Objective: To assess the health-related economic burden attributable to smoking in China for persons aged 35 and older.
Methods: A prevalence-based, disease-specific approach was used to estimate the smoking attributable direct costs, indirect morbidity costs, and costs of premature deaths caused by smoking-related diseases. The primary data source was the 1998 China National Health Services Survey, which contains the smoking status, medical utilisation, and expenditures for 216 101 individuals.
Results: The economic costs of smoking in 2000 amounted to $5.0 billion (measured in 2000, US$) in total and $25.43 per smoker (⩾ age 35). The share of the economic costs was greater for men than women, and greater in rural areas than in urban areas. Of the $5.0 billion total costs, direct costs were $1.7 billion (34% of the total), indirect morbidity costs were $0.4 billion (8%), and indirect mortality costs were $2.9 billion (58%). The direct costs of smoking accounted for 3.1% of China’s national health expenditures in 2000.
Conclusion: The adverse health effects of smoking constitute a huge economic burden to the Chinese society. To reduce this burden in the future, effective tobacco control programmes and sustained efforts are needed to curb the tobacco epidemic and economic losses.
Statistics from Altmetric.com
- GDP, gross domestic product
- ICD-9, International classification of diseases, 9th revision
- NHSS, National Health Services Survey
- PVLE, present value of lifetime earnings
- RR, relative risk
- SAF, smoking attributable fraction
- SAI, smoking attributable indirect morbidity cost
- THE, total health care expenditure
- TIE, total indirect morbidity expenditure
- YPLL, years of potential life lost
China is the world’s largest tobacco consumer, accounting for more than one fourth of the world’s tobacco consumption annually. In 2002, about 350 million adults aged 15 and older were ever smokers and 300 million were current smokers.1 China is considered in an early stage of a tobacco epidemic. The ever-smoking rates in 2002 were 66.0% for men and 3.1% for women of ages 15 years and older, slightly lower than the rates in 1996 of 66.9% for men and 4.2% for women. The prevalence of smoking in Chinese men seems to have levelled off but not yet dropped. The average cigarette consumption per smoker was the same in 2002 as that in 1996—14.8 cigarettes per day.
Cigarette smoking is the leading cause of preventable deaths in the majority of high-income nations and increasingly in low- and middle-income nations,2 and it also causes disability and productivity losses because of premature deaths.3 One way to document and assess the adverse health effects of smoking on a society is to translate smoking-caused illnesses, premature mortality, and productivity losses into economic terms. In other words, the cost of smoking serves as a universal marker for measuring the adverse effects of smoking.
Only two studies have estimated the economic costs of smoking in China. Chen et al,4 using a medical cost accounting method to calculate the disease-specific medical costs, estimated the total medical costs attributable to smoking in 1988 as 2.3 billion Yuan (US$0.3 billion using an exchange rate of 8.2784 Yuan to US$1). Using a prevalence-based method, Jin et al5 estimated the total economic burden of smoking in 1989 at 27.1 billion Yuan (US$3.3 billion), including 6.9 billion Yuan (US$0.8 billion) in direct medical costs and 20.1 billion Yuan (US$2.4 billion) in indirect morbidity and mortality costs. During the decade since these two studies, medical expenditures have risen dramatically, and smoking prevalence rates have also increased. To improve public health and reduce resources spent for smoking-caused illness, we must monitor the progression of the economic burden caused by cigarette smoking and update the estimates for the costs of smoking.
The objective of this study is to assess the economic burden of diseases and deaths attributable to smoking on Chinese society in 2000. The results may be used to justify the need for stronger tobacco control policies in China and guide health policy and planning.
