Early detection of Critical Congenital Heart Disease (CCHD) is crucial for reducing infant mortality. Pulse oximetry (POX) is widely utilised for screening CCHD in high-resource settings; however, its diagnostic accuracy in low-resource environments, such as sub-Saharan Africa, remains under-researched. This study aimed to assess the diagnostic accuracy of POX in screening Tanzanian newborns for CCHD. This prospective cohort study was conducted in two hospitals in Dar es Salaam, Tanzania. We used pre- and post-ductal saturation (SpO2) readings prior to discharge. A positive screen was defined as SpO2 < 90%; two pre- and post-ductal SpO2 readings <95%; and/or a pre- or post-ductal difference that exceeded 3%. All newborns with positive screening tests underwent echocardiography, while those with negative tests were followed for six months. The primary outcome was POX diagnostic accuracy. The study adhered to STARD guidelines for reporting diagnostic accuracy studies. Between October 2020 and June 2023, 10,630 newborns were screened. The majority (5,721; 54.0%) were male, resulting in a male-to-female ratio of 1.2. The median birth weight was 3.0 (IQR: 2.6-4.4) kg. A total of 51 (0.5%) newborns tested positive on POX, of which 18 (35.3%) had congenital heart disease (CHD), and 15 (83.3%) were classified as critical, leading to a CCHD prevalence of 1.41 per 1,000 live births (95% CI: 0.70-2.13), which increased to a cumulative prevalence of 3.27 per 1,000 live births (95% CI: 2.29-4.67) at six months. With a follow-up rate of 86.7% (9,170/10,574), POX demonstrated a sensitivity of 50.0% (95% CI: 32.1-67.9), a specificity of 99.5% (95% CI: 99.4-99.7), a false-positive rate of 0.4%, and an overall accuracy of 99.5% (95% CI: 99.2-99.5). Screenings conducted between 48 and 72 hours exhibited the highest diagnostic performance, AUC 0.79 (95% CI: 0.64-0.93), with a significant odds ratio (OR) of 5.31 (95% CI: 2.45-11.49, p = 0.00001). Newborns with a birth weight <2.5 kg were less likely to have CCHD detected by POX, OR 0.403 (95% CI: 0.19-0.87, p = 0.021). POX demonstrated lower sensitivity but higher specificity and diagnostic accuracy after 48 hours. The timing of screening and birth weight affected its accuracy, indicating a need for protocol adjustment.
POX demonstrated a sensitivity of 50.0% (95% CI: 32.1-67.9), a specificity of 99.5% (95% CI: 99.4-99.7), a false-positive rate of 0.4%, and an overall accuracy of 99.5% (95% CI: 99.2-99.5)
differential verification and incomplete follow-up leave false-negative ascertainment uncertain
Screenings conducted between 48 and 72 hours exhibited the highest diagnostic performance, AUC 0.79 (95% CI: 0.64-0.93), with a significant odds ratio (OR) of 5.31 (95% CI: 2.45-11.49, p = 0.00001)
few CCHD events and workflow-related timing confounding limit inference about optimal screening time
Newborns with a birth weight <2.5 kg were less likely to have CCHD detected by POX, OR 0.403 (95% CI: 0.19-0.87, p = 0.021)
the small-event observational subgroup analysis does not establish a reliable basis for birth-weight-specific protocols
Screening timing, delivery method, and birth weight influenced diagnostic performance.
delivery method was nonsignificant at p=0.234, while the remaining subgroup findings rest on few cases
74.5% of cases flagged by POX in our study required urgent medical attention for non-cardiac issues, highlighting POX’s broader potential as a screening tool for serious unrelated conditions.
the printed breakdown gives 21/51, or 41.2%, rather than 74.5%, without comparative evidence of broader screening benefit
Derived from the full evaluation — not a separate score.
Strengths
The study screened 10,630 newborns and followed screen-negative infants for six months, adding setting-specific evidence from an under-researched region.
↳ Abstract; Results, Demographic characteristics and Echocardiography evaluation results
Screening reached 89% of eligible newborns and used a specified device and prespecified oxygen-saturation thresholds. Positive screens were assessed by paediatric cardiologists.
