Hearing-Targeted Congenital Cytomegalovirus Screening: An 11-Year Experience

Study Design and Setting

We employed an interrupted time series (ITS) design to compare the cCMV diagnoses rates before and after the implementation of hearing-targeted cCMV surveillance within a large health care delivery network in Connecticut. This quasi-experimental design allowed us to assess the impact of a population-level policy change by examining changes in trends over time. The study was conducted within the Yale New Haven Health System (YNHHS), a health care network that includes several hospitals with obstetric delivery services, major maternity centers, and over 130 ambulatory sites, all connected through a single electronic health record (EHR) platform. The YNHHS primary service area is a 55-town region in southern Connecticut that includes 1.9 million people, representing approximately 60% of the state’s population (Figure S1). The Institutional Review Board at the Yale School of Medicine approved the study and waived the requirement for informed consent (Human Investigation Committee: 2000027310). This study follows the Strengthening the Reporting of Observational Studies in Epidemiology reporting guidelines for observational studies.

Study Population

The source population included all patients who received care at YNHHS and were born within its primary service area between January 1, 2013, and December 31, 2023. Limiting inclusion to this 55-town geographic area ensured a well-defined catchment population and enabled calculation of population-based diagnosis rates and trends.

Case Ascertainment and Definition

Potential cCMV cases were identified through a combination of laboratory results and diagnostic codes. Laboratory criteria for case ascertainment included the detection of CMV using polymerase chain reaction (PCR) or culture from urine, saliva, whole blood, or cerebrospinal fluid (CSF). We also used International Classification of Diseases, Ninth Revision and Tenth Revision (ICD-9/10) and Systematized Nomenclature of Medicine Clinical Terms (SNOMED-CT) codes as supplementary data sources for identifying potential cCMV cases for manual review (Table S1).

Medical records for all potential cases were manually reviewed, and data were abstracted using a standardized form, including initial clinical findings, treatments administered, and subsequent audiologic and developmental outcomes. Cases were classified using the 2024 Council of State and Territorial Epidemiologists (CSTE) cCMV case definitions.²² Patients were defined as having had a “confirmed cCMV infection” if CMV was detected by PCR or culture from urine, whole blood, or CSF samples collected within 21 days of life. Infants who had a confirmed cCMV infection plus clinical findings compatible with cCMV were classified as “confirmed cCMV disease.” Clinical findings compatible with cCMV included microcephaly, hepatosplenomegaly, petechiae or purpura, SNHL, visual or neurologic impairments, or neuroimaging abnormalities. Infants who had compatible clinical findings but only presumptive laboratory evidence of cCMV infection (eg, testing performed between 22–42 days of life or saliva testing without urine or blood confirmation) were classified as “probable cCMV disease.”

Clinical Severity and Long-Term Outcomes

We classified disease severity using the American Academy of Pediatrics Red Book definitions.²³ Briefly, “asymptomatic” infection was defined as having normal hearing (based on diagnostic audiological evaluations) and no abnormalities on physical examination, neuroimaging, or laboratory tests. Infants were classified as having cCMV with “isolated SNHL” if their only attributable clinical sign was abnormal hearing. We defined “mildly symptomatic” as having 2 or fewer clinical manifestations (eg, thrombocytopenia, petechiae, or mild hepatitis). Cases with “moderately to severely symptomatic” cCMV included those with multiple (>2) clinical manifestations, severe organ dysfunction, or neurologic involvement (eg, microcephaly, seizures, or intracerebral calcifications). Further details on our study definitions can be found in Supplementary Table S1.

Long-term outcomes were followed through June 30, 2025. Neurodevelopmental outcomes, such as developmental delays and neurologic impairments, were identified from EHR problem lists, diagnostic codes, and clinician notes. All available audiology reports were reviewed to document the presence, laterality, severity, and progression of SNHL. Hearing loss was classified as mild (26–40 dB), moderate (41–70 dB), severe (71–90 dB), or profound (>90 dB) based on thresholds in the more severely affected ear.²⁴ In instances in which specific audiometric thresholds were not reported, the audiologist’s qualitative classification (ie, mild, moderate, severe, or profound) was used. For the subset of infants with follow-up hearing evaluations after 12 months of age, we compared their baseline result (within the first 6 months of life) to their most recent evaluation to classify hearing progression as improved, stable, or worsened. We also documented the need for hearing aids or cochlear implants.

