OHDSI Home | Forums | Wiki | Github

Mapping visits in CPRD


(Seamus) #1

Hello,

I’m working within the EHDEN project on using the OMOP CDM and OHDSI tools for health technology assessment. An important component of this is modelling healthcare utilisation.

As part of a use case in COPD, I am trying to understand how visits with primary healthcare professionals are represented in OMOP. Since I am familiar with source CPRD data and will be using the mapped version for this work, I have currently focused on this.

I have read the detailed data mapping report which includes some information on the representations of visits but I have some further questions I would really appreciate help with. I am coping in @amatcho and @jennareps because your names are on the mapping report. Any help would be greatly appreciated!

By way of framing my questions I will outline my understanding of the CPRD data and the current mapping to OMOP.

Within CPRD all clinical entries are recorded in the ‘Consultation’ file regardless of the nature of that entry, i.e. they do not all present GP visits but could represent contacts with other healthcare professionals or administrative tasks. Characterising consultations requires two variables: staff role and consultation type. There are about 70 distinct staff roles and 50 contact types. A Lancet study into trends in primary care consultations in England categorised staff roles as GP, practice nurse, other clinicians, or administrative, and consultation types as face-to-face clinic visit, telephone, home visit, or administrative.

In the CPRD to OMOP v5.2 mapping report, it is stated that “All visits from the GOLD data are considered to be outpatient, as all encounters occur in the practice”. ‘Consultations’ from CPRD are mapped to ‘Visits’ in OMOP. Details of the staff role are given in the ‘Provider’ table. I can see no reference to consultation type which would/could presumably be in the ‘Care Site’ table of OMOP.

So my main questions are:

  • Are all ‘consultation’ records in CPRD mapped to Outpatient visits in the ‘Visit_occurrence’ table?
  • Or are some ‘Consultation’ records excluded because they relate to, for example, administrative tasks?
  • Why are visits classified as ‘Outpatient’ rather than another existing concept like ‘Office visits’? For healthcare utilisation studies it is important to distinguish primary care visits from hospital outpatient (i.e. specialist) appointments. Would new concepts specifically relating to primary care, e.g. ‘GP’ or ‘Family doctor’, be valuable? Or with even more detail like ‘Practice nurse’?
  • Should consultation type in CPRD be mapped to the ‘Care_site’ table of OMOP CDM?

While I recognise that the current approach to mapping consultations to visits is perfectly adequate for the main uses of the data to date, for studies of healthcare utilisation (which is a common use of CPRD data) I am concerned indiscriminate use of the data would give misleading estimates and limit the reliability of cross dataset/country analyses. In addition, because not all information from the source data is mapped to the CDM (e.g. type of consultation, as far as I can see), it would not be possible to perform desired analysis using source concepts within the CDM. But that would anyway detract from the advantages of data standardisation.

I am sure similar issues are present in the mapping of other EHR databases.

Sorry for the long message, but I hope the detail will make my questions clearer.

Many thanks,

Seamus


t