OpenEvidence Enterprise RBAC

Overview

OpenEvidence delivers AI-powered clinical decision support using peer-reviewed medical literature. Enterprise RBAC controls access to clinical queries, PHI-adjacent data, and research datasets. Clinicians query evidence with full access. Researchers access de-identified datasets and can create study cohorts. Admins manage institutional access, SSO configuration, and compliance settings. HIPAA requires strict audit logging of every clinical query, PHI access event, and data export. Institutional access agreements define which evidence libraries each organization can query.

Role Hierarchy

Permission Check

async function checkClinicalAccess(userId: string, resource: string, accessLevel: string): Promise<boolean> {
  const response = await fetch(`${OE_API}/v1/institutions/${INSTITUTION_ID}/permissions`, {
    headers: { Authorization: `Bearer ${OE_API_TOKEN}`, 'Content-Type': 'application/json' },
  });
  const perms = await response.json();
  const user = perms.members.find((m: any) => m.id === userId);
  if (!user) return false;
  const allowed = ROLE_ACCESS[user.role];
  return allowed?.resources.includes(resource) && allowed.levels.includes(accessLevel);
}

Role Assignment

async function assignInstitutionalRole(email: string, role: string, library: string): Promise<void> {
  await fetch(`${OE_API}/v1/institutions/${INSTITUTION_ID}/members`, {
    method: 'POST',
    headers: { Authorization: `Bearer ${OE_API_TOKEN}`, 'Content-Type': 'application/json' },
    body: JSON.stringify({ email, role, libraryAccess: library, hipaaAcknowledged: true }),
  });
}

async function revokeAccess(email: string): Promise<void> {
  await fetch(`${OE_API}/v1/institutions/${INSTITUTION_ID}/members/${email}`, {
    method: 'DELETE',
    headers: { Authorization: `Bearer ${OE_API_TOKEN}` },
  });
}

Audit Logging

interface OpenEvidenceAuditEntry {
  timestamp: string; userId: string; role: string;
  action: 'clinical_qu

OpenEvidence Hello World

Overview

Minimal working examples demonstrating core OpenEvidence API functionality.

Instructions

Step 1: Clinical Query

const result = await client.query({
  question: 'What is the recommended treatment for acute migraine in adults?',
  context: 'emergency_department',
  evidence_level: 'high',  // Filter by evidence quality
  max_citations: 10
});

console.log('Answer:', result.answer);
console.log('Confidence:', result.confidence);
result.citations.forEach(c =>
  console.log(`  [${c.journal}] ${c.title} (${c.year}) — ${c.evidence_level}`)
);

Step 2: Drug Interaction Check

const interactions = await client.interactions.check({
  medications: ['metformin', 'lisinopril', 'atorvastatin'],
  patient_context: { age: 65, conditions: ['diabetes', 'hypertension'] }
});

interactions.forEach(i =>
  console.log(`${i.drug1} + ${i.drug2}: ${i.severity} — ${i.description}`)
);

Step 3: Guideline Lookup

const guidelines = await client.guidelines.search({
  condition: 'hypertension',
  source: ['ACC/AHA', 'ESC'],
  year_min: 2023
});
guidelines.forEach(g =>
  console.log(`${g.source}: ${g.title} (${g.year})`)
);

Error Handling

Resources

• OpenEvidence API Docs

Next Steps

See openevidence-local-dev-loop.

OpenEvidence Migration Deep Dive

Migration Strategies

1. Parallel run: Run old and new systems simultaneously
2. Strangler fig: Gradually route traffic to OpenEvidence
3. Big bang: Switch all at once (risky)

Migration Checklist

• [ ] API mapping documented
• [ ] Data migration plan
• [ ] Rollback procedure
• [ ] Performance baseline
• [ ] Team training complete

Resources

• OpenEvidence Migration Guide

Next Steps

Start with openevidence-install-auth.

OpenEvidence Observability

Overview

OpenEvidence delivers clinical evidence queries where response accuracy and freshness have direct patient safety implications. Monitor query response times to ensure clinicians get timely answers, track evidence freshness to catch stale citations, and audit every query for compliance. Observability must also verify citation accuracy and maintain complete audit logs for regulatory requirements (HIPAA, clinical decision support standards).

