Uncovering the role of preventive health services utilization in protecting older adults from heat-related depressive risks Authors: Shaowei Liu, Junyao Zheng, Fan Yang, Wenjie Li, Yiyan Zhou, Shihan Lu, Yifan Zhang Published: 2026 ABSTRACT AbstractBackgroundResearch on extreme heat has predominantly focused on acute heatwaves, obscuring the mental health risks of cumulative heat exposure for aging populations. As a proactive public health intervention, preventive health services represent a critical but underexamined dimension of community resilience to climate-related health risks. This study examines the relationships between cumulative heat exposure and depressive symptoms among older adults and explores whether preventive health services utilization attenuates this risk.MethodsWe analyzed longitudinal data from 7,464 older adults (yielding 21,383 observations) in the China Longitudinal Aging Social Survey (2016–2018-2020), linked with city-level meteorological data. Cumulative heat exposure was measured by the annual count of extreme heat days. Depressive symptoms were assessed using the CES-D-9 scale. Two-way fixed-effects regression models were employed. To mitigate confounding by indication and multiple-testing inflation, Inverse Probability of Treatment Weighting (IPTW) and Benjamini–Hochberg False-Discovery-Rate (FDR) adjustments were applied.ResultsCumulative heat exposure significantly exacerbated depressive symptoms among older adults (β = 0.013, p < 0.001). Preventive health services utilization provided substantial protection. Specifically, health education (β = –0.036, p < 0.001) and health records management (β = –0.015, p = 0.034) demonstrated robust attenuating associations, whereas the moderating role of health screening was marginally significant. Heterogeneity analyses indicated that the buffering effects of preventive services were concentrated in older men, rural residents, and the middle region.ConclusionCumulative heat exposure poses a significant risk to older adults’ mental health, but community-based preventive health services can serve as vital institutional adaptation capital to buffer this risk. Prioritizing targeted health education to bridge the knowledge-action gap, alongside continuous health records monitoring in resource-constrained rural and middle regions, offers an equitable strategy for enhancing climate resilience in aging societies.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s12877-026-07828-6.Keywords: Cumulative heat exposure, Depressive symptoms, Preventive health services Introduction Unprecedented warming coincides with accelerating population aging, creating a critical public health challenge as older adults face disproportionate vulnerability to climate hazards due to compromised thermoregulation, chronic health conditions, and limited socioeconomic resources [1]. Investigating the link between heat exposure and health outcomes among older adults is important, as it quantifies their elevated susceptibility to heat stress, thereby guiding evidence-based policies to avert excess mortality and strengthen climate resilience. Depression affects over one-third of older adults globally, with prevalence expected to increase as populations age [2]. Emerging evidence demonstrates that heat exposure might exacerbate depression through multiple pathways, including triggering neuroinflammatory responses, disrupting sleep patterns, limiting outdoor mobility, and fragmenting social networks [3]. This issue is particularly concerning in China, where 26.4% of older adults experience clinically significant depressive symptoms, and rising temperatures have been associated with worsening mental health outcomes [4]. As both climate change and population aging accelerate, identifying effective community-based climate adaptation strategies has become an urgent global priority [5]. Most heat-related morbidities and mortalities are preventable through improved preparedness and appropriate plans [6]. Preventive health services represent a promising yet underexplored mechanism for enhancing climate resilience among vulnerable populations [7]. Over the past decade, China has implemented substantial primary healthcare reforms, establishing a comprehensive three-tier system that emphasizes community-based care and preventive services [8]. These reforms align with the global trend of shifting healthcare from reactive treatment to proactive prevention and health management [2]. While extensive research has documented the physiological impacts and mortality associated with acute heat events [4], two critical areas remain under-examined regarding older adults. First, mental health implications, particularly depression, represent a significant yet underrecognized burden [9]. Second, traditional research predominantly focused on episodic emergencies [10, 11] has obscured the cumulative health risks of exposure to increasingly dangerous extreme heat [12, 13]. Addressing these gaps is crucial for understanding and intervening in heat-induced health risks in aging populations. By exploring the association between cumulative heat exposure and depressive symptoms among older adults and determining whether preventive health services utilization attenuates this association, the present study extends the evidence base in three respects. First, we measure exposure as the annual count of city-specific extreme-heat days rather than as a binary heatwave event indicator. Second, we examine one-, two-, and three-year cumulative exposure windows, allowing us to observe the temporal dynamics of the heat–mental-health relationship. Third, we shift the outcome from mortality and acute morbidity to depressive symptoms, an under-examined but high-burden mental-health endpoint among older adults. The findings contribute to the understanding of cumulative climate risk and climate adaptation grounded in a social-ecological framework through three ways: First, it shifts the empirical focus from acute heatwave events to seasonal cumulative heat exposure, a more realistic operationalization of long-run climate change. Second, it positions preventive health services as a community-scale form of climate adaptation, complementing the dominant focus on individual behavioral adjustment. Third, it provides the evidence that existing primary health care infrastructure can be leveraged as institutional adaptation capital to protect the mental health of older adults. Literature review Cumulative heat exposure and depressive symptoms among older adults Existing literature suggests that extreme heat exposure may be associated with adverse mental-health outcomes among older adults, plausibly operating through multiple interacting pathways. Among these