Study population

We conducted a retrospective cohort study of infants born in the United States between July 1, 2016 and February 29, 2020. We identified live births occurring within this time window and constructed retrospective birth cohorts in each of three de-identified insurance claims datasets—Merative MarketScan Commercial Claims and Encounters ® (CCAE), Multi-State MarketScan Medicaid ® (MDCD), and Optum’s de-identified Clinformatics ® Data Mart (CDM). These datasets capture different geographic distributions of infants and implicitly capture different socioeconomic strata (commercial CCAE/CDM datasets versus Medicaid). We included infants in the current analysis if they had a qualifying MA RSV LRTI diagnosis during their first RSV season. We defined their first such diagnosis as the “index diagnosis”, which triggered the start of their first acute MA RSV LRTI episode. We followed infants for 7 days after this index diagnosis to evaluate their healthcare utilization, assuming no loss to follow-up during this short time window.

Comorbidity groups

We classified infants into one of three mutually exclusive comorbidity groups based on gestational age and comorbidity status, as described in prior work [11]. Briefly, infants were categorized as follows:

• A. Healthy term infants: infants with no indication of having been delivered preterm, based either on diagnosis related group (DRG) codes identifying a full-term neonate without major problems or the absence of any DRG or International Classification of Diseases-Tenth Revision-Clinical Modification (ICD-10-CM) codes indicating preterm birth or gestational age less than 37 weeks. We categorized infants without any such DRG or ICD-10-CM codes (i.e., infants with unknown gestational age) to be full term, as infants born preterm would likely have an accompanying code.
• B. Palivizumab-eligible: infants who would be eligible to receive palivizumab, either because they had a gestational age less than 29 weeks, a qualifying combination of preterm birth and CLD, or, HS-CHD.
• C. Other comorbidities: infants of any gestational age with comorbidities (e.g., lower-risk congenital heart disease, respiratory conditions) that might predispose them to severe RSV-related outcomes.

For comorbidity groups B and C, we included infants with relevant ICD-10-CM codes if that code appeared at any time prior to the onset of their first MA RSV LRTI episode. S1 and S2 Tables in S1 File list the diagnostic codes underlying these categories.

Acute RSV episodes

As in our prior work [11], we identified each infants’ first RSV season by calculating beginning and end dates specific to the Census division in which the infant was born, according to the “retrospective slope 10” procedure described by Midgley et al. [13, 14]. Briefly, this method calculates a normalized 5-week moving average of RSV detections and assigns the season’s start date to the second of two consecutive weeks in which the increase in the moving average exceeds 10. The season offset date is then set as the last week in which the 5-week moving average remains above that during the week of season onset.

We defined an “acute MA RSV LRTI episode” as the 0–7 days (inclusive) following the occurrence of an index MA RSV LRTI diagnosis, where day 0 is the date of this index diagnosis. Hereafter we refer to the first acute MA RSV LRTI episode during an infants’ first RSV season as the “first MA RSV LRTI episode”. We focused on infants’ first MA RSV LRTI episode because our preliminary analyses preparatory to research suggested that second MA episodes within the same RSV season were rare.

Also similar to our prior work [11], we used two alternative definitions of the index MA RSV LRTI diagnosis to reflect uncertainty in the ascertainment of MA RSV LRTI stemming from a lack of systematic laboratory data on RSV infection in the insurance claims databases. Both definitions were based on combinations of ICD-10-CM diagnosis codes (S3 Table in S1 File):

• The specific definition included ICD-10-CM codes that explicitly mention RSV (B974, J121, J205, and J210), where a diagnosis of B974 (“RSV as the cause of diseases classified elsewhere”) attached to an outpatient or emergency department (ED) visit qualified only if the infant had another respiratory condition diagnosed within ± 5 days.
• The sensitive definition included all codes from the specific definition, plus 2 additional codes capturing unspecified bronchiolitis (J218 and J219). The CCAE database includes the first 4 diagnosis positions for outpatient and ED visits and the first 15 diagnosis positions for inpatient hospital stays; we limited diagnosis ascertainment to these positions in the CDM and MDCD data to harmonize analyses across the three databases.

