Maternal and Child Health Journal

, Volume 15, Issue 7, pp 1088–1096

Preterm Birth During an Extreme Weather Event in Québec, Canada: A “Natural Experiment”

Authors

    • Institut national de santé publique du Québec
    • Research Centre of the University of Montréal Hospital Centre
    • Department of Social and Preventive MedicineUniversity of Montréal
  • Erica Kuehne
    • Department of Environmental and Occupational HealthUniversity of Montréal
  • Marc Goneau
    • Institut national de santé publique du Québec
  • Mark Daniel
    • Research Centre of the University of Montréal Hospital Centre
    • Department of Social and Preventive MedicineUniversity of Montréal
    • School of Health SciencesUniversity of South Australia
Article

DOI: 10.1007/s10995-010-0645-0

Cite this article as:
Auger, N., Kuehne, E., Goneau, M. et al. Matern Child Health J (2011) 15: 1088. doi:10.1007/s10995-010-0645-0

Abstract

To clarify the relationship between preterm birth (PTB) and extreme weather events, we evaluated PTB during a January 1998 ice storm that had led to a provincial emergency in the middle of winter in the province of Québec, Canada. Singleton live births for three periods (1993–1997, 1998, 1999–2003) were obtained (N = 855,320). PTB was defined as gestational age <37 completed weeks. Births in the Triangle of Darkness, the area most strongly affect by the storm, were geocoded. Multivariate logistic regression was used to calculate the likelihood of PTB for the Triangle relative to metropolitan Montréal, adjusting for maternal age, education, civil status, maternal birthplace, and previous deliveries. Associations for 1998 relative to other periods were evaluated. Short-term (January–February) and long-term (March–October) exposure periods were examined. The proportion PTB for 1998 January–February births in the Triangle (8.7%) was high compared with 1998 March–October births (6.0%) and with the corresponding proportions for 1993–1997 (6.2%) and 1999–2003 (6.9%). Covariate-adjusted odds of PTB for January–February 1998 were 27% higher for the Triangle relative to metropolitan Montréal, though precision was low. Furthermore, adjusted odds were 28% higher for 1998 relative to 1999–2003, despite increasing rates of PTB over time. Odds were not elevated over a long-term exposure period. This study suggests a weak association between PTB and exposure to extreme weather for the two months following an ice storm, but not for later periods after the storm.

Keywords

ClimateEnvironment and public healthPremature birthStressful eventsStress, physiological

Introduction

The impact that climate change may have on human health is increasingly acknowledged [13]. There are, however, significant challenges for research attempting to address the relationship between climate change and health [25], particularly in documenting the chronic or long-term impacts on population health as factors unrelated to climate change also influence health. One way to address this problem is to examine the acute effects of extreme weather events on health outcomes [6]. Extreme weather events are predicted to increase as climate change becomes more pronounced [2, 3, 7]. The relationship with population health is not well understood. Determining the impact of extreme weather events can inform public health and emergency services planning for the future.

The influence on health of extreme weather is straightforward to evaluate in pregnancy as the exposure or induction period is better defined than exposures that occur across a life span. Exposures during pregnancy are limited to a nine month period, not decades which is the case for many chronic disease outcomes. Preterm birth (PTB) is an adverse outcome potentially sensitive to extreme weather events, and little research has examined how extreme weather may pattern PTB. The determinants of PTB are poorly understood [8, 9], and it is concerning that PTB, an important contributor to infant morbidity and mortality, is increasing in many countries [1014]. A small study on the impact of Hurricane Katrina on PTB in the US found no statistically significant association [15]. Nonetheless, evidence suggests that anxiety and stressful life events may influence PTB [1620], and it is plausible that stress mechanisms may operate during extreme weather events. Larger, better powered studies are therefore warranted, and needed.

We evaluated PTB during an environmental crisis in the province of Québec, Canada. In January 1998, a severe ice storm struck parts of southern Québec, causing widespread destruction and leaving areas without power for up to four weeks in the middle of winter [2123]. This extreme weather event that led to a provincial emergency [23, 24] provides a rare opportunity to evaluate climate-related determinants of PTB [25], a perinatal health outcome widely used in population health research. The objective of this study was to determine, for the areas affected by the ice storm, whether the likelihood of PTB was elevated relative to other periods under stable environmental conditions.

