Create script to analyze created annotations

[tamslo]

Preparation

• Get guidelines and resolve bricks
• Refactor to have checks in single files and only configuration in main analyze file
• Organize log to put failed checks first

Introduce checks, potentially correct, and describe in README

• Consult text (also correct)
• Trim leading and trailing whites paces in drug brand names (also correct)
• Use "much" keyword only if explicitly mentioned in guideline text (usually "significantly")
• Warning levels
  • "up to a certain dose" > yellow (but might be same as side effects)
  • “if #drug-name isn’t working for you…” > at least yellow
  • "increased risk for side effects" > at least yellow
• Metabolization before consequence
• Special guideline sanity:
  • Any fallback has only one guideline
  • Any and any not handled should only have one lookup value in lookup key
• Warn if annotated but not staged (drug with guidelines)
• Remove “You have a normal risk of side effects.” if not in external data
• Non-metabolizers or ignored: warn if "break down" or "activate" in implication (DPYD, G6PD, SLCO1B1, MT-RNR1, HLA-B, HLA-A)
• Single drug brand names should not contain ", " (automatically correct [split])
• Check "slow titration" > "cautiously and slowly"
• Check for unsused bricks
• Check that same indication/recommendation result in same bricks (per guideline)
• Add rule checking per medication: Guidelines should either use break down or activate

Checks to maybe implement in the future

• Check for "still" cases if contradictory (e.g., normal but change dose)
• Check risk severity
  • "may increase risk" used if "may" near to "risk" in guideline
  • Check "may have increased risk" in “increased risk” cases
  • Check "much" cases