Secondary Logo

Journal Logo

Original Studies

Hospital-Based Case-Control Study of Bronchiectasis in Indigenous Children in Central Australia

Valery, Patricia C. MD, MPH, PhD*†; Torzillo, Paul J. MBBS, FRACP; Mulholland, Kim MD, FRACP#; Boyce, Naomi C. BN**; Purdie, David M. BSc, PhD*‡; Chang, Anne B. MBBS, FRACP, PhD§∥

Author Information
The Pediatric Infectious Disease Journal: October 2004 - Volume 23 - Issue 10 - p 902-908
doi: 10.1097/01.inf.0000142508.33623.2f
  • Free

Abstract

Chronic suppurative lung disease not related to underlying causes such as cystic fibrosis (CF), ciliary dyskinesia, immunodeficiency or focal lung abnormalities has largely disappeared in affluent countries.1 However, it is still common in less affluent countries2 and within disadvantaged groups in affluent countries such as among Australian Indigenous children,3 Alaskan children in the United States,4 Maori and Pacific Islanders in New Zealand5 and where the burden of acute respiratory illness is high. In Australia, the burden of respiratory disease remains high in the Indigenous population6 and recent data report that deaths from respiratory illness remain the most common preventable defined cause of death in Indigenous Australians in the 0–1 year age group, with rates 5 times that of non-Aboriginal Territorians.7 In Central Australia, the prevalence of high resolution computed tomography (HRCT)-proved8,9 bronchiectasis in children (younger than 15 years old) is 147/10,000 Aboriginal children (also referred here as Indigenous), which far exceeds the prevalence of cystic fibrosis in non-Indigenous Australian children (35/100,000).3

Although children with non-CF bronchiectasis have increased episodes of past pneumonia,3,4 there are few published data on the epidemiology of bronchiectasis in the last 30 years, and medical knowledge and investigations have significantly changed during this period. Most of the 83 children in Maxwell’s10 retrospective review of Central Australian Indigenous children with bronchiectasis (diagnosed by chest radiograph and/or bronchography) had a preceding lower respiratory tract infection (47% had bronchiolitis, 37% had pneumonia, 11% had unspecified recurrent chest infection). In another study, Maxwell et al11 reported that 83% of children with undefined respiratory illness had a previous respiratory infection. Data from the developing world cannot be appropriately applied to children in affluent countries because measles, tuberculosis and pertussis, common forerunners of childhood bronchiectasis in less affluent countries,2 are relatively rare in Australian Indigenous communities. Although childhood pneumonia is strongly associated with bronchiectasis later in life,12 there has been no case-control study that has specifically examined this (PubMed Search, September 2003).

We present the findings of the first case-control study of bronchiectasis. It examined Indigenous children with and without bronchiectasis, living in a remote region of Australia, with particular emphasis on previous hospitalized episode(s) of pneumonia.

METHODS

Study Design and Subjects.

A case-control study of bronchiectasis in Indigenous children was conducted at the Alice Springs Hospital (ASH). The ASH covers a 1 million-km2 catchment area. For reasons of geographic isolation, it provides a wide range of services not normally seen in a hospital of its size (170 beds), with 4–5 full time consultant pediatricians and a pediatric respiratory pulmonologist at the time of the study. The study was approved by the human ethics committee of the region to undertake a chart review of cases and controls. Because there was no contact with patients or their families, individual informed consent was not sought.

The medical records of 79 children (all Indigenous) admitted for clinical chronic suppurative lung disease at the ASH during June 1991 and March 2002 were examined. Eligible cases (61) were Indigenous children age 0–17 years, residents of Central Australia, who were admitted with chest HRCT-proved bronchiectasis (performed during a nonacute state13) according to standard criteria as previously reported.14 All children were investigated for the usual causes of bronchiectasis (cystic fibrosis and immunodeficiencies). Exclusion criteria included: no HRCT available; child had a known cause of bronchiectasis (immunodeficiency or cystic fibrosis); no radiologic confirmation (HRCT) of bronchiectasis; and not resident of Central Australia (hospital catchment area). Controls (3/case) were matched to cases by gender, date of birth and year of diagnosis and had to be hospitalized in the same calendar year as cases. They were Indigenous children who were hospitalized in the same calendar year as matched case for injury, cardiovascular or renal conditions, gastroenteritis, abscess or failure to thrive. Whole chart review was undertaken to ensure controls were not suspected to have bronchiectasis. For both cases and controls, the time frame for chart review was the child’s lifetime. Cases were matched to the closest age-matched control available; all cases younger than 24 months were age-matched within 12 months, with 67% of controls being within ±24 months of cases date of birth.

