Skip to main content
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2025 Jun 23.
Published in final edited form as: Environ Res. 2021 Mar 6;196:110924. doi: 10.1016/j.envres.2021.110924

The immediate effects of winter storms and power outages on multiple health outcomes and the time windows of vulnerability

Shao Lin a,*, Wangjian Zhang a, Scott Sheridan b, Melanie Mongillo c, Stephen DiRienzo d, Neil A Stuart d, Eric K Stern e, Guthrie Birkhead f, Guanghui Dong g, Shaowei Wu h, Shahriar Chowdhury i, Michael J Primeau j, Yuantao Hao k, Xiaobo X Romeiko a
PMCID: PMC12184809  NIHMSID: NIHMS2080609  PMID: 33689823

Abstract

Background:

While most prior research has focused on extreme heat, few assessed the immediate health effects of winter storms and associated power outages (PO), although severe storms have become more frequent. This study evaluates the joint and independent health effects of winter storms and PO, snow versus ice-storm, effects by time window (peak timing, winter/transitional months) and the impacts on critical care indicators including numbers of comorbidity, procedure, length of stay and cost.

Methods:

We use distributed lag nonlinear models to assess the impacts of winter storm/PO on hospitalizations due to cardiovascular, lower respiratory diseases (LRD), respiratory infections, food/water-borne diseases (FWBD) and injuries in New York State on 0–6 lag days following storm/PO compared with non-storm/non-PO periods (references), while controlling for time-varying factors and PM2.5. The storm-related hospitalizations are described by time window. We also calculate changes in critical care indicators between the storm/PO and control periods.

Results:

We found the joint effects of storm/PO are the strongest (risk ratios (RR) range: 1.01–1.90), followed by that of storm alone (1.02–1.39), but not during PO alone. Ice storms have stronger impacts (RRs: 1.04–3.15) than snowstorms (RRs: 1.03–2.21). The storm/PO-health associations, which occur immediately, and some last a whole week, are stronger in FWBD, October/November, and peak between 3:00–8:00 p.m. Comorbidity and medical costs significantly increase after storm/PO.

Conclusion:

Winter storms increase multiple diseases, comorbidity and medical costs, especially when accompanied by PO or ice storms. Early warnings and prevention may be critical in the transitional months and afternoon rush hours.

Keywords: Winter storm, Power outage, Health risk, Critical care indicators

1. Introduction

While extreme weather events are becoming more frequent and intense, with increasing impacts on infrastructure and public health (Zarzycki, 2018), few know that one of the most impactful extreme events across the northeastern United States (US) is the winter storm, such as snowstorm or ice storm. A variety of strategies are currently under investigation to improve the climate resiliency of the power infrastructure. Particularly, recent reviews indicate that emerging technologies such as photovoltaics (Wang et al., 2021), energy storage (Piana et al., 2020; Xiao et al., 2021), distributed energy resource integration (Lin et al., 2018), intentional islanding (Oboudi et al., 2019), and microgrids (Hussain et al., 2019) are gaining global popularity to enhance the resiliency of power grids against extreme weather events. Nevertheless, for guiding the adoption of these technologies to reduce health burdens, it is critical to understand the health impacts of extreme weather related power outage (PO).

Among the limited studies related to winter storms, several studies found significant increases in emergency department (ED) visits or ambulance transport for injuries after snow or ice storms (Gevitz et al., 2017; Morency et al., 2012; Piercefield et al., 2011; Rajaram et al., 2016; Smith and Nelson, 1998) and deaths after major snowstorms (Burstein et al., 2015). Compounding the direct impacts of snowstorms, these events often lead to PO as well. There have been studies that reported a compounding PO effect on total mortality and respiratory diseases during heat waves (Dominianni et al., 2018; Conlon et al., 2011; Gasparrini et al., 2015), but few have assessed the impact of a PO during winter storms. Degelia et al. (2016) reviewed the infrastructure and health impacts of ice storms in the US, while Call et al. (Call, 2009) reported that the disruption of the electrical grid as a PO can persist for several weeks after the ice has melted. More recently, Dominianni et al. (Dominianni et al., 2018) and Phung et al. (2016) found that localized, cold weather POs were positively associated with all-cause mortality and cardiovascular hospitalization, but negatively associated with respiratory hospital admission, in their assessment of the impacts of citywide and localized PO in New York City.

Although public health agencies usually evaluate the impacts of natural disasters including hurricanes and earthquakes, evaluation of the health impacts of winter storms are rarely reported, especially those storms occurring in transitional months, largely because they are difficult to categorize. Our knowledge is sparse regarding the health impacts of winter storms on a spectrum of biologically plausible disease outcomes; the independent and synergistic effects of snowstorms, ice storms and PO on health; the modifying effects of season, and window of peak timing of the disease occurrences. Given the substantive and long-lasting impact that extreme winter storms can have, the lack of research and public health attention to the associated winter-storm-health risk is surprising. Further studies are needed to inform evidence-based preparedness activities to reduce the related health burden.

To address these knowledge gaps, we 1) evaluated the individual and joint effect of winter storms and POs on hospital admissions for multiple diseases for New York State; 2) compared the health impacts of snowstorms with ice storms; 3) identified burden on other clinical indicators and peak hours of the effects; 4) assessed the storm-health associations by winter months and its transitional months.

2. Methods

2.1. Study population and health data

2.1.1. Health data source

Information on hospital admissions was obtained from the New York State (NYS) Department of Health Statewide Planning and Research Cooperative System (SPARCS), which we have used previously (Fitzgerald et al., 2014; Garcia et al., 2010; Hopke et al., 2020; Zhang et al., 2018). As a legislatively mandated database, SPARCS covers 95% of hospitals in NYS and collects personal and medical information for each patient including gender, age, race, ethnicity, admission and discharge date, health insurance and the ICD code for a principal diagnosis and up to 24 comorbidities. Information on admission timing, length of stay at hospital, medical procedures used and total costs were also obtained from the SPARCS.

2.1.2. Outcome definition

We retained hospitalizations with at least one of the following five principal diagnoses between 1/1/2001–12/31/2013: cardiovascular diseases (CVD, ICD9: 393–396,401–405,410–415,427–428,430–434,436–438), lower respiratory diseases (LRD, ICD9: 490–496), respiratory infectious diseases (ICD9: 480–488), food and water borne diseases (FWBD, ICD9: 003–009) or injuries (ICD9: E880-E910). These diseases were selected because they have been reported to be related to winter events based on prior literature (Burstein et al., 2015; Centers for Disease Control and Prevention, 1998; Dominianni et al., 2018; Rajaram et al., 2016) or biologically plausible. We geocoded the residential address of each case and assigned it to the boundary map of 1742 operating divisions across NYS, as defined by NYS Department of Public Service based on the spatial variation in electricity providers, and matched it with the operating division that it was located in (Figure S1). The health outcome was defined as the daily number admitted cases reported for each disease. Additionally, other critical care indicators, such as co-morbidity, length of stay, medical procedures used and total hospital costs were also used as time of admission.

