International Journal of Infectious Diseases (2007) 11, 536—543

http://intl.elsevierhealth.com/journals/ijid

The epidemiology of hospitalized pneumonia in rural : the potential of surveillance data in setting public health priorities§

Jeffrey A. Tornheim a, Ayub S. Manya b, Norbert Oyando c, Stewart Kabaka c, Robert F. Breiman a, Daniel R. Feikin a,* a International Emerging Infections Program, Centers for Disease Control and Prevention, Unit 64112, APO, AE 09831, Kenya b Field Epidemiology and Laboratory Training Program, Centers for Disease Control and Prevention, Kenya c Bondo District Ministry of Health, Kenya Ministry of Health, Kenya

Received 20 December 2006; accepted 17 March 2007 Corresponding Editor: William Cameron, Ottawa, Canada

KEYWORDS Summary Pneumonia; Introduction: Surveillance data from inpatient health facilities can be useful for prioritization of Epidemiology; public health initiatives, but often are not collected or analyzed in developing countries. We Surveillance evaluated data on hospitalized patients diagnosed with pneumonia in rural western Kenya to characterize pneumonia epidemiology and mortality. Methods: Data were obtained from admission registers of all inpatient facilities from 2001 to 2003 in Bondo District (estimated 2003 population: 255901), which is holoendemic for malaria and has high HIV rates. Inpatients with diagnoses compatible with acute pneumonia were included, and census data (1999) were used to calculate incidence rates by age, sex, season, and residence. Results: From 2001 to 2003, a total of 2466 patients diagnosed with pneumonia were hospi- talized with 282 deaths (11.4%). Incidence peaked at 698 per 100 000 person-years among children <5 years of age. A second peak occurred among 20—29 year-olds at 356 per 100 000 person-years;ratesweretwiceashighinwomenasmeninthisagegroup(p < 0.001). The incidence in persons >65 years was 121 per 100 000 person-years. Pneumonia incidence peaked during the twice-yearly high malaria seasons, 1—2 months after peak rainfall. Rates of pneumonia decreased with increasing distance of residence from the district hospital ( p < 0.0001). Discussion: In Bondo District, the pneumonia burden is greatest among young children and middle-aged adults, the latter peak reflecting the area’s HIV epidemic. Access to care likely

§ Disclaimer: This information is distributed solely for the purpose of pre-dissemination peer review under applicable information quality guidelines. It has not been formally disseminated by the Centers for Disease Control and Prevention. It does not represent and should not be construed to represent any agency determination or policy. * Corresponding author. Tel.: +254 7222 00075. E-mail address: [email protected] (D.R. Feikin).

1201-9712/$32.00. Published by Elsevier Ltd on behalf of International Society for Infectious Diseases. doi:10.1016/j.ijid.2007.03.006 The epidemiology of hospitalized pneumonia in rural Kenya 537

influenced hospital utilization and thus pneumonia rates, particularly among the elderly. Our findings show that hospital-based data can provide useful information for public health priority setting, such as the introduction of new pneumonia vaccines for children and accelerating the introduction of antiretroviral medications. Published by Elsevier Ltd on behalf of International Society for Infectious Diseases.

