A fast, resource efficient, and reliable rule-based system for COVID-19 symptom identification.

Area of Science:

• Medical Informatics
• Natural Language Processing in Healthcare
• Public Health Surveillance

Background:

• Current annotation systems struggle with scalability and resource demands, hindering real-time integration with clinical decision support systems (CDS).
• The COVID-19 pandemic highlighted the need for efficient, rapidly deployable annotation tools for clinical data.
• A novel rule-based gazetteer presents a potential solution to these limitations.

Purpose of the Study:

• To evaluate the performance, resource utilization, and runtime of a rule-based gazetteer for COVID-19 symptom extraction.
• To compare the gazetteer against established annotation systems for clinical data integration.
• To assess the suitability of the gazetteer for real-time symptomatology identification and CDS integration.

Main Methods:

• A rule-based gazetteer was developed and its performance benchmarked.
• The gazetteer's efficiency was compared with five other annotation systems: BioMedICUS, cTAKES, MetaMap, CLAMP, and MedTagger.
• Key metrics included precision, recall, f1-score, resource utilization (processor and memory), and runtime.

Main Results:

• The rule-based gazetteer exhibited the fastest runtime and lowest resource footprint among the evaluated systems.
• It achieved comparable performance (weighted microaverage and macroaverage precision, recall, and f1-score) to existing annotation systems.
• The system successfully facilitated real-time COVID-19 symptomatology identification and integration into CDS.

Conclusions:

• The rule-based gazetteer overcomes critical limitations in scalability and resource utilization for clinical data annotation.
• It is well-suited for large-scale deployment in healthcare settings for acute COVID-19 symptom surveillance and prognostic modeling.
• Ongoing work focuses on performance enhancement through lexical rule fine-tuning and distributed computing, with current application in monitoring post-acute sequelae of COVID-19 (PASC).