Quality of life, out-of-pocket expenditures, and indirect costs among patients with the central nervous system tumors in Thailand

Data source and target population

A prospective cohort study that consecutively followed patients >18 years of age undergoing first CNS tumor surgery at an 850-bed academic tertiary care hospital in Songkhla Province, Thailand, between September 2018 and August 2019 was conducted. Patients were excluded if they were unable to complete a self-reported questionnaire. All included patients gave their informed consent. The study was approved by the Hospital’s Research Ethics Committee.

Disease definition, tumor response, and operational definition

Progressive disease^[14] is a condition in which the sum of the diameters of target lesions has increased by at least either 5 mm or 20%, when compared to the smallest mass before therapy. Moreover, the emergence of one or more new lesions is also considered as progression.

An eloquent area is tumor in area that specifically involved motor cortex, sensory cortex, visual center, speech center, basal ganglion, hypothalamus, thalamus, brainstem, and/ or dentate nucleus.^[15,16] The extent of resection was assessed according to Bloch et al.^[17]

In all instances, imaging was conducted and evaluated by a neuroradiologist at the pre-operative and post-operative periods. Magnetic resonance images (MRI) of the brain were reviewed to estimate tumor size, tumor location, and other characteristics of the tumor. The post-operative residual tumor was evaluated by post-operative MRI or contrast-enhanced computerized tomography of the brain.

Data collection

Patients were interviewed during hospitalization for undergoing first surgery. Within 6 months, they were interviewed once again if the disease continued to progress. An interview schedule consists of three parts: (1) Demographic information and clinical status, for example, Karnofsky performance status (KPS) and neurologic status, (2) the EuroQol five-Dimensional five-levels (EQ-5D-5L), and (3) costs incurred by patients and family including OOP expenditures, informal care (unpaid caregivers), and productivity loss. Clinical, radiologic, treatment, pathologic data, and pre-operative complications were retrieved from the hospital database.

The HRQoL were collected prospectively using the Thai version of the EQ-5D-5L self-reported questionnaire.^[18] According to the Thai value set, the EQ-5D-5L health utility scores were calculated and ranged between −0.283 and 1. The scores “1,” “0,” and negative values represent the state of perfect health, death, and worse than death, respectively.

The cost analysis was carried out from a patient perspective. All costs were annualized, extracted in Thai currency, and then converted to 2019 United States dollars (in August 2019, 30.65 Baht per dollar).^[19] The costs of illness can be divided into three categories: (1) Direct medical expenses, which are typically covered by national public insurance. However, some direct medical costs (i.e., costs of seeking health-care outside patients’ public health insurances, which were clinic visits, other medications, and alternative medicines) require patients to pay out of pocket (OOP expenditures); (2) Direct non-medical costs incurred by patients and family, namely, costs for transportation, food, accommodation, home modification, nutrition supplements, and caregiver’s salary (OOP expenditures); and (3) Indirect costs include expenditures for informal care (unpaid caregivers) and productivity loss.

Informal care, which is unpaid assistance given by someone with whom they have a social bond, such as a family or other non-kin,^[20] was measured using the amount of time that accompanying family members spent on outpatient visits and hospitalizations. Productivity loss is indirect costs incurred with paid and unpaid production loss due to illness, disability, and premature death of productive individuals.^[21] Productivity loss estimation was conducted using patient’s time spent at the hospital (consultation and hospitalizations) and time unable to work due to CNS tumor. These indirect costs were calculated by multiplying amount of time-loss by patients’ individual wage. Daily minimum local wage (320 Thai Baht^[22]) was applied to unemployed patients and all family members for estimating productivity loss and informal care.

Statistical analysis

Descriptive statistics were used to describe the baseline characteristics and clinical status. The Chi-squared test was used to examine categorical variables, but if expected counts were low, the Exact test was utilized instead. For continuous variables (such as age, patient income, and time loss), means and standard deviations were determined. In addition, differences in means were evaluated using independent sample t-test. The Shapiro–Wilk test revealed that all costs and most of utility scores were skewed (P < 0.001). We expressed costs as mean since other measures (median costs and log transformed costs) are not informative for health policy decisions at the population level.^[23] Nonetheless, when making inferences about means for heavily skewed data like costs, a bias-corrected and accelerated non-parametric bootstrap technique was applied to perform the 95% confidence intervals (CI) and t-test. Since this method avoids the assumptions of normality that constrain other approaches, it is a more adjustable way of comparing mean costs between groups.^[24]

For dealing with these skewed data, a generalized linear model (GLM)^[25] was used to investigate the effects of disease severity (malignancy, progressive disease, KPS score, and histology) and costs (OOP, informal care, productivity loss, and total costs). We compared two different distributions (e.g., Gamma and Inverse Gaussian distribution) with either identity or log link function. When the results were presented with identity link, change in mean per unit increase in a covariate was demonstrated. On the other hand, those with log link indicated ratio of means per unit increase in the covariate.^[25] The model performance of GLMs were investigated using Akaike Information Criterion (AIC) and graphical analyses.^[25-27] Analyses were performed with SPSS software version 22.0.^[28] P < 0.05 was considered statistical significance.

Baseline characteristics

A total of 131 patients were interviewed between September 2018 and August 2019. On 97 of the 131 cases, total resection was performed, while the remaining 34 received subtotal resection. Eight patients, however, had a spinal cord tumor and were therefore excluded from the analysis. As a result, the data analysis comprised a total of 123 patients with intracranial CNS tumors.