Introduction Data collection

National surveillance for notifiable infectious diseases has We reviewed records on hospitalized acute pneumonia at all been a cornerstone of the public health system in many eight medical facilities that admitted patients in Bondo developed countries,1 but such surveillance systems have District between 2001 and 2003. Admission registers were been harder to establish and maintain in developing coun- reviewed for all patients with diagnoses consistent with tries. Establishment of effective surveillance is precluded in acute pneumonia including the following terms: acute many countries by scarcity of resources and inadequate respiratory infection, pneumonia, aspiration pneumonia, infrastructure, as well as by an imperative to focus on other bronchopneumonia, lobar pneumonia, chest infection, acute basic services, such as immunization. Yet it is precisely in chest infection, respiratory tract infection, and lower developing countries, where infectious diseases still respiratory tract infection. Diagnoses using the following account for the majority of morbidity and mortality,2 that terms were not considered to be acute pneumonia: cough, such surveillance systems can be most valuable today respiratory distress, upper respiratory tract infection, bron- in recognizing outbreaks, tracking emerging diseases, and chitis, tuberculosis, and chronic respiratory illness. Admis- setting public health priorities for important infectious sion diagnoses were made most often by clinical officers, diseases. clinicians with 3 years of post-secondary school training who Pneumonia is a serious public health problem in serve as the sole clinicians in most medical facilities in Kenya. Kenya,3,4 and its prevention and control could benefit In some facilities, nurses made the admitting diagnoses. from a functional national surveillance system. Acute Clinical officers and nurses were trained according to Kenya pneumonia is responsible for 19—23% of deaths among Ministry of Health clinical guidelines.15 In the guidelines, children under 5 years old in developing countries.5,6 As pneumonia is defined according to Integrated Management HIV predisposes to pneumonia, the HIV epidemic in coun- of Childhood Illness (IMCI) criteria in children aged 2 months tries like Kenya is likely to cause an excess of acute to 4 years of age.15,16 In adults, pneumonia is defined as a pneumonia.7 The epidemiology of pneumonia in Kenya constellation of clinical criteria including the following: has not been thoroughly characterized yet, so incidence, breathlessness, cough with or without sputum which may mortality, and risk factors are not well defined. National be rust-colored, fever, pleuritic chest pain, bronchial breath- surveillance data on pneumonia can be useful in the ing, reduced chest movements, reduced breath sounds, implementation of strategies that can prevent pneumonia, tachypnea, and crackles.15 During the study period, Bondo such as new vaccines,8,9 zinc replacement,10 and hand- District had 2—3 doctors, who were all based at the district washing programs.11 hospital. The following variables were available from the In order to better define the epidemiology of hospitalized admission registers: date of admission, age, sex, residence pneumonia in western Kenya and to show the utility of and co-diagnoses on admission. We attempted to find the including such data in a national surveillance system, we medical record for each pneumonia admission in the admis- reviewed pneumonia inpatient records from all admitting sion register to determine outcome and length of stay. medical facilities in Bondo District, Kenya from January 1, If a patient’s specific age was not listed, but it was 2001 to December 31, 2003. recorded that they were ‘adult’ or the wife of another person, this data was collected. Patients with no recorded sex were assigned the gender of their first names. Because Methods ‘HIV’ or ‘AIDS’ was rarely written in the record, the diagnoses of AIDS related complex (ARC) and immunosuppression were Site considered to indicate HIV disease. Outcomes were recorded as discharged, died, referred, or absconded. Patients with no Bondo District is a rural district in in western explicit indication of death noted in the medical record were Kenya bordering . The projected 2003 popula- assumed to have been discharged alive. The death register at tion of the district was 255 901, with 20.4% under 5 years of Bondo District Hospital was later reviewed for patient iden- age. The majority of residents is ethnically Luo and earns a tification numbers matching those in the database. Patients living through agriculture or fishing. It is a poor district, considered alive from the medical record who were listed in where average household wealth is estimated at $600— the death register with dates consistent with the admission 700, the infant mortality rate is 120 per 1000, and life for pneumonia were reclassified as dead in a second analysis expectancy at birth is approximately 40 years.12—14 Rates specific to Bondo District Hospital. Rainfall data were of HIV disease in Nyanza Province are the highest in Kenya, obtained from the Kenya Meteorology Department, which reported as 15.1% in the 2003 demographic and health were collected at Airport approximately 50 kilo- survey.3 meters from Bondo town. 538 J.A. Tornheim et al.

