Association Between Low Temperature During Winter Season and Hospitalizations for Ischemic Heart Diseases in New York State
Abstract
Most prior research investigating the health effects of extreme cold has been limited to temperature alone. Only a few studies have assessed population vulnerability and compared various weather indicators. The study described in this article intended to evaluate the effects of cold weather on hospital admissions due to ischemic heart disease, especially acute myocardial infarction (AMI), and to examine the potential interactive effects between weather factors and demographics on AMI. The authors found that extremely low universal apparent temperature in winter was associated with increased risk of AMI, especially during lag4-lag6. Certain demographic groups such as the elderly, males, people with
Introduction
With climate change, extreme weather events such as heat waves and cold spells will become more frequent and longer in duration (
Even fewer studies have evaluated the association between cold temperature and 1HD hospital admissions as an indicator of morbidity. A comprehensive review by Bhaskaran and co-authors (2009) reported a statistically significant short-term increased risk of myocardial infarction (MI) at lower temperatures in winter in 8 of 12 studies; relative risks (RR) ranged from 1.01 to 1.40 and the strongest association was observed at lags of 2-7 and 8-14 days (Bhaskaran et al., 2009, 2010). Other research observed a seasonal variation for acute myocardial infarction (AMI) hospitalization that peaked in winter (Lee et al, 2010; Marchant, Ranjadayalan, Stevenson, Wilkinson, & Timmis, 1993; Spielberg, Falkenhahn, Willich, Wegscheider, & Voller, 1996). Old age and previous heart conditions increased vulnerability to cold effects (Analitis et al., 2008; Barnett et al, 2005; Bhaskaran et al, 2010; Danet et al., 1999; Morabito et al., 2005;
Gaps identified in scientific literature include lack of sensitive weather indicators of low temperature on health, lack of consideration of other weather factors in addition to temperature, relative paucity of data on cardiovascular morbidities, no assessment of population vulnerability, and no assessment of interaction between low temperature and air pollution. Our study fills these knowledge gaps by evaluating the effects of various cold indicators and temperature ranges in the winter on IHD hospitalizations, cold temperatures individual and cumulative lag effect, and identifying potential vulnerable population groups to help inform climate adaptation planning.
Materials and Methods
Study Population and Study Design
The study population consisted of subjects hospitalized for IHD, which included AMI (International Classification of Disease 9th Revision [ICD9] codes, ICD9 410), angina pectoris (AP) (ICD9 413), chronic IHD (CIHD) (ICD9 414), and other IHD (OIHD) (ICD9 411, 412) from 1991 to 2004 in
Sources of Data
Hospital admission data were obtained from the
Exposure Definition
The cold weather season was defined as
UAT = -2.7 + 1.04*T + 2.0*VP - 0.65*W,
where T is measured in °C, VP in kPa, and W in m*s'1. VP was derived from DP:
VP = 0.6105 * EXPU7.27 * DP[°C] / (237.7 + DP[°C]).
Maximum, average, and minimum UAT values were calculated (Steadman, 1984). UAT is used in our study as it combines temperature, humidity, and wind, which more accurately represents the overall winter atmosphere and environment than temperature alone (Madrigano et al., 2013).
Fourteen weather regions were assigned in NYS. These regions were created by overlaying and merging the
We created three temperature indicators: mean of the daily average UAT (UATavg), three-day moving average of UATavg, and extreme temperature (10th percentile of UATavg distribution). We also assessed the effects of daily minimum UAT (UATmin) and daily maximum UAT (UATmax) on IHD hospitalizations. Since temperature did not have a linear relationship with IHD hospitalizations, 5°F interval groups were created for the mean of UATavg and the three-day moving UATavg, UATmin, and UATmax. For the stratified analysis we grouped temperature in 10°F intervals to avoid small sample size problems. In both analyses the temperature group closest to and above water's freezing point (32°F) was selected as the reference.
