Thirty-three studies measured morbidity or mortality as an outcome due to various climate events. Twenty-eight of the 33 (85%) studies examined the relationship between temperature extremes, either defined as a heatwave or within a certain percentile of mean ambient temperature, or temperature changes and morbidity/mortality. The remaining five studies investigated the relationship between air pollutants or river water levels and morbidity/mortality.
Sixteen studies examined how exposure to temperature extremes influences presentations and admissions to hospital, presenting results either as (i) exposure to a temperature extreme and the risk of admission to hospital, or (ii) the impact on admissions to hospital due to an increase in ambient temperature. While highly variable among studies, exposure to a temperature extreme > 95th percentile of average daily temperature increased the odds ratio of paediatric hospital admission from 1.17 (95% confidence interval: 1.12–1.21) to 1.27 (1.12–1.44),26, 27 and 3.6 (1.4–9.5) for paediatric renal admission to hospital28. A 1.011–1.013 odds ratio of paediatric emergency-department presentations was reported after exposure to a temperature extreme > 99th percentile of average daily temperature29.
Measuring a health outcome based on an absolute increase in temperature, often during a predefined ‘hot season’, was adopted by many studies30. A 7o C increase in consecutive maximum daily temperatures raised the odds ratio of paediatric admissions to hospital to 1.022–1.031, and all-causes emergency department visits to 1.004–1.022 for a 5.5o C increase31, 32. Socioeconomic level is an important mediating factor for the impact of temperature changes on child morbidity, with a 1.095–1.125 odds ratio increase versus 1.043–1.067 increase in risk of hospitalisation for children in low- and middle-income cities versus high-income cities33. Increased temperature variability is also associated with increased hospitalisations34.
Exposure to temperatures < 5th percentile of average daily temperature increased35 the all-causes mortality odd ratio to 1.117–1.17, and 1.03–1.78 increase in 18 + only/all-causes mortality upon exposure to temperatures > 95th percentile36. An increase of 1o C during hot periods in Chinese cities37 raised the odds ratio of all-causes mortality anywhere from 1.009–1.056 to 1.015–1.097. Children < 5 years of age are generally more sensitive than the rest of the population to high temperatures, but not to low extremes38. Five studies37–41 identified children as a vulnerable group to temperature extremes and increases in average temperature, while one study42 found children were the least vulnerable group.
Forty-nine studies explored the impacts of climate changes on various perinatal outcomes, of which 39 examined temperature changes, and 10 examined air pollutants. The main health outcome measured was preterm birth, followed by changes in birthweight and pregnancy loss. Of the 39 studies examining the impact of temperature changes on perinatal outcomes, preterm birth (births occurring up to week 37 of gestation) was the most common health outcome (29 studies), followed by low birthweight, gestational age changes, premature rupture of membranes, and pregnancy loss. Of the 29 preterm-birth studies, 21 examined exposure to extreme temperature, and the remaining 8 investigated increases in ambient temperature.
Exposure to a temperature extreme > 90th percentile of maximum daily temperature results in an odds ratio of preterm birth43 of 1.25. Exposure to a temperature extreme > 95th percentile also resulted in large range of increased risk of preterm birth (1.028 to 3.171)44–50. Exposure to a temperature extreme > 97.5th percentile increased the odds ratio of preterm birth51 to 1.014–1.039, and to 1.290–1.586 after exposure to a temperature extreme > 99th percentile52–54.
Some studies reported risk after exposure to extremes of cold. One cohort study reported a 1.187–2.493 odds ratio of preterm birth when exposed to a temperature extreme < 1st percentile of mean daily temperature55, and another reported a maximum preterm birth odds ratio of 1.192–1.744 following the 2008 cold spell in two sub-tropical cities in Guangdong Province, China56. Nineteen studies concluded that exposure to temperature extremes (with assorted definitions) was associated with an increased risk of preterm birth43–49, 51–63. In terms of absolute increases in temperature, the odds ratio of preterm births was 1.043–1.246 following an increase of 5.5o C in weekly mean temperature64–66, similar to other studies67–71 identifying a positive relationship between increases in mean ambient temperature and preterm birth.
Exposure to temperatures > 97th percentile of maximum daily temperature over two consecutive days at any point during gestation increased the risk of low birthweight (< 2500 g) by 1.023–1.106, with the first trimester of pregnancy being the most vulnerable72. In Queensland, Australia, exposure to a daily maximum temperature > 30o C during the last week of gestation reduced birthweight73. But other studies suggest an opposite relationship, with a 41.8 g (0.6–82.9 g) increase in birthweight occurring with every 1o C rise in daily average temperature during the third trimester in Sub-Saharan Africa74.
