Healthcare utilization following SARS-CoV-2 infection in children and adolescents with chronic conditions: An EHR-based Cohort Study from the RECOVER Program ============================================================================================================================================================= * Nathan M Pajor * Vitaly Lorman * Hanieh Razzaghi * Abigail Case * Priya Prahalad * Seuli Bose Brill * Qiong Wu * Yong Chen * Jason Block * Payal B Patel * Suchitra Rao * Asuncion Mejias * Deepika Thacker * Ravi Jhaveri * L. Charles Bailey * Christopher B. Forrest * Grace M. Lee ## Abstract **Background** Chronic medical conditions are a risk factor for moderate or severe COVID-19 in children, but little is known about post-acute sequelae of SARS-CoV-2 infection (PASC) in children with chronic medical conditions (CMCs). To understand whether SARS-CoV-2 infection led to potential exacerbation of underlying chronic disease in children, we explored whether children with CMCs had increased healthcare utilization in the post-acute (28 days after infection) period compared to children with CMCs without SARS-CoV-2 infection. **Methods** We conducted a retrospective, matched-cohort study using electronic health record data collected from 8 pediatric health care systems participating in the PEDSnet network. We included children <21 years of age with a wide array of chronic conditions, defined by the presence of diagnostic codes, who were diagnosed with COVID-19 between March 1, 2020 and February 28, 2022. Cohort entry was defined by presence of a positive SARS-CoV-2 PCR test (polymerase chain reaction or antigen) or diagnostic codes for COVID-19, PASC or MIS-C. A comparison cohort of patients testing negative or without these conditions was matched using a stratified propensity score model and exact matching on age group, race/ethnicity, institution, test location, and month of cohort entry. A negative binomial model was used to examine our primary outcome: composite and setting-specific (inpatient, outpatient, ED) utilization rate ratios between the positive and comparison cohorts. Secondary outcomes included time to first utilization in the post-acute period, and utilization stratified by severity at cohort entry. **Results** We identified 748,692 patients with at least one chronic condition, 78,744 of whom met inclusion criteria for the COVID-19 cohort. 96% of patients from the positive cohort were matched. Cohorts were well-balanced for chronic condition clusters, total number of conditions, time since first diagnosis, baseline utilization, cohort entry period, age, sex, race/ethnicity and test location. We found that among children with chronic medical conditions, those with COVID-19 had higher healthcare utilization than those with no recorded COVID-19 diagnosis or positive test, with utilization rate ratio of 1.21 (95% CI: 1.18-1.24). The utilization was highest for inpatient care with utilization rate ratio of 2.03 (95% CI: 1.85-2.23) but the utilization was increased across all settings. Hazard ratios estimated in time-to-first-utilization analysis mirrored these results. Patients with severe or moderate acute COVID-19 illness had greater increases in utilization in all settings than those with mild or asymptomatic disease. **Conclusions** We found that care utilization in all settings was increased following COVID-19 in children with chronic medical conditions in the post-acute period, particularly in the inpatient setting. Increased utilization was correlated with more severe COVID-19. Additional research is needed to better understand the reasons for higher care utilization by studying condition-specific outcomes in children with chronic disease. ## Introduction Throughout the coronavirus disease 2019 (COVID-19) pandemic, children with chronic medical conditions (CMCs) have shown increased rates of severe disease compared to children without chronic conditions [1-4]. Previously, we described increased moderate and severe acute COVID-19 in children with asthma, cancers, cardiac disorders, diabetes mellitus, immunologic conditions, mental health conditions, obesity, prematurity, renal disorders and sickle cell disease[5]. Far less is known about the long-term outcomes of children with CMCs after SARS-CoV-2 infection. In particular, it is not clear whether underlying CMC control or trajectory worsens following SARS-CoV-2 infection. These changes in children with CMCs may represent unique post-acute sequelae of SARS-CoV-2 infection (PASC), defined as ongoing, relapsing, or new symptoms, or other health effects occurring after the acute phase of SARS-CoV-2 infection (i.e., present four or more weeks after the acute infection). Prior studies have found exacerbation of specific CMCs during the acute phase of COVID-19 that could have implications on future control. There is evidence that adults and children with established type 1 and type 2 diabetes mellitus, particularly those with Black race, have a higher incidence of diabetic ketoacidosis during COVID-19 illness[6-8]. There is mixed evidence in asthma with multiple studies reporting worsening asthma symptoms and rates of hospitalization during COVID-19 illness [9-13] while other studies report a protective or neutral effect[14-18]. Obesity predisposes children to increased severity of COVID-19 and is a risk factor for critical illness in adolescents with COVID-19[19, 20]. Pediatric patients with sickle cell disease (SCD), especially those with comorbidities, have experienced increased hospitalization and mortality rates compared to the general population[21]. Numerous case series and case reports have documented vaso-occlusive episodes and acute chest syndrome among children with SCD and COVID-19[22-26]. However, these studies do not describe the continued impact on chronic disease control in the post-acute illness period. Given the lack of understanding around PASC in children with chronic conditions and the wide variety of chronic conditions impacted by SARS-CoV-2 infection, we evaluated whether the presence of any CMC led to increased healthcare utilization in the post-acute period, 28 days and beyond, following COVID-19. We hypothesized that COVID-19 may exacerbate CMCs and that these exacerbations will manifest as increased