Author: Rob Eikelboom, Ear Science Institute Australia
There is no doubt that the move away from paper-based health records to electronic health records has been important for many reasons. Over time the amount of data, also in ear and hearing health, has grown immensely. However, when researchers have started to examine these data, a number of challenges have become apparent. These challenges include:
(i) Ceiling values in audiometry and in non-adaptive speech perception tests. For example, how do we handle the thresholds that are recorded at the limit of audiometers?
(ii) Accessing data in proprietary data databases. For example, there is public and private data in NOAH that is available when viewing clients, but multiple records cannot be exported.
(iii) Linking records of clients that reside in different databases. For example, in the absence of a common identification code, variations and errors in the recording of names and other identifying information hamper linking records.
(iv) Limitations in the available data fields to answer key questions. Large databases are attractive to big data/AI projects, but are predictive models improving?
(v) Correct application of statistical tools; and
(vi) Ethics of using client data. How do researchers deal with the growing awareness of data ownership and confidentiality, and satisfy the requirement of legislation and ethics committees?
The presentation will illustrate each of these challenges with real-life examples, and potential strategies to overcome these.