A retrospective cohort assessing practices of hypoxemia management for inpatients of Rwandan public hospitals

Abstract

Background

Hypoxemia, low blood oxygen levels, poses significant risks to inpatients. Effective management relies on pulse oximetry diagnosis and prompt, high-quality oxygen therapy. In Rwanda, following recent national investments in medical oxygen systems, we aimed to evaluate hypoxemia management for inpatients in hospital settings.

Methods

A retrospective cohort was conducted in public hospitals in Rwanda from July to September 2022, using data extracted from medical records for all inpatient admissions to examine pulse oximetry use, hypoxemia prevalence and oxygen administration on the day of admission, and subsequent patient outcomes.

Results

Pulse oximetry on admission was documented for 87.7% (5025/5731, 95% CI 86.8–88.5%) of patients who were admitted to 39 Rwandan hospitals. Of patients that had a following day of admission, this fell to 75.3% (3702/4919, 95% CI 74.0-76.5%). The prevalence of hypoxemia (SpO₂<90%) amongst those screened on admission was found to be 9.4% (472/5025, 95% CI 8.6–10.2%), particularly affecting neonates at 27.9% (163/585, 95% CI 24.4–31.6%). 21.0% of patients hypoxemic on admission died (99/472, 95% CI 17.5–24.9), compared to 1.7% of those not found to be hypoxemic (78/4538, 95% CI 1.4–2.1%). Among patients identified as hypoxemic on admission, 93.4% (441/472, 95% CI 90.8–95.4%) received oxygen therapy.

Conclusion

Pulse oximetry and oxygen therapy use, was common on admission to hospital in Rwanda. There were some disparities across wards, especially maternity. While most patients found to be hypoxemic received oxygen the same day, improvements in use of pulse oximetry to monitor patients throughout the course of the admission are required.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12982-026-02273-7.

Introduction

Hypoxemia (low blood oxygen level) is a common complication of many conditions affecting populations such as neonates, young children, and critically ill patients [1]. Studies have shown that the risk of death increases approximately four-fold when illness is accompanied by hypoxemia [2–4]. Hypoxemia remains underdiagnosed and inadequately treated in many low- and middle-income countries (LMICs) [5–7].

As recommended by the World Health Organization, a pulse oximeter is a non-invasive, affordable tool for healthcare workers to detect hypoxemia and guide clinical care [8]. However, pulse oximeter access and use remain an issue in LMICs [9, 10].

In Rwanda, as in other sub-Saharan African countries, a shortage of clinicians is not uncommon [11], resulting in sub-optimal care for some patients, including basic services like hypoxemia screening and treatment. Administering oxygen requires specific skills, tools, and delivery interfaces [8]. Missing any of these elements can lead to untreated hypoxemia. Incorrect or over prescription can cause harm such as oxygen toxicity, particularly in preterm and low-birth-weight babies [12].

Rwanda has made substantial investments during and after the COVID-19 pandemic to guarantee continuous and sustainable access to oxygen. By 2022, the country had 28 hospitals with on-site oxygen plants producing medical oxygen at scale, and distributing oxygen cylinders to nearby hospitals [13]. This study aimed to assess pulse oximetry use, hypoxemia prevalence and management with oxygen therapy on admission to hospitals in Rwanda in order to understand coverage of this essential medicine and identify any areas of improvement required.

Methods

Study setting

Rwanda is a mountainous country with average altitude of 1700 m, ranging between 900 and 4707 m [14]. Rwanda has 55 public hospitals, which include eight referral hospitals, four provincial hospitals, three specialized hospitals, and 40 district hospitals. For this study, 40 (72.7%) public hospitals were purposively selected for inclusion in the study, including five referral, three specialized, one provincial and 31 district hospitals, located between 1033 and 2215 m altitude. The majority of hospitals in the country were assessed, however 15 (27.3%) were excluded due to having recently received government-led interventions on oxygen systems strengthening and hypoxemia management, and this study accompanied expansion of these interventions.

District hospitals offer comprehensive medical care, including general surgery, cesarean sections, and both inpatient and outpatient services, alongside laboratory diagnostics; provincial hospitals are larger and better equipped than district hospitals, they serve as referral centers for their provinces; referral hospitals provide more advanced specialized services such as complex surgeries and educational opportunities. Specialized hospitals serve as national referral centers for specific needs, focusing on non-acute conditions such as orthopedics, psychiatry, and physiotherapy.

