Data Extraction - Alto Health

Overview

Alto Health’s data extraction automatically converts unstructured medical referrals into structured, actionable data. Our AI Assistant processes documents in any format and extracts clinical entities with high accuracy.

Example Response

When extraction completes, Alto Health sends a webhook notification with the extracted data. Here’s an example of the webhook payload:

{
  "event": "task.completed",
  "data": {
    "job_id": "62f42de2-2490-4b91-bc61-5dd1cffd6af7",
    "status": "completed",
    "record_id": "692c5c41a0e641f56952aeb9",
    "entities": {
      "entities": [
        {
          "entityName": "patientName",
          "entityValue": "Rosa Brown",
          "boundingBox": {
            "left": 0.1166,
            "top": 0.3684,
            "width": 0.4419,
            "height": 0.0171,
            "page": 1,
            "original_page": 1
          },
          "confidence": "0.987"
        },
        {
          "entityName": "patientGender",
          "entityValue": "female",
          "boundingBox": {
            "left": 0.1168,
            "top": 0.4745,
            "width": 0.7099,
            "height": 0.0177,
            "page": 1,
            "original_page": 1
          },
          "confidence": "0.870"
        },
        {
          "entityName": "referrer",
          "entityValue": "",
          "boundingBox": {},
          "confidence": "1.000"
        }
      ]
    },
    "error": null,
    "extracted_metadata": null,
    "chunks": [
      {
        "text": "Dear Dr. Montgomery",
        "grounding": [
          {
            "page": 1,
            "boundingBox": {
              "left": 0.119755,
              "top": 0.315407,
              "width": 0.178067,
              "height": 0.016624
            },
            "polygon": [
              {"x": 71.482, "y": 265.572},
              {"x": 177.502, "y": 266.443},
              {"x": 177.394, "y": 279.569},
              {"x": 71.374, "y": 278.698}
            ]
          }
        ],
        "chunk_type": "text",
        "chunk_id": "block_OpAaG8eWJW-97LHR9fPon",
        "rotation_angle": 0.0
      },
      {
        "text": "Re: Referral for Rosa Brown [Date of Birth: 30/01/1952]",
        "grounding": [
          {
            "page": 1,
            "boundingBox": {
              "left": 0.119778,
              "top": 0.346499,
              "width": 0.453744,
              "height": 0.016137
            }
          }
        ],
        "chunk_type": "text",
        "chunk_id": "block_umlvx3IvACuYOIKznNms9",
        "rotation_angle": 0.0
      }
    ],
    "extracted_markdown": "Dear Dr. Montgomery\n\nRe: Referral for Rosa Brown [Date of Birth: 30/01/1952]\n\nI am referring Rosa to your clinic for further evaluation and management.\n\n### 1. Introduction\n\nRosa is a 72 years old female who lives alone and is ECOG 2, and presented in my clinic with redness and swelling in right breast for 4 months\n\n### 2. Clinical Details\n\nRosa presented with redness and swelling in right breast since 4 months..."
  }
}

Response Fields

Field	Type	Description
`event`	string	Event type - `task.completed` when extraction finishes
`data.job_id`	string	Unique identifier for the extraction job
`data.status`	string	Job status - `completed`, `processing`, or `failed`
`data.record_id`	string	Record ID from the document upload
`data.entities`	object	Extracted clinical entities
`data.entities.entities`	array	Array of extracted entity objects
`data.entities.entities[].entityName`	string	Name of the extracted entity (e.g., `patientName`, `patientGender`)
`data.entities.entities[].entityValue`	string	Extracted value for the entity
`data.entities.entities[].boundingBox`	object	Location of the entity in the document (coordinates)
`data.entities.entities[].confidence`	string	Confidence score (0-1) - higher is more confident
`data.chunks`	array	Document chunks with text and location information
`data.chunks[].text`	string	Extracted text from the chunk
`data.chunks[].grounding`	array	Location information (page, bounding box, polygon)
`data.chunks[].chunk_type`	string	Type of chunk - `text`, `section_heading`, etc.
`data.extracted_markdown`	string	Full document content in markdown format
`data.error`	string/null	Error message if extraction failed, otherwise null

Confidence Scores: Values closer to 1.0 indicate higher confidence. A score of 1.000 for an empty value typically means Alto Health is confident the entity is not present in the document.

Bounding Boxes: Each extracted entity includes coordinates showing where it was found in the document. Use these for visual verification or highlighting in your UI.