Linking precipitation-driven flooding events to acute respiratory illness in New York city Authors: Laura T Cabrera-Rivera, Andrew Delgado, Andrew Maroko, Suresh A Sethi, Vishal Midya, Maanal Chowdhury, Brett F Branco, Ilias Kavouras, Ismail Nabeel Published: 2026 ABSTRACT AbstractBackgroundHuman-induced climate change is increasing extreme precipitation events, potentially affecting respiratory health and healthcare access in urban areas. This study examined the relationship between precipitation-driven flooding events and emergency department (ED) visits for acute respiratory illnesses in New York City (NYC) communities.MethodsWe conducted a retrospective ecological study across 14 postal ZIP codes in New York City's five boroughs from October 2020 to November 2022. Using NYC Micronet sensor network data, we analyzed ED visits for influenza-like illness and pneumonia (n=53,534), categorizing flooding events as minor (<1.5 inches), moderate (1.5-2.5 inches), or major (>2.5 inches). Associations between flooding exposure and ED visits were analyzed using Poisson mixed effects models with a 6-day lag, controlling for socioeconomic status and COVID-19 rates.ResultsPrecipitation-driven flooding was associated with an average 2.8% decrease in respiratory ED visits after 6 days (IRR 0.972, 95% CI 0.955-0.989). Minor flooding showed a 2.4% reduction in visits, while moderate and major flooding were associated with 17.9% and 12.9% reductions, respectively. These associations remained significant after controlling for confounders.ConclusionsAcute urban flooding events were associated with decreased respiratory ED visits, suggesting possible disruptions to healthcare access or deferred care. These findings affect climate change adaptation and infrastructure planning to ensure continued healthcare accessibility during extreme weather events.Keywords: Flooding, Public health, New York city, Emergency visits, Climate change Introduction The accelerating frequency and intensity of human-induced climate change weather events, including heatwaves, wildfires, droughts, tropical storms, and floods, pose mounting threats to global health [[1], [2], [3]]. Flooding has been associated with adverse respiratory outcomes, including influenza and pneumonia, through mechanisms such as mold exposure, crowding, infrastructure disruption, and exacerbation of pre-existing conditions [[4], [5], [6], [7], [8], [9]]. In addition to direct health effects, flooding can disrupt transportation networks and access to healthcare facilities, delaying critical medical care and potentially altering patterns of emergency department (ED) utilization [[10], [11], [12]]. As extreme precipitation events intensify under climate change, understanding how flooding interacts with healthcare access and respiratory health in densely populated urban areas vulnerable to infectious diseases and flooding [13]. New York City (NYC), characterized by dense development, extensive impervious surfaces, and highly modified drainage systems, faces significant vulnerability to precipitation-driven flooding [14]. Heavy rainfall events in NYC have caused substantial economic losses, infrastructure disruptions, and environmental conditions that promote mold growth, which is strongly associated with respiratory morbidity [15]. These factors highlight the need for improved understanding of how short-term flooding events influence health service utilization in ultra-urban settings, where large populations depend heavily on functioning transportation and healthcare systems [16,17]. Evaluating the association between flooding and respiratory illness is challenging due to the high baseline burden and seasonal variability of respiratory infections, including influenza, COVID-19, pneumonia, and respiratory syncytial virus (RSV) [[18], [19], [20], [21]]. NYC has experienced significant respiratory disease burden during both seasonal influenza outbreaks and the COVID-19 pandemic, with documented disparities in exposure, healthcare access, disparities receiving vaccination and outcomes across socioeconomic groups [[22], [23], [24]]. Disentangling the effects of precipitation-driven flooding from these concurrent drivers remains an important research gap, particularly in dense urban settings where environmental and social vulnerability intersect. While previous studies have examined large-scale flooding events and infectious disease to globally [7], fewer have investigated how short-term precipitation-driven flooding influences respiratory-related ED utilization in highly urbanized environments. This study addresses this gap by examining the association between precipitation-driven flooding events and respiratory-related ED visits across 14 NYC ZIP codes from 2020-2022. We hypothesized that flooding events would be associated with measurable changes in ED utilization following short lag periods, potentially reflecting increased respiratory burden or disruptions in healthcare access. By focusing on a dense urban setting, this study provides insights relevant to climate adaptation, healthcare continuity planning, and infrastructure resilience in flood-prone metropolitan areas. Methodology Selection of study areas The study areas comprised 14 NYC ZIP codes across the five boroughs with Micronet sensors installed to obtain localized weather data. The 14 ZIP codes were selected