Innovation into practice

Big data for newborn health

The NHS generates volumes of data but hasn’t always used it as productively as it might. Hospital Episode Statistics have the merit of accessibility but they are administrative data, full of gaps and errors, and entered by coding departments with little clinical oversight.

Much better, potentially, are the patient records used in day-to-day care and increasingly captured electronically. “Where we are fortunate in newborn medicine in the UK is that virtually all the neonatal units in the country are using the same electronic patient record,” says Neena Modi, professor of neonatal medicine at Imperial College. “But what’s really important is that embedded within those records are our defined 400-odd variables that comprise the neonatal dataset. So even if they were using different versions they would still capture that dataset.”

Making use of these records meant clearing many hurdles, both regulatory and ethical. “We had huge difficulties,” Modi says. “It meant making our case to the regulatory authorities about what we wanted to do, because it was completely new. We must have made a good case, because we got approval to extract our 400 data variables.”

The result has been a national database of high quality to which all neonatal units contribute and from which inferences can be drawn. Parents are given the opportunity to opt out, but none has done so. In effect this is a longitudinal birth cohort, created at much lower cost than a conventional cohort study and without imposing any new data gathering duties on the …