↳ Methods, Enrollment and Echocardiography evaluation; Results, Demographic characteristics
The Discussion compares the observed sensitivity with high-income benchmarks and other low-resource studies, while identifying limits for lower-level facilities and high-altitude regions.
↳ Discussion; Strengths and limitations
Limitations
Negative screens were not routinely verified by echocardiography, and 13.3% lacked follow-up. Two positive-screen newborns who died before echocardiography were excluded, further limiting the accuracy analysis.
↳ Methods, Echocardiography evaluation; Results, POX screening results and Echocardiography evaluation results
The Conclusion says delivery method influenced performance, but the Results report OR=1.613, 95% CI 0.73–3.54 and p=0.234. The analysis does not establish the stated influence.
↳ Results, Accuracy of the pulse oximetry test and the associated factors; Conclusion
Timing and birth-weight analyses rest on only 30 confirmed CCHD cases, with timing also affected by hospital workflow and discharge. These findings provide questions for validation rather than established grounds for protocol tailoring.
↳ Results, POX screening results and associated factors; Discussion
The prospective cohort supplies useful local evidence through systematic nursery recruitment and a fixed screening protocol. The main methodological constraint is unequal diagnostic verification: positives received echocardiography, while negatives relied on incomplete six-month follow-up. The timing and birth-weight findings therefore need validation before supporting tailored protocols. Reporting precision is also reduced by conflicting counts and intervals, and the delivery-method conclusion contradicts the nonsignificant analysis. The study remains relevant to local screening development, but it does not establish deployment-ready protocol changes.
Nabu’s assessment, alongside the field’s view.
Are you an author of this paper?
Limited2.8
Confidence mediumThe prospective cohort supplies setting-specific diagnostic evidence from 10,630 Tanzanian newborns, addressing a documented regional evidence gap. Its advance is principally local validation rather than a new screening capability.
Systematic recruitment, fixed screening thresholds, and cardiologist assessment support the design, but differential verification, incomplete follow-up, and unspecified blinding substantially constrain accuracy inference. The delivery-method contradiction is surfaced separately through the Red reliability flag rather than a reliability cap.
The conventional structure makes the main findings followable, but incompatible counts and intervals reduce reporting precision. The conclusion presents delivery method as an established influence despite its nonsignificant result.
The Discussion compares high-income and low-resource evidence and acknowledges limits on precision and geographic scope. It does not adequately explain differential verification bias or carry subgroup uncertainty through to tailored-protocol recommendations.
Concerns4 of 4 checks
The Methods and Limitations assign incompatible missing-data mechanisms to follow-up losses. Ethics approval is declared, but a published protocol alone does not establish prospective registration.
Conflict of interest: not stated · Ethics approval: declared
35 references in manuscript 35 of 35 checkable references found in an index
References not verified automatically are resolved manually within 24 hours.
No retraction or expression of concern found in Crossref (includes Retraction Watch).
Sources: Crossref (incl. Retraction Watch) ✓
Where this paper’s evidence sits on the path from initial observation to real-world use.
Screening was evaluated within hospital workflow, establishing a practical application pathway. Protocol changes still require validation, implementation evaluation, and cost-effectiveness evidence.
AI-generated, human-governed. Something look off? Contact us to request a review.
Keep reading
Other papers Nabu has evaluated blind, starting with the same field.
Nakamura, Yamazaki, Namikawa, et al. · 2026 · Clinical medicine
ORR by BICR was 22.6% (7/31), and the disease control rate was 54.8% (17/31).
SupportedDance, Weyman, Adams, et al. · 2026 · Health sciences
Systemic constraints, related to limitations of data availability and access and organisational capacity for prevention, were perceived by participants to affect the development of public health interventions to reduce alcohol use.
SupportedZhang, Ploner, Sengpiel, et al. · 2026 · Clinical medicine
Quadrivalent human papillomavirus vaccination before pregnancy was associated with a lower risk of adverse pregnancy outcomes, particularly among women vaccinated at earlier ages
Supported