The Real-World Benefits of Machine Learning in Healthcare

The real-world benefits of machine learning in healthcare

It is safe Medicine has many manual procedures. During the training, I hand wrote lab values, diagnoses, and other chart notes on paper. I always knew this was the place where technology could help improve my workflow and hopefully it also improves patient care. Since then, the progress of electronic medical records has been remarkable, but the information they provide is no better than replacing old paper charts. If technology is to improve care in the future, then the electronic information provided to doctors must be enhanced by the power of analytics and machine learning.

Using these types of advanced analytics, we can provide better information to doctors on the point of patient care. With easy access to blood pressure and other important signals, I have seen my patients become regular and expectant. Imagine how much more useful it would be if I could read the last 500 blood pressure readings, laboratory test results, race, gender, family history, socioeconomic status, and recent clinical trial data for my patient's stocks, coronary artery disease, and kidney failure.

We need to provide more information to clinicians so that they can make better decisions about the patient's diagnosis and treatment options while understanding the potential consequences and costs for each. The value of machine learning in healthcare is the ability to process huge datasets far beyond the scope of human capability, and then transform the analysis of data into clinical insights that help physicians plan and provide care, ultimately leading to better results, at a lower cost. Care, and increased patient satisfaction.

Machine learning implemented in healthcare

Machine learning in medicine has just made headlights. Google has developed a machine-learning algorithm to help identify cancerous tumors on mammograms. Stanford is using an in-depth learning algorithm to identify skin cancer. A recent JAMA article reported the results of an in-depth machine-learning algorithm that was able to diagnose diabetic retinopathy in retinal images. It is clear that machine learning puts another arrow in the clinical decision-making process.

Still, machine learning lends itself to some of the best processes. Algorithms can provide immediate benefits in discipline with reproductive or standardized processes. Also, those with large image datasets, such as radiology, cardiology, and pathology, are tough candidates. Machine learning can be trained to look at images, identify abnormalities, and show areas that need attention, which improves the accuracy of all these processes. In the long run, machine learning benefits family entrepreneurs or bedside interns. Machine learning can offer an objective opinion to improve efficiency, reliability, and accuracy.

At Health Catalyst, we use a proprietary platform to analyze data and loop it in real-time to help physicians make clinical decisions. At the same time a physician sees the patient and enters the symptoms, data, and test results into the EMR, the behind-the-scenes machine looks at everything about that patient, and prompts the doctor with useful information for the diagnosis, ordering the test. , Or a prevention screening suggestion. In the long run, the capabilities will reach all aspects of medicine as we get more usable, better-integrated data. We are able to contain large sets of data that can be analyzed and compared in real-time to provide all kinds of information to the provider and the patient.

Ethics of using algorithms in healthcare

It has been said before that the best machine learning tool in healthcare is the brain of the doctor. Can physicians there view McCain Learning as an unnecessary second opinion? At one point, autoworkers feared robotics would take their jobs. Similarly, there may be physicians who fear the beginning of a machine learning process that may present them obsolete. But this is the art of medicine that can never be changed. Patients always need a humane touch, and a loving and kind relationship with the people who care for them. Neither machine learning, nor any future technologies in medicine will eliminate it but will become the tools that clinics use to improve ongoing care.

The focus should be on how to use machine learning to enhance patient care. For example, if I'm testing a patient for cancer, then I want the highest-quality biopsy results I can possibly get. A machine learning algorithm that can review pathology slides and assist pathologists with diagnosis is valuable. If I can get results with an accuracy of a variation of time, then, ultimately it will improve patient care and satisfaction (this is because my own mother is anxiously waiting for her test because of weeks).

Healthcare needs to move from thinking of machine learning as a concept of the future to seeing it as a real-world tool that can be used today. If machine learning is to play a role in healthcare, we need to take an incremental approach. We must find specified use cases in which the capabilities of machine learning provide value from a specific technical application (e.g. Google and Stanford). It is a step-by-step approach to incorporating daily analytical, machine learning, and prediction algorithms into daily clinical practice.

Initially, our goals need to match our capabilities. Training a machine-learning algorithm to identify skin cancer from a large set of skin cancer images is something that most people understand. If we learn that algorithms are changing radiologists, then people will easily hesitate. It should be a bridge over time. Radiologists are never obsolete, but future radiologists will do the observation and review work that the machine initially read. They employ machine learning such as collaborative partners that identify specific areas of attention, illuminate noise, and help focus on high-potential areas of interest.

How do we get to the point where we need to believe in machine learning? Medical research and treatment is a method to prove that they are safe and effective. It is a long process of trial and error - and decision-based on evidence. We need similar processes in place as we look at machine learning to ensure its safety and effectiveness. We need to understand the ethics involved in handing over part of what we do in a machine.

A “What if” scenario on the potential of machine learning

A few months ago, I gave a presentation on the future of analytics and its potential impact on clinical care. On my slides, I was running a hypothetical EMR running prediction algorithm while a doctor was examining his patient. A pop-up box showcased real-time diagnosis, pathological outcomes, and treatment options, as well as the potential effectiveness of each option and the cost for this patient.

