Changes in pediatric service utilization for depression and anxiety during the COVID-19 pandemic

Abstract

Background

Evaluation of service utilization for common mental health disorders among pediatric population during the COVID-19 pandemic may help identify potential barriers to service use among different groups. This study investigates the utilization of health care services for depression and anxiety among 3-17-year-olds before and during the COVID-19 pandemic.

Methods

The monthly cases of depression and anxiety disorders in 2019 and 2020 were identified using Medicaid claims data from the largest care management organization serving Georgia, US. Service utilization rates were compared between pre-COVID and during/post-COVID using negative binomial regression in an interrupted time series study design. Subgroup analyses were done by age group, gender, and government benefits program participation status.

Results

The average monthly encounters per 1,000 beneficiaries for depression and anxiety disorders decreased by 13.2% and 4.4% during the pandemic. There were statistically significant reductions in month-to-month service utilization for both depression and anxiety in the post-pandemic period across subgroups.

Discussion

Utilization of pediatric health care services for anxiety and depression of the population studied decreased during the pandemic. This may have led to an increase in mental health conditions post pandemic.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12913-026-14644-6.

Background

The COVID-19 pandemic is an extraordinary stressor that has accompanied a variety of negative mental health outcomes among children. Before the COVID-19 pandemic, from 2016 to 2019 in the United States (US), 9.4% and 4.4% of children aged 3–17 years had diagnosed anxiety and depression respectively [1]. A review of global post-pandemic studies, which used self- or parent-reported measures, found the pooled prevalences of clinically elevated depression and anxiety were 28% and 21%, respectively among children and adolescents in the US [2]. The pandemic has exacerbated children and adolescents’ risks for anxiety and depression and has simultaneously worsened access to health care services, including mental health care [3].

While several studies have reported increases in cases of anxiety and depression among children and adolescents following the COVID-19 pandemic [4, 5], relatively few have focused on health service utilization [6]. Much of the existing research has focused on the determinants of youth mental health during the pandemic [3, 4], or the overall prevalence of mental health conditions at a population level [3, 5], and is limited to use of the emergency department visits as the measure for utilization [4, 7, 8]. However, health service utilization is more commonly represented by the number of encounters, visits, stays, or similar events per beneficiary (or a larger group size), such definition is consistent with what is used by the Centers for Medicare and Medicaid Services (CMS) (public insurance providers in the US) [9]. Furthermore, in the US, there have been reports of large declines in pediatric mental health care utilization [3, 7]. From February to October 2020, a 50% drop (145 per 1,000 beneficiaries to 72 per 1,000) was reported in the number of Medicaid and Children’s Health Insurance Program (CHIP) beneficiaries receiving mental health care services [3]. Similar declines have been reported for privately insured patients [10].

No studies have examined changes in the utilization of pediatric mental health care services during the pandemic for clinically diagnosed depression and anxiety based on administrative data. Disease condition and clinical diagnosis information available in administrative data provide an ideal opportunity to study service utilization amidst the pandemic [11]. Changes in utilization are important to study, particularly in the context of the COVID-19 pandemic, as they inform policymakers and health care providers of the most critical community needs in improving access to care. For Georgia, reduced utilization of mental health services during the pandemic is particularly concerning given that it is currently ranked last of all 50 US states in access to these services [12].

In the US, primary public insurance for children in low-income families is Medicaid, each state Medicaid program contracts private managed care entities referred to as Care Management Organizations (CMOs) to administer the insurance program. CMOs receive capitated payments from the state per enrollee and are responsible for coordinating care, including behavioral health services for low-income children. Medicaid is the single largest payer of mental health services in the US and covers approximately 40% of all children nationwide [13]. It plays a critical role in ensuring access to behavioral health services for children from low-income households, who are at elevated risk for depression and anxiety due to social and economic stressors—risks that were further exacerbated during the COVID-19 pandemic due to school closures, parental job loss, and increased household instability.

This burden is especially pronounced among children in the foster care system. Foster care youth are automatically eligible for Medicaid, regardless of family income, under federal law. Because of prior exposure to trauma, instability, and frequent changes in placement, children in foster care experience higher rates of behavioral health conditions and are at increased risk of service disruptions [14]. CMOs may contract with providers offering dedicated behavioral health carve-outs or wraparound services for this population, acknowledging their complex and sustained care needs. Understanding patterns of service use during the pandemic for this vulnerable subgroup is vital for informing targeted interventions.

