- Email: [email protected]
Population Aging, Technological Innovation, and the Growth of Health Expenditure: Evidence From Patients With Type 2 Diabetes in Taiwan
-
Contents lists available at sciencedirect.com Journal homepage: www.elsevier.com/locate/vhri
Population Aging, Technological Innovation, and the Growth of Health Expenditure: Evidence From Patients With Type 2 Diabetes in Taiwan Ya-Ming Liu, PhD* Department of Economics, College of Social Sciences, National Cheng Kung University, Tainan, Taiwan.
A B S T R A C T Objectives: As populations are growing older, the prevalence of chronic diseases such as diabetes mellitus is rapidly increasing. Meanwhile, many new drugs are introduced each year as a result of technological advances. This study uses diabetes as an example to investigate the relative importance of population aging and technological innovation in accounting for the growth of health expenditures. Methods: The retrospective cohort study was conducted based on claims data covering 1997 to 2006 taken from Taiwan’s National Health Insurance. Patients were selected based on whether they received antidiabetic drugs. Growth in health expenditure was decomposed into 3 parts: number of patients, mean treatment cost, and the interaction between the change in the mean treatment cost and the change in the number of patients. Results: The results indicated that 75% of the growth in expenditures for treating diabetic patients is attributable to the effect of population aging, as reflected by the increase in the diabetes prevalence rate (45%) and disease severity (30%). Technological innovation, in the form of treatment substitution (10%) and treatment expansion effects (15%), accounted for only about 25% of the growth in expenditures for treating diabetic patients. Conclusions: Population aging plays a more significant role than technological innovation in driving up health expenditures for the treatment of diabetic patients. This suggests that population aging may contribute significantly to the future growth of the healthcare sector in Asian countries such as Taiwan. Keywords: diabetes, health expenditure, new drug, population aging, technological innovation. VALUE IN HEALTH REGIONAL ISSUES. 2020; 21(C):120–126
Introduction Health systems around the world face 2 dynamic challenges that result in higher healthcare costs: technological change in medicine (“technological innovation” hereafter) and demographic change in the age composition of global populations (“population aging” hereafter). The literature has long proposed the consensus view that technological innovation is the major driver of health expenditure growth.1 The question of whether population aging is also a major cause of higher healthcare costs has been intensely debated.2-7 Some argue that population aging, as reflected by the increasing share of the elderly in the total population, will cause a major increase in health expenditure given the evidence of a strong positive association between age and healthcare costs.2 Others suggest that the effect of aging on health expenditure is
relatively small after life expectancy or remaining lifetime is controlled for.3-5 For example, a French study found that increases in healthcare expenditures are only slightly attributable to aging because the savings from health improvements offset the increase in health expenditures due to aging. The main driver of health expenditures is the change in practices due to technological innovation, and this effect is increasing with age.3 Most studies that investigate the effect of aging on health expenditure have used data obtained from Western countries. The experiences of aging in Asian countries, the demographic structures of which differ significantly from those of Western countries, have been researched much less frequently. Whether results of analyses on the effect of population aging on health expenditure growth vary across countries with different demographic structures is an issue worth exploring.
Funding: This work was supported by the Ministry of Science and Technology in Taiwan (MOST 105-2410-H-006 -003 -MY). Publication of the study results was not contingent upon the sponsors’ approval. Conflict of interest: The author has indicated that she has no conflicts of interest with regard to the content of this article. Ethical approval: Not required. * Address correspondence to: Ya-Ming Liu, PhD, Department of Economics, College of Social Sciences, National Cheng Kung University, 1 University Rd, Tainan, 70101 Taiwan. Email: [email protected] 2212-1099/$36.00 - see front matter ª 2019 ISPOR–The professional society for health economics and outcomes research. Published by Elsevier Inc. https://doi.org/10.1016/j.vhri.2019.07.012
--
In Taiwan, spending on diabetic patients accounts for more than 10% of total National Health Insurance (NHI) expenditures.8 The estimated prevalence rate of diabetes mellitus in Taiwan was about 10% from 1995 to 2003, and about half of the cases remained undiagnosed.9 Figure 1 shows that the percentage of Taiwan’s population older than 65 years increased from 8.06% in 1997 to 10% in 2006, and the prevalence rate of diabetes mellitus increased from 1.57% to 4.20% during the same period, suggesting that the growth in the prevalence of chronic diseases might have increased at a rate faster than that of population aging. This study contributes to the growing body of research on this debated issue by using data obtained from Taiwan, which has a population younger than the populations of most high-income Western countries but which is also facing a population aging trend. This study differs from existing studies in 2 specific ways. First, this study empirically investigates the effect of population aging on health expenditure growth by targeting a specific disease, diabetes mellitus. Targeting a specific disease allows decomposing the channels through which population aging is driving up health expenditures. Second, this study not only examines the effect of population aging but also compares population aging to technological innovation in terms of their importance for the growth of health expenditures, because population aging is an important engine of technological innovation.10
Conceptual Framework A conceptual framework was first established to explain how population aging and technological innovation drive up health expenditures, which in turn provides a basis on which to decompose the sources of expenditure growth in this empirical study. It was assumed that the increase in health expenditures arises from 2 dynamic factors: (1) population aging and (2) technological innovation. The other factors that can affect health expenditures in the real world—such as incomes, lifestyles, practice styles, health insurance, and payment systems—are assumed to be static (i.e., remain constant over time). Technological innovation and population aging are the major influential factors in health costs. In this framework, population aging drives up health expenditures for treating specific diseases such as diabetes through 2 channels: an increase in the number of patients being treated (a change in quantity) and an increase in the mean cost of treatment per patient (a change in price).11,12 The diagnosis of diabetes was recognized as a mature technology by the 1990s. Thus, I assume that the sole cause of the increase in the number of diabetic patients is population aging, given the evidence that the incidence of diabetes mellitus is positively associated with age.13 By contrast, the increase in the mean cost of treatment per patient has 2 plausible explanations. On one hand, diabetes mellitus is a chronic disease, the severity of which is likely to increase with age, as shown in Figure 1. Thus, as the mean age of diabetic patients increases, the mean cost of treatment per patient also increases. On the other hand, technological innovation in the treatment of diabetic patients also contributes to the increase in the mean cost of treatment per patient, insofar as the new technology is cost-increasing instead of cost-saving. The adoption of new technology in medicine affects the cost of healthcare through 2 primary channels: a treatment substitution effect and a treatment expansion effect.14 In the context of pharmaceuticals, the treatment substitution effect describes the substitution of new drugs for older ones in the treatment of
121
Figure 1. Population aging and DM prevalence rate, 1997–2006. DM indicates diabetes mellitus.
established patients. The treatment expansion effect occurs when new drugs are included in the formulary and new patients are treated for the disease. In practice, the treatment substitution effect may lead to either an increase or decrease in health expenditure. However, because of the high costs of pharmaceutical innovation, which are partly reflected in high on-patent prices, the price of a new drug is usually higher than that of an older drug, at least for drugs that are no longer patented. Thus, the treatment substitution effect will typically lead to increased pharmaceutical expenditures. On the other hand, the increase in pharmaceutical expenditures may be offset by savings in other components of healthcare services. For example, a new drug may serve as a substitute for a surgical procedure and hence reduce hospital spending or help to maintain functioning and prevent complications for patients with chronic diseases. Whether the saving of other health expenditures is greater than the increased expenditure on new drugs is an empirical question. One study found that the adoption of new drugs tends to reduce all types of nondrug health expenditure and that this effect dominates the increased expenditure on new drugs.15 However, 2 more recent studies found the opposite result: that the adoption of new drugs tends to increase overall health expenditures.16,17 The treatment expansion effect may lead to an increase in health expenditure because more people are receiving care as the result of the adoption of the new technology. Treatment expansion would reduce spending only if the new technology were to prevent later costly complications, thus saving overall lifetime healthcare spending for an individual patient. Although the treatment substitution and expansion effects may lead to reduced health expenditure in some special cases, adoption of new technology has typically increased overall healthcare expenditures.17 Thus, it was hypothesized that the adoption of new technology for treating diabetic patients increases the mean cost of treatment per patient. The decomposition approach described below was used to explain the extent to which the change in the mean cost of treatment could be attributable to technological innovation. This study’s decomposition of expenditure growth is shown in Figure 2. In the base year, the mean cost of treatment per patient and the number of patients are represented by P0 and Q0, respectively. In the current period, the mean cost of treatment per patient and the number of patients are represented by P1 and Q1,
122
- 2020
VALUE IN HEALTH REGIONAL ISSUES
Figure 2. Decomposition of health expenditure growth.
