Associations between physical diseases and subsequent mental disorders: a longitudinal study in a population‐based cohort

Abstract

People with physical diseases are reported to be at elevated risk of subsequent mental disorders. However, previous studies have considered only a few pairs of conditions, or have reported only relative risks. This study aimed to systematically explore the associations between physical diseases and subsequent mental disorders. It examined a population‐based cohort of 7,673,978 people living in Denmark between 2000 and 2021, and followed them for a total of 119.3 million person‐years. The study assessed nine broad categories of physical diseases (cardiovascular, endocrine, respiratory, gastrointestinal, urogenital, musculoskeletal, hematological and neurological diseases, and cancers), encompassing 31 specific diseases, and the subsequent risk of mental disorder diagnoses, encompassing the ten ICD‐10 groupings (organic, including symptomatic, mental disorders; mental disorders due to psychoactive substance use; schizophrenia and related disorders; mood disorders; neurotic, stress‐related and somatoform disorders; eating disorders; personality disorders; intellectual disabilities; pervasive developmental disorders; and behavioral and emotional disorders with onset usually occurring in childhood and adolescence). Using Poisson regression, the overall and time‐dependent incidence rate ratios (IRRs) for pairs of physical diseases and mental disorders were calculated, adjusting for age, sex and calendar time. Absolute risks were estimated with the Aalen‐Johansen estimator. In total, 646,171 people (8.4%) were identified as having any mental disorder during follow‐up. All physical diseases except cancers were associated with an elevated risk of any mental disorder. For the nine broad pairs of physical diseases and mental disorders, the median point estimate of IRR was 1.51 (range: 0.99‐1.84; interquartile range: 1.29‐1.59). The IRRs ranged from 0.99 (95% CI: 0.98‐1.01) after cancers to 1.84 (95% CI: 1.83‐1.85) after musculoskeletal diseases. Risks varied over time after the diagnosis of physical diseases. The cumulative mental disorder incidence within 15 years after diagnosis of a physical disease varied from 3.73% (95% CI: 3.67‐3.80) for cancers to 10.19% (95% CI: 10.13‐10.25) for respiratory diseases. These data document that most physical diseases are associated with an elevated risk of subsequent mental disorders. Clinicians treating physical diseases should constantly be alert to the possible development of secondary mental disorders.

People with various physical diseases – including cardiovascular, endocrine, respiratory and neurological diseases, as well as cancers – have been reported to be at higher risk of developing mental disorders than those without these diseases ¹ , ² , ³ , ⁴ , ⁵ , ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹² . In people with both physical diseases and mental disorders, a lower quality of life ¹³ , ¹⁴ , ¹⁵ and a shorter life expectancy ¹⁶ have been observed than in those with either physical diseases or mental disorders alone.

Exploring the patterns of association between physical diseases and later mental disorders can allow the identification of groups with elevated risk. However, most studies to date have focused on specific physical diseases or mental disorders. This narrow focus may fail to reveal general associations and etiological links. Moreover, although some studies have considered the lifetime association between groups of mental disorders and physical diseases, only few of them have considered the temporal order of occurrence of the disorders of interest ¹ , ³ , ⁴ , ⁶ , ⁹ , ¹⁰ , ¹⁷ , ¹⁸ .

Cross‐sectional studies often rely on self‐reported collected data and, due to issues with potential recall bias, may underestimate past disorders. Additionally, as participants need to be alive to respond to surveys, it is likely that survey‐based studies will underestimate the number of people with potentially fatal conditions (particularly those of greatest severity). Such “survival bias” could lead to lower estimates of the associations between physical diseases and mental disorders ¹⁹ .

Furthermore, studies have rarely considered risks associated with multiple physical diseases, and they have generally focused on relative risks, not reporting risks in absolute terms, which provide information about the actual number of people with diseases. Thus, further research is needed to assess more comprehensively the association between a wide range of physical diseases and the subsequent development of mental disorders.

The use of Danish register‐based data can address the above issues. The analysis of routinely collected health care data covering the whole population and containing dates of diagnoses offers a better chance of capturing the association between physical diseases and mental disorders in a specified direction ²⁰ , reducing the risk of selection or recall bias.

In an earlier study ²¹ , we documented that mental disorders are associated with an increased subsequent risk of many physical diseases. However, we did not consider the reciprocal associations. While previous evidence ¹ , ³ , ⁴ , ⁵ , ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹² indicates that associations may be present in the opposite direction, only a study considering several physical disease ‐ mental disorder pairs can provide a comprehensive picture.

