The association between daily-diagnosed COVID-19 morbidity and short-term exposure to PM1 is larger than associations with PM2.5 and PM10.

Exposure to particulate matter (PM) could increase both susceptibility to SARS-CoV-2 infection and severity of COVID-19 disease. Prior studies investigating associations between PM and COVID-19 morbidity have only considered PM2.5 or PM10, rather than PM1. We investigated the associations between daily-diagnosed COVID-19 morbidity and average exposures to ambient PM1 starting at 0 through 21 days before the day of diagnosis in 12 cities in China using a two-step analysis: a time-series quasi-Poisson analysis to analyze the associations in each city; and then a meta-analysis to estimate the overall association. Diagnosed morbidities and PM1 data were obtained from National Health Commission in China and China Meteorological Administration, respectively. We found association between short-term exposures to ambient PM1 with COVID-19 morbidity was significantly positive, and larger than the associations with PM2.5 and PM10. Percent increases in daily-diagnosed COVID-19 morbidity per IQR/10 PM1 for different moving averages ranged from 1.50% (-1.20%, 4.30%) to 241% (95%CI: 80.7%, 545%), with largest values for exposure windows starting at 17 days before diagnosis. Our results indicate that smaller particles are more highly associated with COVID-19 morbidity, and most of the effects from PM2.5 and PM10 on COVID-19 may be primarily due to the PM1. This study will be helpful for implementing measures and policies to control the spread of COVID-19.

Introduction

The spread of COVID-19 has become one of the most important global health events since World War II and significantly impacts human life worldwide. The outbreak of COVID-19 was initially reported in Wuhan, China in December 2019 (Li et al., 2020; Huang et al., 2020, Wei et al., 2022). It was declared a pandemic and global public health emergency by the World Health Organization (WHO) in March 2020 due to its high infectivity and severe health threat, as well as rapid growth in morbidity and mortality. Globally, the COVID-19 pandemic has resulted in over 400 million confirmed cases with about 5.8 million confirmed deaths through early February 2022 (WHO, 2022).

People with COVID-19 were reported to have a wide range of mild to severe symptoms and clinical characteristics, including, but not limited to cough, shortness of breath or difficulty breathing, and acute respiratory failure (CDC, 2021; Guan et al., 2020). Main transmission pathways of respiratory viruses include inhalation of respiratory droplets between people in close contact (distance less than 6 feet), aerosol transmission in indoor and crowded spaces with poor ventilation, and physical contact with contaminated surfaces or with an infected person carrying the SARS-CoV-2 (CDC, 2021; Qian et al., 2021).

There is evidence that indoor and outdoor particulate matter (PM) short-term air pollution could impair the defense of the upper airways, weaken the immune response, and make humans more prone to be infected by the virus (Zhu et al., 2020; Chen et al., 2020). Some studies suggested labelling PM of air pollution as potential vectors of SARS-CoV-2 (Barakat et al., 2020; Cao et al., 2021; Shao et al., 2021a, 2021b). Thus, exposure to PM could increase both susceptibility to SARS-CoV-2 infection and severity of COVID-19 disease (Marquès et al., 2022). A cohort study in Milan, Italy showed a relationship between the ambient (outdoor) levels of PM2.5 (particulate matter with aerodynamic diameter <2.5 μm) and PM10 (particulate matter with aerodynamic diameter <10 μm) and COVID-19 cases (Zoran et al., 2020). High historical PM2.5 exposures in the United States are also found to be positively associated with higher county-level COVID-19 mortality rates, after accounting for many area-level confounders (PM2.5, age, temperature, etc.), with an increase of just 1 μg/m3 of PM2.5 resulting in a 15% increase in COVID-19 deaths (Wu et al., 2020).