P (mean cost of treatment)
Effects of Population Aging A: Increase in number of patients B: Increase in disease severity Effects of Technological Innovation C: Treatment substitution effect
P1
D: Treatment expansion effect
C
D
αP1
B
P0 A
O
Q0
respectively. a represents the portion of the change in the mean treatment cost (P1 – P0) that is attributable to the increase in disease severity. Thus, 1 – a represents the portion of the change in the mean treatment cost attributable to technological innovation. Therefore, the shaded area A represents the increase in health expenditure attributable to the increase in the number of patients. The shaded area B represents the increase in health expenditure attributable to the increase in disease severity. The summation of areas A and B represents the effect of population aging on health expenditure growth. Similarly, the shaded area C represents the effect of treatment substitution on expenditure growth (i.e., the change in treatment cost for old patients arising from new technology adoption). The shaded area D represents the effect of treatment expansion on expenditure growth (ie, the change in treatment cost for new patients arising from technological innovation). Therefore, the summation of areas C and D represents the effect of technological innovation on health expenditure growth. Based on the above framework, I decompose the source of expenditure growth into 2 stages. First, I decompose the growth in health expenditure into 2 parts: 1) the change in the number of patients and 2) the change in the mean treatment cost. Second, I estimate the percentage change in the mean treatment cost attributable to technological innovation.
Data and Methods Data Source The data used in this study were obtained from a longitudinal data set containing 1 million individuals (about 4.34% of Taiwan’s population) randomly selected from the registry of NHI beneficiaries in 2005. The sampled file was merged with insurance claim files that trace back all the medical utilization records of the same individuals in each year and follow their medical utilization data for subsequent years (the 2005 Longitudinal Health Insurance
Q1
Q (number of patients)
Database [LHID]). This data set was issued and compiled by the National Health Research Institutes in Taiwan during the study period. The LHID comes from the National Health Insurance Research Database, which is compiled by the National Health Research Institute in Taiwan. This data set contains detailed records on the utilization of personal healthcare services, including outpatient visits, hospital admissions, and prescription drugs. The advantage of this data set is that it allows all the medical utilization data to be linked together for the same patient. In addition, this data set is based on national sampling data and serves as a good representation of the population data. Therefore, the findings of this study can be generalized to Taiwan’s population as a whole. I selected diabetes to decompose the sources of health expenditure growth in terms of higher prevalence rates and treatment costs. Estimating the prevalence rate based only on individuals with outpatient visits for diabetes underestimates the prevalence of diabetes mellitus in my data set. However, I can still obtain time-trend information from the data. As can be seen from Figure 1, the growth in the prevalence of chronic diseases might have increased at a rate faster than that of population aging. Given that Taiwan’s proportion of elderly individuals (65 years or older) is expected to reach 16.3% by 2020,18 the changing demographic trend, along with the long-term treatment required for chronic diseases, is likely not only to raise healthcare costs but also to create a large potential market for healthcare services in these therapeutic categories, which in turn provides strong incentives to induce technological innovation. Thus, diabetes provides an ideal setting in which to investigate the relative importance of population aging vis-à-vis technological innovation in accounting for the growth of healthcare expenditures.
Method I defined patients with diabetes mellitus (“DM patients” hereafter) as patients who have taken antidiabetic drugs, namely, oral hypoglycemic agents (OHAs). I selected DM patients based on
--
123
Table 1. Decomposition of health expenditure growth for diabetic patients in Taiwan. Year
Number of patients (Q)
1997 2006 Index of growth between 1997 and 2006 (1997 = 1)
Total health expenditure (E) (in million NT$)
Mean annual treatment cost per patient (P = E/Q) (in NT$)
15 728
290
18 452
42 004
1364
32 469
2.67
4.70
1.76
Note. The dollar value is expressed in terms of 2006 NT dollars. The exchange rate (NT$/US$) was about 32.5 between 1997 and 2006. Diabetic patients are defined as patients who have taken antidiabetic drugs, namely, oral hypoglycemic agents. Source: 2005 Longitudinal Health Insurance Database, calculated by the authors.
whether physicians prescribed antidiabetic drugs to them. This is a more reliable criterion than International Classification of Diseases, Ninth Revision, Clinical Modification codes, because the detailed list of prescription drugs is the key to determining the amount of reimbursement, whereas the accuracy of the International Classification of Diseases, Ninth Revision, Clinical Modification codes is not essential in reimbursements for prescription drugs. Using sampling cohort claims data from the NHI program, I compared the number of DM patients (Q) and total health expenditure (deflated by the 2005 Consumer Price Index) associated with DM patients (E) between 1997 and 2006. I then calculated the mean cost of treatment per patient (P) by dividing total health expenditure by the number of patients (E/Q). I chose 1997 as the base year and 2006 as the current period. I thus normalized all variables (ie, price, quantity, and expenditure) in 1997 to 1 and then calculated the growth index for these variables in 2006. This process allowed me to decompose the growth of health expenditure for treating DM patients into 2 components: (1) the change in the number of patients (OQ) and (2) the change in the mean cost of treatment per patient (OP). Because the growth of the mean treatment cost (OP) may arise from 2 sources—the increase in disease severity and new technology adoption—I further estimated the portion of the change in the mean treatment cost attributable to technological innovation, to estimate the value of 1 – a. Given that OHAs play an important role in the treatment of diabetes, I further decomposed the mean treatment costs into 3 components: 1) nondrug costs, 2) drug costs allocated to OHAs, and 3) drug costs allocated to nonOHAs. I used drug costs allocated to OHAs as the benchmark to quantify the impact of technological innovation on the mean treatment cost. I followed the same approach as that described in the first stage to decompose the sources of OHA expenditure growth into 2 components: 1) the change in price and 2) the change in quantity. Since the number of OHA products the patient receives per visit serves as a good proxy by which to measure the severity of the diabetes per se,19 the increase in the quantity of OHAs consumed per patient over time reflects the increase in disease severity. Therefore, I attributed the change in the quantity consumed to the effect of the increase in disease severity and attributed the change in price to the adoption of pharmaceutical innovation. My data set did not provide enough information to decompose the effects of population aging and technological innovation on the growth of nondrug costs and drug costs allocated to nonOHAs. For simplicity, then, I assumed that the impact of technological innovation was identical across all forms of healthcare input. I assumed that technological changes in the treatment of diabetic patients were neutral in the sense that the forms of technological innovation were not biased toward a particular component of healthcare inputs. Therefore, I applied the
analytical results for OHA costs to nondrug and non-OHA costs as well.
