Trends in Mental Health-Related Pediatric Emergency Visits Among New York City Students

Abstract

Background and Objective

Recent studies highlight an increase in pediatric mental health disorders, amplified by COVID-19. This study examines changes in mental health-related emergency department visits among New York City public school students across the pandemic timeline.

Methods

We employed logistic regression to examine changes in the probability of a student's emergency department visit being mental health-related, and as a secondary outcome, we analyzed the difference in same-day discharge rates between mental health-related visits and other visits. For this analysis, we used the New York City Student Population Health Registry to link public school students' records to emergency department visit data.

Results

No significant linear trends were observed in the average monthly probability of a mental health-related visit before March 2020. From March 2020 through June 2021 there was an increase for all groups except male elementary school students. Female middle and high school students experienced the largest increase (.031 (CI = [.027, .034])) compared to pre-pandemic (0.103 (CI = [.103, .104])). Post-June 2021, all groups experienced a lower probability except for female middle and high school students, who had a .009 (CI = [.007, .011]) higher probability than during the pandemic.

Compared to the pre-pandemic period and non-mental health-related visits, a .043 (CI = [.029, .057]) lower probability of same-day discharge was observed for mental health-related visits during the pandemic period.

Conclusions

The COVID-19 pandemic correlated with a significant increase in mental health-related emergency department visits and longer stays, particularly among female middle and high school students.

1. Introduction

Mental health disorders in children and adolescents affect their immediate health and well-being [1,2] and have long-term implications [3–5]. Over the last two decades, evidence has shown an increase in the prevalence of mental health disorders in the pediatric population. For example, between 2016 and 2020 there was a 27% rise in the percentage of children aged 3 to 17 years with a parent-reported diagnosis of depression [6]. Additionally, youth suicide rates increased from 6.8 per 100,000 in 2007 to 11 per 100,000 in 2021 [7,8].

Evidence suggests an upward trend in pediatric mental health disorders before the start of the COVID-19 pandemic, based on survey data on children's health conditions [6,9] and emergency department (ED) visits data [10–12]. This pre-existing trend was exacerbated by the pandemic, with an increase in children’s mental health disorders tied to the pandemic [13–17]. After statewide school closures, rates of ED visits and hospitalizations due to suicidal thoughts and behavior among adolescent girls increased [18,19].

Current research on the youth mental health consequences of the COVID-19 pandemic has several limitations. First, data focusing solely on the initial months of the pandemic [14,18,19] restricts the ability to differentiate between pre-existing trends and changes marked by public health measures and/or the pandemic. Additionally, it limits the ability to determine whether any observed changes were transitory or permanent.

Second, studies typically relied on self-reported symptoms from surveys rather than clinical diagnoses [13]. Lastly, the quality of demographic information collected in discharge or claims data limits the ability to study subgroups such as race/ethnicity and sex. It also restricts the ability to control for student demographics in statistical analysis [20].

Utilizing a novel dataset that combines student enrollment information from New York City (NYC) public schools with emergency department visit data from September 2016 through May 2023, we addressed these gaps in the literature and describe changes in youth mental health. We explored changes in emergency department visits with a primary diagnosis of a mental health disorder (MHED) before, during, and after the onset of the COVID-19 pandemic. Additionally, we analyzed trends in the proportion of same-day discharges for MHED visits.

2. Methods

2.1. Data Sources and Study Population

The Student Population Health Registry (SPHR) is a comprehensive, longitudinal database developed by linking individual-level administrative records for NYC public school students and multiple health data sources. [21,22]. SPHR includes all students enrolled in kindergarten through twelfth grade in NYC public schools, including those in District 75 that provide specialized instruction, accounting for approximately one million students annually during the period of our study (eTable 1).

This study uses the Statewide Planning and Research Cooperative System (SPARCS) data linked through SPHR. SPARCS provides patient-level data related to services and charges for each hospital visit. Specifically, we examined discharges for students enrolled during the 2016–17 through 2022–23 school years, which run from September to June.

We identified all ED visits for students enrolled in NYC public schools during the study period, including ED visits that resulted in inpatient admissions. eTable 2 provide details on the total count of total hospital discharge records in SPARCS for patients aged 4 to 19, the records linked to NYC public school students, and the ED visits that occur while the student were enrolled.

