therefore, our study aimed to improve the evidence available by:

In our study, we sought to identify factors related to inhospital death in German lung cancer patients. We found that especially the patients’ age and frailty, as measured in the need for care and nursing home residency, were important factors related to not dying in a hospital. In addition, patients with cerebral metastases had a decreased risk of inhospital death. The utilization of outpatient palliative care also reduced the likelihood of inhospital death but receiving inpatient palliative care increased the likelihood of inhospital death. Furthermore, a higher number of days spent in a hospital, CHF and renal disease, as well as receiving chemotherapy in the last 30 days of life were associated with an increased risk of inpatient death.

Total healthcare expenditures were about twice the expenditures of dying elsewhere, with expenditures for hospitalizations predominantly responsible for these differences. Although expenditures for outpatient care and medications were higher in patients not dying in the hospital, the sum of these differences were negligible compared to the differences in inpatient expenditures.

Evidence, in which direction age affects inhospital death is inconclusive: Dasch et al. found a negative association for the general population in a North Rhine-Westphalia-based study (< 60 versus > 80 years: OR 1.41, p-value 0.001) [30], but O’Dowd et al. observed a positive association among lung cancer patients in the UK (> 85 year vs. 70–74 years OR 1.16; 95% CI 1.08–1.24) [14]. Finally, Escobar Pinzón et al. (2011) found no association at all between age and place of death in a population of Rhineland-Palatinate, however this study was not focused on lung cancer patients but referred to the general population [4]. We believe that our observation of negative association is plausible: Younger patients may be subject to more aggressive inpatient therapy. Therefore, their primary clinical relationship is rather with an oncologist than with a palliative care team. Moreover, they might have more social/familial obligations than older patients and death at home may not be desirable if they have a young family at home. Consequently, younger patients have an increased risk of inpatient death.

Longer survival after diagnosis increased the risk of inhospital death. This was a surprising result at first as we assumed that patients with shorter survival have less time to come to terms with impending death, and therefore to organize their final days in a preferred and comfortable place. However, if a patient is diagnosed in a late stage of the disease the focus of care is more likely to be on palliation and organizing end-of-life care rather than curative therapy. In line with that, patients with cerebral metastases at time of diagnosis died in the hospital less frequently.

In our study nursing home residency was associated with a reduced likelihood of dying in a hospital. This might be explained by the fact that continuous nursing care — as often required at end of life — is already available in this setting, whilst it can often not be guaranteed in an ambulatory setting. Even though dying in a nursing home or in a hospice is not the same as dying at home, most patients prefer those places to hospitals [2]. Relatives who are not able to arrange care of the patients at home due to financial or other reasons, thus should be supported more relating organization of accommodation in a hospice or a palliative care facility in order to increase the patient’s QoL during the last days. Wye et al. (2014) describe that arranging care outside the hospital is a very tough task for the relatives [31]. For this reason, the UK has implemented end of life care services, which provide nurses who support relatives with organizational services, as well as assistance with decision-making. These nurses are supposed to offer time for conversations with the patients and the relatives about death, as well as the practical aspects of caring for the dying. This concept works especially for cancer patients who are close to death and should be considered as a developing concept to further improve end of life care in Germany. Actually, at the time this study was conducted, there was study ongoing to evaluate establishing a system of ‘social care nurses’ in cancer therapy in Germany funded by the German Ministry of health.

Our study indicated that patients with lower care levels are more likely to die in a hospital. Escobar Pinzón et al. (2011) found similar results in his study (care level 3 vs no care level: OR (death elsewhere vs inhospital death) 4.95, p-value < 0.0001) [4]. The higher the care level the higher the need for support in various aspects of everyday life. Thus, comprehensively dealing with the organization of the required care has already taken place. If a home-care setting for a person with high nursing needs has been organized previously, it is far easier to enable the patients dying outside a hospital – as hospitalization is necessary less frequently, and relatives are able to handle complications outside a hospital (with assistance of professionals).

According to our study, CHF and renal diseases increased the probability of inhospital death. These comorbid conditions tend to require intense treatments and are associated with a high likelihood of complications. Managing these complications often requires further inpatient hospital care [32]. However, it should be examined whether it is possible to further train formal and informal care-givers to manage these conditions outside the hospital.

If a patient in our study received inpatient palliative care, he was more likely to die in a hospital setting. This finding was unsurprising as inpatient palliative care is administered in a hospital for example on a palliative care ward. Our SA 2 supports this assumption as in this analysis we regrouped patients from the inhospital death group, if they had an inpatient palliative care treatment in the hospital stay of death. Resulting in the effect of inpatient palliative care changing direction and now reducing the likelihood of inpatient death.

Given, that outpatient palliative care was per se provided on a low level among patients of the inhospital death group, a conclusive interpretation of these results is limited to some extent. By trend, our analyses support the study of Purdy et al. (2015), which reported that patients receiving outpatient palliative care for a longer duration were less likely to die in a hospital. Patients and relatives using palliative care have usually already confronted the fact of impending death. Most families feel overburdened in this situation. The providers of palliative care are in a better position to assist the family with decision-making and to help them with enabling the patient’s death in a preferred place – mainly because there is sufficient time and opportunity for such conversations. Furthermore, palliative care providers have a broad base of knowledge about the health care system as well as local supply structures. In addition, engaging with outpatient palliative care may define a population who are more determined to die at home. Purdy et al. (2015) [33] demonstrated that an end of life coordination centre that aims to help patients under palliative care be cared for in their preferred place, can reduce the number of inhospital deaths significantly (OR 0.33, p-value < 0.001).

