Open-source developmental screening engine for children 0-36 months. Created by Harsh Songra and Aafreen Ansari.

Install from npm: mychild-engine

This is NOT a diagnostic tool. This software tracks developmental milestones and helps caregivers notice changes early. It cannot diagnose any medical condition, developmental disorder, or disability. If you have concerns about your child's development, please consult a qualified healthcare professional immediately.

This software is provided for informational and educational purposes only. It is not intended to be a substitute for professional medical advice, diagnosis, or treatment. Never disregard professional medical advice or delay seeking it because of something produced by this software.

• This engine does not perform medical diagnosis of any kind
• All severity outputs (precaution, warning, flag) are screening signals, not clinical determinations
• All thresholds are labeled "Ruleset v0.2 (hypothesis-v2-cdc2022-aligned)" and have not been validated through clinical trials
• The question bank is aligned with CDC 2022 revised milestone checklists (Zubler et al., Pediatrics 2022, 75th percentile standard) and founder clinical research, not on copyrighted validated instruments (ASQ-3, M-CHAT-R/F, Denver)
• No output from this engine should be interpreted as medical advice

This engine is designed for:

• Developmental monitoring and surveillance (helping caregivers track milestones over time)
• Triage to screening and referral (suggesting when to talk to a doctor, not what the doctor should do)
• Research and development of screening tools by qualified developers and clinicians
• Educational purposes (understanding how developmental screening logic works)

This engine is NOT designed for:

• Clinical decision-making
• Replacing validated screening instruments (ASQ-3, M-CHAT-R/F)
• Standalone use without professional guidance in clinical settings
• Any use where outputs could be interpreted as medical diagnosis

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND. The authors and contributors are not liable for any decisions made based on the outputs of this engine. See the LICENSE file for full terms.

The engine runs entirely locally. No child data is transmitted to any server. The demo app stores answers in browser localStorage only. If you build on top of this engine, you are responsible for compliance with applicable privacy regulations (COPPA, GDPR, DPDP Act, HIPAA, etc.) in your jurisdiction.

This project does not reproduce or distribute any copyrighted screening instruments. The question bank is derived from publicly available CDC milestone checklists (which are not copyrighted screening tools and should not be used as substitutes for validated instruments). If you intend to integrate validated instruments like ASQ-3 or M-CHAT-R/F into a product built on this engine, you must obtain appropriate licenses from the rights holders.

An evidence-weighted rules engine for developmental milestone screening. It sits in the gap between free CDC checklists (simple checkboxes) and licensed clinical instruments (ASQ-3, M-CHAT-R/F).

Features:

• 131 caregiver-facing questions across 10 age bands (birth to 36 months), aligned with CDC 2022 revised milestones (75th percentile standard)
• 8 developmental domains: Gross Motor, Fine Motor, Receptive Language, Expressive Language, Social-Emotional, Cognitive, Self-Help/Adaptive, Vision/Hearing
• Evidence-weighted escalation: questions carry weight classes (Low/Medium/High/Red-Flag) that actively adjust grace windows — high-weight items escalate faster, low-weight items get more grace
• Domain scoring with sufficiency gates: won't flag "high concern" without enough independent observations; streak calculation sorted by developmental sequence
• Regression detection: automatically flags milestones previously achieved but now reported as "not yet" — a key clinical red flag
• Corrected age for preterm infants: automatic adjustment until 24 months
• Probe library: follow-up clarifiers (P1-P5) that reduce false positives
• Multilingual support (i18n): translation framework with Hindi skeleton; designed for Kannada, Tamil, Telugu expansion
• Text-to-Speech (TTS): Web Speech API integration with Indian language locale mapping; critical for low-literacy populations
• Rule simulator: replay synthetic child timelines against different threshold settings and see alert diffs
• Explainability: every severity output includes plain-English "why" traces

npm install mychild-engine

import { evaluate, computeChildAge, getDueQuestions } from 'mychild-engine';

// Create a child (7 months old)
const child = { dob: new Date('2025-09-01') };

// Get age info
const age = computeChildAge(child.dob);
console.log(`${age.chronologicalMonths.toFixed(1)} months old`);

// Get due questions
const questions = getDueQuestions(child, []);
console.log(`${questions.length} questions due`);

