A comprehensive analysis of age of onset and cumulative incidence of mental disorders: A Danish register study

The age of onset (AOO), incidence and cumulative incidence of mental disorders are critical epidemiological measures, providing essential insights into the development and course of these disorders across the lifespan. This study aims to provide up‐to‐date estimates of the AOO, age‐specific incidence, and cumulative incidence for a comprehensive range of mental disorders using data from Danish registers.


| INTRODUCTION
Age of onset (AOO) and lifetime risk (i.e., the absolute risk or cumulative incidence proportion for every age) are fundamental features underpinning the epidemiologic landscape of mental disorders.2][3] While understanding the AOO of mental disorders is required in order to match age-appropriate services for the different types of mental disorders, understanding lifetime risks is also important-some disorders are more common than others and this should be reflected in decisions about service planning. 4AOO and lifetime risks are also key inputs for studies interested in quantifying the burden of disease. 5,6y plotting the incidence rates, that is new cases per person-years, for a disorder by narrow age strata, the distribution of incident cases provides high-quality information on AOO distribution for different mental disorders.
Simultaneously, the cumulative incidence provides an estimate of lifetime risk (by a certain age, e.g., 80 years).Four key studies have been published that present these estimates for a range of mental disorders.The World Mental Health Survey group have published two of these, 1,7 with the most recent study presenting the results of community-based mental health surveys in 29 countries.However, many of these studies were based on surveys completed prior to 2007, and these surveys did not include estimates related to schizophrenia and autism spectrum disorders.In addition, Danish researchers have provided sex-specific incidence rates and lifetime risks for a comprehensive range of treated mental disorders 2 and psychotropic drug prescriptions. 8The published literature on AOO was recently summarized in a systematic review of 192 studies. 3This review furnished global epidemiological estimates concerning the AOO for diverse mental disorders, demonstrating that a considerable fraction of mental health conditions commenced by age 25.It underscored the imperative for early intervention in mental health and advocated for a refined strategy toward mental health care provision.This review reported key properties of AOO distributions for a range of mental disorders, however the underlying heterogeneity of the included studies potentially complicates the interpretation of the results.
Recent studies investigating genetic risks have employed AOO distributions in register-based family studies 9 and in genome-wide association studies based on case-control or case-cohort studies. 10The availability of lifetime population-based cumulative incidence and AOO distributions can bolster power in such genomewide association studies. 10here is a need to examine contemporary AOO and cumulative incidence for a range of mental disorders, and to increase transparency and accessibility by providing open access to the data.Although the Danish Psychiatric Central Research Register does not record the age at diagnosis (AOD), we have the opportunity to update the study by Pedersen et al. 2 and provide additional details of disorder-specific AOO and lifetime cumulative distributions (e.g., median and interquartile range for the AOO) for a range of mental disorders.In addition, we have provided an interactive data-visualization site to facilitate the exploration of the data for future genetic studies.

| METHODS
All Danish residents were followed in the period from January 1, 2004 until December 31, 2021 from the 1st, 5th, or 10th birthday, depending on the disorder of interest (Table 1) and until the 80th birthday..The sample was identified from the Danish Civil Registration System, 11,12 which has maintained information on all residents (including sex and date of birth) since 1968.The register has been continuously updated on vital status (e.g., death and emigration), and has a unique personal identification number that can be linked uniquely to all national registries.

| Assessment of mental disorders
Information on mental disorders was obtained from the Danish Psychiatric Central Research Register, 13,14 which contains information on all admissions to psychiatric inpatient facilities since 1969 and visits to outpatient psychiatric departments and emergency departments since 1995.The diagnostic system used was the Danish modification of the International Classification of Diseases, Eighth Revision (ICD-8) from 1969 to 1993, and Tenth Revision (ICD-10) from 1994 onwards.It is important to highlight that the Danish Psychiatric Central Research Register captures information on AOO rather than AOD, which may manifest at a subsequent time.We considered key mental disorders within the F subchapter (F10-F98), excluding organic disorders/dementia (F00-F09  1.Of note, the earliest possible AOO has been set as (a) 1 or 5 years of age for key childhood-onset disorders, (b) 10 years for traditionally adult-onset disorders, and (c) 1 year for the overarching "Any mental disorder" category.For each individual in the study, the date of onset for each disorder was defined as the date of first contact with the secondary psychiatric treatment system (inpatient, outpatient, or emergency visit).

