Recipe catalog / instruction-readability

Grade how easy patient instructions are to follow

How easy are instructions to follow for a general reader, from dense jargon to plain and stepwise?

You are generating or reviewing after-visit instructions, discharge notes, or medication directions and need to catch wording a patient cannot act on.

Explore this recipe interactively · Source and implementation guide

Use instruction-readability in TypeScript

Install with npm install jev-recipes. Requires Node.js 22.9 or newer and ES modules. Set TYPESAFE_API_KEY in your server environment for live calls, which send input to TypeSafe and use API quota. See the installation guide.

import { instructionReadability } from 'jev-recipes/instruction-readability';

const result = await instructionReadability({
  "instructions": "Administer 500 mg PO BID with food x 10 days. Do not discontinue prematurely. If a dose is missed, take it as soon as possible unless the next dose is imminent. Contact the clinic for any rash or swelling.",
  "audience": "Adult patient with no medical background, reading the after-visit summary at home.",
  "minConfidence": 0.8
});
console.log(result);

Input contract

{
  "input": {
    "$schema": "https://json-schema.org/draft/2020-12/schema",
    "type": "object",
    "properties": {
      "instructions": {
        "type": "string"
      },
      "audience": {
        "type": "string"
      },
      "minConfidence": {
        "type": "number",
        "minimum": 0,
        "maximum": 1
      }
    },
    "required": [
      "instructions"
    ]
  },
  "result": {
    "$schema": "https://json-schema.org/draft/2020-12/schema",
    "type": "object",
    "properties": {
      "model": {
        "type": "string"
      },
      "usage": {
        "type": "object",
        "properties": {
          "input_tokens": {
            "type": "integer",
            "minimum": 0,
            "maximum": 9007199254740991
          },
          "output_tokens": {
            "type": "integer",
            "minimum": 0,
            "maximum": 9007199254740991
          }
        },
        "required": [
          "input_tokens",
          "output_tokens"
        ],
        "additionalProperties": false
      },
      "status": {
        "type": "string",
        "enum": [
          "ready",
          "review"
        ]
      },
      "score": {
        "type": "number",
        "minimum": 0
      },
      "level": {
        "type": "integer",
        "minimum": 0,
        "maximum": 9007199254740991
      },
      "confidence": {
        "type": "number",
        "minimum": 0,
        "maximum": 1
      },
      "probabilities": {
        "type": "object",
        "propertyNames": {
          "type": "string"
        },
        "additionalProperties": {
          "type": "number",
          "minimum": 0,
          "maximum": 1
        }
      },
      "readability": {
        "type": "string",
        "enum": [
          "dense",
          "heavy",
          "mixed",
          "plain",
          "clear"
        ]
      }
    },
    "required": [
      "model",
      "usage",
      "status",
      "score",
      "level",
      "confidence",
      "probabilities",
      "readability"
    ],
    "additionalProperties": false
  }
}

Saved example result

This hand-authored response demonstrates the contract. It is not a model accuracy measurement. Run it without an API key: npx jev-recipes demo instruction-readability.

{
  "model": "demo-fixture",
  "usage": {
    "input_tokens": 0,
    "output_tokens": 0
  },
  "status": "ready",
  "score": 1.14,
  "level": 1,
  "confidence": 0.82,
  "probabilities": {
    "0": 0.04,
    "1": 0.82,
    "2": 0.11,
    "3": 0.02,
    "4": 0.01
  },
  "readability": "heavy"
}

Evaluation evidence

Use the evaluation guide to measure this decision on your own labeled cases.

Limitations

• Grades readability of the wording, not medical accuracy, completeness, or whether the instructions are right for the patient.
• Does not compute a formal reading-grade score; use a readability formula in code if you need a metric.

Related recipes

• audience-fit: Use audience-fit to judge whether the whole text suits a named audience beyond how easy the steps are to follow.
• tone-check: Use tone-check to judge the register and tone of the instructions rather than their readability.