Advertisement

Statistics Gone Mad

      To the Editor:
      I read with interest Appel et al's article “Put a Face to a Name: A Randomized Controlled Trial Evaluating the Impact of Providing Clinician Photographs on Inpatients' Recall.”
      • Appel L.
      • Abrams H.
      • Morra D.
      • Wu R.C.
      Put a face to a name: a randomized controlled trial evaluating the impact of providing clinician photographs on inpatients' recall.
      This randomized controlled trial attempted to address an important issue for hospitalized patients worldwide, namely, recognizing and remembering the names and faces of those caring for them.
      The following methodological flaws are apparent:
      • A sample size of 300 patients was calculated. This allowed for α and β errors and a dropout rate of 8%. The actual dropout rate varied from 20% to 30%.
      • Only 19% of 1371 eligible patients were randomized. The reason for not randomizing 60% of eligible patients is not given.
      • The study was stopped with only 256 patients randomized because of lack of resources. I am interested to know what resources are referred to.
      Patients who received names and photographs (group C) correctly identified one clinician's photograph more than those in both the names only group (group B, P = .04) and the control group (group A, P = .001) (3 vs 2).
      Patients in group C remembered 2 clinicians' names, compared with 1 in group B, and those in group A did not remember any (group C vs A, P = .002; group B vs A, P = .01).
      Although the results reach statistical significance, the actual differences are minimal. Calculating statistical significance on a margin of 1 is meaningless and statistics gone mad. It would be better and more relevant if this study was repeated without these flaws.

      Reference

        • Appel L.
        • Abrams H.
        • Morra D.
        • Wu R.C.
        Put a face to a name: a randomized controlled trial evaluating the impact of providing clinician photographs on inpatients' recall.
        Am J Med. 2015; 128: 82-89

      Linked Article