Xu, Fujie MD, PhD*; Gee, Julianne M. MPH†; Naleway, Allison PhD‡; Zangwill, Kenneth M. MD§; Ackerson, Bradley MD∥; Eriksen, Eileen MPH§; Weintraub, Eric S. MPH†; Hutchins, Kathleen BS*; Wei, Feifei PhD¶; Berman, Stuart M. MD, ScM*; Markowitz, Lauri E. MD*
NEONATAL HERPES IS A POTENTIALLY devastating infection caused by herpes simplex virus (HSV) type 1 or type 2.1,2 Newborn infants can acquire HSV infection in the antepartum, intrapartum, or the postpartum period but approximately 90% of transmissions are vertical.1 The majority of women who deliver infants with neonatal HSV have no signs, symptoms or history of genital herpes.3–5 The risk of infection is significantly elevated among infants born to mothers who are seronegative for both HSV viruses, due to high transmission rates from mother to infant if primary HSV infection is acquired in late pregnancy.6,7 Among women with vaginal HSV shedding at the time of labor, cesarean delivery has been shown to reduce the transmission rate from mother to infant by 86%.6
HSV infection in infants usually manifests during the first 2 to 3 weeks and traditionally has been categorized clinically into 3 forms1,8: disease limited to the skin, eye, and mouth (SEM), to the central nervous system (CNS), or disseminated disease. Although a papulovesicular rash and seizures are typical of neonatal herpes, in about one-third of infected infants, skin lesions never develop. Importantly, nonspecific or subtle symptoms can result in delayed and/or underdiagnosis of neonatal herpes.4
When not promptly treated, SEM cases can disseminate to the viscera and/or CNS, and survivors may sustain severe neurologic impairment even with antiviral therapy.9 The clinical diagnosis of HSV infection in neonates is challenging because the signs of disseminated infection often mimic severe bacterial or viral infection with rapid progression of infection and/or death caused by sepsis, progressive liver failure, disseminated intravascular coagulation, respiratory failure, or neurologic deterioration.8–10
In the United States, reporting of neonatal HSV infection to health departments is required in only a few states, and few population and laboratory-based incidence data are available.11 The surveillance of neonatal HSV infection has garnered much attention lately in the United States and Canada, and some researchers have proposed that neonatal herpes should be a nationally reportable disease.11–14 The objectives of our study are to estimate the incidence of neonatal herpes and to assess the utility of different surveillance methods for neonatal herpes in a large population-based sample.
All neonates/infants enrolled during 1997 to 2002 in the Southern California and Northwest (Oregon and Washington) regions of the Kaiser Permanente Health Plan were included in this study. Information available in automated databases includes member demographics and enrollment periods, diagnoses and procedures, laboratory tests, and prescription medications. Standardized subsets of these databases are made available to researchers at the Centers for Disease Control and Prevention as part of the Vaccine Safety Datalink project. The Vaccine Safety Datalink datasets and methods are described elsewhere.15
We used a comprehensive and hierarchical approach, summarized in Figure 1, to ascertain potential cases of neonatal herpes. We started with searching in automated databases for infants with diagnoses most likely to be neonatal herpes: infants who received a International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9) codes of 054.xx (herpes simplex infection) or 771.2 (other congenital infections, including congenital herpes simplex, listeriosis, malaria, toxoplasmosis, and tuberculosis) during hospitalizations in the first 26 weeks (6 months) of life. To identify potential cases with onset of illness during the neonatal period, the infant must have 1 or more disease related hospitalization in the first 4 weeks of life. The 6-month period was designed to capture cases for whom the workup and diagnosis of neonatal HSV infection was delayed.
We then searched for additional potential cases using a set of ICD-9 codes that was compatible with or suggestive of neonatal herpes infection (Table 1). These ICD-9 codes were selected because: (a) neonatal herpes involving the skin, eye, or mouth often progresses to CNS and disseminated infection if untreated, (b) papulovesicular rashes and seizures are common manifestations suggestive of neonatal herpes, and (c) disseminated neonatal herpes can mimic severe nonherpetic infection, pneumonitis, hepatitis, or disseminated intravascular coagulation. As above, we searched for infants with any of these ICD-9 codes during the first 6 months of life, provided the infant also had 1 or more hospitalizations in the first 4 weeks of life.
