Economic burden of COVID-19, China, January–March, 2020: a cost-of-illness study

Abstract Objective To estimate the economic cost of coronavirus disease 19 (COVID-19) in 31 provincial-level administrative regions and in total, in China. Methods We used data from government reports, clinical guidelines and other publications to estimate the main cost components of COVID-19 during 1 January–31 March 2020. These components were: identification and diagnosis of close contacts; suspected cases and confirmed cases of COVID-19; treatment of COVID-19 cases; compulsory quarantine of close contacts and suspected cases; and productivity losses for all affected residents. Primary outcomes were total health-care and societal costs. Findings The total estimated health-care and societal costs associated with COVID-19 were 4.26 billion Chinese yuan (¥; 0.62 billion United States dollars, US$) and ¥ 2646.70 billion (US$ 383.02 billion), respectively. Inpatient care accounted for 44.2% (¥ 0.95 billion/¥ 2.15 billion) of routine health-care costs followed by medicines, accounting for 32.5% (¥ 0.70 billion/¥ 2.15 billion). Productivity losses accounted for 99.8% (¥ 2641.61 billion/¥ 2646.70 billion) of societal costs, which were mostly attributable to the effect of movement-restriction policies on people who did not have COVID-19. Societal costs were most sensitive to salary costs and number of working days lost due to movement-restriction policies. Hubei province had the highest health-care cost while Guangdong province had the highest societal cost. Conclusion Our results highlight the high economic burden of the COVID-19 outbreak in China. The control measures to prevent the spread of disease resulted in substantial costs from productivity losses amounting to 2.7% (US$ 382.29 billion/US$ 14.14 trillion) of China’s annual gross domestic product.


Introduction
Coronavirus disease 2019 (COVID- 19) is an infectious disease which results in substantial morbidity and mortality in some population groups. By September 2020, over 32.7 million cases of COVID-19 had been confirmed worldwide, of which 90 966 were in China. 1 Prevention and treatment of COVID-19 can be expensive. According to Chinese clinical guidelines, 2,3 all confirmed cases of COVID-19 should receive inpatient care. Moreover, patients with critical COVID-19 often require costly treatment such as mechanical ventilation and extracorporeal membrane oxygenation, potentially substantially increasing health-care costs. The societal cost of COVID-19 could be even greater. To prevent disease transmission, a series of emergency measures were implemented by the Chinese government, 4 including isolation of COVID-19 cases, 14-day quarantine for close contacts of COVID-19 cases, lockdown of Wuhan city and adjacent areas, travel restrictions and extension of the Chinese New Year holiday period. While these containment strategies successfully reduced the transmission of COVID-19, 5 they inevitably caused a considerable loss in productivity.
This study assessed the health and societal costs of the COVID-19 outbreak in 31 provincial-level administrative regions in mainland China.

Methods
We conducted and reported our study according to the costof-illness checklist. 6

Study population
The population of interest was all residents in mainland China, which has 31 provincial-level administrative regions -22 provinces, five autonomous regions (Guangxi Zhuang, Inner Mongolia, Ningxia Hui, Tibet and Xinjiang Uyghur) and four municipalities (Beijing, Chongqing, Shanghai and Tianjin). We divided the population into four mutually exclusive patient subgroups, based on their experience of COVID-19: (i) asymptomatic close contacts of suspected or confirmed cases of COVID-19, who were eventually diagnosed as COVID-19 negative; (ii) symptomatic suspected cases with or without close contact history with existing suspected or confirmed cases, who were eventually diagnosed as COVID-19 negative; (iii) confirmed cases of COVID-19, including those previously assessed as close contacts or suspected cases; and (iv) people not considered to have been exposed to COVID-19. We further divided confirmed cases into non-severe, severe and critical COVID-19, according to the disease severity (Box 1). Fig. 1 shows the diagnostic and treatment pathway for each patient subgroup; also described in the data repository. 8

