Analysis of Missing Dementia in an Urban Environment

My friend Richard Laing, a SAR Manager and GIS expert from Ridge Meadows SAR has written an analysis of missing subjects with dementia in the Metro Vancouver area. The report analyses the distance the subjects were found from the Point Last Seen (PLS), and calculated the 25%, 50%, 75% and 95% confidence intervals. You can view the document here.

This work expands on the mandatory reading in Robert J. Koester’s Lost Person Behaviour. Koester wrote the book after creating the International Search and Rescue Incident Database (ISRID) which gathered statistics from many countries and contains over 50,000 records.

Laing’s research applies the same numerical analysis to incidents in the Metro Vancouver for the specific profile of a subject with Dementia to see if the same confidence intervals hold. If you are interested in the specific details of how certain profiles are identified, or what constitutes the Dementia profile, I suggest you read Koester’s book.

When you have a lost person there are three general ways to attempt to narrow down the area to look for them. The first is using logical methods; trails, and barriers that are hard to cross, like creeks, rivers and mountain ridges. The second is theoretical; you multiply how fast the subject could move by the number of hours or days they have been moving and draw a circle of that size on the map. The third method is statistical — where you analyse previous incidents where the lost person fit the same profile, and use those numbers to predict how far a subject might be found.

In the case of a lost subject in an urban area the statistical model seems to work best because the size of the theoretical circle is almost always too large to search since the subject could easily get on public transit. Coupled with the sheer complexity of finding someone in a crowded urban landscape, being able to narrow down the search area is a very good thing.

The value of having an analysis based on local cases is that searches in the same area are often under the same conditions. If there are local variables at play, then using an analysis based on local numbers has very high value.

In the case of Laing’s analysis, the 25%, 50% and 75% confidence intervals seem to be in similar scales as those from the ISRID, while  the 95% interval is half that of the larger data set. It’s possible this is an artifact of Metro Vancouver having several rivers that limit travel distances, or some other factor.