based on the availability of localized precipitation measurements from the Micronet sensor network, which enabled identification of precipitation-driven flooding events (Table 1). Table 1Selected ZIP codes.Table 1 dummy alt textZIP CodeBoroughNeighborhood / Area (General)10001ManhattanChelsea / Hudson Yards10009ManhattanEast Village / Alphabet City10016ManhattanMurray Hill / Kips Bay10314Staten IslandBulls Head / Heartland Village10462BronxParkchester / Morris Park11101QueensLong Island City11105QueensAstoria11201BrooklynBrooklyn Heights / Downtown Brooklyn11207BrooklynEast New York11230BrooklynMidwood11368QueensCorona11378QueensMaspeth11414QueensHoward Beach11420QueensSouth Ozone Park Selected ZIP codes. Although selection was driven by data availability, the included ZIP codes span all five boroughs and represent a heterogeneous mix of neighborhoods with varying demographic, socioeconomic, and flood-risk characteristics. The NYC Micronet is a collaborative project between Consolidated Edison Company of New York Inc. and the NYS Mesonet [25]. This network offers vital real-time weather data for utility operations and climate monitoring. The ZIP codes were selected based on the availability of Micronet sensors, which provided the necessary hyper-local meteorological data crucial for the study's analysis. Emergency department (ED) visits for influenza-like illness and/or pneumonia Our study analyzed a subset of New York Department of Health and Mental Hygiene (DOHMH) data, accessed through the NYOpenData program [26]. Respiratory-related ED visits were identified using the standardized DOHMH syndromic surveillance classification, based on a combination of chief complaint text and discharge diagnosis information rather than individual billing records. The respiratory illness category included influenza-like illness (ILI) and pneumonia, capturing acute respiratory infections such as influenza, COVID-19, respiratory syncytial virus (RSV), and related respiratory syndromes consistent with DOHMH surveillance definitions. Although the complete dataset spans a broader time period, this pilot study focused on October 18, 2020, to November 30, 2022, reflecting the most comprehensive available records of ED visits for influenza-like illness and/or pneumonia aggregated by patient ZIP code of residence (n = 53,534). This period also allowed analysis under consistent conditions during the COVID-19 pandemic. To evaluate the association between precipitation-driven flooding and respiratory illness, we related flooding events to daily counts of ED visits within the corresponding ZIP codes. The outcome variable was the daily number of ED visits for influenza-like illness and/or pneumonia, aggregated by ZIP code of residence and defined using the DOHMH syndromic surveillance classification based on chief complaint and ICD-10 diagnosis codes. To examine temporal patterns, we first estimated incidence rate ratios (IRRs) across multiple lag periods (0, 1, 3, 6, 9, and 12 days post-flooding). Based on observed temporal patterns and biological plausibility related to pathogen incubation periods [[27], [28], [29]], a 6-day lag was selected for the primary model. The key predictor was precipitation-driven flooding status, and geospatial, socioeconomic, and demographic variables at the ZIP code level were incorporated as covariates to control for confounding. The key predictor examined was precipitation-driven flooding (details below). Further, geospatial, socioeconomic, and demographic variables at the ZIP code level were also incorporated as model covariates to control for confounding factors. Meteorological data Precipitation data were obtained from 22 Micronet stations located within the study ZIP codes. Ten stations are ground-based, eleven are on rooftops, and one is on a pier. The Micronet sensors measure various meteorological variables like temperature, humidity, precipitation, wind, and solar radiation at 5-minute intervals, with quality control applied before data is made accessible [25]. We only used sensors with available data, resulting in 15 sensors within 14 ZIP codes. Precipitation-driven flooding events were subsequently categorized based on daily precipitation amounts as: “minor flooding” (less than 1.5 inches), “moderate flooding” (1.5-2.5 inches), and “major flooding” (over 2.5 inches) [30]. For a more comprehensive analysis, we also considered a coarser categorization of 'no flooding' versus 'any flooding' (including minor, moderate, or major flood events). Geospatial factors Additional control variables encompassed two main categories: (1) socioeconomic factors related to the social determinants of health and (2) flood exposure risk. American Community Survey (2021, 5-year estimates) data were acquired via the National Historical Geographic Information System, [31] including socio-economic status and education measures. Residential flood exposure risk was estimated at the ZIP code level to mitigate the heterogeneous nature of these large geographic aggregates (e.g., flooded areas of a particular ZIP code may be less populated than other ZIP codes based on residential distribution vis-à-vis flood-prone areas). Population counts from the 2020 decennial census at the census block level (the smallest geographic unit where these data are available) were acquired from the National Historical Geographic