Lower respiratory tract infection hospitalizations among American Indian/Alaska Native children and the general United States child population

Background The lower respiratory tract infection (LRTI)-associated hospitalization rate in American Indian and Alaska Native (AI/AN) children aged <5 years declined during 1998–2008, yet remained 1.6 times higher than the general US child population in 2006–2008. Purpose Describe the change in LRTI-associated hospitalization rates for AI/AN children and for the general US child population aged <5 years. Methods A retrospective analysis of hospitalizations with discharge ICD-9-CM codes for LRTI for AI/AN children and for the general US child population <5 years during 2009–2011 was conducted using Indian Health Service direct and contract care inpatient data and the Nationwide Inpatient Sample, respectively. We calculated hospitalization rates and made comparisons to previously published 1998–1999 rates prior to pneumococcal conjugate vaccine introduction. Results The average annual LRTI-associated hospitalization rate declined from 1998–1999 to 2009–2011 in AI/AN (35%, p<0.01) and the general US child population (19%, SE: 4.5%, p<0.01). The 2009–2011 AI/AN child average annual LRTI-associated hospitalization rate was 20.7 per 1,000, 1.5 times higher than the US child rate (13.7 95% CI: 12.6–14.8). The Alaska (38.9) and Southwest regions (27.3) had the highest rates. The disparity was greatest for infant (<1 year) pneumonia-associated and 2009–2010 H1N1 influenza-associated hospitalizations. Conclusions Although the LRTI-associated hospitalization rate declined, the 2009–2011 AI/AN child rate remained higher than the US child rate, especially in the Alaska and Southwest regions. The residual disparity is likely multi-factorial and partly related to household crowding, indoor smoke exposure, lack of piped water and poverty. Implementation of interventions proven to reduce LRTI is needed among AI/AN children.

L ower respiratory tract infection (LRTI) is the leading cause of hospitalization in US children aged B5 years (1). Pneumonia and bronchiolitis are the most common causes of LRTI-associated hospitalizations in US children aged B5 years (2). Among American Indian and Alaska Native (AI/AN) children, there have been persistently higher rates of LRTI-associated hospitalizations compared to the general US child population (2Á4). We last examined this disparity for 2006Á2008 and found that the LRTI-associated hospitalization rate for AI/AN children aged B5 years was approximately 1.6 times higher than the corresponding general US child rate, with greater disparities in AI/AN infants (B1 year) and in AI/AN children in Alaska and the Southwest regions (2). Since 2008, the H1N1 outbreak occurred (2009Á2010) and the 13-valent pneumococcal conjugate vaccine (PCV13) replaced 7-valent PCV in March 2010 (5). In the present study, we describe the change in LRTIassociated hospitalization rates in AI/AN children and in the general US child population aged B5 years between 1998Á1999 and 2009Á2011 to compare rates before and after the introduction of the PCV vaccine and the H1N1 influenza outbreak.

Methods
This manuscript was reviewed and approved for publication by the Alaska Native Tribal Health Consortium, Yukon Kuskokwim (YK) Health Corporation, Maniilaq Association and Indian Health Service (IHS).
We performed an analysis of 1998Á2011 hospital discharge data with an LRTI-associated diagnosis among 2 populations: AI/AN children aged B5 years who were hospitalized at an IHS-operated, tribally operated or contract hospital, and the general US child population aged B5 years. All hospital discharge records with an LRTI listed as any one of up to 15 diagnoses were selected from the respective data sets: (a) the IHS Direct and Contract Health Service Inpatient Dataset record for AI/ AN children (6,7) and (b) the Nationwide Inpatient Sample (NIS) for the general US child population (8). Hospital stays for newborns (births) were excluded, and the unit of analysis is a hospitalization. The International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) (9) codes consistent with LRTI were previously defined by Peck et al. as any one of the following diagnoses (ICD-9-CM codes in parentheses): pulmonary tuberculosis (011), pulmonary anthrax (022.1), pulmonary diseases caused by Mycobacterium (031.0), whooping cough (033), respiratory syncytial virus (RSV) (079.6), syphilis of lung (095.1), acute bronchitis (466.0), acute bronchiolitis (466.1), pneumonia (480Á486), influenza (487), influenza due to identified 2009 H1N1 influenza virus (488.1), empyema (510), pleurisy with effusion (511.1), abscess of lung and mediastinum (513), rheumatic pneumonia (517.1) and congenital pneumonia (770.0) (2,3). Acute bronchiolitis, pneumonia, RSV (466.11, 480.1 and 079.6) and influenza (487 and 488.1) were further examined. A bronchiolitis-associated hospitalization without pneumonia listed as a diagnosis and percentage of LRTI-associated hospitalizations due to pneumonia were also analyzed. The data were examined by time period, age group (infant and 1Á4 years), sex, IHS region (for AI/ANs), LRTI category (bronchiolitis, pneumonia, RSV and influenza) and hospital length of stay.