healthcare utilization in the post-acute period following illness. Thus, this study aims to evaluate healthcare utilization among children with CMCs using EHR data and compare children with any CMCs and positive COVID-19 testing with matched controls who had no documented evidence of COVID-19. This study is part of the NIH Researching COVID to Enhance Recovery (RECOVER) Initiative, which seeks to understand, treat, and prevent the post-acute sequelae of SARS-CoV-2 infection (PASC). For more information on RECOVER, visit [https://recovercovid.org/](https://recovercovid.org/) ## Methods ### Setting We obtained EHR data from institutions participating in PEDSnet ([pedsnet.org](http://pedsnet.org)), a national network of pediatric healthcare systems that share clinical data to conduct observational research, clinical trials, and population surveillance[27]. EHR data from inpatient, outpatient and emergency department encounters within each participating health system are standardized to the PEDSnet common data model, an extension of the Observational Medical Outcomes Partnership common data model. This network is part of the RECOVER consortium and the data set is a subsetof the RECOVER EHR data. The participating institutions are: Children’s Hospital of Philadelphia, Cincinnati Children’s Hospital Medical Center, Children’s Hospital of Colorado, Ann & Robert H. Lurie Children’s Hospital of Chicago, Nationwide Children’s Hospital, Nemours Children’s Health System (includes sites in Delaware and Florida), Seattle Children’s Hospital, and Stanford Children’s Health. Annually, these institutions provide service to 3.3% of the nation’s children. The Children’s Hospital of Philadelphia’s institutional review board designated this study as not human subjects research and the need for consent was waived. The PEDSnet COVID-19 Database Version 2022-03-17 was used. ### Study Design This was a retrospective, observational, matched-cohort study. We evaluated all individuals <21 years of age between March 1, 2020 and February 28, 2022 and defined a COVID-19 positive cohort as presence of a positive SARS-CoV-2 PCR test performed in the outpatient, emergency department (ED), or inpatient setting, or presence of any specific diagnosis code for COVID-19 associated with an inpatient or ED encounter. We restricted diagnosis codes to inpatient and ED encounters to limit diagnoses that may not be associated with an active SARS-CoV-2 infection. We also included any diagnosis code of multi-system inflammatory syndrome in children (MIS-C) or post-acute sequelae of COVID-19 associated with any visit type. The cohort entry date was defined as the date of the first positive test or the date of first diagnosis code, whichever occurred earlier. The comparison cohort was identified based on presence of at least one negative SARS-CoV-2 PCR test over the same date range with no positive test results recorded, and no diagnosis code for COVID-19, MIS-C or PASC in any setting. For children with multiple negative tests, a single, randomly selected negative test date was defined as the date of cohort entry. We evaluated a 3-year period prior to cohort entry date to determine the presence of any chronic medical condition. For both the COVID-19 cohort and the comparison cohort, children were only included if they had at least one in-person encounter for a chronic medical condition in the 3 years prior to cohort entry date. Individuals were followed from the date of cohort entry through February 28, 2022. ### Chronic Condition Definition We used a previously developed set of condition clusters, generated by categorizing ICD-10-CM codes into deduplicatd clusters and based on the Agency for Healthcare Research and Quality (AHRQ) Clinical Classification Software Refined[28] 1. The clusters were reviewed by pediatric physicians participating in the RECOVER initiative. Of the 439 clusters created, 205 were categorized as describing chronic conditions and used here. We separately utilized the Pediatric Medical Complexity Algorithm (PMCA) Version 2.0 categorization to identify conditions which the algorithm labels as progressive[29]. ### Matching We developed a propensity score model using covariates that included the presence of diagnoses within our chronic condition clusters and the presence of a progressive chronic condition defined using the PMCA categorization. The propensity score also incorporated the number of chronic condition clusters (1, 2, or 3+), the number of years since first chronic diagnosis, the baseline number of visits per month prior to cohort entry, age group, sex, race/ethnicity and test location. The propensity score model utilized only clusters that contained diagnoses for at least 500 patients; 150 of the 205 clusters were utilized in the propensity score model. Patients were stratified into 5 strata by quantile of propensity score. Within each stratum, we performed exact matching at the patient level based on age group (<1 year old, 1-4, 5-11, 12-15, 16-20), race/ethnicity, institution, test location (inpatient, ED, outpatient office visit, outpatient test only). Additionally, we performed exact matching based on month of cohort entry to control for effect of disease prevalence, variant type, and vaccination status. We assessed balance between cohorts within each stratum using the weighted standardized mean difference. ### Variables Our primary outcome was composite healthcare utilization based on the total number of visits and, separately the number of visits in the inpatient, outpatient, and ED settings from 28 days after cohort entry through the end of the follow-up period. Our secondary outcome was time to any visit and time to first inpatient, ED, and outpatient visit following cohort entry. Covariates included age group (<1, 1-4, 5-11, 12-15, 16-20) at cohort entry, sex, race/ethnicity (white non-Hispanic, Black non-Hispanic, Hispanic, Asian non-Hispanic, multiracial non-Hispanic, other non-Hispanic), COVID-19 status (positive SARS-CoV-2 PCR test or presence of COVID-19 diagnosis code vs. negative SARS-CoV-2 PCR test), cohort entry period (3-month intervals beginning March 1, 2020), test location (inpatient, ED, outpatient office visit, outpatient test only,), measured obesity (defined as age-sex-standardized