The economic burden of smoking comprises three components: direct medical costs of treating smoking-related diseases, indirect morbidity costs of smoking, and indirect mortality costs of premature deaths caused by smoking-related diseases. We used the prevalence-based, disease-specific approach6–9 to measure the costs of smoking-related diseases and deaths in 2000 caused by current and past smoking. Three major categories of smoking-related diseases were considered: cancer (all types of malignant neoplasm (International classification of diseases, 9th revision (ICD-9) codes: 140–208)), cardiovascular diseases (stroke, ischaemic heart disease, rheumatic heart disease, etc (ICD-9 codes: 390–414, 418–459)), and respiratory diseases (chronic obstructive pulmonary disease, respiratory tuberculosis, pulmonary heart disease, etc (ICD-9 codes: 011–012, 018, 415–417, 460–519)). Because the effects of smoking will appear long after its onset, and because most smokers begin before age 20, we included only persons aged 35 and older in the analysis.
Estimation of smoking attributable fraction
For each of the three components of the economic burden of smoking, we calculated a smoking attributable fraction (SAF), also known as population attributable risk, by disease category, rural/urban district, gender, and age. The estimated SAF is then multiplied by each cost measure of interest to derive the smoking attributable cost.10 For example, the product of SAF and total inpatient hospitalisation expenditures is smoking attributable hospitalisation cost; the product of SAF and total morbidity cost is smoking attributable morbidity cost. The SAF is specified by the following epidemiological formula11:
where PN and PS denote the prevalence rate of non-smokers and smokers, respectively; RR denotes the relative risk of mortality for smokers compared to non-smokers; the subscript i is for disease category; j is for rural/urban district; g is for gender; a is for age classified into two groups, 35–64 and 65+. When RR is < 1, the corresponding SAF becomes negative, implying that smokers have a decreased risk of mortality compared with non-smokers. This implication does not seem realistic, given the huge body of evidence linking smoking with multiple life-threatening diseases. Instead of employing negative SAFs, we assume they are zero.
We estimated SAFs using the RRs of mortality and then applied the same SAF estimates to the total measure for all three components of the economic costs.
The primary data source was the Second National Health Services Survey (NHSS) conducted by the Ministry of Health in China in 1998.12 The NHSS is a nationally representative sample of 216 101 individuals from 56 994 households extracted through a multi-stage random sampling method. Face-to-face interviews were conducted to collect detailed information about households and all household members. Such information includes age, gender, education, employment status, disease histories, health risk behaviours, and medical care utilisation and expenditures including cause-specific outpatient visits and self-medication in the two weeks before the date of interview, as well as inpatient hospitalisations in the past 12 months before the date of interview. Self-reported days lost from work due to illness in the past two weeks before interview were also collected.
Smoking prevalence rates by rural/urban district, gender, and age came from the 1998 NHSS in which all the individuals of age 15 and older were asked their smoking status. We classified adults as non-smokers (never smoked) and smokers (current and former). We did not separate current smokers from former smokers because the RRs of mortality, needed in the calculation of SAFs, have not been estimated separately for former and current smokers in China.
The RRs of mortality for smokers were obtained from published estimates by Liu et al,13 population data from the 2000 Population Census,14 2000 mortality rates from China’s Vital Registration system monitored by the Ministry of Health,15 life expectancy from the 2000 China Life Tables reported by the World Health Organization,16 and earnings data from the China Statistics Yearbook.17
Direct costs include all the health care expenditures for treating smoking-related diseases. China has three types of health care services: inpatient hospitalisations, outpatient visits, and self-medications. Hospitalisation and outpatient costs include expenditures for prescribed medicine during the hospital stays or outpatient visits. Self-medications refer to medicine purchased by patients from a drug store without seeing a doctor.
Smoking attributable health care expenditure (SAE) for each subgroup stratified by disease category, urban/rural district, gender, and age (35–64, and 65+) was estimated by multiplying the SAF by the corresponding total health care expenditure (THE) according to the following formula:
where PH is the average expenditure per inpatient hospitalisation; QH is the average number of hospitalisations per person in 12 months; PV is the average expenditure per outpatient visit; QV is the average number of outpatient visits per person in two weeks; PM is the average medication expenditure per person with positive self-medication expenditures in two weeks; QM is the proportion of persons with positive self-medication expenditures in two weeks; and POP is the 2000 population. The notation for the subscripts i, j, g, and a is the same as in equation 1.
