Development and Validation of a Model to Predict Acute Kidney Injury Following Wasp Stings: A Multicenter Cohort Study

serve as promising predictive tools to assess the probability of the AKI following wasp stings. the lactated dehydrogenase (LDH), HB, total bilirubin (TBI), aspartate aminotransferase (AST), urine volume, alanine aminotransferase (ALT), and activated partial thromboplastin time (APTT), creatine kinase (CK) and serum creatinine (Scr) et al, were recorded and analyzed.

At present, the mechanism of AKI following wasp stings is not completely clear. Wasp venoms is thought having direct toxic effect for kidney, in which main effective poisonous components include phospholipase A 2 (PLA 2 ) and melittin [11,16,17]. Besides, hemoglobin (HB) released by hemolysis and myoglobin released by rhabdomyolysis are considered that may cause secondary damage to kidney [18,19].
At present, the AKI following wasp stings is a serious health hazard [17]. Besides, it's worth noting that our previous study indicated that patients with AKI following wasp stings usually died within 72 hours after admission, so that a better understanding of AKI following wasp stings is helpful to early diagnose and treat [12]. Therefore, we performed a multicenter center prospective cohort to identify risk factors associated with AKI following wasp stings and established an individual nomogram and formula of the model to predict the probability of AKI following multiple wasp stings.

Database and Patient Selection
This was a multicenter prospective observational cohort study that was conducted in 18 hospitals in China from Jul 2015 to Dec 2019 and 508 patients with AKI following wasp stings were included. Patients who suffering from stings by wasp or other bee species, whether AKI occurred or not, were included. Patients who met the following criteria were excluded: (1) lack of enough clinical data to analyze; or (2) disapprove of participating. The study was approved by Institutional Research Ethics Committees of our institution (Approval No. of the ethics committee: 2014-156). Written informed consents were obtained from all included patients or their legally authorized representatives.
The data were recorded and collected from the electronic medical records system. Demographic and clinical characteristics, such as age, sex, number of stings, time from sting to admission, Sequential Organ Failure Assessment (SOFA) score et al, and important laboratory measures, such as the lactated dehydrogenase (LDH), HB, total bilirubin (TBI), aspartate aminotransferase (AST), urine volume, alanine aminotransferase (ALT), and activated partial thromboplastin time (APTT), creatine kinase (CK) and serum creatinine (Scr) et al, were recorded and analyzed.

De nition
In the study, the patients with wasp stings were divided into AKI group or non-AKI group. AKI was de ned by the 2012 Kidney Disease Improving Global Outcomes guidelines [20]: (1) Scr increased to ≥ 26.5 µmol/ L (0.3 mg/dL) within 48 hours; (2) Scr increased to 1.5 times within 7 days ; or (3) urine volume < 0.5 ml/ (kg• h) for 6 hours.
In the logistic regression analysis, continuous variables were transformed into categorical variables according to their reference range and common clinical transformation methods.

Statistical Analysis
We excluded some variables that missing more than 15% and interpolated some variables that missing less than 15% according to multiple interpolation method. And some patients were excluded for lack of enough data to analyze. All eligible patients were randomly assigned 3:1 into the training set and the validation set. The training set mainly constructed a nomogram and a predictive formula of the model for predicting the AKI following multiple wasp stings and the validation set was constructed for external validation.
Continuous variables are expressed as the mean (standard deviation [SD]) which were compared by t test or the median (interquartile range [IQR]) which were compared by Wilcoxon test according to whether the variable coincided with a normal distribution. Categorical variables were expressed as proportions, which were compared by the chi-squared test or Fisher's exact test as appropriate.
The univariable logistic regression analysis to detect the risk factors of AKI following wasp stings and the variables with P-value < 0.1 were introduced into the multivariable logistic regression analysis. In addition, the individual nomogram and the predictive formula of the model were constructed according to the result of multivariable logistic regression analysis. The predictive formula was constructed according to the previous study [21]. The internal validation and external validation were performed in the training set and validation set respectively to assess the accuracy of the model by a bootstrap validation method with 200 resamples. The receiver operating characteristic curve (ROC) analysis was performed to assess the model. Cut off of the model was obtained based on Youden Index. The calibration curves were utilized for estimating the consistency between the actual observed outcome and the nomogram predicted AKI probability. Decision curve analysis (DCA) demonstrated the net bene t associated with the use of the nomogram-derived probability for the prediction of AKI following wasp stings.
All statistical analyses were performed by SPSS (SPSS Inc., Chicago, IL) software for Windows version 23.0. and the packages (rms, hmisc, etc.) in R software version 3.6.1 (http://www.r-project.org), with a two-sided P value < 0.05 considered statistically signi cant.

