Exposing Trends

This was an interesting project! I was working as a senior trainee in paediatric liaison psychiatry and was gripped by the problem of demonstrating that our interventions worked. The scenario is frequently one of a young person repeatedly presenting with a somatic concern which the medical team fails to assuage. It is often possible to enable this concern to reduce and even vanish through a series of conversations or activities broadly informed by a “mental health” mindset (but who will establish causality to general and scientific satisfaction?).

I thought that a time series might help in this task of demonstrating benefit, though I did not have any training in the use of time series. I found a young person who was admitted astonishingly frequently and obtained the admission and discharge dates for a period of about three years, during which time he enrolled in the local CAMHS day programme where he played pool a couple of times with a nurse and maybe chatted about seemingly inconsequential things.

Whether or not there was a drop in his admissions (which I think there was) became lost in my fascination with the numbers. In order to overcome the problem of seasonal variation, known to affect admission in his condition, I calculated moving annual totals (of days spent in hospital) on monthly basis. There still appeared to be a rhythmical cycle in the figures, so I repeated the exercise but using random figures, generated by Excel, setting the range between 0 and 30 days in per month. There was still a cycle visible!

I decided that there was something going on that I couldn’t understand and, under the circumstances, I should go back to being a trainee psychiatrist. I did ask around for some ideas. Someone said that some random number generators don’t generate very random series of numbers. My own idea was that an outlier (let’s say 28 days in one month – but see post below!) might dominate the total as it passed through from being a newby, to a has-been, causing a wave.

If anyone wants to take this up and explain it to me, go for it! There is a “leave a reply” box below, or go to HOME and leave a reply there.

back to AFW in print


One thought on “Exposing Trends

  1. Come to think of it, “outlier” doesn’t make sense with random numbers between 0 and 30, though it might have done with the original data. I am trying to think what happens when there is a group of high scores. Presumably they will arise by chance and dominate for a period ( their influence depending on what happens around them). Perhaps three years was not enough to show that this was not a rhythm with a period, but just slow noise. One day I should redo this with values other than 0-30, 12, and 3.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s