A hospitalist is unlikely to collect the sample or run the test. How many of us centrifuge urine or examine blood smears? How many could read a complicated MRI or a PET scan? The busy clinical workflow coupled with the complexity of these tests demands that someone other than the primary caregiver interpret these tests. This also means that we are removed from the primary data and must rely on another practitioner’s interpretation. Even with this separation in the process, we are presented with more clinical data for each patient than ever before.
Too Much Information
Of course, the availability of these data is not without problems. An individual can review, assess, and act upon only so many data points.2 As the volume of data increases, so does the likelihood that a piece of important data will be missed. This setting can make things difficult for the busy hospitalist.
We need to quickly prioritize our time in the hospital. A first step in this process involves a quick review of physiologic studies that suggest levels of patient acuity. This information helps us to see the most critical patients first, and to identify other important issues that need to be addressed (although less urgently).
As more data are collected, this task gets more complicated. Certainly not all of the data collected are equally important for determining patient acuity. Given this, the volume of data contributes to the overall “noise” of the sample and—in some cases—the important data may be overlooked. Critical values (signals) become more likely to get lost in a sea of less important data (noise). More noise means more uncertainty and requires better evidence to make a clinical judgment.3
Information systems developers use various techniques to get around these problems. One way to manage the vast amount of information is to alert practitioners to outliers. Some have proposed that alarms may be the answer to our noise problem.
Alarms Don’t Work
Unfortunately, alerts and alarms can actually add to the noise, especially in ICU monitors. Anyone who has spent time in an ICU knows that alarms are constantly sounding. It has been estimated that false positive alarm rates range from 64%-87% in airway management situations.4
Alarms are often meaningless. Monitors can be so sensitive that they pick up background noise in their measurements, causing false alarms that increase the noise (both literally and figuratively). Anesthetists, recognizing that alarms are non-specific, frequently work without them.5 This is paralleled in the ICU because caregivers seem to ignore many alarms. (Most hospitalists who have spent time in the ICU can attest to this.) Specific problems with alarms include:
- They seldom localize the problem;
- They do not provide predictive information; and
- The diagnostic process is still left to the practitioner.5
Alarms also vary in importance. An intravenous pump that sounds because the fluid bag is empty may not be as important as an apnea alarm on a mechanical ventilator. A single alarm may not be as concerning as multiple simultaneous alarms for a patient with low blood pressure, high heart rate, and apnea. The goal in these cases is to signal a problem and to transmit that signal appropriately. But until this can be done reliably, alarms do not seem to be the answer.
On the other hand, presenting all the data (rather than just the important data) to a clinician may obscure important elements. This can result in missed diagnosis, delayed treatment, or incorrect treatment. So we don’t want to overwhelm the hospitalist with all the data; we just need to highlight and present important data. But how?