COVID-19 in Portugal: a retrospective review of paediatric cases, hospital and PICU admissions in the first pandemic year

Background COVID-19 is considered by WHO a pandemic with public health emergency repercussions. Children often develop a mild disease with good prognosis and the recognition of children at risk is essential to successfully manage paediatric COVID-19. Quality epidemiological surveillance data are required to characterise and assess the pandemic. Methods Data on all reported paediatric COVID-19 cases, in Portugal, were retrospectively assessed from a fully anonymised dataset provided by the Directorate General for Health (DGS). Paediatric hospital admission results were obtained from the DGS vaccine recommendations and paediatric intensive care unit (PICU) admission results from the EPICENTRE.PT group. Reported cases and PICU admissions from March 2020 to February 2021 and hospital admissions between March and December 2020 were analysed. Results 92 051 COVID-19 cases were studied, 50.5% males, average age of 10.1 years, corresponding to 5.4% of children in Portugal. The most common symptoms were cough and fever, whereas gastrointestinal symptoms were infrequent. The most common comorbidity was asthma. A high rate of missing surveillance data was noticed, on presentation of disease and comorbidity variables, which warrants a cautious interpretation of results. Hospital admission was required in 0.93% of cases and PICU on 3.48 per 10 000 cases. PICU admission for Multisystem Inflammatory Syndrome in Children (MIS-C) was more frequent in children with no comorbidities and males, severe COVID-19 was rarer and occurred mainly in females and infants. Case fatality rate and mortality rates were low, 1.8 per 100 000 cases and 1.2 per 1 000 000 cases, respectively. Conclusions The overall reported case incidence was 5.4 per 100 children and adolescents and <1% of cases required hospital admission. MIS-C was more frequent in patients with no comorbidities and males. Mortality and case fatality rates were low. Geographic adapted strategies, and information systems to facilitate surveillance are required to improve surveillance data quality.


Missing Data
The majority of the case records in these national surveillance data are missing information about the presence or absence of specific symptoms. While this fact is stated in the paper, the implications for analysis and interpretation of findings are not sufficiently appreciated by the authors. For example, the finding that children with comorbidities were more likely to have symptoms reported could have easily resulted from reporting bias, not from a true undelying difference in the population, and the authors have no information to allow them to rule out bias as an explanation for their findings.
Similarly, for comorbidities, we are told that approximately 2% of pediatric cases had a reported comorbidity. But the question of whether the remaining 98% of cases did NOT have comorbidities vs. were missing information about comorbidities is unanswered in the paper.
The high level of missingness should entail very cautious interpretation of associations and causation. Because missingness across multiple variables in a data record is not random, spurious associations can emerge. Pediatric patients who were not missing comorbidity data would also be much more likely to NOT be missing symptom data.
These data are not contextualized at all. How complete was the reporting of pediatric hospitalizations? Were these data directly linked to the case surveillance data, or did they come from an independent, unlinked data source? Is child/family socioeconomic position a barrier to hospitalization in Portugal? Does access to medical care vary geographically?
Hospitalization rates should be presented two ways -population rates (with the child population at risk as the denominator) and case rates (with the child cases as the denominator).
Infants should be separated from the 0-4 group, because both hospitalization and death rates are known to be much higher for infants than for children 1-4 years old.
The quality of the death data should be described and discussedwhat is known about out-of-hospital COVID-19 mortality, misclassification of cause of death (especially in the early days of the pandemic), and any urban-rural differences in quality of mortality surveillance?

Data Tables
As stated above, many of these tables should be reformulated with incidence rates, not just numerator case counts.
In Table 1, cases who were missing gender need to be accounted for in the table.
In Table 2, the total number of cases in each age group should be included in the top header row of the table.
In Table 3, the percents reported in the Female and Male columns are "row percents" and they should be "column percents" to match the way the percents are calculated in the Total column. Also include the totals for females and males in the header row.
Supplemental Table 2 is very misleading. The majority of cases with missing symptom reports are not included here. The conservative approach would be to assume those cases were asymptomatic -at the very least they should be included as a separate column. The percents as currently shown are uninformative. They should be replaced with row percents, to answer the question "What percent of cases at this year of age had symptoms reported in the data system?"