Screening Procedures

Before January 1, 2016, cCMV testing at YNHHS was performed at the discretion of individual health care providers. Following the implementation of Connecticut’s hearing-targeted cCMV screening law, standardized protocols were adopted. In addition to testing based on clinical suspicion, all infants who did not pass the NBHS underwent cCMV testing using saliva or urine. A positive result triggered a referral to an infectious disease specialist for a comprehensive diagnostic evaluation, which included confirmatory CMV testing (blood or urine PCR), bloodwork, a head ultrasonography, and referrals for ophthalmologic and audiologic assessments. The complete screening algorithm is shown in Figure S2.

Statistical Analysis

Patient characteristics were summarized using medians and IQRs for continuous variables and frequencies with percentages for categorical variables. Between-group differences were assessed using standardized mean differences (SMDs), defined as the difference in group means divided by the pooled SD. We considered absolute SMDs greater than 0.25 to indicate meaningful imbalance.²⁵

Because screening policy is expected to affect ascertainment rather than true infection prevalence, our primary outcome was the rate of documented diagnoses. To assess the impact of the hearing-targeted screening policy on annual cCMV diagnosis rates, we fit an ITS model for annual case counts (1 series; 11 years) using Poisson regression, with a log link, and an autoregressive correlation structure. The ITS model included an indicator for the postpolicy period to estimate the immediate level change and an interaction between time and the postpolicy indicator to estimate the change in trend. Additional terms were included to capture analogous level and trend changes associated with the COVID-19 pandemic. The logarithm of annual births for the catchment area, derived from official town birth records, was included as an offset to directly model diagnosis rates. The model coefficients were exponentiated to produce incidence rate ratios (IRRs) with corresponding 95% CIs. In a sensitivity analysis, we refit the ITS using Prais-Winsten regression to assess whether conclusions were robust to an alternative outcome scale (counts vs log rate) and treatment of serial correlation (working correlation vs autoregressive errors). A 2-sided P value less than 0.05 was considered statistically significant for all analyses. All analyses were conducted using StataMP version 18 (StataCorp LLC, College Station, TX).

Study Population

Between January 1, 2013, and December 31, 2023, there were 197 177 live births in the catchment area. Seventy-two infants were identified as potential cases based on electronic search criteria (diagnostic codes or laboratory results). Following manual review, 49 infants met the CSTE definition for confirmed cCMV infection or probable cCMV disease and were included in the final study cohort (Figure 1).

Clinical Characteristics and Long-Term Outcomes

The demographic and clinical characteristics of the 49 infants with confirmed or probable cCMV are detailed in Table 1. The cohort was 45% girls, and 47% were Non-Hispanic white. After a complete diagnostic evaluation, 32 of 49 (65%) infants were classified as moderately to severely symptomatic, 2 of 49 (4%) as mildly symptomatic, 4 of 49 (8%) as isolated SNHL, 9 of 49 (18%) as asymptomatic, and 2 of 49 (4%) had insufficient data for classification (Figure S3). The most common findings at birth were small for gestational age or intrauterine growth restriction (25 of 49; 51%), microcephaly (14 of 49; 29%), abnormal muscle tone (11 of 49; 22%), and petechiae (8 of 49; 16%). Laboratory abnormalities included thrombocytopenia (19 of 49; 39%) and hepatitis (12 of 49; 24%). Abnormal neuroimaging was documented in 20 of 49 (41%) infants. Cystic lesions were the most common abnormality (13 of 20; 65%), frequently involving the choroid plexus, caudothalamic groove, or periventricular regions. Other common findings included lenticulostriate vasculopathy or mineralizing angiopathy (6 of 20; 30%) and intracranial calcifications (4 of 20; 20%). Intracranial hemorrhage was identified in 5 of 20 (25%) cases, all occurring in premature infants. SNHL at birth was documented in 16 of 49 (33%), with 9 of 16 (56%) having severe or profound hearing loss.