Key Metrics

Instrumentation

async function trackClinicalQuery(queryType: string, fn: () => Promise<any>) {
  const start = Date.now();
  const traceId = crypto.randomUUID();
  try {
    const result = await fn();
    metrics.histogram('openevidence.query.latency', Date.now() - start, { queryType });
    metrics.increment('openevidence.query.total', { queryType });
    auditLog.record({ traceId, queryType, status: 'ok', latency: Date.now() - start });
    return result;
  } catch (err) {
    metrics.increment('openevidence.query.errors', { queryType, error: err.code });
    auditLog.record({ traceId, queryType, status: 'error', error: err.message });
    throw err;
  }
}

Health Check Dashboard

async function openEvidenceHealth(): Promise<Record<string, string>> {
  const latencyP95 = await metrics.query('openevidence.query.latency', 'p95', '5m');
  const errorRate = await metrics.query('openevidence.query.error_rate', 'avg', '5m');
  const freshness = await openEvAdmin.getMedianEvidenceAge();
  return {
    query_latency: latencyP95 < 3000 ? 'healthy' : 'slow',
    error_rate: errorRate < 0.005 ? 'healthy' : 'degraded',
    evidence_freshness: freshness < 7 ? 'healthy' : 'stale',
  };
}

Alerting Rules

const alerts = [
  { metric: 'openevidence.query.latency_p95', condition: '> 8s', window: '10m', severity: 'warning' },
  { metric: 'openevidence.query.error_rate', condition: '> 0.02',

OpenEvidence Performance Tuning

Overview

OpenEvidence's clinical API handles evidence query response times, citation batch retrieval, and complex multi-condition query optimization. Clinical evidence queries can take 2-5 seconds as the system searches across thousands of medical studies and synthesizes responses. Citation batch retrieval for systematic reviews generates heavy load when fetching 50-200 references per query. Caching evidence responses, batching citation fetches, and optimizing query specificity reduces clinician wait times by 50-70% and keeps complex queries within acceptable latency bounds.

Caching Strategy

const cache = new Map<string, { data: any; expiry: number }>();
const TTL = { evidence: 1_800_000, citations: 3_600_000, queries: 300_000 };

async function cached(key: string, ttlKey: keyof typeof TTL, fn: () => Promise<any>) {
  const entry = cache.get(key);
  if (entry && entry.expiry > Date.now()) return entry.data;
  const data = await fn();
  cache.set(key, { data, expiry: Date.now() + TTL[ttlKey] });
  return data;
}
// Citations are stable (1hr). Evidence summaries update with new studies (30 min).

Batch Operations

async function fetchCitationsBatch(client: any, citationIds: string[], batchSize = 25) {
  const results = [];
  for (let i = 0; i < citationIds.length; i += batchSize) {
    const batch = citationIds.slice(i, i + batchSize);
    const res = await Promise.all(batch.map(id => client.getCitation(id)));
    results.push(...res);
    if (i + batchSize < citationIds.length) await new Promise(r => setTimeout(r, 500));
  }
  return results;
}

Connection Pooling

import { Agent } from 'https';
const agent = new Agent({ keepAlive: true, maxSockets: 6, maxFreeSockets: 3, timeout: 30_000 });
// Moderate socket count — evidence queries are sequential, citations parallel

Rate Limit Management

async function withRateLimit(fn: () => Promise<any>): Promise<any> {
  try { return await fn(); }
  catch (err: any) {
    if (err.status === 429) {
      const retryMs = parseInt(err.headers?.['retry-after'] || '10') * 1000;
      await new Promise(r => setTimeout(r, retryMs));
      return fn();
    }
    throw err;
  }
}

Monitoring

const metrics = { queries: 0, citationFetches: 0, cacheHits: 0, avgLatencyMs: 0, errors: 0 };
function track(op: 'query' | 'citation', startMs: number, cached: boolean) {
  metrics[op === 'query' ? 'queries' : 'citationFetches']++;
  const lat = Date.now() - startMs;
  metrics.avgLatencyMs = (metrics.avgLatencyMs * (metrics.queries - 1) + lat) / metrics.queries;
  if (cached) metrics.cac

OpenEvidence Production Checklist

Overview

OpenEvidence provides clinical decision support backed by peer-reviewed medical literature. A production integration handles Protected Health Information (PHI) subject to HIPAA, serves evidence-based answers where accuracy directly impacts patient outcomes, and must maintain complete audit trails for regulatory review. Misconfigurations can expose PHI in logs, serve stale clinical guidance, or fail compliance audits that shut down your integration entirely. This checklist enforces HIPAA-grade security, citation verification, and the SLA discipline required for healthcare-adjacent systems.

