Multi-turn Conversation and Prompts for Pharmacoeconomics

Data Characteristics

Pharmacoeconomics regulation data primarily originates from policy documents, guidelines, drug catalogs, and evaluation reports published by national and provincial healthcare security administrations and health commissions. Update frequencies for these documents are irregular, influenced by policy changes, new drug approvals, and evaluation results, with annual or quarterly updates being common. Document formats are typically PDF and Word, containing extensive legal provisions, specialized terminology, evaluation methodologies, cost-benefit analysis models, and drug pricing data. Fields include drug generic names, indications, reimbursement scope, payment standards, evaluation metrics (e.g., QALY, ICER), data sources, and calculation formulas. Units encompass monetary units (e.g., RMB), time units (e.g., years), and quality-adjusted life years (QALY).

Constraints Imposed by Data Characteristics on Multi-turn Conversations and Prompts

The irregular update schedule of pharmacoeconomics regulation documents necessitates a knowledge base with flexible data synchronization and version management to ensure multi-turn conversations are based on the latest policy information. The high density of specialized terminology and complex evaluation methodologies requires the model to accurately identify and link definitions and calculation logic when understanding user queries. For example, if a user asks about "reimbursement differences for a specific drug across provinces," the system must extract and compare relevant information from multiple regional policies. Additionally, numerical data and calculation formulas within the documents demand higher accuracy in multi-turn conversations. Prompt design must guide the model to cite specific values and formula sources in its responses, and explain calculation processes when necessary, to avoid vague or incorrect interpretations.

Configuration Guidelines

Configuration Item	Recommended Value	Rationale
`maxContext`	800–1200 characters	Pharmacoeconomics concepts are complex; sufficient context length helps the model understand user intent and policy details.
`Chunk size` (Segment Length)	400–600 characters	Policy provisions and evaluation report paragraphs are often long; this length helps maintain semantic integrity.
`Recall count` (Recall Count)	5–8 entries	Ensures coverage of multiple relevant policy provisions, evaluation standards, or data points, improving answer comprehensiveness.
`Similarity threshold` (Similarity Threshold)	0.75–0.85	Pharmacoeconomics involves much specialized terminology; a higher threshold helps recall more precise and relevant knowledge snippets.
`Rerank result count` (Rerank Return Count)	3–5 entries	Further filters the most core and directly relevant policy bases or data from the initial recall.
`QUERY_REWRITE_ENABLE`	Enabled	For complex queries and multi-turn conversations, query rewriting improves recall accuracy.

Common Pitfalls

Conversation interface responses lack source citations, preventing users from tracing back to original policy documents. This occurs when the recall strategy configuration does not include original document links or identifiers as required return fields.
Refreshing the page shows "No available index model detected," but the service is actually running normally. This typically results from an incorrect INDEX_MODEL_ID configuration or a failure to load the knowledge base index correctly.
When user queries involve specific calculations, the model provides vague or incorrect numerical values. This happens when prompts do not explicitly require the model to cite data sources or provide calculation steps, leading the model to infer independently.

Verification Steps

For typical pharmacoeconomics policy queries, test whether the conversation system accurately cites chapter or clause numbers from original policy documents, and verify citation accuracy.
Simulate multi-turn conversations to confirm the system maintains context understanding for specific drugs or evaluation methods throughout the dialogue and progressively provides in-depth answers.
Use queries containing specific numerical values and formulas to check if the model's returned values match the original documents and if the calculation logic is correct.
Check the knowledge base management interface to confirm all pharmacoeconomics-related policy documents are successfully uploaded and indexed, and that INDEX_STATUS displays "Completed."

The values provided 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.