Results First Stage Table 1 reports the first-stage results that decompose health expenditure growth for diabetic patients in Taiwan. Based on the LHID, I identified 15 728 patients who used OHAs to treat their respective diseases in 1997. The total NHI health costs incurred by these patients in 1997 amounted to NT$290 million, indicating that the mean treatment cost per patient was NT$18 452 in 1997. By 2006, the number of DM patients increased to 42 004, indicating that the index of growth in the number of patients increased from 1 in 1997 to 2.67 in 2006, a growth of 167% within a decade. Following the same reasoning, I found that the growth indices for total healthcare expenditures and mean treatment costs per patient are 4.70 and 1.76, respectively. This suggests that the mean treatment cost per patient increased by 76% and that total health expenditure increased by 370% within a decade. I plotted the results reported in Table 1 into Figure 3, where the horizontal axis represents the number of patients and the vertical axis represents the mean treatment cost per patient. In Figure 2, the white area represents the total health expenditure in 1997, which is equal to 1 based on normalization (P0 3 Q0). The shaded area represents the growth of health expenditure in 2006 relative to that in 1997. According to the data reported in Table 1, the size of the shaded area is equal to 3.70 (calculated from 4.70 – 1). I decomposed this expenditure growth into 3 sources. The first is an increase in the number of patients, represented by OQ 3 P0; the size of this area is 1.67, indicating that the number of patients increased by 167% between 1997 and 2006 and thus that total health expenditure would increase from 1 in 1997 to 2.67 in 2006 if the mean treatment cost per patient were to remain constant. This result suggests that the change in the number of patients alone accounts for about 45% (1.67/3.70) of the expenditure growth. The second source is an increase in the mean cost of treatment, represented by OP 3 Q0; the size of this area is 0.76, indicating that the mean treatment cost increased by 76% between 1997 and 2006 and thus that total health expenditure would increase from 1 in 1997 to 1.76 in 2006 if the number of patients were to remain constant. This result suggests that the change in the mean cost of treatment alone account for about 21% (0.76/3.70) of the expenditure growth. The third source is the interaction effect between the change in the mean treatment cost and the change in the number of patients, represented by OP 3 OQ; the size of this area is 1.27, which is the product of the above 2 shaded areas. This shaded area represents the new patients, as measured by the increase in the
124
- 2020
VALUE IN HEALTH REGIONAL ISSUES
Figure 3. Relative importance of sources of health expenditure growth.
P (mean treatment cost)
1.76
0.76(ΔP × Q0)
1.27(ΔP × ΔQ)
(21%)
(34%)
1(p0) 1
1.67(ΔQ × P0) (45%)
O
1(Q0)
ΔQ
number of patients between 1997 and 2006, who were treated at a higher treatment cost. Thus, the health expenditure in relation to the new patients is higher than that of the old patients, as measured by the number of patients observed in 1997. My analysis shows that this interaction effect accounts for 34% (1.27/3.70) of the expenditure growth. Summing the effects from the second and third sources, the findings suggest that 55% of the expenditure growth is attributable to the increase in the mean treatment cost. In the next decomposition stage, I estimate to what extent this change could be attributed to technological innovation.
Second Stage To help identify the sources of change in the mean treatment cost, the upper panel of Table 2 reports the results of this decomposition. The results suggest 3 important findings. First, drug costs account for about 50% of the treatment costs for diabetic patients. Second, the growth rate of nondrug costs exceeds the growth rate of drug costs. Thus, the share of nondrug costs in total healthcare costs increased over time, from 47% in 1997 to 51% in 2006. Third, spending on OHAs accounts for about 16% of total health costs for diabetic patients, and the growth rate of OHAs is greater than that of non-OHAs. Overall, the analysis indicates that the share of spending on OHA products in total drug costs increased over time, from 30% (2968/9719) in 1997 to 33% (5284/ 15825) in 2006. Given that spending on OHA products accounts for about onethird of drug costs for diabetic patients, the growing prevalence rate of diabetes induced by population aging has created a large potential market for OHAs, which has in turn fostered pharmaceutical innovation in this therapeutic market. Therefore, I focus on OHA spending to further quantify the effect of technological innovation on the increase in mean treatment costs. The lower panel of Table 2 decomposes the sources of OHA expenditure growth into 2 components: (1) the change in price and (2) the change in quantity in defined daily doses (DDD). The
2.67
Q (number of patients)
results show that the annual quantity of OHA consumption per patient increased from 279 DDDs in 1997 to 399 DDDs in 2006, or a 43% increase. Meanwhile, mean OHA costs per patient increased from NT$2968 in 1997 to NT$5284 in 2006, a 78% increase. This result suggests that the change in the quantity of drug consumption per patient alone accounts for about 55% (0.43/0.78) of the expenditure growth. Similarly, the results show that the mean price per DDD increased from NT$10.65 in 1997 to NT$13.24 in 2006, a 24% increase within a decade, suggesting that the change in the mean price per DDD alone accounts for about 31% (0.24/ 0.78) of the OHA expenditure growth. In addition, an interaction effect is observed between the change in price and change in quantity, which in turn accounts for 14% ([0.24 3 0.43]/0.78) of the expenditure growth. Summing these 2 effects together suggests that about 45% of the increase in the cost of OHAs per patient during this 10-year period was attributable to technological innovation, as reflected by the increase in the mean price per DDD and the interaction effect between the increase in price and increase in quantity. Based on this simplified assumption, I conclude that the value of a (as shown in Figure 2) is equal to 0.55 and that the value of 1 – a is equal to 0.45. Therefore, my analysis finds that 55% of the increase in the mean treatment cost is attributable to the increase in disease severity, whereas 45% of the increase in the mean treatment cost is attributable to technological innovation. The sources of expenditure growth for treating diabetic patients in Taiwan are summarized in Table 3, which combines the analytical results of the above 2 stages. As the figure shows, the increase in the number of patients (prevalence rate) accounts for 45% of the expenditure growth, whereas the increase in the mean treatment cost per patient accounts for 55% (21% 1 34%). As mentioned, my analysis concludes that 55% of this effect is attributed to the increase in disease severity. Thus, the increase in disease severity alone accounts for 30% (0.55 3 0.55) of the expenditure growth. These 2 effects combined suggest that the effect of population aging accounts for 75% of the expenditure growth. As a result, the remaining 25% of the expenditure growth
--
125
Table 2. Decomposition of mean treatment costs for diabetic patients and spending on OHA drugs in Taiwan. Decomposition of mean treatment costs for diabetic patients Year
Nondrug
Drug costs (in NT$)
Costs (in NT$)
OHAs
Non-OHAs
Total
1997
8733 (0.47)
2968 (0.16)
6751 (0.37)
9719 (0.53)
2006
16 641 (0.51)
5284 (0.16)
10 541 (0.32)
15 825 (0.49)
1.91
1.78
1.56
1.63
Index of growth between 1997 and 2006 (1997 =1)
Decomposition of spending on OHA drugs Year
Index of growth between 1997 and 2006 (1997 = 1)
Weighted OHAs Price per DDD (P) (in NT$)
Quantity of OHA consumption per patient (Q) (in DDD)
Mean OHA costs per patient (P 3 Q) (in NT$)
10.65
279
2968
13.24
399
5284
1.24
1.43
1.78
DDD indicates defined daily doses; OHA, oral hypoglycemic agent. Note. The dollar value is expressed in terms of the 2006 NT dollar. The exchange rate (NT$/US$) was about 32.5 between 1997 and 2006. Numbers in parentheses indicate the share in total treatment costs. Source: 2005 Longitudinal Health Insurance Database, calculated by the authors.
is attributable to technological innovation, which is in turn the sum of the treatment substitution effect (10%) and treatment expansion effect (15%). The magnitude of the treatment substitution effect was calculated by multiplying the size of OP 3 Q0 (0.21), as shown in Figure 3, by 0.45. Similarly, the magnitude of the treatment expansion effect was calculated by multiplying the size of OP 3 OQ (0.34), as shown in Figure 3, by 0.45.
