Coding the present problem.

Coding of the ‘problem’ in computerised medical records is sine qua non in the informatics community. The ‘problem’ should be formulation of what the clinician thought was the underlying pathological process, ideally not simply a symptom, or procedure (Box 1). Many in informatics take it for granted that coding clinical data is a good thing; without considering if there is an evidence base for this activity. Whilst much of what we know about health and disease comes from routinely recorded computer data, the quality of that data and our ability to extract it without loss in its fidelity limit its usefulness. The informatics community has focused on data quality rather than developing an evidence base about the value and utility of the data collected. Data quality is defined in terms of its completeness and accuracy, currency (i.e. how up-to-date it is), and in terms of its positive predictive value and sensitivity that someone identified by routine data actually has that condition. Data quality has also been described functionally in terms of its ‘fitness for purpose.’ Kalra et al., in their review of the empirical evidence wakes us up to how there is little high-quality evidence for the benefits of clinical coding. There is no evidence of harm, but a dearth of studies providing positive evidence of benefit from coding data. The principal benefit, in terms of outcomes arising from the use of coded data, is in the management of longterm conditions in which prevention or therapeutic intervention reminders are linked to coded data. This same type of linkage is also used to improve patient safety by providing relevant prescribing alerts.