Diagnostic accuracy of tachypnoea for predicting mortality and identifying sepsis in young infants aged 0–59 days: a systematic review and meta-analysis

Abstract

Background

Tachypnoea is a danger sign for predicting mortality and identifying sepsis. We systematically reviewed evidence on the diagnostic accuracy of different tachypnoea thresholds for predicting mortality and sepsis in infants aged 0–59 days.

Methods

We searched MEDLINE, Embase, CINAHL, Global Index Medicus and CENTRAL for studies reporting associations or diagnostic accuracy of tachypnoea (≥60, ≥70 or ≥80 breaths per minute (bpm)) for predicting mortality and culture-confirmed sepsis in infants aged 0–59 days. We followed Cochrane methods for study screening, data extraction and quality assessment. We pooled ORs using random-effects models and sensitivity/specificity using random-effects bivariate models. We used GRADE (Grading of Recommendations, Assessment, Development and Evaluation) to assess the certainty of evidence.

Results

Of 7641 studies identified, 18 were included (n=92 398). Tachypnoea≥60 bpm had a pooled OR of 3.14 (95% CI 1.30 to 7.56; 5 studies; n=10 407) for predicting mortality and a sensitivity of 31% (95% CI 19% to 45%) and a specificity of 89% (95% CI 66% to 97%; 4 studies; n=7104). For culture-confirmed sepsis, tachypnoea≥60 bpm had an OR of 1.26 (95% CI 0.66 to 2.38; 4 studies; n=3996), and a sensitivity of 28% (95% CI 10% to 58%) and a specificity of 67% (95% CI 53% to 78%; 3 studies; n=693). The higher threshold of tachypnoea≥70 bpm was associated with a higher odds of mortality (OR 10.06 (95% CI 2.39 to 42.35); 2 studies; n=12 138) and resulted in higher specificity (99.4% (95% CI 99.1% to 99.6%)) but lower sensitivity (10.9% (95% CI 5.3% to 19.1%)) for mortality (1 study; n=5214). Data were insufficient to assess performance by age subgroup (0–6 vs 7–59 days). The certainty of evidence was very low.

Conclusions

Tachypnoea≥60 bpm is associated with a threefold increased odds of mortality in young infants, supporting its use as a danger sign. Given its relatively low sensitivity alone, tachypnoea should be interpreted alongside other clinical signs to optimise diagnostic accuracy. While higher thresholds (≥70 bpm) demonstrated stronger associations and increased specificity for predicting mortality, low sensitivity may limit clinical utility.

PROSPERO registration number

CRD42023431387.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Tachypnoea is a danger sign commonly used to predict mortality and to identify sepsis; however, the diagnostic and predictive accuracy of this sign in young infants has not been quantified.

WHAT THIS STUDY ADDS

• Tachypnoea≥60 breaths per minute (bpm) in young infants aged 0–59 days was significantly associated with increased odds of mortality but not with culture-confirmed sepsis. While a higher threshold (≥70 bpm) was associated with increased odds and specificity for predicting mortality, the low sensitivity limits its clinical utility.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Tachypnoea remains an important sign to predict mortality and identify sepsis in young infants, but using this single sign alone may miss infants at risk of death or sepsis. Therefore, it should be used alongside other clinical signs to more accurately identify infants at risk.

Introduction

Deaths among newborns aged 0–28 days globally decreased from 5.2 million in 1990 to 2.3 million in 2022,¹ corresponding to a decline in the global neonatal mortality rate from 37 per 1000 live births to 17 per 1000 live births over the same period.² However, significant geographic disparities exist, with sub-Saharan Africa experiencing the highest neonatal mortality rate at 27 per 1000 live births.² Neonatal mortality is mainly caused by premature birth, birth complications, infections and congenital anomalies.^{3 4} Bacterial pathogens are a common cause of newborn infections, which include pneumonia, sepsis and meningitis.^{5 6} Sepsis occurs when an infection causes a local inflammatory response that spreads, resulting in systemic inflammatory response syndrome, including fever or hypothermia, tachycardia, tachypnoea and leucocytosis or leucopenia.^{7 8} The gold standard for the diagnosis of sepsis in young infants is blood culture positivity.⁶ Identifying critically ill infants at risk of sepsis early is crucial to preventing mortality.⁹ In low-resource settings, treatment decisions mostly rely on clinical signs because laboratory investigations are not routinely available.¹⁰

