Background Acute respiratory infections are the commonest cause of mortality and morbidity in children worldwide. A quarter of all deaths occur in India alone. In order to reduce this disease burden, there is a need for better diagnostic criteria, particularly ones allowing early detection of high-risk children.
Methods We enrolled 516 under 5 year olds, in four Indian hospitals, who met WHO age-dependent tachypnoea criteria for pneumonia at presentation. Patients underwent a protocolised examination assessing 29 items, including history, examination, O2 saturation, plus scores for chest X-ray, auscultation and conscious level. Treatment was determined by the emergency room (ER) physician. All children were reviewed at day 4 by a paediatrician and placed into four diagnostic categories: pneumonia, wheezy disease, mixed and non-respiratory.
Results The majority had wheezy diseases (42.8%). The remainder had pneumonia (35.9%), mixed disease (18.6%) and non-respiratory (2.7%). Best diagnostic predictors for wheezy disease were (auscultation/previous similar episodes) and for pneumonia (auscultation/CXR score). Mortality was 1.6%. Best disease severity predictors were conscious level, weight/age z score and respiratory/pulse rates.
Interpretation Current tachypnoea-based algorithms significantly overdiagnose pneumonia in children and underdiagnose wheezy diseases. Diagnostic accuracy can be improved by various combinations of clinical variables, but the best single diagnostic predictor is auscultation. Simple criteria can also be defined that reliably detect which tachypnoeic children are at high risk of death or deterioration. Management plans based on these protocols could reduce unnecessary antibiotic use, improve the management of wheezy diseases and reduce mortality by earlier identification of high-risk children.
- wheezy diseases
- developing countries
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
What is already known on this topic
Acute respiratory infections remain the commonest cause of death in children worldwide—currently killing over a million under 5 year olds each year.
Improving treatment outcomes requires updated diagnostic protocols that can differentiate pneumonia from wheezy diseases and reliably identify children at high risk of death.
What this study adds
WHO's widely used tachypnoea-based diagnostic criteria greatly overdiagnose pneumonia at the expense of underdiagnosing wheezy diseases.
Management based on our updated protocols could reduce unnecessary antibiotic use, improve management of wheezy diseases and reduce mortality by earlier identification of high-risk children.
The successful introduction of oral rehydration therapy in the 1970s,1 moved diarrheal diseases down to second place, behind acute respiratory infections (ARI), on the list of principal causes of death in small children. Despite the introduction of a WHO programme for the control of ARIs in the 1980s,2 and other targeted initiatives in the subsequent three decades,3 that first place ranking has not changed. Mortality has reduced over that period but the fraction due to ARIs has remained constant at roughly 1 in 5 of all deaths.4 As a recent example shows—of the 7.6 million deaths of children under 5 years in 2010, 1.40 million were due to ARIs (18.1% of total, range 1.2–1.6 million).4
Although there have been extraordinary improvements in child mortality over the last 20 years, those gains have not been evenly distributed. The global distribution of ARI deaths reflects the increasing fraction of total child deaths found in the South Asian and Sub-Saharan African regions.5 Currently, half of all ARI deaths occur in only five countries, all of which are within those two zones.6 The National Million Death Survey estimates that ARIs kill over 350 000 children every year in India alone from an estimated total of over 30 million cases!7 Programmes aimed at reducing ARI mortality should clearly be an essential element of any initiative aimed at accelerating progress towards global child mortality targets.8
While prevention of ARIs is obviously better than cure, the necessary public health interventions, such as immunisation, housing and sanitation, are expensive and take time.9 Until then, ARI mortality reduction is dependent on case detection and treatment.10 The WHO's widely used diagnostic and severity criteria for ARIs were developed in the 1980s to guide management by village health workers11 (table 1). They have served that purpose well,12 but are not accurate enough to meet current epidemiological and clinical requirements.13 We organised a multicentre Indian study designed to meet the need for improved diagnostic protocols for ARIs.14 The main aims were to develop clinical criteria, suitable for low-resource settings, able to differentiate pneumonia from wheezy diseases15 and also allow early detection of children at high risk of death.16
This was a prospective observational study run simultaneously in four Indian public hospitals (King George Medical University, Lucknow; Regency Hospital, Kanpur; Vanivilas Hospital, Bangalore; Bowring and Lady Curzon Hospital, Bangalore). The study ran from October 2012 to April 2013 to cover the Indian respiratory viral season. In each centre, the research team consisted of a qualified paediatrician as leader, plus paediatric clinic physicians and a postgraduate research coordinator. The research coordinator was responsible for patient enrolment, data collection, patient follow-up, regular electronic meetings with the Canadian team and data transmission. Because of the use of standardised scoring systems and the need for accurate clinical data collection, a member from the Canadian team spent a week at each centre familiarising local research team members with the study protocol and standardised scoring systems. This was followed by a 1-week trial period of data collection and electronic transmission of files to Canada.
