AI-Complaint-Assistant for document
App Files
Application executable: Media: AI-Complaint-Assistant-Document.zip
Reference file: Media: openFDA device_search_fields.xlsx
Settings.txt CONFIG:OpenFDA-Vertex-V1.0 // identifies schema version (readonly)
https_url1: // openFDA api for source data
limit1: // limit of query results
report_title: // title of output report
data1_fields: // data fields to be analyzed in MDR query results
KEYWORD1-SEARCH: // general keyword search terms (matched in any field)
SEARCH1-FIELDS:SEARCH1-TERMS // query search fields and terms, see openFDA device_search_fields.xlsx for full details of searchable fields
date_received: // example field
device.device_report_product_code: // example field
SORT1-FIELD:SORT1-TERM // query results sorting date_received:desc
COUNT1-FIELD:COUNT1-TERM // query count field/term
AI-ProblemSummaryPrompt: // Prompt that summarizes the product problem
AI-CountSummaryPrompt: // Prompt that counts the instances of items
AI-ProblemSimilarityPrompt: // Prompt that summarizes the problem similarity to a product problem
AI-MDRSimilarityPrompt: // Prompt that analyzes the similarity between input problem and MAUDE results matching query criteria
AI-ReportSummaryPrompt: // Prompt that summarizes the most similar problems
AI-WordsPerReport: // Number of words in each intermediate report (recommended: 1500, range: 100-5000)
AI-ModelTemperature: // LLM Temperature index (recommended: 0.05, range: 0..1f)
AI-ModelTopP: // LLM TOP_P index (recommended: 0.4, range: 0..1f)
AI-ModelTopK: // LLM TOP_K number of words for next word prediction (recommended: 10, range: 1-128)
AI-ModelMaxOutputTokens: // LLM maximum output words (range: 1-2048)
See the sample files page for example settings and reports.
Sample Files: AI-Complaint-Assistant for document
App Description
App Use Case: The AI-Complaints-Assistant streamlines complaint assessments using AI to apply a medical device reportability engine, which offers predictive scoring to determine if a complaint is reportable, sparing manual research of MAUDE records. It optimizes resource usage and prioritizes RA/QA review for timely MDR submissions by assessing the likelihood of FDA reportability based on MAUDE history, rather than just matching data. In addition it also creates pre-populated FDA Form 3500 forms for complaints with high reportability scores, to further expedite the process and improving efficiency. This AI Assistant accepts complaint descriptions from locally provided PDF or Word documents, as well as from network locations. Additionally, it can operate in batch mode, processing one or more documents within a specified folder path defined in the Settings file, automatically including documents from sub-folders, to allow for a retrospective assessment of old complaint records.
App Functionality: The AI-Complaints-Assistant is designed to accept complaint descriptions from specified file source, which may include natural language describing customer complaints and any relevant references to device models, usage settings, troubleshooting information, and other related details. Additionally, it reads a Settings file containing MAUDE query parameters, such as product problem codes, patient problems and outcomes, date ranges, and various other fields, along with prompts used for different processing steps in the similarity evaluation. The predictive reportability scores are influenced by factors such as the device problem code, patient problems and outcomes, the estimated severity of potential harm and other attributes specified in the Settings file.
In cases where there are no similar records found for the complaint input, expanding the MAUDE search to include additional device codes under the same regulation number is possible with a small change in the Settings file. This approach opens up possibilities for broader analysis and leads to more comprehensive insights.
Using the MAUDE query results, the AI Assistant matches the top problems similar to the provided complaint, reporting the estimated similarity for each along with relevant reference information.
Diagram outlining the processing steps of the Complaints Assistant:
A PDF document is generated containing results detailing the top product problems associated with the specified product code, along with additional analytics for each product problem. It also includes the highest matching Medical Device Reports (MDRs) with estimated similarity scores and detailed explanations for their similarity.
Additional information: Another version of this AI Assistant accepts complaint descriptions from a user interface prompt allowing for complaint assessments in real-time or research use. See AI-Complaint-Assistant with prompt for more info.