DEV Community

Cover image for Introduction to FHIR - The Standard for data exchange in healthcare
Rose Kamal Love
Rose Kamal Love

Posted on

Introduction to FHIR - The Standard for data exchange in healthcare

I have recently been accepted as an open source contributor for LibreHealthIO under the
Google Summer of Code Program and my project for this summer is to rewrite the
whole UI of the lh-toolkit as an Open Web App
using the pre-existing lh-toolkit-webcomponents.

These web components are built with Polymer.js and follow what is called the FHIR Standard which is the
ultimate standard when it comes to data exchange in healthcare. My UI re-write is supposed to be compliant with this
standard. So, I figured why not write a blog about everything I learn about FHIR Standard
so that people that come after me have a place to look for, and my understanding solidifies while writing this.

FYI: FHIR is pronounced as Fire, weird I know but it is what it is.

Checkout Intro to FHIR

Before FHIR and what led to its creations

In the ancient times (the 2000s) it was very hard to have applications that we take for granted today, IE5 was the explorer that
people used back then. Developers back then had to solve really basic problems to even do things as basic as creating a
web portal that links two hospitals' data together and shows it on a web page.

Getting access to data, designing asynchronous web applications, and reloading the web page when new data comes in,
we're really crazy problems for developers back then. Think of writing a frontend for IE5 that updates with the data,
yeah we do take a lot of things for granted that developers in the early 2000s had to do themselves.

The technology that the developer from the 2000s used as a protocol known as HL7 V2, is a bit hard to understand
but once you are used to it, HL7 V2 really makes it a lot easier to quickly scan the data and tell what's it trying to tell
you. It was very dense so allows for a lot of information in smaller transfers.

This is what a typical HL7 V2 payload looked like:

HL7 V2 Payload

I know this looks hilarious, but isn't building web apps for IE5 hilarious in itself? Even though this looks very
old fashioned (because it is) but it really was a groundbreaking technology because it allowed for web apps to be built
which could dynamic, asynchronously querying a bunch of backend systems using these v2 queries and dynamically updating
itself when new data comes in.

Me thinking about useEffect while writing this 😆

After the HL7 V2 there was a large effort to replace it with something more modern than something that was built in
the 90s. That led to the creation of a protocol known as the HL7 V3.

This here is an example of HL7 V3's XML Payload

HL7 V3 Payload

Even though this was a great effort with some really good ideas in it, it utterly failed to attract any kind of mind
share. The main reason that I found it on the internet is that it was too complex and intense to understand and the make
something useful out of it. So it never really managed to take off as a useful standard.

Basic Technology as a Barrier

Basic concepts of getting data out of EHRs and agreeing on formats across jurisdictions were completely preventing large
scale interop.

A lot of amazing things were done using the HL7 V2 and V1 but at the same time a lot of amazing things were not
done, just because the technology hadn't caught up, it really wasn't usable by modern approaches to build applications,
things like building apps that work on your phone and tablet, or work across slow distributed mobile networks, there are
all kinds of barriers there.

All of this leads to the creation of 🔥 HL7 FHIR.

FHIR was created as a response to these old technologies, as a
fresh attempt to build standards that could meet modern needs for interoperability and most importantly that it
would be picked up by people who needed to build technology to solve healthcare problems so let's dive in.

Introduction 🔥 HL7 FHIR

I still can't get over FHIR being pronounced as fire 😆

Fast Healthcare Interoperablity Resources the H and the I are fundamentally what they wanted to build. So, no question about why they chose those two letters.

They chose the word fast kind of deliberately, and fast in this context doesn't mean that the data exchange in itself is fast. The idea behind it are that it is fast to adopt which translates to, you don't need tonnes of experience, PHDs or anything like that to get started with the FHIR Standard

It is intended to be a standard that you can pick up, quickly learn, and very quickly become productive as well as proficient in. And I gotta say, they have succeeded in making it fast.