Information was abstracted from the medical charts with the use of a structured data collection form that included demographics, birth-related factors, anthropometry, immunization and medical history, particularly history of hospital-treated episodes of several illnesses. Although the medical chart format was not fully uniform, it was relatively consistent over the period of the study, though the quality of data available varied by physician. The diagnosis of different medical conditions (eg, pneumonia, bronchiolitis) recorded were made by the pediatricians working at the ASH at the time. Children with pneumonia had to have been hospitalized, have received intravenous antibiotics and have had a supporting radiology report of lobar pneumonia or bronchopneumonia.

Statistical Analysis.

Statistical analyses were completed using the SPSS statistical software package (version 11.0). Multiple logistic regression was applied to determine the association between bronchiectasis and variables of interest while adjusting simultaneously for the effect of potential confounding by the others and to analyze the association between lobes involved in childhood pneumonia and lobes involved in bronchiectasis (controls were assigned the reference bronchiectasis site of their matched case). Odds ratios (ORs) presented here, unless specified, have been adjusted for age, sex, calendar year (stratifying factors in the design), resident community (categorized according to distance to Alice Springs because of possible confounding by access to a hospital), pneumonia, otitis media, bronchiolitis and malnutrition. Results are presented using adjusted ORs, 95% confidence intervals (95% CI) and 2-sided P values as appropriate. For nonnormally distributed data (number of days in hospital because of pneumonia, age at first pneumonia episode), nonparametric tests (Mann-Whitney U test) were used to calculate P value for the difference medians.

RESULTS

The characteristics of cases (61) and controls (183) included in the study are shown by diagnosis in Table 1. Cases were 8 months–15 years of age at diagnosis (median age, 5.3 years). The case and control groups were similar with respect to age, sex and year of diagnosis for cases and year of hospitalization for controls. Greater than two-thirds of cases lived in more distant communities (>180 km from Alice Springs) compared with approximately one-half of the controls. Children diagnosed with bronchiectasis were 3 times more likely to reside outside Alice Springs and its surrounding communities (OR 3.1, 95% CI 1.2–8.0).

T1-4
TABLE 1:
Characteristics of Bronchiectasis Cases and Controls Included in This Study

History of Hospitalized Pneumonia.

Of particular note was the excess of pneumonia history in cases. There were 449 episodes of hospitalized pneumonia recorded in the medical charts of 148 children [248 episodes were seen in 58 cases (95.1%) and 201 episodes in 90 controls (49.2%)]. Ninety-six children included in the study (5% cases and 51% of controls) had never been treated for pneumonia at the ASH. Children who had been previously hospitalized for pneumonia were 15 times more likely to develop bronchiectasis (OR 15.2; 95% CI 4.4–52.7). Of the 449 pneumonia episodes among all children, there were 436 episodes of lobar pneumonia and 13 of bronchopneumonia. According to pneumonia site, odds ratios were 7.0 (95% CI 1.7–28.8) for bronchopneumonia (reference “never had bronchopneumonia”) and 11.3 (95% CI 3.8–34.1) for lobar pneumonia (reference “never had lobar pneumonia”). The specific site of lobar pneumonia was not positively associated with bronchiectasis, but there was a nonsignificant relationship between 2 or more and 3 or more episodes of lobar pneumonia with the specific bronchiectasis lobe, odds ratios were 1.3 (95% CI 0.3–1.6) and 2.7 (95% CI 0.7–9.7), respectively (here odds ratios were also adjusted for total number of pneumonias).

About one-third of pneumonia episodes (94 in cases and 31 in controls; P < 0.001) required oxygen: 99 episodes in cases and 30 episodes in controls required mask or nasal cannula; 12 cases and 2 controls required tent or box; and 7 cases and 1 control required positive pressure ventilation. For the majority of the pneumonia episodes (392), there was follow-up recorded in the ASH medical records; of the 34 seen within 2 months of discharge, 25 had a follow-up chest radiograph documented. The median age at first episode was 8.2 months; 7.8 months for cases (mean, 11.2 months) and 9.6 months for controls (mean, 15.8; P = 0.11).

