All 4 books by Edward Tufte now in
paperback editions, $100 for all 4
Visual Display of Quantitative Information
Beautiful EvidencePaper/printing = original clothbound books.
Only available through ET's Graphics Press:
catalog + shopping cart
All 4 clothbound books, autographed by the author $150
catalog + shopping cart
Edward Tufte e-books
Immediate download to any computer:
Visual and Statistical Thinking $2
The Cognitive Style of Powerpoint $2
Seeing Around + Feynman Diagrams $2
Data Analysis for Politics and Policy $2catalog + shopping cart
Edward Tufte one-day course,
Presenting Data and Information
Boston MA, March 4, 5, 6
Boston MA, April 24, 25, 26
Cambridge MA, June 3, 4
I attended your course this very day and was astonished. You are the clearest, finest, and most logical--thus most perfect--teacher that I have ever seen--and I have had many, many classes with very, very excellent teachers. I do not cavil, I am not captious; but I note that on page 111 of Visual Explanations that the value printed in red for "Reg Insulin", which I interpret to signify the most recent value injected that day, is in error. It should read "0" if insulin therapy was discontinued.
3 units of insulin administered to a euglycemic patient would cause certain, horrific & a sadly theatrical death.
-- Mike O'Neil (email)
I have asked my co-author, Dr. Seth Powsner of the Yale Medical School to answer this question. Since the patient remained alive, there was no deadly error. Maybe there is some problem with the meaning of "units".
Seth and I are co-authors of an information architecture for monitoring patient status (and other complex dynamic processes with lots of variables and a need for warning signals if any part of the process is in trouble). There are 3 publications about this architecture:
Edward Tufte, Visual Explanations: Images and Quantities, Evidence and Narrative (Cheshire, Connecicut: Graphics Press, 1997), pages 110-111.
Seth Powsner and Edward Tufte, "Graphical Summary of Patient Status," The Lancet, 344 (August 6, 1994), 386-389.
Seth Powsner and Edward Tufte, "Summarizing Clinical Psychiatric Data," Psychiatric Services, 48 (November 1997), 1458-1461, which shows, compared to the medical version of the architecture, more narrative material and a generational history of mental status.
If anyone would like free copies of these 2 articles, please send or fax or email a request along with your snail-mail address.
-- Edward Tufte
Design conundrum perhaps, deadly error unlikely.
We might have displayed a zero for "Reg Insulin." However, a back-up insulin dose is standard practice for diabetic patients in hospital. Nurses administer insulin as needed for particularly elevated glucose levels, the specific dose guided by the current glucose value (a "sliding-scale" dose). Displaying a zero for insulin could be misinterpreted to mean that no insulin what-so- ever is on order for our patient.
The posted question highlights a problem for clinical data display: What is to be done with older values? Certainly, old values should be plotted to provide historical context. But, should they be highlighted by a separate numeric display, or does that become misleading?
Even simple blood tests, like hemoglobin, present this problem, if more subtly. Suppose Mr Jones had a normal hemoglobin last week as part of a routine physical. Today, he returns with abdominal discomfort. Should his physician see a numeric value for last week's hemoglobin, or just a data point? His hemoglobin today is probably the same as last week. That is, unless Mr Jones adds that he has been vomiting blood all morning: it would make last week's hemoglobin almost irrelevant.
Back to our original patient's clinical situation (pneumonia and diabetes): she was not euglycemic. Her glucose was elevated to 237mg/dl. Standard sliding-scale insulin dose for a glucose of 237mg/dl is 4units subcutaneously, at Yale-New Haven Hospital. A dose of 3 units might have been helpful, though, as noted, not the treatment undertaken.
One of our senior Emergency Medicine faculty, originally an endocrinologist, opines that at least 10units of insulin would be required to kill a normal individual, if injected subcutaneously. A larger dose would likely be necessary. Three units might have a some impact given rapidly by intravenous injection. Perspective is provided by Micromedex's Poisindex Managements (vol. 110, http://www.micromedex.com/products/healthcare/ ). Poisindex reports that "recovery has occurred following up to 3200 units in adults."
-- Seth Powsner (email)
Graphical Diabetic Information.
It would be of great personal significance and utility to see how daily information relevant to newly diagnosed Type I diabetics might be recorded graphically and displayed for use by doctor and patient. The initial diagnostic and treatment period is comprised of puzzle solving, but only after the puzzle pieces have been located, not an easy chore. Diet is key. Additional puzzle pieces can be, specifically, blood sugar levels at every two hours, (with special categories for Fasting or Awakening, Lunch, Supper, Bedtime) carbohydrate totals at each of three meals, insulin dosages in units and at what time (made more complicated when more than one type of insulin is used, which is not uncommon). This is perhaps a simple task for those skilled and schooled in making medical charts understandable in a graphic way. Or are the components too varied to handle in a coherent way...words, numbers, comments, data points?
-- Peter Pehrson (email)
I recently (as in yesterday) attended one of your one-day courses and feel I walked away with an expanded sense of how things can be done better. I also walked away with a feeling of relief in that I was not alone in my interpretation of the situation in regards to presentation of information in general, and in the medical setting in particular. In the medical environment, we as "informatacists" have failed catastrophically to improve the situation. I believe one could easily argue that the electronic medical records have made things worse in many regards. The vast, vast majority of those systems have adopted the ODTAA ("One damn thing after another") method of interface design. A colleague of mine once quipped that what we have here is the product of much development, but not much in the way of design. A catastrophe is another way of putting it.
I have read two of the 3 articles you referenced above and noted with a wry smile that you and your co-author, Dr. Pownser, were able to predict the future in your conclusions in the "Graphical Summary of Patient Status" article:
"Our proposal for a graphical summary should encourage doctors and nurses to reshape, perhaps re-invent, the medical record before computer programmers cast institutional convenience into silicon."
The statement would have been perfect if you had stated it thusly:
"...before computer programmers, hospital administrators, and 3rd party payors cast institutional convenience into silicon."
This is exactly what has happened so far. Perhaps you should add "Clairvoyance" to your CV. Perhaps it was because you had a sparkline and could project into the future based on trend data...
Thank you for taking the time to study, and write down your findings in the field of effective display of data and information. At least some of us are listening and hoping to effect a change.
-- Stephen L. Jones, MD (email)
Fascinating article about how checklists save lives in the Medical field.
-- Bill Paton (email)
^^^ Thanks for that Bill! The accomplishments of many of the famous faces, and institutions in the linked article are significant and substantial. The beauty is the simplicity.
-- Stephen L. Jones, MD (email)
A report from Epocrates showed 1/3rd of physicians plan to purchase the iPad Mini, as the size of the device fits well into a doctor's lab coat. An increasing likelihood (and hope) that the medical interfaces you describe can become real-world examples.
-- Will Oswald (email)