How can we measure behaviour?

Sometimes behaviour is measured routinely. Behaviour can be measured directly or through self-report.  There are obvious benefits to measuring directly: people might not remember what they have done and also might not want to tell you what they have done.  We find that ensuring that responses are anonymous and assuring healthcare professionals of this helps people answer more honestly about their behaviours.

Direct measurement using audit of evidence

Sometimes behaviour leads to the production of evidence. For example, if you prescribe an antibiotic, there is a record of that prescription or if you refer a patient to another service there is a record of that referral.  These data can be used to measure whether behaviours are happening and to see if the number of times they happen change over time.  If these measures are available then they are good to use as they don’t create much extra work for the healthcare professionals.

Observation of behaviour

Sometimes there is no recorded evidence of a behaviour being performed.  You might be able to spend some time observing the healthcare professionals working and make notes of whether they do or don’t do the specified behaviours.  This might be something like observing the use of certain equipment or the use of an open question at the beginning of a consultation.  This is labour intensive, because it involves someone watching healthcare professionals going about their daily work.  However, the benefit is that the observer can get insights into the behaviours and also the situations in which they occur and the antecedents (the determinants of behaviour is covered in section 4).

Self-report behaviour

Sometimes it is impossible to measure evidence or directly observe practice.  We can ask healthcare professionals if they do or don’t do certain behaviours.  There are obvious drawbacks to this: the healthcare professional might not be able to accurately remember and / or they might not want to disclose doing or not doing a behaviour for fear of judgement or repercussions.  There are a number of ways around this.  In terms of forgetting, you can ask people to report within a certain short time frame (perhaps in one day, if that is appropriate for the particular behaviours) and you can ask multiple times and take an average.  In the case of people not wanting to state doing or not doing a behaviour you can reassure people of anonymity and how there is no right or wrong answer.  In general, to get a measure of behaviour we will ask two questions: a) how many times has this behaviour been indicated or possible (this is usually how many patients with a particular indication have you seen) and b) in how many of these cases did you do the specified behaviour.  For example, where the behaviour of interest is antibiotic prescription for upper respiratory tract infections you might ask: a)“how many patients did you see in the last 5 days with upper respiratory tract infections?” and then b) “for how many of these patients did you prescribe antibiotics?”.  Percentage behaviour would then be (b/a) x 100.

Behavioural expectation

Although there is a gap between what people intend to do and what they actually do, behavioural intention is still a good proxy for behaviour in that people who don’t intend to do things do them less than people who do intend to do things.  Expectation is like intention with an added reality check – “I intend to do that every time but I know that things will get in my way so I expect I will do it fewer times than I would like”. 

The benefit of expectation as a measure of the impact of education on practice is that it might change through the course of an education/training intervention so it can be assessed before and after training.  Measures of actual behaviour cannot as of course the healthcare professional won’t have had time to practice again.  Where is not going to be possible to follow up healthcare professionals after the education and training, measuring behavioural expectation can allow assessment of expected future behaviour as an impact of education and training.

Check your application of knowledge

If the behaviour was a nurse in an emergency department of a busy hospital using an ABCDE approach to begin to manage patients who they suspected might be becoming acutely ill, what question could you ask that would measure:

Self-report behaviour


You could ask: a) “how many patients have you seen in the past month* that you suspected might be becoming acutely ill?” and then b) “in how many of these patients did you use an ABCDE approach?”  The self report percentage behaviour would then be (b/a)x100.

a month might not be the right time -period.  It needs to be long enough for these situations to have happened but not so long that people couldn’t possibly remember or too many that people couldn’t make a good estimate. When deciding on the time period, you should think of how many times the situation is likely to have occurred and how likely someone is to remember over that time period.

Behavioural expectation?


You could ask “given 10 patients whom you suspected might be becoming acutely unwell, in how many would you expect to use an ABCDE approach?”  Answer scale would be from 0 to 10.

Are there any sources of evidence for this that you could potentially monitor?  Is observation possible?


There might be notes that detail an approach taken but in this case probably not.  Observation might be possible.

© Copyright - MISC - Manchester Implementation Science Collaboration