To examine diarrhea-associated deaths among very low-birth-weight (VLBW) (<1500 g) infants and low- and normal-birth-weight (LNBW) (≥1500 g) infants at birth and to identify specific interventions to prevent these deaths.
Retrospective analyses of linked infant and birth death data on diarrhea of all causes compiled by the National Center for Health Statistics, Centers for Disease Control and Prevention, Atlanta, Ga.
Infants aged 27 days through 11 months who died with diarrhea.
United States, 1991.
A majority (56%, n=143) of the 257 diarrhea-associated deaths reported among US infants in 1991 occurred among VLBW infants. Compared with LNBW infants, VLBW infants had a 100-fold greater diarrheal mortality (269 deaths per 100000 live births for VLBW infants vs 2.8 deaths per 100000 live births for LNBW infants), died at a younger age, and more often died in the hospital. Diarrhea-associated deaths among VLBW infants were strongly associated with prematurity and a low 1-minute Apgar score whereas African American race, less maternal education, and a low 1-minute Apgar score were associated with increased diarrheal mortality among LNBW infants.
Infants of VLBW are at an increased risk for diarrheal deaths and new efforts are required to understand and improve the diagnosis of and therapy for diarrhea among these infants. For LNBW infants, diarrheal deaths remain a social problem and efforts need to focus on improved education and home-based rehydration therapy for children whose mothers fit the high-risk profile and who may lack adequate access to health care.
IN THE United States, diarrheal diseases cause an estimated 167000 hospitalizations and 300 deaths each year among children younger than 5 years of age.1- 3 From 1968 through 1991, 78% of these deaths occurred in infants aged 1 through 11 months and the number decreased from 1200 deaths per year in 1968 to about 300 deaths per year in 1985. Since 1985, however, diarrhea-associated infant deaths in the United States have not declined substantially. The decrease in diarrhea-associated infant deaths prior to 1985 could not be attributed to a reduction in the incidence of severe disease as the number of hospitalizations did not undergo a similar decline.1 It may be explained in part by improved therapy for children hospitalized with diarrhea and better access to medical care.4,5 A key question unanswered in previous studies is why the decline in diarrhea-associated infant deaths stabilized at 300 deaths per year in 1985, and remained constant at this level for nearly a decade.
The stabilization of diarrhea-associated mortality since 1985 could represent a failure to provide appropriate treatment, inadequate access to health care for selected groups in society, improper diagnosis or coding problems, or the presence of diarrhea as an associated diagnosis unrelated to death rather than as a primary cause of death. When Ho et al6 first examined diarrhea-associated childhood deaths in the United States that occurred between 1973 and 1983, risk factor analyses indicated that many of these deaths were avoidable because diarrhea was listed as the principal cause of death, and the major associated risk factors were social (ie, maternal race, young age, failure to complete high school, and lack of prenatal care), and not biological conditions such as prematurity. Examination of coroner reports of diarrheal deaths in New Mexico confirmed that many of these deaths in American Indian children were associated with social factors such as limited access to care rather than to particularly severe disease.7 Kilgore et al2 updated these analyses to 1991 and found that by 1985, a majority of diarrhea-associated deaths due to simple dehydration had been prevented and the residual deaths represented the more complicated cases. The median age at death had declined from 5 months in the early period (1968 through 1973) to 1.5 months in the later period (1986 through 1991), and prematurity replaced electrolyte imbalance as the most common secondary diagnosis listed on the death certificate of infants who died of diarrhea. However, from this mortality data alone, we could not assess the extent to which social issues such as maternal education, age, and access to prenatal care were still at play or whether diarrhea-associated deaths among premature infants merely reflected their reduced capacity to survive the physiologic stress associated with enteric infections.
The national linked birth and infant death data consist of vital records for the entire cohort of children born in the United States and includes information on key maternal factors associated with diarrhea-associated deaths, eg, maternal age, marital status, educational attainment, and access to prenatal care, and infant characteristics including birth weight, gestational age, and Apgar score at birth.8 We used data for the year 1991, the latest year for which information is available, to better characterize risk factors for diarrhea-associated deaths among US infants and to identify specific intervention strategies to reduce these deaths.