Outcomes
We estimated direct health-care costs, direct non-health-care cost and productivity losses for each region and for mainland China as a whole (Box 2). We calculated all costs in Chinese yuan (¥) at the 2019 value and converted to United States dollars (US$) using the annual exchange rate for 2019: US$ 1.00 = ¥ 6.91. 9

Study period
Although COVID-19 was first identified in China in December 2019, 99.96% (74 648/74 675) of confirmed cases were identified in January and February 2020. 10 From 6 March 2020, the number of new cases a day fell below 100, and no new cases were identified in 29 regions. Therefore, we calculated costs for the period from 1 January to 31 March 2020.

Cost estimation
There are two approaches to estimate the cost of illness: the bottom-up approach and the top-down approach. 6 The bottom-up approach multiplies the average cost of the illness per patient by the prevalence of the illness. The topdown approach uses aggregated data and a population-attributable fraction to assign a percentage of total expenditure to the disease of interest. Because published total expenditure on COVID-19 was lacking (details in the data repository), 8 we used the bottom-up approach. We estimated unit costs, p x , at the patient or individual level for each component, x, of the overall burden of disease. We calculated the overall cost, C, as: where i x is the number of individuals affected.

Epidemiological data
Over the period of interest, the National Health Commission of the People's Republic of China published national data on COVID-19 daily. 11 However, detailed regional information was only published for Hubei province. Therefore, we manually extracted the number of newly identified close contacts, suspected cases and confirmed cases in each region from the daily updates reported by the local health commission of each region (details in the data repository). 8 While all regions reported complete data for the number of confirmed cases and the numbers of deaths of confirmed cases, data were incomplete for the number of close contacts and/or suspected cases. We estimated these missing data either from published reports, or from the reported regional number of confirmed cases, 11 assuming the same ratio between the number of close contacts or suspected cases and confirmed cases across regions.

Direct health-care cost
We used information in the published literature 12,13 and clinical guidelines, 2,3 supplemented with expert opinion where necessary, to estimate the healthcare resources used for close contacts, suspected cases and confirmed cases. Shanghai is one of the few regions in China which reports full unit cost data. 14 To calculate the unit costs for other regions, we calculated a health-care industry salary index (details in the data repository). 8 We calculated a weight (w r ) for each region as: where s r is the ratio of the average health-care industry salary in the region and s s is the average health-care industry salary in Shanghai. 15 We then estimated regional unit costs (p r ) as: where p s is the unit costs derived from Shanghai. 14 According to the State Council, 42 600 front-line health professionals worked with suspected and/or con-Box 1. Definition of close contacts, suspected cases and confirmed cases of COVID-19, China, 2020

Close contact
An asymptomatic person who has had close (less than 1 m), unprotected (without personal protective equipment) contact with suspected cases or confirmed cases (see definitions below), 2 or fewer days before the onset of their symptoms.

Suspected case
A person who has one epidemiological history criteria and meets two clinical symptoms criteria, or has no epidemiological history but meets all three clinical symptoms criteria.
• Epidemiological history. Fourteen days before the onset of the disease, the person has: (i) travelled to or lived in a high-risk region or country; or (ii) had direct contact with confirmed cases (definition below); or (iii) had direct contact with someone with a fever or respiratory symptoms in a high-risk region or country; or (iv) been to a place with disease clustering -defined as two or more cases with fever and/or respiratory symptoms occurring at places such as homes, offices and school classrooms.
• Clinical symptoms. The person has: (i) a fever and/or respiratory symptoms; (ii) the following imaging features of COVID-19 after computerized tomography of their chest -multiple patchy shadows and interstitial changes, particularly at the periphery of the lungs, multiple groundglass opacities and infiltrates in both lungs, or in severe cases, lung consolidation and pleural effusion; (iii) normal or decreased white blood cell count in the early stage of the disease, or normal or decreased lymphocyte count over time.

