The impact of heat on kidney stone presentations in South Carolina under two climate change scenarios

The risk of kidney stone presentations increases after hot days, likely due to greater insensible water losses resulting in more concentrated urine and altered urinary flow. It is thus expected that higher temperatures from climate change will increase the global prevalence of kidney stones if no adaptation measures are put in place. This study aims to quantify the impact of heat on kidney stone presentations through 2089, using South Carolina as a model state. We used a time series analysis of historical kidney stone presentations (1997–2014) and distributed lag non-linear models to estimate the temperature dependence of kidney stone presentations, and then quantified the projected impact of climate change on future heat-related kidney stone presentations using daily projections of wet-bulb temperatures to 2089, assuming no adaptation or demographic changes. Two climate change models were considered—one assuming aggressive reduction in greenhouse gas emissions (RCP 4.5) and one representing uninibited greenhouse gas emissions (RCP 8.5). The estimated total statewide kidney stone presentations attributable to heat are projected to increase by 2.2% in RCP 4.5 and 3.9% in RCP 8.5 by 2085–89 (vs. 2010–2014), with an associated total excess cost of ~ $57 million and ~ $99 million, respectively.

www.nature.com/scientificreports/ Methods Study design. We performed a two-step study to project the impact of heat on future kidney stone presentations in South Carolina. As a model state, South Carolina offers insight into the effect that climate change will have on a region with an already high disease burden. South Carolina lies within the "kidney stone belt" of the Southeast US, a region with a higher incidence of kidney stone disease thought to be secondary to both diet and higher temperatures 11 . First, we performed an aggregated case-crossover study to determine the relationship between historic daily state-wide mean wet-bulb temperatures (WBT) and kidney stone presentations in South Carolina from 1997 to 2014 (adapted from Vicedo-Cabrera et al. 8 ). Next, we projected the heat-related number of kidney stone presentations and associated cost to 2089 based on projected daily WBT under two climate change scenarios.
Exposure. The exposure used in this study is daily mean WBT, a moist heat metric that accounts for both ambient heat and humidity. We previously reported that wet-bulb temperature is the most accurate temperature metric for predicting kidney stone presentations in South Carolina 12 . The reason for this higher performance is thought to be due to the opposing effect of temperature and humidity on insensible water loss. Insensible water losses increase in higher temperatures, but decrease in higher relative humidities 6,12,13 . Observed daily mean WBT was obtained by averaging county-level measures of WBT across South Carolina from the NASA Land Data Assimilation Systems from January 1, 1997 to December 31, 2014 12,14 .
Projections of future WBT were generated by Dr. Ethan Coffel, Department of Geography and the Environment at Syracuse University 15 . Six predictive datasets were considered: three general circulation models (GCMs) (Access 1-0, BCC-esm, and CNRM-cm5) forced by two representative concentration pathways (RCPs) of global climate change (RCP 4.5 and RCP 8.5). The GCMs were selected to capture some variability within each RCP, and were selected in consultation with the climate scientist on the team.
The RCPs represent different pathways of future climate change, depending on human factors such as land use change and efforts to mitigate climate change. RCP 4.5 and 8.5 lead to radiative forcing outcomes of 4.5 and 8.5 W/m 2 by 2100, respectively, with higher values representing a greater greenhouse gas effect. GCMs project different trajectories in approaching that endpoint based on climatic factors including the atmosphere, oceans, and daily variations in weather 16 . Three GCMs were selected at random to capture a range of assumptions regarding day to day variation within each RCP, in consultation with the team's climate scientist. Each projection dataset was calibrated to minimize long-term differences between historic and projection datasets, using methodology adapted from Hempel et al. 17 . RCP 4.5 represents an "intermediate" future, with shifts to lower-emissions sources of energy, the use of carbon capture technology, prices on CO 2 emissions, and an expansion of forest lands 16,18 . RCP 8.5, conversely, represents a future with high population growth, high energy demand, little advancement towards greater efficiency, and a greater dependence on domestic, more easily accessible energy sources like coal which have a higher greenhouse gas footprint 19 . Under RCP 8.5, forest lands decrease and there are far fewer assumed policies to mitigate greenhouse gas emissions 19 . While RCP 4.5 represents more aggressive greenhouse gas mitigation strategies, RCP 8.5 represents a future with mostly uninhibited greenhouse gas emissions 16 . Comparing these two possible futures in this study, we can compare the projected effect that aggressive global climate policies, actions, or inactions could have on kidney stone presentations.
As the projection dataset covered the entire globe with a 2° × 2° grid, we divided South Carolina into five geographic regions (Supplementary Appendix A), and averaged the daily WBT at the center of these regions to generate a single future daily mean WBT projection for the state per projection.
