The emergence and re-emergence of infectious diseases poses significant challenges to global health, necessitating robust strategies for understanding and controlling their spread. Epidemiological models serve as critical tools in this endeavor, providing insights into the dynamics of disease transmission, informing public health interventions, and guiding policy decisions. These models range from simple deterministic frameworks to complex stochastic simulations, each offering unique advantages and addressing different aspects of disease dynamics. The 21^st century has witnessed several notable outbreaks, including SARS, H1N1 influenza, Ebola, Zika, and most recently, COVID-19. Each of these events has underscored the importance of timely and accurate epidemiological modeling. By simulating the spread of pathogens, epidemiological models help to predict outbreak trajectories, to estimate the impact of interventions, and identify potential hotspots for targeted responses. Furthermore, they enable the assessment of various control strategies, such as vaccination, social distancing, and quarantine measures, thus aiding in the optimization of resource allocation and mitigation efforts. Blockchain technology outline a proper process to share information more conveniently without involving any third party. Blockchain is operating as a node or device that contains stored and encoded data permanently and acts as a decentralized database. It syncs latest blockchain data to fix all nodes constantly through a peer-to-peer network. Not only that, Sensitive information of the patients (PII) cannot be disclosed directly with other organizations or individuals as the demand and necessity of healthcare, but patients record always existed for Research and Development of new treatment of diseases.

In this paper, we present an advanced epidemiological model designed to enhance our understanding of infectious disease dynamics. Building upon the foundational Susceptible-Infectious-Recovered (SIR) framework, our model incorporates several critical extensions and also considered certain socio-economic elements. These additions allow for a more nuanced and realistic representation of disease spread, capturing the complexities observed in real-world scenarios. And to resolve the challenge of disclosing a patient data we anticipate and build a prototype system that share unidentified patient data through a decentralized platform. The platform increases the value of data through prevention of accessing any unnecessary information and filter and retain data before displaying to the users.

In the following sections, we first thoroughly study different state-of-the-art model and later deep drive into the technical aspects of the model. Later we show the analytical results of derived from the model. We also share the comparative analysis of the state of the art models and also describe how blockchain technology maintain decentralization and security aspect of the patient records. Our job concludes with a summary, a near-term forecast, and recommendations for further work.

The balance of various factors within society significantly impacts public health, which includes the infrastructure dealing with communicable diseases. Over the past two decades, researchers have concentrated on developing mathematical models for communicable diseases to analyze their intensity and spread 1, 2, 3, 4, 5. For instance, numerous studies have demonstrated that communicable diseases like cancer pose a substantial threat to public health 6, 7. Researchers have explored various genetic, environmental, behavioral, and socio-economic factors contributing to the cause of cancer, uncovering new areas for early diagnosis and suitable treatments 8. Such studies also enable the analysis of risk patterns in large populations. Epidemiological models, such as the Susceptible-Infectious- Recovered (SIR) model, have been pivotal in understanding the effects of demographic and socio-economic factors on the spread of communicable diseases. These models highlight mitigation strategies that can flatten the curve of disease spread 9, 10.

SIR model utilizes three tightly coupled differential equations to represent disease dynamics. Between 1927 and 1933, Kermack and McKendrick further advanced this model, which has since been instrumental in understanding infectious disease spread and evaluating control measures’ potential impact on reducing disease intensity and mortality rates 2, 11, 12, 13. The COVID-19 pandemic has spurred numerous studies worldwide, including in India, where researchers have applied machine learning algorithms, neural networks, deep learning, and other time series forecasting algorithms to predict the disease’s growth and analyze its spread 9, 14, 15. Similarly, other communicable diseases, such as dengue, have also been studied using these advanced techniques 10. The healthcare system increasingly relies on data-driven and information management systems to guide professionals in making informed decisions and improving services 16.