// Evaluate with some answers
const result = evaluate(child, [
  { questionId: questions[0].id, answer: 'achieved', timestamp: new Date() },
  { questionId: questions[1].id, answer: 'not_yet', timestamp: new Date() },
]);

console.log(result.globalStatus.message);
console.log(result.nextActions);

packages/engine/     # mychild-engine - the standalone TypeScript engine
  src/               # Engine source code (13 modules incl. i18n + TTS)
  data/              # Question bank JSON (131 questions) + translations/
  dist/              # Compiled output

apps/demo/           # Next.js 16 + shadcn/ui one-page demo
  src/app/           # App router pages + server actions
  src/components/    # Client components (screening, results, simulator)

docs/                # Reference documents (question bank PDF, threshold worksheets)

Run the full screening evaluation. Returns per-question severity, domain assessments, global status, and recommended next actions.

Compute chronological and corrected age. Corrected age applies for preterm infants (<37 weeks) until 24 months chronological.

Get age-appropriate questions that are due for the child. Filters out achieved milestones and respects re-check intervals.

Run synthetic child timelines through the engine. If a baseline ruleset is provided, generates diffs showing how threshold changes affect alerts.

Get a detailed explanation of why a specific severity was assigned, including input factors, applied rule, and recommended action.

For each answered question, the engine computes severity based on the child's corrected age relative to the milestone's normative age:

• Normal: milestone achieved, or child is within the expected age window
• Precaution: child is beyond the normative age but within the grace period
• Warning: child is beyond the grace period by 1 month
• Flag: child is beyond the grace period by 2+ months

Each of the 8 developmental domains maintains a vector: flag_count, warning_count, precaution_count, streak_missed, critical_milestone_missed, and confidence. Domain status is derived from this vector with an evidence sufficiency gate (requires >= 2 independent observations before escalating beyond "watch").

Questions carry weight classes (L/M/H/RF) that actively adjust grace windows:

• RF (Red Flag): 0 weeks grace — immediate escalation
• H (High): standard grace window (4 weeks infant, 6 weeks toddler)
• M (Medium): standard + 2 weeks extra grace
• L (Low): standard + 4 weeks extra grace

The engine automatically detects milestones previously answered "achieved" that are later reported as "not yet". This achieved→not_yet pattern is escalated to at least "warning" severity regardless of age-based calculation, as developmental regression is a key clinical red flag.

Seven action profiles define response pathways per milestone category:

• AP-STD: standard monitoring (most milestones)
• AP-LANG: language pathway (recommends validated screening at 9/18/30 months)
• AP-MOTOR: motor pathway (clinician discussion for persistent delays)
• AP-SENS: sensory pathway (earlier escalation, hearing/vision checks)
• AP-RF: red-flag pathway (regression triggers urgent referral)
• AP-ADAPT: adaptive/self-help (gentler escalation)
• AP-TOILET: toilet readiness (reassurance, no pressure)

The question bank at packages/engine/data/question-bank.json contains 131 questions aligned with the CDC 2022 revised milestone checklists (Zubler et al., Pediatrics 2022). The 2022 revision shifted milestones from the 50th to the 75th percentile — each listed milestone is one that 75% of children would be expected to achieve by the given age. Each question includes:

• Caregiver-facing wording (plain language, parent-observable)
• Domain tags (which developmental areas it assesses)
• Evidence strength rating (High/Moderate/Low)
• Weight class (L/M/H/RF) for escalation speed
• Probe references (P1-P5 clarifiers to reduce false positives)
• Action profile (response pathway per answer state)

This engine's question bank and scoring logic are grounded in peer-reviewed developmental science:

1. Zubler JM, et al. "Evidence-Informed Milestones for Developmental Surveillance Tools." Pediatrics. 2022;149(3):e2021052138. — The CDC 2022 revised milestones shifted from the 50th to the 75th percentile standard. Each milestone listed is one that 75% of children would be expected to achieve by the given age.
2. Lipkin PH, Macias MM. "Promoting Optimal Development: Identifying Infants and Young Children With Developmental Disorders Through Developmental Surveillance and Screening." Pediatrics. 2020;145(1):e20193449. — AAP policy statement on developmental surveillance and screening.
3. CDC "Learn the Signs. Act Early." milestone checklists (public domain) — the primary source for the question bank's caregiver-facing wording.

