What we know, what we infer, and where the limits are.
A note on limitations
Some cited studies concern junior doctors or non-Australian settings. We pair external evidence with primary interviews conducted by our team and state limitations rather than hiding them.
A
Primary interview research
Project LISA primary research · Audio interview
Interview with Lisa
A first-hand interview conducted by the Project LISA team as part of our problem discovery and validation research.
Project LISA primary research · Audio interview
Interview with Rachel — part 1
Part one of our first-hand interview with Rachel, conducted by the Project LISA team.
Project LISA primary research · Audio interview
Interview with Rachel — part 2
Part two of our first-hand interview with Rachel, conducted by the Project LISA team.
B
The problem is real and measured
BMJ Quality Improvement Reports · 2014
Great Western Hospital, Swindon
95% depend on guidelines daily; mean 133 seconds to locate a protocol; 38% were unable to locate some documents; 20% of searches were abandoned. The paper itself identifies a smartphone app as future work.
Structured mentorship works, but is not reaching everyone
Structured mentorship is effective, yet rural turnover persists—suggesting that effective support is not consistently available where and when it is needed.
NurRAG lifted accuracy from approximately 49% to 76%. LISA’s difference is its response to uncertainty: route an unanswerable question to a human rather than discard it.
LISA’s intended purpose is retrieval of existing approved documents. It does not analyse patient data or generate clinical recommendations. The design is intended to fall within the TGA’s software exclusions; confirming classification is an explicit next step.