Automatic Home Medical Product Recommendation

Abstract

Web-based personal health records (PHRs) are being widely deployed. To improve PHR's capability and usability, we proposed the concept of intelligent PHR (iPHR). In this paper, we use automatic home medical product recommendation as a concrete application to demonstrate the benefits of introducing intelligence into PHRs. In this new application domain, we develop several techniques to address the emerging challenges. Our approach uses treatment knowledge and nursing knowledge, and extends the language modeling method to (1) construct a topic-selection input interface for recommending home medical products, (2) produce a global ranking of Web pages retrieved by multiple queries, and (3) provide diverse search results. We demonstrate the effectiveness of our techniques using USMLE medical exam cases.

Triggers and Monitoring in Intelligent Personal Health Record

“…In particular, the maximum weight recorded in the past month is max_w 30 . The maximum weight recorded in the past six months is max_w 182 .…”

“…These features capture the essential information of the user's medical data. For instance, for Scenario 1 described in Section 2.3, the features extracted from the weight reading of the COPD patient are W, max_w 30 , and max_w 182 .…”

“…When the first sub-condition W<0.95×max_w 30 is satisfied, the entire firing condition is satisfied and hence there is no need to compute the feature max_w 182 used in the second sub-condition W<0.9×max_w 182 . The feature max_w 30 is less expensive to compute than the feature max_w 182 , although various sub-conditions in the firing condition have different selectivities.…”

“…The feature max_w 30 is less expensive to compute than the feature max_w 182 , although various sub-conditions in the firing condition have different selectivities. To reduce the overall cost of checking a firing condition C f , we need to determine a good order of evaluating C f 's sub-conditions based on both their selectivities and the computational costs of the features used in them.…”

“…Our current iPHR system provides three functions: guided search for disease information [26], recommendation of home nursing activities [28], and recommendation of home medical products [30]. It follows the pull model of information distribution.…”