Rostersmith
The Science Behind Better Healthcare Scheduling
There is a version of healthcare scheduling that most practices accept as normal. Someone - usually a senior clinician or practice manager - spends days building a roster in a spreadsheet. They balance requests, coverage requirements and call duties by hand, make a few compromises, email the result to the team and wait for the complaints. The complaints come. They always do. Then the process starts again next month.
Most practices treat this as a workflow problem. Something to manage. But peer-reviewed research across multiple disciplines - sleep science, organisational psychology, patient safety and operations research - tells a different story. The way healthcare professionals are scheduled has measurable effects on how they perform, how they feel about their work and how safe their patients are.
This is not opinion. It is evidence.
Fatigue is not just tiredness
The relationship between working hours and cognitive performance has been studied rigorously since the late 1990s. In a landmark study published in Nature, Dawson and Reid found that after 17 hours of sustained wakefulness, cognitive psychomotor performance dropped to a level equivalent to a blood alcohol concentration of 0.05% - the legal driving limit in most countries. After 24 hours awake, performance deteriorated to the equivalent of a 0.10% blood alcohol concentration.
For healthcare professionals working long shifts, overnight calls or back-to-back rotations, these numbers are not abstract. A clinician finishing a 12-hour night shift and driving home may have been awake for 17 hours or more. A doctor on a 24-hour call who is actively working through the night is functioning, in measurable cognitive terms, as though legally intoxicated.
The Joint Commission's Sentinel Event Alert 48 addressed this directly, documenting the link between healthcare worker fatigue and adverse events. Extended work hours contribute to lapses in attention, impaired judgment, slowed reaction times and diminished empathy. The alert called on healthcare organisations to assess fatigue-related risks and design work schedules that actively minimise them - not merely accommodate them.
A systematic review published in 2024 found that medical errors occurred 12.1% more frequently during night or rotating shifts compared with standard day shifts. The risk of self-reported fatigue-related errors increased with each additional night of disrupted sleep.
None of this is caused by bad clinicians. It is caused by bad schedules.
The burnout connection runs deeper than workload
Healthcare worker burnout has reached levels that the research community now describes as a crisis. A 2024 meta-analysis in The American Journal of Medicine examined 38 randomised trials testing interventions to reduce physician burnout. The findings were sobering: individual interventions such as mindfulness programmes, discussion groups and resilience training produced statistically significant but clinically meaningless improvements. The authors noted that burnout symptom prevalence in physicians rose from 38.2% in 2020 to 62.8% in 2021.
What the research consistently identifies as a structural driver of burnout is not patient care itself but the systems surrounding it. Long working hours, limited control over scheduling and perceived unfairness in shift allocation appear repeatedly in the literature as organisational factors that amplify stress and emotional exhaustion.
A qualitative study of emergency and internal medicine physicians published in BMJ Open identified four dominant themes driving burnout: interruptions and noise, interdepartmental conflict, heavy workload and scheduling, and feeling undervalued by leadership. Scheduling was not a peripheral concern. It was one of the four pillars.
The New England Journal of Medicine drew on a meta-analysis of controlled interventions and concluded that schedule flexibility - giving doctors genuine control over when and how they work - was one of the few system-level solutions with evidence of reducing burnout. Not yoga. Not snack rooms. Flexible, fair scheduling.
Fairness is a science, not a feeling
Ask any group of healthcare workers what bothers them most about their roster and the answer is rarely the total number of shifts. It is the distribution. Why does one person always get the undesirable weekend? Why are some requests honoured and others ignored? Why does the person building the roster seem to have the best schedule?
Research confirms that these perceptions are not petty grievances. They are expressions of deeply held fairness norms that have been studied in organisational psychology for decades.
A 2020 study published in the proceedings of the CHI Conference on Human Factors - the premier venue for human-computer interaction research - explored how healthcare workers perceive fairness in shift scheduling. Uhde, Schlicker, Wallach and Hassenzahl found that fairness in scheduling operates on two levels.
At an abstract level, healthcare workers endorse equality as the governing principle: everyone should bear an equal share of undesirable shifts. But when specific conflicts arise - who works Christmas, who covers a particular weekend, who gets a school-holiday request approved - equality gives way. In every case studied, need-based resolution was preferred over seniority or equity-based allocation. The person with the greater personal need should get the accommodation, regardless of rank.
Most critically, the study found that procedural justice - involvement in the scheduling process itself - mattered more than any specific outcome. Healthcare workers who felt excluded from schedule decisions perceived the result as unfair, even when the distribution of shifts was objectively balanced. Conversely, those who felt heard were more accepting of imperfect outcomes.
This has profound implications for how scheduling systems should work. A roster built behind closed doors by a single administrator, no matter how well-intentioned, will generate perceptions of unfairness simply because the process is opaque. Transparency, preference capture and the ability to see why decisions were made are not optional features. They are, according to the research, the foundations of a schedule that people will accept.