proposed pathways, two are strongly supported by human empirical evidence. First, heat-related sleep disruption links to next-day mood disturbances and depressive symptoms [14]. Second, reduced outdoor mobility exacerbates social isolation among older adults [15]. A third proposed mechanism involves hypothalamic–pituitary–adrenal (HPA) axis hyperactivity and downstream neuroinflammation [16]. However, as this is primarily documented in animal studies, we view it as a plausible but unconfirmed mediator in human populations. Beyond these biological and behavioral channels, systemic inequities, particularly unequal access to cooling infrastructure, disproportionately expose disadvantaged older adults to extreme heat risks [17]. While many empirical studies report adverse associations between extreme heat and mental health, other studies find null or inverse relationships, particularly in cooler climates, in populations with strong adaptation capacity, or after careful adjustment for indoor environment and socio-economic confounders [18, 19]. This heterogeneity suggests that the extreme heat–depressive symptoms association is context-dependent rather than universal. It underscores the importance of examining long-term cumulative exposure and community-level moderating factors, which can elucidate the specific conditions under which the association emerges. Past research on extreme heat and its health risks has prioritized acute and episodic heat emergencies [4, 10, 11], obscuring the profound impact of cumulative exposure to increasingly dangerous extreme heat [12, 13]. This oversight is critical because cumulative heat exposure presents unique risks compared to acute events. The interplay of enduring societal and individual-level factors creates divergent cumulative exposures over seasonal timeframes, leading to varied health outcomes. Such interacting factors are challenging to analyze and attribute using traditional heat-health research practices, consequently leaving the understanding of cumulative heat impacts largely underexplored [13]. Moreover, the disproportionate burden on vulnerable groups, including older adults and those with lower socioeconomic status, highlights a critical climate justice issue. Investigating disparities in cumulative heat exposure across societal subgroups is essential for informing improvements in public services and facilities, thereby promoting more equitable climate-adaptive environments [20]. Climate adaptive role of preventive health services utilization For extreme heat adaptation, most studies focused on individual-level behavioral adjustments, with limited attention towards community-based strategies [3]. As heat-related health risks are preventable, preventive health services integrated within primary healthcare systems represent potential in building climate resilience, particularly for older adults [21]. From a theoretical standpoint, the climate-adaptive potential of preventive health services can be conceptualized within a social-ecological framework, which posits that health and well-being are influenced by interactions across multiple levels [22]. According to the Health Belief Model and Social Cognitive Theory, preventive health services could enhance perceived susceptibility to heat risks and promote self-efficacy in adopting adaptive measures. These theoretical perspectives further suggest that such services can foster social capital and strengthen mutual help networks through promoting public activities and social interactions [23]. Empirical evidence on the climate-protective role of preventive health services is emerging but uneven. Preventive health services encompass key components such as health education, early screening, and health management. These components have been associated with improved protective behaviors and may enable proactive intervention before significant psychological or physical deterioration occurs [24]. For instance, health education lectures not only disseminate crucial information on heat risks and coping strategies but also facilitate older adults’ participation in social activities, thereby promoting support networks [25]. This community-based engagement, echoed in findings from community programs in India and China, has been reported to foster social cohesion and reduce isolation, particularly for older adults living alone [26]. Furthermore, health screenings and health records management may empower older adults to monitor their health conditions and adopt adaptive measures against extreme heat [27]. Beyond individual empowerment, preventive health service facilities often function as vital aid hubs, providing evidence-based guidance and coordinating with emergency services to ensure timely assistance for vulnerable populations [28, 29]. However, evidence specifically linking preventive health service utilization to mental health outcomes under cumulative heat exposure remains scarce, and the mechanisms through which different service components may operate are rarely disentangled [30]. Such research gap is especially important in rapidly aging societies, where preventive health services must balance increasingly complex health demands while addressing emerging climate-related challenges [31]. The present study While existing scholarship has documented the acute implications of heat exposure, the cumulative effects of sustained, seasonal heat exposure on older adults’ depressive symptoms, remain an empirical lacuna. As the prior part of primary healthcare system, the mitigating role of preventive health services against prolonged heat stressors is also unexamined. We adopt a social-ecological framework of climate resilience to organize our analysis. This framework distinguishes three nested levels of influence on older adults’ depressive symptoms: the macro-level (broad environmental stressors, cumulative heat exposure), the micro-level (individual demographic and socio-economic vulnerabilities that modify susceptibility), and the meso-level (community and institutional resources, specifically preventive health services). This nested framework motivates our empirical specifications, in which preventive health service utilization enters not only as a covariate but also as a critical moderator of the macro-to-micro heat–depressive symptoms association. The purpose of this study is to: (1) examine the association between cumulative heat exposure and depressive symptoms among older adults; (2) determine whether preventive health services utilization attenuates this association. We leverage longitudinal health panel data in China, employing two-way fixed-effects regression models complemented by Inverse Probability of Treatment Weighting (IPTW) to address potential confounding by indication. This study enhances the understanding of cumulative heat exposure’ impact on mental health. It provides new insights on strengthening climate resilience by utilizing preventive health services, expanding the climate adaptation based on social infrastructures. Moreover, by conducting heterogeneity analyses across geographical regions, gender, and urban–rural settings, the study further uncovers disparities in climate vulnerability among older adults. These findings provide a crucial empirical basis to design equitable climate adaptation policies and address climate-justice concerns. Method Data and sample This study utilizes data from 7,464 older adults residing in 85 cities across mainland China, drawn from the China Longitudinal Aging Social Survey (CLASS). 