The burden of heatwave-related preterm births and associated human capital losses in China

Frequent heatwaves under global warming can increase the risk of preterm birth (PTB), which in turn will affect physical health and human potential over the life course. However, what remains unknown is the extent to which anthropogenic climate change has contributed to such burdens. We combine health impact and economic assessment methods to comprehensively evaluate the entire heatwave-related PTB burden in dimensions of health, human capital and economic costs. Here, we show that during 2010-2020, an average of 13,262 (95%CI 6,962-18,802) PTBs occurred annually due to heatwave exposure in China. In simulated scenarios, 25.8% (95%CI 17.1%-34.5%) of heatwave-related PTBs per year on average can be attributed to anthropogenic climate change, which further result in substantial human capital losses, estimated at over $1 billion costs. Our findings will provide additional impetus for introducing more stringent climate mitigation policies and also call for more sufficient adaptations to reduce heatwave detriments to newborn.

(95%CI: 1.09-1.29) due to the heatwave exposure in the last week of gestation, which was further used for heatwave-related PTB estimate.

Identifying spatial-temporal dynamics of heatwaves in the past decades
Identifying heatwaves in actual climate during 1979-2020. The 90 th percentile of daily maximum temperatures in warm seasons during 1986-2005 was computed as the threshold to identify heatwaves over long timescales . Based on ERA5 data, the observed heatwaves in warm seasons (May-October) per year were identified by two or more consecutive days with daily maximum temperatures above the threshold of each grid. The heatwaves were identified for each grid (~50km×50km) based on the cell temperature distribution. The gridded heatwave days were then averaged at the national level to display the temporal trend of national heatwave days. Heatwave days in girds were also averaged by 10 years to illustrate the spatial distribution (in grids) of annual average heatwave days during 2010-2020.
Identifying heatwaves from anthropogenic climate change. The additional heatwave days attributed to anthropogenic climate change were derived from the paired runs of ten General Circulation Models for the factual (accounting for natural and anthropogenic forcings) and counterfactual (accounting for natural forcings only) scenarios. Annual heatwave days (in warm seasons) of the two scenarios during 1979-2020 were identified by simulated temperatures respectively. The heatwaves in the factual scenario are defined as two or more consecutive days with daily maximum temperatures above the threshold, which is the 90 th quantile of modelled temperatures for the factual scenario in warm seasons during 1986-2005. Noticeably, the threshold used for the factual scenario was then applied to identify heatwaves in counterfactual scenario. The differences between heatwave days of the two scenarios were computed as additional heatwave days induced by anthropogenic climate change.

Quantifying the human capital consequences of annual attributable PTBs
Three dimensions of human capital. Owing to the broad concept of human capital and the lack of a consistent measurement, previous studies mainly focused on education outcomes as proxies of human capital, such as years of schooling, academic attainment, enrollment rate and illiteracy rate 4-6 . In recent years, several studies has considered the health and non-cognition dimensions of human capital [7][8][9][10][11] , and a few of them considered these dimensions at the same time 9,10 . To better reflect long-term impacts of attributable PTBs on human potential, we eventually summarized three dimensions of human capital, namely health, cognition and non-cognition, according to previous studies 7-9,12 .
Health is an essential element of human capital, referring to health status including mortality, morbidity and disability 5 . Cognition encompasses a range of complex neurological and psychological processes 13 , which cannot be completely reflected in a single test. Currently, two broad but complimentary constructsintelligence quotient (IQ) scores and academic performance-are generally used measures of cognition in research. Non-cognitive skills, which are alternatively called socio-emotional skills, refer to personality, goals, character, motivation and etc. that are highly appreciated in the labor market 14 . Also, there were not constant and standard measurements for non-cognitive skills. Based on current studies relevant to non-cognitive skills 9-11,15,16 , we divide non-cognitive skills into two groups: psychological characteristics (personality traits, self-esteem, motivation, selfconfidence) and behavioral tendencies (behavioral problems and social interaction tendency).
Outcomes selected to measure human capital impacts of PTB. We selected the most common long-term outcomes of PTB for each dimension to comprehensively measure the PTB-related human capital impacts. According to previous reviews of the PTB-related outcomes, PTB is a risk of a range of health impairments including premature death in neonates and young children 17 , respiratory disease such as child asthma, metabolic disease such as glucose intolerance and diabetes mellitus, and cardiovascular disease such as hypertension [18][19][20] . Substantial evidence also found a higher risk of those born prematurely in cognitive ability, academic performance 21,22 and behavior problems such as ASD and ADHD 23 . Considering the quality of evidence and the availability of data, we eventually selected the neonatal death, child asthma and diabetes mellitus (type 1&2) for health dimension, reduced IQ for cognition, and ASD and ADHD to measure impacts in non-cognitive dimension.

Monetizing annual attributable PTBs and associated human capital losses.
The unit economic costs for PTB and human capital outcomes have been reviewed and available in previous studies. We selected a recent systematic review on per-case monetary estimates for multiple child health outcomes to obtain the costs of PTB, child asthma, ASD, ADHD and loss of an IQ point in 2015 USD 24 . The costs for type 1 and type 2 diabetes are supplemented by another study of assessing the long-term outcomes of PTB induced by air pollution, in which the costs of diabetes were reviewed and converted into 2014 USD 25 . However, we did not find an appropriate estimate of the cost of neonatal mortality for analysis, and so it was not incorporated in monetization. All economic costs per case for PTB or specific human capital outcome and relevant references can be seen in Supplementary Table 7. Finally, the total economic costs of human capital impacts, except the neonatal mortality, are summed to compare with the cost of heatwave-attributable PTBs.