Diagnostic definitions and standardised scores
The primary problem facing any study of pneumonia (or asthma) is accurate diagnosis.16 The overlap of clinical and radiological findings between severe viral infections, asthma and bacterial pneumonia makes it impossible to establish unambiguous measurable definitions suitable for research.17 In many low-resource areas, this is further complicated by infectious diseases, such as malaria and dengue, which can have similar presentations.18 Early WHO tachypnoea-based diagnostic protocols were intentionally over-sensitive to ensure that all children with bacterial pneumonia received antibiotics.3 Later attempts were made to improve the detection of wheezy diseases, by adding audible wheeze or acute bronchodilator response to the basic criteria.19 However, these were shown to be imprecise, particularly among the sickest children.20
We based diagnosis on the interpretation of a standardised set of investigations by a well-trained paediatrician. This is not a gold standard but it does represent common clinical practice on paediatric wards everywhere. We expanded the method described by Sachdev et al21 who classified patients into four groups (pneumonia, wheezy disease, mixed and non-respiratory) based on consultant review of a detailed history and examination plus a chest radiograph (CXR). We recorded 29 items from a protocol that included history, examination, CXR and oximetry (table 2). In order to combine results from data collected in different centres, we used standardised scoring systems for conscious level22 and auscultation findings.23 For chest radiographs, we used a modification of the recently updated system recommended by the WHO (table 3).24 During the 1-week preparatory period, the system was explained to all involved ER physicians and paediatricians using examples and practice interpretation.
All children below 5 years who presented to the emergency rooms of the study hospitals with cough or difficulty breathing of less than 5 days were identified. If their initial respiratory rate met WHO criteria for pneumonia (table 1), the study was explained by a native speaker of their primary language and they were invited to enter the study. Families were not paid to enrol but the study covered the cost of a CXR for every child plus travel expenses for outpatients to return for review on day four. After enrolment, 29 features of the child's history, examination and CXR were assessed by the ER physician and recorded by the study coordinator (table 2).
After reviewing the data, the ER physician was asked to place the child into one of four diagnostic categories: pneumonia, wheezy disease (asthma and bronchiolitis), mixed (evidence of pneumonia and wheeze) and non-respiratory (malaria, dengue, etc). The ER physician was solely responsible for subsequent management decisions. All study patients were reviewed 4 days later by a qualified paediatrician who was blinded to the ER physician's CXR interpretation and diagnosis. Based on a review of the clinical data at presentation, plus subsequent course over 4 days and a second examination, the paediatrician placed the patient into one of the four diagnostic categories. This was considered the child's final diagnosis for analysis.
In order to determine which of the 29 variables assessed in the ER had the best ability to predict the paediatrician's diagnosis at day 4, we chose to use multiple logistic regression rather than multiple linear regression. It is the preferred method when the dependent variable is binary (pneumonia yes/no). Once the best predictors were identified using the logistic model, the accuracy of individual and combined variables in predicting paediatrician diagnosis and patient outcome was assessed by calculating sensitivity, specificity plus positive and negative predictive values from conventional 2×2 tables using standard equations. The best cut-off values for continuous variables were established using receiver operating curve analyses. When the predictive value of combining variables was tested, we chose to link them as ‘A and/or B’. This increases sensitivity but usually decreases specificity compared with using ‘A and B’.
All the measured end points consisting of continuous data (pulse, respiratory rate, temperature, weight, age, oxygen saturation) were skewed and did not satisfy the Kolmogorov–Smirnov or Shapiro–Wilk tests for normal distribution. As a result, we expressed the results (both graphically and in the text) as median plus IQR. Multiple comparisons of values between three groups displayed in figures 2 and 3 were made using the non-parametric Kruskal–Wallis test with Dunn's post hoc test. The degree of diagnostic agreement between ER physician on day 1 and paediatrician on day 4 was assessed using Cohen's d statistic with SE.
We enrolled 524 patients: 189 female, 335 male, median age 11 months, IQR 4–27.8 months. A total of 278 patients were admitted to hospital ward, 246 were treated as outpatients. Mortality and morbidity assessed at day 4 were low. Eight children (1.6%) died and 12 (2.3%) worsened despite treatment. The remainder all improved with treatment. As a reflection of the seriousness with which research into paediatric pneumonia is taken in India, study staff managed to follow all inpatients and only eight outpatients were lost to follow-up. Final study total was 516 children.