Fast breathing, or tachypnoea, often defined as a respiratory rate of ≥60 breaths per minute (bpm), among infants aged 0–59 days is a clinical sign of possible serious bacterial infection (PSBI).¹¹ Sepsis may lead to tachypnoea through metabolic acidosis and compensatory hyperventilation¹² or through inflammatory cytokines that stimulate the respiratory centres in the brainstem.¹³ Tachypnoea is an important early warning sign of sepsis or pneumonia because it is an objective assessment that can be ascertained by health workers or even traditional birth attendants before infants are referred to doctors, midwives or nurses.¹⁴,¹⁶ Among infants with possible serious infection, the prevalence of tachypnoea≥60 bpm is 18% in the first 6 days of life and 26% during days 7–59.¹⁷

Existing guidelines, such as the Integrated Management of Childhood Illness (IMCI) algorithm from WHO, include tachypnoea (≥60 bpm) as a danger sign to identify infants with a PSBI.¹⁸ The cut-off of 60 bpm is the most commonly used dichotomous threshold.¹⁸ However, neonatal respiratory rate varies substantially, even within a brief time interval,¹⁹ and the highest respiratory rate variability occurs within the first week after birth.²⁰ Transient tachypnoea of the newborn (TTN) is a normal physiological transition state caused by delayed clearance of amniotic fluid after birth and is present in infants without infections, normally lasting 48–72 hours and potentially longer in infants born via caesarean section.²¹,²⁴ The optimal threshold of respiratory rate to identify young infants at high risk in the early days of life and at different postnatal ages remains unclear. It is important to characterise the optimal respiratory rate threshold in the first week of life to identify high-risk newborns, given both the presence of TTN and the fact that 75% of neonatal deaths occur in this critical early period.³

Assessing the predictive and diagnostic accuracy of tachypnoea as an early warning sign to identify infants at high risk of mortality or sepsis is crucial to provide timely interventions and improve survival. Two previous systematic reviews and meta-analyses by Shah et al²⁵ and Rambaud-Althaus et al²⁶ evaluated the accuracy of tachypnoea and other clinical symptoms for diagnosing radiographic pneumonia in children aged <5 years and in children aged 2 months to 6 years, respectively. A study using analysis of pooled datasets by Rees et al²⁷ also investigated the sensitivity and specificity of age-based tachypnoea and other clinical signs to diagnose radiographic pneumonia in children aged 0–59 months. However, those three studies did not specifically examine tachypnoea in young infants aged 0–59 days for predicting mortality and identifying sepsis and did not explore different thresholds of respiratory rate for predicting mortality and identifying sepsis.

We recently conducted systematic reviews to assess the accuracy of infant clinical sign algorithms to diagnose sepsis and predict young infant mortality^{28 29} to inform the updated WHO guidelines on the management of serious bacterial infections in infants aged 0–59 days.⁶ Stemming from that review, we identified the need to examine in greater detail the clinical sign of tachypnoea to ascertain the optimal threshold for the identification of infants at high risk and examine the timing of mortality and sepsis risks in the early neonatal period versus later (0–6 days vs 7–59 days). Therefore, this review aims to systematically review the evidence by answering the Population, Index test, Reference standard and Diagnosis of interest question: among young infants (0–6 and 7–59 days), what is the diagnostic accuracy of different thresholds of tachypnoea to predict mortality and identify sepsis?

Methods

We conducted a systematic review and meta-analysis following a protocol registered on PROSPERO (CRD42023431387).³⁰ The report adheres to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA),³¹ as detailed in online supplemental table 1. Tachypnoea was defined according to the authors, and we recorded the threshold for the definition (ie, ≥60, ≥70 or ≥80 bpm). The reference standards of diagnosis were mortality, culture-confirmed sepsis and clinical sepsis. We treat mortality and culture-confirmed sepsis as our primary diagnoses of interest. Mortality was defined as death by any cause prior to 60 days of life or until hospital discharge. Sepsis was categorised as culture-confirmed sepsis and clinical sepsis. Several studies defined sepsis using blood culture confirmation or other laboratory testing. We used a threshold of at least 50% of cases having positive blood culture in order to be categorised as culture-confirmed sepsis in our review. Studies were categorised as clinical sepsis if the sepsis diagnosis was based on clinician judgement of suspected or confirmed bacterial infections supported by laboratory examination or imaging (ie, haematological and biochemical blood examination, radiological examination or microbiological cultures). Studies were excluded if they used clinical signs alone, such as IMCI criteria without laboratory examination.