Study design

We conducted a retrospective cohort study to assess clinical practices. All patients admitted to all inpatient wards on 11 randomly selected days between 1st July 2022 to 25th September 2022 were included in the study. Admission registers were used to identify eligible patients, and data on clinical and demographic characteristics, pulse oximetry, oxygen therapy, and admission outcomes were extracted from clinical case notes using a standardized digital survey (see Supplementary Material 1), using SurveyCTO software [15]. Data collectors with clinical backgrounds were used to undertake this data collection and were trained beforehand on the digital survey tool. Identifiable information was not extracted from the patient records, and whilst individual consent was not collected from patients, written consent was provided by all Heads of Facility. The requirement for written consent was waived by the approving ethics committee.

Key definitions

Patients measured with pulse oximetry were defined as any patients with a documented measurement of SpO₂ on the day of admission. Hypoxemia was defined as a SpO₂ reading less than 90%, aligning with Rwanda clinical treatment guidelines [16]. Patients were classified as having received oxygen therapy if it was recorded in their medical files. Age of the patient was categorized by clinical definitions in Rwanda, neonates less than 28 days, pediatrics from 28 days to 14 years, and adults from 15 years onwards. Symptoms and diagnoses are grouped by the system to which they relate, with a list outlined in the survey (Supplementary Material 3).

Study size

We aimed to calculate the sample size needed to detect a change in oxygen administration practices for hypoxemic patients from 50% to 60% across an approximated sample of 45 hospitals with > 80% power to +/-5% accuracy, using Stata’s twoproportions command [17]. To determine the number of sample frame days needed, we used HMIS admission data from the facilities and assuming 10% prevalence of hypoxemia, we estimated that we would require 11 days of admissions data, expected to result in 495 hypoxemic patients. These days were randomly selected between 1st July − 25th Sept 2022 using a random number generator for each hospital. Data collection took place between October 17, 2022, and December 16, 2022.

Missing data

Missing data of age and sex was presented with demographics data. Key outcomes of pulse oximetry and oxygen therapy were treated as not performed if not documented in the case notes. The proportion of patients with admission outcome missing was described, these patients were treated as not died in the mortality analysis, with a sensitivity analysis undertaken excluding these patients.

Statistical methods

We used descriptive summary statistics to describe the patient population. Proportions on key outcomes of interest were calculated with 95% confidence intervals using sample standard errors.

Results

Participation

There were 5766 patients were admitted on the sample frame days and had patient records from which to collect data, however 35 patients were excluded due to being cared for solely in Outpatients, therefore 5731 patients were included analysis. Of the original sample, 39/40 (97.5%) of facilities were included, as one facility was not accessible to participate due to data access policy restrictions.

Baseline characteristics of the patient cohort

Most patients were adults (15 years or above) (3714/5731, 64.8%). Pediatric admissions (29 days – 14 years) made up 17.0% of the cohort (972/5731), and 10.6% were neonates less than 29 days (607/5731). Overall, a higher proportion of patients were female (3706/5731, 64.7%) compared to male (2023/5731, 35.3%), however the cohort included 1725 (30.1%) adult women admitted for obstetric care, excluding these the proportion of inpatients admitted were approximately equal, with 50.5% males (2023/4006) and 49.4% females (n = 1981/4006). Most patients were admitted to the 31 district hospitals (4520/5731, 78.9%) followed by the four Referral/University Teaching Hospitals (900/5731, 15.7%). Patients could be admitted to multiple wards across the course of admission, the most common ward admissions being Emergency (n = 2020, 33.3%), Maternity (n = 1921, 31.7%), and Neonatology (n = 593, 9.8%). Additional data given in Supplementary Material 2.

Patients could present with multiple symptoms or diagnoses on admission. In neonates, the most commonly reported symptom on day of admission was respiratory distress (117, 33.6%), followed by vomiting or difficulty feeding (51, 14.7%) and chest indrawing (51, 14.7%). In pediatric patients, fever emerged as the predominant symptom (265, 24.5%), followed closely by diarrhea (256, 23.7%) and cough (182, 16.8%). Among adults, the most frequently reported symptoms were urogenital or obstetric-related symptoms (1103, 44.2%), followed by pain (491, 19.7%) and fever (145, 5.8%).

