House Prices
My wife and I are presently in the market to buy a house. A recent visit to an open home prompted the discussion on how to fairly price a given property. There is a subjective element of course, but pricing based on location, number of beds, baths and car spaces seem to be the logical and intuitive first step for calculating price. What does the median 4 bed, 2 bath, 2 car house sell for in the same post code? This is an ‘outside view’ way to calculate an uncertain figure. Law of large numbers means a data driven approach gives you the best odds of being in the ballpark.
Inside (the mouth) view
If we were to treat a dental patient the same way as pricing properties, I would gather critical information like age, gender and smoking status. Then I would be able to create a risk profile even prior to you sitting in the dental chair. Alas dentists aren’t taught this way! We’re taught to thoroughly, and systematically look at the case subjectively. It’s a textbook ‘inside view’ first. As Phil Tetlock puts it in his book super forecasting, doing the inside view first has potential to anchor you to a shockingly erroneous point. Going outside view first will likely anchor you to a much more sensible, in the ballpark point. If we valued properties the same way a dentist treatment plans teeth, we’d look at how nice the fittings are, and pay no heed to post codes.
If we valued properties the same way a dentist treatment plans teeth, we’d look at how nice the fittings are, and pay no heed to post codes.
Data in silos
I’ve written a bit about the lack of data driven decision making in dentistry (here). One of the real problems though, is how decentralized dentistry and therefore the data is. Each clinic has several thousand patients. For big data and law of large numbers to thrive, we need to unite the silos to create macro trends and good risk models.
Outside first, then inside
The ideal way is to consider both inside and outside views. The risk profile/model goes first, and then you look inside and adjust according to what you see. When a new patient walks in the door, ideally the dentist should have an idea on how many active cavities should I expect to see someone with characteristics x,y and z. After the base number is calculated, then the subjective inside view should tune for accuracy and common sense.
Better outcomes but also catch outliers
Getting anchored to the outside view will give better outcomes, but as the patient it also helps reduce information asymmetry somewhat. Knowing how far you’re away from the median of your particular risk profile will let you be more sceptical about outlying recommendations. Both false negatives and false positives could be caught by democratising knowledge of outside view risk factors in dentistry.
Perfect for prediction
According to Nassim Taleb, dentistry lies in ‘mediocristan’. You’re never going to see a human tooth 1000x bigger than the previous. These regularities in the environment makes dentistry the perfect field for big data driven prediction and risk profile building. The goal is difficult and lofty, but in a world with such high interconnectedness, the task shouldn’t be impossible.