Georgia is a southern state and one of the states most disproportionately affected by COVID-19 in the nation [15]. Although the prevalence of mental illness among youths in Georgia is comparable to other US states, it consistently ranks towards the bottom in access to mental health care [12]. Several actions were taken in Georgia following the first COVID-19 case identified on March 2, 2020 (Fig. 1) that affected access to mental health care.

Fig. 1.

2019 timeline of COVID-19 response in the state of Georgia, US [16–18]

The April 1, 2020, press conference from the Governor of Georgia included statewide actions to combat the spread of COVID-19 cases. A major directive in the conference was continued closure of all public schools that had remained closed since March 26 for the remainder of the 2019–2020 school year [19]. School closures were followed consistently across the state. Under the federal response Georgia Medicaid members were eligible for continuous coverage during the emergency which facilitated access to healthcare for mental health among Medicaid patients. The provision of telehealth services was also expanded in the state during 2020, which further increased utilization of telehealth services [20]. However, with school closures, early identification of referral to service providers was limited.

The objective of this observational study is to study the effect of COVID 19 induced major events on the rate of monthly outpatient encounters for depression and anxiety among Medicaid youth. We also examined changes in service utilization by participant demographics including age group, sex, and government benefits program participation. This study is important to identify the role of early identification and school referral services on mental health provision to children.

Methods

Data

We conducted a retrospective observational study utilizing individual-level Medicaid claims data (also referred to as “encounter data”) [21] with service dates 1/1/2019 to 12/31/2020. The claims data comes from one of four CMOs in Georgia and includes information on patients with at least one inpatient or outpatient encounter for Mental, Behavioral, and Neurodevelopmental Disorders. This CMO, the largest in the state, insured nearly 400,000 lives in the study, and the majority were children [22]. Amerigroup is also the sole CMO in the study state for providing health insurance coverage to children in foster care [23]. This CMO is a subsidiary of the nation’s largest company solely focused on serving low-income families. Depression and anxiety-related claims were identified using the International Classification of Diseases, Tenth Revision [ICD-10] codes beginning with F; more specifically F32.0-9 and F33.0-9 for depression and F40-42, F43.0, F43.1, F93.0-93.2, and F93 for anxiety disorders. The study sample was restricted to children and adolescents who were between 3 and 17 years of age during the study period.

Study design

We selected an interrupted time series design with segmented regression analyses to compare changes in service utilization for pediatric depression and anxiety during the pre-COVID period and post/during-COVID period. Interrupted time series design is commonly used for causal evaluations and has been used in post-pandemic studies of temporal patterns of mental health, suicide attempts, and substance abuse among adults [24–25]. Analysis of impact of shut down of in-person care during the pandemic under Veterans Administration have used similar analysis [26].

Study measures

For our outcome measure, we assessed health service utilization using the number of encounters per 1,000 beneficiaries per month, which provides an accurate measure of services utilized by Medicaid beneficiaries enrolled in managed care plans [9, 21, 27]. For both depression and anxiety, we categorized encounters as in-person and telehealth encounters. Telehealth encounters were identified using Current Procedural Terminology (CPT codes 99441-3 and 98966-8 for audio-only and modifier codes GT, GQ or 95 for audio-video) and place of service codes (POS code 02). Unlike a visit, encounters are not aggregated by dates of service [28] and are a commonly used data source for reviewing service utilization [21]. Each encounter represents a distinct healthcare service or procedure billed to Medicaid; therefore, a single enrollee may have multiple encounters on the same day if they received more than one service.

Other measures included as covariates and used for subgroup analysis are 1) category of government benefits program participation: (i) Children’s Health Insurance Program (CHIP) Standard, (ii) Temporary Assistance for Needy Family (TANF) Foster Care, or (iii) TANF Standard; 2) gender: (i) male, (ii) female; 3) age group to reflect developmental stage: (i) 3–11 years (children), and (ii) 12–17 years (adolescents).; and 4) encounter setting: (i) telehealth, (ii) in-person. The main difference between TANF Foster Care and TANF Standard is the population they serve and the eligibility criteria. TANF Foster Care provides assistance specifically to families caring for children in foster care, while TANF Standard provides assistance to low-income families with children who meet the program’s eligibility criteria.

