To determine the incidence of neonatal endogenous endophthalmitis in the United States between 1998 and 2006 and to identify associated risk factors.
Retrospective cohort study.
We used the Nationwide Inpatient Sample database, a 20% representative sample of all hospital discharges in the United States, to help refine our understanding of this condition. International Classification of Diseases, ninth edition, codes for endophthalmitis, sepsis, and suspected endophthalmitis risk factors in hospitalized infants and neonates were searched in the database and were tracked over time. The main outcome measure was incidence of neonatal endophthalmitis over the study period.
Of 3.64 million live births in 1998, 317 newborns were identified with endophthalmitis (8.71 cases per 100 000 live births). Of 4.14 million live births in 2006, only 183 newborns were identified with endophthalmitis (4.42 cases per 100 000 live births) by comparison. The incidence of endophthalmitis decreased at a rate of 6% per year ( P = .01130) between 1998 and 2006. Neonates with endophthalmitis were more likely to have systemic bacteremia (odds ratio, 21.114; P < .0001), Candidemia (odds ratio, 2.356; P < .0001), a birth weight of less than 1500 g (odds ratio, 1.215; P < .0001), and retinopathy of prematurity (odds ratio, 2.052; P < .0001).
We objectively validated the commonly held belief that Candidemia, bacteremia, retinopathy of prematurity, and low birth weight are significant risk factors for endophthalmitis development in infants, which seems to have had a decreasing incidence in recent years.
Unlike exogenous endophthalmitis in adults that arises predominantly from antecedent intraocular surgery, neonatal endophthalmitis is overwhelmingly the result of an endogenous source such as neonatal bacteremia or, more commonly, systemic Candidemia. It has been taught classically that clinical factors that increase the risk of neonatal bacterial sepsis include preterm delivery, a prolonged duration of internal monitoring, and having maternal chorioamnionitis, endometritis, or group B streptococcal colonization.
Neonatal candidemia has been reported in 1% of all neonatal intensive care unit (NICU) admissions, with a higher incidence being reported in infants weighing less than 1500 g. Candida species are the third most common blood culture isolate recovered from late-onset sepsis in the NICU. Candidemia may be associated with the progression of retinopathy of prematurity (ROP), with or without concomitant Candida -associated endophthalmitis. The incidence of endophthalmitis in patients with disseminated candidiasis varies in reports from 6% to 50% and increases with prolonged infection ; however, it is unequivocally the principal cause of endogenous endophthalmitis in the newborn population. Candidemia also is associated with a significant mortality risk, accounting for 12% of the mortality in extremely low birth weight infants.
For ophthalmologists performing inpatient consultations, a frequent consultation request is to rule-out endophthalmitis in a NICU patient with known candidemia or septicemia. Unfortunately, most of what we know about the epidemiologic features of neonatal bacterial and fungal endophthalmitis is the result of clinical experience, retrospective studies, anecdotal data, and meta-analyses. The purpose of the present study was to use an objective metric—the Nationwide Inpatient Sample (NIS)—to provide further appraisal of the contemporary incidence and potential associated risk factors for neonatal endophthalmitis.
Study Design and Subjects
We performed a retrospective cohort study using an existing database, the NIS. The NIS is maintained by the Agency for Healthcare Research and Quality as part of the Healthcare Cost and Utilization Project. The NIS is a 20% representative sample of all hospital discharges in the United States and is stratified by geographic region, hospital size, geographic location, and hospital type. The NIS is the largest all-payer administrative database that incorporates discharge data from approximately 1000 hospitals and 5 to 7 million discharges annually. The NIS 20% sample is based on a stratified probability sample of all United States hospitals to provide a national estimate of inpatient health services. The reported values for rate, incidence, and prevalence are weighted values based on the 20% sample. The NIS sampling frame changes almost annually, with more states being added each year and different hospitals selected as part of the subset each year. The NIS hospital data-sampling frame is based on a subset of hospitals that release their data to Agency for Healthcare Research and Quality for research use. The weight of samples depend on their geographical region, hospital type (public versus for profit), rural versus urban location, teaching hospital status, and bed size.