The work resources comes from the HTTP spec, the language that they (HTTP) use describes the entire world wide web in terms of addressable resources.

For example, you go to google and type in your query term in the form input, and then you pull down a page with a bunch of search results on it, what you have actually done is you have downloaded a resource, that resource, in this case, is search results.

The HTTP spec describes the entire web in terms of these addressable , fetchable, resources.
FHIR is taking all of that context and applying it to Health Care

Hl7 is a standards development organization and FHIR is a standard. (V2 and CDA are also separate standards)

Just a heads up almost none of the actual work of standards development happens by HL7 the corporate entity, what it brings to the table is a process and a governance framework and they organize events, the standard itself is created by volunteers who are spread entirely around the globe.

The amazing stuff, all of it is free and open source, this is honestly is great.

So, What is FHIR?

FHIR is a free and open source standard for health interoperability, based on modern approaches. It consists of five things

  1. A robust data model for describing health and administrative data

  2. A RESTful API for interacting with the data using either JSON or XML

  3. A set of open source tools to implement and test FHIR applications

  4. A collection of FHIR Servers around the world that you can interact with

  5. A community of implementers working together

FHIR also has its own free and open source implementation of the standards in the Java programming language.
But if you're not a Java developer, there are open source libraries of excellent quality in all kinds of other languages as well.

The FHIR Manifesto

One of the primary aims of FHIR was to not repeat the mistakes of the past, and a lot of that led to the creat of The FHIR Manifesto which is a set of guiding principles that was key in terms of decision-making, in order to create and maintain a standard that can be implemented quickly.

Here are a few points from that manifesto

Implementer Focus

This sounds trivial I know, a, of course, a technical standard is going to target developers and implementers, but a lot of people who contributed to the FHIR standard aren't Computer Scientists or doesn't come from a computer background at all, they are clinicians, doctors or PHDs. The vocabulary that they use can be a bit hard to comprehend for the "technical" folks.

  • The specification is targeted at developers and implementers (not clinicians)

  • Every detail of the specification is tested at connectathon events

When creating a standard it is crucial to test its implementations because it can and does happen that we come up with standards, think that they are the optimal ones and then they turn out to be unworkable this is why connectathon events help a lot in testing the standards before making them normative

Common Scenarios

The HL7 V3 tried to be comprehensive, it was trying to be the standard that can be usable for every use case. So it becomes bloated with a lot of stuff that one person needed, or a small group of people needed.

So, FHIR places a heavy emphasis on the 80:20 rule

  • 80% - Specification covers only what is needed by 80% of systems worldwide

  • 20% - FHIR is easy to extend for your use case

Be Open

FHIR is free to use and free to contribute to

  • Specification is licensed under Creative Commons CC0
  • Many tools are open source (MIT / The APACHE License (Apache License)
  • Many tutorials are free as well

Be Easy to Adopt

  • Being easy to learn is critical to the success of FHIR
  • Many organizations have adopted FHIR in weeks

Parts of FHIR

FHIR: Data Model

As I mentioned earlier the R in FHIR stands for resource, and resource ofc is that http concept of a resource. In the world of FHIR, an individual resource is a data model for some specific purpose.

  • FHIR defines a set of roughly 150 resources
  • These are the building blocks of the specification, example include:

    • Patient (the person who receives healthcare)
    • Encounter (a doctor's appointment or hospital say)
    • Observation (e.g. a device reading, or lab value)
    • Diagnostic Report (e.g. a whole lab or X-Ray report)
    • Medication Prescription (Rx for meds)

Let's look into the Patient resource a little deeper,

Patient Resource Type

This is the structure of the Patient resource type which you can find on the FHIR Website, this website is in itself a highly comprehensive documentation of everything FHIR and even everything HL7

One of the things in this website/documentation that I think is great but may sound trivial to you is that everything that can be explained has a link attached to it, which you can click to learn more about what it is.

