A fuzzy-match search engine for physician directories.

Area of Science:

• Health Informatics
• Computer Science
• Natural Language Processing

Background:

• Physician search engines are vital for healthcare websites, but current systems often fail due to exact-match requirements.
• Inefficient search results lead to patient frustration and potential loss of website engagement.
• Diverse patient populations in the US necessitate search functionalities that accommodate variations in names, including misspellings and phonetic differences.

Purpose of the Study:

• To design and develop a fuzzy-match algorithm to enhance physician search functionality on the Marshfield Clinic website.
• To address the observed 26% search failure rate caused by the existing exact-match search engine.
• To improve the ease and speed with which users can find physicians.

Main Methods:

• Implemented a fuzzy-match algorithm to identify similar names rather than requiring exact matches.
• Addressed typographic errors, phonetic spelling variations, and nickname discrepancies.
• Utilized a customized Levenshtein distance calculation combined with Soundex coding and a nickname lookup table derived from US census data.

Main Results:

• The fuzzy-match engine achieved 90% accuracy in top-ten searches, a significant improvement over exact match (0%).
• Performance comparison showed fuzzy-match (90%) outperforming Soundex (24%) and Levenshtein distance (59%).
• The fuzzy-match engine-top-one search yielded 71% accuracy.

Conclusions:

• A novel fuzzy-match search engine for physician directories was successfully designed, implemented, and evaluated.
• The developed algorithm effectively handles name variations, improving search accuracy and user experience.
• Open-source code and a demonstration implementation are available for broader adoption and further development.