Predictors of death from severe pneumonia among children

2–59 months old hospitalized in Bohol, Philippines:

implications for referral criteria at a first-level health facility

S. P. Lupisan1, P. Ruutu2, P. Erma Abucejo-Ladesma3, B. P. Quiambao1, L. Gozum1, L. T. Sombrero1, V. Romano4,

E. Herva2, I. Riley5, E. A. F. Simoes6 and ARIVAC Consortium*

1 Research Institute for Tropical Medicine, Manila, Philippines2 National Public Health Institute, Helsinki, Finland3 Gov. Celestino Gallares Memorial Hospital, Bohol, Philippines4 University of the Philippines, Philippine General Hospital, Manila, Philippines5 Australian Centre for International Tropical Health and Nutrition, University of Queensland, Brisbane, Australia6 Department of Pediatrics, Division of Infectious Diseases, The University of Colorado at Denver and Health Sciences Center, and The

Children’s Hospital, Denver, CO, USA

Summary objective To determine predictors of death among children 2–59 months old admitted to hospital

with severe pneumonia.

methods Prospective observational study from April 1994 to May 2000 to investigate serious infec-

tions in children less than 5 years old admitted to a tertiary care government hospital in a rural province

in central Philippines. The quality of clinical and laboratory work was monitored. The WHO classifi-

cation for severe pneumonia was used for patient enrolment.

results There were 1249 children with severe pneumonia and no CNS infection. Thirty children died.

Using univariate analysis, the following factors were significantly associated with death: age

2–5 months, dense infiltrates on chest radiography and presence of definite bacterial pathogens in the

blood. Stepwise logistic regression analysis revealed the following independent predictors of death: age

2–5 months, weight for age z-score less than )2 SD, dense infiltrates on chest radiography and definite

pathogens isolated in the blood. When the results of chest radiographs and blood cultures were not

included to mimic facilities available at first-level facilities, age 2–5 months and weight for age z-score

less than )2 SD remained independent predictors of death.

conclusion When resources are limited, children with lower chest wall indrawing (severe pneumonia)

who are 2–5 months old or moderately to severely malnourished should be referred for immediate

higher-level care.

keywords Predictors of death, severe pneumonia, prospective study, Philippines

Introduction

Although the past decades have seen a decline in child

mortality because of pneumonia, it remains a very

important cause of death in developing countries (Williams

et al. 2002; World Health Organization 2002). In view

of the high ARI mortality in young children in developing

countries, the World Health Organization expanded its

activities in 1977 to include control of respiratory diseases

other than tuberculosis. The central strategy of this ARI

Programme involved case management, using simple clin-

ical signs for patient assessment. Children less than 5 years

old with cough and fast breathing without stridor or

cyanosis or inability to drink (danger signs) or chest wall

indrawing are classified as having non-severe pneumonia

requiring treatment with a safe, effective antibiotic at a

first-level facility (Cherian et al. 1988). Lower chest wall

indrawing in the absence of danger signs is a reliable sign

and currently the best indicator of severe pneumonia

(Shann et al. 1984; Mulholland et al. 1992). Children with

*Members of the consortium who contributed to this work are

Professor Helen Makela and Dr Antti Nissinen for consulting

ARIVAC laboratory work; Grace Esparar, RITM, Manila Philip-pines for laboratory supervision; Ma Felnore Girasol, Gallares

Hospital Bohol for laboratory work; and Professor Gail Williams,

University of Queensland, Brisbane, Australia, for consulting datamanagement.

Tropical Medicine and International Health doi:10.1111/j.1365-3156.2007.01872.x

volume 12 no 8 pp 962–971 august 2007

962 ª 2007 Blackwell Publishing Ltd

lower chest wall indrawing are at higher risk of death from

pneumonia than those without (Shann et al. 1989; Demers

et al. 2000), and should thus be referred for admission to

hospital for parenteral antibiotic treatment and supportive

care. Inability to drink or cyanosis is considered to be

danger signs indicating very severe pneumonia, likewise

requiring urgent referral (Shann et al. 1984; World Health

Organization 2002; Pio 2003). This strategy has been

shown to result in a 20–27% reduction in overall child

mortality and 36–42% reduction in pneumonia-specific

mortality (Sazawal & Black 2003).

By 1995, this case management strategy had been

adopted by 130 developing countries. It was then included

in the Integrated Management of Childhood Illness (IMCI)

Strategy (Gove 1997; Simoes et al. 1997; WHO Division of

Child Health and Development & WHO Regional Office

for Africa 19976 ). The IMCI Strategy is now being

implemented in more than 60 countries (Child and

Adolescent Health and Development 2005). However,

amongst the problems identified with widespread imple-

mentation of the program, two are particularly important

for the ARI component. These are the over-referral of

children with lower chest wall indrawing to hospitals from

first-level facilities and the accurate recognition of hy-

poxemia (World Health Organization 2004; Simoes et al.

2006). The latter becomes more acute when referral is

difficult (Simoes et al. 2003). Thus, for instance, in Uganda

(Peterson et al. 2004) and the Sudan (al Fadil et al. 2003),

where IMCI implementation was well received by the

community, few young children requiring immediate

referral were taken to a hospital within 24 hours. The

major reason for noncompliance with referral was cost.