The average expenditures and health care utilisation rates were estimated from the 1998 NHSS data. In the NHSS questionnaire, respondents were asked to report the expenditures of inpatient hospitalisations and the amount paid by health insurance and by their own. Because of the logic of the questionnaire structure, we consider that the self-reported hospitalisation expenditures represent actual expenditures including payment from insurance and out-of-pocket. For self-medications, the total expenditures were not likely paid by insurance since they were purchased without doctor’s prescription. For two-week outpatient visits, respondents were asked to report the total expenditures without being probed for the out-of-pocket payment or health insurance payment.
To warrant that the self-reported health expenditures are a good proxy for actual expenditures, we undertook an adjustment process. First, we estimated the average per capita health expenditure for each disease category by urban/rural district, gender and age from the 1998 NHSS data. Second, we multiplied this number by the population in 2000 for each subgroup and added up all the expenditures across all subgroups to derive the estimated national health expenditure in 2000. Our estimate is $40.2 billion, which is lower than the published figure at $55.4 billion.18 Finally, we calculated an adjustment factor of 1.38 (55.4/40.2) and applied it to the estimated average expenditures from the NHHS data.
Indirect morbidity costs
Indirect morbidity costs include expenditures for transportation and caregivers due to inpatient hospitalisations and outpatient visits as well as the value of lost productivity caused by smoking-related diseases.
Smoking attributable indirect morbidity cost (SAI) for each subgroup stratified by disease category, urban/rural district, gender, and age (35–64, and 65+) was estimated by multiplying the SAF by the corresponding total indirect morbidity expenditure (TIE) using the following formula:
where PHI is the average expenditure on transportation and caregivers per inpatient hospitalisation; PVI is the average expenditure on transportation per outpatient visit; PW is the average number of days lost from work due to smoking-related illness per employed person in two weeks; QW is the employment-population ratio which is defined as the proportion of the total population who are currently employed; and Y is daily earnings in 2000.
The expenditures for transportation and caregivers and the days lost from work were from the 1998 NHSS data. Since average earnings data were not available by age and gender, we evaluated the work-loss days using two sets of 2000 earnings data: average annual wage of the staff and workers for urban residents, and per capita annual net income of rural households multiplied by the average number of rural residents supported by the rural working population.17 Daily earnings were derived by dividing the annual earnings by 366 (number of days in 2000).
Indirect mortality costs
We used the human capital approach19 to measure the expected value of lost future productivity caused by smoking attributable premature deaths in 2000. Smoking attributable deaths for each subgroup stratified by disease category, urban/rural district, gender, and age (35–64, and 65+) were estimated by multiplying the SAFs by the corresponding total number of deaths. Smoking attributable years of potential life lost (YPLLs) were estimated by the product of smoking attributable deaths and the average number of years of life expectancy remaining at the age of death. Life expectancy was obtained from gender and five-year age-specific life tables for China in 2000.16
Smoking attributable costs of premature deaths were estimated by multiplying smoking attributable deaths by the present value of forgone lifetime earnings (PVLE) for each person at the age of death. The PVLE is estimated by rural/urban district, gender, and five-year age groups based on the approach developed by Max et al20 which takes into account life expectancy for different gender and age subgroups, varying employment-population rates by rural/urban district, gender, and five-year age groups, and a discount rate of 3% to convert a stream of earnings into its current worth. To consider potential growth on future earnings, we assumed an annual productivity growth rate of 8%, approximately the average growth rate of gross domestic product (GDP) in China between 1998 and 2002. The employment-population rate was estimated from the 1998 NHSS data by dividing the number of currently employed persons by the total population. Definitions and calculation of earnings were the same as described above.