Confounders and Effect Modifiers
Potential sociodemographic confounders were controlled for by the study design. Casecrossover design is one kind of time-series analysis and a special type of matched casecontrol design where cases serve as their own controls. This design controls for confounding due to time-invariant variables, by having constant subject characteristics such as sex, race-ethnicity, age, and source of treatment in both exposure and control periods, which also controls for seasonality as both periods occur within the same month (Madure, 1991). Barometric pressure was also included as a confounding factor in the analysis. Stratified analysis was conducted for the following variables: sex, race-ethnicity, age, source of treatment payment, weather region, and particulate matter <
Statistical Analysis
Conditional logistic regression models using the SAS PHREG procedure were employed to analyze the data. Initially, crude analyses were stratified by sex, race-ethnicity, age, source of payment, weather region, and PM,_. Separate multivariate analyses were conducted for each exposure variable (mean of daily average and UATmax, mean of three-day moving average of UATavg, and extreme temperature). These models included potential effect modifiers along with product terms between main exposure variables and potential effect modifiers. Reduced models were built through a backward elimination process and limited to variables with complete information (Agresti, 2007). Effect modification on the multiplicative scale was assessed as deviation from perfect multiplicatively as determined by the likelihood ratio test with an alpha of .05 and results supported in stratified analysis. Odds ratio (OR) estimates were adjusted for atmospheric pressure by including it in the conditional logistic regression model.
Statewide analyses of AMI hospitalizations were conducted for UATavg, UATmax, and UATmin separately on the day of hospitaliza- tion (lagO) and each of the previous six days (lagl-lag6). Similarly, statewide analyses for AP, CIHD, and OIHD were also related to UATavg on the day of hospitalization (lagO) and each of the previous six days (lagl-lagó). The effect of extreme temperatures on hospitalizations of all categories of IHD was also analyzed. Data management and analysis were conducted using SAS version 9.2.
Results
Our analysis included 232,905 IHD hospitalizations (AMI: 83,650 [35.92%], AP: 10,794 [4.63%], CIHD: 100,496 [43.15%], and OIHD: 37,965 [16.3%]). The statewide mean of the UATavg in the NYS cold weather season, 1991-2004, was 18.0°F with a range of -30°F to 60°F, and varied among weather regions with the highest mean in
Table 1 shows the results of the conditional regression analysis that examines the association between low UATavg and hospital admissions due to AMI by lag (Iag0-lag6) in NYS. Compared to the reference temperature group (35°F-40°F), temperatures below 35°F were associated with increased admissions due to AMI in Iag4-lag6. The most elevated association was observed for the -20°F to -15°F temperature group, with statistically significant ORs ranging from 1.25 (95% confidence interval [Ci] 1.00,1.57) to 1.29 (95% Cl 1.03, 1.61) at lag4 and lag5, respectively. For temperatures ranging between -5°F and 35°F, the statistically significant ORs for hospital admissions due to AMI varied between OR = 1.05 (95% CI 1.01, 1.10) for 5°F-10°F and 10°F-15°F temperature groups at lag4, 0°F-5°F and 5°F-10°F temperature groups at lag5, and OR = 1.14 (95% CI 1.06, 1.21) for -5°F to 0°F temperature group at lag4. The most consistent detrimental effect occurred at lag4. The association was not statistically significant for Iag0-lag3.
Table 2 shows results of analysis using the three-day moving average of UATavg. For temperatures below the reference (35°F-40°F), no significant effects were observed for the major- ity of the temperature groups of the moving averages except for the -15°F to -10°F temperature group with OR = 1.36 (95% Cl 1.08, 1.71) at Iag4-lag6 and for the 0°F-5°F temperature group with OR = 1.06 (95% Cl 1.00, 1.12) at Iag2-lag4. Although not statistically significant, ORs at Iag2-lag4 showed detrimental effect with values ranging from 1.01 (95% Cl 0.96, 1.06) to 1.17 (95% Cl 0.91,1.50).