Because an alternative diagnosis could account for the infants' clinical presentations, we excluded infants who also received a specific ICD-9 coded discharge diagnosis of any of the following infections: listeriosis, congenital syphilis, toxoplasmosis, congenital cytomegalovirus infection, congenital rubella, enterovirus, or streptococcus. We further limited our search of infants with sepsis codes (ICD-9 codes 038.9 and 771.8) to those either with a subsequent neurologic impairment in the first year of life, or with “severe” disease, defined here as sepsis requiring a hospital stay ≥10 days in duration.
In addition to identifying potential cases based on discharge diagnoses, we identified infants who died between 4 and 28 days of life (excluding infants whose causes of death were coded as extreme immaturity, chromosome abnormalities or congenital heart diseases), and infants with a positive HSV laboratory test (excluding serology) with specimen collection data at ≤4 weeks of life (Fig. 1).
Medical Record Abstraction
Trained abstractors at each site used a standardized form to collect information from the medical record about HSV signs, symptoms, laboratory tests, and treatments. We reviewed all hospitalizations that occurred within the first 6 months of life, including the birth hospitalization.
We defined a confirmed neonatal HSV case as any infant who presented with signs, symptoms and findings compatible with neonatal herpes within the first 4 weeks of life, and had a positive culture, polymerase chain reaction (PCR), or direct fluorescent antibody test for HSV. We defined a probable neonatal HSV case as any infant who presented with signs, symptoms and findings compatible with neonatal herpes and received an HSV-related diagnosis but did not have an HSV-specific laboratory test or tested negative for HSV. Results of HSV serologic tests were not considered because these tests are of little value in making a diagnosis of HSV infection in neonates due to the frequent presence of maternal antibodies.
Confirmed and probable cases of neonatal herpes were further classified hierarchically into 3 categories based on clinical characteristics using criteria similar to those of Whitley et al.8 Disseminated infections were defined as those with visceral organ involvement, CNS infections as those with neurologic involvement or abnormalities of cerebrospinal fluid, and SEM infections as those with mucosal or skin lesions without evidence of other organ or CNS involvement.
Final Review by Infectious Disease Specialists
All infants considered by the authors to meet the criteria of confirmed or probable neonatal herpes cases were reviewed by 2 pediatric infectious disease (ID) specialists (K.Z., B.A.). Specifically, the final review included all infants who (a) had any positive HSV tests; or (b) wordings of “herp” (herpes or herpetic) or “congenital infection” appeared in discharge diagnoses (including rule out and ruled out); or (c) received an ICD-9 code of 054.xx or 771.2. The 2 ID specialists independently reviewed the completed abstraction forms and copies of the hospital discharge summaries and made a determination as confirmed, probable or not a case based on our case definitions. When there were discrepant results, final adjudication was made after repeated review and consultation with a third reviewer.
We calculated the incidence of neonatal HSV by dividing the number of confirmed HSV cases by the total number of infants with at least 30 days of health plan enrollment since birth. We determined 95% confidence intervals (CI) using the Δ method.16 All analyses were performed in SAS Version 9.0 (SAS Institute, Cary, NC).
The study was reviewed and approved by the institutional review boards at Centers for Disease Control and Prevention, Kaiser Permanente Southern California (KPSC), Kaiser Permanente Northwest (KPNW), and the Los Angeles Biomedical Research Institute at Harbor-UCLA Medical Center.
From 1997 through 2002, 270,703 infants were enrolled for at least 30 days after birth at the 2 study sites. The search by ICD-9 codes of 054.xx and 771.2 yielded a total of 66 infants, and among these, we were able to locate and complete medical chart abstraction for 60 (91%) infants (Table 2). The searches by other ICD-9 coded discharge diagnoses compatible or suggestive of neonatal herpes, neonatal deaths, and positive HSV test identified a total of 671 infants, of which 639 (95%) had medical chart abstraction completed (Table 2). The reasons for no or incomplete medical chart abstraction include missing medical records, denied access to medical records, or the infant was cared for at a non-Kaiser Permanente affiliated hospital.
