Intel Corporation outlines how one healthcare provider has used machine learning and predictive analytics to enhance the quality of care and drive life-changing value from its data.
Healthcare organizations face a significant challenge in the growing volume of data that they collect on a daily basis – data which is increasingly complex and diverse. In addition to traditional sources of electronic health record (EHR) systems, picture archiving and communication system (PACS), new inputs such as social media, Internet of Things (IoT) devices and even patient-generated data from personal fitness trackers, are now joining the mix. This creates an unprecedented opportunity for providers to understand their patients and deliver more tailored care, but only if they can harness all this data and turn it into valuable insight.
The need to delve deeper into their data to drive insight and innovation in quality of care is just one side of the coin. While doing this, providers must also ensure they keep costs down, remain compliant with HIPAA and other regulations, and treat an increasing number of patients.
Traditional methods of insight gathering, which often rely on manual intervention, are no longer feasible in this burgeoning and increasingly complex data environment. The key to driving more proactive, cost-effective and compliant health services now is analytics. With the right technologies, you can turn this data into an innovation driver rather than a swamp for physicians and other decision makers to wade through.
Let’s explore how one healthcare organization – Montefiore Health System in the Bronx, New York – has taken control of its data to drive new insights, transform working practices, and enhance its service delivery.
Introducing the Semantic Data Lake
Montefiore wanted to use all the data it holds about its patients to help it improve care, and decrease patients’ length of stay and readmission rate. It needed to make sense of all this data, and present it to physicians and other care staff in a format they could easily use, and at the time they needed it.
To do this, Montefiore, which is the University Hospital for Albert Einstein College of Medicine, implemented a Semantic Data Lake, a large repository that pulls together all sorts of data – whatever the volume, source or type – into Franz AllegroGraph, a high-performance semantic graph database based on industry standards that enables predictive analytics at scale.. A layer of metadata creates ties between different components within the data, which makes it discoverable when interrogated with analytics algorithms. On top of that is an ontology layer, which contains specific medical terms, clinical trial details, and information about associated medications and diseases – essentially all the semantic characteristics of the data that make it meaningful to hospital staff.
The system constantly runs machine learning algorithms across this data, identifying patterns, testing them and making predictions about future trends. In this way, it creates an extremely broad and deep understanding of all the data held by the hospital, how it all fits together, and where connections exist.
Improving Care for the Sickest Patients
The Montefiore and Einstein launched with its new Semantic Data Lake was to create a clinical prediction score to indicate which patients are at a higher risk of death, or of requiring mechanical ventilation due to respiratory failure in the next 48 hours. Every few hours, the machine learning algorithms scan 42 different data points about each patient to allocate this score, and when they identify an at-risk patient, the system sends an alert to the clinical team to check on them. It also proactively offers support by suggesting questions to ask the patient, relevant clinical trial data, and details of treatment pathways for similar patients in the past. This provides a window of opportunity to intervene before the patient deteriorates, creating a better overall experience and potentially improved care outcomes.
The organization also found that the new solution can help with its compliance responsibilities. The HIPAA regulation provides clear rules around how and where patient data can safely be used. Semantic Data Lake can help ensure these rules are enforced by understanding and effectively managing the data, and it can also apply these rules to any new projects, enabling innovation without risk. For example, Montefiore/Einstein investigators are now looking at how to integrate data from outside its own firewalls, such as public health records or patient-generated fitness tracker data.
Answering the Unasked Questions
With this new approach to data management and analytics, providers are gaining deeper insight into each patient – how have patients with similar conditions been treated in the past? What were the outcomes and what variables contributed to those outcomes? Are there any clinical trials or new medications that may be relevant for this patient? What’s unique about their condition? How might their genetic make-up impact treatment? Not only are the answers to all these questions (and more) available to them at the touch of a button, but they’re available without the question even needing to be asked.
With all this insight at their fingertips, clinicians can reduce the time to diagnosis and treatment, cutting patients’ time in hospitals and saving costs. Of course, nothing will ever replace the clinician at a patient’s bedside, but this solution can offer these busy and stressed professionals support, clarification, and cognitive prompts to make sure they have considered all the options.