In the age of generation, analysis and reanalysis of large multidimensional datasets, shared resources centered on computing infrastructure, data management and analysis have become as essential to the research mission as those focused on wet lab applications. In this edition of the Research Resource Nexus, we introduce the Health Services Research Data Center, a shared resource that maintains a secure computing infrastructure for HIPAA-protected data and the support services to use it effectively. Please note: This facility does not provide statistical analysis.
Health Services Research Data Center
Watch the Health Services Research Data Center introductory video.
The Health Services Research Data Center (HSRDC), a program of Health Sciences IT, provides a secure computing environment where each PI group can store and analyze sensitive datasets.
Dan Ricketts, director of Health Sciences IT, established HSRDC in 2011 as a secure system to store claims data. Over the last 13 years, the core’s services have grown to include
- secure virtual desktop for storage and analysis of sensitive data,
- data management support,
- custom application development and
- ready-to-use cloud environments in Azure and AWS for compute and storage (e.g. synapse, storage accounts, s3 buckets and ai/ml tools).
The HSRDC routinely works with a wide variety of data types including electronic health records (EHR), claims data, administrative data, real time monitor data, datasets with restricted data use agreements or any sensitive data that needs to be protected from hacking and misappropriation.
The secure computing environment is a virtual Windows desktop or a portal into a cloud virtual machine that can be accessed through a browser from any location. Users interact exactly as they do with any other Windows or Linux environment, with the exceptions that, due to security measures, it is not possible to access the internet directly from this desktop and data download is restricted to small results files that must be transferred using a specific application.
When using HSRDC services, the team offers support and assistance for
- data use authorization and institutional review board application submission,
- data management plan development and budgeting,
- data import into the secure environment,
- data availability to authorized users,
- secure data archive or deletion as appropriate,
- solutions architecture consulting and
- server support.
The HSRDC team is composed of IT experts with strong and complementary skills in database administration, data management, application development and infrastructure support. This team has the expertise necessary to help investigators collect and manage the data they need to answer their scientific questions in a robust and reproducible manner.
To initiate a project with HSRDC, email hsrdc@pitt.edu.
Testimonials
Julie Donohue (Department of Health Policy and Management, School of Public Health) has used HSRDC since its inception to house sensitive claims and EHR datasets. Donohue analyzed Medicare and Medicaid claims data to uncover population patterns in opioid use and abuse, addiction treatment and overdose risk. A HIPAA-compliant, secure environment like that provided by HSRDC was required by the government agencies providing access to this protected data. She is currently engaged in similar analyses around the new weight loss drugs and severe mortality and morbidity in maternal health. HSRDC installs and maintains the cutting-edge analytical software her team needs. The ease with which her team members can share data and findings within the secure desktop has fostered fruitful collaborations. "The HSRDC is essential for our work,” says Donohue.
Gilles Clermont (Department of Critical Care Medicine, School of Medicine) develops models to integrate the dense and dynamic realtime data collected by intensive care unit monitors with more static and contained EHR data into decision-support applications for clinicians at the bedside. The size and density of this data creates challenges in
- data transfer and storage,
- processing speed (answers must come quickly enough for clinical decision points), and
- human interpretation (presenting data to physicians so they can understand it quickly and act upon it).
Clermont found HSRDC to be responsive to the unique needs of this work. The team stepped up to the challenge of learning to manage new data types and demonstrated flexibility in adapting to new technologies and new ways to model data.
HSRDC built the infrastructure to merge multiple data types from 14 different institutions in support of Clermont’s work as data acquisition lead with CHoRUS, a National Institutes of Health initiative “[to create] an ethically sourced, AI-ready dataset from different American institutions to support future discoveries in clinical care.”
Sandy Gill (Department of Pharmacy & Therapeutics, School of Pharmacy) is working with collaborators in Florida to develop a machine learning and AI clinical decision assistance model to evaluate 200 variables related to acute kidney injury (AKI) in hospitalized patients. When operational, the system will alert pharmacists to elevated AKI risk so they can adjust medications and reduce that risk. HSRDC has worked with Gill’s collaborators to make the model operational in Pittsburgh. “They have been true collaborators, attending all meetings, communicating about project progress and sharing ideas on how to enhance it,” says Gill. Once software development is complete, the team will embark on a clinical trial to test if it can successfully reduce the risk of AKI. “With the help of HSRDC, the system is built to make adoption easy if the clinical trial results are positive.”