GAS (Goal Attainment Scaling)

Goal Attainment Scaling (GAS) is an individualized outcome measure: progress on each goal set with the client before treatment is rated on a 5-point scale from −2 to +2, and the ratings are combined into a single T-score ^[2].

What does the GAS measure?

GAS was developed by Thomas Kiresuk and Robert Sherman in 1968 at the Hennepin County Mental Health Service in Minneapolis and published in Community Mental Health Journal ^[1]. Unlike standardized questionnaires that assess a fixed set of symptoms or behaviours, GAS is built around individualized goals negotiated between the clinician and the client. This person-centred design makes it applicable across almost any clinical population or treatment approach.

For each goal, the clinician and client define five possible levels of outcome in advance, each anchored by a behaviourally specific descriptor:

• −2: much less than expected
• −1: less than expected
• 0: expected outcome
• +1: more than expected
• +2: much more than expected

The expected level (0) is defined before treatment begins. It is a realistic, agreed goal rather than the most optimistic or pessimistic possibility.

Typically 3–5 goals are set per client. At a pre-specified follow-up point, each goal is rated on the 5-point scale and a composite T-score is calculated with a standard formula.

A T-score of 50 means the client achieved exactly the expected level across all goals. Doing better than expected gives a T-score above 50, and doing worse gives one below 50 ^[2].

GAS is used in rehabilitation medicine, pediatric occupational therapy, neurological recovery and person-centred mental health programs. In mental health care it can sit beside fixed outcome measures such as the WSAS or the ORS, as part of measurement-based care.

Rate each goal on the 5-point GAS scale. Equal weights assumed. Correlation coefficient (ρ) = 0.3.

Reference calculation only. GAS requires individual goals set collaboratively with the clinician before treatment begins.

How do you give the GAS?

Goals must be set prospectively using the SMART framework and anchored to behaviourally specific descriptors before treatment begins.

Step 1, identify priority goals

The clinician and client identify 3–5 goals together that are meaningful to the client and relevant to the treatment focus. Goals should reflect functional activities, participation or behavioural changes rather than abstract aspirations. The client's values and priorities guide goal selection.

Example areas: returning to work, managing daily routines independently, improving sleep hygiene, reducing avoidance behaviour, or resuming a specific social activity.

Step 2, define the expected outcome (level 0)

For each goal, the clinician and client define what a realistic level of achievement looks like at the target follow-up date. This is the level 0 ("expected outcome") anchor.

It should be achievable with the planned treatment but not trivially easy. Level 0 must be set before treatment begins. It reflects clinical judgment about what is realistically achievable given the client's current status and the planned intervention.

Step 3, anchor all 5 scale levels

After defining level 0, the team writes specific, observable behavioural descriptors for levels −2, −1, +1 and +2. These descriptions should be mutually exclusive and measurable. Replace vague language with observable behaviours (for example, "walks 200m without rest" rather than "improved walking").

All five levels should be described in detail before treatment, so the rating is not influenced by hindsight or knowledge of actual outcomes ^[3].

Step 4, rate goals at follow-up and calculate the T-score

At the agreed follow-up point (for example 4, 8 or 12 weeks), each goal is rated on the −2 to +2 scale using the pre-defined anchors. The T-score formula then converts the goal ratings into a composite score.

A T-score of 50 means expected outcomes were achieved; scores above 50 indicate better-than-expected performance. Goals may be weighted (for example, a higher-priority goal receiving a weight of 2) to reflect their importance to the client and their difficulty ^[2].

How do you score and interpret the GAS?

The 5-point scale is applied to each individualized goal. The T-score formula aggregates outcomes across all goals into a single composite score.

T-score formula

T = 50 + 10 × Σ(wᵢ × xᵢ) / √(1 − ρ)Σwᵢ² + ρ(Σwᵢ)²

Here wᵢ is the weight of goal i (1 each when goals are weighted equally), xᵢ is the score for goal i (−2 to +2), and ρ is the assumed inter-goal correlation, conventionally 0.3. The formula yields a T-score with mean 50 and standard deviation of about 10 ^{[1] [2]}.

When all goals are rated 0 (expected outcome achieved), the T-score equals exactly 50 regardless of the number of goals or weights. The T-score rises above 50 when performance exceeds expectations and falls below 50 when performance is below expectations.

Psychometric properties

GAS is a mathematical technique for quantifying goal achievement rather than a fixed questionnaire, so its measurement properties depend on how well each goal scale is written.

• T-score distribution. Using Kiresuk's T-score, the aggregate is normally distributed about a mean of 50 when goals are achieved precisely, with a standard deviation of around 10 when they are over- or underachieved ^[2].
• Variants and analysis. A literature review in rehabilitation notes that several GAS variants exist and urges caution with the T-score; because GAS ratings are ordinal, medians and rank tests are appropriate for analysis ^[3].

Limitations

• Depends on goal quality. Valid results need all five levels described in detail before treatment; vague or post-hoc anchors bias the score ^[3].
• No normative bands. The T-score shows performance relative to each person's own expected outcome, not against published norms, so scores are not severity categories.
• Ordinal data. Treating the T-score as interval data calls for caution ^[3].
• Not a diagnostic or symptom measure. GAS complements, rather than replaces, diagnosis and standardized symptom measures.