This section describes the results of this study. Experiments were conducted on a 3 GHz i5 computer with 4 GB of main memory and a 64-bit Windows 7 operating system. Experiments are conducted using the programming language Python.
Crossover of characteristics among the three vaccines for the target class of patients
From a pharmacovigilance perspective for the COVID-19 vaccines studied, the aim of this study was to find out what the most common side effects were common to each patient category.
Figures 6, 7, and 8 show the crossover of features among the three vaccines for the ‘Dead’, ‘Hospitalized’, and ‘Recovered’ patient classes, respectively. From the figure, we can see that some features are common among her three vaccines, while others are unique to each vaccine.
The commonality of the three vaccines for the target class of “death” patients is characterized.
It features a crossover between three vaccines for the “hospitalized” patient class.
It features a crossover between the three vaccines against a target class of ‘recovered’ patients.
In the “death” class, we find that the three vaccines share seven features (age, erythema, allergy, cardiac arrest, cerebrovascular disease, unconsciousness, and visitation), but for example, chest discomfort. The feature is only Pfizer vaccine.
Relationship between adverse reactions in each category after vaccination with the new coronavirus vaccine
Figures 9, 10, and 11 show the number of side effects based on organ effects in patients who died, were hospitalized, and recovered after three vaccinations of each class. For the PFIEZER vaccine, CNS-related side effects are more common than blood-related side effects across all patient categories. The third most common side effects in deaths, hospitalizations, and recovering patients are side effects of CVS, gastroenteritis, and allergies, respectively. Figure 9. For the JANSSEN vaccine, CNS-related side effects are the most common across all patient categories. Next, CVS-related side effects for the dead and recovered categories, Figure 10.
There are no side effects based on organ effects in patients who died, were hospitalized, or recovered after PFIEZER vaccination.
There are no side effects based on organ effects in patients who died, were hospitalized, or recovered after receiving the JANSSEN vaccine.
There are no side effects based on organ effects in patients who died, were hospitalized, or recovered after MODERNA vaccination.
For MODERNA vaccines, CNS-related side effects are more common than ANS-related side effects in death and recovery across all patient categories. If hospitalized, bleeding is her second most common side effect. Her third side effect, common among deceased and recovered patients, is related to cardiovascular side effects (Fig. 11).
Classify Patients Using a Deep Learning Classifier
Among the various types of deep learning models, recurrent neural networks (RNN) and long short-term memory (LSTM) networks have received a great deal of attention for their ability to process data. In this section, we compare the RNN model and his LSTM model for patient classification using a deep learning classifier. We explore the strengths and limitations of each model, focusing on factors and various performance metrics (accuracy, recall, specificity, precision, F1_score, computation time). Tables 6, 7, and 8 show a comparison of RNN and LSTM models. We categorize the patient’s adverse reactions after the three vaccinations into three target classes: death, hospitalization, and recovery.
The above table and results show that the RNN model outperforms the LSTM in several indices such as accuracy, recall, specificity, precision, and F1 score, demonstrating superior performance in the patient classification task. increase. On the other hand, RNN models are inferior to LSTM in terms of computation time, and this limitation will be considered as a research point for future research.
Classify Patients Using RNN Classifier
The RNN classifier was tested using different values for batch size and epoch. Figures 12, 13, and 14 show a comparison of the adverse reaction performance of the three post-vaccination patients in terms of accuracy, recall, specificity, precision, and three target classes: mortality, hospitalization, and hospitalization. ” in terms of F1 scores. recovered. We find that the performance of the proposed model is excellent in most cases when we use 50 epochs and the batch size is equal to 50. Therefore, Table 6 shows the optimal parameters used for the RNN classifier.
Performance comparison of patient adverse reactions after PFIEZER vaccine using different number of epochs and different patch sizes. where E represents the number of epochs and B represents the batch size. (be) Classification of deceased patients. (b) Classification of inpatients. (c) classification of recovered patients.
Performance comparison of adverse reactions in patients after JANSSEN vaccine using different number of epochs and different patch sizes. E represents the number of epochs and B represents the batch size. (be) Classification of deceased patients. (b) Classification of inpatients. (c) classification of recovered patients.