Table 3 displays the results of the stratified analyses of the association between UATavg and AMI by demographics, weather region, and PM,5 concentration. Since the UATavgAMI hospitalization association was statistically significant and most consistent at lag4, we conducted the stratified analysis at lag4 only. In general, statistically significant results were consistently observed for temperature group 0°F-10°F Temperatures below 30°F increased the odds of hospitalizations among males: OR = 1.07 (95% Cl 1.02,1.11) for 0°F10°F and OR = 1.03 (95% Cl 1.00, 1.07) for 10°F-20°F temperature group; people of races other than whites or blacks: OR = 1.17 (95% Cl 1.05, 1.29) for 0°F-10°F and OR = 1.09 (95% Cl 1.00, 1.20) for 10°F-20°F; among subjects aged 25-64: OR range 1.05-1.14 for 0°F-10°F; and among subjects aged 65-74: OR = 1.06 (95% Cl 1.01, 1.12) for 20°F-30°F temperature group. Stratification by source of payment categories showed increased ORs for hospitalizations among
We further examined the association between extremely low UAT (<10th percentile) and various IHD categories in Iag0-lag6; Figure 1 describes the statistically significant estimates for different types of IHD at lag4lag6. We observed a statistically significant harmful effect on hospital admissions of AMI at Iag4-lag6 and for CIHD at lag6. Conversely, extreme low temperature displayed a protective effect on hospitalization due to AP and OIHD at lagO and lagl, and for CIHD at lagl, lag2, and lag4. We also further examined the associations between UATavg and various IHD subtypes including AP, CIHD, and OIHD. We did not find any significant relationships between low temperature and IHD other than AMI such as AP, CIHD, and OIHD (data not shown).
We analyzed the effect of UATmax on AMI hospital admissions by different lag periods as described in Table 4. Consistently increased ORs were observed in Iag5-lag6. The strongest association was observed at lag6 in the -15°F to -10°F interval (OR = 1.42 [95% Cl 1.03, 1.971). For temperature groups between 0°F and 35°F at lag5 and lag6, the ORs of AMI hospitalization varied between OR = 1.04 (95% Cl 1.00, 1.07) and OR = 1.07 (95% Cl 1.03, 1.11). With respect to UATmin, no increased risk of hospitalization was observed for UATmin <35°F (data not shown).
Discussion
The present study found that low ambient temperature in winter was associated with increased odds of hospital admissions due to AMI. We found that the association was statistically significant at UAT <35°E The effect is consistent when using multiple exposure indicators (UATavg, extreme UATavg, UATmax). The increases in risk of AMI hospital admission for all indicators range from 4% to 29%, which is within the range of findings from previous studies (Bhaskaran et al, 2010; Danet et al, 1999; Hopstock et al., 2012). Bhaskaran and co-authors (2010) examined the short-term relationship between ambient temperature and MI risk in 15 conurbations in
Several explanatory mechanisms have been suggested for the higher occurrence of IHD in winter. Cold causes constriction of skin vessels, which increases blood pressure and consequently increases oxygen demand (Kloner, 2006). Cold may induce an increase in coronary artery resistance that can, in turn, result in coronary artery vasospasm (Kloner, 2006). Additionally, cold causes an increase in red and white blood cell and platelet concentrations, an increase in plasmatic fibrinogen and blood viscosity, and a shift of protein C to the extracellular space (Kloner, 2006; Nayha, 2005). Protein C along with changes in blood pressure and thrombosis may be responsible for acute heart conditions after exposure to cold temperature (Nayha, 2002).
We found that the effect of UATavg on AMI hospitalizations showed a delayed pattern, with an effect occurring four days after cold temperature. Our results are consistent with the findings of Wolf and co-authors (2009), who observed the strongest association between daily average temperature and nonfatal MI for a lag of three days, as well as
When comparing four cold temperature indicators, we found that UATmax has the highest and most significant effect on AMI (OR: 1.42) at lag6, and followed by threeday moving UATavg (highest OR: 1.36). The effects of UATavg (highest OR: 1.29) and extreme cold (<10th percentile, OR: 1.04) on AMI and different IHD are slightly lower. In a review of 12 studies investigating low ambient temperature on MI, most prior studies used mean temperature as the indicator, a few used maximum temperature, and only one or two considered winter wind chill and dew point (Bhaskaran et ah, 2010). The effects of cold temperature on MI seemed to be larger in the studies using maximum temperature than those using mean temperature, which is consistent with our findings. Few or no studies using moving average or extreme indicators are available to compare our results. The different findings among various studies could be due to differences in adaptation to extreme low temperature of people living in different climates, humidity, and wind patterns. Since almost all previous studies examining cold and IHD focused on AMI, no previous literature is available to compare with our findings for other IHD admissions. While low UATmax with 4-5 single day lag had shown strongest AMI effect, we cannot neglect the cumulative effect of low UAT from lag4 to lag6 days on AMI although little to no previous literature is available to compare with our findings. Our study also diverges from other studies in its use of UAT, which includes dew point and wind chill in addition to temperature to represent the real atmosphere in winter.