Confirmed case
A suspected case that meets one of the following criteria: (i) positive result of the nucleic acid test for SARS-CoV-2; (ii) DNA sequencing results indicating high sequence similarity to known SARS-CoV-2 sequences; (iii) positive result for the serum-specific antibodies (IgM and IgG) of COVID-19. Severity of disease in confirmed cases is categorized as follows.
• Non-severe cases: mild cases (mild clinical symptoms with no signs of pneumonia on imaging) and moderate cases (symptoms such as fever and respiratory tract symptoms, and signs of pneumonia on imaging  16 The daily risk subsidy for front-line health professionals was estimated to be ¥ 300.00 per person. 17 We estimated the emergency funds (for construction of temporary emergency buildings and non-routine procurement of additional medical supplies and equipment) based on the budget plans of the Ministry of Finance and the National Development and Reform Commission (data repository). 8 For reusable equipment, we only included the cost attributable to the 3-month period of the study in our analysis. Calculations and results for emergency funds are in the data repository. 8

Direct non-health-care cost
We estimated a daily cost of quarantine in Shanghai to be ¥ 75.00 (US$ 10.85), assuming that 50.0% of people quarantined at home at zero cost and 50.0% quarantined at a designated centre at the cost of ¥ 150.00 (US$ 21.71) a day. We calculated the regional quarantine costs per person (QC re ) per person by category of exposure (e), as: where w r is the regional weight described earlier and d e is the estimated duration of quarantine.
The average cost of quarantine for close contacts and suspected cases was ¥ 1246.00 (US$ 180.32) and ¥ 735.00 (US$ 106.37) per person, respectively. We calculated the overall cost of quarantine (TQC) as: TQC n r e re = × ∑ ∑ (5) where n re is the number of people quarantined by region (r) and exposure (e) category. Details on methods and results are in the data repository. 8

Loss in productivity
We used the human capital approach to estimate productivity losses. For people not considered to have been exposed to COVID-19, we calculated costs by region (CP r ) as: where i r is the mean daily wage rate by region, f is the proportion of the population in employment, h r is the mean number of days lost by region, and q r is the regional population. We obtained regional employment statistics from the China Statistical Yearbook 2019. 15 The national average daily wage was ¥ 271.94 (US$ 39.35), ranging from ¥ 204.67 (US$ 29.62) in Heilongjiang province to ¥ 486.43 (US$ 70.40) in Beijing. The national unemployment rate was 3.0%, ranging from 1.4% in Beijing to 4.0% in Heilongjiang province. Data were not available on the employment status for close contacts, suspected cases and confirmed cases. Therefore, we estimated the employment rate, f, for each patient subgroup at 54.0% based on the age and sex distribution of confirmed cases, the legal working age (16 years) and official retirement age (60 years for men and 50 years for women), and the national unemployment rate (3.0%). Employment rate calculations are in the data repository. 8 We estimated the average number of working days lost due to restrictions on movement for people not considered to have contracted COVID-19 as 23.26 days, based on the Baidu migration index, 18,19 which tracks the proportion of workers returning from their hometowns to work after the Chinese New Year holiday. Close contacts, suspected cases and confirmed cases may have experienced more working days lost due to their quarantine and/or hospitalization. 12,13,20 Working days lost for these people depended on the start and end date of their quarantine and/or hospitalization, and whether these dates overlapped with the extended Chinese New Year holiday and the study period. We limited productivity losses from COVID-19 deaths to the study period in the base case analysis. Calculations of the working days lost for each patient subgroup are in the data repository. 8

Sensitivity analysis
To determine which parameters were key cost drivers we conducted a sensitivity analysis. We identified costs that contributed to 10.0% or more of the total health-care costs and societal costs and varied the parameters for use of resources and unit cost. We used available data or our judgement to inform the ranges for the selected parameters.