We used Danish nationwide register data to provide temporally ordered age‐ and sex‐specific pairwise estimates of the risks of a comprehensive range of physical diseases and subsequent mental disorders. We explored variations of associations over time. Additionally, we calculated the age‐ and sex‐specific cumulative incidence of subsequent mental disorders among people with vs. without physical diseases (by using matched comparison cohorts), which can aid interpretation of the clinical implications of the relative risks.

This research was carried out using a well‐characterized list of 31 physical diseases, developed in Denmark, with input from public health epidemiologists and clinicians, and previously used in Danish multimorbidity research ²² . By combining register data on hospital contacts and prescriptions, diagnoses or treatments for these diseases could be ascertained for all individuals. Furthermore, using previously published methods enabling extraction of person‐level health‐related disability from registers ²³ , we calculated a cumulative disability burden score summarizing the disability associated with combinations of diagnosed physical diseases, and considered its relationship to subsequent mental disorders.

METHODS

Study population and ascertainment of disorders

This population‐based cohort study included all 7,673,978 people living in Denmark between January 1, 2000 and December 31, 2021 (i.e., living in Denmark on January 1, 2000, or born in or immigrating to Denmark after that date), as identified in the Danish Civil Registration System.

Information on physical diseases was collected from 1995 onward, using criteria developed for investigating multimorbidity in Danish registers ²² . These criteria included 31 physical diseases, grouped into nine broad categories: cardiovascular, endocrine, respiratory, gastrointestinal, urogenital, musculoskeletal, hematological and neurological diseases, and cancers. The data on physical diseases came from two sources: a) diagnoses made during inpatient admissions and outpatient clinic visits from the Danish National Patient Registry recorded as ICD‐10 codes, and b) redeemed prescriptions for disease‐specific medications (Anatomical Therapeutic Chemical Classification System codes) in the Danish National Prescription Register. The registration date of a physical disease was the date of the first hospital diagnosis or relevant repeat prescription, whichever occurred first (however, for simplicity, we refer here to “diagnosis”).

Information on mental disorders was obtained from the Danish Psychiatric Central Research Register, which includes admissions to psychiatric inpatient facilities since 1969, and visits to outpatient psychiatric and emergency departments since 1995. The diagnosis date was defined as the discharge date for the first contact. We considered any mental disorder, as well as ten types of mental disorders (ICD‐10 subchapter F and corresponding ICD‐8 diagnoses): organic, including symptomatic, mental disorders; mental disorders due to psychoactive substance use; schizophrenia and related disorders; mood disorders; neurotic, stress‐related and somatoform disorders; eating disorders; personality disorders; intellectual disabilities; pervasive developmental disorders; and behavioral and emotional disorders with onset usually occurring in childhood and adolescence.

Statistical analyses

Each analysis was performed separately for any mental disorder and for each mental disorder group. Follow‐up started on January 1, 2000, and terminated at mental disorder diagnosis, death, emigration from Denmark, or December 31, 2021, whichever occurred first. Because we were interested in only incident mental disorders during the follow‐up period, all people with a diagnosis before January 1, 2000 were considered to have prevalent mental disorders and were excluded from the analyses in which the specific mental disorder was the outcome of interest.

We estimated associations between 99 pairs of broad physical diseases and subsequent mental disorders (any mental disorder and ten types of mental disorder), as well as 341 specific physical disease ‐ mental disorder pairs. First, we compared rates of mental disorder diagnosis between people with vs. without physical diseases (both broad categories and specific diseases) through incidence rate ratios (IRRs), estimated using Poisson regression models adjusting for sex, age (in 0.25‐year intervals), and birth year. Additionally, we investigated whether these associations depended on the time after diagnosis of the physical disease (0‐6 months, >6‐12 months, >1‐2 years, >2‐5 years, >5‐10 years, >10‐15 years, or >15 years). Subsequently, we estimated the cumulative incidence of diagnosis of mental disorders after physical disease diagnosis using the Aalen‐Johansen estimator, which accounts for competing risks (of dying or emigrating).

To compare the cumulative incidence of mental disorders among people with vs. without physical diseases, we generated matched comparison cohorts. For every index person with the physical disease, up to five age‐ and sex‐matched individuals (without a diagnosis of the relevant physical disease at that point) were randomly selected from the entire population. Cumulative incidence proportions were stratified by sex and age (at diagnosis of the physical disease) groups (<35, 35‐<70 and ≥70 years). They provide a measure of risk in absolute terms that can be interpreted as the percentage of individuals among those with vs. without a particular physical disease who develop the mental disorder after a specified time.