For AI/AN children aged B5 years, IHS direct and IHS-contract health service inpatient data were obtained from the IHS National Patient Information Reporting System (2,7,10). These data consisted of all hospital discharge records reported directly from IHS-and tribally operated hospitals and community hospitals that are contracted with IHS or tribes to provide health care services to eligible AI/AN people (6). The IHS administrative areas were classified by region: East (Nashville); Northern Plains (Aberdeen, Bemidji and Billings); Alaska, Southern Plains and Southwest (Albuquerque, Navajo, Phoenix and Tucson). We excluded the IHS California and Portland (Oregon) Administrative Areas because neither had any IHS-or tribally operated hospitals (6,11).
The average annual rates per 1,000 AI/AN children were calculated from corresponding groups using the IHS user population which includes all registered AI/AN people who received IHS-funded health care service at least once during the preceding 3 years (2,6). Influenzaassociated infant hospitalizations were examined for April 2009 to March 2010 to take into account the influenza season using the corresponding IHS user population for that period. The Alaska IHS region was further analyzed for the YK Delta and Maniilaq sub-regions, which had the highest LRTI-associated average annual hospitalization rates, for 2001Á2003 and 2009Á2011 (8,12,13).
For the general US child population aged B5 years, LRTI-associated hospitalizations were examined using the NIS, a national representative sample of hospitals conducted by the Healthcare Cost and Utilization Project (HCUP) (8). Participating hospitals are short-term, nonfederal general and specialty hospitals sampled annually; these data do not include IHS/tribal facilities' hospital discharge data. National estimates of the number of hospitalizations were calculated using the HCUP weighting methodology for the NIS (8). SUDAAN software was used to account for the sampling design of the US nationwide inpatient data and generate standard errors (SEs) of hospitalization estimates (14). Average annual hospitalization rates per 1,000 children with 95% confidence intervals (CIs) from 2009 to 2011 were calculated for corresponding groups with denominators determined from the National Center for Health Statistics bridged race population estimates (8,15). Additionally, an annual infant influenza-associated hospitalization rate per 1,000 infants with a 95% CI from April 2009 to March 2010 was calculated.
Overall, average annual rates for 2009Á2011 were also compared to corresponding previously published average annual rates for 1998Á1999 and 2006Á2008 for LRTI, bronchiolitis, pneumonia and RSV; rate ratios (RRs) with 95% CIs were calculated using Poisson regression analysis for IHS data, (16) and relative SEs for ratio estimators were calculated using SUDAAN for NIS data (17).
The hospital discharge data used for these earlier time periods were available for the analysis (2,3). The average annual percent rate of change between 1998Á1999 and 2009Á2011 rates was also calculated. Median length of stay comparison by period was performed using the Wilcoxon rank-sum test. Statistical significance was considered at the p B0.05 level.
For 2009Á2011, the median length of stay for LRTIassociated hospitalizations was 2.0 days (quartiles 1.0, 3.0 days) for AI/AN children and 1.9 days (quartiles 1.1, 3.3 days) for US children. The median length of stay for AI/ AN children decreased from 3.0 days in 1998Á1999 (quartiles 2.0, 4.0, p B0.01), but did not change for US children (p 00.32).
