Abstract
Objective:
Administrative claims databases are used to study the care of congenital cytomegalovirus (cCMV), yet the use of International Classification of Diseases, Ninth and Tenth Revision, Clinical Modification (ICD-9/10-CM) codes for cCMV have not been validated. This study examines the accuracy of ICD-based codes for cCMV infection.
Methods:
The study population included infants cared for at a quaternary children’s hospital (2013–2023) that had an ICD-based diagnosis for cCMV or CMV Infection at ≤90 days of age. Medical record data was abstracted, including demographics and evidence of cCMV. True Positive cases were defined as those with an ICD code AND clinical and laboratory evidence consistent with a cCMV infection. False Positive cases were defined as those with an ICD code without evidence of a cCMV infection. Positive predictive value (PPV) and sensitivity for each diagnostic code at different age cutoffs were calculated within the cohort. Multinomial regression examined characteristics of the infant with odds of being a True Positive case of cCMV.
Results:
Of the 108 infants with ICD-9/10 codes for cCMV, 35% were false positives. PPV for ICD-9/10-CM codes for cCMV, CMV Infection, and Either code predicting actual cCMV were 0.86, 0.36, and 0.68 at age ≤45 days. PPV was the highest at ≤21 days of age, and for all codes sensitivity increased with patient age. Multinomial logistic regression found the age of the first diagnostic code ≤21 days (vs. >) (OR=4.11, 95% CI 1.45–12.03), having an ICD-9/10-CM diagnostic code of cCMV (vs. CMV Infection) (OR=10.87, 95% CI 3.64–32.47), and having Clinical Signs at Birth (vs. none) (OR=8.4, 95% CI 2.72–25.81) to be associated with greater odds of having a True Positive case of cCMV (vs. Not cCMV).
Conclusions:
Administrative claims case definitions for cCMV were more likely to be accurate when assigned at a younger age. Studies using case definitions for cCMV that include the presence of codes for either cCMV or CMV Infection may be biased given the high proportion of false positives demonstrated in this study.
Keywords: Cytomegalovirus, cytomegalovirus infections, administrative claims, healthcare, International Classification of Diseases, congenital abnormalities, virology
INTRODUCTION
Congenital cytomegalovirus (cCMV) affects 1 in 200 United States infants, yet is infrequently diagnosed, save for the most severe cases, owing to its often silent or subtle clinical presentation[1,2]. In the absence of a routine newborn screening program, studies estimate that only 1–10% of infants with cCMV are diagnosed in infancy[1,3]. Those born with clinical signs of the infection, also known as symptomatic cCMV or cCMV disease, are at heightened risk of long-term neurodevelopmental delays and disabilities and are thought to generally have a much different clinical course than those born without signs or findings on clinical exam, imaging, or laboratory studies (asymptomatic cCMV or cCMV infection) [4–7]. Many studies requiring larger cohorts to examine the clinical course of cCMV and healthcare utilization have relied on administrative claims data [8–15]. using the International Classification of Diseases, Ninth and Tenth Revision, Clinical Modification (ICD-9/10-CM) diagnostic codes to identify cases [8–12]. Yet the accuracy of diagnostic codes for cCMV representing true cases of cCMV remains unknown. For instance, many studies have defined a congenital CMV case as having a diagnostic code for Cytomegalovirus Infection (ICD-9-CMV 078.5, ICD-10-CM B25) or Congenital Cytomegalovirus (ICD-9-CMV 0771.1, ICD-10-CM P35.1) within 45 days or 60 days of birth[8–11,17,18]. However, an infant with a postnatally acquired CMV infection, such as an infant exposed to CMV secreted in mother’s breastmilk, could meet this case definition [19]. Prior work has called for a formal validation of diagnostic codes for cCMV [16], which in general is important for interpreting the results of studies using administrative claims data [20].
Therefore, the objective of this study was to examine the accuracy of ICD-based diagnostic codes for cCMV at a large midwestern health system.
METHODS
Study overview.
We conducted a retrospective review of electronic health records (EHR) of a quaternary children’s hospital and associated health system in the midwestern US. Eligible patients were identified using DataDirect [21], a tool enabling access to clinical data such as diagnoses and encounters on more than 4 million unique patients from across the health system. The study population included infants who were ≤90 days of age from January 1, 2013, to November 1, 2023, with an ICD-9/10-CM diagnostic code corresponding to Cytomegalovirus Infection (ICD-9-CM 078.5, ICD-10-CM B25) or Congenital Cytomegalovirus (ICD-9-CM 0771.1, ICD-10-CM P35.1) within 90 days of birth. These ICD-9/10-CM codes have been used in prior literature to define an administrative claims diagnosis of cCMV up to 60 days of age [8–11,17,18]. The age limitation of ≤90 days was chosen to be more inclusive of infants who may be diagnosed late, as is common with cCMV.
Variables collected included ICD -9/10-CM code on the first encounter it was assigned to a patient, the age at which the diagnostic code was assigned, subject identifier, and date of birth. The authors manually reviewed the EHR of each patient using a standardized protocol with definitions for each variable extracted. The definitions were refined through an iterative process by which the reviewers independently coded a subset of 10 randomly selected records, then compared codes and refined the definitions over a course of three meetings until reliability was established (Cohen’s kappa >0.7 for all variables). Missing data was common given varying clinical circumstances. For instance, not all infants received head imaging or laboratory studies, as these may not have been clinically appropriate. Denominators were adjusted when a subset of the sample was being analyzed (e.g. the number of infants with any head imaging who had abnormal findings). For bivariate variables examining the presence vs. absence of a specific test, if the test result could not be located in the chart, it was recorded as absent. For physical exam findings, if a specific sign was described it was marked as being present, however if a specific sign was not mentioned or if it was explicitly noted as being absent, in both cases it was coded as being absent.