Prerequisites

• Production OpenEvidence API credentials (not trial/sandbox keys)
• Secrets manager configured (Vault, AWS Secrets Manager, or GCP Secret Manager)
• HIPAA-compliant monitoring stack (no PHI in log aggregators without BAA)
• Business Associate Agreement (BAA) executed with OpenEvidence
• Compliance officer sign-off on data flow architecture

Authentication & Secrets

• [ ] API keys stored in vault/secrets manager (never in code, env files, or CI logs)
• [ ] Key rotation schedule configured (every 90 days, with zero-downtime swap)
• [ ] Separate keys for staging vs production (staging keys cannot reach production data)
• [ ] Service account permissions scoped to query-only (no admin endpoints)
• [ ] API key exposure detection automated (scan logs/repos for leaked credentials)

API Integration

• [ ] Base URL points to production endpoint (not sandbox/staging)
• [ ] Request timeout set to 15 seconds for clinical queries (evidence synthesis is compute-heavy)
• [ ] Response time SLA monitored: p95 < 3 seconds per contractual agreement
• [ ] Pagination implemented for evidence result sets (token-based cursor)
• [ ] Clinical query payloads never include patient identifiers (de-identify before sending)
• [ ] Citation URLs in responses validated before displaying to clinicians
• [ ] Fallback behavior defined when evidence confidence score is below threshold (0.7)

Error Handling & Resilience

• [ ] Circuit breaker configured for OpenEvidence API calls (open after 3 consecutive failures)
• [ ] Retry logic with exponential backoff for 429 (rate limit) and 5xx responses
• [ ] Clinical query failures surface explicit "no evidence available" (never silent failure)
• [ ] Timeout responses distinguished from empty-result responses in UI
• [ ] Stale cache clearly labeled with retrieval timestamp when serving cached evidence
• [ ] Degraded mode displays disclaimer: "Results may not reflect latest evidence"
• [ ] All API errors logged with correlation ID (without PHI in the log entry)

Monitoring & Alerting

• [ ]

OpenEvidence Rate Limits

Overview

OpenEvidence's clinical decision support API enforces strict rate limits to ensure reliable evidence retrieval for healthcare applications. Clinical query endpoints are throttled per API key, with lower limits on evidence synthesis calls that involve AI-powered literature analysis. In clinical settings, rate limiting directly impacts patient care workflows, so implementations must prioritize graceful degradation over retry storms. Batch research queries during off-peak hours and cache evidence summaries aggressively since medical literature changes infrequently.

Rate Limit Reference

Rate Limiter Implementation

class OpenEvidenceRateLimiter {
  private tokens: number;
  private lastRefill: number;
  private readonly max: number;
  private readonly refillRate: number;
  private queue: Array<{ resolve: () => void }> = [];

  constructor(maxPerMinute: number) {
    this.max = maxPerMinute;
    this.tokens = maxPerMinute;
    this.lastRefill = Date.now();
    this.refillRate = maxPerMinute / 60_000;
  }

  async acquire(): Promise<void> {
    this.refill();
    if (this.tokens >= 1) { this.tokens -= 1; return; }
    return new Promise(resolve => this.queue.push({ resolve }));
  }

  private refill() {
    const now = Date.now();
    this.tokens = Math.min(this.max, this.tokens + (now - this.lastRefill) * this.refillRate);
    this.lastRefill = now;
    while (this.tokens >= 1 && this.queue.length) {
      this.tokens -= 1;
      this.queue.shift()!.resolve();
    }
  }
}

const queryLimiter = new OpenEvidenceRateLimiter(25);
const synthesisLimiter = new OpenEvidenceRateLimiter(8);

Retry Strategy

async function openEvidenceRetry<T>(
  limiter: OpenEvidenceRateLimiter, fn: () => Promise<Response>, maxRetries = 3
): Promise<T> {
  for (let attempt = 0; attempt <= maxRetries; attempt++) {
    await limiter.acquire();
    const res = await fn();
    if (res.ok) return res.json();
    if (res.status === 429) {
      const retryAfter = parseInt(res.headers.get("Retry-After") || "30", 10);
      const jitter = Math.random() * 2000;
      await new Promise(r => setTimeout(r, retryAfter * 1000 + jitter));
      continue;
    }
    if (res