2.2. Exposure data and definition

2.2.1. Winter storm data and definitions

We obtained the weather event data from all local storm reports of the National Oceanic and Atmospheric Administration (NOAA) Storm Events Database (https://www.ncdc.noaa.gov/stormevents/). The data contained the starting and ending time, location (i.e., coordinates) and the type of different severe weather events including ice storm, snowstorm (defined as heavy snow, blizzard, damaging snow, snow squall, heavy sleet, etc.), hurricane (surge or tropical storm), thunderstorm (lightning or severe thunderstorm), wind event (wind damage, marine-tropical storm related wind event or other wind events) and flooding (coastal flood and flash flood). Although some events are a combination of concurrent events, these events were reported jointly as a single, dominant event. We defined both snowstorms and ice storms as winter storms (https://www.ncdc.noaa.gov/stormevents/details.jsp). We geocoded assigned each event record to the power operating division according to the coordinates of event point or event area (overlapping with the division boundary).

2.2.2. Power outage data and definition

From the NYS Department of Public Service (https://www.dps.ny.gov/), we obtained PO data (2001–2013), including total number of customers in each power operating division, the date and cause of a PO (i.e., caused by a snow or ice storm) and the number of customers affected by the PO. We estimated the coverage of PO by dividing the number of customers affected over the total and then defined the PO occurrence (Yes/No) as the PO coverage exceeding the 50th percentile of the distribution of PO coverage among all operating divisions and dates with PO, which was based on a prior study that detected health impact of PO with similar PO coverage (Dominianni et al., 2018).

For both the weather event and PO data, we only included those reported from October through April where both winter and winter-transitional months (Oct.–Nov. and March–April) were included. As we geocoded the residential address for all cases from the SPARCS data, we can determine if each case was located in the power division and weather event location. Similarly, weather events were also linked to the PO data by overlapping the geocoded coordinates of each event to the power operating division boundary.

2.3. Statistical analysis and confounders

We used the distributed lag nonlinear models (DLNM) (Gasparrini, 2011; Xiao et al., 2017; Zhang et al., 2016a; Zhang et al., 2016b) to assess the potential association of winter storms and associated PO with five health outcomes. For each operating division, we ran DLNM models by including an indicator of days based on PO occurrence and weather events. With the DLNM model, the risk ratio of health outcomes in the operating division affected by the winter storm/PO were compared within the same affected area. Specifically, we compared the health outcomes during the three exposure periods (i.e., days with both winter storm and PO; days with winter storm only, and days with PO only) with the control days (i.e., neither of winter storm nor PO) during the same season in the same division. As the exposure-health associations were evaluated within the same division during different periods, the impacts of population structure, division size and SES were automatically controlled due to the self-comparison.

For each division, air pollution concentrations and weather factors of the nearest simulation grid or monitoring site were assigned to the centroid of the division to represent the values of the division. The temperature and humidity data were obtained from the integrated surface database (ISD) of NOAA, and the PM2.5 concentration data were from the community multiscale air quality (CMAQ) modeling system of EPA.

We specified our models similarly as those used previously (Gasparrini, 2013; Xiang et al., 2017; Zhang et al., 2016a):

Logcase~cbPO&winterstorm+DOW+holiday+PM25+nstime+nsdoy+nstemp+nshumidity

Where case was the daily number of admissions, following a quasi- Poisson distribution to avoid overdispersion; cb (PO&winterstorm) was a cross-basis for the indicator of power outage/winter storm day types, with 3 degrees of freedom (df) for the lag space of 0–6 days; DOW was an indicator of day of the week; holiday was an indicator of holidays; PM25 was the daily concentration of ambient PM2.5; ns (time) and ns (doy) (i.e. day of the year) were splines with 4 df to control for the long-term trend of the time series of cases; ns (temp) and ns (humidity) were splines with 3 df to fit the confounding effect of temperature and humidity. The year of 2001 was excluded from modeling since the PM2.5 concentration data were not available.

We chose to use 0–6 lag days because of the following reasons: 1) This study is focused on acute and severe exacerbation of five health outcomes rather than chronic occurrence of these diseases; 2) Based on the results, we observed that the highest health effect generally occurred within the first three days, but vanished in the later lag days; 3) Some previous studies (Anderson and Bell, 2012; Dominianni et al., 2018; Lin et al., 2011), especially a published study in NYC also examined and found the immediate health effect within the first 3 days following the PO events; and 4) We also started investigating the longer-term effect of extreme weather through other papers. Thus, we considered lags up to 6 days may be enough and appropriate to address the immediate or acute effects of winter storm and PO.

We then pooled the estimates at the operating division level to an overall risk ratio (RR) across the state with random-effect meta-analyses as was commonly used in literature (Rich et al., 2019; Viechtbauer, 2010; Wijngaarden et al., 2020). Specifically, we used the rma function from the metafor package in R. Time-series analyses were run for each health outcome and subtype of winter storm (i.e., snow vs ice), at lag days from 0 to 6. We also compared the critical care indicators between the exposure and non-exposure periods using the mean comparison based on simulation and multivariate regression models as described in Zhang et al. (2020).

2.4. Sensitivity analysis

To confirm that our estimates were robust, we changed the df for splines of temperature, humidity and time trend, and compared changes in both the estimates and Akaike’s information criterion for quasi-Poisson distribution (QAIC), an established indicator for the goodness of model fit. To estimate potential spatial variation, we also stratified our estimates by periods (i.e. 2002 – 2007 VS 2008–2013) to examine the potential changes in the immediate health effect between periods. In the stage of pooling divisional RRs to state averages, we unfortunately could not further adjust for the regional socioeconomic variables since these variables were not available at the divisional level. To estimate the potential modification effects of sociodemographic variables, we conducted another sensitivity analyses by re-examining the exposure-health associations by stratifying multiple sociodemographic variables, including gender, ethnicity, race, age group and insurance at individual level which is available from the SPARCS data and should be more accurate than divisional or regional level data. All data cleaning, analysis and results presentations were accomplished in R 3.5.1 (https://www.r-project.org/).

3. Results

As described in Table S1, the average population at the division level was 2520 (interquartile range (IQR): 1212–6982), with an average number of cases of 476 (IQR: 218–1314) for CVD, 113 (IQR: 53–320) for Respiratory infections, 99 (IQR: 46–278) for LRD, 21 (IQR: 9–65) for FWBD, and 133 (IQR: 55–419) for injuries.

Fig. 1 shows the percentage of weather events (A) and POs (B) by the following causes: thunderstorm, flooding, hurricane, winter storm and wind event in NYS, 2001–2013. There were a total of 17,571 division days reported with winter storms during the study period (Oct–Apr for 2001–2013 inclusive), which accounted for 40.7% of all severe weather events reported during this period to NOAA Storm Events database. There were a total of 19,815 division days reported with POs, among which 8339 (42.4% of all POs) were attributed to winter storms. The average coverage of PO events during the study period was 0.37% (interquartile range (IQR): 0.04%–6.12%), with an average duration of 1.00 (IQR: 1.00–3.00) days. We also found that among all winter storms by month, 38.97% of them was combined with PO in winter (Dec.–Feb.), but 73.58% of the winter storm with PO occurred in transitional months (Oct., Nov., March, April).

Fig. 1.

Fig. 1.

Percentage of weather events (A) and power outages (B) by cause* during the study period (Oct–Apr, 2001–2013), New York State. *The present study only focused on weather causes, but not causes from other sources.

