dose-instruction-parser 2024.1022a0

Tool for parsing free text prescription dose instructions into structured output

dose_instruction_parser: Dose instructions free text parser for Public Health Scotland

Current version: "2024.1022-alpha"

📓 Documentation can be found at https://public-health-scotland.github.io/dose_instruction_parser/

📦 Package is available on PyPI at https://pypi.org/project/dose-instruction-parser/

• The dose_instruction_parser package is for parsing free text dose instructions which accompany NHS prescriptions.
• It draws upon the parsigs package, adapting and expanding the code to the context of data held by Public Health Scotland.
• dose_instruction_parser works by first applying a named entity recogniser (NER) model to identify parts of the text corresponding to different entities, and then using rules to extract structured output.
• The default NER model is named en_edris9. This is an extension of the med7 model, which has the following named entities: form; dosage; frequency; duration; strength; route; drug. en_edris9 has the additional entities as_directed and as_required, and has been further trained on approximately 7,000 gold standard examples of NHS dose instructions which were manually tagged by analysts at Public Health Scotland.
• en_edris9 is not currently publicly available. For interested researchers and colleagues in the NHS, please contact the eDRIS team at Public Health Scotland via [e-mail]. For other users, code to train your own model is available on GitHub.

Contents

1. File layout
2. Setup
3. Usage
4. Development

File layout

📦dose_instruction_parser
 ┣ 📂dose_instruction_parser                  # source code
 ┃ ┣ 📂data                     
 ┃ ┃ ┣ 📜keep_words.txt         # key words which won't be spellchecked
 ┃ ┃ ┣ 📜replace_words.csv      # key words to replace 
 ┃ ┃ ┗ 📜__init__.py
 ┃ ┣ 📂tests                    
 ┃ ┃ ┣ 📜conftest.py
 ┃ ┃ ┣ 📜test_dosage.py         
 ┃ ┃ ┣ 📜test_duration.py
 ┃ ┃ ┣ 📜test_frequency.py
 ┃ ┃ ┣ 📜test_parser.py
 ┃ ┃ ┣ 📜test_prepare.py
 ┃ ┃ ┗ 📜__init__.py
 ┃ ┣ 📜di_dosage.py             # parsing dosage tags
 ┃ ┣ 📜di_duration.py           # parsing duration tags
 ┃ ┣ 📜di_frequency.py          # parsing frequency tags
 ┃ ┣ 📜di_prepare.py            # preprocessing 
 ┃ ┣ 📜parser.py                # dose instruction parser
 ┃ ┣ 📜__init__.py  
 ┃ ┗ 📜__main__.py              # parse_dose_instructions command line function
 ┣ 📜.coveragerc
 ┣ 📜LICENSE
 ┣ 📜MANIFEST.in
 ┣ 📜pyproject.toml
 ┗ 📜README.md

Setup

Basic setup

conda create -n di                              # setup new conda env
conda activate di                               # activate
python -m pip install dose_instruction_parser   # install dose_instruction_parser from PyPI
parse_dose_instructions -h                      # get help on parsing dose instructions

(Optional) Install the en_edris9 model. Contact [e-mail] for access.

Development setup

1. Clone this repository

2. Add a file called called secrets.env in the top level of the cloned repository with the following contents:

   export DI_FILEPATH="</path/to/model/folder>"
   

   This sets the environment variable DI_FILEPATH where the code will read/write models. If you are working within Public Health Scotland please contact [e-mail] to receive the filepath.

3. Create new conda environment and activate:

   conda create -n di-dev
   conda activate di-dev
   

4. Install package using editable pip install and development dependencies:

   python -m pip install -e dose_instruction_parser[dev]
   

[!IMPORTANT] Make sure you run this from the top directory of the repository

5. (Optional) Install the en_edris9 model. Contact [e-mail] for access.

Usage

[!TIP] Run parse_dose_instructions -h on the command line to get help on parsing dose instructions

In the following examples we assume the model en_edris9 is installed. You can provide your own path to an alternative model with the same nine entities.

Command line interface

The simplest way to get started is to use the in-built command line interface. This can be accessed by running parse_dose_instructions on the command line.

A single instruction

A single dose instruction can be supplied using the -di argument.

