key: cord-0427780-9sv03cqa
authors: Evensen, M.; Hart, R. K.; Godoy, A. A.; Hauge, L. J.; Lund, I. O.; Knudsen, A. K.; Grotting, M. W.; Suren, P.; Reneflot, A.
title: Impact of the COVID-19 pandemic on mental healthcare consultations among children and adolescents in Norway: a nationwide registry study
date: 2021-10-11
journal: nan
DOI: 10.1101/2021.10.07.21264549
sha: c25a697987e9e42c577a5875fd81d656a333ba16
doc_id: 427780
cord_uid: 9sv03cqa

ABSTRACT BACKGROUND: There have been widespread concerns about the impact of the COVID-19 pandemic and its associated restrictions on children's and adolescents' mental health. While some studies have found increasing rates of mental health problems during the pandemic, other evidence suggest that mental symptoms and disorders were increasing before the pandemic. This study compared trends in mental healthcare use during the first 15 months of the pandemic with similar pre-pandemic trends. METHODS: Consultations related to mental symptoms and disorders were identified through national registries from primary and specialist healthcare services, including hospitalizations, covering the entire population of children 6-18 years in Norway (N=908 272). The monthly likelihood of having a consultation or hospitalization related to overall mental health problems, and specific diagnoses for depression, anxiety, ADHD and sleep problems were plotted from January 2020 to May 2021 and compared to trends over the same period between 2017-2019 using event study and difference-in-difference designs. FINDINGS: We found reductions in consultations for mental health symptoms and disorders in primary and specialist healthcare during the first weeks of lockdown in 2020. This decline was temporary, and volumes of consultations quickly returned to pre-pandemic levels. However, during fall 2020, consultation volumes related to mental health in primary care increased. This increase persisted into the winter of 2021 but levelled off by the last month of the sample period. The increase in consultation volumes was about 50 % compared to the pre-pandemic period. We did not find increases in consultations in specialist healthcare, except for hospitalizations. CONCLUSION: We found an increase in primary care consultation volumes related to mental health among children that depart from the previously established increases over recent years. We did not see similar increases in consultations in the specialist healthcare. It is unknown whether increases in primary care consultations represent milder cases, which primarily do not need specialist treatment, or if the stability in specialist healthcare consultations reflects capacity problems or timelags in referrals.

The COVID-19 pandemic, declared by the WHO on March 11, 2020 , prompted a range of interventions such as social distancing and stay-at-home orders that affected the everyday routines for children and adolescents, including closing of schools and leisure time activities to slow transmission rates. In Norway, a national eight-week lockdown was implemented from March 12, with gradual re-opening throughout the summer of 2020. However, as the pandemic continued during 2020 and 2021, many social restrictions were sustained and reinforced. There have been widespread concerns amount the impact of these restrictions on children's mental health (1-3).

Childhood and adolescence are a peak time for the onset of common mental health problems such as anxiety, depression, and ADHD (4) . Estimates show that about one in five children and adolescents in Western countries suffer some impairment from mental health problems (5) . If left untreated, mental health problems can have lasting effects into adulthood and are associated with lower education and income (6, 7) .

Several factors could potentially worsen mental health among children and adolescents in the wake of the pandemic. For example, stay-at-home orders, including school-closings and restrictions in leisure activities and social gatherings, could lead to increased loneliness and isolation with potentially harmful consequnces (8, 9) . Uncertainty about the length and scope of the pandemic may also lead to fear and worries (10) . The pandemic was followed by an economic downturn, including job loss and economic uncertainty, known to have adverse effects on children's mental health (11, 12) . Transitions to homeschooling during the pandemic negatively influenced many children's learning outcomes which may spill over to their wellbeing (13) . Furthermore, as mental health problems are more prevalent among children of lower socioeconomic origins, social distancing measures may exacerbate already marked social inequalities in child health (14) . In contrast, there are reports of unintended benefits of the pandemic, such as reduced bullying (15) , reduced parental stress (16) , and increased awareness about mental well-being, which could buffer against some detrimental consequences.