BMI z-score ≥95th percentile based on weight measurement at the time of SARS-CoV-2 testing and height within 60 days of cohort entrance), number of chronic conditions (1, 2, or 3+), years since first chronic diagnosis, and number of visits per month for existing chronic conditions during the 3-year time period prior to cohort entry. ### Statistical Analyses To model our outcomes, which consisted of visit count data, we used a negative binomial model with an offset to account for non-uniform follow-up times across individuals. Within each propensity score stratum, the dispersion parameter estimate was substantially larger than 1, supporting the choice of this model over the corresponding Poisson model. The model was fit on each of the 5 propensity score strata and the resulting estimates were pooled by taking the mean, weighted by the proportion of the number of matched units within each stratum divided by the total number of matched units across all strata. The pooled variance used to estimate pooled confidence intervals was computed using a weighted average (again accounting for non-uniform sample sizes) of the variances in each stratum. Propensity scores were estimated using a logistic regression to estimate probability of COVID-positivity, as described above. Our methods are doubly robust[30], as the covariates used in the negative binomial model estimating effect of COVID-positivity were used in the propensity score model and used in exact matching. The weights which resulted from the exact matching procedure in each stratum were used to fit each model as a weighted regression (with unmatched units receiving weights of zero). Because our model used an offset for follow-up time, our estimates which correspond to the exponentiated coefficient of COVID-positivity in the model are interpreted as visit utilization rate ratios between the positive and negative cohorts. Exponentiated coefficients of categorical variables (for example, age category) are interpreted similarly, as ratios of visit utilization rates between each indicated level and the referent level for that variable, and exponentiated coefficients of continuous variables (for example, years since first chronic diagnosis) are the estimated increase in utilization rate per a one unit increase in the covariate. The hazard ratios in the supplementary appendix were estimated using a Cox proportional hazards model using time to first event as the outcome. The model was fit separately for each stratum (using weights arising from exact matching and the same covariate adjustments as in the negative binomial models) and estimates pooled across strata as described above. To estimate utilization rate ratios for different levels of COVID severity we utilized a previously developed index in which severity was categorized as asymptomatic, mild (symptoms such as abdominal pain, fever, cough), moderate (COVID-19–related conditions such as gastroenteritis, dehydration, and pneumonia), or severe (unstable COVID-19–related conditions such as acute respiratory distress syndrome, acute kidney injury, ICU admission, or mechanical ventilation)[5]. We estimated propensity scores using the same model as discussed above, then used inverse probability of treatment weighting (IPTW) in the same negative binomial model applied to the full cohort, capping weights at the 95th percentile. We evaluated balance using weighted SMD’s. To investigate how the positive/negative utilization rate ratios differ across age, site, race/ethnicity, cohort entry period, and visit type at time of cohort entry, we conducted an exploratory analysis. Using the same propensity score stratification and exact matching approach as in the case of our overall utilization rate ratio estimates, for each covariate we fit a negative binomial model including an interaction term with that covariate. We then used linear contrasts (pooled across propensity score strata as above) to report the utilization rate ratios at each level of the covariates. We additionally conducted exploratory analyses into utilization across existing chronic conditions. To estimate utilization rate ratios for each chronic condition cluster, we fit the negative binomial model with the same covariates (omitting matching due to small sample sizes) once for each sub-cohort defined by patients with that chronic condition. We stratified the conditions into high, medium and low prevalence based on number of encounters for total, inpatient, outpatient and ED divided into terciles. Within each tercile, we ranked based on utilization rate ratio. To evaluate whether utilization related to COVID-19 was unique to the chronic disease population, we repeated our analysis on a separate cohort of patients who had no chronic disease codes recorded. We adjusted for age group, sex, race/ethnicity, cohort entry period, test location, site, and measured obesity. Analyses were conducted using *R* version 2022.02.1 Build 461 (with the libraries *MatchIt, WeightIt, survival, MASS, emmeans, and cobalt*). ## Results ### Study Population From March 1, 2020 through February 28, 2022, we identified 1,042,066 patients 0-20 years of age within PEDSnet with any COVID PCR test, COVID-19 diagnosis, PASC diagnosis or MIS-C diagnosis. Of those, 557,418 patients had at least one chronic condition in the three years prior to cohort entry. Among those with chronic conditions, we identified 78,744 patients with evidence of SARS-CoV-2 infection or COVID-19 related diagnosis and 478,674 without evidence of COVID-19 (Table 1). The most common chronic condition clusters overall were allergies; asthma; sleep-wake disorders; implant, device or graft related encounters; and esophageal disorders. Our matched cohorts were similar. Most patients across both cohorts were tested in the outpatient setting either through a testing facility (42.2%) or outpatient office visit (31.3%). The majority (97.9%) of children in the positive cohort were included based on positive SARS-CoV-2 PCR test with 2% included based on a COVID-19 diagnosis code in the inpatient or ED setting, or MIS-C or PASC diagnosis codes. View this table: [Table 1.](http://medrxiv.org/content/early/2022/07/09/2022.07.08.22276768/T1) Table 1. Demographic and clinical characteristics of children with presence of any chronic medical condition by COVID-19 infection status, prior to matching. 