Reading the Wing Log
A wing log for a single 24-hour period records the following incidents:
| Incident type | Count |
|---|---|
| Assault (detainee on detainee) | 3 |
| Assault (detainee on officer) | 1 |
| Self-harm | 4 |
| Refusal to comply | 9 |
| Property damage | 2 |
| Medical emergency | 5 |
The wing holds 84 detainees. The previous 24-hour period recorded 18 incidents in total. Overtime hours worked on the wing rose from 22 (previous period) to 31 (this period). The average across the previous four periods was 24 incidents per 24 hours, and average overtime was 20 hours.
1. Calculate the total incidents this period, the percentage change from the previous period, and the percentage change from the four-period average.
Read the worked answer
Total this period: 3 + 1 + 4 + 9 + 2 + 5 = 24 incidents.
vs previous period (18): (24 − 18) / 18 = 6/18 = +33.3%.
vs four-period average (24): (24 − 24) / 24 = 0% - this period equals the average.
Overtime: 31 vs 22 previous = (31 − 22)/22 = 9/22 = +40.9%. vs average of 20 = 11/20 = +55%.
The key insight: the incident count is not elevated against the four-period average, only against the immediately preceding period, which was unusually low. The overtime rise, however, is elevated against both baselines.
2. A manager says 'incidents are up a third and overtime is up 40% - the wing is in crisis.' Assess this claim using the data.
Read the worked answer
The claim is half right and overstated.
- 'Incidents up a third' compares 24 to 18 - a single unusually low period. Against the four-period average of 24, this period is exactly average (0% change). One comparison point is a weak basis for a trend.
- 'Overtime up 40%' is more robust: 31 hours is above both the previous period (22) and the average (20), a rise of 41% and 55% respectively. That is a genuine signal worth investigating.
- Composition matters more than the total. The 24 incidents include 4 self-harm and 5 medical emergencies - 9 of 24 are welfare-related. Four self-harm events in 24 hours on an 84-person wing is a specific, actionable concern, distinct from the 9 refusals to comply.
- Per-capita context: 24 incidents across 84 detainees is roughly 0.29 per detainee per day; without a longer baseline, 'crisis' is a label, not a finding.
Better framing: overtime is genuinely elevated and self-harm events warrant immediate review; the total incident count is within normal range.
3. What additional data would you request before drawing conclusions, and why?
Read the worked answer
- A longer time series (e.g., 12 weeks) to distinguish noise from trend - two periods cannot show a trend, and one low period distorts any percentage change.
- Staffing levels per shift, to test whether overtime correlates with incidents or with vacancies/sickness.
- Breakdown by time of day and sub-unit, to locate where incidents cluster - a wing-wide total can hide a single problematic sub-unit or shift.
- Definitions and recording practice, to check whether the rise reflects more events or more consistent recording (a change in reporting behaviour inflates counts without any real change).
- Repeat individuals, since a few detainees may account for many incidents - this changes the response entirely (individual welfare plans vs wing-wide measures).
The principle: percentages from two data points are fragile, and totals hide the pattern that actually guides action.