ADS Capstone Chronicles Revised
11
4.2.2.3 Labels Table. The labels table was extracted from the drug labels API. It consists of free text variables and date of label version (YYYYMMDD). The ingredients list of the drug has almost no missing values (0.05%), warnings have 20.1% missing, and purpose has 40.50% missing, with brand names, generic name, manufacturer name, substance name, ndc code, rxcui code, contraindications, and precautionsallhaving50%ormoremissing values. The text features were transformed withaNLPpipeline.Brandandgenericdrug names were combined into a list of unique drug name tokens. Drug warnings, contraindications, and precautions were combined intoalistofwarningtokens.The following functions were applied to this table: -clean_data -contains_unique_value -nan_info -process_label_text 4.2.2.4 Patient Drugs Table. The patient drugs table was a nested variable derived from the events table and linked by event_id.Thedatawasexpandedasitsown table. The variables include descriptionson drug names, manufacturers, indications, rxcui, and unique ingredient identifiers, which all have 27-30% of total records missing. Drug characterization and medicinal product fields have no missing data. Functions were applied to clean and tokenize drug manufacturer name data.The following functions were applied to this table: -add_sequential_index -nan_info -process_label_text -add_sequential_index -remove_duplicates_nan -plot_character_lengths
reactions dataframe (Figure 5a,b). The following functions were used on this table: -find_matching_drugs -check_reactions -get_matching_reactions -add_sequential_index Figure 5 Document Drugs (a) and ADRs (b) A
B
4.2.2.2 Events Table. The events table was extracted from the drug events API. It consists of adverse event report dates (YYYYMMDD),descriptivevariablesabout the person involved (age, sex) with two nested dictionaries for drug reactions and drugs involved, the reporting authority’s qualification (expedited, occupation), and the patient primary ADR outcome (nonserious, serious, death). The following functions were applied to this table: -convert_to_years -add_sequential_index -nan_info
-examine_text_outliers -plot_character_length
-plot_character_length -examine_text_outliers -clean_manufacturer_text
161
Made with FlippingBook - Online Brochure Maker