The aerosol particles with smaller sizes (less than 5 μm), such as those containing the SARS-CoV-2, are more dangerous to human health since they can suspend for longer periods and penetrate deeper into the lungs than those with larger sizes (Hinds, 1999; Meselson, 2020; Liu et al., 2020; Shao et al., 2021a, 2021b; Port et al., 2022). Marquès and Domingo (2022) systematically reviewed the research advances on the association between the outdoor air pollution and incidence of COVID-19, and pointed out that, the limitation of the prior studies is that the examined pollutants are only confined to PM2.5, PM10, O3, NO2, SO2 and CO, and investigation on other important pollutants is urgently needed. At present, there have been no reports regarding PM1 (particulate matter with aerodynamic diameter <1 μm) and COVID-19. PM1 is composed of extremely fine particles. It has been reported that PM1 is associated with a higher risk for emergency department visits (EDVs) than larger sizes of particulate matter (Liu et al., 2021).

This study aims to investigating the association between PM1 and daily-diagnosed COVID-19 morbidity. Specifically, a two-step analysis (time-series quasi-Poisson analysis for each city and random-effects meta-analysis for overall association) was conducted based on data from 12 cities in China from January to June 2020 to assess the associations between ambient PM1 and daily-diagnosed COVID-19 morbidity.

Methods

Data collection

Our investigation was based on data from 12 cities in China from Jan 13, 2020 to June 30, 2020 (Wuhan, Beijing, Guangzhou, Shanghai, Shenzhen, Nanchang, Hangzhou, Ningbo, Taizhou, Zhengzhou, Shaoxing, and Dongguan). Fig. 1 shows the locations of the 12 cities. The criteria for selecting the cities were: the total diagnosed morbidity of COVID-19 in the city was greater than 90 cases diagnosed during the study period; and the ambient PM1 data was available from China Meteorological Administration. We excluded Feb 04, 2020 in Wuhan, because a large number of cumulative cases instead of the daily diagnosed morbidity was reported on that day (13436 cases).

Fig. 1

The locations of 12 cities included in our study.

Daily morbidities (cases with positive Polymerase Chain Reaction tests) were obtained from the daily published data (https://news.qq.com/zt2020/page/feiyan.htm#/global), which were gathered and reported by the National Health Commission in China. In order to minimize the interference of population flow, we only included the cases identified as local for each city and excluded the cases for those who traveled from abroad or other cities. Cases with positive PCR test were defined as diagnosed COVID-19 morbidity.

The PM1 in situ measurements for the study period were collected from the China Atmosphere Watch Network (CAWNET) of the China Meteorological Administration (Wei et al., 2019). The detailed information for the PM1 monitor stations used in this study is listed in Table S1 in Supplementary Material. PM1 was measured using the GRIMM Model 1.180 Aerosol Spectrometer, which is an optical particle counter that measures and records particle number concentration (counts/cc) in multiple channels based on aerodynamic size and records 5-min average counts (5-min) for each channel. The 5-min values for each channel are used to calculate the mean 24h (daily) particle concentrations. For each size channel the particle number concentrations are converted to particle mass concentrations using the GRIMM protocols. PM1 is then determined by adding the particle mass values for all channels for sizes up to 1 μm. The PM1 determined this way is comparable to the PM2.5 and PM10 measurements we used for this study, obtained from the China National Environmental Monitoring Centre (http://www.cnemc.cn/), measured using the tapered element oscillating microbalance (TEOM) or beta attenuation monitor (BAM), to determine compliance with China's National Ambient Air Quality Standard (Wei et al., 2021). It has been reported that the GRIMM-derived particle mass concentrations are highly consistent with those measured by the TEOM or BAM (Wei et al., 2019; Mukherjee et al., 2017). We assigned the average PM concentrations across all monitor stations in each city as exposure to the cases since we were unable to obtain the addresses of each case.

Daily meteorological data (including temperature and relative humidity) for each city were calculated from hourly data obtained from the NOAA Satellite and Information Service (http://www7.ncdc.noaa.gov/CDO/cdo).

