Routine data: a resource for clinical audit?

The recent review by Martin McKee' makes several important points but also raises one issue which needs further discussion. This concerns the definition of "routinely collected data" which he proposes: "data whose primary reason for collection is other than audit." It is implicit in the paper that the definition refers to data which are collected regularly and over an extended period. However, omitting these concepts from the definition leaves them unqualified. For example, would a system which regularly gathered data on, say, 10% of patients count as routine, or is this definition limited to systems covering all patients? Similarly, how long need the period of collection be: should it be open ended or does the definition include systems planned to collect data for only one or two years? There is a more serious confusion which arises from this definition. It leads to systems such as the Lothian2 and North West Thames3 surgical audits, which have collected data on every patient seen for an extended period, being described as "nonroutine." The problem arises because "routine data collection" is defined in terms which do not relate to the way in which the data are collected but only to the reasons for their collection that is, data which were initially collected for reasons other than audit. As such the definition does not follow the convention of "describing a thing by its properties4" but uses instead the motives of the unseen planners of data gathering. What happens when, as is often the case with large scale data collection systems, there is more than one motive behind the data gathering: how do we decide which is the primary one? Equally, what if several people with different motives are involved in the design of the system: whose motive is to be considered paramount? A further distinction needs to be made about the reasons for data collection, that of ownership. For some systems, such as those in pathology or pharmacy, data are requested, processed, and used by the health professionals involved in their collection. However, for other systems, such as the Korner minimum data set or cancer registry returns, data are provided by health care professionals for official purposes. Thus the purpose for which data are collected is more complex than simply whether they are to be used for audit. There is a need for a standard definition which covers all dimensions of routine data: purpose of collection, ownership, and the features of the collection process.