"When we create digital records and pretend to anticipate all the needs of clinicians, we make mistakes," Reider said. "We think we’ve anticipated all the questions that can be asked but we really can’t and then we destroy the usability."Standardized data input lets users build “incredibly creative things," he said, "but at a granular level we haven’t been explicit about how to capture data.” For example, there are many ways blood pressure is captured “which means its semantic value is variable. If we settle on a core subset, I would argue, explicitly captured in a standardized way, then we can start to grow the complexity of the systems but only then."
It's been said many times (including on this blog) but it bears repeating: today's medical data leads to tomorrow's medical breakthroughs. In this ever expanding age of evidence-based medicine, physicians logging information about the ailments facing patients and the efficacy of various treatments can end up forming the basis for care for doctors across the world. Researchers can take that data and use it to figure out patterns in behavior, disease, demographics, and outcomes in order to anticipate approaching issues and identify the best means to solve them. Meanwhile, physicians can learn from their peers by seeing similar cases, and medical students learn about the most effective way of treating patients based on what current doctors are doing. The easiest way to get this data out to the various personnel that need it is to capture structured data - that is have defined fields in which responses are entered by the physician. That way, medical personnel can look up anyone with diabetes, for example, and see what the common response is to various medicines or diet changes.
Unfortunately, structured data only works when there is an agreed upon standard. When an industry or domain experts determine which bits of information are important to capture, that's helpful. It highlights the elements that will have the most impact on a patient's health. And while it's important to define WHAT is captured, it is equally important to define HOW it is captured. That doesn't mean by touchscreen, or bar code scanner, or dictation or any of those other user interfaces. The HOW in this case is determining the format for that information - will it be a number field, will it be a set of possible responses to choose from, or will it be so utterly unique case-to-case that it will require a free text solution. The most important part in determining the HOW is to make sure the standard bearers' format is comprehensive enough to cover most possibilities, while limiting the occasions for synonymous terms, or erroneous tangents. If those synonyms or incorrect statements are allowed in, then the data becomes harder to collate and takes more time and money to streamline into usable medical evidence.
This means that the onus isn't just on developers to produce more dynamic health IT solutions that can perform more robust searches on available data, but also on the medical community to begin to define these standards of data. There have been initiatives around the world, including work by The College of American Pathologists (CAP) and the Canadian Partnership Against Cancer (CPAC), to properly explain the requirements for reporting on cases (mostly in the realm of cancer treatment). But these efforts need to be expanded - there need to be more discussions and agreement upon standards so that like can be compared to like and so that the evidence is easier for physicians to find when determining their treatments. And — as has happened with HL7, ICD, and CPT — once a standard is agreed upon, it is much easier for Health IT developers to build solutions to incorporate it and to better utilize those standards. Searching will be easier, predictive models will compile faster, drug interactions and outcomes will be delivered to doctors in a way that is easy to understand and easier to implement. The Tower of Babel fell a long time ago in medicine, and it's time for the community to rebuild a unified language.