Shortly after the onset of the pandemic, there were international reports of a possible worsening of mental health among children and adolescents (17) . However, most of these studies were based on convenience sampling, relied on cross-sectional estimates on measures of mental health, and focused on mental health problems during quarantine (18) . Even before the pandemic, rates of mental health problems had been increasing (19, 20) , urging caution in attributing any increase to the pandemic and its associated restrictions. Existing evidence from larger studies comparing measures of mental health collected before the pandemic with data collected during the pandemic is mixed (10, 21, 22) . Two studies of short-term consequences (up to summer 2020) show no substantial changes in mental health (21, 23) . Two studies follow children to fall 2020, an Icelandic study report deterioration in children's mental health while a Norwegian study suggests no substantial changes (24, 25) . Beyond differences in the observation period, the mixed findings may reflect differences in questionnaire scales, age profiles, sample selections and settings. Moreover, previous studies have relied on self-or parent-reported symptoms of mental health problems with less knowledge about healthcare use for mental health problems. The latter is important, as changes in healthcare utilization, particularly specialist healthcare, for mental health problems may indicate a more severe change in children and adolescents' mental health status than can be captured through symptom questionnaires. Finally, the pandemic and its associated restrictions may have had both acute and longer-term impacts on mental healthcare utilization, which may have differed as the pandemic evolved. For example, reduced capacity or fear of contagion may have reduced utilization, while an increased focus on mental health may have lowered the threshold for seeking professional help. In the acute phase, mental healthcare may have been reduced, for instance, due to lockdown or fear of contamination of the virus. This may in turn lead to a longer-term increase in mental healthcare utilization, as a potential result of previously unmet need for mental healthcare utilization, an increase in mental health problems or a reduced threshold for help-seeking.

This study examines changes in consultation volumes related to mental health symptoms and disorders among children 6-19 years old using population-wide data on all primary and specialist healthcare use during the pandemic compared to pre-pandemic years. Our approach allows us to net out seasonal effects and period changes. Further, we examined whether consultation volumes changed more among children with high and low parental SES. In Norway, primary and specialist healthcare is free for all children below 18 years old and mental healthcare for children has been operating at normal capacity.

. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted October 11, 2021 We use data from the Norwegian registry BeredtC19, a national emergency preparedness registry administered by the Norwegian Institute of Public Health (26) . It includes data from the Norwegian Control and Reimbursement Database (KUHR) and the National Patient Registry (NPR) matched with data from the Population Registry (Statistics Norway). Unique (de-identified) personal identifiers allow for linkage between different registries and between children and their parents. The study sample was restricted to all children aged 6-19 in 2018 or 2020 (see Appendix A for details on sample construction).

Diagnoses of mental health problems were taken from two sources: reimbursement data from primary healthcare services (KUHR) and specialist data from the NPR. Primary healthcare comprises services such as consultations with general practitioners (GPs) and emergency room visits. Diagnostic information is registered in KUHR according to the International Classification of Primary Care (ICPC-2) with either a symptom or disorder code (27) .

The NPR is a nationwide registry covering all consultations in specialist healthcare coded in accordance with the 10 th edition of the International Classification of Diseases and Related Health Problems (ICD-10). A referral from the GP is necessary to get specialist treatment (except for acute hospitalizations).

Monthly measures indicating at least one mental health consultation or hospital admission were constructed for: (i) all mental symptoms and disorders registered in primary care and specific diagnoses for ADHD, anxiety, depression, and sleep problems. (ii) all mental disorders in specialist care as well as specific diagnoses for ADHD, anxiety, depression and hospitalizations (see Table 1 for details on coding). Due to the high level of comorbidity between anxiety and depression, we analyzed these disorders jointly (28).

To evaluate how consultation volumes in primary and specialist healthcare for mental symptoms and disorders changed due to the COVID-19 pandemic and its countermeasures, we compare the use of health services from January 2019 to June 2021 for a pandemic cohort and from January 2017 to June 2019 for a pre-pandemic cohort.