Results
During the study period, there were 707 913 close contacts, 126 032 suspected cases and 81 879 confirmed cases in mainland China (Table 1) Table 2 shows the health-care cost per person for each patient subgroup, based on the estimated use of resources and the unit costs from Shanghai. 15 The health-care cost of managing close contacts and suspected cases diagnosed

Box 2. Components of the cost categories used in the COVID-19 costing study, China, 2020 Direct health-care costs
Routine health care: identification, diagnosis, treatment and follow-up of people with suspected or confirmed COVID-19. Non-routine health care: (i) risk subsidy for front-line health professionals who work with suspected and/or confirmed cases; and (ii) emergency funds for construction of temporary emergency buildings (i.e. Huoshenshan and Leishenshan hospitals, and Wuhan mobile cabin hospital), and non-routine procurement of additional medical supplies and equipment (e.g. personal protective equipment).

Direct non-health-care costs
Compulsory quarantine for close contacts and suspected cases. The quarantine cost can be covered by the local government, or by the quarantined individual, or jointly, depending on local policies.

Productivity losses
These losses include: (i) employed close contacts, suspected cases or confirmed cases who lost work time due to their quarantine and/or inpatient care; and (ii) any employed individuals who lost work time due to government policies controlling population movement (these individuals include people not considered to have had COVID-19).  8 We calculated costs of routine health-care services, quarantine and productivity losses, and total healthcare and societal costs (   8 For the confirmed cases, we assumed that 45% of them were identified from close contacts, whilst 55% were identified from suspected cases. Therefore, the cost of identification and diagnosis for all confirmed cases was calculated as the multiplication of the cost per case and 0.45 for close contacts and 0.55 for suspected cases.  Fig. 2 and Fig. 3 show the health-care cost and societal cost for each region, respectively. The health-care cost for Hubei province alone accounted for 66.7% (¥ 2.84 billion/¥ 4.26 billion) of the national health-care cost (Fig. 2). Guangdong province incurred the highest societal cost, followed by Jiangsu province and Beijing (Fig. 3).
The results of the sensitivity analyses are reported in the data repository. 8 The direct health-care cost was most sensitive to the proportion of confirmed cases with severe or critical disease, and the health-care cost per person for treating severe and critical cases. The cost of the loss in productivity was most sensitive to the number of working days lost for people not considered to have had COVID-19, the national average daily salary, and assumptions on the effect of movement restriction policies on worker productivity.