In contrast to receiving palliative care, the administration of chemotherapy in the last 30 days of life was associated with an increased likelihood of inhospital death. Receiving chemotherapy shortly before death can be considered as an aggressive treatment and deemed undesirable by some clinicians [34, 35], even though it needs to be kept in mind that palliative care could to some extent also include chemo-therapeutic measures. That these patients more often die in a hospital could be explained by the close relationship between the patient and the oncologist who is often located in a hospital. Furthermore, death might be associated with the treatment itself while in the hospital.

Similar reasons could play into the finding that patients with a higher number of hospital days also have a higher likelihood of inpatient death. A higher number of hospitalizations may indicate a more aggressive or longer tumour-directed treatment. Therefore, patients have an established relationship with the hospital and its clinicians and associate them rather than a palliative care team with the primary point of contact.

Despite higher comorbidity burden, patients in the inhospital death group were less frail as measured in care level and nursing home residency. Therefore, we can deduct that it’s not the patients’ frailty that has an impact on the place of death as these patients are already cared for at diagnosis, but more importantly specific comorbidities. Therefore, these comorbidities are important aspects to consider when lung cancer is diagnosed.

As expected, we found that expenditures for inpatient care were around double in the inhospital death group. However importantly, we could see that the excess costs in the outpatient sector for patients who did not die in the hospital were almost negligible. From this, we can conclude that not only costs for hospitalizations are much higher, but also additional expenditures for outpatient care are extremely low.

Thus, extending outpatient palliative care services as it currently happens in Germany will probably create a win-win situation for patients and the SHI system, since the expected decline in expenditures for inpatient care would compensate for the additional expenses in the outpatient sector. Our results are supported by the studies of Faßbender (2005) [36] and Smith et al. (2012) [37] who found that palliative care might have a cost-cutting effect because of fewer therapeutic interventions not primarily aiming at improving QoL for terminally ill patients. Thereby the need for inpatient treatments might be reduced and patients remain in a stable condition while still receiving sufficient medical and non-medical care.

Most limitations in our study stem from the use of claims data, which are primarily documented by healthcare providers for billing purposes with health insurance funds. The exact place of death is not important for these parties and is therefore not documented. This is the reason that within this study we have only been able to distinguish between inhospital death and death elsewhere. We do not know if patients in the inhospital death group died on a palliative ward or on a normal ward, nor can we differentiate between various places of death outside a hospital such as a hospice or the family home. Owing to data protection laws, data contained only year and month of death for those who died in an outpatient setting. We used rules to define a more valid date of death but there are still possible differences to the actual date of death of up to 2 weeks for each individual. As O’Malley et al. (2005) described, the coding of ICD- and OPS-codes is a source of error in itself [38] which can e.g. arise out of unintentional and intentional coder errors, such as misspecification, unbundling, and up-coding. This should be considered when interpreting variables based on those codes like Charlson comorbidity groups. For historical reasons, the average income and the family structure of AOK-insured people may differ from the general German population. When there was no free choice of health insurance provider, the AOK as a Local Health Care Funds insured the general population (workers, retirees, family members) in contrast to company-provided health insurance funds, which for example insured predominantly persons with higher incomes and who tends to be healthier. Even now when all members of the population are in the position to choose their own health insurance provider, this structure is still visible. However, Jaunzeme et al. (2013) have shown that these differences are small enough to assume that the population of AOK-insured and herewith our results are representative for the German resident population [39].

Our study represents the first nationwide examination of factors possibly related to inhospital death in German lung cancer patients such as the patients’ care needs, comorbidity burden, lung cancer-related treatments, and regional aspects. More importantly, we could provide expenditures for inhospital death not only as total but also for different aspects of care and therefore provide evidence that investments in palliative care can nevertheless reduce expenditures in total. As an innovative methodological approach, we applied the LASSO selection method which combines the easy interpretability and robustness. LASSO selection has been shown to outperform other selection methods like stepwise selection [40]. Finally, our analyses are automatically based on a multicentre study, whereby biases resulting from specific documentation rules and approaches of single healthcare providers are avoided. Thus, we are convinced that these analyses create a sound starting point for conditions that need to be considered when trying to create an environment supporting terminally ill-patients to make a more self-determined decision on where to spend their last days of life.

This study provides further evidence about factors related to inhospital death in German lung cancer patients. Findings suggest that factors associated with inhospital death often relate to previous contacts with hospitals. This includes prior hospitalizations, tumour-directed treatment and treatment of comorbidities. Additionally, factors associated with dying elsewhere relate to access to other care settings than hospitals where therapy is focused more on palliation. From these results we can implicate that an early or earlier integration of palliative care into tumour-directed therapy might be a useful tool in helping patients to make informed decisions in the last phase of their life, by using established relationships e.g. with the oncologist. Additionally, further expanding the palliative supply network, can be achieved while still reducing costs for SHIs.