If you read much about technology, you have likely heard about machine learning, but may be wondering how it would work in healthcare. Where’s the low-hanging fruit? And how could it help my clinical team?
Throughout healthcare, and many other industries, there are heuristics and established best practices that help people make decisions. A popular example in healthcare is the LACE index, which provides the likelihood of patient 30-day readmission risk. You might have also heard of similar tools like the SOFA Score, Apgar Score, PRISM Score, and the PIM Score.
Like most of these scores, the LACE calculation is fairly simple. It’s based on length of stay, acuity of the admission, patient comorbities, and ED visits within the last six months. In each of these categories, points are assigned—a length of stay of three days equals three points, for example. Then the points from each categories are added up to form the LACE index.
It’s simple and indicative of how healthcare has worked for the last 20-30 years. First, there’s a national study, which eventually leads to guidelines and a simple calculation to help prioritize which patients are most at risk of something.
So what’s wrong with that? Well, the guidelines can only be narrowly applied and even then don’t give impressive results. Think of it—LACE was developed from patients seen in Ontario from 2004 to 2008. Do your patient demographics closely match those in Ontario? Or, do your patient demographics even match your same set from ten years ago? Perhaps not. Another issue is applicability—since LACE requires the patient’s length of stay, the score is only available upon discharge. What if you want a risk score early during their stay?
This is why machine learning is fantastic—it fills these gaps. First, it learns the important relationships in your data on past patients and their outcomes. This means that the model is customized on your data from the last few years–you don’t have to rely on scores made on other populations, 10-20 years ago. Second, machine learning allows you to create a model based on whatever data is available when you need a risk score (i.e., upon admission rather than discharge).
In summary, what does a machine learning model provide? Accurate, timely risk scores, enabling confident and precise resource allocation, leading to lower costs and improved outcomes. As an added bonus, healthcare.ai shows why a risk score was high, so the clinician not only knows which patients are most at risk, but also what can be done to lower that patient’s risk. We’ll detail this ability in a future post.
Thanks, and please reach out with any questions or comments!
We started healthcare.ai in late 2016 to bring machine learning (ML) to the healthcare masses. As we release version 2.0 of the software (on April 20th), it’s worth stepping back to fully understand why we invest in this open-source project, which is freely available to all. Why would a for-profit firm spend time investing in this public good? Since the 2009 HITECH act incentivized EHR adoption, data has become much more ubiquitous in healthcare. Despite all that’s gone wrong in US healthcare, the fact that healthcare data is…
Many vendors deliver machine learning models with different applications in healthcare. But they don’t all deliver accurate models that are easy to implement, targeted to a specific use case, connected to actionable interventions, and surrounded by a machine learning community and support team with extensive, exclusive healthcare experience.
These machine learning qualities are possible only through a machine learning model delivered by a vendor with a unique set of capabilities. There are five differentiators behind effective machine learning models and vendors:
Vendor’s expertise and exclusive focus on healthcare.
Machine learning model’s access to extensive data sources.
Machine learning model’s ease of implementation.
Machine learning model’s interpretability and buy-in.
Machine learning model’s conformance with privacy standards.
These five factors separate the high-value vendors and models from the crowd, so healthcare systems can quickly implement machine learning and start seeing improvement results.
tl;dr: Healthcare needs practical machine learning tools; the focus on deep learning and GPUs doesn’t help the average health system. Background Google just released a paper called “Scalable and accurate deep learning for electronic health records” that has received deserved acclaim in both the machine learning (ML) and healthcare communities. This research comes from the Google Brain group and isn’t their first foray into healthcare. See, for example, their impressive work in diabetic retinopathy. In fact, it’s now common for tech giants to wade into…
Last summer we discussed the simplified interface of the 1.0 CRAN release of healthcare.ai-R, and we’re now thrilled to demo new features related to clinician guidance in the 1.2 version. We’re calling this Patient Impact Predictor (PIP). Understanding an ML model This week we’d like to highlight new functionality that allows one to go a step beyond surfacing predictions to also surface targeted interventions. Risk scores are a great first step, but prescriptive guidance is where the results of machine learning (ML) may actually catch up to the…
Subscribe and get updates delivered to your email.
This project was started by and receives ongoing support from Health Catalyst.