The Institute of Medicine (IOM) developed the concept of a learning healthcare system in 2007 in response to earlier studies, including the 1999 To Err is Human report, which reported that an alarming number of patients were being harmed, rather than helped, by the U.S. healthcare system as a result of preventable medical errors. The To Err is Human report was also the impetus for U.S. Food and Drug Administration (FDA) regulations requiring standards-based auto identification carriers, e.g., barcodes, on the labels of pharmaceuticals and medical devices. The ability to capture data on the drugs and devices used in patient care is an important factor in the effort to establish a body of evidence on their real world performance, which can then be shared as part of a learning healthcare system. A 2012 IOM report, entitled Better Care at Lower Cost, highlighted the fact that there remains very little evidence for many of the care decisions made today. A learning healthcare system is designed to correct that.

While medical errors remain a serious problem, a report recently issued by the Agency for Healthcare Research and Quality found that there were 2.1 million fewer hospital-acquired conditions (HACs) between 2010 and 2014. That’s a 17 percent decline resulting in nearly $20 billion in savings. Forty percent of the reductions were in adverse drug events, followed by a 28 percent reduction in pressure ulcers and 16 percent for catheter-associated urinary tract infections (CAUTI). Less frequent but often more serious central line associated blood stream infections (CLABSI) dropped more than 70 percent.

Government officials could not point to the exact causal factors for the improvements, but they credit payment reforms that no longer reimburse hospitals for HACs and increased use of electronic health records (EHRs). Personally, I believe the former is probably the primary catalyst, but if we are to create a true learning system, we will need to document the specific steps taken by hospitals to reduce adverse events along with the results. I have high hopes that EHRs will help us achieve this but not until there is more interoperability and better clinical supply documentation. I am only surmising, but could the much larger drop in adverse drug events be related to the longer standing pharmaceutical barcode rule and meaningful use requirements related to capture of data on medications in EHRs? If so, then we should watch to see if increased adoption of UDIs supports improvements in adverse events involving medical devices. That’s where standardization of data makes a difference. Standardization of process is also critically important. Ideally, EHRs will be able to tell us how and when changes in clinical practice contribute to better results.