Definitions
Clinical Signs at Birth was determined after manual review of the infants’ documented physical exam, laboratory and imaging studies, and pediatric subspecialty consultants’ examination. Infants with physical signs or findings on studies at birth consistent with possible cCMV were categorized as having Clinical Signs at Birth, based on the criteria for moderate to severe symptomatic cCMV in the International Consensus Statement by Rawlinson et al [6]. This included the presence of any of the following: intrauterine growth restriction (IUGR) or small-for-gestational age (SGA), microcephaly, hepatomegaly, splenomegaly, seizures, petechiae or purpura, jaundice from direct hyperbilirubinemia, thrombocytopenia, transaminitis, chorioretinitis, or brain abnormalities. Isolated and otherwise mild or transient findings such as SGA, hearing loss, physiologic jaundice, and borderline elevation in liver enzymes were not categorized as features of symptomatic cCMV. Infants without findings on physical exam, laboratory or imaging studies consistent with cCMV end organ damage as mentioned above were categorized as No Clinical Signs at Birth.
Reason for CMV testing was determined after manual review of the ordering clinicians’ documentation to determine the rationale for ordering CMV testing or using a cCMV or CMV Infection ICD-9/10-CM diagnostic code. Primary reasons included: prenatal concerns (e.g. abnormal fetal imaging, known maternal exposure), preterm infant, IUGR or SGA, multiple clinical signs at birth, failed newborn hearing screen or sensorineural hearing loss (SNHL), abnormal CMV-related lab (e.g. thrombocytopenia, transaminitis), abnormal brain imaging, or other condition (e.g., respiratory issues, sepsis, transplant evaluation.)
Laboratory evidence of cCMV.
The chart was manually reviewed for all CMV-related laboratory testing performed, including CMV polymerase chain reaction (PCR) of saliva, urine, serum or dried blood spot (DBS), urine viral shell culture and CMV IgM antibody. The date of each test and results were recorded.
Confirmatory laboratory evidence of cCMV was defined as CMV DNA in the infant’s bodily fluids (serum, saliva, urine or DBS) by PCR or viral culture within 21 days of age, AND an absence of negative CMV tests. Inclusion of an expansive list of CMV testing beyond the gold standard of urine and saliva accounted for testing obtained outside the primary hospital system where screening protocols, or lack thereof, may vary. Our laboratory criteria for confirming a case of congenital CMV was similar to the Council of State and Territorial Epidemiologists (CSTE) case definition[7,22], with the exception that we also accepted CMV DNA isolated from saliva PCR without repeat testing with a secondary sample in the setting of clinical symptoms being present. Laboratory evidence of CMV DNA isolated in an infant’s bodily fluids only on or after 22 days of age with features consistent with cCMV (e.g., clinical findings at birth, SNHL, brain abnormalities), but without a confirmatory DBS PCR was defined as Presumptive laboratory evidence.
IgM antibodies against CMV alone without any subsequent CMV DNA isolation was not considered to be either Confirmatory or Presumptive evidence of cCMV.
Final Disease Status represents whether an infant was found to have laboratory evidence confirming a cCMV infection on review of the EHR. A Final Disease Status of cCMV was assigned if the infant had Confirmatory or Presumptive laboratory evidence of cCMV. A Final Disease Status of Not cCMV was assigned if the infant had no laboratory evidence of cCMV (e.g., cCMV was ruled out, or no testing was ordered).
Statistical analysis.
Data analysis was performed using IBM SPSS Statistics Version 28.0 [23]. Descriptive statistics based on sociodemographic and clinical features were calculated. The proportion of cases with ICD-9/10-CM codes for cCMV vs. CMV Infection vs. Either and those with a Final Disease Status of cCMV vs. Not were calculated across these classifications.
Definitions
True Positive was defined as an infant with a Final Disease Status of cCMV in the setting of an ICD-9/10-CM code corresponding to cCMV or CMV Infection.
False Positive was defined as an infant with a Final Disease Status of Not cCMV in the setting of an ICD-9/10-CM code corresponding to cCMV or CMV Infection.
Positive predictive value (PPV) was defined as the number of True Positive cases / True Positives + False Positives for each period and a particular ICD-9/10-CM diagnostic code category (cCMV, CMV Infection, or Either).
Sensitivity was defined as the number of cases with a Final Disease Status of cCMV and a particular ICD-9/10-CM diagnostic code category by a set period divided by the total cases with a Final Disease Status of cCMV and a particular ICD-9/10-CM diagnostic code category (cCMV, CMV Infection, or Either).
PPV and sensitivity within our cohort (not at a population level) were calculated for an ICD-9/10-CM diagnostic code category of cCMV, CMV Infection, or Either by age at which the diagnostic code was issued: ≤21 days, ≤45 days, ≤60 days, and ≤90 days of age.
A stepwise multinomial logistic regression was created to examine predictor variables associated with a Final Disease Status of cCMV (vs. Not). First, basic demographic and clinical variables were modeled, and then categories of age at first diagnostic code were added to the model. Next, the ICD-9/10-CM diagnostic code category was added, and finally, covariates related to clinical findings were added.
RESULTS
Of the 102,217 infants aged ≤90 days cared for in the health system during the study period, 108 cases met inclusion criteria (i.e., with ICD-9/10-CM diagnostic codes for CMV Infection or cCMV). Of these, 56% (60) had an ICD-9/10-CM diagnostic code for cCMV and 44% (48) had an ICD-9/10-CM diagnostic code for CMV Infection (Table 1). Only three infants had ICD diagnostic codes for both cCMV and CMV Infection. Most (78%) of the cohort had an initial Reason for CMV Testing due to Concern for cCMV (versus postnatal CMV, CMV-related sepsis, etc.) and just over half (58%) of those had a Final Disease Status of cCMV, representing the number of True Positive cases in the cohort. Conversely, percentage of False Positive cases was 42% using inclusion criteria of ≤90 days, which improved to 35% and 29% when restricting age to <60 and ≤21 days, respectively.
Table 1.
