Estimating the lifelong health impact and financial burdens of different types of lung cancer

All patients with lung cancer during the period from 1998 to 2007 were abstracted from the National Cancer Registry database. The diagnosis of lung cancer and its subtypes was based on either histology or cytology codes taken from the International Classification of Diseases [12]. The survival status for each patient was verified by linking the patient’s identification information to the National Mortality Registry database. Each patient underwent follow-up from the day of diagnosis until the end of 2010.

After obtaining the survival function of the lung cancer cohort, a method proposed by Huang and Wang [13] was used to extrapolate this curve beyond the end of the follow-up period. This approach assumes that lung cancer generates a constant excess hazard (i.e., mortality) after a certain follow-up period, and its calculation is carried out in three steps. First, we took the hazard functions from the life tables of the National Vital Statistics of Taiwan to generate an age- and sex-matched reference population using the Monte Carlo method and then estimated the survival function. Second, we calculated the survival ratio between the lung cancer cohort and the reference population at each time t and performed a logit transformation of this. Third, the logit transformation of the ratio was fitted by simple linear regression up to the end of the follow-up period. The estimated regression line, together with the survival function of the reference population beyond the follow-up limit, was used to extrapolate the lifetime survival function of the lung cancer cohort, and thus the life expectancy of the lung cancer cohort (up to 600 months) after diagnosis could be estimated. Expected years of life lost in the lung cancer cohort were defined as the survival difference between the cohort and the reference population. The method described above has been demonstrated to be effective using computer simulations [13], proven mathematically [14], and corroborated by several examples of cancer cohorts [15, 16]. The iSQoL statistical package [17] was used to facilitate the computation. To validate the extrapolation method described above, we used the survival data of patients who were diagnosed during the first seven years and then extrapolated them to 13 years. Because these patients were actually followed until the end of 2010, the mean survival duration within the 13-year follow-up period, using the Kaplan-Meier method, was considered as the gold standard. The relative bias was computed to compare the difference in values between the extrapolation and the Kaplan-Meier estimation.

From May 2011 to April 2012, all consecutive patients with lung cancer from the outpatient oncology, chest surgery, and chest medicine departments of NCKUH were invited to participate in this study. To enrich our sample, we also screened patients admitted to the wards between November 2011 and January 2012. The inclusion criteria included the realization of a lung cancer diagnosis in the participants, the absence of malignancy at another site, and the subjects’ ability to understand and answer the questionnaire. In some individuals, measurements were performed repeatedly; however, each measurement took place 3 months apart.

The EuroQol 5-dimension questionnaire (EQ-5D) [18, 19], which is a preference-based generic instrument, was used to estimate the QoL utility values. The five dimensions assessed by EQ-5D are mobility, self care, usual activities, pain/discomfort, and anxiety/depression, each with three levels of severity. Using the scoring function from the United Kingdom [20], these health state parameters were transformed into a utility value ranging from 0 to 1, in which 0 represents death and 1 indicates perfect health.

The duration-to-date for each measurement was defined as the period between the date the lung cancer diagnosis was made and that of the interview. A kernel-smoothing method (i.e., the moving average of the nearby 10%) was used to estimate the mean QoL function [10, 11]. The QoL utility values beyond the follow-up period were assumed to be the same as the average of the last 10% of patients near the end of follow-up.

The lifetime survival function of the lung cancer cohort was adjusted using the corresponding mean QoL function to obtain a quality-adjusted survival curve, in which the sum of the area under the curve was the QALE of the lung cancer patients [10]. The loss-of-QALE was calculated by subtracting the area under the quality-adjusted survival curve of the lung cancer patients from that of the reference population. Because the EQ-5D scoring function for Taiwanese residents is still under development, we tentatively assumed the QoL score of the reference population to be 1 [10, 13]. However, we also tested the assumption in sensitivity analysis assuming the average QoL values of 0.9 and 0.95 for the reference population. We used the iSQoL software [17], which can be downloaded for free, to facilitate the computation.