Statistical analysis

First, using descriptive statistics, we compared encounters for major depressive and anxiety disorders before and during the pandemic by program participation type, age group, gender, and encounter setting. We then plotted a time series of the average monthly encounters per 1,000 beneficiaries for depression and anxiety for the study period. We calculated absolute and relative change in trends of average monthly encounters per 1,000 beneficiaries before and during the pandemic for the study population and by subgroups (age, gender, government benefits program participation, service setting). To assess level and slope changes in the monthly service utilization rate following the onset of the pandemic, we employed an interrupted time series analysis [24]. The analytic model is represented as follows:

where Y is the outcome measure at time point t; T is a continuous measure of time in months passed since January 2019 (ending in December 2020), and with the interruption occurring at time T_I (April 1, 2020), therefore represents the pre-COVID trend; D_t is a dummy variable indicative of the COVID-19 pandemic (i.e., 0 for pre-pandemic and 1 for during-pandemic), and therefore represents the level change in the outcome measure due to the pandemic; is the interaction between the interruption and the measure of time, and therefore represents the change in slope following the pandemic; and ε_t represents the error term at time point t. The time of interruption was set at April 1, 2020, dividing the study period into pre-interruption/pre-pandemic (the 15-month period from January 2019 to March 2020) and post-interruption/during-pandemic (the 9-month period from April 2020 to December 2020) (see Fig. 1). The schools in the state of Georgia generally resume during the first week of August, so the during-pandemic study period captured 2 months of school closure (April to May 2020), 2 months of summer (June to July 2020) and 5 months school in session in 2020 (August to December 2020). The monthly count of encounters (outcome) was modelled using negative binomial regression to allow for over-dispersion in the data revealed during initial analyses. We ran separate models for depression and anxiety for the full sample and each subgroup. The outcomes for each subgroup analysis were the respective numbers of encounters. Subgroup analysis was conducted by program participation type, age group, gender, and encounter setting (telehealth vs. in-person). Autocorrelation was assessed to determine whether monthly encounters were serially correlated over time, which could affect model assumptions if unaccounted for. While we did not observe strong autocorrelation upon visual inspection of autocorrelation plots, we use a 3-month lag of the outcome variable as a covariate. This conservative approach is consistent with prior health services research that used short-term lags to account for serial dependence in healthcare utilization patterns [29]. The log of the total beneficiaries enrolled in the CMO plan each month during the study period was used as an offset term in each model to convert the outcomes into rate i.e., encounters per person per month. The exponentiated coefficient from the negative binomial regression estimates are interpreted as the incidence rate ratios (IRRs) and gives the multiplicative change in the monthly rate of service utilization (encounters). All analyses were conducted using R, version 4.0.3.

Results

Of 48,694 children between the ages of 3 and 17 that comprised our study sample, the mean age was 10.7 years (SD = 4.0 years), 54.5% were male, 56.9% were enrolled in TANF standard, and 34.8% were enrolled in TANF Foster care. The total number of encounters made by children (N = 26,148) and adolescents (N = 26,046) for depression and anxiety during the study period were 203,122 (32.3 encounters per 1,000 beneficiaries) and 212,753 (33.8 encounters per 1,000 beneficiaries) respectively. Of these, 37.3% of the total encounters for depression and 39.5% of the total encounters for anxiety occurred during the pandemic (April 1, 2020 to December 31, 2020). The most common encounters for depression during the study period involved children and adolescents in TANF Standard (50.1%), those aged 12–17 years (84.4%), and females (66.9%). Similarly, the most common encounters for anxiety were among children and adolescents in TANF Foster care (52.3%), those aged 12–17 years (54.5%), and females (59.2%). The total numbers of encounters before and during the pandemic for depression and anxiety by patient characteristics are shown in Table A1 in the appendix.