The most common diagnoses in neonates were low birth weight or prematurity (228, 45.0%); in pediatric patients, pneumonia (195, 22.9%) was the most frequent diagnosis. Adult admissions were mostly due to maternity-related reasons (1269, 38.1%), followed by injury or trauma (461, 13.8%) (See Supplementary Material 3).

Hypoxemia prevalence

Out of 5025 patients who had pulse oximetry documented on day of admission, around 1 in 10 had hypoxemia (SpO₂ <90%), (9.4%, 472/5025, 95% CI 8.6–10.2%) (Supplementary Table 3). Just over a quarter of neonates had documented hypoxemia, (27.9%, 163/585, 95% CI 24.4–31.6%), followed by 12.3% of pediatric patients (108/875, 95% CI 10.3–14.7%), and 5.7% of adults (180/3169, 4.9–6.5%). Excluding adult women admitted to maternity, an increased proportion, 9.6% of adults (175/1821) were documented to be hypoxemic on admission.

Hypoxemia was most commonly found in patients presenting with respiratory symptoms such as cough, in 40.5%, 38.6%, 50.0% of neonates, pediatric and adult patients experiencing cough, respectively (See Supplementary Material 3).

Pulse oximetry usage

Pulse oximetry on admission was frequent, with 87.7% of all patients (5025/5731, 95% CI 86.8–88.5%) with SpO₂ documented on admission (Table 1). It was documented in 96.4% of neonates (585/607, 95% CI 94.6–97.6%), followed closely by pediatric patients at 90.0% (875/972, 95% CI 88.0-91.8%), and 85.3% of adults (3169/3714, 95% CI 84.2–86.4%).

Table 1.

Proportion of patients that were measured with pulse oximetry on the day of their admission, identified with hypoxemia and provided with oxygen therapy if found to be hypoxemic, amongst 5731 inpatients in Rwanda

Patient Characteristics	Patients admitted (N)	Number of patients that received pulse oximetry on admission	Percentage that received pulse oximetry on admission (CI 95%)	Number of patients found to be hypoxemic on admission	Percentage that were hypoxemic on admission of those screened (CI 95%)	Number of hypoxemic patients that were provided oxygen on admission	Percentage of hypoxemic patients that were provided oxygen on admission (CI 95%)
Overall	5731	5025	87.7 (86.8–88.5)	472	9.4 (8.6–10.2)	441	93.4 (90.8–95.4)
Age Group
Neonates (< 29 days)	607	585	96.4 (94.6–97.6)	163	27.9 (24.4–31.6)	159	97.5 (93.6–99.1)
Pediatric (29 days − 14 years)	972	875	90.0 (88.0-91.8)	108	12.3 (10.3–14.7)	99	91.7 (84.7–95.6)
Adult (≥ 15 years)	3714	3169	85.3 (84.2–86.4)	180	5.7 (4.9–6.5)	163	90.6 (85.3–94.1)
Missing Age	438	396	90.4 (87.3–92.8)	21	5.3 (3.5-8.0)	20	95.2 (72.7–99.3)
Sex
Male	2023	1854	91.6 (90.4–92.8)	279	15.0 (13.5–16.8)	262	93.9 (90.4–96.2)
Female	3706	3169	85.5 (84.3–86.6)	193	6.1 (5.3-7.0)	179	92.7 (88.1–95.7)
Missing Sex	2	2	100.0 (NA)	0	0.0 (NA)	0	0.0 (NA)
Facility type
District Hospital (n = 31)	4520	3952	87.4 (86.4–88.4)	387	9.8 (8.9–10.8)	359	92.8 (89.7–95.0)
Provincial Hospital (n = 1)	229	224	97.8 (94.9–99.1)	5	2.2 (0.9–5.3)	5	100.0 (NA)
Referral/University Teaching Hospital (n = 4)	900	824	91.6 (88.6–93.2)	78	9.5 (7.7–11.7)	75	96.2 (88.7–98.8)
Specialized Hospital (n = 3)	82	25	30.5 (21.5–41.2)	2	8.0 (2.0–27.0)	2	100.0 (NA)
Ward*
Emergency	2020	1892	93.7 (92.5–94.7)	219	11.6 (10.2–13.1)	197	90.0 (85.2–93.3)
Maternity	1921	1515	78.9 (77.0-80.6)	11	0.7 (0.4–1.3)	10	90.9 (56.0-98.8)
Neonatology	593	572	96.5 (94.6–97.7)	161	28.1 (24.6–32)	157	97.5 (93.6–99.1)
Surgery	354	297	83.9 (79.7–87.4)	9	3.0 (1.6–5.7)	9	100.0 (NA)
Paediatrics	352	335	95.2 (92.4–97.0)	49	14.6 (11.2–18.8)	46	93.9 (82.6–98)
Internal Medicine	328	307	93.6 (90.4–95.8)	31	10.1 (7.2–14.0)	31	100.0 (NA)
Gynecology	265	241	90.9 (86.8–93.9)	1	0.4 (0.1–2.9)	0	0.0 (NA)
Operations Theatre	93	83	89.2 (81.2–94.1)	3	3.6 (1.2–10.6)	3	100.0 (NA)
Intensive Care Unit	14	13	92.9 (63.0–99.0)	4	30.8 (12.0-59.1)	4	100.0 (NA)