Discussion
Conclusion
Overall, the results indicate that population aging plays a more important role than technological innovation in accounting for the growth of healthcare expenditures on caring for diabetic patients. This result is in contrast to the findings in the literature, which generally suggest a smaller effect of population aging.3-5 This contrast highlights an important difference in the major drivers of rising healthcare costs between high-income Western countries and East Asian countries such as Taiwan. This difference has at least 2 plausible explanations. First, population aging has been faster in East Asian countries than in
Table 3. Summary of sources of expenditure growth for diabetic patients. Sources of expenditure growth Population aging Prevalence rate Disease severity* Subtotal Technological innovation Treatment substitution effect† Treatment expansion effect‡ Subtotal Total *Calculated as (0.21 1 0.34) 3 0.55. † Calculated as 0.21 3 0.45. ‡ Calculated as 0.34 3 0.45.
high-income Western countries over the past decade because of economic growth and demographic transitions. Therefore, the impact of population aging on healthcare costs is greater in Taiwan than in high-income Western countries. Second, Taiwan and other East Asian countries with similar developmental backgrounds have experienced more longevity gains during the past decade than high-income Western countries have, which increases the impact of population aging on healthcare costs.
Contribution rate (%) 45 30 75 10 15 25 100
This study uses retrospective NHI sampling cohort claims data to investigate the relative importance between population aging and technological innovation in accounting for the growth of expenditures for treating diabetic patients in Taiwan. The results suggest that population aging plays a more important role than technological innovation in driving up health expenditures for diabetic patients. Specifically, I found that about 75% of the growth in expenditures for treating diabetic patients is attributable to the effect of population aging, which in turn is reflected by the increase in the prevalence rate and disease severity of diabetes. Technological innovation, working through the treatment substitution and treatment expansion effects, accounts for only about 25% of the growth of expenditures for treating diabetic patients. An important implication of this study is that population aging may contribute much more significantly to the growth of expenditures in the health sector in Asian countries such as Taiwan than is claimed by most of the existing empirical evidence obtained from high-income Western countries.2-7 It is worth noting that the decomposition method used in previous studies differs from that used in this study. Future research using data from Western countries could verify the generalizability of this study’s findings. Given that technological innovation is endogenous in the sense that an increase in the size of the health sector provides strong incentives for innovation,10 the increase in health expenditure resulting from population aging will induce technological innovation, which will in turn provide another important driver for boosting healthcare costs. This suggests that population aging and the interaction effect between population aging and technological
126
- 2020
VALUE IN HEALTH REGIONAL ISSUES
innovation are important drivers of rising healthcare costs in aging societies. In addition, these research findings highlight the importance of health promotion and disease management programs as effective policy tools for controlling increases in healthcare costs. Given the evidence that the increase in the prevalence rate of diabetes accounts for about 45% of the expenditure growth, a health promotion policy that specifically targets diabetes could reduce the prevalence rate of this disease and thus reduce the rate of health expenditure growth. Similarly, a disease management program that helps diabetic patients maintain their health conditions could reduce the progress in disease severity and hence decrease the growth rate of health expenditure, given the evidence that the increase in disease severity per se accounts for about 30% of the growth in expenditures for caring for diabetic patients.
Finally, this study focuses on 1 specific category of disease. Whether this case study can be generalized to other types of diseases is a potential avenue for future study. Given that several high-cost therapeutic areas such as hypertension and oncology also account for a larger share of healthcare expenditures, expanding the research in those areas would be worthwhile.
REFERENCES 1. 2. 3. 4.
Limitations
5.
This study has 6 main limitations. First, I assumed that there were only 2 dynamic factors affecting the growth of healthcare costs. Other factors that may also change over time, such as income, practice styles, and payment systems, were ignored. Second, this study assumed that the sole cause of the increase in the prevalence rate of diabetes is population aging, which ignores the effect of obesity on the incidence of diabetes. Thus, the effect of population aging may be overestimated in this study. Third, this study cannot exclude from its analysis patients who receive OHAs for some other indications, such as polycystic ovary syndrome and ovarian hyperstimulation syndrome. Fourth, I assumed that technological innovation is neutral in the sense that innovation paths are identical across various forms of health inputs when, in fact, innovation may be more intensive toward some forms of healthcare inputs than toward others. For example, product innovation (such as in prescription drugs) may be more intensive than process innovation (such as in physician services), given that the former is rewarded by the patent system. My conclusion on the role of technological innovation may thus be overestimated (ie, 25% would be an upper bound). Fifth, the method of patient inclusion in the analysis was based on the use of OHAs, which excluded early-stage DM patients who rely on lifestyle treatment (eg, diet and exercise). Because the treatment costs for early-stage DM patients are relatively low, this exclusion would not affect the results in a significant way.
6. 7. 8. 9.
10. 11.
12. 13. 14. 15. 16. 17.
18.
19.
Newhouse JP. Medical care costs: how much welfare loss? J Econ Perspect. 1992;16(3):3–21. Denton FT, Gafni A, Spencer BG. Exploring the effects of population change on the costs of physician services. J Health Econ. 2002;21:781–803. Dormount B, Grignon M, Huber H. Health expenditure growth: reassessing the threat of aging. Health Econ. 2006;15:947–963. Zweifel P, Felder S, Meiers M. Aging of population and health care expenditure: a red herring? Health Econ. 1999;8:485–496. Shang B, Goldman D. Does age or life expectancy better predict health care expenditure Health. Econ. 2008;17:487–501. Getzen TE. Aging and health care expenditures: a comment on Zweifel, Felder and Meiers. Health Econ. 2001;10:175–177. Chernichovsky D, Markowitz S. Aging and aggregate costs of medical care: conceptual and policy issues. Health Econ. 2004;13:543–562. Lin T, Chou P, Lai MS, et al. Direct cost-of-illness of patients with diabetes mellitus in Taiwan. Diabetes Res Clin Pract. 2001;54:S43–S46. Chan JCN, Malik V, Jia W, Kadowaki T, Yajnik CS, Yoon KH, Hu FB. Diabetes in Asia: epidemiology, risk factors and pathophysiology. JAMA. 2009;301:2129– 2140. Acemoglu D, Linn J. Market size in innovation: theory and evidence from the pharmaceutical industry. Q J Econ. 2004;119:1049–1090. Addis A, Magrini N. New approaches to analyzing prescription data and to transfer pharmacoepidemiological and evidence-based reports to prescribers. Pharmacoepidemiol Drug Saf. 2002;11:721–726. Berndt ER. Pharmaceuticals in U.S. health care: determinants of quantity and price. J Econ Perspect. 2002;16(4):45–66. Wild S, Rogil G, Green A, et al. Global prevalence of diabetes: estimates for the year 2000 and projections for 2030. Diabetes Care. 2004;27:1047–1053. Cutler DM, McClellan M. Is technological change in medicine worth it? Health Aff. 2001;20:11–29. Lichtenberg F. Are the benefits of newer drugs worth their cost? Evidence from the 1996 MEPS. Health Aff. 2001;20:241–251. Duggan M. Do new prescription drugs pay for themselves? The case of second-generation antipsychotics. J Health Econ. 2005;24:1–31. Liu YM, Hsieh CR. New drugs and the growth of health expenditures: empirical evidence from diabetic patients in Taiwan. Health Econ. 2012;21:496–513. Manpower Planning Department, Council for Economic Planning and Development (CEPD). Estimates of Population in Taiwan from 2010 to 2060. Taipei: CEPD; 2010. Liu YM, Kao Yang YH, Hsieh CR. The determinants of the adoption of pharmaceutical innovation: evidence from Taiwan. Soc Sci Med. 2011;72: 919–927.