Search strategy

A comprehensive search across multiple databases, including MEDLINE, Embase, CINAHL, Global Index Medicus and CENTRAL, was conducted by a medical research librarian (CGW) from inception until 8 May 2023 and updated on 5 September 2024. An umbrella search of systematic reviews on the accuracy of clinical signs for mortality prediction in young infants was also performed in January 2024 to identify additional studies (online supplemental table 2b). The search strategy included terms related to infants, clinical signs and symptoms, tachypnoea, fast breathing, mortality, sepsis and diagnostic accuracy measures (online supplemental tables 2a,b). We also identified additional studies by hand-searching the reference lists of relevant studies.

Study selection

Study selection was performed using Covidence³² after duplicates were removed. Titles and abstracts were independently screened by two reviewers, followed by full-text review. A third reviewer adjudicated and resolved any discrepancies.

Inclusion and exclusion criteria

We included studies that (1) involved infants with a postnatal or chronological age of 0–59 days or a mean or median age of <60 days; (2) evaluated tachypnoea at any threshold (≥60, ≥70 and ≥80 bpm) with physical examination by any type of trained health worker; (3) were conducted in any setting, including hospital, outpatient clinic or community; (4) reported data about mortality, culture-confirmed sepsis or clinical sepsis in young infants aged 0–59 days as reference standards; (5) reported at least one diagnostic accuracy measure (sensitivity, specificity, positive or negative predictive value) or measure of association (OR, risk ratio or HR) for mortality, culture-confirmed sepsis or clinical sepsis and (6) were published in any language. We excluded studies that (1) included infants with specific comorbidities (ie, all infants had congenital heart defects) or (2) were case reports, editorials or conference proceedings, or had no available full text.

Data extraction and management

Two reviewers extracted data from the included studies into an Excel spreadsheet (online supplemental tables 4 and 5). We extracted information on the authors, publication year, study design, country (categorised as low- and middle-income countries (LMICs) or high-income countries (HICs) according to the World Bank classification³³), setting, study population, age group, number of infants, number of events (mortality, culture-confirmed sepsis or clinical sepsis), postnatal age at respiratory rate assessment, respiratory rate threshold for tachypnoea, personnel who assessed respiratory rate, procedure for measuring respiratory rate, definition of reference standard and diagnostic accuracy measures (ie, sensitivity and specificity) or measures of association for reference standards (eg, OR, risk ratio or HR). A third reviewer acted as an adjudicator to resolve any disagreements.

Quality assessment

For studies that reported sensitivity and specificity for mortality only, we used the Quality Assessment of Prognostic Accuracy Studies (QUAPAS)³⁴ tool to evaluate study quality across five domains (eg, patient selection, index test, outcome, flow and timing and analysis) and the concern for applicability. Meanwhile, for studies reporting sensitivity and specificity for sepsis, PSBI or severe illness, we used the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2)³⁵ tool to evaluate the quality across five domains (eg, patient selection, index test, outcome, flow and timing and analysis) and the concern for applicability. We evaluated studies as high, unclear or low risk. For all studies reporting OR, risk ratio or HR, we assessed quality using the Newcastle–Ottawa Scale, which we tailored to our research question.³⁶ We graded studies as good, fair or poor according to Agency for Healthcare Research and Quality standards.³⁶ Two independent reviewers performed quality assessment, and a third reviewer resolved disagreements. The details of the quality assessment are reported in online supplemental tables 6–13. We could not evaluate publication bias with statistical or graphical methods due to the limited number of studies within the same reference standard and respiratory rate threshold in the meta-analysis, as advised by Cochrane.^{37 38}

Certainty of evidence assessment

We used the Grading of Recommendations, Assessment, Development and Evaluation (GRADE) approach for diagnostic tests and strategies to evaluate the certainty of evidence.³⁹ The assessment and grading criteria for risk of bias, indirectness, inconsistency and imprecision are presented in online supplemental table 14.

