Table 3.

Natural language processing (NLP) algorithms applied in the reviewed studies.

Study	NLP algorithm type	Details of NLP algorithms
Ganz et al, 2014 [37]; Nunes et al, 2016 [31]; Nunes et al, 2017 [29]; Loughlin et al, 2018 [32]; Pettus et al, 2019 [33]	Rule-based	Identify terms consistent with hypoglycemia (including alternative or incorrect spellings and abbreviations) Identify descriptive attributes of the hypoglycemia mention (eg, seriousness, duration, and frequency) Identify sentiment of the mention (eg, denial and affirmation, including “has,” “diagnosed,” and “present”) Identify contextual information (eg, note section headers and neighboring text). Sections such as “history of present illness,” “assessment,” “hospital course,” “reason,” “review of symptoms,” and “chief complaint” generally reflected occurrence of hypoglycemia
Li et al, 2019 [34]	Rule-based	A formally defined pattern (regular expression), which identified clinical reports mentioning a “blood sugar word” followed within 5 words by what could be a low blood sugar value represented by a number ranging from 10 to 69
Misra-Hebert et al, 2020 [35]	Rule-based	Split clinical notes into sentences and phrases Filter sentences and phrases to those containing a hypoglycemia-related Unified Medical Language System concept Identify temporal phrases (when the event occurred) Classify polarity (assertion or negation) into no, nonsevere, and severe event
Uzoigwe et al, 2020 [36]	Rule-based	Identify keywords or concepts of interest: symptom-based and nonsymptom-based hypoglycemic events Symptom-based terms: neuroglycopenic and adrenergic symptomology associated with hypoglycemia. Adrenergic symptomology: elevated or irregular heart rate, sweating, tremor, trembling, tingling, or shaking, and vision impairment Neuroglycopenic symptomology: cognitive issues, irritable or anxious, mood or behavior change+NOT substance abuse or alcohol, slurred speech+NOT stroke+NOT substance abuse or alcohol Nonsymptom-based definition: Mention of “hypoglycemia” Relevant medical ontology such as “low glucose” A blood glucose laboratory value ≤70 mg/dL documented