The clinical data of 409 children with CH from 1 January 2015 to 31 December 2020 were obtained from delivery institutions in southern China. CH was diagnosed by obstetricians, pediatricians, or surgeons through ultrasonography, urography, magnetic resonance imaging, and other techniques and clinical manifestations (5). The disease was coded according to the International Classification of Diseases 10 (ICD-10), and the birth defect disease was diagnosed and reported according to the requirements of the “China Birth Defect Monitoring Program” (13). Full-term healthy newborns without any birth defects were used as the control group. Data were collected on several parameters including maternal age, parity, pregnancy disease, past medical history and infant gestational age, fetal sex, and birth time, among others (see Figures 1, 2).

The PM_2.5 data were obtained from the monitoring data of China's ground air pollution monitoring station and the multi-angle atmospheric corrected aerosol optical depth (AOD) provided by the National Aeronautics and Space Administration of the United States. Combined with meteorological data, land use type, road network information, surface elevation, and pollutant emissions and using satellite remote sensing technology and machine learning methods, a stable PM_2.5-AOD conversion relationship was constructed using the space–time extremely randomized trees (STETs) model. The daily average PM_2.5 exposure concentration was determined at a spatial resolution of 1 km. The random 10-fold cross-validation R² value was 0.89 (14, 15).

The original data on ambient temperature were collected from the National Environmental Information Center. By obtaining the daily temperature values of 12,312 meteorological stations, the daily Tmax grid map of the whole country was obtained using the inverse distance weighted interpolation method.

According to the climatic characteristics of the southern region of China, we stipulated that the mothers participating in the study had at least 1 day of exposure to PM_2.5 and temperature within the susceptibility window of CH during the warm season (May–October) (16, 17). According to the distribution of ambient temperature in 2015–2020, the 10th percentile (14.25°C), 25th percentile (17.54°C), 50th percentile (22.58°C), 75th percentile (27.25°C), and 90th percentile (29.13°C) of Tmax were analyzed using the stratified method (18).

Based on the residential address of the participants during pregnancy, we obtained the geographical coordinates and used ArcMap 10.3 software to perform spatial location matching to calculate the pollutant concentration in the residential area of pregnant women. We thus obtained the PM_2.5 exposure level and ambient heat exposure at the corresponding spatial location. Relevant literature shows that the critical period of fetal renal pelvis and ureter development is between 5th week and 7th week of gestation. We selected this period for time matching the exposure concentration (19).

Percentile description and the Chi-squared test were performed considering the following confounding factors: maternal age, parity, disease during pregnancy, past medical history, gestational age, and weight of infants. Using the data exposure assessment method, the exact values of PM_2.5 and temperature were obtained for each participant. A multivariate logistic regression model was established to evaluate the relationship between the risk of CH and PM_2.5 concentration and ambient temperature. Using directed acyclic graphs, we referred to the relevant literature to evaluate confounding factors (20, 21). The following covariates were incorporated into the final model: maternal age, infant sex, pregnancy disease, etc. These confounding factors were analyzed as covariates (22–24) (see Figure 3).

The adjusted odds ratio (aOR) and its respective 95%CI were calculated to represent the association between the risk of CH and exposure to PM_2.5 and ambient temperature. In addition, we stratified PM_2.5 with maternal age, fetal sex, pregnancy disease, and other factors.

Finally, the interaction between PM_2.5 and ambient temperature on CH was evaluated, and the relationship between their combined effects and independent effects was evaluated. We calculated relative excess risk due to interaction (RERI), attributable proportion (AP), and synergy index (S) to evaluate the additive interaction. When RERI and AP were > 0, S was > 1, and p-value was < 0.05, the combined effect of air pollutants and intense heat exposure was greater than the sum of independent effects. When RERI and AP were < 0, S was < 1, and p-value was < 0.05, there was an antagonistic interaction, which means that when they exist at the same time, PM_2.5 and intense heat exposure will reduce the interaction (25–27). The incidence of CH (/10,000) was calculated as the number of CH/the number of perinatal infants × 10,000. A value of p of < 0.05 indicated that the difference was statistically significant. Statistical analysis was performed using the software program R, version 4.3.0.

