Unconsidered Issues of Measurement Noninvariance in Biological Psychiatry: A Focus on Biological Phenotypes of Psychopathology

Abstract

There is increasing appreciation that certain biological mechanisms may not be equally related to all psychiatric symptoms in a given diagnostic category. Research on the biological phenotyping of psychopathology has begun examining the etiological and treatment implications of identified biotypes; however, little attention has been paid to a critical methodological implication of these results: measurement noninvariance. Measurement invariance is the ability of an instrument to measure the same construct across different people or across different time points for the same individual. If what a measure quantifies differs across different people (e.g., those with or without a particular biotype) or time points, it is invalid to directly compare means on said measure. Using a running example of inflammatory phenotypes of depression, we first describe the biological phenotyping of psychopathology. Second, we discuss three types of measurement invariance. Third, we demonstrate how differential biology-symptom associations invariably creates measurement noninvariance using a theoretical example and simulated data (for which code is provided), and how this issue can lead to false conclusions about the broader diagnostic construct. Finally, we provide several suggestions for addressing these important issues to help advance the field of biological psychiatry.