Utilizing Text Mining on Electronic Health Records for COVID-19 Outbreak Surveillance – Physician’s Weekly

The following is a summary of COVID-19 outbreaks surveillance through text mining applied to electronic health records, published in the March 2024 issue of Infectious Disease by Rocha et al.

The COVID-19 pandemic spurred a surge in tech solutions, but more technologies are needed to be helpful in low-resource settings for disease monitoring.

Researchers conducted a retrospective study to address this gap by developing a data science model that uses routinely generated healthcare encounter records to detect potential new outbreaks in real-time.

They developed an epidemiological indicator that served as a proxy for suspected COVID-19 cases, utilizing health records from Emergency Care Units (ECUs) and employing text mining methods. The dataset consisted of 2,760,862 medical records from nine ECUs, each containing patient age, reported symptoms, and admission timestamps. A dataset of 1,026,804 officially confirmed COVID-19 cases was utilized, covering records from January 2020 to May 2022. Models were assessed using sample cross-correlation between two finite stochastic time series.

The results showed that for patients aged 18 years, the time lag () was 72 days with a cross-correlation () of approximately ~0.82 for the first wave, 25 days with a cross-correlation () of around ~0.93 for the second wave, and 17 days with a cross-correlation () of about ~0.88 for the third wave.

Investigators concluded that the model effectively detects signs of potential COVID-19 outbreaks weeks ahead of traditional methods, allowing for earlier public health interventions.

Source: bmcinfectdis.biomedcentral.com/articles/10.1186/s12879-024-09250-y

Read the rest here:

Utilizing Text Mining on Electronic Health Records for COVID-19 Outbreak Surveillance - Physician's Weekly