National linked birth and infant death data for the year 1991, compiled by the National Center for Health Statistics, Centers for Disease Control and Prevention, were used for this study.8 A diarrhea-associated infant death was defined as the death of a US resident infant aged 1 month through 11 months for whom at least 1 of the diarrhea-specific International Classification of Diseases, Ninth Revision (ICD-9) codes was recorded anywhere on the death record.9,10 The diagnosis and ICD-9 codes for diarrhea-associated deaths included those of determined etiology (bacterial [001, 003-005, 008.0-008.5], parasitic [006-006.2, 006.8, 006.9, 007-007.3, 007.8, 007.9], viral [008.6, 008.8]), and undetermined etiology, including those presumed to be infectious (009.0-009.3) and noninfectious (558). Deaths occurring within the first 27 days of life were excluded due to their reclassification in ICD-9 as deaths due to neonatal causes alone, with no mention of diarrhea.9
Diarrhea-associated infant mortality rates were calculated as the number of diarrhea-associated deaths per 100000 live births. Prematurity and low birth weight have been associated with increased infant mortality,11 and may be important cofactors in deaths due to diarrhea and other infectious diseases. We hypothesized that diarrhea-associated deaths among very low-birth-weight (VLBW) (<1500 g) infants could be related to prematurity and nosocomial infections whereas deaths among low- and normal-birth-weight (LNBW) (≥1500 g) infants could be attributed to a previous diarrheal illness among children born to young mothers living in poor, socially disadvantaged, and underserved segments of the US population. To evaluate this hypothesis, we compared characteristics—including the underlying cause of death, age at death, and location at the time of death—between the 2 groups of infants. The comparison of age at death in months was analyzed using the Wilcoxon rank-sum test, and location was analyzed using the χ2 test. We also examined infant mortality rates by birth weight group overall, and by infant characteristics including gestational age, sex, 1-minute Apgar score, and maternal characteristics including age, race, educational level, marital status, and trimester of initiation of prenatal care. Within each birth-weight group, relative risks and 95% confidence intervals were determined using logistic regression analysis.12
Multiple logistic regression analysis was used to further examine infant and maternal characteristics that were significantly associated with diarrheal mortality by fitting a series of hierarchial models. Gestational age was not included in the modeling strategy because it correlates with birth weight and its measurement can be unreliable.13 Because of the large number of survivors in the 1991 birth cohort (>4 million), a 25% random sample of the survivors was used in the multiple logistic regression to obtain estimated relative risks.
During 1991, 257 diarrhea-associated deaths were reported among infants aged 1 to 11 months in the United States. Fifty-six percent (n=143) of these deaths occurred among VLBW infants and 44 % (n=114) occurred among LNBW infants (Table 1). Most (79.7 %) diarrheal deaths among VLBW infants were of presumed infectious etiology; of the remaining, 18.9% were of presumed noninfectious etiology and 1.4% were of viral origin. A majority (57%) of the deaths among LNBW infants were of presumed noninfectious etiology and 25.4% were of presumed infectious origin. For the remaining deaths for which an etiology was recorded, viral diarrheas were most common and accounted for 17.5% of the deaths. Bacterial causes were reported in 3 deaths: 2 deaths due to Escherichia coli and 1 due to cholera.
We compared deaths among VLBW and LNBW infants for the underlying cause and other associated diagnoses, age at death, and location at the time of death. The proportion of infants for whom diarrhea was recognized as the underlying cause of death was similar for the 2 groups (76% for VLBW infants vs 72% for LNBW infants). However, additional diagnoses listed on the death records differed greatly between the groups: prematurity (55%), extreme prematurity with a gestational age less than 28 weeks (26%), and unspecified septicemia (26%) were the most common additional diagnoses listed on the death records of VLBW infants, whereas volume depletion (23%) and unspecified septicemia (20%) were the most common additional diagnoses listed on the death records of LNBW infants. Compared with LNBW infants, VLBW infants died at a younger age (median age, 1 month for VLBW infants vs 3 months for LNBW infants; P<.001) and more often died in the hospital (96% for VLBW infants vs 64% for LNBW infants; P<.001).
We examined the relative importance of individual infant and maternal characteristics associated with diarrheal mortality among VLBW and LNBW infants to determine whether the deaths were related to the inherent biological vulnerability of these infants or to disparities in social conditions and access to health care (Table 2). The overall diarrhea-associated infant mortality rate for VLBW infants was approximately 100 times higher than that for LNBW infants (269 vs 2.8 per 100000 live births; relative risk, 97; 95% confidence interval, 75-123). Biological characteristics—including a gestational age less than 27 weeks and a 1-minute Apgar score less than 8—were strongly associated with an increased risk for diarrhea-associated death among VLBW infants. Low-birth-weight infants with a gestational age less than 34 weeks and 1-minute Apgar score less than 8 also appeared more likely to die of diarrhea, but few deaths (10 and 25 deaths, respectively) were reported among these groups of infants, and the magnitude of the risk associated with a low 1-minute Apgar score was approximately 2 times greater among VLBW infants. In contrast, social characteristics such as young maternal age, less maternal education, and less prenatal care were strongly associated with an increased risk for diarrhea-associated death only among LNBW infants. Other social factors including having a black mother or an unmarried mother were associated with an increased risk for a diarrhea-associated death among both groups of infants, but the magnitude of the risk was greater for LNBW infants.
We used multiple logistic regression analyses to evaluate the relative contribution of infant and maternal characteristics to diarrheal mortality among VLBW infants and LNBW infants (Table 3). Among VLBW infants, a low 1-minute Apgar score remained strongly associated with an increased risk of a diarrhea-associated death. Infants born to black mothers were also at increased risk for a diarrhea-associated death, but the magnitude of the risk was low. Black maternal race, lesser maternal education, and a low 1-minute Apgar score were associated with increased diarrheal mortality among LNBW infants.