Materials and Methods

Data and Variables

Singleton live born infants from the Québec birth file were extracted for the period extending from 1993 to 2003 (N = 855,320). These years correspond to the decade centered on 1998, the year of the ice storm. The storm began on the 4th of January, a Sunday [22, 23]. Births during the first three days of 1998 (446 births from January 1 to 3) were recoded as December 1997 births as they were not affected by the storm. Data were analysed for three periods (1993–1997, 1998 and 1999–2003) corresponding to intervals prior to the storm, the year of the storm, and well after the storm. As PTB is known to be increasing in Québec and elsewhere [10, 12], the pre- and post-storm periods were not pooled but compared to 1998 separately.

The areas most strongly affected by the ice storm were primarily those bounded by the municipalities St-Hyacinthe, Granby and St-Jean-sur-Richelieu in the region of Monterégie, located southeast of Montréal [24]. These areas form a triangle referred to at the time as the “Triangle of Darkness” given a complete lack of electricity during the initial four weeks following the ice storm [24]. The mean temperature in January 1998 was approximately −7°C, with lows passing −23°C [26]. To identify mothers residing in the Triangle, we geocoded 6-digit residential postal codes into latitude and longitude coordinates using the Postal Code Conversion File of Statistics Canada obtained from the Public Health Agency of Canada [27]. Postal codes were missing or invalid for 949 births (0.1%) subsequently excluded from analyses. MapInfo Professional 8.0 Geographic Information System software was used to delineate the triangle formed by lines connecting the outer 5-km radius around the estimated population centroid of the three municipalities, and to identify postal codes situated within the Triangle (Fig. 1). Live births to mothers residing in one of these postal codes were defined as Triangle of Darkness births.
https://static-content.springer.com/image/art%3A10.1007%2Fs10995-010-0645-0/MediaObjects/10995_2010_645_Fig1_HTML.gif
Fig. 1

Location of the Triangle of Darkness, Outer Triangle, and Montréal metropolitan area. Borders of the Montréal metropolitan area are based on 2001 Statistics Canada boundaries

The same procedure was used to identify births within a 10-km buffer surrounding the Triangle, labelled the Outer Triangle. The Outer Triangle was considered separately as it may have been affected by the ice storm to a lesser extent. The metropolitan area of Montréal, also affected by the ice storm though to a much lower degree with less widespread power outages, was also analysed. Other urban and rural areas of Québec were evaluated as separate categories [28]. Births in Montréal, other urban centres and rural areas were identified using the Postal Code Conversion File [27]. Five areas (Triangle of Darkness, Outer Triangle, Montréal, other urban centres, and rural areas) were therefore assessed.

PTB was expressed as dichotomous variable, defined as gestational age less than 37 completed weeks [29]. We did not evaluate more extreme gestational age cut-points (i.e., PTB less than 32 weeks) because of insufficient numbers of such cases in the year 1998 compared with the pre- and post-storm periods. In Québec, gestational age is calculated from dating ultrasound examinations, not recall of last menstruation period [29]. Hence misclassification of PTB based on recall, a major limitation of PTB research in the US [30, 31], is likely to be low for this study. Gestational age was missing for 988 excluded births (0.1%), resulting in a final sample of 853,383 births for analysis.

Covariates included maternal age (<20, 20–34, ≥35 years), education (no high school diploma, high school diploma, some postsecondary, some university or more, unknown), civil status (legally married, cohabiting, single, unknown), maternal birthplace (Canadian-born, foreign-born, unknown), and previous deliveries (0, 1, 2 or more births). These covariates were considered potential confounders based on previous research [8, 9, 29].