Contents lists available at sciencedirect.com Journal homepage: www.elsevier.com/locate/vhri
Population Aging, Technological Innovation, and the Growth of Health Expenditure: Evidence From Patients With Type 2 Diabetes in Taiwan Ya-Ming Liu, PhD* Department of Economics, College of Social Sciences, National Cheng Kung University, Tainan, Taiwan.
A B S T R A C T Objectives: As populations are growing older, the prevalence of chronic diseases such as diabetes mellitus is rapidly increasing. Meanwhile, many new drugs are introduced each year as a result of technological advances. This study uses diabetes as an example to investigate the relative importance of population aging and technological innovation in accounting for the growth of health expenditures. Methods: The retrospective cohort study was conducted based on claims data covering 1997 to 2006 taken from Taiwan’s National Health Insurance. Patients were selected based on whether they received antidiabetic drugs. Growth in health expenditure was decomposed into 3 parts: number of patients, mean treatment cost, and the interaction between the change in the mean treatment cost and the change in the number of patients. Results: The results indicated that 75% of the growth in expenditures for treating diabetic patients is attributable to the effect of population aging, as reflected by the increase in the diabetes prevalence rate (45%) and disease severity (30%). Technological innovation, in the form of treatment substitution (10%) and treatment expansion effects (15%), accounted for only about 25% of the growth in expenditures for treating diabetic patients. Conclusions: Population aging plays a more significant role than technological innovation in driving up health expenditures for the treatment of diabetic patients. This suggests that population aging may contribute significantly to the future growth of the healthcare sector in Asian countries such as Taiwan. Keywords: diabetes, health expenditure, new drug, population aging, technological innovation. VALUE IN HEALTH REGIONAL ISSUES. 2020; 21(C):120–126
Introduction Health systems around the world face 2 dynamic challenges that result in higher healthcare costs: technological change in medicine (“technological innovation” hereafter) and demographic change in the age composition of global populations (“population aging” hereafter). The literature has long proposed the consensus view that technological innovation is the major driver of health expenditure growth.1 The question of whether population aging is also a major cause of higher healthcare costs has been intensely debated.2-7 Some argue that population aging, as reflected by the increasing share of the elderly in the total population, will cause a major increase in health expenditure given the evidence of a strong positive association between age and healthcare costs.2 Others suggest that the effect of aging on health expenditure is
relatively small after life expectancy or remaining lifetime is controlled for.3-5 For example, a French study found that increases in healthcare expenditures are only slightly attributable to aging because the savings from health improvements offset the increase in health expenditures due to aging. The main driver of health expenditures is the change in practices due to technological innovation, and this effect is increasing with age.3 Most studies that investigate the effect of aging on health expenditure have used data obtained from Western countries. The experiences of aging in Asian countries, the demographic structures of which differ significantly from those of Western countries, have been researched much less frequently. Whether results of analyses on the effect of population aging on health expenditure growth vary across countries with different demographic structures is an issue worth exploring.
Funding: This work was supported by the Ministry of Science and Technology in Taiwan (MOST 105-2410-H-006 -003 -MY). Publication of the study results was not contingent upon the sponsors’ approval. Conflict of interest: The author has indicated that she has no conflicts of interest with regard to the content of this article. Ethical approval: Not required. * Address correspondence to: Ya-Ming Liu, PhD, Department of Economics, College of Social Sciences, National Cheng Kung University, 1 University Rd, Tainan, 70101 Taiwan. Email: [email protected] 2212-1099/$36.00 - see front matter ª 2019 ISPOR–The professional society for health economics and outcomes research. Published by Elsevier Inc. https://doi.org/10.1016/j.vhri.2019.07.012
--
In Taiwan, spending on diabetic patients accounts for more than 10% of total National Health Insurance (NHI) expenditures.8 The estimated prevalence rate of diabetes mellitus in Taiwan was about 10% from 1995 to 2003, and about half of the cases remained undiagnosed.9 Figure 1 shows that the percentage of Taiwan’s population older than 65 years increased from 8.06% in 1997 to 10% in 2006, and the prevalence rate of diabetes mellitus increased from 1.57% to 4.20% during the same period, suggesting that the growth in the prevalence of chronic diseases might have increased at a rate faster than that of population aging. This study contributes to the growing body of research on this debated issue by using data obtained from Taiwan, which has a population younger than the populations of most high-income Western countries but which is also facing a population aging trend. This study differs from existing studies in 2 specific ways. First, this study empirically investigates the effect of population aging on health expenditure growth by targeting a specific disease, diabetes mellitus. Targeting a specific disease allows decomposing the channels through which population aging is driving up health expenditures. Second, this study not only examines the effect of population aging but also compares population aging to technological innovation in terms of their importance for the growth of health expenditures, because population aging is an important engine of technological innovation.10
Conceptual Framework A conceptual framework was first established to explain how population aging and technological innovation drive up health expenditures, which in turn provides a basis on which to decompose the sources of expenditure growth in this empirical study. It was assumed that the increase in health expenditures arises from 2 dynamic factors: (1) population aging and (2) technological innovation. The other factors that can affect health expenditures in the real world—such as incomes, lifestyles, practice styles, health insurance, and payment systems—are assumed to be static (i.e., remain constant over time). Technological innovation and population aging are the major influential factors in health costs. In this framework, population aging drives up health expenditures for treating specific diseases such as diabetes through 2 channels: an increase in the number of patients being treated (a change in quantity) and an increase in the mean cost of treatment per patient (a change in price).11,12 The diagnosis of diabetes was recognized as a mature technology by the 1990s. Thus, I assume that the sole cause of the increase in the number of diabetic patients is population aging, given the evidence that the incidence of diabetes mellitus is positively associated with age.13 By contrast, the increase in the mean cost of treatment per patient has 2 plausible explanations. On one hand, diabetes mellitus is a chronic disease, the severity of which is likely to increase with age, as shown in Figure 1. Thus, as the mean age of diabetic patients increases, the mean cost of treatment per patient also increases. On the other hand, technological innovation in the treatment of diabetic patients also contributes to the increase in the mean cost of treatment per patient, insofar as the new technology is cost-increasing instead of cost-saving. The adoption of new technology in medicine affects the cost of healthcare through 2 primary channels: a treatment substitution effect and a treatment expansion effect.14 In the context of pharmaceuticals, the treatment substitution effect describes the substitution of new drugs for older ones in the treatment of
121
Figure 1. Population aging and DM prevalence rate, 1997–2006. DM indicates diabetes mellitus.
established patients. The treatment expansion effect occurs when new drugs are included in the formulary and new patients are treated for the disease. In practice, the treatment substitution effect may lead to either an increase or decrease in health expenditure. However, because of the high costs of pharmaceutical innovation, which are partly reflected in high on-patent prices, the price of a new drug is usually higher than that of an older drug, at least for drugs that are no longer patented. Thus, the treatment substitution effect will typically lead to increased pharmaceutical expenditures. On the other hand, the increase in pharmaceutical expenditures may be offset by savings in other components of healthcare services. For example, a new drug may serve as a substitute for a surgical procedure and hence reduce hospital spending or help to maintain functioning and prevent complications for patients with chronic diseases. Whether the saving of other health expenditures is greater than the increased expenditure on new drugs is an empirical question. One study found that the adoption of new drugs tends to reduce all types of nondrug health expenditure and that this effect dominates the increased expenditure on new drugs.15 However, 2 more recent studies found the opposite result: that the adoption of new drugs tends to increase overall health expenditures.16,17 The treatment expansion effect may lead to an increase in health expenditure because more people are receiving care as the result of the adoption of the new technology. Treatment expansion would reduce spending only if the new technology were to prevent later costly complications, thus saving overall lifetime healthcare spending for an individual patient. Although the treatment substitution and expansion effects may lead to reduced health expenditure in some special cases, adoption of new technology has typically increased overall healthcare expenditures.17 Thus, it was hypothesized that the adoption of new technology for treating diabetic patients increases the mean cost of treatment per patient. The decomposition approach described below was used to explain the extent to which the change in the mean cost of treatment could be attributable to technological innovation. This study’s decomposition of expenditure growth is shown in Figure 2. In the base year, the mean cost of treatment per patient and the number of patients are represented by P0 and Q0, respectively. In the current period, the mean cost of treatment per patient and the number of patients are represented by P1 and Q1,
122
- 2020
VALUE IN HEALTH REGIONAL ISSUES
Figure 2. Decomposition of health expenditure growth.