Information System [31]. Using dasymetric disaggregation, the population was reapportioned to tax lots [32], which are smaller geographic units. This method redistributes data (population counts) from a source zone (census block) to a target zone (tax lot) using an ancillary dataset (in this case, availability of housing). This provides a more granular estimate of population distribution [33,34]. Flood-prone areas were operationalized using two datasets: 100-year flood zones [35] and NYC stormwater flood zones [36]. If a tax lot intersected one zone, the residential population of that tax lot was considered at risk of exposure. These data were then aggregated up to the ZIP code level for analysis, where the proportion of the area of each ZIP code in flood-prone zones and the estimated proportion of the ZIP code population in flood-prone zones were calculated. Statistical analysis Descriptive statistics were reported for ZIP code characteristics. Distributions were examined for normality, results for all continuous variables are presented as mean (SD), and the median (min, max) are also reported for variables with non-normal distributions. All p-values were derived from two-sided hypothesis tests with statistical significance at 0.05. Note that the p-values presented are False Discovery Rate-corrected p-values. All statistical analyses were conducted using the R statistical programming environment version 4.1.1 [37]. To clarify our modeling assumptions and identify an appropriate adjustment set, we developed a directed acyclic graph (DAG) representing hypothesized relationships between precipitation-driven flooding (exposure), respiratory-related ED visits (outcome), and potential confounders (Fig. 1). In this framework, season and area-level socioeconomic and flood-risk characteristics were treated as confounders influencing both flooding exposure and healthcare utilization. Weekly ZIP code–level COVID-19 case rates were included as a time-varying covariate to account for pandemic-related variation in background respiratory illness and healthcare-seeking behavior. We acknowledge that COVID-19 activity could also plausibly lie on the causal pathway between flooding and respiratory ED visits; therefore, results were interpreted with consideration of potential partial overadjustment. Directed acyclic graph (DAG) describing hypothesized causal relationships between precipitation-driven flooding (exposure) and respiratory emergency department visits (outcome). We used Poisson generalized linear mixed effects models with a log link function to quantify precipitation-driven flooding events' association with respiratory illness-related ED visits [38]. The dependent variable was the number of respiratory illness-related ED visits (6 days). We first examined candidate exploratory variables for collinearity using pairwise correlations and excluded highly related ones. The fixed effects ultimately included in the model were the primary exposure precipitation-based acute flood event status for a given day within a particular ZIP code (categorical variable: no flooding, any flooding), and the primary cofounders were: season (categorical: summer (June 1 to August 31), fall (September 1 to November 30), winter (December 1 to February 28), spring (March 1 to May 31)), weekly ZIP code–level COVID-19 case rate (numeric), obtained from the New York City Department of Health and Mental Hygiene via the NY Open Data platform and included as a contemporaneous time-varying covariate to account for pandemic-related variation in respiratory illness and healthcare utilization, socioeconomic variables (median household income (numeric), the proportion of residents that are Hispanic (numeric), the proportion of residents lacking a high school degree (numeric)), and two flood risk exposure metrics (proportion of a given ZIP code area in a flood zone (numeric) and proportion of a given ZIP code’s population living in flood zones (numeric)). Random effects included ZIP codes to allow for different baseline rates of ED visits across areas and to provide a spatial blocking factor to accommodate potential unexplained spatial variation in the data not addressed by other covariates. Subsequently, to explore relationships with the degree of acute precipitation-driven flood events (i.e., flooding severity, categorical variable: no flooding, minor, moderate, major), the main model was refit using this flooding severity variable. Incidence response rates were calculated from fitted model parameters and provided the predicted ED visits ratio relative to a no-flood baseline. To facilitate policy-relevant interpretation, we translated the primary incidence rate ratio (IRR) into an approximate absolute changes in ED visits using a model-based counterfactual approach. Let Ntotal denote the total number of respiratory-related ED visits during the study period, Dtotal the number of zip-code-days observations, and Df the number of ZIP-code days classified as flooding events for each severity category. The mean daily ED visit rate was calculated as The estimated number of visits deferred or suppressed per flood zipcode-day was calculated as And the total estimated deffered visits across flooding days was approximated as Results Table 1 provides a demographic and socio-economic profile of selected ZIP codes in NYC. Notable findings include