1Standardized Mean Difference (SMD), computed as the difference in means between the positive and negative cohorts divided by the pooled standard deviation. The SMDs reported here prior to propensity score stratification and exact matching. An SMD less than 0.1 in absolute value is often taken to indicate a negligible difference in means between the two groups[33] Prior to matching, the positive cohort (vs. comparison) was more likely to be older, female, Black non-Hispanic, and tested in the ED setting. We also found differences between groups in cohort entry period due to the variation in disease prevalence (Figure 1). The largest number of patients entered the positive cohort between January and February 2022 when the omicron variant (B.1.1.529) was most prevalent in the U.S. This resulted in a median duration of follow-up of 0.43 years (IQR: 0.15-1.11) in the positive cohort and 0.78 years (IQR: 0.38-1.27) in the comparison cohort. These differences were addressed through matching. ![Figure 1:](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2022/07/09/2022.07.08.22276768/F1.medium.gif) [Figure 1:](http://medrxiv.org/content/early/2022/07/09/2022.07.08.22276768/F1) Figure 1: Timing of Cohort entry for COVID-19 positive and negative patients. ### Matching Following propensity score estimation, we found excellent balance between the COVID positive and comparison cohorts for chronic conditions, even before exact matching, with nearly all clusters demonstrating standardized mean difference (SMD) of ±0.1. Exceptions were obesity, allergies, and mental health which showed slightly higher rates in the positive cohort vs. comparison cohort, but with standard mean differences of ±<0.2. Following stratification and exact matching the SMD was ±0.1 in each stratum for all covariates. As each chronic disease cluster contains several different diagnosis groups, we developed condition subgroups of the 5 most common clusters and found that the balance between groups persisted even at the subgroup level. Additionally, the groups were well balanced (SMD ±0.1) for the baseline number of visits per year, years since first diagnosis of chronic condition and total number of chronic disease clusters. After matching, 96% of individuals were matched in the positive cohort and 70% were matched in the comparison cohort. Similarly, in the IPTW model used to estimate utilization rate ratios for different levels of COVID-19 severity, we found that our weights, estimated from the propensity score model and capped at the 95th percentile, resulted in excellent balance (weight SMD ±0.1) across almost all covariates (for baseline visits per year, the weighted SMD was 0.22). ### Health Services Utilization Overall and by Setting The rate ratio of total visits between the positive cohort and the comparison cohort, in any context, from 28 days after cohort entry through the end of the available follow-up period was 1.21 (95% CI: 1.18-1.24) (Figure 2 and Table s1). The ratio was highest for inpatient visits of 2.03 (1.85-2.23) and lowest for outpatient visits 1.15 (1.11-1.18). These effects were estimated for each propensity score stratum; the results listed here and in subsequent figures represent the pooled totals. The absolute utilization rates in visits/year estimated by our model were 3.00 (95% CI: 2.85-3.16) in the COVID-positive cohort and 2.48 (95% CI: 2.37-2.60) in the comparison cohort for total visits (absolute difference: 0.52), 2.43 (2.31-2.57) in the positive cohort and 2.12 (2.02-2.23) in the comparison cohort for outpatient visits (absolute difference: 0.31), 0.31 (0.28-0.34) in the positive cohort and 0.20 (0.18-0.22) in the comparison cohort for ED visits (absolute difference: 0.11), and 0.12 (0.10-0.14) in the positive cohort and 0.06 (0.05-0.07) in the comparison cohort for inpatient visits (absolute difference: 0.06). Similar to the overall utilization, we found that the hazard ratio representing time to first utilization was 1.04 (1.01-1.06), suggesting that the positive cohort had a higher risk of first utilization after 28 days than the comparison group (Figure s1). As with the total utilization findings, the hazard ratios, incorporating time to events, were highest for first inpatient visit, 1.82 (1.68-1.97), and first ED visit, 1.36 (1.30-1.44). The hazard ratio was not significant for outpatient utilization. ![Figure 2.](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2022/07/09/2022.07.08.22276768/F2.medium.gif) [Figure 2.](http://medrxiv.org/content/early/2022/07/09/2022.07.08.22276768/F2) Figure 2. Health services utilization overall and by setting for children with chronic medical conditions, comparing the COVID-19 positive cohort to the COVID-19 negative cohort in matched analyses. One site without ED visits in their data was excluded from the ED and total utilization analyses. We also conducted sensitivity analyses excluding this site’s data from the inpatient and outpatient analyses; results were similar to the analyses including this site (inpatient [2.15 (1.95, 2.38)], overall [1.21 (1.18, 1.24)]); thus, we included their data in the inpatient and outpatient utilization ratios. For all patients from both the positive and comparison cohort, independent of COVID-19 diagnosis, we found that baseline utilization (number of visits per year prior to cohort entry) and the presence of 2 or more chronic condition clusters was associated with increased utilization (Table s1). Hispanic, Black, Asian and those listing multiple racial or ethnic groups had slightly lower utilization over all but increased ED care utilization. Females had higher utilization than males. And utilization tended to increase overall compared to the first months of the pandemic. ### Utilization Related to COVID-19 Severity To better understand the utilization differences, we conducted additional analyses focused on the presenting illness at cohort entry and this relationship to post-acute utilization. First, we utilized our previously developed severity criteria which evaluates severity of acute SARS-CoV-2 infection based on specific diagnosis and treatment codes independent of hospitalization. The utilization ratio for inpatient