Analysis as above for age-specific CFRs. We adjusted the overall CFR using the proportion of patients at Bondo District Hospital Data were entered and analyzed in Epi Info (version 3.3.2, who were registered as discharged alive in their medical CDC, Atlanta, GA, USA). Further analyses of rates were records, but registered as deceased in the hospital death performed in PEPI (version 4.0, Sagebrush Press), and dis- register. The death rate from hospitalized pneumonia was tance analysis in SAS for windows (version 8.2, SAS Institute, calculated as an incidence rate of death, and a Chi-square Cary, NC, USA). Patients residing outside of Bondo District test was employed to determine differences between age were excluded from analyses (5.5% of patients). To calculate groups and sexes. incidence rates, we used annual population projections for 2001—2003 by age, sex, and location, from the Kenyan Results Central Bureau of Statistics, which used the 1999 census results and assumed constant rates of birth, death, and During the 3-year study period, 2466 Bondo District residents migration. Confidence intervals around incidence rates were were hospitalized with pneumonia. The number of pneumo- calculated using the exact binomial method. Fisher’s exact nia patients was relatively stable between years, with 827 in two-tailed method was used to calculate the significance of 2001, 782 in 2002, and 857 in 2003. Overall, 43.7% of pneu- differences between incidence rates. To calculate incidence monia admissions were in children under 5 years of age, and rates by age, the patients in our database with missing age 58.3% were female. Of all admissions, 1604 (65%) were were redistributed between age groups in proportion to the admitted to Bondo District Hospital. distribution of patients with known age, and those listed as Pneumonia admissions accounted for 12.8% (95% confi- ‘adult’ or ‘spouse’ were distributed only among patients 10 dence intervals (CI) 12.3—13.3%) of all admissions in the years and older. To calculate incidence rates by location, district, ranging by facility from 7.8% to 19.3%. The percen- patients with missing residence were redistributed in propor- tage of admitted children under 5 years old who had pneu- tion to the distribution of patients with known residence; this monia was 21.6% (95% CI 21.1—22.2%) compared with 9.8% was done separately for each hospital. (95% CI 9.4—10.2%) of patients 5 years and older (13.5% of Distances to the facilities were calculated by taking the non-maternity admissions). The median length of hospital straight-line distance from the centroid point of each sub- stay among the cases with this information available was 3 location in the district to the facility as defined by GIS using days (range 0—242 days), and the mode was 2 days. ArcView software (version 3.2). The association between rates by sub-location and distance was evaluated using Pois- Incidence rates son regression with rates as the dependent variable, using the events/trials syntax, and distance to facility in kilometers as The overall annual incidence rate of hospitalized pneumonia the independent variable (PROC GENMOD). Overdispersion in Bondo District was 324 per 100 000 persons (95% CI 311— was corrected using the ‘‘pscale’’ command. 337), but rates varied significantly by age. The annual inci- Overall case—fatality ratios (CFR) were calculated and dence was 698 (95% CI 656—739) and 229 (95% CI 217—241) deceased patients with missing exact ages were redistributed per 100 000 persons for those <5 years old and 5 years old,

Figure 1 Average annual incidence and case—fatality ratio (CFR) of hospitalized pneumonia in Bondo District by age group, 2001— 2003. The epidemiology of hospitalized pneumonia in rural Kenya 539

Table 1 Incidence of hospitalized pneumonia per 100 000 person-years by age group and sex, 2001—2003

Age (years) Incidence in females (95% CI) Incidence in males (95% CI) Rate ratios (female/male, 95% CI) 0—4 693 (637—749) 826 (758—893) 0.84 (0.75—0.95) 5—9 85 (61—109) 88 (61—114) 0.98 (0.64—1.52) 10—14 87 (56—119) 56 (30—83) 1.51 (0.81—2.88) 15—19 262 (204—320) 95 (58—132) 2.81 (1.77—4.60) 20—29 442 (396—488) 188 (154—222) 2.35 (1.90—2.92) 30—39 337 (288—385) 245 (203—288) 1.37 (1.09—1.73) 40—49 321 (257—384) 187 (141—233) 1.71 (1.24—2.39) 50—64 211 (148—274) 218 (158—279) 0.96 (0.63—1.48) 65+ 119 (58—180) 104 (46—163) 1.18 (0.52—2.76) All ages 356 (324—388) 287 (257—318) 1.24 (1.14—1.34)