Among the 699 infants with complete medical chart abstraction (Table 2), 73 were included in the final review by the 2 ID specialists. The categorizations (confirmed, probable, or not a case) from their independent review agreed for all except 7 infants. The different opinions were resolved for 6 infants after repeated review; a third specialist was consulted to resolve the difference in categorization for the remaining one. After final adjudication, we identified a total of 35 confirmed and probable cases, 24 (69%) of which were confirmed cases (Table 2). Among the 24 confirmed cases, 20 (83%) had at least 1 positive test by culture, and 4 were positive by PCR only.
The incidence of laboratory-confirmed neonatal HSV infection was 8.9 (95% CI 5.4–12.4) per 100,000 live births. When the probable cases are included, the incidence was 12.9 (95% CI 8.6–17.2) per 100,000 live births. Based on the annual birth cohort of approximately 4.1 million in the United States, the projected number of cases of neonatal HSV would be 529 (95% CI 353–705) per year.
The incidence of confirmed neonatal herpes was 18.1 per 100,000 live births (6 cases of 33,112 live births) at KPNW and 7.6 per 100,000 live births (18 cases of 237,591 live births) at KPSC (P = 0.06). When probable cases were included, the incidence increased to 24.2 per 100,000 live births (8 cases of 33,112 live births) at KPNW and 11.4 per 100,000 live births (27 cases of 237,591 live births) at KPSC (P = 0.09).
Clinical History and Presentation of Neonatal HSV Infection
Most cases with confirmed or probable neonatal herpes were of normal birth weight and were delivered vaginally (Table 3). However, 7 confirmed cases and 1 probable case were delivered by cesarean section. Only 1 confirmed case had a maternal history of neonatal herpes.
Among all cases of neonatal herpes, 66% were admitted to the hospital within the first week of life or never discharged after birth, and only 17% were admitted after week 2. Among the confirmed and probable cases of neonatal herpes, 57% had SEM infection, 23% had CNS infection, and 20% had disseminated infection (Table 3). One of the confirmed cases died. Among the 17 cases with known HSV type, 10 were HSV-2 and 7 were HSV-1.
Use of ICD-9 Codes for Identifying Cases of Neonatal Herpes
We evaluated the performance of using the 2 HSV ICD-9 codes alone as the method to ascertain cases of neonatal herpes (Table 4). The ICD-9 code 054.xx identified 31 infants, and of these, 16 were confirmed cases, 6 were probably cases, and 9 were not a case based on the final review (Table 4). When both ICD-9 codes 054.xx and 771.2 were used, a total of 60 infants were identified, and of these, 22 were confirmed, 9 were probable cases, and 29 were not a case (Table 4). Together, the 2 codes identified 31 (89%) of 35 confirmed or probable neonatal HSV infection and 22 (92%) of the 24 confirmed neonatal HSV infection.
The positive predictive value of the 2 codes (054.xx and 771.2) is fairly low, especially for the ICD-9 code of 771.2. Among the 31 infants who received the code 054.xx, 22 (71%) were confirmed or probable cases of neonatal herpes. Of the remaining 9 infants, 7 had a diagnosis of herpes or neonatal herpes “rule out” or “ruled out,” and in 2, coding errors may have occurred. Among the 29 infants who received the code of 771.2 without the code 054.xx, only 9 (31%) were confirmed or probable cases of neonatal herpes.
Four cases of neonatal herpes (2 confirmed and 2 probable) did not receive ICD-9 codes 054.xx or 771.2 (Table 2). Among these, 2 cases were captured in our study because they received a diagnosis coded as “neonatal hepatitis” or “seizures,” and these codes made them eligible for chart abstraction. These 2 also had herpes or neonatal herpes listed among discharge diagnoses but omission or miscoding might have occurred. The remaining 2 were diagnosed and coded as “viral encephalitis,” but neonatal herpes was mentioned in the discharge summary.
