The Vanishing Near-Miss Reports

Divergence between indicator families is the first signal. Reporting volume collapsed (41 to 6 per month) while minor injuries requiring first aid nearly tripled (12 to 29). A genuine risk reduction lowers both injury frequency and near-miss frequency; here the two moved in opposite directions, which is the classic signature of reporting suppression or classification drift, not hazard elimination.

Structural incentives explain the pattern. Supervisor bonuses tied to zero recordables create a direct financial reason to under-record and to discourage reports. Observation cards requiring the observer's name suppress candid reporting in a workforce where the observer may be reporting on a peer or a supervisor. Losing two technicians removed the people who processed reports and classified cases, so fewer reports get logged and more get pushed into the first-aid column.

Tests to run (triangulate, don't trust one stream):

• Reconcile the OSHA 300 log against workers' compensation first reports, clinic billing, and absence data for the same period; gaps reveal unlogged cases.
• Pull a sample of 25 to 30 cases classified as first aid and re-score them against the regulatory recordability criteria (medical treatment beyond first aid, days away/restricted, loss of consciousness, significant diagnosis).
• Review 301 descriptions for vagueness and for repeat body parts and repeat workers.
• Interview 30 hourly employees off-shift with supervisors absent.
• Compare the first-aid-to-recordable ratio over three years; a healthy organization runs roughly 8:1 to 12:1. An inverted ratio is diagnostic.
• Check the hours-worked denominator and whether work moved to contractors, whose injuries would not appear in your rate.

Reasoning principle: a rate is a product of both events and behavior about events. Before crediting a program, prove the denominator and the numerator are honest.

Six metrics, each from a system of record rather than a safety-department spreadsheet:

1. Hazard reports per 100 employees per month (target rising toward 8), reported by area. Pair it with percent closed within 14 days and percent closed as 'duplicate/no action' so volume cannot be manufactured or dumped.
2. Aging of safety-flagged maintenance work orders - count older than 90 days and percent completed on schedule, pulled directly from the CMMS.
3. Critical control verification - percent of scheduled checks of engineered safeguards (machine guards, interlocks, local exhaust flow, eyewash and shower activation) completed and percent found functioning. Completion alone is gameable; function is not.
4. Competency verification - percent of employees whose training was confirmed by observed skill demonstration rather than an LMS completion click.
5. Exposure and condition monitoring - sampling results as a percentage of the applicable limit, plus the count of tasks with uncharacterized exposure.
6. Contractor and high-hazard work oversight - percent of permitted high-risk jobs with a documented pre-job verification.

Anti-gaming design:

• Every metric comes from a system the safety function does not hand-enter (CMMS, LMS, industrial hygiene lab, permit system).
• Report distributions by area and shift, not plant averages; suppression becomes visible as a low-reporting outlier.
• Pair every volume metric with a quality metric so a number cannot be inflated without cost.
• Audit a random 5 percent sample monthly and publish the audit error rate.
• Remove recordable counts from supervisor pay entirely and replace with verified control reliability and timely hazard closure.
• State openly to leadership that any single metric will be optimized against once it becomes a target; the set only works as a set.

Reframe the request from a staffing argument to a risk-acceptance decision. 'Make it work' is not an engineering statement; it is a decision to accept a defined set of exposures, and it should be documented as one. Do not argue about headcount. Deliver a one-page decision memo that says what will stop being done at 1.5 FTE, and what that buys in risk:

• Contractor oversight at the gate: an expired scaffold tag is already in the plant. This is the highest-consequence exposure you currently own and the least defensible in an inspection.
• Industrial hygiene sampling and exposure characterization.
• Investigation depth: the backlog of 62 safety-flagged work orders goes unreviewed.
• Case classification and recordkeeping review - which is precisely the control that would catch the pattern you just described.

Include the arithmetic: fully loaded cost of one lost-time injury (wage replacement, overtime coverage, investigation, potential penalty exposure per instance) against the $180,000. Include the regulatory consequence of a willful or repeat characterization if under-recording is later found.

Offer three options with consequences attached: full staffing; 2.0 FTE with a reduced scope; 1.5 FTE with a signed risk-acceptance by the plant manager and a named executive, listing the deferred activities.

If he proceeds: comply with the lawful direction, document your recommendation and the decision in writing, keep a dated copy outside the plant system, and set explicit tripwires (any serious injury, any safety work order past 90 days, any regulatory contact) that trigger escalation to corporate counsel. Never sign a document, certification, or audit response that asserts a level of assurance you cannot actually provide at that staffing level. The professional line is not 'I refuse to work here'; it is 'I will not misrepresent what this program can and cannot do.'