This site is about funding, it’s about how we fund healthcare, and about how we fund hospitals. It’s about how we measure and analyze the effect that different funding methods have on the healthcare so that, in the end, we can design better systems.
To do this, we analyze vast amounts of data. We often make use of hospital clinical administrative databases. These databases contain rich clinical and demographic data for every hospital discharge. We rely heavily on the diagnostic data captured in administrative data to adjust our comparisons for differences in the complexity of patients both over time and across hospitals. A crucial question for us and other health services researchers is, “how reliable are the data we use?”
In an article that looks at how diagnostic coding errors affect estimates of adverse event costs, Gavin Wardle and colleagues analyzed the results of a large reabstraction study performed in Ontario. Wardle reports that diagnostic coding disagreements are extensive. For example, the original and reabstracting coders agreed on only about half of important comorbidities and adverse events. Of importance to researchers interested in comparing hospitals, the rate of coding disagreement varied extensively by hospital.
Wardle’s study was of Ontario data, but BC’s data are subject to the same risks for coding error. Variation in the reliability of diagnostic data elements can be caused by:
- Differences in the rate of chart completion and intensity of documentation by clinicians.
- Different interpretation of coding standards established by the Canadian Institute for Health Information (CIHI).
- Willful violation of CIHI’s coding standards. The type and extent of error can depend on how the data are used for funding and performance measurement.
Ontario’s experience, and that of other jurisdictions, has been that some hospitals change their coding practice to improve their funding under models that rely on administrative data. As BC increases its use of administrative data for funding, researchers and policymakers will need to closely monitor the response of hospital coding practice.
What options are available to monitor the quality of hospital data in BC? Some jurisdictions rely on random chart auditing programs to ensure that hospitals do not miscode data to optimize funding. Since these programs must be routine and large scale to ensure the reliability of data, they are resource intensive. Another way to eliminate hospital level variation is to centralize the coding function and randomly assign charts to coders. This would remove hospital-level effects on coding and randomize coder-level effects across hospitals.
Since the accuracy of hospital performance measurement and some reimbursement depends on the quality of data, BC will have to increase its focus on data quality measurement and surveillance.