Characteristics of Eligible Patients
In the study, 547 patients with wasp stings were screened, and 39 patients were excluded due to missing necessary dada (as shown in the Fig. 1). Finally, 508 eligible patients were randomly assigned into the training set (n = 381) and the validation set (n = 127). There was no signi cant difference between the training set and the validation set in the distributions of demographic and disease characteristics such as year, gender, number of AKI, condition of wasp stings (see Table A1 in Additional le 1). In the training set, of the 381 patients, 118 (31.0%) patients with AKI and 263 (69.0%) without AKI. The demographic and clinical characteristics between the AKI group and non-AKI group in the training set were presented in the Table 1. The proportion of patients received renal replacement therapy (RRT) were higher and the time from stings to admission was longer in the AKI group than those in the non-AKI group (all P value < 0.05). Meanwhile, the proportion of wasp stings, sting at head and face, number of stings ≥ 30, area of stings ≥ 25% and the level of ALT, AST, LDH, TBI, CK in the AKI group were higher than those in the non-AKI group (all P value < 0.05). To determine the risk factors for AKI following wasp stings, the univariable and multivariable logistic regression analysis were performed in the training set. Variables with P < 0.1 in the univariable logistic regression analysis introduced into multivariable logistic regression analysis (see Table A2 in Additional le 2). Subsequently, multivariable logistic regression analysis identi ed ve variables as independence risk factors for AKI following wasp stings: number of stings, HB < 110 g/dl, TBI > 34 mg/dl, ALT > 40 U/L and APTT > 47 s (as shown in the Table 2). In the formula, P indicated the predicted AKI probability, X indicated the variables that included in the model (X1-X6 represented that number of stings is 15-50, number of stings > 50, HB < 110 g/dl, TBI > 34 mg/dl, ALT > 40 U/L, APTT > 47 s respectively). X was assigned as 1 while patient was consistent with the variable, otherwise, X was assigned as 0. The probability of AKI following multiple wasp stings could be calculated according to the formula based on the patient's individual clinical characteristics. When calculated probability > 0.338 (the cut off of the model), the AKI was predicted to occur, and when calculated probability ≤ 0.338, the AKI might not occur.
Of note, as shown in the Table 3 Table 3 and Figure A1 in Additional le 3).

Construction of a Nomogram and Validation
The individual nomogram was constructed based on the result of the multivariable logistic regression analysis to predict AKI for wasp sting patients in the training set too. As shown in the Fig. 2 and Table A3 in Additional le 4, scores were signed for each variable. A total point could be calculated by adding all points based on patient's individual clinical characteristics, which was lower meaning a lower probability of AKI. The nomogram was validated internally in the training set and were validated externally in the validation set. The calibration curves for AKI prediction showed excellently accordance between predictions of the nomogram and the actual observations in the training set (as shown in the Fig. 3) and validation set (see Figure A2 in Additional le 5). DCA veri ed the net bene t associated with the use of the nomogram-derived probability for AKI following wasp stings in the training set (as shown in the Fig. 3) and validation set (see Figure A3 in Additional le 6).

