HL7 Parser for Elixir

Overview

Health Level 7 (HL7) is a protocol (and an organization) designed to model and transfer health-related data electronically.

This parser has support for the HL7 version 2.x syntax. It was tested using v2.4-compliant data, but it should also work with any v2.x messages. It doesn't support the XML mappings that were created for HL7 v3.x, though.

It also has support for custom segment and composite field definitions though an easy-to-use DSL built on top of Elixir macros. A set of preset default segments and composites are available to be used with the parser.

The parser was designed to make the interaction with HL7 as smooth as possible, but its use requires at least moderate knowledge of the HL7 messaging standards.

Requirements

This application waslast updated and tested using Elixir 1.9.4 (and Erlang 22.2) but there shouldn't be any special dependency that prevents it from working with older or later versions.

There are no runtime dependencies on external projects. The parser will make use of the Logger application included in Elixir to output warnings when reading or writing to fields that are not present in the corresponding segment's definition.

Installation

You can use ex_hl7 in your projects by adding it to your mix.exs dependencies:

def deps do
  [{:ex_hl7, "~> 1.0.0"},
end

Contributing

Only a small subset of the HL7 segments and composite fields are included in the project. You can always roll your own definitions in your project, but if you feel your changes would help others, please fork the repository, add whatever you need and send a pull-request.

Encoding Rules

An HL7 message in its v2.x wire-format is actually a collection of concatenated segments, each terminated by a carriage-return (0x0d) character. Each segment is a collection of fields separated by a custom separator character (| by default). Depending on the type of the field, each field can have multiple optional repetitions (separated by ~ by default), can be made out of multiple components (separated by ^ by default) where each of them can also have subcomponents (separated by & by default).

This structure maps nicely to a k-ary tree. For example, given the following segment:

OBX|1|CE|71020&IMP|1|.61^RUL^ACR~.212^Bronchopneumonia^ACR\r

We could represent it as the following subtree within a message:

segment                            OBX
                                    |
fields        [1]--[2]--------[3]---+------[4]------------[5]
              /     |          |            |               \
             1     "CE"        |           "1"               |
                               |                             |
components                    [1]                [1]--------[2]---------[3]
                               |                 /           |            \
                               |               0.61        "RUL"         "ACR"
                               |              0.212  "Bronchopneumonia"  "ACR"
subcomponents             [1]--+--[2]
                          /         \
                      "71020"      "IMP"

The field on sequence 5 contains two repetitions of a composite field.

Note: the indexes used for the fields adn other items are 1-based because this value is actually the sequence number assigned by HL7 to identify the field.

The input and output of the high level functions used to read or write a message (e.g. HL7.read/2, HL7.write/2) is affected by a boolean argument named trim. This value changes the input and output from the lower level functions of the parser. If set to true, some trailing optional items and separators will be omitted from the decoded or encoded message.

For example, a field that was originally read as:

504599^223344&&IIN&^~

Would be written in the following way when trim is set to true:

504599^223344&&IIN

Both representations are correct, given that HL7 allows trailing items that are empty to be omitted.

Single Value Fields

HL7 supports many types of single value (scalar, non-composite) fields. This parser maps all of them (including those that are identifiers in a table) to a few data types:

• nil: the null value from HL7 ("\"\"").
• :string: text value with no conversion performed on it. If the text contains characters that may overlap any message delimiter, it should be modified following the HL7 escaping rules (see HL7.escape/2 and HL7.unescape/2).
• :integer: integer number
• :float: floating-point number with a dot (.) as decimal point; its text representation can be that of an integer (i.e. with no decimal point).
• :date: date in the YYYYMMDD format that is represented as a Date struct.
• :datetime: date/time in the YYYYMMDD[hhmm[ss]] format represented as a NaiveDateTime struct. If the time is not present, the hour, minutes and seconds will be set to 0.

Composite Fields

HL7 supports many types of composite fields and not all of them are included in this project, so to simplify their use there are some macros in the HL7.Composite.Spec module that help you easily define new ones.

This parser allows you to define composite fields as modules and, given the following definition from the HL7 standard:

2.9.3 CE - coded element

<identifier (ST)> ^ <text (ST)> ^ <name of coding system (IS)> ^ <alternate identifier (ST)> ^ <alternate text (ST)> ^ <name of alternate coding system (IS)>

They can be defined in the following way:

use HL7.Composite.Spec

defmodule HL7.Composite.CE do
  composite do
    component :id,                type: :string
    component :text,              type: :string
    component :coding_system,     type: :string
    component :alt_id,            type: :string
    component :alt_text,          type: :string
    component :alt_coding_system, type: :string
  end
end

Using this information the parser will build an internal specification for the composite that will be used to resolve the fields referencing the composite during compile-time.

Each component has a name represented by an atom with the following properties:

• type: atom corresponding to the data type of the value (see single value fields) or to a composite field's module name (e.g. HL7.Composite.Default.CE).
• default: optional default value; if not defined it will be set to an empty string ("").