Statistical analysis

We constructed 2-by-2 tables for each study (including true positives, false positives, true negatives and false negatives) when data were available. We calculated the OR, sensitivity and specificity from the data provided in the study. We contacted the authors of studies that reported incomplete data. We estimated pooled ORs and 95% CI using a DerSimonian–Laird random-effects model⁴⁰ for each reference standard, stratified by respiratory rate threshold. We also pooled the sensitivity and specificity and 95% CI among studies within the same reference standards, stratified by respiratory rate threshold, using hierarchical summary receiver operating characteristic curves, which account for the correlation between sensitivity and specificity if there were at least four studies.⁴¹ When fewer than four studies were available or convergence was not achieved, we used random-effects bivariate models.⁴¹

In addition, we reported pooled likelihood ratios (LRs) across the same reference standards, stratified by respiratory rate threshold, which summarises the information contained in both sensitivity and specificity.⁴² LRs determine how much more likely a particular test result is to occur among those with the condition than in those without the condition.⁴³ The positive LR (LR+) was calculated by sensitivity/(1−specificity), while the negative LR (LR−) was calculated by (1−sensitivity)/specificity. Both LR+ and LR− are equal to one if a diagnostic test has no discriminating ability. A test is considered good when LR+ exceeds 5.0 and LR− falls below 0.2.^{44 45}

The I² statistic was used to evaluate statistical inconsistency, with an I² exceeding 50% considered to have substantial inconsistency.^{38 46} A 95% prediction interval was calculated to determine the range of predicted effects in an individual new study when I²>50%, as recommended by IntHout et al.^{47 48}

When there were adequate data, we performed subgroup analysis by postnatal age groups (<7 days and ≥7 days). We were unable to perform subgroup analyses by clinical setting (community vs hospital) and personnel assessing respiratory rate (community health worker vs clinician) due to inadequate data. All statistical analyses were conducted with Stata (BE V.18.0, StataCorp, College Station, Texas, USA)⁴⁹ using the metadta⁵⁰ and meta⁴⁸ commands.

Results

Study characteristics

Of 7641 studies retrieved from the databases and 28 from reference lists of relevant studies, 642 full-text articles were assessed for eligibility, and 18 studies met the inclusion criteria (figure 1).⁵¹,⁶⁸ Reasons for exclusion during the full-text review are detailed in online supplemental table 3. The characteristics of the included studies are summarised in online supplemental table 4. We also summarised information about infant characteristics, including weight or birth weight, gestational age, nutritional status and catheterisation status across studies in online supplemental table 4.

Figure 1. PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) flow diagram of included studies in the systematic review.

Results of individual studies are detailed in online supplemental table 5. Study sample sizes ranged from 60 to 63 017 infants, totalling 92 398 infants. 4 studies⁵¹,⁵⁴ were conducted in a community-based non-hospital setting, 4 studies⁵⁵,⁵⁸ took place in an outpatient ambulatory setting and 10 studies⁵⁹,⁶⁸ were conducted in a hospital (neonatal intensive care unit and admitted setting). 14 studies⁵¹,^{6163 67 68} were conducted in LMICs, while 4 studies⁶²⁶⁴,⁶⁶ were in HICs.

Quality assessment of included studies

The QUAPAS assessment was used for four studies,^{51 54 61 67} shown in online supplemental table 6. Three studies^{51 54 67} were graded as serious, while one study⁶¹ was graded as not serious. For QUADAS-2 assessments (online supplemental table 8), three studies^{55 60 64} were deemed serious, while three studies^{59 62 63} were assessed as not serious.

The Newcastle–Ottawa Scale for cohort studies was completed for 13 studies⁵¹,⁵³⁵⁶ and graded as good quality for 11 studies⁵¹,⁵³⁵⁶ and fair for 2 studies^{60 68} (online supplemental table 10). Additionally, the Newcastle–Ottawa Scale for case–control studies was performed in one study,⁶⁷ demonstrating good quality (online supplemental table 12).

Study results and data synthesis

Studies used different thresholds of tachypnoea: ≥60 bpm (12 studies),⁵¹⁵³,^{56 58} ≥70 bpm (2 studies)^{52 54} and ≥80 bpm (1 study).⁵⁴ Respiratory rates were assessed by medical clinicians (7 studies),⁵⁶⁶²,⁶⁷ clinical/research staff (2 studies),^{57 60} community/village health workers (7 studies)⁵¹,^{5558 61} or mothers (1 study).⁵¹ Nine studies5154 57 59 62 64 66,⁶⁸ assessed infants aged 0–28 days, two studies^{53 63} focused on infants aged 0–3 days, three studies^{53 58 60} on infants aged 0–6 days, one study⁵² on infants aged 0–10 days and four studies^{53 55 58 60} on infants aged ≥7 days. The GRADE assessment indicated very low certainty of evidence for tachypnoea in predicting mortality and diagnosing culture-confirmed or clinical sepsis. The detailed GRADE certainty of evidence tables are shown in online supplemental table 14.

