Daily Mean Temperature and Kidney Stone Presentation
Daily Mean Temperature and Kidney Stone Presentation
Between 2005 and 2011, 60,433 patients enrolled in insurance plans that were contained in MarketScan sought medical attention for kidney stones in Atlanta, Chicago, Dallas, Los Angeles, and Philadelphia (Table 1). With the exception of Atlanta and Los Angeles (p = 0.09), mean annual temperatures of each city were different (p < 0.001).
Associations between mean daily temperature and kidney stone presentation were not monotonic, and there was variation in the shape of the exposure–response curves and the strength of associations at different temperatures. However, in most cases, RRs increased for temperatures above the reference value of 10°C (Figure 1). RRs for a kidney stone presentation cumulated over a 20-day period associated with a mean daily temperature of 30°C compared with 10°C were 1.38 in Atlanta [95% confidence interval (CI): 1.07, 1.79], 1.37 in Chicago (95% CI: 1.07, 1.76), 1.36 in Dallas (95% CI: 1.10, 1.69), 1.11 in Los Angeles (95% CI: 0.73, 1.68), and 1.47 in Philadelphia (95% CI: 1.00, 2.17). The temperatures > 10°C at which statistically significant associations were first observed varied among the cities (Table 2). Heterogeneity was also noted at the limits of the temperature ranges. In Dallas, the excess RR of kidney stone presentation stabilized at 36–39% as temperatures increased > 30°C. For Atlanta, Chicago, and Philadelphia, the risk of kidney stone presentation increased throughout the upper temperature range of each city. Kidney stone presentations also were positively associated with temperatures < 2°C in Atlanta, and < 10°C in Chicago and Philadelphia. Relative humidity was not a statistically significant predictor of the risk of kidney stone presentation (data not shown).
(Enlarge Image)
Figure 1.
Overall RRs of kidney stone presentation cumulated over a 20-day lag period associated with mean daily temperature (°C) relative to 10°C in Atlanta (A), Chicago (B), Dallas (C), Los Angeles (D), and Philadelphia (E) from 2005 through 2011. The estimated RRs of kidney stone presentation associated with mean daily temperature cumulated over a 20-day lag period using distributed lag nonlinear models are shown for each city. Two spline knots were placed at equal intervals over the range of temperatures for each city. Locations of temperature knots were as follows: Atlanta (6.7°C, 18.9°C), Chicago (–8.9°C, 6.1°C), Dallas (6.5°C, 21.4°C), Los Angeles (13.0°C, 20.6°C), and Philadelphia (3.7°C, 18.4°C). Four spline knots were placed at equal intervals in the natural log scale of lags (2, 3, 4, and 7 days) to increase sensitivity for shorter lags. The solid blue line is the point estimate at each temperature, and the surrounding gray area the 95% CI.
We estimated bimodal increases in the RR of kidney stone presentation for days in which the temperature was 30°C relative to days with mean temperatures of 10°C. The strongest association between kidney stone presentation and a daily mean temperature of 30°C versus 10°C was estimated for lags ≤ 3 days and a second peak was estimated at 4 to 6 days (Figure 2). Periods of increased risk were followed immediately by days of lower risk. The RRs of kidney stone presentation after hot days at 10–20 days lag were heterogeneous. A trend of increased risk was found in Philadelphia from 10–20 days, and in Atlanta and Chicago from 15–20 days, whereas the risk in Dallas and Los Angeles varied around the null after 10 days.
(Enlarge Image)
Figure 2.
Lag response between a 30°C (mean) day and kidney stone presentation relative to 10°C over a 20-day period in Atlanta (A), Chicago (B), Dallas (C), Los Angeles (D), and Philadelphia (E) from 2005 through 2011. For each city, the estimated RRs of kidney stone presentation in association with a daily mean temperature of 30°C (relative to 10°C) for each lag day from the temperature exposure during a 20-day period are shown. We used distributed lag nonlinear models to estimate the RRs and placed two spline knots at equal intervals over the range of temperatures for each city. Locations of temperature knots were as follows: Atlanta (6.7°C, 18.9°C), Chicago (–8.9°C, 6.1°C), Dallas (6.5°C, 21.4°C), Los Angeles (13.0°C, 20.6°C), and Philadelphia (3.7°C, 18.4°C). We placed four spline knots at equal intervals in the natural log scale of lags (2, 3, 4, and 7 days) to increase sensitivity for shorter lags. The solid blue line is the RR at each lag day from the exposure, and the surrounding gray area the 95% CI.
We constructed three-dimensional graphs to demonstrate simultaneously the relationships along temperature and lag (Figure 3). Consistent trends of increasing RR of kidney stone presentation were observed within 7 days of high temperatures across cities. However, CIs cannot be represented in these figures, and therefore the precision of the estimates cannot be appreciated.
(Enlarge Image)
Figure 3.
