End-of-life cost for lung cancer patients in Greece: a hospital-based retrospective study

A retrospective study was conducted based on data extracted from a hospital-based registry kept in the Oncology Unit of ‘Sotiria’ Hospital, in Athens, Greece with all new lung cancer cases who took care in this unit from 2015 onwards. The population of the present study consisted of patients diagnosed with non-small-cell lung cancer, specifically adenocarcinoma and squamous, or small-cell-lung cancer (SCLC) and died between 1 January 2015 and 31 December 2018, with at least 6 months of follow-up before death occurred. All patients died in this time interval were accounted, irrespectively of the place of death. The information maintained in this cancer registry was extracted from patients’ medical records.

For the purposes of the present study the following data were extracted: patients’ demographic characteristics such as gender and age and lifestyle characteristics such as smoking status at the time of diagnosis (i.e., current, former or nonsmoker). Moreover, cancer related characteristics at the time of diagnosis like the histological type, the stage and the performance status were also extracted. As for staging, the 7th edition of TNM classification for patients diagnosed earlier than 1 January 2017 was used, while for those with a diagnosis after the mentioned date, 8th edition was used. Patients were classified into those diagnosed at an early stage (stage I or II), those diagnosed at locally advanced (stage III) and metastatic stage (stage IV).

Healthcare resource utilization data, such as hospitalization, therapy (i.e., chemotherapy/immunotherapy, radiotherapy and surgery), and laboratory/imaging tests during the last 6 months before death were also extracted. More specifically, the data extracted regarding medication therapy (i.e., chemotherapy, targeted therapy, immunotherapy, supportive therapy like zoledronic acid, denosumab, granulocyte colony-stimulating factors (G-CSFs) and erythropoietin) included the name of the active substance, the total dose per administration, the route of administration, the number of days of administrations per cycle, and the number of cycles per line. Hospitalization data included the reason of hospitalization, and the duration of hospitalization, while the type of surgery (if any) was recorded, as well. The type of radiotherapy along with the number of fractions were recorded, and the number of laboratory and imaging tests to which patients were subjected were recorded. Finally, the date of diagnosis along with the date of death was used to estimate the overall survival.

To estimate EOL care cost from payer perspective, healthcare resource utilization data were combined with the corresponding local unit costs (in €, 2019), as reimbursed by the public payer (National Organization for Healthcare Services Provision [EOPYY]). To be more specific, to estimate systematic therapy cost and the supportive care cost during the EOL period, the total dose per regimen per cycle was combined with the reimbursed unit costs and the number of cycles per line of therapy administered during the last 6 months. The cost of administration of intravenous agents was included to the cost of drug therapy; this cost was extracted from the Official Government Gazette, law 2150/B/2011. The reimbursed drug cost was calculated on the grounds of the ex-factory prices, as they were published in the drug bulletin issued by the Greek Ministry of Health [16] after applying the relevant discounts provided in the corresponding legislation (Official Government Gazette, law 115/7 August 2017). The type of radiotherapy along with the number of fractions and the corresponding unit costs as extracted from the official website of EOPYY was used to estimate the radiotherapy cost. Similarly, the laboratory/imaging cost was calculated, combining the frequency of laboratory/imaging tests used during the EOL period with the corresponding unit costs as obtained from the official website of EOPYY under the assumption that all tests were performed privately (Supplementary Table 1). The cost for inpatient services including hospitalization for reasons relevant to the disease setting as well as surgeries was calculated using the corresponding Diagnostic Related Groups tariff issued by the Greek Ministry of Health [17] (Supplementary Table 2). Similarly, the 3 and 1-month EOL was calculated, as well.

All continuous variables were summarized with median, 1st and 3rd qualtiles as their distributions were skewed, while all categorical variables were summarized with frequencies (n) and percentages (%).

