FHIR Client

A modern, lightweight FHIR client for TypeScript/JavaScript with support for R3, R4, and R5.

Features

• Multi-version support: R3, R4, R5.
• TypeScript: Fully typed.
• Simple API: read, search, create, update, delete.
• Pagination: Automatically fetch all pages of search results.
• Caching: Built-in configurable LRU caching.
• Auth: Basic Auth support (extensible).

Installation

npm install @outburn/fhir-client

Usage

Initialization

import { FhirClient } from '@outburn/fhir-client';

const client = new FhirClient({
  baseUrl: 'https://hapi.fhir.org/baseR4',
  fhirVersion: 'R4',
  timeout: 30000, // Optional: Request timeout in ms (default: 30000)
  auth: {
    username: 'user',
    password: 'password',
  },
  cache: {
    enable: true,
    ttl: 60000, // 1 minute
  },
});

Read

const patient = await client.read('Patient', '123');
console.log(patient);

// Bypass cache and fetch fresh data from server
const freshPatient = await client.read('Patient', '123', { noCache: true });

Get Capabilities

// Fetch server capabilities (CapabilityStatement)
const capabilities = await client.getCapabilities();
console.log(capabilities.fhirVersion);
console.log(capabilities.format);

Search

// Simple search
const bundle = await client.search('Patient', { name: 'John' });

// Search with query string
const bundle2 = await client.search('Patient?active=true', { name: 'John' });

// Fetch all pages (with safeguard to prevent OOM)
const allPatients = await client.search('Patient', { active: true }, { fetchAll: true });

// Override the default max limit for this search
const manyPatients = await client.search('Patient', {}, { fetchAll: true, maxResults: 50000 });

// Search via POST with form-urlencoded (useful for long query strings or server requirements)
const postResults = await client.search('Patient', { name: 'John' }, { asPost: true });

// Bypass cache and fetch fresh data from server
const freshData = await client.search('Patient', { _id: '123' }, { noCache: true });

// Combine options
const freshAllPatients = await client.search(
  'Patient',
  { active: true },
  { fetchAll: true, noCache: true, asPost: true }
);

Create

const newPatient = await client.create('Patient', {
  resourceType: 'Patient',
  name: [{ family: 'Doe', given: ['John'] }],
});

Process Transaction

// Process a transaction bundle
const transactionBundle = {
  resourceType: 'Bundle',
  type: 'transaction',
  entry: [
    {
      request: {
        method: 'POST',
        url: 'Patient',
      },
      resource: {
        resourceType: 'Patient',
        name: [{ family: 'Doe', given: ['John'] }],
      },
    },
    {
      request: {
        method: 'PUT',
        url: 'Patient/123',
      },
      resource: {
        resourceType: 'Patient',
        id: '123',
        active: false,
      },
    },
  ],
};

const response = await client.processTransaction(transactionBundle);
console.log(response.type); // 'transaction-response'

Process Batch

// Process a batch bundle
const batchBundle = {
  resourceType: 'Bundle',
  type: 'batch',
  entry: [
    {
      request: {
        method: 'GET',
        url: 'Patient/123',
      },
    },
    {
      request: {
        method: 'GET',
        url: 'Patient/456',
      },
    },
  ],
};

const response = await client.processBatch(batchBundle);
console.log(response.type); // 'batch-response'

Update

// Read, modify, and update
const patient = await client.read('Patient', '123');
patient.active = false;
const updatedPatient = await client.update(patient);

// Or update directly with a resource object (must include resourceType and id)
const updatedPatient = await client.update({
  resourceType: 'Patient',
  id: '123',
  active: false,
});

Delete

await client.delete('Patient', '123');

Resolve Single Resource

These methods help you find and work with single resources using search criteria. They automatically filter out informational entries (like OperationOutcome with search.mode !== 'match') and only count actual resource matches.

toLiteral

Searches for a resource and returns its literal reference (resourceType/id). Throws an error if zero or multiple matches are found.

