I found a discussion of it in Michael Gregg’s book, Field Epidemiology, chap 6, The Field Investigation. That section was written by Bob Fontaine and Rick Goodman. (btw - Rick is one of the folks who came up with the idea of a Disease Detectives event in the Science Olympiad.) They say “The choice of the time interval used is critical. Intervals that are too short (like hours for diseases with long incubations) may overemphasize random noise in the underlying pattern and hinder interpretation. Intervals that are too long (like weeks for diseases with short incubation periods) will group many cases into a few intervals and obscure the real pattern. As a general rule, intervals between one-fourth to one-half an incubation/latency period work best at revealing the time pattern of an epidemic. As numbers increase, shorter intervals will reveal more detail to the pattern.” (p 88 2nd edition)scio444 wrote:Okay, so on epi-curves- I know that the unit of time should be 1/3-1/4 of the typical incubation period. But I have been noticing that only multiplying the time unit by 3 is giving me an incubation period that falls within the range on the answer keys. Any thoughts on this? What numbers do you guys use?
The incubation period is the time period between exposure and onset of an infectious disease. The term “latency period” is similar but often used with chronic diseases. The idea here is the scale you use for the X-axis affects the usefulness of the curve. Let’s say you have a salmonella outbreak where 50 people get sick over a 3 day period. If you mark off the X axis in minutes, your X axis would be at least 60min/hrX24hr/dayX3days or 4320 intervals long. There would be a lot of blank space between the 50 cases and few, if any, points would have more than one case. If you were to mark off the X axis in weeks, all 50 cases would be in the same space. That does not tell you much either. If you use day, you will have data for 3 points and you could probably show a start, peak and end. If you were to break in down into 8 or even 12 hour intervals, your graph would become more interesting.
While I might use the incubation period as the basis for setting my X-axis intervals, I would not put much faith in using X-axis interval as the basis for determining the incubation period for an outbreak where I did not know the bug that was causing it. I would much rather use distance between peaks if I had them.