Reason for cytomegalovirus testing and Final Disease Status by diagnostic code, N=108
| Any cCMV/ CMV Infection ICD-9/10-CM Diagnostic Code n (%) |
ICD-9/10-CM Diagnostic Code of Congenital CMV n (%) |
ICD-9/10-CM Diagnostic Code of CMV Infection n (%) |
||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Age in days at first dx code | ≤90 N=108 |
≤60 N=85 |
≤45 N=78 |
≤21 N=68 |
≤90 N=60 |
≤60 N=52 |
≤45 N=50 |
≤21 N=44 |
≤90 N=48 |
≤60 N=33 |
≤45 N=28 |
≤21 N=24 |
|
| ||||||||||||
| Reason for CMV testing | ||||||||||||
| Concern for cCMV | 84 (77.8) | 71 (83.5) | 69 (88.5) | 62 (95.4) | 55 (93.2) | 49 (94.2) | 48 (96.0) | 43 (97.7) | 29 (60.4) | 22 (66.7) | 21 (75.0) | 19 (86.4) |
| Concern for postnatal CMV | 14 (13.0) | 8 (9.4) | 3 (3.8) | 2 (3.1) | 1 (1.7) | 1 (1.9) | 0 (0.0) | 0 (0.0) | 13 (29.5) | 7 (21.2) | 3 (10.7) | 2 (9.1) |
| Acute care /Sepsis | 5 (4.6) | 2 (2.2) | 2 (1.6) | 1 (1.5) | 3 (5.1) | 1 (1.9) | 1 (2.0) | 0 (0.0) | 2 (4.5) | 1 (3.0) | 1 (3.6) | 1 (4.5) |
| Other/Unclear | 5 (4.6) | 4 (4.7) | 4 (5.1) | 3 (4.4) | (0.0) | 1 (1.9) | 1 (2.0) | 1 (2.3) | (0.0) | 3 (9.1) | 3 (10.7) | 2 (8.3) |
|
| ||||||||||||
| Final Disease Status | ||||||||||||
| cCMV | 63 (58.3) | 55 (64.7) | 53 (67.9) | 48 (70.6) | 49 (81.7) | 44 (84.6) | 43 (86.0) | 38 (86.4) | 14 (29.2) | 11 (33.3) | 10 (35.7) | 10 (41.7) |
| Not cCMV | 45 (41.7) | 30 (35.3) | 25 (32.1) | 20 (29.4) | 11 (18.3) | 8 (15.4) | 7 (14.0) | 6 (13.6) | 34 (70.8) | 22 (66.7) | 18 (64.3) | 14 (58.6) |
Abbreviations: ICD-9/10-CM -International Classification of Disease 9th and 10th Edition Clinical Modification; cCMV -congenital cytomegalovirus; CMV -cytomegalovirus.
Characteristics of the cohort by ICD-9/10-CM diagnostic code category and Final Disease Status are shown in Tables 2a–c. The cohort was 50% female, mostly White (64%) and/or non-Hispanic (95%), mean gestational age at birth of 36 4/7 weeks (SD 5.6 weeks) and less than half (37%) were inborn. Nine infants (8%) were deceased at the time of chart review (median age of death = 100 days). Most infants were cared for by the newborn nursery/well baby service (36%) or the neonatal intensive care unit (37%). The mean age at first diagnostic code was 26.3 days (SD = 29.5 days). The number of total administrative diagnoses made in the health system each year increased over the study period (Figure 1).
Table 2a.
Demographic and clinical characteristics (N= 108) by ICD-9/10-CM Diagnostic Code Category and Final Disease Status
| ICD-9/10 Diagnostic Code of cCMV n (%) |
ICD-9/10 Diagnostic Code of CMV Infection n (%) |
Final Disease Status of cCMV n (%) |
Final Disease Status of Not cCMV n (%) |
Any cCMV/ or CMV Infection ICD-9/10-CM Diagnostic Code n (%) | |
|---|---|---|---|---|---|
| N=60 | N=48 | N=63 | N=45 | N=108 | |
| Infant is female | 30 (50.0) | 24 (50.0) | 30 (47.6) | 24 (53.3) | 54 (50.0) |
|
| |||||
| Infant is deceased at time of chart review | 4 (6.7) | 5 (10.6) | 3 (4.8) | 6 (13.6) | 9 (8.4) |
|
| |||||
| If deceased, median age of death (Q1, Q3); days | 23.5 (19.0, 250.8) | 103 (54.4, 174.5) | 22.0 (9, .) | 128.0 (79.5, 228.5) | 100.0 (20.0, 174.5) |
|
| |||||
| Primary race | |||||
| Black | 10 (16.9) | 5 (10.4) | 11 (17.7) | 4 (8.9) | 15 (14.0) |
| White | 42 (71.2) | 27 (56.3)* | 45 (72.6) | 24 (53.3)** | 69 (64.5) |
| Declined/Unknown/Unlisted | 2 (3.4) | 7 (14.6) | 4 (5.0) | 7 (15.6) | 9 (8.4) |
| Other (Asian, American Indian, Native American) | 5 (8.5) | 9 (18.8) | 3 (4.8) | 7 (15.6) | 14 (13.1) |
|
| |||||
| Hispanic ethnicity (vs. not) | 3 (5.0) | 2 (4.2) | 3 (4.8) | 2 (4.4) | 5 (4.6) |
|
| |||||
| Gestational age (weeks); mean (SD) | 37.3 (3.5) | 35.2 (7.3) | 37.4 (3.7) | 35.0 (7.3) | 36.4 (5.6) |
|
| |||||
| Inborn (vs. transferred in) | 20 (33.3) | 20 (41.7) | 21 (33.3) | 19 (42.2) | 40 (37.0) |
|
| |||||
| Service caring for infant/assigning diagnostic code: | |||||
| Newborn Nursery | 29 (48.3) | 10 (20.8) | 32 (50.8) | 7 (15.6) | 39 (36.1) |
| NICU | 25 (41.4) | 15 (31.1) | 25 (38.1) | 16 (35.6)*** | 40 (37.0) |
| ENT/Audiology | 3 (5.0) | 7 (14.6) | 3 (4.8) | 7 (15.6) | 10 (9.3) |
| Other | 2 (4.0) | 15 (31.3) | 4 (6.3) | 13 (28.9) | 17 (15.7) |
| Unclear | 1 (1.7) | 1 (2.1) | 0 (0.0) | 2 (4.4) | 2 (1.9) |
|
| |||||
| Age at first CMV dx code (days); mean, SD | 19.3 (25.3) | 35.0 (32.2) | 18.4 (24.1) | 37.3 (29.5)** | 26.3 (29.5) |
Table 2c.
