Research and evaluation methodologies

We have developed a range of methods to research and evaluate health data and computerised medical record systems.

• Extracting primary care data. The mainstay of the groups work over the last decade has been the extraction of routinely collected health data from primary care. The group has a detailed understanding of the differences between brands of computer systems and has developed new levels of sophistication in its data extraction and analyses. These include: sophisticated processing of demographic and ethnicity data. We can provide information about deprivation at the individual patient level; we can map ethnicity codes to any census category group; we collect data about encounter type to help differentiate incidence from prevalence. We can obtain data about when people join and leave a practice and death data. Our largest data set was collected from 2.4 million patients, or largest single dataset had over 3,000 variables.

• Linking to survey data. We can identify people who meet inclusion and exclusion criteria for studies from routine data using pseudonyms so these patients are only every identified in their own practice.

• Calculating risk outside the clinical database. In a similar way we can stratify risk (e.g. of stroke risk in atrial fibrillation) using pesudonymised data extracted from the CMR. The identity of these patients is only made know to the clinical team directly involved in their care, by decoding the pseudonym for that team. We can subsequently re-audit that any quality improvement intervention has taken place.

• Linking health data – open and secure linkage. We have now linked primary care, clinic (with proprietary system), screening database, and hospital data in studies.

• Secure and Private Record Linkage (SAPREL). We have also demonstrated – working with a commercial partner – that we can pseudonymise, and separately encrypt individual strong identifiers, and then use fuzzy logic to link health data from multiple sources. The advantage for governance is that the researcher never sees any patient strong identifiers.

• Observing how the computer is used in the consultation (ALFA open source toolkit). We have developed an open source toolkit to look at the usability of GP computer systems as well as explore how consultations about a particular condition are abstracted into a clinical record.

• Decision and information support. We also have expertise in linking from clinical records or observations to decision and information support.

• Ethics, governance and information security. We are familiar with, compliant with the processes of research and information governance – including relevant legislation and information security. We hold our anonymised data on our own secure servers which are entirely separate from any other networks at St. George’s originally, and now at Surrey University. We have successfully steered project through research ethics committees, through local R&D, arranged information sharing agreements with practices, and successfully received approval (and commendation) from National Information Governance Board for Health and Social Care (NIGB), when constituted as the Patient Information Advisory Group (PIAG).