Merged in bugfix/reimb-primary (pull request #834)

Bugfix/reimb primary

* prompt update for reimb primary

* reimb primary prompt update to extract more info

* prompt update at to priotrize $ amounts

* prompt update for split tbales

* updated valid entries for reimbs

* addition to reimb primary prompt

* reimb primary pormpt update

* Merge remote-tracking branch 'origin/main' into bugfix/reimb-primary
merhing latest main in reimb-primary

* Merged main into bugfix/reimb-primary

* fix- PR comment


Approved-by: Katon Minhas

fieldExtraction/src/prompts/prompt_templates.py

@@ -377,20 +377,22 @@ SERVICE_TERM: Describes the medical service or procedure that is being reimburse
 REIMB_TERM: Describes the method by which the price of the Service is calculated. It can be a simple dollar value (like '$100 per unit') or a percentage of some other value (like '100% of the Medicare Fee Schedule').
 
 [IMPORTANT INSTRUCTIONS]
-- Number one rule: ERR ON THE SIDE OF INCLUSION!
+- Number one rule: ERR ON THE SIDE OF INCLUSION! Return whatever relevant information is available.
 - Ensure that the COMPLETE REIMB_TERM values are captured EXACTLY as they are written in the contract. Return only EXACT word-for-word text from the contract.
 - Sometimes, the SERVICE_TERM may not be explicitly written and instead simply implied. When this is the case, default the SERVICE_TERM to "Covered Services". If needed, add additional relevant detail (e.g. "Inpatient Covered Services", "Hospital Covered Services", etc.)
 - NEVER return a partial sentence - include the entire sentence even if only part of the sentence is necessary.
 - It's ok if the SERVICE_TERM shows up in the REIMB_TERM. Prioritize maintaining complete sentences.
-- Do not leave out any relevant information from either the SERVICE_TERM or REIMB_TERM even if it is not present in the vicinity. e.g. answer "$100 per visit" rather than just "$100". If table headers provide ANY relevant information (like code type, provider type, hospital name, etc), include that in the SERVICE_TERM or REIMB_TERM as appropriate.
+- Do not leave out any relevant information from either the SERVICE_TERM or REIMB_TERM even if it is not present in the vicinity. e.g. answer "$100 per visit" rather than just "$100". If table headers OR table rows provide ANY relevant information (like code type, provider type, hospital name, provider name, Tax ID, legal name, etc), include that in the SERVICE_TERM or REIMB_TERM as appropriate.
 - Include all SERVICE_TERM and REIMB_TERM pairs mentioned, even if they are redundant or overlapping. There should be at least one entry for each SERVICE_TERM and REIMB_TERM pair mentioned.
 - ALWAYS include all relevant percentages and dollar values. If there's even a slight chance that the number is relevant, include it.
 - For each REIMB_TERM, include any additional information mentioned that will be useful in determining reimbursement for the service, e.g., "payment amount will be determined by multiplying it by the DRG weight". It should be appended to relevant REIMB_TERM and not added as a separate entry.
 - If different reimbursement percentages or dollar values are mentioned for different effective dates, create separate entries for each effective date.
+- For the tables: Extract EVERY ROW as a separate entry, including all line items, subtotals, totals, adjustments, charges, and calculated ratios. When rows involve different providers/entities, create separate entries for EACH provider/entity and include the provider/entity name in the SERVICE_TERM (e.g., “Provider Name – Service Category”).
 
 [EXCLUSIONS]
 - NEVER return reimbursement terms that are labeled as Examples or appear in an EXAMPLE section. These are not actual reimbursements and should not be included in the output.
 - NEVER return reimbursements from the Reciprocity Agreements section of the contract. These are not actual reimbursements and should not be included in the output.
+- NEVER return Payment methodology metrics or reference-only tables (e.g., DRG weights, relative weights, LOS, thresholds). These are not actual reimbursements and should not be included in the output.
 
 [SPECIAL CASES]
 Some types of Reimbursements are special and require additional thought.
@@ -1143,13 +1145,14 @@ Determine if the pair of SERVICE TERM and REIMBURSEMENT TERM represent a COMPLET
 
 VALID entries - return YES:
 - Specific payment rates (percentages, dollar amounts)
-- Fee schedule references with rates
+- Fee schedule references (including references to named/specific fee schedules without explicit rates)
 - Calculation formulas
 - Payment limitations with amounts
 - "Lesser of" or "greater of" statements with clear and specific rates
 - Capitation rates (PMPM, PMPY, per member costs)
 - Flat rates per service or episode
 - These Special cases: OUTLIER, STOP_LOSS, SEQUESTRATION, DISCOUNT, PREMIUM, RATE_ESCALATOR, FACILITY_ADJUSTMENT
+IMPORTANT: If a REIMBURSEMENT TERM contains a specific dollar amount or percentage rate, return YES regardless of the program type referenced in the SERVICE TERM. Specific payment amounts always constitute a complete reimbursement methodology.
 
 INVALID entries - return NO unless there is VALID language as well:
 - 'Empty' Reimbursements - have the same structure as a valid reimbursement but doesn't actually reference any specific rate or fee schedule (e.g. "Covered Services will be paid at the rates stated in the next section")
@@ -1164,7 +1167,7 @@ HARD INVALID entries - return NO even if there is VALID reimbursement language p
 - Any reference to Clean Claims in SERVICE TERM only
 - Audit and Time-based language describing how many claims must be paid in a certain timeframe (e.g. "... pay 85% of claims within a 30-day window")
 - Risk-sharing arrangements, surplus/deficit distributions, settlement payments
-- Performance bonuses and incentive payments, or administrative compensation
+- Performance bonuses and incentive payments, or administrative compensation.
 - Non-reimbursement statements ("not covered", "excluded", "not payable")
 - Charge Description Master (CDM) or "chargemaster" language.