The volume of total encounters for depression remained generally stable throughout the study period, with 8,492 average monthly encounters in the pre-pandemic period and 8,416 during the pandemic. For anxiety, the average monthly encounters were higher during the pandemic by 767 (8,577 vs. 9,344). When accounted for number of beneficiaries, the average monthly encounters for depression and anxiety per 1,000 beneficiaries decreased by 4.5 (34.1 to 29.6; p = 0.001) and 1.5 (34.4 to 32.9; p = 0.310) encounters respectively during the pandemic. While TANF Foster care saw the largest absolute and relative decrease in average monthly encounters for depression (16% lower encounters per 1,000 beneficiaries; p = 0.002) during the pandemic, TANF Standard saw the largest decrease for anxiety (7% lower encounters per 1,000 beneficiaries; p = 0.107). Similarly, children aged 3–11 years and males saw the largest relative decrease in average monthly encounters for depression (33% (p < 0.001) and 19% (p = 0.001) lower encounters per 1,000 beneficiaries respectively) and anxiety (7% (p = 0.132) and 10% (p = 0.033) lower encounters per 1,000 beneficiaries respectively) compared to their counterparts during the pandemic (Table 1).

Table 1.

Average monthly encounters (per 1,000 Beneficiaries) for depression and anxiety pre- and during the pandemic

Characteristics	Depression			Anxiety
	Pre-pandemic	During pandemic	% Change [95% CI]; P	Pre-pandemic	During pandemic	% Change [95% CI]; P
Overall	34.1	29.6	-13.2% [-19.6%, -6.6%]; 0.001	34.4	32.9	-4.4% [-11.6%, 3.5%]; 0.310
Program type
CHIP Standard	2.4	2.3	-4.2% [-17.0%, 5.3%]; 0.320	2.0	2.1	+ 3.2% [-3.8%, 10.2%]; 0.377
TANF Foster Care	14.7	12.3	-16.3% [-23.5%, -7.7%]; 0.002	17.9	17.3	-3.4% [-11.6%, 5.9%]; 0.534
TANF Standard	17.0	15.0	-11.8% [-19.2%, -4.1%]; 0.006	14.5	13.5	-6.9% [-14.0%, 1.1%]; 0.107
Age group
3–11 years	5.8	3.9	-32.8% [-39.0%, -24.1%]; < 0.001	15.8	14.7	-7.5% [-15.0%, 1.6%]; 0.132
12–17 years	28.3	25.7	-9.2% [-16.1%, -2.0%]; 0.019	18.6	18.2	-2.2% [-9.1%, 5.6%]; 0.645
Sex
Male	11.6	9.4	-19.0% [-28.3%, -9.2%]; 0.001	14.3	12.9	-9.8% [-17.7%, -1.6%]; 0.033
Female	22.5	20.3	-10.2% [-15.9%, -3.8%]; 0.004	20.1	20.0	-0.5% [-7.5%, 7.4%]; 0.981
Encounter Setting
In-person	32.9	15.0	-54.4% [-60.8%, -48.2%]; < 0.001	33.1	13.1	-60.4% [-67.0%, -54.3%]; < 0.001
Telehealth	1.2	14.6	+ 1167% [1007%, 1359%]; < 0.001	1.3	19.8	+ 1423% [1211%, 1709%]; < 0.001

Note: Pre-pandemic: Jan 2019 to March 2020, During pandemic: April 2020 to December 2020; ICD-10 codes for depression: F32.0-9 and F33.0-9; ICD-10 codes for anxiety: F40-42, F43.0, F43.1, F93.0-93.2, and F93. The P-values obtained from t-test and 95% confidence intervals obtained using delta method

Figure 2 shows the trend in overall encounters and by encounter setting (telehealth vs. in person) over the study period. During the pandemic, the monthly encounters for depression and anxiety per 1,000 beneficiaries decreased from 32.5 to 39.6 in April 2020 to 26.4 and 28.6 in December 2020 respectively. In contrast, during the pre-pandemic period, the monthly encounters for depression and anxiety per 1,000 beneficiaries increased from 34.5 to 34.5 in January 2019 to 36.3 and 36.6 in March 2020 respectively. Telehealth utilization peaked in April 2020, with a concurrent dip in in-person encounters. Between April 2020 and December 2020 (during-pandemic period), the average monthly in-person encounters for depression and anxiety increased by 1.9 (12.8 to 14.7) and 2.5 (10.3 to 12.8) encounters per 1,000 beneficiaries respectively while average telehealth encounters decreased by 8.0 (19.7 to 11.7) and 13.4 (29.2 to 15.8) encounters per 1,000 beneficiaries respectively.