Statistical analysis

We employed a two-step analysis to estimate the associations between ambient PM1 and daily-diagnosed COVID-19 morbidity.

Time-series quasi-Poisson analysis

For the first step, we used a time-series quasi-Poisson model (R package “mgcv”) to estimate the association between ambient PM1 and daily-diagnosed COVID-19 morbidity for each city. We included time as a variable to control for long-term trends (Hamilton, 1994). We also controlled for meteorological parameters, i.e., daily temperature (T) and relative humidity (RH) 5 days before diagnosis, and day of week (if the day is weekend or not), as previous studies found associations between these factors and morbidity (Ma et al., 2020). We selected T and RH 5 days before diagnosis as T and RH at this lag time had strongest associations with morbidity compared to associations for other lag times, according to our analysis as well as Ma et al.’s (2020) research. The model can be expressed as:where, E(Y) is the daily-diagnosed COVID-19 morbidity; PM 1 is the PM1 exposures for 22 different moving averages starting at 0 through 21 days before the day of diagnosis respectively; lag5T is the daily temperature 5 days before diagnosis; lag5RH is the daily relative humidity 5 days of diagnosis; weekend is a binary variable which equals to 1 for weekend and 0 for weekdays; s(time, df = 2 per-month) is a natural spline with 2 degrees of freedom per month, which represents time series; β 0 is the intercept; β 1, β 2, β 3, and β 4 are the regression coefficients; e is the standard error.

Random-effects meta-analysis

Subsequently, we applied meta-analysis (R package “metaphor”) to evaluate the overall associations in all 12 cities. We used random-effects meta-analysis as there was significant heterogeneity between the results for different cities (I2 > 50%). The overall study error variance in a random-effects meta-analysis includes two components: within-study and between-study errors (Borenstein et al., 2009). The observed mean effect of city i (ʘ ) can be expressed as follows:where, μ is the grand mean effect of PM1 on daily-diagnosed COVID-19 morbidity; ϵ is the within-study error and £ is the between study error; i represents the city ID (i = 1, 2, …12).

The combined association between PM1 and daily-diagnosed COVID-19 morbidity in all 12 Chinese cities (β) is given by weighted observed mean effects (ʘ ) for all cities. The weight for city i is the reciprocal of the error variance (Zanobetti and Schwartz, 2009).

We used equation (3) to calculate the relative percent changes (PC) of diagnosed morbidity per IQR/10 (IQR denotes interquartile range) increase of exposure to ambient PM1:where, β is the overall coefficient of PM1 for the 12 cities.

Sensitivity analysis

As the daily-diagnosed COVID-19 morbidity cases in Wuhan were much higher than other cities, this may skew the overall results for the 12 studied cities. Therefore, we conducted a sensitivity analysis on the associations for other 11 cities to exclude the interference of high morbidity factors in Wuhan cases, and compared the results with the overall results for 12 cities.

All the above statistical analyses were performed using R software. We consider results with p < 0.05 as significant and 0.05 ≤ p < 0.1 as marginally significant.

Results

Characteristics of daily-diagnosed morbidity and exposures

Our study included 170 days with daily-diagnosed COVID-19 morbidity and environmental data in 12 cities. The mean daily-diagnosed COVID-19 morbidity was 18 cases/day. The highest mean daily-diagnosed morbidity was found in Wuhan (196 cases/day), where COVID-19 was initially reported. Summary statistics of COVID-19 morbidities for all 12 cities together and for each city separately are shown in Table 1 .

Table 1

Summary of daily COVID-19 morbidities in the 12 study cities.