For each acute MA RSV LRTI episode, we identified outpatient visits, ED visits, and inpatient hospital stays attached to a sensitive RSV diagnosis occurring during the episode (regardless of the index diagnosis definition). To reduce the impact of coding errors in claims and to avoid overestimating healthcare utilization, we allowed for a maximum of one outpatient and one ED visit per day; therefore, infants could have up to 8 outpatient and/or ED visits recorded during the episode, inclusive of the index date. We counted inpatient stays if they overlapped with any part of the 7-day episode window. If an inpatient stay began on the same day that an outpatient or ED visit occurred, we counted each event as a separate instance of healthcare utilization in order to characterize the different settings infants contact during the acute RSV episode. In the CDM database, we identified inpatient stays using a confinement identification code available in the base dataset. For a subset of inpatient events without associated confinement codes, we chained events with contiguous dates of service together into a single inpatient stay. In all databases, a new inpatient stay was triggered only when we found a gap of at least one day between the discharge date of one stay and the admission date for the next. S4 Table in S1 File lists the place of service codes used to classify events as outpatient, ED, or inpatient.

Descriptive analysis

We summarized the number of healthcare contacts at each place of service by summing individual outpatient, ED, and inpatient events and calculating the mean number of visits/stays across acute MA RSV LRTI episodes. We calculated the proportion of episodes involving 0, 1, ≥ 1, and/or ≥ 2 contacts with a given place of service and quantified the uncertainty due to random error around these estimated proportions using approximate 95% confidence intervals [15]. In addition to estimating marginal quantities within each claims database, we stratified estimates by comorbidity group.

Ethics approval

This study involved only secondary analysis of de-identified insurance claims data and is not considered human subjects research for that reason. No institutional review board approval was required.

Descriptive characteristics

In the CCAE database, we identified 25,409 acute MA RSV LRTI episodes under the specific RSV definition and 69,068 under the sensitive definition. In the MDCD database, we identified 67,357 and 170,744 such episodes, respectively. In the CDM database, we identified 12,402 and 31,363 such episodes. Geographic distributions and insurance plan types differed between the commercial databases. Across data sources and RSV definitions, approximately 3% of infants had HS-CHD, < 1% had CLD, and 9–13% had a chronic condition of any kind. See Table 1 for demographic and clinical characteristics among infants with acute RSV episodes under each definition.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics of infants with an acute MA RSV LRTI during their first RSV season, by insurance claims database and index diagnosis definition.

https://doi.org/10.1371/journal.pone.0313573.t001

Participants

In the final CCAE analytic sample 76.7% of infants were classified as comorbidity group A (healthy term), 3.8% comorbidity group B (palivizumab-eligible), and 19.4% comorbidity group C (other comorbidities). In the final MDCD analytic sample, these percentages were 77.7%, 3.9%, and 18.3%, respectively, while in the CDM data they were 79.4%, 3.3%, and 17.2%.

Healthcare utilization during acute RSV episodes

Specific definition.

Under the specific definition, infants averaged 1.11, 0.9, and 1.57 outpatient visits in the CCAE, MDCD, and CDM databases, respectively. They averaged 0.36, 0.52, and 0.4 ED visits, and 0.24, 0.22, 0.24 inpatient stays per acute RSV episode. Eighty percent, 68%, and 94% of episodes involved at least one outpatient visit in the CCAE, MDCD and CDM databases, respectively. Thirty-three percent, 45%, and 36% had at least one ED visit, while 22%, 20%, and 24% had at least one inpatient stay (Fig 1). Across the databases, 43–63% of episodes involved ≥ 2 visits to any healthcare setting, varying by database and comorbidity group (Fig 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Percentage of acute MA RSV LRTI episodes involving 0, 1, or ≥ 2 visits to a given healthcare setting, stratified by RSV index diagnosis definition and insurance claims database.

Fill colors depict the share of each bar accounted for by a given comorbidity group.

https://doi.org/10.1371/journal.pone.0313573.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Overall percentage of acute MA RSV LRTI episodes involving ≥ 2 visits to any healthcare setting(s), stratified by comorbidity group and insurance claims database.

Line ranges depict 95% confidence intervals.

https://doi.org/10.1371/journal.pone.0313573.g002

Healthcare utilization by comorbidity group

Healthy term infants (comorbidity group A) accounted for the largest number of visits to all healthcare settings during acute MA RSV LRTI episodes, followed by infants with other comorbidities (comorbidity group C) and palivizumab-eligible infants (comorbidity group B) (Table 2, Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Percentage of MA RSV LRTI episodes that involved 0, 1, or 265 2 visits to a given healthcare setting, stratified by insurance claims database, index diagnosis definition, and comorbidity group.

https://doi.org/10.1371/journal.pone.0313573.t002

Healthy term infants had the highest number of outpatient visits across claims databases and RSV definitions and the highest proportion of MA RSV LRTI episodes involving the outpatient setting. Infants with other comorbidities had comparable outpatient usage in most cases. Palivizumab-eligible infants consistently exhibited the highest proportion of MA RSV LRTI episodes involving an inpatient stay, followed by infants with other comorbidities and then healthy term infants.

