pathling

Python API for Pathling

This is the Python API for Pathling. It provides a set of tools that aid the use of FHIR terminology services and FHIR data within Python applications and data science workflows.

View the API documentation →

Installation

Prerequisites:

• Python 3.8+ with pip

To install, run this command:

pip install pathling  

Encoders

The Python library features a set of encoders for converting FHIR data into Spark dataframes.

Reading in NDJSON

NDJSON is a format commonly used for bulk FHIR data, and consists of files (one per resource type) that contains one JSON resource per line.

from pathling import PathlingContext

pc = PathlingContext.create()

# Read each line from the NDJSON into a row within a Spark data set.
ndjson_dir = '/some/path/ndjson/'
json_resources = pc.spark.read.text(ndjson_dir)

# Convert the data set of strings into a structured FHIR data set.
patients = pc.encode(json_resources, 'Patient')

# Do some stuff.
patients.select('id', 'gender', 'birthDate').show()

Reading in Bundles

The FHIR Bundle resource can contain a collection of FHIR resources. It is often used to represent a set of related resources, perhaps generated as part of the same event.

from pathling import PathlingContext

pc = PathlingContext.create()

# Read each Bundle into a row within a Spark data set.
bundles_dir = '/some/path/bundles/'
bundles = pc.spark.read.text(bundles_dir, wholetext=True)

# Convert the data set of strings into a structured FHIR data set.
patients = pc.encode_bundle(bundles, 'Patient')

# JSON is the default format, XML Bundles can be encoded using input type.
# patients = pc.encodeBundle(bundles, 'Patient', inputType=MimeType.FHIR_XML)

# Do some stuff.
patients.select('id', 'gender', 'birthDate').show()

Terminology functions

The library also provides a set of functions for querying a FHIR terminology server from within your queries and transformations.

Value set membership

The member_of function can be used to test the membership of a code within a FHIR value set. This can be used with both explicit value sets (i.e. those that have been pre-defined and loaded into the terminology server) and implicit value sets (e.g. SNOMED CT Expression Constraint Language).

In this example, we take a list of SNOMED CT diagnosis codes and create a new column which shows which are viral infections. We use an ECL expression to define viral infection as a disease with a pathological process of "Infectious process", and a causative agent of "Virus".

result = pc.member_of(csv, to_coding(csv.CODE, 'http://snomed.info/sct'),
                      to_ecl_value_set("""
<< 64572001|Disease| : (
  << 370135005|Pathological process| = << 441862004|Infectious process|,
  << 246075003|Causative agent| = << 49872002|Virus|
)
                      """), 'VIRAL_INFECTION')
result.select('CODE', 'DESCRIPTION', 'VIRAL_INFECTION').show()

Results in:

CODE	DESCRIPTION	VIRAL_INFECTION
65363002	Otitis media	false
16114001	Fracture of ankle	false
444814009	Viral sinusitis	true
444814009	Viral sinusitis	true
43878008	Streptococcal sore throat	false

Concept translation

The translate function can be used to translate codes from one code system to another using maps that are known to the terminology server. In this example, we translate our SNOMED CT diagnosis codes into Read CTV3.

result = pc.translate(csv, to_coding(csv.CODE, 'http://snomed.info/sct'),
                      'http://snomed.info/sct/900000000000207008?fhir_cm='
                      '900000000000497000',
                      output_column_name='READ_CODE')
result = result.withColumn('READ_CODE', result.READ_CODE.code)
result.select('CODE', 'DESCRIPTION', 'READ_CODE').show()

Results in:

CODE	DESCRIPTION	READ_CODE
65363002	Otitis media	X00ik
16114001	Fracture of ankle	S34..
444814009	Viral sinusitis	XUjp0
444814009	Viral sinusitis	XUjp0
43878008	Streptococcal sore throat	A340.

Subsumption testing

Subsumption test is a fancy way of saying "is this code equal or a subtype of this other code".

For example, a code representing "ankle fracture" is subsumed by another code representing "fracture". The "fracture" code is more general, and using it with subsumption can help us find other codes representing different subtypes of fracture.

The subsumes function allows us to perform subsumption testing on codes within our data. The order of the left and right operands can be reversed to query whether a code is "subsumed by" another code.

# 232208008 |Ear, nose and throat disorder|
left_coding = Coding('http://snomed.info/sct', '232208008')
right_coding_column = to_coding(csv.CODE, 'http://snomed.info/sct')

result = pc.subsumes(csv, 'IS_ENT',
                     left_coding=left_coding,
                     right_coding_column=right_coding_column)

result.select('CODE', 'DESCRIPTION', 'IS_ENT').show()

Results in:

CODE	DESCRIPTION	IS_ENT
65363002	Otitis media	true
16114001	Fracture of ankle	false
444814009	Viral sinusitis	true

Retrieving properties

Some terminologies contain additional properties that are associated with codes. You can query these properties using the property_of function.