Risk of kidney stone presentation relative to 10°C along temperature and a 20-day lag period in Atlanta (A), Chicago (B), Dallas (C), Los Angeles (D), and Philadelphia (E) from 2005 through 2011. The three-dimensional relationships include temperature (x-axis), lag (z-axis), and RR of kidney stone presentation (y-axis). The point estimate of the RR of kidney stone presentation at each point along the temperature range and lag window is shown using 10°C as the reference temperature. We used distributed lag nonlinear models to estimate the RRs and placed two spline knots at equal intervals over the range of temperatures for each city. Locations of temperature knots were as follows: Atlanta (6.7°C, 18.9°C), Chicago (–8.9°C, 6.1°C), Dallas (6.5°C, 21.4°C), Los Angeles (13.0°C, 20.6°C), and Philadelphia (3.7°C, 18.4°C). We placed four spline knots at equal intervals in the natural log scale of lags (2, 3, 4, and 7 days) to increase sensitivity for shorter lags.
The shapes of the overall cumulative exposure–response and lag response relationships between temperature and nephrolithiasis using quadratic splines were similar to the natural cubic splines and did not reveal any nonlinear relationships at temperature range limits. This suggests that these relationships were adequately captured by natural cubic splines, which were chosen for the main model because of their more conservative smoothing of data and the gain of a degree of freedom. The patterns of the overall cumulative exposure–response relationship and the distribution of risk across time in the lag response estimations were consistent whether mean, maximum, or minimum daily temperatures were used. However, compared with mean daily temperature, the precision of the estimates for maximum temperatures were lower and the AIC values were larger for minimum temperatures (data not shown). Increasing the lag window to 30 days also decreased the precision of the temperature–nephrolithiasis association as evidenced by the wider CIs in the overall cumulative exposure–response curves (see Supplemental Material, Figure S1 http://ehp.niehs.nih.gov/wp-content/uploads/122/10/ehp.1307703.s001.508.pdf). With a lag window of 30 days, a slight decrease in the RR of kidney stone presentation was observed at lags of 20–25 days from a 30°C day in Dallas, Los Angeles, and Philadelphia (see Supplemental Material, Figure S2 http://ehp.niehs.nih.gov/wp-content/uploads/122/10/ehp.1307703.s001.508.pdf); however, the RR of nephrolithiasis over this lag subperiod was < 3% (95% CI: 1.00, 1.05).
Results
Between 2005 and 2011, 60,433 patients enrolled in insurance plans that were contained in MarketScan sought medical attention for kidney stones in Atlanta, Chicago, Dallas, Los Angeles, and Philadelphia (Table 1). With the exception of Atlanta and Los Angeles (p = 0.09), mean annual temperatures of each city were different (p < 0.001).
Overall Cumulative Exposure–Response Relationship
Associations between mean daily temperature and kidney stone presentation were not monotonic, and there was variation in the shape of the exposure–response curves and the strength of associations at different temperatures. However, in most cases, RRs increased for temperatures above the reference value of 10°C (Figure 1). RRs for a kidney stone presentation cumulated over a 20-day period associated with a mean daily temperature of 30°C compared with 10°C were 1.38 in Atlanta [95% confidence interval (CI): 1.07, 1.79], 1.37 in Chicago (95% CI: 1.07, 1.76), 1.36 in Dallas (95% CI: 1.10, 1.69), 1.11 in Los Angeles (95% CI: 0.73, 1.68), and 1.47 in Philadelphia (95% CI: 1.00, 2.17). The temperatures > 10°C at which statistically significant associations were first observed varied among the cities (Table 2). Heterogeneity was also noted at the limits of the temperature ranges. In Dallas, the excess RR of kidney stone presentation stabilized at 36–39% as temperatures increased > 30°C. For Atlanta, Chicago, and Philadelphia, the risk of kidney stone presentation increased throughout the upper temperature range of each city. Kidney stone presentations also were positively associated with temperatures < 2°C in Atlanta, and < 10°C in Chicago and Philadelphia. Relative humidity was not a statistically significant predictor of the risk of kidney stone presentation (data not shown).
(Enlarge Image)
Figure 1.
Overall RRs of kidney stone presentation cumulated over a 20-day lag period associated with mean daily temperature (°C) relative to 10°C in Atlanta (A), Chicago (B), Dallas (C), Los Angeles (D), and Philadelphia (E) from 2005 through 2011. The estimated RRs of kidney stone presentation associated with mean daily temperature cumulated over a 20-day lag period using distributed lag nonlinear models are shown for each city. Two spline knots were placed at equal intervals over the range of temperatures for each city. Locations of temperature knots were as follows: Atlanta (6.7°C, 18.9°C), Chicago (–8.9°C, 6.1°C), Dallas (6.5°C, 21.4°C), Los Angeles (13.0°C, 20.6°C), and Philadelphia (3.7°C, 18.4°C). Four spline knots were placed at equal intervals in the natural log scale of lags (2, 3, 4, and 7 days) to increase sensitivity for shorter lags. The solid blue line is the point estimate at each temperature, and the surrounding gray area the 95% CI.