Fig. 2.

Monthly encounters (total, in-person, telehealth) per 1,000 beneficiaries for depression (top) and anxiety disorders (bottom), 2019–2020

When comparing the utilization rates in respective quarters in the pre-pandemic and during-pandemic period (Q2 of 2020 with Q2 2019 and so on), Q4 2020 (October 2020 - December 2020) saw the largest relative decrease in average number of encounters for both depression (17% lower encounters per month per 1,000 beneficiaries) and anxiety (9% lower encounters per month per 1,000 Beneficiaries) compared to Q4 2019 (Table A2).

Table 2 displays the IRR for pre-pandemic, immediate (level change) and longer-term (during-pandemic) slopes derived from the negative binomial model. Overall, we observed statistically significant reductions in month-to-month service utilization for both depression (IRR: 0.971, 95% CI: 0.947, 0.996] and anxiety disorders (IRR: 0.962, 95% CI: 0.942, 0.982) during the pandemic. In-person month-to-month encounters showed a significant level change for both depression and anxiety immediately as the pandemic hit and showed increasing but non-significant trends during the pandemic. In contrast, telehealth encounters increased significantly for depression and anxiety as the pandemic hit and showed a significantly decreasing trend afterwards. The significantly increasing pre-existing trend for telehealth encounters is attributed to spike in telehealth use in March 2020 due to COVID-19 lockdown measures, including the declaration of a public health emergency in Georgia.

Table 2.

IRR [95% CI] for pre-pandemic trend, level-change, and post-pandemic trend for depression and anxiety encounters

Characteristics	Depression			Anxiety Disorders
	Pre-pandemic trenda, IRR (95% CI)	Level-changeb, IRR (95% CI)	Post-pandemic trendc, IRR (95% CI)	Pre-pandemic trenda, IRR (95% CI)	Level-changeb, IRR (95% CI)	Post-pandemic trendc, IRR (95% CI)
Overall	1.006 [0.991, 1.019]	0.969 [0.822, 1.142]	0.971 [0.947, 0.996]	1.008 [0.996, 1.019]	1.034 [0.893, 1.198]	0.962 [0.942, 0.982]
Overall, in-person	0.999 [0.984, 1.015]	0.406 [0.342, 0.482]	1.019 [0.984, 1.056]	0.997 [0.980, 1.013]	0.327 [0.273, 0.393]	1.027 [0.987, 1.070]
Overall, telehealth	1.151 [1.083, 1.225]	8.871 [4.210, 19.615]	0.822 [0.692, 0.979]	1.212 [1.129, 1.302]	9.570 [3.993, 24.684]	0.776 [0.642, 0.942]
CHIP Standard	0.989 [0.974, 1.005]	0.834 [0.711, 0.978]	1.056 [1.026, 1.087]	1.002 [0.986, 1.018]	1.041 [0.880, 1.232]	0.996 [0.965, 1.027]
TANF Foster Care	1.002 [0.991, 1.013]	1.056 [0.927, 1.204]	0.953 [0.934, 0.972]	1.005 [0.994, 1.015]	1.135 [0.998, 1.291]	0.953 [0.936, 0.972]
TANF Standard	1.011 [0.993, 1.028]	0.906 [0.737, 1.113]	0.977 [0.946, 1.008]	1.012 [0.998, 1.026]	0.945 [0.796, 1.121]	0.967 [0.941, 0.993]
Age 3–11	0.993 [0.977, 1.010]	0.819 [0.681, 0.986]	0.974 [0.942, 1.007]	1.006 [0.995, 1.018]	1.019 [0.872, 1.192]	0.959 [0.938, 0.981]
Age 12–17	1.008 [0.993, 1.023]	0.992 [0.829, 1.185]	0.971 [0.946, 0.996]	1.009 [0.997, 1.020]	1.050 [0.911, 1.210]	0.965 [0.945, 0.986]
Males	1.016 [1.002, 1.031]	0.958 [0.798, 1.150]	0.936 [0.910, 0.964]	1.005 [0.993, 1.017]	1.002 [0.858, 1.170]	0.963 [0.942, 0.983]
Females	0.999 [0.985, 1.013]	0.976 [0.836, 1.139]	0.991 [0.967, 1.015]	1.010 [0.999, 1.022]	1.054 [0.913, 1.216]	0.962 [0.942, 0.982]