(di-dev)$ parse_dose_instructions -di "take one tablet daily" -mod en_edris9 

Logging to command line. Use the --logfile argument to set a log file instead.
2024-05-28 07:45:49,803 Checking input and output files
2024-05-28 07:45:49,803 Setting up parser
2024-05-28 07:46:34,205 Parsing single dose instruction

StructuredDI(inputID=None, text='take one tablet daily', form='tablet', dosageMin=1.0, dosageMax=1.0, frequencyMin=1.0, frequencyMax=1.0, frequencyType='Day', durationMin=None, durationMax=None, durationType=None, asRequired=False, asDirected=False)

Multiple instructions

Multiple dose instructions can be supplied from file using the -f argument, where each line in the text file supplied is a dose instruction. For example, if the file multiple_dis.txt contains the following:

daily 2 tabs
once daily when required

then you will get the corresponding output:

(di-dev)$ parse_dose_instructions -f "multiple_dis.txt" -mod en_edris9

Logging to command line. Use the --logfile argument to set a log file instead.
2024-05-28 07:47:56,270 Checking input and output files
2024-05-28 07:47:56,282 Setting up parser
2024-05-28 07:48:18,003 Parsing multiple dose instructions
Parsing dose instructions                                                                                               
Parsed 100%|██████████████████████████████████████| 2/2 [00:00<00:00, 79.78 instructions/s]

StructuredDI(inputID=0, text='daily 2 tabs', form='tablet', dosageMin=2.0, dosageMax=2.0, frequencyMin=1.0, frequencyMax=1.0, frequencyType='Day', durationMin=None, durationMax=None, durationType=None, asRequired=False, asDirected=False)
StructuredDI(inputID=1, text='once daily when required', form=None, dosageMin=None, dosageMax=None, frequencyMin=1.0, frequencyMax=1.0, frequencyType='Day', durationMin=None, durationMax=None, durationType=None, asRequired=True, asDirected=False)

Where you have a lot of examples to parse you may want to send the output to a file rather than the command line. To do this, specify the output file location with the -o argument. If this has .txt extension the results will be presented line by line like they would on the command line. If this has .csv extension the results will be cast to a data frame with one entry per row.

(di-dev)$ parse_dose_instructions -f "multiple_dis.txt" -mod en_edris9 -o "out_dis.csv"

The contents of out_dis.csv is as follows:

inputID,text,form,dosageMin,dosageMax,frequencyMin,frequencyMax,frequencyType,durationMin,durationMax,durationType,asRequired,asDirected
0,daily 2 tabs,tablet,2.0,2.0,1.0,1.0,Day,,,,False,False
1,once daily when required,,,,1.0,1.0,Day,,,,True,False

[!NOTE] Sometimes a dose instruction really contains more than one instruction within it. In this case the output will be split into multiple outputs, one corresponding to each part of the instruction. For example, "Take two tablets twice daily for one week then one tablet once daily for two weeks"

$ parse_dose_instructions -di "Take two tablets twice daily for one week then one tablet once daily for two weeks"

Logging to command line. Use the --logfile argument to set a log file instead.
2024-06-21 08:35:41,765 Checking input and output files
2024-06-21 08:35:41,765 Setting up parser
2024-06-21 08:35:59,572 Parsing single dose instruction

StructuredDI(inputID=None, text='Take two tablets twice daily for one week then one tablet once daily for two weeks', form='tablet', dosageMin=2.0, dosageMax=2.0,  frequencyMin=2.0, frequencyMax=2.0, frequencyType='Day', durationMin=1.0, durationMax=1.0, durationType='Week', asRequired=False, asDirected=False)
StructuredDI(inputID=None, text='Take two tablets twice daily for one week then one tablet once daily for two weeks', form='tablet', dosageMin=1.0, dosageMax=1.0, frequencyMin=1.0, frequencyMax=1.0, frequencyType='Day', durationMin=2.0, durationMax=2.0, durationType='Week', asRequired=False, asDirected=False)

Providing input IDs

The inputID value helps to keep track of which outputs correspond to which inputs. The default behaviour is:

• For a single dose instruction, set inputID=None
• For multiple dose instructions, number each instruction starting from 0 by the order they appear in the input file

You may want to provide your own values for inputID. To do this, provide input dose instructions as a .csv file with columns