Discussion
In our study, number of stings, HB < 110 g/dl, TBI > 34 mg/dl, ALT > 40 U/L and APTT > 47 s were identi ed as independent risk factors for AKI following wasp stings according to univariable and multivariable logistic regression analysis in the training set. And the predictive formula and the individual nomogram that included those independent risk factors was developed and validated to predict the probability of the AKI following wasp stings. Those were demonstrated having the su cient accuracy and good predictive capability based on the internal validation and external validation in the training set and validation set respectively. In addition, the predictive formula and individual nomogram both have clinical signi cance to assess the probability of the AKI following wasp stings and make a decision for therapy by those easy, convenient and effective methods.
At present, wasp stings are reported frequently, especially in rural areas where patients may have low incomes [22][23][24], which is a common challenge to society. In the present study, the incidence rate of AKI is 30.5% (155/508) in patients with wasp stings, which is nearly equal to results in the previous studies that reported the incidence rate of AKI is 20-25% [11,25,26]. In addition, the mortality in patients with wasp stings is 5.7% (29/508), however the mortality (17.4% [27/155]) in the AKI group is apparently higher than that (0.6% [2/353]) in the non-AKI group (P < 0.001 as shown Figure A4 in Additional le 7). Those indicated that it is signi cant to understand, risk predict and early diagnose the AKI following wasp stings. Of note, 55.6% (15/27) patients with AKI following wasp stings died within 72 hours after admission, that is similar to the result of previous study [14]. At present, AKI following wasp stings with relatively high mortality rate and rapid onset were called the "Silent Killer" because it threated to human public health. Therefore, early detection and diagnosis should be performed promptly to help clinicians make therapeutic decisions for obtaining a good prognosis.
However, the prediction model of AKI following wasp stings is rarely reported in previous studies. We established the predictive formula and the individual nomogram based on the independent risk factor, which predicted the AKI of patients undergoing wasp stings with a good validation. As reported in previous studies, nomogram is extensively used to predict the probability of a disease or a clinical outcome based on multiple variables [27][28][29][30]. In the present study, the visual nomogram could calculated the speci c probability of AKI following wasp stings based on the sum of the scores of each risk factor, that is the most user-friendly tool to judge the speci c situation of each patient [31]. The nomogram is intuitive and easy-to-understand not only clinicians but for patients as well, which might make it easy for communication between clinicians and patients. Of note, nomograms have never ever been reported for AKI following wasp stings to our knowledge, we conduct the rst nomogram to predict the AKI following wasp stings.
Besides, the predictive formula also is conducted to assess whether occur AKI or not in patients with wasp stings, that is judged according to whether the calculated probability > the cut off of the model (0.338) or not. One of the two methods could be selected according to the habits and preferences of clinicians, or both to mutually detect and support the results. Those are composed of common clinical parameters, that are easy to obtain from laboratorial blood tests. In addition, the su cient accuracy and good predictive capability of the model are veri ed by AUC of ROC (0.912 in the training set and 0.936 in the validation set), and the net bene t is veri ed by DCA. Therefore, the model might provide a clinical assistance in early recognition, detection, diagnosis and intervention of AKI following wasp stings.
According to previous studies, AKI is induced in wasp stings based on direct toxicity of the venom components, hypotension, intravascular hemolysis and rhabdomyolysis [32]. The venom components, such as PLA 2 that is mainly included in wasp venom and melittin that is the mainly included in the bee venom, both have strong hemolytic toxicity and direct toxic effect for inducing the apoptosis of renal tubule epithelial cells [16,33]. Hypotension might lead to ischemic renal lesion, which is induced by main components of bee venom such as hyaluronidase, apamin and substances induced by those venom themselves such as histamine, serotonin, bradykinin.
Besides, rhabdomyolysis and hemolysis induced AKI by renal vasoconstriction, formation of intratubular deposits of myoglobin and direct cytotoxicity of myoglobin and HB that are release from muscle and red blood cells. However, there is no full understanding in the mechanism through which renal damage occurs. Nonetheless, it is certain that rhabdomyolysis and hemolysis are thought play important roles in AKI following wasp stings.
