The Clinical Global Impression scale (CGI) was developed by William Guy at the National Institute of Mental Health and published in 1976 in the ECDEU Assessment Manual for Psychopharmacology. It has become the most widely used outcome measure in psychiatry, not because it is sophisticated but because it is short, intuitive, and clinically meaningful. A trained clinician can complete it in less than a minute. The FDA accepts it as a co-primary or secondary endpoint in nearly every psychiatric trial.
The CGI is a one- to three-item clinician rating of overall illness severity (CGI-S) and treatment-related change (CGI-I), each on a 1–7 scale.
The three components
- CGI-S — Severity of Illness. "Considering your total clinical experience with this particular population, how mentally ill is the patient at this time?" Rated 1 (normal, not at all ill) to 7 (among the most extremely ill patients).
- CGI-I — Global Improvement. "Compared to the patient's condition at admission to the project, this patient's condition is..." Rated 1 (very much improved) through 4 (no change) to 7 (very much worse).
- CGI-Efficacy Index. A two-axis combination of therapeutic effect and side effects. Rarely used today.
The first two — CGI-S and CGI-I — are the workhorses.
Why such a simple scale survives
The CGI does something no other instrument does: it asks an experienced clinician for a single, holistic judgement. This is exactly the kind of judgement a senior psychiatrist makes every day in the clinic. By asking it explicitly and recording it numerically, the CGI captures clinical wisdom in a form that can be aggregated across trials and across centres. The trade-off is that it depends entirely on rater experience — a green clinician's CGI-S of 4 may not be the same as a 30-year veteran's.
Anchoring CGI to other scales
One of the most cited papers in modern schizophrenia methodology is by Stefan Leucht and colleagues (2005, Neuropsychopharmacology), which mapped CGI-S to PANSS and BPRS thresholds. Their estimates, often called the "Leucht conversions":
- CGI-S 3 (mildly ill) ≈ PANSS total ~58
- CGI-S 4 (moderately ill) ≈ PANSS total ~75
- CGI-S 5 (markedly ill) ≈ PANSS total ~95
- CGI-S 6 (severely ill) ≈ PANSS total ~116
This mapping is what allows researchers to translate continuous PANSS change into clinically intuitive labels.
How it is administered
The CGI is completed by the treating clinician (or a trained rater) at every assessment point — typically baseline and follow-up visits. There is no patient interview specific to the CGI; it is a synthesis of everything the rater knows about the patient. For trials, it is usually rated immediately after a longer assessment such as the PANSS, drawing on that interview content.
What "very much improved" means in practice
The CGI-I anchor descriptions tie loosely but consistently to other measures. In schizophrenia trials, "much improved" (CGI-I 2) typically corresponds to a 25–30% reduction in PANSS total; "very much improved" (CGI-I 1) corresponds to 50% or more. This is one reason both scales are usually reported together — the CGI-I gives a global feel, the PANSS gives the numbers.
Strengths
- Extraordinarily brief — under a minute per rating
- Captures holistic clinical judgement
- Cross-condition: used in schizophrenia, depression, anxiety, ADHD, bipolar, dementia
- FDA-accepted endpoint in regulatory submissions
- Public domain, free to use
Limitations
- Depends on rater experience. Inter-rater reliability is acceptable but not as strong as for anchor-rich scales like the PANSS.
- Single-number summary. Two patients with the same CGI-S of 4 may have very different clinical pictures. The CGI does not break out symptom domains.
- Scale anchors are vague. "Borderline mentally ill" and "moderately ill" are interpreted differently by different raters.
- Memory bias on CGI-I. Long intervals between ratings make accurate change ratings harder.
Variants and adaptations
Several disorder-specific CGI versions exist, including the CGI-BP (bipolar disorder) and the CGI-SCH (schizophrenia, with separate ratings for positive, negative, depressive, and cognitive symptom domains). The CGI-SCH is useful when a rater wants the brevity of the CGI but a domain-level breakdown.
What this means for patients and families
The CGI is one of the rating scales most likely to actually appear in your clinical chart, because it is so quick. If your psychiatrist's note says "CGI-S = 4, CGI-I = 3 (minimally improved)," they are recording a clinical impression in a standardised way. You can ask what would constitute a "much improved" CGI-I rating at the next visit, which is a useful structuring conversation for shared decision making.
How regulators think about it
The FDA, the European Medicines Agency, and most national regulators accept CGI-I as a secondary endpoint in psychiatric trials. Some early trials and accelerated-approval contexts have used CGI as a co-primary endpoint. The FDA Drugs portal includes briefing documents that frequently cite CGI changes alongside scale-specific outcomes like PANSS.
This article is for educational purposes only and is not medical advice, diagnosis, or treatment. Always consult a qualified mental health professional. If you or someone you know is in crisis, call or text 988 in the US, or your local emergency number.