Skip to main content
Download PDF
Download ePub

Investigating the impact of extreme weather events and related indicators on cardiometabolic multimorbidity

Archives of Public Health volume 82, Article number: 128 (2024) Cite this article

Abstract

Background

The impact of weather on human health has been proven, but the impact of extreme weather events on cardiometabolic multimorbidity (CMM) needs to be urgently explored.

Objectives

Investigating the impact of extreme temperature, relative humidity (RH), and laboratory testing parameters at admission on adverse events in CMM hospitalizations.

Designs

Time-stratified case-crossover design.

Methods

A distributional lag nonlinear model with a time-stratified case-crossover design was used to explore the nonlinear lagged association between environmental factors and CMM. Subsequently, unbalanced data were processed by 1:2 propensity score matching (PSM) and conditional logistic regression was employed to analyze the association between laboratory indicators and unplanned readmissions for CMM. Finally, the previously identified environmental factors and relevant laboratory indicators were incorporated into different machine learning models to predict the risk of unplanned readmission for CMM.

Results

There are nonlinear associations and hysteresis effects between temperature, RH and hospital admissions for a variety of CMM. In addition, the risk of admission is higher under low temperature and high RH conditions with the addition of particulate matter (PM, PM2.5 and PM10) and O3_8h. The risk is greater for females and adults aged 65 and older. Compared with first quartile (Q1), the fourth quartile (Q4) had a higher association between serum calcium (HR = 1.3632, 95% CI: 1.0732 ~ 1.7334), serum creatinine (HR = 1.7987, 95% CI: 1.3528 ~ 2.3958), fasting plasma glucose (HR = 1.2579, 95% CI: 1.0839 ~ 1.4770), aspartate aminotransferase/ alanine aminotransferase ratio (HR = 2.3131, 95% CI: 1.9844 ~ 2.6418), alanine aminotransferase (HR = 1.7687, 95% CI: 1.2388 ~ 2.2986), and gamma-glutamyltransferase (HR = 1.4951, 95% CI: 1.2551 ~ 1.7351) were independently and positively associated with unplanned readmission for CMM. However, serum total bilirubin and High-Density Lipoprotein (HDL) showed negative correlations. After incorporating environmental factors and their lagged terms, eXtreme Gradient Boosting (XGBoost) demonstrated a more prominent predictive performance for unplanned readmission of CMM patients, with an average area under the receiver operating characteristic curve (AUC) of 0.767 (95% CI:0.7486 ~ 0.7854).

Conclusions

Extreme cold or wet weather is linked to worsened adverse health effects in female patients with CMM and in individuals aged 65 years and older. Moreover, meteorologic factors and environmental pollutants may elevate the likelihood of unplanned readmissions for CMM.

Peer Review reports

Contributions to the literature

• The study deepens understanding of how air pollutants and various weather events affect cardiometabolic multimorbidity (CMM).

• It highlights increased risk for specific populations, especially females and the elderly, to inform targeted health strategies.

• The study identifies key laboratory parameters associated with unplanned readmissions, which helps in rational clinical decision making.

• Using employing machine learning models such as XGBoost, the study demonstrates innovative methods for predicting health outcomes associated with environmental factors.

Background

Cardiometabolic multimorbidity (CMM) is defined as the coexistence of two or three cardiometabolic diseases (CMDs), including type 2 diabetes mellitus (T2DM), ischemic heart disease (IHD) and stroke [1,2,3]. Among cardiovascular diseases, IHD is the most prevalent and is also considered a major threat to sustainable development in the 21st century [4, 5]. At the same time, diabetes is the leading risk factor for IHD [6, 7], with care costs at least 3.2 times higher than per capita healthcare expenditures, rising to 9.4 times higher when complications occur [8].

In contrast to single CMDs occurring on their own, combination of multiple CMDs has been found to be associated with multiplicative increase in a substantial reduction in life expectancy [9]. In recent years, the rapid global popularity of CMM has made it an emerging research focus for public health professionals [10]. It has been shown that elevated cholesterol and triglycerides are significantly associated with a progressively higher risk of cardiometabolic multimorbidity, particularly the progression of IHD to the multimorbidity of IHD and T2DM [11]. The results of the meta-analysis showed a significant association between high levels of triglycerides (TGs), low-density lipoprotein cholesterol (LDL-C) and fasting plasma glucose (FPG), and low levels of high-density lipoprotein cholesterol (HDL-C) with the progression of CMM, especially those related to lipid metabolism [12]. In addition several studies have reported fasting glucose, which is particularly associated with the coexistence of T2DM with vascular disease [13, 14]. Notably, metabolic risk factors have emerged as major drivers of IHD [15]. Despite the established associations between lifestyle factors or relevant indicators and individual CMD, the number of studies on the association between them and CMM is limited [16, 17].

With the dramatic changes in global climate in recent years, the impact of extreme weather on human health has also received widespread attention [18]. Several studies have shown that large fluctuations in extreme ambient temperatures due to climate change play a crucial role in diseases affecting the cardiovascular diseases and metabolic diseases [19,20,21,22]. The results of the study showed a nonlinear (J-shaped) relationship between temperature and IHD admissions [23]. It has been demonstrated that exposure to cold causes vasoconstriction and tachycardia, which can have a deleterious effect on people with IHD [24, 25] and T2DM [26]. Additionally, there may also be a 30-day lag in cold-related IHD deaths [27]. Patients with diabetes mellitus who experience an acute myocardial infarction at hypothermia had a sharply elevated risk of admission, with stronger effects than nondiabetic patients [28]. From a clinical perspective, patients with multiple conditions often require more comprehensive and integrated medical management than patients with only one or no chronic conditions [29]. Therefore, it is useful to investigate the association between ambient exposure and the odds of cause-specific hospitalizations in people with multimorbidity.

The valley topography of Urumqi surrounded by mountains on three sides and the unique natural environment resulting in the region’s winter inversion layer [30], create conditions that impede the horizontal flow of urban air pollutants and their accumulation over the city. This reduces the potential for diffusion and dilution [31]. Additionally, the large temperature difference between day and night as well as long, cold winters in Urumqi increase the severity of the cardiovascular disease burden in the Urumqi region [32]. Given that patients with CMM often require more comprehensive and integrated medical management, and that unplanned readmissions account for approximately $17 billion in annual Medicare spending, providers are increasingly interested in which patients are at risk for readmission [33]. In summary it is particularly important to examine the relationship between environmental exposures and biomarkers on the risk of hospitalization in patients with CMM.

This study utilized records of CMM patients hospitalized in Urumqi, China, from 2014 to 2022, and identified environmental factors and biomarkers associated with high hospitalization risk and burden by separately exploring the associations between environmental exposures and biomarkers on hospitalization risk in the population suffering from CMM and incorporating both into a prediction model for unplanned readmission risk in CMM patients. In a climate and environment that is becoming increasingly complex, conducting the above-related research can help identify sensitive populations so that healthcare resources can be appropriately allocated. Furthermore, it can provide new pathways and strategies for the local area to prevent current climate-related medical conditions.

Materials and methods

Research data source

The study included patients with ischemic heart disease (ICD: I20-I25) and type 2 diabetes mellitus (ICD-10: E11) who were recorded in the First Affiliated Hospital of Xinjiang Medical University, Xinjiang Uygur Autonomous Region Center for Disease Control and Prevention, and Urumqi Center for Disease Control and Prevention from 2014 to 2022. The data set primarily encompassed the following variables: gender, age, time of admission, chief complaint, and permanent residence, as well as the admission status. The indicators of each laboratory test at the time of admission include serum calcium, serum potassium, serum chloride, blood urea nitrogen (BUN), creatinine (Cr), fasting plasma glucose (FPG), total cholesterol (Tc), serum total bilirubin, high-density lipoprotein (HDL), albumin (ALB), alanine aminotransferase (ALT), aspartate aminotransferase/alanine aminotransferase ratio (AST/ALT), gamma-glutamyltransferase (GGT), etc.