GENERAL COMMENTS
This is an important documentation. However, no information was available for a large number of patients (>62%). These patients have also been included in denominator for calculating the percentage of patients having various symptoms. This is not appropriate. Some recalculations are needed.
More information about patients with MISC and death would be better.
In Table-2, provide total number of patients studied in each age group.
In Table-3, the calculations of percentages are confusing. For example, the number of co-morbidities under the column Total are 2% but it becomes 42.6% for Females and 57% for Males.

REVIEWER
Reviewer name: Dr. Mariana Poppe Institution and Country: Hospital Beatriz Angelo, Portugal Competing interests: None REVIEW RETURNED 21-Apr-2022

GENERAL COMMENTS
GENERAL COMMENT: The theme of the article is very relevant. Publication of national data is helpful to understand the "bigger picture" of this pandemic and might help to ground future measures on solid scientific evidence. The title reflects the content of the article appropriately. The abstract is well-constructed, needing small adjustments. The background section correctly presents the context of the study. The objectives are clearly defined. More information concerning the methods and analysis is required for better understanding and interpretation of the results of the study. Further statistical analysis could enhance the quality of the study. The presentation of the results could be improved in order to be coherent with the abstract and discussion. A major limitation concerning the study results hinders the conclusions of the article. The discussion is well written, although with a few statements without the appropriate grounding that need to be revised. The limitations of the study are portrayed. The conclusion responds to the proposed objectives, needing small adjustments.
ASTRACT -Page 2, line 16: The time period to which the study refers is not mentioned, it could enrich the abstract to include that information.
-Page 4, line 19: it reads "dyspnoea", while in the abstract the authors used "dyspnea" -the authors should use one or the other spelling coherently throughout the article.
BACKGROUND -Page 5, lines 27-29: "the role of the pediatric population in SARS-CoV-2 spread in the community have been considered of major interest" -is this worth mentioning in the background, since your research does not approach this any further nor is it ever mentioned again during the article? -Page 5, lines 33-35: "A small number of children need hospital admission, and of these, a minority required ICU admission either due to severe COVID-19 or multisystem inflammatory disease in children (MIS-C)" -the verb 'need' is in the present, the verb 'required' in the past tense. -Page 7, line 47: The major limitation of the study is portrayed here -62.5% of cases without information regarding symptoms. This limitation is utterly relevant, as it has the potential to hinder all the other conclusions of the study concerning the presenting symptoms -one could argue that the article only analyses 37.5% of cases in Portugal when approaching symptoms. It could be that gastrointestinal symptoms were predominant in the two thirds of patients without reported symptoms, or that any other information collected on such a large proportion of the sample could change the conclusions of this study. Given the relevance of this matter, one could argue that the analysis of the study should focus only on the data from completely filled-out forms on the SINAVE platform, excluding the incomplete forms, to avoid misleading interpretations. I stand with great doubt about whether this is information should be disclosed right from the start in the abstract to avoid misleading the reader.
-Page 7, line 40: CLINICAL PRESENTATION. The gastrointestinal symptoms are barely referred in the results, however they are enhanced in the abstract. If they are mentioned in the 'results' part of the abstract ("Gastrointestinal symptoms were infrequent."), this information should be contained in the results section of the article.
-Page 8, line 20: COMORBIDITIES. Were the comorbidities an obligatory answer while filling out the SINAVE form, or is it possible that they are also under-reported due to incomplete SINAVE forms?
-Page 8, line 37: "Patients with comorbidities presented symptoms more frequently than patients with no comorbidities." Is there a statistically significant difference? Were comparison tests performed?
-Page 9, line 24: "Two deaths caused by COVID-19 were reported by the DGS during this study." These deaths were not in PICU admitted patients?

DISCUSSION
-Page 10, line 34: "significant differences between sex were observed mainly between 16 and 17, and predominantly in females." This information belongs in the results section, it was not mentioned there. 'Significant differences' implies that comparison tests were performed, but they are not mentioned. Or is it based only on the information in -Page 10, line 41: "geographic" should read 'geographically'.
-Page 10, line 47: "Though" does not make sense within the sentence, revise the phrase.
-Page 10, line 52: "diverse backgrounds Brazil, Angola" should read 'diverse backgrounds such as Brazil,' … or the names of countries in brackets.