OpenEvidence Reference Architecture

Overview

Production architecture for clinical decision support integrations with OpenEvidence. Designed for healthcare platforms needing evidence-based query processing, citation-backed clinical answers, and full audit logging for regulatory compliance. Key design drivers: HIPAA-compliant data handling, deterministic citation pipelines for clinical accuracy, query audit trails for malpractice risk mitigation, and sub-second response times for point-of-care workflows where clinicians need answers during patient encounters.

Architecture Diagram

Clinician UI ──→ API Gateway (auth + HIPAA) ──→ Query Service ──→ OpenEvidence API
                        ↓                            ↓             /query
                   Audit Logger ──→ Audit DB    Cache (Redis)      /citations
                        ↓                            ↓
                   Analytics ──→ Usage Dashboard  Citation Store ──→ Evidence DB

Service Layer

class ClinicalQueryService {
  constructor(private oe: OpenEvidenceClient, private cache: CacheLayer, private audit: AuditLogger) {}

  async queryEvidence(query: ClinicalQuery): Promise<EvidenceResponse> {
    await this.audit.log({ type: 'query_submitted', clinicianId: query.clinicianId, queryText: query.text, timestamp: new Date() });
    const cacheKey = `evidence:${this.hashQuery(query.text)}`;
    const cached = await this.cache.get(cacheKey);
    if (cached) { await this.audit.log({ type: 'cache_hit', cacheKey }); return cached; }
    const response = await this.oe.query(query.text, { specialty: query.specialty });
    await this.storeCitations(response.citations);
    await this.cache.set(cacheKey, response, CACHE_CONFIG.evidence.ttl);
    await this.audit.log({ type: 'query_completed', citationCount: response.citations.length });
    return response;
  }

  async getCitationChain(citationId: string): Promise<Citation[]> {
    return this.evidenceDb.getCitationWithReferences(citationId);
  }
}

Caching Strategy

const CACHE_CONFIG = {
  evidence:   { ttl: 86400, prefix: 'evidence' },  // 24 hr — clinical evidence changes slowly
  citations:  { ttl: 604800, prefix: 'cite' },     // 7 days — published citations are stable
  queryHist:  { ttl: 3600, prefix: 'qhist' },      // 1 hr — recent query dedup for same clinician
  guidelines: { ttl: 43200, prefix: 'guide' },      // 12 hr — clinical guidelines update infrequently
  audit:      { ttl: 0, prefix: 'audit' },          // never cached — every audit entry must persist
};
// New guideline publication events invalidate evidence cache for affected specialties

Event Pipeline

class ClinicalEventPipeline {
  private queue = new Bull('clinical-events

OpenEvidence SDK Patterns

Overview

Production-ready patterns for the OpenEvidence clinical evidence API. OpenEvidence provides REST endpoints for querying medical literature, retrieving clinical guidelines, and generating evidence-based recommendations. The API authenticates via OPENEVIDENCE_API_KEY and returns structured clinical data with citation provenance. A singleton client enforces consistent auth, handles healthcare-specific errors, and preserves citation chains for audit compliance.

Singleton Client

const OE_BASE = 'https://api.openevidence.com/v1';
let _client: OpenEvidenceClient | null = null;
export function getClient(): OpenEvidenceClient {
  if (!_client) {
    const apiKey = process.env.OPENEVIDENCE_API_KEY;
    if (!apiKey) throw new Error('OPENEVIDENCE_API_KEY must be set — get it from openevidence.com/developer');
    _client = new OpenEvidenceClient(apiKey);
  }
  return _client;
}
class OpenEvidenceClient {
  private headers: Record<string, string>;
  constructor(apiKey: string) { this.headers = { 'Authorization': `Bearer ${apiKey}`, 'Content-Type': 'application/json' }; }
  async query(question: string, opts: { specialty?: string; maxResults?: number } = {}): Promise<EvidenceResponse> {
    const res = await fetch(`${OE_BASE}/query`, { method: 'POST', headers: this.headers,
      body: JSON.stringify({ question, specialty: opts.specialty, max_results: opts.maxResults ?? 10 }) });
    if (!res.ok) throw new OEError(res.status, await res.text()); return res.json();
  }
  async getCitation(citationId: string): Promise<Citation> {
    const res = await fetch(`${OE_BASE}/citations/${citationId}`, { headers: this.headers });
    if (!res.ok) throw new OEError(res.status, await res.text()); return res.json();
  }
}

