Most of us have a very linear approach when it comes to dealing with data. If A and B are above a definite number, C has to be a certain value. Not just sometimes, but all the time. And I think it’s safe to say that healthcare data isn’t that way. It’s complex and at times inconsistent, making it difficult to perform a linear analysis.
Healthcare sector is sitting on a goldmine of data- structured, unstructured, disparate, dated and whatnot. True, we have electronic medical records and the progress made with them has been remarkable, but the information they provide isn’t very different or groundbreaking as compared the old paper records.
Not data, let’s bring information
Almost every physician I have come across would prefer application-ready information- something they can use at the point of care. Physicians would appreciate having easy access to a patient’s vitals without having to scroll through hundreds of records. Processed and analyzed information can assist providers in making better decisions based on patient diagnoses, their vital signs, and the available treatment options. Not only a better understanding, but good analytics can also help providers determine possible outcomes of the care plans they prescribe.
The bottom line is that a reliable transformation of raw data into clinical insights will not only help providers plan and provide care; it would ultimately lead to better outcomes, lower costs of care and an improved patient experience. However, the question of essence remains- how do we begin? How do we pull data together? How do we make these processes efficient?
Extensive analysis and its challenges
Analysis in healthcare is challenging- not similar to how other industries leverage it. The less daunting and easier one is the very nature of healthcare data. It comes from all over the organization- EMRs, lab records, claims, HRs, different departments and aggregating it all into a single, holistic record is a challenge. Almost 80% of healthcare data is unstructured and inconsistent- which means processing and analyzing data would be like creating order out of chaos.
The second and the more difficult one is the fact that healthcare is dynamic, and we are constantly coming across newly determined knowledge. Analytics, in many sectors, works on binary prediction- either the answer is ‘Y’ or ‘N.’ Most of the attributes like readmissions, no-shows, ED visits could be sorted by either a ‘Y’ column or an ‘N’ column. However, as our understanding changes, our goals do too. For example, a HbA1c level of 6.5% or higher is diagnosed as Type 2 diabetes. It’s possible that if providers discover other valuable insights, this agreement might change.
It’s not just creating order out of chaos- it’s trying to hit a moving bull’s eye, only that the bull’s eye is moving unpredictably. That’s why managing population health needs something far more sophisticated than other industries like retail or manufacturing.
Bringing machine learning to care
Agreed, robots and technology can never completely replace physicians and nurses. However, we have to entertain the fact that machine learning and AI hold the potential of improving outcomes, providing more accurate care and basically transforming healthcare as we know it.
Imagine a patient walking into her PCP’s office about pain. After listening to her symptoms, the PCP inputs them into his computer, pulling up her last test records and understanding the diagnosis and previous treatments. The physician would take a look at the patient’s vitals, her allergies and her responses to previous treatments with algorithms helping him detect any anomaly too small for a human to see. Finally, the physician looks at the patient’s medical history and family history and suggests a treatment that is specially cut out for the patient.
The advent of machine learning was based on the initial thought that it would be the answer to all data-related questions. However, in the past decade, healthcare has really progressed on making data accessible, and now providers are turning to machine learning.
Not only machine learning in healthcare is mature enough to improve diagnostics and predict medical events, but it has also only just begun to scratch the surface of personalized care. Several renowned companies and research facilities are using machine learning to identify cancerous tumors, diagnose diabetic retinopathy and identify events of readmissions. Still, machine learning is better suited for some processes than others, some areas yet unexplored:
Improving or replacing processes that are standardized or reproducible in nature.
Reading through large, image-based datasets like radiology or cardiology and identify trends and anomalies, making these processes efficient and accurate.
Understanding the high-level information as well as predicting what might be the best course of action.
An Intelligence System for Healthcare
We at Innovaccer recently launched the world’s first Care Intelligence SystemTM for Healthcare, with a practical machine learning perspective. The Care Intelligence SystemTM strikes a balance between traditional analytics and machine learning. We use a Hadoop-based Integrated Data Lake to integrate and analyze data and to derive useful information about diagnoses, care coordination and suggests preventive measures. In case of any anomalies, the algorithm prompts physicians with alerts and loops back the information in real time to improve clinical decision-making. Considering the requirements of healthcare providers all over the country, CISTM offers some features such as:
- Real-time data integration and management
- Massive data handling capabilities
- Point-of-care decision support
- Heuristic Analytics
- Highly-reduced manual work
The road ahead
In no way will machine learning will ever replace the human torch in healthcare. It’s all about improving outcomes, enabling accurate care delivery and augmenting patient-centric care. If data from all over the place integrates into a single record, it will allow a clear understanding of patients and on a higher level- the population. If the results and outcomes are shared across the network in real-time and made available at the point of care, that will ensure faster decision making. If a physician can obtain results in a fraction of time, with an ideal degree of accuracy, that will ultimately improve patient care.
Someday, it will be common to have machine learning embedded in various processes to understand what’s going on with patients in real time. Sounds fairly futuristic. Innovaccer’s Care Intelligence SystemTM bridges this gap and brings the future to you, in the hopes that intelligent data-driven insights can improve the health of millions.
For more updates, Subscribe
If you want to see our efforts in the area, Schedule a quick demo
For further information about Datashop, Join Innovaccer team at Becker’s Hospital Review 3rd Annual Health IT + Revenue Cycle Conference, Booth #408.