Outcome. The outcome was the number of index kidney stone presentations in South Carolina emergency rooms. For the first stage in which the temperature dependence of kidney stone presentations was modeled, the counts of kidney stone presentations from January 1, 1997 to December 31, 2014 was obtained from the South Carolina Medical Encounter Data and Financial Reports all-payer claims database. This database includes all emergency department visits, surgical procedures, and hospital admissions across South Carolina, with individual IDs and accompanying demographic data. Audits are regularly performed on this dataset to ensure 99.9% accuracy of diagnostic codes and 99.5% completion 20 . Kidney stone cases were identified as emergency room visits associated with a primary International Classification of Diseases, 9th and 10th Revision codes for kidney and urinary tract stones (ICD-9 codes 592.0, 592.1, 592.9, and 274.11; ICD-10 N20). For individuals with multiple encounters with these codes, only the earliest encounter was considered so as to not double-count multiple encounters related to a single kidney stone event. Cost per presentation was simplified as the mean total charges incurred by all patients presenting through the emergency room with a primary presentation of kidney stone disease from January 1, 1997 to December 31, 2014, without adjusting for inflation. This average cost per patient between within these dates was $9525.95. Statistical analysis. The first step consisted of the estimation of the exposure-response curve between daily WBT and kidney stone presentations using historical data and the method and modelling specifications described in Vicedo-Cabrera et al. Briefly, conditional quasi-Poisson regression with distributed lag non-linear models (DLNMs) was used to determine the non-linear and delayed association accounting up to 10 days after the exposure. We fitted a natural spline with two internal knots equally spaced in the log scale in the exposurelag dimension. For the exposure-response dimension, we used a natural spline function, instead of the original quadratic b-spline, with internal knots at the 50th and 90th percentile of the temperature distribution. The use of natural splines allowed for the log-linear extrapolation of risk beyond the observed range of temperatures in www.nature.com/scientificreports/ the historical dataset, a critical consideration when projecting into years with temperatures higher than those ever observed in the historical period 21 .
In the second step, the overall cumulative exposure-response association was used to estimate the number of WBT-related kidney stone presentations in future periods using each of the six daily projections of WBT. The method used is described in detail in Vicedo-Cabrera et al. 21 . The total number of kidney stone presentations attributed to heat in each 5-year period between 2025 and 2089 (above the mean from 2010 to 2014) was computed by summing the daily temperature-related number of cases on days with WBT > 6.7 °C (defined as the temperature of minimum risk estimated from the exposure-response curve) (see Vicedo-Cabrera et al. 21 ). Estimates were reported as average ensemble estimates across the 3 GCMs for each RCP. This modeling approach assumes no changes in population demographics or vulnerability (i.e., adaptation). That is, estimates reflect the impact on today's population in South Carolina if exposed to a warmer climate consistent with these two climate change scenarios. As measures of uncertainty, 95% empirical confidence intervals (eCI) were estimated using Monte Carlo simulations with 1000 samples of coefficients of the association over each of the GCM-specific series, assuming a normal distribution of the estimates.

Results
Both RCP 4.5 and 8.5 project an increase in temperatures from the 2010s to the 2080s, with RCP 8.5 projecting a larger increase in temperatures secondary to greater greenhouse gas emissions compared to RCP 4.5 (Fig. 1). Tables 1 and 2 describe characteristics of the historic and projected temperature datasets for both RCPs. RCP 4.5 projects a 2.3°C increase in mean temperature per 5-year period from 2010-2014 to 2085-2089, while RCP 8.5 projects a 3.6°C increase in the same time frame ( Table 2). The number of days exceeding the historic maximum daily mean from 1997 to 2014 (26.0 °C) are also expected to increase over the course of the century, with far more days exceeding this historic maximum in RCP 8.5 ( Table 2).