The associations between winter storm and PO, jointly and separately, with multiple outcomes are shown in Table 1. Compared to PO or winter storm alone, we identified stronger associations between the combination of a PO and a winter storm and hospital admissions for all outcomes examined, with the RRs ranging from 1.04 to 1.90 (all p < 0.01) and the highest risk found in FWBD. The significantly increased risks for all outcomes occurred on lag 0–2 days after winter storm/PO, but the independent effect of winter storms was significantly associated with all health outcomes (RRs ranged from 1.05 to 1.39) on different lag days. However, we did not find significant associations between the independent effect of PO and most outcomes (CVD, LRD and respiratory infection admissions), except for injury and FWBD admission with significantly increased risks on lag days 1 and 3, respectively. When results were stratified by periods, we still found significant immediate effect estimates with the 95% confidence intervals of the estimates between the two periods overlapped with each other (Table S2). Our primary findings did not change, although some estimates decreased to non-significant due to reduced sample size. Furthermore, as described in Figure S2, we found RR variations between some of the demographic groups, but their 95% CIs generally overlapped with each other, indicating no statistically significant modified effects of socioeconomic variables on the storm/PO-health relationships.

Table 1.

The joint and independent effectsa of winter storms and power outage (PO) on hospital admissions for multiple health outcomes by lag days in New York State, 2001–2013.

Outcome Lag Winter storm/PO joint effect
Winter storm effect
PO effect
Case Risk ratio (%) Case Risk ratio (%) Case Risk ratio (%)
CVDb 0 792,070 1.25 (1.18, 1.33) 804,655 1.09 (1.04, 1.15) 705,942 0.99 (0.91, 1.08)
1 783,290 1.22 (1.13, 1.30) 790,309 1.16 (1.10, 1.22) 701,940 0.97 (0.88, 1.07)
2 808,205 1.08 (1.02, 1.15) 810,716 1.09 (1.04, 1.14) 724,685 0.95 (0.87, 1.03)
3 793,802 1.04 (1.00, 1.07) 831,392 1.05 (1.02, 1.07) 729,669 0.99 (0.95, 1.05)
4 826,831 1.09 (1.04, 1.13) 839,551 1.06 (1.03, 1.10) 755,601 1.01 (0.95, 1.07)
5 842,349 1.04 (1.01, 1.07) 839,340 1.02 (1.00, 1.04) 769,783 1.00 (0.96, 1.04)
6 833,147 1.12 (1.06, 1.18) 824,437 1.06 (1.02, 1.11) 744,411 1.00 (0.93, 1.08)
LRDc 0 144,081 1.54 (1.36, 1.75) 154,339 1.16 (1.06, 1.28) 125,528 1.03 (0.86, 1.23)
1 131,123 1.33 (1.14, 1.54) 152,613 1.25 (1.13, 1.38) 114,660 1.08 (0.89, 1.32)
2 133,490 1.24 (1.08, 1.43) 154,428 1.10 (1.00, 1.20) 116,641 1.12 (0.93, 1.35)
3 142,653 1.07 (0.99, 1.15) 160,985 1.06 (1.01, 1.11) 128,851 0.96 (0.86, 1.07)
4 144,771 1.05 (0.95, 1.16) 167,786 1.14 (1.07, 1.21) 131,722 0.92 (0.81, 1.05)
5 149,348 0.96 (0.90, 1.03) 167,102 1.09 (1.05, 1.13) 132,896 0.88 (0.80, 0.97)
6 145,665 1.04 (0.92, 1.17) 164,709 1.16 (1.08, 1.24) 130,909 0.91 (0.77, 1.07)
Respiratory infection 0 131,242 1.55 (1.36, 1.77) 159,303 1.30 (1.18, 1.43) 114,286 1.10 (0.91, 1.32)
1 126,990 1.41 (1.20, 1.64) 150,250 1.36 (1.22, 1.51) 106,118 1.03 (0.83, 1.28)
2 131,133 1.13 (0.97, 1.31) 157,065 1.16 (1.06, 1.28) 120,887 0.97 (0.80, 1.17)
3 140,903 1.08 (1.00, 1.17) 161,324 1.06 (1.01, 1.12) 130,636 0.92 (0.82, 1.02)
4 144,885 1.13 (1.03, 1.24) 161,575 1.13 (1.07, 1.21) 128,501 0.97 (0.85, 1.10)
5 144,522 1.04 (0.97, 1.11) 165,272 1.13 (1.09, 1.18) 129,532 0.91 (0.83, 1.00)
6 142,304 1.19 (1.05, 1.34) 165,406 1.31 (1.22, 1.41) 126,800 0.87 (0.74, 1.02)
FWBDd 0 9971 1.71 (1.06, 2.76) 17,123 1.39 (1.09, 1.78) 9020 1.34 (0.74, 2.43)
1 8873 1.67 (1.01, 2.78) 16,399 1.36 (1.04, 1.78) 6677 1.10 (0.55, 2.18)
2 9248 1.90 (1.21, 3.00) 16,235 1.13 (0.88, 1.46) 5972 1.92 (0.93, 3.97)
3 10,561 1.32 (1.00, 1.74) 18,233 1.05 (0.91, 1.21) 7881 1.51 (1.01, 2.27)
4 10,249 1.30 (0.92, 1.83) 18,883 1.19 (1.01, 1.40) 9140 1.16 (0.74, 1.81)
5 11,947 1.17 (0.94, 1.44) 19,591 1.13 (1.01, 1.26) 11,187 1.04 (0.77, 1.41)
6 11,818 1.44 (0.97, 2.13) 18,226 1.27 (1.05, 1.54) 10,413 1.22 (0.74, 2.03)
Injury 0 184,917 1.42 (1.28, 1.58) 205,966 1.16 (1.07, 1.26) 157,267 1.11 (0.95, 1.28)
1 172,825 1.42 (1.25, 1.61) 204,540 1.18 (1.08, 1.29) 156,344 1.22 (1.03, 1.44)
2 177,136 1.25 (1.12, 1.40) 202,285 1.14 (1.05, 1.23) 154,504 1.03 (0.88, 1.19)
3 189,664 1.04 (0.97, 1.10) 209,965 1.04 (1.00, 1.09) 166,532 1.02 (0.93, 1.11)
4 197,623 1.04 (0.96, 1.12) 217,988 1.07 (1.01, 1.12) 173,599 1.01 (0.91, 1.12)
5 190,645 1.01 (0.96, 1.06) 216,705 1.08 (1.04, 1.12) 171,728 0.96 (0.89, 1.03)
6 192,651 1.25 (1.14, 1.38) 215,862 1.22 (1.14, 1.30) 169,883 0.96 (0.85, 1.10)
a

compared to non-storm and non-PO periods; the potential confounders controlled in the model included temperature, humidity, time trend and indicators of holiday and day of the week.

b

CVD: Cardiovascular diseases.

c

LRD: Lower respiratory diseases.

d

FWBD: Food and water-borne diseases.

