Of Hospital Alarms and the Lessons for Population Health Alerts

This hot-off-the-presses JAMA article by Vineet Chopra Laurence McMahon describes how the cacophony of alarm beeps and bells from hospitals' intravenous pumps, remote EKG monitoring, call buttons and other monitoring devices are paradoxically threatening inpatient safety.  While intent of all that noise is good, its effectiveness is compromised by a high false negative rate, making it difficult for docs and nurse to distinguish between simple vs. a life-threatening problems.

It seems researchers are finally discovering what distracted providers have known all along: that the all beeping buzzing and binging interferes with focusing on their duties and workflows. As a result, it's not unusual for some nurses and docs to resort to disabling the alarms.

The authors offer up some simple suggestions that the former hospitalist Disease Management Care Blog agrees with.  But since the DMCB is also interested in population health, it couldn't help but think of the parallel lessons about "alerts" that also imbue population health management. 

As PHM workers know, it's not unusual for vendor and medical home care managers to receive lists of patients who have missed appointments or have tests results that are outside of optimum range.  Instead of a noisy work environment, these health care professionals have a cluttered in-box.

To wit, the lessons are to be less one dimensional:

1. Establish priorities based on risk: Just as an alarm triggered by ventricular fibrillation should be distinguishable from an alarm from an innocuously malfunctioning intravenous pump, patients with test results that are slightly outside the optimum range should be distinguishable from patients with more pressing care needs.

Depending on the population, the prevalence of disease and available resources, clinical algorithms and predictive modeling should be able to prioritize the patients with the greatest needs. Start by assigning priority alerts to those patients and work your way down.

2. Accommodate patient priorities and clinic work flows: Just as an IV pump could best signal a problem only when the nurse enters the patient room, so must population health anticipate how to best accommodate a patient need based on what the patient wants and what the doctor can provide. That not only means putting the informed and engaged patient in control, but also understanding the network and having a working familiarity with each clinic's characteristics and culture. 

3. Use multiple signals:  One outcome measure that is out of compliance may mean little to a real world patients and the real providers who take care of him or her.  A blood test that is out of range is far more pressing if it also accompanied by problems involving other organ systems and may be more actionable if  it is the result of not being able to afford a medication co-pay.  Achieving a particular outcome may be less important than other issues, such as other medical needs or bigger problems such as food and shelter that fall outside traditional health care.  No medical problem can be fixed in an information vacuum.
 
The DMCB took the JAMA article's last paragraph and reworded it slightly:

Simple outcome measurements no longer establish an umbrella of quality. The scope and design of these measurement systems and their alerts must shift from the status quo to a biologically valid, clinically relevant, patient-centered model. Existing technology allows integration and intelligent assessment of patient data to create advanced alert systems. Changes to design and implementation of alerts are necessary to improve patient outcomes.

Image from Wikipedia