P (mean cost of treatment)
Effects of Population Aging A: Increase in number of patients B: Increase in disease severity Effects of Technological Innovation C: Treatment substitution effect
P1
D: Treatment expansion effect
C
D
αP1
B
P0 A
O
Q0
respectively. a represents the portion of the change in the mean treatment cost (P1 – P0) that is attributable to the increase in disease severity. Thus, 1 – a represents the portion of the change in the mean treatment cost attributable to technological innovation. Therefore, the shaded area A represents the increase in health expenditure attributable to the increase in the number of patients. The shaded area B represents the increase in health expenditure attributable to the increase in disease severity. The summation of areas A and B represents the effect of population aging on health expenditure growth. Similarly, the shaded area C represents the effect of treatment substitution on expenditure growth (i.e., the change in treatment cost for old patients arising from new technology adoption). The shaded area D represents the effect of treatment expansion on expenditure growth (ie, the change in treatment cost for new patients arising from technological innovation). Therefore, the summation of areas C and D represents the effect of technological innovation on health expenditure growth. Based on the above framework, I decompose the source of expenditure growth into 2 stages. First, I decompose the growth in health expenditure into 2 parts: 1) the change in the number of patients and 2) the change in the mean treatment cost. Second, I estimate the percentage change in the mean treatment cost attributable to technological innovation.
Data and Methods Data Source The data used in this study were obtained from a longitudinal data set containing 1 million individuals (about 4.34% of Taiwan’s population) randomly selected from the registry of NHI beneficiaries in 2005. The sampled file was merged with insurance claim files that trace back all the medical utilization records of the same individuals in each year and follow their medical utilization data for subsequent years (the 2005 Longitudinal Health Insurance
Q1
Q (number of patients)
Database [LHID]). This data set was issued and compiled by the National Health Research Institutes in Taiwan during the study period. The LHID comes from the National Health Insurance Research Database, which is compiled by the National Health Research Institute in Taiwan. This data set contains detailed records on the utilization of personal healthcare services, including outpatient visits, hospital admissions, and prescription drugs. The advantage of this data set is that it allows all the medical utilization data to be linked together for the same patient. In addition, this data set is based on national sampling data and serves as a good representation of the population data. Therefore, the findings of this study can be generalized to Taiwan’s population as a whole. I selected diabetes to decompose the sources of health expenditure growth in terms of higher prevalence rates and treatment costs. Estimating the prevalence rate based only on individuals with outpatient visits for diabetes underestimates the prevalence of diabetes mellitus in my data set. However, I can still obtain time-trend information from the data. As can be seen from Figure 1, the growth in the prevalence of chronic diseases might have increased at a rate faster than that of population aging. Given that Taiwan’s proportion of elderly individuals (65 years or older) is expected to reach 16.3% by 2020,18 the changing demographic trend, along with the long-term treatment required for chronic diseases, is likely not only to raise healthcare costs but also to create a large potential market for healthcare services in these therapeutic categories, which in turn provides strong incentives to induce technological innovation. Thus, diabetes provides an ideal setting in which to investigate the relative importance of population aging vis-à-vis technological innovation in accounting for the growth of healthcare expenditures.
Method I defined patients with diabetes mellitus (“DM patients” hereafter) as patients who have taken antidiabetic drugs, namely, oral hypoglycemic agents (OHAs). I selected DM patients based on
--
123
Table 1. Decomposition of health expenditure growth for diabetic patients in Taiwan. Year
Number of patients (Q)
1997 2006 Index of growth between 1997 and 2006 (1997 = 1)
Total health expenditure (E) (in million NT$)
Mean annual treatment cost per patient (P = E/Q) (in NT$)
15 728
290
18 452
42 004
1364
32 469
2.67
4.70
1.76
Note. The dollar value is expressed in terms of 2006 NT dollars. The exchange rate (NT$/US$) was about 32.5 between 1997 and 2006. Diabetic patients are defined as patients who have taken antidiabetic drugs, namely, oral hypoglycemic agents. Source: 2005 Longitudinal Health Insurance Database, calculated by the authors.
whether physicians prescribed antidiabetic drugs to them. This is a more reliable criterion than International Classification of Diseases, Ninth Revision, Clinical Modification codes, because the detailed list of prescription drugs is the key to determining the amount of reimbursement, whereas the accuracy of the International Classification of Diseases, Ninth Revision, Clinical Modification codes is not essential in reimbursements for prescription drugs. Using sampling cohort claims data from the NHI program, I compared the number of DM patients (Q) and total health expenditure (deflated by the 2005 Consumer Price Index) associated with DM patients (E) between 1997 and 2006. I then calculated the mean cost of treatment per patient (P) by dividing total health expenditure by the number of patients (E/Q). I chose 1997 as the base year and 2006 as the current period. I thus normalized all variables (ie, price, quantity, and expenditure) in 1997 to 1 and then calculated the growth index for these variables in 2006. This process allowed me to decompose the growth of health expenditure for treating DM patients into 2 components: (1) the change in the number of patients (OQ) and (2) the change in the mean cost of treatment per patient (OP). Because the growth of the mean treatment cost (OP) may arise from 2 sources—the increase in disease severity and new technology adoption—I further estimated the portion of the change in the mean treatment cost attributable to technological innovation, to estimate the value of 1 – a. Given that OHAs play an important role in the treatment of diabetes, I further decomposed the mean treatment costs into 3 components: 1) nondrug costs, 2) drug costs allocated to OHAs, and 3) drug costs allocated to nonOHAs. I used drug costs allocated to OHAs as the benchmark to quantify the impact of technological innovation on the mean treatment cost. I followed the same approach as that described in the first stage to decompose the sources of OHA expenditure growth into 2 components: 1) the change in price and 2) the change in quantity. Since the number of OHA products the patient receives per visit serves as a good proxy by which to measure the severity of the diabetes per se,19 the increase in the quantity of OHAs consumed per patient over time reflects the increase in disease severity. Therefore, I attributed the change in the quantity consumed to the effect of the increase in disease severity and attributed the change in price to the adoption of pharmaceutical innovation. My data set did not provide enough information to decompose the effects of population aging and technological innovation on the growth of nondrug costs and drug costs allocated to nonOHAs. For simplicity, then, I assumed that the impact of technological innovation was identical across all forms of healthcare input. I assumed that technological changes in the treatment of diabetic patients were neutral in the sense that the forms of technological innovation were not biased toward a particular component of healthcare inputs. Therefore, I applied the
analytical results for OHA costs to nondrug and non-OHA costs as well.
