Monthly Nurse Roster — Holding Burnout, Preference and Skill Together

Editor’s note

In everyday speech this problem is called the “nurse rota”, the “monthly schedule” or the “ward roster”. In the academic literature its name is the Nurse Rostering Problem (NRP). Burke, De Causmaecker, Vanden Berghe and Van Landeghem (2004) is the canonical survey; De Causmaecker and Vanden Berghe (2011) is the modern categorisation; Cheang, Li, Lim and Rodrigues (2003) the early bibliographic reference.

This page should not be confused with generic shift scheduling (#003): #003 is the weekly shift decision in restaurants, retail and contact centres — legal hour caps, skill fit and preferences are there, but patient-safety fatigue constraints are not, the seniority mix is weak, and there is no accreditation burden. NRP carries healthcare-specific complexity: ICU vs surgery vs general-ward skill split, max-consecutive-night caps (typically 3 or 4 — grounded in patient-safety research), 11-hour minimum rest between shifts (a European Working Time Directive lineage), very heavy preference weight (nurse turnover is sensitive to burnout), and mandatory fairness reporting (accreditation rule). Call-centre workforce scheduling (#023), operating-room scheduling (#008) and patient flow (#038) are close relatives but the mathematical skeleton differs. Building NRP on a #003 engine usually fails — the extra constraints overload the engine.

Most-skipped point in the sector: patient-safety (fatigue + minimum rest) constraints. Plain shift schedules treat “weekly hour cap” as enough; modern NRP literature models night-shift fatigue (Burke 2004 framework), max consecutive nights (typically 3 or 4), minimum shift-to-shift rest (11 hours from night to morning), and weekly night-shift distribution equity as hard or soft constraints. International health-services research reports that bad rosters raise patient adverse-event rates by 20-40% and measurably lift nurse burnout scores (Burke, Curtois, Qu and Vanden Berghe 2010-related field work). In any demo, ask the vendor for a scenario where a 4th consecutive night cannot be assigned after 3 — an engine that cannot do this is not NRP-ready. The second skipped point is seniority mix. Every ward-shift should carry a hard rule of “at least 1 senior + at least 1 mid + new graduates allowed only with senior cover”; without it, the night shift may run with a lone new graduate and patient-safety risk grows.

A step-by-step path for an SMB hospital

Stage 1 — Measure first, plan second. For at least 12 months, log nurse work data: each nurse’s monthly shift distribution (morning/evening/night counts), consecutive work days, weekend counts, weekly hour total, leave days, breach count (consecutive nights, rest period). Skill matrix: per nurse, certification list (ICU, OR, maternity, paediatric certifications, first-aid, IV-therapy authority). Seniority matrix: senior/mid/new graduate split. Demand pattern: historical nurse need per ward-shift-day.

Stage 2 — Write down the knowledge capital. Which nurse can work which ward (skill fit — no ICU without ICU certification). Contract type (full-time 40h, part-time 24h, on-call). Union rules (weekly hour cap, consecutive nights, mandatory rest). Move verbal preferences to written forms — every nurse fills a “5 preferred days + 5 refused days” form at least once a month.

Stage 3 — Pilot. 8-12 weeks. Run the NRP module in parallel with manual planning on a subset (e.g. monthly roster for a university-hospital ICU, or for a single surgical ward). The decision still belongs to the charge nurse; the system advises. Success criterion in writing, before start: “in 60 days, consecutive-night breach -50%, fairness standard deviation -30%, nurse satisfaction score +15%.” If the bar is missed, the pilot ends — exit right in the contract.

Stage 4 — Rollout. Full hospital + all wards over 6-12 months. Nurse training 1-2 weeks (the mobile app is intuitive). Charge-nurse training is heavier — skill-matrix maintenance, preference scoring, fairness-report reading. Monthly operations review: breach report, fairness distribution, preference-hit rate, replacement time.

Risks — what can go wrong

1. Contract or union rule change. An annual collective agreement or regulatory change requires hard-constraint updates in the system. If the solver hard-codes the rule, the update becomes a project — and the old plan starts breaching in the meantime. The contract must include a rule-update guarantee.
2. Sudden illness + replacement. Sick leave (especially during pandemic surges) breaks the roster daily. Without a replacement-suggestion engine, the charge nurse spends hours on the phone hunting cover. A replacement candidate-list (3-5 backups per ward) must be prepared in advance.
3. Preference breach → burnout + attrition. When soft constraints (preferences) are weighted too low, the system breaches them; the nurse leaves 3-4 months later. This must be measured in exit interviews under “roster unfairness”. Preference weights re-calibrate every 6 months.
4. Single-supplier (WFM software) lock-in. Without a contract clause for “annual standard-format export (CSV, JSON) of the skill matrix, preference history, roster archive and fairness reports”, leaving the system means losing years of accumulated skill + preference + seniority data.

Solution method — technical view

Main approaches in the NRP literature:

Problem sub-type to formulation mapping:

• Cyclic NRP: Fixed monthly pattern, each nurse on a weekly cycle. Small-mid wards. Low preference load.
• Non-cyclic NRP: Re-solved each month. Heavy preference weight. Burke 2004 framework.
• Preference-based NRP: Preference-max objective. JFB-Purnomo 2005 column generation.
• Demand-based NRP: Patient-volume forecast as input. Demand × coverage strict.
• Self-scheduling NRP: Nurse fills in own preferences first, the system fills gaps. High satisfaction, more complex optimisation.

Objective function options:

• Weighted soft-constraint violation sum minimum: The most common formulation (Burke 2004).
• Preference score maximum + violations minimum (lexicographic): Nurse satisfaction first.
• Fairness (night-shift, weekend SD) minimum: Equity-driven.
• Total work cost (overtime + replacement) minimum: Budget-driven.

Most field deployments use a weighted blend of all four; the weights revisit every 6 months.

Close relatives:

• Generic Shift Scheduling (#003): Personnel scheduling without skill + fatigue + accreditation load.
• Call Center Workforce Scheduling (#023): Forecast-driven, continuous-arrival call process.
• Operating Room Scheduling (#008): OR-time scheduling; doctor + nurse + room combination.
• Patient Flow Optimization (#038): Bed occupancy and patient-flow planning.
• Workforce Tour Scheduling (#079 sibling): Broader workforce tour planning.

NRP, with healthcare-specific constraints (skill + fatigue + accreditation + preference weight + contract + union), falls into a distinct mathematical class from other shift problems; an engine that does not recognise this class fails at pilot stage.

Academic references

Listed in the page frontmatter under sources. Burke, De Causmaecker, Vanden Berghe and Van Landeghem (2004) is the canonical survey; De Causmaecker and Vanden Berghe (2011) is the modern categorisation; Cheang, Li, Lim and Rodrigues (2003) is the early bibliographic reference. For column generation see J.F. and Purnomo (2005); for metaheuristics see Burke, Curtois, Qu and Vanden Berghe (2010). NRP has been one of the most-published problems in Journal of Scheduling, European Journal of Operational Research and Journal of the Operational Research Society over the last 20 years.