City	Total diagnosed morbidity	Daily range	Mean daily diagnosed
Wuhan	30706	0–1985	196
Beijing	864	0–121	5
Guangzhou	554	0–38	4
Shanghai	543	0–112	3
Shenzhen	457	0–60	3
Nanchang	225	0–21	1
Hangzhou	207	0–19	1
Ningbo	170	0–27	1
Taizhou	162	0–24	1
Zhengzhou	153	0–13	1
Shaoxing	109	0–13	1
Dongguan	96	0–9	1

Table 2 shows the descriptive information of the environmental data on the day of diagnosis for all 12 cities together and for each city separately, including ambient exposures to PM1, PM2.5 and PM10, temperature, and relative humidity. The Pearson correlations between PM1, PM2.5, and PM10 used in this study are shown in Fig. S1. The mean outdoor PM1, PM2.5 and PM10 exposure for all cities was 18.2, 29.8, and 49.1 μg/m3, respectively. The mean PM1, PM2.5 and PM10 exposure for Wuhan was 17.5, 34.3, and 52.1 μg/m3, respectively. Among the 12 cities, Zhengzhou had the highest PM exposure. The lowest PM1 and PM10 levels were observed in Dongguan, while the lowest PM2.5 level was observed in Guangzhou. The mean temperature for all cities was 18.0 °C and the mean relative humidity was 72%.

Table 2

Descriptive information for environmental data on the day of diagnosis.

City	PM1 (μg/m3)				PM2.5 (μg/m3)				PM10 (μg/m3)				Temperature (oC)				RH
	Mean	SD	Min	Max	Mean	SD	Min	Max	Mean	SD	Min	Max	Mean	SD	Min	Max	Mean	SD	Min	Max
Wuhan	17.51	12.69	1.86	83.71	34.40	17.87	8.50	108.79	52.07	23.96	11.67	121.63	17.30	7.94	0.68	31.83	0.72	0.14	0.40	0.98
Beijing	23.18	21.58	2.13	118.98	40.85	36.22	4.00	206.38	61.78	37.53	8.63	223.17	13.34	10.39	−5.98	31.06	0.45	0.18	0.10	0.96
Guangzhou	14.31	7.52	1.91	36.55	20.85	10.84	3.50	63.04	38.53	18.56	5.58	116.71	22.56	5.56	9.28	31.70	0.73	0.12	0.39	0.93
Shanghai	17.08	9.57	3.72	58.13	35.03	21.65	8.29	127.96	42.68	20.16	6.96	103.96	16.20	7.03	2.74	29.37	0.72	0.15	0.37	0.96
Shenzhen	9.83	6.47	1.78	50.08	17.50	9.25	4.00	61.17	31.07	14.62	6.17	96.67	23.45	4.65	10.82	30.54	0.79	0.09	0.40	0.93
Nanchang	14.95	8.67	1.83	46.49	31.20	16.09	6.92	76.63	54.21	30.96	10.96	134.33	18.37	7.62	3.71	31.79	0.76	0.15	0.40	0.98
Hangzhou	19.64	10.28	5.10	80.25	29.47	15.68	6.75	114.46	53.52	26.80	10.33	167.75	16.32	7.76	2.63	31.36	0.73	0.16	0.35	0.99
Ningbo	15.46	9.35	3.25	54.89	24.22	14.72	3.79	95.46	40.06	21.55	7.38	120.29	15.86	6.39	2.61	27.91	0.82	0.13	0.35	1.00
Taizhou	17.34	10.39	3.65	55.86	26.61	14.74	2.88	84.21	46.61	23.46	8.50	124.79	16.02	5.85	3.91	26.25	0.86	0.11	0.42	1.00
Zhengzhou	41.89	30.44	3.53	157.73	47.80	34.98	8.79	238.96	90.01	40.34	13.58	255.79	16.06	9.12	−1.91	32.13	0.55	0.20	0.14	0.97
Shaoxing	18.49	7.87	5.65	48.77	27.23	11.61	6.17	71.47	45.34	20.74	10.00	116.67	17.01	7.42	2.56	31.79	0.73	0.17	0.27	0.98
Dongguan	8.24	5.16	0.95	25.99	22.20	13.11	1.29	74.96	33.05	16.08	3.29	98.21	23.51	4.64	10.82	30.54	0.79	0.10	0.27	0.93
All cities	18.17	15.79	0.95	157.73	29.79	21.58	1.29	238.96	49.09	29.78	3.29	255.79	17.99	7.88	−5.98	32.13	0.72	0.18	0.10	1.00

We presented the PM1 exposure and daily-diagnosed COVID-19 morbidity by date in Wuhan, where the diagnosed morbidity was the highest, in Fig. S2 in Supplementary Material.

