# A signing bonus—but how many patients? Version 3

Publisher: Oh My Job Research. Collection: September 9, 2026 UTC.

## Scope and results

This descriptive audit covers 1,677 retained US hospital RN postings from 44 employer groups. Of 398 postings explicitly advertising recruitment bonuses, 383 (96.2%) included no identified numeric patient-load figure. These bonus postings came from 30 groups. Across the entire panel, 80 postings contained a numeric patient-load figure.

This is a purposively selected sample of accessible employer boards, not a nationally representative survey. The 15 bonus-plus-load cases are disclosure counts, not a comparison cohort for compensation, working conditions or patient safety. Absence of information does not establish concealment, unsafe staffing or what is communicated in interviews.

## Expansion and selection

The frozen [version 2 methodology](../v2/methodology.md) documents the original 1,575 records, which remain unchanged. Version 3 adds DCH Health System (22 records), Baptist Health Alabama (Orlando Health, 30) and OSF HealthCare (50). All captures remain within a single 24-hour window; exact timestamps are in the JSON. This was an iterative expansion, not a preregistered study or an independent replication.

Employer admission required at least 50 non-managerial RN titles and at least 20 titles satisfying the existing, narrower inpatient/clinical RN screen. An initial 50-report-eligible-title threshold was revised after listing-only feasibility checks, before descriptions were reviewed for bonus or ratio outcomes. Every admitted employer was retained, including Baptist, which contributed no qualifying bonus offers. Failed endpoints and below-threshold employers are retained in private verification logs. Search results alone were not accepted as proof of active inventory.

The unchanged title screen requires RN, registered nurse, clinical nurse or staff nurse wording plus specified inpatient, ICU, medical-surgical, telemetry, stepdown, emergency or maternity terms. Managerial, advanced-practice, outpatient and support titles are excluded. This misses relevant roles whose titles do not identify their clinical setting. It may affect employers differently.

Candidates are ranked by SHA-256 using the unchanged seed `oh-my-job-hospital-rn-disclosure-2026-09-v2`, before inspecting descriptions, with at most 50 postings per employer. No replacements were made. All 102 selected new descriptions were retrieved, with explicit US location checks and no collection errors. Smaller caps are nested sensitivity checks, not independent replications. Group boards are not split into separate facilities. Baptist's board is labeled as part of Orlando Health and is not treated as a separate parent from another Orlando board.

## Sources and completeness

The panel contains 32 Workday groups, nine iCIMS groups, two SmartRecruiters groups and Tenet's career site. OSF's iCIMS career API is used for listing enumeration only; its public JobPosting page supplies the description. DCH and Baptist use iCIMS search pages and public JobPosting descriptions. No employer text is rewritten. Full descriptions, URLs and hashes are retained privately; public downloads contain short evidence excerpts.

Numeric patient-load figures appeared in 13 of 333 bonus-advertising Workday postings and 2 of 43 iCIMS postings, with the latter coming from DCH and OSF. None were identified in the 5 SmartRecruiters or 17 Tenet bonus-advertising postings. Positive cases are no longer exclusive to Workday, but 13 of 15 still come from that source. We cannot rule out that description completeness varies by platform. This audit does not distinguish employer disclosure practices from generally more detailed descriptions or differences in capture completeness. It does not estimate an independent platform effect. Employer weighting and the sensitivity checks do not resolve this limitation.

The panel is not matched on description length, clinical unit or general information richness. The two new positive cases establish that the collection and classification can identify numeric patient loads outside Workday; they do not establish measurement equivalence across platforms.

## Coding and review

The unchanged bonus rules require an explicit recruitment offer in the title or description, including sign-on, signing, hiring, employment and combined sign-on/relocation bonuses. Conditional candidate eligibility is not a guaranteed payout. Generic eligibility-only wording (46 records) and ambiguous or unspecified bonuses (107) are excluded from the main denominator. Referral, annual, retention and completion bonuses alone do not qualify.

Numeric ratios, ranges and equivalent patients-per-nurse statements count, including written-out numbers. A guaranteed maximum is not required. Bed counts, training ratios, identifiers and generic promises of favorable staffing do not count. Existing version 2 review is preserved. Numeric-match passages and broader workload candidates from the new employers received contextual review; this is not independent double-coded annotation or proof of perfect recall.

## Sensitivity and missing data

- Two largest contributors: 19.3% of bonus postings.
- Equal weight across the 30 contributing employer groups: 93.7% without an identified numeric patient load. Small-denominator groups affect this average substantially; DCH contributes one bonus posting, which contains a ratio.
- Removing each possible pair of groups: 95.4%–97.9% without numeric patient load.
- Caps of 10, 25 and 50: 98.9%, 96.8% and 96.2%, respectively.
- Conservative employer/title grouping: 94.9% of 293 groups without a numeric load; this is not a count of unique vacancies.
- Including ambiguous bonuses: 96.0% of 505 postings without numeric load.
- Assigning all six unresolved original Tenet records to bonus-plus-ratio: 94.8% without numeric load. This addresses those records only, not all unobserved vacancies.

Identified locations span 39 states; 702 original records still have no resolved state. No geographic weighting or regional estimates are provided. HCA remains absent because its search access was blocked. Accessible boards do not form a national sampling frame, and these ranges are not national confidence intervals.

## Reproduction and archives

Filter `bonusStatus` to `advertised` to reproduce 398, then `numericRatio` to false to reproduce 383. CSV and JSON contain identical row values. Source summaries, employer counts, timestamps, collection hashes and expansion rules are included in the JSON. `scripts/verify-nurse-board-expansion.ts` records candidate boards and verification evidence; `scripts/extend-nurse-sector-panel.ts` creates a new version without replacing archived files or writing to the jobs database.

Version 1 and version 2 remain archived. Differences between versions reflect changes in scope and coverage, not a time trend.