encounters occurring 28 days or more after cohort entry was far higher for those with acute COVID-19 infections associated with a severe [3.79 (95% CI 3.67-3.92)] or moderate [3.65 (95% CI 3.58-3.79)] acute COVID-19 presentation compared to the comparison cohort of patients without a positive COVID test result or diagnosis); modestly higher utilization was also noted in mild and asymptomatic cases (Figure 3). We similarly found a correlation between moderate or severe COVID-19 at presentation and higher utilization for ED and outpatient care in the 28 days or more after cohort entry. Second, we conducted an exploratory stratified analysis by test location at cohort entry. We found that patients tested in the inpatient and ED settings had higher utilization ratios than those tested in an outpatient setting. ![Figure 3:](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2022/07/09/2022.07.08.22276768/F3.medium.gif) [Figure 3:](http://medrxiv.org/content/early/2022/07/09/2022.07.08.22276768/F3) Figure 3: Utilization rate ratios by COVID-19 severity, overall and by setting ### Exploratory Analyses We conducted a number of additional exploratory analyses to better elucidate the degree to which specific covariates associated with increased utilization were associated with COVID-19 status. We found that increases in utilization were similar across all sites (Figure s2a) with inpatient utilization consistently showing the highest rate ratio, consistent with our primary outcome. We noted a trend in utilization stratified by age group (Figure s2b) with highest utilization rate ratios in the youngest age group (<1, 1-4) for all care settings. The utilization increases in the positive cohort were similar in all racial and ethnic groups (Figures s2c-d). We did not see significant differences in utilization rate ratios when stratifying by cohort entry period (Figure s2d). Patients tested in an inpatient or ED setting had higher utilization rate ratios than those tested outpatient (Figure s2e). To explore utilization by baseline chronic conditions, we estimated utilization rate ratios for each chronic condition cluster separately. In Tables s2a-d we report the 10 highest utilization rate ratios for each of low, medium, and high utilization clusters. In our analysis of patients without chronic conditions we found that the general pattern of increased utilization in the post-acute (28+ day) time-period was present in the non-chronic groups: the COVID-19 positive patientswithout baseline chronic disease had greater utilization than the comparison cohort without chronic disease with inpatient and ED utilization higher than outpatient utilization. However, the utilization ratio was greater in the chronic cohorts for all settings. Finally, given the potential difficulties of detecting COVID-19 cases during January-February 2022, we conducted a sensitivity analysis where we excluded patients who joined the cohort during this period. The utilization rate ratios did not change substantially: inpatient [2.03 (1.85, 2.23)], ED [1.53 (1.44, 1.62)], outpatient [1.19 (1.16, 1.23)], and overall [1.23 (1.19, 1.26)]. ## Discussion Children with chronic medical conditions are more likely to develop moderate or severe COVID-19, and SARS-CoV-2 infection can lead to acute exacerbation of chronic condition. However, little is known about the relationship between CMCs and long-term outcomes associated with PASC. We evaluated post-acute COVID-19 healthcare utilization in children with CMCs to identify general patterns and help facilitate further investigation into specific diseases and trajectories. We found that in the post-acute time-period ≥28 days after COVID-19 infection, the number of inpatient, outpatient and ED visits were increased compared to children without COVID-19 infection. The time to first event within the post-acute period was shorter for COVID-19 patients, mirroring the primary results. The increase was greatest among those who had moderate or severe COVID-19 at initial diagnosis, and we saw the largest effect in inpatient utilization. There are multiple strengths of this study: We described a large, geographically diverse cohort of patients from large, tertiary care, pediatric medical centers. The use of EHR data allowed for more robust assessment of elements such as PCR test results and measured obesity that may not be available through administrative databases, and we were able to evaluate utilization in all settings across the healthcare systems. Our use of a well-matched control cohort who had no recorded COVID-19 infection is also a significant strength as it allows for a better understanding of COVID-19 impact on a chronic condition population. The overall patterns showing increased utilization for children with higher number of conditions, and those with higher baseline utilization rates fit expected utilization patterns and provide face validity for further evaluation of the COVID-19 impact. Additionally, non-white racial and ethnic groups were more likely to utilize ED care. These associations align with existing literature suggesting racial and ethnic disparities in care access and utilization[31]. Our primary findings support our hypothesis that COVID-19 is related to increased utilization in patients with chronic medical conditions. However, there are many possible drivers of these findings. First, the correlation with severity at initial diagnosis is not unexpected as severe acute illness likely leads to prolonged recovery. This has been described in other viral infections[32]. It is challenging to determine whether the further increase seen in COVID-19 patients is due to true pathology or the result of uncertainty and escalation of care because of close monitoring by providers, perhaps out of an abundance of caution by providers. Our severity classification is weighted toward respiratory symptoms and may misclassify severe COVID-19 cases that manifest as chronic disease exacerbation. Additionally, there is the possibility that poorer control of chronic disease stems from the COVID-19 pandemic environment separate from an underlying pathophysiologic process, but we would have expected similar resulting utilization in our comparison cohort. The pronounced increase in inpatient utilization was