The 2009Á2011 average annual pneumonia-associated hospitalization rate for AI/AN infants was highest in Alaska (67.6 per 1,000), and approximately 6 times higher than the general US infant rate (Table I)

Discussion
The average annual rate of LRTI-associated hospitalizations in AI/AN children and in the general US child population aged B5 years declined significantly from 1998Á1999 to 2009Á2011. However, there remains a persistent disparity in the rate of LRTI-associated hospitalizations between AI/AN children and the general US child population. The 2009Á2011 AI/AN rate was 1.5 times higher than the general US child population rate, which was similar to the disparity from 2006 to 2008 (1.6 times higher) but slightly less than the disparity from 1998 to 1999 (1.9 times higher) (2,3). The disparity in LRTI-associated hospitalization rates in 2009Á2011 was greatest for AI/AN infants (2.1 times higher than the general US child population). LRTI-associated hospitalizations accounted for about half of all hospitalizations in AI/AN children compared to about one-quarter of all hospitalizations for US children. The rate disparity was  Other studies have reported that AI/AN children have higher rates of morbidity mortality from pneumonia compared to the general US child population. Groom et al. (4) reported a mortality rate for pneumonia and influenza 4.8 times higher for AI/AN infants than US white infants during 1990Á2009. Bronchiectasis secondary to childhood pneumonia is a rare lung condition in US children, but it is commonly diagnosed in AI/AN children living in Alaska (14Á20 per 1,000 births) similar to Australian and New Zealand Indigenous children (18). Pneumonia with empyema and invasive pneumococcal disease have also been more common among AI/AN children (12,19,20).
Various environmental exposures may in part explain this disparity. Household crowding, indoor smoke, lack of piped water and poverty have been associated with LRTIs and are more common in the Alaska and Southwest regions (21Á24). During the study period, approximately 27% of households in rural regions of Alaska did not have access to in-home water service (24). Lack of access to piped water and having less than 2 sinks in the home have been shown to be risk factors for LRTIs in children, factors which could be related to an inability to perform adequate hand hygiene (24,25). Increasing access to piped water reduced rates of LRTI-related clinic encounters in AI/ANs in Alaska (26).
Crowding was more common among AI/AN households; in the 2000 census, household crowding ( !1 person per room) occurred in 5.8% of US households overall compared with 19% of households in US Censusdesignated AI/AN, Hawaiian areas (15), and 39% households in Alaska's YK Delta sub-region (25). Indoor smoke from woodstoves and tobacco smoke was common in AI/AN homes, particularly in the rural Alaska and Southwest regions (23,25,27,28). In the Southwest region, presence of a wood stove increased the odds of LRTI by 4.9 times (27). Passive tobacco smoke exposure occurred in 46Á75% of AI/AN households compared to 18% of US households (23,29). Decreasing indoor smoke from cook stoves reduced physician-diagnosed episodes of childhood pneumonia in the RESPIRE study in Guatemala (30).
The decline in pneumonia-associated hospitalization rates is likely in part due to the introduction of the PCV7 in 2000. Griffin et al. (31) estimated that by 2007Á 2009, the pneumonia-associated hospitalization rate in US children B2 years fell by 43%. We found similar reductions of pneumonia hospitalization rates in AI/AN infants (34%) and children aged 1Á4 years (9.9%) from 1998Á1999 to 2009Á2011; however, our measured reduction did not specifically quantify the impact of PCV7 on pneumonia hospitalizations since our analysis included all listed pneumonia diagnoses, while Griffin et al. (31) limited pneumonia hospitalizations to the first-listed discharge diagnosis or pneumonia with a first-listed diagnosis of meningitis, septicaemia or empyema. There was a disparity in completion rates of the 4 dose PCV vaccine between AI/AN and the US non-Hispanic white child population noted in 2009 (71% vs. 83%, respectively) and 2011 (75% vs. 85%, respectively); (32,33). However, the 2010 PCV coverage rate was similar for AI/AN children compared to the white US population (32). PCV13 was introduced in March 2010, and coverage increased to !70% of the US birth cohort by June 2011 (5). In one region of Alaska, 91% of AI/AN children aged B5 years had received at least one dose of PCV13 by August 2011 (34). Griffin et al. (35) observed a 27% decline in pneumonia-associated hospitalizations in children aged B2 years after introduction of PCV13. We observed a decrease in AI/AN pneumonia-associated hospitalizations during 2011 compared to 2009Á2010 (Fig. 2); however, we could not assess the impact of PCV13 on pneumonia hospitalizations due to limited time since introduction.
Other factors such as changing hospitalization practices, improved socioeconomic conditions, community education and improved household and environmental factors may have contributed to the decline in pneumonia-associated hospitalization rates in AI/AN children, before and after introduction of PCV (2,3). Evaluation of these practices could elucidate the differing trends in rates among Alaska sub-regions.
