Outcome, process, and balancing measures
A project can be implemented exactly as planned without improving the outcome. It can also improve one part of a service while creating a problem elsewhere. You need measures that distinguish these possibilities.
IHI describes three complementary types of measures. An outcome measure addresses the result you want. A process measure shows whether relevant steps occurred. A balancing measure looks for an unwanted consequence of the change. [1]
Think through one project
Hypothetical example. A clinic introduces a referral tracking workflow to reduce the time patients wait for confirmation that a referral was received. Staff check a shared queue each workday and contact the receiving service when information is missing.
| Measure type | Question in this project | Possible measure |
|---|---|---|
| Outcome | Is referral confirmation happening sooner? | Days from referral submission to documented receipt by the receiving service. |
| Process | Are staff carrying out the new workflow? | Percentage of eligible workdays on which the queue review is completed. |
| Balancing | Is the new workflow creating another problem? | Staff time spent managing the queue each workday. |
These choices are illustrative, not a required measurement set. Select a small set you can collect reliably and use to make decisions. Add measures because they answer an important question, not because a longer list appears more rigorous.
Classification depends on the aim
The same indicator can play different roles. Timely queue review is a process measure when the aim is faster referral confirmation. If the immediate project aim is to establish dependable queue review, it may be the main implementation outcome. State the level of the outcome rather than presenting workflow completion as a patient health benefit.
A balancing measure should follow a plausible concern. For this example, extra staff time is connected to the intervention. An unrelated annual satisfaction score would be less informative, especially if it is unavailable during the project.
Match the numerator to the denominator
Suppose the queue was reviewed on 16 of 20 eligible workdays. Completion is 80%. Dividing the number of reviews by the number of patients would answer a different question and could exceed 100%.
For any percentage, the numerator must be a subset of the denominator. Decide whether you are counting patients, visits, shifts, tasks, or opportunities. A person may generate several opportunities, so “per patient” and “per opportunity” are not interchangeable.
Read the measures together
High process completion with little outcome change raises questions about the intervention, its timing, or the outcome measure. Low process completion means the intended intervention may not have been adequately delivered. An improved outcome with a growing burden raises a sustainability question. None of these patterns can be understood well from one number alone.
Your next decision
Draw a simple chain from the work you will introduce to the result you expect. Place a process measure on the work, an outcome measure on the result, and a balancing measure beside the most plausible unintended consequence. See Did the intervention actually happen? Measuring fidelity for fidelity and Building and reading a run chart for tracking change over time.
Sources
[1] Institute for Healthcare Improvement. (n.d.). Model for Improvement: Establishing measures.