*Patients can be admitted to multiple wards across the course of admission and therefore could be counted twice. Non-acute wards not shown, see Supplementary Material 4 for these wards

A higher proportion of male patients had pulse oximetry documented on day of admission, 91.6% (1854/2023, 95% CI 90.4–92.8%), compared to 85.5% (3169/3706, 95% CI 84.3–86.6%) of female patients. However, excluding adult women admitted to maternity, this was more equal with 91.9% women screened (1821/1981).

Pulse oximetry usage on admission varied across different hospital wards, with Neonatology screening 96.5% (572/593, 95% CI 94.6–97.7%) of patients on admission, whereas 83.9% (297/354, 95% CI 79.7–87.4%) patients admitted to Surgery had SpO₂ documented that day.

Provincial and referral/university hospitals showed the highest rate of pulse oximetry measurement on admission at 97.8% (224/229, 95% CI 94.9–99.1%) and 91.6% (824/900, 95% CI 88.6–93.2%), respectively, followed by 87.4% (3952/4520, 95% CI 86.4–88.4%) of patients at district hospitals. Patients admitted to specialized hospitals had the lowest rate of pulse oximetry usage on admission, at 30.5% (25/82, 95% CI 21.5–41.2%).

Examining pulse oximetry usage across the course of the admission, on the day of admission (Day 0), 87.7% of patients (5025/5731, 95% CI 86.8–88.5%) received pulse oximetry. Of 4919 patients that had a following day of admission (Day 1), 75.3% had a record of an SpO₂ measurement on this day (3702/4919, 95% CI 74.0-76.5%). Of the 4203 patients with longer admissions, 75.8% (3186/4203, 95% CI 74.5–77.1%) had at least one SpO₂ measurement at some point in the remaining time from Day 2 until the end of admission (Fig. 1).

Fig. 1.

Percentage of patients with recorded pulse oximetry measurement by time point in admission with 95% confidence intervals. **Day 2 – End of Admission indicates at least one pulse oximetry measurement at any point between Day 2 and end of admission

By age group, adult women admitted to maternity had lower documentation of pulse oximetry overall and a steeper reduction in subsequent days after admission to just 54.9% (625/1139, 95% CI 52.0-57.7%), whilst neonates had little decline in measurement across the admission – with 92.6% receiving a measurement between Day 2 to the end of admission (476/514, CI 95% 90.0-94.6%) (See Supplementary Material 5).

Oxygen therapy provision

Most of the 472 hypoxemic patients, 93.4%, had oxygen therapy documented on the same day (441/472, 95% CI 90.8–95.4%). Coverage ranged from 97.5% of hypoxemic neonates (159/163, 95% CI 93.6–99.1%) to 90.6% of hypoxemic adults (163/180, 95% CI 85.3–94.1).

