| HN Mirror

As a physician I can tell you it's not about data entry and data retrieval. It takes time to abstract into words the cognitive model in a physician's brain that represents the patient's condition and progress. Words are static and the format we are forced into documenting is also arthritic and linear. It does not convey the fluid and fuzzy nature of a patient's condition and progress.

Treating a sick patient is a little like reverse engineering a binary. Disassembly into blocks takes some guesswork. Pattern recognition plays a role- therefore you already need to know design patterns before you even approach a binary. Properly codifying your guesses can be helped by using something like IDA but sometimes that slows you down and you can go just as fast using ollydbg/gdb/whatever. This is because the construct is in your brain and you are using these software tools to gather data points, not construct some model on paper that stands like a sculpture or perhaps a Rube-Goldberg device that is a rough approximation of the original source code.

Treating patients begins with such a puzzle. You need to know what diseases look like, what the key distinguishing features of look-alike diseases are. Then you also need to have a feel for the time course of these processes. On this mental scaffold you pin data points gleaned from the tests you order. Trying to communicate this whole mental structure along with the relevant data points and also pointing out the confusing chaff can be a tremendous waste of time. Verbal communication is a lot faster. The abstraction/deabstraction bottleneck is avoided. Instead, approximations are presented so the model is quickly understood by the next physician. The next level of detail is then overlaid. Then another level of detail is overlaid and so on, like building a fractal image. The advantage of this approach is that the big picture is conveyed first without a lot of detail that could be misleading if it receives the wrong emphasis. Also, the receiving physician can interrogate the process early in the big picture phase to verify major diagnostic and treatment decisions.

Writing text is a linear process- it is a serial data stream. However, puzzle solving requires a constant zooming and zooming out as well as non linear jumping around, Puzzle solving is not well suited to a linear data stream. But writing and reading text is all that we are taught since grade school. Even the attempt to structure knowledge into an 'outline' is too restrictive. Interestingly, when we teach children how to use the this linear tool of text writing, we do not use linear methods. Rather, a sentence is approached as a whole entity and then broken down until the meaning is understood. We use this game of 'diagramming sentences' to break down, analyze and communicate the parts of our sentences. The most meaningful parts are the subject and verb. Then we sometimes add an object. Adjectives and adverbs are modifiers. Prepositions are key fulcrums as well. The similarity to medicine is that a patient's condition is analogous to a sentence in English.

When we discuss a patient's condition, there is a similar pattern-the principal parts are presented first. Additional levels of detail are then added to embellish and refine the picture. The model morphs and morphs again as certain facts are emphasized and others de-emphasized. Finally it locks into one of the 'disease patterns' we recognize.

Electronic medical records are not any different than the notes we used to write. They are a skeumorphic sham that do not permit better or faster comprehension. Computers have permitted an amazing use of graphics and improved design patterns that have changed the content on the net. If you saw a web site designed in 1990 you would recognize it immediately. Today's sites are much more dynamic. Font choices, formatting, parallax page movement, drop downs, animation and sound focus our attention to communicate information quickly. Hyperlinks permit non linear navigation. Electronic medical records make no use of that. Graphic primitives, hyperlinks or audio cues are not built in to these programs. Instead we are forced to use relics from a previous era. Why? That's a WHOLE 'NOTHER can 'o worms. There are over 450 electronic medical record systems most of which do not talk to each other. It feels like I am using wordstar. Remember that program??

Improving patient care needs to begin with the patient. We need to be able to give a person an iPad with a picture of a body on it. As they point and click and expand relevant body parts, they need to be given choices of typical symptoms. The whole patient centered part can be in any language- it would be mostly pictures with a few words. As they continue to point and click, the iPad is generating the text that represents their choices. The text could be in the physician's language. A person could do this while waiting for the physician. So when the physician gets to the patient, you just saved half an hour. and maybe avoided some unnecessary tests. and maybe gotten them the care they they needed. faster.