While the patient in this condition may be imaginary, it was sampled after my father passed away from prostate cancer many years ago. I chose this scenario to demonstrate the results that machine learning would have been available at that time.

In my father's real case, his doctor initially allowed him to live for two years. It was based on a combination of the doctor's experience with similar patients and the treatment options available at the time. Didn't he know that I would take an active role in overseeing my father's care? I researched clinical trials and new treatment options. I studied the side effects and test results. Did the treatment make the person live longer? If so, was it a few weeks, a few months, or longer?

My father was a great oncologist, but he cared for thousands of patients with different types of cancer. He never put the time and effort he needed to learn all the new drugs and treatment options that would come out for all these cancers. Many times, I presented treatment options and clinical trials that my father's doctor did not know.

With my years of training and expertise, I can cull literature and recommend the best option for my dad. In other words, I'm the human algorithm, the doctor's mind, which I meant, and most importantly, the motivation to work in concert with my father's doctor and to develop an optimal plan for the time, which eventually extended my father's life to nine years.

With the analytics platform and machine learning running in the background, the human algorithm - the extra layer of back-up therapist - is not needed. The analytics engine will have unlimited data that any one person can process. It will be with a library of patients like my father, with his diagnosis and tissue type. It will provide treatment options with predictions of how effective they will be, mortality, side effects, and cost. Despite all the efforts by human carers, the analytics platform can work unlimitedly behind the scenes and deliver decisive information to physicians in real-time.

Data drive machine education

The more data available, the better information we have available to patients. Predictive algorithms and machine learning provide a good mortality prediction model that doctors can use to educate patients.

But machine learning requires some amount of data to generate an effective algorithm. Most machine learning initially comes from organizations with large datasets. Health Catalyst is developing Collective Analytics for Excellence (CAFÉ ™), an application built into application data warehouses (EDWs) and national de-identification repositories of health care data from third-party data sources. It enables comparative effectiveness, research, and the production of unique, powerful machine learning algorithms. CAFÉ is a great collaboration between our healthcare system partners.

Another possibility for small organizations is their ability to merge their data into larger systems. At some point, we may see regional data hubs with datasets optimized for geographic, environmental, and socioeconomic factors, giving health systems of all sizes access to more data.

As larger datasets begin to drive machine learning, we can improve care in a specific way for each area. And considering rare diseases with low data volumes, it should be possible to merge regional data into national sets to measure the segment required for machine learning.

Move machine learning from theory to clinical reality

We've seen machine learning applications in healthcare that are taking medicine to new heights. It's fun to think about where it can go. Someday, it is common to have embedded machine learning expertise that analyzes not only what is happening to patients in real-time, but what is happening to similar patients in a multi-healthcare system, what clinical trials are underway, and the effectiveness and cost of new treatment options. This may sound like the future, but the analytics engine that can present all this information in place of care.

Machine learning: life-saving technology that transforms healthcare

Health Catalyst believes that machine learning is a life-saving technology that transforms healthcare.

This technology challenges traditional and reactionary approaches to healthcare. In fact, it is the exact opposite: predictive, proactive, and preventive - life-saving qualities that make it a critically necessary ability in every health system.

Health catalysts to save lives by creating machine learning activities to help the health system, through actionable, and comprehensive through (models built into every health catalyst application), healthcare.y free (free, open-source software), and its Healthcare Analytics Platform (JUG).

There are endless opportunities for machine learning in healthcare

One may question whether this is just a technology or whether it offers the right value in healthcare. Health Catalyst believes that the introduction and widespread use of machine learning in healthcare will be one of the most important, life-saving technologies ever launched. We believe that the opportunities to improve technology and accelerate clinical, workflow, and financial outcomes are limitless. Here are some examples:

Reduce reading. Machine learning can reduce reading in a targeted, effective, and patient-centered way. Clinicians can receive daily guidance on which patients are most likely to be referred and how they are able to reduce that risk.

Prevent hospital-acquired infections (HAIs). The health system can reduce HAIs, such as Central Line Related Blood Transfusion Infections (CLABSIs) - CLBSI patients die by percent0 percent - Central line patients predict CLABSI development. Clinicians can monitor high-risk patients and intervene to reduce that risk by focusing on patient-specific risk factors.

Reduce hospital length-stay (LOS). The health system can reduce LOS and improve other outcomes by identifying patients as patients who are more likely to have increased LOS and then ensuring that best practices are followed.

Predict chronic disease. Machine learning helps hospital systems identify patients with undiagnosed or misdiagnosed chronic disease, predicts the likelihood of patients developing chronic disease, and presents patient-specific preventive interventions.

Reduce 1-year mortality. The health system can reduce mortality by 1 year by predicting the likelihood of death within one year of discharge and reconciling patients with appropriate patients, care providers, and support.

Prediction-of-payment. Health systems can determine who needs reminders, who need financial support, and how the likelihood of payment changes over time and special events.
Predict a no-show. The health system can create scheduled scheduling models, with each scheduled appointment, improving the risk of a no-show, ultimately patient care, and efficient use of resources.

