The Rounds Journal · Health data
How to choose measures for a hospital operations dashboard
Design a small set of measures around decisions a service team actually makes.
Design a small set of measures around decisions a service team actually makes.
Start with who will use it
Ask what the team reviews, how often it meets and what action it can take. A clinic reviewing appointment access might want to see whether waits differ by appointment type, while a team checking records may need a completeness measure.
Write a definition people can reproduce
For a fictional waiting-time measure, identify the start event, end event, unit, period and excluded rows. Show how many records are missing either timestamp. State who owns the definition and where users can ask about changes.
Give the viewer enough context
Show trends and the number of records behind the measure. If a service changes how it records arrivals, annotate the change before comparing periods. Remove measures that nobody uses for a decision and check whether readers can interpret the remaining ones without help.
Write a measure card
For an illustrative turnaround metric, use the elapsed hours from receipt of a complete request to confirmation. Record how to treat weekends and reopened requests. Show the number measured and the number excluded. Assign someone to confirm the definition before publishing it.
Choose a display that matches the question
A trend can show whether turnaround is changing. A breakdown can show where the differences occur. A single weekly average may hide long waits in one team, so check the distribution before deciding which summary to display. Give small groups enough context to avoid overinterpreting a handful of records.
Test with a real review question
Ask a team member to explain what changed this week, which records the measure covers and what action they would consider. If they cannot tell when the data was refreshed or whether a zero means no work or missing data, fix those labels before adding more charts.
Sketch a dashboard for one meeting before building it
Imagine a fictional outpatient service holds a weekly meeting about administrative delays. Its team can redirect routine requests between queues and ask for a closer review of exceptions. It cannot change staffing budgets in that meeting. Start with measures the people in the room can act on: how many requests are waiting, how long completed requests took and how often forms are returned for missing information. If a number only matters to a different decision-maker, consider putting it in a different report.
Draw a simple page. Put the reporting period and last refresh date in a consistent location. Show the backlog for the present week and the trend for previous weeks. Put the completeness measure next to it so people can ask whether repeated work is contributing to delays. Add a space for an explanatory note when a change in recording or a period of missing data makes a comparison uncertain. Keep labels understandable to someone who was not involved in building the dashboard.
Work through the decision behind each measure
For backlog, agree what counts as open at the reporting cut-off and how a request transferred to another team is handled. For turnaround, decide whether you are looking at the time from first receipt or from the point at which all required information is available. These answer different questions. A manager interested in what the requester experiences may care about time from first receipt. A team studying its own processing capacity may also need the time from complete receipt. Show both only if people have a reason to use each.
For first-review completeness, ask what counts as a complete form, who applies that rule and which request types are included. Display the numerator and denominator or make them available beside the percentage. If a team sends only a few forms in a week, explain what an individual request can do to the result before comparing it with a team that sends hundreds. If the measure misses a type of work, note that scope near the chart.
Prepare for missing data and changing definitions
Test what the dashboard shows when a source fails to refresh. A missing count should be distinguishable from a confirmed zero. Agree who reviews the data before release and what they do if a measure cannot be reproduced. Keep a short record of the extraction dates, definition versions and corrections. If users have routine questions about a measure, record the person who can answer rather than leaving them to guess from the chart.
Watch someone use it against a real decision
Give a colleague a simple scenario: the backlog rose and the completeness percentage fell. Ask them which records they would look at first and what action they would consider. If they jump to a conclusion about staffing without noticing that a new request category was introduced, make the scope change easier to see. At the next meeting, ask which measure actually informed a discussion. If a chart is repeatedly ignored, find out whether it is poorly labelled, has no owner or answers a question the group does not have.
Review the design again after several meetings. Keep a small set of measures people can explain, and document when a new decision justifies adding another one.