Using Spatially Contiguous Remote Sensing Reanalysis Data to Estimate Policy-Relevant Health Effects of Extreme Heat Exposures in New York State

Background

Regional National Weather Service (NWS) heat advisory criteria in New York State (NYS) were based on frequency of heat events estimated by sparse air monitor data. These may not accurately reflect temperatures at which specific health risks occur in specific areas within the state. North American Land Data Assimilation System (NLDAS) derived from the North American Regional Reanalysis (NARR) datasets produce hourly meteorological parameters that can be used to calculate daily air temperatures, heat index and other extreme heat metrics at a 12-km spatial resolution. The NLDAS-derived uniform exposure surfaces provide spatially resolved estimates that can be linked with health data to improve accuracy and precision for exposure-response functions especially in large rural areas where air monitoring is inadequate.

Objectives

To characterize health risks related to summertime heat exposure using remotely sensed data and estimate the temperatures at which excessive risk of heat-related adverse health effects occurs in NYS. We also evaluate the need to adjust current heat advisory threshold and messaging using exposure data with a higher temporal and spatial resolution.

Methods

The NLDAS which combines remote sensing in situ and model data to provide fine scale temperature metrics at the surface level were used to obtain exposure (temperature) data. In the absence of information on the actual place of exposure, the temperature metrics in the NLDAS dataset were assigned based on the grid that included the patient’s residential address. This dataset was then spatially merged with the datasets containing fine particulate matter (PM_2.5) and ozone estimates using the geographic coordinates of the patients’ residences. Lagged exposure values were created for all models to assess the immediate (on admission day, i.e. lag 0) and delayed (exposure up to 4 days before admission i.e. lag 1, 2, 3, and 4, respectively) effects of the temperature metrics on the health outcomes. Effect of cumulative exposures were assessed using moving average for 1 – 2, 1 – 3 and 1 – 4 days of exposure. Daily summer hospital admissions and emergency department (ED) visits (May – September) in NYS of heat stress, dehydration, acute renal failure (ARF) and cardiovascular diseases (CVD) were acquired from the New York State Department of Health’s Statewide Planning and Research Cooperative System (SPARCS) inpatient and outpatient datasets.

We conducted a case-crossover analysis using a semi-symmetric bidirectional, time-stratified design to assess the effect of summertime temperature on hospitalizations and ED visits in NYS. This method compares the temperature metrics on the day of hospital admission /visit (case/exposure day) with the temperature metrics, before and/or after (control period), within the same pre-specified stratum window of time, when the subject is not hospitalized or in the ED. Each case served as his or her own control on all measured and unmeasured subject non-time-varying characteristics. A one-month stratum window of exposure was used to compare cases with the control (referent) period of ±7, ±14, or ± 21 days. We controlled for time-varying variables such as PM_2.5 and ozone in the model and assessed effect modification due to demographics, rurality and air-pollution using stratification and interaction terms. We calculated risk ratios for a 5°F change in temperature using a conditional logistic regression analysis. We used a piecewise linear spline regression to assess the shape of the temperature-outcome association. Knots defining slope changes were sequentially selected and postfitting analysis was used to plot predicted probabilities and 95% confidence limits.

Results

We observed an increased risk of heat stress (Risk ratio (RR) = 2.38, 95% confidence interval (CI): 2.29, 2.48) and dehydration (RR = 1.07, 95% CI: 1.06, 1.08) for every 5°F increase in maximum temperature on the day of exposure. There was a 4 – 6% increased risk of ARF at lag 1; and a 0.1 – 0.5% increased risk for CVD at lag 4.

The risk of heat related illness starts increasing at temperatures much lower than the current NWS criteria. The risk of illness persists up to 4 days after a heat event. There was a significant non-linear association between temperature and heat stress. The risk of heat illness only showed small increases below 80°F but had much steeper slopes at higher temperatures. The slope leveled off after 105°F, a reflection of the low number of heat events above those temperatures in the study area.

Age was a significant effect modifier in the associations between temperature and heat stress; and temperature and dehydration. Females were less likely to be hospitalized or visit the emergency room for dehydration-related health issues. Although the highest risk of heat stress occurred in the summer months of June, July and August, the cooler shoulder months of May and September show elevated risk albeit smaller in magnitude. We observed a complex relationship between exposures to particulate matter, ozone and high temperature. The risk for effects of heat exposure was highest on days with low ozone and high PM_2.5 and lowest on days with high ozone and low particulate matter exposures. Rural areas of NYS are at as high a risk of heat-related illness as urban areas.

Discussion

In our exploration of the effects of summer temperature on morbidity in NYS, we observed excess risk for various health conditions at lower temperatures than the NWS threshold criteria for the region and that heat awareness messaging should be initiated early in the summer. We were able translate research into policy through the lowering of the heat advisory criteria for NYS by four NWS offices in the region. NYS has revised heat messaging and has developed infographics and county specific heat health profiles to aid local health departments in their heat emergency planning.

Conclusion

Remote sensing data provide refined exposure-response functions for health research, in areas with sparse monitor observations. Heat advisory definitions should be revisited to reflect region-specific risk thresholds. Remote sensing data aids development of outreach materials that describe local climate trends and health risks for use by local health departments and the public.