We classified each ED visit either as a mental health-related ED (MHED) visit or a non-MHED visit. MHED visits were identified based on ICD-10 codes for the primary diagnosis using the Child and Adolescent Mental Health Disorders Classification System (CAMHD-CS) [23,24]. CAMHD-CS aligns ICD-10-CM [25] diagnosis codes with Diagnostic and Statistical Manual of Mental Disorders (Fifth Edition) (DSM-5) [26] diagnoses. While this method may introduce some misclassification, ICD-10 has improved diagnostic specificity, especially in distinguishing mental health conditions from physical health issues. However, some visits involving mental health concerns, particularly those presenting as somatic symptoms or injuries, may not be captured in the primary diagnosis field. Overall, any misclassification likely results in an underestimation of mental health-related ED visits rather than an overcount.

2.2. Study Variables and Statistical Analysis

Our first objective is to examine how the probability of a MHED visit has changed between 2016 and 2023. For each student, we create a binary variable for each month of the study period that takes the value one if the student has an emergency department (ED) visit during that month and zero otherwise. We also create another binary variable that takes the value one if the ED visit is classified as a MHED and zero otherwise for each month of the study period.

Using these variables, we calculated the monthly probability and 95% confidence interval (CI) of students having an ED visit classified as mental health-related stratified by the student's sex (as recorded by DOE based on parent’s self-reporting at enrollment) and grade level (elementary school (ES) vs. middle/high school (MS/HS)). This approach allows us to account for changes in the number of students with ED visits. Accounting for these changes is crucial, as the early months of the COVID-19 pandemic saw many individuals avoiding or being unable to access care [27].

Then, we conduct a logistic regression on students who had an ED visit, using a binary variable to indicate whether the ED visit is a MHED visit as the dependent variable. To test for linear trends, the model includes a continuous variable for the year-month of the observation. It also includes a categorical variable for periods: “Pre-Pandemic” (September 2016 to February 2020) as the reference category, “Pandemic” (March 2020 to June 2021), and “Post-Pandemic” (July 2021 to May 2023).

We include a categorical variable for student sex and grade level, with the categories being “Elementary School Female (ES Female),” “Elementary School Male (ES Male),” “Middle School/High School Female (MS/HS Female),” and “Middle School/High School Male (MS/HS Male).” “ES Female” is used as the reference category. To test for differences in trends and intercepts across groups and periods, the model includes two-way interaction terms between the categorical variables for sex and grade level with the periods of the study, between the continuous year-month variable and the categorical variable for periods, and a three-way interaction term between the continuous year-month variable, the categorical variable for periods, and the sex and grade level classification.

We also included a categorical variable for self-reported race/ethnicity (Black, Asian, Hispanic, and White), using “White” as the reference category. The model accounts for seasonality with fixed effects for each month of the year. It also includes fixed effects for the NYC borough where the school is located (Manhattan, Bronx, Queens, Staten Island, and Brooklyn).

Based on the estimation results of the logistic regression, we calculate the average change in the predicted probability of a MHED over a change of one unit of the continuous month variable. This allows us to present our results in terms of change in the probability of a MHED, conditional on the student having had an ED visit.

We also estimate the model without the continuous year-month variable to focus on differences in the average monthly probability of MHED visits across periods for each group. This approach allows us to directly estimate the average monthly probability for each group within each period, without extrapolating between linear trends and period effects. Consequently, we gain a clearer understanding of the impact of each period on the probability of MHED visits, providing insight into how MHED visits differ across groups and periods.

Finally, we examine changes in MHED visit severity using a linear probability model. To approximate the severity of each visit, we used a dummy indicator to differentiate between same-day discharge and next-day discharge or inpatient admission. This approach was chosen because same-day discharge generally indicates less severe cases, while next-day discharge or inpatient admission suggests more severe conditions requiring extended care.

Independent variables include the relative month of the ED visit (with February 2020 as the reference), a binary variable for MHED, and their interaction. The parameters of interest are the interaction terms, which capture the differences in outcomes for MHED visits relative to non-MHED visits for each period. This approach accounts for system-wide differences affecting all ED visits.

Additionally, we include all demographic covariates from the previous model, specifically student sex and grade, race/ethnicity, and school borough, as defined earlier. We also added fixed effects by hospital. Furthermore, we test whether the average difference between MHED and non-MHED visits is different from zero for each of the periods defined earlier: “pre-pandemic,” “pandemic,” and “post-pandemic,” based on the weighted linear combination of the estimates.

This research protocol was reviewed by the Institutional Review Board of the first author's institution and was determined to meet the criteria for exemption under 45 CFR 46.104(d). Analyses were conducted using R version 4.3.1. Reporting followed the STROBE reporting guidelines for cohort studies.