Every enrolled patient fulfilled WHO's diagnostic criteria for pneumonia at time of presentation (figure 1). Those criteria still appear to be a reliable screening test for the presence of respiratory disease since only 2.7% had non-respiratory diseases (8 dengue, 4 typhoid/paratyphoid, 1 malaria, 1 congenital heart disease). However, as many others have reported, tachypnoea alone is not an accurate indicator of the type of respiratory disease.13 Only 35.9% were diagnosed with pneumonia at day 4 review by a paediatrician. Of the remainder, 42.8% had wheezy diseases and 18.6% showed mixed findings of wheezy disease and pneumonia. Although the paediatrician was blinded to the ER physician's original diagnosis and CXR interpretation, the agreement between their clinical opinions was close (figure 1). Agreement was better for wheezy diseases (Cohen's d 0.87, SE 0.02) than for pneumonia (Cohen's d 0.68, SE 0.03).
Value of chest radiographs
The study paid for a chest radiograph for every enrolled patient in the expectation that CXRs would prove to be of major diagnostic value in tachypnoeic children. However, under practical conditions, problems with film quality, film interpretation and interobserver variability combined to make chest radiographs less reliable than we had initially assumed. Our modified six-part reporting system (table 3) was unnecessarily complex. For pneumonia, the best predictive CXR findings were major patches and/or pleural effusions and/or lobar changes (table 4). This corresponds to ‘end point consolidation’ and/or effusions using the WHO scoring system (table 3).24 Chest radiograph findings had no predictive value for wheezy diseases.
While children diagnosed with pneumonia tended to have higher fever and lower oxygen saturations compared with those with wheezy disease or mixed cases, there was considerable overlap (figure 2). The best positive and negative predictive values for pneumonia were given by crackles or bronchial breathing on auscultation (table 4). In low-resource areas, sensitivity can be improved by combining auscultation with fever >38.6° but this reduces positive predictive value. Where chest radiography is available, a greater increase in sensitivity can be gained by combining auscultation with CXR findings (major patches, pleural effusion or pleural fluid). Again, this is at the expense of positive predictive value (table 4) (figure 2). Auscultation alone remains the single most useful test.
Predicting wheezy disease
The best positive and negative predictive values for wheezy diseases were given by wheeze on auscultation (table 4) (figure 2). In keeping with previous studies,21 ,25 we found that a history of two or more similar past episodes had good specificity but low sensitivity for predicting a diagnosis of wheezy disease. Combining these two indicators added little to the predictive power of auscultation alone other than small increases in sensitivity and negative predictive value (table 4). No other variable, including hyperinflation reported on CXR, had any predictive value for the diagnosis of wheezy disease. As with pneumonia, the single best diagnostic predictor was auscultation.
Diagnostic uncertainty and mixed cases
WHO's tachypnoea-based diagnostic criteria correctly predicted pneumonia in little more than 1 in 3 cases and had no predictive value for asthma. Auscultation-based criteria offered a significant improvement with best positive and negative predictive values around 80% for both asthma and pneumonia (table 4). This is an improvement but it is still far from perfect. Given the clinical and radiological overlaps between asthma and pneumonia, cases of diagnostic confusion will exist even under the best circumstances. We found that 18% of cases did not fit into either of the two main diagnostic categories. They were classified as ‘mixed’.
Predicting death or deterioration
All eight children that died were classified as severe or very severe using WHO criteria (three pneumonia, five mixed). There were no deaths in the wheezy disease group. At day 4 review, a further 12 children were considered to have deteriorated (8 pneumonia, 3 wheezy, 1 mixed). Although the WHO's severity criteria, developed in the 1980s,26 are still relevant, we found that the AVPU score (table 2) was easier to define and standardise than the WHO's use of ‘lethargy or poor feeding’. The most sensitive screen for risk of death was the conscious level score (table 4). Tachypnoea (>66/min), tachycardia (>166/min) and malnutrition (weight for age z score >−3) were the best combined predictors of death or deterioration (figure 3). Combinations of these clinical measures provide good screening criteria for the sickest children with sensitivities around 80% and negative predictive powers greater than 95% (table 4).
Our study is the first to provide estimates of the common causes of tachypnoea in under 5-year-old children in a low-resource region. It confirms the frequently repeated suggestion that tachypnoea-based criteria overdiagnose pneumonia at the expense of significantly underdiagnosing wheezy diseases.15 ,20 Our diagnoses were based on review of carefully collected clinical data by well-qualified paediatricians. They are not gold standards but are designed as a formalised version of common clinical practice on paediatric wards everywhere. Based on our results, we would suggest that global estimates of childhood pneumonia are too high—perhaps by a factor of two or more. There were no deaths in the wheezy disease groups so it is possible that current estimates of mortality due to pneumonia are more reliable than estimates of its incidence. We estimate that pneumonia case fatality is in the 2–4% range rather than current 1–2% estimates.7
Our findings suggest that WHO's tachypnoea-based criteria are more valuable when used as a screening test for respiratory disease rather than a diagnostic test for pneumonia. Their predictive accuracy for pneumonia is low and they have no discriminatory value for defining wheezy diseases. However, they accurately predicted the presence of respiratory diseases, at time of presentation, in over 97% of cases. We would suggest a two-stage diagnostic process for managing children presenting with tachypnoea—an initial screen using WHO criteria, followed by selection into management groups based on careful examination and validated diagnostic criteria. Given the limited literature in this area, our own approach cannot be viewed as anything more than a preliminary study that needs further confirmation. However, we found that patient selection using age-dependent tachypnoea criteria, combined with auscultation findings, was able to identify pneumonia and wheezy diseases with a predictive value of 80%. Given the predictive limitations of current clinical and radiological tests, some diagnostic uncertainty is inevitable, even under the best conditions. The best management of cases showing mixed signs will require more study, but caution would dictate the use of antibiotics in this group.
