Global Association of Cold Spells and Adverse Health Effects: A Systematic Review and Meta-Analysis

Abstract

Background: There is substantial evidence that mortality increases in low temperatures. Less is known about the role of prolonged cold periods denoted as cold spells.

Objective: We conducted the first systematic review and meta-analysis to summarize the evidence on the adverse health effects of cold spells in varying climates.

Data sources and extraction: Four databases (Ovid Medline, PubMed, Scopus, Web of Science) were searched for all years and languages available. "Cold spell" was defined as an event below a temperature threshold lasting for a minimum duration of 2 days. Of 1,527 identified articles, 26 satisfied our eligibility criteria for the systematic review, and 9 were eligible for meta-analyses. The articles were grouped by the three main study questions into Overall-effect Group, Added-effect Group, and Temperature-change-effect Group.

Data synthesis: Based on random-effects models in the meta-analyses, cold spells were associated with increased mortality from all or all nonaccidental causes (summary rate ratio = 1.10; 95% CI: 1.04, 1.17 based on 9 estimates from five studies), cardiovascular diseases (1.11; 95% CI: 1.03, 1.19; 12 estimates from eight studies), and respiratory diseases (1.21; 95% CI: 0.97, 1.51; 8 estimates from four studies). Estimated associations were stronger for people ≥ 65 years of age (1.06; 95% CI: 1.00, 1.12) than for people 0-64 years of age (1.01; 95% CI: 1.00, 1.03). Study-specific effect estimates from a limited number of studies suggested an increased morbidity related to cold spells, but it was not possible to quantitatively summarize the evidence.

Conclusions: Cold spells are associated with increased mortality rates in populations around the world. The body of evidence suggests that cold spells also have other adverse health effects. There was substantial heterogeneity among the studies, which should be taken into account in the interpretation of the results.

Citation: Ryti NR, Guo Y, Jaakkola JJ. 2016. Global association of cold spells and adverse health effects: a systematic review and meta-analysis. Environ Health Perspect 124:12-22; http://dx.doi.org/10.1289/ehp.1408104.

Figure 1

Flow diagram of study selection process. Exclusion of articles was done in three consecutive steps based on a) irrelevance of the title and/or the abstract, b) the general eligibility criteria after close inspection of the full text, and c) the meta-analysis eligibility criteria (meta-analysis only).

Figure 2

Forest plots showing the association between cold spells and mortality from all nonaccidental causes (A), cardiovascular diseases (B), and respiratory diseases (C) for all ages. Effect estimates are displayed on a logarithmic scale. The three independent effect estimates for Xie et al. (2013) represent cold spells in three different cities. The three independent effect estimates for Huynen et al. (2001) represent cold spells occurring in different years. Additional details about the study-specific effect estimates are provided in the Supplemental Material, “Part 1” and Tables S1–S3. Abbreviations: df, degrees of freedom; I², total heterogeneity / total variability; Q, Q-statistic; RR, mortality rate ratio. ^aThe effect estimate was calculated using the effect estimates for nonaccidental mortality in the age strata 30–64 and ≥ 65 in the two-stage meta-analysis. ^bThe effect estimates are for mortality from all causes instead of all nonaccidental causes. ^cThe effect estimate was calculated using the effect estimates for nonaccidental mortality for females and males in the age stratum ≥ 25 years in the two-stage meta-analysis. ^dThe effect estimate was calculated using the effect estimates for ischemic heart disease and cerebrovascular diseases in the age strata 30 to ≥ 65 years in the two-stage meta-analysis. ^eThe effect estimate was calculated using the effect estimates for cardiac disease and cerebrovascular accident in the two-stage meta-analysis. ^fThe effect estimate is for mortality from COPD instead of all respiratory diseases.

Figure 3

Forest plots showing the association between cold spells and mortality from all non-accidental causes by sex for males (A) and females (B), all ages. Effect estimates are displayed on a logarithmic scale. The three independent effect estimates for Xie et al. (2013) represent cold spells in three different cities. Additional details about the study-specific effect estimates are provided in the Supplemental Material, “Part 1” and Tables S4–S5. Abbreviations: df, degrees of freedom; I², total heterogeneity / total variability; Q, Q-statistic; RR, mortality rate ratio. ^aThe effect estimate is for mortality from all causes instead of all nonaccidental causes.

Figure 4

Forest plots showing the association between cold spells and mortality from all non-accidental causes by age: 0–64 years (A) and ≥ 65 years (B). Effect estimates are displayed on a logarithmic scale. The three independent effect estimates for Xie et al. (2013) represent cold spells in 3 different cities. The two independent effect estimates for Revich and Shaposhnikov (2008) represent 2 different cold spells. Additional details about the study-specific effect estimates are provided in the Supplemental Material, “Part 1” and Tables S6–S7. Abbreviations: df, degrees of freedom; I², total heterogeneity / total variability; Q, Q-statistic; RR, mortality rate ratio. ^aThe effect estimate was calculated using the effect estimates for non-accidental mortality in the age strata 0–4, 5–44, and 45–64 years in the two-stage meta-analysis. ^bThe effect estimate is for the age stratum 30–64 years. ^cThe city-specific effect estimates were calculated using the city-specific effect estimates for nonaccidental mortality in the age strata 65–74 and ≥ 75 years in the two-stage meta-analysis. ^dThe effect estimate was calculated using the effect estimates for mortality from all causes in the age strata 65–74, 75–84, 85–94, and > 95 years in the two-stage meta-analysis. ^eThe cold spell-specific effect estimates are for the age stratum ≥ 75 years.