-Page 10, last paragraph: When referring to the study's results, verbs are used in the present tense (are, reflects, etc) and in the past tense (were) -revise throughout the article.
-Page 10, line 57: "the pediatric and adult population peaks cooccurred simultaneously". Lacking a reference for this statement.
-Page 11, line 23: "such has been reported elsewhere" -'as has been reported in other studies'.
-Page 11, line 31: "As found in our results (38.1%), the combination of symptoms, fever and cough 30% has also been documented." Revise formulation of the sentence, hard to understand.
-Page 11, line 43: "Unknown clinical presentations were frequent and while the majority of these are likely to be asymptomatic patients in which no symptoms were denied on the digital platform SINAVE". Based on which grounds (or references) can the authors state that these are likely asymptomatic patients? One could claim that is a possibility, but not affirm it without objective data supporting the affirmation. This is the main limitation of the study, and thus should be portrayed objectively. Once more, this evidences why characterizing the SINAVE platform in the methods section is important, otherwise readers will not be able to fully understand what is being discussed in this section.
-Page 12, line 23: "Lisbon Hospital". Clarifying it is a level 3 hospital and one of the two reference centres for paediatric Covid-19 patients during the early stage of the pandemic might shed light on the high percentage of risk factors found.
-Page 13, line 11: "as less frequent than" -'is less frequent than' -Page 13, line 52: "The authors found unusual the lack of surveillance data in the analyzed dataset" -'The authors found the lack of surveillance data in the analysed dataset unusual' CONCLUSIONS -Page 14, line 23: "analyses" -'analysis'.
-Page 14, line 35: "Case fatality rates and mortality rates were extremely low when comparing with other countries". The authors compared the fatality rate of the study with data from England (which was higher) and Spain (which was lower) -thereby this conclusion is not accurate.
-The improvement in the reporting system could also be mentioned in the conclusions section, being an important report of this study.
TABELES AND FIGURES -

VERSION 1 -AUTHOR RESPONSE
Dear Editor-in-Chief of the BMJ Paediatrics Open, We now resubmit the original manuscript entitled "COVID-19 in Portugal: a retrospective review of paediatric cases, hospital and PICU admissions in the first pandemic year" to be considered for publication in the BMJ Paediatrics Open. Firstly, the authors would like to thank the editor-in-chief and the reviewers for their assessment and recommendations regarding our manuscript. Paediatric COVID-19 national studies and results are extremely relevant to fully characterize and preserve our knowledge of this pandemic. The authors set out with very specific objectives to characterize this disease in children and to compare our results with those from other countries. However, while trying to answer these questions, others rose, connected to data quality of two variables we had extensively analyzed. At the time of submission, the authors did not fully appreciate this data limitation. However, we have taken upon reflection all the reviews received and have updated our manuscript accordingly. In particular, the authors have curtailed their analysis, comparisons and conclusions connected to these variables. We removed comparisons on disease presentation and the presence of comorbidities and removed the mention of comorbidity from the section "What this study adds". Furthermore, without taking the focus of our study away from a paediatric COVID-19 case description and analysis, and as recommended by one of the reviewers, we have conducted a missingness analysis. This analysis can prove an important tool to improve surveillance data in our country. Major changes affecting all sections and tables of the manuscript have been taken in-keeping with peer-review suggestions. Given the lengthy revision, the authors highlighted revised segments rather than add line-by-line tracking. Point-by-point answers to the reviewers are provided in separate word documents. We reaffirm this work has neither been published nor is currently submitted for publication elsewhere in whole or in part and all the listed authors have contributed significantly to this work and approved it. Altogether, we are confident that our contribution is suitable for publication in the BMJ Paediatrics Open, and we look forward to hearing back in due course. Further statistical analysis could enhance the quality of the study. Response: The authors have updated all sections of the manuscript. As suggested, the authors have added information to the methods section and, given the high level of missingness on two variables extensively studied in the previous version, have tailored their analysis, limiting comparions and conclusions from these variables, in accordance with reviewers. A missingness analysis was similarly conducted.