Lag Response
We estimated bimodal increases in the RR of kidney stone presentation for days in which the temperature was 30°C relative to days with mean temperatures of 10°C. The strongest association between kidney stone presentation and a daily mean temperature of 30°C versus 10°C was estimated for lags ≤ 3 days and a second peak was estimated at 4 to 6 days (Figure 2). Periods of increased risk were followed immediately by days of lower risk. The RRs of kidney stone presentation after hot days at 10–20 days lag were heterogeneous. A trend of increased risk was found in Philadelphia from 10–20 days, and in Atlanta and Chicago from 15–20 days, whereas the risk in Dallas and Los Angeles varied around the null after 10 days.
(Enlarge Image)
Figure 2.
Lag response between a 30°C (mean) day and kidney stone presentation relative to 10°C over a 20-day period in Atlanta (A), Chicago (B), Dallas (C), Los Angeles (D), and Philadelphia (E) from 2005 through 2011. For each city, the estimated RRs of kidney stone presentation in association with a daily mean temperature of 30°C (relative to 10°C) for each lag day from the temperature exposure during a 20-day period are shown. We used distributed lag nonlinear models to estimate the RRs and placed two spline knots at equal intervals over the range of temperatures for each city. Locations of temperature knots were as follows: Atlanta (6.7°C, 18.9°C), Chicago (–8.9°C, 6.1°C), Dallas (6.5°C, 21.4°C), Los Angeles (13.0°C, 20.6°C), and Philadelphia (3.7°C, 18.4°C). We placed four spline knots at equal intervals in the natural log scale of lags (2, 3, 4, and 7 days) to increase sensitivity for shorter lags. The solid blue line is the RR at each lag day from the exposure, and the surrounding gray area the 95% CI.
RR Along Exposure–Response Curve and Lag
We constructed three-dimensional graphs to demonstrate simultaneously the relationships along temperature and lag (Figure 3). Consistent trends of increasing RR of kidney stone presentation were observed within 7 days of high temperatures across cities. However, CIs cannot be represented in these figures, and therefore the precision of the estimates cannot be appreciated.
(Enlarge Image)
Figure 3.
Risk of kidney stone presentation relative to 10°C along temperature and a 20-day lag period in Atlanta (A), Chicago (B), Dallas (C), Los Angeles (D), and Philadelphia (E) from 2005 through 2011. The three-dimensional relationships include temperature (x-axis), lag (z-axis), and RR of kidney stone presentation (y-axis). The point estimate of the RR of kidney stone presentation at each point along the temperature range and lag window is shown using 10°C as the reference temperature. We used distributed lag nonlinear models to estimate the RRs and placed two spline knots at equal intervals over the range of temperatures for each city. Locations of temperature knots were as follows: Atlanta (6.7°C, 18.9°C), Chicago (–8.9°C, 6.1°C), Dallas (6.5°C, 21.4°C), Los Angeles (13.0°C, 20.6°C), and Philadelphia (3.7°C, 18.4°C). We placed four spline knots at equal intervals in the natural log scale of lags (2, 3, 4, and 7 days) to increase sensitivity for shorter lags.
Sensitivity Analyses
The shapes of the overall cumulative exposure–response and lag response relationships between temperature and nephrolithiasis using quadratic splines were similar to the natural cubic splines and did not reveal any nonlinear relationships at temperature range limits. This suggests that these relationships were adequately captured by natural cubic splines, which were chosen for the main model because of their more conservative smoothing of data and the gain of a degree of freedom. The patterns of the overall cumulative exposure–response relationship and the distribution of risk across time in the lag response estimations were consistent whether mean, maximum, or minimum daily temperatures were used. However, compared with mean daily temperature, the precision of the estimates for maximum temperatures were lower and the AIC values were larger for minimum temperatures (data not shown). Increasing the lag window to 30 days also decreased the precision of the temperature–nephrolithiasis association as evidenced by the wider CIs in the overall cumulative exposure–response curves (see Supplemental Material, Figure S1 http://ehp.niehs.nih.gov/wp-content/uploads/122/10/ehp.1307703.s001.508.pdf). With a lag window of 30 days, a slight decrease in the RR of kidney stone presentation was observed at lags of 20–25 days from a 30°C day in Dallas, Los Angeles, and Philadelphia (see Supplemental Material, Figure S2 http://ehp.niehs.nih.gov/wp-content/uploads/122/10/ehp.1307703.s001.508.pdf); however, the RR of nephrolithiasis over this lag subperiod was < 3% (95% CI: 1.00, 1.05).
