Data-mining initiatives to improve patient outcomes
Posted Feb 02 2011 2:13pm
by Keeley Wray
Hospitals are under enormous pressure to cut costs to meet their own bottom line demands, yet stay open to patients.
Many are looking closely at all aspects of their business to find new ways to keep quality high and costs low.
In their historical performance and patient outcomes data, hospitals can draw upon a hidden treasure trove of information that can guide quality improvement efforts. The science of large-scale collection and analysis of patient data to inform medical best practices and increase efficiencies is in its embryonic stage, which is why it is more important than ever to have the appropriate infrastructure in place to capture this data now. That way, when these analysis techniques mature, the data will be available to analyze.
What types of data am I talking about? How do you generate cost savings and improvements in clinical care from data? The following categories are important to capture for every patient: financial, demographic, physiological, phenotypic, genotypic, environmental exposure and lifestyle.
Each of these categories would have multiple data types but unless they are all captured in a systematic and easily mineable way from each patient, hospitals will risk losing a priceless patient care tool. With sophisticated pattern recognition software you could use this data to inform clinical decisions to prevent adverse events or improve therapeutic approaches.
The path to using this data is long and arduous given the current lack of systems in place to effectively capture rich data sets in a systematic way so that that they can be analyzed later to inform patient care. Here are a few examples of initiatives at Children's Hospital Boston where we are trying to improve our data capture with the end goal of improving the lives of our patients:
Pioneering work in Electronic Health Records - Right now we have implemented the infrastructure for a personally controlled health record to give patients access to their data and down the road, the opportunity to enter information themselves related to their condition. Also, we have the ability to extract information from our existing physician-entered medical record data using Natural Language Processing algorithms embedded in our Informatics for Integrating Biology and the Bedside (i2b2) system. I2b2 allows researchers to view health data abstracted across patient groups. Right now a second-generation EHR platform under development will allow independent developers to write applications that communicate with the centralized record.
The Gene Partnership Program - The ultimate goal of this initiative is to sequence every patient that is seen at Children's Hospital Boston and follow his or her progress longitudinally, sending updates about relevant research findings to patients and families who elect to receive them. This initiative will tie highly detailed, ordered phenotypic information to genetic results, allowing us to make correlations between genetic background and health outcomes to be abstracted from the data.
Information Management in the ICU - Our Cardiac ICU is building a new system for information management of vital sign data. The system involves data capture, improved visualization and predictive alerts based on machine learning algorithms.
Standardizing Clinical Assessment and Management Plans (SCAMPs) - Standardized management plans have been developed based on known data and plausible outcomes for a number of conditions. One aim is to standardize practices and resource utilization where possible. Data is collected prospectively for patients entered into each SCAMP and the cohort analyzed to make feedback adjustments to the SCAMPs based on outcomes.
With these new initiatives, I am confident that we are securing our future as well as that of our patients. We are fortunate that we are a research hospital with informaticians who ensure that data get captured efficiently. This requires significant innovation given that it is only now that the appropriate conditions (technology, market forces and otherwise) enable disruptive approaches. The importance of ensuring that this first phase of data capture is done well cannot be discounted, because of its ability to inform clinical decisions.
Granted, there are challenges associated with our goals. Integration with centralized databases is a cumbersome process requiring creative programming. Determining the type of data that need to be entered by whom and in what format has yet to be immutably decided: it is currently an iterative quality improvement process of finding what "works."
Finally, ensuring that medical practitioners comply with data entry demands is a human motivational problem that stumps even the brightest thought leaders in interface design. But despite the challenges, I believe these new and continued hospital initiatives are right on track to lead us to clinical decisions that lead to improved outcomes.
Keeley Wray is technology marketing specialist at Children's Hospital Boston's Technology and Innovation Development Office.