Antiviral therapy was given to 35 of 49 (71%) infants, 80% (28 of 35) of whom had moderate to severe symptomatic disease. Among treated infants, 27 of 35 (77%) completed the planned 6-month course. Treatment-associated adverse events were documented in 11 of 35 (31%), most commonly neutropenia (9 of 11; 82%). Fourteen infants did not receive antivirals; most (10 of 14; 71%) were not treated because they were either asymptomatic, mildly symptomatic, or had isolated SNHL. One infant was not treated because they died within the first few days of life. One infant with moderate to severe disease presented later in infancy and was not treated because available evidence did not clearly support benefit for treatment initiation at that age. For 2 infants, the reason antivirals were not offered was not documented.

Follow-up beyond 12 months was documented for 42 of 49 (86%) children, with a median follow-up duration of 72 months (IQR, 41–88 months). At the most recent audiologic assessment, 18 of 42 (43%) had SNHL, of which 2 (11%) presented as delayed-onset hearing loss. Among the 16 infants with SNHL at the initial diagnostic evaluation, hearing worsened in 13 (81%), remained stable in 2 (13%), and improved in 1 (6%) (Figure 2). All children with SNHL eventually required hearing aids or underwent cochlear implantation. At the end of the follow-up period, only 9 of 42 children (21%) had no sequelae attributable to cCMV. Among the 33 children with sequelae, deficits most often involved speech and language or motor function domains (Table S2).

Impact of Targeted Screening on Detection Rates

The estimated effects of the 2016 targeted screening law on annual cCMV diagnosis rates are shown in Figure 3. Before implementation of Connecticut’s 2016 hearing-targeted screening law, the mean annual cCMV diagnosis rate was 1.08 per 10 000 births (95% CI, 0.22–1.95), with no significant change year-to-year (IRR, 0.92; 95% CI, 0.49–1.74; P = .80). After the 2016 law, there was an immediate and statistically significant increase in the rate of cCMV (IRR, 4.32; 95% CI, 1.09–17.01; P = .037). The trend of new diagnoses remained stable from 2016 to 2019, averaging 4.23 new diagnoses per 10 000 births per year (95% CI, 2.72–5.74). In 2020, rates dropped relative to the postlaw mean (IRR, 0.29; 95% CI, 0.18–0.50; P = .001). Notably, the interval between birth and NBHS remained stable throughout the study period (median age, 1 day; IQR, 1–2) (Table S3). After 2020, the trend reversed and began increasing toward prepandemic levels (IRR, 1.27; 95% CI, 1.05–1.53; P = .013). Accounting for this transient decrease, the mean postpolicy cCMV diagnosis rate was 3.03 per 10 000 births (95% CI, 2.13–3.94). On sensitivity analysis, results from the primary ITS models were robust across alternative modeling specifications, with consistent direction, magnitude, and statistical inference. The only exception was the post-2020 slope change, for which statistical significance was not consistent across models (Table S4).

Case ascertainment in the postpolicy period reflected both hearing-targeted screening and clinician-initiated testing. Postpolicy, 42% (18 of 43) of cCMV cases were infants who passed their NBHS and were tested based on clinical suspicion or a known maternal CMV history. Additionally, among the 22 infants who did not pass the NBHS postpolicy, 5 (23%) had CMV testing ordered before the hearing results were known.

The spectrum of disease severity also shifted after implementation of targeted screening (Figure 4). Before 2016, none of the detected cases were asymptomatic at birth. In the postpolicy period, 12 of 43 (28%) cases identified were asymptomatic or had isolated SNHL, of which 75% later developed significant sequelae.