- [Comment 4] The presentation of the results could be improved in order to be coherent with the abstract and discussion. A major limitation concerning the study results hinders the conclusions of the article. Response: The authors reviewied the abstract and discussion and added information to the results session, which was lacking. We reviewed our major limitation, associated with the high level of missingness on 2 variables studied in our analysis, and decided to scaleback our comparisons from these variables. Concomitantly, we added a missingness analysis which provides an insight on the causes and potencial improvements to be achieved with the reporting system. Nevertheless, the authors tried to maintain their main goal to describe peaditric COVID-19 in Portugal.
-[Comment 5] The discussion is well written, although with a few statements without the appropriate grounding that need to be revised. The limitations of the study are portrayed. 2 Response: The authors have extensively revised the discussion and segments needing referencing were edited in accordance. In particular, the discussion on missing epidemiological surveillance data was replaced with the discussion of the missingness analysis. -[Comment 6] The conclusion responds to the proposed objectives, needing small adjustments. Response: The authors have ehnanced the conclusion session considering the need to improve surveillance data. ABSTRACT -[Comment 7] Page 2, line 16: The time period to which the study refers is not mentioned, it could enrich the abstract to include that information. Response: The authors would like to thank the reviewer for the suggestion and have updated the methods section of the abstract accordingly. -[Comment 8] Page 4, line 19: it reads "dyspnoea", while in the abstract the authors used "dyspnea" -the authors should use one or the other spelling coherently throughout the article. Response: The authors would like to thank the reviewer for the correction. Although this wording has been removed from the abstract, a consistent spelling throughout the manuscript has been attempted. BACKGROUND -[Comment 9] Page 5, lines 27-29: "the role of the pediatric population in SARS-CoV-2 spread in the community have been considered of major interest" -is this worth mentioning in the background, since your research does not approach this any further nor is it ever mentioned again during the article? Response: The authors would like to thank the reviewer for this comment, with which we agree, and have removed this sentence from the background section accordingly. The major limitation of the study is portrayed here -62.5% of cases without information regarding symptoms. This limitation is utterly relevant, as it has the potential to hinder all the other conclusions of the study concerning the presenting symptoms -one could argue that the article only analyses 37.5% of cases in Portugal when approaching symptoms. It could be that gastrointestinal symptoms were predominant in the two thirds of patients without reported symptoms, or that any other information collected on such a large proportion of the sample could change the conclusions of this study. Given the relevance of this matter, one could argue that the analysis of the study should focus only on the data from completely filled-out forms on the SINAVE platform, excluding the incomplete forms, to avoid misleading interpretations. I stand with great doubt about whether this is information should be disclosed right from the start in the abstract to avoid misleading the reader. Response: As stated previously, by the authors, the high level of missingness in our analysis is an important constraint. To minimize these limitations the authors have curtailed their analysis, comparisons and conclusions regarding disease presentation and the presence of comorbidity variables. We reviewed our analysis to exclude the missing cases from percentages calculi. Furthermore, without taking the focus of our study away from a paediatric COVID-19 case description and analysis we have conducted a missingness analysis. Information regarding the level of missingness has been included in the abstract and also in the beginning of the discussion. The authors were able to reduce missing information by recoding the disease presentation and comorbiditys variables, as described in the method section. Were the comorbidities an obligatory answer while filling out the SINAVE form, or is it possible that they are also under-reported due to incomplete SINAVE forms? Response: The authors would like to thank the reviewer for this pertinent comment, as stated previously, the high level of missingness in the disease presentation and the presence of comorbidities variables is a major limitation which the authors did not fully appraise initially. None of these variables are mandatory in the SINAVE collected information. Comorbidities were reported (as present or absent) in 19,680 cases (21.4%) which is clearly inadequate. From our missingness analysis, reported comorbidity information decreased as the pandemic progressed and varied between districts and sexes. In light of this, the authors excluded missing information from our percentages and abstained from comparing results obtained from these variables. Information forewarning to the level of missingness has been included in the abstract and also at the beginning of the discussion. The authors would like to apologize for this inatention. This analysis was performed at a previous version of the manuscript that was removed and the text referring to it was not updated. 