Clinical course data (N= 108) by ICD-9/10-CM Diagnostic Code Category and Final Disease Status
| ICD-9/10 Diagnostic Code of cCMV n (%) |
ICD-9/10 Diagnostic Code of CMV Infection n (%) |
Final Disease Status of cCMV n (%) |
Final Disease Status of Not cCMV n (%) |
Any cCMV/ or CMV Infection ICD-9/10-CM Diagnostic Code n (%) | |
|---|---|---|---|---|---|
| N=60 | N=48 | N=63 | N=45 | N=108 | |
|
| |||||
| Physical exam signs at birth (vs. none) * | 36 (61.7) | 17 (35.4) | 43 (68.3) | 11 (24.4)*** | 53 (50.0) |
| If present, which signs: | |||||
| Petechiae or purpura | 4 (11.1) | 1 (5.9) | 5 (11.9) | 0 (0.0) | 5 (9.4) |
| Hepatomegaly or splenomegaly | 1 (2.8) | 0 (0.0) | 1 (2.4) | 0 (0.0) | 1 (1.9) |
| Jaundice from direct hyperbilirubinemia | 4 (11.1) | 1 (5.9) | 4 (9.5) | 1 (9.1) | 5 (9.4) |
| Microcephaly (<10%ile) | 2 (5.6) | 1 (5.9) | 2 (4.8) | 1 (9.1) | 3 (5.6) |
| IUGR or SGA (<10%ile) | 19 (52.8) | 13 (76.5) | 25 (59.5) | 7 (63.6) | 32 (60.4) |
| Seizures or abnormal tone | 3 (8.3) | 0 (0.0) | 2 (4.8) | 1 (9.1) | 3 (5.7) |
| Other | 3 (8.3) | 1 (5.9) | 3 (7.1) | 1 (9.1) | 4 (7.5) |
|
| |||||
| Any abnormal CMV-related lab | 25 (41.7) | 14 (29.2) | 18 (28.6) | 20 (44.4) | 38 (35.2) |
| Unknown/Unavailable | 13 (21.7) | 18 (37.5) | 16 (25.4) | 15 (33.3) | 31 (28.7) |
|
| |||||
| If abnormal, which labs: | |||||
| Thrombocytopenia <100k | 6 (10.5) | 7 (15.6) | 19 (30.2) | 6 (13.3)* | 13 (12.7) |
| ALT ≥70 | 7 (12.3) | 0 (0.0)* | 7 (11.7) | 0 (0.0)* | 7 (6.9) |
| AST ≥70 | 9 (15.8) | 4 (8.9)* | 17 (27.0) | 3 (6.7)** | 13 (12.7) |
| Direct bilirubin >1.0 | 2 (8.0) | 1 (7.1) | 0 (0.0) | 0 (0.0) | 3 (7.7) |
|
| |||||
| Brain imaging performed (vs. not) | 55 (91.7) | 33 (68.8)* | 59 (93.7) | 29 (64.4)*** | 88 (81.5) |
|
| |||||
| Any abnormal brain imaging (vs. not) | 36 (60.0) | 18 (37.5)* | 39 (61.9) | 15 (33.3)** | 54 (50.0) |
| CUS abnormal (vs. not) | 34 (56.7) | 16 (33.3) | 37 (58.7) | 13 (28.9)*** | 50 (46.3) |
| MRI abnormal (vs not) | 13 (21.7) | 8 (16.7) | 12 (19.0) | 9 (20.0) | 21 (19.4) |
|
| |||||
| Ophthalmologic exam performed (vs. not) | 50 (83.3) | 29 (60.4) | 55 (87.3) | 34 (75.6) | 79 (73.1) |
|
| |||||
| Newborn hearing screen results | |||||
| Passed bilaterally | 33 (55.0) | 28 (58.3) | 36 (57.1) | 25 (55.6) | 61 (56.5) |
| Referred bilaterally | 13 (21.7) | 9 (18.8) | 14 (22.2) | 8 (17.8) | 22 (20.4) |
| Referred unilaterally | 9 (15.0) | 5 (12.5) | 10 (15.9) | 5 (11.1) | 15 (13.9) |
| Screening not complete/unknown | 5 (8.3) | 5 (10.4) | 3 (4.8) | 7 (15.6) | 10 (9.3) |
|
| |||||
| Sensorineural hearing loss diagnosed to date | 23 (38.7) | 13 (31.7) | 26 (41.3) | 10 (27.8) | 36 (36.4) |
|
| |||||
| Valganciclovir/ ganciclovir treatment (N=71) | 35 (70.0) | 10 (47.6) | 42 (66.7) | 3 (37.5) | 45 (63.4) |
<0.5
<.01
<.001
Confirmatory Laboratory Evidence:[22]
• Absence of a negative test (CMV DNA by Nucleic Acid Amplification Test (NAAT) or culture) on a urine specimen collected within 21 days of life, AND
• Detection of CMV DNA by NAAT from urine, whole blood (including dried blood spot [DBS]), or cerebrospinal fluid (CSF) collected from an infant within 21 days of life, OR
• Detection of CMV DNA by NAAT from amniotic fluid specimen, OR
• Isolation of CMV in viral culture from urine, whole blood, or CSF collected from an infant within 21 days of life, OR
• Isolation of CMV in viral culture from amniotic fluid specimen, OR
• Demonstration of CMV antigen in an autopsy specimen by immunohistochemistry (IHC), OR
• Detection of CMV antigen by antigenemia test in whole blood collected from an infant within 21 days of life.
