Hospitalization for child maltreatment and other types of injury during the COVID-19 pandemic

Background The possibility that child maltreatment was misclassified as unintentional injury during the COVID-19 pandemic has not been evaluated. Objective We assessed if child maltreatment hospitalizations changed during the pandemic, and if the change was accompanied by an increase in unintentional injuries. Participants and setting This study included children aged 0–4 years who were admitted for maltreatment or unintentional injuries between April 2006 and March 2021 in hospitals of Quebec, Canada. Methods We used interrupted time series regression to estimate the effect of the pandemic on hospitalization rates for maltreatment, compared with unintentional transport accidents, falls, and mechanical force injuries. We assessed if the change in maltreatment hospitalization was accompanied by an increase in specific types of unintentional injury. Results Hospitalizations for child maltreatment decreased from 16.3 per 100,000 (95 % CI 9.1–23.4) the year before the pandemic to 13.2 per 100,000 (95 % CI 6.7–19.7) during the first lockdown. Hospitalizations for most types of unintentional injury also decreased, but injuries due to falls involving another person increased from 9.0 to 16.5 per 100,000. Hospitalization rates for maltreatment and unintentional injury remained low during the second lockdown, but mechanical force injuries involving another person increased from 3.8 to 8.1 per 100,000. Conclusions Hospitalizations for child maltreatment may have been misclassified as unintentional injuries involving another person during the pandemic. Children admitted for these types of unintentional injuries may benefit from closer assessment to rule out maltreatment.


Introduction
Child maltreatment is a leading cause of pediatric morbidity, but this type of violence was poorly documented during the COVID-19 pandemic (Cappa & Jijon, 2021). There is growing concern that underreporting of child maltreatment worsened during lockdowns (Martinkevich et al., 2020;Thomas et al., 2020). Lockdowns were associated with financial hardship and unemployment and may have affected caregivers enough to increase the incidence of child maltreatment (Lawson et al., 2020;Schneider et al., 2017). Reduction in social services and home support programs, which are usually protective against maltreatment (National Center for Injury Prevention and Control, 2016), may have made it challenging to identify child maltreatment in community settings. Social distancing measures, daycare closures, and stay-at-home orders may have compounded the problem. As maltreatment became more difficult to detect, greater reliance was placed on healthcare systems to report suspected cases (Cappa & Jijon, 2021;Martinkevich et al., 2020). However, it is not clear if hospitals were able to adequately capture cases of child maltreatment during the pandemic, despite the suspected increase in maltreatment rates.
Some studies suggest that hospitals detected a growing number of cases of child maltreatment during the first wave of the pandemic (Cappa & Jijon, 2021;Collings et al., 2022;Kovler et al., 2021;Negriff et al., 2022), while others have found a decrease or no meaningful change compared with prepandemic levels (Amick et al., 2022;Chaiyachati et al., 2022;De Boer et al., 2022;Lewit et al., 2022;Russell et al., 2022;Sanford et al., 2021). Early reports from the U.S. and the U.K. suggest that hospitals treated more patients with maltreatment injuries (Cappa & Jijon, 2021). Yet, emergency department visits for child maltreatment declined by up to 19 % in the U.S. between April 2020 and March 2021 (Chaiyachati et al., 2022). The decline was greatest for mild or unclear maltreatment injuries, suggesting that the pandemic may have made it difficult to detect less obvious cases of maltreatment. Because resources were scarce during the pandemic (Alami et al., 2021), capacity to determine the intent of injuries may have decreased and caused a proportion of cases to be misclassified as unintentional. However, studies have not compared maltreatment injuries with unintentional injuries during the pandemic, nor evaluated the possibility of misclassification (Amick et al., 2022;Chaiyachati et al., 2022;De Boer et al., 2022;Lewit et al., 2022;Russell et al., 2022). We assessed whether changes in hospitalization rates for child maltreatment were accompanied by an increase in unintentional injuries in Quebec, the part of Canada that was hardest hit by COVID-19 the first year of the pandemic (Urrutia et al., 2021).