Associations between PM1 exposures for different moving averages and daily-diagnosed morbidity

Overall associations for all 12 cities together

We assessed the associations between daily-diagnosed COVID-19 morbidity and ambient PM1 starting with the day of diagnosis (day0) to moving averages through 21 days before diagnosis (day21), controlling for daily temperature and relative humidity 5 days before diagnosis, weekend/weekday, and long-time trends. Values of regression coefficients of PM1 (β) and p-values for the 12 cities for different moving averages are presented in Table S2 in the Supplementary Material. We found that PM1 exposures for moving averages starting from 21 days before the day of diagnosis were all positively associated with the increase of daily-diagnosed COVID-19 morbidity, and most of the associations were significant. Overall percent increases of daily-diagnosed COVID-19 morbidity with per IQR/10 PM1 for different moving averages ranged from 1.50% (95%CI: −1.20%, 4.30%) to 241% (95%CI: 80.7%, 545%) as shown in Fig. 2 . The largest association between diagnosed morbidity and PM1 is for average exposure from 17 days before diagnosis.

Fig. 2

Overall percent changes in daily-diagnosed COVID-19 morbidity per IQR/10 PM1 for 12 cities for different exposure windows.

Associations for 11 cities excluding Wuhan

We excluded Wuhan and conducted analysis on the associations for other 11 cities for sensitivity analysis. Values of regression coefficients of PM1 (β) and p-values for 11 cities for different moving averages are presented in Table S3 in the Supplementary Material. We found that the results were very similar with the associations for 12 cities as shown Fig. 3 . It indicated that, though daily-diagnosed COVID-19 morbidity cases in Wuhan were much higher than other cities, including Wuhan in our meta-analysis did not skew the overall results for the 12 studied cities.

Fig. 3

Overall percent changes in daily-diagnosed COVID-19 morbidity per IQR/10 PM1 for 11 cities excluding Wuhan for different exposure windows.

Associations for Wuhan

The regression coefficients for PM1 (β 1) for Wuhan, the first outbreak site of COVID-19 and the most severely affected area in China, are shown in Table S4 in the Supplementary Material. Percent changes in daily-diagnosed COVID-19 morbidity per IQR/10 p.m.1 for Wuhan for different moving averages are shown in Fig. 4 . According to our results, the pattern of the associations was similar with the overall results. Percent changes were all positive except for the exposure on the diagnosis day. The percent changes of daily-diagnosed COVID-19 morbidity per IQR/10 PM1 in Wuhan were larger than the overall values, ranging from −1.4% (95%CI: −3.1%, 0.4%) to 32.7% (95%CI: 20.7%, 45.8%). The values of percent changes were lower for Wuhan compared with the overall percent changes, because the IQR value in Wuhan was much lower.

Fig. 4

Percent changes in daily-diagnosed COVID-19 morbidity per IQR/10 PM1 for Wuhan with different exposure windows.

Associations between PM2.5 and PM10 and daily-diagnosed morbidity

In order to compare the effects of PM1 with those of the two larger particle sizes, we applied the two-step method as described in the Methods section to examine the associations between daily-diagnosed COVID-19 morbidity with both PM2.5 and PM10. The time-series quasi-Poisson model for each city is:where, Pollutant is the exposure of PM2.5 or PM10 with different moving averages starting from 21 days before diagnosis.