21,383 observations across three waves (2016, 2018, and 2020) [32] were included. Daily meteorological data were sourced from the National Meteorological Information Center of China (NMICC), which maintains comprehensive weather records from more than 2,400 monitoring stations nationwide [33]. The primary climatic variable of interest was the daily mean temperature (°C) for each city, with historical records spanning from 1971 to 2020 to capture climate trends. City socioeconomic indicators were extracted from the China City Statistical Yearbook. City data were linked to individual records based on household addresses. This approach allowed for statistical adjustment of both environmental covariates and demographic characteristics relevant to older adults. We restricted initial sample to older adults aged 60 and above who participated across three survey waves. Missing data were confined to four city-level covariates (GDP, green space, hospitals, and medical beds) due to gaps in municipal yearbooks; individual-level variables had no missing values (Additional file 1, Table S2). Baseline comparisons between the retained and excluded samples demonstrated that most characteristics, including depressive symptoms, were well balanced, with only minor differences in age and marital status (Additional file 1, Table S3). This indicates that listwise deletion did not introduce material selection bias. Further sensitivity analyses addressing missing data are detailed in Sect. 4.4 and Additional file 1, Table S4. Measurement Depressive symptoms Depressive symptoms were assessed using a psychometrically validated nine-item version of the Center for Epidemiologic Studies Depression Scale (CES-D-9), demonstrating cross-cultural reliability in detecting non-psychotic mental disorders among Chinese older adults [34]. The scale comprises four domains: (1) three positive affective symptoms: happiness, life enjoyment, pleasure; (2) two negative affective symptoms: loneliness, distress; (3) two somatic symptoms: appetite loss, sleep disturbances; and (4) two cognitive symptoms: perceived uselessness, purposelessness. Each item employed a 3-point scale (1 = rarely/never; 2 = occasionally; 3 = frequently). Subsequent reverse coding of positive affect items ensured higher aggregate scores corresponded to elevated depressive symptoms (Cronbach’s α = 0.709). Cumulative heat exposure Cumulative heat exposure for each participant was measured by the annual number of extreme heat days. Given that the CLASS survey is administered annually between January and April, cumulative heat exposure was operationalized using climate data from the previous calendar year (i.e., 2015, 2017, and 2019), ensuring temporal alignment between cumulative heat exposure and subsequent health outcomes. This approach supports causal inference by linking prior exposure to post-exposure assessments of depressive symptoms, following the methodology of Larson et al. [35], Falchetta et al. [36]. To account for China’s substantial regional climatic variation, we applied a percentile-based relative threshold approach using the 1971–2000 period as the climatological baseline, as recommended by prior studies [37]. For each city, the 90th percentile of historical daily maximum temperatures was calculated for every calendar day across the baseline years and used to define local extreme heat thresholds. An extreme heat event was recorded when the daily maximum temperature exceeded this benchmark. Each participant’s cumulative heat exposure was thus defined as the yearly frequency of extreme heat days in their residential city. Two-year and three-year cumulated heat exposure before the survey year of CLASS were calculated to further examine the impacts of cumulative heat exposure. To assess the robustness of the main findings, we conducted a supplementary analysis using an alternative threshold: the 95th percentile of daily mean temperatures based on a 2010–2015 climatological baseline. Cumulative heat exposure was recalculated accordingly. Preventive health services utilization Preventive health services utilization was assessed based on three service types offered by public health institutions based on community, including health screenings, health records management, and health education. Each service was coded as a binary variable (1 = ever used; 0 = never used), and a composite score reflecting overall preventive health services utilization was computed by summing the number of services ever used. Detailed measurement procedures and operational definitions are presented in the additional file 1 Table S1. Covariates This study was guided by the framework proposed by Lundgren et al. [38] and accounted for both socio-environmental and individual-level characteristics. Key demographic covariates included age, gender (0 = female, 1 = male), marital status (1 = married, 2 = widowed, 3 = divorced, 4 = never married), and education (on a 7-point scale from 1 = no formal education to 7 = college or above). Following the IPCC [3], four indicators were used to represent personal adaptive capacity: financial assets, financial support from children, social support, and chronic disease burden. Financial assets were assessed by counting the number of non-housing assets (range: 0–6). Financial support from children was measured on a 9-point ordinal scale, with higher scores indicating greater levels of support. Social support was evaluated using seven items reflecting interpersonal connectedness, with higher scores indicating stronger perceived support. Chronic disease burden was proxied by the total number of self-reported chronic conditions based on a 23-item checklist. At the city level, contextual variables included GDP (per 100 million CNY), green space (in hectares), number of hospitals, and number of medical beds. To address potential collinearity between age and survey year in the longitudinal models, a quadratic