Study design and population
We carried out an observational study of hospitalization rates for child maltreatment and other injuries in Quebec between April 1, 2006 and March 31, 2021. We included hospitalizations among children between 0 and 4 years of age and obtained the data from the Maintenance and Use of Data for the Study of Hospital Clientele registry. The registry contains discharge summaries for all pediatric admissions in the province (Ministry of Health and Social Services, 2022). We excluded birth trauma and injuries associated with osteogenesis imperfecta.

Exposure
The main exposure measure in this study was the pandemic. COVID-19 was declared a public health emergency on March 13, 2020 by the government, marking the beginning of the pandemic in Quebec (Institut national de santé publique du Québec, 2022; Urrutia et al., 2021). We divided the pandemic into three phases based on the timing of lockdowns, including the (1) first lockdown (March 13, 2020 to June 23, 2020), (2) reopening after the first lockdown (June 24, 2020 to December 14, 2020), and (3) second lockdown (December 15, 2020to March 31, 2021. Lockdown measures included interruption of nonessential services, closure of daycares, schools and offices, and restrictions on social gatherings (Institut national de santé publique du Québec, 2022). Curfews were added during the second lockdown.
We used equivalent time periods before the pandemic for comparison, including (1) March 13, 2019 to June 23, 2019, (2) June 24, 2019 to December 14, 2019, and (3) December 15, 2019 to March 12, 2020. These dates match the calendar periods preceding the first lockdown, reopening, and second lockdown, thus accounting for seasonal trends in injuries that could bias the results.

Outcome
The main outcome measure was hospitalization for maltreatment injuries between 0 and 4 years of age. In this age group, suspected maltreatment has a low threshold for admission and usually requires in-hospital investigation (Chauvin-Kimoff et al., 2018). As we hypothesized that maltreatment injuries may have been misclassified as unintentional, we included unintentional injury hospitalizations as a secondary outcome. Unintentional injuries included transport accidents, falls, and mechanical force injuries.
We classified injuries by type (superficial, open, fracture, other), anatomical site (head and neck, spine, thorax and pelvis, upper limb, lower limb), and place of occurrence (home, other location). We investigated the main causes of unintentional injury, including transport accidents (cycling, car, other), falls (involving another person, play-related, other), and mechanical force (involving another person, animal, sharp object, other). We used codes from the 10th revision of the International Classification of Diseases (ICD-10) to identify injury hospitalizations (Table S1). We included codes for maltreatment, assault, abuse, and neglect in the main outcome measure, following previous research that has established that maltreatment hospitalizations can be captured in Quebec data using this approach (Auger et al., 2021(Auger et al., , 2022. For other injuries, we included codes for trauma of unintentional intent.

Patient characteristics
We examined patient characteristics at the time of hospitalization, including sex, rural residence, socioeconomic disadvantage, and neighborhood ethnic concentration. Socioeconomic disadvantage included the two most materially deprived quintiles of the population based on an index of neighborhood income, education, and employment (Pampalon et al., 2012). High ethnic concentration corresponded to the two most marginalized quintiles of a neighborhood-level index for the proportion of visible minorities, immigrants, and refugees (Centers for Disease Control and Prevention, 2020).

Data analysis
We examined characteristics of children hospitalized for maltreatment and unintentional injury (n, %). For each phase of the pandemic, we computed hospitalization rates for maltreatment, transport accidents, falls, and mechanical force injuries per 100,000 children. We calculated rates according to cause, type, anatomical site, and place of injury occurrence. We obtained population estimates for the denominator of rates from the Statistics Institute of Quebec (2022).
We used interrupted time series analysis to assess monthly trends in hospitalization rates for maltreatment and unintentional injury. Interrupted time series analysis is a quasi-experimental method of evaluating changes in outcome rates in a population over time (Penfold & Zhang, 2013). This approach uses interaction terms to model the effect of interruptions, such as lockdowns, on a time series. In this study, the time series consisted of hospitalization rates for every month between 2006 and 2021. As we began the time series in 2006, we had enough time before the pandemic to accurately estimate temporal trends in the short and long-term. We used an autoregressive model to analyze the time series, as this method accounts for seasonality, secular trends, and month-to-month correlation in rates (Penfold & Zhang, 2013). We included three interruptions in the model to account for the start of the pandemic lockdown, reopening without lockdown, and second lockdown.
We performed sensitivity analyses using a broader definition of child maltreatment (Schnitzer et al., 2011). We conducted the analysis in SAS version 9.4 (SAS Institute Inc., Cary, NC) and assessed statistical significance with 95 % CIs. The data were anonymized, and the institutional review board of our research centre provided an ethics waiver. Informed consent was not required to carry out the analysis.