Measures of association

Pooled ORs of different tachypnoea thresholds for mortality and sepsis are shown in figures2 3, respectively. The pooled ORs for tachypnoea ≥60 bpm were 3.14 (95% CI 1.30 to 7.56; 95% prediction interval 0.12 to 81.97; I²=92.02%; 5 studies; n=10 407) for mortality, 1.26 (95% CI 0.66 to 2.38; 95% prediction interval 0.08 to 19.04; I²=76.06%; 4 studies; n=3996) for culture-confirmed sepsis and 3.07 (95% CI 1.68 to 5.62; 95% prediction interval 0.40 to 23.73; I²=83.81%; 5 studies; n=10 124) for clinical sepsis. Tachypnoea≥70 bpm had 10.06 times greater odds of mortality (95% CI 2.39 to 42.35; I²=80.08%; 2 studies; n=12 138). Only one study reported data on tachypnoea ≥80 bpm, which had an OR of 44.4 (95% CI 13.8 to 142.6; n=5185) for mortality. Data on ORs for tachypnoea ≥70 bpm and ≥80 bpm for culture-confirmed or clinical sepsis diagnosis were not available. One study by Darmstadt et al⁵³ reported HRs for mortality, which were not included in the meta-analysis with other studies that reported ORs. Darmstadt et al⁵³ reported an adjusted HR (aHR) of 5.5 (95% CI 4.2 to 7.2; n=63 017) for the association between tachypnoea ≥60 bpm and mortality in infants aged <60 days (online supplemental table 5).

Figure 2. Forest plot of pooled OR for the association between tachypnoea and young infant mortality, stratified by respiratory rate thresholds. bpm, breaths per minute.

Figure 3. Forest plot of pooled OR for the association between tachypnoea and sepsis (culture-confirmed and clinical) by respiratory rate thresholds. bpm, breaths per minute; YICSG, Young Infants Clinical Signs Study Group.

Diagnostic accuracy measures

The diagnostic accuracy of different thresholds of tachypnoea to predict mortality and diagnose culture-confirmed and clinical sepsis in young infants aged 0–59 days is summarised in table 1. We pooled the sensitivity and specificity of tachypnoea at the ≥60 bpm threshold only, since only this threshold had a sufficient number of studies across all reference standards (figure 4). Tachypnoea≥60 bpm had a pooled sensitivity of 31% (95% CI 19% to 45%) and a specificity of 89% (95% CI 66% to 97%) for predicting infant mortality by 59 days (4 studies; n=7104). For diagnosing culture-confirmed sepsis, tachypnoea ≥60 bpm had a sensitivity of 28% (95% CI 10% to 58%) and a specificity of 67% (95% CI 53% to 78%) across three studies (n=693). For clinical sepsis, tachypnoea ≥60 bpm demonstrated a sensitivity of 49% (95% CI 31% to 67%) and a specificity of 70% (95% CI 58% to 80%) across four studies (n=1235). For tachypnoea ≥70 bpm, only one study (n=5214) reported a sensitivity of 10.9% (95% CI 5.3% to 19.1%) and a specificity of 99.4% (95% CI 99.1% to 99.6%) for predicting infant mortality. Meanwhile, for the highest threshold (≥80 bpm), one study (n=5185) reported a sensitivity of 5.7% (95% CI 1.9% to 12.9%) and a specificity of 99.9% (95% CI 99.7% to 99.9%) for mortality. No studies reported sensitivity and specificity data for tachypnoea ≥70 bpm or ≥80 bpm in diagnosing culture-confirmed or clinical sepsis.

Table 1. Summary of pooled accuracy measures of tachypnoea to predict mortality and diagnose sepsis in young infants aged 0–59 days, for each reference standard and stratified by respiratory rate threshold.

