The denominator is where the adoption math lives. Aletta screened 1,743 participants and could not identify a suitable vein in 110 of them, who were referred for manual phlebotomy and excluded from the first-stick calculation [10]. Apply the reported 94.5% to the 1,633 patients who stayed in [9] and you get roughly 1,543 successful first sticks [1] and about 90 misses [2]. Add the screen-outs back, and 200 of the 1,743 people who walked up to the machine, 11.5%, ended with a person holding the needle [3]. End to end, the robot finished 88.5% of arrivals on the first attempt [4]. Call it one draw in nine routed to a human [5].
The subgroups say the stick itself is solid. First-stick success was 92.7% among 411 patients reporting difficult venous access and 93.4% among 695 patients aged 65 or older [11], one to two points off the pooled figure [7]. The sensitivity is upstream, at screening. Aletta stops before inserting if it cannot identify an appropriate vein [7], which protects the numerator by design: a harder patient mix pushes more people into the referral bucket and leaves 94.5% largely intact. A device that declines the cases it does not like and then reports its rate on the ones it accepted is at least being explicit about its denominator, which is more than most benchmark tables manage.
The refusal path is where the craft shows. Onboard sensors can pause a procedure when they detect an unsafe condition, and excessive patient movement detaches the needle automatically [8].
The FDA's three-devices-per-phlebotomist figure [6] is a ceiling on fleet ratio, not a throughput promise. Split the human work up. A trained supervisor initiates each session, remains available during it, checks that tubes were collected in the correct order and verifies they hold enough blood [4]. A trained professional then cleans the machine between patients [5]. Availability amortizes across three devices. Session start, order-of-draw check, fill check and cleaning do not; they scale with draws. Whichever of those runs longest sets the ratio you actually get.
Then the site variance. Pooled hemolysis was 0.3% among samples with available measurements [13], but 5.2% at site 1 and 0% at sites 2 and 3 [14], roughly seventeen times the pooled rate at one location [6]. The robot is the same machine at all three sites. The source names no cause, and I would not guess at one from a single table, but the spread has to come from something the sites do not share, which puts post-collection handling on the list ahead of the needle. Hemolysis can render a sample unusable or distort results, and the repeat draw is precisely the cost this system is meant to remove [15].
On labor, the 2024 American Society for Clinical Pathology vacancy survey found a 9.4% vacancy rate in phlebotomy departments and 9.9% for staff positions, with night and overtime shifts described as difficult to fill [16]. Aletta does not cover a night shift alone, because a trained supervisor still initiates every session [4]. What it changes is how many routine outpatient draws one present phlebotomist can carry while skilled staff take the patients needing closer attention [20]. The published clinical research assessed performance and safety, not labor costs or throughput economics [17]. The savings case remains the buyer's to model.