The administrative cost of doing it by hand
There is a less visible cost to manual scheduling: the hours consumed by the process itself.
Research presented at the American Society of Anesthesiologists' ADVANCE 2022 conference documented one department's experience in detail. Before implementing an automated system, creating the monthly schedule for 60 anaesthesiologists at Ochsner Health took between 60 and 75 hours per month. A senior clinician or administrator was spending the equivalent of nearly two full working weeks every month building a spreadsheet.
After implementing an optimisation-based scheduling system, the same schedule was generated in 14 hours - an 80% reduction. But the time saving was not the most significant finding. Within six months, physician engagement scores on the Press Ganey survey rose from 3.3 to 4.2 out of 5, representing one of the largest improvements in the health system. The AI-based system granted more holiday requests, reduced ungranted leave days and provided greater flexibility and predictability than the manual approach ever could.
The physicians understood something important: a long week would be reciprocated with a shorter one in the future. That kind of systematic fairness is nearly impossible to achieve manually, because the human mind simply cannot hold enough variables simultaneously.
This is not a peripheral administrative concern. Administrative duties consume an average of 15.5 hours per week for physicians across specialties, according to a study of nearly 1,800 academic physicians. Physicians who reported higher proportions of time spent on administration had lower career satisfaction, higher burnout levels and were more likely to consider reducing their clinical workload. Every hour spent wrestling with a spreadsheet roster is an hour not spent on patient care - or on rest.
What mathematical optimisation changes
The research converges on a set of requirements for healthcare scheduling that no manual process can consistently meet: fatigue-aware shift distribution, equitable allocation of undesirable duties, accommodation of individual preferences, compliance with regulatory hour limits and transparency in decision-making.
These requirements are not new. Operations researchers have studied the healthcare rostering problem for decades. The ANROM (Advanced Nurse Rostering Model), developed by Burke, Curtois, Qu and Vanden Berghe, provides the most comprehensive published constraint taxonomy for healthcare scheduling: three hard constraints and 26 soft constraints across categories covering hospital rules, work regulations and personal preferences. The model has been validated in over 40 Belgian hospitals through a commercial scheduling system, demonstrating that mathematical approaches can handle the real-world complexity that defeats manual methods.
The hard constraints are straightforward: no double-booking, qualified personnel only and minimum staffing levels met. The soft constraints are where the nuance lives - minimum rest between shifts, maximum consecutive working days, weekend distribution patterns, day-off requests and shift-type preferences, each weighted according to organisational priorities.
What mathematical optimisation does is evaluate thousands of possible schedule configurations simultaneously, balancing all of these constraints against each other and returning a solution that satisfies as many as possible while violating none of the non-negotiable rules. A human scheduler working in a spreadsheet evaluates one configuration at a time, making trade-offs based on memory, habit and who complained loudest last month.
Where Rostersmith fits in this picture
Rostersmith was built on these research foundations. Not as an academic exercise but as a working scheduling platform for healthcare practices that need to solve these problems every week.
The scheduling engine uses mathematical optimisation to generate compliant, fair rosters across multiple locations and disciplines. It enforces hard constraints that protect patient safety - minimum rest periods, maximum consecutive shifts, qualification-based assignment. It optimises soft constraints that protect clinician wellbeing - equitable distribution of nights, weekends and public holidays, accommodation of individual preferences and leave requests.
Every scheduling decision is explainable. Rostersmith does not just produce a roster; it tells each practitioner why they were assigned what they were assigned. This directly addresses what the research identifies as the most important dimension of scheduling fairness: procedural justice. When people can see the reasoning, they trust the result.
The system captures preferences, processes leave and generates the entire roster in seconds rather than days. Clinicians get flexibility and predictability. Administrators get hours back. Practices get a defensible, evidence-based process that replaces the annual argument about who builds the roster and whether they do it fairly.
This is not about replacing human judgment. It is about giving human judgment better tools. The roster can be manually adjusted after generation. Shifts can be locked, swapped and reassigned. But the starting point is a schedule that has already accounted for every rule, every preference and every constraint - something no human mind can do reliably across 20 or 30 practitioners and multiple locations.
The evidence is clear
The research does not leave much room for ambiguity. Fatigue from poor scheduling impairs clinical performance to the level of alcohol intoxication. Burnout is driven more by system design than by individual resilience. Fairness in scheduling depends on transparent processes, not just balanced outcomes. Manual scheduling consumes administrative hours that practices cannot afford to lose. And mathematical optimisation - the same class of technology that powers logistics, airline scheduling and financial modelling - produces measurably better results when applied to healthcare rosters.
These are not theoretical claims. They are findings from Nature, the New England Journal of Medicine, the Joint Commission, the CHI Conference on Human Factors and the American Society of Anesthesiologists.
If the way your practice builds its roster has not changed in years, the science suggests it should.
Request a demo to see how Rostersmith applies this research to real healthcare scheduling.