Coefficients estimated from negative binomial regression; IRR: Incidence rate ratio; ^a Estimates of changes in service utilization for time period from Jan 1, 2019 to March 31, 2020; ^b Estimates of changes in service utilization rates during the first month following the announcement of public schools closure on April 1, 2020; ^c Estimates of changes in service utilization rates for time period from April 1, 2020 to December 31, 2020

Subgroup analyses revealed that there were small but statistically significant decreases in month-to-month encounters for depression during the pandemic for TANF Foster care (IRR: 0.953, 95% CI: 0.934, 0.972), 12-17-year-olds (IRR: 0.971, 95% CI: 0.946, 0.996) and males (IRR: 0.936, 95% CI: 0.910, 0.964). In contrast, CHIP Standard participants, which account for the lowest proportion of beneficiaries, saw 5.6% (IRR: 1.056, 95% CI: 1.026, 1.087) increases in monthly encounters during the pandemic. There was a statistically significant immediate decrease in monthly encounters for depression among CHIP Standard participants (IRR: 0.834, 95% CI: 0.711, 0.978) and 3-11-year-olds (IRR: 0.819, 95% CI: 0.681, 0.986). Similarly, for anxiety, we observed statistically significant decreases in month-to-month encounters during the pandemic for TANF Foster care (IRR: 0.953, 95% CI: 0.936, 0.972), TANF Standard (IRR: 0.967, 95% CI: 0.941, 0.993), 3-11-year-olds (IRR: 0.959, 95% CI: 0.938, 0.981), 12-17-year-olds (IRR: 0.965, 95% CI: 0.945, 0.986), males (IRR: 0.963, 95% CI: 0.942, 0.983), and females (IRR: 0.962, 95% CI: 0.942, 0.982).

Discussion

Our results suggest that compared to the pre-pandemic period, the average monthly encounters for depression and anxiety per 1,000 beneficiaries decreased during the pandemic irrespective of encounter setting (in-person vs. telehealth). There was a significant decrease in healthcare utilization for depression among children aged 3–11 years old during and after the pandemic. We found a decrease in utilization rate (number of encounters per patient) during the pandemic despite the stable total number of encounters which may have been due to the entry of newly diagnosed patients with depression and anxiety. This means that, from the provider’s perspective, the cases remained stable. Whereas from the patient’s perspective, access to care during the pandemic could have been limited or the decline could have been due to other reasons such as services provided by faith organizations or non-profits. However, multiple studies support the notion that access to health care was severely impacted by the pandemic, and that this lack of access was particularly detrimental to individuals with existing mental illness and to younger groups [30, 31]. The consequences of this enlarged gap in access to care may have been detrimental to mental health outcomes for the patients in the system. The lower monthly encounters per 1,000 beneficiaries further suggest that the frequency of care for both existing and new patients was lower during the pandemic. The statistically significant decrease in telehealth encounters and modest increase in in-person encounters in the post-pandemic period potentially suggests a slow rebound in health service utilization over time.

Even though we expected health service utilization to increase based on reports of increased anxiety and depression, lack of increase in utilization could be suggestive of lack of access to healthcare services and is consistent with the lower access to mental health services in the study state [12]. Service delivery via telehealth may not offset the decline in in-person care utilization due to several barriers such as limited access to digital services and internet, lack of experience with service delivery via telehealth, etc [32]. However, effective tools and resources are available to assist healthcare professionals to provide appropriate mental health services to pediatric patients [33]. An analysis of claims data from the CMS showed that Medicaid and CHIP beneficiaries during the pandemic showed a 34% (14 million) decline in number of mental health services utilized by children under age 19 [34]. The same analysis also reported that rebound in mental health care utilization had been slower compared to other services such as dental services. It would be worthwhile for future research to study the trends in the utilization of mental health services beyond the time point covered by this study. Access issues to mental health care among children and adolescents likely persist beyond the pandemic.

