How AI and machine learning can help predict SDOH needs
DISCLAIMER: This blog is fully automated, unmonitored, and does not reflect the views of Garett MacGowan. The ML model may produce content that is offensive to some readers.
This blog post was generated with a (potentially) real article title as the only prompt. A link to the original article is below.
Photo by Scott Graham on Unsplash
Generated: 1/23/2022
How AI and machine learning can help predict SDOH needs
What does the future of public health look like? We all know that in the near future it will consist of many different technologies, but in the long run, a more sustainable future might be the one where artificial intelligence (AI), machine learning (ML) and predictive analytics (PA) play a more important and bigger role.
There are already several healthcare AI startups that are disrupting the industry. The biggest players have been AI’s biggest challenger in the industry since 2017, when IBM, Microsoft, Salesforce and Google were the leaders in developing AI for healthcare. IBM has also gone back to their previous successes after losing their lead in 2017. IBM Watson, created in 2010 by the company’s cognitive computing unit, took a step back from their previous success with Jeopardy! to create their cognitive healthcare unit in 2013. IBM spent four years trying to sell their vision of what “cognitive healthcare” could do for patients and doctors by showing how Watson could make healthcare systems much more productive. IBM recently announced the launch of Watson in Health™ to show that IBM plans to bring their technologies and solutions to healthcare across industries, including the healthcare industry.
These two companies are trying out AI in a variety of different ways. IBM is taking advantage of technologies such as Watson’s APIs, data, and analytics to bring together different data sets and machine learning algorithms to build AI that may be able to predict the next step in an emergency situation, such as an ER patient that has shown potential to cause harm to the patient or caregiver. On the other hand, IBM is leveraging the data sets that it acquired from healthcare tech startup Aledade in partnership with Aledade to combine Watson’s data sets with machine learning to power Watson in Health. Watson in Health can be used for billing, operations, quality and risk management as part of the IBM HealthCloud. This may be a good place to try out AI-driven solutions.
Microsoft, on the other hand, has been focused on its healthcare AI startup HealthFlow in a different way, where the company has built its own healthcare AI and predictive modeling engine inside of its Azure cloud platform. One of the key features of this platform is that it makes it possible to bring AI into any kind of domain by using Microsoft Data Science Virtual Machines (DSVMs) to bring in and train these machines with data sets that are stored in the Azure SQL Data Lake. An example that Microsoft is utilizing in the healthcare market is their own healthcare AI project called, Bipartite Attention (BA). BA builds on top of AI technologies from Microsoft’s previous healthcare projects like MSR Infer.com, which uses AI to explore possible biological mechanisms leading to various diseases by using textual analysis of PubMed studies and public health data sets like the CDC’s U.S. Mortality report and U.S. Vital Statistics report. HealthFlow uses the same BA AI engine that powers and develops healthcare AI technologies. Microsoft has brought in and trained ML and predictive modeling capabilities from BA to power their Healthcare Data Science Virtual Machine (HCDSVM) in HealthFlow. As an added benefit to using BA as a training engine, their ML and predictive modeling can be trained using a variety of health data sources, which is not a possible option if using an AI engine like Jeopardy’s AI.
Aledade and Amazon’s AWS are also using their own unique approach toward AI in the healthcare space, where the company has decided to go with a smaller, more specialized form of AI to power different functions in the company’s new Healthcare Artificial Intelligence Platform. Aledade is focused on creating AI that may one day be able to tell someone if they have a serious condition like a disease or other condition that needs immediate attention from the doctor. The company has built their AI using deep neural networks (DNN). DNN is a form of AI that can make use of different types of data to create predictive models. In this case, Aledade is using DNN to train themselves what kind of diseases and conditions show possible signs when someone has stopped taking their daily medication. Aledade has already had success in combining this AI with a device that is connected to the medical device data that is sent through the internet to build a healthcare AI that can be used in the future to create a form of automated AI that can be placed as a medical device at home to send alerts to family members when a patient stops taking their medicine. This may bring personalized medicine to the home, providing a sense of safety to families who may not be able to stay home with their patients.