Table 2 shows the comparison of immediate health effects between snowstorms versus ice storms on all five outcomes. Ice storms had overall greater health impacts on all outcomes with RRs ranging from 1.04 to 3.15, compared to impacts by snowstorms with RRs ranging from 1.03 to 2.21. The highest point estimate of ice storm impact was on lower respiratory diseases (LRD, RR: 2.05, 95% CI: 1.41, 2.99), followed by injury hospitalization (RR: 1.98, 95% CI: 1.32, 2.97) and respiratory infection (RR: 1.87, 95% CI: 1.28, 2.73). The greatest impact by snow was on FWBD (RR: 2.21, 95% CI: 1.35, 3.62), followed by LRD (RR: 1.71, 95% CI: 1.49, 1.97) and respiratory infection (RR: 1.62, 95% CI: 1.39, 1.89). The increased risks for all five diseases occurred immediately after ice storms (0–2 lag days), but the impacts of snowstorm lasted a week for CVD and respiratory infection. As described in Fig. 2, higher proportions of cases for most diseases were admitted between 3:00–8:00 p.m. after the storm compared to similar time window without events. The peak of CVD cases admission was at 11:00 a.m. The trends were generally higher among days with ice storms compared to snowstorms. The secondary peaks for respiratory infection and FWBD were around 3:00–4:00 a.m.

Table 2.

The associationa between snowstorm/PO versus ice storm/PO and hospital admissions for multiple health outcomes by lag days, New York State, 2001–2013.

Outcome Lag Snow
Ice
Case Relative rate (%) Case Relative rate (%)
CVDb 0 751,774 1.31 (1.23, 1.40) 119,131 1.57 (1.31, 1.88)
1 731,752 1.21 (1.12, 1.31) 116,308 1.56 (1.26, 1.94)
2 768,755 1.08 (1.01, 1.16) 116,350 1.27 (1.04, 1.56)
3 750,227 1.03 (0.99, 1.07) 123,240 1.10 (0.97, 1.24)
4 776,993 1.09 (1.04, 1.14) 123,432 1.13 (0.97, 1.32)
5 799,847 1.05 (1.02, 1.08) 123,477 1.09 (0.97, 1.22)
6 790,584 1.15 (1.09, 1.22) 119,914 1.33 (1.08, 1.64)
LRDc 0 137,840 1.71 (1.49, 1.97) 19,500 2.05 (1.41, 2.99)
1 124,726 1.35 (1.14, 1.60) 15,645 1.72 (1.06, 2.81)
2 126,877 1.19 (1.02, 1.40) 11,540 1.84 (1.12, 3.02)
3 135,570 1.05 (0.97, 1.15) 12,296 1.08 (0.79, 1.49)
4 138,663 1.08 (0.97, 1.21) 14,493 1.22 (0.86, 1.74)
5 144,135 1.01 (0.94, 1.08) 11,935 0.87 (0.67, 1.13)
6 138,098 1.10 (0.96, 1.25) 11,680 1.24 (0.76, 2.03)
Respiratory infection 0 122,199 1.62 (1.39, 1.89) 20,735 1.87 (1.28, 2.73)
1 116,051 1.43 (1.20, 1.70) 13,040 1.72 (0.97, 3.03)
2 120,940 1.20 (1.01, 1.41) 12,050 1.06 (0.60, 1.87)
3 129,606 1.11 (1.01, 1.21) 9481 0.86 (0.58, 1.27)
4 131,570 1.16 (1.04, 1.30) 14,220 1.64 (1.09, 2.45)
5 134,402 1.06 (0.99, 1.14) 14,184 1.41 (1.01, 1.97)
6 134,586 1.23 (1.07, 1.40) 12,263 1.36 (0.73, 2.57)
FWBDd 0 8547 1.84 (1.05, 3.23) 1230 3.15 (0.91, 10.96)
1 6421 1.74 (0.93, 3.23) 801 2.66 (0.51, 14.02)
2 8675 2.21 (1.35, 3.62) 198 2.08 (0.22, 19.40)
3 9794 1.22 (0.88, 1.68) 596 1.60 (0.44, 5.81)
4 9460 1.32 (0.89, 1.95) 1006 0.84 (0.27, 2.64)
5 10,843 1.13 (0.89, 1.44) 1411 1.29 (0.45, 3.68)
6 9848 1.66 (1.05, 2.61) 808 2.39 (0.51, 11.22)
Injury 0 175,864 1.49 (1.33, 1.67) 31,019 1.81 (1.30, 2.50)
1 162,358 1.48 (1.28, 1.70) 24,847 1.98 (1.32, 2.97)
2 167,012 1.22 (1.07, 1.38) 28,543 1.88 (1.28, 2.74)
3 176,381 1.04 (0.97, 1.11) 27,106 1.12 (0.87, 1.43)
4 185,965 1.08 (0.99, 1.18) 26,696 1.25 (0.94, 1.67)
5 180,220 1.04 (0.98, 1.10) 23,590 1.21 (0.97, 1.50)
6 181,275 1.27 (1.15, 1.42) 28,942 1.80 (1.28, 2.54)
a

compared to non-storm and non-PO periods; the potential confounders controlled in the model included temperature, humidity, time trend and indicators of holiday and day of the week.

b

CVD: Cardiovascular diseases.

c

LRD: Lower respiratory diseases.

d

FWBD: Food and water-borne diseases.

Fig. 2.

Fig. 2.

Peak admission hours by health outcomes, New York State, 1/1/2001–12/31/2013.

Additionally, average cost during the hospital stay significantly increased for CVDs, injuries and LRD in the aftermath of winter storm/POs (Table 3), particularly for CVDs admissions (cost increased by $1420 per stay). The number of comorbidities for all five diseases significantly increased with the highest increase in FWBD (1.4 per case). However, the number of procedures decreased for all health outcomes after storm/PO.

Table 3.

Comparison of critical care indicators for hospital admissions between the winter storms/PO and non-storm/non-PO periods in New York State between 1/1/2001–12/31/2013.

CVDb (95%CI) LRD c (95%CI) Respiratoryinfection (95% CI) FWBD d (95%CI) Injury (95%CI)
Length of stay (days) Snow/POa 5.3(5.2, 5.5) 4.7(4.6, 4.9) 6.0(5.7, 6.2) 5.5(4.8, 6.2) 6.2(6.0, 6.5)
Control 5.3(5.3, 5.3) 5.0(5.0, 5.0) 6.3(6.3, 6.3) 5.3(5.2, 5.3) 6.8(6.7, 6.8)
Difference 0.0(−0.1, 0.1) −0.3(−0.4, −0.1) −0.3(−0.6, −0.1) 0.3(−0.5, 0.9) −0.5(−0.8, −0.3)
Regressione Slope: −0.100 (P = 0.504) Slope: −0.460 (P = 0.021) Slope: −0.392 (P = 0.219) Slope: −0.055 (P = 0.933) Slope: −0.677 (P = 0.036)
Cost ($) Snow/PO 35436 (34443, 36399) 18542 (17787, 19241) 23503 (21725, 24910) 21366 (15860, 24246) 31273 (29777, 32595)
Control 34017 (33962, 34067) 18361 (18306, 18413) 22825 (22754, 22892) 20086 (19946, 20221) 30466 (30381, 30550)
Difference 1420 (429, 2383) 181 (−558, 905) 678 (−1109, 2082) 1280 (−4177, 4214) 806 (−682, 2144)
Regression Slope: 0.078 (P < 0.001) Slope: 0.059 (P = 0.036) Slope: 0.038 (P = 0.266) Slope: −0.017 (P = 0.822) Slope: 0.062 (P = 0.022)
Number of comorbidities Snow/PO 8.8(8.7, 8.9) 7.5(7.3, 7.7) 7.9(7.7, 8.1) 8.4(7.9, 8.9) 8.0(7.9, 8.2)
Control 7.9(7.9, 7.9) 6.4(6.4, 6.4) 7.6(7.5, 7.6) 7.0(7.0, 7.1) 7.6(7.6, 7.7)
Difference 0.9(0.8, 1.0) 1.1(0.9, 1.3) 0.4(0.2, 0.5) 1.4(0.9, 1.9) 0.4(0.2, 0.5)
Regression Slope: 1.005 (P < 0.001) Slope: 0.896 (P < 0.001) Slope: 0.442 (P = 0.003) Slope: 0.896 (P = 0.007) Slope: 0.449 (P < 0.001)
Number of procedures Snow/PO 2.5(2.4, 2.5) 0.7(0.6, 0.7) 1.0(0.9, 1.1) 0.9(0.7, 1.0) 1.7(1.6, 1.7)
Control 2.7(2.7, 2.7) 1.0(1.0, 1.0) 1.2(1.2, 1.2) 1.0(1.0, 1.0) 1.9(1.9, 1.9)
Difference −0.3(−0.3, −0.2) −0.3(−0.4, −0.3) −0.2(−0.3, −0.1) −0.1(−0.3, 0.1) −0.3(−0.3, −0.2)
Regression Slope: −0.287 (P < 0.001) Slope: −0.421 (P < 0.001) Slope: −0.287 (P < 0.001) Slope: −0.191 (P = 0.206) Slope: −0.349 (P < 0.001)
a