Results First Stage Table 1 reports the first-stage results that decompose health expenditure growth for diabetic patients in Taiwan. Based on the LHID, I identified 15 728 patients who used OHAs to treat their respective diseases in 1997. The total NHI health costs incurred by these patients in 1997 amounted to NT$290 million, indicating that the mean treatment cost per patient was NT$18 452 in 1997. By 2006, the number of DM patients increased to 42 004, indicating that the index of growth in the number of patients increased from 1 in 1997 to 2.67 in 2006, a growth of 167% within a decade. Following the same reasoning, I found that the growth indices for total healthcare expenditures and mean treatment costs per patient are 4.70 and 1.76, respectively. This suggests that the mean treatment cost per patient increased by 76% and that total health expenditure increased by 370% within a decade. I plotted the results reported in Table 1 into Figure 3, where the horizontal axis represents the number of patients and the vertical axis represents the mean treatment cost per patient. In Figure 2, the white area represents the total health expenditure in 1997, which is equal to 1 based on normalization (P0 3 Q0). The shaded area represents the growth of health expenditure in 2006 relative to that in 1997. According to the data reported in Table 1, the size of the shaded area is equal to 3.70 (calculated from 4.70 – 1). I decomposed this expenditure growth into 3 sources. The first is an increase in the number of patients, represented by OQ 3 P0; the size of this area is 1.67, indicating that the number of patients increased by 167% between 1997 and 2006 and thus that total health expenditure would increase from 1 in 1997 to 2.67 in 2006 if the mean treatment cost per patient were to remain constant. This result suggests that the change in the number of patients alone accounts for about 45% (1.67/3.70) of the expenditure growth. The second source is an increase in the mean cost of treatment, represented by OP 3 Q0; the size of this area is 0.76, indicating that the mean treatment cost increased by 76% between 1997 and 2006 and thus that total health expenditure would increase from 1 in 1997 to 1.76 in 2006 if the number of patients were to remain constant. This result suggests that the change in the mean cost of treatment alone account for about 21% (0.76/3.70) of the expenditure growth. The third source is the interaction effect between the change in the mean treatment cost and the change in the number of patients, represented by OP 3 OQ; the size of this area is 1.27, which is the product of the above 2 shaded areas. This shaded area represents the new patients, as measured by the increase in the
124
- 2020
VALUE IN HEALTH REGIONAL ISSUES
Figure 3. Relative importance of sources of health expenditure growth.
P (mean treatment cost)
1.76
0.76(ΔP × Q0)
1.27(ΔP × ΔQ)
(21%)
(34%)
1(p0) 1
1.67(ΔQ × P0) (45%)
O
1(Q0)
ΔQ
number of patients between 1997 and 2006, who were treated at a higher treatment cost. Thus, the health expenditure in relation to the new patients is higher than that of the old patients, as measured by the number of patients observed in 1997. My analysis shows that this interaction effect accounts for 34% (1.27/3.70) of the expenditure growth. Summing the effects from the second and third sources, the findings suggest that 55% of the expenditure growth is attributable to the increase in the mean treatment cost. In the next decomposition stage, I estimate to what extent this change could be attributed to technological innovation.
Second Stage To help identify the sources of change in the mean treatment cost, the upper panel of Table 2 reports the results of this decomposition. The results suggest 3 important findings. First, drug costs account for about 50% of the treatment costs for diabetic patients. Second, the growth rate of nondrug costs exceeds the growth rate of drug costs. Thus, the share of nondrug costs in total healthcare costs increased over time, from 47% in 1997 to 51% in 2006. Third, spending on OHAs accounts for about 16% of total health costs for diabetic patients, and the growth rate of OHAs is greater than that of non-OHAs. Overall, the analysis indicates that the share of spending on OHA products in total drug costs increased over time, from 30% (2968/9719) in 1997 to 33% (5284/ 15825) in 2006. Given that spending on OHA products accounts for about onethird of drug costs for diabetic patients, the growing prevalence rate of diabetes induced by population aging has created a large potential market for OHAs, which has in turn fostered pharmaceutical innovation in this therapeutic market. Therefore, I focus on OHA spending to further quantify the effect of technological innovation on the increase in mean treatment costs. The lower panel of Table 2 decomposes the sources of OHA expenditure growth into 2 components: (1) the change in price and (2) the change in quantity in defined daily doses (DDD). The
2.67
Q (number of patients)
results show that the annual quantity of OHA consumption per patient increased from 279 DDDs in 1997 to 399 DDDs in 2006, or a 43% increase. Meanwhile, mean OHA costs per patient increased from NT$2968 in 1997 to NT$5284 in 2006, a 78% increase. This result suggests that the change in the quantity of drug consumption per patient alone accounts for about 55% (0.43/0.78) of the expenditure growth. Similarly, the results show that the mean price per DDD increased from NT$10.65 in 1997 to NT$13.24 in 2006, a 24% increase within a decade, suggesting that the change in the mean price per DDD alone accounts for about 31% (0.24/ 0.78) of the OHA expenditure growth. In addition, an interaction effect is observed between the change in price and change in quantity, which in turn accounts for 14% ([0.24 3 0.43]/0.78) of the expenditure growth. Summing these 2 effects together suggests that about 45% of the increase in the cost of OHAs per patient during this 10-year period was attributable to technological innovation, as reflected by the increase in the mean price per DDD and the interaction effect between the increase in price and increase in quantity. Based on this simplified assumption, I conclude that the value of a (as shown in Figure 2) is equal to 0.55 and that the value of 1 – a is equal to 0.45. Therefore, my analysis finds that 55% of the increase in the mean treatment cost is attributable to the increase in disease severity, whereas 45% of the increase in the mean treatment cost is attributable to technological innovation. The sources of expenditure growth for treating diabetic patients in Taiwan are summarized in Table 3, which combines the analytical results of the above 2 stages. As the figure shows, the increase in the number of patients (prevalence rate) accounts for 45% of the expenditure growth, whereas the increase in the mean treatment cost per patient accounts for 55% (21% 1 34%). As mentioned, my analysis concludes that 55% of this effect is attributed to the increase in disease severity. Thus, the increase in disease severity alone accounts for 30% (0.55 3 0.55) of the expenditure growth. These 2 effects combined suggest that the effect of population aging accounts for 75% of the expenditure growth. As a result, the remaining 25% of the expenditure growth
--
125
Table 2. Decomposition of mean treatment costs for diabetic patients and spending on OHA drugs in Taiwan. Decomposition of mean treatment costs for diabetic patients Year
Nondrug
Drug costs (in NT$)
Costs (in NT$)
OHAs
Non-OHAs
Total
1997
8733 (0.47)
2968 (0.16)
6751 (0.37)
9719 (0.53)
2006
16 641 (0.51)
5284 (0.16)
10 541 (0.32)
15 825 (0.49)
1.91
1.78
1.56
1.63
Index of growth between 1997 and 2006 (1997 =1)
Decomposition of spending on OHA drugs Year
Index of growth between 1997 and 2006 (1997 = 1)
Weighted OHAs Price per DDD (P) (in NT$)
Quantity of OHA consumption per patient (Q) (in DDD)
Mean OHA costs per patient (P 3 Q) (in NT$)
10.65
279
2968
13.24
399
5284
1.24
1.43
1.78
DDD indicates defined daily doses; OHA, oral hypoglycemic agent. Note. The dollar value is expressed in terms of the 2006 NT dollar. The exchange rate (NT$/US$) was about 32.5 between 1997 and 2006. Numbers in parentheses indicate the share in total treatment costs. Source: 2005 Longitudinal Health Insurance Database, calculated by the authors.
is attributable to technological innovation, which is in turn the sum of the treatment substitution effect (10%) and treatment expansion effect (15%). The magnitude of the treatment substitution effect was calculated by multiplying the size of OP 3 Q0 (0.21), as shown in Figure 3, by 0.45. Similarly, the magnitude of the treatment expansion effect was calculated by multiplying the size of OP 3 OQ (0.34), as shown in Figure 3, by 0.45.