57.3% (441/769) of all patients receiving oxygen on day of admission were hypoxemic (SpO₂<90%), with 41.2% (317/769) having SpO₂ 90–100% on day of admission and the remaining 1.4% (11/769) prescribed oxygen on admission with no documented pulse oximetry that day.

Mortality

Mortality was more frequent in patients found to be hypoxemic on day of admission; 21.0% (99/472, 95% CI 17.5–24.9) of these inpatients died, compared to 1.7% (78/4538, 95% CI 1.4–2.1%) of inpatients not found to be hypoxemic when screened on day of admission (Table 2). Neonates hypoxemic on day of admission experienced heightened mortality 27.6% (45/163, CI 95% 21.3–35.0%) died, compared to 3.8% of those not found to be hypoxemic (16/420, CI 95% 2.4–6.1%). 24.0% (42/175, 18.3–30.9%) of adult patients not admitted for maternity care died when hypoxemic on admission, compared to 3.1% (51/1643, 95% CI 2.4–4.1%) when not hypoxemic on admission.

Table 2.

Mortality of inpatients found to be hypoxemic compared to those not found to be hypoxemic on admission, amongst 5731 inpatients in Rwanda

Demographics	Patients not found to be hypoxemic on admission who died		Patients found to be hypoxemic on admission who died
	n	% (95% CI)	n	% (95% CI)
Neonate (< 28 days)	16	3.8 (2.4–6.1)	45	27.6 (21.3–35.0)
Pediatric (29 days – 14 years)	3	0.4 (0.1–1.2)	9	8.3 (4.4–15.3)
Other adults (≥ 15 years)	51	3.1 (2.4–4.1)	42	24.0 (18.3–30.9)
Women admitted in maternity (≥ 15 years)	1	0.1 (0.0-0.5)	0	0.0 (NA)
Unknown Age	7	1.9 (0.9–3.9)	3	14.3 (4.7–36.1)
Overall	78	1.7 (1.4–2.1)	99	21.0 (17.5–24.9)

*721 patients not screened are not shown here, see Supplementary Material 7. This analysis includes 46 patients with unknown admission outcomes categorised as not died, to see sensitivity analysis excluding these patients, see Supplementary Material 6

Of the remaining patients in the cohort, 88.7% (5085/5731) were discharged, 4.5% (256/5731) were referred to another health facility, 0.6% (32/5731) absconded or left against medical advice, and 1.9% (108/5731) were still admitted at the end of the study window, and 1.1% (64/5731) had unknown outcomes of admission.

Discussion

Key results

The study highlights a considerable utilization of pulse oximetry (87.7%) among inpatients on admission, with higher rates in neonates (96.4%) and pediatric patients (90.0%) compared to adults (85.3%). Hypoxemia was found in 9.4% of the patients on admission who were screened, with a higher prevalence amongst neonates (27.9%). Patients found to be hypoxemic during admission were more likely to die. Among those experiencing hypoxemia, 93.4% of patients received oxygen therapy the same day.

Interpretation

Pulse oximetry use on admission was found to be common practice across inpatient settings in Rwanda, in 87.7% of patients overall, with slightly more comprehensive practices in younger age groups, which could reflect an appreciation of the higher vulnerability of neonates and children to hypoxemia. Specialized hospitals, that treat non-acute conditions such as mental health and physiotherapy, had much lower pulse oximetry use, in 30.5% of patients admitted. Whilst hypoxemia prevalence is likely to be lower in these settings, it represents a missed opportunity to identify serious illness or deterioration. Pulse oximetry use fell to 75.3% on the second day of admission, yet patient outcomes can be improved by ensuring that all inpatients receive regular monitoring during their admission. A prospective cohort study undertaken in two Rwandan hospitals found 90.0% of adults were screened with pulse oximetry [18], aligning closely with our findings. The study also revealed gaps in use of target SpO₂ values for oxygen therapy, emphasizing that dosage and duration of oxygen therapy should be guided by ongoing monitoring with pulse oximetry.