Amazon’s approach to healthcare AI comes at it from the business side, where the company is focusing on being a data provider for AI and machine learning companies. Amazon Web Services (AWS), of course, is one of the most well known business services that the company offers. Its data centers are well known as being one of the best places that companies may choose to store their data sets, and being one of the first to market with its private cloud data centers enabled by AI.
Amazon is working with the three companies they founded, Batch.ai, Fauna, and Lumin, with their own healthcare AI platform called Manticore for building AI. Manticore is different than the other three because it is focused on healthcare data sets. This is where Amazon is hoping to have its success with AI and predictive analytics. AWS is offering a wide variety of data sets that other companies may not be able to use. They have also decided to use a smaller form of deep learning to build AI that will be easier to deploy for clinical decision support. Amazon’s vision is to provide AI and PA on the massive scale of its business where the company may not be able to provide the best healthcare solutions for patients if they could only rely on AI technologies from a single company when the market is trying to provide more than 50+ different companies that may provide AI based technologies that can be used in the healthcare system.
The companies listed above are all working in the healthcare AI space, but what about the other more traditional use cases for artificial intelligence? How are they leveraging AI and machine learning across their business lines?
We are at an all-time high state of AI adoption across multiple industries such as finance, marketing, retail, agriculture, insurance, and so on. But what about healthcare? Yes, AI has been very successful at predicting many things about our lives; from the weather to what we will get at the grocery store, to what our phone may be doing in the middle of the night, AI is everywhere, and we are all using it everyday. But healthcare? That is a whole new world altogether. Healthcare AI adoption rates have seen a dip according to IDC, from a high of 34% in the U.S., down to 22% in 2018. According to a study done in the U.K. by Accenture and WMG Consultants titled, "A Roadmap for Artificial Intelligence in Healthcare," the study found that, “Although adoption of AI in healthcare is on the rise, there are considerable knowledge barriers preventing AI practitioners from getting the confidence that they require from data scientists and data scientists from other disciplines to make intelligent advances in their daily work.”
Many healthcare executives that are implementing AI within their businesses have started to feel frustrated by having to find solutions that only make sense on a case-by-case basis. The reason for this is that there is typically no single-purpose technology that only does one thing. Each form of AI requires an understanding and knowledge of different kinds of data and analytics, and each requires the use of data sets and analytic techniques from different industries and fields of experience. AI technologies can only be successful if they have access to the right data sets to train their algorithms, and AI technologies that have not been able to find these right type of data sets cannot use AI to solve any complex problems that come to them.
What is driving the decline in AI adoption rate in the healthcare industry? Well, there have been three major barriers that have prevented many AI solutions for healthcare from being developed in spite of the availability of AI, PA and analytics. The first three challenges that have prevented the adoption of AI in the healthcare industry are:
Data is not Big enough
Big data, analytics, and AI technologies rely on a lot of data set that are not typically available within healthcare for the reason above. This is one of the reasons that many AI technologies that companies such as Aledade are using the smaller forms of data and analytics to build their AI because the data that would normally be required for them is either not available or too expensive to use for health IT purposes.
Data standards or integration standards are expensive
According to McKinsey & Company, “Healthcare is a $3 trillion industry that is fragmented and inefficient, with only a few of the large players controlling a substantial part of the value. As a result, healthcare spend is anemic relative to its potential and its results are poor relative to the amount of money invested. Healthcare needs to be digitized and standardized. We show how companies at the center of this digital transformation could generate $22 trillion over the next decade through an integrated customer approach.” The most important aspects of digitize-and-standardize are standards and data standards. Because of the lack of these standards, most AI technologies are at best being able to run at 60%, meaning that in most cases they are not working as expected. There are also companies that have chosen not to use these standards unless they have full market control over the data that they own, which isn’t realistic because other companies have access to that same data. A good example of this is AI.js, a common JavaScript library that has allowed developers to use AI at scale. AI.