Error Wrapper

export class OEError extends Error {
  constructor(public status: number, message: string) { super(message); this.name = 'OEError'; }
}
export async function safeCall<T>(operation: string, fn: () => Promise<T>): Promise<T> {
  try { return await fn(); }
  catch (err: any) {
    if (err instanceof OEError && err.status === 429) { await new Promise(r => setTimeout(r, 5000)); return fn(); }
    if (err instanceof OEError && err.status === 401) throw new OEError(401, 'Invalid OPENEVIDENCE_API_KEY');
    if (err instanceof OEError && err.status === 422) throw new OEError(422, `${operation}: query rejected — check terminology`);
    throw new OEError(err.status ?? 0, `${operation} failed: ${err.message}`);
  }
}

Request Builder

class OEQueryBuilder {
  private body: Record<string, any> = {};
  question(q: string) { this.body.question = q; return this; }
  specialty(s: string) { this.bod

OpenEvidence Security Basics

Overview

OpenEvidence provides AI-powered clinical evidence synthesis that processes protected health information (PHI), patient queries, and medical literature references. Integrations must comply with HIPAA requirements for PHI handling, audit logging, and access controls. A breach exposes patient health questions, clinical recommendations, and potentially identifiable medical conditions. Every API interaction must be treated as a HIPAA-regulated transaction.

API Key Management

function createOpenEvidenceClient(): { apiKey: string; baseUrl: string } {
  const apiKey = process.env.OPENEVIDENCE_API_KEY;
  if (!apiKey) {
    throw new Error("Missing OPENEVIDENCE_API_KEY — store in HIPAA-compliant secrets manager");
  }
  // PHI-adjacent access — enforce audit logging on every request
  console.log("OpenEvidence client initialized (key suffix:", apiKey.slice(-4), ")");
  return { apiKey, baseUrl: "https://api.openevidence.com/v1" };
}

Webhook Signature Verification

import crypto from "crypto";
import { Request, Response, NextFunction } from "express";

function verifyOpenEvidenceWebhook(req: Request, res: Response, next: NextFunction): void {
  const signature = req.headers["x-openevidence-signature"] as string;
  const secret = process.env.OPENEVIDENCE_WEBHOOK_SECRET!;
  const expected = crypto.createHmac("sha256", secret).update(req.body).digest("hex");
  if (!signature || !crypto.timingSafeEqual(Buffer.from(signature), Buffer.from(expected))) {
    res.status(401).send("Invalid signature");
    return;
  }
  next();
}

Input Validation

import { z } from "zod";

const ClinicalQuerySchema = z.object({
  query_id: z.string().uuid(),
  clinical_question: z.string().min(10).max(2000),
  specialty: z.enum(["oncology", "cardiology", "neurology", "general", "pediatrics", "emergency"]).optional(),
  evidence_level: z.enum(["systematic_review", "rct", "cohort", "case_report", "expert_opinion"]).optional(),
  include_guidelines: z.boolean().default(true),
});

function validateClinicalQuery(data: unknown) {
  return ClinicalQuerySchema.parse(data);
}

Data Protection

const OPENEVIDENCE_PHI_FIELDS = ["patient_name", "date_of_birth", "mrn", "clinical_question", "diagnosis", "medication_list"];

function redactOpenEvidenceLog(record: Record<string, unknown>): Record<string, unknown> {
  const redacted = { ...record };
  for (const field of OPENEVIDENCE_PHI_FIELDS) {
    if (field in redacted) redacted[field]

OpenEvidence Upgrade & Migration

Overview

OpenEvidence is a clinical AI platform that provides evidence-based medical answers and clinical decision support. The API exposes endpoints for clinical queries, evidence retrieval, and citation management. Tracking API changes is critical because OpenEvidence evolves its evidence grading schema, citation format, and clinical query response structure — and breaking changes in a healthcare context can surface outdated medical evidence, alter confidence scores, or remove critical safety disclaimers that downstream clinical applications depend on.

