Table 6.

Summary of articles related to Case Identification

Article	Search method	Reaction analyzed	Sample size	Summary of findings
Unstructured data (n=3)
Epstein et al. (2013) 136	RxNorm and natural language processing (NLP)	Adverse drug events	N/A	A high performing algorithm was used to identify medication allergies with a specificity of 90.3% and 85% in the training and testing data respectively. Accuracy, precision, recall, and F-measure for medication allergy matches were all above 98% in the training dataset and above 97% in the testing dataset for all allergy entries
Wolfson et al. (2019) 134	Free text keyword search of allergy module	DRESS syndrome*	69	Of 538 hypersensitivity reactions identified, 69 patients (2.18 in 100,000 patients) had DRESS syndrome.
DeLozier et al. (2021) 135	Text processing system	SJS/TEN* and torsades de pointes	138	The automated recruitment system resulted in the capture of 138 true cases of drug induced rare events, improving recall from 43% to 93%
Structured data (n=2)
Davis et al. (2015) 137	ICD-9 codes	SJS/TEN	475-875	Patients with the ICD-9 codes introduced after 2008 were more likely to be confirmed as cases (OR 3.32; 95%CI 0.82, 13.47) than those identified in earlier years. Likelihood of case status increased with length of hospitalization. Applying the probability of case status to the 56 591 potential cases, we estimated 475-875 to be valid SJS/TEN cases.
Saff et al. (2019) 138	ICD-9 codes and E codes	Allergic drug reactions	409	Specific ICD-9 codes can identify patients with allergic drug reactions, with antibiotics accounting for almost half of true reactions. Most patients with codes 693.0, 995.1, 708, and 995.0 had allergic drug reactions, with 693.0 as the highest yield code. An aggregate of multiple specific codes consistently identifies a cohort of patients with confirmed allergic drug reactions.
Combination of structured and unstructured data (n=4)
Kim et al. (2012) 142	Procedure codes and International classification of nursing practice terms	contrast-media-induced hypersensitivity reactions	266	An EHR-based electronic search method was highly efficient and reduced the charts that needed to be reviewed by 96% (28/759)
Cahill et al. (2017) 140	ICD-9 codes and an informatics algorithm	Aspirin-exacerbated respiratory disease (AERD)	593	An informatics algorithm can successfully identify both known and previously undiagnosed cases of AERD with a high positive predictive value. Involvement of an allergist/immunologist significantly increases the likelihood of an AERD diagnosis.
Fukasawa et al. (2019) 141	ICD-10 codes and informatics algorithms using clinical course and medical encounters	SJS/TEN	N/A	One algorithm, consisting of a combination of clinical course for SJS/TEN, medical encounters for mucocutaneous lesions from SJS/TEN, and items to exclude paraneoplastic pemphigus, but not ICD-10 codes, showed a sensitivity of 76.9%, specificity of 99.0%, positive predictive value of 40.5%, negative predictive value of 99.8%, and diagnostic odd ratio of 330.00.
Banerji et al. (2020) 139	ICD-9 codes and NLP	Allergic drug reactions	335	Among the 335 confirmed positive cases, NLP identified 259 true cases, resulting in a recall/sensitivity of 77% (range: 26%-100%). Among the 390 negative cases, NLP achieved a specificity of 89% (range: 69%-100%).

Abbreviations: DRESS, Drug Reaction with Eosinophilia and Systemic Symptoms; SJS, Stevens Johnson Syndrome; TEN, Toxic Epidermal Necrolysis; AERD, Aspirin Exacerbated Respiratory Disease; NLP, Natural Language Processing