3. Results

3.1. Trends in ED visits related to Mental Health

Figure 1 shows the number of students admitted to an ED between September 2016 (beginning of the 2016–17 school year) and May 2023 (latest month with available data). We note the cyclical nature of ED visit, which fluctuates within the school year [28]. Additionally, we do not observe differences based on the sex of the student during the study period. The number of students with an ED visit annually was 214,015 in 2016–17 and 150,086 in 2022–23 (eTable 1).

Figure 1:

Number of Students with a visit to an Emergency Department

Note: Each point of the time series represents the number of unique students in a specific month-year that had a recorded emergency department record in SPARCS. The shaded areas correspond to the months when schools are in session (“School Year”), the months when schools are not in session (“Summer Vacation”), and the months between March 2020 and June 2020 when NYC was affected by the state-level stay-at-home orders and all schools were closed for in-person instruction.

Figure 2 shows the monthly probability of a student with an ED visit that is as MHED, by student sex and grade level. We observe a higher monthly probability for MS/HS male and female students and ES male students during the 2017–18 and 2018–19 school years, compared to the 2016–17 school year. We note an increase in the probability of a MHED visit for male and female MS/HS students following March 2020. However, this sharp increase does not persist over time for male MS/HS students after the 2020–21 school year. For female MS/HS students, the probability of a MHED visit appeared to decrease during the 2022–23 school year compared to the 2019–20 and 2020–21 school years.

Figure 2:

Proportion of Students with Mental Health-Related Emergency Department Visits Among Those with Any ED Visit, by Gender and Grade Level

Note: Each point of the time series represents the proportion of unique students with an emergency department visit that has an ICD-10 code as primary diagnosis that can be classified as a mental health disorder in a specific month-year by student sex and grade level. The shaded around each estimated proportion line represent the 95% confidence interval. The confidence interval is calculated using the critical values of a standardized normal distribution, and the estimated standard error of the proportion.

Table 1 presents the changes in the monthly probability that a student's ED visit is classified as MHED visit. Columns (1) to (3) shows the estimated linear trend in the probability that a student with an ED visit was admitted for a MHED conditional on observable characteristics and seasonal patterns. During the “Pre-Pandemic” period (column (1)), we observe either no significant changes or changes smaller than .01 percentage points in the probability of a student's ED visit being classified as a MHED visit.

Table 1:

Changes in the Monthly Probability that a Studenťs ED Visit is Classified as MHED

	Trend by Period					Difference to Pr(MHED|Pre-Pandemic)
	(1)	(2)	(3)		(4)	(5)	(6)
Group	Pre-Pandemic	Pandemic	Post-Pandemic		Pr(MHED|Pre-Pandemic)	Diff Pandemic	Diff Post-Pandemic
ES Female	0.000	−0.000	−0.000		0.020	0.005	−0.006
	[0.000, 0.000]	[−0.001, 0.000]	[−0.001, −0.000]		[0.020, 0.020]	[0.003, 0.007]	[−0.007, −0.005]
ES Male	−0.000	−0.001	−0.001		0.036	−0.009	−0.015
	[−0.000, 0.000]	[−0.002, −0.000]	[−0.001, −0.000]		[0.036, 0.036]	[−0.011, −0.007]	[−0.016, −0.014]
MS/HS Male	0.000	−0.006	−0.000		0.092	0.012	−0.010
	[0.000, 0.000]	[−0.008, −0.005]	[−0.001, −0.000]		[0.091, 0.092]	[0.009, 0.015]	[−0.012, −0.008]
MS/HS Female	0.000	−0.001	−0.001		0.103	0.031	0.009
	[0.000, 0.000]	[−0.002, 0.000]	[−0.001, −0.000]		[0.103, 0.104]	[0.027, 0.034]	[0.007, 0.011]

Note: Columns (1) to (3) are derived from a logistic regression model that includes a continuous linear trend variable. Each column shows the average linear increase over time for each group, based on the sex and grade of the student, for the periods "Pre-Pandemic," "Pandemic," and "post-pandemic." Columns (4) to (6) are derived from a logistic regression model without a linear trend component. Column (4) shows the estimated average probability that a student with an ED visit has a visit classified as MHED for each group. Columns (5) and (6) present the difference in the average probability of a student with an ED visit being classified as MHED between the "Pandemic" and "Post-Pandemic" periods compared to the "Pre-Pandemic" period, based on the interaction of indicator variables by groups and periods. All marginal effects are calculated conditional on the mean of all other independent variables (race/ethnicity, Charter, District 75, month fixed effects, school borough fixed effects). Confidence intervals are calculated at the 95% level of significance.

