One of the hallmarks of modern technology is the avalanche of information available about seemingly every subject, every aspect of our lives. Despite the wealth of data available, it can be difficult to use when the pieces of information needed reside in different databases or formats that are not easily combined. Medical records present a perfect example of the problem of multiple datasets and formats. At The Ohio State University College of Medicine, Albert Lai, PhD, research assistant professor in the Department of Biomedical Informatics, devotes his research to “information fusion,” or the integration of information from multiple, disparate sources to enhance the information signal.
With a family heavily involved in medicine, Lai always wanted to combine his education in computer science with his long-standing interest in medicine. As a graduate student in biomedical informatics, his research focused on the use of telemedicine to support chronic-disease management. Since arriving at Ohio State in 2009, he has focused on creating a cohesive and accurate picture of a patient’s health history. Lai hypothesizes that it is possible to use semantic and temporal data in the health record to create a health history timeline.
“Information fusion involves processing computer data stored in uncoded, narrative text fields, extracting structured data from unstructured data, merging multiple data sets, and binding episodic events together to create a medical portrait for each patient,” says Lai. He is working to create a virtual timeline that integrates inpatient and outpatient health events that can be applied to a patient’s medical record and leveraged for other purposes, such as research or the improvement of computer-aided clinical decision support alerts.
One of the major challenges to fusing data is the conversion of unstructured data in narrative text fields to structured data. A large proportion of the important data in an electronic medical record (EMR) resides only in narrative text fields, so it is not a trivial problem. Differentiation of similar health-related events is another challenge.
“We hypothesize that we can utilize both semantic and temporal information from multiple sources to help glue the data together in cases where an incident may appear to occur numerous times. Referencing the temporal information contained in the patient’s medical record will allow us to see whether events are the same or unique,” says Lai. In other words, if successful, he will be able to construct a timeline that shows that a subject has had two heart attacks from data that mention the two events 10 times throughout a patient’s chart.
Lai’s studies may lead to advances in clinical research across Ohio State’s Medical Center. Clinical trials are essential to the development of new therapies, and one of the most difficult steps in conducting a clinical trial is patient enrollment. Recruiting and enrolling patients into clinical trials is tedious, and screening study participants is especially time consuming.
“We want to create a cohesive timeline of a patient’s medical history that summarizes all events in order to accelerate the pace of recruiting patients to clinical trials. Cleaning up the data will allow us to leverage technology to query electronic medical records, expedite the eligibility screening and patient matching process and reduce research study coordinators’ workloads,” Lai adds.
In addition to enabling researchers to screen records more efficiently, Lai is working to help physicians point individual patients toward relevant clinical trials. Electronic medical records alerts programmed with clinical trial eligibility criteria can identify patients who meet the criteria and therefore promote physician referrals to study coordinators for assessment.
Lai’s goal of a fused medical history may also improve computerized clinical decision support. Best Practice Alerts (BPAs), such as recommendations for vaccinations, medication dose reminders, and alerts to changes in medical treatment, are alerts provided to physicians in real time in the electronic medical record. Unfortunately, though, BPAs utilize only structured data within the electronic medical record and do not reference unstructured data available only as narrative text. The inability to utilize the entire medical record in the implementation of BPAs leads to false alarm alerts, and the resulting “alarm fatigue” leads to the dismissal of all alerts as false, even when they are not. “We need clean data to work from in order to provide the most accurate alerts and feedback to clinicians, hence the importance of ongoing research,” he adds.
Lai further utilizes his expertise in biomedical data to assist biomedical researchers across the country via the caGrid Knowledge Center. caGrid is a software infrastructure that allows researchers to access and share data across and within institutions. The Knowledge Center, an NIH-supported function for which he is the principal investigator, is tasked with supporting researchers as they use caGrid. Lai provides scientific direction for the development of tools to securely share data so that they may be used across platforms.
In addition to his research in information fusion, Lai also realized that the Medical Center’s October 2010 launch of the Integrated Healthcare Information System (IHIS) represented a “once-in-a-lifetime opportunity to use OSUMC as a living laboratory” – a chance to study how workflow and time spent on tasks change during and after an electronic medical record system is implemented. Conducted in two intensive care units, Lai’s research examines efficiency before, during and post launch, observing the length of time that was required to perform tasks and the manner in which technical support was provided to staff. Lai is interested in the evolution of knowledge transfer and improvements and interactions impacting workflow efficiency and patient safety, and the overall affect of IHIS implementation on staffing during launch. “We hope it is clear how much staffing is sufficient to support such a tremendous project versus how much can be provided,” Lai says.
As Lai’s interest in using computers to solve problems in health care developed, he realized that, despite his initial plans for entering the industry, he was better suited to life as a researcher. Lai’s interest in research topics so close to home should lead to benefits for the entire Ohio State community, and he is hopeful that advances in his field geared toward gathering and sharing information more efficiently will help increase patient safety and lead to the best quality care.Click here to view a video of Albert Lai discussing his research.