In terms of uptake, at PHC level, 253/254 (99.6%) and 895/897 (99.8%) samples were tested by nurses and laboratory technicians respectively, and at referral hospitals, all samples were tested. There was no significant difference of detection yield between nurses and laboratory technicians. Although the proportion of invalid and error results was overall very low (< 5%) it was significantly higher when testing was done by nurses vs laboratory technicians at both PHC and referral hospitals (Table 1).

All PHCs participated in both EQA panels but one due to a missing EQA panel. There were 2/9 (33.3%) PHCs with nurses that had results > 87.5% in all EQA panels vs 9/14 (64.3%) for PHC with laboratory technicians, p=0.487. At referral hospital, 1/4 sites with Ultra testing done by nurse and 3/11 sites with testing done by laboratory technician did not participate in any EQA panel because panels had expired when the sites were randomized to the intervention. Since the TB-Speed pneumonia study had a stepped wedge design, the number of EQA panels per site depended on the time of randomization to the intervention of each site. There were 2/3 (66.7%) sites that had all EQA panel results > 87.5% vs 6/8 (75.0%) in sites with nurses or laboratory technicians (P=0.640), respectively (Figure 1).

The median TAT of Ultra testing from sample reception to Ultra results at PHC levels, was 1.13 hour (h) [IQR 1.08; 1.79] for nurses and 1.10h [1.07; 1.45] for laboratory technicians, p < 0.001. The TAT was below 1h30 for 158/252 (62.7%) samples tested by nurses vs 677/893 (75.8%) for samples tested by laboratory technicians, p<0.001. At referral hospitals, the TAT from sample collection to results delivery was 1.95h [1.5; 2.87] for nurses and 2.87h [1.97; 4.81] for laboratory technicians, p <0.001. The TAT was within 3h for 201/258 (77.9%) samples for nurses vs 464/874 (53.1%) for laboratory technicians, p<0.001.

The median performance score results ranged between 4 and 5 out of 5 points for all performance indicators at PHC levels for sites with nurses and laboratory technicians (Table 2).

Difficulties were mostly related to the occurrence of technical errors when running the tests. During interviews, nurses reported that, during the first few weeks, performing Ultra testing was not so easy, but that they had felt more confident with time and practice: “First, the process is difficult, long process for it. For other work process, if we do more and more, we get faster, but for this process, it is related to a lot technical, like computer and typing. It takes time it is difficult” (Cambodia-PHC Nurse). Among the challenges, nurses reported experiencing technical issues with scanning barcode of the cartridge: “Sometimes, scanning barcode doesn’t match with cartridge. So, the cartridge will be broken, and we need to change new one. When scanning doesn’t match, we disconnect the cable from computer and then connect and scan again. Sometime, we put it in machine but computer also errors. It ran for a while then it can run backward. I had met this challenge. Then, I tried to put it again and call to Mr. XXXX and he suggested to put it again” (Cambodia-PHC Nurse); inserting the cartridge in the GeneXpert module: “With the machine, what we often have difficulty with is the last stage where the machine has to open, the part where you put the cassette in, and it often refuses, so you have to start again several times. (…) So that’s the difficulty. Otherwise, when it works, there are no problems” (Cote d’Ivoire PHC Nurse); and using the computer for documentation and navigating the platform: “My dear, we don’t know how to handle computers, that’s our problem (laughs). We don’t know how to handle computers. I myself had to…it’s because I had to force it…force it a little bit, that’s my little worry, technology is not my forte. But as I wanted. I was curious, so I forced it” (Cote d’Ivoire, ward nurse).) Another challenge more specific to PHC nurses was the lack of practice, due to few samples to test each week, “If I did not do it long time, I will forget it again. … I get confused. If I do it more often I will remember it.” (Cambodia-PHC nurse). On the other hand, some nurses reported the issue of increase in workload, especially when they have to re-run a test due to error result: “….it’s an extra activity yeah, …. it’s the lab to run except the RDT, now in this case it me to wait for probably 87 minutes for it to run, if it gives me error then I need to retrieve the bio-banked sample I start preparing, I has given me a work overload” (Zambia, Ward Nurse). Nurses highlighted several experiences of facilitators that included the training and training materials that they could consult after training and the support during supervision visits: “When I take the NPA sample, I mark everything because you have to mark the time of sampling. I go to my laboratory because it’s a small lab there, I go to the lab, the procedure for doing the technique is there, I look at the procedure, I technique the NPA and then I put it into the Xpert and I wait 87 minutes and then the result comes out.” (Cote d’Ivoire PHC Nurse). Nurse explained that the technical challenges could be overcome by contacting laboratory technicians: “Sometimes when we have such errors we usually call for a meeting with the lab so that we help and understand why we are having such errors, so they usually attend.” (Zambia, ward nurse). They appreciated the battery operated equipment: “(…) the machine itself runs on an autonomous battery, and we have up to two batteries. So even if there’s a power cut, we can work”. (Cote d’Ivoire PHC Nurse). Additional results can be found in Supplementary material Table S3.