Version Detection

const OPENEVIDENCE_BASE = "https://api.openevidence.com/v1";

async function detectOpenEvidenceVersion(apiKey: string): Promise<void> {
  const res = await fetch(`${OPENEVIDENCE_BASE}/status`, {
    headers: { Authorization: `Bearer ${apiKey}`, "Content-Type": "application/json" },
  });
  const version = res.headers.get("x-openevidence-api-version") ?? "v1";
  console.log(`OpenEvidence API version: ${version}`);

  // Test clinical query response schema
  const queryRes = await fetch(`${OPENEVIDENCE_BASE}/query`, {
    method: "POST",
    headers: { Authorization: `Bearer ${apiKey}`, "Content-Type": "application/json" },
    body: JSON.stringify({ question: "What is the recommended treatment for hypertension?", max_citations: 1 }),
  });
  const data = await queryRes.json();
  const hasStructuredCitations = data.citations?.[0]?.evidence_grade !== undefined;
  console.log(`Structured citations: ${hasStructuredCitations}`);
  const hasDisclaimers = data.disclaimers !== undefined;
  console.log(`Disclaimers field present: ${hasDisclaimers}`);
}

Migration Checklist

• [ ] Review OpenEvidence release notes for API schema changes
• [ ] Verify clinical query response structure (answer, citations, confidence)
• [ ] Check evidence grading scale — letter grades vs. numeric scores may change
• [ ] Validate citation format (PMID references, DOI links, journal metadata)
• [ ] Test disclaimer and safety warning fields in query responses
• [ ] Update clinical specialty filters if taxonomy was expanded
• [ ] Check rate limits for clinical query endpoints (may differ by plan tier)
• [ ] Verify streaming response format if using real-time query mode
• [ ] Update evidence date range filters if temporal query syntax changed
• [ ] Run clinical validation suite against known question-answer pairs

Schema Migration

// OpenEvidence query response: flat answer → structured evidence with grading
interface OldQueryResponse {
  answer: string;
  citations: Array<{ title: string; url: string; source: string }>;
  confidence: number;
}

interface NewQueryRespo

OpenEvidence Webhooks & Events

Overview

OpenEvidence delivers webhook notifications for clinical evidence retrieval workflows. Subscribe to events when queries complete, evidence bases are updated, new citations are added, or clinical reviews are flagged. Use these webhooks to keep clinical decision support systems current and trigger downstream audit workflows in real time.

Webhook Registration

const response = await fetch("https://api.openevidence.com/v1/webhooks", {
  method: "POST",
  headers: {
    "Authorization": `Bearer ${process.env.OPENEVIDENCE_API_KEY}`,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    url: "https://yourapp.com/webhooks/openevidence",
    events: ["query.completed", "evidence.updated", "citation.added", "review.flagged"],
    secret: process.env.OPENEVIDENCE_WEBHOOK_SECRET,
  }),
});

Signature Verification

import crypto from "crypto";
import { Request, Response, NextFunction } from "express";

function verifyOpenEvidenceSignature(req: Request, res: Response, next: NextFunction) {
  const signature = req.headers["x-openevidence-signature"] as string;
  const expected = crypto.createHmac("sha256", process.env.OPENEVIDENCE_WEBHOOK_SECRET!)
    .update(req.body).digest("hex");
  if (!crypto.timingSafeEqual(Buffer.from(signature), Buffer.from(expected))) {
    return res.status(401).json({ error: "Invalid signature" });
  }
  next();
}

Event Handler

import express from "express";
const app = express();

app.post("/webhooks/openevidence", express.raw({ type: "application/json" }), verifyOpenEvidenceSignature, (req, res) => {
  const event = JSON.parse(req.body.toString());
  res.status(200).json({ received: true });

  switch (event.type) {
    case "query.completed":
      deliverResults(event.data.query_id, event.data.evidence_count); break;
    case "evidence.updated":
      refreshClinicalCache(event.data.topic_id, event.data.revision); break;
    case "citation.added":
      indexCitation(event.data.citation_id, event.data.pubmed_id); break;
    case "review.flagged":
      escalateReview(event.data.review_id, event.data.flag_reason); break;
  }
});

Event Types