Storm: winter storms, i.e., snowstorm or ice storm; control: days with neither winter storms nor POs.

b

CVD: Cardiovascular diseases.

c

LRD: Lower respiratory diseases.

d

FWBD: Food and water-borne diseases.

e

multivariate regression with gender, ethnicity, race and temperature, humidity and PM2.5 concentration on the day of admission.

Table 4 shows the comparisons of monthly average number of hospital admissions of all five diseases with-/without-winter storms/PO in winter months and transition months. The number of cases unexpectedly decreased during traditional winter months for all diseases (all p < 0.05), except for January with slight increases. On the other hand, the admissions for all outcomes significantly increased during the transition months of October and November (except for FWBD, which was not significant), followed by April (significantly increased for LRD, respiratory infection and injury) and March (significantly increased for injury).

Table 4.

Comparison of the monthly average number of hospital admissions between winter storm/PO period and non-storm/non-PO period and their differences by months in New York State, 1/1/2001–12/31/2013.

Oct Nov Dec Jan Feb Mar April
CVDb Snowa/PO 15.5 14.8 7.9 11.0 8.8 9.6 10.8
Control 9.8 9.5 9.4 10.1 10.1 10.2 10.2
Difference (95% CI) 5.7 (4.4, 6.8) 5.3 (3.4, 7.1) −1.5 (−1.9, −1.0) 0.9 (−0.2, 2.0) −1.2 (−1.9, −0.7) −0.6 (−1.9, 0.8) 0.6 (−0.3, 1.4)
LRD c Snow/PO 5.1 4.4 2.3 2.9 2.9 2.9 2.2
Control 2.3 2.2 2.4 2.6 2.5 2.5 2.4
Difference 2.8 (2.2, 3.2) 2.2 (1.5, 2.8) −0.1 (−0.3, 0.1) 0.3 (−0.2, 0.7) 0.4 (0.1, 0.7) 0.5 (−0.3, 1.0) −0.2 (−0.5, 0.1)
Respiratory infection Snow/PO 3.2 2.9 1.8 2.2 1.9 1.9 3.4
Control 2.1 2.2 2.7 3.0 2.8 2.6 2.4
Difference 1.2 (0.8, 1.5) 0.6 (0.1, 1.1) −0.9 (−1.1, −0.7) −0.8 (−1.2, −0.5) −0.9 (−1.1, −0.6) −0.7 (−1.2, −0.2) 1.1 (0.7, 1.4)
FWBD d Snow/PO 0.4 0.5 0.4 0.8 0.5 0.6 0.5
Control 0.4 0.4 0.5 0.5 0.5 0.6 0.5
Difference 0.1 (−0.1, 0.2) 0.1 (−0.1, 0.2) −0.1 (−0.2, −0.0) 0.3 (0.0, 0.5) 0.0 (−0.1, 0.1) −0.0 (−0.3, 0.2) −0.1 (−0.2, 0.0)
Injury Snow/PO 7.4 6.7 3.2 4.2 3.2 4.0 3.5
Control 2.9 2.8 3.0 3.1 3.1 2.9 2.8
Difference 4.4 (3.6, 5.0) 3.8 (2.6, 4.9) 0.2 (−0.0, 0.4) 1.1 (0.4, 1.6) 0.1 (−0.2, 0.5) 1.1 (0.3, 1.8) 0.7 (0.3, 1.0)
a

Storm: winter storms, i.e., snowstorm or ice storm; control: days with neither winter storms nor POs.

b

CVD: Cardiovascular diseases.

c

LRD: Lower respiratory diseases.

d

FWBD: Food and water-borne diseases.

4. Discussion

4.1. Winter storm and health

We found that hospital admissions due to all five groups of diseases, including CVD, LRD, respiratory infections, FWBD and injury, significantly increased during winter storm with PO in NYS, with the magnitudes of the RRs ranged from 1.04 to 1.90. These findings are similar to other reports in the literature (Dominianni et al., 2018; Conlon et al., 2011; Liu et al., 2021). Our sensitivity analysis by comparing the associations in early and later periods also showed similar results. Dominianni et al. (Dominianni et al., 2018) and Conlon et al. (2011) found that cold-weather outages in NY City were significantly associated with all-cause mortality (cumulative risk ratio (CRR) = 1.06) and CVD admissions (CRR = 1.14), but not in warm-weather outages. By comparing “high-fall days” with control days within the same winter season in Philadelphia, Gevitz et al. (2017) stated that snow was a major predictor of a high-fall day with an adjusted odds ratio (OR) of 13.4 (95% CI: 2.9–61.5), which was significantly higher than rain days (OR: 7.1, 95% CI: 0.6–87.4) and fog days (OR: 2.3, 95% CI: 0.6–8.9), respectively. However, the large confidence interval of the risk estimate implies small sample size. In addition, Burstein et al. (2015) examined mortality in Erie County, NY after receiving unprecedented amounts of snowfall in November 2014, and found that all 11 deaths occurred in areas that received the most snow. Therefore, our study provided additional, robust and consistent evidence of the adverse effects of winter storm and PO on multiple health endpoints.

A question that might be important is: were the immediate health effects we identified from the winter storms attributed to PO, snow/ice storms or both? We found that the health effects from both winter storms and PO combined were much greater than those from winter storms alone or PO alone, occurring in all five groups of diseases. Winter storms alone were significantly associated with most health outcomes, but there were no significant associations between PO alone and most health endpoints. Unfortunately, no prior research was available in the same areas for comparison, although a previous study found a significant increase of respiratory admissions after the joint effects of the Northeast blackout and heatwave in 2003, but not heat-wave alone (Lin et al., 2011). The concurrence of both winter storms and PO jointly may indicate the biggest winter storms, that produced PO and were strong enough to cause many environmental and health problems, but those PO alone events may represent isolated, small PO incidences. One important issue is that many winter storms contain mixed precipitation that can lead to substantial ice buildup on trees and power lines, along with strong winds. These factors could lead to much more widespread difficulties in restoring power.