A total of 472,241 newborns from 1 January 2015 to 31 December 2020 were considered for the study. Out of them, there were 409 children with CH, and the average incidence of CH was 8.82/10,000 persons. In the case group, 142 pregnant women (34.72%) were older than 30 years and 267 pregnant women (65.28%) were younger than 30 years. The Chi-squared test value (X²) of 20.49 (P < 0.05) indicated that the difference between different ages was statistically significant. There were 377 cases (68.21%) of full-term infants with gestational age≥37 weeks. The Chi-squared test value (X²) of 12.15 (P < 0.05) indicated that there was a statistical significance between different gestational weeks. There were 115 cases (6.51%) of gestational diabetes during pregnancy. The Chi-squared value (X²) of 44.77 (P < 0.05) indicated that gestational diabetes during pregnancy was statistically significant. There were 40 cases (5.91%) of hepatitis B during pregnancy. The Chi-squared test value (X²) of 2.31 (P > 0.05) indicated that hepatitis B during pregnancy was not statistically significant (see Table 1).

The incidence of CH was 5.57/10,000 in 2015, 8.20/10,000 in 2016, 7.45/10,000 in 2017, 11.21/10,000 in 2018, 10.71/10,000 in 2019, and 9.75/10,000 in 2020. In the case group from 2015 to 2020, the highest temperature was recorded in 2020 (22.18°C) and the highest PM_2.5 concentration was recorded in 2016 (33.05 μmol/L) (see Figure 4). The average temperature to which the CH group was exposed to was 21.3 ± 5.45°C, and the average temperature to which the healthy control group was exposed to was 22.43 ± 5.01°C. The average concentration of PM_2.5 to which the case group was exposed to was 28.31 ± 7.97 μmol/L, and the average concentration to which the control group was exposed to was 26.59 ± 7.19 μmol/L (P < 0.05), indicating that the difference was statistically significant (see Table 2).

A Pearson correlation analysis was performed on exposure to PM_2.5, meteorological factors, and CH throughout pregnancy. The results showed that there was a negative correlation between temperature and humidity (P < 0.05), and the correlation coefficient was −0.77. There was a negative correlation between precipitation and temperature (P < 0.05), and the correlation coefficient was −0.79. We found that precipitation was positively correlated with humidity (P < 0.01), and the correlation coefficient was 0.99. There was no correlation between PM_2.5 and meteorological factors (P > 0.05) as well as between CH and meteorological factors (see Table 3).

Fetal sex, maternal age, gestational diabetes, and other factors were used for the stratification analysis. The results showed that the presence of thyroid disease and hepatitis B during pregnancy were risk factors for CH in newborns. Pregnant women without thyroid disease and hepatitis B during pregnancy carried a lower risk of CH in newborns [β = 0.25 and 0.34, 95%CI = (0.15, 0.41) and (0.21, 0.55)]. No correlation was found between age, fetal sex, gestational diabetes, and CH (see Figure 5).

The critical period of renal pelvis and ureter development is between 5th and 7th weeks of gestation. The correlation between PM_2.5, temperature, and CH was calculated by using this period of time as the exposure window (19). The results showed that, in the single pollutant model, for every 10 μg/m³ increase in PM_2.5 at the 6th week of gestation, the incidence of CH increased by 1% [OR = 1.01,95%CI = (0.98, 1.10)]. Taking maternal age, fetal sex, pregnancy disease, and other factors as covariates into the analysis, the obtained aOR was 1.02 and the 95%CI was (0.99, 1.05) (P > 0.05). In addition, the results revealed that no significant relationship between PM_2.5 and CH exists.