The engine includes a built-in verification pipeline that runs 108 synthetic developmental profiles (typical development, speech delay, motor delay, global delay, regression patterns, and borderline/boundary cases) through the full scoring logic and compares outputs against hand-written ground truth labels. This is internal consistency verification, not clinical validation. It demonstrates that the engine behaves predictably and consistently across known developmental trajectories, including edge cases near classification boundaries.

Run it yourself:

npx mychild-engine validate --profiles data/synthetic-profiles.json --format markdown

All thresholds in this engine are hypothesis-level (Ruleset v0.2, CDC 2022 aligned). They have not undergone clinical validation (prospective pilot, sensitivity/specificity against a clinical gold standard, or regulatory review). The internal consistency verification above demonstrates engine behavior, not clinical accuracy.

If you are a clinician or researcher interested in validating these thresholds, please open an issue. Contributions from the clinical community are welcome and encouraged.

The engine includes a translation registry that supports multiple Indian languages:

import { translationRegistry } from 'mychild-engine';
import hiBundle from 'mychild-engine/data/translations/hi.json';

// Register Hindi translations
translationRegistry.register(hiBundle);

// Get a translated question
const q = translationRegistry.getQuestion('rf_01', 'hi');
console.log(q?.text); // क्या आपके बच्चे ने कोई ऐसा कौशल खो दिया है...

// Available locales: en, hi (Hindi), kn (Kannada), ta (Tamil), te (Telugu)

Currently includes a Hindi skeleton (5 red-flag questions, all probes, UI strings). Community contributions for additional translations are welcome.

For low-literacy populations, the engine provides TTS support via the Web Speech API:

import { createTTSProvider } from 'mychild-engine';

const tts = createTTSProvider(); // auto-detects browser vs Node
if (tts.isSupported()) {
  await tts.speak('क्या आपका बच्चा तेज़ आवाज़ पर प्रतिक्रिया करता है?', { locale: 'hi' });
}

Locale mapping uses Indian English and Indian language variants (en-IN, hi-IN, kn-IN, ta-IN, te-IN).

git clone <this-repo>
cd mychild-engine
npm install
npm run build --workspace=packages/engine
npm run dev --workspace=apps/demo

Open http://localhost:3000 to see the demo.

The demo includes:

• Screening tab: enter child DOB, answer age-appropriate milestone questions, see real-time severity badges
• Results tab: domain status cards across 8 developmental domains, global status, recommended actions
• Simulator tab: replay 3 built-in trajectories (normal development, speech delay, motor concern with regression), tune thresholds with sliders, compare alert diffs

Run releases from the repo root:

npm run release:engine:check
npm run release:engine:version:patch
npm run release:engine:publish

Notes:

• release:engine:check runs lint, tests, and npm pack --dry-run for the engine workspace.
• release:engine:version:patch creates the npm version bump for packages/engine. Use release:engine:version:minor or release:engine:version:major when needed.
• release:engine:publish publishes mychild-engine with --access public.
• Make sure you are logged into npm before publishing.

Contributions are welcome, especially from:

• Clinicians: threshold validation, question bank review, clinical accuracy
• Developers: engine improvements, test coverage, additional integrations
• Translators: multi-language question bank adaptations
• Researchers: retrospective analysis, sensitivity/specificity studies

Please open an issue before submitting large PRs.

• Engine code: Apache-2.0
• Question bank data: CC BY-SA 4.0

See LICENSE for full terms.

Harsh Songra — Forbes 30 Under 30 (India & Asia), TEDx speaker, World Economic Forum Agenda Contributor. Diagnosed with dyspraxia at age 11 after 9 years of misdiagnosis. Built the first version of MyChild App at 18. Website | TED Talk

Aafreen Ansari — Forbes 30 Under 30 (Asia), TED speaker, co-founder of We, Included. Dropped out of college at 19 to co-found MyChild App. Led product development and raised seed funding from 500 Startups and Singapore Angel Network. LinkedIn

MyChild App raised $100K in seed funding from 500 Startups and angel investors including Samir Bangara, Pallav Nadhani, and Singapore Angel Network. The startup ran its course. The problem didn't go away. 10 years later, the core technology is being open-sourced.

• CDC "Learn the Signs. Act Early." program for publicly available milestone checklists
• AAP developmental screening guidelines
• The original My Child App team and early users who informed the question bank design