specification of age was employed. Statistical analysis This study employed two-way fixed-effects regression models to examine the impact of cumulative heat exposure on depressive symptoms among older adults, with a focus on the moderating role of preventive health services utilization. The econometric models are specified as follows: Equation (1) represents the baseline regression model, where the dependent variable, denotes depressive symptoms of each older adult, and represents cumulative heat exposure. Equation (2) is the model for examining the moderating effect, where ​ is the moderating variable, and denotes the interaction term between cumulative heat exposure and the moderating variable, with capturing the moderating effect. refers to the control variables, represents the individual fixed effect, represents the year fixed effect, ​ is the random error term. ​ are the constant terms, and ​ are the regression coefficients, measuring the effects of the respective variables. We estimated all moderation models using Inverse Probability of Treatment Weighting (IPTW). Propensity scores for any preventive health service use were estimated by logistic regression on all individual and city-level covariates at the 2016 baseline wave, and stabilized weights were truncated at the 99th percentile to limit the influence of extreme observations. The resulting interaction estimates are interpreted as exploratory associations rather than causal effects. Selection bias from missing city-level covariates was further addressed through a missing-data sensitivity analysis. Standardized Mean Differences were calculated to compare baseline characteristics between the retained and excluded samples. Multiple Imputation by Chained Equations using predictive mean matching was then applied, generating five imputed datasets to restore the full eligible sample, with pooled estimates compared against alternative sample specifications. Results Descriptive statistics Table 1 reports the descriptive statistics for the analytic sample, comprising 7,464 older adults aged 61 to 98 years, contributing a total of 21,383 observations. The average CES-D-9 score was 14.830 (SD = 4.331), indicating moderate levels of depressive symptoms. On average, participants were exposed to 58.090 extreme heat days per year (SD = 15.720). Preventive health services utilization was generally low (mean = 0.454, SD = 0.794). Among the three service categories, health screening utilization was the most frequently used (mean = 0.290, SD = 0.454), while health records utilization (mean = 0.106, SD = 0.307) and health education utilization (mean = 0.058, SD = 0.234) were seldom utilized. Table 1Descriptive statisticsVariableObsMeanStd. DevMinMaxDepressive symptoms21,38314.8304.331027Heat exposure21,38358.09015.72022104Preventive services21,3830.4540.79403Health screening21,3830.2900.45401Health records management21,3830.1060.30701Health education21,3830.0580.23401CovariatesAge21,38370.6006.4236198Gender21,3830.5070.50001Education21,3832.8571.32917Marriage21,3831.2860.50314Financial assets21,3830.7660.68503Financial support from children21,3839.9016.283056Social support21,38314.596.020034Chronic disease21,3831.5561.567016GDP_std21,3830.0021.003−0.7933.079Green space_std21,3830.0031.003−0.6453.182Hospitals_std21,383−0.0090.995−0.9745.221Medical beds_std21,3830.0011.001−1.0203.203Obs observations, std standard deviation, GDP gross domestic product Descriptive statistics Obs observations, std standard deviation, GDP gross domestic product The impact of cumulative heat exposure on depressive symptoms among older adults Table 2 presents the relationships between cumulative heat exposure and depressive symptoms among older adults. Cumulative exposure to extreme heat in the previous year significantly increased depressive symptoms (β = 0.013, p < 0.001), indicating that each additional extreme heat day was associated with a 1.370% increase in the standard deviation of depressive symptom scores. The effect diminished over longer exposure periods: the two-year cumulative exposure still had a significant but smaller effect (β = 0.008, p < 0.001), while the three-year cumulative exposure showed no statistically significant influence (β = 0.001, p = 0.800), suggesting temporal attenuation of the heat effect. Table 2The impact of cumulative heat exposure on depressive symptoms among older adultsVariablesdepressive symptomsdepressive symptomsdepressive symptomscoef (SE)coef (SE)coef (SE)Heat exposure0.013***(0.003)Heat exposure20.008***(0.002)Heat exposure30.001(0.002)Age_sq0.013***0.013***0.014***(0.001)(0.001)(0.001)Gender−0.635−0.580−0.624(1.009)(1.009)(1.010)Education−0.209**−0.211**−0.223***(0.082)(0.082)(0.082)Marriage0.0920.0770.088(0.168)(0.168)(0.168)Financial assets−0.079−0.094−0.085(0.058)(0.058)(0.058)Financial support from children−0.002−0.0010.000(0.009)(0.009)(0.009)Social support−0.010−0.009−0.010(0.007)(0.007)(0.007)Chronic disease0.163***0.171***0.165***(0.027)(0.027)(0.027)GDP_std−0.774***−0.735***−0.709***(0.243)(0.243)(0.244)Green space_std−0.939***−1.165***−0.964***(0.336)(0.339)(0.337)Hospitals_std−0.184***−0.152**−0.166**(0.071)(0.071)(0.071)Medical beds_std0.859*0.846*0.5350.013***(0.480)(0.476)2018.year−2.962***−2.919***−2.890***(0.363)(0.362)(0.363)2020.year−6.320***−6.299***−6.363***(0.719)(0.720)(0.720)Constant−48.240***−48.217***−50.045***(6.006)(6.005)(5.997)Observations21,38321,38321,383R-squared0.0530.0530.052Participants7,4647,4647,464Standard errors in parentheses: *** p < 0.010, ** p < 0.050, * p < 0.100. GDP_std = standardized city gross domestic product using z-score normalization; green space_std = standardized green space using z-score normalization; hospitals_std = standardized hospitals using z-score normalization; medical beds_std = standardized medical beds using z-score normalization; age_sq = age*age; coef = coefficient, SE = standard error The impact of cumulative heat exposure on depressive symptoms among older adults Standard errors in parentheses: *** p < 0.010, ** p < 0.050, * p < 0.100. GDP_std = standardized city gross domestic product using z-score normalization; green space_std = standardized green space using z-score normalization; hospitals_std = standardized hospitals using z-score normalization; medical beds_std = standardized medical beds using z-score normalization; age_sq = age*age; coef = coefficient, SE = standard error City-level characteristics exerted meaningful contextual effects. Higher GDP reduced depressive symptoms (β range: –0.709 to –0.774, p = 0.001–0.004), and greater green space had a similarly protective effect (β range: –0.939 to –1.165, all p < 0.004). Health infrastructure produced mixed results: a greater number of hospitals