To assess the effects of complex (multi)morbidity patterns, we calculated the IRR of any mental disorder diagnosis (and of each specific mental disorder diagnosis) as a function of the “disability burden score”, i.e. a summary score accounting for health‐related disability associated with each type of physical disease. Specifically, we used the disability weights developed by the Global Burden of Disease Study to quantify the health loss associated with combinations of diseases ²⁴ and adapted in our previous work for the physical diseases considered in the present study ²³ . This approach allowed us to determine scores for our specific list of 31 physical diseases ²⁵ . Disability weights represent the severity of health loss associated with a given disorder. The weights are measured on a scale of 0 to 1, where 0 indicates full health, and 1 a health state equivalent to death.

For each individual, we calculated a time‐varying disability burden score based on the observed diagnoses and disability weights. Calculation of the disability burden score was carried out using a specific formula (see supplementary information). Individuals with no physical diseases were assigned a disability burden score of 0. As an example, an individual with a physical disease linked to a disability weight of 0.3 was assigned a total disability burden score of 0.3, while an individual with two physical diseases, with respective disability weights of 0.3 and 0.5, had a total disability burden score of 1‐(1‐0.3)×(1‐0.5) = 0.65.

To simplify the presentation of results, we focus here on the associations between any mental disorder and the nine broad physical disease categories. Results for all included pairs (i.e., all mental disorder groupings and the 31 specific physical diseases) are provided in the supplementary information and in an interactive website (see Acknowledgements).

The study was registered with the Danish Data Protection Agency via Aarhus University (no. 2016‐051‐000001‐2587) and was approved by Statistics Denmark and the Danish Health Data Authority. According to the Danish law, informed consent or ethical approval is not required for register‐based studies conducted in the country. All data were pseudonymized and not recognizable at an individual level, and were analyzed on the secured platform of Statistics Denmark via remote access. A pre‐specified analysis plan was uploaded on the Open Science Framework before the analyses started ²⁶ .

RESULTS

The cohort consisted of 7,673,978 Danish residents (50.0% female), followed up for a total of 119.3 million person‐years. During the follow‐up, 1.19 million people died, and 0.92 million emigrated. The baseline cohort characteristics are presented in Table 1.

Table 1.

Baseline characteristics of the study population

Sex, N (%)
Men	3,838,695 (50.0)
Women	3,835,283 (50.0)
Birth year, N (%)
<1900	484 (<0.1)
1900‐1909	30,102 (0.4)
1910‐1919	178,562 (2.3)
1920‐1929	364,596 (4.8)
1930‐1939	483,548 (6.3)
1940‐1949	749,288 (9.8)
1950‐1959	768,167 (10.0)
1960‐1969	887,277 (11.6)
1970‐1979	879,585 (11.5)
1980‐1989	891,834 (11.6)
1990‐1999	940,511 (12.3)
2000‐2009	732,669 (9.5)
2010‐2021	767,355 (10.0)
Age at start of follow‐up, years, median (IQR)	28 (7‐49)
Age at end of follow‐up, years, median (IQR)	46 (24‐68)

IQR – interquartile range

The number of prevalent cases (between 1969 and 1999), the population defined as being at risk at the start of follow‐up, and the number of incident cases during follow‐up (between 2000 and 2021) for any mental disorder and for the individual groupings of mental disorders are shown in Table 2. In total, 646,171 people (8.4%) were identified as having any mental disorder during follow‐up (302,506 males and 343,665 females).

Table 2.

Frequencies of prevalent cases, persons at risk, and incident cases for each mental disorder, in the total cohort

