Which test?

#1
Hi,

I'm a clinician who hasn't done much stats since my undergrad (won't say how long ago that was but some people really wanted to zigga zig ah at the time). I could use some guidance on which test to use to assist in a service evaluation we will be conducting. We are measuring the number of mental health consultations carried out by the GPs in a cluster of surgeries for a snapshot of a week. We will measure this at baseline, 3 months, 6 months, 9 months and 12 months. We hope that the introduction of a new service should reduce load on GP and improve service for patients. How would I best measure what is happening over time with the number of consultations? Any advice gratefully received.
 
#3
Hi Karabiner

We have 16 GPs involved and around 20% of all consultations in a week concern mental health as the primary focus.

Kind regards
 

Karabiner

TS Contributor
#4
Is the proportion of MH consultation among all consultations your outcome variable,
or is the number of MH consultations the outcome variable? If the latter, how often
do they typically carry out mental health consultations?

In addition, will a MH consultation be identified and recorded as such in exactely the
same way, throughout the whole study? If boundaries are blurred and GPs decide what
counts as MH consultation and what not, then a program could introduce changes in
everyday definition of the outcome, but maybe not in the outcome itself.

With kind regards

Karabiner
 
#5
Hi Karabiner

The number of MH consultations is the outcome variable. GP's use codes to represent the focus of consultations and the same codes will be used to identify MH consultations for all GPs at all time points.

Kind regards
 

Karabiner

TS Contributor
#6
Karabiner said:
How often do they typically carry out mental health consultations?
Please?

GP's use codes to represent the focus of consultations and the same codes will be used to identify MH consultations for all GPs at all time points.
The codes will be the same, but how is it assured and controlled that exactely the
same consultation will not be coded as MH at baseline and non-MH at 6 months
(or vice versa)? There are probably cases where MH and non-MH issues and problems
overlap, and it might be decided by the GP how to code it. If that can happen, you'd
normally need some quality assurance of your outcome measurement, in order to rule
out bias.

With kind regards

Karabiner
 
#7
Hi,

Frequency and level of mental health consultations will vary depending on what patients have booked appointments that day. However, it is very rare that a GP will have a day without at least one consultation where mental health is the primary problem.

GPs will be given guidance and where mental health is present it is coded and treated as a primary problem. Despite this the team will have to take on board there will be a level of bias as we are relying on others to record the data.

Kind regards
 

Karabiner

TS Contributor
#8
Frequency and level of mental health consultations will vary depending on what patients have booked appointments that day. However, it is very rare that a GP will have a day without at least one consultation where mental health is the primary problem.
You gave a very precise estimate (20%) of the share of MH consultations.
This now is a bit vague, but I'd suppose that at least 5 MH consultations
per week will take place at baseline.

This is a repeated measures design with n=16 subjects and a repeated-measures
factor with 5 levels. You can perform a global test of whether there were
changes over time using the repeated-measures analysis of variances, or
alternatively the Friedman test, especially if number of MH consultations per
week is generally low.

With kind regards

Karabiner