Reference standard	Threshold of respiratory rate	Number of studies for pooled sensitivity and specificity	Sensitivity (95% CI)	Specificity (95% CI)	Certainty of evidence sensitivity	Certainty of evidence specificity	Number of studies for pooled OR	OR (95% CI)
Mortality	≥60	4	31.0% (19.0% to 45.0%)	89.0% (66.0% to 97.0%)	⨁◯◯◯ Very low	⨁◯◯◯ Very low	5	3.14 (1.30 to 7.56)
	≥70	1	10.9% (5.3% to 19.1%)	99.4% (99.1% to 99.6%)	NA	NA	2	10.06 (2.39 to 42.35)
	≥80	1	5.7% (1.9% to 12.9%)	99.9% (99.7% to 99.9%)	NA	NA	1	44.40 (13.80 to 142.60)
Culture-confirmed sepsis	≥60	3	28.0% (10.0% to 58.0%)	67.0% (53.0% to 78.0%)	⨁◯◯◯ Very low	⨁◯◯◯ Very low	4	1.26 (0.66 to 2.38)
	≥70	0	NA	NA	NA	NA	0	NA
	≥80	0	NA	NA	NA	NA	0	NA
Clinical sepsis	≥60	4	49.0% (31.0% to 67.0%)	70.0% (58.0% to 80.0%)	⨁◯◯◯ Very low	⨁◯◯◯ Very low	5	3.07 (1.68 to 5.62)
	≥70	0	NA	NA	NA	NA	0	NA
	≥80	0	NA	NA	NA	NA	0	NA

Certainty of the evidence:

NA, not available.

Figure 4. Forest plot of pooled sensitivity and specificity of tachypnoea≥60 breaths per minute by reference standards.

The overall LR+ and LR− of tachypnoea ≥60 bpm for predicting infant mortality were 2.82 and 0.78, respectively, across four studies (n=7104). For diagnosing culture-confirmed sepsis, tachypnoea ≥60 bpm had an LR+ of 0.85 and an LR− of 1.07 across three studies (n=693). The 95% CI for the LRs could not be estimated for mortality and culture-confirmed sepsis at the threshold of ≥60 bpm due to sparse data. For clinical sepsis, tachypnoea ≥60 bpm had an LR+ of 1.64 (95% CI 1.15 to 2.33) and an LR− of 0.73 (95% CI 0.54 to 0.99) across four studies (n=1235). For tachypnoea ≥70 bpm, only one study (n=5214) reported an LR+ of 17.96 (95% CI 9.08 to 35.53) and an LR− of 0.90 (95% CI 0.83 to 0.96) for predicting infant mortality. Meanwhile, for the highest threshold (≥80 bpm), one study (n=5185) reported an LR+ of 41.86 (95% CI 13.55 to 129.30) and an LR− of 0.94 (95% CI 0.90 to 0.99) for mortality. For tachypnoea ≥70 bpm or ≥80 bpm in diagnosing culture-confirmed or clinical sepsis, the LR data were not available.

Subgroup analysis by age (<7 vs 7–59 days)

In the early neonatal period, Darmstadt et al,⁵³ Darmstadt et al⁵² and Rosenberg et al⁶³ reported associations between tachypnoea and our primary outcomes (mortality and culture-confirmed sepsis); however, we were unable to perform a pooled analysis. Darmstadt et al⁵³ reported the association between tachypnoea≥60 bpm and mortality, with an aHR of 6.9 (95% CI 4.7 to 10.3; n=56 889) in infants aged from 0 to <3 days, an aHR of 6.9 (95% CI 5.0 to 9.6; n=62 922) in infants aged from 0 to <7 days and an aHR of 4.2 (95% CI 2.6 to 6.9; n=59 704) in infants aged from 7 to <60 days. At the threshold of ≥70 bpm, Darmstadt et al⁵² reported an OR of 4.60 (95% CI 1.60 to 13.00) for mortality in infants aged 0–10 days. Rosenberg et al⁶³ reported an OR of 1.35 (95% CI 0.56 to 3.29; n=193) for culture-confirmed sepsis in infants aged from 0 to <3 days, though no respiratory rate threshold was specified.

Tachypnoea (≥60 bpm) in the first week was not significantly associated with clinical sepsis (OR 2.35; 95% CI 0.59 to 9.31; I²=88.25%; 2 studies; n=3474), while in the later period (≥7 days), tachypnoea≥60 bpm had an OR of 5.31 (95% CI 2.76 to 10.20; I²=70.62%; 2 studies; n=6358) (figure 5). The test of group differences was not significant (p=0.30).

Figure 5. Subgroup analysis of pooled OR for the association between tachypnoea≥60 breaths per minute (bpm) and clinical sepsis, stratified by age categories. YICSG, Young Infants Clinical Signs Study Group.

Discussion

Our systematic review and meta-analysis confirm the significant association between tachypnoea≥60 bpm and mortality and clinical sepsis in the first 59 days of life. Higher tachypnoea thresholds (≥70 or ≥80 bpm) corresponded to higher odds of death and higher specificity for mortality; however, the low sensitivity limits its clinical utility for risk screening as a standalone sign. There was inadequate data to assess the performance of tachypnoea in the first week of life to predict mortality or diagnose culture-confirmed sepsis.