There is also a display function that can be used to retrieve the preferred display term for each code.

# Get the parent codes for each code in the dataset.
parents = csv.withColumn(
    "PARENTS",
    property_of(to_snomed_coding(csv.CODE), "parent", PropertyType.CODE),
)
# Split each parent code into a separate row.
exploded_parents = parents.selectExpr(
    "CODE", "DESCRIPTION", "explode_outer(PARENTS) AS PARENT"
)
# Retrieve the preferred term for each parent code.
with_displays = exploded_parents.withColumn(
    "PARENT_DISPLAY", display(to_snomed_coding(exploded_parents.PARENT))
)

Results in:

CODE	DESCRIPTION	PARENT	PARENT_DISPLAY
65363002	Otitis media	43275000	Otitis
65363002	Otitis media	68996008	Disorder of middle ear
16114001	Fracture of ankle	125603006	Injury of ankle
16114001	Fracture of ankle	46866001	Fracture of lower limb
444814009	Viral sinusitis	36971009	Sinusitis
444814009	Viral sinusitis	281794004	Viral upper respiratory tract infection
444814009	Viral sinusitis	363166002	Infective disorder of head
444814009	Viral sinusitis	36971009	Sinusitis
444814009	Viral sinusitis	281794004	Viral upper respiratory tract infection
444814009	Viral sinusitis	363166002	Infective disorder of head

Retrieving designations

Some terminologies contain additional display terms for codes. These can be used for language translations, synonyms, and more. You can query these terms using the designation function.

# Get the synonyms for each code in the dataset.
synonyms = csv.withColumn(
    "SYNONYMS",
    designation(to_snomed_coding(csv.CODE), Coding.of_snomed("900000000000013009")),
)
# Split each synonyms into a separate row.
exploded_synonyms = synonyms.selectExpr(
    "CODE", "DESCRIPTION", "explode_outer(SYNONYMS) AS SYNONYM"
)

Results in:

CODE	DESCRIPTION	SYNONYM
65363002	Otitis media	OM - Otitis media
16114001	Fracture of ankle	Ankle fracture
16114001	Fracture of ankle	Fracture of distal end of tibia and fibula
444814009	Viral sinusitis (disorder)	NULL
444814009	Viral sinusitis (disorder)	NULL
43878008	Streptococcal sore throat (disorder)	Septic sore throat
43878008	Streptococcal sore throat (disorder)	Strep throat
43878008	Streptococcal sore throat (disorder)	Strept throat
43878008	Streptococcal sore throat (disorder)	Streptococcal angina
43878008	Streptococcal sore throat (disorder)	Streptococcal pharyngitis

Terminology server authentication

Pathling can be configured to connect to a protected terminology server by supplying a set of OAuth2 client credentials and a token endpoint.

Here is an example of how to authenticate to the NHS terminology server:

from pathling import PathlingContext

pc = PathlingContext.create(
    terminology_server_url='https://ontology.nhs.uk/production1/fhir',
    token_endpoint='https://ontology.nhs.uk/authorisation/auth/realms/nhs-digital-terminology/protocol/openid-connect/token',
    client_id='[client ID]',
    client_secret='[client secret]'
)

Installation in Databricks

To make the Pathling library available within notebooks, navigate to the "Compute" section and click on the cluster. Click on the "Libraries" tab, and click "Install new".

Install both the pathling PyPI package, and the au.csiro.pathling:library-api Maven package. Once the cluster is restarted, the libraries should be available for import and use within all notebooks.

By default, Databricks uses Java 8 within its clusters, while Pathling requires Java 17. To enable Java 17 support within your cluster, navigate to Advanced Options > Spark > Environment Variables and add the following:

JNAME=zulu17-ca-amd64

See the Databricks documentation on Libraries for more information.

Spark cluster configuration

If you are running your own Spark cluster, or using a Docker image (such as jupyter/pyspark-notebook) , you will need to configure Pathling as a Spark package.

You can do this by adding the following to your spark-defaults.conf file:

spark.jars.packages au.csiro.pathling:library-api:[some version]

See the Configuration page of the Spark documentation for more information about spark.jars.packages and other related configuration options.

To create a Pathling notebook Docker image, your Dockerfile might look like this:

FROM jupyter/pyspark-notebook

USER root
RUN echo "spark.jars.packages au.csiro.pathling:library-api:[some version]" >> /usr/local/spark/conf/spark-defaults.conf

USER ${NB_UID}

RUN pip install --quiet --no-cache-dir pathling && \
    fix-permissions "${CONDA_DIR}" && \
    fix-permissions "/home/${NB_USER}"

Pathling is copyright © 2018-2023, Commonwealth Scientific and Industrial Research Organisation (CSIRO) ABN 41 687 119 230. Licensed under the Apache License, version 2.0.