For example, what do you think that little sigma symbol in the 2nd grid represents? Well, click on it and find out.

One other thing that I have noticed myself is that the documentation is written in a way that I (a developer) don't have a hard time understanding. It does be following the manifesto XD

Data Type: Primitives

Much like any other programming language, FHIR has defined its own set of data types. These data types are very similar to the ones used in general programming languages so no worries there.

Types Example
string Patient is awake
boolean true
date 2016-02-19
decimal 12.347000
integer 500
base64 rwr39o9h=
dateTime 2015-01-26t15:33:13-05:00
instant 2015-01-26t15:33:13.099-05:00

In the FHIR spec, uris and urls are very widely used hence there is a whole data type for that only, base64 here is used for things like images and all.

There are two types of dates in the FHIR spec, the dateTime represents the human time, which can have variable precision. instant on the other hand stores the system time, with the precision of seconds/milliseconds.

Data Type: General Purpose (composite)

General purpose datatypes are mini structures that are reused in a different part of the models

Here's an example of a general purpose data type: Human Name


FHIR Data Model - Resources

  • Meta Data

    • Resource ID
    • Resource Version
    • Tags
    • Profile
  • Extensiosn

    • Any information that doesn't fit into the 80% (Remember the 80:20 rule?)
  • Narrative

    • Human-readable version of the content (optional but encouraged)
    • CDA has taught them that this is very important
  • Body

    • Raw Structured Data

FHIR Encodings

FHIR has defined as of today 3 encodings, 3 means of serializing data and sending it across the wire,

  • JSON: The most popular with modern internet APIs
  • XML: Still very popular with great tools available
  • RDF/Turtle: Only generally used in some research contexts

A JSON encoding example of the Patient resource

  "resourceType" : "Patient",
  // from Resource: id, meta, implicitRules, and language
  // from DomainResource: text, contained, extension, and modifierExtension
  "identifier" : [{ Identifier }], // An identifier for this patient
  "active" : <boolean>, // Whether this patient's record is in active use
  "name" : [{ HumanName }], // A name associated with the patient
  "telecom" : [{ ContactPoint }], // A contact detail for the individual
  "gender" : "<code>", // male | female | other | unknown
  "birthDate" : "<date>", // The date of birth for the individual
  // deceased[x]: Indicates if the individual is deceased or not. One of these 2:
  "deceasedBoolean" : <boolean>,
  "deceasedDateTime" : "<dateTime>",
  "address" : [{ Address }], // An address for the individual
  "maritalStatus" : { CodeableConcept }, // Marital (civil) status of a patient
  // multipleBirth[x]: Whether patient is part of a multiple birth. One of these 2:
  "multipleBirthBoolean" : <boolean>,
  "multipleBirthInteger" : <integer>,
  "photo" : [{ Attachment }], // Image of the patient
  "contact" : [{ // A contact party (e.g. guardian, partner, friend) for the patient
    "relationship" : [{ CodeableConcept }], // The kind of relationship
    "name" : { HumanName }, // A name associated with the contact person
    "telecom" : [{ ContactPoint }], // A contact detail for the person
    "address" : { Address }, // Address for the contact person
    "gender" : "<code>", // male | female | other | unknown
    "organization" : { Reference(Organization) }, // C? Organization that is associated with the contact
    "period": { Period } // The period during which this contact person or organization is valid to be contacted relating to this patient
  "communication" : [{ // A language which may be used to communicate with the patient about his or her health
    "language" : { CodeableConcept }, // R!  The language which can be used to communicate with the patient about his or her health
    "preferred" : <boolean> // Language preference indicator
  "generalPractitioner" : [{ Reference(Organization|Practitioner|
   PractitionerRole) }], // Patient's nominated primary care provider
  "managingOrganization" : { Reference(Organization) }, // Organization that is the custodian of the patient record
  "link" : [{ // Link to another patient resource that concerns the same actual person
    "other" : { Reference(Patient|RelatedPerson) }, // R!  The other patient or related person resource that the link refers to
    "type" : "<code>" // R!  replaced-by | replaces | refer | seealso
Enter fullscreen mode Exit fullscreen mode

Model Concepts

To work with the understanding we should get a basic understanding of some concepts that are defined by the folks who create FHIR, so that's what we
re gonna do it.