Few studies examined the relationship between temperature and premature rupture of membranes, stillbirth, hypertensive diseases of pregnancy, or low Apgar scores at birth. One U.S. study found that exposure to a 1° C increase in average temperature during the week before delivery in the warm season resulted in an odds ratio of 1.03–1.06 for premature rupture of membranes75. That study also reported a decreased risk of premature rupture of membranes with a 1° C increase in average temperature during the cold season75. Increased risk of stillbirth was observed upon exposure to a temperature extreme > 90th percentile of whole-pregnancy average temperature76. Exposure to heat extremes is associated with an increased risk of maternal hypertensive disorders and low Apgar scores in the neonate77, 78.
Ten studies explored relationships between airborne pollutants and various perinatal outcomes (preterm births [n = 7] and birth weight [n = 3]), all using various measures of particulate matter: PM1, PM2.5, PM10. One cross-sectional study in Africa found that an interquartile range increase (33.9 µg m− 3) in the concentration of PM2.5 was associated79 with an odds ratio of preterm birth of 1.01–1.16, and another in China observed an increased risk of preterm birth (1.094–1.105) with an increase PM1 concentration of 10 µg m− 3 during the entire pregnancy80. Others62, 81 reported no relationship between PM2.5 or PM10 and the risk of preterm birth. A 1.014–1.018 increase in the odds ratio of preterm birth was observed after exposure to a 10 mg m− 3 rise in SO282.
The three studies examining reductions in birthweight81, 83, 84 found that airborne pollutants due to fire exposure reduced birth weight by 0.56–3.77 g per 1 µg m− 3 increase and a 1.01–1.049 odds ratio of low birth weight83, a 1.01–1.12 increase in the odds ratio of low birth weight with a 1 µg m− 3 increase81 in PM10, but the magnitude of the effect depends on temperature84. One study in South Asia found different increases in the risk of stillbirth per 1 µg m− 3 rise in average PM2.5 depending on the source of the pollutant85. A 1 µg m− 3 increase in fire-sourced PM2.5 increased the stillbirth odds ratio85 by 1.035–1.067, but non-fire PM2.5 by 1.011–1.016.
Forty-three studies examined the way in which various climate variables influence child respiratory health (mainly, emergency-department presentations, asthma incidence, and risk of infectious respiratory disease), 26 of which explored the deleterious impacts of temperature and 16 of which focussed on air quality and pollution. Five studies measured the influences of temperature changes on rates of admission to hospital for generalised respiratory paediatric illnesses. One cohort study (n = 796,125) in Fuzhou, China found that lower mean temperatures (< 25th percentile of the mean) had a higher risk (0.998–1.045) of admission due to respiratory disease among children86. Similar impacts were reported in Beijing, China, were cold extremes increased child respiratory presentations to the emergency department, but the impacts on respiratory health had a longer period of influence for children (< 15 years) than for older people (> 65 years)87. Another study in China (Cangnan) found children < 4 years of age did not experience greater respiratory disease due to heatwaves88.
Exposure to temperature extremes > 95th percentile of the mean results in an odds ratio of 1.287–2.737 for childhood asthma emergency-department presentations89, 90, whereas an increase in temperature from the 25th to 75th percentiles results91, 92 in an odds ratio of 1.019–1.086. The risk for childhood asthma emergency department presentations following a 5o C rise in diurnal temperature range93 increased to 1.124–2.381, and there was a 0.986–1.009 greater risk following a 1o C increase in lower daily mean temperature during thunderstorm days. The consensus is that exposure to increased temperatures and/or high temperature extremes is associated with a variable increased risk of childhood asthma89–95.
Thirteen studies examined the influences of temperature changes on infection-related respiratory disease, with ten exploring respiratory viruses. Exposure to lower temperatures is associated with an increase in risk of human parainfluenza virus, respiratory syncytial virus, influenza A virus, coronavirus and viral-induced childhood bronchitis/bronchiolitis96–103, but ‘lower temperatures’ are defined differently among studies. An increase in 10o C in diurnal temperature range was associated with an odds ratio of 1.012–1.103 for upper respiratory tract infections in children ≤ 5 years old104, and 1.004–1.131 for lower respiratory tract infections among children aged 6–18 years104. Every 1o C increase in weekly average diurnal temperature range results in an odds ratio of respiratory syncytial virus infection of 1.71–7.23105. However, variation in diurnal temperature range did not affect the risk of influenza A virus101 or childhood pneumonia106. Likewise, weather conditions did not affect the risk of paediatric intensive care unit admission due to respiratory syncytial virus infection107.