In the present study, compared with non-AKI group, we actually nd the levels of CK, ALT, AST elevate in the AKI group, which might be associated with rhabdomyolysis. We also nd LDH increase and anemia (HB in the red blood cells decrease) in the AKI group, which might be is concerned with hemolysis. They also are the risk factors of AKI following wasp sting according to univariable logistic regression analysis.
Continuous variables are transformed into categorical variables according to their reference range, which is bene cial to full use the results through constructing the model.
According to multivariable logistic regression analysis, number of stings, HB < 110 g/dl, TBI > 34 mg/dl, ALT > 40 U/L and APTT > 47 s are considered as independence risk factors and could help to predict AKI in patients with wasp stings. Rhabdomyolysis and hemolysis could induce the level of indirect bilirubin to elevate, which could be one explanation of the increasing of TBI in AKI following wasp stings. According to the study reported, disseminated intravascular coagulation (DIC) might be a possible factor that contributes to AKI, described in rhabdomyolysis. While DIC occurred, thromboplastin is released and micro thrombi is formed in the glomeruli, which causing the consequent glomerular ltration rate reduction [34]. DIC might induce APTT prolonged by increasing the consumption of coagulation factors. We think those might explain why prolonged APTT is an independence risk factors in AKI following wasp stings. In addition, we also nd that wasp compared with other bee species, sting at head and face compared with other locations, the greater number of stings and large area of sting might be associated with AKI following wasp stings.
However, a recent study with a retrospective cohort study involving 112 patients conducted by Hai Yuan et al, which showed that elevated leukocytes, high myoglobin, high urinary monocyte chemotactic protein-1 (MCP-1) are the independence risk factors of AKI induced by multiple wasp stings [25]. In fact, our results also nd that, while compared with non-AKI, the level of leukocytes is higher in the patients with AKI following wasp stings. However, we do not think elevated leukocytes is an independence risk factor of AKI following wasp stings based on the multivariable logistic regression analysis. There exists a difference between the two studies. It's worth noting that the results are obtained based on the single center as well as the small sample size in the Hai Yuan's study. In addition, 12 variables included in the multivariable logistic regression analysis to nd out the independence risk factors based on only 54 patients who occurred AKI. We think those might restrict external validity and cause the different results in the Hai Yuan's study. Besides, Hai Yuan's study only reported some independence risk factors, which did not construct a model to predict AKI for fully utilization the result. We construct the rst model based on a large data from the multicenter prospective cohort study and validate it.
There exist some limitations in the present study. First, there are some variables missing too much so that we have to exclude them such as cystatin -C and urine protein, although we already try our best to collect the complete data of each patient. Second, although data were collected from multicenter, the enrolled patients all were of the same ethnicity, which might limit the scalability of the model. Third, we select the variable by forward stepwise in multivariable logistic regression analysis, that might induce the nal model that contain terms of little values. Besides, the validation was performed by bootstrapping technology, however there also need further external validation in further.

Conclusion
In conclusion, we proved that number of stings, HB < 110 g/dl, TBI > 34 mg/dl, ALT > 40 U/L and APTT > 47 s were independence risk factors for AKI following multiple wasp stings. And the predictive formula and the individual nomogram were established and validated based on those predictive factors to predict AKI following wasp stings. Of note, those could serve as promising predictive tools to assess the probability of the AKI following wasp stings and help make a decision for therapy easily and conveniently.

Declarations
Ethics approval and consent to participate All enrolled patients approved to participant and written informed consents were obtained from all enrolled patients or their legally authorized representatives.

Consent for publication
All authors approved of publication.

Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request

Competing interests
The authors declare that they have no competing interests Funding 1·3·5 project for disciplines of excellence-Clinical Research Incubation Project, West China Hospital, Sichuan University (18HXFH018 and ZYGD18027). Those funded in the design of the study and collection, analysis, and interpretation of data and in writing the manuscript.