4.2. Comparing snowstorm vs. ice-storm

Although snowstorms were eight times as frequent as ice storms, we found that ice storms had much higher health risks than snowstorms with the RRs ranging from 1.04 to 3.15 for all the outcomes studied. Consistent with our findings, Rajaram et al. (2016) found significant increased risks of ED visits due to ice storms in Canada (RR:2.52, 95% CI: 2.21–2.87). Similarly, Smith and Nelson (1998) found significantly increased fractures and other injuries from falls after an ice storm in Indiana, US. A study by Morency et al. (2012) demonstrated an increase in outdoor falls in a densely populated urban setting in Laval and Montreal Island of Canada, after freezing rain. The possible mechanism of ice storms affecting multiple health outcomes may be due to icy roads or sidewalks, which may increase frequency and seriousness of traffic-related accidents and injuries. Sidewalk cleaning or snow shoveling may also increase the risk of heart attack or asthma episodes (Alter, 2017; Boulet and O’Byrne, 2015).

4.3. Winter storms-health associations by diseases and lag

We found significant association between winter storm/PO and all five outcomes studied, with the higher effect in FWBD, followed by LRD and respiratory infection. Consistent with our findings, Rajaram et al. (2016) found significantly increased RR of ED visits due to injuries after the 2013 December ice storm in Toronto and Ottawa, Canada. Furthermore, Dominianni et al. (Dominianni et al., 2018) identified significantly increased risks of CVD and respiratory hospitalizations during cold-weather in NYC, USA. The study by Gevitz et al. (2017) reported that snow was a predictor of increased injuries due to accidental falls. Unfortunately, there is little prior literature regarding FWBD and respiratory infection after winter-storm/PO to compare with our study. Compared to pre-event period, we found increase of FWBD due to bacterial poisoning and intestinal infection aftermath of storm/PO. The major causes for injuries increase after winter storm include accidental falls and accidents due to weather.

In terms of lag days, our study found that the increased health effects showed immediately or the same day effect after storm/PO for all diseases (Lag 0–2 days), especially during ice storms. Similarly, Dominianni et al. (Dominianni et al., 2018) found that the highest effect on respiratory diseases and CVD from winter PO occurred on the same day or one day later. More importantly, we also found a significantly lasting effects of winter storms or snowstorm without PO associated with most diseases, e.g., 0–6 day.

4.4. Time windows of increased vulnerability

We interestingly found that the largest increases in hospital admissions after winter-storms/PO occurred in winter transitional months, such as October and November, and then March and April, but not in typical winter months (Dec.–Feb.). To the best of our knowledge, no prior studies evaluated similar issues, resulting in the inability to compare our results. The health risks of winter storms on multiple diseases by season may be explained by population adaptability behaviors. During winter months, the human body has adjusted to cold temperatures or snow, and people most likely stayed indoors, kept heating on, or dressed appropriately due to winter storms or the result of Winter Warnings. However, if snow or ice storms come early (October/November) or late (April) when there are leaves on the trees, it can cause PO because of snow accumulating on the leaves. The weight of ice on power lines and trees is the main cause of PO in ice storms, and wind is a secondary cause for additional PO. Additionally, people may not turn on heating or wear enough for sudden cold exposure (during beginning or end of winter months), resulting in increased susceptibility to winter storms impact. More interestingly, we found that peak admission time occurred from 3:00–8:00 p.m., that provides crucial information to clinical facilities for hospital bed planning and the general public for altering work schedules during snow days.

4.5. Potential biological plausibility

The environmental hazards that occur after the storms might have integrated with each other and triggered the disease episodes among patients with existing diseases, which is consistent with our findings regarding the increased co-morbidities and medical costs after winter storms. More specifically, severe ice storms or snowstorms may down power lines, causing PO which could further limit drinking water supply, and may shutdown sewage removal and treatment system. This was particularly true for elderly residents living in an urban high-rise apartment, many of whom were unable to descend stairways and became trapped for days or even weeks as a result of nonfunctioning elevators during periods of prolonged PO (Urban Green Building Coucil, 2014). Furthermore, cold weather is associated with increased heart rate, systemic vascular resistance, blood pressure, platelet aggregation, red blood cell counts and plasma cholesterol (Cheng and Su, 2010; Hong et al., 2003; Lin et al., 2016), especially among elderly (Ikaheimo et al., 2014). Oxygen demand, which is critical for asthma or COPD, is also elevated due to enhancing blood pressure caused by cold stress that increases systemic vascular resistance (Cheng and Su, 2010).

4.6. Strengths and limitations

To our knowledge, this may be the first study evaluating the immediate effects of winter storms on multiple health outcomes, especially in a large state like New York with over 2.6 million hospitalization cases annually in multiple biological systems, multiple sub-climate regions and a diverse racial/ethnic population; this increases the generalizability of our findings. Our study may contribute to new knowledge such as assessing the individual and joint effects of winter storms and PO, separating the effects of snowstorms from ice storms, understanding both lag and cumulative effects of winter storms/PO on multiple diseases and new indicators, and identifying admission peaks during winter, transitional months and peak timing. In terms of potential limitations, this study has only included the most severe hospital admission cases, the tip of the iceberg. Therefore, the total immediate effects of winter storms on all five groups of cases could have been underestimated. However, acute episodes of most diseases studied, including CVD, severe respiratory diseases (asthma, COPD and pneumonia) and injuries, require immediate medical attention and treatment through hospital admission. Therefore, hospitalization records could be a good data source of assessing urgent care for these cases. Another concern is the potentially meteorological confounding effects, but temperature, relative humidity, seasonal effect and various time-varying variables have been controlled in DLNM model. Additionally, potential effect modifications by spatial differences or temporal changes in sociodemographic characteristics on the associations examined may be a concern. However, the multiple comparisons used, with-/without-winter storms/PO comparison during the same season, and the comparison within the same area, should have controlled for the confounding or interactive effects of regional SES characteristics, seasonal factors, and PO effect at the divisional level. Nevertheless, we could not further adjust for the SES variations while pooling divisional RRs at state level because the information on these variables were not available at the divisional level. We have used the following strategies to address this limitation: 1) We have checked the NYS census data and found that the SES distribution has not significantly changed over the study period; 2) We conducted a sensitivity analyses of re-assessing the exposure-health associations by stratifying multiple SES variables at individual level, and found that the average of the RRs may not be significantly affected by SES as the associations in various SES groups were not significantly different; 3) Further, as our stratification analysis between earlier and later periods showed similar results, this may suggest that the temporal impact of SES change is minimal; and 4) In case if certain residual influences by SES occurred in some areas, which should not be related to winter storm or PO, their impacts should be non-differential, and the bias would be towards the null. We also excluded the 13-days of Hurricane Sandy period and all relevant cases occurring during that period from our study to minimize confounding bias. Other confounders such as personal activity pattern and heating use were not available, which could affect the validity of this study. However, more people may stay indoors during a winter storm, and potential change of personal activity pattern may make the bias toward the null, i.e., underestimated the actual association found.