The daily average temperature (°C), daily minimum temperature (°C), daily maximum temperature (°C), daily average wind speed (m/s), daily average air pressure (hpa) and daily average relative humidity (RH, %) data are from meteorological station in Urumqi. The daily average concentrations of particulate matter (PM, PM2.5 and PM10) and the maximum 8-hour average concentration of O3 (O3_8h) are mainly obtained from the national air quality monitoring station in Urumqi. The daily average exposure level for each patient was measured by neighborhood matching based on the station location and the district where the patient permanently resided. The study area and the distribution of stations are shown in Figure S1. The study was approved by the Ethics Review Committee of Xinjiang Medical University (Approval No.: XJYKDXR20240314058).

Definition of indicators

In this study, extreme low temperature and extreme low relative humidity (RH) were defined as the 1st percentile (P1), low temperature and low RH were defined as the 10th percentile (P10), high temperature and high RH were defined as the 90th percentile (P90), and extreme high temperature and extreme high RH were defined as the 99th percentile (P99). The daily temperature range (DTR) was defined as the difference between the daily maximum and minimum temperatures [34]. The temperature changes between neighboring days (TCN) were defined as the difference between the daily average temperature of the previous day and the next day [35, 36]. TCN and DTR have been identified in small studies as significantly associated with an increased risk of hospitalization and outpatient visits due to cardiovascular disease, metabolic disease, or other systemic conditions [37,38,39,40,41]. Given that Urumqi has a typical large temperature difference between day and night, it is necessary to consider it as an environmental variable when assessing its impact on CMM.

Statistical analysis

The study utilized various environmental and laboratory examination indicators to estimate the occurrence of CMM-related adverse events (admission volume or unplanned readmissions), and then constructed multiple machine learning predictive models to investigate the risk of unplanned readmission for CMM in Urumqi, China. A dataset related to CMM admission in Urumqi, China from 2014 to 2022 was collected, including temperature, RH, DTR and TCN, and laboratory examination indicators at admission. The research process is illustrated in Fig. 1.

Fig. 1
figure 1

Flow chart for inclusion and exclusion of patients with cardiometabolic multimorbidity (co-morbidity of ischemic heart disease and type 2 diabetes mellitus)

Full size image

Association between the meteorological conditions and CMM

A two-stage approach was used to explore this relationship. The first stage, using a time-stratified case-crossover design to investigate the association between environmental factors and CMM admissions, with a further focus on quantifying their interactive effects. This design has been widely used to assess the short-term effects of environmental exposures on a variety of health outcomes. In this design, each subject serves as his or her own control, and exposures on the day of admission are compared to exposures before and after the date of admission [42]. For each subject, the date of admission is defined as the case day, and all other dates that match the case day in terms of year, month, and day of the week are selected as corresponding control days. This design allows for comprehensive control of non-time-varying factors within the time stratum (such as age, gender, socioeconomic status, lifestyle, or chronic comorbidities), and seasonality [43]. In this stage, a distributed lag non-linear model (DLNM) was employed to study the relationship between exposure to temperature or RH and CMM admissions, while also considering the effects of TCN and DTR on CMM admissions. The model formula for the first stage is as follows:

$$Log[E({Y_t})] = \alpha + cb({A_j},{\beta _j}) + ns(B,df) + Stratum$$
(1)

Log[E(Yt)] represents the model linked by the “log” function; Yt is the dependent variable, representing the number of CMM admissions on the t-th day, where t is the observation date. j denotes the lag day; A is the independent variable, representing temperature, RH, TCN or DTR; βj represents the corresponding coefficients; B are the confounding factors, including daily average wind speed and daily average air pressure; cb represents the cross-basis function, selecting the natural spline function (ns) as the basis function; Based on the Quasi-Akaike information criterion (QAIC) [44] and previous research [45,46,47], the maximum lag days were determined to be 31 days. df represents the degrees of freedom for other confounding factors (such as wind speed, PM, or O3_8h), set at 3 [48]. Stratum is used to match the same day or week of the same month in the same year.

The second stage, based on the thresholds determined in the first stage for temperature and RH, the temperature and RH are analyzed together to explore the impact of dry cold or humid cold on CMM admissions. In this stage, the focus is on examining the association between dry-cold or wet-cold and CMM through analyses of relative risk (RR), attributable risk, and interaction effects. Firstly, the cold season is selected (from November to March each year). Secondly, temperatures below the 25th percentile (P25) are defined as cold, and RH above the 55th percentile (P55) is defined as humid [49]. Therefore, dry cold is defined as temperatures below the P25 and RH below the P55, and so on. Using the forward attribution fraction (f-AF), the backward attribution fraction (b-AF), and the total attribution fraction (AFtotal) to calculate the daily CMM admissions attributable to meteorological exposure, where the total attribution fraction is obtained as a weighted average of daily actual admission numbers [50]. The calculation formulas are as follows:

$$\eqalign{& b - A{F_{x,t}} = 1 - \exp ( - \sum\nolimits_{I = {I_0}}^L {{\beta _{{x_{t - 1}}}},l} ), \cr & f - A{F_{x,t}} = 1 - \exp ( - \sum\nolimits_{I = {I_0}}^L {{\beta _x}} ) \cr}$$
(2)

b-AFx, t represents the attribution to the lagged exposure x at time t-l0, …, t-L at time t. f-AFx, t represents the proportion of the single exposure x occurring at time t that is attributable to future times t + l0, …, t + L.

This study evaluates the additional interaction effects of temperature and RH using the following three metrics: Relative Excess Risk due to Interaction (RERI), Attributable Proportion (AP), and Synergy Index (SI). These metrics represent the effect caused by the interaction, the proportion of the effect attributable to the interaction, and the ratio between the combined effect and the individual effects, respectively. The formulas for calculating these three metrics are as follows:

$$\eqalign{& RERI = R{R_{11}} - R{R_{10}} - R{R_{01}} + 1, \cr & AP = RERI/R{R_{11}}, \cr & SI = (R{R_{11}} - 1)/[(R{R_{01}} - 1)] + (R{R_{10}} - 1) \cr}$$
(3)

Where RR11 represents the relative risk of low temperature and high RH; RR10 represents the relative risk of low temperature and low RH; RR01 represents the relative risk on non-cold and non-humid days.

Association between laboratory test indicators at admission and unplanned readmission for CMM

This phase of the study examined the risk between biological indicators at admission and unplanned readmission for the same illness within 30 days in patients with CMM. To address potential imbalances in the sample, this study employed 1:2 propensity score matching. This method is frequently employed to address the issue of exposure and non-exposure groups being unable to be grouped in a manner that is equivalent to randomization. Covariates are not balanced between groups, which affects the results of the analyses. The propensity score method, PSM, helps to control for the imbalance of confounders [51]. Our matching method employs nearest neighbor matching. The matching process involves pairing each individual in the observation group with the nearest individual in the control group until each individual in the observation group has been matched. In this study, the caliper value was set to 0.2 based on the results of Austin’s Monte Carlo simulation. The risk factors for unplanned readmission associated with CMM were explored by conditional logistic regression (CLR) of laboratory indicators at admission. The optimal CLR model demonstrated a nearly 20% improvement in classification accuracy over the straightforward logistic regression model [52].

Predictive model for readmission risk

The predicted outcome for this component was the risk of unplanned readmission. The inclusion and exclusion criteria for the predictive model are as follows:

Inclusion criteria

Patients with a discharge diagnosis with ICD-10 codes I20-I25 and a concurrent diagnosis of T2DM (ICD-10: E11).

Exclusion criteria

(1) Death during admission (discharge disposition = 5); (2) Transfer to another hospital during hospitalization; (3) Unclear discharge status.

In the variable screening phase, in addition to demographic characteristics, factors of environmental exposure and indicators of laboratory tests during the first 24 h of each admission were evaluated. To avoid overfitting, variables were selected and filtered using the least absolute shrinkage and selection operator (LASSO) [53, 54].

In the model-development phase, we developed nine machine learning models to predict the risk of unplanned readmission for CMM patients. In addition to eight commonly used models [55] such as Logistic Regression (LR), Random Forest (RF), k Nearest Neighbors (KNN), Support Vector Machines (SVM), naive Bayesian (NB), Decision Tree (D-tree), and Bagging Tree [56]. we introduced Extreme Gradient Boosting (XGBoost). XGBoost approximates the value of the loss function by using second-order Taylor series and further reduces the possibility of overfitting through regularization [57]. Subsequently, Logistic regression and other machine learning models were established and 5-fold cross-validation was performed hyperparameter tuning on the above algorithms.