Performance comparison of patient side effects after MODERNA using different epoch numbers and different patch sizes. where E represents the number of epochs and B represents the batch size. (be) Classification of deceased patients. (b) Classification of inpatients. (c) classification of recovered patients.
Using the RNN classifier parameters presented in Table 9, the proposed model exhibited the highest accuracy, recall, F1_score, specificity, Giving an accuracy score, the accuracy value is observed to be 96.03%. During JANSSEN vaccination, Figure 16 shows that the ‘hospitalized’ subject class performed best, with a read accuracy of 94.7%. Finally, as shown in Figure 17, this model performed best in the ‘recovery’ class of MODERNA vaccination with an accuracy of 97.794%.
Comparison of patient performance regarding side effects after each class of PFIEZER vaccination.
Performance comparison of adverse reactions in patients after each class of JANSSEN vaccination.
Performance comparison of side reactions in patients after each class of MODERNA vaccination.
A comparison of the loss of the validation and training datasets in patient adverse reactions after PFIEZER, JANSSEN, and MODERNA vaccination for each class is shown in Figures 1 and 2. They are 18, 19 and 20 respectively. A similar accuracy comparison of the validation and training datasets of patient adverse reactions following PFIEZER, JANSSEN, and MODERNA vaccination for each class is shown in Figures 1 and 2. They are 21, 22 and 23 respectively.
Comparison of loss in validation and training datasets in patient adverse reactions following PFIEZER vaccination. (be) Model loss for deceased patients. (b) model loss for inpatients. (c) model loss for recovered patients.
Comparison of losses in validation and training datasets in patient adverse reactions after JANSSEN vaccination. (be) Model loss for deceased patients. (b) model loss for inpatients. (c) model loss for recovered patients.
Comparison of loss in validation and training datasets for patient side reactions after MODERNA vaccination. (be) Model loss for deceased patients. (b) model loss for inpatients. (c) model loss for recovered patients.
Validation of side effects in patients after PFIEZER vaccination and accuracy comparison of training data sets. (be) Accuracy for deceased patients. (b) model accuracy for inpatients. (c) Accuracy of models in recovered patients.
Validation of adverse reactions in patients after JANSSEN vaccination and accuracy comparison of training data sets. (be) Accuracy of deceased patients. (b) model accuracy for inpatients. (c) Accuracy of models in recovered patients.
Validation of side effects in patients after MODERNA vaccination and accuracy comparison of training data sets. (be) Accuracy of deceased patients. (b) model accuracy for inpatients. (c) Accuracy of models in recovered patients.
Last but not least, in the development of vaccines with limited initial supply, such methods may help identify high-risk patients in primary vaccination campaigns. Educating the public about vaccine safety is critical to public health and current and future large-scale vaccination campaigns. The results obtained will inform pharmacovigilance and drug safety approaches to select the most suitable vaccine based on patient history.
statistical analysis
A Wilcoxon signed rank test was performed to determine if there was a statistically significant difference between the proposed model using RNN and the LSTM model. The Wilcoxon signed rank test is a nonparametric statistical test used to compare two related samples. It is commonly used to determine if there is a significant difference between two of her methods or treatments applied to the same group of subjects. This test is especially useful when the normality assumption is not met or when the sample size is small. The null hypothesis for the test is that there is no difference between the population medians of the two samples, and the alternative hypothesis is that the medians are not equal. If the p-value is less than the significance level (usually 0.05), reject the null hypothesis and conclude that there is a significant difference between the two samples.31. Descriptive statistics of the two models showed that the average accuracy of RNN was higher than his LSTM model. The LSTM model had an average accuracy of 86.197778 (SD = 6.2431678) and the RNN model had an average accuracy of 91.941778 (SD = 4.9613537). Shown in Table 10.
A Wilcoxon signed-rank test showed that the average rank of the RNN model (average rank = 3.00) was higher than the LSTM model, indicating a significant difference between the two models. Furthermore, there was a large difference in accuracy between the LSTM and RNN models (Z = −2.312, p = 0.021 two-sided). A negative Z value indicates that Accuracy_RNN is statistically significantly lower than Accuracy_Lstm. A p-value of 0.021 indicates that there is a 2.1% chance of observing such a large difference between the two models by chance, and that this difference is statistically significant at the 0.05 level. Table 11: Summarizes the results of the Wilcoxon signed-rank test for RNN and LSTM models.
