|
Abstract
|
We thank Kartik Singhai and Naresh Nebhinani for their suggestions, which have provided us with an opportunity to clarify some concerns pertinent to register-based epidemiological research.
Singhai and Nebhinani raise questions regarding our choice of diagnostic criteria, possible discrepancies between the ICD-9 and ICD-10 classifications, and the lack of information on who made the diagnosis. In response, we wish to highlight two considerations affecting research based on electronic health records: (1) the context of the practice (in this case, Hong Kong has adopted ICD-10 for coding hospital diagnosis in a phased approach since 2001, with use of the clinically modified version of ICD-9 [ICD-9, Clinical Modification] continued for compatibility); and (2) the issue of measurement, as unlike in controlled trial environments (which face other challenges such as selection bias and small samples and statistical power), some level of noise is inevitable in a real-world setting.
In this regard, we could consider not just a mental health professional versus general practitioner dichotomy, but also other potential factors that might affect the accuracy of diagnosis, such as years of practice, clinician time available, and the screening or triage system. When health records are used, we are interested in how real-world diagnoses predict particular outcomes, without holding assumptions about the factors affecting these diagnoses.
Another concern was the balance between adjusting for confounders and over-adjustment. In our study we only included well established risk factors, instead of all known potential confounders.
Some information is seldom available in health records (eg, personality traits). Patients provide clinicians with the necessary information for their care and treatment, which can also be used to advance research without further intrusion. However, Singhai and Nebhinani's inquiries about non-clinical information are a valid reminder of this intrinsic limitation for any individuals involved in the generation and use of research findings based on electronic health records.
Finally, our study focused on the first record of self-harm only, as the complexity of repeated self-harm ...
Language: en |