In the data segmentation phase, the data was then divided into training and test sets in a 70–30% ratio. In the model-comparison phase, we examined and compared the performances of the nine predictive models through accuracy, F1 score, and by area under curves (AUC) of the receiver operating characteristic curves (ROC). Accuracy is the ratio of the number of correctly predicted samples to the total number of samples. F1 score is the reconciled average of precision and recall, which takes into account the precision and recall of the model. F1 score ranges from 0 to 1, and the larger the value, the better the model’s performance [18]. The AUC is used to measure the overall performance of the model at different thresholds, and the value of AUC ranges from 0 to 1, with larger values indicating better model performance [58].

In the interpretation process, the concept of SHapley Additive Explanations (SHAP) was introduced. As a model-independent interpretation method, SHAP helps to explain predictive models. SHAP typically visualizes machine learning predictions graphically for better presentation. SHAP typically visualizes machine learning predictions graphically for better presentation. For example, the SHAP variable importance graph succinctly demonstrates the contribution of each feature to the prediction performance: the larger the value, the higher the contribution [59]. The SHAP dependency graph describes the distribution of SHAP values of features across individuals. Since the SHAP values of the features vary from individual to individual, the predictions of the corresponding feature mappings of the individuals are also different [60].

The results were expressed in terms of relative risk (RR), cumulative relative risk (Cum RR) and 95% confidence intervals (CI). These measures were used to estimate the daily CMM admission risk associated with meteorological factors. The risk of readmission was evaluated as a hazard ratio (HR) and 95% CI. A greater HR indicates a higher risk of readmission occurring.

All statistical analyses were mainly conducted using the R statistical packages “dlnm”, “reportReg”, “MatchIt”, and “ggrcs” (version 4.3.3), and the level of statistical significance was set at 0.05.

Results

Descriptive analysis results

This study collected a total of 90,331 IHD patients from Urumqi, China, from 2014 to 2022, including the First Affiliated Hospital of Xinjiang Medical University, Xinjiang Center for Disease Control and Prevention and Urumqi Center for Disease Control and Prevention. After excluding (Fig. 1), a total of 29,993 hospitalized patients with IHD and comorbid type 2 diabetes mellitus (T2DM) were included in the association study between CMM and meteorological conditions.

From 2014 to 2022, the mean age of patients with cardiometabolic multimorbidity (CMM) in Urumqi city was 62.9 ± 10.7 years (mean ± standard deviation), refence to Table 1. Male patients accounted for 21,108 (70.4%) of admissions. 21,427 (71.4%) had a history of hypertension, and 17,825 (59.4%) had a history of smoking. The time series of meteorological factors and pollutants in Urumqi city from 2014 to 2022 are shown in Figure S2. It can be seen that all variables exhibit periodic patterns, and O3_8h concentrations are higher during the warm season. Correlations between various meteorological factors and pollutants are shown in Figure S3. Temperature is positively correlated with O3_8h and negatively correlated with RH and PM.

Table 1 Characteristics of cardiometabolic multimorbidity in Urumqi, 2014-2022
Full size table

The impact of meteorological conditions on admission for CMM

Association of temperature and relative humidity with CMM admission risks

By plotting contour plots (Fig. 2A and B), it was evident that the risk of admission for CMM was higher at low temperatures and high RH with a lag time of 31 days. Consequently, separate plots were created for extreme low temperature, low temperature, extreme high RH, and high RH at lags of 0–31 days for both single lagged relative risks (RR) (Fig. 2C to F) and cumulative relative risks (Cum RR) (Fig. 2G to J). As the lag days increased, the cumulative relative risk for CMM also increased. Additionally, extreme low temperature (Fig. 2C) had the highest relative risk of admission at lag 12 days (RR = 1.055, 95% CI: 1.020 ~ 1.090), while extreme high RH (Fig. 2E) had the highest relative risk of admission at lag 31 days (RR = 1.051, 95% CI: 1.011 ~ 1.094).

Fig. 2
figure 2

Contour plots of temperature and relative humidity (RH) for cardiometabolic multimorbidity and single-lagged and cumulative relative risks (Cum RR). (A) Contour plots of temperature against CMM; (B) Contour plots of RH against CMM; (C) Single-lag relative risk of extreme low temperature (P1: -17.5 °C) against CMM; (D) Single-lag relative risk of low temperature (P10: -10.9 °C) against CMM; (E) Single-lag relative risk of extreme high RH (P99: 87.5%) against CMM; (F) High RH (P90: 77.0%) single-lag relative risk to CMM; (G) extreme low temperature cumulative lag relative risk to CMM; (H) low temperature cumulative lag relative risk to CMM; (I) extreme high RH cumulative lag relative risk to CMM; (J) high RH cumulative lag relative risk to CMM.

Full size image

Table 2 described the impact of temperature and RH at different percentiles on the risk of hospitalization for CMM. Compared to the main model with only meteorological factors (temperature, RH), the cumulative lagged relative risks were higher after adjusting for PM and O3_8h. The results from the adjusted models showed that extreme low temperatures (P1: -17.5℃) exhibit the highest cumulative relative risk at a lag of 0–31 days (Cum RR = 2.460, 95% CI: 1.656 ~ 3.655). However, the cumulative relative risk for extreme high RH (P99: 87.5%) decreased slightly in the adjusted for PM and O3_8h, with a Cum RR of 1.957 at a lag of 0–31 days (95% CI: 1.162 ~ 3.297).

Table 2 Cumulative relative risk (cum RR) of different percentiles of temperature and relative humidity (RH) on hospitalizations for cardiometabolic multimorbidity (CMM) at different lag days
Full size table

Association between DTR and TCN with CMM admission risk

Plotting in 3D showed that the greater the difference in DTR (Figure S4A) or negative TCN (Figure S4B), the greater the risk of admission for patients with CMM. At different lag times, TCN and the risk of CMM admission exhibited an “S-shaped”, with the greatest effect observed at a TCN of -19℃ at lag 20 days (RR = 2.089, 95% CI: 1.722 ~ 2.534). DTR and the risk of CMM admission showed an “inverse L-shaped”, with the risk of CMM admission having the greatest effect at lag 20 days when the DTR was 23℃ (RR = 1.606, 95% CI: 1.452 ~ 1.776).

Association between relative humidity and CMM admission risk during the cold season

Exposure to accumulated lagged 31-day CMM hospitalization risk in dry-cold and wet-cold is illustrated in Figure S5A, where both dry and wet had an increased risk of CMM admission compared to normal RH, but the relative risk was higher for wet-cold (RR = 1.2828, 95% CI: 1.0616 ~ 1.5501). Stratified analyses revealed that female (RR = 1.4254, 95% CI: 1.0874 ~ 1.7635) and adults over 65 (RR = 1.2601, 95% CI: 1.0421 ~ 1.4781) were more sensitive to wet-cold. Figure S5B explains that 8.921% (f-AF) to 9.213% (b-AF) of CMM hospitalizations can be attributed to wet-cold. The interaction between low temperature and high RH is depicted in Figure S5C and Table S1. The results indicate a positive interaction effect between low temperature and high RH on the risk of hospitalization for CMM (RERI and AP > 0, and 95% CI does not include 0; SI > 1).

The interaction of low temperature and high humidity had a greater effect on female patients (RR = 1.247,95% CI: 1.094 ~ 1.424) and elderly patients over 65 years of age (RR = 1.225, 95% CI: 1.086 ~ 1.382).

The impact of laboratory testing indicators on unplanned readmissions for CMM

After screening, it was found that there were 908 patients who were unplanned readmissions, and 24,503 patients who were not unplanned readmissions. Due to the large imbalance between positive and negative samples, we used 1:2 PSM to balance the positive and negative samples. Table 3 shows a 1:2 PSM of all CMM patients without missing values. It can be observed that after matching, the differences between the various indicators were significantly reduced (P > 0.05).