Baseline characteristics are demonstrated in [Table 1]. Mean age was 50.65 years, when categorized by histology, meningioma, glioma, and pituitary adenoma were commonly found in 45 (36.6%), 37 (30.1%), and 15 (12.2%) cases, respectively. In addition, ten of the 15 patients in our study with pituitary adenoma had a functional adenoma, and eight of these ten had visual field defects. Within 6 months after the first brain surgery, 18 patients had disease progression. In addition, 94.4% of these patients had the pre-operative KPS score <80. Average time to disease progression was 4 months. There were 83 and 40 patients classified into benign and malignant tumor, respectively. In addition, the proportion of patients developed to progressive disease and length of hospital stay was significantly worsened (P < 0.001) in patients with malignant tumor.

HRQoL

The EQ-5D utility scores categorized by histology and disease severity are shown in [Table 2]. The mean EQ-5D scores at pre-operative and progressive disease periods were 0.711 (95% CI: 0.662–0.760) and 0.261 (95% CI: 0.144– 0.378), respectively. All 18 patients with progressive disease were determined to have glioma. In addition, glioblastoma afflicted 11 of these patients. Glioma patients had a low HRQoL of 0.584 (95% CI: 0.487–0.681) before surgery, and their HRQOL worsened to 0.261 (95% CI: 0.144–0.378) as the disease progressed. There was no statistically significant difference between patients having benign and malignant tumor, (P = 0.072 at first surgery and P = 0.557 at progressive disease).

Costs incurred by patients with CNS tumor

Between benign and malignant tumors, there was no significant difference in any sorts of patient expenditures, including OOP expenditures, indirect costs, and total costs (P = 0.121, 0.449, and 0.152, respectively) [Table 3].

The mean annual OOP expenditures in patients with benign tumors were $290.02 (95% CI: $211.54–$368.50, median $180.12), while the mean OOP expenditures were $508.19 (95% CI: $240.68–$775.71, median $187.95) in those with malignant tumors. Since three of the 40 patients with malignant tumor reported a mean cost for caregiver’s salary of $2,827.99 (95% CI: $351.59–$5,304.39), the caregiver became the primary cost driver for OOP expenses in patients with malignant tumor. In patients with benign tumors, however, there was no report on costs for caregiver’s salary. Food and transportation were the main influences for costs incurred in these patients [Table 3].

The average length of hospital stay was 11.09 and 16.83 days for first brain surgery and progressive disease period, respectively [Table 4]. In addition, the average length of hospital stay was 11.22 and 18.40 days, when patients were classified as having benign and malignant tumors, respectively. The average annual costs of informal care for patients with benign and malignant tumors were $240.29 (95% CI: $167.55–$313.02) and $295.76 (95% CI: $155.80–$435.73), respectively.

Because of the CNS tumor, 79 of the 123 patients were unable to work for an average of 2.05 months. Fifty-six of these 79 patients had benign tumors and had lost an average of 1.80 months (on average), while the remaining 23 patients had malignant tumors and had lost an average of 2.65 months due to CNS tumors. As a result, productivity loss due to inability to work among patients having benign and malignant tumors was $486.71 (95% CI: $311.74–$661.68) and $507.00 (95% CI: $291.47–$722.53), respectively.

Annual total costs incurred by patients with CNS tumor

Indirect expenditures were the primary cost driver, accounting for 73.81% of total costs borne by CNS tumor patients. In addition, almost half of these indirect costs were productivity loss due to inability to work (35.79% of total cost). The productivity loss due to time spent in the hospital (19.29% of total cost) and informal care (18.74% of total cost) was the second and third cost drivers, respectively.

Costs expressed as a proportion of the patient’s annual income

OOP expenditures accounted for 8.27%, 19.12%, and 11.30% of the reported patient’s annual income for employed patients with benign tumors (67 of the 83 cases), malignant tumors (26 of the 40 cases), and all patients who reported their income (93 of the 123 cases), respectively.

On the other hand, the disease severity was associated with increases in the proportions of annual indirect costs on the reported patient’s annual income. Indirect costs accounted for 40.69% in malignant tumor patients, which was 1.6 times higher than benign tumor patients (25.00%). Similarly, the proportion of total costs incurred by malignant tumors (59.81%) was nearly twice that of benign tumors (33.27%).

The influence of disease severity and baseline characteristics on patient expenditures

[Table 5] demonstrated results of GLM analyses for patient expenditures. GLMs with the inverse Gaussian distribution and the log link were found to be the most applicable models for patient expenditures, according to the AICs and graphical analyses. Gender was not a predictor in any sorts of patients’ expenditures, whereas older age (over 50 years old) was a significant predictor of greater OOP expenditures (P = 0.009).

When all indicators of disease severity were taken into account, disease progression was found to be significantly associated to rises in all cost categories. Progressive disease was significantly increases in the OOP expenses (P = 0.001), the indirect costs (P = 0.013), and the total annual costs (P = 0.001), with a ratio of mean costs of 3.85 (95% CI: 1.74–8.52), 2.46 (95% CI: 1.21–4.98) and 2.91 (95% CI: 1.51–5.61), respectively, when compared to patients without progressive disease. Although malignant tumors were found to be a significant predictor of decreased OOP expenditures (P = 0.021) when compared to benign tumors, no significant difference in total costs was detected. On the other hand, a KPS score was found to have no statistical significance in predicting costs of any kind. On the other hand, a KPS score < 80 was not a predictor in any costs. Patients with glioma had a significantly higher OOP expenditure (0.015).