Center for Innovation in Health
Transformational Research and Development in Digital Health
The Center for Innovation in Health (CIH), based in Carnegie Mellon University's School of Computer Science, hosts the expertise needed to dramatically advance digital health research and technology. With more than 39 renowned scientists engaged with the center and representation across numerous disciplines, CIH is focused on improving the effectiveness and efficiency of digital healthcare.
CIH-related research thrusts range from digital bioscience and informatics, forecasting and generative AI to VR-based therapies, medical robotics, and clinical-care technologies and telehealth.
Our expertise domains include:
- Data-Driven Biomarker Discovery.
- Data Management Security, Privacy and Sharing.
- Health Analytics and Diagnostic AI.
- Patient Safety Healthcare Logistics.
- Public-Health Forecasting.
- Remote Monitoring and Telehealth.
- Sensors and Wearables.
- Preventative Care Strategies via Social Media.
CIH-research innovations advance many domains. Virus forecasting improvements have altered the course of public health response to disease, while rethinking and establishing new machine learning methodologies for advanced genomics will accelerate diagnostic capabilities. Similarly, using AI in image and sensor applications has revolutionized pathology, basic molecular biology and remote monitoring of patients.
Initiative for Patient Safety Research (IPSR)
Carnegie Mellon's Initiative for Patient Safety Research (IPSR) is a partnership and grant between CMU and the Jewish Healthcare Foundation (JHF) to build and engage a multidisciplinary community of researchers to analyze data with the goal of:
- Detecting medication errors and developing proof-of-concept innovations to reduce them.
- Developing new computational and analytical methods to identify and define medication errors within electronic medical record data.
- Identifying trends associated with the errors to provide an understanding of the precursors to, and potential causes of, medication errors.
The IPSR will foster and support research across the patient safety landscape, with a first phase of research that will collect and analyze data on medication errors using foundational data-driven concepts focused on data availability, pattern recognition and assessment. The ultimate goal of IPSR is to create a better health system and a healthcare environment dedicated to safety. Addressing medication errors is just the first step in using data-driven, systems-based solutions to anticipate medical errors and prevent them before they occur.
Our Research and Projects
Senior Care Delivery — Aging in Place
Seniors represent one of our most vulnerable populations, yet care for people aging in place remains substandard. Aging in place allows seniors to remain in their homes or communities as they age, with support services provided as needed. The ability to live in one’s home safely, independently and comfortably as one ages provides autonomy and familiarity. But it also requires specialized services and careful planning, including home modifications, flexible services and integrated technologies to help seniors live independently longer. Ensuring these services is not only a societal benefit but an economic one, reducing the cost associated with in-place specialized care for the elderly.
CMU has an opportunity to work with the Jewish Healthcare Foundation (JHF), Jewish Association of Aging (JAA) and potentially the PA Department of Human Services (DHS)/Area Association on Aging (AAA) to understand and assess needs, identify opportunities to leverage data collection and analysis, and apply machine learning/AI applications to deliver targeted and personalized meals/nutrition to seniors. The research goal is to develop more efficient operations and service delivery, provide higher levels of reimbursement around critical care delivery, develop value-added services, and enable more mechanisms to allow for socialization. The researchers will interview multiple service providers to understand operational capacity and needs while working with volunteers to determine issues and initiate ways to design pilot projects to provide assessment, adoption and use to increase stakeholder value.
Augmenting Electronic Health Records for Adverse Event Detection
Adverse events (AEs) resulting from medical interventions are significant contributors to patient morbidity, mortality and healthcare costs. Prediction of these events using electronic health records (EHRs) can facilitate timely clinical interventions. However, effective prediction remains challenging due to severe class imbalance, missing labels, and the complexity of EHR records. Classical machine learning approaches frequently underperform due to insufficient representation of minority adverse event classes and limited capacity to capture interactions among patient demographics, administered medications and associated complications. Gün Kaynar, a Ph.D. student from the Computational Biology Department in CMU’s School of Computer Science, is working on a method to introduce TASER-AE, a novel data augmentation pipeline tailored for structured EHR data, coupled with transformer-based classification. TASER-AE addresses these issues through an NLP-inspired data augmentation framework adapted for EHR, enabling effective minority-class representation in sparse and imbalanced clinical datasets.
Medication Reconciliation To Improve Patient Safety With AI and OR
Rema Padman and Ph.D. student Xinyu Yao are continuing work on their research project “Medication Reconciliation To Improve Patient Safety With Artificial Intelligence (AI) and Operations Research (OR) Methods.”
The medication reconciliation process aims to reduce adverse patient health outcomes resulting from incomplete information about a patient's current set of medications. This has been a long-standing patient safety challenge for which process-oriented measures and metrics have been proposed in the practice and policy settings, but are considered to be onerous, time-consuming and inefficient in meeting the goals of creating accurate medication lists. These approaches also do not leverage the vast amount of patient medication data currently available in clinical information systems such as EHR and CPOE applications that have been widely deployed in practice. Providing real-time access to clinicians and patients about medication discrepancies in a patient's medication list at the point of prescribing, based on scientific knowledge about drugs and diseases, clinical guidelines and patient-specific data, has the potential to significantly improve the medication reconciliation process and resultant patient safety.
2023 Capstone Project
Four students from the Heinz College of Information Systems and Public Policy worked on a CMU capstone project led by faculty advisers Rema Padman, Ari Lightman and Alan Scheller-Wolf. Assisting the CMU team was the Department of Biomedical Informatics (DBMI) team at the University of Pittsburgh, led by Professor Richard Boyce. The Pitt team provided access to the Medication Error at Regional Scale (MEARs) database, guidance in understanding the observational health data sciences and informatics (OHDSI) framework, and assistance in defining a representative patient cohort for the analysis.
The capstone project aimed to explore and validate signals of adverse drug events (ADE), adverse drug reactions (ADR) or medication errors with electronic health record (EHR) data. Doing so required the students to build an understanding of the patient journey in predefined cohorts through visualization and analysis, as well as for novel cohorts that they defined. The students used two data sources: EHR data (MEARS) from the University of Pittsburgh and adverse event reporting data (FAERS/AEOLUS). The CMU team, along with Padman and Boyce, identified a representative cohort (patients taking colchicine) and further narrowed that cohort to patients taking colchicine and clarithromycin (an antibiotic). This was done to focus on a known drug-drug interaction and provide a smaller dataset suitable for mining and analysis. The student team used several algorithmic methods for analysis but settled on association rules to uncover patterns in prescribing practices within the EHR data. They also worked with subject matter experts to identify rationale and patterns within the EHR data.
The capstone project was successful in identifying high-risk patient groups based on a high degree of confidence within the association rules identified, including patients taking colchicine and metaprolol (a well-known beta blocker). Other important achievements of the project were: introducing students to the OHDSI framework; strengthening our partnership with the University of Pittsburgh; developing techniques for extracting and processing data for analysis; identifying algorithmic methods to assess EHR data for indicators for ADR/ADE; and defining rules to determine levels of confidence in signal detection. We believe the methods and procedures developed by the team can be used, modified and expanded to undertake several student- and faculty-run projects on detecting ADR/ADE through EMR data to develop risk indicators, and eventually in applications for reducing the incidence of ADEs and severity of ADRs.
Faculty Spotlight in Digital Health
CMU Experts: What If Infections Could be Treated Without Antibiotics?
CMU is Revolutionizing Research Through this MRI Facility