Statistical Analysis

Descriptive statistics were computed. Multivariate logistic regression was used to calculate odds ratios (OR) and 95% confidence intervals (CI) for the association between area, period and PTB. Initial models adjusted for area and period, and subsequent models adjusted for area, period, education, age, civil status, maternal birthplace, and previous deliveries. Period was examined with 1999–2003 as the referent because the prevalence of PTB was expected to be higher at the end of the study period [10, 12]; hence, an elevated OR for 1998 relative to 1999–2003 would be more meaningful than an OR for 1998 relative to 1993–1997. Area was examined with Montréal as the referent, given the large number of births in this category. We did not use time series analytic methods as we did not evaluate counts of PTB cases, but rather the occurrence of PTB relative to term births (a dichotomous outcome). Compared with individual-level analyses, time series analyses may underestimate effects and are less easily to adapt to evaluate long-term effects of exposures [32].

Extreme weather events may have immediate (short-term) or delayed (long-term) effects on PTB. We evaluated short-term effects for births during the 8-week span after the start of the storm (January–February), and long-term effects for births during the subsequent 8-month span (March–October). Analyses were also repeated for a slightly longer span covering March–December births. Although the later births (November–December) reflect conceptions that occurred after January 1998, the pathways linking extreme weather to PTB may begin pre-conception. Final models were re-run after excluding extreme PTB cases defined as gestational age less than 22 weeks (n = 375) [33].

Statistical analyses were performed using SAS version 9.1 (SAS Institute Inc, Cary, North Carolina) and SPSS version 16.0 (SPSS Inc., Chicago, Illinois). This study was conducted under the Québec population health surveillance plan mandated by the health ministry and approved by the Public Health Ethics Committee.

Results

Just over 6% of births overall were preterm (Table 1). PTB proportions varied little across areas with the exception of the Outer Triangle where the proportion was 1% lower relative to other areas. The proportion PTB was greater for 1998 (6.4%) than for 1993–1997 (6.0%; P < 0.001), but not for 1998 relative to 1999–2003 (6.3%; P = 0.3).
Table 1

Proportion preterm birth (PTB) according to maternal characteristics, singleton infants, Québec, 1993–2003

 

% PTB

N

Area

 Triangle of Darkness

6.4

22,650

 Outer Triangle

5.3

18,530

 Montréal metropolitan area

6.1

418,945

 Other urban centres

6.4

227,191

 Rural area

6.3

166,067

Period

 1993–1997

6.0

424,066

 1998

6.4

73,423

 1999–2003

6.3

355,894

Category of education

 No high school diploma

7.9

112,382

 High school diploma

6.9

91,542

 Some postsecondary

6.2

193,780

 Some university or more

5.2

403,024

Age

 <20 years

8.5

38,284

 20–34 years

5.9

710,043

 ≥35 years

7.0

105,056

Civil status

 Married

5.5

390,099

 Cohabiting

6.4

374,928

 Single

8.5

71,687

Maternal birthplace

 Canadian-born

6.2

716,435

 Foreign-born

6.3

122,472

Previous deliveries

 0

7.0

383,614

 1

5.2

303,515

 2 or more

6.1

166,254

Totala

6.2

853,383

aMay not consistently sum to total given missing data

Table 2 shows PTB proportions over time according to area. PTB proportions tended to be slightly higher in 1998 compared with 1999–2003 for all three areas affected by the ice storm (Triangle of Darkness, Outer Triangle, Montréal). This was not the case for areas less touched by the storm (other urban and rural areas) where PTB proportions were lower in 1998 relative to 1999–2003. This general pattern held for births during the first two months of the year (January–February), as well as for later months (March–October). The highest PTB proportion was 8.7% in the Triangle of Darkness during the first two months of 1998 (the year of the ice storm)—the proportion was lower (6.0%) for births from March–October. The proportion was also relatively higher in the Outer Triangle during the first two months of 1998 (6.5%), compared with births from March–October (5.1%).
Table 2

Proportion preterm birth (PTB) according to area and period for January–February and March-October births

 