4.7. Implications of our findings

As winter storms are the most prevalent weather event accompanying PO, understanding their health impacts may help clinical facility and public health agencies for preparedness and planning early intervention. Although further confirmation is needed, our findings regarding the immediate and lagged impact of winter storms on health outcomes, the magnitudes of the risks of different diseases and other critical care indicators (e.g., comorbidities and medical cost) may provide new insight to clinical facilities for hospital beds and treatment planning. Physicians should caution their patients about the potential risks associated with shoveling snow, driving or walking in icy conditions or staying outdoors during winter storms. Additionally, there should be greater awareness of having sufficient food, water and medication in advance of possible winter storms in case there is a loss of power. Furthermore, the higher health burdens during an ice storm and with PO, peak at late afternoon and stronger transitional season effects could be important for public health and forecasting agencies in order to target high risk time window and issuing early warnings. Additionally, education on preventive behaviors, such as keeping extra clothing and blankets in vehicles and avoid driving during rush or peak hour should be provided to the general public and vulnerable populations such as elderly.

5. Conclusion

We found that winter storms had significantly increased hospital admissions for multiple health outcomes, comorbidities and medical costs which were stronger when accompanying PO and during an ice storm. The highest health risks mostly occurred on the same day, but adverse impacts persisted for a whole week. The health burdens peaked at 3:00–8:00 p.m., higher in transitional months. While further research is needed, this study provides useful information to clinical facilities, public health agency and the general public on behavior modification and disease prevention during winter storm events.

Supplementary Material

supplemental materials

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.envres.2021.110924.

Acknowledgement

This work was supported by Grant # 1R15ES02800001A1 from the National Institute of Environmental Health Sciences of United States. We thank the New York State Department of Health for providing the comprehensive health data (data sharing protocol number: 1509-01 A) and thank the New York Department of Public Service for providing the statewide power outage data.

Footnotes

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.

Credit author statement

Dr. Shao Lin has full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Concept and design: Lin, Romeiko, Sheridan, Hao. Acquisition, analysis, or interpretation of data: Lin, Zhang, Mongillo, Chowdhury. Drafting of the manuscript: Lin, Zhang, Mongillo. Critical revision of the manuscript for important intellectual content: Birkhead, DiRienzo, Stuart, Stern, Primeau, Dong, Wu. Statistical analysis: Zhang. Supervision: Lin.