During the “Pandemic” period (column (2)), between March 2020 and June 2021, we observe a monthly reduction in the probability of student with an ED visit to be mental-health related by −.001 (CI = [−.002, −.000]) for ES male and −.006 (CI = [−.008, −.005]) for MS/HS male. We find no significant changes over time during this period for female students.

During the “Post-Pandemic” period, between July 2021 and May 2023, we observe a change in the outcome of −.001 (CI = [−.001, −.001]) for ES male students and −.001 (CI = [−.001, −.000]) for MS/HS female students, which was accompanied by an increase in the intercept for ES female and MS/HS male students (eTable 3). However, during this period all groups have an estimated intercept higher compared to the “Pre-Pandemic” period.

Table 1 columns (4) to (6) presents the changes of the probability that a student with an ED visit has a mental health disorder as primary diagnosis based on the model without a linear trend component. Column (4) presents the estimated probability of the outcome for each group during the period before March 2020, and the differences in probability for the “Pandemic” and “Post-pandemic” period to the “Pre-Pandemic period’.

We observe significant differences between the “Pre-Pandemic” and “Pandemic” periods. For ES female students, the probability increased by .005 (CI = [.003, .007]) from a “Pre-Pandemic” probability of .02 (CI = [.02, .02]). For ES male students, the probability decreased by −.009 (CI = [−.011, −.007]) from a “Pre-Pandemic” probability of .036 (CI = [.036, .036]). For MS/HS male students, the probability increased by .012 (CI = [.009, .015]) from a “Pre-Pandemic” probability of .092 (CI = [.091, .092]). For MS/HS female students, the probability increased by .031 (CI = [.027, .034]) from a “Pre-Pandemic” probability of .103 (CI = [.103, .104]).

For the average difference in the monthly probability of a MHED between the “Post-Pandemic” and “Pre-Pandemic” periods, we estimate differences of −.006 (CI = [−.007, −.005]) for ES female, −.015 (CI = [−.016, −.014]) for ES male, −.010 (CI = [−.012, −.008]) for MS/HS male, and .009 (CI = [.007, .011]) for MS/HS female students.

3.2. Changes in Severity

We observe a consistent decrease in the percentage of both MHED and non-MHED visits that ended in either same day discharged or inpatient admission (eTable 4). Throughout the study period, the proportion of non-MHED visits admitted to inpatient care was relatively small (2%) compared to MHED visits (9%) during the 2016–17 school year.

Figure 3 shows the estimated results from the linear probability model of the interaction between the relative month and the indicator variable for MHED visits and eTable 5 presents the average of those coefficients by period. Compared to February 2020, we observe that the average difference in the monthly probability of a same-day discharge was .027 percentage points lower (CI = [.041, .014]) for MHED visits compared non-MHED visits during the “Pre-Pandemic” period.

Figure 3:

Changes in Probability of Next-Day Discharge or Inpatient Admission for MHED Visits Relative to Non-MHED Visits by Month

Note: This figure plots the coefficients of the interaction between the relative month (with zero being February 2020) and the indicator variable for whether the ED visit is classified as MHED or not. The dependent variable is one if the ED visit results in next-day discharge or inpatient admission, and zero otherwise. Each coefficient captures the differences in outcomes for MHED visits relative to non-MHED visits for each period. Confidence intervals are calculated at the 95% level of significance. The shaded areas indicate the periods designated as “School Year,” “Summer Vacation,” and “Stay-at-Home Order and School Closures,” when all students were assigned to remote learning in the second half of the 2019–20 school year.

During the “Pandemic” period, we observe a significant difference in the probability of next-day discharge or being admitted to inpatient between MHED and non-MHED visits, with an average difference of .043 (CI = [.029, .057]). In the “post-pandemic” period, there was a non-significant difference between the two types of visits of .013 (CI = [−.001, .027]).

4. Discussion

The main contribution of this paper is documenting the trends in MHED visits three years after the start of the pandemic, presenting evidence on how ED visits associated with mental health diagnoses evolved before, during, and after the COVID-19 pandemic. After the onset of COVID-19 pandemic at the end of March 2020 [29], there was an increase in the monthly probability of students with a MHED visit. The timing of this increase coincided with the implementation of public health measures such as mobility restrictions and school closures. Although these increases moderated after June 2020, compared to the pre-pandemic period we observe a higher probability of a MHED visit among MS/HS students during the 2020–21 school year, when 61% of students remained under fully remote instruction [30]. Consistent with previous studies [19,31], our results show that MS/HS female students experienced the highest increase in MHED visits. These results align with national-level research for this period, which primarily focuses on the immediate post-COVID period [11].