Composite fields can also be nested, and you can do it in the following way:

use HL7.Composite.Spec

alias HL7.Composite.CE

defmodule HL7.Composite.CQ do
  composite do
    component :quantity,          type: :integer
    component :units,             type: CE
  end
end

Segments

As with composite fields, not all HL7 segments are provided with the project, so there is also a set of macros in the HL7.Segment.Spec module that help define new segments.

Segments are also exposed as structs and can be defined in this way:

use HL7.Segment.Spec

defmodule HL7.Segment.OBX do
  alias HL7.Composite.CE

  segment "OBX" do
    field :set_id,                          seq:  1, type: :integer, len: 4
    field :value_type,                      seq:  2, type: :string, len: 10
    field :observation_id,                  seq:  3, type: {CE, :id}, len: 14
    field :observation_coding_system,       seq:  3, type: {CE, :coding_system}, len: 8
    field :observation_sub_id,              seq:  4, type: :string, len: 20
    field :observation_value_id,            seq:  5, type: {CE, :id}, len: 14
    field :observation_value_coding_system, seq:  5, type: {CE, :coding_system}, len: 8
    field :observation_status,              seq: 11, type: :string, len: 1
  end
end

This segment will be exposed as the following struct:

defstruct :set_id, :value_type, :observation_id, :observation_coding_system, :observation_sub_id,
          :observation_value_id, :observation_value_coding_system, :observation_status

Each field has a name represented by an atom and has the following properties:

• seq: sequence (1-based index) of the field in the segment.
• type: atom corresponding to the data type of the value (see single value fields) or to a composite field's module name (e.g. HL7.Composite.Default.CE). The composite fields will be resolved during compile-time and their specification will be added to the corresponding field specification that can be accessed through the segment module's spec/0 function.
• len: maximum length of the serialized field.
• default: optional default value; if not defined it will be set to an empty string ("") for all types when creating a new segment.

Note: not all of the fields need to be defined in a segment. Segments can be "sparse" and the fields can be defined in an order that is not their sequence order. This means that if a segment containing an undefined field is parsed, that field will be lost when writing/serializing the segment back to its wire-format.

Messages

A parsed HL7 message is represented as a list of segment structs, so you can use the functions from the Enum and List modules to retrieve data or modify them.

The HL7 module has several functions that can be used with messages. The examples below assume that the following HL7 message is being used:

buffer =
  "MSH|^~\\&|BLAKEMD|EWHIN|MSC|EWHIN|19940110105307||RQA^I08|BLAKEM7898|P|2.4|||NE|AL\r" <>
  "PRD|RP|BLAKE^BEVERLY^^^DR^MD|N. 12828 NEWPORT HIGHWAY^^MEAD^WA^99021| ^^^BLAKEMD&EWHIN^^^^^BLAKE MEDICAL CENTER|BLAKEM7899\r" <>
  "PRD|RT|WSIC||^^^MSC&EWHIN^^^^^WASHINGTON STATE INSURANCE COMPANY\r" <>
  "PID|||402941703^9^M10||BROWN^CARY^JOE||19600309||||||||||||402941703\r" <>
  "IN1|1|PPO|WA02|WSIC (WA State Code)|11223 FOURTH STREET^^MEAD^WA^99021^USA|ANN MILLER|509)333-1234|987654321||||19901101||||BROWN^CARY^JOE|1|19600309|N. 12345 SOME STREET^^MEAD^WA^99021^USA|||||||||||||||||402941703||||||01|M\r" <>
  "DG1|1|I9|569.0|RECTAL POLYP|19940106103500|0\r" <>
  "PR1|1|C4|45378|Colonoscopy|19940110105309|00\r"

You can read/parse a message from a binary in the following way:

{:ok, message} = HL7.read(buffer)

Retrieve a specific repetition of a segment:

alias HL7.Segment.Default.PRD

%PRD{role_id: role_id} = prd = HL7.segment(message, "PRD", 1)
"PRD" = HL7.segment_id(prd)
"RT" = role_id

Insert segments:

alias HL7.Segment.Default.{AUT, PR1}
alias HL7.Composite.Default.{CE, EI}

pr1 = HL7.segment(message, "PR1")
aut = %AUT{plan_id: "PPO", company_id: "WA02",
           effective_date: ~D[1994-01-10],
           expiration_date: ~D[1994-05-10],
           authorization_id: "123456789"}
message = HL7.insert_before(message, "PR1", 0, [pr1, aut])
message = HL7.insert_after(message, "PR1", 1, aut)

Replace segments:

message = HL7.replace(message, "PR1", 0, %PR1{pr1 | set_id: 2})

Delete segments:

message = HL7.delete(message, "PR1", 1)
message = HL7.delete(message, "AUT", 1)

Write a message into the HL7 wire format:

iobuf = HL7.write(message, output_format: :wire, trim: true)

