29 July 2013

How not to assess a service


Well I often moan about the state of statistical communications and point out that people, although not wantonly misleading people, are at the very least implying a different meaning than might be appropriate. After the last post I thought I’d delve a little deeper into why I felt it was a bit daft in my opinion to assess the effectiveness of the IAPA (improving access to talking therapies) in this manner. As I pointed out it is a little bit involved.(full paper here)
First off the figures they quoted as using for their ‘effectiveness value’ were calculated in this instance with this aim
A more robust and transparent measure of the intervention’s effectiveness may entail use of the larger denominator of all those who entered therapy rather than only those who completed it.
patients categorised as ‘moving to recovery’ (KPI 6a) as a proportion of those who
entered psychological therapy (KPI 4).

This will include those who left without completing the therapy (although there is no clear KPI which is defined as “leaving the programme during the quarter” within the IAPT dataset). ‘Entered psychological therapies’ is defined as:
attending first therapeutic session, which may be during the same appointment as initial assessment.’
Please note not in format in the paper found here and that the notations kpi are different depending on the dataset collection data you can now online here. 

Which is basically a percentage of all people who were seen according to the questionnaires to not need therapy at the end of their treatment over all those who came to the service.
Understanding the system.

People are referred either by themselves or from a GP or other medical group/service to talking therapies. They make their case for why they need it and are then offered an assessment appointment for ‘therapy.’ There should be more than one type of therapy on offer which is why they need an assessment to decide which one is most or best suited to them. They will have been sent/given an assessment questionnaire to fill in and bring with them to their appointment to assess ‘caseness’ (baseline indicator questionnaire on wellness). 

This assessment period can take more than one appointment depending on the person and what type of therapy they are being offered. If they agree to therapy they will then be referred to the person who will conduct the therapy, which may not be the person who assessed them. And then they will start their therapy.

 At the end of their therapy they will cease appointments and depending on the provider they will be sent/given another questionnaire to complete and send back. At any point in this process a person can choose to not return to the service or therapy be refused.
Now it seems to me if you are going to look at effectiveness of a service you need to consider the difficulties of each part of the process and although ball parks can give a quick overall assessment they really don’t help understand the problem.

In this instance the assessment period seems quite different to the overall effectiveness of giving treatment. So why assess the overall process when splitting it up would make more sense?
For my mind I would split it into how effective treatment is when adherence is good, how good is adherence and how effective is the assessment process in getting people to adhere to treatment.

Effectiveness of treatment.

I think there are several good reasons not to do it the way these authors have. If you look at the data for IAPTA  you will see one of the datasheet contains
Number of service requests with completed treatment where the service user was not at caseness at initial assessment
i.e. not at the level at which they deemed treatment necessary.
Total number for England 6416 people treated.

Oh dear it looks like the NHS assessed and treated 6416 when there was no apparent need.

 So why were they treated?

Well they were treated because after a face to face assessment these people where seen to not only need but would benefit from having therapy. Now you could argue two things here, either the assessment questionnaire does not give a good overall picture of ill health and should be changed for something else or you should exclude these people from the overall effectiveness figures because they could not improve under these conditions and therefore would always return a negative to the statistics ie reduce the % seen as recovering.

Secondly adherence to treatment is important and the more people attend the sessions the more improvement they see. ie the more effective the treatment is. 

Stands to reason right, so currently completion of treatment is to have undertaken more than 3 sessions of therapy but most courses of therapy are at least 15 sessions if not 20-25 and in some cases of group therapy upwards of 60 sessions. Well I think anyone can see that attending 3 sessions of a course that lasts more than 60 sessions is not going to produce a good indicator of whether that therapy helped. Even 3 sessions of a 15 is pretty bad (20% attendance) A more realistic outlook would come from putting completion of therapy as at least 50% percentage of the overall number of session usually given for that therapy. If people are only going to attend a very small percentage of sessions how can they have completed therapy in any meaningful way and therefore should be excluded from the recovery rates %. Currently there is no published information on this but maybe there should be.

Adherence to treatment.

This is people who start but don’t finish and in my opinion should be a separate figure as this is about the services ability to assess and allocate people to the right kind of therapy. For this calculation I would exclude all those people who were assessed and sign posted to other services within psychological services as these people are in the wrong place and this is not a problem of effectiveness of this service but of the referral process (see later).
Assessment process
What you really want to assess is how effective are the staff at getting people to the right therapy. ie that they adhere to their treatment and improve. And I’ve used the word improve deliberately because it is not always that easy to see improvement. So you want to know how many people come into your service and adhere to treatment of those who are suitable for it. Now I would exclude people who are referred to other services and those who choose not to take the therapy for other reasons. And I would also change the way recovery is counted for instance, did the score on the questionnaire go down by a certain amount. Not, did they not meet caseness but was it less than the previous one as people coming in will vary in the degree to the problems they have and some may score a lot higher than others.

Referral process

The biggest problem I have is in working out whether the right people are being referred to the service. Is there any way to pre-select people based on a telephone call or whether an assessment is necessary for all. I feel there should be a way to assess whether reselection is working but the only thing I can think of is how many people get referred to other services. However I am not sure that is a good indicator as this is a self referral scheme and there may be many people coming to the service who are on the cusp of problems. Hence the figure from earlier of people not making caseness and therefore not seemingly ill although I fear that is more a limitation of the questionnaire rather than the people not needing help.



So after all that do I think any of it could be implemented?

 Well some of it already is but I’m not sure it’s practical to do the rest nationally. Maybe on a local PCT level yes but nationally I think they’d struggle to get all the returns in.
And I certainly don’t think it appropriate to lump all the different parts together in the way that those authors did and if those figures were used in the papers it would undermine a service without just cause. 

If there’s just cause then feel free but that assessment to my mind doesn’t really support it to be fair it doesn't really support much at all and neither does what they are doing.

No comments:

Post a Comment