Today’s critical problem in patient safety is the prevalence of diagnostic errors including failure to diagnose, missed diagnosis and delayed diagnosis of disease.
Threshold Model of Diagnostics
-
Introduction
As the 20th Century "golden age" of medicine brought forth new discoveries (beginning with the identification of the three blood types and ending with the identification of monoclonal antibodies) and new technology (including the electron microscope, non-invasive imaging, the pulse oximeter), more clinical information was available than ever before. It was now possible to quantify things like white blood cells, platelets, and molecules in the blood. But what did those numbers mean?Statistics provided a 20th Century solution with the introduction of thresholds. A threshold is essentially a predefined value for a system that is used to determine an action, including diagnosis or treatment. In medicine, there are threshold applications for tests, monitors, and diagnostic and treatment models.
-
How it Works
Threshold-based tests attempt to define a "normal" value or a predicted "normal" value for a particular physiologic measure. An actual obtained value that is within this predicted range is considered "normal." An obtained value that is above or below the threshold range would be considered "abnormal." For example, if a patient without known infection had a white blood cell count of 12K, this would be considered "abnormal" and further examination may be warranted. Contemporary threshold devices typically provide an alarm function that alerts the healthcare worker when an obtained value breaches the threshold value. Sometimes, multiple thresholds are added together to produce a super threshold. An example of this is the Modified Early Warning Score (MEWS). -
The Problem
Threshold models rely on a single data-and-time “snap shot,” fragment, or group of fragments to assign probabilities about the likelihood of a disease or condition. However, many clinical conditions such as sepsis rapidly change in relational morphology over time. Bright-line thresholds or super thresholds are far too simplistic to detect these dynamic patterns in the data. In the era of threshold-based monitoring, this leads to a high incidence of false alarms and failure to alarm of a patient’s adverse clinical condition. In hospital wards, frequent false alarms contribute to the problem of alarm fatigue—often setting in before the occurrence of a true adverse clinical event. When threshold settings of the alarms are increased to prevent the so-called false alarms, the alarm’s silence a second problem—a false sense of security among hospital staff, which can delay care and adversely affect the patient’s health.