This potential unnecessary referral of children with lower

chest wall indrawing adds to the caretaker cost of

management with little or no benefit to the sick child.

Chest indrawing is the only easily recognizable clinical

sign that indicates severe pneumonia, but not all children

with cough and lower chest wall indrawing have severe

pneumonia (Shann et al. 1984; Cherian et al. 1988; Shann

et al. 1989; Mulholland et al. 1992). One-third of children

with cough and lower chest wall indrawing have wheeze

(Simoes & McGrath 1992), one-third of whom respond

well to bronchodilators (Hazir et al. 2004). However, even

audible wheezing is not easily recognized by health

workers (Simoes & McGrath 1992). So, using wheeze or

bronchodilator, response may still not be the best methods

to reduce unnecessary referrals. Another 10–15% of

children with lower chest wall indrawing have only an

upper respiratory tract infection (Mulholland et al. 1992).

As there are no other readily recognizable signs and

symptoms that identify a child with severe pneumonia

(Shann et al. 1989; Demers et al. 2000), we tried to

identify other demographic and nutritional risk factors in

infants and young children with severe pneumonia that

predict death. This is an attempt at developing a triage

system to identify severely ill children with pneumonia,

who undoubtedly require referral in situations where

referral is difficult or impossible.

Materials and methods

Study site

We conducted a prospective observational study from

April 1994 to May 2000 to investigate serious infections in

children less than 5 years old admitted to a district

government hospital in a rural province on the island of

Bohol in central Philippines. Of the million inhabitants,

about 66000 live in the provincial capital, Tagbilaran City.

The Gov. Celestino Gallares Memorial Hospital (Bohol

Regional Hospital, BRH) serves as the primary centre for

paediatric and adult patients for the island, but the

majority are walk-in patients. One-third of admitted

patients come from Tagbilaran City and the rest from the

surrounding municipalities.

Study design and patients

In 1994, a clinical and laboratory research unit was

established at BRH for surveillance of serious childhood

infections. Dedicated study personnel included a paediat-

rician, two study nurses and three laboratory staff. The

staff of the Pediatric Department was trained in WHO ARI

case management protocols. Quality of clinical and

laboratory work was maintained through monthly visits by

consultants from the Research Institute of Tropical Medi-

cine (RITM) and semi-annual visits from the Finnish

National Public Health Institute (KTL) (Herva et al. 1999).

The overall study design has been described earlier (Herva

et al. 1999; Lupisan et al. 2000). Children with cough and/

or difficult breathing < 3 weeks’ duration were evaluated

for ARI according to the WHO classification (World

Health Organization 1990), outlined previously and the

signs and symptoms prospectively recorded in study forms

by the admitting paediatric resident, were confirmed by a

project physician.

Children 2–59-months old seeking care at the BRH were

enrolled if they had WHO defined severe or very severe

pneumonia (World Health Organization 1990). Meningitis

was suspected in infants and children with combinations of

the following signs: major signs/symptoms – convulsion,

abnormal sleepiness or difficulty in awakening, bulging or

tense anterior fontanel, Kernig’s sign, Brudzinski sign,

nuchal rigidity and lateralizing signs; minor signs were

Tropical Medicine and International Health volume 12 no 8 pp 962–971 august 2007

S. P. Lupisan et al. Predictors of death from severe pneumonia among children1

ª 2007 Blackwell Publishing Ltd 963

irritability, fever > 38 �C, vomiting and headache (in the

older child). Subjects were defined as having suspect

meningitis if they had two major signs, one major and

minor, or three minor signs.

Radiological methods

Antero-posterior and lateral chest radiographs were

requested for all infants and children with severe or very

severe pneumonia. Chest radiographs obtained were read

by an experienced paediatric study radiologist. Infiltrates,

when present, were described as dense (localized) or diffuse.

Bacteriological methods

Blood was collected for culture from all study patients,

and CSF specimens from patients with suspected meningitis

for culture and antigen detection after obtaining written

informed consent (as this was not part of routine care).

Bacterial growth was preliminarily identified in the BRH

Research Laboratory and confirmed in the RITM micro-

biology laboratory using standard methods (Herva et al.

1999). Streptococcus pneumoniae, Staphylococcus aureus,

Haemophilus influenzae, Salmonella typhi, other

salmonella, Escherichia coli, Pseudomonas aeruginosa and

other enterobacteriae were defined as definite pathogens.

Coagulase negative staphylococci, Bacillus, Diphtheroids

and Gram negative non-fermenting organisms were con-

sidered contaminants. Quality assurance was implemented

by confirming the identity of all isolates of S. pneumoniae

and H. influenzae at KTL (Herva et al. 1999).

Nutritional status

The sex-specific weight for age was compared with the

National Center for Health Statistics/Centers for Disease

Control reference population values (Dibley et al. 1987)

and z-scores were generated using the EPINUT program

(Epi InfoTM 6.04, Centers for Disease Control and

Prevention, Atlanta, Georgia, USA). Individuals with

weight for age z-score below )2 were defined as mal-

nourished.