In the single-pollutant model, we performed a logistic regression analysis of ambient temperature and CH. The results showed that, at 6th and 7th weeks of gestation, for every 1°C increase in temperature, we obtained unadjusted odds ratio (cOR) values of 1.18 [95%CI = (1.05, 1.32)] and 1.12 [95%CI = (1.02, 1.22)], respectively, for CH. The covariate factors such as maternal age, fetal sex, and pregnancy disease were considered in the analysis. The aOR values were 1.18 and 1.13 at the 6th and 7th weeks of gestation, respectively. The 95%CI values at 6th and 7th weeks of gestation were (1.05, 1.33) and (1.03, 1.24), respectively. There was a significant relationship between temperature and CH (see Table 4).

In the two-pollutant model, the environmental temperature and PM_2.5 were analyzed together with CH. It was found that, at the 6th and 7th weeks of pregnancy, for every 1°C increase in temperature, the cOR was 1.20 and 1.14, respectively. The 95%CI was (1.06, 1.35) at the 6th week of pregnancy and (1.04, 1.26) at the 7th week of pregnancy. Considering the relevant covariates for the analysis, the aOR values were 1.21 and 1.13 at the 6th and 7th weeks of pregnancy, respectively. The 95%CI values were (1.04, 1.41) and (1.02, 1.24) at the 6th and 7th weeks of pregnancy, respectively. There was a significant relationship between temperature and PM_2.5 on the occurrence of CH (see Table 4).

Table 4 shows that the 10th, 25th, 50th, 75th, and 90th of Tmax distribution are used as thresholds for hierarchical analysis. At the 5th week of pregnancy and at the 50th, 75th, and 90th of Tmax distribution, the RERI values were 0.22, 0.46, and 0.60 and the 95%CI values were (0.05, 0.40), (0.01, 0.91), and (0.03, 1.52), respectively, indicating that they had positive additive interaction. The S and 95%CI values were 1.04 (1.03, 1.05), 1.07 (1.04, 1.09), and 1.09 (1.04, 1.14) for the 50th, 75th, and 90th of Tmax distribution, respectively, indicating that individuals with both intense heat exposure and PM_2.5 risk factors had a higher risk of CH than the sum of the risk of single risk factor exposure. At the 6th week of pregnancy, the p-value of RERI was < 0.05, indicating that the intense heat exposure and PM_2.5 had a synergistic effect on the occurrence of CH at the 6th week of pregnancy.

When the Tmax distribution was less than 50th percentile, the 95%CI value of RERI and AP was 0, and the 95%CI of S was 1 (P > 0.05). PM_2.5 exhibited no synergistic effect on the occurrence of CH. From these results, it can be indicated that the effect of PM_2.5 on CH will be enhanced by intense heat exposure (see Table 5).

This study has some strengths. In this study, CH was professionally managed. The trained nurses were asked about the prenatal exposure factors and the professional doctors used the relevant scales to carry out the preliminary classification of the disease and the reclassification of the severity (42, 43). Inspection and analysis were carried out by relevant statisticians. This study had a large sample size and covered a wide area, encompassing a large population in southern China, which is conducive to the assessment of risk factors. The large sample size also improves the reliability and representativeness of this study. The personal history, disease history, and past history of the mother and the infant were used as covariates to adjust and reduce the interference of other factors. We used spatial remote sensing technology combined with machine learning methods to accurately map PM_2.5 exposure during pregnancy to the residential address of each pregnant woman, thereby generating accurate air pollutant exposure values. Finally, this study included the medical history data of live births and stillbirths.

Our current research also has some limitations. First of all, our research is based on the living address of pregnant women, but pregnant women spend most of their time at home, and there may be movement during pregnancy, which will affect the accuracy of the results (44). Second, in addition to meteorological factors and air pollution, the influencing factors of CH also include living environment, diet, exercise, and other factors during pregnancy (45–47). However, these factors were not collected for correlation analysis. In terms of statistics, our study was limited by data, and the onset time was limited to the critical period of renal pelvis and ureter development, which limited the application of research duration and related statistical models. Finally, since this study was retrospective in nature, there is a possibility of some recall bias.