From clinical observations to clinimetrics: a tribute to Alvan R. Feinstein, MD.

This issue of the Journal is dedicated to Alvan R. Feinstein, the father of clinical epidemiology [1]. We like to join those who honor his memory and acknowledge his invaluable contribution to methodology in clinical research. The debate is still open and calls for reassessment of the field. As to psychometrics, the basis of measurement in psychiatry, Fava et al. [2] argue in this issue that it is an inadequate method that persists in the field despite its many flaws. Their observations and musings have profound implications, because many clinical trials have faltered on the shores of the insensitive Hamilton Depression Rating Scale (HAMD). With insufficient sensitivity and an inability to detect small changes, the HAMD has caused investigators to fail to reject the null hypothesis (to accept that no difference exists), with the result that potentially effective antidepressants are kept out of the US market. An alternative, clinimetrics, is proposed as the basis to assess clinical phenomena, diagnostics, prognostics, and therapeutics. Psychometrics is the managed care of methods; all variables have the same weight, just as all physicians have the same value, regardless of experience, expertise, or judgement. Clinimetrics arises from clinical observation, from clinical methods, and, most important, from clinicians, with different variables having different weights. The acceptance of psychometrics may be a simple lack of clinical research training, an inferiority complex such that clinicians feel the need to adapt the methods of their better-trained psychometrician colleagues. Per Bech [3] discusses the item response theory of Rasch, a theory that uses a sensible method to assess symptoms based on their prevalence in those with a disorder (clinical coherence) and the importance of those symptoms for clinicians to define severity (weighting of symptoms). Consistent with the item response theory, several versions of a foreshortened HAMD have been found more sensitive to change than the full HAMD. Carlo Faravelli [4] questions the mathematical assumptions behind rating scales: that individual item scores can be summed to produce a total that has a linear relationship with the clinical phenomenon it was designed to measure. He notes that the same rating scale score can reflect a broad array of severity, i.e., many different individual item scores can add up to the same sum. To calculate the number of combinations and permutations of individual items that can sum to a HAMD of 30 is to realize the complexity of this total score. Furthermore, he argues convincingly that not only is a 50% reduction in baseline scores insufficient as a criterion of response, but even remission, the new standard for outcomes, falls short