| Study design
For each specific disorder, follow-up began either on January 1st 2004 or the earliest age at which a person might develop the specific disorder (Table 1), whichever came later.Follow-up was terminated at disorder onset, at age 80 years, death, emigration from Denmark, or December 31, 2021, whichever came first.We excluded disorder-specific prevalent cases at the start of the follow up period, based on all available ICD-8 and ICD-10 diagnoses in the Danish Psychiatric Central Research Register (i.e., prevalent cases since April 1969 were "washed-out"), and we included only people resident in Denmark at initiation of follow-up.This washout procedure was applied independently for each disorder under investigation.

| Statistical analyses
We utilized the counting process approach in our study to handle delayed entry (January 1, 2004), administrative censoring (December 31, 2021), and competing risks from death or emigration. 15Given that not all members of the study population were followed across their entire lifespan, cumulative incidences were created through age-specific incidence rates.For example, those born in year 2000 contributed information for the estimates at ages 3-21 years, while those born in 1990 contributed information for the estimates at ages 13-31 years, and so on.This approach ensured accurate data analyses and reliable interpretation of the results.
We employed Poisson generalized linear models incorporating a cubic spline with knots every 8 years to estimate the incidence rate as a function of age for each specific type of mental disorder, as well as for broader ICD-10 F-chapter categories, as presented in Table 1. 16eparate estimates are provided for males and females.The incidence rate represents the number of individuals receiving initial treatment for the respective mental disorder per 10,000 person-years at risk.
We utilized the Aalen-Johansen estimator of the cumulative incidence function, which accounts for the fact that persons are simultaneously at risk for a mental disorder, death, or emigration from Denmark. 17escriptive statistics for the AOO distributions (median, first quartile, third quartile, mean, 95% confidence interval) were provided for: (a) the cases only, and (b) the full population based on cumulative incidence curves.We present the lifetime risk, which is the absolute risk by age 80 years, as well as the absolute risk at 40 years of age.The full population median age was defined as the age at which half of the lifetime risk was reached.The quartiles were analogously defined.The population mean AOO was calculated using the area under the cumulative incidence curve. 18The precision of population medians and means surpasses that of caseonly estimates, as the former rely on the full population data rather than individual cases alone.These two types of estimates carry distinct interpretations.The population medians and means can be regarded as projected estimates at the initialization of follow-up.Conversely, the case-only medians and means can be interpreted as estimates conditioned upon the occurrence of the disorder in the future. 19

| Ethics and data approvals
The Danish Data Protection Agency and the Danish Health Data Authority approved this study.According to Danish law, informed consent is not required for registerbased studies.All data are de-identified and not recognizable at an individual level.

| Code and data availability
All programming code and summary outcomes is publicly available for downloading https://github.com/CBeck96/AOO2023.An interactive data-visualization website is available here https://csievert.shinyapps.io/mental-aoo-danish/.The age-and sex-specific data underlying the figures can be downloaded from the datavisualization platform and the GitHub website.We used R version 4.2.1 for the analysis.The protocol was preregistered on the Open Science Framework. 20

| RESULTS
A total of 6,222,682 individuals aged 1-80 years were followed for up to 91.6 million person-years.The number of cases ranged from 2963 for schizoaffective disorder to 463,057 for any disorder (Table 1).
Table 2 also shows the sex-specific median AOO for the mental disorder types.Negligible sex differences emerged: the median AOO was approximately age 22 years (median and interquartile range; males 21.37, (11.85-36.00);females 22.55 (16.31-36.08)).For both sexes, three quarters developed their disorder by age 36 years.Figure 1 shows that the cumulative risk of any mental disorder is higher in males in the first two decades of life, after which females have a higher cumulative risk.The age distributions of incidence rates peaks earlier in males (12 years) compared with females (18 years).
Figure 2 shows four of the broad (pooled) categories.Here we see striking sex differences, with mood disorders (panel A), and neurotic, stress-related and somatoform disorders (panel B) being more common in females compared with males.In contrast, mental and behavioral disorders due to psychoactive substance use (panel C), and schizophrenia (panel D) are more common in males.Of interest, the age of peak incidence rates for these four disorder groupings is comparable between males and females.In addition, within the cumulative incidence curves, the median AOO for these broad disorders are similar in males and females (shown by the colored dots).Mood disorders show an increased incidence in later life (after age 60 years) for both males and females.Comparable figures for the other pooled disorders (eating disorders, specific personality disorders, intellectual disabilities, pervasive developmental disorders, and behavioral and emotional disorders with onsets usually occurring in childhood and adolescence) can be found in Figures S1-S5.
Figure 3 shows comparable figures for the 12 specific disorders.Note that each of the 12 panels in Figure 3 can be found in full size in Figures S6-S17.Again, most of the disorders show marked sex differences in incidence rates across the lifespan, and cumulative incidence by age 80.This is most marked for anorexia nervosa where there is a marked female excess.Of interest, the two specific personality disorders also show marked sex differences: Antisocial personality disorder is more common in males, while Borderline personality disorder is more common in females.As expected, intellectual disabilities, childhood autism and ADHD peaks in early life, and are more common in males compared with females.Analogous figures are shown for eating disorders, specific personality disorders, intellectual disabilities, pervasive developmental disorders, and behavioral and mental disorders with onset usually occurring in childhood and adolescence in Figures S1-S5.Finally, we found that when AOO distributions are based on cases only, versus population-based estimates, the later estimates are lower but more precise than the case-only estimates (Table S1 and S2).Table S3 shows sex-specific cumulative incidences at age 40 and 80 years.