ZIP code 11368 (Corona, Queens) having the highest percentage of Hispanic/Latino residents at 75.62% and a high proportion without a high school degree at 35.62%. ZIP code 10009 (East Village, Manhattan) stands out, with 26% of its area and 27.82% of its population located in flood-prone areas according to FEMA flood maps, indicating significant flood risk. Conversely, ZIP codes such as 10462 (Parkchester/Morris Park, Bronx) and 11230 (Midwood, Brooklyn) show no estimated population within these flood-prone areas (Fig. 2). Table 1aDemographic and socio-economic characteristics of selected NYC zip code tabulation areas (ZCTAs) with flood risk exposure.Table 1a dummy alt textZIP code (ZCTAa)PopulationbPopulation without Highschool Degree (%)Hispanic/Latino (%)Area per ZIP code in FIRMc (%)Estimated ZIP code population in FIRM (%)ZIP code Median HH IncomePrecipitation-driven Flooding EventsdTotal Rainfallf (cm)1000126,0237.2218.4416.231.44101409289269.241000959,95311.6124.4726.0027.8277551283264.161001655,4273.4311.967.388.82136431269247.141031492,66310.1315.1422.450.4494578279266.701046277,92617.947.921.250.0059032299284.4811101e33,41410.1920.8218.1813.0698920333276.86287281.941110538,27411.421.5614.182.1793200266256.541120165,2284.7811.829.814.78148282296271.781120796,86019.0934.440.930.0045616302253.231123091,29216.4510.930.000.0061017314244.6011368114,76235.6275.6233.50.0062254293264.161137839,75714.1535.214.930.0085464296254.001141430,71911.2624.9117.6613.0186113289236.981142051,62221.7522.910.000.0089027292235.20aZCTA= ZIP Code Tabulation AreabTotal Population according to the 2021 American Community SurveycFIRM= Flood Insurance Rate MapdFlooding events include minor, moderate or major events as defined by the daily total precipitation.fTotal precipitation in the period of October 18, 2020, to November 30, 2022eZIP code 11101 contained two monitoring stations Demographic and socio-economic characteristics of selected NYC zip code tabulation areas (ZCTAs) with flood risk exposure. ZCTA= ZIP Code Tabulation Area Total Population according to the 2021 American Community Survey FIRM= Flood Insurance Rate Map Flooding events include minor, moderate or major events as defined by the daily total precipitation. Total precipitation in the period of October 18, 2020, to November 30, 2022 ZIP code 11101 contained two monitoring stations Flood-prone areas (map on the left), and total precipitation (map in the middle) ED visits for respiratory-related diseases (map on the right) across different ZIP code tabulation areas by season. The first map on the left highlights flood-prone areas in red, and study ZIP codes are outlined in white. The middle map shows total precipitation in inches for different seasons, represented by a bar graph overlay on each ZIP code area. The colors of the bars represent different seasons: green for spring, yellow for summer, orange for fall, and light blue for winter. The third map on the right indicates the number of ED visits with the same color coding for seasons as the middle map. To examine temporal patterns, incidence rate ratios (IRRs) were first estimated across multiple lag periods (0, 1, 3, 6, 9, and 12 days post-flooding) (Fig. 3). IRRs generally decreased and plateaued after several days, suggesting a temporary reduction in ED visits following flooding events. Major flooding showed a transient increase at shorter lags before stabilizing. Based on these temporal patterns and biological plausibility, a 6-day lag was selected for the primary analysis. Incidence Rate Ratios (IRR) of Respiratory Emergency Department Visits by Flood Severity Category across different lag periods (0-day, 1-day, 3-day, 6-day, 9-day, and 12-day lags). Results from the main model revealed a significant association between the “any flooding” and respiratory-related ED visits after a 6-day lag (Fig. 4). The IRR for “any flooding” versus “no flooding” indicates a 2.8% decrease in the average number of respiratory-related ED department visits following a precipitation-driven flooding event after accounting for the 6-day lag (IRR [95%CI] = 0.972 [0.955-0.989], p = 0.002). Incidence rate ratios (IRRs; reference baseline = “no flooding”) for the effects of precipitation-driven flood events on respiratory-related ED visits lagged 6 d in NYC study ZIP code areas. Fitted models controlled for season and other ZIP code-specific sociodemographic factors. Points represent IRR estimates, with horizontal lines depicting 95% confidence intervals. Significant findings are denoted with asterisks: *p < 0.05, **p < 0.01. The reference line at IRR = 1 indicates no effect. An IRR greater than 1 suggests increased visits following the precipitation event, while an IRR less than 1 suggests a decreased rate. Examining the IRR results further using the model with the flooding severity (minor, moderate, major) categorical variable found that “minor flooding” was associated with a 2.4% decrease (IRR [95%CI] = 0.976 [0.959-0.993], p = 0.007), “moderate flooding” was associated with a 17.9% decrease (IRR [95%CI] = 0.821 [0.730-0.923], p = 0.002), and “major flooding” was associated with a 12.9% decrease (IRR [95%CI] = 0.871 [0.763-0.994], p = 0.04) in the average number of respiratory-related ED visits after a 6-day lag. Parameter estimates for all models generated can be found in the supplementary materials. To contextualize