Only inpatient data found in a discharge abstract are available in the NIS. The NIS does not have unique patient identifiers, and therefore, patients cannot be followed up longitudinally. We studied these variables across the period spanning approximately the last decade, from January 1, 1998, through December 31, 2006, the most recent year the data were available to us.
In this retrospective cohort study, cases were identified by the International Classification of Diseases, ninth revision (ICD-9), diagnostic code for the following entry variables: V30 newborn ICD-9 codes to captures all live births, ICD-9 code 360.00 for endophthalmitis, and the ICD-9 codes and corresponding diagnoses that have been suspected to be associated risk factors for endophthalmitis development: birth weight < 1500 g, fungemia, bacteremia, sepsis, prematurity, ROP, birth trauma, hypoxia, having had a blood transfusion, having had retinal laser photocoagulation, hemolytic anemia, necrotizing enterocolitis, intraventricular hemorrhage, respiratory disorder, perinatal infection, fetal hemolysis, and gender. The database was queried for elective, nontransfer patients with an age younger than 19 years (ie, patients who were born in one hospital and not transferred to another institution for a higher level of care). We evaluated data from 1998 through 2006. The incidence rates of endophthalmitis were weighted according to the total number of live births over the study period. The join point was a logarithmic scale. The teaching status of the hospital was determined by hospital affiliation with either a medical school or an Accreditation Council for Graduate Medical Education residency program.
The main outcome measure of this study was the change in incidence of endophthalmitis over each of the years during the study period. Secondary outcome metrics included the identification of factors associated with endophthalmitis development. We also sought to determine the relationship between these associated factors and the risk of mortality by comparing those neonates with endophthalmitis with a reference population of those newborns without endophthalmitis. Mortality was determined by the method of inpatient reporting. No postdischarge collection of data on mortality was obtained.
Dichotomous and continuous variables were examined by the Student t test and chi-square analyses. Linear and logistic regression analyses were applied to endophthalmitis and mortality variables to correct for potential confounders. In addition, logistic regression analyses identified candidate risk factors for endophthalmitis development in newborns. A P value of less than .05 was considered statistically significant. The statistical software package used for database analysis was SAS version 8.1 (SAS Institute, Cary, North Carolina, USA). The logistic regression model was tested by application of the Hosmer and Lemeshow goodness-of-fit test and by evaluating the area under the receiver operating characteristic curve. Sample design and weights were used for all analyses, including the regression analysis. Variables were included in the models only if they showed a positive association in a univariate analysis (results not shown) or if there was a strong known clinical association. Collinearity was addressed through multicollinearity diagnostic statistics.
Patients and Database Characteristics
Using the NIS, it was extrapolated that between 1998 and 2006, there were 35.49 million live births in the United States ( Table 1 ). Among these, 1959 cases of endophthalmitis were observed for a cumulative incidence of 5.52 cases per 100 000 live births per year ( Table 2 ). The incidence was observed to be declining at a average rate of 6% per year ( P = .01130) over the study period ( Figure ), with an incidence of 8.71 per 100 000 being reported in 1998 and only 4.42 per 100 000 being reported in 2006 ( Figure and Table 2 ). For those born with endophthalmitis, there was no greater tendency toward male gender, white race, being a Medicaid beneficiary, or being born at a nonprofit hospital ( Table 1 ). For those newborns with endophthalmitis, however, there was a greater likelihood to have been born in a teaching hospital ( P < .0001) compared with newborns without endophthalmitis. Additionally, patients with endophthalmitis had a mean total charge of $43 684, compared with $4680 for patients who did not have endophthalmitis during their postnatal admission ( P < .0001).