The risk of kidney stone presentations increases non-linearly during the 10 days following a daily WBT greater than the referent temperature of 6.7 °C with no evidence of a ceiling effect at higher temperatures (Fig. 2a). Risk increases with higher temperatures and is greatest in the 2 days following a temperature event (Fig. 2b). The increase in heat-related kidney stone events (above the baseline from 2010 to 2014) from 2025 to 2089 is 5938

Discussion
This analysis generated a projection of emergent kidney stone presentations and related costs in South Carolina under two different scenarios of climate change, one scenario with high future greenhouse gas emissions (RCP 8.5) and one path assuming more aggressive future climate mitigation policies (RCP 4.5). This analysis suggests a total increase of 5938 emergent kidney stone presentations attributed to heat with an associated total cost of ~ $57 million from 2025 to 2089 in RCP 4.5, compared to a total projected increase of 10,431 emergent kidney stone presentations attributed to heat and a total cost of ~ $99 million from 2025 to 2089 in RCP 8.5. While it is impossible to predict with certainty how future policies will slow or hasten greenhouse gas emission and anthropogenic climate change, and to know exactly what future daily temperatures will be, this analysis suggests that an increased burden of kidney stone disease on healthcare systems attributed to climate warming  Figure 2. (a) The overall cumulative exposure-response relationship between daily WBT cumulated over 10-day lag period and kidney stone presentations, relative to 6.7 °C. (b) The lag-response relationship shows the distribution of risk over 10 days following a daily WBT at the 99th percentile, relative to the referent temperature of 6.7 °C. www.nature.com/scientificreports/ is very likely. This analysis additionally demonstrates how much of that increased burden could be curbed with moderately aggressive climate action. A landmark 2008 study by Brikowski, et al. estimated that the number of Americans living in a "high risk" zone for kidney stones will increase from 40% in 2000 to 56% in 2050. Since 2008, we have a greater understanding of the temperature dependence of nephrolithiasis, which allows for more precise estimates of the impact of climate change on kidney stone presentations. While the Brikowski study applied hypothesized linear and non-linear relationships between temperature and kidney stone disease to the entire US, it has since been demonstrated that the relationship between temperature and kidney stone disease varies by geographic location, and that neither the hypothesized linear nor non-linear models accurately reflect the true exposure risk curve 5 . In addition, it has been demonstrated that moist-heat metrics such as WBT model the relationship between climate and kidney stone presentation better than dry bulb temperature alone, with or without corrections for relative humidity 5,8 , as temperature and humidity contribute differently to the extent of evaporative water loss 12 . Though more limited in geographic scope, the current study improves on this seminal 2008 study in several ways. In this study, we empirically defined the precise relationship between heat, humidity, and kidney stone presentation using historic data congruent with the study location. We also used WBT projections that account for future day-to-day variation in heat and humidity to define and project kidney stone disease risk.
There are several important assumptions to consider when interpreting the results of this study. First, we assumed a constant population at risk of kidney stone events from year to year; in reality, South Carolina's population will change over time. While population increase is likely (and thus our estimates may be an underestimate), it is also likely that population will adapt better as hot and humid days become more common, e.g. by spending more time in climate-controlled environments and drinking more fluids. We chose not to consider adaptation as efforts accurately model this adaptation are often overly simplistic 21,22 . In addition, the future effect of adaptation would likely vary by population characteristics and has not been well quantified 23 . We also applied a single risk curve for the entire population although the risk of temperature related kidney stone presentations is sexually dimorphic 8 . It is likely, however, that our population-wide estimates are accurate assuming a consistent population-wide distribution between males and females. Finally, South Carolina has recently experienced an observed rapid increased incidence in kidney stone disease, with a ~ 1% increase in mean annual incidence of kidney stone disease from 1997 to 2012, and a particularly rapid rise among adolescents 2 . We did not project future increases in non-heat related presentations given the uncertainty of future shifts in the epidemiology of the disease.
Limitations to this study include the uncertain causal inference in the temperature-dependence of kidney stone presentations. Second, our estimates of total costs were based only on the medical costs incurred by a patient at their emergency department visit with a kidney stone. We did not consider inflation or indirect costs such as loss of work. Thus, the actual "cost" to society of the increased incidence of kidney stone disease are likely far greater. Third, we only considered one geographic area. Our group previously reported that the shape of the exposure response curve differs by location. These results, therefore, should not be extrapolated to other parts of the United States or other areas of the globe. As more data on kidney stone presentations and other heat-related health outcomes become available, these projections will be possible at a larger scale.
Despite these limitations, our analysis is a model to conceptualize how the burden of kidney stone disease is expected to progress with climate change, and also how mitigations to greenhouse gas emissions can offset some of this burden. With climate change a modern reality, the US can expect warmer days overall and more frequent extreme heat events 24 . A recent analysis of the proposed resolutions at the 2021 United Climate Change Table 3. Increase in the number of kidney stones attributed to heat (defined as WBT above 6.7 °C) and associated charges projected under RCP 4.5 and RCP 8.5 by 5-year interval between 2025 and 2089, versus 2010-2014 baseline. www.nature.com/scientificreports/ Conference indicated a likely increase in mean global temperature to below 2 °C by 2100 with adherence to these resolutions 25 , an outcome just below the RCP 4.5 pathway represented in this study. As our analysis shows, under even a relatively conservative projection of climate change we can expect a higher incidence and cost of symptomatic kidney stones, particularly in the near future before any compensatory adaptive efforts. As the research community continues to elucidate the impacts that climate change will have on the environment and human well-being, it is important to explore the burden of climate change on human health via physiologic stressors. Improving our understanding of this relationship will help health systems better prepare and encourage the public and policymakers to prioritize sustainable behaviors and policies to mitigate climate change.