these relative estimates for policy audiences, we applied a model-based counterfactual approach to approximate the absolute number of respiratory ED visits that would have occurred in the absence of flooding. Based on a mean of 4.61 ED visits per zipcode-day (Ntotal = 53,534; Dtotal = 11,610 zipcode-days), had the 4,387 flood-event zipcode-days not occurred, an estimated 583 additional respiratory ED visits would have been expected across the 14 study ZIP codes over the 25-month study period, approximately 23 additional visits per month. The effect was concentrated in moderate flooding events, where each moderate flood zipcode-day was associated with approximately one additional suppressed ED visit. These estimates are consistent with the interpretation that flooding events led to deferred or forgone care rather than a true reduction in respiratory disease burden. Discussion These findings in NYC's highly ultra-urban environments add a complex layer to understanding how climate change-induced disasters interact with public health. By covering a range of respiratory illnesses, including COVID-19, influenza, and RSV, our data provide a comprehensive view of the potential health impacts in the aftermath of precipitation-driven flooding events. However, the causal pathways linking flooding and respiratory outcomes remain complex and require further investigations; therefore, these findings should be interpreted as evidence of association rather than definite causation. NYC's dense population and coastal location make it well-suited for studying climate change impacts on health. Our study documents a 2.8% decrease in the average number of respiratory-related ED visits after a 6-day lag for precipitation-based flooding events (i.e., “any flooding” category). Moreover, moderate and major precipitation-based flooding decreases ED visits by 17.9% and 12.9%, respectively. Other studies have found that flooding can impede hospital access, negatively impacting provider staffing availability, disposition, and transfer processes, and resource consumption [39,40], but also suggest the increase in ED visits during Hurricane Sandy in NYC [12]. In recent years, in NYC, we have seen how heavy rains shut down much of the city’s transportation because of flooding, restricting residents’ access to care. According to the most recent U.S. Census, only 45% of the population in NYC owns a car, as most residents rely heavily on the city’s public transportation system [41]. Together, these findings contribute to our understanding of the potential health impacts associated with urban flooding events. Because NYC represents a highly urbanized environment characterized by dense populations, extensive impervious surfaces, aging drainage systems, and heavy reliance on public transportation, these findings may be informative for other major metropolitan areas facing increasing precipitation-driven flooding under climate change. The observed association between flooding and reduced emergency department utilization suggests that short-term disruptions in healthcare access may occur in similar urban contexts, highlighting the importance of strengthening healthcare continuity and infrastructure resilience in flood-prone cities. While our study did not directly evaluate the role of infrastructure, the results suggest the importance of further examining modifiable factors, such as the adequacy of drainage infrastructure, to mitigate the health risks posed by increased precipitation events in densely populated areas. Implementing nature-based flood control solutions could improve resilience by offering co-benefits like urban greening and water management, thereby supporting public health [[42], [43], [44]]. Our study shows that even minor precipitation-based flooding events can contribute to decreased ED visits for acute respiratory illness, a finding not previously known in the literature. Reduced ED utilization following flooding does not necessarily indicate a true decline in respiratory illness. Instead, it may reflect temporary disruptions in healthcare access, transportation barriers, or altered healthcare-seeking behavior. Because the study period overlaps with the COVID-19 pandemic, when ED use declined broadly, the observed reductions likely represent healthcare disruption rather than decreased disease incidence. While this reduction in ED visits could potentially represent avoided or deferred health care needs during flood events, it also raises questions about whether it reflects foregone or delayed use of hospital resources that could lead to increases in preventable acute illness due to lack of timely access to care. To put this in concrete terms, a model-based counterfactual estimate suggests that approximately 583 respiratory ED visits, roughly 23 per month, were suppressed or deferred across the 14 study ZIP codes over the study period, with moderate flooding events accounting for the largest per-event impact (∼1 visit per flood zipcode-day). This issue is particularly critical for respiratory illnesses, which can exacerbate quickly. If not addressed promptly, delayed care may lead to more severe health complications and increased healthcare costs. Prior research has shown that older patients with multiple indicators of delayed or missed care experienced higher mortality and marked declines in healthcare