Predictive from reactive: how machine learning saves lives

Many health organizations across the country have begun to improve outcomes and save lives by partnering with Health Catalyst and using its Catalyst.Y-powered analyzer.

Example: Using machine learning to reduce hospital-acquired infections

Health Catalyst Machine Education has been deployed on many customer sites with promising results. A recent article published by the Hospital and Health Network shares the perspectives and results of an integrated health system that is using analysts to predict its CLABSI prevention initiatives. They want to be able to predict which patients are more likely to have CLABSI so clinicians can intervene more quickly and provide maximum care for her or the patient.

They use Health Catalyst Data Warehouse (EDW) and analysis to bridge information gaps in its EMR data and develop a complete picture of patients' CLABSI risk. Models in the data were developed and tested using algorithms such as Logistic Regression and Irregular Jungle the Work Horse of the Machine Learning World. This led to the development of a CLABSI risk forecasting model built into a unit-level dashboard that nurses use to identify patient-level care gaps. In addition to the risk score, the top three risk factors are displayed for each high-risk patient, providing immediate insight into specific actions that may reduce CLABSI risk for those patients.

According to a recent report by the Scottsdale Institute on this project, the risk-forecasting tool - which consumes daily EHR - has successfully predicted CLABSI from 85 percent to 0 percent.

The results in the first six months were impressive, including:

. 87% predictive accuracy with only one false positive rate.

20% reduction in CLABSIs.

Harm0% overall in loss events.

Problem: Machine education is not regular in healthcare

Machine education is part of the daily lives of many Americans, from navigation applications to Internet shopping, and is widely used in other industries, such as retail and banking. But it is not routine in healthcare because of the complexity and limited availability of data - the lack of access to highly skilled data scientists and teams - to make meaningful improvements to the data.

Solution: Create machine learning routines, actionable, and comprehensive in healthcare

The Health Catalyst has three main solutions for broadening technology routines, procedures, and healthcare:

1.— our model and strategy for building machine learning in each health catalyst product. machine learning models are appearing in applications across all health catalyst product lines and our strategy is to build models on each single health catalyst product.

The health catalyst is a results company, so we focus on how the models are used, what kind of workflow they affect, and what decisions they make. When we developed the predictive model for Central Line-Associated Blood Transfusion Infections (CLABSIs), it was not enough to just show which patients were at higher risk for CLABSI development. The forecast was displayed along with modifiable factors that were advancing the risk, giving the nurses procedural information on how to reduce the risk.

In addition to our CLABSI risk forecasting model, Health Catalyst has taken advantage of the catalyst to modify numerous predictive models to support clinical, financial, and operational decision making:

Health Catalyst has developed more than a dozen prognostic models to support clinical, financial, and operational decision making.


CHF reading risk

Diabetes future risk

Diabetes pure potential complication

Forecast spent but no claim/year-end expense reported

Possibility to pay the risk

Meetings are no risk

There are more predictive models of health catalysts planned for the first quarter of 2017:

Clinical and operational decision support for bowel surgery

Hip and knee surgery

Heart disease

Geospatial network referrals and leakage prevention

Early detection of CAUTI and sepsis

Expected mortality rate and length of stay
And more.

2. Health care. ai Our way of encouraging the adoption of machine learning in healthcare.

Healthcare.EI is a community resource for education and support in machine learning in healthcare, weekly blogs, and broadcasts where users can get to know and interact with the Health Catalyst Data Science team. It also includes software packages that automate tasks and reduce access barriers when adopting this technology. Healthcare.EE is free and open-source because we believe the faster we can get the system to adopt machine learning, the faster they can use analytics to improve care and save lives. Healthcare.E.E. With, the health system has people right now to do the data. There are three main benefits of Healthcare.EE:

Free access to machine learning educational and support resources.

Free, open-source machine learning software that democratizes machine learning by reducing barriers to entry.

Makes deployment easy and ensures product quality.

Health. Healthcare Analytics Platform - Our platform is not the second most complex machine learning problem.

The effectiveness of is closely linked to the proven ability to integrate the high-data data available from almost all internal and external sources of health catalyst. Because multiple sources of data are needed to develop models for machine learning that drives predictive analytics, the more effective this technology is the more data there is.

Many companies that offer machine learning solutions require customers to figure out how to connect more than 100 different data sources in order for the technology to work. In contrast, Health Catalyst's Healthcare Analytics platform integrates 120 different data sources, including electronic health records (EHR), claims, financial, operational, and patient satisfaction systems.

Healthcare providers know how to apply technology to deliver results

Most importantly, Health Catalyst combines these three techniques with our proven improvement methods to produce meaningful results improvement results.

Many existing machine learning solution vendors offer educationally appealing, standalone models without an understanding of how they can be translated into meaningful, scalable results. As a result, there are some concrete examples of improvements in comprehensive machine-learning collaborative outcomes in healthcare.

Health catalysts, on the other hand, know how to improve outcomes. We know how physicians use data to make decisions. We understand the context in which machine learning insights must be understood and the exact timing and modality must be understood. That knowledge is built on, scaling machine learning, and results from improvement for use by virtually any organization. The bottom line is that a health system can save more lives and care more while saving money at the same time.