Statistical analysis

Data were double entered using Epi-info version 6 (Centers

for Disease Control, Atlanta, Georgia). All statistical tests

and descriptive analyses were done using SAS statistical

software (SAS Institute, Carry, NC, USA). Statistical

significance of differences in rates was calculated using

Chi-square test or Fisher exact test where indicated. P

value (<0.05) was considered significant. STATA was used

for multivariable analysis of risk factors for outcome. The

outcome was defined as ‘Died’ if the patient died during

hospitalization, or as ‘moribund’ when discharged home

against medical advice. The outcome was defined as

‘Survived’ if the patient’s condition had improved war-

ranting normal discharge, or as ‘improving’ when dis-

charged against medical advice. For logistic regression, the

first model included all clinical, radiological and laboratory

variables with a P < 0.15 in the univariate analysis. In the

second model, only clinical indicators were used to mimic

the situation at a first-level facility.

Ethical assessment

The study was approved by the Institutional Review Board

of the Research Institute for Tropical Medicine, Depart-

ment of Health, Philippines.

Results

From 1 April 1994 to 31 May 2000, 1670 infants and

children 2–59 months old were admitted and treated for

pneumonia: 1289 (77.2%) for severe pneumonia and 381

(22.8%) for very severe pneumonia. There were 55

children with severe pneumonia and suspected meningitis,

of whom 17 were assessed to have had no CNS infection,

three had definite CNS infection and 35 were not reliably

classifiable for the presence of CNS infection in the absence

of CSF examination (Table 1). Included in the subsequent

analyses of this paper are subjects with severe pneumonia

(cough and lower chest wall indrawing) (N ¼ 1232), and

those with severe pneumonia with suspected meningitis on

admission, but assessed not to have had CNS infection

(N ¼ 17), excluding two cases transferred to another

hospital, leaving a total of 1249 cases of severe pneumonia

for further analyses.

Table 1 Distribution of Pneumonia Admissions, BRH, April1994–May 2000

Disease classification Frequency Per cent

Severe pneumonia only 1234 95.7

Severe pneumonia with suspect meningitisbut no CNS infection

17 1.3

Severe pneumonia with suspect meningitis

and with CNS infection

3 0.2

Severe pneumonia with suspect meningitisnot classifiable for CNS infection�

35 2.7

Total 1289

�No CSF examination.

Tropical Medicine and International Health volume 12 no 8 pp 962–971 august 2007

S. P. Lupisan et al. Predictors of death from severe pneumonia among children1

964 ª 2007 Blackwell Publishing Ltd

Patient profile

Almost two-thirds of the subjects were less than 1 year old;

31.7% were 2–5 months old and 32.7% were

6–11 months of age. A total of 74.1% had symptoms for

< 7 days (Figure 1), 51.6% had a temperature ‡ 38.5 �C,

97.4% had fast breathing, and wheezing was observed in

36.0%. Almost one-third of the infants and children were

malnourished (Figure 1). Only 1197 children (93.6%) had

chest radiographs. There were infiltrates in 47%, dense

ones in 30.0% and diffuse infiltrates in 17.0%. A bacterial

pathogen grew in blood culture in 32 patients (2.6%).

S. pneumoniae (Sazawal & Black 2003), H. influenzae

(Cherian et al. 1988), S. typhi (Williams et al. 2002) and

other salmonella (Williams et al. 2002) were the most

common pathogens isolated. Parents/guardians reported

prior antibiotic use in 32.3% of the study population.

Cotrimoxazole (11.0%) and amoxycillin (10.3%) were

more common antibiotics used at home prior to admission.

Risk factors for death

There were 30 deaths, with a case fatality rate of 2.4%.

Using univariate analysis, age between 2 and 5 months, the

presence of dense infiltrates on chest radiograph, and the

presence of definite pathogens in blood were significantly

associated with death (P < 0.05) (Figure 1).

In the multivariate model, no interaction was noted in

forward, backward and stepwise logistic regression. The

variables which were independently significantly associated

Figure 1 Severe pneumonia: Risk variables and outcome.

Tropical Medicine and International Health volume 12 no 8 pp 962–971 august 2007

S. P. Lupisan et al. Predictors of death from severe pneumonia among children1

ª 2007 Blackwell Publishing Ltd 965

with death were age 2–5 months old, weight for age z-score

less than )2 SD, the presence of dense infiltrates on chest

radiograph and isolation of definite pathogens in blood

(Figure 2). The Hosmer and Lemeshow test (P ¼ 0.807)

was used to test goodness of fit of the final model.

In the multivariate model, where the results of chest

radiograph and blood culture were not included, age

2–5 months (OR ¼ 3.82, 95% CI 1.61, 9.05, P ¼ 0.002)

and weight for age z-score less than )2 SD (OR ¼ 3.45,

95% CI 1.44, 8.26, P ¼ 0.005) remained the only

independent predictors of death.

When tested for validity, the independent predictors in

both models had high sensitivity. Both sets of independent

predictors had a higher specificity, 30.3% and 41.4%,

respectively, compared with lower chest wall indrawing

alone. The combination of independent predictors also had

higher positive and negative predictive values compared with

chest indrawing alone (Table 2). Of the 1249 infants and

children who would have been referred with lower chest wall

indrawing from a first-level facility, 396 were infants

2–5 months of age and 344 were older children with

malnutrition. This group included 86.7% of the 30 whodied.