What does the future of public health look like? We all know that in the near future it will consist of many different technologies, but in the long run, a more sustainable future might be the one where artificial intelligence (AI), machine learning (ML) and predictive analytics (PA) play a more important and bigger role.
There are already several healthcare AI startups that are disrupting the industry. The biggest players have been AI’s biggest challenger in the industry since 2017, when IBM, Microsoft, Salesforce and Google were the leaders in developing AI for healthcare. IBM has also gone back to their previous successes after losing their lead in 2017. IBM Watson, created in 2010 by the company’s cognitive computing unit, took a step back from their previous success with Jeopardy! to create their cognitive healthcare unit in 2013. IBM spent four years trying to sell their vision of what “cognitive healthcare” could do for patients and doctors by showing how Watson could make healthcare systems much more productive. IBM recently announced the launch of Watson in Health™ to show that IBM plans to bring their technologies and solutions to healthcare across industries, including the healthcare industry.
These two companies are trying out AI in a variety of different ways. IBM is taking advantage of technologies such as Watson’s APIs, data, and analytics to bring together different data sets and machine learning algorithms to build AI that may be able to predict the next step in an emergency situation, such as an ER patient that has shown potential to cause harm to the patient or caregiver. On the other hand, IBM is leveraging the data sets that it acquired from healthcare tech startup Aledade in partnership with Aledade to combine Watson’s data sets with machine learning to power Watson in Health. Watson in Health can be used for billing, operations, quality and risk management as part of the IBM HealthCloud. This may be a good place to try out AI-driven solutions.
Microsoft, on the other hand, has been focused on its healthcare AI startup HealthFlow in a different way, where the company has built its own healthcare AI and predictive modeling engine inside of its Azure cloud platform. One of the key features of this platform is that it makes it possible to bring AI into any kind of domain by using Microsoft Data Science Virtual Machines (DSVMs) to bring in and train these machines with data sets that are stored in the Azure SQL Data Lake. An example that Microsoft is utilizing in the healthcare market is their own healthcare AI project called, Bipartite Attention (BA). BA builds on top of AI technologies from Microsoft’s previous healthcare projects like MSR Infer.com, which uses AI to explore possible biological mechanisms leading to various diseases by using textual analysis of PubMed studies and public health data sets like the CDC’s U.S. Mortality report and U.S. Vital Statistics report. HealthFlow uses the same BA AI engine that powers and develops healthcare AI technologies. Microsoft has brought in and trained ML and predictive modeling capabilities from BA to power their Healthcare Data Science Virtual Machine (HCDSVM) in HealthFlow. As an added benefit to using BA as a training engine, their ML and predictive modeling can be trained using a variety of health data sources, which is not a possible option if using an AI engine like Jeopardy’s AI.
Aledade and Amazon’s AWS are also using their own unique approach toward AI in the healthcare space, where the company has decided to go with a smaller, more specialized form of AI to power different functions in the company’s new Healthcare Artificial Intelligence Platform. Aledade is focused on creating AI that may one day be able to tell someone if they have a serious condition like a disease or other condition that needs immediate attention from the doctor. The company has built their AI using deep neural networks (DNN). DNN is a form of AI that can make use of different types of data to create predictive models. In this case, Aledade is using DNN to train themselves what kind of diseases and conditions show possible signs when someone has stopped taking their daily medication. Aledade has already had success in combining this AI with a device that is connected to the medical device data that is sent through the internet to build a healthcare AI that can be used in the future to create a form of automated AI that can be placed as a medical device at home to send alerts to family members when a patient stops taking their medicine. This may bring personalized medicine to the home, providing a sense of safety to families who may not be able to stay home with their patients.