Cost data are presented as mean and 95% CI using the bias-corrected and accelerated CI (adjusted for ties) as obtained from 5000 nonparametric bootstrapped resamples. To determine factors associated with EOL cost we applied generalized linear models with gamma family and a log link function, which best fit the expenditure data according to the Box-Cox test and the modified Park test [18]. Cost data were additionally checked if they fit the gamma distribution using Villasenor and Gonzalez-Estrada parameter estimators [19]. Robust standard errors were computed to account for the fact that clinicians treating patients in the same hospital may share the same clinical practices. Gamma co-efficients generated by the regression model were exponentiated to retransform them into relative cost estimates. As potential risk factors the number of comorbidities, the number of metastases, the histological type, the duration from diagnosis to death, the cancer stage, the gender and the age during diagnosis were tested using univariate models. Factors which had a p-value < 0.2 in the univariate analyses were included in the multivariate model. The significance level was set to a = 0.05. Statistical analysis was performed using STATA software (version 13.0, STATA Corp, College Station, TX, USA) while gamma distribution fitness was checked using R version 3.6.1.

A total of 122 patients met the inclusion criteria, almost half of whom were diagnosed with adenocarcinoma. The place of death was reported for the 44% of the patients. Among them, the 59% had died in a hospital, while the rest in their home. The mean age at diagnosis was 68 years with 78.7% of patients being male and 55.0% diagnosed at stage IV. Only 9.0% of patients had no comorbidities, while more than 50.0% were current smokers at the time of cancer diagnosis. Almost all patients except for two had metastases at the time of diagnosis (Table 1). The median (1st–3rd quartile) overall survival of these patients was 10.8 (8–13) months. Almost 60.0% had died within 12 months since diagnosis.

Table 2 presents the resource utilization during the EOL periods. During the last 6 months before death, 88.5% of patients received medication therapy, while this percentage reduced to 74.6% at 3 months and 41% at 1 month before death. Almost two-third of patients that received medication therapy over the last 6 months of their life, received only chemotherapy while almost 20% of drug treated patients received immunotherapy even the last month before death. Platinum-based therapies were the most frequently prescribed treatment as first-line therapy (∼75% of those received chemotherapy as first-line), and nivolumab was the most frequently prescribed immunotherapy as second-line treatment (data not shown). Moreover, during the last 6 months of patients’ life, more than a third was subjected to radiotherapy, while this figure was estimated to be about 10% the last month before death. Only 16% of patients were hospitalized for reason other than drug administration during the last 6 months before death, with this percentage falling to 3% in the last month of patient life. Patients required hospitalization mainly due to respiratory system disorders (in 60% of patients during the last 6 months of life; data not shown).

The mean total cost per patient from payer perspective were €7665, €3351 and €1009 in the last 6, 3 and 1 month before death, respectively. Pharmaceutical cost was the key driver of the total 6, 3 and 1-month EOL cost (75, 65 and 58%, respectively) followed by radiation therapy (16.2, 22.8 and 31.1%, respectively; Table 3). The key drivers for the 6, 3 and 1-month pharmaceutical cost were the cost of G-CSFs (31, 28 and 30%, respectively), the immunotherapy cost (25, 30 and 32%, respectively) and the chemotherapy cost (23, 20 and 20%, respectively).

The 6-month EOL cost was found to be statistically significantly higher among patients diagnosed with adenocarcinoma compared with those diagnosed with squamous and SCLC, while 1-month EOL was found to be statistically significantly different only between patients diagnosed with adenocarcinoma and those diagnosed with squamous (Figure 1). Moreover, patients who were from 75 to 85 years old incurred statistically significantly less expenditures during last 6 months of their life compared with those who were from 65 to 75 years old (p =.0014) or younger than 65 years (p =.0036, Table 4).

The multiple generalized linear model in which patients’ gender, postdiagnosis survival, number of metastases at the time of diagnosis, age, stage and histological type were entered as independent factors, revealed that the 6-month EOL cost was almost 51% higher in males compared with females (p = 0.004) after controlling the rest of independent variables. Moreover, a 10-year increase in patients’ age at diagnosis was found to be associated with an almost 16% lower cost (p = 0.020), and patients diagnosed with SCLC were found to be associated with a 32% lower cost compared with those diagnosed with adenocarcinoma (p = 0.023). The time from diagnosis since death, the number of metastases and the stage of cancer during diagnosis were not associated with the 6-month EOL cost (Table 5). A model that included patients’ gender, number of metastases at the time of diagnosis, age and histological type as independent factors, revealed that a unit increase in the number of metastases was associated with almost 15% higher 3-month EOL cost (p = 0.047), while the rest independent factors were not associated with the 3-month EOL (data not shown). The 1- month EOL cost was not associated with any of the variables tested, as such a multivariate model was not constructed (data not shown).