respectively (rate ratio (RR) = 3.04, 95% CI 2.81—3.30, 707—820) per 100 000 person-years and lowest in Mageta Figure 1). This is equivalent to one out of every 143 children Island in Lake Victoria with a rate of 89 (95% CI 38—139) in the district being hospitalized for pneumonia each year, per 100 000 person-years. The incidence rate of pneumonia and one out of every 437 adults. The incidence of hospitalized decreased significantly with distance from Bondo District pneumonia was highest in infants less than 1 year old at 1370 Hospital (Figure 2, p < 0.0001). For each one kilometer of (95% CI 1259—1481) per 100 000 person-years. Incidence increased distance, the incidence rate decreased by 5% (95% decreased substantially after the age of 5 years, increasing CI 4—7%). There was also a marginally significant decrease in again in older adolescents aged 15—19 years to reach a incidence rate with increased distance to the nearest inpa- second peak of 356 (95% CI 325—387) per 100 000 person- tient health facility (Figure 2, p < 0.05). years for people 20—29 years old. Incidence for persons over Pneumonia incidence varied by season. Monthly incidence 65 years was lower than for middle-aged adults at 121 per among children under 5 years old ranged from 253 per 100 000 person-years (95% CI 76—165). 100 000 in December 2003 to 1175 per 100 000 in July 2003 There were also significant differences in pneumonia (Figure 3). Monthly incidence among people 5 years and older incidence between the sexes. In children less than 5 years ranged from 95 per 100 000 in May 2003 to 356 per 100 000 in of age, the incidence was lower among females (RR = 0.84, July 2002. Incidence in both age groups had peaks each year 95% CI 0.75—0.95, Table 1). Conversely, among adults aged in January—February and June—August, 2—3 months after 15—49 years, the incidence was significantly higher among each rainy season. females (RR = 1.92, 95% CI 1.68—2.20). Patients with co-diagnoses of malaria, tuberculosis, and Pneumonia incidence by location of residence was found HIV represented 60.7%, 14.9%, and 1.5% of hospitalized to be highest in Bondo Township with a rate of 763 (95% CI pneumonia cases, respectively (Figure 4). Pneumonia with

Figure 2 Incidence of hospitalized pneumonia by proximity to inpatient facilities in Bondo District, 2001—2003. Straight lines are linear regression lines. Incidence rate decreased significantly with increasing distance from sublocation of residence to facility for both Bondo District Hospital (Figure 2a, p < 0.0001) and nearest inpatient facility (Figure 2b, p < 0.05) using Poisson regression analysis (see methods). 540 J.A. Tornheim et al.

Figure 3 Monthly rainfall and incidence of hospitalized pneumonia in Bondo District. malaria was most common in children under the age of 5 monia with malaria). Among persons 20—49 years of age, years, among whom 73.3% had a malaria co-diagnosis. The compared with pneumonia alone, the risk of fatality increased percentage of pneumonia inpatients with malaria co-diag- with a co-diagnosis of either tuberculosis (15.5%, RR = 1.59, noses was highest in the same months that the incidence 95% CI 1.14—2.21) or HIV (22.2%, RR = 2.28, 95% CI 1.19—4.37). rates of pneumonia peaked, which are also the months that The total death rate from hospitalized pneumonia in malaria peaks in this area.17 Tuberculosis and HIV co-diag- Bondo District was 32 deaths per 100 000 person-years noses were concentrated in the 20—49 year-old patients. (95% CI 28.2—36.4). The death rates for children <5 years The overall case—fatality ratio (CFR) was 9.9% (95% CI 8.8— and patients 5 years were 65 and 24 per 100 000 person- 11.2%), which increased to 11.4% (95% CI 10.2—12.7%) when years, respectively (RR = 2.76, 95% CI 2.11—3.60). There was using the death register. (The remainder of this report uses no difference in death rates between sexes. CFRs derived only from medical records.) The CFR was highest at the extremes of life — 11.0% for those under five and 20.0% Discussion for adults 65 years and older — and lowest from 10 to 14 years of age (4.8%, Figure 1). CFRs did not vary with malaria This study reveals important epidemiologic findings about co-diagnosis (9.8% for pneumonia alone versus 8.6% for pneu- pneumonia in western Kenya. Hospitalized pneumonia was