Amazon’s approach to healthcare AI comes at it from the business side, where the company is focusing on being a data provider for AI and machine learning companies. Amazon Web Services (AWS), of course, is one of the most well known business services that the company offers. Its data centers are well known as being one of the best places that companies may choose to store their data sets, and being one of the first to market with its private cloud data centers enabled by AI.
Amazon is working with the three companies they founded, Batch.ai, Fauna, and Lumin, with their own healthcare AI platform called Manticore for building AI. Manticore is different than the other three because it is focused on healthcare data sets. This is where Amazon is hoping to have its success with AI and predictive analytics. AWS is offering a wide variety of data sets that other companies may not be able to use. They have also decided to use a smaller form of deep learning to build AI that will be easier to deploy for clinical decision support. Amazon’s vision is to provide AI and PA on the massive scale of its business where the company may not be able to provide the best healthcare solutions for patients if they could only rely on AI technologies from a single company when the market is trying to provide more than 50+ different companies that may provide AI based technologies that can be used in the healthcare system.
The companies listed above are all working in the healthcare AI space, but what about the other more traditional use cases for artificial intelligence? How are they leveraging AI and machine learning across their business lines?
We are at an all-time high state of AI adoption across multiple industries such as finance, marketing, retail, agriculture, insurance, and so on. But what about healthcare? Yes, AI has been very successful at predicting many things about our lives; from the weather to what we will get at the grocery store, to what our phone may be doing in the middle of the night, AI is everywhere, and we are all using it everyday. But healthcare? That is a whole new world altogether. Healthcare AI adoption rates have seen a dip according to IDC, from a high of 34% in the U.S., down to 22% in 2018. According to a study done in the U.K. by Accenture and WMG Consultants titled, "A Roadmap for Artificial Intelligence in Healthcare," the study found that, “Although adoption of AI in healthcare is on the rise, there are considerable knowledge barriers preventing AI practitioners from getting the confidence that they require from data scientists and data scientists from other disciplines to make intelligent advances in their daily work.”
Many healthcare executives that are implementing AI within their businesses have started to feel frustrated by having to find solutions that only make sense on a case-by-case basis. The reason for this is that there is typically no single-purpose technology that only does one thing. Each form of AI requires an understanding and knowledge of different kinds of data and analytics, and each requires the use of data sets and analytic techniques from different industries and fields of experience. AI technologies can only be successful if they have access to the right data sets to train their algorithms, and AI technologies that have not been able to find these right type of data sets cannot use AI to solve any complex problems that come to them.
What is driving the decline in AI adoption rate in the healthcare industry? Well, there have been three major barriers that have prevented many AI solutions for healthcare from being developed in spite of the availability of AI, PA and analytics. The first three challenges that have prevented the adoption of AI in the healthcare industry are:
Data is not Big enough
Big data, analytics, and AI technologies rely on a lot of data set that are not typically available within healthcare for the reason above. This is one of the reasons that many AI technologies that companies such as Aledade are using the smaller forms of data and analytics to build their AI because the data that would normally be required for them is either not available or too expensive to use for health IT purposes.
Data standards or integration standards are expensive
According to McKinsey & Company, “Healthcare is a $3 trillion industry that is fragmented and inefficient, with only a few of the large players controlling a substantial part of the value. As a result, healthcare spend is anemic relative to its potential and its results are poor relative to the amount of money invested. Healthcare needs to be digitized and standardized. We show how companies at the center of this digital transformation could generate $22 trillion over the next decade through an integrated customer approach.” The most important aspects of digitize-and-standardize are standards and data standards. Because of the lack of these standards, most AI technologies are at best being able to run at 60%, meaning that in most cases they are not working as expected. There are also companies that have chosen not to use these standards unless they have full market control over the data that they own, which isn’t realistic because other companies have access to that same data. A good example of this is AI.js, a common JavaScript library that has allowed developers to use AI at scale. AI.