JMIR Publications

ABSTRACT

Background:

Medicine 2.0 creates the need for applications that find similar patients based on a patient's electronic health record (EHR). At the same time, there is an increasing number of longitudinal EHR datasets with rich information, which are essential to fulfill this need.

Objective:

We evaluate the hypothesis that we can leverage similar EHRs to predict possible future medical concepts (e.g. disorders) in a patient’s EHR.

Methods:

We represent patients’ EHRs using time-based prefixes and suffixes, where each prefix or suffix is a set of medical concepts from a medical ontology. We compare the prefixes of other patients in the collection with the state of the current patient using various inter-patient distance measures. The set of similar prefixes yields a set of suffixes, which we use to determine probable future concepts for the current patient's EHR.

Results:

We evaluated our methods on the MIMIC II dataset of patients, where we achieved precision up to 56.1% and recall up to 69.5%. For a limited set of clinically interesting concepts, we found that 86.9% of the true positives are clinically useful, that is, these procedures were actually performed later on the patient, and only 4.7% of true positives are totally irrelevant.

Conclusions:

Our results show that this is a promising direction of research. Effectively predicting medical concepts can have several applications, such as managing resources in a hospital.