Furthermore, we found that these changes did not persist for most students once they returned to in-person instruction during the 2021–22 school year [32] and executive orders during the public health emergency expired [33]. However, MS/HS female students continued to experience a higher probability of a MHED visit, though at a reduced level compared to the peak during the pandemic.

Our findings offer a long-term perspective on how the COVID-19 pandemic and related policies have affected mental health-related ED visit trends among children and adolescents. While most students show similar or lower probabilities of MHED visits compared to the period before March 2020, our results emphasize the need for continued support for specific groups, such as MS/HS female students. Both our study and previous research highlight their increased vulnerability.

When looking at our results for changes in severity of MHED visits, previous studies have used length-of-stay (LOS) to study changes in severity [34,35]. While we do not have information on length of stay, we examine changes in the proportion of visits admitted to inpatient care and overnight stays for MHED and ED visits. The overall decrease in same-day discharge for MHED visits, compared to non-MHED visits, suggests a rise in the prevalence of severe mental health disorders requiring ED visits, or it may indicate that the threshold for seeking ED care, which presumably increased during the COVID-19 pandemic, has remained high.

Factors such as resource availability could explain the decrease in inpatient admissions from ED visits and the increase in next-day discharges during the study period. The reduction in the number of MHED visits derived to inpatient care could result from the declining number of certified psychiatric beds in NYC, which decreased by 8.5% between 2012 and 2018 according to the New York State Institutional Cost Reports [36]. If the decrease in same-day discharges is due to an increase in the severity of MHED visits, and if this is offset by an increase in overnight stays and lower inpatient admissions due to resource constraints, it could raise concerns about the adequacy of resources.

5. Limitations

One limitation of our study is the changing number of students with ED visit changes over time. Specifically, the number of students with an ED visit significantly dropped during the initial wave of the COVID-19 pandemic, a trend observed across the US [37,38], especially for pediatric ED visits [39]. This variation could be influenced by changes in the student population, as our study is an open cohort. By the 2022–23 school year, the number of students with an ED visit decreased by 23% compared to the 2018–19 school year. However, during the same period, the size of the NYC student population decreased by only 8.7%. This This enrollment decline is consistent with national trends [40,41]. Therefore, while it partially explains the reduction in ED visits, it does not account for the entire decrease.

Another potential explanation for the decrease in the number of students with an ED visit could be the rise in telemedicine usage following the onset of the pandemic. The expansion of telehealth services might deter patients from visiting the ED, particularly those seeking primary care or services that could be provided by virtual urgent care. Since our data only covers ED discharges from hospitals, we are limited in capturing these virtual interactions.

One limitation of our analytical approach is the assuming independence of the observations over time in the case of students with multiple ED and MHED visits. However, as shown in eTable 6, the number of repeated visits over the study period is limiting our ability to account for within-student correlation in our estimation strategy.

While our study does not identify specific factors driving our results, it highlights an opportunity for future research. The trends likely result from cumulative effects of unobserved influences, such as changes in healthcare-seeking behavior, school policies, and student health. Further investigation is needed to understand these factors. Additionally, the increase in longer ED stays for mental health-related issues underscores the need to expand access to care and preventive support for NYC students.

6. Conclusions

Over the past decade, there has been growing concern about the rise in mental health disorders among students in NYC schools and worldwide. Based on the regression results on the probability of a MHED visit conditional on attending an ED visit, we found evidence of an increase in the monthly average proportion of students with an ED visit where their primary diagnosis was associated with a mental health disorder for all students except male ES students. Additionally, when examining the average monthly probability four years after the start of the COVID-19 pandemic, we found a higher probability compared to the period before March 2020, specifically for MS/HS female students.

These results, along with the reduction in the probability of same-day discharge for MHED compared to non-MHED visits, particularly during periods of increased demand for these services, highlight the importance of building an appropriate infrastructure to support the mental health needs of youth.

Abbreviations

CAMHD-CS

Child and Adolescent Mental Health Disorders Classification System

ED

Emergency Department

ES

elementary school

MHED

mental health-related ED

MS/HS

middle/high school

NYC

New York City

SPHR

Student Population Health Registry

SPARCS

Statewide Planning and Research Cooperative System

Data Sharing Agreement

Deidentified individual participant data will not be made available. The data supporting this study's findings are available from the New York State Department of Health. Restrictions apply to the availability of these data, which were used under license for this study. Data are available with the permission of the New York State Department of Health. Links between the Statewide Planning and Research Cooperative System and New York City Department of Education enrollment data are unavailable due to legal reasons.