Fourteen hand lacerations: investigating a trend, not an event

Scope: both, with the aggregate as the lead.

No individual laceration is a rich enough event to reveal system causes - a single cut has too many possible explanations and will end in a plausible but untested story. Fourteen events sharing a mechanism over 18 months give you variation to compare: what changed, when, and for whom. That is a trend investigation.

But aggregate data alone gives you pattern without mechanism. So select two or three recent cases for full case investigation - the most recent on the deboning line and one box-opening case - to establish how the cuts actually happen: which hand, which knife motion, which glove, what the worker was reaching for. The trend tells you where to look; the cases tell you what to look at.

Data to pull

• All 14 register entries with date, time, shift, task, body part and treatment; plot them on a timeline.
• First-aid log and any A&E or GP attendances not in the register, to test whether reporting has changed rather than injury rate.
• Glove specification and issue records, and the date of the C-to-A change; knife purchase, sharpening schedule and any sharpness checks; training and induction records; agency versus permanent status and time in job.
• Production volume per month, overtime and shift patterns; near-miss report texts, not just counts.

Dealing with 'carelessness' Treat it as a coding artefact. 'Carelessness' is a conclusion that stops inquiry, and nine such entries tell you about the coder, not the work. Re-code all 14 against a fixed, neutral taxonomy (task, mechanism, equipment, glove condition, tenure) with two people coding independently and reconciling differences - this also tests how reliable the register is as a data source. State in the report that the original coding is unreliable and why.

Hypotheses and tests

1. Glove downgrade from cut level C to A. Test: plot laceration dates against the change date. If cuts cluster after it, and the pre-change rate was stable, that is strong. Check which gloves were actually being worn at the time of each injury (issue records and interviews) - a specification change that workers ignore changes nothing.
2. In-house sharpening producing poorly maintained blades. Test: sharpening records, any sharpness checks, and whether cuts cluster in the 14 months since the change; ask workers whether knives 'drag' or are used with more force. Dull blades increase the force needed and can slip.
3. New or agency staff with less experience. Test: time-in-role at the date of each injury; compare agency and permanent rates per hours worked, not raw counts. High turnover predicts a cohort effect; if cuts are spread across experienced staff, this weakens.
4. Volume and pace increase. Test: monthly injuries against monthly throughput and overtime; look for cuts late in shifts and at end-of-shift box-opening, when pace and fatigue combine.
5. Reporting artefact. Test: compare first-aid log, near-miss reports and any external treatment data. If the register rise coincides with the new supervisor and near-miss reporting also rose, part of the increase may be better reporting of an unchanged underlying rate - which is good news about culture, not about risk, and must not be misread.
6. Task-specific equipment, e.g. box-opening with knives rather than safety cutters. Test: the case investigations - what tool was in hand.

Analysis discipline Plot by month and mark the glove change, sharpening change and supervisor start as annotations. Multiple changes within 18 months mean the effects are confounded - say so, and avoid claiming a single cause. Where two explanations remain plausible, recommend controls that address both rather than choosing the more convenient one.

Why the toolbox talk is weak as the primary control It assumes the problem is a knowledge or attitude deficit. The evidence points elsewhere: a deliberate reduction in glove cut resistance for cost reasons, a change in who sharpens knives, high turnover among agency staff and a volume increase. A talk does not change any of those. It also sits at the bottom of the hierarchy of control - it relies on every worker, every time, compensating for equipment and conditions that are not fit for the task, and it will not protect a new agency worker in their first week.

There is a second problem: the phrase 'carelessness' in the register suggests the committee has already decided the cause is worker behaviour. A talk based on that premise will be heard as blame and will suppress the near-miss reporting that is currently improving - the site's best source of early warning.

What to recommend instead Specify gloves by task risk rather than one glove for all food handling - reintroduce a higher cut level where the risk assessment supports it, and verify that the specified glove is actually worn. Set a knife sharpness standard with checks and records. Provide safety cutters for box-opening. Strengthen induction and buddying for agency staff on the deboning line. Keep the toolbox talk, but as reinforcement of the controls, not a substitute for them.

Measurement plan

• Leading: percentage of glove compliance observed at unannounced checks; sharpness checks completed against schedule; near-miss reports per 100 workers per month (watch for a fall, which would signal fear); time-in-role at induction completion.
• Lagging: lacerations per 100,000 hours worked, split by task and by agency/permanent; severity (first aid versus medical treatment).
• Design: baseline now, review at three and six months, with defined success criteria (for example, a 50% reduction in deboning-line lacerations at six months with no fall in near-miss reporting). Report whether the controls were implemented before judging whether they worked - otherwise a failed implementation is misread as a failed control.