Furthermore, our study found an increase in in-person visits as schools opened in August 2020. Mental health services are widely available and provided in schools, responding to students’ healthcare needs [35]. School counseling plays a major role in identifying depression and anxiety and referring children to healthcare services for appropriate care and follow up through school-based health centers. A meta-analysis showed notable reductions in anxiety and depression associated with counselor and school professional interventions [36]. Another study highlighted the impact of counselors on mental health outcomes during COVID-19, further supporting the idea that they may facilitate a genuine decrease in anxiety and depression [37]. However, some schools, especially those located in rural areas, do not have health practitioners and funding to offer mental health services [38]. Many schools began providing counseling and behavioral health services virtually during the pandemic which may have contributed to the stable number of total encounters throughout the study period but an increase in in-person visits as schools started re-opening.

This study has a few limitations. Unlike other studies using a similar approach [39], a control group was not available, and the study was set up as a before-after design. Lack of control could also have been addressed by using a different outcome; however, our data was limited to mental health patients with at least one inpatient/outpatient encounter for mental health. In our approach, the interrupted time series has multiple pre- and post-observations to account for the underlying trends, and confounding due to within-group characteristics is rare, including confounding from comorbidities [24, 40, 41]. The study period in general was short, with 15 months before and 9 months during the pandemic; however, the sample size per time point (month) was large and ranged from 6,900 to 9,700 encounters for depression and 7,407 to 10,391 encounters for anxiety. In a time-series analysis, sample size per time point has a larger impact on power [42]. Furthermore, future studies should also separate inpatient and outpatient encounters to identify differences in trend pre- and post- to further analyze utilization for different types of services separately. Further, using administrative claims data presents some limitations. The ICD-10 codes in the claims data are codes used by healthcare professionals to classify diagnosis; therefore, we are unable to report the approaches used for diagnosis. The data available is accurate but only represents care delivered and therefore in our findings the increasing rates of anxiety and depression observed from self-reported data did not translate to increased healthcare utilization, which could be suggestive of lack of access to needed mental health services. Due to limitation in number of years of data availability we were unable to identify new cases vs. existing patients, however, the dataset contained data from patients who were continuously enrolled in the CMO plan during the study period. Medicaid recipients enrolled in one specific CMO were included, therefore though likely it cannot be concluded that other Medicaid recipients enrolled in other CMOs experienced similar trends. Likewise, findings cannot be generalized children who are not Medicaid recipients or children from other states. Accurately capturing telehealth services during the early months of the pandemic when billing codes/modifiers were being expanded may have led to underestimation of telehealth encounters during that period. Further, we reported service utilization based on encounter counts, which may not fully reflect changes in access among unique patients. Finally, the seasonal nature of mental illnesses may be masking the effect of the pandemic [26]. Our descriptive quarter-by-quarter comparison (Table A2) suggested a positive effect that school openings may have had on children and adolescents’ mental health and in managing early symptoms of depression and anxiety. Future studies must leverage multiple years of data before and after the pandemic to utilize the natural experiment of school closures to study the role of schools on promoting mental health among children and adolescents, through early diagnosis and timely provision to care.

Conclusions

We observed decreases in health care service utilization for depression and anxiety disorders among children and adolescents during the COVID-19 pandemic. This was expected given the consequences the pandemic had on health systems and mobility. With the increase in the prevalence of mental health disorders and the decrease in access to mental health care, the gap in access to care may have been widened. The COVID-19 pandemic increased the need for mental health services and simultaneously increased barriers to access care. More research is needed to understand the impact of limited access to mental health care on children’s mental health outcomes.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

KT– conceptualization, formal analysis, writing original draft; JRT– conceptualization, methodology, writing– review & editing; YA– assisting with formal analysis, writing original draft, writing– review & editing; DS & ES – critical feedback, writing– review & editing; EAV – critical feedback, supervision, writing– review & editing. All authors read and approve the final manuscript.

Funding

We acknowledge support from Resilient GA Inc. for providing the funds to conduct this study.

Data availability

Restrictions apply to the availability of the data, which were used under license for the current study, and so are not publicly available. Data are however, available from the authors upon reasonable request.

Declarations

Ethics approval and consent to participate

This study was reviewed and approved by the University of Georgia’s Institutional Review Board. The review determined the study as Not Human Subjects Research because the dataset did not contain any protected health information and was de-identified.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.