|Patient Characteristics||Newborns with Endophthalmitis (n = 1959)||Newborns without Endophthalmitis (n = 35.49 million)||P Value|
|Inpatients with routine discharge||89.43%||95.32%||<.0001|
|Inpatients with complications||16.76%||36.71%||<.0001|
|Inpatient mortality rate||0.44%||0.28%||<.0001|
|Mean age (days)||14.78||0.2||<.0001|
|Mean length of stay (days)||14.98||2.89||<.0001|
|Mean total charges during stay||$43 684||$4680||<.0001|
|Year||Newborns with Endophthalmitis||Newborns without Endophthalmitis||Incidence (Cases per 100 000)|
|1998||317||3 638 152||8.71|
|1999||238||3 725 897||6.39|
|2000||241||3 971 941||6.07|
|2001||215||3 878 041||5.54|
|2002||213||4 014 225||5.31|
|2003||165||3 946 724||4.18|
|2004||171||4 095 901||4.18|
|2005||216||4 082 480||5.29|
|2006||183||4 139 586||4.42|
|Total||1959||35 492 947||5.52|
Effect of Birth Weight on Endophthalmitis
It should be noted that only 8% of all database entries included actual birth weights, so 92% of birth weights in this study were unknown and could have represented birth weights that were either more or less than 2500 g. Therefore, we do not report the incidence of neonatal endophthalmitis stratified by birth weight because of an insufficient dataset with respect to this variable.
Mortality in Newborns with or without Endophthalmitis
Multivariate logistic regression analysis revealed predictive factors of mortality ( Table 3 ). Newborn patients with fungemia had nearly a 30-fold increased risk of mortality (odds ratio [OR], 29.74; 95% CI, 22.1 to 39.98; P < .0001). By contrast, patients with endophthalmitis had a significantly decreased likelihood of mortality compared with those without endophthalmitis (OR, 0.575; ; 95% CI, 0.56 to 0.59; P < .0001). Similarly, female gender (OR, 0.850; 95% CI, 0.84 to 0.86; P < .001) and viremia (OR, 0.557; 95% CI, 0.475 to 0.65; P < .001) also had a decreased likelihood for mortality ( Table 3 ). An increased likelihood for mortality also was observed for patients with birth weight of less than 1500 g, bacteremia, Candidemia, cytomegalic viremia, being born in a teaching hospital, and white race ( Table 3 ).
|Odds Ratio||Confidence Interval||P Value|
|Viremia||0.557||0.475 to 0.65||<.0001|
|Endophthalmitis||0.576||0.56 to 0.59||<.0001|
|Female gender||0.850||0.84 to 0.86||<.0001|
|White race||1.003||1.001 to 1.003||<.0001|
|Birth weight < 1500 g||1.145||1.14 to 1.15||<.0001|
|Teaching hospital||1.21||1.21 to 1.22||<.0001|
|Bacteremia||2.644||2.61 to 2.67||<.0001|
|Candidemia||3.553||3.54 to 3.56||<.0001|
|Cytomegalic viremia||9.654||6.13 to 15.19||<.0001|
|Fungemia||29.74||22.13 to 39.98||<.0001|
Comorbidities in Newborns with or without Endophthalmitis
Compared with those without endophthalmitis, univariate analysis demonstrated an increased likelihood for newborns with endophthalmitis to have perinatal infection ( P < .0001), respiratory disorder ( P < .0001), and blood transfusion ( P < .0001), among other variables ( Table 4 ). ROP was more common (2.43%) among those with endophthalmitis compared with those without endophthalmitis (0.12%; P < .0001).
|Newborns with Endophthalmitis (n = 1959)||Newborns without Endophthalmitis (n = 35.49 Million)||P Value|
|Retinal laser photocoagulation||0.28%||0.02%||.0003|
|Retinopathy of prematurity||2.43%||0.12%||<.0001|
With respect to concomitant systemic infection classification, newborns with endophthalmitis were significantly more likely to also harbor candidemia (1.81%) compared with those without endophthalmitis (0.07%; P < .001; Table 5 ). Similarly, those with endophthalmitis were significantly more likely to have a diagnosis of bacteremia (0.53%) compared with those without endophthalmitis (<0.01%; P < .001). Viremia, cytomegalovirus infection, and fungemia demonstrated no increased or decreased incidence in either cohort ( Table 5 ).
|Coincident Systemic Infection||Newborns with Endophthalmitis (n = 1959)||Newborns without Endophthalmitis (n = 35.49 Million)||P Value|