IN 2009, THE US CONGRESS PASSED THE HEALTH INFORmation Technology for Economic and Clinical Health (HITECH) Act, which offers nearly $30 billion in financial incentives to physicians and hospitals that adopt and choose to meaningfully use electronic health records (EHRs). The act is meant to help a health care system that consumes $2.5 trillion each year and produces health care that is below the standards of safety, quality, and efficiency that should be expected in the United States. There is broad consensus among US policy makers that EHRs will play a key role in transforming health care into a safer, more effective, and more efficient system. Despite the promise of EHRs (often referred to as electronic medical records or EMRs), recent data on their benefits have been disappointing. Although studies have consistently shown that EHRs can help clinicians adhere to guideline-based care and reduce medication errors, beyond these narrow benefits, there is little evidence that EHRs improve patient outcomes and even less evidence that they improve the efficiency of care. The lackluster data on the benefits of EHRs have led to a marketplace where EHR adoption has been underwhelming: based on the latest estimates, only a third of ambulatory care physicians and an even smaller minority of US hospitals are using EHRs (broadly defined as electronic systems that incorporate electronic prescribing, clinical notes, results management, and basic clinical decision support). Because of the slow adoption of EHRs, the US Congress included incentives in HITECH. In this sea of disappointing data about EHRs comes some good news. In an innovative study published in this week’s JAMA, Murff and colleagues push beyond the traditional uses of the EHR by demonstrating that natural language processing, when applied to electronic data, can help clinicians track adverse events after surgery. To many readers, the topic may appear esoteric, but its significance should not be underestimated. Instead, these findings suggest that EHRs can transform health care delivery. Until now, much of the benefits from EHRs have appeared to come from decision support capabilities, such as offering advice on avoiding 2 drugs with serious drug-drug interactions. Decision support is essentially a set of rules applied to structured data such as laboratory test results or a list of active medications. These rule-based capabilities are low-hanging fruit because they rely on what electronic systems do best—store and run algorithms on structured data. Yet there is so much more that EHRs could and should be able to do. Electronic health records will create greater value for clinicians when they allow clinicians and quality managers to reliably identify adverse events and track them over time. Their value as quality measurement tools will improve substantially when EHRs can automatically generate quality measures that account for the reasons guideline-driven care is adhered to or, if not, why not. Currently, few EHR systems can do these things reliably, primarily because much of the required information resides in “unstructured” form within clinicians’ notes. These notes are rich in detail about signs and symptoms of patients’ conditions, their priorities for clinical care, and their willingness to take some medications but not others. The notes often offer insights into why the clinician chose one medication over another, how patients responded to treatment, and other specifics key to understanding the care patients receive. Clinical notes have to be read manually to extract these details, which limits the ability of clinicians or researchers to examine large numbers of clinical encounters quickly and efficiently. Natural language processing has the potential to alter the landscape by analyzing the context of words and phrases in medical records making them available for computer processing, resulting in the ability to automatically interpret EHRs. Although no consensus definition of natural language processing exists, it is widely used to describe a field of computational linguistics that allows computers to understand human language. Natural language processing has been pursued for half a century, and although it is used in other in-