Model Concepts: Indentifiers

Identifies in FHIR have two important parts:

  • The system is a URI indicating the kind of identifier
  • The value is the actual identifier
  "system": "",
  "value": "0123456789"
Enter fullscreen mode Exit fullscreen mode

Model Concepts: Codes

  • In health we often draw values from a set of allowable codes (e.g. standards like LOINC and local code systems)

  • Codes in FHIR have two important parts:

    • The system is a URI indicating the kind of code
    • The code is the actual code
    • The display is a human-readable display name
  "system": "",
  "code": "718-7",
  "display": "Hemoglobin mass/volume in blood"
Enter fullscreen mode Exit fullscreen mode

Model Concepts: Extensions

  • Extensions are the way we include out 20%
  • An extension is really just a simple key-value pair
    • The key is a URL
    • The value is an FHIR datatype or more extensions
  • Extensions can also include child extensions which are also called Child Extensions
  "url": "",
  "valueDateTime": "2020-01-13T10:12:00+08:00"
Enter fullscreen mode Exit fullscreen mode

Resource URLs is a Resource URL, if we look closely the pattern here is that the URL starts with a base URL in this case followed by the type of the resource, Patient here followed by the ID which here is example

REST API: Overview

The FHIR REST API is very expansive and covers many many use cases and features, I will be covering a few of them in this blog post but if you want to learn more you can do that by going to the following links

Here's an example of CRUD for the patient resource

Operation Verb Base URL Type ID
Read GET /Patient/ /123
Create POST /Patient/
Update PUT /Patient/ /123
Delete DELETE /Patient/ /123

The request body in POST/PUT is FHIR resources, in FHIR DELETE doesn't destroy the data, instead it flags the data as deleted

FHIR Searching (the Search API)

One of the very powerful features of the FHIR spec is the search API which allows us to get data using various searchParams. If we're building any applications that are being used for a purpose, it gotta have search functionality, and FHIR provides that.

Verb Base URL Type Search Parameter
Get /Patient? /Patient? name=smith /Patient? birthdate=gt1970

The search parameter can even be empty, which means to the server that we want all it sends the data of all the patients, not all servers support this however because this is a security concern.

That's cool and all but how do we know what all search parameters are available for a particular resource? Well, every resource definition has a table of default search parameters at the bottom of the resource definition.

You can just go to the FHIR website and look for a resource, then at the bottom of the resource definition table, you will find the Search Parameters. There are a couple of columns for every search param table.

The first one is the part that goes in the URL so, baseUrl/Patient?birthDate=2020 here, the birthDate is the name that will be present in the first column of every resource.

The second column is an expression, and an expression is a "FHIRPath" expression determining what is indexed, often times these are very intuitive and one-to-one.

Search parameters have datatypes, which are related to (but NOT the same as) regular datatypes

FHIR Search Datatypes: String

The search param datatypes determine the semantics
Enter fullscreen mode Exit fullscreen mode

FHIR Search Datatypes: Token

The search param index identifier and coded fields|1234|
Enter fullscreen mode Exit fullscreen mode

FHIR Search Datatypes: Date

Date parameters index dates and times

Enter fullscreen mode Exit fullscreen mode


So, that's FHIR for you in a lot more than 100 seconds, I will next be blogging about consuming an FHIR API (HAPI FHIR to be exact) which I will be doing in a typeScript React.js app will be using Axios + ReactQuery for data fetching.

Until then, bye 👋🏻

Top comments (1)

krlz profile image

Oh this is great to see, not much content about Healthcare interoperability, I really would like to see more of your content, thanks for sharing