utilization during the pandemic [45]. Other examples, following Hurricanes Irma and Maria in Puerto Rico and the US Virgin Islands, most healthcare infrastructure, including hospitals and pharmacies, was impacted due to flooding-related structural damage and staff shortages [[11], [46], [47]]. A study by Calo et al. (2022) analyzing these hurricanes found that disruptions in oncology care were common, with challenges in communication between providers and patients. It underscored how patient resilience and community support were key to resuming care [48]. Additionally, such disruptions in access to critical health services can have severe consequences, as evidenced by findings emphasizing the need to incorporate medical care continuity into disaster preparedness plans [48]. Beyond clinical impacts, healthcare disruptions during extreme weather events may also carry important economic implications. Acute respiratory illnesses impose substantial direct and indirect costs on healthcare systems, including medical expenditures, productivity losses, and broader societal impacts [49]. Although this study did not directly estimate economic burden, deferred or disrupted care during flooding events may contribute to increased disease severity at presentation, higher hospitalization rates, and greater downstream healthcare costs. Framing flood-related healthcare disruption within its economic context strengthens the case for investing in infrastructure resilience and healthcare continuity planning as components of climate adaptation strategies. In addition to respiratory illnesses, flooding can increase the risk of infectious disease transmission through displacement, overcrowding, disruptions to sanitation, and interruptions to healthcare services [50]. These mechanisms highlight the broader health risks associated with flooding and reinforce the need for comprehensive disaster preparedness strategies addressing both immediate and longer-term health impacts. Rapid urban development, the effects of human-induced climate change, and shifting rain patterns have exacerbated flooding problems in urban areas. The implications of our findings are important for community health, as this pilot study demonstrates the complex association between precipitation-driven flooding and a decrease in ED visits for acute respiratory illness, a link not established in previous studies. Improved environmental monitoring and multidisciplinary research collaborations are needed to advance the understanding of urban flood impacts on respiratory health. Unfortunately, conventional “stream gage” flood stage monitoring has proven ineffective in heavily modified landscapes such as ultra-urban and population-dense metropolises. Accurate flood detection in ultra-urban areas will require more precise technologies. For example, NYC is currently working to implement a city-wide flood sensor network called “FloodNet” to enhance flood detection [51]. Expanding the number of FloodNet sensors will provide improved environmental data spatial coverage across NYC communities. It will generate more precise flooding information to support healthcare planning, infrastructure planning, and delivery of community education related to respiratory health. However, distributed sensor systems such as the FloodNet are novel tools and require further testing and validation across locations and weather conditions. These findings may help guide green and gray infrastructure investments for flood control and inform prioritization of mitigation strategies to reduce respiratory health risks in flood-prone communities. Limitation Several limitations should be considered when interpreting these findings. First, age-stratified analyses were not possible because the publicly available emergency department dataset provides aggregated counts at the ZIP code level and does not include individual-level demographic information. Consequently, we were unable to directly evaluate whether flooding-related healthcare disruptions disproportionately affected vulnerable populations such as older adults or children. Similarly, stratified analyses by race and ethnicity were not feasible because individual-level racial and ethnic identifiers were not available. To partially account for structural inequities, we incorporated area-level socioeconomic and demographic indicators into the models, which capture spatial patterns of social vulnerability. However, these ecological measures cannot substitute for individual-level data, and future research is needed to better understand disparities in healthcare access and health outcomes during flooding events. Although socioeconomic and demographic variables were included to control for confounding, this study was not designed to identify predictors of differential disruption in healthcare access. Given the relatively small number of ZIP codes and the ecological design, interaction analyses between flooding and social determinants of health could produce unstable estimates. Future studies using larger spatial samples and individual-level clinical data are needed to determine whether flooding-related healthcare disruptions disproportionately affect socially vulnerable populations and to inform equitable climate adaptation and healthcare continuity planning. The study period (October 2020-November 2022) largely coincided with the COVID-19 