The reimbursement database at NCKUH was used to obtain spending details on medical services for lung cancer patients between January 2005 and December 2012. As these data included not only medical costs paid by the NHI but also out-of-pocket money, most direct medical costs attributable to lung cancer could thus be obtained to calculate the total monthly healthcare expenditures. These were then divided by the effective sample sizes, namely, the number of lung cancer patients who survived that month, to obtain the average monthly healthcare expenditures per case. These values were subsequently multiplied by the corresponding monthly survival probabilities and summed up to obtain the lifetime healthcare expenditures per case. All the payments in different calendar years were adjusted based on the related consumer price indices [21] and made equivalent to those in 2012. To discount costs in future years, we also adjusted the healthcare expenditures, using an annual discount rate of 3%. Because patients might incur expenses outside the hospital, we compared the costs we estimated with the costs reimbursed in the national database by using the same group of patients (2005–2007). The proportion of reimbursement at NCKUH was estimated. A total of 1,002 patients ever cared at the NCKUH were linked with the reimbursement database of NHI, of which 11 cases could not be successfully connected. There were 4 cases (2 cases of small cell lung cancer (SCLC) and 2 cases of non-small cell lung cancer (NSCLC)) with the difference on the date of diagnosis larger than 1,000 days and were not included for the above calculation.

After multiplying the survival probability with the mean QoL at each time t (duration-to-date), we obtained the quality-adjusted survival curve, as shown in Figure 2. The sum of the shaded area under the curve represents the QALE. Assuming that the QoL of age- and sex-matched referents was 1, the difference between the quality-adjusted survival curve of patients and that of the referents was the loss-of-QALE (Figure 3). The QALE for patients with SCLC and NSCLC was 1.21 and 2.65 QALY, respectively, while the loss-of-QALE for the two groups was 13.69 and 14.08 QALY, respectively. Compared with NSCLC patients with squamous cell carcinoma (SqCC), those with adenocarcinoma had a longer QALE (3.03 vs. 2.37 QALY, p < 0.001, Table 2). Moreover, we found that the loss-of-QALE of adenocarcinoma patients was significantly greater than that of SqCC patients (15.03 vs. 12.22 QALY, p < 0.001), probably because of the younger mean age at diagnosis (65.60 vs. 69.35 years, p < 0.001).

Using the kernel-smoothing method, the average healthcare expenditures per case of patients with SCLC and NSCLC are depicted in Figure 4. It shows that the medical costs reimbursed by the NHI decreased gradually after the diagnosis; however, the out-of-pocket medical costs in the terminal phase of NSCLC increased. Compared NHI-reimbursed costs of the 987 patients in our hospital with those in the national database, we found that healthcare costs of the NCKUH for SCLC and NSCLC patients accounted for 74.3% and 84.1%, respectively, of the total costs reimbursed by the NHI.

Table 2 also shows the lifetime healthcare expenditures per case for patients with different NSCLC subtypes, adjusted for medical care inflation rates up to the year 2012 and for annual discount rates in future years. Compared with SqCC patients, adenocarcinoma patients had greater lifetime healthcare expenditures per case (US$ 36,771 ± 1,998 versus US$ 20,599 ± 1,787), especially in regard to out-of-pocket medical costs (US$ 18,620 ± 1,857 versus US$ 6,350 ± 741). Moreover, we found that both the annual healthcare expenditures and cost-per-QALY attributable to adenocarcinoma were higher than those for SqCC.

There were 4,425 patients with SCLC and 38,509 patients with NSCLC diagnosed during the first seven years, between 1998 and 2004, of which the survival curves were extrapolated to 2010 and compared with the Kaplan-Meier estimates based on the 13-year follow-up (Table 3). The relative biases of the extrapolation ranged between + 7.71% (p = 0.122) and + 1.29% (p = 0.551).