To separate effects from the pandemic from other temporal trends, we followed the prepandemic cohort over the same time span and ages, albeit two calendar years earlier. We . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted October 11, 2021. ; https://doi.org/10.1101/2021.10.07.21264549 doi: medRxiv preprint show bivariate trends for the pre-pandemic and the pandemic cohorts. We fit multivariate event study models with controls for month and time in years to formally test, month by month, whether the use of healthcare services in the period 2019-2021 differs from two years earlier. Data for the first part of the period is used to assess whether trends were comparable in the two cohort groups before the onset of the pandemic. Data up to February 2020 for the pandemic sample, and February 2018 for the pre-pandemic sample, are used for this purpose.

Then, we assess whether diverging trends in consultations emerged at the onset of and during the pandemic and its associated restrictions. To quantify the magnitude of the effects, we also estimate difference-in-difference models, where we group the months into four periods (see methodological details in Appendix A). Table 1 shows the distribution of our covariates for a data set of person-months for the entire study. In the pandemic cohort in the period before lockdown (column January 2019-February 2020), 1.2 percent of the children had a mental health consultation in primary care in any given month, and 1.9 percent had a mental health consultation in specialist care.

[ Table 1 about here] [ Figure 1 about here] First, the graphs document marked seasonal variations in consultations, with large decreases in July each year (school holiday) and small peaks in January for some outcomes. Up to March 2020 (March 2018), the trends are comparable in the pandemic and pre-pandemic cohorts. Second, there is a weak increase in consultations over time, so the pandemic cohort is often at a higher level than the pre-pandemic cohort until March 2020. Third, the share of consultations dropped sharply around the lockdown in the pandemic cohort but increased rather quickly to pre-lockdown levels. Finally, from September 2020, the number of . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted October 11, 2021. ; https://doi.org/10.1101/2021.10.07.21264549 doi: medRxiv preprint consultations started to increase faster in the pandemic cohort compared to the pre-pandemic cohort. The increase pertains to all primary care outcomes (Panels a-d). For specialist care (panels e-h), there is a tendency of a faster increase from January 2021 for the age group 13-15 years.

To formally test whether the healthcare utilization of the pandemic cohort differed from that of the pre-pandemic cohort, we netted out shared seasonal differences and secular change over time. Thus, using event study models, we plot monthly deviations from pre-pandemic trends.

[ Figure 2 about here] Figure 2 , panel a shows that the monthly probability of having any primary healthcare visits related to mental health decreased sharply at the start of lockdown . The dip is largest for anxiety and depression consultations, and there is no decline for sleep disorders (panels b-d, and). After the lockdown period, we see a quick rebound and levelling off, with a slight increase (for all mental health consultations, sleep disorders and ADHD) or no change (for anxiety and depression, except a fall in the oldest age group) in summer relative to the prepandemic cohort. As indicated by the bivariate plots ( Fig. 1) , the share of children with a primary care consultation (Fig. 2 , Panel a), as well as for all diagnostic groups in primary care (Panels b-d) starts to increase faster around August 2020 for the two youngest age groups, levelling off at a substantially higher level in 2021. We see no increases in consultation volumes in specialist healthcare when all mental health consultations are considered jointly (Panel e) or for anxiety/depression (Panel g). However, there is a significant increase among 13-15-year-old children for hospitalizations (Panel h).

To get a sense of the magnitude of these effects we show difference-in difference estimates (ie., average monthly coefficients) across 4 time periods (lockdown, summer, fall and winter)

in Table 2 .

[ Table 2 about here]

. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted October 11, 2021. ;  The estimates for all primary mental health consultations during lockdown suggest a 30.5 percent reduction among 16-19-year-olds, 22.3 percent fall for those aged 13-15, and 9.7 percent reduction for those aged 6-12, all statistically significant at the 5 percent level. During winter 2021, however, the share of children with any mental health consultation is 47.2 percent higher in the pandemic cohort compared to the pre-pandemic cohort for 13-15-yearold children (relative to pre-pandemic baseline 1.1 percent), and 31.7 percent higher for 6-12-year-old children (baseline 0.7 percent) On a relative scale, effects are strongest for anxiety/depression and sleep disorders, and weakest for ADHD. Except for a small increase in ADHD consultations, there is no increase for the oldest age group (Fig. 2) . The increase in hospitalizations among 13-15-year-olds, is estimated to 21 percent in the difference-indifference models, relative to a baseline of 0.1 percent . ADHD consultations (panel f) increase in both age groups, estimated to 8.1 percent from a baseline of 0.6 percent among the 6-12-year-old children and 15.2 percent from a baseline of 0.7 percent among the 13-15-yearold children.