significantly reduced depressive symptoms (β range: –0.152 to –0.184, p = 0.006–0.023), whereas medical beds had inconsistent and statistically fragile effects (β range: 0.535 to 0.859, p = 0.048–0.204). Although the per-day coefficient is modest, the policy-relevant exposure unit is the heat season rather than a single day. We translated the per-day estimate into a seasonal-exposure scale by comparing the 10th and 90th percentiles of annual extreme-heat days, which yielded a substantively meaningful increase in depressive symptoms. Non-linearity was further examined through quadratic and restricted-cubic-spline specifications. Both rejected linearity and indicated a rising-then-saturating dose–response. Detailed results, including the joint tests and the dose–response curve, are reported in Supplementary Table S9 and Supplementary Figure S1. The moderating role of preventive health services utilization Table 3 evaluates the exploratory moderating effect of preventive health services utilization on the relationship between cumulative heat exposure and depressive symptoms. In Model (1), cumulative heat exposure remained positively associated with depressive symptoms (β = 0.009, p = 0.006), while the composite preventive health services utilization index significantly attenuated this association (β = −0.009, p = 0.001). Table 3Inverse Probability of Treatment Weighting (IPTW) fixed-effects models for the exploratory moderation of preventive services(1)Main models(2)Health screening(3)Health recordsmanagement(4)Health Education(5)Overall preventive health servicesdepressive symptomsdepressive symptomsdepressive symptomsdepressive symptomsdepressive symptomscoef (SE)coef (SE)coef (SE)coef (SE)coef (SE)Heat exposure0.009*** (0.003)0.014*** (0.004)0.011*** (0.003)0.011*** (0.003)0.016*** (0.004)Preventive health ServiceHealth screening0.650* (0.333)Health records1.222*** (0.452)Health Education2.593*** (0.628)Preventive services0.694*** (0.185)Interaction (Service×heat exposure)Health screening × heat exposure−0.009* (0.005)Health records × heat exposure−0.015** (0.007)Health education × heat exposure−0.036*** (0.009)Preventive services × heat exposure−0.009*** (0.003)Age_sq0.015*** (0.002)0.015*** (0.002)0.015*** (0.002)0.015*** (0.002)0.015*** (0.002)Chronic disease0.144*** (0.040)0.143*** (0.040)0.144*** (0.040)0.144*** (0.040)0.143*** (0.040)Other CovariatesYESYESYESYESYESYear Fixed EffectsYESYESYESYESYESIndividual Fixed EffectsYESYESYESYESYESConstant−57.225*** (8.969)−56.978*** (8.969)−57.755*** (8.987)−56.750*** (8.941)−57.075*** (8.966)Observations20,28220,28220,28220,28220,282Number of individuals6,8956,8956,8956,8956,895Robust standard errors clustered at the individual level are presented in parentheses. To address potential confounding by indication, models were weighted using Inverse Probability of Treatment Weighting (IPTW). Propensity scores were estimated at the baseline (2016) wave using individual and city-level covariates, with weights truncated at the 99th percentile. All models include individual and city-level controls, as well as year and individual fixed effects, which are omitted from the table for brevity*** p < 0.01, ** p < 0.05, * p < 0.1.age_sq = age*age; coef = coefficient, SE = standard error Inverse Probability of Treatment Weighting (IPTW) fixed-effects models for the exploratory moderation of preventive services Interaction (Service × heat exposure) Robust standard errors clustered at the individual level are presented in parentheses. To address potential confounding by indication, models were weighted using Inverse Probability of Treatment Weighting (IPTW). Propensity scores were estimated at the baseline (2016) wave using individual and city-level covariates, with weights truncated at the 99th percentile. All models include individual and city-level controls, as well as year and individual fixed effects, which are omitted from the table for brevity *** p < 0.01, ** p < 0.05, * p < 0.1.age_sq = age*age; coef = coefficient, SE = standard error Further analysis of specific preventive health services in Models (2), (3), and (4) revealed varied moderating associations. The interaction term for health education demonstrated the strongest attenuating effect (β = −0.036, p < 0.001), followed by health records management (β = −0.015, p = 0.034). However, after adjusting for baseline confounding via IPTW, the interaction between health screening and heat exposure was not statistically significant at the 0.05 level (β = −0.009, p = 0.064), suggesting that its unadjusted effect may partially reflect underlying health-seeking behaviors or pre-existing health needs. Robustness test Table 4 presents the results of robustness tests for the impact of cumulative heat exposure on depressive symptoms, along with the moderating effects of preventive health services utilization. These results provide additional insights into the consistency and reliability of the initial findings reported in Table 4. Table 4Robustness test of cumulative heat exposure’s impact on depressive symptoms and the moderating effect of preventive services utilizationVariables(1)(2)(3)withdepressdepressive symptomsdepressive symptomscoef (SE)coef (SE)coef (SE)Heat exposure0.004*0.017***(0.002)(0.003)Heat exposure950.011***(0.002)Bi-preventive services0.971***(0.263)Bi-preventive services × heat exposure−0.013***(0.004)Control variablesYesYesYesYear fixedYesYesYesIndividual fixedYesYesYesConstant-−52.864***−48.661***-(6.166)(6.004)Observations10,70320,88221,383R-squared-0.0550.054Participants3,6287,4197,464Standard errors in parentheses: *** p < 0.010, ** p < 0.050, * p < 0.100. withdepress = a binary metric of depressive symptoms (1 = “depressive symptoms > 15”, 0 = “depressive symptoms < = 15”); heat exposure95 = heat exposure measured by the 95th percentile temperature threshold; Bi-preventive services = binary overall preventive services utilization (1 = any historical use, 0 = never used); coef = coefficient, SE = standard error. a Meteorological records for certain cities contain data gaps during the baseline period, samples with missing data was deleted Robustness test of cumulative heat exposure’s impact on depressive symptoms and the moderating effect of preventive services utilization Standard errors in parentheses: *** p < 0.010, ** p < 0.050, * p < 0.100. withdepress = a binary metric of depressive symptoms (1 = “depressive symptoms > 15”, 0 = “depressive symptoms < = 15”); heat exposure95 = heat exposure measured by the 95th percentile temperature threshold; Bi-preventive services = binary overall preventive services utilization (1 = any historical use, 0 = never used); coef = coefficient, SE = standard error. a Meteorological records for certain cities