Decision Support

A team of scientists trained a neural network to evaluate electrocardiograms to predict which patients were likely to develop an irregular heartbeat.

The companies say their approach to remote patient monitoring is uniquely modular and customizable, enabling provider organizations to implement pieces of an end-to-end system that best meet their particular needs.

In addition to gathering opinions on vendors including Health Catalyst, IBM Watson and Jvion, the research clears up some misconceptions about artificial intelligence systems and offers some best practices for safe and effective deployment.

At the HIMSS Asia Pac 19 Conference in Bangkok, Thailand in early October, Omar Sunna, Director, Global Product Management, GE Healthcare, shared his insights on some of the data and workflow-related challenges currently faced by healthcare providers and how some of GE’s solutions and applications can help tackle or address some of these issues. Consuming the right amount of algorithm that creates desired levels of outcomes Omar said that one of the big challenges for healthcare providers is knowing which algorithm is the most appropriate for the required workloads. Additionally, there is no single vendor that is going to provide all the algorithms and the challenge is to create an ecosystem that allows the provider to work on the algorithms regardless of vendors/developers that can be infused into the workflow. “We see this as a tremendous opportunity and a challenge today. It is an opportunity for us to partner with customers and solve this issue in a way that is compatible for the workflow, in a way that controls the costs, increases productivity and promote the right levels of access for the patients. Above all, empower for improved outcomes and higher quality care through leveraging intelligent applications and devices,” he explained. One of the solutions that can tackle this challenge is GE’s Edison AI platform, which focuses on two things. The first is to break the silos of data within the healthcare organisation, to orchestrate the data within the organisation and create a foundation for aggregating the data, making sure that they can look at the patient holistically. The second is to create an ecosystem so that GE developed algorithms as well as ones developed by partners of third parties (such as startups) are able to integrate it into medical devices and leverage Edison's data lake, so the algorithm continues to grow, even though it was developed by a third party company.  “Continuous learning is a big focus because the more you feed the algorithm with data, the more precise and accurate it gets, and that’s the goal with Edison.” A potential implementation of AI-embedded workflow One for the opportunities that Omar sees in which AI can help in improving workflows is in the area of radiology, such as to notify the radiologist that there is potentially an urgent need that needs to be done. There are also opportunities to show preliminary findings to the radiologist so that he or she is able to interpret it, review it, look at patient-specific trends based on prior imaging and make the determination of the right diagnosis for that patient in the most efficient manner.  He provided an example of the workflow orchestration maturity model of an imaging interpretation which can be divided into three steps: 1) How to determine the right imaging exam 2) The diagnosis and treatment plan 3) The study interpretation and reporting In his experience, most of the customers have a fragmented and ad hoc approach and there is an opportunity for quality improvement in terms of the pre-scan algorithm and which protocols are used but most of that are dictated by their jobs in a manual process. “There’s an opportunity where the data can be leveraged so you need multiple sources of data. Today, a lot of healthcare providers are struggling with being able to access that data from multiple sources.” Data and data management There is a data explosion in healthcare – data within the hospital setting in growing at 48% per year and data aggregation is growing as well. From a healthcare technology provider perspective, Omar sees great opportunity to provide solutions for GE’s customers. Within GE’s enterprise solutions business unit, its software applications are divided into three categories: solving for diagnostic speed and accuracy, fostering collaboration outside of the hospital network and focusing on productivity around the health system. In terms of solving for diagnostic speed and accuracy, one of the software applications and tools is the Centricity Cardio Enterprise, which helps the cardiologist aggregate all the data that is relevant for the patient. The cardiologist can access the ECG, the echo, the cardiac CT and arrive at a diagnosis in a fast time with a higher level of competence. For fostering collaboration outside the hospital network, such as consulting with specialists and sub-specialists, there needs to be a way to share reports and imaging with them easily and in a secure manner and this is where solutions such as GE’s vendor neutral archive (VNA) and Centricity 360 come into play. Lastly, GE’s analytics solutions that are embedded within the workflow that provides the right level of insights can help improve efficiency of clinicians, help them make the right decisions and lead to an increase in productivity. Omar believes that the vendor neutrality of GE solutions brings about several benefits. This means that GE’s VNA can adhere to industry standards such HL7 which has the capability of FHIR APIs and the ability to leverage IHE-XDS standards.  “Another aspect of neutrality is being able to leverage the IT ecosystem. Being able to reside in the customer's own hardware, being able to leverage on-prem storage provided by the hospital and being able to open it up to different vendors, as long as we validate the standards of that technology. We also leverage cloud-based storage. All of that is key for us to be able to build scale,” Omar concluded.

Data from the accelerometers allowed researchers to correctly rank the mortality risk using 30-40 percent more accuracy than when using data about smoking status or a patient's stroke or cancer history.

Partnering with two firms focused on lab analytics and building a High-Value Care program, the health system within 15 months reduced unnecessary potassium tests by 63%, unnecessary magnesium tests by 40%, and unnecessary lipid panels by 40%.

Real-world evidence drawn from unstructured clinical notes was more accurate in algorithmic prediction of coronary artery disease than structured data, a new study in JAMIA shows.

Genomic medicine researchers at the laboratory have been using artificial intelligence, developed as part of Microsoft's Project Hanover, to help manage the vast amount of research data needed to power its precision oncology initiatives.

The leading-edge technology could provide vastly faster power and processing speeds, and enable fundamentally different algorithmic search and data homogenization strategies.

Experienced professionals in the area of nursing informatics from Singapore recently shared their lessons learnt at the Hillrom Dialogue Series.