• inputID specifying the input ID
• di specifying the dose instruction

For example, using test.csv with the following contents:

inputID,di
eDRIS/XXXX-XXXX/example/001,daily 2 caps
eDRIS/XXXX-XXXX/example/002,daily 0.2ml
eDRIS/XXXX-XXXX/example/003,two mane + two nocte
eDRIS/XXXX-XXXX/example/004,2 tabs twice daily increased to 2 tabs three times daily during exacerbation chest symptoms
eDRIS/XXXX-XXXX/example/005,take one in the morning and take two at night as directed
eDRIS/XXXX-XXXX/example/006,1 tablet(s) three times daily for pain/inflammation
eDRIS/XXXX-XXXX/example/007,two puffs at night
eDRIS/XXXX-XXXX/example/008,0.6mls daily
eDRIS/XXXX-XXXX/example/009,to be applied tds-qds
eDRIS/XXXX-XXXX/example/010,take 1 tablet for 3 weeks then take 3 tablets for 4 weeks
eDRIS/XXXX-XXXX/example/011,one to be taken twice a day  if sleepy do not drive/use machines. avoid alcohol. swallow whole.
eDRIS/XXXX-XXXX/example/012,1 tab take as required
eDRIS/XXXX-XXXX/example/013,take one daily for allergy
eDRIS/XXXX-XXXX/example/014,2x5ml spoonfuls with meals
eDRIS/XXXX-XXXX/example/015,one per month
eDRIS/XXXX-XXXX/example/016,1 cappful every four weeks
eDRIS/XXXX-XXXX/example/017,take two every 4-6hrs for pain
eDRIS/XXXX-XXXX/example/018,up to qid prn
eDRIS/XXXX-XXXX/example/019,one or two tabs dissolved in a glass of water at night
eDRIS/XXXX-XXXX/example/020,bid-tid
eDRIS/XXXX-XXXX/example/021,change every 2 weeks
eDRIS/XXXX-XXXX/example/022,take every fortnight

yields the corresponding output

inputID,text,form,dosageMin,dosageMax,frequencyMin,frequencyMax,frequencyType,durationMin,durationMax,durationType,asRequired,asDirected
eDRIS/XXXX-XXXX/example/001,daily 2 caps,capsule,2.0,2.0,1.0,1.0,Day,,,,False,False
eDRIS/XXXX-XXXX/example/002,daily 0.2ml,ml,0.2,0.2,1.0,1.0,Day,,,,False,False
eDRIS/XXXX-XXXX/example/003,two mane + two nocte,,2.0,2.0,2.0,2.0,Day,,,,False,False
eDRIS/XXXX-XXXX/example/004,2 tabs twice daily increased to 2 tabs three times daily during exacerbation chest symptoms,tablet,2.0,2.0,5.0,5.0,Day,,,,False,False
eDRIS/XXXX-XXXX/example/005,take one in the morning and take two at night as directed,,3.0,3.0,1.0,1.0,Day,,,,False,False
eDRIS/XXXX-XXXX/example/006,1 tablet(s) three times daily for pain/inflammation,tablet,1.0,1.0,3.0,3.0,Day,,,,False,False
eDRIS/XXXX-XXXX/example/007,two puffs at night,puff,2.0,2.0,1.0,1.0,Day,,,,False,False
eDRIS/XXXX-XXXX/example/008,0.6mls daily,ml,0.6,0.6,1.0,1.0,Day,,,,False,False
eDRIS/XXXX-XXXX/example/009,to be applied tds-qds,,,,3.0,3.0,Day,,,,False,False
eDRIS/XXXX-XXXX/example/010,take 1 tablet for 3 weeks then take 3 tablets for 4 weeks,tablet,1.0,1.0,,,,3.0,3.0,Week,False,False
eDRIS/XXXX-XXXX/example/010,take 1 tablet for 3 weeks then take 3 tablets for 4 weeks,tablet,3.0,3.0,,,,4.0,4.0,Week,False,False
eDRIS/XXXX-XXXX/example/011,one to be taken twice a day  if sleepy do not drive/use machines. avoid alcohol. swallow whole.,,1.0,1.0,2.0,2.0,Day,,,,False,False
eDRIS/XXXX-XXXX/example/012,1 tab take as required,tablet,1.0,1.0,,,,,,,True,False
eDRIS/XXXX-XXXX/example/013,take one daily for allergy,,1.0,1.0,1.0,1.0,Day,,,,False,False
eDRIS/XXXX-XXXX/example/014,2x5ml spoonfuls with meals,ml,10.0,10.0,3.0,3.0,Day,,,,False,False
eDRIS/XXXX-XXXX/example/015,one per month,,1.0,1.0,1.0,1.0,Month,,,,False,False
eDRIS/XXXX-XXXX/example/016,1 cappful every four weeks,capful,1.0,1.0,1.0,1.0,Month,,,,False,False
eDRIS/XXXX-XXXX/example/017,take two every 4-6hrs for pain,,2.0,2.0,1.0,1.0,4 Hour,,,,True,False
eDRIS/XXXX-XXXX/example/018,up to qid prn,,,,0.0,4.0,Day,,,,True,False
eDRIS/XXXX-XXXX/example/019,one or two tabs dissolved in a glass of water at night,tablet,1.0,2.0,1.0,1.0,Day,,,,False,False
eDRIS/XXXX-XXXX/example/020,bid-tid,,,,2.0,3.0,Day,,,,False,False
eDRIS/XXXX-XXXX/example/021,change every 2 weeks,,,,1.0,1.0,2 Week,,,,False,False
eDRIS/XXXX-XXXX/example/022,take every fortnight,,,,1.0,1.0,2 Week,,,,False,False