Discussion

We observed in our prospective study in the Philippines on

infants and children admitted to hospital with severe

pneumonia, that young age of 2–5 months and malnutri-

tion were significant independent predictors of death, and

could be used as additional criteria for referral from a first-

level facility to a hospital in developing countries with

limited resources.

The case fatality rates from acute lower respiratory

infections in developing countries vary (Douglas 1991).

The 2.4% CFR in our relatively large study population is

rather low compared with case fatality rates of 3.4–

16.6% (Tupasi et al. 1988,1990a; Suwanjutha et al.

1994; Agrawal et al. 1995; Sehgal et al. 1997; Banajeh

1998) observed in other epidemiological studies in

developing countries including a study in an urban

population in the Philippines (Tupasi et al. 1988). The

observed low rate in our study may be due to the use of a

standardized ARI case management protocol by trained

and supervised physicians and nurses, the provision of the

recommended antibiotics, ancillary care including oxygen

Figure 2 Severe pneumonia: Clinical and laboratory predictors of death (stepwise backward and forward regression).

Table 2 Validity of risk variables to predict death

Risk variables Sensitivity Specificity

Positive

predictivevalue

Negativepredictive value

Lower chest wall indrawing� only 100% (85.9, 100) 0.0% (0.0, 0.4) 2.4 undefinedLower chest wall indrawing� and (Age 2–5 months or

weight for age z-score less than )2 SD or dense infiltrates

on chest radiograph or definite pathogens present in blood)

96.7% (80.9, 99.8) 30.3% (27.7, 33.0) 3.3 99.7 (98.3, 100)

Lower chest wall indrawing� and (Age 2–5 months orweight for age z-score less than )2 SD)

86.7% (68.4, 95.6) 41.4%(38.7, 44.3) 3.5 99.2 (97.9, 99.7)

�Criterion for classifying severe pneumonia in a child with cough or difficult breathing; present in all study children included in this analysis.

Tropical Medicine and International Health volume 12 no 8 pp 962–971 august 2007

S. P. Lupisan et al. Predictors of death from severe pneumonia among children1

966 ª 2007 Blackwell Publishing Ltd

therapy when indicated. This was supported by a general

acceptance of the ARI case management in the province

which was introduced into the Philippines in 1989 (Dayrit

1999).

The maximum impact of an ARI control programme is

seen in countries with a high infant mortality rate (IMR) of

‡40 or more infant deaths per 1000 live births (World

Health Organization 1994–1997). In 1994, when the

surveillance started, the IMR in Bohol was estimated to be

43.2, which gradually declined to 34.4 by 2000 (Bohol

Provincial Health Office statistics). The ARI control

program may have contributed to this decline in IMR in

Bohol and may partly account for the low CFR in our

hospital surveillance. The low CFR may also be partly

attributed to socio-economic progress and improvements

in the standards of living (Kohli & Al-Omaim 1983).

The mortality predictors found included blood culture

positivity and presence of dense infiltrates on chest radio-

graph. These findings are not surprising as blood culture

positivity indicates and a dense infiltrate is a strong proxy

for bacterial infection. The bacterial isolation rate was low,

despite the considerable effort made to obtain reliable

bacteriological results. A blood culture positive for a

definite pathogen, indicative of bacterial pneumonia, was

similarly observed as a predictor of death in a local study in

an urban community in Alabang, Philippines (Tupasi et al.

1988). Currently, the best available method for diagnosing

pneumonia is the chest radiograph. At one end of the

spectrum of radiological appearances, severe lobar con-

solidation, which would have been described as dense

infiltrate, is highly associated with bacterial pneumonia

(Cherian et al. 2005).

However, blood cultures and chest radiographs are not

available in first-level health care facilities in most devel-

oping country settings, and thus, are not useful indicators

for immediate referral. Hence, we performed a second

multivariable analysis using demographic information and

clinical signs and symptoms which showed that age

2–5 months and weight for age z-score less than )2 were

significant predictors of death.

Malnutrition has always been reported as a risk factor

for both childhood pneumonia morbidity and mortality

(Handayani et al. 1983; Tupasi et al. 1988, b; Berman

1991; Huffman & Martin 1994; Suwanjutha et al. 1994;

Victora et al. 1994; Agrawal et al. 1995; Sehgal et al.

1997; Banajeh 1998; Deb 1998; Victora et al. 1999; West

et al. 1999; Demers et al. 2000). Interestingly, long-term

survival of children admitted to hospital for severe pneu-

monia depends more on nutritional status than on being

hypoxemic (Victora et al. 1999). Tupasi et al. noted that

Filipino children with acute lower respiratory infection

(ALRI) with weight for age z-scores of less than )2 had a

twofold higher risk of dying (Tupasi et al. 1990b). Our

study showed that weight for age z-score less than )2 was a

significant predictor of death with an odds ratio of 3.57

(95% CI 1.33, 9.56). The current IMCI guidelines include

a z-score less than )3, to identify infants and children at

risk for death in the next month and for whom nutrition

counselling is advised (World Health Organization 2000).

