To the Editor: The Garling report highlights gaps in the New South Wales health system and establishes a framework for change.1 We support Garling’s view that there is a need for greater measurement of the health system to assess quality of care, and we believe that clinical networks have an important role to play in establishing, interpreting and actioning quality indicator data.2 But the statement that “There is already a great reservoir of information available within NSW Health from which data are available to measure the quality of patient care ...”1 is unhelpful in moving this work forward.

While it is true that vast amounts of data are collected and classified using the ICD (International classification of diseases) coding system, in their current state these data have limited ability to accurately measure quality and drive change at a clinical level, for a number of reasons. First, data are not recorded in the medical record in a planned and systematic manner using agreed definitions, making findings susceptible to reporting bias. Second, the coding system fails to accurately account for differences in patient casemix, which is needed to enable appropriate risk adjustment to occur. This was recently highlighted in the United Kingdom when administrative data were compared with registry data to assess outcomes after cardiac surgery: admitted episode data underestimated the number of patients receiving treatment by approximately 20%, and captured only between 27% and 78% of deaths at 30 days in nine of the 11 centres undertaking cardiac surgery.3

Others have gone down the path of investing huge sums of money to measure quality using administrative data and reward high performers, with moderate success at best.4 To emulate this approach in the absence of further investigation would be premature.

To measure quality of care we need to:

• invest in developing high-quality clinical registries across a broad range of high-cost, high-impact diseases and procedures. Where these have been introduced, they have had demonstrable impact on improving patient outcomes.5 These registries may be populated in part by data elements from existing data sources, where these are of high quality.

• develop data fields for the electronic medical record to ensure that data are epidemiologically sound. This work must include the adoption of unique identifiers to enable monitoring of treatment across care boundaries.

• validate and test proposed indicators, which might be collected from existing data sources.

It is only through assuring clinicians that the data used to measure quality of care are reliable and clinically relevant that they will have any positive impact on changing practice.