Critical limits represent quantitative decision thresholds for drugs that require immediate clinician notification and potential life-saving intervention. United States hospitals lack a national standard for drug critical limits. We collected critical limits from 417 US hospitals across all 50 states and Washington, D.C.; of these, 411 maintained drug critical limit lists. We classified hospitals by US Census division and network-affiliated vs. independent status. We applied non-parametric statistical analyses, examined critical limits for 111 drugs, and observed significant inter-institutional variability. Listing frequencies were highest for digoxin 99.8% (410/411), theophylline 72.3% (297/411), lithium 94.9% (390/411), and acetaminophen 86.9% (357/411). Non-therapeutic measurands also appeared, led by ethanol 47.2% (194/411). Kruskal-Wallis analysis revealed highly significant differences (P < 0.01) across census divisions. Mann-Whitney U analysis comparing network vs. independent hospitals yielded significant differences (P < 0.05) in 16 analytes. Qualitative critical values were listed for volatile alcohols and methotrexate. All digoxin critical limits exceeded levels associated with increased hazard ratios. Only 33.9% of hospitals aligned with consensus guidelines for acetaminophen poisoning. Vancomycin and aminoglycoside critical limits showed wide ranges, overlapping peaks and troughs, and random values inconsistently aligned with peak and troughs. Immunosuppressant critical limits often exceeded thresholds associated with toxicity. Psychotropic drugs, including antiepileptics, tricyclic antidepressants, and lithium, demonstrated variability and misalignment relative to consensus guidelines. Drugs of abuse did not appear in critical notification lists. Results provide hospitals with references that will enhance hospital notification practices and patient safety. We encourage sharing of critical notification lists to foster future research efforts and enhance standards of care.
Kuang et al. (Thu,) studied this question.