Development and performance characteristics of novel code-based algorithms to identify invasive Escherichia coli...

Area of Science:

• Medical Informatics
• Infectious Disease Epidemiology
• Health Data Science

Background:

• Invasive Escherichia coli disease (IED) poses a significant public health challenge.
• Accurate identification of IED in healthcare databases is crucial for epidemiological surveillance and resource allocation.
• Existing methods for identifying IED in electronic health records may lack precision.

Purpose of the Study:

• To develop and evaluate novel code-based algorithms for identifying patients with invasive Escherichia coli disease (IED) within healthcare databases.
• To compare the performance characteristics of different algorithmic approaches for IED detection.

Main Methods:

• Six algorithms were designed using diagnosis and drug exposure codes from inpatient records (2016-2020).
• Algorithms were assessed against a reference standard derived from microbiology data.
• Performance metrics included positive predictive value (PPV) and sensitivity.

Main Results:

• Among 97,194 records, 26.0% were classified as IED.
• Algorithm 1 showed high PPV (96.0%) but low sensitivity (60.4%).
• Algorithm 4 achieved a balanced performance (PPV: 93.6%, sensitivity: 78.1%, F1 score: 85.1%), outperforming others.
• Antibiotic exposure data had minimal impact on algorithm performance.

Conclusions:

• Algorithm 4 provides a robust and balanced approach for identifying IED in large healthcare datasets.
• This algorithm can serve as a valuable tool for estimating the true burden of IED.
• Improved case ascertainment facilitates better understanding and management of IED.