A linear effect in the number of pneumonia episodes was seen (Table 2): bronchiectasis was 3 times more likely in children with 4–5 episodes of pneumonia and 21 times more likely if they had 6 or more episodes (P for trend < 0.01). In addition, more severe pneumonia episodes requiring oxygen, with longer hospital stay, were strongly associated with bronchiectasis. The median number of days in the hospital [total number of days for pneumonia episode(s)] for cases was 43 days (range, 2–608 days; mean, 64 days) and 11 days for controls (range, 3–109 days; mean, 17 days; P < 0.01). Within the case series, there was one outlier with a total of 608 days in the hospital as a result of 9 episodes. Excluding this child who had long episodes of hospitalization involving social issues as well as medical complications, the median number of days for cases was 43 days (mean, 55 days). The median number of days in the hospital per episode for cases was 9.5 days (range, 2–70 days; mean, 14.0 days) and 6.0 days for controls (range, 3–28 days; mean, 17.6 days; P < 0.01). Children who developed bronchiectasis were twice as likely to have required oxygen during 1 episode. There was also a strong association with the presence of atelectasis (OR 11.9; 95% CI 3.1–45.9). Equal numbers of cases and controls had pleural effusion (1 case and 1 control), and 5 cases had signs of hyperinflation on their chest radiograph. Age at first episode was not associated with bronchiectasis later in life.

T2-4
TABLE 2:
History of Hospitalized Pneumonia Episodes Experienced by Cases and Controls (Adjusted Odds Ratios and 95% Confidence Intervals)

The total number of pneumonia episodes in the first year of life did not appear to be associated with the development of bronchiectasis later in life. However, more severe episodes early in life with longer hospital stay, requiring oxygen, were strongly associated with bronchiectasis. A linear effect was seen: bronchiectasis was 14 times more likely (OR 14.5; 95% CI 1.1–29.6) in children who required oxygen more than once in the first year of life (P for trend, <0.01) and 5.7 (95% CI 2.2–96.1) times more likely if they were hospitalized for 5 weeks or longer (P for trend, 0.17). With the very first episode, the average number of days in hospital for cases was twice that of controls, with a median of 9 days for cases and 6 days for controls (P < 0.01); 31 cases and 14 controls required oxygen during this episode (OR 7.8; 95% CI 2.9–21.4) (reference group was participants who had at least 1 episode of pneumonia but did not require oxygen). Although there was an association between the site of the first episode of lobar pneumonia and the site of bronchiectasis (OR 1.8; 95% CI 0.8–4.1), this was not statistically significant. Other radiographic features showed that 12 cases and 1 control had atelectasis, 3 cases had hyperinflation, 1 case had a cystic lesion and 1 case had both hyperinflation and atelectasis, as reported on the chest radiograph.

There was no difference in the history of hospitalized pneumonia episodes between controls with different diagnosis (P = 0.33).

Selected Medical History.

Immunization coverage was very good for this group of children; all cases and controls had >90% of the vaccines considered adequate for age; 84.2% of cases and 80.6% of controls had complete vaccine coverage for age (P = 0.54). In particular, cases and controls had 100% coverage for measles and pertussis vaccine. Few statistically significant differences in histories of medical conditions existed between cases and controls. A history of having at least 1 episode of hospitalized pneumonia and being in hospital to treat malnutrition occurred more frequently in case subjects. Chronic suppurative otitis media, developmental delay and urinary tract infection were also more frequent in cases but were consistent with chance findings. Measles, tuberculosis and whooping cough (no cases and 2 controls) were potentially of interest; however, both the numbers were small or were more frequent in controls, and results were consistent with chance findings. With regard to bronchiolitis recorded anywhere in the medical charts, one-half the number of cases had been hospitalized with bronchiolitis compared with one-fourth of controls (OR 1.8; 95% CI 0.9–3.8). Age at the first episode of bronchiolitis was similar in cases and controls (average, 6.1 months for cases and 5.7 months for controls; median, 4.8 months for both groups).

The most common comorbidity before diagnoses was malnutrition (53), followed by otitis media (acute or chronic; 36 diagnoses), anemia (26) and scabies (9). Cases and controls did not differ significantly. Cases and controls were very similar with regard to chronic suppurative otitis media diagnosis: 61% were diagnosed in the first 12 months; 33% had unilateral perforation (28% cases and 36% controls); and 65% had bilateral perforation (67% cases and 64% controls).

Selected Pregnancy and Neonatal Related Factors.