Presumptive Laboratory Evidence:
• Absence of a negative test (CMV DNA by NAAT or culture) on a urine specimen collected within 21 days of life, AND
• Detection of CMV DNA by NAAT from saliva collected from an infant within 42 days of life§, OR
• Isolation of CMV in viral culture from saliva collected from an infant within 42 days of life§, OR
• Detection of CMV DNA by NAAT from urine, whole blood, or CSF collected from an infant within 22–42 days of life¶, OR
• Isolation of CMV in viral culture from urine, whole blood, or CSF collected from an infant within 22–42 days of life¶.
Abbreviations: CMV- cytomegalovirus; cCMV- congenital cytomegalovirus; dx- diagnostic; NICU- neonatal intensive care unit; PCR- polymerase chain reaction; IUGR- intrauterine growth restriction; SGA- small-for-gestational-age; ALT- alanine transaminase; AST- aspartate aminotransferase; CUS- cranial ultrasound; MRI- magnetic resonance imaging.
Figure 1.
Frequency of different ICD-9/10 diagnostic codes for congenital cytomegalovirus used over time at a tertiary children’s hospital.
Infants with an ICD-9/10-CM diagnostic code for cCMV were more likely to be of Black or White race, have elevated liver enzymes on laboratory studies, and have had brain imaging performed (91.7% vs. 68.8%, p <.01), which reported abnormal findings more often (60% vs. 37.5%, p >0.05) as compared to those with an ICD-9/10-CM diagnostic code for CMV Infection. There was no difference between gestational age, primary reason for CMV testing, clinical symptoms at presentation, newborn hearing screening result, or CMV laboratory test type used between those with an ICD-9/10-CM diagnostic code of cCMV vs. CMV Infection. Infants with a Final Disease Status of cCMV were more likely to be White (p >0.01), cared for by the newborn nursery (p<.001), younger at first diagnostic code (18.4 days vs. 37.3 days, p <.01), have petechiae at birth (p<.001), and have abnormal brain imaging (61.9% vs. 33.3 %, p <.01) as compared to those with a Final Disease Status of Not cCMV. There were no differences in reason for testing, newborn hearing screening result, or any hearing loss diagnosis to date between infants with a Final Disease Status of cCMV vs. Not cCMV.
The performance of the ICD-9/10-CM diagnostic codes in predicting cases by infant age is shown in Table 3. Overall, using any qualifying ICD-9/10-CM diagnostic code, the PPV at ≤90 days of age was 0.58 and the sensitivity was 1.0. PPV increased when using a more specific ICD-9/10-CM diagnostic code (cCMV vs. CMV infection) and with decreasing age. For instance, the PPV for an ICD-9/10-CM diagnostic code of cCMV at ≤21 days of age was 0.86 vs. 0.22 at ≤90 days of age for an ICD-9/10-CM diagnostic code of CMV Infection. On the other hand, the sensitivity increased with increasing infant age and using both diagnostic codes together.
Table 3.
Positive predictive value and sensitivity estimates for final clinical status of congenital CMV by diagnostic codes and timing of diagnostic codes
| Age in days at first dx code | ||||||||
|---|---|---|---|---|---|---|---|---|
| ≤90 days | ≤60 days | ≤45 days | ≤21 days | |||||
| PPV | Sensitivity | PPV | Sensitivity | PPV | Sensitivity | PPV | Sensitivity | |
| Congenital CMV dx code | 0.82 | 0.71 | 0.85 | 0.70 | 0.86 | 0.68 | 0.86 | 0.60 |
| CMV infection dx code | 0.22 | 0.21 | 0.33 | 0.21 | 0.36 | 0.16 | 0.29 | 0.16 |
| Any cCMV/CMV dx code | 0.58 | 1.00 | 0.65 | 0.87 | 0.68 | 0.84 | 0.71 | 0.76 |
Definitions:
Congenital cytomegalovirus CMV diagnostic codes: ICD-9/10-CMV codes corresponding to Congenital Cytomegalovirus (771.1 or P35.1)
CMV infection diagnostic codes = ICD-9/10-CM codes corresponding to Cytomegalovirus Infection 0.78.5, ICD-10 B25.x
Any cCMV/CMV diagnostic codes = ICD-9/10-CM codes corresponding to either Congenital CMV (771.1 or P35.1) or Cytomegalovirus Infection (0.78.5, ICD-10 B25.x)
Abbreviations: PPV- positive predictive value; CMV-cytomegalovirus; cCMV- congenital cytomegalovirus; dx- diagnostic.
In the multinomial regression, the statistically significant variables retained in the model included abnormalities on newborn exam. Multinomial logistic regression found the age of the first diagnostic code ≤21 days (vs. >21 days) (OR=4.11, 95% CI 1.45–12.03), having an ICD-9/10-CM diagnostic code of cCMV (vs. CMV Infection) (OR=10.87, 95% CI 3.64–32.47), and having Clinical Signs at Birth (vs. none) (OR=8.4, 95% CI 2.72–25.81) to be associated with greater odds of having a Final Disease Status of cCMV (vs. Not cCMV)(Table 4).
Table 4.
Multinominal stepwise logistic regression model examining variables associated with final disease status of congenital cytomegalovirus (vs. not).
| Odds Ratio | Standard Error | 95% CI | Wald | p-values | |
|---|---|---|---|---|---|
| Age at first diagnostic code ≤21 days (>21 days of age) | 4.11 | .548 | 1.45–12.03 | 6.634 | .010 |
| ICD-9/10-CM code for congenital CMV (vs. CMV Infection) | 10.87 | .558 | 3.64–32.47 | 18.278 | <.001 |
| Clinical signs of congenital CMV at birth (yes vs. no) | 8.38 | .574 | 2.72–25.81 | 13.735 | <.001 |
Abbreviations: ICD-9/10-CM - International Classification of Disease 9th and 10th Edition Clinical Modification; cCMV -congenital cytomegalovirus; CMV-cytomegalovirus; CI- confidence interval.
