Goal Attainment Scaling
- Introduced by
- Thomas J. Kiresuk & Robert E. Sherman
- Year
- 1968
- Type
- Individualized, criterion-referenced outcome measurement procedure
- Family
- Individualized goal measurement
What does it do?
Goal Attainment Scaling (GAS) is a method for measuring the outcomes of an individualized intervention by writing, in advance, a small set of client-specific goals and defining for each a graded scale of possible outcomes from much worse than expected to much better than expected.
After the intervention, the actual outcome on each goal is scored on this scale and the scores are combined into a single standardized index, allowing idiosyncratic, personally meaningful goals to be aggregated and compared across clients and programs. It was introduced by Thomas Kiresuk and Robert Sherman in 1968 to evaluate community mental health programs.
The intuition behind the method
Standardized scales ask the same questions of everyone, which can miss what actually matters to a particular client. GAS flips this: it lets each client (with the worker) name the goals that matter to them, but disciplines that flexibility by forcing everyone to describe, before treatment, exactly what 'no change', a 'somewhat better' outcome, and a 'much better' outcome would concretely look like — a five-point ladder centered on the most likely result. Because every client's idiosyncratic ladder is anchored to the same -2-to-+2 metric, the otherwise incomparable goals can be added up and converted to a common score, marrying individualization with aggregation.
History
Thomas Kiresuk and Robert Sherman developed Goal Attainment Scaling in the late 1960s at the Hennepin County Mental Health Center in Minneapolis, publishing the foundational paper in the Community Mental Health Journal in 1968 as a way to evaluate diverse community mental health programs against individualized objectives. Over the following decades GAS spread well beyond mental health into rehabilitation medicine, geriatrics, education, and social services. Kiresuk, Smith, and Cardillo's 1994 edited volume consolidated the theory, measurement properties, and applications, and GAS remains a widely cited individualized outcome method.
When to use it
Use Goal Attainment Scaling when intervention goals are genuinely individualized and a fixed, off-the-shelf scale would fail to capture what success means for a given client — common in rehabilitation, case management, mental health, geriatrics, and complex social work cases. It is valuable when you nevertheless need to aggregate outcomes across heterogeneous clients for program evaluation. It is less appropriate when goals are uniform across clients (a standardized measure is then simpler and more comparable), when staff cannot be trained to write balanced, realistic scales, or when the act of goal-setting itself would bias the outcome being measured.
Limitations and considerations
- Validity depends heavily on the skill and consistency with which staff write the five outcome levels; poorly calibrated scales bias scores up or down.
- Setting the expected ('0') level too low or too high systematically inflates or deflates attainment, and there is no external benchmark to catch this.
- The T-score formula assumes a fixed inter-goal correlation and treats ordinal levels as interval data, assumptions that are debated.
- It measures attainment of the goals that were chosen, not whether the right goals were chosen, so a client can score well on trivial goals.
Source conditions do not replace independent scientific review or validation by an analysis engine.
Strengths
- Captures client-specific, personally meaningful goals while still producing a quantitative score that can be aggregated across very different clients.
- Responsive to change because the scale is built around the individual's expected trajectory, often detecting improvement that fixed instruments miss.
- The act of collaboratively defining goal levels is itself clinically useful, clarifying expectations and engaging the client in treatment planning.
- Applicable across an enormous range of fields and outcomes, from rehabilitation and mental health to education and organizational change.
Applications
- Case managers measuring progress toward individualized care-plan goals across a caseload of clients with very different needs.
- Rehabilitation teams quantifying functional goal attainment for patients recovering from stroke or injury.
- Mental health programs evaluating overall effectiveness by aggregating individualized treatment goals into a comparable score.
- Geriatric and dementia care, where standardized outcome measures often fail to capture meaningful, person-centered goals.
- Program evaluation in social services that must demonstrate outcomes across a heterogeneous client population to funders.
How it works
How it works
Worker and client jointly identify a small number of goals (commonly three to five) that are important, realistic, and likely to be influenced by the intervention. Each goal must be framed so that a range of outcomes — not just success or failure — can be observed.
For each goal, five mutually exclusive, behaviorally specific outcome levels are written in advance and mapped to the values -2 (much less than expected), -1 (somewhat less), 0 (the expected/most likely level), +1 (somewhat more), and +2 (much more than expected). Writing the 0 level as the realistically expected outcome is what keeps scores interpretable.
Formula
\{-2, -1, 0, +1, +2\}Each goal may be given a non-negative weight reflecting its relative importance to the client or its difficulty. Equal weights are common and often perform comparably; differential weights let a more important goal contribute more to the overall score.
Formula
w_i \ge 0
At follow-up, the client's actual outcome on each goal is matched to one of the five predefined levels, yielding a score x_i between -2 and +2. Because the levels were written before the outcome was known, this scoring is criterion-referenced rather than a retrospective impression.
Formula
x_i \in \{-2,-1,0,+1,+2\}The weighted level scores are aggregated into a single standardized T-score, scaled so that fully meeting expectations across goals yields 50 with a standard deviation of 10. The denominator includes an assumed inter-goal correlation ρ (Kiresuk and Sherman used 0.30); the result places an individualized outcome on a common metric that can be averaged across clients or programs.
Formula
T = 50 + \frac{10 \sum_i w_i x_i}{\sqrt{(1-\rho)\sum_i w_i^2 + \rho\left(\sum_i w_i\right)^2}}Common pitfalls
- Writing the expected outcome level so conservatively that almost any result lands at +1 or +2, producing artificially high attainment scores.
- Letting the same person who delivers the intervention set the goals and score the outcome without any check, inviting optimistic bias.
- Defining the five levels in vague or overlapping language so that two raters would place the same outcome at different levels.
- Treating the resulting T-scores as if they were measured on a true interval scale and applying statistics that the ordinal, formula-derived metric does not support.
Frequently asked questions
How many goals should a GAS follow-up guide contain?
Most applications use between two and five goals. Too few goals make the overall score volatile — a single goal can dominate — while too many become burdensome to define and score reliably. Three is a common compromise that balances breadth of coverage against the practical effort of writing five well-calibrated outcome levels for each goal.
Why are outcomes scored on a -2 to +2 scale instead of a simple met/not-met judgment?
A binary met/not-met judgment loses information and is sensitive to where the threshold is set. The five-level scale, centered on the realistically expected outcome (0), records both shortfalls (-1, -2) and over-attainment (+1, +2), making the measure more sensitive to graded change and allowing the standardized T-score to reflect how far above or below expectation the client actually landed.
Is the GAS T-score a valid interval measurement?
This is debated. The T-score formula converts ordinal attainment levels into a 50/10 standardized scale and assumes a fixed correlation among goals, so strictly it is a transformation of ordinal data rather than a true interval measure. Many evaluators report the mean level score or use the T-score descriptively, and treat any parametric analysis of GAS outcomes with appropriate caution.
Related methods
Sources
Kiresuk, T. J., & Sherman, R. E. (1968). Goal attainment scaling: A general method for evaluating comprehensive community mental health programs. Community Mental Health Journal, 4(6), 443–453.
Open DOI referenceCopied.Kiresuk, T. J., Smith, A., & Cardillo, J. E. (Eds.). (1994). Goal Attainment Scaling: Applications, Theory, and Measurement. Lawrence Erlbaum Associates.
ISBN: 9780805814040Copied.