17 Proposed a method for preserving data security and privacy at every stage of data transaction.18 introduced a data-sharing platform based on consortium blockchain, ensuring decentralization and patient privacy through anonymization techniques for Personally Identifiable Information (PII). This approach facilitates knowledge sharing among stakeholders without compromising confidential information. Blockchain-based medical decision support systems have also been developed to ensure reliable and secure medical information management, focusing on areas like tissue donation, transplantation, and infertility treatment 17, 19. Named Entity Recognition (NER) using machine learning was introduced in 2015 with models like Hidden Markov Models and Conditional Random Field Models 20. Deep learning advancements have since enhanced these models’ efficiency. In 2021, anonymization of sensitive data was performed using Natural Language Processing (NLP) 21. NLP processes unstructured data to identify and anonymize sensitive information. Other studies have used Large Language Models (LLMs) like BERT for NER on medical records, exploring anonymization possibilities 22. A comparative study by23 demonstrated that transformer-based models, particularly BERT, outperform non-transformer-based models in NER tasks 23.

Figure 9 helps to identify the spikes generated at different time intervals during the COVID-19 pandemic.

Table 1 represents a comparison of the overall accuracy level between the conventional SIR model and thefeedback-based SIR model.

In Figure 10 we show how feedback based SIR model outperforms the traditional SIR model when we plot the data derived from the each of the model along with the actual time series data. We found that the curve generated by the feedback based SIR model is much nearer to the actual data curve compared to the curve generated by traditional SIR model.

Table 2 shows the how the proposed ML model dominate the state of the art time series machine learning model.

Not only that, the proposed model is highly computationally efficient (Table 3) compared to that models.

The anonymization technique, i.e., fine-tuned BERT model, that we follow in our overall Decision Support System ensures a high accuracy level. The comparison with the over techniques enlisted in Table 4.

The Figure 11 shows the overall correlation among the socio-economic and socio-demographic factors. Also, we found the correlation among the various variants that we experience during the pandemic time frame.

From a survey conducted in sample population we find the following feedback for the DSS model.

Our proposed model not only give you an overall picture of current pandemic scenario related to communicable disease 27, 28, 29, it also provides a comprehensive view for both communicable and non-communicable disease, Our model also suggest about the influence of multiple socio-economic and socio-demographic factors not sticked to common factors like vaccination of population 28. The proposed ML framework helps in predicting future pandemic taking into consideration of those different socio-economic and socio-demographic factors by maintaining an accuracy level of 93% compared to 85% in 28. Like other study [30], we have also used a traditional epidemiological model. But we further modify that by adding feedback to that which approximately doubled the efficiency level compared to the traditional one. As we may face multiple stakeholders, we decentralized the overall data storage and retrieval phase using Blockchain which is still now only used in sharing specific Electronic Health Records 30. On top of that as we are dealing with highly sensitive data, we used an additional anonymization technique for protecting those PII data.

To combat communicable and non-communicable diseases like COVID-19 and cancer, we need to be more organized and effective in both medical and technological perspectives. The threat posed by epidemiological circumstances to public health can be mitigated with proper analysis and prediction. Data analysis and prediction will aid authorities and public health personnel in situation evaluation and decision-making. The SIR model is one of the most broadly used epidemiological models in public health. We have aimed to address certain limitations of the conventional SIR model and refine it to increase its efficiency. By comparing real data with both the conventional and our modified SIR model, we demonstrate that the modified model has superior predictive efficiency. Our innovation extends to evaluating the influence of various impulsive factors associated with public health in real-time and how those factors are correlated with each other. The ML model that we used is also computationally more efficient com- pared to state-of-the-art approach. For anonymization and security of sensitive data like health records, we have used Blockchain Technology. This also helps in sharing reports in a secured and decentralized way.

We have accounted for virus reinfection or reaction to improve the accuracy of forecasting and prediction. Our data and other calculation factors are based on a specific locality. In future, our planning is to expand our model to more global aspects with additional decisive parameters. We will also enhance the anonymization efforts for unstructured data, such as image data and textual documents. The data search and retrieval techniques used in the blockchain can be fine-tuned to increase the model’s efficiency.