Uncategorized

Predictive Analytics in Healthcare

Predictive analytics and machine learning in healthcare are rapidly becoming some of the most-discussed, perhaps most-hyped topics in healthcare analytics. Machine learning is a well-studied discipline with a long history of success in many industries. Healthcare can learn valuable lessons from this previous success to jump start the utility of predictive analytics for improving patient care, chronic disease management, hospital administration, and supply chain efficiencies. The opportunity that currently exists for healthcare systems is to define what “predictive analytics” means to them and how can it be used most effectively to make improvements.

However, predictions made solely for the sake of making a prediction are a waste of time and money. In healthcare and other industries, prediction is most useful when that knowledge can be transferred into action. The willingness to intervene is the key to harnessing the power of historical and real-time data. Importantly, to best gauge efficacy and value, both the predictor and the intervention must be integrated within the same system and workflow where the trend occurs.

The following Health Catalyst® paper, “Using Predictive Analytics in Healthcare: Technology Hype vs Reality” is a good summary of both the hype and hope of predictive analytics in healthcare

 

How To Get Started With Predictive Analytics and Machine Learning

Given the many pitfalls to avoid in healthcare predictive analytics, then where do you get started?  The most important starting point is to establish a fundamental data and analytic infrastructure upon which to build. Deliberately but quickly move your organization up the levels of the Healthcare Analytics Adoption Model. This model draws upon lessons learned from the HIMSS EHR Adoption Model and describes a similar approach for assessing the adoption of analytics in healthcare. This model starts a level 1 foundation of an integrated, enterprise data warehouse combined with a basic set of foundational and discovery analytic applications.

1.  Start With an Integrated Data Warehouse and Analytics Platform

gearsEnterprise Data Warehouse

You need data across the entire continuum of care to manage patient populations. This requires an enterprise data warehouse (EDW) platform. An EDW is the central platform upon which you can build a scalable analytics approach to systematically integrate and make sense of the data.

Health Catalyst® deploys a unique Late-Binding™ Data Warehouse that enables healthcare organizations to automate extraction, aggregation, and integration of clinical, financial, administrative, patient experience, and other relevant data and apply advanced analytics to organize and measure clinical, patient safety, cost, and patient satisfaction processes and outcomes.

2.  Use the Three Basic Steps of Predictive Modeling

predictive-modeling

The following is a simple schematic of the predictive modeling process.  For predictive analytics to be effective, Lean practitioners must truly “live the process” to best understand the type of data, the actual workflow, the target audience and what action will be prompted by knowing the prediction.

  1. The first step is to carefully define the problem you want to address, then gather the initial data necessary and evaluate several different algorithm approaches.
  2. Step two refines this process by selecting one of the best performing models and testing with a separate data set to validate the approach.
  3. The final step is to run the model in a real world setting.

The more specific term is prescriptive analytics, which includes evidence, recommendations and actions for each predicted category or outcome.  Specifically, prediction should link carefully to clinical priorities and measurable events such as cost effectiveness, clinical protocols or patient outcomes. Finally, these predictor-intervention sets are best evaluated within that same data warehouse environment.

So many options exist when it comes to developing predictive algorithms or stratifying patient risk. This presents a daunting challenge to health care personnel tasked with sorting through all the buzzwords and marketing noise.  Healthcare providers need to partner with groups that have a keen understanding of the leading academic and commercial tools, and the expertise to develop appropriate prediction models.

Follow 4 Key Lessons Learned for Adopting Predictive Analytics and Machine Learning in Healthcare

Given that predictive analytics are listed as level 7 out of the 8 possible levels on the Healthcare Analytics Adoption Model, there are many keys and pitfalls that can occur at such a level if not properly prepared. Fortunately for healthcare, there are numerous existing models from other industries that can be combined with past healthcare examples to ease some of the potential pains and pitfalls. Highlights of some those key lessons include:

  1. Don’t confuse more data with more insight: While many solid scientific findings may be interesting, they do little to significantly improve current clinical outcomes.
  2. Don’t confuse insight with value: While many solid scientific findings may be interesting, they do little to significantly improve current clinical outcomes.
  3. Don’t overestimate the ability to interpret the data: Sometimes even the best data may afford only limited insight into clinical health outcomes.
  4. Don’t underestimate the challenge of implementation: Leveraging large data sets successfully requires a health system to be prepared to embrace new methodologies; this, however, may require a significant investment of time and capital and alignment of economic interests.

The following Health Catalyst Executive Report, “4 Essential Lessons for Adopting Predictive Analytics in Healthcare”  expounds more in detail around each of these 4 lessons:

 

In order to be successful, we feel that clinical event prediction and subsequent intervention should be both content driven and clinician driven.  Importantly, the underlying data warehouse platform is key to gathering rich data sets necessary for training and implementing predictors.  Notably, prediction should be used in the context of when and where needed—with clinical leaders that have the willingness to act on appropriate intervention measures.

In the end, the overall goal is to leverage historical patient data to improve current patient outcomes. Predictive analytics is a powerful tool in this regard.

Leave a Reply

Your email address will not be published.