How Depression Rating Scales Are Scored and Why the Cutoffs Are Debated

The short answer

Depression rating scales convert a set of symptoms into a single number by scoring individual items and adding them up. The two most common in research are the Hamilton Rating Scale for Depression (HAM-D) and the Montgomery-Asberg Depression Rating Scale (MADRS). A long-standing convention treats a three-point difference between a drug and placebo on the HAM-D as the mark of a clinically meaningful effect. That number is the heart of a genuine scientific dispute, because studies that anchor scale points to what clinicians can actually perceive suggest the smallest noticeable improvement corresponds to roughly seven points, not three. This article explains how the scoring works and why reasonable experts disagree about where the line should sit. It is educational and not medical advice.

How the scales turn symptoms into a score

The HAM-D, published by Max Hamilton in 1960, is a clinician-administered instrument. Its most-used form, the HAM-D-17, rates 17 items covering depressed mood, guilt, suicidal thoughts, insomnia, anxiety, agitation, appetite, and several physical symptoms. Some items are scored 0 to 4, others 0 to 2, and the points are summed. Higher totals mean more severe illness. Because several items measure sleep, weight, and somatic complaints, a portion of the total reflects symptoms that overlap with ordinary physical states and with medication side effects.

The MADRS, developed by Montgomery and Asberg in 1979, takes a different design approach. It has 10 items, each scored 0 to 6, and it deliberately concentrates on the mood and cognitive core of depression, with less weight on sleep and physical symptoms. Both scales are administered through a structured interview, and both depend on a trained rater's judgment. Because the two instruments have different item counts and point ranges, a given raw-score change on one does not translate one-to-one onto the other.

What "clinically important difference" actually means

A statistically significant difference in a trial only tells you the effect is unlikely to be pure chance. It says nothing about whether a patient or clinician would notice or care. The concept meant to fill that gap is the minimal clinically important difference: the smallest change in score that corresponds to a real, perceptible improvement in a person's condition.

The problem is that a scale point has no inherent meaning. Two points on the HAM-D could represent a small shift in sleep quality or a meaningful lift in mood, depending on which items move. To give the numbers external meaning, researchers anchor them to something clinicians already interpret, most often the Clinical Global Impression scale, on which a rater simply judges a patient as unchanged, minimally improved, much improved, and so on.

The most cited anchoring study is by Leucht and colleagues, published in the Journal of Affective Disorders in 2013. Using data pooled from dozens of drug trials, they linked HAM-D-17 scores to clinician global ratings and found that a judgment of "minimally improved," the faintest detectable benefit, corresponded to a HAM-D change of roughly seven points. A rating of "much improved" did not map to a fixed number of points but rather to a reduction of more than half the baseline score. They also found the relationship shifted with baseline severity: in less severely ill patients a minimal improvement mapped to about a six-point change, and in the most severe cases to about eight. In other words, the same raw-score movement means different things at different starting points.

Why the three-point cutoff is contested

For years, guidance from the UK's National Institute for Health and Care Excellence used a three-point HAM-D difference between drug and placebo as a working boundary for clinical significance. The critique that followed is straightforward. If the smallest change a clinician can even perceive is around seven points, then a three-point threshold labels as "clinically important" a difference too small for anyone to detect at the bedside. By that logic, the conventional cutoff is too lenient, and many antidepressant trials that clear it may still be producing changes below the threshold of noticeability.

This matters because average drug-placebo gaps in antidepressant trials tend to land in exactly this contested zone. A 2020 meta-analysis by Hengartner and colleagues in PLOS ONE pooled results from more than 130 placebo-controlled trials, examining the HAM-D-17 and MADRS separately rather than head to head. The standardized effect sizes were nearly identical, about 0.27 on the HAM-D-17 and 0.30 on the MADRS, and the raw drug-placebo differences were roughly two points on the HAM-D and three on the MADRS. Both sat below the minimal-improvement benchmarks the authors adopted, which is why they described the average benefit as of questionable importance for the typical patient.

The counter-argument

The debate does not end there, and it would be misleading to present one side as settled. A 2020 paper by Hieronymus, Jauhar, Ostergaard, and Young in the Journal of Psychopharmacology argues that judging antidepressants by the full HAM-D sum-score understates their effect. Their reasoning is that the scale bundles core depressive symptoms together with items measuring sleep, somatic complaints, and effects that overlap with medication side effects, adding noise. When they analyze symptom-specific items or shorter core subscales such as the HAM-D-6, the drug-placebo separation grows, and effects on core features like depressed mood look larger than the full-scale average suggests. On this view, the instrument itself, not the treatment, is partly responsible for the small headline numbers.

Both positions can be read as pointing at the same underlying issue. A single summed number compresses a heterogeneous illness into one figure, and the meaning of any threshold depends on how the scale was built and which symptoms drive the total. That is why a three-point cutoff cannot be evaluated in isolation from the anchoring evidence, the choice of scale, and the severity of the patients being studied.

How to read these numbers as a reader

When you encounter a trial reporting an X-point improvement, three questions clarify what it means. Which scale was used, since HAM-D and MADRS points are not interchangeable? Was the difference anchored to any external judgment of benefit, or only reported as statistically significant? And how severe were the patients at baseline, given that the same change means more in milder illness and less in the most severe? None of this settles whether a given treatment helps a given person. It does make the reported number easier to interpret honestly, which is the point of learning how the scoring works.