a hospital running a 10-robot fleet for drug and specimen delivery
Hospital Drug and Specimen Delivery Robots Cut Delivery Time 32 to 36 Percent
A peer reviewed study of a hospital robot fleet found 32 to 36 percent faster drug and specimen delivery and 7.3 times more trips than 19 manual staff.
- 32 to 36%
- faster delivery than manual runs (p below 0.001)
- 7.3x
- more delivery trips than 19 manual workers over six months
- 0.004%
- robot omission rate, versus 0.056% manual
- 6.928M RMB
- modeled 10 year net savings (projection, not observed cash)
Based on a documented real-world deployment. Figures are from public reporting; the organization is not named.
Staff time pulled into the corridors
Drug and specimen runs in a busy tertiary hospital quietly consume skilled hours. Each trip takes a staff member away from bedside or lab work, and the timing of a dispensed medication or a specimen can matter as much as the item itself. Those runs also compete for the same corridors and elevators as everything else in the building.
The study behind this case was designed to test whether a fleet of autonomous mobile robots could carry that load without losing accuracy. It ran for six months at a single center, with robots and manual staff working in parallel so the two groups could be compared over the same period.
A separate 2026 feasibility study of medication delivery robots at a tertiary hospital found performance sensitive to elevator congestion, which is a reminder that hallway traffic shapes results as much as the robot does.
- Medication and specimen runs sharing corridors and elevators with other traffic
- Delivery timing that feeds directly into clinical and lab workflows
- Verification and omission risk at every handoff
- Staff hours that could go to patient-facing work instead
Navigation, coordination and a controlled comparison
The robot group used a layered navigation stack. Dijkstra's algorithm handled global path planning, while A* combined with the Dynamic Window Approach managed local navigation around people and obstacles. A queuing model coordinated the ten robots so that jobs were assigned across the fleet rather than piling onto one unit.
The study design is the real backbone of this case. Ten autonomous mobile robots ran alongside a manual delivery group of 19 staff, and outcomes covered delivery time, trip volume, omission and verification accuracy, item integrity, and satisfaction among nurses and patients. The robot system was supplied by a vendor the paper does not need readers to know by name.
The published material focuses on the system and its measurements. It does not describe a field service model, training program, or go live sequence, so anyone planning a similar fleet should ask their integrator for those details in writing.
- Ten autonomous mobile robots on a fixed navigation stack
- A matched manual delivery group of 19 staff for comparison
- Six months of parallel measurement at one tertiary hospital
- Outcome tracking for time, trips, omission, verification, integrity and satisfaction
Speed, volume, accuracy and the cost model
Robot delivery cut delivery time by 32 to 36 percent compared with manual delivery, a difference reported at p below 0.001. The ten robot fleet completed 7.3 times as many delivery trips as the 19 manual workers over six months. Those are the two headline figures, and both come straight from the journal's reporting.
Accuracy moved in the same direction. The robot group reached 100 percent verification accuracy and item integrity, against 97 percent verification accuracy and 99 percent integrity for manual delivery. Omission fell from 0.056 percent manually to 0.004 percent with robots, a significantly lower rate at p equals 0.011.
Satisfaction also rose. Nurse satisfaction was 97 percent in the robot group against 94 percent manually, and patient satisfaction was 93 percent against 90 percent, both significantly higher at p below 0.05.
The ten year cost figures are a model, not a ledger. The study projects net savings of 6.928 million RMB over a ten year lifecycle, with sensitivity runs between 5.36 and 8.61 million RMB depending on labor cost growth of 2, 3 or 4 percent a year. The baseline model compares 12.63 million RMB of labor cost with 5.705 million RMB of robot system cost over the same decade.
Trip counts alone do not settle the staffing question. A 2026 arXiv paper that cites this study argues that staff outcomes should rest on scheduled work and released capacity, not on trips completed.
What the lifecycle around the robots requires
The study shows that the robots performed. It cannot tell a hospital how to match a platform to its floor plan, how to map the building, or how to keep a fleet running after go live. Those questions sit between the paper and the ward, and that is where Service Robot Co. works.
Service Robot Co. is a full service, OEM neutral commercial robot integrator for US businesses. We select the right robots across manufacturers, then lease, deploy, integrate, train and service every unit through a nationwide US engineer network, so one vendor covers the whole lifecycle. We lease robots only, through financing on a three year lease with one fixed payment quoted after a site assessment.
This hospital is a documented real world example we analyze, not a Service Robot Co. client, and the study's robots were not ours. Elevator traffic, lab schedules and corridor widths differ from building to building, which is why the site assessment comes first. The lesson for a planning team is practical: measure the workflow in your own building before you trust any headline percentage.