Figure 4 Percentage of hospitalized pneumonia patients with other diseases by age group, Bondo District 2001—2003. HIV included diagnoses of HIV, AIDS, AIDS-related complex, and immunosuppression in the admission register. The epidemiology of hospitalized pneumonia in rural Kenya 541 common and often resulted in death. As shown in studies inpatients) and toward HIV care (such as widespread intro- elsewhere, the rate of pneumonia was highest in young duction of highly active antiretroviral treatment) would have children, especially among infants who had an annual rate beneficial effects on hospitalized pneumonia incidence. of 1370 cases per 100 000 children.4 Although lower than the Moreover, such data as produced in this study could contri- rate of hospitalized pneumonia among infants found in bute towards monitoring the impact on morbidity by imple- The Gambia (5270 cases/100 000), the rate found among mentation of such strategies. infants was comparable to a study from Mali (998 cases/ Another interesting variant in the age distribution of 100 000).18,19 Mortality among children under 5 years pneumonia in Bondo District is the relatively low rate of admitted with pneumonia (11%) was similar to rates found pneumonia in the elderly. Studies in developed countries in several developing countries (12—15%),19—21 but high com- have found hospitalized pneumonia incidence rates to rise pared to hospitalized pediatric pneumonia mortality in sev- in the elderly with rates of over 1000 per 100 000 person- eral other developing countries, as well as in developed years, which is at least 10 times the incidence we found.27,30— countries.4,22,23 Death registry data were only found in the 33 The lower rates of pneumonia among the elderly might be district hospital. We therefore relied upon medical records to because older Kenyans are not as comfortable with Western calculate overall case—fatality ratios, but this is likely to medicine and often prefer to go to traditional healers or slightly underestimate true fatality rates, since at the district remain at home. Additionally, travel to facilities in most of hospital the case—fatality ratio increased from 9.9% to 11.4% Bondo District is difficult due to a lack of paved roads and when we included death registry data. Though treatment limited means of transport. Older people, especially if ill, regimens were not analyzed, the local standard of care for could be less likely to travel these distances. It is also possible inpatient pneumonia is penicillin plus gentamicin, which that the rate of pneumonia in the elderly was falsely low due should have been consistent for all pneumonia cases in the to our methodological assumption that patients without district. documented ages had the same age distribution as those The excess pneumonia found among boys has been pre- of known age. If adult patients without a documented age viously described elsewhere;4,18,24 it is unclear if this is due were more likely to be elderly, we might have under-repre- to variability in susceptibility between boys and girls or to sented cases among those over 65 years of age, resulting in an selection bias from preferential care-seeking. In either case, erroneously low rate in this age group. the data clearly show that children carry a considerable Our study had several important limitations. We relied on burden of pneumonia morbidity and mortality in Kenya, the admission registers to define cases of pneumonia. Admit- making them a group that would benefit significantly from ting diagnoses were usually made only on clinical grounds by existing preventive interventions. One such intervention is a the clinical officers or nurses working in the hospitals’ out- conjugate vaccine against Streptococcus pneumoniae, the patient clinics. Though the clinical officers were trained with leading cause of pneumonia deaths among children.25,26 It a standard curriculum for pneumonia diagnosis, some might has been shown to decrease overall mortality by 16% in have varied the criteria used to diagnose pneumonia.15 For Gambian children and is also effective in HIV-infected chil- instance, although trained in the World Health Organization’s dren.25,26 IMCI criteria, we could not verify that these criteria were The second relevant age-specific finding of our study was consistently applied to diagnose pneumonia in children dur- the high incidence of hospitalized pneumonia among young ing the period reviewed.16 Only the district hospital had an X- adults. After reaching a low among children aged 5—14 years, ray machine, which was rarely used for acute pneumonia the incidence rate rose again, peaking in 20—29 year-olds. cases. As discharge diagnoses were rarely found in the med- This pattern of pneumonia is very different from that ical records, it was difficult to assess how often the