When leaders ask how long residents wait for help after pressing the call light, the answer is often a single average. It is easy to compute and easy to report, and it is nearly always misleading. Averages blur together quick answers and long waits, and it is the long waits that residents and families remember.
A more useful picture uses at least two numbers: the typical response and the slow end. Together they tell you how the building usually performs and how it performs on its hardest moments.
Median: the typical experience
The median is the middle value: half of calls are answered faster, and half slower. It is less affected by a handful of extreme cases than an average. If the median response time is steady, your typical resident is getting a typical experience.
Think of it as the answer to "what usually happens?"
The 90th percentile: the slow end
The 90th percentile is the response time that nine out of ten calls beat. The remaining one in ten takes longer. This number captures the tail, the waits that stand out in a resident's day.
It answers "how bad does it get on a bad day?" If your median is steady but your 90th percentile is rising, something is happening in specific moments, such as shift change, mealtime, or a short-staffed stretch, that the typical view hides.
A hypothetical illustration
Picture a hypothetical building where the median response is steady from month to month, yet the 90th percentile response has climbed from 12 to 18 minutes. An average-only report might show a small, unremarkable increase. The percentile view shows that a meaningful share of residents are waiting noticeably longer, and it invites the question: when, and where?
The numbers here are purely illustrative; your building's values will differ, and what counts as acceptable is a decision for your leadership and clinical teams.
Add the longest waits
Beyond percentiles, keep a short list of the longest individual waits each week. Reviewing the specific events, with time of day, unit and staffing context, often reveals causes that no aggregate would show: a unit with a single aide covering for a meal break, a call routed to a station nobody was watching, or a device problem.
Slice before you conclude
Response data becomes actionable when you divide it.
- By shift and hour. Shift change and meal periods are common pressure points.
- By unit and hall. Layout, acuity and staffing differ.
- By call type, if your system distinguishes them, such as bathroom assistance, pain, or emergency alerts.
- By day of week. Weekends often differ from weekdays.
Look for consistent patterns rather than one-off blips.
Beware of data quirks
Call-light data has quirks worth knowing. Calls canceled at the bedside, calls answered at the door versus completed at the bedside, and test calls can all affect the numbers. Agree with your nurse call vendor and your clinical leaders on how events are counted, document it, and apply it consistently. A metric nobody trusts will not be used.
Use the data supportively
Staff can reasonably worry that response data is a stick. Frame it as a way to see where the system is stressed, not who is at fault. If response slows on a hall during mealtimes, the answer may be a staffing assignment change, not a lecture. When staff see the data lead to practical support, they are far more willing to engage.
Build a simple routine
- Review median and 90th percentile weekly, by unit and shift.
- Look at the five longest waits and note what was going on.
- Choose one pattern to test with a small change, such as adjusting break coverage.
- Check the same view the following weeks to see whether it helped.
Small, specific experiments beat sweeping initiatives.
Where CarePulse fits
CarePulse reads call-light data from your nurse call system and presents median, tail and longest-wait views by unit, shift and hour, so floor leaders and administrators are looking at the same facts. If you would like to see your own response patterns laid out this way, we would be glad to show you in a demo.