	Prevalent cases before follow‐up	Persons at risk at start of follow‐up	New cases during follow‐up
Any mental disorder
Total	270,730	7,328,530	646,171
Males	119,132	3,681,297	302,506
Females	151,598	3,647,233	343,665
Organic disorders
Total	24,728	4,611,275	103,856
Males	10,035	2,274,720	43,183
Females	14,693	2,336,555	60,673
Substance use disorders
Total	71,206	6,898,341	95,864
Males	43,927	3,433,032	62,534
Females	27,279	3,465,309	33,330
Schizophrenia and related disorders
Total	50,590	6,918,957	67,421
Males	23,824	3,453,135	35,127
Females	26,766	3,465,822	32,294
Mood disorders
Total	79,161	6,890,386	222,564
Males	27,095	3,449,864	86,209
Females	52,066	3,440,522	136,355
Neurotic, stress‐related and somatoform disorders
Total	98,167	7,207,158	321,228
Males	34,906	3,614,529	133,145
Females	63,261	3,592,629	188,083
Eating disorders
Total	5,329	7,593,931	27,266
Males	280	3,800,149	1,836
Females	5,049	3,793,782	25,430
Personality disorders
Total	76,433	6,893,114	75,588
Males	30,834	3,446,125	24,516
Females	45,599	3,446,989	51,072
Intellectual disabilities
Total	5,890	7,593,370	24,083
Males	3,290	3,797,139	14,714
Females	2,600	3,796,231	9,369
Developmental disorders
Total	3,470	7,595,790	54,913
Males	2,723	3,797,706	38,033
Females	747	3,798,084	16,880
Behavioral disorders
Total	14,474	7,584,786	116,163
Males	9,620	3,790,809	72,104
Females	4,854	3,793,977	44,059

“Behavioral disorders” is an abbreviation for “behavioral and emotional disorders with onset usually occurring in childhood and adolescence”

The risks of any subsequent mental disorder diagnosis in people with vs. without each physical disease are shown in Table 3 and Figure 1. For the nine broad pairs of physical diseases and mental disorders, the median point estimate of IRR was 1.51 (range: 0.99‐1.84; interquartile range, IQR: 1.29‐1.59). The IRRs were 1.58 (95% CI: 1.57‐1.60) for cardiovascular diseases; 1.29 (95% CI: 1.28‐1.31) for endocrine diseases; 1.30 (95% CI: 1.29‐1.30) for respiratory diseases; 1.59 (95% CI: 1.57‐1.61) for gastrointestinal diseases; 1.27 (95% CI: 1.25‐1.29) for urogenital diseases; 1.84 (95% CI: 1.83‐1.85) for musculoskeletal diseases; 1.68 (95% CI: 1.65‐1.71) for hematological diseases; 0.99 (95% CI: 0.98‐1.01) for cancers; and 1.51 (95% CI: 1.50‐1.52) for neurological diseases.

Table 3.

Incidence rate ratios (IRRs) for any subsequent mental disorder in people with vs. without a physical disease diagnosis

	IRR (95% CI)
	Total	Males	Females
Cardiovascular diseases	1.58 (1.57‐1.60)	1.70 (1.68‐1.73)	1.49 (1.47‐1.50)
Endocrine diseases	1.29 (1.28‐1.31)	1.39 (1.37‐1.41)	1.24 (1.22‐1.25)
Respiratory diseases	1.30 (1.29‐1.30)	1.27 (1.26‐1.28)	1.30 (1.30‐1.31)
Gastrointestinal diseases	1.59 (1.57‐1.61)	1.75 (1.72‐1.78)	1.47 (1.45‐1.49)
Urogenital diseases	1.27 (1.25‐1.29)	1.35 (1.32‐1.37)	1.27 (1.21‐1.33)
Musculoskeletal diseases	1.84 (1.83‐1.85)	1.90 (1.88‐1.92)	1.78 (1.76‐1.80)
Hematological diseases	1.68 (1.65‐1.71)	1.76 (1.71‐1.81)	1.62 (1.59‐1.66)
Cancers	0.99 (0.98‐1.01)	1.03 (1.00‐1.05)	0.98 (0.96‐1.00)
Neurological diseases	1.51 (1.50‐1.52)	1.64 (1.62‐1.66)	1.42 (1.41‐1.43)

Figure 1.

Incidence rate ratios for a diagnosis of any mental disorder after the diagnosis of a physical disease

The IRRs were higher for males than females for almost all broad physical diseases (see Table 3). The IRRs for broad physical diseases and specific types of mental disorders are shown in Figure 2 (see also supplementary information).

Figure 2.

Incidence rate ratios for each mental disorder diagnosis after a diagnosis within each broad physical disease category. “Neurotic disorders” is an abbreviation for “neurotic, stress‐related and somatoform disorders”. “Behavioral disorders” is an abbreviation for “behavioral and emotional disorders with onset usually occurring in childhood and adolescence”.