Discussion
We estimated the health-care and societal costs associated with the COVID-19 outbreak in China for the first 3 months of 2020 to be ¥ 4.26 billion (US$ 0.62 billion) and ¥ 2646.70 billion (US$ 383.03 billion), respectively. Although the health-care cost per person for con- Includes risk subsidy for health-care staff and emergency funds for construction of temporary emergency buildings and non-routine procurement of additional medical supplies and equipment. c Column total. We could not assign the cost of non-routine health care to any specific individual patient group so we only report the total cost of non-routine health care in the last column; therefore, the cost of non-routine health care is not reflected in the total societal cost for each patient subgroup (last row).
firmed cases was high, 99.9% of the societal cost was attributable to productivity losses in people not considered to have had COVID-19. These findings reflect the overall number of employed people in China (416.5 million), which is much larger than the number of confirmed cases (81 879 cases). Our estimated cost of productivity losses -¥ 2641.61 billion (US$ 382.29 billion) -is comparable to the decrease in the Chinese GDP for the first quarter of 2020 compared with the same period in 2019: ¥ 1506.68 billion (US$ 218.04 billion). 22 Hubei province, where most confirmed cases were identified, accounted for two thirds of the national health-care cost. The productivity loss was greatest for those regions with the highest number of employed people and/or the highest daily salary, such as Guangdong province (57.7 million employed people, ¥ 296.37, US$ 42.89, daily salary), Jiangsu province (42. We did not identify any cost-ofillness studies for COVID-19 in our rapid review of the literature. Evidence on cost of illness is available for severe acute respiratory syndrome (SARS). [23][24][25][26][27] To facilitate comparison of results, we inflated costs from the literature to 2019 values using a local consumer price index and converted to US$ using the annual exchange rate. 28 Three studies [23][24][25] reported the cost of managing patients with SARS; the health-care cost per case ranged from US$ 4151.00 in mainland China 24 to US$ 362 700.00 in Canada. 23   29 Another study used a simulation model to estimate the societal cost of SARS in 30 countries. 30 The cost in mainland China was 1.03% (¥ 0.12 trillion/¥ 11.69 trillion) of GDP, 30 which is comparable to our estimate of the societal cost of COVID-19 (2.7% of China's GDP in 2019). 22 The societal cost of COVID-19 is substantial and greatly outweighs the health-care cost. Our analysis, which demonstrates the effect of COVID-19 beyond the health-care system, justifies the redirection of resources from other sectors of the economy to strengthen health systems, as the potential productivity losses caused by a pandemic may far exceed the health-care cost. Despite a lack of evidence on their cost-effectiveness, unprecedented controls on people's movements and ability to work have been imposed in several countries in an attempt to reduce the spread of COVID-19. Future work will examine the cost-effectiveness of these policies. Our data can help inform these analyses by providing the cost of identifying, diagnosing and treating patients with suspected or confirmed COVID-19. Our analysis underlines the importance of action to strengthen health systems, particularly the capacity to test for infection and trace contacts, which has been identified as one of the most cost-effective policy responses. 31 Effective disease mitigation action will require international cooperation and considerable investment. Underinvestment in strengthening the capacity of health systems to tackle future pandemics could prove to be far costlier than the additional investment required.
Our study has several strengths. This study fills an important evidence gap by presenting the first cost-of-illness study of COVID-19. The study identified the cost of the COVID-19 pandemic in different sectors of the economy; such data are necessary to inform planning of services and the prioritization of research. Our data also provide important information for future economic evaluations of interventions for CO-VID-19. We accessed detailed data on use of resources in the 31 regions of mainland China, including incidence of close contacts, suspected cases and confirmed cases, from the local health commission of each region. We applied unit cost data adjusted to reflect relative price differences across provinces, and used clinician input from Shanghai and Hubei province to check the use of resources for each subgroup (close contacts, suspected and confirmed cases). We estimated productivity costs for close contacts, suspected cases and confirmed cases based on the duration of quarantine and/or treatment, and regional migration patterns after the end of the extended Chinese New Year holiday period.
Our analysis also has some limitations. First, we only covered the first 3 months of the epidemic and there-fore could not capture the long-term economic effects of COVID-19. Future research is needed to assess the longterm economic impact of COVID-19 on the health-care system (e.g. for management of chronic diseases) and on society (e.g. reduced international trade and increased unemployment rates). Second, due to a lack of data, we could not include some cost components, such as productivity losses for carers of suspected and confirmed cases and out-of-pocket payments for travel to hospitals and over-the-counter medicines. Third, because of a shortage of nucleic acid tests in China in January 2020, not all patients suspected of having COVID-19 were tested. 3 Therefore, the reported number of confirmed cases is likely to be an underestimate, especially in Hubei province. Fourth, our estimate of the number of working days lost, which we based on migration data, may have overestimated losses for people who worked from home. Fifth, we lacked some data on the incidence, demographic information and prognosis for close contacts and suspected cases, and had to estimate these data based on published literature and/or expert opinion. Finally, some positive effects of the restrictive measurements have been reported, such as reductions in crime rates, 32 environmental improvements 33 and a rapid increase in e-commerce. 34 Analysis of the effects of these factors was beyond the scope of our study.
The results of our study highlight the substantial economic burden of the COVID-19 outbreak. Research is needed on the cost-effectiveness of different policies to control infectious diseases and developing capacity to limit the spread of disease while minimizing the impact on everyday life. ■