Sixteen studies explored the relationship between air quality or aeroallergens and childhood respiratory health (mainly, risk of asthma), with PM2.5 being the most studied airborne pollutant, followed by aeroallergens, PM10, O3, and NO2. The risk of asthma can increase by 1.058–1.105 when PM2.5 (due to fire smoke) increases 1 µg m− 3 108, although this result is not consistent across studies109. An increase of 1 µg m− 3 PM2.5 also raises the risk of combined respiratory disease (1.012–1.031)108. An interquartile range increase in SO4 concentration was associated with a greater risk (1.343–4.365) of all-cause hospitalisations among children with asthma109, and other studies110–112 report a greater risk of childhood asthma from increased O3 and NO2, but not PM2.5. A 10 grains m− 3 increase in total aeroallergen concentration also raised the incidence of hospitalisation due to asthma (1.011–1.098)113–115. PM2.5 also increases the risk of childhood pneumonia presentations to hospital116, 117. Respiratory outcomes due to pollution increases are worse when temperature decreases118.
Increased PM2.5 and PM10 concentration can also increase the risk of all-cause childhood respiratory presentations to the emergency department108, 119–121. For example, fires associated with drought conditions in the Brazilian Amazon increase childhood respiratory hospitalisations122, whereas another study in the same region observed an odds ratio of 1.266 for childhood respiratory disease during the dry season compared to the wet season123. Fire-related air pollution also increases the risk of hospitalisations124.
Standardisation was not possible for all climate and health variables due to a lack of studies available. For those variables with adequate data, the standardisation procedure allowed us to compare different categories directly to assess the relative magnitude of effect sizes. Overall, there was a higher effect of temperature extremes on the risk of preterm birth (Fig. 2a) compared to the risk of respiratory disease (Fig. 2b) or mortality/morbidity (Fig. 2c). Temperature changes caused an average increase in preterm birth of 60% but this belies high uncertainty and variability across studies.
However, 8 of the 27 (30%) studies for which an odds ratio of effect of temperature change on preterm birth risk could be calculated indicated no effect (lower confidence limit overlapping or near 1; Fig. 2a). For the risk of respiratory disease from exposure to temperature extremes, at least 6 of the 14 (43%) studies with comparable data indicated no effect (Fig. 2b), and 18 of the 32 (56%) studies assessing the effects of temperature extremes on the risk of morbidity/mortality indicated no effect (Fig. 2c). For the latter, there was no discernible difference in the odds ratio combined for different categories of age (children versus all ages) or outcome (mortality versus morbidity) (Fig. 2c, 4 category combinations at the top of the graph).
When we examined the relationship between effect size (odds ratio) and either the percentile change (Fig. 3, top panel) or absolute magnitude (Fig. 3, bottom panel) of temperature change, there was no discernible relationship for any health outcome category (Fig. 3).
One might expect a U-shaped relationship for the temperature percentile relationships, where extreme cold or extreme heat would be expected to increase the risk of the health outcome in question, but the data were too variable and there were not enough results available for mid-range temperatures to test this hypothesis (Fig. 3a). Neither did increasing magnitudes of temperature increase affect effect sizes, mainly given the lack of sufficient sample size to detect a trend (Fig. 3b).
The odds ratios for the effects of different pollutants on health outcomes were considerably smaller than for temperature effects, but with most (16/20 = 80%) studies indicating at least a weak effect (Fig. 4).
While the odds ratios were on average slightly higher for respiratory disease compared to perinatal outcomes, the range of effect sizes were approximately the same for both categories (Fig. 4). When we combined the different outcomes by pollution variable (irrespective of health outcome measured), there was a higher impact of NOx pollution on health compared to any particulate matter- or SOx-focussed studies, although sample sizes were consistently small (Fig. 5). Appendix 7 provides additional details of studies examining other health outcomes not summarised above.
Local climate and socioeconomic factors are important to consider when exploring health-climate relationships. Most studies included in the meta-analysis explore populations in the Northern Hemisphere (84.5%). Fifty-six percent of these studies were done in high-income nations/regions, 39% in upper-middle income nations/regions, 2% in lower-middle income nations/regions, and 2% in low-income nations/regions. There is no obvious clustering of studies geographically by intensity of effect (odds ratio) or health measure (Fig. 6).