Table 3 Characterization of cardiovascular metabolic multimorbidity before and after propensity score matching
Full size table

The selected indicators were divided into four groups (Q1, Q2, Q3, Q4) based on quartiles, and their results for three Logistic models on unplanned readmissions for CMM are shown in Fig. 3. In Model 3 (Wald test = 74.4, P = 0.004), after adjusting for other factors, compared to the Q1 group, the Q4 group showed a positive correlation between serum calcium (HR = 1.3632, 95% CI: 1.0732 ~ 1.7334), Creatinine (HR = 1.7987, 95% CI: 1.3528 ~ 2.3958), fasting plasma glucose (HR = 1.2579, 95% CI: 1.0839 ~ 1.4770), aspartate aminotransferase/ alanine aminotransferase ratio (HR = 2.3131, 95% CI: 1.9844 ~ 2.6418), alanine aminotransferase (HR = 1.7687, 95% CI: 1.2388 ~ 2.2986), and gamma-glutamyltransferase (HR = 1.4951, 95% CI: 1.2551 ~ 1.7351) and unplanned readmissions after CMM. On the other hand, serum bilirubin and HDL showed a protective effect (HR < 1).

Fig. 3
figure 3

Association of serum metabolism and other indicators with cardiometabolic multimorbidity using conditional logistic regression modeling. Note: Model 1 is unadjusted; Model 2 adjusted for age and gender; Model 3 adjusted for lifestyle factors such as smoking, drinking, sleep disorders, and cancer based on Model 2. BUN: Blood urea nitrogen; Cr: Creatinine; FPG: Fasting plasma glucose; Tc: Total cholesterol; HDL: High-density lipoprotein; ALB: albumin; ALT: Alanine aminotransferase; AST/ALT: Aspartate aminotransferase/Alanine aminotransferase ratio; GGT: Gamma-glutamyltransferase; Q1: First Quartile(Reference); Q2: Second Quartile; Q3: Third Quartile; Q4: Forth Quartile

Full size image

Predictive analysis results of unplanned readmissions for CMM

Variable selection

Using LASSO regression analysis, non-zero coefficient predictive variable selection was performed on 36 variables (Figure S7). Vertical lines were drawn at the minimum lambda value (λ = 0.0059) and at 1 SE minimum (λ = 0.029), and the optimal lambda value with the smallest average error was chosen. Nineteen non-zero coefficient predictive variables were selected, including serum calcium, Cr, FPG, HDL, serum bilirubin, ALT, AST/ALT ratio, GGT, history of smoking, history of hypertension, temperature at 31 days lag, relative humidity at 31 days lag, DTR at 20 days lag, TCN at 20 days lag, O3_8h at 31 days lag, PM at 31 days lag, sex, and age.

Model evaluation

Different machine learning models were established and compared with the traditional Logistic regression model, and the results of model evaluation based on Accuracy and F1 Score are shown in Fig. 4. The results show that after adjusting for meteorological, pollutant factors and their lag terms (Model 3), the AUC of each model has increased (Fig. 4A). XGBoost demonstrated higher predictive performance (AUC = 0.767, 95% CI:0.7486 ~ 0.7854), with an accuracy of 77.7% and an F1 Score of 0.894 (Fig. 4B).

Fig. 4
figure 4

Different models comparison of indicators. RF: Random Forest; KNN: k-nearest neighbor classification; D-tree: decision tree; SVM: support vector machines; Model 1: a model that includes only relevant laboratory examination indicators (serum calcium, Cr, FPG, HDL, serum bilirubin, ALT, AST/ALT ratio, GGT), totaling 8 indicators; Model 2: based on Model 1, adjusts for history of smoking, history of hypertension, sex and age, totaling 12 indicators; Model 3: based on Model 1 and Model 2, further adjusts for environmental factors and their lag times (temperature, RH, DTR, TCN, PM, and O3_8h), totaling 19 indicators; AUC: area under the receiver operating characteristic (ROC) curves

Full size image

Analysis of the contributions of model features

Figure S8 shows the features importance values based on the XGBoost model, reflecting the contribution of the features to the model. By analyzing the contribution of each input feature to the model, the top five significant features in order: DTR, temperature, RH, PM10, and age. We found that weather events or pollutants play a larger role in adverse outcomes for unplanned readmissions for CMM.

Individual sample analysis

Figure 5 shows the SHAP force plot for two randomly selected instances from the actual estimated results. In the figure, E[f(x)] represents the model’s predicted expected value given the input x, indicating the average predicted output for a specific input. Meanwhile, f(x) represents the model’s predicted output for the input x. If the model’s output is to the right of the predicted expected value, there may be a risk of unplanned readmission for CMM. Red arrows indicate input variables that increase the risk of readmission, and blue arrows indicate input variables that are detrimental to readmission. The size (number) of areas occupied by the variable in each arrow indicates the degree of influence of the variable. An 83-year-old man who had an unplanned readmission during the study period (Fig. 5A) had a higher risk of unplanned readmission as predicted (f(x) = 0.940); a 49-year-old man who had no readmissions during the follow-up study period (Fig. 5B) had a lower predictive score (0.234).

Fig. 5
figure 5

SHAP-based force plot of the XGBoost explainer for two selected patients. AST.ALT: AST/ALT ratio; temp: Temperature; rhum: relative humidity

Full size image

Feature dependence analysis

The SHAP dependency plot for individual variables was analyzed to observe the relationship between feature importance and SHAP values in XGBoost model estimation, demonstrating how individual features influenced the model’s predictive outcomes (Figure S9). The SHAP value greater than 0 indicates a positive association with the risk of unplanned readmission to CMM. The results were similar to the restricted cubic spline results (Figure S6), FPG and CMM admissions were predominantly “J-shaped”. When the temperature was below 12.23℃, the risk of readmission for CMM increased with decreasing temperature, reaching a maximum at -21℃ (HR = 2.135, 95% CI = 1.278 to 3.566). Humidity in the range of 53.45–64.56% was a possible reduction in the risk of readmission for CMM, which was more affected by excessive relative humidity. When the Cr concentration was greater than 72.73µmol/L, the readmission risk of CMM was positively correlated with Cr.

Discussion

Many factors can lead to unplanned readmission, such as environmental or meteorological, complications due to surgical treatment, biological indicators and other factors, which will bring a great economic burden to patients [33]. Environment or weather often has a lagging effect on the occurrence of disease [61, 62], so this study firstly explores the nonlinear and lagging relationship of environment on the risk of CMM admission. Secondly, we considered the association between indicators at admission and readmission risk in patients with CMM. Finally, different classification machine learning models were developed to predict the incidence of unplanned readmission rates for CMM. Weather (temperature, relative humidity), pollutants (PM, O3_8h), and biological metrics at the time of admission were utilized to combine to predict the incidence of unplanned readmission rates for CMM. We found that the optimal lag time for temperature, RH, PM, and O3_8h for CMM patient admission was 31 days, and the optimal time after for DTR and TCN was 20 days. Low temperature and high RH had a greater effect on CMM admission, so we tried to explore the effect of wet and dry cold on the risk of CMM admission, and found that the effect of wet and cold was greater, especially among women and adults older than 65 years of age. The XGBoost model has better performance in predicting the risk of readmission in CMM patients and can predict the risk of readmission well [63, 64].

In this study, we found a lagged effect of temperature and RH on CMM admissions, with extreme low temperature (-17.5 °C) and extreme high RH (87.5%) having the highest relative risk of CMM admissions. After adjusting for PM and O3_8h, we found that temperature increased the cumulative relative risk of CMM admission, while high RH decreased the cumulative relative risk. This may be due to the fact that high humidity promotes deposition of pollutants to the ground or other surfaces more quickly, reducing the concentration of airborne pollutants in the air [65]. This study therefore builds on this result to investigate the relationship between RH and CMM admissions during the cold season. The results found that there was a significant relative risk of wet and cold on CMM admissions, especially for female and adults over 65. A study in Lanzhou, China [23] showed that adverse effects increased with lag time, and that the risk of hospitalization for IHD was significantly increased at -12ºC in females and in people > 65 years of age, which is in consistent with the results of the present study. Another study showed that due to a powerful and rapid thermogenesis mechanism that is shivering, women preferred warmer environments and felt less comfortable in cold environments than men did, given the same temperature conditions [66]. We observed adverse effects of negative TCN, which is consistent with the results of another study in China [35]. An additive interaction between low temperature and high RH on the risk of hospital admission for CMM was found in the interaction analysis. A Chinese study found that wet and cold events accounted for 2.99% of excess deaths, with higher burdens in the elderly and cardiovascular disease, in addition to a synergistic additive interaction between low winter temperatures and high humidity [67]. This may be due to intolerance of metabolic and motor responses to wet and cold conditions [68]. Therefore, reducing outings during the cold season in women and older CMM patients would potentially reduce the risk of hospital admission.