[!NOTE] In this example, eDRIS/XXXX-XXXX/example/010 has been split up into two dose instructions

Usage from Python

For more adaptable usage you can load the package into Python and use it within a script or on the Python prompt. For example, using iPython:

In [1]: import pandas as pd
   ...: from dose_instruction_parser import parser

In [2]: # Create parser
   ...: p = parser.DIParser("en_edris9")

In [3]: # Parse one dose instruction
   ...: p.parse("Take 2 tablets morning and night")
Out[3]: [StructuredDI(inputID=None, text='Take 2 tablets morning and night', form='tablet', dosageMin=2.0, dosageMax=2.0, frequencyMin=2.0, frequencyMax=2.0, frequencyType='Day', durationMin=None, durationMax=None, durationType=None, asRequired=False, asDirected=False)]

In [4]: # Parse many dose instructions
   ...: parsed_dis = p.parse_many([
   ...:     "take one tablet daily",
   ...:     "two puffs prn",
   ...:     "one cap after meals for three weeks",
   ...:     "4 caplets tid"
   ...: ])

In [5]: print(parsed_dis)
[StructuredDI(inputID=0, text='take one tablet daily', form='tablet', dosageMin=1.0, dosageMax=1.0, frequencyMin=1.0, frequencyMax=1.0, frequencyType='Day', durationMin=None, durationMax=None, durationType=None, asRequired=False, asDirected=False), StructuredDI(inputID=1, text='two puffs prn', form='puff', dosageMin=2.0, dosageMax=2.0, frequencyMin=None, frequencyMax=None, frequencyType=None, durationMin=None, durationMax=None, durationType=None, asRequired=True, asDirected=False), StructuredDI(inputID=2, text='one cap after meals for three weeks', form='capsule', dosageMin=1.0, dosageMax=1.0, frequencyMin=3.0, frequencyMax=3.0, frequencyType='Day', durationMin=3.0, durationMax=3.0, durationType='Week', asRequired=False, asDirected=False), StructuredDI(inputID=3, text='4 caplets tid', form='carpet', dosageMin=4.0, dosageMax=4.0, frequencyMin=3.0, frequencyMax=3.0, frequencyType='Day', durationMin=None, durationMax=None, durationType=None, asRequired=False, asDirected=False)]

In [6]: # Convert output to pandas dataframe
   ...: di_df = pd.DataFrame(parsed_dis)

In [7]: print(di_df)
inputID                                 text     form  dosageMin  dosageMax  frequencyMin  frequencyMax frequencyType  durationMin  durationMax durationType  asRequired  asDirected
      0                take one tablet daily   tablet        1.0        1.0           1.0           1.0           Day          NaN          NaN         None       False       False
      1                        two puffs prn     puff        2.0        2.0           NaN           NaN          None          NaN          NaN         None        True       False
      2  one cap after meals for three weeks  capsule        1.0        1.0           3.0           3.0           Day          3.0          3.0         Week       False       False
      3                        4 caplets tid   carpet        4.0        4.0           3.0           3.0           Day          NaN          NaN         None       False       False

Development

1. Please open a new branch for any change and submit a pull request for merging to main
2. If you have ideas for an improvement, or spot a bug, please open an issue
3. Remember to include tests for any changes you might make, where appropriate