Our analyses showed subjects age 65 and older had higher odds of AMI hospitalization, which is consistent with previous findings by Bhaskaran and co-authors (2010), Hopstock and co-authors (2012), and Morabito and co-authors (2005), showing the highest risks and a same day effect in the elderly population compared to delayed lag in other age groups. The increased risk in older age can be explained by the deterioration of effective body thermoregulation. Our findings of slight but persistent increased AMI risks of cold temperature among males compared to females is supported by the studies of
Study Strengths
This is one of the few studies examining the association between cold weather conditions and IHD morbidity, especially AMI, over a long period in NYS, a state with multiple subclimate regions. The unique contributions of this project include comparing the IHD effects using different cold weather indicators, in different temperature ranges during winter, by different lag periods, short-term vs. cumulative effects, and assessing different subgroups of IHD. Using UAT allowed us to consider multiple weather factors including temperature, dew point, and wind simultaneously. Since both health outcome and exposure data (weather and air pollution data) are objective, routinely collected datasets, recall bias is unlikely. By examining potential interactions between low temperature and pollution/sociodemographics, we were able to identify vulnerable populations, which is useful for future adaptation planning.
Study Limitations
This study used hospital admission data as health endpoints and thus may have captured only the most severe IHD cases. To minimize case under-ascertainment, we focused on AMI, a condition that requires immediate medical attention and treatment through hospital admission. Another limitation is that meteorological data were obtained from fixed monitoring sites, which assumes uniformity across regions. Although some factors such as daily activity patterns, family history of IHD, seasonal weight gain, and sociodemographic factors are important confounders, the case-crossover design automatically controls for these factors by using cases as their own controls. In addition, some factors such as quality of housing, heating use, insulation, and influenza are potential confounders that are unavailable in our current study and will be examined through surveys in the future.
Conclusion
The present study found that extreme low UAT in winter was associated with increased risk of AMI, especially during Iag4-lag6, but not associated with other IHD subgroup diseases. We also found that low UATmax in winter was a more sensitive indicator to AMI than mean and UATmin. This study also found that certain demographic groups such as the elderly, males, people with
Acknowledgements/Disclaimers: This work was supported by a grant from the
Prepublished online
References
Agresti, A. (2007). An introduction to categorical data analysis (2nd ed., p. 139).
Analitis, A., Katsouyanni, K., Biggeri, A., Baccini, M., Forsberg, B., Bisanti, L., Kirchmayer, U., Balleste, F, Cadum, E., Goodman, P.G., Hojs, A., Sunyer, J., Tiittanen, R, & Michelozzi, P. (2008). Effects of cold weather on mortality: Results from 15 European cities within the PHEWE project.
Bhaskaran, K., Hajat, S., Haines, A., Herrett, E., Wilkinson, R, & Smeeth, L. (2009). Effects of ambient temperature on the incidence of myocardial infarction. Heart, 95(21), 1760-1769.
Bhaskaran, K., Hajat, S., Haines, A., Herrett, E., Wilkinson, R, & Smeeth, L. (2010). Short-term effects of temperature on risk of myocardial infarction in
Carder, M., McNamee, R., Beverland, I., Elton, R., Cohen, G.R., Boyd, J., & Agius, R. (2005). The lagged effect of cold temperature and wind chill on cardiorespiratory mortality in
Chinery, R., & Walker, R. (2009). Development of exposure characterization regions for priority ambient air pollutants. Human and Ecological Risk Assessment, 15(5), 876-889.
Danet, S.,
Guttman, N.B., & Quayle, R.G. (1996). A historical perspective of
Hajat, S., & Haines, A. (2002). Associations of cold temperatures with GP consultations for respiratory and cardiovascular disease among the elderly in
Hopstock, L.A., Fors, A.S., Bonaa, K.H., Mannsverk, J., Njolstad, I., & Wilsgaard, T. (2012). The effect of daily weather conditions on myocardial infarction incidence in a subarctic population: The Tromso study 1974-2004.