contain data gaps during the baseline period, samples with missing data was deleted In Model (1), depressive symptoms were measured by a binary indicator (withdepress, 1 = “depressive symptoms > 15”, 0 = “depressive symptoms < = 15”), cumulative heat exposure still significantly increases depressive symptoms (β = 0.004, p = 0.075), supporting the original conclusion. Model (2) further evaluates the robustness of the findings by introducing a different measure for cumulative heat exposure (heat exposure95), which still significantly affects depressive symptoms (β = 0.011, p < 0.001). In Model (3), cumulative heat exposure remains a significant predictor of depressive symptoms (β = 0.017, p < 0.001), and preventive health services utilization measured by a binary indicator (β = –0.013, p < 0.001) continues to negatively moderate the effect of cumulative heat exposure on depressive symptoms. Full estimation details and extended robustness checks are reported in the additional file 1 Table S-robustness-test. We conducted four sets of sensitivity analyses. First, we examined potential selection bias arising from missing data through three complementary checks: an item-level missingness audit, a baseline comparison of standardized mean differences between retained and excluded samples, and Multiple Imputation by Chained Equations restoring the full eligible sample (Supplementary Tables S2–S4). Second, to address potential omitted-variable bias from unmeasured housing and household-composition factors, we constructed three CLASS-derived proxies: building type, floor of residence, and a living-alone indicator, and re-estimated the main model under five sequential specifications, complemented by a set of interaction analyses examining whether housing characteristics moderate the heat–depressive symptoms association (Supplementary Tables S5–S6). Third, to evaluate sensitivity to multicollinearity among the city-level covariates, we re-estimated the main model under five alternative treatments, including a principal-component reduction (Supplementary Table S7). Fourth, we re-estimated the main model with cluster-robust standard errors at three alternative aggregation levels (city, province, and two-way city × year; Supplementary Table S8). Across all four sets of sensitivity analyses, the cumulative heat-exposure coefficient remained stable in magnitude and statistically significant, supporting the robustness of the main findings. Detailed coefficients, standard errors, and full diagnostic results are provided in Supplementary Tables S2–S8. Exploratory heterogeneity tests To evaluate heterogeneity of the moderation effects across subgroups while accounting for baseline confounding and multiple testing, we estimated IPTW-weighted models and applied the Benjamini–Hochberg False-Discovery-Rate (FDR) adjustment (q = 0.05). The full results of the 28 exploratory subgroup tests are detailed in Supplementary Table S10. Of the 12 nominally significant interactions, 11 retained significance after FDR adjustment. Regarding gender, three preventive services significantly attenuated the heat–depressive-symptoms association among older men, health education (FDR p < 0.001), the overall preventive health services utilization index (FDR p < 0.001), and health screening (FDR p = 0.039), whereas none of the four moderators reached FDR significance among older women. Regarding residential location, all four preventive services significantly attenuated the heat–depressive-symptoms association among rural older adults (all FDR p < 0.001), whereas none reached FDR significance among urban older adults. Regarding geographic region, the middle region exhibited significant moderation by three preventive services: health education (FDR p < 0.001), the overall preventive health services utilization index (FDR p < 0.001), and health records management (FDR p = 0.045). The western region showed significant moderation only by health screening (FDR p = 0.018), while no moderator reached FDR significance in the eastern region. Detailed regression coefficients for the subgroup analyses are provided in Supplementary Tables S11–S13. Discussion Our findings indicate that cumulative heat exposure is significantly positively associated with depressive symptoms among older adults in China, highlighting the particular vulnerability of aging populations to climate-related depressive risks. These results are consistent with growing evidence suggesting that heat exposure adversely affects psychological well-being in later life [15, 16]. Notably, we observed a temporal decline in effect size: heat exposure in the most recent year had a stronger association with depressive symptoms than cumulative multi-year exposure. The attenuation of the extreme heat–depressive symptoms association over longer exposure windows suggests three complementary mechanisms identified in prior literature. First, psychological habituation may occur, as subjective distress often diminishes when individuals develop adaptive coping strategies to repeated environmental stressors [39]. Second, selective survival (mortality displacement) could mechanically dampen the estimated association if heat-vulnerable older adults experience adverse health events during earlier exposures, leaving a comparatively resilient surviving cohort [40]. Third, measurement attenuation may arise because extending the cumulative window inherently introduces exposure noise relative to the point-in-time survey assessments. These interpretations align with the saturating, non-linear dose–response curve observed in our supplementary analysis (Supplementary Figure S1). Disentangling these complementary mechanisms remains an important direction for future longitudinal research. By embedding our findings within a social-ecological model of climate resilience, this study elucidates how meso-level community resources interact with individual vulnerabilities to mitigate environmental stressors. Our results confirm that preventive health services function effectively as institutional adaptation capital. Rather than acting solely as passive healthcare deliverables, these services equip vulnerable older adults with the actionable knowledge and institutional support necessary to buffer against heat-related mental health deterioration. The protective associations were predominantly driven by health education and health records management, while the moderating role of health screening became marginally significant. Clinical services, such as screening, are highly susceptible to confounding by indication, meaning that older adults with pre-existing health vulnerabilities or pre-symptomatic conditions are more likely to seek out these services [5]. Once this