DISCUSSION
This study assessed the accuracy of ICD-9/10-CM diagnostic codes for identifying cCMV. Among infants with an ICD-9/-10 CM code for cCMV, nearly 1 in 3 did not have evidence of cCMV. As such, use of case definitions for cCMV that include the presence of codes for either cCMV or CMV Infection may be biased. Further work is necessary to identify an accurate administrative claims-based algorithm to identify cCMV. This algorithm would enable large-scale assesments of prevalence and quality of care among this population.
Studies using administrative claims data to identify cCMV cases, such as ours, can only measure the prevalence of diagnoses[24], which is lower than the known true population prevalence of cCMV due to underascertainment and limitations relating to lack of universal screening both in our cohort and in our country at large[8]. In other words, infants meeting the ICD case definition for cCMV only represent a fraction of the true cases present in the population. A recent study by Campione et al demonstrated this fact noting that only one in ten infants with laboratory-confirmed CMV infection had a corresponding CMV diagnostic code in their EHR, for a diagnostic code prevalence of 0.89 cases per 10,000 infants[25]. The authors hypothesized the low sensitivity may be due to assignment variation and missing diagnostic code data[25]. Additionally, of those meeting ICD case criteria with accompany diagnostic codes, there is risk of misclassification. As noted in our cohort, just over half were correctly identified as true positives with corresponding Final Disease Status of cCMV. Our study found the accuracy of ICD-9/10-CM diagnostic codes for a Final Disease Status of cCMV to be greatest when limiting the codes to those for cCMV only at ≤21 and ≤45 days of age. Conversely, the sensitivity increased as expected by including more infants in the sample up to ≤90 days of age and for both sets of diagnostic codes together. The sensitivity of all cCMV diagnostic codes in predicting a Final Disease Status of cCMV in our study ranged 76–87% (excluding inherent 100% sensitivity at 90 days of age). It should be noted that we examined the sensitivity of the diagnostic code and timing within our study cohort, not at the population level, which others have done[25]. As such, comparisons between our calculated sensitivity, and that calculated by Campione et al,[25] cannot be drawn. In a logistic regression model, we also found that the age of the first diagnostic code ≤21 days, a diagnostic code of cCMV (vs. CMV Infection), and having clinical findings on the newborn exam (vs. not) were associated with greater odds of having cCMV Final Disease Status.
To ensure the distinction between congenital and postnatally acquired infection, and minimize false positives, most research using administrative claims data has defined cases as those with diagnostic codes within 30–60 days of age[8–11,17,18]. This also allows for the inclusion of cases with delayed testing or claims submission beyond the 21-day testing window for directly collected specimens. However, our results suggest that PPV in our sample was highest for all CMV-related codes similarly at ≤21 and ≤45 days of age, which raises the question of whether an ICD case designation should use this narrower cut-off for improved accuracy. In general, the accuracy (as measured by PPV) of all CMV-related diagnostic codes ranged from 58–71% in our study but improved to 82–86% when isolated to the cCMV code alone. We hypothesize that accuracy may lessen over time due to the inclusion of rule-out cases that, upon further review, do not clinically have confirmed disease or have been incorrectly assigned the code. Cases within this category may include those with neonatal respiratory failure or sepsis with multiorgan dysfunction requiring critical care. It may be important for researchers to consider the frequency of false positives from studies completed to date using the current ICD case definition that extends beyond 45 days. By narrowing the age cut-off closer to 21 days, future work may be more accurate in identifying true cases, which has implications for studying clinical characteristics, opportunities for antiviral treatment, and long-term monitoring of developmental outcomes with the added benefit of cost analysis using large-scale administrative claims data. Adjustments to administrative claims data coding may also occur within the changing landscape of screening protocols, as more infants are diagnosed through expanded targeted or universal screening policies that afford fewer missed cases. However, standardization of assigned codes is needed when considering per-child cost to avoid under- or over-estimation[8].
Our logistic regression model did not find many of the classic features of clinically significant infection, such as sensorineural hearing loss, abnormal laboratory findings, or abnormal brain imaging, as predictive of true disease state. This contrasts with other studies, such as Fowler et al[26], which found that failed/referred hearing screening was more likely to occur in cCMV than in the general population. This illustrates the variability of disease presentation and further reaffirms the difficulty of predicting cCMV based on these findings alone. This may provide evidence to support universal screening protocols, as referred newborn hearing screen or abnormal serologies/imaging are common criteria under consideration for expanded targeted screening programs[27,28]. Variables that were predictive were age of first diagnostic code ≤21 days, a diagnostic code of cCMV and having clinical signs at birth.
This study has several potential limitations. The first is lack of generalizability given that the study was carried out at a single quaternary center. Second, the study was conducted using a retrospective chart review and infants were not subject to universal CMV screening, or standardized testing protocols. As such, we could not identify true or false negative cases and results cannot be utilized in calculating cCMV prevalence nor portrayed as a prevalence study. Additionally, given lack of standardized testing and screening protocols for cCMV across institutions, diagnostic evaluations both within the health system and through referred outside facilities could not be controlled. Third, the study occurred during the transition from ICD-9 to ICD-10-CM codes. This change moved the code for congenital CMV (P35.1) within the perinatal conditions chapter, allowing more clear differentiation between congenital and acquired CMV cases. It is possible that accuracy of ICD codes improved over this change. In addition, identified cases were reliant on ICD codes associated with the chart through patient encounters; there is a is a risk of underascertainment of cases when screening results or treatment is not associated with a disease-specific ICD code, or the code is applied at subsequent encounters. However, it should be noted that we extended the age of inclusion to 90 days, which is almost double that of most studies on cCMV, to avoid misclassification and encompass the majority of cases if diagnostic codes were applied later in a patient’s clinical course[8–11]. Additionally, our cohort is likely not representative of the full spectrum of cCMV cases, and lacks in measurement of False Negatives, especially considering those with clinically inapparent presentations that may be unrecognized and undiagnosed, with the use of administrative claims data alone[8].
