The algorithm uses Structured Query Language to identify AAA cases, controls, and excludes from the Electronic Medical Record. AAA cases were defined as meeting at least one of three criteria: had a AAA repair procedure (Case Type 1), had at least one vascular clinic encounter with a diagnosis of ruptured AAA (Case Type 2), or had at least two vascular clinic encounters with a diagnosis of unruptured AAA (Case Type 3).
ACE-I induced cough is a common side effect of use of ACE inhibitors, one of the most common class of antihypertensives. The frequency of ACEI-induced cough varies based on ancestry. A GWAS of ACEI cough using this algorithm in the eMERGE Network identified KCNIP4 as associated with this phenotype, which was validated in two replication cohorts.
Cases are those with ACEI cough. Controls are those exposed to ACEI without adverse events noted and not switched to angiotensin receptor blockers (ARBs).
Algorithm validated - December 12, 2012.
Appendicitis is one of the most common acquired surgical conditions of childhood. Diagnosis of appendicitis remains difficult. Much work has been done on validation of clinical scores to reduce the number of unnecessary surgeries and radiographic tests while maintaining a high sensitivity for disease. However, no score performs well enough in practice to reduce these risks (Kulik et al., 2013). It is also known that appendicitis has a familial predominance, but little is known about the genetic factors that may increase a certain child's risk for the condition (Oldmeadow et al., 2009).
Atrial Fibrillation phenotype algorithm for the DNA Demonstration project. The algorithm selects cases based on atrial fibrillation but no presence of a heart transplant. Controls select for records with no evidence of atrial flutter, atrial fibrillation, or atrial tachycardia but with at least one ECG.
Algorithm to select subjects with "normal" electrocardiograms. Subjects do not have heart disease, interfering medications, or abnormal electrolytes at the time of the normal ECG. Individuals may, however, develop abnormalities later in life.
Hypothetical timeline for a single patient:
We used a multi-modal strategy consisting of structured database querying, natural language processing on free-text documents, and optical character recognition on scanned clinical images to identify cataract subjects and related cataract attributes.
Note: Attached documents contain full case definition and two different control definitions. One is for controls with 2 years of follow up, the other for controls with 1 year of follow up. All available controls with 2 years of follow up were used in Vanderbilt's study. The control population was supplemented by controls with only 1 year of follow up. At the time of study, many of the available controls had experienced their qualifying events somewhat recently and 2 years had not yet passed for full follow up.
Algorithm to identify patients with diabetic retinopathy.
An algorithm for finding patients with diverticulosis, and of those, patients who also have diverticulitis, and to also find control patients. Control patients will have had a colonoscopy but have no evidence of diverticula.
Simple NLP (a portable program is posted here, with instructions, and support is availabe from NU as needed) of colonoscopy reports is the gold standard algorithm, but if the text of colonoscopy reports is not available, an alternate algorithm using CPT & ICD-9 codes can be used, which is also posted.