| Autor: |
A. Myall, I. Venkatachalam, C. Philip, M. Yin, D. Koon, S. Arora, Y. Yue, R. Peach, A. Weiße, P. Tambyah, A. Chow, J. Price, A. Cook, A. Holmes, M. Barahona |
| Jazyk: |
angličtina |
| Rok vydání: |
2023 |
| Předmět: |
|
| Zdroj: |
International Journal of Infectious Diseases, Vol 130, Iss , Pp S31- (2023) |
| Druh dokumentu: |
article |
| ISSN: |
1201-9712 |
| DOI: |
10.1016/j.ijid.2023.04.071 |
| Popis: |
Intro: Understanding infectious disease transmission dynamics is key to optimising infection prevention and control (IPC) strategies. Yet, standard epidemiological modelling tools do not fully utilise large datasets of time- evolving spatial interactions, which could lead to improved descriptions of complex transmission dynamics, as exhibited by multi-drug resistant organisms (MDRO). Here, we showcase an approach accounting for spatial-temporal dynamics using machine learning and graph theory to characterize nosocomial MDRO transmission. Methods: We performed a retrospective cohort study using all inpatient admissions to a 1200-bed Singaporean hospital from 2018-01-01 to 2021-12-31. We employ a graph-based machine learning methodology to learn transmission dynamics using spatial-temporal patient interactions. Taking Methicillin-resistant Staphylococcus aureus (MRSA) as a proof-of-principle, we examine the real-world efficacy using routine surveillance and clinical cultures obtained during acute care. Findings: There were 149,352 inpatient admissions and 3.21 million contact-interactions during the study period; 7,753 (5.2%) of these patients tested positive for MRSA, with 1,585 (1.1%;) being likely hospital acquisitions. The probability of onward transmission was largest on the day a culture-positive result was reported (11.2% of all transmissions). However, high onward-transmission persisted beyond (weighted-mean 7.8 days). Through spatial-temporal analsysis, we found hospital rooms with comparatively high acquisition-rates (36/665 wards accounting for 50.2% of hospital-acquisitions); these hotspots were found to be central in network of patient transfers, and interconnected with each other. Conclusion: With extensive and detailed datasets becoming available, new methodologies that leverage their size and detail can provide insights into transmission processes. Here, we demonstrated their use to learn the transmission dynamics of an MDRO and showed their ability to make inferences on spatial-temporal profiles of transmission. Furthering this research, we will engineer spatial-temporal features to predict pathogen transmission dynamics and expand to additional MDROs. |
| Databáze: |
Directory of Open Access Journals |
| Externí odkaz: |
|
