How providers could achieve returns on data through AI
Healthcare organizations are under pressure to make better use of information and their IT infrastructures, but many key questions must be considered first.
Healthcare delivery organizations face stiff competition for business, and in light of that, administrators are focusing on strategies like cost containment, consumerism and vertical integration. Given significant recent investment in digitization, CIOs are increasingly being asked to obtain a return on those investments.
Even so, many organizations are still seeking the value that they expected with digitalization. One reason why is perhaps because they have not applied the capabilities that come with digitization to processes and services or products in healthcare. When that happens, it’s likely that healthcare will see the increased efficiency, consumer-centricity and more profitable business models that have benefited industries like airlines and banking.
Turning digitized data into value-added information sounds simple enough, but the complexities of healthcare data, government incentives and technology advancements have given rise to a relatively new sector of the health care IT industry: analytics and intelligence. Hundreds of companies are vying to help healthcare organizations become data-driven through artificial intelligence, machine learning and digital transformation.
I tend to think about the journey to becoming data-driven in simpler terms: as questions that inform the strategies needed to get the answers. Here are key questions to answer.
What happened, and why did it happen?
Describing what has occurred and diagnosing how it happened has been the work of informaticists, process engineers and data analysts for a long time. Reactive data is used to summarize history and inform shifts in clinical, financial and operational processes. Technology advancements in data warehouses have increased diagnostic capabilities, because data from multiple sources can be normalized, combined and summarized, providing incremental value and informing new insights. Optimization of processes based on history, both the good times and the rough times, is becoming easier with data accessibility and modern visualization tools.
What will happen next?
The complexities and consequences of the actions of patients and care providers make predicting the future difficult. That said, digitized data and the democratization of statistical analysis tools are enabling data scientists and analysts to identify trends across healthcare and the factors that either show causation or correlation with those trends. That’s helping predict the future based on historical data—particularly in health system operations, as the nexus of clinical, operational, situational and environmental data sets is leading to increased accuracy in the prediction of patient volumes, transitions of care and staffing needs. The predictive algorithms developed today are trained against real data sets with known outcomes. Advancements in machine learning capabilities are enabling algorithms to increase in accuracy based on outcomes and results, with limited to no human intervention.
What should I do about it?
There are a lot of definitions of artificial intelligence (AI), which has led the concept to become a buzzword for companies providing data-driven solutions. Many of those claiming to provide AI are really offering simple algorithms requiring human interventions. The definition of AI that I prefer is perhaps the simplest: It is “the capability of a machine to imitate intelligent human behavior.” The computer is analyzing as much data as needed, simulating an outcome and identifying the optimal path, ideally in near real-time. Still, in healthcare, imitating human behavior might not be good enough.
In 1950, medical knowledge doubled every 50 years. By 1980, it took 7 years for medical knowledge to double. It is estimated that by 2020, the doubling time will drop to 73 days. Global knowledge is rapidly outpacing the individual’s ability to assimilate and apply it in a timeframe that should be acceptable for patient care. AI applied to medical knowledge may expand the opportunities for physicians by linking the most recent research to the specific patient situation and recommending pathways and treatment options. I look forward to the day when AI becomes more than a marketing buzzword in health care: The opportunities for advancing patient care, access and quality are limitless.
How do I optimize for the future?
Analytics should inform organizational strategic decision making. The key is to build accurate assumptions in models that exist somewhere in the confluence between the outcomes of current daily decisions and the way future investments will affect those outcomes. In other words: One must simulate reality as much as possible to predict the future.
The tools used for simulation run the gamut between Microsoft Excel and expensive proprietary software. The tool is less important than the validity of the assumptions and the accurate understanding of how multiple variables work together to form an outcome. Simulation advancements combined with increased processing power are fueling AI’s ability to aid in decision making at the point-of-care delivery or in the moment decisions need to be made, as well as in the longer term, informing strategic decision making.
The application of these types of questions to problem statements could begin to build the near-and-next technology and process strategies for organizations as they seek to become data-driven. There are a couple of more important decisions to consider, however, as the journey to become data-driven is complex.
Consideration 1: The delivery vehicle for information matters
Should key information be delivered in a report that is auto-generated and sent, on a large screen that is always on, inside of workflow applications or via a text? The person receiving the information and the context in which they receive it matters. A data strategy for a centralized process only works if the work to centralize that process has been done. By contrast, if an organization is trying to drive federated decision making based on a set of the rules, it is better to consider delivery of that information inside of near or future-state workflows so as not to add burden to already busy staff.
Consideration 2: The form that information comes in is important to drive decisions
We all know how we like to receive information used in our daily work. What if, all of a sudden, that information doubled or tripled and was based on machine-driven intelligence? Information overload, relevance and targeted trend analysis must be considered as we think through smart alerting and visualization.
Consideration 3: Relevant and accurate insights come from clean data
The inputs to analytical tools and algorithms must be “clean” for the algorithms to gain new insights. It’s important that any organization on this journey has a strategy to ensure the data is validated for accuracy prior to incorporating advanced analytics in the workflow.
On the surface, it may seem like AI, machine learning and predictive intelligence are a distant future state in healthcare. However, right now, there is a tremendous amount of investment going into how information will be leveraged to increase care coordination, the quality of care, cost optimization and consumer-focused strategies.
Perhaps the nearest term, least-controversial opportunity for predictive analytics and AI in healthcare is to drive decision making aimed at optimizing care access and throughput across the continuum of care. In the U.S., like many countries, many have felt there is capacity in today’s healthcare infrastructure to ensure timely access for every member of the population. Daily decisions based on decision support tools and AI have the potential to improve throughput through optimization of care venues, transitions of care, discharge and turn-around times. If the capacity is in the system, perhaps more intelligent workflows are the key to unlocking that capacity.