Time‐dependent IRRs for receiving a diagnosis of any mental disorder, according to the time after the first diagnosis of each broad physical disease, are shown in Figure 3 (see supplementary information for all IRR values). For all nine broad physical diseases, the rate of diagnosis of any mental disorder was highest in the initial 0‐6 months after a physical disease diagnosis and subsequently decreased to varying extents (e.g., relatively steep decreases were observed for cardiovascular and hematological diseases at 6‐12 months).

Figure 3.

Incidence rate ratios for any mental disorder diagnosis after a diagnosis within a broad physical disease category, according to the timing of the physical disease diagnosis

As shown in Figure 4, the IRR of any mental disorder diagnosis generally increased with increasing disability burden score (see supplementary information for estimates concerning the various mental disorders).

Figure 4.

Incidence rate ratios for any mental disorder diagnosis as a function of the disability burden score. The gray shading indicates the 95% CIs.

The cumulative incidence of any mental disorder for people previously diagnosed with a physical disease, and for people within comparison cohorts, is presented in Table 4 and Figure 5. Within 15 years after the physical disease diagnosis, the highest cumulative incidence of any mental disorder was observed for the broad group of respiratory diseases (10.19%, 95% CI: 10.13‐10.25). For the corresponding matched comparison cohort (i.e., people without a respiratory disease diagnosed at the time of matching), the cumulative incidence within 15 years was 7.68% (95% CI: 7.66‐7.71).

Table 4.

Cumulative incidence (after 15 years) of any mental disorder in people with vs. without a prior physical disease diagnosis

	Cumulative incidence, % (95% CI)
	Physical disease cohort	Comparison cohort
Cardiovascular diseases	6.73 (6.68‐6.78)	4.46 (4.44‐4.48)
Endocrine diseases	7.18 (7.10‐7.26)	5.59 (5.56‐5.62)
Respiratory diseases	10.19 (10.13‐10.25)	7.68 (7.66‐7.71)
Gastrointestinal diseases	8.31 (8.20‐8.42)	5.92 (5.88‐5.96)
Urogenital diseases	6.14 (6.04‐6.25)	5.27 (5.23‐5.32)
Musculoskeletal diseases	8.23 (8.18‐8.28)	4.46 (4.44‐4.48)
Hematological diseases	6.58 (6.47‐6.70)	6.52 (6.47‐6.57)
Cancers	3.73 (3.67‐3.80)	5.72 (5.69‐5.76)
Neurological diseases	8.92 (8.85‐8.98)	5.94 (5.92‐5.97)

Figure 5.

Cumulative incidence of any mental disorder diagnosis after the diagnosis of a broad physical disease category. The bold lines indicate the incidence in people diagnosed with a physical disease; the dashed lines indicate the incidence in matched comparison cohorts.

For almost all broad groups of physical diseases, the cumulative incidence of any mental disorder within 15 years was higher in people with the prior diagnosis of interest than in the respective matched comparison cohorts, with two exceptions. First, the incidence of any mental disorder was initially higher in people with a hematological disease than in the matched comparison group (IRR=1.68, 95% CI: 1.65‐1.71); however, similar levels were seen in the two groups within 15 years after diagnosis (6.58%, 95% CI: 6.47‐6.70 vs. 6.52%, 95% CI: 6.47‐6.57). Second, the incidence of any mental disorder was initially similar in people with vs. without a cancer diagnosis (IRR=0.99, 95% CI: 0.98‐1.01), but became higher in the comparison cohort starting approximately 2 years after diagnosis (2.33% vs. 1.87% at 5 years; 4.26% vs. 2.94% at 10 years; 5.72% vs. 3.73% at 15 years) (see supplementary information for estimates concerning the various mental disorders, also by sex and age).

DISCUSSION

This population‐based study, comprising 7.7 million people, provides a detailed picture of the emergence of mental disorders in people with a broad range of prior physical diseases. For most pairs, the risk of subsequent mental disorders was higher among people with vs. without the physical disease in question. However, for some pairs, the opposite association was observed (e.g., the risk for any mental disorder and for several specific mental disorders was lower among people with vs. without a prior diagnosis of cancer).

Different pairs had different temporal patterns. For example, the risk of any mental disorder was much higher in the 0‐6 months after hospital contact for a cardiovascular disease and decreased as time went on. In contrast, for any mental disorder after respiratory diseases, the association remained somewhat flat over time.

Within 15 years after most physical diseases, the proportion of people subsequently diagnosed with mental disorders was higher than among people in the comparison cohorts. However, this was not the case for any mental disorder following cancer, which was higher in the comparison cohort than among people with a prior cancer diagnosis.

