mhealthgen

mhealthgen - Random health data generator

mhealthgen is a javascript library for generating realistic health-related information about a person in JSON format. The generated data include data types commonly recorded by modern consumer mhealth devices, like Apple Watch, blood pressure cuffs, and patient diaries.

This includes blood pressure, calories burned, caloric intake, heart rate, height, weight, sleep, life satisfaction, prescriptions, clinic visits, diagnoses, and symptoms.

Example portion of some random data generated using mhealthgen:

{
  "firstName": "Charles",
  "lastName": "Wirth",
  "birthdate": "1963-02-24T13:42:52.344Z",
  "sex": "male",
  "bloodType": "a+",
  "normalHeight": 150,
  "normalWeight": 90,
  "normalSleepHours": 8,
  "normalBurn": 2770,
  "restingHeartRate": 60,
  "nodes": [
    { "date": "2016-10-27T15:23:58.320Z", "type": "hr", "source": "iwatch-hr", "value": 61 },
    { "date": "2016-10-28T00:13:53.190Z", "type": "sleep", "source": "fitbit-sleep", "value": 12.7 },
    { "date": "2016-10-28T11:23:58.320Z", "type": "hr", "source": "iwatch-hr", "value": 81 },
    { "date": "2016-10-28T11:24:53.223Z", "type": "hr", "source": "iwatch-hr", "value": 76 },
    { "date": "2016-10-28T11:25:46.610Z", "type": "hr", "source": "iwatch-hr", "value": 68 },
    { "date": "2016-10-28T12:07:29.737Z", "type": "intake", "source": "fitbit - intake", "value": 2513 },
    { "date": "2016-10-29T22:19:42.538Z", "type": "weight", "source": "scales", "value": 95.5 },
    { "date": "2016-10-31T15:25:43.758Z", "type": "bp_diastolic", "source": "valuemed-bp", "value": 80 },
    { "date": "2016-10-31T15:25:43.758Z", "type": "bp_systolic", "source": "valuemed-bp", "value": 112  }
  ],
  "sources": [
    { "id": "fitbit - intake", "name": "Fitbit App", "types": ["intake"] },
    { "id": "iwatch-hr", "name": "Apple Watch", "types": ["hr"], "precision": 1 },
    { "id": "valuemed-bp", "name": "ValueMed Blood Pressure Cuff", "types": ["bp_diastolic", "bp_systolic"], "precision": 2 },
    { "id": "scales", "types": ["weight"], "precision": 0.5 },
    { "id": "fitbit-sleep", "device": "Fitbit Surge", "types": ["sleep"], "precision": 0.1 }
  ],
  "types": [
    { "id": "bp_diastolic", "name": "Diastolic blood pressure", "unit": "mmHg" },
    { "id": "bp_systolic", "name": "Systolic blood pressure", "unit": "mmHg" },
    { "id": "intake", "name": "Caloric intake", "unit": "kcal" },
    { "id": "hr", "name": "Heart rate", "unit": "bpm" },
    { "id": "sleep", "name": "Sleep", "unit": "hours" },
    { "id": "weight", "name": "Body weight", "unit": "kg" }
  ]
}

This tool is not intended to be used for medical purposes, but may help to test tools for working with health data. The normal ranges of health measurements mostly uses UK population averages.

Installation

For use within node.js, install via npm:

npm install mhealthgen

For use within browser via CDN:

<script src="https://unpkg.com/mhealthgen"></script>

Or, install with bower:

bower install mhealthgen

Usage

To generate a health data for a random person to person.json using CLI:

node mhealthgen person.json

Full CLI usage:

node mhealthgen [options] <outfile>

Options:

• person - existing person profile (this can be a json file previously generated)
• days - number of days to generate (default 30)
• associations - adjust values based on associated preceding events (default true)
• fluctuations - adjust values based on random fluctuations (default true)

To use the API (options match the CLI options):

let healthdata = require('healthdata');
healthdata.generate();
 
let dennis = require('people/dennis.json');
healthdata.generate({ person: dennis });

Implementation

Values are sampled using sources - these are the devices or techniques that the fictional person used to take measurements, such as an Apple Watch or blood pressure cuff. The frequency that the person uses this varies from person to person. The values are generated in three stages:

1. Initial: An initial value is computed, based on, for example, a population average for patient's age/sex
2. Associations: The value is adjusted based on any associated preceding events. For example, physical acitivity in the past hour will increase blood pressure.
3. Fluctuations: To account of other factors which haven't been specifically considered in the generator, simplex noise is used to create random fluctuations in the value.

Health datatypes

ID	Name	Unit
weight	Body weight	kg
sleep	Sleep	hours
satisfaction	Life satisfaction
procedure	Clinical procedure
prescription	Drug prescription
palpitations	Palpitations	severity
medicate	Dosage	%
inr	International Normalized Ratio (INR)
height	Height	cm
hr	Heart rate	bpm
diagnosis	Diagnosis
intake	Caloric intake	kcal
caloric_burn	Caloric burn	kcal
breathlessness	Breathlessness	severity
bp_systolic	Systolic blood pressure	mmHg
bp_diastolic	Diastolic blood pressure	mmHg

Sources

ID	Name	Description	Types	Precision
palpitations	Symptom diary	Person recorded an episode of palpitations.	palpitations
breathlessness	Symptom diary	Person recorded an episode of breathlessness.	breathlessness
scales	Weighing scales	Consumer analog weighing scales.	weight	0.5
medicate	Medication diary	The person took their medication and recorded it.	medicate	0.01
inr_reader	INR reader	Consumer INR reader.	inr
height manual	Tape measure.	Tape measure.	height	0.5
fitbit-step-counter	Fitbit Surge	A wearable device which records steps through motion sensors.	caloric_burn
fitbit-sleep	Fitbit Surge	A wearable device which records sleep through motion sensors.	sleep	0.1
fitbit - intake	Fitbit App	An app which can be used with the Fitbit device to record diet.	intake
daylio - mood	Daylio app	A mood diary app.	satisfaction
clinical-visit	Clinical visit	Visit to a clinic or hospital	diagnosis, procedure, prescription
valuemed-bp	ValueMed Blood Pressure Cuff	A cheap consumer blood pressure cuff.	bp_diastolic, bp_systolic	2
iwatch-hr	Apple Watch	Apple Watch smartwatch heart rate monitor. The heart rate is read using light sensors.	hr	1

Todo

• Associations
• Descriptions of provenance in nodes
• Use of existing ontologies/schemas: SNOWMED, LOINC, openmhealth