Conditional logistic regression of laboratory indicators at admission for patients hospitalized with CMM revealed that serum calcium, Cr, FPG, ALT, AST/ALT ratio, and GGT were risk factors for unplanned readmission with CMM. Epidemiologic studies had shown that elevated GGT activity was associated with a variety of CMM risk factors, including traditional cardiovascular risk factors, systemic inflammation, oxidative stress burden, and a variety of co-morbidities, which may negatively impacted patient risk and prognosis [69]. In the present study, serum bilirubin and HDL were found to be protective effect against CMM. Previous studies had suggested that higher bilirubin may be a potential target for reducing the risk of CMM, that it is positively correlated with HDL [70], and that bilirubin may favorably regulate lipid oxidation [71, 72] and inflammatory markers that modulate obesity and glucose sensitivity [73]. The results of a systematic study showed a significant negative correlation between HDL and the progression of CMM [12], which was consistent with the results of the present study. In addition, there was a J-shaped relationship between patients with CMM and adverse outcome events, consistent with our results. FPG as a traditional diagnostic biomarker of T2DM, FPG levels may provide additional information on health status and may also serve as an indicator of T2DM deterioration and a potential biomarker of CMM [74, 75]. One study found that in an adjusted model, subjects in the fourth quartile of GGT and ALT/AST ratio had a 3.29- and 2.94-fold increased risk of developing metabolic syndrome, respectively, compared with those in the first quartile [76], which is similar to the results of the present study. Previous studies have linked elevated ALT to T2DM [77], meanwhile previous meta-analyses had shown that elevated levels of ALT and GGT were associated with an increased risk of T2DM, and GGT has also been found to be a significant risk factor relative to ALT [78].

Adjusting the environmental factors and their lag terms in the XGBoost model for prediction was found to have a greater AUC (0.767) and higher accuracy, with an F1 score of 0.894, which indicates a better performance of the model. A previous study correlated personal electronic health data with environmental data as well as time-resolved climate, pollution, allergen, and influenza case data, which were well predicted by XGBoost (AUC = 0.730–0.742) [79]. A machine learning-based study of predicting unplanned readmissions for cardiovascular disease in cancer hospitalized patients found XGBoost to have the best performance [80]. This allows for a more comprehensive assessment of patient risk and allows for the influence of environmental factors on predictive outcomes to be taken into account when interpreting model results [81], and enables physicians and researchers to better understand how patient risk is shaped. However, there are some limitations to this study. The first is that the effect of persistent cold temperatures or persistent high humidity on CMM hospitalizations was not adequately considered, which will also be explored in subsequent studies. The second is that this study did not externally validate the predictive model despite adjusting it for different variables, and future studies will conduct validated predictive model analysis.

In summary, low temperature and high relative humidity, with an interaction effect, have a large effect on adverse events in female with CMM and inpatients aged 65 years and older. Incorporation of environmental factors into a predictive model for unplanned readmissions in CMM resulted in a significant improvement in model performance. We found that weather events or pollutants may play a large role in unplanned readmissions for CMM. Age, gender, serum calcium, Cr, FPG, high-density lipoprotein, serum bilirubin, ALT, AST/ALT ratio, history of smoking, and history of hypertension were risk factors contributing to the risk of CMM admission. Finally, by using population-based data, this study provides insights that may help guide clinical decisions related to unplanned readmission prevention strategies for hospitalized CMM patients. This will enable relevant authorities such as the Urumqi government, healthcare organizations, and environmental or meteorological departments to integrate environmental changes and healthcare resources more rationally and develop more appropriate prevention strategies.

Data availability

The data involved in this study are not publicly available due to privacy but are available from the authors on reasonable request.

Abbreviations

AF:

Attributable fraction

ALB:

Albumin

ALT:

Alanine aminotransferase

AN:

Attributable number of cases

AST/ALT:

Aspartate aminotransferase/alanine aminotransferase ratio

AUC:

Average area under the receiver operating characteristic curve

BUN:

Blood urea nitrogen

Cr:

Creatinine

Cum RR:

Cumulative relative risks

CVD:

Cardiovascular disease

DLNM:

Distributed lag non-linear models

Dow:

Day-of-the-week

DTR:

Daily temperature range

FPG:

Fasting plasma glucose

GAM:

Generalized additive model

GGT:

Gamma-glutamyltransferase

HDL:

High-density lipoprotein

ICD-10:

The International Classification of Diseases, tenth edition

IHD:

Ischemic heart disease

O3_8h:

The maximum daily 8-hour average ozone concentration

PM:

Particulate matter

PM10 :

Particulate matter with ≤ 10 μm in aerodynamic diameter

PM2.5 :

Particulate matter with ≤ 2.5 μm in aerodynamic diameter

QAIC:

Quasi-likelihood for Akaike information criterion

RH:

Relative humidity

RR:

The relative risk

Tc:

Total cholesterol

TCN:

Temperature changes between neighboring days

T2DM:

Type 2 diabetes mellitus

References

  1. Saxon DR, Reiter-Brennan C, Blaha MJ, Eckel RH. Cardiometabolic Medicine: development of a New Subspecialty. J Clin Endocrinol Metab 2020, 105(7).

  2. Zhang D, Tang X, Shen P, Si Y, Liu X, Xu Z, Wu J, Zhang J, Lu P, Lin H, et al. Multimorbidity of cardiometabolic diseases: prevalence and risk for mortality from one million Chinese adults in a longitudinal cohort study. BMJ Open. 2019;9(3):e024476.

    Article  PubMed  PubMed Central  Google Scholar 

  3. Ralston J, Nugent R. Toward a broader response to cardiometabolic disease. Nat Med. 2019;25(11):1644–6.

    Article  CAS  PubMed  Google Scholar 

  4. Khan MA, Hashim MJ, Mustafa H, Baniyas MY, Al Suwaidi S, AlKatheeri R, Alblooshi FMK, Almatrooshi M, Alzaabi MEH, Al Darmaki RS, et al. Global epidemiology of Ischemic Heart Disease: results from the global burden of Disease Study. Cureus. 2020;12(7):e9349.

    PubMed  PubMed Central  Google Scholar 

  5. Roth GA, Mensah GA, Johnson CO, Addolorato G, Ammirati E, Baddour LM, Barengo NC, Beaton AZ, Benjamin EJ, Benziger CP, et al. Global Burden of Cardiovascular diseases and Risk factors, 1990–2019: Update from the GBD 2019 study. J Am Coll Cardiol. 2020;76(25):2982–3021.

    Article  PubMed  PubMed Central  Google Scholar 

  6. Global regional, national burden of diabetes. From 1990 to 2021, with projections of prevalence to 2050: a systematic analysis for the global burden of Disease Study 2021. Lancet. 2023;402(10397):203–34.

    Article  Google Scholar 

  7. Global burden. Of 369 diseases and injuries in 204 countries and territories, 1990–2019: a systematic analysis for the global burden of Disease Study 2019. Lancet. 2020;396(10258):1204–22.

    Article  Google Scholar 

  8. Khan MAB, Hashim MJ, King JK, Govender RD, Mustafa H, Al Kaabi J. Epidemiology of type 2 diabetes - global burden of Disease and Forecasted trends. J Epidemiol Glob Health. 2020;10(1):107–11.

    Article  PubMed  PubMed Central  Google Scholar 

  9. Skou ST, Mair FS, Fortin M, Guthrie B, Nunes BP, Miranda JJ, Boyd CM, Pati S, Mtenga S, Smith SM. Multimorbidity. Nat Rev Dis Primers. 2022;8(1):48.