Our findings are supported by a systematic review by Opiyo and English,⁶⁹ which found that a respiratory rate of ≥60 bpm, whether observed alone or alongside other clinical signs, could identify severe illness in young infants (0–59 days), defined as any cases that required either referral or hospitalisation, with ORs ranging from 1.5 to 3.1. Our findings also align with a previous systematic review by our group, Driker et al⁷⁰ that showed tachypnoea (not stratified by threshold) to be significantly associated with mortality, but not with culture-confirmed sepsis. While we found similar odds (threefold increased odds for mortality and clinical sepsis), we found that, as an individual sign, tachypnoea had low sensitivity for predicting mortality risk and diagnosing culture-confirmed and clinical sepsis in young infants. This finding aligns with a 1990 study by Morley et al, which found that respiratory rate alone is not very useful for indicating illness severity in infants.⁷¹ Given the relatively low sensitivity of tachypnoea as a single sign, tachypnoea should be interpreted in combination with other clinical signs (such as severe chest indrawing) for improved diagnostic accuracy.¹⁸

We did not review pneumonia outcomes since other systematic reviews and meta-analyses by Shah et al,²⁵ Rambaud-Althaus et al²⁶ and a pooled analysis by Rees et al²⁷ have already investigated this. Shah et al²⁵ reported that age-related tachypnoea also had low sensitivity (54% (95% CI 23% to 82%)) and low specificity (64% (95% CI 35% to 86%)) to diagnose pneumonia in children aged <5 years. In older children, Rambaud-Althaus et al²⁶ investigated that age-related tachypnoea had low sensitivity (62% (95% CI 26% to 89%)) and low specificity (59% (95% CI 29% to 84%)) to diagnose pneumonia in children aged from 2 months to 6 years. In those reviews,^{25 26} the authors defined age-related tachypnoea according to WHO criteria for pneumonia, that is, respiratory rate>60 bpm in infants aged <2 months, >50 bpm in children aged 2–11 months and >40 bpm in children aged 12–59 months. Additionally, a pooled analysis by Rees et al²⁷ reported that age-based tachypnoea had 92% sensitivity and 22% specificity to diagnose pneumonia in children aged 0–59 months. Rees et al²⁷ also found that among infants aged <2 months without wheezing, tachypnoea≥60 bpm showed 60% sensitivity and 39% specificity for pneumonia diagnosis, while a threshold of ≥70 bpm had 18% sensitivity and 83% specificity to diagnose pneumonia. Our review examined all-cause mortality and sepsis rather than pneumonia specifically. This all-cause mortality outcome encompasses deaths from both respiratory and non-respiratory causes.

At the ≥60 bpm threshold, culture-confirmed sepsis demonstrated the lowest sensitivity, which might be explained by stricter diagnostic criteria requiring positive bacterial cultures that are often difficult to obtain in young infants.⁷² Conversely, clinical sepsis showed the highest sensitivity at the same threshold, as it uses symptom-based diagnostic criteria that capture a wider spectrum of sepsis cases. Clinical sepsis showed the highest sensitivity and a relatively high OR. These findings should be interpreted with caution due to potential incorporation bias⁷³ since tachypnoea may be included as part of the clinical judgement component of the reference standard for clinical sepsis. In our systematic review, we excluded studies if they used clinical signs alone to diagnose clinical sepsis (ie, laboratory, X-ray or other corroborating data were required); however, the potential for incorporation bias still is possible, and thus we treat mortality and culture-confirmed sepsis as our primary diagnoses of interest in this manuscript.

TTN commonly occurs shortly after birth and typically resolves within 72 hours of life.²⁴ Our subgroup analysis demonstrated that tachypnoea≥60 bpm was not significantly associated with odds of clinical sepsis in young infants aged <7 days; however, odds of clinical sepsis were higher among young infants aged ≥7 days of life. This finding suggests that tachypnoea is more predictive of adverse outcomes in late young infants (≥7 days) since TTN is common in the early neonatal period. Higher thresholds of tachypnoea (≥70 or ≥80 bpm) have also been used in the early neonatal period, as demonstrated by Darmstadt et al,⁵² who applied the threshold of ≥70 bpm in neonates aged 0–10 days. In our review, which identified only two studies examining this early period, higher thresholds of tachypnoea (≥70 or ≥80 bpm) showed higher mortality odds but reduced sensitivity for mortality, capturing fewer infants at risk. Additional evidence of higher thresholds of tachypnoea in the early neonatal period is needed to understand how TTN may influence the performance of clinical sign algorithms such as IMCI.

