|
Abstract:
|
Reliability is a key metric of the suitability of a quality measure for evaluating medical providers. It describes how well we can confidently distinguish the performance of one medical provider from others. The inter-unit reliability (IUR) is commonly adopted for assessing the reliability of measures. If the measure under evaluation were a simple average across individuals in the medical providers, the usual approach for determining measure reliability would be a one-way analysis of variance, in which the between and within provider variation in the measure is determined. Unlike the individual component measures, the standardized measures or composite measures are not simple averages. Thus, the simple approach of an analysis of variance does not apply directly. We propose new approaches to compute IUR. The proposed approaches are computationally feasible and can be applied to large-scale biomedical data with complex data structures. The methods are applied to national dialysis data.
|