Our findings indicate that diarrheal deaths among the 1991 birth cohort were reported among 2 distinct groups of infants. A majority of diarrhea-associated infant deaths were reported among VLBW infants, and results from this study suggest that these deaths may not be preventable by conventional strategies. First, the secondary diagnoses most commonly reported on the death certificates of these infants were biological conditions, such as prematurity and septicemia, rather than complications of diarrhea (ie, dehydration, electrolyte imbalance, and shock), suggesting that diarrhea may be a contributory factor rather than the principal cause of death. However, the fact that dehydration is difficult to diagnose clinically among VLBW infants may also be responsible, in part, for this observation. Second, risk factors most significantly associated with these deaths were neonatal conditions such as a low Apgar score and not social factors, such as maternal age, race, level of education, and access to prenatal care. Finally, the occurrence of deaths at a very young age and in hospital wards (96%) suggests that some of these may result from nosocomial infections that may differ in etiology and be harder to prevent than community-acquired infections. Moreover, these patients die despite the fact that 96% are in the hospital and have full access to care. Some may represent patients with necrotizing enterocolitis that is common in these children and may not be adequately diagnosed at death.
On the other hand, dehydration was the most common secondary diagnosis reported on the death certificate of LNBW infants who died of diarrhea. Many diarrheal deaths among these infants occurred at an older age and outside of hospitals, suggesting that the underlying diarrheal episodes possibly resulted from community-acquired infections. In addition, these deaths were more related to social and behavioral factors, and not biological conditions such as prematurity. This profile of the LNBW infant at risk for death due to diarrhea is similar to that reported in previous studies,6,7 and suggests that these deaths are potentially preventable through intervention strategies targeted to high-risk mothers (ie, young, black, unmarried mothers with less than a high school education) who receive limited prenatal care, and who may have limited access to medical care for their infants with diarrhea.
The observed differences in the risk profile and potential preventability of diarrheal deaths among VLBW and LNBW infants might explain the lack of decline in diarrhea-associated infant mortality beyond 1985. Ho et al6 reported that a majority of diarrhea-associated deaths in the 1980 birth cohort occurred among infants who weighed 2500 g or more at birth, whereas we found that most diarrheal deaths in the 1991 birth cohort were reported among VLBW infants. It is likely that prevention of deaths among LNBW infants by improved access to medical care and appropriate rehydration therapy accounted for the decline in diarrhea-associated mortality between 1980 and 1985. As a result, deaths among VLBW infants have emerged as the major contributor to overall diarrhea-associated infant mortality since 1985, and failure to recognize and specifically address the high risk for mortality among these infants could explain the lack of a further decline in mortality.
The marked differences in the characteristics and risk profile of VLBW and LNBW infants who died of diarrhea indicates the need for targeted intervention strategies to reduce diarrheal deaths among the 2 groups of infants. Many diarrheal episodes that led to the death of VLBW infants were presumed to be of infectious etiology, but the responsible microbial agent was rarely identified. A greater understanding of diarrhea in the context of complicated illness of VLBW infants in the hospital, including identification of the etiologic agent, may help reduce diarrheal deaths among these infants. Effective vaccines against rotavirus are likely to be licensed in the near future,14,15 and studies are needed to determine whether rotavirus is a major contributor to diarrheal mortality among VLBW infants who may lack their full complement of maternal antibody to rotavirus and other childhood infections.
At the same time, efforts to provide timely rehydration therapy to infants with diarrhea born to mothers in socially disadvantaged groups with limited access to health care are likely to further reduce diarrheal mortality among LNBW infants. An improvement in the accuracy of diagnosis at the time of death or a thorough investigation of the circumstances surrounding the death of all patients who die due to diarrhea may provide further insights to determine what preventive measures might benefit these infants.16 However, it is important to recognize that the persistent deaths due to diarrhea likely represent the more "stubborn" cases that are hard to prevent. Although marked reductions in diarrheal deaths are unlikely, recognition of the high risk for VLBW infants for death due to diarrhea provides a new prevention opportunity that needs to be explored further.
Accepted for publication September 16, 1997.
We thank John O'Connor for editorial assistance during preparation of the manuscript.
Editor's Note: The findings of this study provide new insights for determining strategies to prevent mortality for diarrheal diseases in infants. Is anyone out there working on a vaccine against social problems?—Catherine D. DeAngelis, MD
Reprints: Umesh D. Parashar, MBBS, MPH, Viral Gastroenteritis Section, Mailstop-G04, Centers for Disease Control and Prevention, Atlanta, GA 30333.
Parashar UD, Kilgore PE, Holman RC, Clarke MJ, Bresee JS, Glass RI. Diarrheal Mortality in US InfantsInfluence of Birth Weight on Risk Factors for Death. Arch Pediatr Adolesc Med. 1998;152(1):47-51. doi:10.1001/archpedi.152.1.47