Mapping Clinical Documents to the Logical Observation Identifiers, Names and Codes (LOINC) Document Ontology using Electronic Health Record Systems Structured Metadata

Abstract

As Electronic Health Record (EHR) systems increase in usage, organizations struggle to maintain and categorize clinical documentation so it can be used for clinical care and research. While prior research has often employed natural language processing techniques to categorize free text documents, there are shortcomings relative to computational scalability and the lack of key metadata within notes' text. This study presents a framework that can allow institutions to map their notes to the LOINC document ontology using a Bag of Words approach. After preliminary manual value- set mapping, an automated pipeline that leverages key dimensions of metadata from structured EHR fields aligns the notes with the dimensions of the document ontology. This framework resulted in 73.4% coverage of EHR documents, while also mapping 132 million notes in less than 2 hours; an order of magnitude more efficient than NLP based methods.

Figure 1:

The figure shows the process of going from a clinical document to LOINC Code. Each of the pink blocks are a metadata fields for each individual clinical document, mapping to a green block representing the LOINC dimensions. The distinct combination of the LOINC dimension points to a LOINC Code, which will connect back to the clinical document.

Figure 2:

Shows a example note’s metadata, its possible BOW as well as a few LOINC BOW’s. In this example, the LOINC Code that will be selected will be 3A due to it being a complete match to the note’s BOW. Note that while 3B is also a full match, 3A is given priority due to covering more LOINC Dimensions.

Figure 3:

The figure shows a few different LOINC Codes, each with their corresponding part numbers and part names defining their dimensions. The PartTypeName column specifies the exact dimension name for the given row.

Figure 4:

This figure visualizes the process of reducing the computational challenge of the pipeline. We start out with over 130 Million individual notes and get unique combinations of columns and remove irrelevant terms to decrease the computational complexity.

Figure 5:

This figure shows two graphs: The left graph shows the logarithmic growth of the count of Unique BOW with respect to the actual document count and the right graph shows the linear runtime of the algorithm based on the count of Unique BOW.