Beyond the unit score: a proposal for an alternative correction method for psychometric instruments

Abstract

Background: Measurement in mental health often uses unit sums (such as Likert scales), assuming that all items have the same equidistance and explanatory weight. This may fail to account for the fact that different symptoms have varying impacts on complex constructs, thereby limiting diagnostic accuracy. Method: In this methodological reflection article, three alternative weighted correction methods are proposed, using the PHQ-9 depression questionnaire as a methodological model: a) Bartlett’s approach: based on principal component analysis, assigning linear multipliers; b) Partial Credit Model (PMC): based on Item Response Theory, using non-linear parameters and category thresholds; c) Clinical interpretation: based on the categorical diagnostic criteria of the DSM-5. Results: The Bartlett approach corrects unitary bias, although its effectiveness depends on the calibration sample. The PMC approach provides the test with sample independence, limiting the patient’s actual psychometric ceiling. The clinical model provides predictive value and triggers necessary clinical alerts. Conclusion: Moving away from the traditional sum of points in favour of weighting systems may improve the explanatory accuracy of the tests. Although it requires recalibration and the use of decimals, current computational capabilities may justify this methodological advance.