1. IntroductionCongenital cytomegalovirus (cCMV) is the most common congenital infection in the United States (US) affecting approximately 3–6 per 1000 live births [1,2]. Approximately 90% of infants with cCMV are born with asymptomatic infections, defined as not having any clinical signs on physical exam, cranial imaging or laboratory tests (e.g., thrombocytopenia or elevated liver enzymes) [3]. The remaining 10% are born with symptomatic disease, which is characterized by visible signs at birth, laboratory abnormalities and/or intracranial involvement [3]. Yet, in many infants the visible signs of symptomatic cCMV disease may be subtle, often non-specific (e.g., jaundice or small-for-gestational age), and/or may occur in isolation [4,5]. The International Congenital Cytomegalovirus Recommendations Group (ICCRG) defined disease severity based on the number of manifestations attributed to cCMV and whether they occurred in isolation or not, thus differentiating mild disease from moderate to severe symptomatic disease [3]. Other critical markers of symptomatic disease, such as laboratory abnormalities and intracranial lesions, are not always outwardly noticeable and are only identified on further work-up after clinical suspicion is raised [3,4].As such, it is likely that many infants born with cCMV go undiagnosed when testing relies on the clinical recognition and suspicion of providers alone, although how many infants are being missed is a point of some debate. One study from British Columbia, Canada reported that only 10% of infants with symptomatic cCMV were being diagnosed, or 1% of all infants with cCMV [6], which is significantly lower than the estimated national prevalence of 3–6 cases per 1000 live births [1]. Other work using administrative claims data from US hospital discharges identified cCMV diagnosis codes in approximately 2 per 10,000 infants, presumably all coded for symptomatic infections [7]. Thus, nearly 50% of all expected cases of symptomatic cCMV, or 5% of all cases, were identified, which is 5x greater than the Canadian study [7]. As there is a greater urgency to diagnose, evaluate, and initiate treatment for infants with symptomatic cCMV, if half are already being diagnosed based on clinical symptoms alone, then the utility of a newborn screening program may not be as dramatic.. However, asymptomatic infants are likely to continue to go unrecognized and undiagnosed.A timely diagnosis of cCMV is critical for access to early interventions, which may include antiviral medication, hearing monitoring and amplification as well as developmental therapies [3,8,9]. Given the clinical challenge of and narrow window for diagnosing cCMV, enthusiasm for neonatal cCMV screening programs has grown, but at present they are only implemented in a handful of states [10]. To date, little is known about US healthcare providers’ practices surrounding cCMV testing in infants. While neonatal cCMV screening programs have yet to be endorsed by the American Academy of Pediatrics, a better understanding of cCMV testing practices when based on clinical suspicion alone, including the indications that prompt providers to test for cCMV, may inform the need for and the content of future guidelines.

Therefore, this study sought to examine cCMV testing practices at a quaternary US children’s hospital based on clinical suspicion alone over a six-year period. Specifically, the objectives of this study were to examine: (1) the circumstances surrounding each cCMV test ordered (number tested, test result, type of testing modality, clinical unit that ordered the testing, and the indications for testing); (2) the differences in the sociodemographic and clinical characteristics of infants based on whether they were tested for cCMV and the test results; and finally, (3) the prevalence of cCMV diagnoses made in the study population as compared to the true prevalence of cCMV based of epidemiologic estimates. This study focused on infants less than or equal to 31 days old, and who were all cared for at the same hospital, which did not have a systematic cCMV screening program in place.

3. ResultsA flowchart of infants tested for cCMV based on the ordering unit, testing indications, and results is presented in Figure 1. Of the 178 infants tested for cCMV, 12 had positive results confirming a diagnosis of cCMV (percent positive 6.7%, 95% CI 3.9–11.4), only 10 of whom were identified based on the clinical suspicion of providers at the study institution (percent positive 5.6%, 95% CI 3.1–10.0). Two of the 12 infants diagnosed with cCMV were transferred from a referring institution’s Neonatal Intensive Care Unit (NICU) to the study institution’s NICU secondary to an escalation of acuity and/or need for sub-specialty consultation due to a suspicion for congenital infections. At the time of transfer both infants had dedicated CMV testing at the referring institution. As these results were not yet available at the time of admission, the study institution ordered additional CMV PCR testing (all tests eventually came back positive). As such, we consider these two infants as being identified by their referring institution. All 12 infants had “congenital CMV” or “congenital CMV infection” associated with International Classification of Diseases 9 or 10 codes (771.1 and/or P35.1) [21,22] in their clinical records.Quantitative CMV PCR (urine or plasma/serum) was the most used testing modality (142/178) as depicted in Figure 2. However, early in the study period, viral culture was more commonly ordered and then it phased out to exclusive PCR testing by 2016. Antibody testing was used in five infants in 2014, and once in 2018. Of note, CMV antibody testing is not recommended for congenital CMV, as maternal antibodies are found in the neonatal circulation. Dried blood spot or CSF CMV PCR were not used to test any infant in the study sample.