CONCLUSION
In conclusion, the PPV of ICD-9/10-CM codes for cCMV was greatest in infants with younger age at first diagnosis and with the Congenital Cytomegalovirus diagnostic code in particular. PPV and sensitivity were low for the Cytomegalovirus Infection code at all ages. Inclusion of both codes and age up to 90 days increased the sensitivity, but decreased PPV. Our results also suggested greater odds of Final Disease Status of cCMV in infants with clinical signs at birth. Future studies should examine the PPV and sensitivity of cCMV diagnostic codes in larger, multi-center cohorts.
Supplementary Material
Table 2b.
CMV laboratory data (N= 108) by ICD-9/10-CM Diagnostic Code Category and Final Disease Status
| ICD-9/10 Diagnostic Code of cCMV n (%) |
ICD-9/10 Diagnostic Code of CMV Infection n (%) |
Final Disease Status of cCMV n (%) |
Final Disease Status of Not cCMV n (%) |
Any cCMV/ or CMV Infection ICD-9/10-CM Diagnostic Code n (%) | |
|---|---|---|---|---|---|
| N=60 | N=48 | N=63 | N=45 | N=108 | |
|
| |||||
| Primary reason for CMV testing | |||||
| Clinical signs at birth | 3 (5.0) | 0 (0.0) | 2 (3.2) | 1 (2.2) | 3 (2.8) |
| SGA or IUGR | 10 (16.7) | 5 (10.4) | 13 (20.6) | 2 (4.4) | 15 (13.9) |
| Failed newborn hearing screen or SNHL | 11 (18.3) | 10 (20.8) | 12 (19.9) | 9 (20.0) | 21 (19.4) |
| Prenatal concerns | 13 (21.7) | 6 (12.5) | 9 (14.3) | 10 (22.2) | 19 (17.6) |
| Abnormal brain imaging | 3 (5.0) | 2 (4.2) | 4 (6.3) | 1 (2.2) | 5 (4.6) |
| Abnormal CMV-related lab | 11 (18.3) | 8 (16.7) | 13 (20.6) | 6 (13.3) | 19 (17.6) |
| Preterm | 0 (0.0) | 2 (4.2) | 0 (0.0) | 2 (4.3) | 2 (1.9) |
| Other (e.g., respiratory issues, sepsis) | 9 (15.0) | 15 (31.3) | 10 (15.9) | 14 (31.1) | 24 (22.2) |
|
| |||||
| CMV testing performed (First test performed) | |||||
| Urine PCR | 0 (0.0) | 1 (2.1) | 0 (0.0) | 1 (2.2) | 1 (0.9) |
| Saliva PCR | 5 (8.3) | 5 (10.4) | 5 (7.9) | 5 (11.1) | 10 (9.4) |
| Serum PCR | 22 (37.9) | 20 (41.7) | 26 (41.2) | 16 (35.6) | 42 (39.6) |
| Dried blood spot | 19 (31.7) | 8 (16.7) | 18 (30.0) | 9 (20.0) | 27 (25.5) |
| Viral shell culture | 1 (1.7) | 1 (2.1) | 2 (3.2) | 0 (0.0) | 2 (1.9) |
| Serologies (IgM) | 9 (15.0) | 11 (22.9) | 9 (14.3) | 11 (24.4) | 20 (18.9) |
| No CMV testing performed | 3 (5.0) | 3 (6.3) | 3 (4.8) | 3 (6.7) | 6 (5.5) |
|
| |||||
| CMV testing performed (Final test performed) | |||||
| Urine PCR | 21 (35.0) | 17 (35.4) | 26 (42.1) | 12 (26.7) | 38 (35.2) |
| Saliva PCR | 7 (11.7) | 6 (12.5) | 6 (9.5) | 7 (15.6) | 13 (12.0) |
| Serum PCR | 22 (36.7) | 12 (25.0) | 22 (34.9) | 12 (26.7) | 34 (31.5) |
| Dried blood spot | 3 (5.0) | 4 (8.3) | 3 (4.7) | 4 (8.9) | 7 (6.5) |
| Viral shell culture | 3 (5.0) | 1 (2.1) | 3 (4.7) | 1 (2.2) | 4 (3.7) |
| Serologies (IgM) | 1 (1.7) | 6 (12.5) | 1 (1.6) | 6 (13.3) | 7 (6.5) |
| No CMV testing performed | 3 (5.0) | 2 (4.2) | 2 (3.2) | 3 (6.7) | 5 (4.6) |
|
| |||||
| Laboratory Evidence of cCMV | |||||
| Confirmed laboratory evidence | 48 (80.0) | 15 (31.3) | 61 (96.8) | 2 (4.4)*** | 63 (58.3) |
| Presumptive laboratory evidence | 2 (3.3) | 1 (4.2)*** | 2 (3.3) | 2 (4.4) | 4 (3.7) |
| No laboratory evidence | 10 (16.7) | 31 (64.6) | 0 (0) | 41 (91.1) | 41.0 (38.0) |
Acknowledgements:
Thank you to Scott Grosse, PhD, Jessica Leung MPH, Tatiana Lanzieri PhD, Julie Sturza MPH for their insights into the design of this study.
Funding:
MHP received funding from the Gerber Foundation and the NICHD K23 HD108278.
Conflicts of interests:
MHP serves on the board of the National CMV Foundation (unpaid) and is a paid consultant for Moderna and Medscape/WebMD. These entities did not contribute in any way to the planning, data gathering, analysis, or writing of the manuscript for this project.