7 -[Comment 29] Page 11, line 23: "such has been reported elsewhere" -'as has been reported in other studies'. Response: The authors would like to thank the reviewer for the correction and have updated the manuscript accordingly. -[Comment 30] Page 11, line 31: "As found in our results (38.1%), the combination of symptoms, fever and cough 30% has also been documented." Revise formulation of the sentence, hard to understand. Response: The authors would like to thank the reviewer for the correction, this sentence has been removed from the manuscript. -[Comment 31] Page 11, line 43: "Unknown clinical presentations were frequent and while the majority of these are likely to be asymptomatic patients in which no symptoms were denied on the digital platform SINAVE". Based on which grounds (or references) can the authors state that these are likely asymptomatic patients? One could claim that is a possibility, but not affirm it without objective data supporting the affirmation. This is the main limitation of the study, and thus should be portrayed objectively. Once more, this evidences why characterizing the SINAVE platform in the methods section is important, otherwise readers will not be able to fully understand what is being discussed in this section. Once more, this evidences why characterizing the SINAVE platform in the methods section is important, otherwise readers will not be able to fully understand what is being discussed in this section. Response: We thank the reviewer for this comment and to better assess and analyze the missingness observed, the authors conduced a missingness analysis. This shows missing information on disease presentation and comorbidities across districts, time and sexes were not random and significant differences were observed between groups. As previously stated, missing information, not selected from the SINAVE platform, increased as the pandemic progresses and differences were observed between districts, this data was composed of negative and positive answers that were not selected. To minimize this limitation the authors have update the analysis and manuscript thoroughly. Missing cases have been removed from denominator calculations, the analysis and comparisons of these variables has been curtailed and conclusions have been scaledback. : "Case fatality rates and mortality rates were extremely low when comparing with other countries". The authors compared the fatality rate of the study with data from England (which was higher) and Spain (which was lower) -thereby this conclusion is not accurate. Response: We thank the reviewer for this comment, case fatality rates were only compared with the english results. Case-fatality rates in England were higher than the ones we report. Mortality rates were compared with spanish, english, US results and with combined results from 7 countries. -[Comment 39] The improvement in the reporting system could also be mentioned in the conclusions section, being an important report of this study. 9 Response: The authors would like to thank the reviewer for this suggestion and have updated the manuscript accordingly. Improving epidemiological surveillance information must be a priority and this information should be included in the conclusion section. ] The topic of this paper, the descriptive epidemiology of COVID-like to thank the reviewer for pointing our critical methodology flaws. Care has been taken to improve our analysis and manuscript by addressing these concerns. Our main goal was to, enthusiatically, describe peaditaric COVID-19 presentations and outcomes of patients in our country. Unfortunately, the surveillance data utilized in this analyzis had high levels of missingness which greatly hampered part of the analysis conducted. We did not fully appreciate the ramifications of this in our initial analysis, in particular when describing and comparing variables with high levels of missingness. We have reflected upon your comments, have partially reanalyzed the data, added the missingness analysis and have made major manuscript editing. This analysis can prove an important tool to improve surveillance data in our country. We hope this revised version is inaccordance with your comments and suggestions. 1. Missing Data -[Comment 3] The majority of the case records in these national surveillance data are missing information about the presence or absence of specific symptoms. While this fact is stated in the paper, the implications for analysis and interpretation of findings are not sufficiently appreciated by the authors. For example, the finding that children with comorbidities were more likely to have symptoms reported could have easily resulted from reporting bias, not from a true undelying difference in the population, and the authors have no information to allow them to rule out bias as an explanation for their findings. Response: The authors would like to thank the reviewer for these comments which greatly impacted our assessment and analysis. The high level of missingness is a major limitation and, as you stated, interconnecting symptoms and comorbidities to obtain results from this data was not appropriate and likely introduced bias to our analysis. To minimize this limitation the authors have curtailed their analysis, comparisons and conclusions regarding disease presentation and the presence of comorbidity variables. We reviewed our analysis to exclude the missing cases from percentages calculi. Furthermore, without taking the focus of our study away from a paediatric COVID-19 case description and analysis we have conducted a missingness analysis, as you suggested. Important information was obtained and statistical differences between district and month were observed. Information regarding the level of missingness has been added throughout the manuscript and in particular in the abstract and at the beginning of the discussion. The need for improved surveillance information has also been highlighted in the abstract, discussion and conclusion. Given our study focuses on reported cases, hospital and PICU admissions and deaths with this review we have decreased our