| DISCUSSION
Our results show that about one in three individuals in Denmark will be treated for at least one mental disorder in a psychiatric department by age 80 years.Of those with any mental disorder, half will have their first onset by age 22 years, and three quarters will have their first onset by age 36 years.The peak of the incidence rates for many specific types of mental disorder occurs in childhood or adolescence.
The findings are congruent with the results from the World Mental Health Survey, 1,7 which consistently indicates early ages of onset across a wide spectrum of mental disorders.Additionally, our findings are generally consistent with the first fully epidemiological and largest metaanalysis on age at onset of mental disorders globally, 3 and with our detailed comprehensive nationwide register-based study on the incidence rate and lifetime risk for mental disorders. 2These findings have important implications for service planning, and lend weight to the need for services that are optimized for children and young adults. 21The fact that disabling mental disorders, such as depression, anxiety disorders, and psychosis (a) often emerge in childhood and young people, and (b) are also often persistent or recurrent, contributes to the substantial global burden of disease associated with these disorders, in particular, years lived with disability. 6,22e observed striking differences in the cumulative incidence of mental disorders between males and females.Disorders like depression and anxiety-related disorders-often clustered under the heading "internalizing disorders"-are much more prevalent in females. 1,7I G U R E 1 Incidence rate per 10,000 person-years and cumulative incidence with 95% confidence intervals by age and sex for any mental disorder.Left y-axis curves, with dashed blue lines for males and red for females, are incidence rate curves.These are calculated per year of age per 10,000 person-years.Right y-axis curves, with solid blue lines for males and red for females, are cumulative incidence curves per 100 persons.On these curves, we show the median age of onset as the colored dots.Shaded areas are 95% confidence intervals are shown.
In contrast, neurodevelopmental disorders, such as autism, ADHD, and schizophrenia are more common in males compared with females, as are substance use disorders.Nevertheless, our findings indicate a comparable AOO for schizophrenia among males (26.50 years, IQR 21.68-35.22)and females (24.57years, IQR 20.14-36.82),which contrasts with certain previous studies that report an approximately 5-year earlier onset in males. 3,23,24hile the incidence and absolute risk vary between the sexes across the lifespan, it is noteworthy that (a) the peak incidence rates are broadly comparable in males and females, and (b) the median AOO (the preferred measure of central tendency for the non-Gaussian distributions seen in mental disorder AOOs) often co-occur.This suggests that while men and women have different risks for developing different types of mental disorders, the ages at which they are most at risk are similar.
Strengths of our study include the large and contemporary sample, the use of entire population longitudinal registers, and the coverage of a comprehensive range of mental disorders.Health care is publicly funded in Denmark, thus access to secondary care is based on need rather than on income.Data are collected prospectively, in contrast to survey data, which are vulnerable to recall bias and may miss people admitted to a hospital and imprisoned at the time of the survey.6][27][28][29] We have reported case-only and population-based estimates for cumulative incidence distributions.Finally, case-only estimates exhibit an upward bias (i.e., older medians and means) due to their reliance on the occurrence of the disorder in the future, which is sometimes called conditioning on the future. 19In contrast, the population estimates use F I G U R E 2 Incidence rate per 10,000 person-years and cumulative incidence with 95% confidence intervals by age and sex for mood disorders, neurotic, stress-related and somatoform disorders, mental and behavioral disorders due to psychoactive substance abuse, and schizophrenia and related disorders.Left y-axis curves, with dashed blue lines for males and red for females, are incidence rate curves.These are calculated per year of age per 10,000 person-years.Right y-axis curves, with solid blue lines for males and red for females, are cumulative incidence curves per 100 persons.On these curves, we show the median age of onset as the colored dots.95% confidence intervals are shown (however for many of these figures, they are so precise as to be hidden under the main curves).the entire population distribution of the data including those at risk for the disorders and are thus more precise.As a result, these estimates exhibit higher precision compared with case-only estimates.