Results
During the first year of the pandemic, 67 children aged 4 years or less were hospitalized for maltreatment, and 1249 were hospitalized for unintentional injuries (Table 1). In the preceding year, there were 82 maltreatment and 1355 unintentional injury hospitalizations. During the pandemic, maltreatment hospitalizations were more frequent in boys, socioeconomically disadvantaged children, and neighborhoods with high ethnic concentration, compared with the previous year.
The interrupted time series analysis indicated that monthly hospitalization rates for child maltreatment declined sharply following the first month of lockdown, compared with the period before February 2020 ( Fig. 1). However, temporal trends in unintentional injury differed depending on the cause. Hospitalizations for transport accidents tended to increase between April 2020 and June 2020. Hospitalizations for falls or mechanical force injuries were mostly stable during the pandemic.
Closer examination of hospitalization rates confirmed that maltreatment hospitalizations were lower than expected during each phase of the pandemic compared with the previous year (Table 2). During the first lockdown, 13.2 children per 100,000 were hospitalized for maltreatment (95 % CI 6.7-19.7 per 100,000) versus 16.3 per 100,000 (95 % CI 9.1-23.4 per 100,000) in the previous year. The decrease was most prominent for maltreatment resulting in fractures (9.0 vs. 4.1 hospitalizations per 100,000) and upper limb injuries (8.1 vs. 2.5 hospitalizations per 100,000). Maltreatment injuries of the upper limbs also decreased during the second lockdown. Hospitalization rates for maltreatment injuries that occurred at home increased during each phase of the pandemic. In sensitivity analyses using a broader definition of child maltreatment, hospitalization rates also appeared to decrease during the first lockdown (16.3 vs. 14.0 per 100,000), reopening (22.6 vs. 16.1 per 100,000), and second lockdown (27.6 vs. 21.8 per 100,000) relative to the prepandemic period.
Overall, hospitalization rates for unintentional falls and mechanical force injuries decreased or remained stable during the  Monthly hospitalization rates before and during the pandemic. Vertical lines at the right of each panel refer to, from left to right, the first lockdown (March 1, 2020), reopening (July 1, 2020), and second lockdown (January 1, 2021). Trends before 2015 are not shown, but extend backwards linearly to 2006.
pandemic, compared with the previous year (Table 3). However, transport accident rates increased during the first lockdown (7.3 vs. 12.4 hospitalizations per 100,000), and rates for falls tended to be higher during the second lockdown (111.4 vs. 117.8 hospitalizations per 100,000). Unintentional injuries occurring at home increased throughout the pandemic, especially during the second lockdown.
There was no apparent change in rates of unintentional injury to the head and neck, spine, thorax and pelvis, upper limbs, and lower limbs. There was also no significant change in the type of unintentional injury during the pandemic. Hospitalizations for cycling accidents and injuries involving another person saw the greatest increase among children 0-4 years during the pandemic (Table 4). Rates for cycling accidents went from 1.6 to per 100,000 before the pandemic to 4.9 per 100,000 during the first lockdown. Rates for falls involving another person also increased during the first lockdown, reaching 16.5 hospitalizations per 100,000 children compared with 9.0 per 100,000 in the preceding year. Rates for mechanical force injuries involving another person increased during the second lockdown, reaching 8.1 hospitalizations per 100,000 children compared with 3.8 per 100,000 in the preceding year.