Period

1993–1997

1998

1999–2003

% PTB

N

% PTB

N

% PTB

N

January–February

 Triangle of Darkness

6.2

1699

8.7

299

6.9

1368

 Outer Triangle

5.3

1346

6.5

247

5.9

1134

 Montréal metropolitan area

6.8

30,676

6.8

5377

6.4

26,119

 Other urban centres

6.5

17,127

6.9

3002

7.2

14,122

 Rural areas

6.4

12,659

6.1

2168

7.0

10,370

March–October

 Triangle of Darkness

6.1

8253

6.0

1512

6.3

6819

 Outer Triangle

4.8

6790

5.1

1203

5.4

5550

 Montréal metropolitan area

5.8

150,391

6.2

26,507

6.0

129,254

 Other urban centres

6.1

83,450

6.2

14,258

6.4

67,882

 Rural areas

5.7

60,207

6.1

10,588

6.4

49,036

Table 3 shows results from non-stratified regression models for the association between period and PTB, and area and PTB. The odds of PTB increased over time, and tended to be lower for the Outer Triangle (adjusted ORMar-Oct 0.88, 95% CI 0.82–0.94) and higher for other urban areas (adjusted ORMar-Oct 1.05, 95% CI 1.03–1.07) relative to Montréal. The odds of PTB in the Triangle of Darkness were similar to Montréal after accounting for covariates.
Table 3

Association between period, area and preterm birth for January–February and March-October births

 

Partially adjusted OR (95% CI)a

Fully adjusted OR (95% CI)b

JanuaryFebruary

Area

 Triangle of Darkness

1.01 (0.88–1.17)

0.96 (0.83–1.11)

 Outer Triangle

0.84 (0.71–1.00)

0.85 (0.72–1.01)

 Montréal metropolitan area

1

1

 Other urban centres

1.03 (0.98–1.08)

1.02 (0.97–1.08)

 Rural area

1.00 (0.94–1.06)

0.97 (0.91–1.03)

Period

 1993–1997

0.97 (0.93–1.02)

0.99 (0.94–1.03)

 1998

1.00 (0.92–1.09)

1.00 (0.93–1.09)

 1999–2003

1

1

MarchOctober

Area

 Triangle of Darkness

1.05 (1.00–1.11)

1.01 (0.95–1.06)

 Outer Triangle

0.86 (0.80–0.91)

0.88 (0.82–0.94)

 Montréal metropolitan area

1

1

 Other urban centres

1.05 (1.03–1.07)

1.05 (1.03–1.07)

 Rural area

1.03 (1.00–1.05)

1.00 (0.98–1.03)

Period

 1993–1997

0.95 (0.93–0.96)

0.95 (0.94–0.97)

 1998

1.00 (0.97–1.03)

1.00 (0.97–1.04)

 1999–2003

1

1

aOdds ratios (OR) and 95% confidence intervals (CI) adjusted for area and period

bORs and 95% CIs adjusted for area, period, education, age, civil status, maternal birthplace, and previous deliveries

Associations between PTB and period shown for areas separately are given in Table 4. When only January–February births were considered, no statistically significant associations were present for 1998 relative to 1999–2003. However, ORs for 1998 relative to 1999–2003 tended to be elevated in areas affected by the ice storm, especially the Triangle of Darkness where the odds were 28% higher for the year of the storm compared with the last period. This finding contrasts with a 12% lower odds of PTB for 1993–1997 relative to 1999–2003. ORs for 1998 relative to 1999–2003 were more weakly elevated for the Outer Triangle and Montréal. Patterns for other urban centres and rural areas less affected by the storm indicated lower odds of PTB in both 1998 and 1993–1997 relative to 1999–2003. Elevated ORs were not present for March–October births in any area.
Table 4

Association between period and preterm birth according to area for January–February and March–October births

 

Triangle of Darkness

Outer Triangle

Montréal

Other urban centres

Rural area

OR (95% CI)

OR (95% CI)

OR (95% CI)

OR (95% CI)

OR (95% CI)

January–February

 1993–1997

0.88 (0.65–1.19)

0.92 (0.64–1.31)

1.07 (1.00–1.15)

0.90 (0.82–0.98)

0.93 (0.83–1.03)

 1998

1.28 (0.81–2.02)

1.13 (0.64–2.00)

1.07 (0.95–1.20)

0.96 (0.82–1.12)

0.88 (0.72–1.06)

 1999–2003

1

1

1

1

1

March–October

 1993–1997

1.01 (0.88–1.16)

0.88 (0.74–1.04)