Hypoxemia was found with declining prevalence in increasing age groups, from 27.9% of neonates (< 28 days), 12.3% of pediatrics (29 days – 14 years) and 5.7% of adults (15 or more years). In a global meta-analysis, a similar trend was found of 24.5% neonates (< 28 days), 12.1% of pediatrics (1 month – 17 years), and 10.8% of adults (18 or more) [21]. Paediatric hypoxemia could be driven by high prevalence of respiratory infections in children, particularly in LMICs. Neonates require specific considerations in management of blood oxygen, as they are transitioning from fetal hemoglobin (HbF) to adult (HbA) and undergoing rapid cardiopulmonary changes at birth, which in turn affects interpretation of pulse oximetry readings, particularly in low-birthweight and preterm neonates [19]. Clinicians should consider interpretation of pulse oximetry relative to birth timing, gestational age and the pulse oximeter probe placement, pre- or post-ductal. Importantly, target SpO₂ should be adjusted to prevent hyperoxia from oxygen therapy, which is associated with increased risk of retinopathy of prematurity and permanent lung damage [12]. It was found that 30.8% of patients in adult ICU were hypoxemic (SpO₂<90%) on day of admission, compared to 54% found in a global ICU point-prevalence study who defined hypoxemia using direct measurement of oxygen dissolved in arterial blood (PaO₂) and Fraction of Inspired Oxygen (FiO₂) [20]. This suggests further research exploring diagnostic practices in critical care in Rwanda is needed to ensure hypoxemia is not missed.

Pulse oximeters detect the proportion of haemoglobin bound to oxygen, which has a non-linear relationship to blood oxygen (PaO₂); the steep portion of the oxyhaemoglobin dissociation curve means small drops in PaO₂ cause large falls in SpO₂ below 90%. Some physiological conditions such as poor perfusion, anemia and jaundice can affect that relationship, shifting the curve and therefore the interpretation of SpO₂ values. Whilst pulse oximeters are practical, affordable and useful for detecting hypoxemia, healthcare workers should be aware how these conditions impact recommended clinical care. In addition, some pulse oximeters have been found to underestimate hypoxemia in people with darker skin pigments [21], which requires global efforts to improve the technology to provide equitable care.

In patients with hypoxemia, the majority, 93.4%, were documented to receive oxygen therapy the same day with no significant gaps seen between age groups or wards. This is an improvement compared to previous studies of neonatal and pediatric patients in Rwanda in 2021, where only half (48.3%) of patients were prescribed oxygen therapy on the same day [22]. However, quality of oxygen therapy can be examined across different dimensions [23]. Oxygen therapy should be timely, with minimum delay between a hypoxemic patient presenting and initiation of therapy. Treatment should be safe and effective, with appropriate duration and dose of treatment, therefore guided by regular or continuous monitoring with pulse oximetry.

Some findings indicate there are potential gaps in the quality of therapy, such as the finding that 41.2% of patients who received oxygen had SpO₂ values of 94–100%. Whilst some patients might require oxygen for other reasons such as trauma, perioperative care or anticipatory use, it does suggest further research is needed to ensure there is no over-prescription. People living in high altitudes will have lower normal SpO₂ levels, with the standard hypoxemia definition of < 90% being applicable up to altitudes of 2500 m, and < 85% being recommended at 3200 m [24]. Clinicians should be aware of this to prevent overprescription of oxygen therapy for patients who live at high altitudes above 2500 m who might have lower normal SpO₂ saturations.

Rwanda has demonstrated considerable access to pulse oximetry and oxygen therapy in all inpatient age groups. This could be due to the implementation of clinical treatment guidelines and job aids for the management of children and adults [16, 25]. The country’s national oxygen strategy, published after this study, highlights a multifaceted approach that combines increased national oxygen production and key policy reforms with a strong focus on capacity building within the healthcare system [13].

Yet, high mortality rates among hypoxemic patients indicates there is still progress to be made in the quality of care. Oxygen therapy and respiratory care should be integrated with other aspects of care and delivered to a high standard, including adhering to clinical guidelines. As a medical emergency, hypoxemia should be detected rapidly and treated immediately. The national strategy prioritizes decentralizing diagnosis and treatment of hypoxemia to primary levels of care, which could show additional benefits in reducing severity of presenting cases through early detection, and more research is recommended to measure the impact of this strategy [13].

