F Flahaut, La Parole intermédiaire, París, 1978, p 37.

Twenty-six out of 27 GPs expressed their strong support for the use of EHRs to collect outcome data for trials within the database (Table 17). Primarily, this approach is regarded as an efficient use of an existing, highly valuable resource of patient information within the primary care database to facilitate much needed research on real patients within real primary care settings. GPs across all groups prefer the remote

electronic method for data collection over clinician-led questionnaires to reduce practice workload: It’s the way forward.

GP 1017 (recruiter)

primary care has been computerised for about 30 years so there’s a huge volume of information in GP

computer systems.

GP 1019 (recruiter)

I certainly felt EHR was a sensible and interesting way of getting useful outcome data of real world stuff. We need to be using our data from our patients, we have a strong data base, we have a strong history of computerised records and reasonably good data which we really should be making better use of.

GP 1036 (incomplete set-up)

EHR is intriguing as a modern method to replace subjectivity of patient self-report with hard

end-points as well as reducing the clinician’s workload.

GP 1031 (incomplete set-up)

TABLE 17 GP support for outcome data collection via EHRs

GP by eLung status Support for EHR Lack of support for EHR

Decliner (n= 9) 9 0

Withdrawal (n_{= 3)} 3 0

Incomplete set-up (n= 8) 7 1

Recruiters (n_{= 7)} 7 0

Total (n= 27) 26 1

QUALITATIVE PROCESS EVALUATION (QUEAN)

NIHR Journals Library www.journalslibrary.nihr.ac.uk

EHR is an interesting method and patient questionnaires have value for different purposes too.

GP 1005 (incomplete set-up)

It’s still in its infancy but it looks to have great potential.

GP 1037 (recruited)

EHR is a preferred method as its difficult to get questionnaires back and is a lot of administrative work to chase them up.

GP 1024 (incomplete set-up)

Use of EHR could help to achieve participation in research from other GPs as it’s less work and

tick boxing.

GP 1011 (incomplete set-up)

The use of EHR was one of the things we liked about it.

GP 1048 (withdrawal)

It [EHR] sounds like a good idea as it would reduce the workload.

GP 1029 (decliner)

It’s [EHR] much better than questionnaires.

GP 1032 (decliner)

An integrated approach using routine data is more efficient.

GP 1004 (decliner)

I have no concerns with EHR and it’s less work for GPs.

GP 1052 (recruited)

EHR sounds efficient as the amount of time spent extracting data for study questionnaires is demanding and burdensome.

GP 1025 (recruiter)

EHR is massively advantageous as it removes the need for the GP to be involved.

GP 1030 (recruiter)

One GP decliner particularly liked the potential for longitudinal data collection using routinely collected EHRs and one GP who had not completed set-up for eLung liked the scope for hard data end points compared with patient-reported outcome measures. One GP believed that EHRs would motivate other GPs to participate, suggesting this might be an important methodological element in potential education and/or recruitment strategies.

The one GP (incomplete set-up) who did not support use of EHRs to collect trial outcome data was very concerned about confidentiality of patient data and similar potential concerns by patients:

If the patient is fine and they’re OK for the data to go [to CPRD], I don’t have any problems with that.

But if the patient comes all the time and the data is being downloaded [by CPRD], if they have a

problem with that, then I don’t know about EHR. As a GP if you need any information and the patient

consents, that’s fine but if the patient says no, that’s it. I don’t know what misconceptions they might

have. As long as it specifies what is the reason it’s being collected and what benefits it is and what

future outcomes it will be and it’s just for this condition and its confidential and all that. And as long

as it says all this and the patient agrees, I don’t see any reason why it can’t be done. But if the patient

says no, I have reservations, I don’t want my data being collected and if he’s not very sure about what

it involves, then the explanation might take time, it might take time so as long as the leaflet is clear.

This GP identified the need for patient information to be explicit about how and when the data would be accessed and by whom. Two GP decliners who were supportive of EHR in principle also expressed some concerns regarding patient confidentiality. One GP suggested data should be anonymised at all times and to exclude patients who have signed a disclaimer and another GP highlighted the need for more

information from the EHR database on this issue if they had taken part in the pilot trial.

General practitioners across all groups of eLung status generally agreed that most patient data would be available to search remotely via free text with a small number of GPs in each group being willing to enter study data into pre-defined study-specific codes. Free text was generally considered more appropriate to increase access to non-practice data, such as out-of-hours and hospital discharge data which are scanned rather than coded into patient practice records. Most GPs believed that the EHR database would need to consult with individual GPs on a case-by-case basis to assist with interpretation of ambiguous clinical data:

The quality of coding and variability of coding is a major concern. I disagree with up to 50% of hospital codes and valid interpretation is tricky, for example, was it a chest infection, exacerbation of COPD, chest pain, an upper respiratory traction infection? The headline code from the hospital is often not what seems to have been the major reason why the person is in hospital so you have to look at the sub-coding and treatment pathway to try to work it out. CPRD have assessed their coding of episodes per prescription at 95% accurate but 85% for acute prescriptions. This may be a problem statistically and there is huge variation between practices. I would recommend we keep the coding as simple as possible or use free text to avoid missing data on searching.

GP 1005 (incomplete set-up)

This is a result of variation in codes and quality of coding between practices combined with the need for clinical expertise to interpret more complex cases. The need to avoid double entry into the patient and trial records and a preference for using codes consistent with local and national guidance was stated by several GPs across several subgroups. Several GPs flagged the need for financial remuneration to compensate for back-fill of GP time to provide this role. One GP recruiter (GP 1030) suggested a funding formula could be built into the remuneration package for every query answered to achieve quality data in a way that does

not waste people’s time.