0.96 (0.93–0.99)

0.94 (0.90–0.98)

0.89 (0.84–0.93)

 1998

0.96 (0.76–1.22)

0.94 (0.71–1.25)

1.02 (0.97–1.08)

0.97 (0.90–1.05)

0.94 (0.86–1.02)

 1999–2003

1

1

1

1

1

Odds ratios (OR) and 95% confidence intervals (CI) adjusted for education, age, civil status, maternal birthplace and previous deliveries

Table 5 shows associations between area and PTB for each period separately. Associations were not statistically significant for 1998; however, the odds of PTB were 27% higher for the Triangle of Darkness relative to Montréal for January–February (no other ORs were comparably as high). The associations disappeared when births from March–October were considered. Importantly, the odds of PTB were not elevated in the Triangle of Darkness for other periods.
Table 5

Association between area and preterm birth according to period for January–February and March–October births

 

1993–1997

OR (95% CI)

1998

OR (95% CI)

1999–2003

OR (95% CI)

January–February

 Triangle of Darkness

0.91 (0.74–1.11)

1.27 (0.83–1.93)

0.97 (0.78–1.21)

 Outer Triangle

0.81 (0.63–1.03)

0.98 (0.58–1.66)

0.88 (0.69–1.14)

 Montréal metropolitan area

1

1

1

 Other urban centres

0.97 (0.89–1.05)

1.03 (0.86–1.24)

1.09 (1.00–1.19)

 Rural area

0.94 (0.86–1.02)

0.89 (0.72–1.10)

1.03 (0.94–1.14)

March–October

 Triangle of Darkness

1.04 (0.95–1.15)

0.95 (0.76–1.18)

0.99 (0.89–1.09)

 Outer Triangle

0.86 (0.76–0.96)

0.86 (0.66–1.12)

0.91 (0.80–1.02)

 Montréal metropolitan area

1

1

1

 Other urban centres

1.06 (1.02–1.10)

1.03 (0.94–1.12)

1.08 (1.04–1.12)

 Rural area

0.97 (0.93–1.01)

0.96 (0.87–1.06)

1.05 (1.00–1.09)

Odds ratios (OR) and 95% confidence intervals (CI) adjusted for education, age, civil status, maternal birthplace and previous deliveries

Regression models evaluating slightly longer-term effects for March–December births yielded similar results, as did models run for data excluding extreme PTB cases (data not shown).

Discussion

This study takes advantage of a rare opportunity to investigate the potential impact on PTB of a severe ice storm that led to prolonged blackouts, cold temperatures and a provincial emergency in the middle of winter [23, 24]. We found slightly higher odds of PTB in areas most strongly affected by the ice storm. While this study was underpowered in terms of number of births in areas most affected by the storm, our data are compelling because associations for 1998 relative to later years were elevated despite the increase in PTB over time. Furthermore, odds of PTB were elevated particularly during the first 2 months after the start of the storm, but not when more prolonged periods after the start of the storm were examined. This study suggests that extreme weather events may weakly influence PTB in their immediate aftermath, but that delayed (or longer-term) effects are unlikely.

Studies evaluating extreme climatic events have mainly concentrated on the effects of heat waves on mortality [34, 35]. Although some studies have examined the relationship between air pollution and PTB [36, 37], few have evaluated extreme climatic events such as severe storms. Research on the health effects of Hurricane Katrina has tended to focus on mental health issues including post-traumatic stress disorder and depression [15, 38, 39] or autism [40]. Just one study has examined the relationship between Hurricane Katrina and adverse birth outcomes [15]. That study, like ours, found the odds of PTB to be higher, but not statistically significantly so, for mothers with more pronounced hurricane exposure relative to those without such exposure [15]. Interestingly, one study found that neural tube defects were more common following Hurricane Gilbert in Jamaica, which the authors attributed to decreased folic acid intake from poor nutrition [41]. Similar factors might be related to PTB during storms (including the ice storm) when nutrition may be less adequate [42, 43], given that PTB has been linked to decreased folic acid intake [4446], though data do not always support this relationship [47]. Other research has linked flood disasters with spontaneous abortion [48], and earthquakes with birth defects [49, 50] and PTB [50].