Our study suggests that using a cutoff z-score )2 SD would

be useful to identify children with severe pneumonia at

high risk for dying from the acute infection. We chose to

use international standards for calculating weight for age

z-scores and not Philippine standards so that international

comparisons and extrapolations of data can be made. This

recommendation can easily be implemented, as most

IMCI-related programmes include a weight for age cutoff

line at z-score )2 as well as the standard )3 z-score cutoff.

Most deaths in infants and children with acute lower

respiratory tract infections occur in those less than 1 year

of age (Agrawal et al. 1995; Sehgal et al. 1997; Demers

et al. 2000). Our study further specifies 2–5 months of age

as a higher risk group. A combination of age 2–5 months

old or weight for age z-score less than )2 or presence of

dense infiltrates on chest radiograph or presence of definite

pathogens in the blood in a child with cough and chest

indrawing is a very sensitive screening tool, which will

enable health providers to pick up 96.7% of children with

severe pneumonia who are most likely to die. However,

this combination is not realistic at a first-level facility. The

more practical combination of age 2–5 months old or

weight for age z-score less than )2 in a child with cough

and chest indrawing may be less sensitive (86.7%), but

nonetheless more sensible as a screening tool in the field.

With this combination, we may have reduced by 41% the

proportion of potentially unnecessary referrals. When

compared with lower chest wall indrawing alone, both

statistically significant combinations are more specific. The

higher positive and negative predictive values of both

combinations compared with chest indrawing only further

underscores the clinical significance of these criteria in

children 2–59 months old with severe pneumonia. When

these predictors are absent, the patient will most likely not

die when treated appropriately.

The limitation of our study is that all children with

severe pneumonia were admitted to the hospital (following

WHO guidelines) and received parenteral antibiotics. They

were also closely monitored and were given oxygen when

deemed clinically necessary (a pulse oximeter was not

available at the time of this study at the BRH), hence we

may have prevented death in children by our intervention.

However, these results may not be generalizable to a

first-level facility. We recognize that the risks we identified

with potential prognostic relevance for deaths (in this

Tropical Medicine and International Health volume 12 no 8 pp 962–971 august 2007

S. P. Lupisan et al. Predictors of death from severe pneumonia among children1

ª 2007 Blackwell Publishing Ltd 967

hospital-based study) does not mean that all children who

did not present with these factors could be treated safely

at a first-level facility. However, it does generate a

hypothesis that could be tested. Current international

research initiatives are looking into measures that can

improve the referral of severe pneumonia and effective

management of severe pneumonia at first-level hospitals

(Rasmussen et al. 2000; World Health Organization 2004;

Simoes et al. 2006). This study presents evidence which can

be tested to decisively refine the IMCI guidelines for severe

pneumonia in children 2–59 months old and for possible

field-testing prior to inclusion in a guideline for manage-

ment of severe pneumonia in a first-level facility.

For a child assessed to have severe pneumonia at a first-

level facility, age 2–5 months or moderate to severe

malnutrition are both sensitive and specific independent

predictors of death. Thus, when there is poor access to

health care or referral is simply difficult, this group of

children may be prioritized for immediate referral to a

higher level of care. The strategy of using an oral

medication such as oral amoxicillin to treat severe pneu-

monia at a first-level facility preventing a referral to a

hospital in older non-malnourished infants and children

may have to be field tested and closely monitored. Oral

amoxicillin has been shown to be as safe and efficacious as

injectable antibiotics in a hospital-based study, where

oxygen was provided to hypoxemic patients and close

monitoring of patients occurred in this study; (Yobo et al.

2004) however, the safe use of oral amoxicillin has not

been demonstrated at a first-level facility, where subjects go

home, do not get oxygen and are not monitored. Where

highly penicillin-resistant pneumococci are prevalent,

treating children with 90 mg/kg/day of amoxicillin orally

should be considered as an alternative (Klugman & Madhi

2004).

Proper management of children presenting in first-level

facilities and hospitals with respiratory symptoms is the

cornerstone of pneumonia control. When resources both of

patients and health systems are limited, prioritizing which

children with severe pneumonia will be managed at the

first-level facility and which children will be referred for

immediate care, is a necessary step in delivering cost-

effective health care. Our study suggests a strategy that

could be tested in the field setting: in infants and children

with lower chest wall indrawing and no danger signs

presenting to a health worker at a first level facility with

limited access to a hospital or where referral is difficult, i.e.

for referral, give priority to infants 2–5 months of age and

older children less than )2 SD weight for age. A trial of

oral medications (amoxicillin 90 mg/kg/day) could be

administered to other infants and children with close

monitoring and follow-up.

Acknowledgements

We thank the parents of the children enrolled in this study

without whose support this study would not have been

possible and other members of the ARIVAC consortium

for their contribution to the project. In particular, we

express thanks to the ARIVAC laboratory staff for

laboratory work, the study nurses for patient enrolment

and follow-up and the staff of Gallares Hospital, Bohol, for

their contribution (the Paediatric and emergency room staff

for patient enrolment and management, Dr Jose Arcay for

supervising the radiographic examination and maintaining

quality control, the radiology staff for performance of chest

radiographs, Dr Juanita Arcay for support in establishing

the ARIVAC laboratory and acting as liaison between the

hospital and the ARIVAC project), Dr Romel Besa and

Carol Malacad for assisting in the review of literature and

Tarja Kaijalainen at KTL for quality assurance of pneu-

mococcal and H. influenzae bacteriology.

