- On November 6, 2018
1. It takes a lot of people and a sophisticated health system to use triggers to identify and reduce patient harm.
This is one of the most common misconceptions encountered by Pascal Metrics, and it is held by veteran researchers and young practitioners alike. First, clinical review is required for an epidemiologically sound process that generates clinically validated EHR-based adverse events and, without exception, the health systems in Pascal’s Community Collaborative have demonstrated the process assisted by automation to be efficient and scalable.
Second, most health systems are recognizing that as they identify and reduce harm, they are also reducing harm-related cost. Therefore, this approach — what we call “value-based patient safety” — incurs not only a cost but creates financial benefit, all of which resulting in value. Contrast this with many traditional views on patient safety, which hold that investments in this area are required but do not improve economic sustainability.
Third, most health systems are also recognizing that, if EHR-based harm identification is more clinically efficient and effective than voluntary event/incident reporting (or even administratively coded data), then they have the opportunity to reallocate FTE time reviewing event reports to conducing clinical review. What’s more, clinicians reviewing automated harm detection data find more opportunity faster, as compared to traditional methods.
Takeaway: Simply put, the trigger-based method is being used from large hospitals to hospitals well under 100 beds, with the latter showing equal if not — in some cases — superior performance in identifying and reducing patient harm.
2. A trigger with a low PPV means that its use is clinically ineffective and financially costly.
Those who hold this misconception assume that the use of one trigger in a one-to-one manner detects one harm; if this were the case, this view would not be a misconception but a valid criticism. The reality is that clinically effective use of this method relies on the use of multiple triggers in a guided management workflow reviewed with clinical judgment.
For those in the field becoming familiar with the application of predictive technologies using machine learning and AI-assisted advanced analytics, the utility of using multiple predictors is becoming recognized and appreciated. For example, in Pascal’s Global Safety Risk (GSR) Score predicting all-cause harm, hundreds of predictors are used as demonstrated our methods paper. Likewise, when it comes to detection, the use of multiple triggers — or analogously, “detectors” — in a comprehensive system as described in the “PSAM” described in the November 2018 Health Affairs paper similarly makes the PPV of a single trigger or “detector” not the relevant metric of evaluation.
Therefore, it is not each trigger that should be evaluated in isolation but, rather, it is the performance of a comprehensive patient safety management & quality improvement system that delivers outcomes. The outcomes in Pascal’s Community Collaborative are this: the benefits generated by the system exceeds the costs (both direct and indirect) incurred to deliver those benefits — all of which resulting in value.
Takeaway: The value of a trigger is not a function of the trigger but, rather, the system in which it is operated.
3. “We’ve used the IHI Global Trigger Tool. I likely won’t find anything I don’t already know.”
First of all, while the Pascal used the Global Trigger Tool (GTT) triggers as a foundational set in our trigger library, the Pascal solution has taken health systems in our Community Collaborative far beyond the GTT. While referring to an “automated GTT” is a useful first step in educating those who are new to what we do, the idea of simply automating the GTT and stopping there is tantamount to “paving the cowpath.”
More importantly, those who implement Pascal’s solution, which is comprised of a real-time software service and our expert clinical program, have invariably found patient harm not detected by any other system. This includes opportunities both to conduct concurrent intervention and high frequency cycles of improvement.
With respect to the GTT, users are manually pulling and reviewing about 20 patient charts per month. The number of patient records and associated data being fired upon by real-time triggers as well as being reviewed by clinical review (for those records brought on an exceptional basis to clinical attention) is a fraction of what an automated harm detection method covers on a continuous basis.
Therefore, expect to find far more harm for actionable intervention and improvement using an automated trigger method than with the GTT.
Takeaway: Dr. Don Berwick, founder of of the Institute for Healthcare Improvement and former CMS Administrator, pointed out that using EHR data provides a unique “lens” into safety vulnerabilities that boards and “all hospitals” should embrace.