This blog post originally published on Cerner.com. Click here to view the original content.
Even so, many organizations are still seeking the value that they expected with digitalization. One reason why is perhaps because they have not applied the capabilities that come with digitization to processes and services or products in healthcare. When that happens, it’s likely that healthcare will see the increased efficiency, consumer-centricity and more profitable business models that have benefited industries like airlines and banking.
Turning digitized data into value-added information sounds simple enough, but the complexities of healthcare data, government incentives and technology advancements have given rise to a relatively new sector of the health care IT industry: analytics and intelligence. Hundreds of companies are vying to help healthcare organizations become data-driven through artificial intelligence, machine learning and digital transformation. I tend to think about the journey to becoming data-driven in simpler terms: as questions that inform the strategies needed to get the answers. Here are key questions to answer.
What happened, and why did it happen?
Describing what has occurred and diagnosing how it happened has been the work of informaticists, process engineers and data analysts for a long time. Reactive data is used to summarize history and inform shifts in clinical, financial and operational processes. Technology advancements in data warehouses have increased diagnostic capabilities, because data from multiple sources can be normalized, combined and summarized, providing incremental value and informing new insights. Optimization of processes based on history, both the good times and the rough times, is becoming easier with data accessibility and modern visualization tools.
What will happen next?
The complexities and consequences of the actions of patients and care providers make predicting the future difficult. That said, digitized data and the democratization of statistical analysis tools are enabling data scientists and analysts to identify trends across healthcare and the factors that either show causation or correlation with those trends. That’s helping predict the future based on historical data—particularly in health system operations, as the nexus of clinical, operational, situational and environmental data sets is leading to increased accuracy in the prediction of patient volumes, transitions of care and staffing needs. The predictive algorithms developed today are trained against real data sets with known outcomes. Advancements in machine learning capabilities are enabling algorithms to increase in accuracy based on outcomes and results, with limited to no human intervention.
What should I do about it?
There are a lot of definitions of artificial intelligence (AI), which has led the concept to become a buzzword for companies providing data-driven solutions. Many of those claiming to provide AI are really offering simple algorithms requiring human interventions. The definition of AI that I prefer is perhaps the simplest: It is “the capability of a machine to imitate intelligent human behavior.” The computer is analyzing as much data as needed, simulating an outcome and identifying the optimal path, ideally in near real-time. Still, in healthcare, imitating human behavior might not be good enough.
In 1950, medical knowledge doubled every 50 years. By 1980, it took 7 years for medical knowledge to double. It is estimated that by 2020, the doubling time will drop to 73 days. Global knowledge is rapidly outpacing the individual’s ability to assimilate and apply it in a timeframe that should be acceptable for patient care. AI applied to medical knowledge may expand the opportunities for physicians by linking the most recent research to the specific patient situation and recommending pathways and treatment options. I look forward to the day when AI becomes more than a marketing buzzword in health care: The opportunities for advancing patient care, access and quality are limitless.
How do I optimize for the future?
Analytics should inform organizational strategic decision making. The key is to build accurate assumptions in models that exist somewhere in the confluence between the outcomes of current daily decisions and the way future investments will affect those outcomes. In other words: One must simulate reality as much as possible to predict the future.
The tools used for simulation run the gamut between Microsoft Excel and expensive proprietary software. The tool is less important than the validity of the assumptions and the accurate understanding of how multiple variables work together to form an outcome. Simulation advancements combined with increased processing power are fueling AI’s ability to aid in decision making at the point-of-care delivery or in the moment decisions need to be made, as well as in the longer term, informing strategic decision making.
The application of these types of questions to problem statements could begin to build the near-and-next technology and process strategies for organizations as they seek to become data-driven. There are a couple of more important decisions to consider, however, as the journey to become data-driven is complex.
Consideration 1: The delivery vehicle for information matters
Should key information be delivered in a report that is auto-generated and sent, on a large screen that is always on, inside of workflow applications or via a text? The person receiving the information and the context in which they receive it matters. A data strategy for a centralized process only works if the work to centralize that process has been done. By contrast, if an organization is trying to drive federated decision making based on a set of the rules, it is better to consider delivery of that information inside of near or future-state workflows so as not to add burden to already busy staff.
Consideration 2: The form that information comes in is important to drive decisions
We all know how we like to receive information used in our daily work. What if, all of a sudden, that information doubled or tripled and was based on machine-driven intelligence? Information overload, relevance and targeted trend analysis must be considered as we think through smart alerting and visualization.
Consideration 3: Relevant and accurate insights come from clean data
The inputs to analytical tools and algorithms must be “clean” for the algorithms to gain new insights. It’s important that any organization on this journey has a strategy to ensure the data is validated for accuracy prior to incorporating advanced analytics in the workflow.
On the surface, it may seem like AI, machine learning and predictive intelligence are a distant future state in healthcare. However, right now, there is a tremendous amount of investment going into how information will be leveraged to increase care coordination, the quality of care, cost optimization and consumer-focused strategies.
Perhaps the nearest term, least-controversial opportunity for predictive analytics and AI in healthcare is to drive decision making aimed at optimizing care access and throughput across the continuum of care. In the U.S., like many countries, many have felt there is capacity in today’s healthcare infrastructure to ensure timely access for every member of the population. Daily decisions based on decision support tools and AI have the potential to improve throughput through optimization of care venues, transitions of care, discharge and turn-around times. If the capacity is in the system, perhaps more intelligent workflows are the key to unlocking that capacity.
This blog post originally published on Cerner.com. Click here to view the original content.
More for you
Loading data for hdm_tax_topic #better-outcomes...