We utilised routinely collected data from Scottish national hospital registries for the period from 1 July 2017 to 30 June 2023, for this retrospective analysis. We defined the annual respiratory season as the period between 1 July of one year and 30 June of the following year, to capture all hospitalisations during the winter period.

Hospitalisation data were obtained from the Scottish Morbidity Records 01 (SMR01), an episode-based patient record covering all in-patient and day-case episodes from non-obstetric and non-psychiatric specialities in Scotland [9]. We included all SMR01 records with at least one RTI-related ICD-10 diagnostic code (primary or secondary; up to six codes were available for each episode). The list of RTI ICD-10 codes is provided in Supplementary Material S1 [10]. Routine admissions were identified using the ‘type of admission’ variable, and admissions recorded as ‘routine’ were excluded because these represent scheduled admissions unlikely to be related to an acute RTI episode. We excluded day cases because these involve planned clinical care with clinician review and supervised recovery at the treatment facility without an overnight hospital stay. Inpatient admissions lasting <2 days (not involving an overnight stay) were, however, retained, as these may reflect short acute hospitalisations.

Laboratory data for hMPV and RSV were obtained from the Electronic Communication of Surveillance Scotland (ECOSS), a Public Health Scotland system that collects patient-level laboratory results from all National Health Service (NHS) laboratories in Scotland [11]. Only positive test results were available; data on negative tests or the total number of tests performed were not available. Testing was conducted as part of routine clinical care, and policies varied across health boards. In Lothian, respiratory virus testing using the LUMINEX Respiratory Virus Panel (including hMPV and RSV) was recommended for all RTI hospital admissions. In contrast, testing practices in other health boards were unclear. Laboratory-confirmed hMPV or RSV hospital episodes were defined as RTI hospitalisations with a positive test from a specimen obtained within 7 days before and up to 3 days following admission [12]. Co-infections with other respiratory viruses were not excluded but were not analysed separately. A 28-day washout period was applied at the patient level to laboratory tests so that multiple positive tests within 28 days were considered a single positive test result.

Length of stay (LOS) was defined as the difference between discharge and admission dates, and admissions with same-day discharge were assigned a LOS of 1 day.

Intensive care unit (ICU) or high-dependency unit (HDU) admission data were obtained from the Scottish Intensive Care Society Audit Group (SICSAG) [13]. ICU admission was defined as admission to either ICU or HDU within RSV or hMPV RTI hospital admission episodes. ICU admission rate was defined as the proportion of RSV or hMPV RTI admissions requiring ICU (or HDU) admission and was expressed as a percentage. The mortality data were extracted from the National Records of Scotland (NRS) [14]. The in-hospital case fatality rate (CFR) was defined as the proportion of in-hospital deaths among all RSV or hMPV RTI admissions.

Data from SMR01, ECOSS, SICSAG and NRS mortality records were linked using a pseudonymised unique patient identifier available across all datasets. Linkage was performed by the study team within the national Trusted Research Environment (TRE). Previous evaluations of Scottish national data linkage systems have demonstrated high linkage quality, with strong agreement between routinely linked datasets, suggesting minimal linkage error [15]. The data linking procedures for linking all datasets are illustrated in Supplementary Material S2.

Population data for the estimation of incidence rate ratios were obtained from the open-access NRS [16, 17]. For each season, we used the mid-year population estimates corresponding to the end of that season.

The Public Benefit and Privacy Panel for Health and Social Care (PBPP-HSC) approved data access to the required data in the National Safe Haven (Approval number: 2223-0019).

This research examined pseudonymised data that did not require patients’ active consent. The data were collected routinely through NHS Scotland, with patients informed of potential use and their rights through various PHS and NHS Scotland privacy notices.

All statistical analyses were conducted in RStudio (4.2.1). All analyses were conducted at hospital admission (episode) level. Incidence rate ratios (IRRs) comparing RSV and hMPV were estimated using Poisson regression with a log link and population offset, stratified by age group and annual respiratory season. IRRs were obtained by exponentiating virus coefficients, with 95% confidence intervals calculated using profile-likelihood methods to account for sparse data. Strata with zero laboratory-confirmed RSV or hMPV RTI hospital admissions were excluded, as IRRs could not be estimated in such cases.

To account for potential undertesting outside Lothian, analyses were repeated using extrapolated counts. Age- and month-specific virus positivity proportions among RTI admissions in Lothian were applied to RTI admissions in other health boards to estimate virus counts, which were then combined with observed Lothian counts to generate national estimates. These age- and month-specific virus positivity proportions among RTI admissions in Lothian are provided in Supplementary Material S3. The extrapolation procedure is described in detail in Supplementary Materials S4 and S5. IRRs were recalculated using the same modelling approach mentioned above.

To provide context for the interpretation of regional estimates, overall population characteristics for Lothian and Scotland were described separately and are presented in Supplementary Material S6. In addition, as a sensitivity analysis, laboratory-confirmed IRR estimates were reported separately for Lothian.

We also compared laboratory-confirmed RSV- and hMPV-associated RTI hospitalisations by clinical presentation and disease severity. Clinical presentation was described using RTI-related diagnostic codes, while severity was assessed using LOS, ICU admissions and in-hospital mortality.

First, we calculated the proportion of laboratory-confirmed RSV and hMPV hospital admissions across all ages with primary or secondary diagnostic codes for bronchitis/bronchiolitis, pneumonia, unspecified lower respiratory tract infection (LRTI), upper respiratory tract infection (URTI) and pathogen-specific diagnoses to describe the common clinical presentations of each virus (Supplementary Material S1).

Secondly, LOS was summarised descriptively for each virus by age groups. For each group, we reported the mean LOS with standard deviation and the median LOS with interquartile range (IQR). Because LOS distributions were typically right-skewed, both the mean and median LOS were reported. To further describe differences in prolonged hospitalisation between the two viruses, we also calculated the proportion of admissions with prolonged LOS within each age group. Prolonged LOS was defined as >5 days [18].

Thirdly, we estimated the proportion of laboratory-confirmed RSV and hMPV hospital admissions requiring ICU care (including HDU) and resulting in in-hospital death (CFR), expressed as percentages with 95% CI derived using non-parametric bootstrap resampling.

Owing to the rarity of these events, hospitalisations were aggregated across seasons, and broader age groups were used to improve estimate robustness for the severity analyses: young children (<5 years), older children and adolescents (5–17 years), young to middle-aged adults (18–59 years) and older adults (≥ 60 years). Owing to the small number of events (fewer than five in several groups) and variable testing policies across health boards, formal statistical comparisons between viruses were not performed for severity outcomes to avoid potentially misleading results. Instead, these outcomes were presented descriptively, using percentages with 95% CI to illustrate the observed burden.

We assessed temporal trends in monthly laboratory-confirmed hospitalisation episodes of RSV and hMPV. Seasonality was evaluated descriptively using time series plots of monthly case counts.

Seasons for each virus were defined using the Annual Average Percentage (AAP) method, which identifies the smallest set of months accounting for 75% of total seasonal cases [19]. Further details are provided in Supplementary Material S7. AAP analyses were additionally stratified by age group (<5 years and ≥ 60 years) as the highest overall burden (hMPV or RSV) was observed in these groups during the study period.