Janes, H., Sheppard, L., & Lumley, T. (2005). Overlap bias in the case-crossover design, with application to air pollution exposures. Statistics in Medicine, 24(2), 285-300.
Kloner, R.A. (2006). Natural and unnatural triggers of myocardial infarction. Progress in Cardiovascular Diseases, 48(4), 285-300.
Kloner, R.A.,
Kysely, J., Pokorna, L., Kyncl, J., & Kriz, B. (2009). Excess cardiovascular mortality associated with cold spells in the
Lee, J.H., Chae, S.C., Yang, D.H., Park, H.S., Cho, Y, Jun, J.E., Park, W.H., Kam, S., Lee, W.K., Kim, K.S., Hur, S.H., & Jeong, M.H. (2010). Influence of weather on daily hospital admissions for acute myocardial infarction (from the Korea Acute Myocardial Infarction Registry).
Lin, S., Bell, E.M., Liu, W, Walker, R.J., Kim, N.K., &
Loughnan, M.E., Nicholls, N., & Tapper, N.J. (2008). Demographic, seasonal, and spatial differences in acute myocardial infarction admissions to hospital in
Madure, M. (1991). The case-crossover design: A method for studying transient effects on the risk of acute events.
Madure, M., & Mittleman, M.A. (2000). Should we use a case-crossover design? Annual Review of Public Health, 21, 193-221.
Madrigano, J., Mittleman, M.A., Baccarelli, A., Goldberg, R., Melly, S.,
Marchant, B., Ranjadayalan, K., Stevenson, R., Wilkinson, P, & Timmis, A.D. (1993). Circadian and seasonal factors in the pathogenesis of acute myocardial infarction: The influence of environmental temperature.
Morabito, M.,
Nayha, S. (2002). Cold and the risk of cardiovascular diseases. A review.
Nayha, S. (2005). Environmental temperature and mortality.
Panagiotakos, D.B., Chrysohoou, C, Pitsavos, C, Nastos, P, Anadiotis, A., Tentolouris, C, Stefanadis, C, Toutouzas, P, & Paliatsos, A. (2004). Climatological variations in daily hospital admissions for acute coronary syndromes.
Schwartz, J.,
Sheth, T, Nair, C, Muller, J., & Yusuf, S. (1999). Increased winter mortality from acute myocardial infarction and stroke: The effect of age.
Spielberg, C, Falkenhahn, D., Willich, S.N., Wegscheider, K., & Voller, H. (1996). Circadian, day-of-week, and seasonal variability in myocardial infarction: Comparison between working and retired patients.
Steadman, R.G. (1984). A universal scale of apparent temperature.
Wichmann,J., Ketzel, M., Ellermann, T, & Loft, S. (2012). Apparent temperature and acute myocardial infarction hospital admissions in
Wolf, K., Schneider, A., Breitner, S.,
Shao Un, MD, PhD
Health Science
University at
of Public Health
Aida Soim, MD, PhD
and Biostatistics
University at
of Public Health
Syni-An Hwang, PhD
and Biostatistics
University at
of Public Health
Corresponding Author:
E-mail: [email protected]
Colombia Reinsurance Market 2016 Industry Trends, Growth, Demand, Size and Analysis to 2019
Advisor News
- When new investment trends emerge, Gen Z is most likely generation to be first in
- Could ‘plain English’ become an advisor’s secret weapon?
- IRI urges Senate action on 403(b) parity legislation
- Three estate planning ideas to protect your clients and their wealth
- What advisors must know about accessible client documents
More Advisor NewsAnnuity News
- NUNN INTRODUCES BILL TO CUT RED TAPE, GIVE IOWANS CLEARER INSURANCE INFORMATION
- NAIC working group pressed to accelerate annuity illustration overhaul
- State Auditor James Brown Kicks Off Life Insurance Awareness Month With Policy Locator Tool
- Wink: Annuity sales post strong Q2, led by MYGAs and structured products
- Legacy Marketing Group partners with Malibu Life USA for annuity launch
More Annuity NewsHealth/Employee Benefits News
Life Insurance News