Telehealth EMR/EHR Desktop System Development Using WPF

Client: Medha Hospital, 8 public hospitals and 23 outpatient clinics

Megha Hospital operates 8 Public Hospitals and 23 OPD Clinics. The legacy EMR system was web-based but had performance and reliability issues, both due to rural broadband limitations and system failure during peak.

Security logs were also not complete, which made it difficult for Megha Hospitals to conduct a GDPR audit.

The client's requirement was an Offline-Friendly and Desktop-Based EMR Solution , Automated Diagnosis via AI along with Vitals Data from IoT Devices being sent to the EMR System.

Selected Technology

• WPF (Dot Net 9) using MVVM
• Entity Framework Core (Offline: SQLite - Online: SQL Server)
• HL7/FHIR Gateways
• Real-Time Telehealth Video

During the initial phase of discussions, we had a meeting with the CIOs, and they told us:

"We don't need another React dashboard. We need a system that works when the internet fails. We need every mouse click logged for compliance. And we need doctors to finish notes before they leave the room."

Web EMRs caused:

• A 3-5 second delay in the search for patients
• Loss of prescribed medications due to network jitter
• No access to lab results offline

Consequently, WPF was chosen for the following reasons:

• Direct access to hardware such as fingerprint readers, barcode readers and signing pads
• Consistent performance with older PCs (up to 5 years old) in hospitals
• Automatic uploads of newly added items (data) when the WiFi is functioning again

Management of Patient Records

• Unified View of Timeline: All treatment, lab & imaging reports, and nursing notes are displayed in chronological order from newest to oldest.
• Fast Text Search Ability: Using Lucene.NET indexing prevents lengthy returns of patient record searches; however searches for 2 million records have an average return time of <0.2 seconds.
• Demographic Information and Emergency Contacts: Can call/send a text message from the desktop to an emergency contact.

Lab Reports & Imaging

• DICOM Viewer Integrated into the Application: Embedded in a WPF image control with custom pan/zoom capabilities for displaying x-rays and MRIs.
• Lab Trends: Shows one-year trend of lab tests, such as HbA1c or creatinine levels, with ScottPlot (lightweight plotting solution without needing webview).
• Abnormal Results Flags: Runs AI model (ONNX Runtime) locally, notifying users of any abnormal test values when doctor first opens lab report.

E-Prescriptions

• Linkage to national drug databases - Kanta in Finland, G-Standard in the Netherlands, barcode scanning for medications.
• Drug interaction check - Real-time, cross-reference patient current medications against local drug database (updated weekly).
• One-click refill - See latest prescription and also include dosage change; complete audit trail
• Print-out and QR-code - Patient scans QR code to see medication instructions in their own language.

Insurance & Billing

• Multi-payer engine - Public insurance systems (DRG), Private insurers, Self-Pay
• Automatic coding suggestion - AI auto-extracts ICD-11 codes from the physicians' standardized documentation.
• Real-time eligibility check - Background thread queries the insurer's API without freezing the user interface.
• Claim status tracking - Claims that were rejected are highlighted along with the reason codes.

Compliance logs (similar to HIPAA and GDPR) have the following characteristics:

• Creation of an unmodifiable audit trail. Every access and modification (or lack thereof) performed on the local write-once SQLite database will be logged in the database until they are transferred to an unchangeable ledger in the cloud at a later time.
• Creation of logs via "break-glass" access protocol will include an emergency contact PIN for anyone requiring emergency access to the patient's file while the patient is unconscious. The log will remain stored in an unchanging manner.
• Creation of screenshots (via background) for any attempt made by the user to take the patient file to a USB drive will be stored in the audit folder as a record of the user's wrongdoing.

Beyond the requirements, we added:

• Voice-to-text SOAP notes – Offline speech recognition (Windows Dictation API) for doctors. Reduces typing by 60%.
• Smart consent forms – Patient signs on capacitive signing pad; hash of signature stored, PDF generated instantly.
• Video telehealth – Embedded WebRTC via WPF WebView2. Call recording opt-in only, with automatic transcript added to EMR.
• Role-based UI – Doctor sees prescriptions + notes; nurse sees vitals + tasks; biller sees only financial data.
• Task reminder engine – "HbA1c due in 2 weeks" or "Prior authorization expiring" – desktop toast notifications.
• Bulk operations – Vaccination campaigns: select 200 patients → generate batch e-prescriptions → print all.