China had 1.243 billion people in 2000—427 million aged 35–64 years and 88 million aged 65+.14 Among the 515 million people aged 35+, 323 million (63%) lived in rural areas and 192 million (37%) lived in urban areas. The prevalence of cigarette smoking in China is high. According to the 1998 NHSS data, 38% of the adults aged 35+ were ever smokers including 34.5% as current smokers and 3.5% as former smokers. The ever-smoking rates among men in the urban areas were 75.3% for ages 35–64 and 61.0% for ages 65+, lower than the corresponding rates in the rural areas of 84.8% and 73.4%, respectively. The smoking rates among women in urban areas were higher than those in rural areas for both age groups (6.1% and 14.8% versus 5.3% and 10.3%). The smoking rates for men were higher among ages 35–64 than among the older group; the opposite was found for women.
Based on our estimates of the smoking prevalence rates and the RRs estimated by Liu et al,13 SAFs were calculated (table 1). The SAFs are substantially smaller for women than men, resulting from the much lower smoking prevalence among women. Cancer showed the highest disease-specific SAFs for men, while for women, respiratory diseases showed the highest SAFs. Note that the SAFs for rural female smokers were zero for cardiovascular diseases because the estimated RRs were less than one.13
Table 2 shows the SAEs of inpatient hospitalisations, outpatient visits, and self-medications by disease category, rural/urban district, gender, and age. The total SAE amounted to $1705.7 million. Among the three disease categories, treating respiratory diseases cost the most for both urban smokers ($604.1 million) and rural smokers ($274.0 million)—$878.2 million, or 51% of the total SAE. The results from disaggregated analyses by types of health care services show similar patterns except that for inpatient hospitalisation, treating cancer cost the most for ages 35–64 for rural men and urban women and treating cardiovascular diseases cost the most for urban men aged 35–64, and that for self-medication, treating cardiovascular diseases cost the most for urban men aged 65+. The SAEs for treating cancer and cardiovascular diseases were $300.2 million (18%) and $527.3 million (31%), respectively. In general, the middle age group incurred higher SAEs than the older group.
China had 688 512 deaths attributed to smoking in 2000—599 723 for men and 88 789 for women; 248 808 in urban areas and 439 704 in rural areas—accounting for 8.3% of all deaths. Table 3 shows that the YPLLs for these smoking attributable deaths were estimated as 9.7 million years—8.6 million years for men and 1.1 million years for women; 3.4 million years for urban smokers and 6.3 million years for rural smokers. The disaggregated analyses by underlying causes of death show that older women had higher total smoking attributable YPLLs than the middle age women except for the deaths caused by cancer in rural areas. The opposite is true for men in that the middle age group had higher smoking attributable YPLLs than the older group except for the deaths caused by respiratory diseases in both urban and rural areas. The average years of life lost per death was 14.1 years—14.4 years for men and 12.2 years for women; 13.9 for urban smokers and 14.2 for rural smokers.
The estimated PVLEs for an average Chinese in 2000 by rural/urban district, gender, and age are shown in the Appendix. Table 4 shows the value of lost productivity caused by smoking-related deaths in 2000: $2935.5 million—$2816.3 million for men and $119.2 million for women; $790.4 million in urban areas and $2145.1 million in rural areas. The value of lost productivity averaged $4264 per death—$4696 for men and $1342 for women; $3177 for urban smokers and $4878 for rural smokers. The leading cause of smoking attributable mortality costs is cancer at $1903.1 million (65%), followed by respiratory diseases at $534.4 million (18%), and cardiovascular diseases at $498.0 million (17%).
Table 5 shows that the smoking attributable transportation and caregivers costs were $122.5 million, much smaller than the direct medical costs of smoking. Productivity losses due to absence from work caused by smoking-related illness amounted to $270.6 million. Total indirect morbidity costs of smoking in 2000 amounted to $393.0 million—$344.5 million for men and $48.5 million for women; $175.8 million in urban areas and $217.2 million in rural areas.