We ran the models separately by sex (Appendix Figure A .1). Across outcomes, the increase starting around August 2020 is stronger among females than males, but confidence intervals overlap. The increases in ADHD consultations in specialist care (Panel g) and hospitalizations (Panel h) is found among females only. We also estimated the sex-specific models for 13 -15- year-old children only, the age group where we have seen the largest effects (Appendix Figure   A .2). For this group, the sex differences in increases are even larger, and the 95% confidence interval for the event study estimates are no longer overlapping for all primary care (Panel a), anxiety and depression (Panel b), all specialist care (Panel e) and ADHD specialist care (Panel g).

We also split the sample by parent's socioeconomic status using the information on parental occupation from Statistics Norway (see appendix for details). Results were similar across socioeconomic groups ( Figure A.3 ).

Finally, we tested whether the effects differed between the Capital area (regions Oslo and Viken), which had by far the strictest restrictions, and the rest of Norway (Appendix Figure   A .4). We find no evidence that the effects are restricted to the Capital area.

. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted October 11, 2021. ; https://doi.org/10.1101/2021.10.07.21264549 doi: medRxiv preprint

Using population-wide data on mental healthcare for the first 15 months of the pandemic, we found a pronounced increase in primary care consultation volumes related to mental health symptoms and disorders among children that depart from previously established increases over recent years. Although the number of consultations for mental health declined sharply during the initial period of lockdown, consultation volumes returned to pre-pandemic levels by June 2020. However, our models uncovered a gradual increase in the number of primary consultations related to mental health during fall 2020 and winter 2021, corresponding to the second and third waves of infections and their associated social distancing mandates. We did not find similar increases for mental disorders in the more selective specialist healthcare, except for hospitalizations.

Compared to pre-pandemic years, primary care mental health consultations increased by 47.5 percent in the first five months of 2021. Both on an absolute and relative scale, the increase was highest for anxiety and depression, with a 65.9 percent increase in 2021. ADHD and sleep disorders increased with 29.4 and 52.3 percent, respectively, for the same period and age groups. The increase was found for both sexes but was most pronounced among girls.

The increase in primary care consultations was less pronounced for older adolescents (16) (17) (18) (19) years old). It could be that the oldest adolescents are better at coping with the pandemic and its associated restrictions. This would be in line with some previous Norwegian studies, but slightly at odds with other studies that have reported largest deteriorations in mental health among older children (24, 29) . A policy change regarding absences for upper secondary school students, which might have affected consultations volumes in this age group only. Before the pandemic, absences from upper secondary to illness or injury were required doctor's certification. This requirement was lifted shortly after the beginning of the pandemic, reduceding the demand for primary care consultations in this age group (30).

The lack of increases in specialist care could suggest that the increase in mental health consultations seen in primary care to a large extent reflect an increase mostly in milder symptoms that does not require specialist treatment, or it could reflect a greater propensity to seek treatment. In Norway, less severe mental health problems will generally be treated by a physician in primary care while more severe cases will be referred to specialist treatment by the physician and treated by psychologists and psychiatrists. Alternatively, the lack of . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted October 11, 2021. ; https://doi.org/10.1101/2021.10.07.21264549 doi: medRxiv preprint increase in specialist treatment could reflect constraints in treatment. Already before the pandemic, specialist healthcare had long wait lines and delayed access to treatment (31) , meaning that an increased demand for specialist healthcare would not necessary show up in our data as an increase in consultations. We did try to address these challenges by analyzing acute psychiatric hospitalizations, which are not affected by capacity constraints in the same way, and for this outcome we found a 21 per cent increase. However, hospitalizations are rare among children (0.1 per cent), which limits the generalizability of these findings to overall consultation volumes in specialist care.