Discussion
Conclusion
Overall, the results indicate that population aging plays a more important role than technological innovation in accounting for the growth of healthcare expenditures on caring for diabetic patients. This result is in contrast to the findings in the literature, which generally suggest a smaller effect of population aging.3-5 This contrast highlights an important difference in the major drivers of rising healthcare costs between high-income Western countries and East Asian countries such as Taiwan. This difference has at least 2 plausible explanations. First, population aging has been faster in East Asian countries than in
Table 3. Summary of sources of expenditure growth for diabetic patients. Sources of expenditure growth Population aging Prevalence rate Disease severity* Subtotal Technological innovation Treatment substitution effect† Treatment expansion effect‡ Subtotal Total *Calculated as (0.21 1 0.34) 3 0.55. † Calculated as 0.21 3 0.45. ‡ Calculated as 0.34 3 0.45.
high-income Western countries over the past decade because of economic growth and demographic transitions. Therefore, the impact of population aging on healthcare costs is greater in Taiwan than in high-income Western countries. Second, Taiwan and other East Asian countries with similar developmental backgrounds have experienced more longevity gains during the past decade than high-income Western countries have, which increases the impact of population aging on healthcare costs.
Contribution rate (%) 45 30 75 10 15 25 100
This study uses retrospective NHI sampling cohort claims data to investigate the relative importance between population aging and technological innovation in accounting for the growth of expenditures for treating diabetic patients in Taiwan. The results suggest that population aging plays a more important role than technological innovation in driving up health expenditures for diabetic patients. Specifically, I found that about 75% of the growth in expenditures for treating diabetic patients is attributable to the effect of population aging, which in turn is reflected by the increase in the prevalence rate and disease severity of diabetes. Technological innovation, working through the treatment substitution and treatment expansion effects, accounts for only about 25% of the growth of expenditures for treating diabetic patients. An important implication of this study is that population aging may contribute much more significantly to the growth of expenditures in the health sector in Asian countries such as Taiwan than is claimed by most of the existing empirical evidence obtained from high-income Western countries.2-7 It is worth noting that the decomposition method used in previous studies differs from that used in this study. Future research using data from Western countries could verify the generalizability of this study’s findings. Given that technological innovation is endogenous in the sense that an increase in the size of the health sector provides strong incentives for innovation,10 the increase in health expenditure resulting from population aging will induce technological innovation, which will in turn provide another important driver for boosting healthcare costs. This suggests that population aging and the interaction effect between population aging and technological
126
- 2020
VALUE IN HEALTH REGIONAL ISSUES
innovation are important drivers of rising healthcare costs in aging societies. In addition, these research findings highlight the importance of health promotion and disease management programs as effective policy tools for controlling increases in healthcare costs. Given the evidence that the increase in the prevalence rate of diabetes accounts for about 45% of the expenditure growth, a health promotion policy that specifically targets diabetes could reduce the prevalence rate of this disease and thus reduce the rate of health expenditure growth. Similarly, a disease management program that helps diabetic patients maintain their health conditions could reduce the progress in disease severity and hence decrease the growth rate of health expenditure, given the evidence that the increase in disease severity per se accounts for about 30% of the growth in expenditures for caring for diabetic patients.
Finally, this study focuses on 1 specific category of disease. Whether this case study can be generalized to other types of diseases is a potential avenue for future study. Given that several high-cost therapeutic areas such as hypertension and oncology also account for a larger share of healthcare expenditures, expanding the research in those areas would be worthwhile.
REFERENCES 1. 2. 3. 4.
Limitations
5.
This study has 6 main limitations. First, I assumed that there were only 2 dynamic factors affecting the growth of healthcare costs. Other factors that may also change over time, such as income, practice styles, and payment systems, were ignored. Second, this study assumed that the sole cause of the increase in the prevalence rate of diabetes is population aging, which ignores the effect of obesity on the incidence of diabetes. Thus, the effect of population aging may be overestimated in this study. Third, this study cannot exclude from its analysis patients who receive OHAs for some other indications, such as polycystic ovary syndrome and ovarian hyperstimulation syndrome. Fourth, I assumed that technological innovation is neutral in the sense that innovation paths are identical across various forms of health inputs when, in fact, innovation may be more intensive toward some forms of healthcare inputs than toward others. For example, product innovation (such as in prescription drugs) may be more intensive than process innovation (such as in physician services), given that the former is rewarded by the patent system. My conclusion on the role of technological innovation may thus be overestimated (ie, 25% would be an upper bound). Fifth, the method of patient inclusion in the analysis was based on the use of OHAs, which excluded early-stage DM patients who rely on lifestyle treatment (eg, diet and exercise). Because the treatment costs for early-stage DM patients are relatively low, this exclusion would not affect the results in a significant way.
6. 7. 8. 9.
10. 11.
12. 13. 14. 15. 16. 17.
18.
19.
Newhouse JP. Medical care costs: how much welfare loss? J Econ Perspect. 1992;16(3):3–21. Denton FT, Gafni A, Spencer BG. Exploring the effects of population change on the costs of physician services. J Health Econ. 2002;21:781–803. Dormount B, Grignon M, Huber H. Health expenditure growth: reassessing the threat of aging. Health Econ. 2006;15:947–963. Zweifel P, Felder S, Meiers M. Aging of population and health care expenditure: a red herring? Health Econ. 1999;8:485–496. Shang B, Goldman D. Does age or life expectancy better predict health care expenditure Health. Econ. 2008;17:487–501. Getzen TE. Aging and health care expenditures: a comment on Zweifel, Felder and Meiers. Health Econ. 2001;10:175–177. Chernichovsky D, Markowitz S. Aging and aggregate costs of medical care: conceptual and policy issues. Health Econ. 2004;13:543–562. Lin T, Chou P, Lai MS, et al. Direct cost-of-illness of patients with diabetes mellitus in Taiwan. Diabetes Res Clin Pract. 2001;54:S43–S46. Chan JCN, Malik V, Jia W, Kadowaki T, Yajnik CS, Yoon KH, Hu FB. Diabetes in Asia: epidemiology, risk factors and pathophysiology. JAMA. 2009;301:2129– 2140. Acemoglu D, Linn J. Market size in innovation: theory and evidence from the pharmaceutical industry. Q J Econ. 2004;119:1049–1090. Addis A, Magrini N. New approaches to analyzing prescription data and to transfer pharmacoepidemiological and evidence-based reports to prescribers. Pharmacoepidemiol Drug Saf. 2002;11:721–726. Berndt ER. Pharmaceuticals in U.S. health care: determinants of quantity and price. J Econ Perspect. 2002;16(4):45–66. Wild S, Rogil G, Green A, et al. Global prevalence of diabetes: estimates for the year 2000 and projections for 2030. Diabetes Care. 2004;27:1047–1053. Cutler DM, McClellan M. Is technological change in medicine worth it? Health Aff. 2001;20:11–29. Lichtenberg F. Are the benefits of newer drugs worth their cost? Evidence from the 1996 MEPS. Health Aff. 2001;20:241–251. Duggan M. Do new prescription drugs pay for themselves? The case of second-generation antipsychotics. J Health Econ. 2005;24:1–31. Liu YM, Hsieh CR. New drugs and the growth of health expenditures: empirical evidence from diabetic patients in Taiwan. Health Econ. 2012;21:496–513. Manpower Planning Department, Council for Economic Planning and Development (CEPD). Estimates of Population in Taiwan from 2010 to 2060. Taipei: CEPD; 2010. Liu YM, Kao Yang YH, Hsieh CR. The determinants of the adoption of pharmaceutical innovation: evidence from Taiwan. Soc Sci Med. 2011;72: 919–927.