Evidence-Based Medicine and the Hospitalist

Many accepted practices in hospital medicine, ranging from electrolyte management to the components of a syncope workup, may have scant or equivocal evidence to support their use. Leonard Feldman, MD, FAAP, FACP, SFHM, of Johns Hopkins Bayview Medical Center, spoke to several of these practices at the Society for Hospital Medicine 2013 annual conference, held May 17-19 at the Gaylord Nelson Convention Center at the National Harbor in Ft. Washington, MD. He began by emphasizing that unnecessary tests and treatments, driven by providers’ limited understanding of costs, learned behaviors, and even economic incentives and defensive medicine, may account for up to $375 billion per year in wasted health care costs.

Feldman examined the utility of routine carotid ultrasound (US) as part of a syncope workup. Neurovascular causes of syncope are uncommon, yet carotid US is a frequent component of the syncope workup. A 2009 Yale University study by Mendu et al examined the evaluation of syncopal episodes in older (>65 years old) patients. Carotid US affected management or helped define etiology in only one to two percent of patients studied. Conversely, if positional blood pressure (PBP) readings were obtained, they affected management or clarified etiology in 25 to 30 percent of patients. Costs incurred in order for carotid US to affect outcome in one patient would be approximately $20,000, versus $17 for PBP.

In 2005, Schnipper et al studied a subset of inpatients with an admission diagnosis of syncope who were then referred for neurovascular testing. Neurovascular US clarified diagnosis in only 2% of this “sickest of the sick” cohort, which comprised just 3% of syncopal admissions. Focal neurologic signs and carotid bruits were both associated with an increased likelihood of US results being diagnostically meaningful. Carotid US is not recommended for syncopal patients without focal neurologic signs or carotid bruits.

Perioperative seizure prophylaxis with phenytoin or, increasingly, levetiracetam is routinely ordered for patients undergoing resection of intracranial masses. Limited data support this practice, and research in the field is sparse. Additionally, phenytoin can have many adverse effects, and levetiracetam is less well studied. A 2013 prospective, randomized controlled trial by Wu et al studied the efficacy and safety of perioperative phenytoin prophylaxis versus observation in patients undergoing resection for glioma or cerebral metastasis resection. Seizure was the primary outcome, and harm from phenytoin was the secondary outcome. Interestingly, only about 8%, rather than the expected 30%, of untreated patients experienced any seizure activity. The number needed to treat for harm to occur (NNH) was only six for any adverse event and 20 for a major event; conversely, the number needed to treat for benefit to be seen (NNT) was 20.

Additionally, a 2008 Cochrane review found evidence neither for nor against seizure prophylaxis in people with brain tumors, and a NNH of three. In 2008, the American Academy of Neurology reaffirmed its recommendation against routine anti-seizure prophylaxis in individuals with newly diagnosed brain tumors.

In the setting of acute myocardial infarction, many physicians have been trained to maintain serum potassium in the 4.0-5.0 mmol/l range in order to minimize risk of ventricular fibrillation (VF), ventricular tachycardia (VT), and death. Until recently, these practices had primarily been based on small studies from the mid-1980s that used arrhythmias, rather than mortality, as primary end-points, and that primarily examined serum potassium on admission.

A much larger 2012 JAMA study by Goyal et al used retrospective cohort analysis to examine more than 38,000 records. Researchers obtained baseline and mean post-admission serum potassium levels, and delineated relationships between these values and in-hospital mortality, ventricular arrhythmias, and cardiac arrest. Inpatient mortality was lowest when potassium was between 3.5 and 4.5 mmol/l. Similar patterns were seen for VF/VT and arrest. Feldman emphasized that more large trials are needed, in light of the findings of this large observational study.

Finally, Feldman addressed the question of whether a proton pump inhibitor (PPI) or H2 blockade should be used for prophylaxis of GI bleeding in non-ICU patients. A large 2011 cohort-matched study by Herzig et al saw a very small benefit to acid suppression via H2 blockade or PPI versus no prophylaxis. However, the NNT was 770 to prevent one GI bleed. By contrast, when considering the risk of Clostridium difficile infection, the NNH was 533; for hospital acquired pneumonia, the NNH was only 111. Feldman said he does not recommend routine use of prophylactic acid suppressing medication for non-ICU inpatients.