Infants tested for cCMV were cared for by five different units/divisions: the NICU, Newborn Nursery, Pediatric Intensive Care Unit (PICU), Pediatric Cardiac Intensive Care Unit (PCTU) and the inpatient general care Pediatric service. Infants cared for by the NICU represented the majority (72%) of those tested for cCMV. The percent positive for tests ordered by the NICU was 2.3% (95% CI 0.8–6.7) or 0.8% (95% CI 0.14–4.4) after the removal of the two infants transferred with pending cCMV testing. In comparison, the percent positive of the tests ordered by providers in a non-ICU setting was 27.6% (95% CI 14.7–45.7) and 7.7% (95% CI 1.4–33.3) for the Newborn Nursery and Pediatric general care units, respectively.

The clinical indications documented by healthcare providers that prompted testing are displayed in Figure 3. Small-for-gestational age (SGA), defined as a birthweight for sex 12,13] (53/178) and intracranial abnormalities (15/178) were the most documented testing indications. A failed newborn hearing screen only prompted five providers to order cCMV testing. While SGA was the most common indication that prompted testing, it only yielded three positive cCMV diagnoses, all of which were cared for by providers in the Newborn Nursery, as illustrated in Figure 1. Prenatal concern for CMV due to positive maternal serologies was the testing indication with the highest yield for positive diagnoses (4/12).The sociodemographic characteristics of the cohort are presented in Table 1a. Most infants tested for cCMV were male (54.5%) and of Caucasian race (66.9%); these proportions did not statistically differ from the larger cohort. Infants tested for cCMV disproportionately had public (vs. private or other) insurance as compared to those not tested for cCMV (ppThe birth characteristics of the cohort are shown in Table 1b. Infants tested for cCMV were more likely to have been inborn (vs. at an outside hospital), born preterm (vs. term), and have a lower birthweight and head circumference percentile for gestational age, than those not tested. Infants who tested positive for cCMV (vs. negative) were less likely to be born premature and had a lower mean birthweight percentile for gestational age (17.0%ile vs. 31.5%ile, ppp3].

The overall cCMV testing rate in this study was 4.5 tests per 1000 infants, with a resulting prevalence rate of diagnosed cCMV of 0.32 cases per 1000 infants cared for at the study institution. However, 2 of the 12 infants were transferred to the NICU with cCMV testing in process due to the referring provider’s clinical suspicion for congenital infection, resulting in only 10 infants with cCMV whose testing was initiated based on the study provider’s clinical suspicion. As such, the prevalence rate of diagnosed cCMV based on the clinical suspicion of providers at the study institution was 10/40,091 or 0.2 cases per 1000 infants.