// Find patient by identifier and get their reference
const ref = await client.toLiteral('Patient', { identifier: 'http://system|12345' });
console.log(ref); // "Patient/abc-123"

// Even if the server includes an OperationOutcome in the Bundle, it will be ignored
// and only the actual Patient match will be counted

resourceId

Same as toLiteral but returns only the ID part.

const id = await client.resourceId('Patient', { identifier: 'http://system|12345' });
console.log(id); // "abc-123"

resolve

Hybrid method that can:

• Read a resource using a literal reference: resolve('Patient/123')
• Search for a single resource and return it: resolve('Patient', { identifier: '...' })

Throws an error if search returns zero or multiple matches.

// Using literal reference (equivalent to read)
const patient1 = await client.resolve('Patient/123');

// Using search criteria (must return exactly one match)
const patient2 = await client.resolve('Patient', { identifier: 'http://system|12345' });

// With options
const patient3 = await client.resolve('Patient', { name: 'Doe' }, { noCache: true });

Search Options

The search method accepts an optional third parameter with the following options:

fetchAll

Automatically fetches all pages of results by following next links in the Bundle. Returns an array of resources instead of a Bundle.

const allPatients = await client.search('Patient', { active: true }, { fetchAll: true });
// Returns: Patient[] instead of Bundle

maxResults

Maximum number of resources to fetch when using fetchAll. Overrides the client-level maxFetchAllResults config for this specific search.

const patients = await client.search('Patient', {}, { fetchAll: true, maxResults: 5000 });

asPost

Use HTTP POST with application/x-www-form-urlencoded instead of GET. Useful when:

• Query strings are too long for GET requests
• Server requires POST for search operations
• Working with servers that have URL length limitations

const results = await client.search('Patient', { name: 'John' }, { asPost: true });
// POSTs to: Patient/_search with form data

noCache

Bypass the cache and fetch fresh data from the server, even if a cached response exists.

const freshData = await client.search('Patient', { _id: '123' }, { noCache: true });

All options can be combined as needed.

Configuration

Timeout

All HTTP requests have a configurable timeout to prevent indefinite waiting:

const client = new FhirClient({
  baseUrl: 'https://hapi.fhir.org/baseR4',
  fhirVersion: 'R4',
  timeout: 30000, // Default: 30000ms (30 seconds)
});

If a request takes longer than the specified timeout, it will be aborted and throw an error. You can customize this value based on your server's expected response times.

Fetch All Results Limit

To prevent out-of-memory (OOM) errors when using fetchAll: true, the client enforces a maximum limit on the number of resources that can be fetched:

const client = new FhirClient({
  baseUrl: 'https://hapi.fhir.org/baseR4',
  fhirVersion: 'R4',
  maxFetchAllResults: 10000, // Default: 10000 resources
});

When the limit is exceeded, an error is thrown. You can:

• Increase maxFetchAllResults in the client configuration (applies to all searches)
• Override the limit per-search using the maxResults option (see Search examples above)
• Use regular pagination instead of fetchAll for very large result sets

Caching

The FHIR client includes built-in caching using an LRU (Least Recently Used) cache. Caching significantly improves performance by storing responses from GET requests and reusing them for identical subsequent requests.

How It Works

• Only GET requests are cached: read(), search(), and getCapabilities() operations are cached. Mutations (create, update, delete) are never cached.
• Cache key: Each request is cached based on the full request configuration (URL, query parameters, headers).
• Automatic eviction: Items are automatically removed when:
  • The TTL (time-to-live) expires
  • The cache reaches its maximum size and needs to make room for new items (LRU policy)

Caching Configuration Options

Disabled (Default)

By default, caching is disabled. You must explicitly enable it:

const client = new FhirClient({
  baseUrl: 'https://hapi.fhir.org/baseR4',
  fhirVersion: 'R4',
  // No cache config = caching disabled
});

Basic Caching (TTL Only)