Finally, we observed that the mental disorder risk was generally higher with greater disability burden from physical diseases, in line with previous findings suggesting that the overall burden of physical morbidity is associated with poor mental health with a dose‐response pattern ²⁷ , ²⁸ , ²⁹ .

Our findings for cancer were a notable exception from the general pattern of increased risk of mental disorders following physical diseases. In fact, the occurrence of several mental disorders was lower among people with vs. without a prior diagnosis of cancer. We did, however, find an increased rate of mood and neurotic disorders following cancer (IRR=1.11, 95% CI: 1.09‐1.14; and IRR=1.19, 95% CI: 1.17‐1.22, respectively). The observed reduced risk for certain mental disorders after cancer is likely due to the competing risk of death, as premature mortality is particularly pronounced for cancer ¹⁶ .

Several mechanisms may underlie the associations observed in our study. For example, illness‐related stress, pain, restriction and disability due to physical diseases ³⁰ , ³¹ , ³² , ³³ have been demonstrated to increase the risk of onset of mental disorders. Moreover, shared environmental and/or genetic risk factors may also underlie the observed findings ³⁴ . On the other hand, detection bias (hospital contacts for physical diseases leading to referral to psychiatric services) may also be at work. In some cases, specific etiological mechanisms may link pairs of physical diseases and subsequent mental disorders. However, the increase in the risk of virtually all mental disorders after diagnosis of almost any physical disease observed herein suggests that more general mechanisms (such as those mentioned above) are likely to be involved, and should be investigated further.

This register‐based study comprised a large population, and had no susceptibility to recall or self‐reporting bias. Data were available for the entire population, thereby minimizing selection bias. Because Danish citizens have free and equal access to health care ³⁵ , any effects associated with the ability to afford private insurance/access to health care are likely negligible. In Denmark, all hospitals must report discharge diagnoses, and all pharmacies must report redemptions of drug prescriptions to the central registers providing data for this study. We considered a broad range of disorders, and also assessed cumulative disability due to physical diseases as a risk factor for subsequent mental disorders.

However, this study has some limitations. First, physical diseases and mental disorders might have been incompletely ascertained. Although register‐based psychiatric diagnoses in Denmark have been found to be generally valid for a range of disorders ³⁶ , ³⁷ , ³⁸ , ³⁹ , ⁴⁰ , the validity varies to some extent across diagnoses. Given the lack of data on people who did not seek treatment, and on diagnoses assigned by general practitioners and other private practitioners, there may have been some underdetection. For certain physical diseases, this limitation was partially offset by the use of prescription data, in combination with hospital diagnoses; however, prescriptions were not used to identify mental disorders, because of the lack of specificity of many psychotropic medications ⁴¹ . Thus, cases of both mental and physical diseases may have been biased toward relatively high severity.

Since we were interested in only incident mental disorders during follow‐up (starting on January 1, 2000), a “wash‐out” period in the preceding years (1969‐1999) excluded prevalent mental disorder cases from their respective analysis. However, our procedures might have not identified all prevalent mental disorders. Furthermore, details on the date of onset of illness are unavailable in the registers, and the dates of diagnosis or prescription were used as proxies. In some cases, incorrect temporal ordering of conditions might have occurred.

Studies based on health care registers are prone to surveillance and diagnostic bias; i.e., people who are in contact with the health care system and diagnosed with a physical disease might be more likely to be diagnosed with additional conditions ⁴² . The time‐dependent analyses helped us assess the extent of this bias.

Some of the physical disease categories studied were very broad and may warrant more focused/stratified investigation. Other prior diseases/events (e.g., highly acute diseases, surgeries, infections and accidents) may also be of interest, but were outside the scope of this study. Finally, our findings may not be generalizable outside of Denmark, as patterns of morbidity and comorbidity vary across countries and may be different in those with health care systems and socioeconomic structures that differ from Denmark.

Our analyses considered many physical disease ‐ mental disorder pairs, which were likely to be linked by different pathways, as discussed above. We hope that our findings at the more general level will, in turn, support specific hypothesis‐driven research into some of these pairs.

In conclusion, our data document that most physical diseases are associated with an elevated risk of subsequent mental disorders. This risk is pervasive in both relative and absolute terms. Clinicians treating physical diseases should constantly be alert to the possible development of secondary mental disorders.