Reported estimates of the prevalence of cCMV in the US range from 3 to 6 cases per 1000 live births [1,2]. Using a median estimate of 4.5 cases per 1000 live births, 180 cases of cCMV, including 18 symptomatic and 162 asymptomatic infants, would be expected among the cohort of 40,091 infants. With those estimates in mind, the rate of cCMV diagnosis in this study was 15-fold, 2-fold and 81-fold lower than the expected prevalence of all cCMV, symptomatic cCMV, and asymptomatic cCMV infections, respectively. 4. DiscussionThis study found a low rate of cCMV testing and diagnosis in a cohort of over 40,000 infants cared for in a quaternary US children’s hospital without a systematic cCMV screening program in place. The indications that prompted testing were multifarious. While isolated SGA was the most common single indication for testing, positive maternal prenatal CMV serologies was the most common indication that was associated with a positive result. Of the infants found to have cCMV with testing prompted by positive maternal prenatal serologies, two of the four infants had no clinical signs of disease at birth and would have likely otherwise gone undiagnosed. As cCMV testing was infrequent, the diagnostic prevalence of cCMV in this cohort was low. The prevalence of cCMV in the United States is estimated to be 4.5 per 1000 live births [1]; this present study found the diagnosed prevalence to be 0.3 per 1000 live births in total (including outside hospital transfers for likely cCMV), or 0.2 per 1000 live births based on clinical suspicion of the providers at the study institution alone. Our findings of low rates of cCMV testing and diagnosis echo those of prior work [6], and it is likely that most cases of cCMV go undiagnosed when a systematic cCMV testing or screening program is not in place [6,23,24,25]. We hypothesize that the low testing rate may be due to low cCMV awareness among clinicians, and the often subtle or inapparent clinical presentation of cCMV [26,27,28].In this study, there did not seem to be a uniform or best practice in terms of clinical indications for cCMV testing. While isolated SGA was the most documented testing indication, it did not have a high yield association with a positive cCMV result. Providers in the NICU predominantly tested infants based on this indication and yielded zero positive cases, whereas the Newborn Nursery tested far less and yielded three positive cases. While this may raise the question about the utility of testing for cCMV based on a clinical indication of SGA in the NICU, it must be noted that not all infants with SGA cared for in the NICU were tested for cCMV, and as such caution must be used in generalizing these results beyond the tested cohort. Our findings parallel those of Smiljkovic et al. [29], who reported screening indications for children diagnosed with cCMV by retrospective review of the EHR in Quebec, Canada. Smiljkovic et al. reported infant clinical, laboratory or imaging findings concerning for cCMV, which they termed “symptomatic infant” as being the most common cCMV testing indication (51%) among 47 infants with cCMV, followed by maternal serologies [29]. Three quarters of the infants with cCMV in our study were tested due to clinical, laboratory or imaging findings, therefore 75% of the primary screening indications were due to being a “symptomatic infant”, and 25% were due to maternal serologies. Although the definitions used for specific clinical findings varied between studies, our findings regarding the proportion of infants with cCMV that had specific clinical signs were largely similar to prior studies using review of the clinical medical record: evidence of CNS involvement (75% vs. 52–58%), intrauterine growth restriction/SGA (33% vs. 38–48%) and thrombocytopenia (50% vs. 36%) [29,30,31]. The proportion of infants with cCMV with these clinical signs in our study is higher than estimates based on administrative claims data. For example, microcephaly was found in 58% of infants with cCMV in our study vs. 6–11% in two studies of administrative data reliant on diagnostic codes [14,32]. It is likely that administrative claims analyses underestimate the prevalence of clinical signs, as they are reliant on the provider documenting the sign as a separate diagnosis, rather than in the clinical note alone [33].Prenatal maternal serologies accounted for 25% of the indications for cCMV testing which were associated with a positive cCMV diagnosis. This may indicate that prenatal serologies could be important for cCMV detection at birth. At present, routine prenatal CMV serologic testing is not recommended by US professional organizations due to the challenge of interpreting results [34]. In other countries, especially where seroprevalence is low, prenatal CMV serologic testing is common practice in order to monitor for maternal seroconversion throughout the pregnancy as a marker of cCMV risk [35,36,37,38]. While the risk of a screening test in pregnancy should not be minimized [39], especially in the absence of clinically proven prenatal treatments for cCMV, maternal serologic monitoring may provide an opportunity for targeted neonatal testing and early diagnosis. This is an important area of future research as cCMV screening programs gain traction.Of those infants tested for cCMV in our study, many had non-specific clinical findings that can also be seen in symptomatic cCMV (e.g., microcephaly, petechiae, small-for-gestational age). As a result, infants with symptomatic disease made up the majority (83%) of those diagnosed with cCMV in the cohort, as opposed to the minority (~10%) as has been found in population-based studies [2,40]. It is likely that provider’s clinical suspicion for cCMV is not raised in the case of asymptomatic infections, due to the absence of visible findings, and therefore these infants are even more likely to go undiagnosed. Of note, although not visible, a failed newborn hearing screen is a finding that clinical providers are made aware of, but this rarely prompted cCMV testing in the study cohort. It may be that some clinical providers lack awareness about the prevalence and presentation of cCMV, as has been suggested by recent work [28]. Infants with cCMV who go undiagnosed likely miss opportunities for early interventions including close auditory follow-up, treatment of hearing loss and developmental surveillance [3,8,41,42,43].The results from this study highlight the critical need for systematic cCMV screening and testing programs to diagnose most infants with cCMV. Examples of such programs include hearing-targeted cCMV testing, in which all infants who fail their newborn hearing screening are then tested for cCMV, and universal/routine screening programs, in which all infants are screened for cCMV regardless of apparent risk factors [40,44,45]. Such programs have been found to be effective and acceptable in several studies [46,47,48,49,50], although the debate about cost-effectiveness persists [51]. As a more nuanced understanding is being gained about the possible long-term outcomes of cCMV [43,50,52], as well as the effectiveness of early interventions [53,54,55], the risk–benefit analysis may be more clearly weighted towards systematic screening.

This study is not without limitations. In line with the design of a retrospective study, data collection relied on the identification of infants who were tested for CMV and the extraction of clinically significant variables of interest. The identification of clinical indications for testing and clinical findings relied on the documentation by medical providers in the clinical record for the group tested for CMV, which may not accurately or completely capture the physical exam findings. For the study cohort, the provider’s documentation was often vague, allowing for potential selection and misclassification bias, especially for the clinical indication that prompted CMV testing. Furthermore, this study used an electronic query tool to create the cohort, the definitions from which have not been validated. It may be that some infants tested for cCMV were missed. In the larger cohort, ICD-9 and ICD-10 diagnostic codes were used to identify infants with clinical features, which likely underestimates the presence of such signs for the larger cohort. Study results may not be generalizable to other healthcare settings or geographic areas.