Mean weight at birth for female cases was 2.7 kg compared with 3.0 kg for controls (adjusted P = 0.12); male cases on average weighed 2.8 kg compared with 3.2 kg for male controls (adjusted P = 0.04). The data show a pattern of increasing risk of bronchiectasis with decrease in weight; however, when adjusting for confounding factors, including gestational age, this trend was not statistically significant (P values for trend were 0.10 for girls and 0.28 for boys). Cases and controls did not differ in body length at birth. Children who were small for gestational age and those who required ventilation as a neonate were likely (but nonsignificantly) to develop bronchiectasis [OR 2.1 (95% CI 0.7–6.1) and OR 2.3 (95% CI 0.7–7.4), respectively]. A linear effect was seen with gestational age. Bronchiectasis was 5 times more likely (OR 5.2; 95% CI 0.8–34.0) in children born with <31 weeks gestation (P value for trend = 0.03). History of breast-feeding was a protective factor (OR 0.2; 95% CI 0.1–0.7).

DISCUSSION

Although many groups have reported an association between lower respiratory infections in childhood and later respiratory morbidity,12,15,16 this is the first case-control study of childhood pneumonia and radiologically proved bronchiectasis. The most striking observation in this study is the strong and significant association between hospitalized pneumonia and bronchiectasis, in particular recurrent (>1) hospitalized pneumonia and more severe pneumonia episodes with longer hospital stay and requirement for oxygen. In addition, the overall number of pneumonia episodes rather then the site of pneumonia was associated with bronchiectasis. Children with bronchiectasis were also nearly 3 times more likely to have been diagnosed with malnutrition. Less precise estimates raised the possibility of an effect of perinatal factors. Being born premature and being small for gestational age was more common among cases. Among the few inverse associations observed, breast-feeding was the only protective effect.

Childhood respiratory infection, a leading cause of mortality and morbidity in developing countries, is of lesser concern in affluent countries.17 However, within populations of some affluent countries, appreciable social inequities and difficulties within health care services are present,18 and the age distribution19 and some childhood illness patterns reflect those of third world countries. In Australia, remote Indigenous Australians are the most severely disadvantaged group,18 and within their communities pneumonia and gastroenteritis are the most common causes of hospitalization in childhood.20 Our unique clinical setting of medical resources at the level of affluent countries, combined with “third world” disease profile of childhood pneumonia in the absence of malaria and human immunodeficiency virus infection, allowed us to perform the first case-control study of childhood bronchiectasis and pneumonia. This study adds stronger evidence than did previous reports on the association between childhood pneumonia and bronchiectasis.12

Bronchiectasis can result from a single severe pneumonia episode,21 and it is biologically plausible that recurrent pneumonic events increase the risk of developing bronchiectasis, as found in this study. Chronic lung damage may have occurred early and predisposed these children to recurrent pneumonia before their bronchiectasis was diagnosed. However, the odds ratio for development of bronchiectasis was similar for both lobar pneumonia and bronchopneumonia. In addition, there was no relationship between lobar pneumonia site and bronchiectasis site. Thus it is more likely that the same etiologic factors that predispose the child to both lobar and bronchopneumonia also predispose the child to bronchiectasis. Other speculative possible reasons for the absence of relationship between lobar pneumonia and development of bronchiectasis as well as similar odds ratio found for lobar pneumonia and bronchopneumonia are: spillover effect of inflammation from infection in an adjacent lobe leading to bronchiectasis22; effect of nonhospitalized pneumonia episodes treated in the community (thus episodes that most likely to lead to bronchiectasis were not identified); and selective interaction between infectious agent, lobe and genetics (hence immunologic and inflammatory response) of child.

It is increasingly apparent that some respiratory diseases have their its roots in early childhood and for disease processes such as bronchiolitis, premorbid lung function abnormality has been demonstrated.23 Although this was not a large study and had limited power to detect small differences between cases and controls with certainty, significant findings were present; in particular, children previously hospitalized with pneumonia had a highly significant risk (15 times) of developing bronchiectasis. Although the results presented here could have been affected by chance because of small numbers, we believe that they are unlikely to have been biased. Information about exposure was abstracted from medical records compiled before diagnosis, making it unlikely that quality of notes differed for cases and controls. Further, although the research nurse who traced and abstracted the medical notes was not “blind” to case-control status, great care was taken to ensure that such knowledge did not result in biased data collection, the tightly structured abstraction forms and coding procedures being designed specifically to avoid this type of error.