Overall values of regression coefficients of PM1 (β) for PM2.5 and PM10 are presented in Table S5 and Table S6 in Supplementary Material. Overall percent changes in daily-diagnosed COVID-19 morbidity per IQR/10 PM2.5 and PM10 with different moving averages are shown in Fig. S3 and Fig. S4 in the Supplementary Material, respectively. Compared with PM1, the overall percent changes with PM2.5 and PM10 per IQR/10 were smaller. Overall percent increases of daily-diagnosed COVID-19 morbidity per IQR/10 PM2.5 for different moving averages ranged from 2.00% (95%CI: 0.300%, 3.80%) to 90.8% (95%CI: 36.3%, 167%), as shown in Fig. S3. For PM10, the percent changes ranged from 2.80% (95%CI: 0.700%, 5.00%) to 52.7% (95%CI: 23.6%, 88.6%) (Fig. S4).

Discussion

This study investigated the associations between short-term exposure to ambient PM1 and daily-diagnosed COVID-19 morbidity based on the data from 12 cities in China using a two-step analysis. Prior studies of the associations between PM and COVID-19 morbidity generally focused on PM2.5 or PM10 (Zoran et al., 2020; Wu et al., 2020; Zhou et al., 2020a, 2020b; Conticini et al., 2020; Travaglio et al., 2021; Liang et al., 2020). To the best of our knowledge, our study provides the first attempt to examine the association between PM1 and COVID-19 morbidity.

Our results show, for the 12 investigated cities together, daily-diagnosed COVID-19 morbidity was significantly associated with PM1 exposures for different moving averages starting from 1 day before diagnosis (day1) to day21 (p-values < 0.05), except for day18. While for Wuhan city, daily-diagnosed COVID-19 morbidity was significantly associated with ambient PM1 exposures starting from day3 to day19 (p-values < 0.05), except for day11. Specifically, with per IQR/10 increase of PM1 concentration for day17, the overall percent change of daily-diagnosed COVID-19 morbidity was 241% (95%CI: 80.7%, 545%) for the 12 cities together, and the percent change was 32.7% (95%CI: 20.7%, 45.8%) for Wuhan. This is a remarkable increase, indicating that PM1 exposure is highly correlated with daily-diagnosed COVID-19 morbidity.

Our results also indicate that the associations between PM1 and daily-diagnosed COVID-19 morbidity increased with the length of exposure windows for exposure windows starting from day0 to day17, and then decreased. For PM1 exposure starting from day17, the percent change (PC) of daily-diagnosed COVID-19 morbidity increases about 160-fold compared to PC for exposure on diagnosis day. While for Wuhan city, the PC for exposure starting from day17 increases about 29-fold compared to PC for day1. The main reason may be that for longer exposure windows, the coronavirus can transmit wider and farther, resulting in more significant cumulative impact. Under the exposure condition of longer time, people have more possibility and chance to be infected by the virus via various pathways, e.g., inhalation of respiratory droplets or aerosols or direct touch, which makes the morbidity rate rise rapidly. In addition, the possibility of a lag between exposure and when there is sufficient response to infection to test positive also contributes the higher morbidity in longer time.

Since many studies attempted to analyze the impact of PM2.5 and PM10 on the COVID-19 morbidity, it is necessary to make a comparison among the contributions for particles of different sizes (i.e., PM1, PM2.5 and PM10), to identify which size of particles shows the highest effect. Therefore, we also examined associations of diagnosis of COVID-19 with PM2.5 and PM10. We found that the association between diagnosis of COVID-19 and PM1 exposure was larger than the associations with PM2.5 and PM10 exposures. For example, at moving average starting from day17, the PC of PM2.5 (90.8% (95%CI: 36.3%, 167%)) and PM10 (29.5% (95%CI: 16.6%, 43.8%)) for the 12 cities together were about 62%–88% lower than that of PM1 (241% (95%CI: 80.7%, 545%)). This analysis suggests that PM1 concentration may contribute more significantly than PM2.5 and PM10 concentrations, which have been used as exposure indices in prior studies. Because PM1 has a smaller size range, it may travel farther and suspend longer than PM2.5 and PM10, and it can penetrate deeper into the lungs, and lead to more severe health effects. This indicates that future studies of PM health effects should include this size range. In order to examine which size range of the particles may influence the COVID-19, we included PM1 together with two residuals: the residual of PM2.5 and PM1 (represents PM from 1 μm to 2.5 μm), and the residual of PM10 and PM1 (represents PM from 1 μm to 10 μm), respectively, in our models. However, no significant associations were found between either of these residuals and diagnosed morbidity. This result suggests that most of the effects from PM2.5 and PM10 on COVID-19 may be primarily due to the PM1.