Enable caching with just a time-to-live. Uses default max size of 100 items:

const client = new FhirClient({
  baseUrl: 'https://hapi.fhir.org/baseR4',
  fhirVersion: 'R4',
  cache: {
    enable: true,
    ttl: 60000, // Items expire after 60 seconds (1 minute)
  },
});

Defaults when enable: true:

• max: 100 items
• ttl: 300000 ms (5 minutes)

Full Configuration (TTL + Max Size)

Configure both time-to-live and maximum cache size:

const client = new FhirClient({
  baseUrl: 'https://hapi.fhir.org/baseR4',
  fhirVersion: 'R4',
  cache: {
    enable: true,
    max: 500,      // Store up to 500 items
    ttl: 120000,   // Items expire after 120 seconds (2 minutes)
  },
});

Minimal TTL Configuration

For very short-lived caches (e.g., during a single workflow):

const client = new FhirClient({
  baseUrl: 'https://hapi.fhir.org/baseR4',
  fhirVersion: 'R4',
  cache: {
    enable: true,
    ttl: 10000, // 10 seconds
    max: 50,    // Small cache for temporary use
  },
});

Long-Lived Cache

For read-heavy applications with relatively stable data:

const client = new FhirClient({
  baseUrl: 'https://hapi.fhir.org/baseR4',
  fhirVersion: 'R4',
  cache: {
    enable: true,
    ttl: 3600000,  // 1 hour
    max: 1000,     // Large cache size
  },
});

Expected Behavior

const client = new FhirClient({
  baseUrl: 'https://hapi.fhir.org/baseR4',
  fhirVersion: 'R4',
  cache: {
    enable: true,
    ttl: 60000,
    max: 100,
  },
});

// First call - hits the server
const patient1 = await client.read('Patient', '123');

// Second call within 60 seconds - returns cached result (no server request)
const patient2 = await client.read('Patient', '123');

// After 60 seconds - cache expired, hits the server again
await sleep(61000);
const patient3 = await client.read('Patient', '123');

// Different resource - hits the server (different cache key)
const patient4 = await client.read('Patient', '456');

// Mutations always hit the server (never cached)
await client.update('Patient', '123', updatedPatient); // Always hits server
await client.create('Patient', newPatient);            // Always hits server
await client.delete('Patient', '789');                 // Always hits server

Testing

The project includes both unit tests and integration tests.

Unit Tests

Run unit tests with coverage:

npm test

Unit tests are located in tests/ and provide comprehensive coverage of the client's functionality using mocked HTTP responses.

Integration Tests

Integration tests run against a real HAPI FHIR R4 server using Docker Compose. They verify the client works correctly with an actual FHIR server.

Prerequisites:

• Docker and Docker Compose installed and running
• Port 8080 available

Run integration tests:

npm run test:integration

Run all tests (unit + integration):

npm run test:all

Integration tests cover:

• Server metadata and capabilities
• CRUD operations on Patient, Encounter, and Observation resources
• Pagination with fetchAll (tests with 250+ resources)
• Bundle operations (transaction and batch)
• Error handling

For more details, see tests/integration/README.md.

Development

Prerequisites

This project uses ESLint and Prettier for code quality. We recommend installing the following VS Code extensions (they will be suggested automatically when you open the project):

• ESLint
• Prettier

Scripts

# Run tests
npm test

# Run integration tests (requires Docker)
npm run test:integration

# Run all tests
npm run test:all

# Lint code
npm run lint

# Lint and auto-fix
npm run lint:fix

# Format code
npm run format

# Build
npm run build

Pre-commit Checklist

Before committing, ensure:

1. npm run lint passes with no errors
2. npm test passes
3. Code is formatted with Prettier

The prepublishOnly script will automatically run linting and build before publishing.

License

MIT
© Outburn Ltd. 2022–2025. All Rights Reserved.

Disclaimer

This project is part of the FUME open-source initiative and intended for use in FHIR tooling and development environments.