OCR App Scanning For Medical Invoices & Reimbursements in Germany

How does Pia work?

Pia Health app offers a smartphone-based solution to digitize invoice processing for health insurance reimbursements in Germany. The app allows users to scan physical invoices using their smartphone camera, enhanced by an integrated contrast filter to improve legibility. It also supports multi-page document scanning. The captured images are analyzed by Optical Character Recognition (OCR) algorithm, which extracts only the metadata, that is, key fields like recipient name, invoice amount, invoice date, due date, and banking details (IBAN).

This metadata is then processed through a locally-hosted Large Language Model (LLM), which further parses and organizes the data. All this processing is performed on servers that are hosted in Germany. No US-based Services were used. OCR and LLM analysis are essentially powered by open source solutions. The structured output is displayed in a backend interface resembling a spreadsheet (Excel-like) for user clarity and processing efficiency. Pia Health app then compiles the data into a clean, converted, readable PDF and attaches it to a pre-filled email including user contact information and insurance number, ready for direct submission to the user’s health insurer in Germany.

Pia Health app eliminates manual form-filling, physical mailing, and post office visits, significantly reducing administrative friction and improving the speed and accuracy of reimbursement claims. All communication is done digitally via email and supports nationwide compatibility with German private health insurers using the app.

From a compliance and cost-efficiency standpoint, Pia Health app adheres strictly to GDPR regulations. For development, on-prem EU data residency for OCR and LLM operations are performed exclusively on servers located within Germany, with no data transmission to third-party services or servers outside the EU. The solution uses open-source software for both OCR and LLM components, running entirely in-house. This approach avoids high licensing fees associated with commercial OCR tools in Germany and ensures that no external parties have access to sensitive medical or financial data, all maintaining full data sovereignty.

In short, Pia Health app is a privacy-first, cost-effective, and fully automated health invoice submission system designed for the German market, leveraging smartphone OCR, LLMs, and open-source tech, strictly within regulatory bounds.

Evaluating OCR and LLM solutions that are GDPR-Compliant, scalable and budget-friendly

From a product and technical standpoint, we, as a development team, had to identify the most effective OCR and LLM integrations that offered cutting-edge performance, were fully compliant with GDPR regulations and cost-efficient. This required picking an OCR capable of handling document scan variability while being either open-source or licensed for local, on-premise deployment within EU-based infrastructure. At the same time, choosing an appropriate LLM involves balancing performance accuracy, context length, and processing speed with the constraints of in-house hardware and limited budgets. 

Commercial models like ChatGPT, while powerful, are hosted externally, rely on large-scale infrastructure, and are not viable under GDPR for our built, Pia Health app’s use case. The core issue meant delivering the best possible results within the shortest processing time, using available resources, without compromising compliance. Noteworthy, open-source LLMs are released and are updated by the day, and the app’s backend infrastructure must remain adaptive. This means regularly benchmarking new models to assess whether they can deliver faster or more accurate results with the same or lower computational overhead. 

As a development team, we are on a constant dial in to ensure the technical stack remains both competitive and efficient. This includes creating a framework to test OCR and LLM results against actual use cases, assess cost-to-performance ratios based on current infrastructure, and guide our B2B client on an educated, informed, and timely decision on upgrading components. 

To address these quality issues, the development team evaluated multiple EU vendors offering a proprietary camera Software Development Kit (SDK). This licensed SDK was then integrated into the app, enabling sophisticated document detection and image optimization.

This improvement allowed for a much better OCR accuracy, even under suboptimal conditions. Better OCR output led to improved LLM parsing performance, with data extraction accuracy reaching > 90% and text processing times of less than 20 seconds, depending on document volume.

With improved OCR results, documents are automatically recognized, contrast enhanced, and poor lighting corrected. More so, manual entry was largely eliminated. The system automatically verifies key metadata fields for example, IBANs are mathematically validated and marked with a confirmation check. Users then proceed to the next step, selecting the intended recipient. 

Nevertheless, while the OCR and LLM pipeline achieves high accuracy, users are still prompted to review and verify extracted data. A confirmation screen then follows that allows for manual corrections or additional input to ensure complete accuracy before submission.