baseline health-seeking bias was balanced, the direct buffering effect of screening might diminish. Concurrently, health records management potentially buffers mental health risks by ensuring continuity of care and supporting tailored follow-ups across recurrent heat events [27]. Notably, health education emerged as a robust moderator. This finding underscores the importance of addressing the well-documented knowledge–action gap in health-behavior research [41]. The literature suggests that cognitive awareness of heat risk does not automatically translate into protective behavior, primarily due to structural and economic barriers (e.g., the recurring costs of running cooling appliances, age-related mobility limitations, and social isolation) [42]. Effective community-based health education appears to bridge this knowledge–action gap not merely by disseminating information, but by imparting actionable, low-cost behavioral adaptation strategies (e.g., proper hydration routines, optimizing natural ventilation) [43]. This proactive behavioral empowerment proves highly effective in mitigating long-term heat-related distress, contrasting with passive clinical diagnostic services [44]. These protective associations were evident despite relatively low overall utilization rates, indicating substantial untapped potential for broader impact through the expansion of accessible public health services [2, 5]. The underutilization of such services remains a pervasive challenge. On the demand side, individual willingness and health literacy significantly influence engagement [45], where lower educational attainment may impede awareness of available community resources [46]. On the supply side, while health records systems exist at the institutional level, their data are often underutilized in proactive community outreach [47]. By addressing these modifiable barriers, the public health sector could more effectively leverage these social infrastructures to buffer the adverse psychological impacts of extreme heat [48]. Furthermore, our heterogeneity analyses revealed that the protective mechanisms of preventive health services are highly context-dependent, aligning with frameworks of resource dependence and socio-spatial vulnerability [48]. Regarding the urban–rural divide, rural older adults derived broad-spectrum benefits from preventive service utilization, all four moderators significantly attenuated the heat–depressive-symptoms association in this subgroup, whereas none reached FDR significance in urban older adults. This pattern likely reflects rural elders' elevated baseline heat vulnerability arising from limited cooling infrastructure, sparser built-environment heat refuges, and reduced access to advanced clinical facilities [49], conditions under which formal preventive services yield disproportionate marginal protective returns. Regarding gender, older men experienced robust benefits from multiple preventive services, health education, the composite service index, and health screening, while no moderator reached FDR significance among older women. This pattern may reflect older men's stronger reliance on formal community-based structures, given their comparatively weaker informal kin-based support networks relative to older women [50]. Finally, regional variations underscore the joint effects of climatic intensity and adaptive capacity. The middle region, which endures some of the most severe summer wet-bulb temperatures, uniquely showed significant moderation by three complementary services, suggesting that multi-faceted preventive support is required to buffer extreme heat stress in this high-exposure context [51]. The western region’s narrower significant pattern (only screening) and the eastern region's null pattern likely reflect lower heat-burden baselines in eastern coastal areas and possibly different service-delivery infrastructure in the west. The substantial heterogeneity observed across these demographic and geographic dimensions holds profound implications for climate justice. In the Chinese context, rural residents lacking cooling infrastructure, middle-region older adults enduring severe climatic intensity [25], and older men with weaker kin-based support networks [47] derive the most significant marginal benefits from formal preventive health services. Addressing these disparities highlights that expanding accessible, non-pharmaceutical primary healthcare is not only a public health imperative but also a fundamental tool for advancing equitable climate adaptation for aging populations. Based on the observed associations, our study offers insights for public health resource allocation. First, in rural townships, where all four preventive services significantly attenuate heat-related depressive risks, scaling up the comprehensive primary-care preventive service package, including screening, records management, education, and integrated delivery, should be prioritized prior to the summer season. Second, in the middle region (highest cumulative heat-exposure burden), brief heat-risk counseling (health education) should be formally integrated into the existing annual health assessments under the National Basic Public Health Service Program, complemented by health records management and integrated service delivery. Third, grassroots channels such as village doctors and community-level networks should be mobilized to conduct male-targeted outreach, compensating for the generally weaker informal kin-based support networks among older men. Fourth, heat-protection messaging and community welfare checks should be specifically directed toward older adults residing in bungalow-type housing, as they represent a particularly vulnerable subgroup to heat-related depressive symptoms compared to apartment residents. Several limitations should be acknowledged. First, although the fixed-effects modeling approach helps control for time-invariant confounding, residual bias from unmeasured time-varying factors cannot be ruled out. Second, regarding potential omitted-variable bias, while our measure of cumulative heat exposure is granular, the CLASS instrument does not capture critical microclimatic variations or indoor environments (e.g., indoor temperature, air-conditioner usage, daily time spent outdoors). Future research integrating indoor-temperature monitoring or time-use diaries is needed. Third, the reliance on self-reported preventive health services utilization introduces potential recall bias, as the binary operationalization prevents us from distinguishing between high-frequency and low-frequency users. Furthermore, there is an imperfect temporal misalignment between the recall-based service utilization measure (typically covering a 12-month