This work was supported by grants from The Academy of

Finland (contracts No 2041057 and 5569), Directorate

General Research of the European Union (contracts No.

18CT950025, IC18-CT97-0219, ICA4-CT1999-10008,

ICA4-CT-2002-10062), the Public Health Research and

Development Committee (PHRDS) of the National Health

& Medical Research Council (MHMRC), Australia (Project

No. 964161) and the Finnish International Development

Agency to Physicians for Social Responsibility, Finland.

References

Agrawal PB, Shendumikar N & Shastri NJ (1995) Host factors

and pneumonia in hospitalized children. Journal of the Indian

Medical Association 93, 271–272.

Banajeh SM (1998) Outcome for children <5 hospitalized with

severe acute lower respiratory tract infections in Yemen: a

5 year experience. Journal of Tropical Pediatrics 44, 343–346.

Berman S (1991) Epidemiology of acute respiratory infections in

children of developing countries. Reviews of Infectious Diseases

13 (Suppl. 6), 454–462.9

Cherian T, John TJ, Simoes EAF, John M, Steinhoff MC & John

M (1988) Evaluation of simple clinical signs for the diagnosis of

acute lower respiratory tract infections. Lancet 2, 125–128.

Cherian T, Mulholland EK, Carlin JB et al. (2005) Standardized

interpretation of paediatric chest radiographs for the diagnosis

of pneumonia in epidemiological studies. Bulletin of the World

Health Organization 83, 353–359. Epub 24 June 2005.

Child and Adolescent Health and Development (2005) Integrated

management of childhood illness. http://www.who.int/child-

adolescent-health/integr.htm.

Dayrit ESN (1999) A national program for control of acute res-

piratory tract infections: the Philippine experience. Clinical

Infectious Diseases 28, 195–199.

Tropical Medicine and International Health volume 12 no 8 pp 962–971 august 2007

S. P. Lupisan et al. Predictors of death from severe pneumonia among children1

968 ª 2007 Blackwell Publishing Ltd

Deb SK (1998) Acute respiratory disease survey in Tripura in case

of children below 5 years of age. Journal of the Indian Medical

Association 96, 111–116.

Demers AM, Morency P, Mberyo-Yaah F, Jaffar S, Blais C, Somse

P, Bobossi G & Pepin J (2000) Risk factors for mortality among

children hospitalized because of acute respiratory infection in

Bangui, Central African Republic. The Pediatric Infectious

Disease Journal 19, 424–432.

Dibley MJ, Goldsby JB, Staehling NW & Trowbridge FL (1987)

Development of normalized curves for international growth

reference: historical and technical considerations. The American

Journal of Clinical Nutrition 46, 736–748.

Douglas RM (1991) ARI in children in the developing world.

Seminars in Respiratory Infections 6, 217–224.

al Fadil SM, Alrahman SH, Cousens S et al. (2003) Integrated

management of childhood illnesses strategy: compliance with

referral and follow-up recommendations in Gezira State, Sudan.

Bulletin of the World Health Organization 81, 708–716.

Gove S (1997) Integrated management of childhood illness by

outpatient health workers: technical basis and overview. The

WHO Working Group on Guidelines for Integrated Manage-

ment of the Sick Child. Bulletin of the World Health

Organization 75, 7–24.

Handayani T, Mujiani, Hull V & Rhode JE (1983) Child mortality

in a rural Javanese village: a prospective study. International

Journal of Epidemiology 12, 88–92.

Hazir T, Qasi S, Nisar YB et al. (2004) Assessment and manage-

ment of children aged 1–59 months presenting with wheeze, fast

breathing, and/or lower chest indrawing; results of a multicentre

descriptive study in Pakistan. Archives of Disease in Childhood

89, 1049–1054.

Herva E, Sombrero L, Lupisan SP, Arcay J & Ruutu P (1999)

Establishing a laboratory for surveillance of invasive bacterial

infections in a tertiary care government hospital in a rural

province in the Philippines. The American Journal of Tropical

Medicine and Hygiene 6, 1035–1040.

Huffman SL & Martin L (1994) Child nutrition, birth spacing and

child mortality. Annals of the New York Academy of Sciences

709, 236–248.

Klugman KP & Madhi SA (2004) Oral antibiotics for the treat-

ment of severe pneumonia in children. Lancet 364, 1104–1105.

Kohli KL & Al-Omaim M (1983) Infant and child mortality in

Kuwait. Journal of Biosocial Science 15, 339–348.

Lupisan S, Herva E, Sombrero L et al. (2000) Invasive bacterial

infections of children in a rural province in the Central Philip-

pines. The American Journal of Tropical Medicine and Hygiene

62, 341–346.

Mulholland EK, Simoes EA, Costales MO, McGrath EJ, Manalac

EM & Gove S (1992) Standardized diagnosis of pneumonia in

developing countries. The Pediatric Infectious Disease Journal

11, 77–81.

Peterson S, Nsungwa-Sabiiti J, Were W et al. (2004) Coping with

paediatric referral—Ugandan parents’ experience. Lancet 363,

1955–6.