admitting observed in most developed and developing countries with- diagnosis changed after review during the hospital course. out a high prevalence of HIV, where the incidence of pneu- Moreover, the clinical staff changed over time, leading to monia remains lower throughout most of young and middle possible lack of diagnostic consistency even within each adulthood before rising among the elderly.27—30 Moreover, we hospital. The difficulty in distinguishing malaria and pneu- found that women had rates of pneumonia approximately monia on clinical grounds alone is a well-known phenom- twice those of men in this age group. The most likely expla- enon,34,35 and such lack of specificity in diagnosis of nation for the high rates in this age group and the variation pneumonia could explain part of the observed overlap of between the sexes is the impact of the HIV epidemic in malaria and pneumonia seasonality in Bondo District. Nyanza Province, where in 2003 the rates of HIV (15.1%) Another limitation was that in calculating incidence rates were over double the national average (6.7%), and within we only used hospital admissions in Bondo District. Though Nyanza Province the rates were higher in women (18.3%) than we excluded non-Bondo residents from our rate calculations men (11.6%).3In both developed and developing countries, (5.5% of admissions), we did not include pneumonia admis- pneumonia is 25 times more common in HIV-infected persons sions for Bondo residents who were hospitalized outside of than non-infected persons, and HIV infection occurs in 28— the district. This would have underestimated the true inci- 95% of African patients with pneumococcal bacteremia.7 dence rate of pneumonia. Reviews of admission logbooks Despite the apparent association between pneumonia admis- from the district hospital in neighboring District and sions and HIV infection among middle-aged adults in this part the Provincial Hospital in Kisumu, however, showed that of Kenya, the HIV status of most pneumonia patients remains patients admitted with pneumonia rarely had Bondo District unknown. Given the relationship between pneumonia and residences. HIVand the epidemiology of pneumonia in Bondo District, it is Another issue of concern is that people in Bondo District reasonable to conclude that strategies aimed toward HIV might not access health facilities when they have pneumonia. diagnosis (such as diagnostic counseling and testing among The district is largely rural with few paved roads. The 542 J.A. Tornheim et al. decrease in incidence rate with distance to the district 8. Mulholland K, Hilton S, Adegbola R, Usen S, Oparaugo A, Omo- hospital suggests that poor access to care may indeed have sigho C, et al. Randomised trial of Haemophilus influenzae type- lowered overall incidence rates of hospitalized pneumonia in b tetanus protein conjugate vaccine [corrected] for prevention the district. Distance to hospital has been shown before to of pneumonia and meningitis in Gambian infants. Lancet 1997;349:1191—7. influence rates of pneumonia among Gambian children.18 In 9. Black SB, Shinefield HR, Ling S, Hansen J, Fireman B, Spring D, conducting facility-based surveillance in African settings, the et al. Effectiveness of heptavalent pneumococcal conjugate effects of distance and access to care need to be factored in vaccine in children younger than five years of age for prevention when interpreting incidence rates. of pneumonia. Pediatr Infect Dis J 2002;21:810—5. Lastly, Bondo District might not be representative for all of 10. Brooks WA, Santosham M, Naheed A, Goswami D, Wahed MA, Kenya in terms of pneumonia incidence and epidemiology. As Diener-West M, et al. Effect of weekly zinc supplements on discussed, the rates of HIV disease are higher in this part of incidence of pneumonia and diarrhoea in children younger than Kenya than elsewhere in the country, which should affect the 2 years in an urban, low-income population in Bangladesh: rate and distribution of pneumonia admissions.3 Malaria rates randomized controlled trial. Lancet 2005;366:999—1004. are also very high in Bondo, which likely decreased specificity 11. Luby SP, Agboatwalla M, Feikin DR, Painter J, Billhimer W, Altaf A, et al. Effect of handwashing on child health: a randomized of a pneumonia diagnosis compared to areas with lower 34,35 controlled trial. Lancet 2005;366:225—33. prevalence of malaria. 