Use of Laboratory Databases for Identifying Cases of Neonatal Herpes
We also evaluated the performance of using laboratory tests alone as the method to ascertain cases of neonatal herpes (Table 4). We identified 28 infants with a positive test by culture, PCR or direct fluorescent antibody, and 24 were confirmed cases of neonatal herpes after the final review (Table 4). Of the 23 infants with positive cultures, 19 were confirmed cases. Of the remaining 4, the specimens positive for HSV were collected at ≤1 day after birth due to maternal history of genital herpes (n = 2) or maternal lesions suspected of genital herpes during labor (n = 2). All culture positive specimens were surface specimens (conjunctival, nasopharyngeal, or pharyngeal swabs). Because these 4 infants never developed any symptoms related to neonatal herpes infection, they were not counted as cases in this study because HSV contamination could not be ruled out. All 4 infants were treated with acyclovir and evaluated for, but not diagnosed with CNS involvement (3 had lumbar puncture, 2 had magnetic resonance imaging, and 1 had cranial ultrasound along with electroencephalography); all 4 received a discharge diagnosis coded as 054.xx or 771.2.
Despite our attempt to exclude cases with only positive HSV serology from our chart review, the laboratory databases maintained by commercial or hospital labs sometimes did not specify clearly the type of the HSV test. The search of laboratory data identified 5 potential cases that had not been identified by ICD-9 codes (Fig. 1, Table 2). After chart review, 4 of the 5 had a positive anti-HSV IgG test, and 1 had an indeterminate serologic test. In addition, 4 of the 5 had 1 or more negative HSV cultures and 3 had a negative IgM test.
Investigation of Neonatal Deaths
After excluding infants whose causes of deaths were coded as extreme immaturity, chromosome abnormalities, and cardiovascular system defects, our search identified 68 infants who died between 4 and 28 days of life (Fig. 1). Of these, 62 had complete medical chart information. The majority of the infants (54 or 87%) died after a single hospital admission, including 41 infants (66%) who were never discharged from the hospital after birth. For the 21 infants who were discharged and readmitted, the mean age at admission was day 7 of life. HSV cultures were obtained from 10 infants (one of whom also had a PCR test of cerebrospinal fluid); all results were negative. Only 1 infant received acyclovir treatment. Among the 51 infants who were not tested or treated for HSV, 48 died of noninfectious disease causes, 2 had an infection other than HSV (streptococcus), and 1 had an unexplained death. No case of neonatal herpes was identified.
In our large, population-based sample, the incidence of laboratory-confirmed neonatal HSV infection was 8.9 per 100,000 live births. When probable cases were included, the incidence was 12.9 per 100,000 live births.
There have been 3 large studies about the incidence of neonatal herpes in the United States. The only large prospective study was from the Seattle area, where 15 cases of neonatal herpes were identified among 48,390 births in the 1980s and 1990s, yielding an incidence of 31.0 per 100,000 live births.6 Even when probable cases are included, the overall incidence of neonatal herpes from our study is substantially lower than the incidence observed in the study in Seattle (P <0.0001). When probable cases are included, the incidence in our study is similar to that reported in California, where the incidence of neonatal herpes (based on ICD-9 code 054) was estimated at 12.1 per 100,000 live births during 1995 to 2003.17 The third study, in a managed care population mainly from mid-Atlantic and Northeast regions, used various ICD-9 codes (054.xx, 286.6, 695.1, 771.2, and 790.8) to identify possible cases of neonatal herpes and reported a higher incidence (178 cases of 233,487 infants or 76 per 100,000 live births).18
Several reasons may contribute to these differences in incidence of neonatal herpes. First, the 3 studies employed different methodologies, with the latter 2 based on ICD-9 code(s) only. There are also possible differences in HSV-1 and/or HSV-2 seroprevalence in the study populations. Although not statistically significant, the estimated incidence of neonatal herpes in our study was higher in KPNW (Oregon and Washington) than in KPSC (California). Interestingly, the incidence at KPNW (24.2 per 100,000 live births) was not statistically different from the disease burden reported from Seattle area (31.0 per 100,000 live births, P = 0.4). The proportion of pregnant women who are seronegative for HSV is likely to be higher in Washington/Oregon than in California because of differences in racial/ethnical compositions of the populations in these states (the proportion of non-Hispanic white women is higher in Washington/Oregon than in California) and because non-Hispanic white women are more likely to be seronegative compared with women of other race/ethnicity.13 In addition, the incidence of neonatal herpes will vary with the proportion of pregnant women undergoing cesarean section, the management of women in labor and the management of infants with unspecific signs and symptoms compatible with neonatal herpes.