References

  1. Alter DA, 2017. The eco-biological-behavioural perfect storm that follows heavy snowfall. Cmaj 189, E225–E226. 10.1503/cmaj.170022. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Anderson GB, Bell ML, 2012. Lights out: impact of the August 2003 power outage on mortality in New York, NY. Epidemiology 23, 189–193. 10.1097/EDE.0b013e318245c61c. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Boulet LP, O’Byrne PM, 2015. Exercise-induced bronchoconstriction in athletes. N. Engl. J. Med 327 10.1378/chest.128.6.3966. [DOI] [PubMed] [Google Scholar]
  4. Burstein GR, Blank J, Chalmers TF, Mahar T, Mahoney MC, 2015. Morbidity and mortality weekly report notes from the field. MMWR (Morb. Mortal. Wkly. Rep.) [DOI] [PubMed] [Google Scholar]
  5. Call DA, 2009. Changes in ice storm impacts over time: 1886–2000. Weather. Clim. Soc 2, 23–35. 10.1175/2009wcas1013.1. [DOI] [Google Scholar]
  6. Centers for Disease Control and Prevention, 1998. Community Needs Assessment and Morbidity Surveillance Following an Ice Storm — Maine, January 1998 47. [PubMed] [Google Scholar]
  7. Cheng X, Su H, 2010. Effects of climatic temperature stress on cardiovascular diseases. Eur. J. Intern. Med 21, 164–167. 10.1016/j.ejim.2010.03.001. [DOI] [PubMed] [Google Scholar]
  8. Conlon KC, Rajkovich NB, White-Newsome JL, Larsen L, O’Neill MS, 2011. Preventing cold-related morbidity and mortality in a changing climate. Maturitas 69, 197–202. 10.1016/j.maturitas.2011.04.004. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Degelia SK, Christian JI, Basara JB, Mitchell TJ, Gardner DF, Jackson SE, Ragland JC, Mahan HR, 2016. An overview of ice storms and their impact in the United States. Int. J. Climatol 36, 2811–2822. 10.1002/joc.4525. [DOI] [Google Scholar]
  10. Dominianni C, Lane K, Johnson S, Ito K, Matte T, 2018. Health impacts of citywide and localized power outages in New York city. Environ. Health Perspect. 126, 67003. 10.1289/EHP2154. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Fitzgerald EF, Pantea C, Lin S, 2014. Cold Spells and the Risk of Hospitalization for Asthma, pp. 947–954. 10.1007/s00408-014-9645-y. New York, USA: 1991–2006. Lung 192. [DOI] [PubMed] [Google Scholar]
  12. Garcia VC, Gego E, Lin S, Pantea C, Rappazzo K, Wootten A, Rao ST, 2010. An evaluation of transported pollution and respiratory–related hospital admissions in the state of New York. Atmos. Pollut. Res 2, 9–15. 10.5094/apr.2011.002. [DOI] [Google Scholar]
  13. Gasparrini A, 2013. Distributed Lag Linear and Non-linear Models for Time Series Data. http//143107.2013 212. Doc. Is Available R Proj. Https//cranR-Projectorg/web/packages/dlnm/. (Accessed 4 May 2015). [Google Scholar]
  14. Gasparrini A, 2011. Distributed lag linear and non-linear models in R: the package dlnm. J. Stat. Software 43, 1–20. [PMC free article] [PubMed] [Google Scholar]
  15. Gasparrini A, Guo Y, Hashizume M, Lavigne E, Zanobetti A, Schwartz J, Tobias A, Tong S, Rocklöv J, Forsberg B, Leone M, De Sario M, Bell ML, Guo YLL, Wu CF, Kan H, Yi SM, De Sousa Zanotti Stagliorio Coelho M, Saldiva PHN, Honda Y, Kim H, Armstrong B, 2015. Mortality risk attributable to high and low ambient temperature: a multicountry observational study. Lancet 386, 369–375. 10.1016/S0140-6736(14)62114-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Gevitz K, Madera R, Newbern C, Lojo J, Johnson CC, 2017. Risk of fall-related injury due to adverse weather events, Philadelphia, Pennsylvania. Publ. Health Rep 132, 53S–58S. 10.1177/0033354917706968,2006-2011. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Hong Y-C, Rha J-H, Lee J-T, Al E, 2003. Ischemic stroke associated with decrease in temperature. Epidemiology 14, 473–478. 10.1097/01.ede.0000078420.82023.e3. [DOI] [PubMed] [Google Scholar]
  18. Hopke PK, Croft DP, Zhang W, Lin S, Masiol M, Squizzato S, Thurston SW, van Wijngaarden E, Utell MJ, Rich DQ, 2020. Changes in the hospitalization and ED visit rates for respiratory diseases associated with source-specific PM2.5 in New York State from 2005 to 2016. Environ. Res 181, 108912. 10.1016/j.envres.2019.108912. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Hussain A, Bui V-H, Kim H-M, 2019. Microgrids as a resilience resource and strategies used by microgrids for enhancing resilience. Appl. Energy 240, 56–72. 10.1016/j.apenergy.2019.02.055. [DOI] [Google Scholar]
  20. Ikaheimo TM, Lehtinen T, Antikainen R, Jokelainen J, Nayha S, Hassi J, Keianen-Kiukaanniemi S, Laatikainen T, Jousilahti P, Jaakkola JJK, 2014. Cold-related cardiorespiratory symptoms among subjects with and without hypertension: the National FINRISK Study 2002. Eur. J. Publ. Health 24, 237–243. 10.1093/eurpub/ckt078. [DOI] [PubMed] [Google Scholar]
  21. Lin S, Fletcher BA, Luo M, Chinery R, Hwang S-AA, 2011. Health impact in New York city during the northeastern blackout of 2003. Publ. Health Rep 126, 384–393. 10.1177/003335491112600312. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Lin S, Soim A, Gleason KA, Hwang S-A, 2016. Association between low temperature during winter season and hospitalizations for ischemic heart diseases in New York state. J. Environ. Health 78, 66–74. [PubMed] [Google Scholar]
  23. Lin Y, Bie Z, Qiu A, 2018. A review of key strategies in realizing power system resilience. Glob. Energy Interconnect. 1, 70–78. 10.14171/j.2096-5117.gei.2018.01.009. [DOI] [Google Scholar]
  24. Liu X, He Y, Tang C, Wei Q, Xu Z, Yi W, Pan R, Gao J, Duan J, Su H, 2021. Association between cold spells and childhood asthma in Hefei, an analysis based on different definitions and characteristics. Environ. Res 195, 110738. 10.1016/j.envres.2021.110738. [DOI] [PubMed] [Google Scholar]
  25. Morency P, Voyer C, Burrows S, Goudreau S, 2012. Outdoor falls in an Urban context: winter weather impacts and geographical variations. Can. J. Public Health 103, 218–222. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Oboudi MH, Mohammadi M, Rastegar M, 2019. Resilience-oriented intentional islanding of reconfigurable distribution power systems. J. Mod. Power Syst. Clean Energy 7, 741–752. [Google Scholar]
  27. Phung D, Thai PK, Guo Y, Morawska L, Rutherford S, Chu C, 2016. Ambient temperature and risk of cardiovascular hospitalization: an updated systematic review and meta-analysis. Sci. Total Environ 550, 1084–1102. 10.1016/j.scitotenv.2016.01.154. [DOI] [PubMed] [Google Scholar]
  28. Piana G, Ricciardi M, Bella F, Cucciniello R, Proto A, Gerbaldi C, 2020. Poly (glycidyl ether)s recycling from industrial waste and feasibility study of reuse as electrolytes in sodium-based batteries. Chem. Eng. J 382, 122934. 10.1016/j.cej.2019.122934. [DOI] [Google Scholar]
  29. Piercefield E, Wendling T, Archer P, Mallonee S, 2011. Winter storm-related injuries in Oklahoma, January 2007. J. Saf. Res 42, 27–32. 10.1016/j.jsr.2010.11.004. [DOI] [PubMed] [Google Scholar]
  30. Rajaram N, Hohenadel K, Gattoni L, Khan Y, Birk-Urovitz E, Li L, Schwartz B, 2016. Assessing health impacts of the december 2013 ice storm in Ontario, Canada. BMC Publ. Health 16. 10.1186/s12889-016-3214-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Rich DQ, Zhang W, Lin S, Squizzato S, Thurston SW, van Wijngaarden E, Croft D, Masiol M, Hopke PK, 2019. Triggering of cardiovascular hospital admissions by source specific fine particle concentrations in urban centers of New York State. Environ. Int 126, 387–394. 10.1016/J.ENVINT.2019.02.018. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Smith RW, Nelson DR, 1998. Fractures and Other Injuries from Falls after an Ice Storm. [DOI] [PubMed] [Google Scholar]
  33. Urban Green Building Coucil, 2014. Baby it ‘ S Cold inside. [Google Scholar]
  34. Viechtbauer W, 2010. Conducting meta-analyses in R with the metafor package. J. Stat. Software 36, 1–48. [Google Scholar]
  35. Wang P, Zhang S, Pu Y, Cao S, Zhang Y, 2021. Estimation of photovoltaic power generation potential in 2020 and 2030 using land resource changes: an empirical study from China. Energy 219, 119611. 10.1016/j.energy.2020.119611. [DOI] [Google Scholar]
  36. Wijngaarden E. van, Rich DQ, Zhang W, Thurston SW, Lin S, Croft DP, Squizzato S, Masiol M, Hopke PK, 2020. Neurodegenerative hospital admissions and long-term exposure to ambient fine particle air pollution. Ann. Epidemiol [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Xiang J, Hansen A, Liu Q, Liu X, Tong MX, Sun Y, Cameron S, Hanson-Easey S, Han G-S, Williams C, Weinstein P, Bi P, 2017. Association between dengue fever incidence and meteorological factors in Guangzhou, China, 2005–2014. Environ. Res 153, 17–26. 10.1016/j.envres.2016.11.009. [DOI] [PubMed] [Google Scholar]
  38. Xiao J, Niu B, Song Q, Zhan L, Xu Z, 2021. Novel targetedly extracting lithium: an environmental-friendly controlled chlorinating technology and mechanism of spent lithium ion batteries recovery. J. Hazard Mater 404, 123947. 10.1016/j.jhazmat.2020.123947. [DOI] [PubMed] [Google Scholar]
  39. Xiao X, Gasparrini A, Huang J, Liao Q, Liu F, Yin F, Yu H, Li X, 2017. The exposure-response relationship between temperature and childhood hand, foot and mouth disease: a multicity study from mainland China. Environ. Int 100, 102–109. 10.1016/J.ENVINT.2016.11.021. [DOI] [PubMed] [Google Scholar]
  40. Zarzycki CM, 2018. Projecting changes in societally impactful northeastern. U.S. Snowstorms. Geophys. Res. Lett 45 (12) 10.1029/2018GL079820, 067–12,075. [DOI] [Google Scholar]
  41. Zhang Wangjian, Du Z, Zhang D, Yu S, Hao Y, 2016a. Quantifying the adverse effect of excessive heat on children: an elevated risk of hand, foot and mouth disease in hot days. Sci. Total Environ 541, 194–199. 10.1016/j.scitotenv.2015.09.089. [DOI] [PubMed] [Google Scholar]
  42. Zhang W, Du Z, Zhang D, Yu S, Huang Y, Hao Y, 2016b. Assessing the impact of humidex on HFMD in Guangdong Province and its variability across social-economic status and age groups. Sci. Rep 6 10.1038/srep18965. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Zhang W, Lin S, Hopke PK, Thurston SW, van Wijngaarden E, Croft D, Squizzato S, Masiol M, Rich DQ, 2018. Triggering of cardiovascular hospital admissions by fine particle concentrations in New York state: before, during, and after implementation of multiple environmental policies and a recession. Environ. Pollut 242, 1404–1416. 10.1016/j.envpol.2018.08.030. [DOI] [PubMed] [Google Scholar]
  44. Zhang W, Sheridan SC, Birkhead GS, Croft DP, Brotzge JA, Justino JG, Stuart NA, Du Z, Romeiko XX, Ye B, Dong G, Hao Y, Lin S, 2020. Power outage: an ignored risk factor for COPD exacerbations. Chest 158, 2346–2357. 10.1016/j.chest.2020.05.555. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

supplemental materials

RESOURCES