Detecting Simulated Nosocomial Disease Outbreaks with Sequential Monte Carlo Methods
Introduction: Nosocomial diseases, or healthcare-associated infections, pose significant challenges to patient safety and public health. In this study, we propose a novel approach that extends the state-of-the-art for detecting nosocomial disease outbreaks using Sequential Monte Carlo methods, also...
Saved in:
| Main Author: | |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-03-01
|
| Series: | International Journal of Infectious Diseases |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S1201971224006192 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Introduction: Nosocomial diseases, or healthcare-associated infections, pose significant challenges to patient safety and public health. In this study, we propose a novel approach that extends the state-of-the-art for detecting nosocomial disease outbreaks using Sequential Monte Carlo methods, also known as particle filters (PFs). PFs have emerged as powerful tools for disease surveillance and have been widely used to estimate the state of a disease based on noisy observations. Observations could coincide with results of diagnostic tests such as PCR tests, antigen tests or serological tests for the disease. Early detection of outbreaks within healthcare facilities is crucial for effective control and prevention measures. Methods: Following the methodology of the susceptible, infected, recovered (SIR) disease transmission model, an agent-based model is utilised to simulate interactions between healthcare workers and patients. These individuals are categorised into susceptible or infected compartments. By fixing the rates of infection and recovery, determined by transmission and recovery parameters, respectively, we simulate a time-series of the number of infected individuals spanning 150 days. To introduce outbreaks, we double the transmission parameter during random intervals, e.g. days 50-65 and 105-110. This would correspond to a 15 and a 5 day outbreak, respectively. Increasing the transition parameter accelerates the spread of the disease within the population, leading to more susceptible patients coming into contact with infected patients. The non-outbreak time-period is defined as Model 1 and the outbreak period as Model 2.Traditionally, as data becomes available at each increment of time t, PFs provide an estimate of the probability distribution over the states of a disease model via a set of weighted particles (samples). These states coincide with the unobservable S and I compartments of the transmission model. We extend the PF to make daily inferences on whether Model 1 or Model 2 was used in the data generation process. Results: We demonstrate the effectiveness of our approach through simulations that use multiple runs of synthetic data with random outbreaks included within the time-series of infected individuals. Using ROC curves and the AUC metric, we assess the performance of the proposed method of disease outbreak detection. We show that algorithm can detect true outbreaks with minimal false alarms. Discussion: The proposed algorithm successfully detects outbreaks in a timely manner. A drawback of the scenarios considered is that the algorithm runs on synthetic data. Future work will involve applying the algorithms to real scenarios, where outbreaks have occurred in the hospital setting. Conclusion: By detecting nosocomial disease outbreaks early, healthcare facilities can implement timely interventions to mitigate the spread of infections, protect vulnerable patients and optimize resource allocation. Our methodology offers a valuable tool for improving infection control practices and enhancing patient safety in healthcare settings. |
|---|---|
| ISSN: | 1201-9712 |