Table 5 also summarises the estimates for all the components of the economic burden of smoking. In 2000, the total economic cost of smoking in China was $5034.2 million. A larger share of the total economic cost was borne by men than women ($4609.2 million versus $425.0 million), and in rural areas than in urban areas ($2842.1 million versus $2192.1 million). Total annual economic losses averaged $25.43 per smoker among the 198 million smokers aged 35 and older in 2000—$31.46 for urban smokers and $22.16 for rural smokers. Given that China’s cigarette consumption in 2000 was 30.76 million cases (that is, 76.9 billion packs),21 the economic cost of smoking averaged $0.07 per pack.
Lost productivity caused by premature deaths accounted for the largest proportion of the total economic costs at $2935.5 million (58% of the total), followed by the direct costs at $1705.7 million (34%), and then indirect morbidity costs at $393.0 million (8%). Smoking attributable direct costs accounted for 3.1% of China’s national health care expenditures in 2000 at $55.4 billion.18 Outpatient visits accounted for the highest proportion of smoking attributable direct costs at $898.9 (53%), followed by inpatient hospitalisations at $654.4 (38%), and self-medication at $152.3 (9%).
The results of this study indicate that the adverse health effects of smoking cause a huge economic burden to the Chinese society, reaching $5034.2 million in 2000. Results also show that $1705.7 million, 3.1% of the national health care expenditures, were spent on treating smoking-related diseases.
Compared to Western countries such as the USA where smoking attributable direct costs accounted for 6–8% of national personal health care expenditures,9 our estimates seem low. Our estimates are likely conservative for several reasons.
First, we used the RRs from a retrospective mortality study of one million deaths by Liu et al13 and their estimates were much lower than those from Western countries. For example, the RRs of lung cancer among men aged 35 and older were 2.98 for urban smokers and 2.57 for rural smokers in China compared to 22.36 for current smokers and 9.36 for former smokers in the USA.22 The RRs of coronary heart disease among men aged 35–64 were 1.28 for both urban and rural smokers in China versus 2.81 for current smokers and 1.75 for former smokers in the USA.22
Few studies have assessed the RRs of smoking for the Chinese population.13,23–26 We chose the estimates by Liu et al13 because their study was based on a large national sample and was the only study providing the RR estimates by gender and rural/urban district. In general, the RRs estimated by Liu et al were comparable to those from other Chinese prospective studies and case–control studies.13 However, Lam et al26 have argued that most studies on smoking-related mortality in China tend to underestimate the RRs of smoking because of the short follow-up. They demonstrated that as the duration of follow-up period increased, RRs also increased. Based on 20 years of follow-up, their estimated RRs were similar to those in the USA and the UK. We conducted a sensitivity analysis by using the RRs estimated by Lam et al26 while keeping other data unchanged. The total economic costs of smoking more than doubled, from $5034 million to $12 151 million, and the proportion of smoking attributable direct costs to the national health expenditures increased to 7.7%, comparable to that in the USA.
Second, when estimating PVLE, we assumed earnings were the same for all subgroups by age and gender because of the data limitation. As shown in the Appendix, the PVLE is the highest for persons < 1 year and decreases with age. Max et al20 found that the American males and females aged 20–24 have the highest PVLE at a discount rate of 3% given the varying earnings for each five-year age group. If the pattern of earnings by age in China is similar to that in the USA, our PVLE estimates may be overestimated for persons younger than age 20 and underestimated for persons ⩾ 20; therefore, our estimated indirect costs of smoking for persons aged 35 and older may be underestimated. Also, we assumed 8% as the annual growth rate in productivity that is lower than the average growth rate of GDP in China during 1979–2002 at 9.4%.17 According to China’s National Bureau of Statistics, the annual rates of GDP growth in 2003 and 2004 reached 10%. The estimated costs of smoking would be higher if the growth rate of 9.4% is assumed. However, if we assume the growth rate in China will be reduced to 5% beginning in 2006, the indirect mortality costs of smoking would be reduced from $2.9 billion to $2.2 billion.