Challenges in measuring respiratory rate in young infants

Measuring respiratory rate in infants presents challenges requiring clinicians or health workers to account for multiple factors, including environment, timing, measurement technique and underlying conditions. An infant’s respiratory rate is highly variable and may fluctuate frequently depending on circumstances, such as body temperature, sleeping position and the infant’s state (calm and awake, feeding, sleeping and agitated).⁷¹⁷⁴,⁷⁷ A study by Simoes et al⁷⁵ reported that among different ages of children, respiratory rate measurement in infants aged under 2 months was most difficult to obtain reliably. Young infants also spend most of their time in active sleep when their respiratory rate is higher and more irregular than when in quiet sleep.^{78 79} Therefore, tachypnoea alone is not sufficient to predict mortality and diagnose sepsis in young infants.

Implication for clinical practice and policy

Quantifying the accuracy of tachypnoea in predicting mortality and diagnosing sepsis in young infants is important for examining the validity of algorithms used to identify infants at high risk of adverse outcomes. Identifying infants at highest risk may aid resource allocation by prioritising infants who need urgent care and may guide referral and treatment decisions. In the current WHO IMCI algorithm, tachypnoea≥60 bpm in an infant aged <60 days initiates referral and intramuscular/intravenous antibiotics, and isolated tachypnoea in infants aged 7–59 days is used to indicate the need for oral antibiotics.¹⁸ Receipt of antibiotics may reduce the risk of mortality for infants with serious bacterial infections.

While we found that tachypnoea≥70 bpm and ≥80 bpm were associated with increased odds of death compared with the current WHO threshold of ≥60 bpm, this higher threshold demonstrated reduced sensitivity, which could miss infants at risk of death. Therefore, our findings support maintaining the current WHO threshold of ≥60 bpm for tachypnoea to maximise sensitivity for case detection, accepting lower specificity as a trade-off.

Strengths and limitations

To our knowledge, this is the first systematic review and meta-analysis of the diagnostic accuracy of tachypnoea identifying infants at risk of mortality and sepsis. Our review involved many research databases without language restrictions, which captured a diverse population across many countries and settings. Our study used three reference standards (mortality, culture-confirmed sepsis and clinical sepsis) to provide a comprehensive evaluation of infants at high risk. Additionally, we investigated different postnatal age periods of young infants (<7 days vs ≥7 days), given that TTN is common in the early neonatal period.

However, our review was limited by the lack of data on the timing of respiratory rate assessment, age group (<7 days vs ≥7 days, which was only available for clinical sepsis with a threshold≥60 bpm) and measurement of respiratory rate. In addition, only a few studies examined tachypnoea thresholds above 60 bpm or respiratory rate in the first week of life. We were also unable to assess publication bias due to the limited number of studies. Although inconsistency was present in our meta-analysis, this supports the generalisability of our findings to different clinical settings. The GRADE assessment indicated very low certainty of evidence for tachypnoea≥60 bpm in predicting mortality and diagnosing culture-confirmed or clinical sepsis. While this review provides valuable insights, findings regarding higher tachypnoea thresholds should be interpreted cautiously.

Implications for future research

Given the limited number of studies in our review, future studies should investigate the accuracy of different thresholds of tachypnoea to predict infant mortality and to diagnose sepsis, particularly within the first 48 hours of life and 0–6 days, when transient tachypnoea is very common in newborns. Most of the studies in this review were also conducted in a health facility setting, mainly inpatient care; thus, further studies in community settings may provide a more comprehensive understanding of the accuracy of tachypnoea to predict mortality and to diagnose sepsis in young infants in population-based settings.

Conclusions

Tachypnoea was significantly associated with increased odds of mortality and clinical sepsis, but not with culture-confirmed sepsis. Tachypnoea had low sensitivity as a standalone sign to predict mortality and to diagnose culture-confirmed and clinical sepsis in young infants aged 0–59 days. Although the higher threshold of respiratory rate (≥70 bpm) was associated with significantly greater odds of mortality, it was at the expense of sensitivity. Combining tachypnoea with other infant clinical signs may improve the accuracy of identifying the infants at highest risk. Future studies are needed to assess tachypnoea thresholds in the first week of life.

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information.