focus on the reported cases presentation and description and refocused on the other components of the analysis. -[Comment 4] Similarly, for comorbidities, we are told that approximately 2% of pediatric cases had a reported comorbidity. But the question of whether the remaining 98% of cases did NOT have comorbidities vs. were missing information about comorbidities is unanswered in the paper. Response: The authors would like to thank the reviewer for this comment. As stated previously, the high level of missingness in the disease presentation and in the presence of comorbidities variables is a major limitation. From an analysis point, we recoded the comorbidity variable to include the general comorbidity question and specific disease questions, and concluded comorbidities were reported (as present or absent) in 19,680 cases (21.4%) and were missing in 72,371 (78.6%) of cases. Furthermore, patients who were not missing comorbidity were, like you pointed out, more likely to also not be missing symptoms. In retrospect, this is clearly inadequate and the authors amended their objectives, methodology, analysis and consequent conclusions to minimize these limitations. The missingness analysis showed statistical significant differences between sexes, districts and months in comorbidity reporting and reported comorbidity information decreased as the pandemic progressed. This confirms comorbidities are underreported in our study. In light of the high rate and pattern of missingess observed, the authors abstained from comparing results obtained from these variables and excluded missing information from the calculated percentages. Information forewarning to the level of missingness has been included in the abstract and also at the beginning of the discussion. -[Comment 5] The high level of missingness should entail very cautious interpretation of associations and causation. Because missingness across multiple variables in a data record is not random, spurious associations can emerge. Pediatric patients who were not missing comorbidity data would also be much more likely to NOT be missing symptom data. Response: The authors would like to thank the reviewer for this comment. Upon reflection over our previous analysis, and as previously recognized by the authors, spurious associations can become apparent and probably have... Thank you for your thorough review and comments on this issue. We updated our study and manuscrpt to minimize this bias and opted to simply describe these variables as results, avoid comparisons between them and concomitantly describe our limitations. -[Comment 6] Furthermore, the authors should conduct and report a systematic analysis of the pattern of missingness by age and gender, and possibly by geographic area and time period as well. Did the quality of surveillance data reporting improve over time? Was it higher or lower in major cities vs. rural areas? Response: The authors would like thank the reviewer for suggesting a systematic analysis of the pattern of missingness which we have included in our manuscript. This analysis has protrayed very interesting results, showing statistical differences between month and district for disease presentation and between month, district and sex for the presence of comorbidities. Two districts presented the highest recorded rates of information: Lisbon, a mainly urban district and Faro. The districts with the lowest recorded information were mainly rural. Furthermore, as the pandemic progressed the missingness rates increased, as discussed in the manuscript. The authors are grateful for this suggestion and hope this analysis can prove an important tool to improve surveillance data in our country. 2. Case counts vs. incidence -[Comment 7] The authors conflate numerator counts of cases with "incidence." Incidence must be expressed as a rate, with an appropriate census-derived population denominator. This is a very elementary error. The authors describe incidence increasing by single years of age in the text, when the data only show increasing numerator counts -there are no incidence rates presented anywhere in the paper. Similarly, the map and discussion of geographic patterns is based on numerators only. Obviously there will always be a larger number of cases in cities than in rural areas. The authors have not investigated whether the incidence RATES are higher in urban vs. rural areas. Response: The authors would like to thank the reviewer for this comment. We have enhanced the analysis to include a census-derived population denominator when characterizing the cases sex and age, we have also calculated hospital and PICU admissions similarly. Census derived population information is, unfortunately, not available to compare with reported case counts geographically. SINAVE reports geographic locations based on district information, which is the portuguese tradicional and commonly used location referencing, whereas Portugal Census geographic information is collected as NUTS. NUTS were created by Eurostat with the aim at harmonizing statistical data between european countries for statistical purposes. Nevertheless we have updated the manuscript to include incidence rates, as suggested, with important quality gains in our analysis. 3. Hospitalization and mortality analyses and discussion -[Comment 8] These data are not contextualized at all. How complete was the reporting of pediatric hospitalizations? Were these data directly linked to the case surveillance data, or did they come from an independent, unlinked data source? Is child/family socioeconomic position a barrier to hospitalization in Portugal? Does access to medical care vary geographically? Response: The authors would like to thank the reviewer for these comments. We did not have access to this data or database, we only had access as results published by the National