Increasingly detailed electronic health records are available and being used to try to work out how to improve management and outcomes in hospitals. However, these data are not always recorded accurately, and the impact of this on how they should best be used is unknown.
We plan to look at commonly recorded measurements of patient status related to infection, including temperature (fever), how well the lungs are getting oxygen into the blood (called “oxygen saturation”), and blood pressure, to see whether these are always recorded accurately. Examples when they might not be include more than expected numbers of temperatures equal to 36.0C (normal is 36.4C; 36 may be used to indicate the patient is “well”), or blood pressure measurements that are multiples of 10 or 5 (so called “digit-preference”).
If we find that these kind of inaccuracies are occurring, we will try to find out when they are more common, e.g. at particular times of the day or days of the week, in particular kinds of hospital areas, after patients have been admitted for a while and are doing better. We will use these findings to make recommendations about how these measurements should best be used in analyses of electronic health records in future.