The investigation described here was specifically designed to examine the relation between hospitalized pneumonia (or recurrent pneumonia) and subsequent development of bronchiectasis in childhood, its success depending on the ability to link past history of pneumonia with correct diagnosis of bronchiectasis. Because the ASH is the only hospital in the region, we can be sure that most children, specifically Indigenous children residing in Central Australia, if diagnosed with moderate to more severe pneumonia, would have been referred to this hospital for treatment. Therefore, very few episodes of more severe pneumonia would have been missed and in particular those with recurrent pneumonia would have been hospitalized at the ASH for at least some of the episodes. As access to health services (here measured as resident community, Alice Springs versus other places) could have influenced diagnosis as well as mode of treatment of pneumonia and other medical conditions (affecting time of starting of antibiotics and delays in appropriate case), all odds ratios were adjusted for resident community. Indeed there was a tripling of risk for children living outside Alice Springs. This is potentially an important public health issue that may have potential for intervention and should be examined further. The possible speculative reasons for the distance effect are many and include differences in environmental, genetic, socioeconomic and medical services. The effect of medical services may be a 2-way interaction, because a higher level of medical service to the remote communities may increase case identification and appropriate referral whereas lesser services may lead to inadequate treatment and increased severity and frequency of pneumonia.

The identification of the appropriate study base from which to select controls is an important issue in the design of case control studies. In a hospital-based study such as this one, cases are all patients diagnosed with the study disease at that hospital, whereas controls are all subjects who would be diagnosed at that hospital had they developed the study disease.24 Given that the ASH is the only referral hospital in the region, we can be sure no selection bias was introduced by using this particular study design. The most serious problem with hospital controls is that choosing subjects with other diseases may jeopardize the assumption of representativeness of exposure (the distribution of the exposures under study in the controls is the same as that in the random sample from the base that produced the cases).25 There should be no relation between exposures under study and the diagnoses used to determine inclusion of controls. Even though the potential for this type of bias exists, we took steps to reduce it by excluding controls who, during relevant hospital admission, were treated for pneumonia, measles, whooping cough and tuberculosis. In addition, we included as controls patients with several diseases to minimize bias if anyone turned out to be related with exposures of interest. This study does, however, have certain weakness, although it is not clear how our findings would have been influenced by them. Although we know that the controls were alive, did not have bronchiectasis and had no serious anomalies diagnosed before discharge, hospitals controls may be more “case-like” than population controls. Therefore, by using such a control group, effects may be underestimated, particularly for the perinatal factors, because they can predispose to adverse health outcomes.26 Case-selection bias because of missed bronchiectasis cases could be another possible source of bias in this study. However, we do not believe case-selection bias here was differential, because we have no reason to think that cases not included in the study are different in any systematic way to our case series. Because HRCT is the most sensitive and is the standard for identifying bronchiectatic airways,8 we can be sure misclassification of case status was not an issue in this study.

Poor housing, nonpneumonia lower respiratory infections and environmental exposures such as biomass combustion and cigarette smoking, which could contribute to the development of bronchiectasis in children, were not measured in this study. Given the limitations of retrospective data collection for these factors in our setting, we did not attempt to collect these potential influences. Although the circumstances in developing countries are different from ours in several ways, this study could have implications for the treatment of children with severe pneumonia in developing countries as well as disadvantaged children in affluent countries. In particular, if the treatment duration for acute lower respiratory infection does influence the likelihood of developing chronic lung disease, then the trend to short course antibiotics27 might require longer term follow-up. We conclude that, although we cannot fully answer the question of why bronchiectasis is much more common in Indigenous children in Central Australia, we have provided the first case-control study evidence of an association between bronchiectasis and the severe and recurrent lower respiratory infections in infancy and childhood. Therefore prevention and careful treatment and follow-up of children with the identified risk factors can have potential for prevention of bronchiectasis in children living in a setting similar to ours.

ACKNOWLEDGMENTS

We thank Valerie Logan for technical support.