There have been concerns that the pandemic might increase the already large social inequalities in the prevalence of mental health disorders between children from high and lowincome families (14) . However, we found that the increase in mental health consultations was largely similar among children of parents with high and low occupations. This suggests that the pandemic has not exacerbated social inequality related to mental health consultations.

To our knowledge, ours is the first study to examine possible increases in consultation volumes related to children's mental health with a research design that handles both age change and (linear) period trends. We established empirically that the pre-pandemic and the pandemic cohorst had similar trends consultation volumes before March 2020, and then showed that the trends diverge markedly over time. Thus, our results suggest that the pattern of consultation volumes increased beyond what we would expect based on previous trends.

The increase in consultation volumes became visible 6 to 8 months into the pandemic, suggesting that most children (and parents) coped with changes in the short run. Still, in the long run, consultation volumes increased.

To the extent that our findings reflect a worsening of underlying mental health, this in line with an Icelandic study indicating that self-reported depression increased during the pandemic (24) . In contrast, Norwegian survey-based studies indicate no increase up to fall 2020 (25) .

However, the conflicting findings may reflect timing of data collection since our study has a considerably longer follow-up (May 2021), and the largest effects are found toward the end of the period. The decline in consultations during the initial lockdown happened in a period where surveys suggest no worsening of mental health. This suggests that our estimates are driven by a temporary change in healthcare utilization rather than an improvement in underlying health.

. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted October 11, 2021. ; https://doi.org/10.1101/2021.10.07.21264549 doi: medRxiv preprint

There are multiple aspects of the pandemic that could plausibly lead to deteriorating mental health, including diminishing social support networks (32) and unpredictability and disruption in daily routines (33). Even as restrictions eased, public health measures such as social distancing and attempts to reduce mixing of student across cohorts severely limited social gatherings (34). However, the increase seen in consultations could be due to greater media coveage and awareness of mental health problems. This could lead parents or physicians to rate children to have more symptoms now compared to pre-pandemic years. Increased family time during the pandemic (16) could make parents more responsive to their children's symptoms. If fears of contracting COVID-19 or overburdening the healthcare system increased the threshold for seeking help, the worsening of mental health would be larger than the increase in consultations suggest. Changes in consultation practice, such as more online consultations, could also affect our results by lowering the threshold to contact physicians.

However, findings from a sensitivity analysis examining consultations volumes for overall healthcare did not support this (cf. Appendix Fig.2 ).

There are strengths and limitations to our study. Unlike prior studies, that use self-report data, we likely capture clinically relevant symptoms and conditions causing distress in everyday life. Our inclusion of hospitalizations also means that we can examine changes among vulnerable children, which are likely not included in survey-based studies. However, relying on healthcare data also means that we only examine a small proportion of children with mental disorders. To the extent that one is interested in underlying mental health, it is a limitation with our study that our results can also be influenced by other changes affecting health service use, as discussed above. Future studies should examine underlying prevalence trends and using direct mental health measures in an equally robust design would be particularly valuable.

As for the validity of diagnoses in primary care, a previous study compared interview-based diagnoses for depression and anxiety with diagnoses taken from KUHR and NPR and found that registry-based diagnoses have moderate sensitivity and excellent specificity, with very few false positives (35). While we are not able to rule out that physician evaluations may have changed during the pandemic, we consider it unlikely that the increase is caused by sudden changes in diagnostic practice.

. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. 

We found that consultations related to mental health symptoms and disorders in primary care increased during fall and winter 2020-2021 over and above increases that occurred in recent years prior to the pandemic. We found no corresponding increase in consultations in specialist healthcare, except for hospitalizations. It is paramount to understand whether the increase in consultations reflects a worsening of child and adolescent mental health and if so, how this can be mitigated.

. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted October 11, 2021. ; https://doi.org/10.1101/2021.10.07.21264549 doi: medRxiv preprint Author approval:

For confidentiality reasons data cannot be shared.

None.

The study has been approved by the Norwegian Regional Committees for Medical and Health

Research Ethics (REC), approval number 2021/267200. We confirm that all administrative permissions have been granted to access and use the data for this study.

. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted October 11, 2021. . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted October 11, 2021. ; https://doi.org/10.1101/2021.10.07.21264549 doi: medRxiv preprint TABLES AND FIGURES: Diagnoses are based on ICPC-2 codes, Chapter P for primary care, and ICD-10, Chapter F for specialist care (see Table A1 ).

. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted October 11, 2021. ; https://doi.org/10.1101/2021.10.07.21264549 doi: medRxiv preprint Figure 1 : Percent of children with at least one consultation for mental health problems/disorders in primary and specialist healthcare in a given month. Diagnoses are based on ICPC-2 codes Chapter P for primary care, and ICD-10 Chapter F for specialist care (see Table A is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted October 11, 2021. ; Figure 2 : Results from separate event study models for three age groups. Complete lines show coefficients, and shaded areas their 95% confidence intervals. Coefficients and confidence intervals are scaled to the pre-lockdown level in the main sample (see Table 1 ). The outcome is the monthly propensity to have at least one consultation of the type mentioned in the panel headers. Diagnoses are based on ICPC-2 codes Chapter P for primary care, and. ICD-10 Chapter F for specialist care (see Table A .1). The x-axis refers to the measurement time for the main sample. For the comparison sample, all measurements are taken 24 months earlier. Age group 13-15 includes 16-year-olds for specialist care. Models control for duration in years, sex, municipality, month and easter holidays.

. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted October 11, 2021. ; https://doi.org/10.1101/2021.10.07.21264549 doi: medRxiv preprint Note: Results from Difference-in-Differences models, estimated separately by age group. For specialist care, age group 13-15 includes 16 year olds. Coefficients and confidence intervals are scaled to the pre-lockdown level in the main sample (see Table 1 ). The outcome is the monthly propensity to have at least one consultation of the given type. Diagnoses are based on IPCD codes Chapter P for primary care, and. ICD-10 Chapter F for specialist care (see Table A .1). For the comparison sample, all measurements are taken 24 months earlier. Models control for duration in years, sex, municipality, month, age category and easter holiday.

.

It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint To formally compare the trend development in the intervention and comparison group, we estimate event study models, taking the following form:

Where t0 refers to the first month of lockdown, and k is month number. The expression For 2019, most person-month records will be included in both the control-and intervention cohort (albeit at different durations). We test whether the results are sensitive to this by reducing the observation period, so that no person month is included in both the main and comparison sample. The results are not sensitive to this ( Figure A.6) .

. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted October 11, 2021. ; https://doi.org/10.1101/2021.10.07.21264549 doi: medRxiv preprint category), a dummy for being male (unless models are separate by sex) and a variable running from 0 to 1 showing the proportion of easter falling into the given month in the given year.

If pre-trends are parallel, i.e., the β k coefficients for before lockdown (negative t's) should be insignificant and close to zero. Effects of (prolonged) lockdown should then become emergent no earlier than t0. Note that the direction of effects and their drivers may vary over time: while the access to health services was restricted in the immediate lockdown, they were generally accessible in the prolonged period of social distancing that followed.

We also estimate difference-in-difference models (i.e.., average monthly coefficients) for the same outcomes, including the same control variables. In these models, we collapse the duration variables into periods. We group the months into four periods (with measurements in 

Note that for the analysis of specialist healthcare, we restrict our sample to children aged [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] years in the starting year. This different restriction is due to that at age 17 children are transitioned from child to adult specialist healthcare age out of parts of the specialist care system at age 17. We follow the children for a year after the starting year, and to obtain a balanced panel, we need to make sure that they do not "age out" of specialist health care.

Mean age is therefore slightly lower in the specialist care sample due to these restrictions.

Parental occupations were registered in accordance with the International Standard Classification of Occupations, ISCO-88 (21) . Using information on the parent with the lowest first digit in the ISCO code, corresponding approximately to the highest occupational status, we distinguish between three main parental class categories: upper white collar, lower white collar and blue collar. Parental occupations was measured January 1 2020 for the pandemic cohort, and January 1 2018 for the pre-pandemic cohort.

. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted October 11, 2021. ; https://doi.org/10.1101/2021.10.07.21264549 doi: medRxiv preprint . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted October 11, 2021. ; Figure A .1: Results from separate event study models for males and females. Complete lines show coefficients, and shaded areas their 95% confidence intervals. Coefficients and confidence intervals are scaled to the prelockdown level in the main sample (see Table 1 ). The outcome is the monthly propensity to have at least one consultation of the type mentioned in the panel headers. Diagnoses are based on ICPC-2 codes Chapter P for primary care, and ICD-10 Chapter F for specialist care (see Table A .1). The x-axis refers to the measurement time for the main sample. For the comparison sample, all measurements are taken 24 months earlier. Models control for duration in years, age category, municipality, month and easter.

. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted October 11, 2021. ;

Results from separate event study models for males and female, ages 13-15 for primary and 13-16 for specialist care. Complete lines show coefficients, and shaded areas their 95% confidence intervals. Coefficients and confidence intervals are scaled to the pre-lockdown level in the main sample (see Table 1 ). The outcome is the monthly propensity to have at least one consultation of the type mentioned in the panel headers. Diagnoses are based on ICPC-2 codes Chapter P for primary care, and ICD-10 Chapter F for specialist care (see Table A .1). The x-axis refers to the measurement time for the main sample. For the comparison sample, all measurements are taken 24 months earlier. Models control for duration in years, age category, municipality, month and easter.

. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted October 11, 2021. ; Figure A .3: Results from separate event study models by parents' social background. Complete lines show coefficients, and shaded areas their 95% confidence intervals. Coefficients and confidence intervals are scaled to the pre-lockdown level in the main sample (see Table 1 ). The outcome is the monthly propensity to have at least one consultation of the type mentioned in the panel headers. Diagnoses are based on ICPC-2 codes Chapter P for primary care, and ICD-10 Chapter F for specialist care (see Table A .1). The x-axis refers to the measurement time for the main sample. For the comparison sample, all measurements are taken 24 months earlier. Age group 13-15 included 16 year old children for specialist care. Models control for duration in years, sex, municipality, month, age category and easter.

. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted October 11, 2021. ; https://doi.org/10.1101/2021.10.07.21264549 doi: medRxiv preprint Figure A.4: Results from separate event study models for capital area (Oslo and Viken counties) and the rest of Norway, all age groups. Complete lines show coefficients, and shaded areas their 95% confidence intervals. Coefficients and confidence intervals are scaled to the pre-lockdown level in the main sample (see Table 1 ). The outcome is the monthly propensity to have at least one consultation of the type mentioned in the panel headers. Diagnoses are based on ICPC-2 codes Chapter P for primary care, and ICD-10 Chapter F for specialist care (see Table A .1). The x-axis refers to the measurement time for the main sample. For the comparison sample, all measurements are taken 24 months earlier. Models control for duration in years, sex, age category, municipality, month and easter.

. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted October 11, 2021. ; https://doi.org/10.1101/2021.10.07.21264549 doi: medRxiv preprint

A concern is that the pandemic and the associated consequences changed all primary health care utilization, so all health care utilization increased, not only that related to mental health.

To provide a robustness check of the results for primary health care service use, we show the development for all primary care consultations in Appendix Figure A.5. Figure A.5: Trends and event study models for any primary care consultation. In Panel b, complete lines show coefficients, and shaded areas their 95% confidence intervals. Coefficients and confidence intervals are scaled to the pre-lockdown level in the main sample (see Table 1 ). Diagnoses are based on IPCD codes Chapter P for primary care, and. ICD-10 Chapter F for specialist care. The x-axis refers to the measurement time for the main sample. For the comparison sample, all measurements are taken 24 months earlier. Separate models by three age groups, primary school (ages 6-12), secondary (ages 13-15, 13-16 for specialist health care) and high school (ages [17] [18] [19] . Models control for duration in years, sex, municipality, month and easter.

As for mental health, there is a lockdown-dip followed by a recuperation for this outcome.