12. Meltzer MI, Terlouw DJ, Kolczak MS, Odhacha A, ter Kuile FO, These limitations — the lack of specificity of the pneumo- Vulule JM, et al. The household-level economics of using per- nia diagnosis, hospitalization outside the district, and poor methrin-treated bed nets to prevent malaria in children less access to care — along with the variability and difficulty in than five years of age. Am J Trop Med Hyg 2003;68(4 Suppl): diagnosing pneumonia in most developing country settings, 149—60. make comparison of the rates of pneumonia we found with 13. Ministry of Health, Government of Kenya. Bondo District 2002 –— rates from other studies an imprecise exercise. A review of year health report. Republic of Kenya: Ministry of Health; 2003. childhood pneumonia found that incidence rates obtained 14. Adazu K, Lindblade KA, Rosen DH, Odhiambo F, Ofware P, Kwach from active surveillance at the home were up to 10 times J, et al. Health and demographic surveillance in rural western Kenya: a platform for evaluating interventions to reduce mor- higher than incidence of severe pneumonia likely to be 36 bidity and mortality from infectious diseases. Am J Trop Med Hyg hospitalized. For these reasons, our study likely under- 2005;73:1151—8. estimates the true figures for severe pneumonia in Bondo 15. Ministry of Health, Government of Kenya. Clinical guidelines for District. Nevertheless, the data from this study are internally diagnosis and treatment of common hospital conditions in consistent and therefore are useful in defining epidemiologic Kenya. , Kenya: The Regal Press Kenya Ltd; 1994. patterns and in monitoring trends in pneumonia within Bondo 16. World Health Organization, Department of Child and Adolescent District. These data can be used by the Kenyan Ministry of Health and Development, and UNICEF. Handbook: IMCI inte- Health in setting public health priorities, such as the intro- grated management of childhood illness. Geneva, Switzerland: duction of new vaccines for childhood pneumonia, and WHO; 2005. increased access to antiretroviral therapy for HIV-infected 17. Phillips-Howard PA, Wannemuehler KA, ter Kuile FO, Hawley WA, Kolczak MS, Odhacha A, et al. Diagnostic and prescribing prac- persons. Moreover, ongoing surveillance can monitor the tices in peripheral health facilities in rural western Kenya. Am J impact of these interventions over time, and a functioning Trop Med Hyg 2003;68(4 Suppl):44—9. national surveillance system for pneumonia could help detect 18. Weber MW, Milligan P,Sanneh M, Awemoyi A, Dakour R, Schneider and define emerging infectious disease epidemics in Kenya as G, et al. An epidemiological study of RSV infection in the they arise. Gambia. Bull World Health Organ 2002;80:562—8. Conflict of interest: No conflict of interest to declare. 19. Campbell JD, Sow SO, Levine MM, Kotloff KL. The causes of hospital admission and death among children in Bamako, Mali. J Trop Pediatr 2004;50:158—63. References 20. Djelantik IG, Gessner BD, Sutanto A, Steinhoff M, Linehan M, Moulton LH, et al. Case fatality proportions and predictive 1. Centers for Disease Control and Prevention. Guidelines for eval- factors for mortality among children hospitalized with severe uating surveillance systems, MMWR Morb Mortal Wkly Rep pneumonia in a rural developing country setting. J Trop Pediatr 1988;37(S-5):1—18. 2003;49:327—32. 2. World Health Organization. The world health report 2003: 21. Nathoo KJ, Nkrumah FK, Ndlovu D, Nhembe M, Pirie DJ, Kowo H. shaping the future. Geneva, Switzerland: WHO; 2003. Acute lower respiratory tract infection in hospitalized children in 3. Central Bureau of Statistics (CBS) Kenya, Ministry of Health Zimbabwe. Ann Trop Paediatr 1993;13:253—61. (MOH) Kenya, and ORC Macro. Kenya demographic and health 22. Yin CC, Huah LW, Lin JT,Goh A, Ling H, Moh CO. Lower respiratory survey 2003: key findings. Calverton, Maryland, USA: CBS, MOH, tract infection in hospitalized children. Respirology 2003;8:83— and ORC Macro; 2004. 9. 4. Selwyn BJ. Epidemiology of acute respiratory tract infection in 23. Mufson MA, Stanek RJ. Bacteremic pneumococcal pneumonia in young children: comparison of findings from several developing one American city: a 20-year longitudinal study, 1978—1997. Am countries. Rev Infect Dis 1990;12(8 Suppl):S870—88. J Med 1999;107:34S—43S. 5. Black RE, Morris SS, Bryce J. Where and why are 10 million 24. Henrickson KJ, Hoover S, Kehl KS, Hua W. National disease children dying every year? Lancet 2003;361:2226—34. burden of respiratory viruses detected in children by polymerase 6. Williams BG, Gouws E, Boschi-Pinto C, Bryce J, Dye C. Estimates chain reaction. Pediatr Infect Dis J 2004;23(1 Suppl):S11—8. of world-wide distribution of child deaths from acute respiratory 25. Cutts FT, Zaman SM, Enwere G, Jaffar S, Levine OS, Okoko JB, infections. Lancet Infect Dis 2002;2:25—32. et al. Efficacy of nine-valent pneumococcal conjugate vaccine 7. Feikin DR, Feldman C, Schuchat A, Janoff EN. Global strategies against pneumonia and invasive pneumococcal disease in The to prevent bacterial pneumonia in adults with HIV disease. Gambia: randomized, double-blind, placebo-controlled trial. Lancet Infect Dis 2004;4:445—55. Lancet 2005;365:1139—46. The epidemiology of hospitalized pneumonia in rural Kenya 543

