As data quality improves, GP extraction services become available and doctors become more adept at commissioning, it is likely more commissioning decisions will rely on the integration of primary and secondary care datasets.



Analysis of prescribing data is of particular interest used in conjunction with risk assessment and long-term condition registries. It can help commissioners pinpoint gaps in provision and quality. Integrating secondary care usage makes it possible to analyse the relationship between medication use and admission to hospital.

We looked at the prescribing data of several GP surgeries in the North West and adjusted our analysis for segmented risk using the combined predictive model, which uses variables pulled from primary and secondary care. This let us analyse whole population risk against a common outcome: unplanned admission.

The first chart shows the whole population “risk pyramid” for patients in this group of practices. The population is segmented into groups of very high, high, moderate and low relative risk categories.

Alongside the risk pyramid you see the percentages of patients in the very high risk group taking one to four, five to nine, or 10 or more unique prescription drugs in any month.

For example, 40.7 per cent of this top risk segment were taking 10 or more different prescriptions a month.

Routine review

There may be good clinical reasons for this but it is a significant variable in risk prediction: a care and quality action might then be to provide a routine pharmacy review to assess potential drug interactions.

Patients in the moderate risk segment had been found to be about twice as likely as the average person to be admitted to hospital. But almost 7,000 patients are an unmanageable number for case management intervention. Unlike with the very high risk group, commissioners will have to be more selective and committed to further work with data.

For example, the second chart shows that of those 6,945 moderate risk patients, 9.8 per cent had coronary heart disease. Of those, 42.1 per cent did not appear to be taking beta blockers (excluding people who had contraindications to beta blockers). Again, this may be for good clinical reasons, but it shows how linking primary and secondary care data can show up potential care and quality gaps and help commissioners fine-tune and target their interventions.