The Measurement : Medicine Interface

The dominant scientific paradigm expects advancement of Medicine only from progresses in Biology. By \u201cBiology\u201d here is meant the ensemble of any vital phenomenon concerning \u201cparts\u201d of the person, however small, from organs to molecules. Blood circulation as well as cellular trans-membrane electric potential are \u201cbiological\u201d phenomena in this sense. By contrast, walking, speaking, suffering pain, being dependent on another one\u2019s help cannot be ascribed to any parts of the person: these \u201cwhole-person\u201d phenomena postulate the existence of a unitary \u201cself\u201d. From the measurement perspective, either class of phenomena has different requirements. Both manifest themselves and can be measured through observable representative events. \u201cEvents\u201d can be thought of as any exchange of energy or information: within the body, for biological events (e.g. nerve eletric conduction leading to muscle contraction); between the person and the environment, for person events (e.g. verbal communication). Person \u201cevents\u201d, in this view, can be also defined as behaviours. Biological exchanges a) are mostly deterministic: a given cause entails a predictable effect, and b) they are directly observable (e.g. blood pressure peaks through manometers, cellular membrane potentials through micro-electrodes). These events are thus either valid or quantitative indices of the represented phenomena. \u201cMore\u201d mPa or \u201cmore\u201d mV represent, and are proprtional to, a higher blood pressure or a higher electric transmembrane potential, respectively. By contrast, person behaviours a) require a probabilistic approach since unpredictable self-initated behaviours may interact with, or substitute for, predictable stimulus-response circuits, and b) they are loosely representative of their generating phenomenon. For instance crying may represent pain as well as depression or sudden happiness. In fact, person phenomenon are also said to be \u201clatent\u201d traits, which can only be inferred by \u2013not observed through- behaviours. Let\u2019s accept that crying does represent pain: \u201cmore\u201d crying may indicate \u201cmore\u201d pain, but how much more remains undetermined. Clearly, an original paradigm for measurement is needed whenever person \u2013rather than biological- phenomena have to be addressed. This has been acknowledged for over 150 and for 80 years for psychological (EH Weber) and sociological (LL Thurstone) sciences, respectively. By contrast, only in the latest 15 years the problem was acknowledged for Rehabilitation (BD Wright). This is not surprising. Rehabilitation as a medical specialty is deeply rooted into biology (mostly in movement sciences), yet it is more and more recognized that its outcomes are behavioural in nature (independence, pain, fatigue, balance, continence, quality of life and the like). The Rasch statistical model (originally proposed by G Rasch in 1960) is perhaps the most exhaustive solution to the specific problems of measurement raised by person phenomena. Counts of observations (e.g. raw ordinal scores on questionnaires encompassing various behaviours) can be transformed into continuous linear measures telling you \u201chow much\u201d of the trait (independence, pain, fatigue etc.) is latent in the subject. The confidence one can place in the measure can be formally estimated mostly from a) the homogeneity of the scored behaviours (do they represent the same trait?) and b) the precision of the instrument. Are there enough questions? Are the questions \u201con target\u201d.i.e. neither too easy nor too difficult for a given subject ? . Rasch measures can thus compete with biological measures with respect to fundamental metric properties such as validity,precision, reliability. This paves the way for the advancement of Research in Rehabilitation. Compared to the dominant (and successful) biological paradigm, Rehabilitation still suffers from a lag in the development of a) specific statistical approaches and b) specific trial designs. But proper measurement theory is now at hand: most of the gap has been filled up. ----------------------------------------------- References 1. Andrich D. \u2018Rasch models for measurement.\u2019 1998. Sage Publications, Newbury Park. 2. Cook TD, Campbell DT. Quasi-experimentation. Design and analysis issues. 1979, Houghton Mifflin College. 3. Gray ML, Bonventre JV. Training PhD researchers to translate science to clinical medicine: closing the gap from the other side. Nature Medicine 2002;8:433-43. 4. Kuhn TS. The structure of scientific revolution. The University of Chicago Press 1970 (original edition 1962). 5. Tesio L. Measuring person\u2019s behaviours and perceptions: Rasch analysis as a tool for rehabilitation research. Journal of Rehabilitation Medicine 2003;35:105-115. 6. Tesio L. Bridging the Gap Between Biology and Clinical Medicine. Some help from Rasch measurement theory. J appl Meas 2004, submitted