reference window) and the precise calendar-year cumulative heat exposure. Fourth, despite the use of Inverse Probability of Treatment Weighting (IPTW), our design cannot fully eliminate dynamic confounding or reverse causality. The observed initial main effect of service utilization likely reflects higher pre-existing health needs (confounding by indication), and unobserved time-varying confounding may persist. Consequently, all moderation findings in this study should be interpreted as exploratory associations rather than strict causal relationships. Finally, the CLASS dataset exclusively samples community-dwelling older adults; thus, institutionalized or multiply marginalized subpopulations are not represented, and extending these findings to primary-care systems with different organizational structures should be done with caution. Conclusions This study demonstrates the adverse mental health impacts of cumulative heat exposure and the protective role of community-based interventions. It contributes to the literature in three ways. First, it shifts the focus from acute heatwaves to cumulative heat exposure to better represent long-term climate change. Second, it positions preventive health services as a community-level climate adaptation strategy, moving beyond individual behavioral adjustments. Third, it provides longitudinal evidence that existing primary healthcare infrastructures can serve as institutional adaptation capital to attenuate the psychological impacts of climate stressors. The analyses reveal that the protective effects of these services are context-dependent. Proactive interventions, including health education and continuous health records management, attenuate heat-related depressive symptoms more effectively than passive clinical screening. Furthermore, the heterogeneity findings highlight the necessity of climate justice. Institutional support is particularly important for subpopulations with heightened exposure or weaker informal support networks. Specifically, rural residents, who face elevated heat vulnerability due to limited cooling infrastructure, derived broad-spectrum benefits across all four preventive services; older adults in the central region, where cumulative heat exposure is most severe, benefited most from health education, records management, and integrated service delivery; and older men, who typically rely more on formal community-based structures than on informal kin support, showed consistent benefits from multiple preventive services. As climate change and population aging progress concurrently, expanding tailored primary health care is essential for equitable climate adaptation. Policymakers should integrate community-based preventive health services into national climate resilience strategies to protect vulnerable older adults from extreme heat. Supplementary Information Additional file 1: Supplementary Table S1. Specific survey questions of variables in the China Longitudinal Aging Social Survey. Supplementary Table S-Robustness test of cumulative heat exposure’s impact on depressive symptoms and the moderating effects of preventive services utilization. Supplementary Table S2. Distribution of missing data at the item level. Supplementary Table S3. Baseline characteristics comparison between retained and excluded samples. Supplementary Table S4. Sensitivity analyses for missing-data handling and Multiple Imputation (MI). Supplementary Table S5. Sequential inclusion of housing and household-composition proxies. Supplementary Table S6. Cumulative heat exposure× housing/composition interactions. Supplementary Table S7. VIF sensitivity analysis. Supplementary Table S8. Sensitivity to alternative cluster-robust standard errors Supplementary Table S9. Non-linear specifications. Supplementary Table S10. Multiple-testing correction for IPTW-weighted heterogeneity analyses (Benjamini-Hochberg FDR). Supplementary Table S11. IPTW-weighted fixed-effects models for the moderating effects of preventive health services by gender. Supplementary Table S12. IPTW-weighted fixed-effects models for the moderating effects of preventive health services by residential location (urban vs. rural). Supplementary Table S13. IPTW-weighted fixed-effects models for the moderating effects of preventive health services by geographic region. Figures. Supplementary Figure S1. Non-linear dose-response relationship between cumulative heat exposure and depressive symptoms. Acknowledgements Not applicable. Statement on ethics approval This study utilizes publicly available secondary data from the China Longitudinal Aging Social Survey (CLASS), which is a nationally representative dataset that collects information on older adults in China. As the data used in this research is secondary, publicly accessible, and anonymized, it does not require individual consent for participation. Additionally, since the CLASS dataset is available for public use and does not involve any direct interaction with participants, ethical approval from an institutional review board (IRB) or ethics committee is not required. All data analysis conducted in this study adheres to the ethical guidelines for the use of secondary data and ensures that the privacy and confidentiality of participants are respected. Authors’ contributions **Fan Yang**: Conceptualization; Data curation; Formal analysis; Funding acquisition; Investigation; Supervision; Writing-original draft; Writing-review & editing. **Shaowei Liu**: Data curation; Formal analysis; Investigation; Methodology; Software; Visualization; Writing-original draft; Writing-review & editing. **Junyao Zheng**: Data curation; Validation; Writing-original draft; Writing-review & editing. **Wenjie Li**: Data curation; Validation; Writing-review & editing. **Yiyan Zhou**: Data curation; Writing-review & editing. **Shihan Lu**: Data curation; Writing-review & editing. **Yifan Zhang**: Writing-review & editing. Funding This work was supported by the National Natural Science Foundation of China under Grant No. 72174118; the Hong Kong University Grants Committee General Research Fund (GRF) under Grant No.17603923. Data availability The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request. The data has not been previously presented orally or by poster at scientific meetings. Declarations Ethics approval and consent to participate Not applicable. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Shaowei Liu and Junyao Zheng are contributed equally to this work and should be considered as co-first author. References Associated Data Data Availability Statement The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request. The data has not been previously presented orally or by poster at scientific meetings.