Write a message as text to standard output:

IO.puts(HL7.write(message, output_format: :text, trim: true))

Example

This is a basic example of a pre-authorization request with referral to another provider (RQA^I08) that shows how to use the parser. For more information, please check the rest of the sections above.

defmodule Authorizer do
  alias HL7.Segment.Default.{AUT, MSA, MSH, PID, PRD}
  alias HL7.Composite.Default.{CE, CM_MSH_9, CP, EI, MO}

  def authorize(req) do
    message_type = HL7.segment(req, "MSH").message_type
    authorize(req, message_type.id, message_type.trigger_event)
  end

  def authorize(req, "RQA", "I08") do
    msh = HL7.segment(req, "MSH")
    msa = %MSA{
            ack_code: "AA",
            message_control_id: msh.message_control_id
          }
    msh = %MSH{msh |
            sending_app_id: msh.receiving_app_id,
            sending_facility_id: msh.receiving_facility_id,
            sending_facility_universal_id: msh.receiving_facility_universal_id,
            sending_facility_universal_id_type: msh.receiving_facility_universal_id_type,
            receiving_app_id: msh.sending_app_id,
            receiving_facility_id: msh.sending_facility_id,
            receiving_facility_universal_id: msh.sending_facility_universal_id,
            receiving_facility_universal_id_type: msh.sending_facility_universal_id_type,
            message_datetime: NaiveDateTime.utc_now(),
            # RPA^I08
            message_type_id: "RPA",
            message_control_id: Base.encode32(:crypto.rand_bytes(10)),
            accept_ack_type: "ER",
            app_ack_type: "ER"
          }
    aut = %AUT{
            plan_id: "PPO",
            company_id: "WA02",
            company_name: "WSIC (WA State Code)",
            effective_date: ~D[1994-01-10],
            expiration_date: ~D[1994-05-10],
            authorization_id: "123456789",
            reimbursement_limit: 175.0,
            requested_treatments: 1
          }
    req
    |> HL7.replace("MSH", msh)
    |> HL7.insert_after("MSH", msa)
    |> HL7.insert_after("PR1", 0, aut)
  end

  def patient(%PID{last_name: last_name, first_name: first_name}) do
    "Patient: #{first_name} #{last_name}"
  end
  def patient(_pid) do
    nil
  end

  def practice([dg1, pr1]) do
    """
    Diagnosed with: #{dg1.description}
    Treatment: #{pr1.description}
    """
  end

  def providers(prds), do:
    providers(prds, [])

  def providers([%PRD{} = prd | tail], acc) do
    info = """
    #{role_label(prd.role_id)}:
      ##{prd.first_name} #{prd.last_name}
      #{prd.street}
      #{prd.city}, #{prd.state} #{prd.postal_code}
    """
    providers(tail, [info | acc])
  end
  def providers([_prd | tail], acc) do
    providers(tail, acc)
  end
  def providers([], acc) do
    Enum.reverse(acc)
  end

  def role_label("RP"), do: "By"
  def role_label("RT"), do: "And referred to"
end

import Authorizer

buf =
  "MSH|^~\\&|BLAKEMD|EWHIN|MSC|EWHIN|19940110105307||RQA^I08|BLAKEM7898|P|2.4|||NE|AL\r" <>
  "PRD|RP|BLAKE^BEVERLY^^^DR^MD|N. 12828 NEWPORT HIGHWAY^^MEAD^WA^99021| ^^^BLAKEMD&EWHIN^^^^^BLAKE MEDICAL CENTER|BLAKEM7899\r" <>
  "PRD|RT|WSIC||^^^MSC&EWHIN^^^^^WASHINGTON STATE INSURANCE COMPANY\r" <>
  "PID|||402941703^9^M10||BROWN^CARY^JOE||19600309||||||||||||402941703\r" <>
  "IN1|1|PPO|WA02|WSIC (WA State Code)|11223 FOURTH STREET^^MEAD^WA^99021^USA|ANN MILLER|509)333-1234|987654321||||19901101||||BROWN^CARY^JOE|1|19600309|N. 12345 SOME STREET^^MEAD^WA^99021^USA|||||||||||||||||402941703||||||01|M\r" <>
  "DG1|1|I9|569.0|RECTAL POLYP|19940106103500|0\r" <>
  "PR1|1|C4|45378|Colonoscopy|19940110105309|00\r"

{:ok, req} = HL7.read(buf, input_format: :wire)

# Print authorization request data
req |> HL7.segment("PID") |> patient() |> IO.puts()
req |> HL7.paired_segments(["DG1", "PR1"]) |> practice() |> IO.puts()
req |> Enum.filter(&(HL7.segment_id(&1) === "PRD")) |> providers() |> IO.puts()

# Create an authorized response and print it
req |> authorize() |> HL7.write(output_format: :text, trim: true) |> IO.puts()