Abbreviations and Acronyms:
- cCMV
congenital cytomegalovirus
- CMV
cytomegalovirus
- ICD-9/10-CM
International Classification of Diseases, Ninth and Tenth Revision, Clinical Modification
- PPV
positive predictive value
- EHR
electronic health record
- SGA
small for gestational age
- SNHL
sensorineural hearing loss
- PCR
polymerase chain reaction
- DBS
dried blood spot
References:
- 1.Wilson KL, Shah K, Pesch MH. Inconsistent provider testing practices for congenital cytomegalovirus: Missed diagnoses and missed opportunities. Int J Neonatal Screen 2022;8(4):60. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Pesch MH, Schleiss MR. Emerging concepts in congenital cytomegalovirus. Pediatrics. 2022;150(2). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Sorichetti B, Goshen O, Pauwels J, et al. Symptomatic congenital cytomegalovirus infection is underdiagnosed in British Columbia. J Pediatr 2016;169:316–317. [DOI] [PubMed] [Google Scholar]
- 4.Pesch MH, Lauer CS, Weinberg JB. Neurodevelopmental outcomes of children with congenital cytomegalovirus: A systematic scoping review. Pediatr Res 2024. (95):418–435. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Bartlett AW, McMullan B, Rawlinson WD, et al. Hearing and neurodevelopmental outcomes for children with asymptomatic congenital cytomegalovirus infection: A systematic review. Rev Med Virol 2017. Sep 6. [DOI] [PubMed] [Google Scholar]
- 6.Rawlinson WD, Boppana SB, Fowler KB, et al. Congenital cytomegalovirus infection in pregnancy and the neonate: consensus recommendations for prevention, diagnosis, and therapy. Lancet Infect Dis 2017;17(6):e177–e188. [DOI] [PubMed] [Google Scholar]
- 7.Prevention CfDCa. Congenital Cytomegalovirus (cCMV) Infection and Disease 2024 Case Definition. National Notifiable Diseases Surveillance System; 2024. [Google Scholar]
- 8.Grosse SD, Leung J, Lanzieri TM. Identification of congenital CMV cases in administrative databases and implications for monitoring prevalence, healthcare utilization, and costs. Curr Med Res Opin 2021;37(5):769–799. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Leung J, Grosse SD, Hong K, et al. Changes in valganciclovir use among infants with congenital cytomegalovirus Diagnosis in the United States, 2009–2015 and 2016–2019. J Pediatr 2022;246:274–278.e2. [DOI] [PubMed] [Google Scholar]
- 10.Leung J, Dollard SC, Grosse SD, et al. Valganciclovir use among commercially and Medicaid-insured infants with congenital CMV infection in the United States, 2009–2015. Clin Ther 2018;40(3):430–439.e1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Leung J, Kennedy JL, Haberling DL, et al. Congenital CMV-coded diagnosis among American Indian and Alaska Native infants in the United States, 2000–2017. J Immigr Minor Health. 2020;22(5):1101–1104. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Meyers J, Sinha A, Samant S, et al. The economic burden of congenital cytomegalovirus disease in the first year of life: A retrospective analysis of health insurance claims data in the United States. Clin Ther 2019;41(6):1040–1056.e3. [DOI] [PubMed] [Google Scholar]
- 13.Lopez AS, Ortega-Sanchez IR, Bialek SR. Congenital cytomegalovirus-related hospitalizations in infants <1 year of age, United States, 1997–2009. Pediatr Infect Dis J 2014. Nov;33(11):1119–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Candrilli S, Trantham L. The economic burden of congenital cytomegalovirus-related hospitalizations in the United States. Value in Health. 2017;20(9):A784–A785. [Google Scholar]
- 15.Inagaki K, Blackshear C, Palmer A, et al. Risk factors, geographic distribution, and healthcare burden of symptomatic congenital cytomegalovirus infection in the United States: analysis of a nationally representative database, 2000–2012. J Pediatr 2018;199:118–123.e1. [DOI] [PubMed] [Google Scholar]
- 16.Grosse SD, Dollard SC, Ortega-Sanchez IR. Economic assessments of the burden of congenital cytomegalovirus infection and the cost-effectiveness of prevention strategies. Semin Perinatol 2021:151393. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Leung J, Grosse SD, Hong K, et al. Changes in valganciclovir use among infants with congenital cytomegalovirus diagnosis in the United States, 2009–2015 and 2016–2019. J Pediatr 2022. [DOI] [PubMed] [Google Scholar]
- 18.Lanzieri TM, Leung J, Caviness AC, et al. Long-term outcomes of children with symptomatic congenital cytomegalovirus disease. J Perinatol 2017. Jul;37(7):875–880. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Alford CA, Pass RF, Stagno S. Chronic congenital infections: common environmental causes for severe and subtle birth defects. Birth Defects Orig Artic Ser 1983;19(5):87–96. [PubMed] [Google Scholar]
- 20.Kuang A, Xu C, Southern DA, et al. Validated administrative data based ICD-10 algorithms for chronic conditions: A systematic review. Journal of Epidemiology and Population Health. 2024;72(4):202744. [DOI] [PubMed] [Google Scholar]
- 21.DataDirect [Internet]. The University of Michigan. 2021. Available from: https://datadirect.med.umich.edu/. [Google Scholar]
- 22.Council of State and Territorial Epidemiologists. Standardized Surveillance Case Definitions for Congenital Cytomegalovirus (cCMV) Infection and Disease. In: Infectious Disease Committee, editor. 2024. [Google Scholar]
- 23.IBM Corp. IBM SPSS Statistics for Windows. 28.0. Armonk, NY: IBM Corp; 2024. [Google Scholar]
- 24.Gianfrancesco MA, Goldstein ND. A narrative review on the validity of electronic health record-based research in epidemiology. BMC medical research methodology. 2021;21(1):234. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Campione A, Lanzieri TM, Ricotta E, et al. Missing diagnoses of congenital cytomegalovirus infection in electronic health records for infants with laboratory-confirmed infection. Current medical research and opinion. 2022;38(2):273–275. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Fowler KB, McCollister FP, Sabo DL, et al. A targeted approach for congenital cytomegalovirus screening within newborn hearing screening. Pediatrics. 2017;139(2):e20162128. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Akiva MH, Hyde-De Sousa H, Lamarre V, et al. Identifying Clinical Criteria for an Expanded Targeted Approach to Screening for Congenital Cytomegalovirus Infection—A Retrospective Study. International Journal of Neonatal Screening. 2023;9(3):40. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Suarez D, Nielson C, McVicar SB, et al. Analysis of an expanded targeted early cytomegalovirus testing program. Otolaryngology--head and neck surgery : official journal of American Academy of Otolaryngology-Head and Neck Surgery. 2023. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.