Discussion
In this study of children 0 to 4 years of age in Canada, hospitalization rates for maltreatment decreased during two COVID-19 lockdowns. The decrease coincided with a decline in most types of unintentional injury hospitalization. However, hospitalizations for cycling accidents and falls involving another person increased during the first lockdown, while hospitalizations for mechanical force injuries involving another person increased during the second lockdown. The findings raise the possibility that some cases of child maltreatment may have been misclassified as unintentional cycling accidents, falls, or mechanical force injuries involving another person during the pandemic. A more thorough evaluation of preschool children admitted for cycling accidents or unintentional injuries involving another person may be warranted during a pandemic, as these may be indicative of maltreatment.
COVID-19 lockdowns disrupted sectors commonly involved in identifying child maltreatment. Daycare closures, as well as stay-athome orders imposed by lockdowns, drastically reduced interactions between children and professionals who usually report the majority of suspected child maltreatment (Cappa & Jijon, 2021;Martinkevich et al., 2020). Early studies indicate that referrals to child protective services decreased by up to 50 % and that fewer police reports of child abuse were issued during the pandemic (Cappa & Jijon, 2021;Rapoport et al., 2021). It was suspected that the drop in the number of cases was due to greater difficulty diagnosing child maltreatment during the pandemic, rather than a true reduction. However, it is not clear if hospitalizations for child maltreatment also decreased. Around 20 % of maltreatment cases are identified by hospital personnel and other healthcare professionals (Martinkevich et al., 2020). Thus, hospital data may be a helpful alternative when traditional routes of detecting maltreatment are disrupted.
The accuracy of hospital data for capturing maltreatment injuries during the pandemic has been questioned (Amick et al., 2022). Three U.S. studies reported no significant change in the number of children assessed for maltreatment in hospital centers (Lewit et al., 2022;Russell et al., 2022;Sanford et al., 2021). An analysis of 47,385 children presenting to U.S. trauma centers found that the proportion of injuries attributed to maltreatment remained stable during the pandemic (Collings et al., 2022). Other data conflict with these findings, including multicenter studies of over 30,000 children that suggest there was a significant decline in the number of children presenting to hospitals for maltreatment (Amick et al., 2022;Chaiyachati et al., 2022;De Boer et al., 2022). Conversely, a study of 407,228 emergency department visits reported that the proportion of children assessed for child maltreatment doubled following March 2020 stay-at-home orders (Negriff et al., 2022). However, methodological differences and lack of population-level denominators in these studies make it difficult to compare findings. In our study, hospitalization rates for child maltreatment decreased during both lockdowns. Although the reason for the decrease is difficult to confirm, our results align with the possibility that hospitals missed cases of child maltreatment (Chaiyachati et al., 2022;De Boer et al., 2022). Staffing shortages affected most hospitals in Quebec during the pandemic, as pediatric resources were diverted to adult hospitals (Alami et al., 2021). These changes may have limited the ability of social workers and healthcare professionals to identify and diagnose child maltreatment during the pandemic. It is also possible that patients were referred to outpatient clinics for follow-up.
Some cases of maltreatment may have been misclassified as unintentional injuries. Data from different countries indicate that the number of children presenting to hospitals with bicycle-related injuries increased by up to 48 % during the pandemic (Shack et al., 2022;van Oudtshoorn et al., 2021). In our study, hospitalization rates for bicycle injuries tripled during the first lockdown. Bicycle injuries are usually more prevalent in children aged 9 to 10 years (Shack et al., 2022;van Oudtshoorn et al., 2021), and are normally rare between 0 and 4 years. Preschoolers are expected to be slow riders, wear safety equipment, and use tricycles or balance bikes that Transport accident Any 7.3 (2.5-12.1) 12.4 (6.1-18.6) 14.9 (9.7-20.2) 10.3 (5.9-14.6) 6.7 (1.7-11.6) 5.6 (1.5-9.8) Cycling 1.6 (0.0-3.9) 4.9 (1.0-8.9) 4.3 (1.5-7.2) 1.5 (0.0-3.1) --Car 2.4 (0.0-5.2) 1.6 (0.0-3.9) 2.9 (0.6-5.2) 2.9 (0.6-5.3) 4.8 (0.6-8.9) -Other 3.3 (0.1-6.4) 5.8 (1.5-10.1) 7.7 (3.9-11.5) 5.9 (2.5-9.2) 1.9 (0.0-4.5) 5.6 (1. are easy to maneuver. Cycling injuries in preschoolers usually result in mild abrasions or soft tissue lesions that do not require hospitalization (Shack et al., 2022). However, young children may be passengers on adult bicycles traveling at greater speeds that are more susceptible to high impact collisions (Oxley et al., 2016). Cycling injuries in outdoor settings are also more likely to be witnessed by bystanders