By Zina Moukheiber, Contributor A study published today in Nature Clinical Pharmacology and Therapeutics showcases the potential power of sophisticated data analytics when applied to electronic health records on a large scale. By predicting harmful drug side effects two years before an alert from the Food and Drug Administration, the technology can complement the FDA’s Adverse Event Reporting System, which collects information from doctors, pharmacists, and patients once a drug is on the market. Less than three years ago, bioinformaticians at Stanford University set out to capture information buried, not just in insurance claims, lab tests or medical codes, but in the rich narrative of 10 million clinical notes gleaned from 1.8 million patients on Epic Systems electronic health records. “Ninety percent of patient data comes from clinical notes,” says senior author Nigam Shah, who teaches biomedical informatics at Stanford. Along a binary matrix, Shah and his colleagues plotted de-identified patients, diagnoses, drugs and outcomes, among other variables. Patterns emerged detecting which patients, for example, with rheumatoid arthritis took Vioxx, and subsequently suffered a heart attack, as well different variations of the same variables. Drug dosages were not taken into account. Shah and his Stanford colleagues tested their model on nine drugs that were slapped with strong warnings after deadly side effects emerged, including Vioxx and Avandia. Their algorithms were able to detect adverse effects in six of them, two years before the FDA’s alert. (Stanford had 15 years of patient data). Shah says they didn’t have enough patients to get statistically significant information on three drugs, but that will change as more patient data gets stored in electronic health records.
Source: FULL ARTICLE at Forbes Latest
