What is the Patient Safety Market?

The patient safety market traditionally has been comprised of EHR vendors and legacy patient safety software companies that serve the governance, risk, and compliance (GRC) market. More recently, private equity-backed predictive analytics vendors have emerged, many promising to solve the patient safety problem by predicting adverse events.

While the focus of this blog – which assumes that data-driven patient safety is the best way forward – there are also other non-data analytics companies that offer products and services to improve patient safety. Those vendors are not the subject of this blog.

While companies (or even non-profits, which are a distinctive feature of the healthcare industry) focusing on generating data analytics related to culture, employees, and experience also claim to improve patient safety, this blog also assumes – to be addressed in a separate blog – that cultural interventions, while necessary, are inadequate to measure, manage, and improve patient safety.

Why Has the Patient Safety Market Historically Failed?

The answer is simple:  the different categories of product and service providers have not been outcomes-driven.  Instead, the major categories have variously assumed accessing high volume of patient data would work; relying on compliance with regulatory requirements would be sufficient; or being able to target higher profile specific harms with advanced technology would lead the industry.

Therefore, to be clear, the market has been successful in supporting compliance requirements, but it has not been successful in making patients safer.

Let’s take each category one by one.

  1. EHR Vendors. Addressing just the inpatient setting first, the EHR market is now dominated by two vendors, Epic and Cerner. While these vendors offer health systems the ability to query EHR data, the capabilities required to monitor, measure, manage, and reduce adverse events are vastly greater than sold. More importantly, the EHR vendors neither offer the capability of generating clinically validated adverse event outcomes using health IT data (AE Outcomes) nor support intervention and improvement with these AE Outcome data.
  2. GRC Vendors. The data analytics of GRC vendors that pervade entire platforms offered are based on voluntary event reporting (VER) data. Patient safety, peer review, quality improvement, risk management, and much else relies on unreliable VER data. VER data identifies only approximately 5% of patient harm, which by definition means VER is also capable only of identifying patterns which exist within that 5% of events – excluding not only 95% of the events but also the patterns of harm in that 95% of events invisible to VER. Research, regulation, and health system adoption is obsolescing this legacy approach. Most importantly, the GRC vendors neither offer the capability of generating clinically validated adverse event outcomes using health IT data (AE Outcomes) nor support intervention and improvement with these AE Outcome data.
  3. Predictive Analytics Vendors. The data analytics of predictive analytics vendors are more technically impressive than other vendors to date, but these vendors building on more modern technology and using more advanced techniques suffer from three problems. First of all, and (again!) most importantly, the predictive analytics vendors neither offer the capability of generating clinically validated adverse event outcomes using health IT data (AE Outcomes) nor support intervention and improvement with these AE Outcome data. Second, and consequently, predictive analytics companies which do not hold sizable data sets of AE Outcomes are relegated to training predictive adverse event models with claims data, coarse-grained EHR data (such as mortality and morbidity data), or “proxy” outcomes data; but to predict “X” outcome, the model designed to predict “X” should be trained with “X” outcomes – not even very close “proxy” outcomes. Third, with respect to performance improvement strategy, a health system cannot predict its way to performance; measurement is vital, and knowing what just happened is a critical input – along with any predictive insight – for making a decision in any domain; a comprehensive analytic continuum is required.

While we’re on the topic, many in healthcare do not appreciate that not all retrospective data is created equal. For many, data that are weeks, months, or years old are treated equally as “retrospective data” as are “retrospective data” that are minutes, hours, and days old. They are different. The latter – which can often be described as “concurrent” data – can be effectively used in clinical operations to support interventions and high-frequency cycles of improvement (versus the months-long “rapid cycles” historically referenced in the patient safety community). The data analytic continuum is properly thought of not as retrospective and prospective but, rather, as retrospective, concurrent, and prospective.

Finally, predictive analytics companies suffer from other limitations in addressing adverse event market needs that will be the subject of a future blog.

The Roadmap to Patient Safety Value

How do we move from the traditional commercial offerings for patient safety data analytics to the optimal approach? The answer is “outcomes”, and specifically AE Outcomes.

There are three key requirements for generating and using these AE Outcomes that will result in what Pascal termed many years ago as value-based patient safety.

  1. Scientific Validation. Any adverse event (used interchangeably with patient safety here) data analytic methodology purporting to include adverse event outcomes data – the foundation of measuring and increasing value – must be evidence-based and validated by peer-reviewed publication. Ideally, especially in light of the need to standardize adverse event definitions in highly complex health IT environments, real-world implementation and evidence are important sources of further validation.
  2. Clinical Credibility. Data analytics using AE Outcomes should be developed in a multi-disciplinary manner with appropriate clinical leadership and guidance. The benefit of real-world implementation and evidence is maximizing the probability of clinical adoption, which is further enhanced by an appropriately cautious “First, do no harm” attitude – often requiring many iterations and more time to refine how these analytics work in practice.
  3. Financial Justification. Crucial to achieving value in patient safety is demonstrating a CFO-grade business case. Business cases in patient safety historically have not been robust and proven quite weak. How can we expect otherwise if we don’t have outcomes, and specifically AE Outcomes, to link to financial outcomes? We cannot. That is why demonstrating CFO-grade ROI requires AE Outcomes to link to the appropriate financial outcomes, showing how, on a directly linked basis, the reduction of harm or the substantiated avoidance of an adverse event results in an improved financial outcome on a patient-specific level.

That is also why (i) the three vendor types above struggle to deliver value-based patient safety why (ii) Pascal’s VPS solution does.