Pio A (2003) Standard case management of pneumonia in children

in developing countries: the cornerstone of the acute respiratory

infection programme. Bulletin of the World Health Organiza-

tion 81, 298–300.

Rasmussen Z, Pio A & Enarson P (2000) Case management of

childhood pneumonia in developing countries: recent relevant

research and current initiatives. The International Journal of

Tuberculosis and Lung Disease 4, 807–826.

Sazawal S & Black RE (2003) Pneumonia case management trials

group: effect of pneumonia case management on mortality in

neonates, infants, and preschool children—a meta-analysis of

community-based trials. The Lancet Infectious Diseases 3,

547–56.

Sehgal V, Sethi GR, Sachev HP & Satyanarayana L (1997) Pre-

dictors of mortality in subjects hospitalized with acute lower

respiratory tract infections. Indian Pediatrics 34, 213–219.

Shann F, Hart K & Thomas D (1984) Acute lower respiratory

tract infections in children: possible criteria for selection of

patients for antibiotic therapy and hospital admission. Bulletin

of the World Health Organization 62, 749–753.

Shann F, Barker J & Poore P (1989) Clinical signs that predict

death in children with severe pneumonia. The Pediatric

Infectious Disease Journal 8, 852–858.

Simoes EA & McGrath EJ (1992) Recognition of pneumonia by

primary health care workers in Swaziland with a simple clinical

algorithm. Lancet 340, 1502–1503.

Simoes EA, Desta T, Tessema T, Gerbresellasie T, Dagnew M &

Gove S (1997) Performance of health workers after training in

integrated management of childhood illness in Gondar, Ethi-

opia. Bulletin of the World Health Organization 75, 43–53.

Simoes EA, Peterson S, Gamatie Y et al. (2003) Management

of severely ill children at first level health facilities in sub-

Saharan Africa when referral is difficult. Bulletin of the

World Health Organization 81, 522–531.

Simoes EA, Cherian T, Chow J, Shahid-Salles SA, Laxminarayan R

& John T (2006) Acute respiratory diseases in children. In:

Disease Control Priorities in Developing Countries (2nd ed.)

(eds. DT Jamison, JG Breman, AR Measham, G Alleyne,

M Claeson, DB Evans, P Jha, A Mills and P Musgrove).

New York: The World Bank and Oxford University Press.

Suwanjutha S, Ruangkanchnasetr S, Chantarojanasiri T &

Hotrakitya S (1994) Risk factors associated with morbidity and

mortality of pneumonia in Thai children under 5 years. Southeast

Asian Journal of Tropical Medicine and Public Health 25, 60–66.

Tupasi TE, Velmonte MA, Abraham L et al. (1988) Deter-

minants of morbidity and mortality due to acute respiratory

infections: implications for intervention. The Journal of

Infectious Diseases 157, 615–623.

Tupasi TE, Lucero MG, Magdangal DM et al. (1990a) Etiology of

acute lower respiratory tract infection in children from Alabang,

Metro Manila. Reviews of Infectious Diseases 12 (Suppl. 8),

S929–S939.

Tupasi TE, Mangubat NV, Sunico ES et al. (1990b) Malnutrition

and acute respiratory tract infections in Filipino children.

Reviews of Infectious Diseases 12, S1047–S1054.

Victora CG, Fuchs SC, Flores JA, Fonseca W & Kirkwood B

(1994) Risk factors for pneumonia among children in a Brazilian

metropolitan area. Pediatrics 93, 977–985.

Tropical Medicine and International Health volume 12 no 8 pp 962–971 august 2007

S. P. Lupisan et al. Predictors of death from severe pneumonia among children1

ª 2007 Blackwell Publishing Ltd 969

Victora CG, Kirkwood BR, Ashworth A et al. (1999) Potential

interventions for prevention of childhood pneumonia in devel-

oping countries: improving nutrition. The American Journal of

Clinical Nutrition 70, 309–320.

West TE, Goetghebuer T, Milligan P, Mulholland EK & Weber

MW (1999) Long term morbidity and mortality following

hypoxemic lower respiratory tract infection in Gambian chil-

dren. Bulletin of the World Health Organization 77, 144–148.

WHO Division of Child Health and Development & WHO

Regional Office for Africa (1997) Integrated management of

childhood illness: field test of the WHO/UNICEF training course

in Arusha, United Republic of Tanzania. Bulletin of the World

Health Organization 75, 55–64.

Williams BG, Gouws E, Boschi-Pinto C, Bryce J & Dye C (2002)

Estimates of world-wide distribution of child deaths from acute

respiratory infections. The Lancet Infectious Diseases 2, 25–32.

World Health Organization (1990) Programme for the control of

acute respiratory infections: acute respiratory infections in

children: case management in small hospitals in developing

countries. Geneva: WHO/ARI/90.5.

World Health Organization (1999) Health Situation in the South-

East Asia Region (1994–1997). SEA/HS/209. New Delhi: World

Health Organization.

World Health Organization (2000) IMCI model handbook.

WHO/FCH/CAH/00.12. Geneva.

World Health Organization (2002) Evidence base for the

management of pneumonia. WHO/FCH/CAH/02.23. WHO,

Sweden.