Under the specific index diagnosis definition, infants did not appear to differ meaningfully by comorbidity group in their usage of the ED. Under the sensitive definition in commercial claims, palivizumab-eligible infants and those with other comorbidities exhibited higher ED utilization than healthy term infants. The magnitudes of the differences in healthcare utilization between comorbidity groups varied by claims database, although less so for ED visits.

Infants with other comorbidities were the most likely to have ≥ 2 visits to healthcare settings during their first MA RSV LRTI episode, followed by palivizumab-eligible infants and healthy term infants, respectively. Regardless of comorbidity group, infants’ RSV episodes frequently involved multiple visits to healthcare settings (Table 2, Fig 2).

Limitations

The following limitations should be kept in mind when interpreting the results of our study. First, we focus only on acute MA RSV LRTI episodes, which we defined as the 0–7 days following the index diagnosis. However, healthcare utilization associated with the onset of infection may extend beyond this initial episode. A study by Doucette et al. found that up to one-third of infants hospitalized for RSV are readmitted to the hospital within 3 months [21].

Second, insurance claims databases generally do not contain information on laboratory testing for RSV infection, and so we used diagnostic codes as proxies. Misclassification of RSV status varies by healthcare setting, with inpatient diagnoses being most accurate [22]. We addressed the inherent limitations of code-based proxies by implementing both a specific and a sensitive definition for the index RSV diagnosis, “bounding” our estimates under extreme assumptions regarding the accuracy of ICD-10-CM codes. Given that a substantial proportion of non-RSV-coded bronchiolitis diagnoses are probably RSV-related [23], our specific definition likely underestimates the true RSV burden, while our sensitive definition assumes all unspecified bronchiolitis diagnoses are RSV-related. Under the sensitive definition, episodes were less likely to involve an ED visit or inpatient hospital stay, suggesting this definition captures milder MA respiratory episodes.

Third, our definition of palivizumab eligibility only approximates current eligibility guidelines [7], due again to the limitations of insurance claims data. Infants’ gestational age and/or comorbidity status may have been misclassified, possibly understating differences in healthcare utilization patterns by comorbidity group. Similarly, we assume that infants with unknown gestational age (24–27%) indicate a full-term birth. The elicited proportions of full-term births across insurance claims databases ranged from 83.6% (MDCD) to 88% (CDM), in good agreement with recent data from the National Center for Healthcare Statistics (just under 90%) [24], though we would not expect Medicaid infants to be representative of the broader population.

Fourth, we did not account for potential selection bias or loss to follow-up [25, 26]. Selection bias could have occurred if infants who lost insurance eligibility before or during their first RSV season differed from infants who were observed until the end of their first season. In our prior work, weighting for loss of insurance eligibility had modest effects on most estimates of MA RSV LRTI burden [11].

Fifth, we observed notable differences between the commercial databases (CCAE and CDM) in the proportion of acute RSV episodes without an outpatient visit and in the proportion of episodes involving ≥ 2 visits to a healthcare setting. We speculate that these differences could be explained by variation in the geographic distribution or payer types between the data sources but cannot rule out subtle differences in how analyses were implemented across databases.

Sixth, we cannot rule out that some infants in our analytic sample appear in multiple insurance claims databases. We suspect the overlap to be negligible, particularly over the timeframe of infants’ first MA RSV LRTI episode.

Seventh, our acute RSV episode definition might have missed healthcare utilization related to the RSV infection but which preceded an infant’s meeting the index RSV diagnosis criteria. Healthcare utilization associated with the RSV episode may also have occurred beyond the 7-day episode window we investigated. Therefore, our estimates may prove to be slightly conservative.

Finally, the SARS-CoV-2 pandemic markedly altered RSV transmission patterns among infants [27–29]. We ended our birth cohort to February 29, 2020 to focus on healthcare utilization under typical seasonal RSV dynamics, and our estimates may not apply to healthcare utilization occurring during atypical waves of RSV infection.

Prior presentation

Portions of the data reported in this paper were presented at IDWeek 2022, October 19–23, 2022, Washington, D.C.