REFERENCES

1. Nikolaizik WH, Warner JO. Aetiology of chronic suppurative lung disease. Arch Dis Child. 1994; 70: 141–142.
2. Karakoc GB, Yilmaz M, Altintas DU, Kendirli SG. Bronchiectasis: still a problem. Pediatr Pulmonol. 2001; 32: 175–178.
3. Chang AB, Grimwood K, Mulholland EK, Torzillo PJ. Bronchiectasis in indigenous children in remote Australian communities. Med J Aust. 2002; 177: 200–204.
4. Singleton R, Morris A, Redding G, et al. Bronchiectasis in Alaska Native children: causes and clinical courses. Pediatr. Pulmonol. 2000; 29: 182–187.
5. Smith AH, Pearce NE. Determinants of differences in mortality between New Zealand Maoris and non-Maoris aged 15–64. N Z Med J. 1984; 97: 101–108.
6. Australian Bureau of Statistics. Mortality of Aboriginal and Torres Straits Islander Australians. In: Australian Bureau of Statistics; 2000.
7. Dempsey KE, Condon JR. Mortality in the Northern Territory 1979–1997. Darwin, Australia: Darwin Territory Health Services; 1999.
8. Grenier P, Maurice F, Musset D, Menu Y, Nahum H. Bronchiectasis: assessment by thin-section CT. Radiology. 1986; 161: 95–99.
9. Currie DC, Cooke JC, Morgan AD, et al. Interpretation of bronchograms and chest radiographs in patients with chronic sputum production. Thorax. 1987; 42: 278–284.
10. Maxwell GM. Chronic chest disease in Australian Aboriginal children. Arch Dis Child. 1972; 47: 897–901.
11. Maxwell GM, Elliott RB, McCoy WT, Langsford WA. Respiratory infections in Australian Aboriginal children: a clinical and radiological study. Med J Aust. 1968; 2: 990–993.
12. Field CE. Bronchiectasis: third report on a follow-up study of medical and surgical cases from childhood. Arch Dis Child. 1969; 44: 551–561.
13. Chang AB, Masel JP, Boyce NC, Wheaton G, Torzillo PJ. Non-CF bronchiectasis: clinical and HRCT evaluation. Pediatr Pulmonol. 2003; 35: 477–483.
14. Chang AB, Boyce NC, Masters IB, Torzillo PJ, Masel JP. Bronchoscopic findings in children with non-cystic fibrosis chronic suppurative lung disease. Thorax. 2002; 57: 935–938.
15. Cooreman J, Redon S, Levallois M, Liard R, Perdrizet S. Respiratory history during infancy and childhood, and respiratory conditions in adulthood. Int J Epidemiol. 1990; 19: 621–627.
16. Johnston ID, Strachan DP, Anderson HR. Effect of pneumonia and whooping cough in childhood on adult lung function. N Engl J Med. 1998; 338: 581–567.
17. UNICEF. We the children: Meeting the promises of the World Summit for Children. Available at: http://www.unicef.org/specialsession/about/sg-report.htm; 2001.
18. Eades SJ. Reconciliation, social equity and indigenous health. Med J Aust. 2000; 172: 468–469.
19. Fisher D, Ruben A. Funding of Northern Territory public hospitals. Aust Health Rev. 2002; 25: 189–205.
20. d’Espaignet E, Kennedy K, Paterson B, MEasey M. From infancy to young adulthood: health status in the Northern Territory. Darwin, Australia: Darwin Territory Health Services; 1998.
21. Chang AB, Masel JP, Masters B. Post-infectious bronchiolitis obliterans: clinical, radiological and pulmonary function sequelae. Pediatr Radiol. 1998; 28: 23–29.
22. Gutierrez JP, Grimwood K, Armstrong DS, et al. Interlobar differences in bronchoalveolar lavage fluid from children with cystic fibrosis. Eur Respir J. 2001; 17: 281–286.
23. Turner SW, Young S, Landau LI, Le Souef PN. Reduced lung function both before bronchiolitis and at 11 years. Arch Dis Child. 2002; 87: 417–420.
24. Wacholder S, McLaughlin JK, Silverman DT, Mandel JS. Selection of controls in case-control studies, I: principles. Am J Epidemiol. 1992; 135: 1019–1028.
25. Wacholder S, Silverman DT, McLaughlin JK, Mandel JS. Selection of controls in case-control studies, II: types of controls. Am J Epidemiol. 1992; 135: 1029–1041.
26. West D, Schuman K, Lyon J, Robison L, Allred R. Differences in risk estimations from a hospital and a population-based case-control study. Int J Epidemiol. 1984; 13: 235–239.
27. Group M-PMASCTps. Clinical efficacy of 3 days versus 5 days of oral amoxicillin for treatment of childhood pneumonia: a multicentre double-blind trial. Lancet. 2002; 360: 835–841.
Keywords:

bronchiectasis; children; epidemiology; risk factors

© 2004 Lippincott Williams & Wilkins, Inc.