The trimester of exposure may be related to adverse birth outcomes. A study of 40 pregnant women exposed to an earthquake found no association with length of gestation among mothers exposed during later trimesters of pregnancy, but shorter gestational lengths for those exposed during the first trimester [51]. This finding contrasts with our results as we found a higher odds of PTB in the two months immediately following the start of the ice storm (when mothers in all trimesters were exposed) and no evidence of elevated odds in later months (when mothers in the first and second trimesters were exposed). Four analyses of a cohort of women exposed during pregnancy to the 1998 Québec ice storm (with sample sizes below 150) reported problems in cognitive development of offspring among mothers exposed during the first and second trimesters of pregnancy, but no association for exposures during the third trimester [5255]. Similar findings were found with respect to dermatoglyphic asymmetry (a minor physical anomaly of fingerprints) in a small sample of children exposed to the same storm during the first trimester [56]. Although these studies indicate that maternal stress from extreme weather events during early pregnancy could influence fetal development, our findings suggest that the influence on gestational length may be weak.

Pregnant women exposed to the 1998 Québec ice storm have reported elevated levels of stress [57] either from direct exposure to cold weather or from indirect effects of coping with power outages. Stress is known to cause the release of corticotropin-releasing hormone leading to elevate cortisol levels that may increase PTB [58, 59]. Maternal catecholamine levels, another substance known to increase with stress, have also been associated with PTB [60]. However, it is unclear whether exposure to direct or indirect effects of extreme weather events could generate biologic effects in pregnant women [61]. Most research on PTB and stress focuses on psychosocial stressors including depression, anxiety, or financial insecurity [1618, 58, 62] for which associations are conflicting [19, 20]. Although stress-related studies on the Québec ice storm have tended to differentiate objective stress related to material effects from subjective stress related to hyperarousal [5257], there is little support for extending such observations to effects on PTB—the associations observed in this study were weak. The literature on war and terrorism, where stress mechanisms may be similar to those of extreme weather events, suggests no association with PTB [63, 64].

There are other routes by which extreme weather could influence health. Inadequate housing, lack of electrical power/telephone access, limited prenatal care, or poor nutrition due to decreased access to healthy food sources were problems during the ice storm [2124], and also during Hurricane Katrina [42, 43]. The use of heating sources such as wood stoves during the ice storm is known to have led to cases of carbon monoxide poisoning [21, 22, 24], and research on air pollution indicates that carbon monoxide poisoning can result in PTB [37]. Last, infection due to contamination of water supplies associated with power failures could also influence maternal-child health [21, 48, 65]. Once again, however, our results suggest that as a whole, the impact of these exposures on PTB may be small.

This study was limited by the relatively small number of births that occurred during the ice storm in the Triangle of Darkness (this being a function of the small population in the area) which limited statistical power. We could not account for characteristics not recorded in the birth file, including prenatal care, co-morbidity, nutrition or tobacco use, which may have influenced our results. We used postal codes to identify births to mothers living in the Triangle of Darkness, and nondifferential misclassification of exposure may have attenuated associations towards the null. Other parts of southern Québec that may have been less strongly affected by the ice storm could not be identified. We cannot be certain that mothers were residing at home during the ice storm. Last, the generalizability of our results to other settings is unclear. Populations less accustomed to severe winters might experience stronger effects than observed for Québec, where winters are usually severe.

Conclusions

While research on the health consequences of extreme weather events is in its infancy, climate change is occurring and its impact on public health, including perinatal health, has the potential to be large. PTB rates are at the same time increasing, and we do not understand the basis of this increase. Although our results suggest but a weak association between extreme weather events and PTB, this study bridges a gap in the literature by linking the environment to prematurity. Additional research is necessary to assess how climate-related environmental changes may influence PTB and other maternal-child health outcomes, in order to develop appropriate public health responses to extreme environmental emergencies and improve PTB rates in general. In addition, our findings point to the need to more specifically consider the needs of pregnant women in disaster planning.

Acknowledgments

This study was not funded by an external source.

Copyright information

© Springer Science+Business Media, LLC 2010