When leaders ask about call-light performance, the answer is often a single number: the average response time. It is easy to compute and easy to report. It is also easy to misread, because averages can hide exactly the experiences that residents and families remember.
Understanding the difference between averages, medians, and the long tail of slower calls leads to better reporting and better decisions.
The problem with averages
An average adds up all response times and divides by the number of calls. If most calls are answered quickly but a few take a long time, the average is pulled upward, yet it still may not reflect any resident's actual experience. Conversely, a handful of very fast calls can pull it down and make things look better than they feel.
Consider a hypothetical unit where most calls are answered within a few minutes but a small number wait considerably longer. The average could look acceptable, even though the residents behind those long waits had a very different experience.
What the median adds
The median is the middle value: half of calls are answered faster, half slower. It is not distorted by a few extreme values, so it describes a typical call better than the average. Reporting the median gives you a more stable picture of normal performance.
On its own, though, the median says nothing about the slow end. A building could have a good median and still have troubling waits.
Look at the long tail
The slow end of the distribution is where much of the resident experience problem lives. Useful ways to look at it include:
- A high percentile, such as the response time that 90 percent of calls beat, which shows how long the slowest tenth of calls wait
- The share of calls over a set threshold, chosen by your team, such as calls taking longer than a standard you define
- The longest waits per week, reviewed individually to see what happened
You choose the thresholds. There is no need to borrow someone else's number. What matters is that your team agrees on a standard that reflects the experience you want residents to have.
Show all three together
A simple report might include:
- The median response time
- A high-percentile response time
- The share of calls above your threshold
- Trends over several weeks, by shift and unit
Together, they provide a fuller description than any one number could.
Make sure the clock is measured consistently
Call-light data comes with definitional choices. Does response time end when someone enters the room or when the light is cancelled? Are calls cleared from the hallway panel counted as responses? Some systems can distinguish these events and others cannot. Understand what your system records and document the definition, so numbers are interpreted the same way each month.
Consider the type of call
Not all calls are equal. Some residents press the button for assistance with mobility or toileting, others for a question, others for a bathroom emergency. If your system can classify calls, or if you can group by room or unit, you may see that certain types take longer. Use this carefully and keep conversations clinical staff-led where the content is about resident needs.
Use it to improve, not to rank
When staff see response data, they may worry it will be used to evaluate individuals. Be clear about the intent: to find system problems such as assignment layout, timing, or staffing, and to support teams. Share unit and shift views broadly, with resident-level detail limited to those who need it.
A hypothetical example
Suppose a hypothetical building shows a stable median across all shifts, but the high-percentile view reveals that long waits cluster in one hallway during meal service. A closer look finds that aides on that hallway are assisting with dining at the same time lights are going off. The team staggers tasks and adds a floating helper for that period. The median barely changed, but the slowest waits shortened. Only the long-tail view showed the problem.
Mistakes to avoid
- Reporting only the average. It can mislead in either direction.
- Choosing a threshold arbitrarily. Involve staff and leaders.
- Ignoring definitions. Make sure everyone knows what is being measured.
- Overreacting to single events. Look for patterns.
- Skipping the human follow-up. The data tells you where to ask questions.
Keep it simple for the floor
Leaders may want percentiles, but frontline teams need simple language. Present results in plain terms, such as "most calls are answered in about X minutes, and the slowest calls tend to happen at Y time." Then ask what would make it easier.
CarePulse Analytics presents call-light response as medians, high percentiles, and threshold counts by unit and shift. If you would like to see your own long tail, a brief demo can show you.