    Article  PubMed  PubMed Central  Google Scholar 

  10. Freisling H, Viallon V, Lennon H, Bagnardi V, Ricci C, Butterworth AS, Sweeting M, Muller D, Romieu I, Bazelle P, et al. Lifestyle factors and risk of multimorbidity of cancer and cardiometabolic diseases: a multinational cohort study. BMC Med. 2020;18(1):5.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  11. Zhao Y, Zhuang Z, Li Y, Xiao W, Song Z, Huang N, Wang W, Dong X, Jia J, Clarke R, et al. Elevated blood remnant cholesterol and triglycerides are causally related to the risks of cardiometabolic multimorbidity. Nat Commun. 2024;15(1):2451.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  12. Jin Y, Xu Z, Zhang Y, Zhang Y, Wang D, Cheng Y, Zhou Y, Fawad M, Xu X. Serum/plasma biomarkers and the progression of cardiometabolic multimorbidity: a systematic review and meta-analysis. Front Public Health. 2023;11:1280185.

    Article  PubMed  PubMed Central  Google Scholar 

  13. Sarwar N, Gao P, Seshasai SR, Gobin R, Kaptoge S, Di Angelantonio E, Ingelsson E, Lawlor DA, Selvin E, Stampfer M, et al. Diabetes mellitus, fasting blood glucose concentration, and risk of vascular disease: a collaborative meta-analysis of 102 prospective studies. Lancet. 2010;375(9733):2215–22.

    Article  CAS  PubMed  Google Scholar 

  14. Cullen P, von Eckardstein A, Souris S, Schulte H, Assmann G. Dyslipidaemia and cardiovascular risk in diabetes. Diabetes Obes Metab. 1999;1(4):189–98.

    Article  CAS  PubMed  Google Scholar 

  15. Wang W, Hu M, Liu H, Zhang X, Li H, Zhou F, Liu YM, Lei F, Qin JJ, Zhao YC, et al. Global burden of Disease Study 2019 suggests that metabolic risk factors are the leading drivers of the burden of ischemic heart disease. Cell Metab. 2021;33(10):1943–56. e1942.

    Article  CAS  PubMed  Google Scholar 

  16. Han Y, Hu Y, Yu C, Guo Y, Pei P, Yang L, Chen Y, Du H, Sun D, Pang Y, et al. Lifestyle, cardiometabolic disease, and multimorbidity in a prospective Chinese study. Eur Heart J. 2021;42(34):3374–84.

    Article  PubMed  PubMed Central  Google Scholar 

  17. Jin Y, Liang J, Hong C, Liang R, Luo Y. Cardiometabolic multimorbidity, lifestyle behaviours, and cognitive function: a multicohort study. Lancet Healthy Longev. 2023;4(6):e265–73.

    Article  PubMed  Google Scholar 

  18. Alahmad B, Khraishah H, Royé D, Vicedo-Cabrera AM, Guo Y, Papatheodorou SI, Achilleos S, Acquaotta F, Armstrong B, Bell ML, et al. Associations between Extreme temperatures and Cardiovascular cause-specific mortality: results from 27 countries. Circulation. 2023;147(1):35–46.

    Article  PubMed  Google Scholar 

  19. Faye M, Dème A, Diongue AK, Diouf I. Impact of different heat wave definitions on daily mortality in Bandafassi, Senegal. PLoS ONE. 2021;16(4):e0249199.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  20. Geirinhas JL, Russo A, Libonati R, Trigo RM, Castro LCO, Peres LF, Magalhães M, Nunes B. Heat-related mortality at the beginning of the twenty-first century in Rio De Janeiro, Brazil. Int J Biometeorol. 2020;64(8):1319–32.

    Article  PubMed  Google Scholar 

  21. Shi W, Schooling CM, Leung GM, Zhao JV. Early-life exposure to ambient air pollutants and kidney function in adolescents: a cohort study based on the ‘Children of 1997’ Hong Kong birth cohort. Public Health. 2024;230:59–65.

    Article  CAS  PubMed  Google Scholar 

  22. Li N, Ma J, Liu F, Zhang Y, Ma P, Jin Y, Zheng ZJ. Associations of apparent temperature with acute cardiac events and subtypes of acute coronary syndromes in Beijing, China. Sci Rep. 2021;11(1):15229.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  23. Meng X, Jin J, Han X, Han B, Bai M, Zhang Z. Effect of Meteorological Factors, Air Pollutants on Daily Hospital Admissions for Ischemic Heart Disease in Lanzhou, China. Cardiology 2023.

  24. Houdas Y, Deklunder G, Lecroart JL. Cold exposure and ischemic heart disease. Int J Sports Med. 1992;13(Suppl 1):S179–181.

    Article  PubMed  Google Scholar 

  25. Pan R, Okada A, Yamana H, Yasunaga H, Kumazawa R, Matsui H, Fushimi K, Honda Y, Kim Y. Association between ambient temperature and cause-specific cardiovascular disease admissions in Japan: a nationwide study. Environ Res. 2023;225:115610.

    Article  CAS  PubMed  Google Scholar 

  26. Luo J, He G, Xu Y, Chen Z, Xu X, Peng J, Chen S, Hu J, Ji G, Liu T, et al. The relationship between ambient temperature and fasting plasma glucose, temperature-adjusted type 2 diabetes prevalence and control rate: a series of cross-sectional studies in Guangdong Province, China. BMC Public Health. 2021;21(1):1534.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  27. Imai C, Barnett AG, Hashizume M, Honda Y. The role of Influenza in the Delay between Low Temperature and Ischemic Heart Disease: evidence from Simulation and Mortality Data from Japan. Int J Environ Res Public Health 2016, 13(5).

  28. Lam HCY, Chan JCN, Luk AOY, Chan EYY, Goggins WB. Short-term association between ambient temperature and acute myocardial infarction hospitalizations for diabetes mellitus patients: a time series study. PLoS Med. 2018;15(7):e1002612.

    Article  PubMed  PubMed Central  Google Scholar 

  29. Xu Z, Yi W, Bach A, Tong S, Ebi KL, Su H, Cheng J, Rutherford S. Multimorbidity and emergency hospitalisations during hot weather. EBioMedicine. 2024;104:105148.

    Article  PubMed  PubMed Central  Google Scholar 

  30. Wang L, Zhu S, Liu Z, Lu J, Xiang Z, Lan J, Liu J, Yu M, Chen Y, Chen J. Characterization of particulate matter and its extinction ability during different seasons and weather conditions in Sinkiang, China: two case studies. Environ Sci Pollut Res Int. 2020;27(18):22414–22.

    Article  CAS  PubMed  Google Scholar 

  31. Zhu L, Lu Y, Cheng F, Zhang L, Yusan A, Alifu X, Alimasi A, Alemujiang X. Association between atmospheric pollutant levels and oxidative stress in pregnant women and newborns in Urumqi. BMC Public Health. 2023;23(1):330.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  32. Meng X, Wu Y, Pan Z, Wang H, Yin G, Zhao H. Seasonal characteristics and particle-size distributions of Particulate Air pollutants in Urumqi. Int J Environ Res Public Health 2019, 16(3).

  33. Swain MJ, Kharrazi H. Feasibility of 30-day hospital readmission prediction modeling based on health information exchange data. Int J Med Inf. 2015;84(12):1048–56.

    Article  Google Scholar 

  34. Phosri A, Sihabut T, Jaikanlaya C. Short-term effects of diurnal temperature range on hospital admission in Bangkok, Thailand. Sci Total Environ. 2020;717:137202.

    Article  CAS  PubMed  Google Scholar 

  35. Zha Q, Chai G, Zhang ZG, Sha Y, Su Y, Wu T. Impact of temperature changes between neighboring days on cardiovascular disease hospital admissions among suburban farmers in Qingyang, Northwest China. Int J Biometeorol. 2022;66(6):1233–45.

    Article  PubMed  Google Scholar 

  36. Lei X, Liu L, Chen R, Liu C, Hong J, Cao L, Lu Y, Dong X, Chen X, Qiu X, et al. Temperature changes between neighboring days and childhood asthma: a seasonal analysis in Shanghai, China. Int J Biometeorol. 2021;65(6):827–36.