Prompt recognition and diagnosis and appropriate antiviral therapy can limit the morbidity and mortality from neonatal HSV infections. Because no single constellation of symptoms definitively distinguishes neonates with HSV infections, a high index of suspicion for herpes infection must be maintained and appropriate laboratory tests aggressively utilized. Although the number of cases in our study is small, we found the proportion of cases having SEM infection was 57%, higher than ∼40% reported in the literature.1,4,9 These findings suggest that earlier recognition of HSV infection and earlier treatment may affect the distribution of disease categories. In addition, we found 4 infants who had a positive culture but never had HSV-related signs or symptoms; early antiviral treatment might have prevented these exposed infants from developing clinical disease.
Changes in the epidemiology of HSV-1 and HSV-2 likely will have complex effects on the incidence and etiology of neonatal herpes. Of the small number of cases in which the type of virus was known, we found the majority (10 of 17) were typed as HSV-2. With the seroprevalence of HSV-1 and HSV-2 decreasing in the United States, more pregnant women will be seronegative, and thus, at risk for acquiring both HSV-1 and HSV-2 during pregnancy.19 Changes in sexual behavior, such as the practice of oral sex, can increase the risk of genital herpes caused by HSV-1,19 and may lead to an increase in neonatal herpes caused by HSV-1. In the recent study in Canada, the majority (62%) of HSV infections in infants ≤2 months of age were typed as HSV-1.14
Some HSV researchers and public health authorities have suggested that neonatal herpes should be a nationally reportable disease.11,12 Critical to the identification of cases is an understanding of the utility of existing tools for surveillance. Previous studies have used the ICD-code of herpes (054.xx) to identify cases of neonatal herpes.17,18,20 Our data suggest that ICD-9 codes 054.xx combined with 771.2 are fairly sensitive (the yields from all other ICD-9 codes were low). Together these 2 codes can identify the majority of cases of neonatal herpes. But these codes are not specific in identifying cases of neonatal herpes; the positive predictive value of code 054.xx is higher than that code 771.2. To increase the sensitivity and specificity, some modifications to the ICD-9 codes for neonatal herpes, especially the code of 771.2, and training of coders to exclude “rule out” or “ruled out” cases are needed. Monitoring the trends in neonatal herpes using discharge databases without chart review will be limited until such changes are instituted. A laboratory-based reporting has been proposed as a surveillance approach for neonatal herpes.12 In this study, we found only 69% of cases were confirmed by definitive laboratory results. Therefore, any laboratory-based surveillance systems, which are essentially dependent on physician clinical practice patterns, will underestimate the burden of disease. A laboratory-based surveillance system is not 100% specific because false-positive can occur due to contamination. In addition, trends and variations among populations from laboratory-based surveillance system may be difficult to interpret because provider behaviors regarding HSV testing can change and vary. Despite these limitations, the need for ongoing surveillance for neonatal HSV disease is clear, given the incidence of disease, its potential morbidity, and the availability of antiviral therapy.
This study has several limitations. First, up to 9% of potential cases had no or incomplete chart review. Second, case ascertainment was confined to hospitalized cases, which probably led to the preferential identification of infants with more severe neonatal herpes infection. Another limitation is that we excluded cases with other common neonatal infections. As a result, neonatal herpes cases with a dual infection would have been missed. In addition, it is possible the denominator for our incidence calculation was overestimated because we did not require infants to be enrolled continuously beyond the first 30 days of life. These limitations may have contributed to an underestimate of disease incidence in our study. Our incidence estimates likely represent the minimum burden of disease.
Our study exposed methodological issues and some limitations of the use of large administrative databases for surveillance of neonatal herpes. National strategies for neonatal herpes surveillance and improved surveillance tools should be developed and evaluated in an attempt to increase our capacity to identify new cases and possibly to decrease morbidity and mortality of this serious disease.