pandemic, during which domestic and international travel to NYC was substantially reduced relative to pre-pandemic levels. Because NYC is a major global travel hub, reduced population mobility during this period may have influenced baseline respiratory virus circulation and healthcare utilization patterns, potentially limiting generalizability to less connected settings. This study included data from 14 ZIP codes with available Micronet precipitation measurements, which may limit broader geographic representativeness. Additionally, the ecological and aggregated nature of the dataset prevented adjustment for individual comorbidities and clinical history, which may influence susceptibility to respiratory illness and healthcare-seeking behavior. Finally, characterization of precipitation-driven flooding events remains challenging. We used publicly available meteorological data combined with localized Micronet sensor measurements; however, the coarse spatial resolution of some meteorological data, the limited spatial coverage of Micronet sensors, and the complex hydrological dynamics of urban environments introduce uncertainty in the identification of flooding events. Not all precipitation events necessarily result in localized flooding, and misclassification may have occurred. Future research using individual-level clinical data, broader geographic coverage, and improved flood detection systems is needed to better understand the pathways linking flooding, healthcare access, and respiratory health outcomes, as well as to identify populations most vulnerable to healthcare disruptions during extreme weather events. Conclusion This pilot study demonstrates an association between decreased respiratory ED visits and precipitation-driven flooding in dense urban areas like NYC. While preliminary, these findings uniquely identify the importance of ensuring healthcare access during extreme and non-extreme precipitation events to prevent exacerbated health issues. The results underscore the need for infrastructure adaptations and strategies to maintain healthcare continuity in response to intensifying climate change impacts on public health. Declarations Author ethics approval and consent to participate Icahn School of Medicine Institutional Review Board deemed this work not human subject research. Consent for publication Not applicable. The manuscript does not include any individual person’s data in any form. Availability of data and materials The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request. Funding This research was supported by the New York State Mesonet, whose original funding was provided by the Federal Emergency Management Agency (FEMA) under the New York State Mesonet buildup grant FEMA-4085-DR-NY. The National Mesonet Program, the University at Albany, Federal and private grants, and other contributors support the continued operation and maintenance of the New York State Mesonet. Additional funding for this study was provided by Grant #ES023515 (P30), the Short-Term Training Program for Minority Students (5 R25 HL108857), and the City University of New York Pilot Award. This project also benefited from support from the National Institute on Minority Health and Health Disparities under grants 5 T37 MD001452 and 5 R25 HL108857. CRediT authorship contribution statement Laura T. Cabrera-Rivera: Writing – review & editing, Writing – original draft, Software, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Andrew Delgado: Writing – review & editing, Validation, Formal analysis, Data curation. Andrew Maroko: Writing – review & editing, Formal analysis, Conceptualization. Suresh A. Sethi: Writing – review & editing. Vishal Midya: Writing – review & editing, Visualization, Methodology, Conceptualization. Maanal Chowdhury: Writing – review & editing. Brett F. Branco: Writing – review & editing, Resources, Methodology, Investigation, Funding acquisition, Conceptualization. Ilias Kavouras: Writing – review & editing, Methodology, Investigation, Conceptualization. Ismail Nabeel: Writing – review & editing, Writing – original draft, Visualization, Supervision, Software, Resources, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization. Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Acknowledgments The authors would like to express their gratitude to the New York State Mesonet team for their invaluable support and collaboration, as well as to the Federal Emergency Management Agency (FEMA) for their foundational funding. Special thanks to the City University of New York for their Pilot Award support and to the National Institute on Minority Health and Health Disparities for their contributions through training and research grants (5 T37 MD001452 and 5 R25 HL108857). The authors also acknowledge the Short-Term Training Program for Minority Students and the Center for Environmental Health at the University at Albany for their role in advancing this research. This work was presented at the Urban Transitions Elsevier Conference in November 2024, where the findings contributes to discussions on urban resilience, climate change, and health disparities. Footnotes Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.joclim.2026.100697. Supplementary materials References Associated Data Supplementary Materials Data Availability Statement The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request.