**Background:** Preterm birth and resulting respiratory failure are leading causes of neonatal death worldwide, with the majority occurring in resource-constrained settings. Continuous positive airway pressure (CPAP), particularly bubble CPAP (bCPAP), is standard of care for respiratory distress syndrome and reduces mortality in premature infants. Commercial bCPAP devices cost several thousand US dollars, placing them out of reach for many facilities in low-resource settings. The World Health Organization provides guidance on improvised bCPAP devices assembled from available components, but these have variable performance and most lack oxygen blending capability, forcing clinicians to deliver 100% oxygen—which WHO strongly advises against due to risks of retinopathy of prematurity, chronic lung disease, and brain injury. Commercial oxygen blenders cost approximately US$1,000 each and require pressurized air, which is rarely available in resource-constrained settings. To address this gap, PATH and a multidisciplinary team developed a very low-cost bCPAP device with two fixed-ratio oxygen blenders (37% and 60%) that do not require pressurized air, functioning via Bernoulli's principle to entrain room air. This study assessed the feasibility, usability, and acceptability of this device in a clinical setting; it was not designed to evaluate device effectiveness.
**Methods:** This early feasibility study was conducted from October 2021 through January 2022 at the Kiwoko newborn care unit, a level two unit in rural Uganda that serves as a referral center for three districts. The unit had used improvised bCPAP since 2012 with a historical mortality rate of 34% among bCPAP-treated patients. Prior to the study, patients had intermittent access to blended oxygen via Y-connection from limited air compressors that frequently failed during continuous use. Fourteen neonates with respiratory failure requiring bCPAP were enrolled; 57 patients received bCPAP during the study period, with 27 not enrolled due to consent not being available within 24 hours, 5 declining consent, and 11 treated during an enrollment pause. Healthcare workers received two days of intensive training before the study. Research assistants were present 24/7 to collect data through observation, case report forms, Likert-type usability scales completed by nurses after device setup, and semi-structured interviews with 15 healthcare workers (13 nurses/midwives with median 5 years of experience, and 2 doctors). Data analysis used descriptive statistics (frequencies, means, medians, IQR) for quantitative data and descriptive coding with content analysis for qualitative interview data.
**Key Results:** Fourteen neonates were treated with the study bCPAP device. Patients had a median gestational age of 31.5 weeks (IQR 26–32), median birth weight of 1310 g (IQR 1005–1760), 57% were born onsite, and 50% were female. Median Apgar scores were 7 (IQR 7–8) at 1 minute and 8 (IQR 8–9) at 5 minutes. Prior to bCPAP initiation, 11 of 14 patients were on nasal cannula oxygen, and 10 of 12 patients with recorded data had saturations >95%. All patients were started at 5 cm H₂O CPAP pressure; median peak pressure was 6 cm H₂O (IQR 5–6). Twelve of 14 patients (85.7%) were treated with the blenders during their course; the 37% blender was in use 90% of the time. Median treatment length was 3 days (IQR 2–6). bCPAP was stopped due to clinical improvement in 12 of 14 patients (86%) and death in 2 of 14 (14%). No patient stopped due to complications. All 14 patients experienced episodes of hyperoxia (saturations >95% lasting >5 minutes). No occurrences of nasal irritation, abdominal distention, nasal breakdown, feeding intolerance, or pneumothorax were reported. No device-related adverse events occurred. The mortality rate during the study (14%) was below the unit's historical bCPAP mortality rate of 34%. Median device setup time was 15 minutes (IQR 12–20); initial setup was done correctly 64% of the time, with common difficulties including finding the correct cannula size (8 of 11) and securing the cannula to the face (3 of 11). Median time to change the blender was 15 seconds (IQR 12–27). All 9 nurses who completed Likert surveys reported the device was easy to set up and the blenders were easy to learn to use; overall satisfaction was 8.5 out of 10 (IQR 6.5–9.5). Among 5 nurses who used the device more than once, all believed it improved the quality of care. All 15 healthcare workers interviewed expressed a desire to continue using the device after the study, and 14 of 15 found it easy to use. Key themes included appreciation for the ability to deliver less than 100% oxygen, desire for additional blender options below 37%, and the need for included accessories such as quality adhesive, nasal foam pads, and a sizing guide.
**Clinical Implications:** This study demonstrates that a very low-cost bCPAP device with oxygen blenders is feasible and acceptable in a resource-constrained level two newborn unit with existing bCPAP experience. The setup time of 15 minutes and blender change time of 15 seconds are comparable to commercial devices and practical for busy units with high patient-to-nurse ratios (median 9.2 at bCPAP initiation). The device addresses the critical problem of unblended 100% oxygen delivery, which causes retinopathy of prematurity and other oxygen-related injuries. All healthcare workers interviewed wanted to continue using the device. Limitations include the single-center design at a site with prior bCPAP and blending experience, the lack of heated humidification, and the use of a non-sealing RAM cannula interface. The study was not designed to evaluate effectiveness. The authors note that the availability of bCPAP and oxygen blenders is crucial for achieving Every Newborn Action Plan 2030 targets, but successful implementation also requires robust thermoregulation, nutritional support, infection management, training, and adequate staffing.