    Article  PubMed  Google Scholar 

  37. Zhou CL, Lv LS, Jin DH, Xie YJ, Ma WJ, Hu JX, Wang CE, Xu YQ, Zhang XE, Lu C. Temperature change between neighboring days contributes to years of life lost per death from respiratory disease: a multicounty analysis in Central China. Int J Environ Res Public Health 2022, 19(10).

  38. Zhang Q, Kan H. Author response: Effect of temperature changes between neighboring days on acute aortic dissection in non-heating periods. Lancet Reg Health West Pac. 2023;30:100659.

    PubMed  Google Scholar 

  39. Zhan Z, Zhao Y, Pang S, Zhong X, Wu C, Ding Z. Temperature change between neighboring days and mortality in United States: a nationwide study. Sci Total Environ. 2017;584–585:1152–61.

    Article  PubMed  Google Scholar 

  40. Song X, Wang S, Hu Y, Yue M, Zhang T, Liu Y, Tian J, Shang K. Impact of ambient temperature on morbidity and mortality: an overview of reviews. Sci Total Environ. 2017;586:241–54.

    Article  CAS  PubMed  Google Scholar 

  41. Lim YH, Hong YC, Kim H. Effects of diurnal temperature range on cardiovascular and respiratory hospital admissions in Korea. Sci Total Environ. 2012;417–418:55–60.

    Article  PubMed  Google Scholar 

  42. Wu Y, Li S, Guo Y. Space-Time-Stratified Case-Crossover Design in Environmental Epidemiology Study. Health Data Sci 2021, 2021:9870798.

  43. Li X, Zhao Y, Zhang D, Kuang L, Huang H, Chen W, Fu X, Wu Y, Li T, Zhang J, et al. Development of an interpretable machine learning model associated with heavy metals’ exposure to identify coronary heart disease among US adults via SHAP: findings of the US NHANES from 2003 to 2018. Chemosphere. 2023;311(Pt 1):137039.

    Article  CAS  PubMed  Google Scholar 

  44. Gasparrini A, Guo Y, Hashizume M, Lavigne E, Zanobetti A, Schwartz J, Tobias A, Tong S, Rocklöv J, Forsberg B, et al. Mortality risk attributable to high and low ambient temperature: a multicountry observational study. Lancet. 2015;386(9991):369–75.

    Article  PubMed  PubMed Central  Google Scholar 

  45. Guo Y, Gasparrini A, Armstrong BG, Tawatsupa B, Tobias A, Lavigne E, Coelho MS, Pan X, Kim H, Hashizume M, et al. Temperature variability and mortality: a multi-country study. Environ Health Perspect. 2016;124(10):1554–9.

    Article  PubMed  PubMed Central  Google Scholar 

  46. Aklilu D, Wang T, Amsalu E, Feng W, Li Z, Li X, Tao L, Luo Y, Guo M, Liu X, et al. Short-term effects of extreme temperatures on cause specific cardiovascular admissions in Beijing, China. Environ Res. 2020;186:109455.

    Article  CAS  PubMed  Google Scholar 

  47. Zhai G, Qi J, Zhang X, Zhou W, Wang J. A comparison of the effect of diurnal temperature range and apparent temperature on cardiovascular disease among farmers in Qingyang, Northwest China. Environ Sci Pollut Res Int. 2022;29(19):28946–56.

    Article  PubMed  Google Scholar 

  48. Song X, Wang S, Li T, Tian J, Ding G, Wang J, Wang J, Shang K. The impact of heat waves and cold spells on respiratory emergency department visits in Beijing, China. Sci Total Environ. 2018;615:1499–505.

    Article  CAS  PubMed  Google Scholar 

  49. Fu SH, Gasparrini A, Rodriguez PS, Jha P. Mortality attributable to hot and cold ambient temperatures in India: a nationally representative case-crossover study. PLoS Med. 2018;15(7):e1002619.

    Article  PubMed  PubMed Central  Google Scholar 

  50. Gasparrini A, Leone M. Attributable risk from distributed lag models. BMC Med Res Methodol. 2014;14:55.

    Article  PubMed  PubMed Central  Google Scholar 

  51. Kane LT, Fang T, Galetta MS, Goyal DKC, Nicholson KJ, Kepler CK, Vaccaro AR, Schroeder GD. Propensity score matching: a statistical method. Clin Spine Surg. 2020;33(3):120–2.

    Article  PubMed  Google Scholar 

  52. Zhu K, Lou Z, Zhou J, Ballester N, Kong N, Parikh P. Predicting 30-day hospital readmission with publicly available administrative database. A conditional logistic regression modeling Approach. Methods Inf Med. 2015;54(6):560–7.

    Article  CAS  PubMed  Google Scholar 

  53. Ranalli MG, Salvati N, Petrella L, Pantalone F. M-quantile regression shrinkage and selection via the Lasso and Elastic Net to assess the effect of meteorology and traffic on air quality. Biom J. 2023;65(8):e2100355.

    Article  PubMed  Google Scholar 

  54. Li Y, Lu F, Yin Y. Applying logistic LASSO regression for the diagnosis of atypical Crohn’s disease. Sci Rep. 2022;12(1):11340.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  55. Jiang T, Gradus JL, Rosellini AJ. Supervised machine learning: a brief primer. Behav Ther. 2020;51(5):675–87.

    Article  PubMed  PubMed Central  Google Scholar 

  56. Wang K, Tian J, Zheng C, Yang H, Ren J, Liu Y, Han Q, Zhang Y. Interpretable prediction of 3-year all-cause mortality in patients with heart failure caused by coronary heart disease based on machine learning and SHAP. Comput Biol Med. 2021;137:104813.

    Article  PubMed  Google Scholar 

  57. Hou N, Li M, He L, Xie B, Wang L, Zhang R, Yu Y, Sun X, Pan Z, Wang K. Predicting 30-days mortality for MIMIC-III patients with sepsis-3: a machine learning approach using XGboost. J Transl Med. 2020;18(1):462.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  58. Namdar K, Haider MA, Khalvati F. A modified AUC for training convolutional neural networks: taking confidence into account. Front Artif Intell. 2021;4:582928.

    Article  PubMed  PubMed Central  Google Scholar 

  59. Lundberg SM, Erion G, Chen H, DeGrave A, Prutkin JM, Nair B, Katz R, Himmelfarb J, Bansal N, Lee SI. From local explanations to Global understanding with explainable AI for trees. Nat Mach Intell. 2020;2(1):56–67.

    Article  PubMed  PubMed Central  Google Scholar 

  60. Yi F, Yang H, Chen D, Qin Y, Han H, Cui J, Bai W, Ma Y, Zhang R, Yu H. XGBoost-SHAP-based interpretable diagnostic framework for alzheimer’s disease. BMC Med Inf Decis Mak. 2023;23(1):137.

    Article  Google Scholar 

  61. Nhung NTT, Amini H, Schindler C, Kutlar Joss M, Dien TM, Probst-Hensch N, Perez L, Künzli N. Short-term association between ambient air pollution and pneumonia in children: a systematic review and meta-analysis of time-series and case-crossover studies. Environ Pollut. 2017;230:1000–8.

    Article  PubMed  Google Scholar 

  62. Landguth EL, Knudson J, Graham J, Orr A, Coyle EA, Smith P, Semmens EO, Noonan C. Seasonal extreme temperatures and short-term fine particulate matter increases pediatric respiratory healthcare encounters in a sparsely populated region of the intermountain western United States. Environ Health. 2024;23(1):40.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  63. Ru B, Tan X, Liu Y, Kannapur K, Ramanan D, Kessler G, Lautsch D, Fonarow G. Comparison of Machine Learning Algorithms for Predicting Hospital readmissions and worsening heart failure events in patients with heart failure with reduced ejection fraction: modeling study. JMIR Form Res. 2023;7:e41775.

    Article  PubMed  PubMed Central  Google Scholar 

  64. Zhao P, Yoo I, Naqvi SH. Early prediction of unplanned 30-Day hospital readmission: Model Development and Retrospective Data Analysis. JMIR Med Inf. 2021;9(3):e16306.