AI Assistants (e.g., ClinicoCopilot)

• Create summaries of discharge plans (10 pages to 3 bullet points - admission reasons, procedure descriptions, discharge medication lists.)
• Create risk scores based on vitals, age, and lab trends to predict the likelihood that a patient will be readmitted within 30 days. Displayed as red-yellow-green gauge on the computer screen.
• Help physicians develop differential diagnoses: Patient has chest pain, shortness of breath, and leg swelling → top 3 possible diagnoses with references.

All AI is executed on the hospital's computer and does not include any patient data leaves the hospital.

Predictive Scheduling (Time Series)

• Predicts the likelihood that a patient will no-show for their appointment, using historical attendance data, as well as the weather and day of the week. Then automatically double-books those patients who are unlikely to show.

We connected medical devices (IoT) to a WPF Desktop Application without using the cloud.

Bedside Monitors Wirelessly Transmitting Vital Signs

Vitals – The WPF application receives vital signs through:

• Bluetooth LE – Using Nonin Pulse Oximeters and Bluetooth-enabled Omron Blood Pressure cuffs.
• USB Serial – The WPF application reads byte streams of data from older GE patient monitors and displays this data live.
• Alerts – If a patient's SpO2 level falls below 90% and a nurse has not documented within the last 2 minutes a pop-up alert will be displayed as well as a message will be displayed on the hallway screen.

Smart Medication Dispensing Systems

The nurse scans the wristband of the patient and then the blister pack of medications. The WPF application verifies that the "Five Rights" have been followed (i.e. right patient, right drug, right dose, right route, right time).

If the "Five Rights" have not been followed, the drawer in the dispensing cabinet will lock and the WPF application will display the incorrect message on the screen.

Ambient Room Sensors (Pilot initially implemented in two wards)

• Occupancy sensor – recognizes whether someone is in bed or using the bathroom; EMR indicates an icon to demonstrate lower fall risk.
• Door Open Counter – Infection Control Dashboard: Logs the number of times each isolation room door is opened and closed.

IoT architecture:

WPF app subscribes to local MQTT broker (Mosquitto) on hospital VLAN.

Device → Raspberry Pi gateway → MQTT → WPF subscriber → EMR update.

Latency under 100 ms.

The client also demanded no subscription fees. They paid a perpetual license + 15% annual maintenance. WPF made this model viable.

• Deployment Timeline: 9 months (5 months minimum viable product)
• Pilot Project: one hospital, consisted of 60 doctors and 200 nurses, duration of the pilot project was 6 weeks (1st 3 weeks were to create baseline data).
• Onboarding through Adoption Process (after 3 months): 94% daily active users using the new online EMR versus only 47% of the old web EMR.
• Reduction in the average time from 9 minutes to 4 minutes for documentation time per patient due to the use of speech recognition with AI summarization.
• Reduction in prescription errors of 82% due to use of interaction checker and barcode scanning.
• Reduction in patient no-show rate by 23% due to use of predictive reminders.

• Challenge: Offline sync conflict when same patient edited at two clinics.
  Solution: Last-write-wins with field-level version stamps. UI shows "Remote update detected – review changes" diff view.
• Challenge: Telehealth video + local AI together consumed 3.5 GB RAM.
  Solution: Unload AI models after 30 seconds idle. Used GPU for inference, CPU for video.
• Challenge: 80+ different devices (printers, label makers, scanners) – all needed to work.
  Solution: Device abstraction layer + driver fallback to generic USB HID.

• Hospital savings: reduced transcription + fewer prescription errors + less no-show loss
• Payback period: 6 months

"Finally – an EMR that doesn't freeze during a blackout. Our doctors actually close the browser now. And the IoT vitals coming straight into the patient chart? That's not science fiction. That's Tuesday morning."

— Dr. Lena Vinterberg, Chief Medical Informatics Officer

The development of an EMR/EHR desktop application using WPF demonstrates that it is possible to create enterprise-quality offline-first integrated medical software using AI and IoT while providing better overall performance than web-only solutions.

In addition, WPF provides access to native devices, consistent performance, and provides an enjoyable user experience for clinicians.

Hospitals require a desktop solution that has both artificial intelligence and internet of things capabilities, rather than simply creating another cloud-based dashboard.