We have presented the predictive values of several combinations of clinical variables (table 4), but it is important to emphasise that the single most reliable indicator (both for wheezy diseases and pneumonia) was auscultation. We also defined simple clinical signs that allow early detection of children at risk for death or deterioration. While these criteria are within the resources of urban public hospitals, our next step will be to examine their applicability for rural India where deaths from respiratory diseases are greatly over-represented.7 Auscultation was not included when WHO's diagnostic criteria were first developed for village health workers in the 1980s.2 ,3 We believe the time has come when auscultation should be taught at all levels, along with the ability to collect accurate vital signs.27 If midwives can detect a fetal heart rate using a primitive fetoscope, then well-trained rural health workers of equivalent experience should be trusted to differentiate wheeze from crackles using a modern stethoscope.
Technology (chest X-ray and pulse oximetry) proved of limited added benefit over clinical skills. Oxygen saturation, measured at presentation, had no predictive ability either for disease type or disease severity. Chest radiography had no predictive value for wheezy disease or disease severity. While CXRs had value in predicting pneumonia, their accuracy was no greater than auscultation (table 4). The absolute requirement for any predictive system based on clinical signs is accurate measurement—this cannot be emphasised enough. Predictive cut-off values have little meaning unless they are applied to carefully taken measurements. Unfortunately, ‘simple’ clinical signs are not so simple to measure accurately in a busy emergency room.28 We believe the most relevant place for technology in low-resource areas is a cheap and accurate monitor to measure vital signs.
Our findings concerning the underdiagnosis of wheezy diseases are in line with the limited available literature concerning misdiagnosis and underdiagnosis of bronchiolitis and asthma in poor countries. In studies from Kenya29 and South Africa,30 over 30% of the children who met WHO's severe or very severe pneumonia criteria tested positive for RSV. Similarly, studies of asthma prevalence in Tanzania,31 Brazil32 and Uganda33 all showed unexpectedly high rates of 10–15% among local under-5-year children. In the absence of clear clinical definitions for bronchiolitis and asthma, it is difficult to subdivide wheezy diseases into diagnostic groups but patient age distribution offers some clues. Of the 42.8% that were diagnosed with wheezy diseases in our study, almost half were less than 12 months of age. If this is accepted as a crude dividing line between bronchiolitis in the younger and asthma in the older, then numbers of both these common conditions are considerably underestimated in India and likely in other low-resource regions.
Since pneumonia is the commonest single cause of death in small children worldwide, developing better treatment protocols and defining more accurate diagnostic criteria for ARI are essential steps in accelerating progress against avoidable childhood deaths.14 We believe that management plans for tachypnoeic children, based on our diagnostic protocols, could reduce unnecessary antibiotic use,34 improve the management of wheezy diseases and reduce mortality by earlier identification of high-risk children. We hope they prove to be of value to others working in this field.
Contributors SA and AV (Lucknow), RK and AA (Kanpur), DS, BG and NP (Bangalore, Vanivilas), PV (Bangalore, Lady Curzon) were responsible for collecting patient data in their respective hospitals and transmitting all records to Canada. MS and VG stored and analysed all the collected data. All listed physicians were then equally responsible for preparation of the final manuscript. The corresponding author had full access to all study data and had final responsibility for the decision to submit the finished manuscript for publication.
Funding The study was funded by the Judi Bowden Memorial Fund for Respiratory Research. Apart from providing money, the funding source had no other role in any aspect of the study.
Competing interests None.
Ethics approval Because of the joint nature of the study, it was reviewed extensively by the research ethics boards of all involved institutions. The study was performed in accordance with current research standards of ethics and human rights. The study protocol was reviewed and given formal research approval by (1) University of BC's research ethics committee. (2) BC's Children's Hospital research ethics committee. (3) Research ethics board of the King George's Medical University, Lucknow. (4) Institutional research ethics committee of Regency Hospital, Kanpur. (5) Research ethics review board of Bangalore Medical College and Research Centre. (6) India's National Epidemiology Network Research Ethics Committee (INDIAClen).
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement No further data is available.