However, compared to mental health consultations, the share of any primary care consultation displays a much more modest increase after the lockdown period for the two youngest age groups. For the age group 16-19, the total number of consultations falls throughout the school year 2020-202. Difference-in-difference estimates (Table 2) suggest that as of 2021, primary care consultations were unchanged for children aged [13] [14] [15] year and had fallen by 3.2 percent in the youngest age group, and 25.8 percent in the oldest age group. One explanation for the sharp fall in the oldest age group, is that prior to lockdown, sickness absence from high school . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted October 11, 2021. ; https://doi.org/10.1101/2021.10.07.21264549 doi: medRxiv preprint (which this age group attends) beyond a low threshold had to be doctor certified. After lockdown, this requirement was removed, potentially changing the need for primary care services in this group quite substantially. As discussed in the main text, this is also likely to influence the results for mental health consultations. For the two youngest age groups, absence is certified by parents rather than doctors.

. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted October 11, 2021. ; https://doi.org/10.1101/2021.10.07.21264549 doi: medRxiv preprint Figure A.6: Results from separate event study models for three age groups. Shortened observation window. Complete lines show coefficients, and shaded areas their 95% confidence intervals. Coefficients and confidence intervals are scaled to the pre-lockdown level in the main sample (see Table 1 ). The outcome is the monthly propensity to have at least one consultation of the type mentioned in the panel headers. Diagnoses are based on IPCD codes Chapter P for primary care, and. ICD-10 Chapter F for specialist care (see Table A .1). The x-axis refers to the measurement time for the main sample. For the comparison sample, all measurements are taken 24 months earlier. Age group 13-15 included 16-year old's for specialist care. Models control for duration in years.

. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted October 11, 2021. ; https://doi.org/10.1101/2021.10.07.21264549 doi: medRxiv preprint

Coronavirus Disease 2019 (COVID-19) and Mental Health for Children and Adolescents

Mental health effects of school closures during COVID-19. Lancet Child Adolesc Health

Growing up in the shadow of COVID-19. Lancet Child Adolesc Health

Adolescent mental health-Opportunity and obligation

Lifetime prevalence of mental disorders in US adolescents: results from the National Comorbidity Survey Replication-Adolescent Supplement (NCS-A)

Adolescent mental health and earnings inequalities in adulthood: evidence from the Young-HUNT Study

Adolescent depression and educational attainment: results using sibling fixed effects

Rapid Systematic Review: The Impact of Social Isolation and Loneliness on the Mental Health of Children and Adolescents in the Context of COVID-19. J Am Acad Child Adolesc Psychiatry

The effects of social deprivation on adolescent development and mental health. Lancet Child Adolesc Health

Stress and Mental Health among Children/Adolescents, Their Parents, and Young Adults during the First COVID-19 Lockdown in Switzerland

Impact of the Great Recession on mental health, substance abuse and violence in families with children: a systematic review of the evidence

How do economic downturns affect the mental health of children? Evidence from the National Health Interview Survey. Health Econ

Learning loss due to school closures during the COVID-19 pandemic

Socioeconomic inequalities and mental health problems in children and adolescents: a systematic review

The COVID-19 Pandemic Disrupted Both School Bullying and Cyberbullying

Risk of stress/depression and functional impairment in Denmark immediately following a COVID-19 shutdown

Child and adolescent mental illness during COVID-19: A rapid review

Mental Health Status Among Children in Home Confinement During the Coronavirus Disease

Increase in diagnosis of depressive disorders contributes to the increase in antidepressant use in adolescents

Annual research review: secular trends in child and adolescent mental health

The impact of COVID-19 related lockdown measures on self-reported psychopathology and health-related quality of life in German adolescents

The impact of COVID-19 on the lives and mental health of Australian adolescents. Eur Child Adolesc Psychiatry

Adolescents' symptoms of anxiety and depression before and during the Covid-19 outbreak -A prospective population-based study of teenagers in Norway. Lancet Reg Health -Eur

Depressive symptoms, mental wellbeing, and substance use among adolescents before and during the COVID-19 pandemic in Iceland: a longitudinal, populationbased study

Self-reported Mental and Physical Health Among Norwegian Adolescents Before and During the COVID-19 Pandemic

Emergency preparedness register for COVID-19 (Beredt C19)