More information can be found at mgh-cdtl.org.

BPCi is an application for real-time, bedside clinical decision support. The system aims to provide automated guidance for reducing the duration of hypotension in ICU patients on vasopressors. The BPCi system is comprised of a server and a user interface. The UI is an evolution of the APPRAISE project.

As the sole developer on the project, I built the server and UI. The server application is written in Python and is backed by a PostgreSQL database. The server aggregates all clinical data and calculates several algorithms including the Clinical Decision Support messages (CDS) and the Blood Pressure Control Index (BPCi). The UI is written with a web technology stack (JS/HTML/CSS) and utilizes the React/Redux libraries as well as D3 for plotting. The UI is packaged as a native application using Electron.js. Communication between the server and UI(s) is accomplished using ZeroMQ. Docker is leveraged for automating the build process.

Additionally, I created the 4th iteration of the BPCi algorithm (originally developed at U of Maryland). The algorithms are implemented in Python using Numpy/SciPy. In order to develop a new iteration of the BPCi algorithm, I also built a MATLAB data analysis pipeline. Timeseries vital signs from several hundred subjects, including from MIMIC II, were analyzed. A modified autoregressive model was developed to predict the likelihood of sustained hypotension. This model was analyzed in MATLAB and ported to Python in the application server. This project is awaiting IRB approval for testing in the ICU.

APPRIASE is an application for real-time, bedside clinical decision support. The system provides automated guidance for clinicians during a trauma resuscitation by monitoring data from the bedside monitor and user controls. I built a prototype UI to view the patient’s vital sign trends, the APPRAISE risk strata, and the clinical decision support notifications. The APPRAISE algorithm is a logistic regression model that outputs a risk strata. I developed a new iteration of the model based on a larger dataset, TraumaDB. I assembled TraumaDB by meticulously reconciling all existing trauma project datasets into a standardized schema in PostgreSQL. APPRAISE is IRB approved for use in the MGH ED and is awaiting study deployment.

More information can be found at openice.info and mdpnp.org.

OpenICE is a prototype clinical ecosystem for connecting medical devices and clinical applications. A multidisciplinary team worked to instantiate a standards-based distributed system and implement proprietary medical device comm protocol translation.

Personal responsibilities included requirements management, designing and developing project website (www.openice.info), development of demonstration software and conference exhibits, medical device integration testing and validation, wrote and maintained user facing documentation, user trainings, and community site implementation and management.

More information can be found at opencommons.uconn.edu.

Modern hospitals are plagued by excessive alarms generated by patient monitoring technologies with very high sensitivity and low specificity. These excessive alarms create high rates of false and clinically irrelevant alarms. Multiple studies have shown that these false and clinically irrelevant alarm rates can negatively impact patient care and lead to "alarm fatigue". For my Master's thesis, I employed a quantitative, database driven approach to alarm management in the acute care and medical/surgical environment with the intent of identifying and implementing technological, clinical, educational, and workflow practice changes to curtail excessive alarming.