i Care TracerHospital performance indicators — how they are built and read
Most hospital indicator dashboards fail not from too few numbers but from too many, poorly built. This page is about constructing an indicator that survives questioning — its six elements, its four families, and the measurement rules that should never be broken.
Why most dashboards fail
Open an indicator dashboard in almost any hospital and the symptoms are the same: forty indicators of which three are ever discussed, numbers nobody can trace precisely, and measures that have not moved in a year because nobody actually collects them.
The cause is always the same: the dashboard was built bottom-up — what data do we have? — instead of top-down — what decisions do we need to make, and what number would change them?
An indicator that changes no decision is worth nothing however precise it is. That is the first test to put to any indicator before adopting it: if this number moved, who would do something differently?
Anatomy of a sound indicator — six elements
A number without these six is an observation, not an indicator. Any measure in your facility missing one of them will collapse at the first serious question about it.
| Element | The question it answers | What its absence costs |
|---|---|---|
| Definition | What exactly are we measuring? | Each unit computes it its own way, and comparison is meaningless |
| Numerator | What are we counting? | One thing is counted and another is read |
| Denominator | Out of what? | The most consequential element — a number without one compares to nothing |
| Data source | Where does the number come from? | It cannot be verified or reproduced |
| Frequency | How often is it measured? | Intermittent measurement makes no trend |
| Owner | Who is responsible for it? | An unowned indicator is neither collected, interpreted, nor moved |
Four families — and a sound dashboard carries all four
A dashboard measuring one family gives a misleading picture. A hospital excellent on efficiency and weak on safety looks successful on a purely financial dashboard.
| Family | What it answers | Examples |
|---|---|---|
| Safety | Are we harming anyone unintentionally? | Healthcare-associated infection · falls · pressure injuries · medication errors |
| Clinical effectiveness | Are we treating correctly? | Unplanned readmission · clinical pathway adherence · reviewable mortality |
| Efficiency and flow | Are we using resources well? | Length of stay · waiting time · laboratory turnaround · cancelled operations |
| Patient experience | How was it from their side? | Experience score · recommendation measure · complaint closure time |
Process and outcome indicators — and why you need both
An outcome indicator says what happened to the patient: an infection rate, a readmission rate. A process indicator says whether we did what we should: hand hygiene compliance, prevention-bundle application.
The practical difference between them is time. Outcome indicators are honest but slow — a meaningful change may take months. Process indicators are fast but insufficient alone: perfect compliance with the wrong bundle lowers nothing.
So the rule is to pair them: for every outcome indicator you are working on, at least one process indicator that explains its movement. When the infection rate falls you know why; when it does not, you know whether the reason is that the bundle is not being applied at all.
Five rules that should not be broken
1 · Definitions are not changed undocumented
Changing a definition severs the time series. If the change is necessary, document it with its date and annotate the point on the chart. A sudden drop caused by a new definition is misread as improvement, and decisions get built on it.
2 · No comparison without a comparable denominator
Comparing a surgical unit with a medical unit on falls without case-mix adjustment produces an arithmetically correct and entirely wrong result. Compare with yourself over time first, and with others second and carefully.
3 · Small cells are suppressed
A unit with three cases a month: one case moves the rate by thirty-three per cent. Do not present percentages on small counts — show the raw count or aggregate a longer period. In patient-experience measures there is a second reason: a small cell can point to an identifiable individual.
4 · No target without a source
A target chosen in a meeting because it sounds reasonable is managed by negotiation rather than by improvement. Give every target a written source: your own historical performance, a published reference, or a regulatory requirement.
5 · Every indicator has an owner and a review cadence
An indicator not discussed in a minuted recurring meeting stops being updated within months, then sits on the dashboard as a dead number misleading whoever reads it.
How many indicators do you need?
The wrong question is how many indicators should we measure. The right one: how many can the quality committee genuinely discuss each month in a discussion that ends in a decision?
In most facilities the answer is far smaller than what their dashboards carry. A few indicators that are discussed and moved are worth immeasurably more than dozens displayed and never questioned.
There is also a difference between what you measure and what you display. Measure what you need; display at leadership level what warrants a decision. The rest stays at unit level, visible to whoever owns it.
From indicator to action — otherwise there is no point
An indicator displayed with nothing following from it teaches the team that display is the purpose. Bind every indicator to written trigger rules before you need them:
- A threshold breached → a corrective action with an owner and a date.
- An adverse trend across three consecutive periods → a formal review even if no threshold was breached. The trend matters more than the point.
- A serious incident → root cause analysis, not a direct corrective action.
- An unexplained sudden improvement → verify the data before celebrating. Most sudden improvements come from a change in collection rather than in performance.
Frequently asked
What is the difference between an indicator and a target?
The indicator is what you measure; the target is the value you are aiming for. An indicator remains valid even when its target changes — but changing the indicator's own definition severs its time series and must be documented.
Should we use standard indicators or design our own?
Start from known, published definitions wherever they exist: they permit external comparison and spare you an internal argument about definitions. Design your own only when measuring something no existing definition covers.
Why not aggregate indicators into one quality score?
Because aggregation hides exactly what you need to see. A facility excellent on nine indicators and collapsing on one safety measure looks good in a composite. The right decision needs the collapsing measure visible, not dissolved into an average.
How many periods before we can claim improvement?
One point is not improvement. The practical minimum is three consecutive periods in the same direction with a documented intervention between them. Without the documented intervention you cannot attribute the change to your work at all.
What do we do with an indicator we cannot collect accurately?
Either fix its data source or drop it. An indicator collected by estimation is worse than no indicator, because it creates a false sense of knowledge that a decision then rests on.
Related guides
What does CBAHI require from hospitals?
What actually has to be in place before an accreditation survey — and the six mistakes that most often cost facilities points, written from the field rather than from the manual.
Ten recurring accreditation findings — and their real cause
Every recurring finding has a systemic cause, not an individual one. Ten findings by domain, each with its root cause, its fix, and how closure is proven.
Measurement is part of a larger picture
The readiness check assesses every domain of your facility — not measurement alone — in five minutes, with no account.
Start the readiness check