As expected, our results are consistent with the findings from our earlier article. 2The current sample overlaps with our previous study, but the added benefit of recent follow-up contributes to increased statistical power.It is important to note that our study is based on psychiatric hospital treated cases, thus people with mental disorders who do not seek help, or who are only treated by their general practitioners will be under-represented in our sample.People who receive treatments in hospital settings may have more severe types of illnesses compared with those detected in community surveys.This may result in a bias in the AOO distributions (rightshifted with older ages of first onset) compared with survey studies.In addition, the cumulative incidences would be downwardly biased, because of those who are not included in the psychiatric health registers.This is especially true for anxiety and depression, which are often treated by general practitioners or in somatic wards. 30owever, it is important to acknowledge that numerous mental disorders, including schizophrenia, 31 bipolar disorder, 32 and obsessive-compulsive disorder, 33 may present with an insidious onset characterized by a prolonged prodromal phase, where symptoms emerge years prior to formal diagnosis.Studies that combine registers and community-based surveys will be better able to quantify these biases. 34Studies based on Danish registers have noted that incidence rates for some mental disorders have increased in Denmark since 1970, with age of diagnosis shifting downwards. 35We selected the relatively brief follow-up period from January 1, 2004, to December 31, 2021, to strike a balance between period effects and birth cohort influences.While some of these secular F I G U R E 3 Incidence rate per 10,000 person-years and cumulative incidence with 95% confidence intervals by age and sex for mental and behavioral disorders due to alcohol use, mental and behavioral disorders due to cannabis use, schizophrenia, schizoaffective disorder, bipolar disorder, single and recurrent depressive disorder, obsessive compulsive disorder, anorexia nervosa, borderline personality disorder, antisocial personality disorder, childhood autism, and ADHD.The left y-axis curves, represented by dashed blue lines for males and red for females, depict incidence rate curves.These rates are calculated per year of age per 10,000 person-years.On the right y-axis, the curves, shown as solid blue lines for males and red for females, are cumulative incidence curves per 100 persons.Colored dots on these curves indicate the median age of onset.The shaded areas are 95% confidence intervals, though, for many of these figures, they are so precise that they are concealed beneath the main curves.changes may be the results of administrative differences, earlier detection and increased reporting of child-onset conditions may also contribute these findings. 36These time trends could result in an overestimating of the cumulative incidence estimates 37 ; we plan to explore these issues in future studies related to age, birth cohort, and period effects related to the AOO of mental disorders.While the key findings of this study may be broadly generalizable to other Nordic nations, this may not be the case for nations without universal health coverage, nor low-and middle-income countries.Furthermore, the availability of services (e.g., early intervention programs from children and young adults) can influence the shape of AOO curves.
In conclusion, our study provides important insights into the impact of mental disorders across the life course.Our new estimates of incidence rates and cumulative incidence of disorders can guide future mental health policy and service development, and inform studies related to genetic epidemiology 38 and the burden of mental disorders. 39 Panel A: Mood disorders.Panel B: Neurotic, stress-related, and somatoform disorders.Panel C: Mental and behavioral disorders due to psychoactive substance abuse.Panel D: Schizophrenia and related disorders.
Panels: Panel A: Mental and behavioral disorders due to alcohol use.Panel b: Mental and behavioral disorders due to cannabis use.Panel C: Schizophrenia.Panel D: Schizoaffective disorder.Panel E: Bipolar disorders.Panel F: Single and recurrent depressive disorder.Panel G: Obsessive-compulsive disorder.Panel H: Anorexia nervosa.Panel I: Borderline personality disorder.Panel J: Antisocial personality disorder.Panel K: Childhood autism.Panel L: ADHD.
T A B L E 2 Sex-specific lifetime risks by age 80 years and 95% confidence intervals, and median age of onset (AOO) and interquartile range for specific and broad types of mental disorders.