Introducing ParkinSync: A Household Care-Log Research Pipeline

ParkinSync is a single-household research prototype for organizing structured paper care logs alongside weather and indoor-environment data. It is not a public care service, a diagnostic tool, or a validated predictor of symptoms.

Keep the paper workflow, check the digital result

A paper record and an OCR result are different kinds of evidence. A successful extraction request does not establish that a handwritten value was read correctly.

The current handler automatically imports normalized rows into Google Sheets, then asks an operator to review them. This is a responsibility to review after import: the code does not enforce approval before writing to the ledger, and it does not establish that every imported row has been checked. An operator should compare the imported values with the source before using them for exploratory analysis.

What the pipeline does

An S3 upload triggers an AWS Lambda handler. Amazon Textract extracts table cells from the structured form; the pipeline adds weather context and normalizes the result into a 25-column schema before appending rows to Google Sheets. A separate scheduled Lambda collects indoor temperature and humidity from SwitchBot.

The spreadsheet makes the output available for inspection and correction. That ability is useful, but it is not an implemented approval gate or a guarantee of data quality.

What the observations can mean

The repository includes scripts for preparing data and exploratory analysis. Those tools can help inspect missingness, timing, and environmental context. They do not turn a household observation into a clinical conclusion, establish causation, or provide treatment advice.

This introduction does not claim a measured OCR accuracy, a clinically validated temperature relationship, or a demonstrated operating cost for a multi-patient service. Those claims would need evidence and a defined measurement scope of their own.

What remains to build and verify

Stronger review controls, including explicit review states and excluding unchecked rows from analysis by default, remain development work. General availability and clinical validation are separate from having an import handler and a test suite.

The product page describes the current scope. Source code, tests, and deployment notes are in the ParkinSync repository. The pipeline walkthrough explains the components in more detail.