The bronchiolitis-associated hospitalization rate increased in US infants from 1980 to 1996, thought to be due to more medically complex children, increased oximetry use, increased daycare use and increased RSV virulence (36,37). In the present study, the bronchiolitisassociated hospitalization rate for 2009Á2011 decreased since 1998Á1999 for both AI/AN and the general US infant population. Another study described a similar decrease for US children aged B2 years during 2000Á2009 (38). Possible causes of decreased bronchiolitis-associated hospitalizations could include a decrease in disease incidence, differing physician practices regarding hospitalization, decrease in passive smoke exposure, differing child care practices, improved hygiene practices and PCV (29,38,39).
The AI/AN infants in the Alaska and Southwest regions had the highest RSV-coded hospitalization rates. The RSV ICD-9-CM diagnosis coding may vary based on specimen collection and coding practices; however, during 18 years of RSV surveillance in Alaska's YK Delta RSV testing procedures remained stable and RSV-associated hospitalization was associated with crowding, lack of piped water, prematurity, and other underlying chronic medical conditions and lack of breastfeeding (22,24,40). Exclusive breastfeeding rates remain relatively low in AI/ AN infants through 3 months (40%) and 6 months (12%) of age, similar to the national average (37% and 17%, respectively) (41). One study suggested that exclusive breastfeeding for 4 months decreased the odds of a doctor-attended LRTI visit by approximately 50% for children less than 12 months of age (42). During this RSV surveillance period, improvements in piped water and decreased household crowding were associated with declines in RSV-and LRTI-associated hospitalization (43).
During the 2009Á2010 H1N1 influenza pandemic, AI/AN persons experienced higher morbidity and mortality than the general US population (43,44). In this analysis, the influenza-associated hospitalization rate was 3 times higher in AI/AN infants than in US infants. In the Alaska region, the AI/AN infant hospitalization rate was approximately 12 times greater. Influenza vaccine coverage remained relatively low in AI/AN children; in 2010Á2011, 67% of AI/AN children aged from 6 months to 4 years were vaccinated (45). Although the influenza vaccine coverage rate in AI/AN children was similar to the US general child population, given the high burden of influenza in AI/AN children, increasing coverage could decrease influenza-associated morbidity and mortality (45,46). One report suggested that requiring influenza vaccination in child care and pre-school settings increased vaccination coverage and reduced rates of influenzaassociated hospitalizations in children (46). Administering the influenza vaccine to pregnant women was also effective in reducing influenza-associated hospitalizations in infants in both the general US population and in AI/AN populations, yet maternal influenza vaccination rates remain low (47Á49).
The present study has some limitations. Diagnoses may be miscoded or incomplete and may vary by hospital and/or area. Hospital admission criteria, ICD-9-CM diagnostic coding, and/or pathogen-specific diagnostic workups for RSV or influenza, and sensitivity or specificity of diagnostic tests may vary between hospitals. The LRTI trends during the H1N1 influenza pandemic are not representative of long-term trends (50). No standard criteria for pneumonia diagnosis are in use nationally, so pneumonia rate comparisons should be made with caution (51,52). The IHS data set only includes hospital discharge data from IHS/tribal and IHS-contract hospitals; records for AI/AN people hospitalized outside of the IHS health care system were not included. The NIS data set may include hospitalizations of AI/AN children who were hospitalized outside of the IHS/tribal health care system (6). The population used to calculate the hospitalization rates for AI/AN children was an estimate of all children using the IHS/tribal health care system, which may not be representative of the total AI/AN childhood population. The number of states reporting to HCUP during 1998Á 1999 was less than that reporting during 2009Á2011; however, the NIS from 1993 onward has a large enough frame to produce reliable national estimates (8).
While LRTI-associated hospitalization rates in AI/AN children decreased from 1998Á1999 to 2009Á2011, the rate remained persistently higher in AI/AN children compared to the general US child population and varied by IHS region. Analysis of data from laboratory-confirmed invasive pneumococcal disease, RSV and other respiratory viruses in the Alaska and Southwest IHS regions demonstrate similar trends in these endpoints (12,20,53). The disparity of LRTI-associated hospitalizations in AI/ AN children is likely multi-factorial and in part related to increased exposure to indoor smoke, household crowding, lack of piped water and poverty. Interventions to address these environmental and socioeconomic factors as well as promoting immunization, handwashing and breastfeeding could help in reducing the disparity.