    Article  Google Scholar 

  65. Lou C, Liu H, Li Y, Peng Y, Wang J, Dai L. Relationships of relative humidity with PM(2.5) and PM(10) in the Yangtze River Delta, China. Environ Monit Assess. 2017;189(11):582.

    Article  PubMed  Google Scholar 

  66. Kaikaew K, van den Beukel JC, Neggers S, Themmen APN, Visser JA, Grefhorst A. Sex difference in cold perception and shivering onset upon gradual cold exposure. J Therm Biol. 2018;77:137–44.

    Article  PubMed  Google Scholar 

  67. Li M, Fang W, Meng R, Hu J, He G, Hou Z, Zhou M, Zhou C, Zhu S, Xiao Y, et al. The comparison of mortality burden between exposure to dry-cold events and wet-cold events: a nationwide study in China. Sci Total Environ. 2023;904:166859.

    Article  CAS  PubMed  Google Scholar 

  68. Thompson RL, Hayward JS. Wet-cold exposure and hypothermia: thermal and metabolic responses to prolonged exercise in rain. J Appl Physiol (1985). 1996;81(3):1128–37.

    Article  CAS  PubMed  Google Scholar 

  69. Ndrepepa G, Colleran R, Kastrati A. Gamma-Glutamyl transferase and the risk of atherosclerosis and coronary heart disease. Clin Chim Acta. 2018;476:130–8.

    Article  CAS  PubMed  Google Scholar 

  70. Seyed Khoei N, Wagner KH, Sedlmeier AM, Gunter MJ, Murphy N, Freisling H. Bilirubin as an indicator of cardiometabolic health: a cross-sectional analysis in the UK Biobank. Cardiovasc Diabetol. 2022;21(1):54.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  71. Hinds TD Jr., Stec DE. Bilirubin, a Cardiometabolic Signaling Molecule. Hypertension 2018, 72(4):788–795.

  72. Frazier-Wood AC, Ordovas JM, Straka RJ, Hixson JE, Borecki IB, Tiwari HK, Arnett DK. The PPAR alpha gene is associated with triglyceride, low-density cholesterol and inflammation marker response to fenofibrate intervention: the GOLDN study. Pharmacogenomics J. 2013;13(4):312–7.

    Article  CAS  PubMed  Google Scholar 

  73. Gordon DM, Neifer KL, Hamoud AA, Hawk CF, Nestor-Kalinoski AL, Miruzzi SA, Morran MP, Adeosun SO, Sarver JG, Erhardt PW, et al. Bilirubin remodels murine white adipose tissue by reshaping mitochondrial activity and the coregulator profile of peroxisome proliferator-activated receptor α. J Biol Chem. 2020;295(29):9804–22.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  74. Lee SH, Han K, Kim HS, Cho JH, Yoon KH, Kim MK. Predicting the Development of myocardial infarction in Middle-aged adults with type 2 diabetes: a risk model generated from a Nationwide Population-based Cohort Study in Korea. Endocrinol Metab (Seoul). 2020;35(3):636–46.

    Article  CAS  PubMed  Google Scholar 

  75. Lee JH, Han K, Huh JH. The sweet spot: fasting glucose, cardiovascular disease, and mortality in older adults with diabetes: a nationwide population-based study. Cardiovasc Diabetol. 2020;19(1):44.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  76. Kohsari M, Moradinazar M, Rahimi Z, Pasdar Y, Shakiba E. Liver Enzymes and Their Association with Some Cardiometabolic Diseases: Evidence from a Large Kurdish Cohort. Biomed Res Int 2021, 2021:5584452.

  77. Abro MUR, Butt A, Baqa K, Waris N, Khalid M, Fawwad A. Association of serum liver enzyme Alanine Aminotransferase (ALT) in patients with type 2 diabetes. Pak J Med Sci. 2018;34(4):839–43.

    Article  PubMed  PubMed Central  Google Scholar 

  78. Ko SH, Baeg MK, Han KD, Ko SH, Ahn YB. Increased liver markers are associated with higher risk of type 2 diabetes. World J Gastroenterol. 2015;21(24):7478–87.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  79. Hurst JH, Zhao C, Hostetler HP, Ghiasi Gorveh M, Lang JE, Goldstein BA. Environmental and clinical data utility in pediatric asthma exacerbation risk prediction models. BMC Med Inf Decis Mak. 2022;22(1):108.

    Article  Google Scholar 

  80. Han S, Sohn TJ, Ng BP, Park C. Predicting unplanned readmission due to cardiovascular disease in hospitalized patients with cancer: a machine learning approach. Sci Rep. 2023;13(1):13491.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  81. Liu W, Dai Q, Bao J, Shen W, Wu Y, Shi Y, Xu K, Hu J, Bao C, Huo X. Influenza activity prediction using meteorological factors in a warm temperate to subtropical transitional zone, Eastern China. Epidemiol Infect. 2019;147:e325.

    Article  PubMed  PubMed Central  Google Scholar 

Download references

Acknowledgements

This study was supported and funded by grants from the National Natural Science Foundation of China. We thank the Center for Disease Control and Prevention of Xinjiang Uygur Autonomous Region and the First Affiliated Hospital of Xinjiang Medical University for their invaluable assistance with administration, and data collection.

Funding

This work was supported by National Natural Science Foundation of China (Grant Nos. 72163033, 72064036, 72174175).

Author information

Authors and Affiliations

  1. School of Public Health, Xinjiang Medical University, Urumqi, China

    Di Wu & Yu Shi

  2. Center for Disease Control and Prevention of Xinjiang Uygur Autonomous Region, Urumqi, China

    ChenChen Wang

  3. The First Affiliated Hospital of Xinjiang Medical University, Urumqi, China

    Cheng Li

  4. Center for Disease Control and Prevention of Urumqi, Urumqi, China

    Yaoqin Lu

  5. School of Public Health, Nanjing Medical University, Nanjing, China

    Chunfang Wang

  6. School of Computer and Information Engineering, Xinjiang Agricultural University, Urumqi, China

    Weidong Zhu

  7. School of Agriculture, Xinjiang Agricultural University, Urumqi, China

    Tingting Sun

  8. School of Nursing and Public Health, Yangzhou University, Yangzhou, China

    Junjie Han

  9. School of Medical Engineering and Technology, Xinjiang Medical University, Urumqi, China

    Yanling Zheng & Liping Zhang

Authors
  1. Di Wu
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Yu Shi
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. ChenChen Wang
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Cheng Li
    View author publications

    You can also search for this author in PubMed Google Scholar

  5. Yaoqin Lu
    View author publications

    You can also search for this author in PubMed Google Scholar

  6. Chunfang Wang
    View author publications

    You can also search for this author in PubMed Google Scholar

  7. Weidong Zhu
    View author publications

    You can also search for this author in PubMed Google Scholar

  8. Tingting Sun
    View author publications

    You can also search for this author in PubMed Google Scholar

  9. Junjie Han
    View author publications

    You can also search for this author in PubMed Google Scholar

  10. Yanling Zheng
    View author publications

    You can also search for this author in PubMed Google Scholar

  11. Liping Zhang
    View author publications

    You can also search for this author in PubMed Google Scholar

Contributions

Conception and design of study：D. Wu ; acquisition of data: C.C. Wang, C. Li, Y.Q. Lu; analysis and/or interpretation of data: D. Wu, Y. Shi and W.D. Zhu,; Drafting the manuscript: D. Wu, C.F. Wang and J.J. Han; Revising the manuscript critically for important intellectual content: T.T. Sun, Y.L. Zheng, L.P. Zhang; Approval of the version of the manuscript to be published: D. Wu, Y. Shi, C.C. Wang, C. Li, Y.Q. Lu, C.F. Wang, W.D. Zhu, T.T. Sun, JJ Han, Y.L. Zheng, L.P. Zhang.

Corresponding author

Correspondence to Liping Zhang.

Ethics declarations

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Supplementary Material 1

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Wu, D., Shi, Y., Wang, C. et al. Investigating the impact of extreme weather events and related indicators on cardiometabolic multimorbidity. Arch Public Health 82, 128 (2024). https://doi.org/10.1186/s13690-024-01361-x

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s13690-024-01361-x

Keywords

Archives of Public Health

ISSN: 2049-3258