Limitations

Despite the valuable insights provided by the study findings, some limitations were noted. Reliance on patient records may introduce bias due to incomplete documentation, where pulse oximetry or oxygen therapy is performed but not recorded. However, documentation of SpO₂ and oxygen therapy is itself essential for high quality of care. Trained data collectors with a clinical background were used to minimize errors in collecting and interpreting clinical case notes.

The pulse oximetry reading assessed in the study could have been performed when the patient is breathing room air (FiO₂ of 0.21) or whilst receiving oxygen therapy with a higher FiO₂, which would make the resulting pulse oximetry reading have different implications for care. This was not examined in this study, nor other physiological measurements such as bilirubin or perfusion status, which may impact interpretation of hypoxemia prevalence on admission. This should be explored in future research to understand if pulse oximetry is used to titrate oxygen treatment, and how FiO₂ and these physiological measurements are used to inform oxygen therapy.

The study included 39 out of 55 hospitals in Rwanda, which is a large proportion of hospitals in the country and therefore highly representative. However, the study excluded 15 hospitals which had previously received hypoxemia management strengthening activities from the government, it is likely therefore these hospitals excluded might have improved outcomes, potentially biasing our findings to be an underestimate of pulse oximetry screening and oxygen treatment coverage in Rwanda. Days were randomly selected across different study sites, improving the representativeness of the sample of patients included, although there might be seasonal variation outside of the study window that was not captured. The study was focused on inpatient clinical care and therefore did not address outpatient, policy environment, financial investments, or supply chain complexities, all crucial for lasting improvements in healthcare delivery.

Conclusion

This study found high levels of pulse oximetry use and oxygen therapy amongst inpatients of all ages in Rwandan public hospitals. Despite this, some findings indicate the remaining gaps in high quality hypoxemia management, that indicates further research and action are needed to ensure high-quality care. These findings demonstrate the positive impact of increased investments in oxygen systems in recent years in LMICs like Rwanda and the value of continued efforts to reduce preventable mortality through quality respiratory care.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

F.M. led the conceptualization, methodology, data curation, data analysis, and supervision, and drafted and revised the manuscript. H.H.W., H.M., and C.K. contributed to the study design, data analysis, and drafting and reviewing the manuscript, with H.H.W. and H.M. providing visualisation and supervisory support. A.T., A.N., B.K., N.C., and A.R. contributed to conceptualization, supervision, and manuscript preparation. N.S. supported manuscript writing and review. J.B.N., S.M., and L.T. contributed to conceptual design and review. R.S., M.B., and H.R.G. played key roles in study conceptualisation, validation, and supervision, and critically reviewed the manuscript. F.L. contributed to conceptualization, investigation, methodology, supervision, validation, and manuscript review. C.K.N. contributed to conceptual design, data analysis, and manuscript preparation. All authors reviewed and approved the final version of the manuscript and agree to be accountable for its content.

Funding

This study was funded by the Bill and Melinda Gates Foundation under grant reference number INV-043011. We sincerely appreciate their support, which was vital for the completion of this research. This initiative is part of ongoing efforts to enhance the oxygen ecosystem in low- and middle-income countries (LMICs).

Data availability

Data are only available upon reasonable request, as the datasets contain personaldata, and data transfer requires a written agreement approved by Rwanda NationalEthics Committee and Rwanda Ministry of Health. Data access requests can be sent toFelix Musabirema (fmusabirema@clintonhealthaccess.org).

Declarations

Ethics approval and consent to participate

This study received approval from the Rwanda National Ethics Committee (RNEC) (Approval Number 456/RNEC/2002) and was conducted in accordance with the Declaration of Helsinki. Informed consent was obtained from heads of facility, who were fully informed of the study’s objectives, and procedures. Due to the study design using anonymized data extracted from routine clinical notes retrospectively, patient informed consent was not gathered and a waiver was granted by the Ethics Committee. We ensured strict confidentiality and anonymity, adhering to ethical standards in the treatment of human subjects.

Consent for publication

Not applicable.

Competing interests

FM, HHW, HM, CK, AT, NS, NC, AN, BK, and FL declare being employees of the Clinton Health Access Initiative which received payments from Bill and Melinda Gates Foundation and ELMA Foundation for the implementation of a program to improve oxygen access. AR, JBN and CKN are or were employees of the Ministry of Health of Rwanda which aims to improve oxygen access in Rwanda.