and therefore not due to maltreatment. We also found an increase in hospitalization rates for falls and mechanical force injuries involving another person. These unintentional injuries have received little attention during the pandemic. An analysis of 2730 pediatric trauma encounters found that fallrelated injuries decreased in patients aged 0-18 years (Sanford et al., 2021), while a separate study of 1061 children the same age found no difference compared with prepandemic levels (Sephton et al., 2021). A European study of 38 hospital centers reported a decrease in the number of children and adolescents evaluated for fractures and head injuries during the pandemic (Nijman et al., 2022). However, none of these studies specified if the injury involved another person (Nijman et al., 2022;Sanford et al., 2021;Sephton et al., 2021). A nonnegligible number of supposedly unintentional fractures and head injuries may involve another person and be hidden cases of maltreatment. When the involvement of another person cannot be hidden, caregivers who intentionally inflict maltreatment may be more likely to report an unintentional injury than a voluntary one.
Unintentional injuries occurring at home also increased. In the U.S., pediatric fractures that occurred at home went up at the start of the pandemic (Bram et al., 2020). The rise is expected as children spent more time at home and may have had more opportunities for accidents, some of which may have resulted in fractures. Yet, the possibility that some home injuries were due to maltreatment has not been assessed. Lockdown measures and stay-at-home orders isolated children with their families, and reduced interactions with other caregivers (Cappa & Jijon, 2021;Thomas et al., 2020). Social or financial stress may have made caregivers less patient and more likely to inflict injuries on young children at home (Cappa & Jijon, 2021;Lawson et al., 2020). Home injuries are also less likely to be witnessed, making it hard to verify the injury mechanism. For these reasons, children hospitalized for injuries sustained at home are another group that could benefit from a more thorough evaluation by healthcare providers to rule out maltreatment during lockdowns.
Several strengths and limitations of this study deserve consideration. We included all hospitalizations for maltreatment and unintentional injuries among children aged 0-4 years in Quebec, a province that has publicly-funded healthcare. Thus, selection bias is unlikely. However, hospitalizations for maltreatment represent the most severe cases, and we could not capture suspected cases of maltreatment or unintentional injury that were not documented or were recorded in outpatient settings. We could not examine children aged <1 year separately from older children. As we used administrative data, nondifferential misclassification due to coding errors may have occurred. The small number of maltreatment cases may have limited our ability to detect statistically significant changes during the pandemic. Further studies are needed to distinguish the extent to which changes resulted from lockdown measures, other aspects of the pandemic, secular trends, or a combination of factors. Quebec is culturally diverse and the study findings likely apply to similar populations, but may not be generalizable to other types of settings.
In this population-based study, hospitalization rates for child maltreatment decreased during two COVID-19 lockdowns. The decrease was accompanied by an increase in unintentional injuries, especially falls and mechanical force injuries involving another person. The findings raise the possibility that child maltreatment injuries during the pandemic may have been missed by hospital surveillance systems, and potentially misclassified as unintentional injuries involving another person. Children presenting with these types of injuries may benefit from closer assessment to rule out maltreatment. Strategies to improve the diagnosis of child maltreatment in hospitals are needed, especially during a pandemic when community surveillance systems are disrupted.

Ethics statement
The data were anonymized, and the institutional review board of our research centre provided an ethics waiver. Informed consent was not required to carry out the analysis.

Funding
This study was funded by the Quebec Population Health Research Network (266659), Canadian Institutes of Health Research (WI2-179928) and Fonds de Recherche du Québec-Santé (296785).

CRediT authorship contribution statement
NA had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. GCC, É B, and NA conceived the study. É B performed statistical analysis with input from GCC, TML, AL, AA, CB, and NA. All authors contributed to analysis and interpretation of data. GCC, É B, and NA drafted the manuscript, with critical revision from TML, AL, AA, and CB. All authors read and approved the final manuscript.

Declaration of competing interest
None.

Data availability
The dataset supporting the conclusions of this study is available from the Quebec Statistics Institute following standard access procedures (https://statistique.quebec.ca/recherche/#/accueil).