26. Klugman KP, Madhi SA, Huebner RE, Kohberger R, Mbelle N, outcome in the United States. Am J Respir Crit Care Med Pierce N, et al. A trial of a 9-valent pneumococcal conjugate 2002;165: 766—72. vaccine in children with and those without HIV infection. N Engl J 32. National Center for Health Statistics. Health, United States, Med 2003;349:1341—8. 2004 with chartbook on trends in the health of Americans. Table 27. Marston BJ, Plouffe JF, File Jr TM, Hackman BA, Salstrom SJ, 95. Rates of discharge and days of care in non-federal short-stay Lipman HB, et al. Incidence of community-acquired pneumonia hospitals, according to sex, age, and selected first-listed diag- requiring hospitalization. Results of a population-based active noses: United States, selected years 1990—2002. Hyattsville, surveillance study in Ohio. Arch Intern Med 1997;157: Maryland: NCHS; 2004. 1709—18. 33. Jackson ML, Neuzil KM, Thompson WW, Shay DK, Yu O, Hanson 28. Monge V, San-Martin VM, Gonzalez A. The burden of community- CA, et al. The burden of community acquired pneumonia in acquired pneumonia in Spain. Eur J Public Health 2001;11: seniors: Results of a population-based study. Clin Infect Dis 362—4. 2004;39:1642—50. 29. Kanlayanaphotporn J, Brady MA, Chantate P, Chantra S, Siasir- 34. English M, Punt J, Mwangi I, McHugh K, Marsh K. Clinical overlap iwattana S, Dowell SF, et al. Pneumonia surveillance in Thailand: between malaria and severe pneumonia in Africa children in current practice and future needs. Southeast Asian J Trop Med hospital. Trans R Soc Trop Med Hyg 1996;90:658—62. Public Health 2004;35:711—6. 35. O’Dempsey TJ, McArdle TF, Laurence BE, Lamont AC, Todd JE, 30. Marrie TJ, Carriere KC, Jin Y, Johnson DH. Factors associated Greenwood BM. Overlap in the clinical features of pneumonia with death among adults <55 years of age hospitalized for and malaria in African children. Trans R Soc Trop Med Hyg community-acquired pneumonia. Clin Infect Dis 2003;36: 1993;87:662—5. 413—21. 36. Rudan I, Tomaskovic L, Boschi-Pinto C, Campbell H. Global 31. Kaplan V, Angus DC, Griffin MF, Clermont G, Scott Watson R, estimate of the incidence of clinical pneumonia among children Linde-Zwirble WT. Hospitalized community-acquired pneumonia under five years of age. Bull World Health Organ 2004;82:895— in the elderly: age- and sex-related patterns of care and 903.