Third, we estimated the smoking prevalence rates by using the 1998 NHSS data. Yang et al27 estimated that the prevalence rate of ever-smokers for persons aged 15–69 was 66.9% for men and 4.2% for women using the 1996 National Prevalence Survey. Applying the same age criteria, the corresponding rate was 57.5% for men and 3.8% for women from the 1998 NHSS data. In other words, our estimate was about 15% lower for men and 10% lower for women than the estimates by Yang et al. This could be due to an underreporting of smoking status among respondents in the 1998 NHSS. We conducted another sensitivity analysis by increasing the smoking prevalence rate by 15% for each age group for men while keeping other things unchanged. The total economic costs of smoking increased by about 10%, from $5034 million to $5569 million.
Fourth, our calculation for the economic burden of smoking only considers the three major categories of smoking-related diseases. We did not include peptic ulcer diseases and liver cirrhosis, two diseases estimated to account for 14–19% of smoking attributable medical costs, YPLLs, and mortality costs according to two recent studies in Asia.28,29 Additionally, Yang et al29 found that kidney diseases caused the most smoking attributable medical costs. Further research is needed to examine the effects of smoking on the morbidity and mortality resulting from peptic ulcer, liver cirrhosis, and kidney diseases and their economic costs among the Chinese population.
Finally, we did not count the days lost from work by the relatives taking care of patients with smoking-related illness, the health care costs of environmental tobacco smoke, and the deaths resulting from cigarette-caused fires and passive smoking. For persons younger than age 35, passive smoking is the main risk for diseases such as bronchitis and low birth weight. Yang et al1 estimated that more than 50% of Chinese non-smokers aged 15 and older were passive smokers in 2002. Therefore, the true economic burden of smoking in China could be much higher than the estimates in this study, which only focuses on the evaluation of active smoking.
This study is based on the conventional epidemiological approach6,7,29 which estimates the smoking attributable risks of mortality caused by the three major smoking-related diseases, and applies the attributable risks of mortality to the morbidity toll. In an alternative epidemiological approach,8,28 the smoking attributable risks of healthcare utilisation for smoking-related diseases are estimated and applied to healthcare expenditures. Several recent studies in the USA have developed the econometric approach10,30–32 which examines the relationship between smoking and healthcare expenditures in a multi-equation framework by adjusting other risk factors, health status, and individuals’ taste for consuming health care. These econometric studies estimate smokers’ total excess expenditures over those of non-smokers by including all types of diseases. Warner et al9 anticipate that the econometric approach will become the norm in cost-of-smoking analyses. Due to the limitation of our data, neither the smoking attributable risks of healthcare utilisation nor the econometric models are estimated in this study.
In conclusion, our very conservatively estimated results suggest that cigarette smoking costs an enormous economic burden in China through a huge number of preventable diseases, health care costs, premature deaths, and productivity losses. The total economic costs had grown from $3.3 billion in 19895 to $5.0 billion in 2000. Since the health effects of smoking on morbidity and mortality are cumulative, China will bear a much heavier economic burden from cigarette smoking in the future if the current trends in smoking behaviour continue. The economic burden of smoking in the rural areas will be of special concern because of the large proportion of population, the high smoking prevalence rate, and poverty. To reduce the economic burden of smoking to the society, effective tobacco control programmes and sustained efforts are needed in China to curb the tobacco epidemic. We hope that the results of this study will motivate policymakers in China to formulate a comprehensive health and economic agenda with adequate investment in research, public education, health providers’ training, and intervention programmes aimed at tobacco prevention and control.
What this paper adds
This paper provides an updated and comprehensive estimate of the economic costs of smoking in China. Our findings justify the need for stronger tobacco control policies in China.
This study was supported by a grant from the Fogarty International Center (grant R01-TW05938) National Institutes of Health. The authors are grateful for comments provided by Professors Ming-Chin Yang, Gregor A Franz, and Dorothy P Rice. However, the authors alone are responsible for the findings.
Competing interests: None
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.