Data Characteristics in this Category
Telemedicine pharmacovigilance data primarily originates from remote patient consultation records, electronic prescriptions, wearable device monitoring data, and patient-reported symptom logs. This data has a high update frequency. Consultation records and prescriptions are typically generated in real-time. Monitoring data may upload minute-by-minute. Symptom logs are updated by patients daily or weekly. Document structures vary. Consultation records are often unstructured text, containing doctor diagnoses and patient descriptions. Electronic prescriptions are structured data, including drug names, dosages, and usage instructions. Wearable device data is time-series, with fields like heart rate, blood pressure, and blood oxygen. Symptom logs are semi-structured text or structured tables, describing adverse event times, symptoms, and severity. Common field units include mg, ml, bpm, and mmHg.
Constraints Imposed by these Characteristics on Citation and Traceability
The high update frequency of telemedicine data requires the knowledge base to rapidly synchronize the latest information. This prevents misjudgments due to outdated data. Unstructured consultation records and semi-structured symptom logs require robust text parsing capabilities to accurately identify key information and adverse event descriptions. Diverse data structures mean that citations may need to aggregate data fragments from different sources. An example is combining prescription information and symptom logs to trace an adverse reaction. Time-series data requires consideration of time windows and trend analysis during citation. This ensures that cited data points are representative. Additionally, due to high data sensitivity, the precision and traceability of citation sources are critical. The system must clearly indicate which patient, at what time, and from which record the information originated.
Configuration Settings
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
Chunk Size | 300–500 characters | Balances contextual completeness and retrieval efficiency, accommodating short sentences and key information in consultation records |
Recall Count | Top 8 | Considers that adverse reactions in telemedicine may involve multiple concurrent factors, requiring more context |
Similarity Threshold | 0.75–0.85 | Ensures the relevance of recalled content, filtering out irrelevant consultation records or monitoring data |
Rerank Return Count | Top 3 | Focuses on the most relevant evidence, reduces interference from irrelevant information, and improves traceability efficiency |
Citation Variable Format | {{doc.title}} - {{doc.content}} | Clearly displays the source document title and relevant content snippets for quick localization |
Citation Display Mode | Display citation links only | Prioritizes providing conclusions directly in the response, offering detailed citation links as needed to avoid redundancy |
Three Common Mistakes
- The citation variables
{{doc.title}}or{{doc.content}}are empty. This may be due to original document parsing failure, resulting in missing metadata. - The system reports
quote type error. This typically occurs when attempting to cite an unsupported variable type, such as non-text data. - The response does not include the complete citation file address. This may be because
Citation Display Modeis not enabled or incorrectly configured in the knowledge base settings.
How to Confirm Correct Configuration
- Submit a consultation record containing a typical adverse reaction description. Check if the response accurately cites relevant prescription information and symptom log snippets.
- Test whether citation variables parse and display correctly. For example, check if
{{doc.source}}displays the original data source identifier. - Simulate a drug adverse reaction event. Check if the system provides a clear timeline and relevant data points during traceability. An example is the link from electronic prescriptions to patient symptom logs.
- Verify that clicking a citation link in the response correctly navigates to the corresponding original data record or knowledge base document.
The values given are common starting points and should be measured against the reader's own samples.
Question material comes from public community discussions. Configuration values are common starting points and should be measured against your own samples. Verified on 2026-09-21.