World Health Organization (2004) Consultative meeting to review

evidence and research priorities in the management of acute

respiratory infections (ARI). WHO/FCH/CAH/03.9. WHO,

Geneva.

Yobo EA, Chisaka N, Hassan M et al. (2004) Oral amoxicillin

versus injectable penicillin for severe pneumonia in children

aged 3–59 months: a randomized multicentre equivalency study.

Lancet 364, 1141–1148.

Corresponding Author Eric A.F. Simoes,7 1056 East 19th Ave., Box B055, The University of Colorado Health Sciences Center and The

Children’s Hospital, Section of Pediatric Infectious Diseases, Denver, CO 80218, USA. Tel.: +1-303-861-6977; +1-303-764-8117

E-mail: eric.simoes@uchsc.edu

Facteurs predictifs de la mort par pneumonie severe chez les enfants de 2 a 59 mois hospitalises a Bohol aux Philippines: Implications pour les criteres

de transfert dans un service de sante de premier niveau

objectif Determiner les facteurs predictifs de la mort chez les enfants de 2 a 59 mois admis a l’hopital avec une pneumonie severe.

methodes Etude prospective d’observation d’avril 1994 a mai 2000 investiguant les infections serieuses chez les enfants de moins de 5 ans admis dans

un hopital gouvernemental de soin tertiaire dans une province rurale du centre des Philippines. La qualite des activites cliniques et de laboratoires a ete

surveillee. La classification de l’OMS pour la pneumonie a ete utilisee pour le recrutement des patients.

resultats 1249 enfants avec une pneumonie grave et sans infection du SNC ont ete recrutes. 30 enfants sont decedes. L’analyse univariee a revele que

les facteurs suivants etaient significativement associes a la mort: l’age de 2 a 5 mois, des infiltrations denses sur la radiographie de poitrine et la presence

de pathogenes bacteriens dans le sang. L’analyse de regression logistique par etapes a revele les facteurs independants suivants, predictifs de la mort:

l’age de 2 a 5 mois, poids-age Z-score <)2 ecart-type, les infiltrations denses sur la radiographie de poitrine et les pathogenes isoles du sang. Lorsque les

resultats des radiographies de poitrine et les cultures de sang ne sont pas inclus pour simuler les equipements disponibles dans les services de premier

niveau, l’age de 2 a 5 mois et le poids-age Z-score <)2 ecart-type demeuraient les facteurs independants predictifs de la mort.

conclusion Lorsque les ressources sont limitees, en plus pour des enfants avec une faible retraction des muscles sous-costaux, la pneumonie severe

chez les enfants ages de 2 a 5 mois qui sont moderement a severement sous-alimentes devrait etre transferee pour des soins immediats de plus haut

niveau.

mots cles Facteurs predictifs de la mort, pneumonie severe, etude prospective, Philippines

Tropical Medicine and International Health volume 12 no 8 pp 962–971 august 2007

S. P. Lupisan et al. Predictors of death from severe pneumonia among children1

970 ª 2007 Blackwell Publishing Ltd

Predictores de muerte por neumonıa severa en ninos de 2–59 meses, hospitalizados en Bohol, Filipinas. Implicaciones para criterios para derivacion en

un centro sanitario de primer nivel

objetivos Determinar los predictores de muerte en ninos de 2 a 59 meses de edad, admitidos en un hospital con neumonıa severa.

metodos Estudio observacional y prospectivo, entre Abril de 1994 y Mayo de 2000, para investigar infecciones graves en ninos menores de 5 a nos,

admitidos en un hospital gubernamental de tercer nivel en una provincia rural de Filipinas central. Se monitorizo la calidad del trabajo clınico y de

laboratorio. Se utilizo la clasificacion de neumonıa de la OMS al momento de incluir los pacientes en el estudio.

resultados Habıa 1249 ninos con neumonıa severa y sin infeccion del sistema nervioso central. 30 ni nos murieron. Utilizando un analisis uni-

variado, los siguientes factores estaban significativamente asociados con la muerte: la edad de 2 a 5 meses, infiltrados densos en la placa de torax, y la

presencia de patogenos bacterianos en la sangre. El analisis de regresion logıstica escalonada revelo los siguientes predictores independientes de muerte:

edad de 2 a 5 meses, un puntaje <)2 SD para el indicador z-score de peso-por-edad, infiltrados densos en la placa de torax, y el aislamiento de patogenos

de la sangre. Cuando no se incluıan los resultados de las placas de torax y los hemocultivos, con el fin de imitar las facilidades disponibles en un centro

sanitario de primer nivel, la edad de 2–5 meses y el puntaje del z-score (peso-por-edad) <)2 SD continuaban siendo predictores independientes de

muerte.

conclusion Cuando los recursos son limitados, ademas de los ni nos con tiraje intercostal bajo, los ni nos con neumonıa severa que tienen entre 2 y 5

meses o estan moderada o severamente mal nutridos, deberıan ser referidos inmediatamente a centros con un mayor nivel de cuidados.

palabras clave Predictores de muerte, neumonıa severa, estudio prospectivo, Filipinas

Tropical Medicine and International Health volume 12 no 8 pp 962–971 august 2007

S. P. Lupisan et al. Predictors of death from severe pneumonia among children1

ª 2007 Blackwell Publishing Ltd 971