A possible feature of the PM1 that may be responsible for the higher effects per μg/m3 (or IQR) is that each μg of PM1 particles has a far greater number of particles than for the same mass for either PM2.5 or PM10. This suggests that part of why there are higher effects from PM1 than the other two is that the particle number concentration influences the outcome, rather than the mass concentration of each of these size ranges, since each individual particle will cause cell damage (See and Balasubramanian, 2008). Table 2 shows that for all 12 cities, the mean mass concentration (μg/m3) increases with increasing size range going between PM1 and PM10. So it is clear that the differences in effects are not related to mass concentrations for these three sizes. This reason suggests that future studies should include direct measurements of particle number concentration, using the conventional condensation nuclei counter (CNC) method (McMurry, 2005).

For PM2.5 and PM10, the percent increases of daily-diagnosed COVID-19 morbidity in the present study are higher than the results in a prior study. Zang et al. (2022) applied the meta-analysis to quantitatively explore the association between the short-term exposure to air pollutants and COVID-19 risk. They obtained that, with a 1 μg/m3 increment in the PM2.5 and PM10 concentrations, the COVID-19 morbidity increases 0.3% and 0.5%, respectively. The reasons for the difference are probably due to that: (1) the present study uses a different index, i.e., the percent change of daily-diagnosed COVID-19 morbidity per IQR/10; (2) we perform calculation and analysis under different moving averages from day0 to day21. However, the present results are comparable to the results in Lu et al. (2021)'s study based on data of 41 Chinese cities, which indicated that, a 10 μg/m3 increase in the PM2.5 concentrations was positively associated with a 5% percent change in the COVID-19 morbidity.

Our study has some limitations. Firstly, confirmed cases may not capture all morbidity and the number of observation days is limited (170 days), which will result in some bias and reduce the precision and power. However, previous study indicates that the precision and power depend only in the useable variation of exposure and the total number of disease events, instead of the number of the days (Armstrong et al., 2020). Secondly, the official report of morbidity is the number of cases with positive PCR tests and the test capacity varies greatly at the beginning or end of the pandemic, which will impact the number of cases reported each day. Thirdly, some potential confounding factors in our statistical models may be missing, such as public health-related policies, the shifting populations between cities, or UV radiation (Chaudhry et al., 2020; Hopman et al., 2020; Badr et al., 2020; Zhou et al., 2020a, 2020b; Ianevski et al., 2019). We tried to include some policies such as “the Wuhan lockdown” in our model, but did not find significant results. Moreover, the present study doesn't include analysis on individual-level features (age, gender, etc.) of morbidity. Despite the above limitations, our present study is the first to incorporate PM1 exposure as an important factor for the morbidity of COVID-19 and may be helpful for implementing measures and policies related to community mitigation and personal protection (e.g., use of masks that could capture small particles <1 μm efficiently) to control the spread of COVID-19.

Conclusions

This is the first study to identify a significantly positive association between short-term exposure to PM1 and daily-diagnosed COVID-19 morbidity. Our results also indicate that smaller particles are more highly associated with COVID-19 morbidity, and most of the effects from PM2.5 and PM10 on COVID-19 may be primarily due to the PM1. Therefore, researchers and policy makers should account for PM1 as a potential risk factor of COVID-19.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.