unexpected. The limited available evidence on PASC in children suggests prolonged, though less severe symptoms, that are often self-limited. In our initial assessment of PASC phenotypes in children which included all patients, the most common syndromic findings were generally less acute symptoms that likely would not warrant inpatient admission such as changes in small and taste, loss of smell, hair loss. However, there were systemic features of PASC such as MIS-C, Addison disease, acute respiratory distress syndrome, that may lead to a higher acuity population, with perhaps those with chronic conditions experiencing those conditions more commonly. We explored severity, stratification by test setting, and stratification across multiple additional covariates to better understand the increase in inpatient utilization. It is possible that misclassification of visit types may exist within the EHR data however, the increased utilization was consistent across all but 1 site and the result held in our time-to-event analysis which should mitigate the possibility of “over-counting” visits classified as inpatient. Moreover, we expect that any systematic error in visit categorization should apply equally to our case and control cohorts given our cohort’s balance across a variety of potential confounders including existing chronic conditions, cohort entry period, and baseline utilization. Possibilities for this increased utilization include the aforementioned tendency toward caution practiced by providers when dealing with a poorly understood entity of post-acute COVID-19 or increased utilization resulting from complications or increased vigilance following exposure to unique treatments of COVID-19. One limitation of the present study which cast a broad look at all chronic conditions, is our inability to identify specific reasons for increased health utilization given the heterogeneity of the underlying diagnoses experienced by our population. We explored the chronic conditions with the utilization rate ratios segmented into those with high, medium and low prevalence based on total number of patients (Table s2a-d). There are a wide range of conditions clusters represented and it is important to note that there were significant utilization rate ratios for the majority of condition clusters. We do see numerous cancers and other disorders or diagnoses impacting immune function; however, patients with underlying mental health conditions had highest inpatient utilization. These preliminary results suggest important areas of focus for further investigation. It is not clear to what extent this effect is due to COVID-19 or broader care delivery changes related to the pandemic. There are additional limitations which warrant discussion: We primarily utilized EHR-based diagnoses, symptoms and signs recorded by providers but may have missed relevant laboratory, imaging and procedural findings. Additionally, the provider attribution of utilization to COVID-19 cannot be interpreted from the discrete EHR data alone. Our condition clusters were developed pragmatically and we identified clusters as chronic with the goal of broadly including chronic conditions. These clusters can be further refined and stratified based on severity or baseline utilization in future work, however they were adequate to define cohort entry for these analyses. Our aim was to evaluate utilization broadly to better understand whether PASC could present differently in chronic populations through exacerbation of their existing disease. We described a trend of increased utilization in a non-chronic population identified through exploratory analyses that is less marked than what we observed in the chronic population. Further investigation to understand whether the reasons for increased utilization are common among all children or whether, and to what degree, underlying mechanisms of specific chronic diseases drive our observed trends is needed. ## Conclusion In this broad, multi-center evaluation of the post-acute effects of COVID-19 on healthcare utilization in a pediatric chronic disease population, we found that care utilization in all settings was increased following COVID-19, particularly in the inpatient setting. The increased utilization was more marked in children with moderate or severe COVID-19 at initial presentation and the reasons for utilization appeared more related to the underlying diagnosis than to new onset symptoms or signs described in existing PASC literature. Further study, particularly at the disease-specific level, is required to better understand whether unique PASC phenotypes exist in the chronic disease population. ## Supporting information Supplement [[supplements/276768_file06.docx]](pending:yes) ## Data Availability All data produced in the present study are available upon reasonable request to the authors [https://github.com/PEDSnet/PASC](https://github.com/PEDSnet/PASC) ## Footnotes * * Co-first authorship * The above authors submit this on behalf of the RECOVER Consortium. * **Funding Source:** This research was funded by the National Institutes of Health (NIH) Agreement OT2HL161847-01 as part of the Researching COVID to Enhance Recovery (RECOVER) program of research. * **Disclosures:** Dr Brill received support from Novartis and Regeneron Pharmaceuticals within in the last year Dr. Chen receives consulting support from GSK. Dr. Patel reports funding from the National Institute of Health and Bayer Pharmaceuticals Dr. Rao reports prior grant support from GSK and Biofire and is a consultant for Sequiris. Dr. Mejias reports funding from Janssen and Merck for research support; Janssen, Merck and Sanofi-Pasteur for Advisory Board participation, and Sanofi-Pasteru and AstraZeneca for CME lectures. Dr. Jhaveri is a consultant for AstraZeneca, Seqirus, Dynavax, receives an editorial stipend from Elsevier and Pediatric Infectious Diseases Society and royalties from Up To Date/Wolters Kluwer. Dr. Lee serves on the PASC Advisory Board for United Health Group. All other authors have nothing to disclose. * 1 Codeset available at [https://github.com/PEDSnet/PASC](https://github.com/PEDSnet/PASC). * Received July 8, 2022. * Revision received July 8, 2022. * Accepted July 9, 2022. * © 2022, Posted by Cold Spring Harbor Laboratory The copyright holder for this pre-print is the author. All rights reserved. The material may not be redistributed, re-used or adapted without the author's permission. ## References 1. [1]. R. C. Woodruff et al., “Risk Factors for Severe COVID-19 in Children,” (in eng), Pediatrics, Dec 22 2021, doi: 10.1542/peds.2021-053418. [FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6MzoiUERGIjtzOjExOiJqb3VybmFsQ29kZSI7czoxMDoicGVkaWF0cmljcyI7czo1OiJyZXNpZCI7czoxODoicGVkcy4yMDIxLTA1MzQxOHYxIjtzOjQ6ImF0b20iO3M6NTA6Ii9tZWRyeGl2L2Vhcmx5LzIwMjIvMDcvMDkvMjAyMi4wNy4wOC4yMjI3Njc2OC5hdG9tIjt9czo4OiJmcmFnbWVudCI7czowOiIiO30=) 2. [2]. B. Zhou et al., “Risk profiles of severe illness in children with COVID-19: a meta-analysis of individual patients,” (in eng), Pediatric research, vol. 90, no. 2, pp. 347–352, Aug 2021, doi: 10.1038/s41390-021-01429-2. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1038/s41390-021-01429-2&link_type=DOI) 3. [3]. L. Kompaniyets et al., “Underlying Medical Conditions Associated With Severe COVID-19 Illness Among Children,” (in eng), JAMA Netw Open, vol. 4, no. 6, p. e2111182, Jun 1 2021, doi: 10.1001/jamanetworkopen.2021.11182. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1001/jamanetworkopen.2021.11182&link_type=DOI) 4. [4]. L. C. Bailey et al., “Assessment of 1351]794 Pediatric Patients Tested for Severe Acute Respiratory Syndrome Coronavirus 2 Across the United States,” (in eng), JAMA Pediatr, vol. 175, no. 2, pp. 176–184, Feb 1 2021, doi: 10.1001/jamapediatrics.2020.5052. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1001/jamapediatrics.2020.5052&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=33226415&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2022%2F07%2F09%2F2022.07.08.22276768.atom) 5. [5]. C. B. Forrest et al., “Severity of Acute COVID-19 in Children <18 Years Old March 2020 to December 2021,” Pediatrics, vol. 149, no. 4, 2022, doi: 10.1542/peds.2021-055765. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1542/peds.2021-055765&link_type=DOI) 6. [6]. C. Demeterco-Berggren et al., “Age and Hospitalization Risk in People With Type 1 Diabetes and COVID-19: Data From the T1D Exchange Surveillance Study,” (in eng), J Clin Endocrinol Metab, vol. 107, no. 2, pp. 410–418, Jan 18 2022, doi: 10.1210/clinem/dgab668. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1210/clinem/dgab668&link_type=DOI) 7. [7]. O. Ebekozien et al., “Inequities in Diabetic Ketoacidosis Among Patients With Type 1 Diabetes and COVID-19: Data From 52 US Clinical Centers,” (in eng), J Clin Endocrinol Metab, vol. 106, no. 4, pp. e1755–e1762, Mar 25 2021, doi: 10.1210/clinem/dgaa920. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1210/clinem/dgaa920&link_type=DOI) 8. [8]. J. S. Stevens et al., “Diabetic ketoacidosis and mortality in COVID-19 infection,” Diabetes & Metabolism, vol. 47, no. 6, p. 101267, 2021/11/01/ 2021, doi: [https://doi.org/10.1016/j.diabet.2021.101267](https://doi.org/10.1016/j.diabet.2021.101267). 9. [9]. K. E. J. Philip et al., “Impact of COVID-19 on people with asthma: a mixed methods analysis from a UK wide survey,” (in eng), BMJ Open Respir Res, vol. 9, no. 1, Jan 2022, doi: 10.1136/bmjresp-2021-001056. [Abstract/FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6NzoiYm1qcmVzcCI7czo1OiJyZXNpZCI7czoxMToiOS8xL2UwMDEwNTYiO3M6NDoiYXRvbSI7czo1MDoiL21lZHJ4aXYvZWFybHkvMjAyMi8wNy8wOS8yMDIyLjA3LjA4LjIyMjc2NzY4LmF0b20iO31zOjg6ImZyYWdtZW50IjtzOjA6IiI7fQ==) 10. [10]. S. C. Leftin Dobkin, J. M. Collaco, and S. A. McGrath-Morrow, “Protracted respiratory findings in children post-SARS-CoV-2 infection,” (in eng), Pediatric pulmonology, vol. 56, no. 12, pp. 3682–3687, Dec 2021, doi: 10.1002/ppul.25671. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1002/ppul.25671&link_type=DOI) 11. [11]. C. I. Bloom, P. Cullinan, and J. A. Wedzicha, “Asthma Phenotypes and COVID-19 Risk: A Population-based Observational Study,” (in eng), Am J Respir Crit Care Med, vol. 205, no. 1, pp. 36–45, Jan 1 2022, doi: 10.1164/rccm.202107-1704OC. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1164/rccm.202107-1704OC&link_type=DOI) 12. [12]. K. Gaietto et al., “Asthma as a risk factor for hospitalization in children with COVID-19: A nested case-control study,” (in eng), Pediatric allergy and immunology : official publication of the European Society of Pediatric Allergy and Immunology, vol. 33, no. 1, p. e13696, Jan 2022, doi: 10.1111/pai.13696. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1111/pai.13696&link_type=DOI) 13. [13]. A. P. Sunjaya, S. M. Allida, G. L. Di Tanna, and C. R. Jenkins, “Asthma and COVID-19 risk: a systematic review and meta-analysis,” (in eng), Eur Respir J, vol. 59, no. 3, Mar 2022, doi: 10.1183/13993003.01209-2021. [Abstract/FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6MzoiZXJqIjtzOjU6InJlc2lkIjtzOjEyOiI1OS8zLzIxMDEyMDkiO3M6NDoiYXRvbSI7czo1MDoiL21lZHJ4aXYvZWFybHkvMjAyMi8wNy8wOS8yMDIyLjA3LjA4LjIyMjc2NzY4LmF0b20iO31zOjg6ImZyYWdtZW50IjtzOjA6IiI7fQ==) 14. [14]. G. de Boer, G. J. Braunstahl, R. Hendriks, and G. Tramper-Stranders, “Asthma exacerbation prevalence during the COVID-19 lockdown in a moderate-severe asthma cohort,” (in eng), BMJ Open Respir Res, vol. 8, no. 1, May 2021, doi: 10.1136/bmjresp-2020-000758. [Abstract/FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6NzoiYm1qcmVzcCI7czo1OiJyZXNpZCI7czoxMToiOC8xL2UwMDA3NTgiO3M6NDoiYXRvbSI7czo1MDoiL21lZHJ4aXYvZWFybHkvMjAyMi8wNy8wOS8yMDIyLjA3LjA4LjIyMjc2NzY4LmF0b20iO31zOjg6ImZyYWdtZW50IjtzOjA6IiI7fQ==) 15. [15]. S. Ghosh et al., “A review on the effect of COVID-19 in type 2 asthma and its management,” (in eng), Int Immunopharmacol, vol. 91, p. 107309, Feb 2021, doi: 10.1016/j.intimp.2020.107309. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/j.intimp.2020.107309&link_type=DOI) 16. [16]. P. A. Palmon, D. J. Jackson, and L. C. Denlinger, “COVID-19 Infections and Asthma,” (in eng), J Allergy Clin Immunol Pract, vol. 10, no. 3, pp. 658–663, Mar 2022, doi: 10.1016/j.jaip.2021.10.072. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/j.jaip.2021.10.072&link_type=DOI) 17. [17]. M. Liuzzo Scorpo, G. Ferrante, and S. La Grutta, “An Overview of Asthma and COVID-19: Protective Factors Against SARS-COV-2 in Pediatric Patients,” (in eng), Front Pediatr, vol. 9, p. 661206, 2021, doi: 10.3389/fped.2021.661206. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.3389/fped.2021.661206&link_type=DOI) 18. [18]. N. Rai, J. A. Cornett, P. Zachariah, L. Quittell, and S. Lovinsky-Desir, “Severe respiratory viral infections in children with history of asymptomatic or mild COVID-19,” (in eng), Pediatric pulmonology, vol. 57, no. 2, pp. 361–366, Feb 2022, doi: 10.1002/ppul.25752. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1002/ppul.25752&link_type=DOI) 19. [19]. B. V. Guzman, B. Elbel, M. Jay, M. J. Messito, and S. Curado, “Age-dependent association of obesity with COVID-19 severity in paediatric patients,” (in eng), Pediatr Obes, vol. 17, no. 3, p. e12856, Mar 2022, doi: 10.1111/ijpo.12856. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1111/ijpo.12856&link_type=DOI) 20. [20]. B. K. Tsankov et al., “Severe COVID-19 Infection and Pediatric Comorbidities: A Systematic Review and Meta-Analysis,” (in eng), Int J Infect Dis, vol. 103, pp. 246–256, Feb 2021, doi: 10.1016/j.ijid.2020.11.163. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/j.ijid.2020.11.163&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=http://www.n&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2022%2F07%2F09%2F2022.07.08.22276768.atom) 21. [21]. A. K. Clift, D. Saatci, C. A. C. Coupland, H. Dambha-Miller, and J. Hippisley-Cox, “Sickle Cell Disorders and Severe COVID-19 Outcomes: A Cohort Study,” (in eng), Ann Intern Med, vol. 174, no. 10, pp. 1483–1487, Oct 2021, doi: 10.7326/m21-1375. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.7326/m21-1375&link_type=DOI) 22. [22]. T. S. Vilela, J. A. P. Braga, and S. R. Loggetto, “Hemoglobinopathy and pediatrics in the time of COVID-19,” (in eng), Hematol Transfus Cell Ther, vol. 43, no. 1, pp. 87–100, Jan-Mar 2021, doi: 10.1016/j.htct.2020.11.002. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/j.htct.2020.11.002&link_type=DOI) 23. [23]. L. Mucalo et al., “Comorbidities are risk factors for hospitalization and serious COVID-19 illness in children and adults with sickle cell disease,” (in eng), Blood Adv, vol. 5, no. 13, pp. 2717–2724, Jul 13 2021, doi: 10.1182/bloodadvances.2021004288. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1182/bloodadvances.2021004288&link_type=DOI) 24. [24]. B. Sayad, M. Karimi, and Z. Rahimi, “Sickle cell disease and COVID-19: Susceptibility and severity,” (in eng), Pediatr Blood Cancer, vol. 68, no. 8, p. e29075, Aug 2021, doi: 10.1002/pbc.29075. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1002/pbc.29075&link_type=DOI) 25. [25]. A. Appiah-Kubi et al., “Varying presentations and favourable outcomes of COVID-19 infection in children and young adults with sickle cell disease: an additional case series with comparisons to published cases,” (in eng), Br J Haematol, vol. 190, no. 4, pp. e221–e224, Aug 2020, doi: 10.1111/bjh.17013. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1111/bjh.17013&link_type=DOI) 26. [26]. G. M. Elia et al., “Acute chest syndrome and COVID-19 in sickle cell disease pediatric patients,” (in eng), Hematol Transfus Cell Ther, vol. 43, no. 1, pp. 104–108, Jan-Mar 2021, doi: 10.1016/j.htct.2020.11.005. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/j.htct.2020.11.005&link_type=DOI) 27. [27]. C. B. Forrest et al., “PEDSnet: a National Pediatric Learning Health System,” (in eng), J Am Med Inform Assoc, vol. 21, no. 4, pp. 602–6, Jul-Aug 2014, doi: 10.1136/amiajnl-2014-002743. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1136/amiajnl-2014-002743&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=24821737&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2022%2F07%2F09%2F2022.07.08.22276768.atom) 28. [28]. S. Rao et al., “Clinical features and burden of post-acute sequelae of SARS-CoV-2 infection in children and adolescents: an exploratory EHR-based cohort study from the RECOVER program,” medRxiv - Pre-print, p. 2022.05.24.22275544, 2022, doi: 10.1101/2022.05.24.22275544. [Abstract/FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6NzoibWVkcnhpdiI7czo1OiJyZXNpZCI7czoyMToiMjAyMi4wNS4yNC4yMjI3NTU0NHYxIjtzOjQ6ImF0b20iO3M6NTA6Ii9tZWRyeGl2L2Vhcmx5LzIwMjIvMDcvMDkvMjAyMi4wNy4wOC4yMjI3Njc2OC5hdG9tIjt9czo4OiJmcmFnbWVudCI7czowOiIiO30=) 29. [29]. T. D. Simon, W. Haaland, K. Hawley, K. Lambka, and R. Mangione-Smith, “Development and Validation of the Pediatric Medical Complexity Algorithm (PMCA) Version 3.0,” (in eng), Academic pediatrics, vol. 18, no. 5, pp. 577–580, Jul 2018, doi: 10.1016/j.acap.2018.02.010. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/j.acap.2018.02.010&link_type=DOI) 30. [30]. E. A. Stuart, “Matching Methods for Causal Inference: A Review and a Look Forward,” Statistical Science, vol. 25, no. 1, pp. 1–21, 21, 2010. [Online]. Available: [https://doi.org/10.1214/09-STS313](https://doi.org/10.1214/09-STS313). [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1214/09-STS313&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=20871802&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2022%2F07%2F09%2F2022.07.08.22276768.atom) 31. [31]. T. Whitfill, M. Auerbach, D. J. Scherzer, J. Shi, H. Xiang, and R. M. Stanley, “Emergency Care for Children in the United States: Epidemiology and Trends Over Time,” (in eng), J Emerg Med, vol. 55, no. 3, pp. 423–434, Sep 2018, doi: 10.1016/j.jemermed.2018.04.019. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/j.jemermed.2018.04.019&link_type=DOI) 32. [32]. D. Buonsenso et al., “Long-term outcomes of pediatric infections: from traditional infectious diseases to long Covid,” Future Microbiology, vol. 17, no. 7, pp. 551–571, 2022, doi: 10.2217/fmb-2022-0031. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.2217/fmb-2022-0031&link_type=DOI) 33. [33]. P. C. Austin, “An Introduction to Propensity Score Methods for Reducing the Effects of Confounding in Observational Studies,” (in eng), Multivariate Behav Res, vol. 46, no. 3, pp. 399–424, May 2011, doi: 10.1080/00273171.2011.568786. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1080/00273171.2011.568786&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=21818162&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2022%2F07%2F09%2F2022.07.08.22276768.atom) [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=000291533400002&link_type=ISI)