Association of Income With the Incidence Rates of First Psychiatric Hospital Admissions in Finland, 1996-2014

This cohort study compares income-specific trends in the incidence rates of first psychiatric hospital admissions and evaluates whether an income gradient exists in the incidence rates at all levels of household income among adults in Finland.

Key Points

Question

Is household income associated with the incidence rates of first hospital admissions for mental disorders?

Findings

In this nationwide open cohort study of more than 6.2 million persons, a clear income gradient was observed at all levels of income among adults, with adjusted incidence rate ratios varying from 2.94 to 4.46 in the lowest compared with the highest income deciles. This association varied over time, and a continuous decrease in the annual incidence rates emerged only in the high-income groups.

Meaning

Household income appears to be an important risk factor for first hospital-treated mental disorders at all levels of income, and mechanisms linking income and mental health may be located partly within the health care system itself.

Abstract

Importance

The association between income and mental health has long been a question of interest. Nationwide register data provide means to examine trends and patterns of these associations.

Objectives

To compare income-specific trends in the incidence rates of first psychiatric hospital admissions and to evaluate whether an income gradient exists in the incidence rates at all levels of household income.

Design, Setting, and Participants

This population-based open cohort study used linked registry data from nationwide Finnish Hospital Discharge and Statistics Finland population registers to determine annual incidence rates of first psychiatric hospital admissions. All Finnish citizens (N = 6 258 033) living in the country at any time from January 1, 1996, through December 31, 2014, contributed to 96 184 614 person-years at risk of first inpatient treatment for mental disorders. The analyses were conducted from August 1, 2018, through September 30, 2019.

Exposures

Equivalized disposable income, sex, age group, reduction in income decile in the previous 3 years, urbanicity, educational level, and living alone status.

Main Outcomes and Measures

Annual percentage changes in the age-standardized incidence rates and incidence rate ratios (IRRs).

Results

Altogether, 186 082 first psychiatric inpatient treatment episodes occurred (93 431 [50.2%] men), with overall age-standardized incidence rates per 1000 person-years varying from 1.59 (95% CI, 1.56-1.63) in 2014 to 2.11 (95% CI, 2.07-2.15) in 2008. In the highest income deciles, a continuous mean decrease per year of 3.71% (95% CI, 2.82%-4.59%) in men and 0.91% (95% CI, 0.01%-1.80%) in women occurred throughout the study period, in contrast to the lowest deciles, where the trends first increased (1.31% [95% CI, 0.62%-2.01%] increase in men from 1996 to 2007 and 5.61% [95% CI, 2.36%-8.96%] increase in women from 1996 to 2001). In the adult population, an income gradient was observed at all levels of household income: the lower the income decile, the higher the adjusted IRRs compared with the highest decile. The IRRs in the lowest decile varied from 2.94 (95% CI, 2.78-3.11) to 4.46 (95% CI, 4.17-4.76). In other age groups, the gradient did not persist at the highest income deciles. Diagnosis-specific income gradient was steepest in schizophrenia and related psychotic disorders, with estimated IRRs of the lowest income decile of 5.89 (95% CI, 5.77-6.02).

Conclusions and Relevance

In this cohort study, clear negative income gradient in the incidence rates of first hospital-treated mental disorders was observed in the adult population of Finland. These findings suggest that reduction in the use of inpatient care has not taken place equally between different income groups.

Introduction

The negative association between household income and different mental health outcomes has been observed repeatedly,^{1,2,3,4,5,6,7,8} but not always.^9,10,11,12 Significant differences in health and mortality are found between all socioeconomic levels of society, which is often referred to as the social gradient in health.^13,14 This pattern has its roots in childhood^15,16 and has been observed in some studies of common mental disorders.^8,17 However, some studies^18,19,20 argue that, after basic needs are met, additional income is not associated with increases in well-being. At present, whether there is an income gradient in the incidence rates of first hospital-treated mental disorders is unknown.

Explanations for the association between income and mental health fall into 2 categories.^21,22,23,24 According to social causation, income influences mental health through psychosocial pathways. The selection hypothesis, on the other hand, posits that mental disorders may cause downward social mobility within and across generations. These mechanisms are not mutually exclusive, but they vary between disorders and contexts and during the life course.^22,25,26,27 In addition to academic and ethical interest, these associations have practical implications for social policy in general and for the design of accessible health services.^28,29,30

Psychiatric hospital treatment has been reduced in the era of deinstitutionalization.^31,32 In Finland, the annual population rates of psychiatric hospital care have decreased from 6.2 to 4.7 per 1000 inhabitants, and mean length of stay in inpatient care has reduced from 67 to 31 days from 1996 to 2014.³³ First psychiatric hospital admission rates are rarely reported^34,35 but present an overall perspective on the annual emergence of severe mental health conditions and can thus facilitate comparisons between income groups. Hence, whether the incidence of first-time hospital-treated mental disorders has been reduced in the era of outpatient-centered services and, if so, whether the reduction has occurred equally within population are unknown to date.

Using national individual-level register data, we investigated the associations between household income and the incidence rates of first hospital admissions for mental disorders in Finland from January 1, 1996, through December 31, 2014. We hypothesized that an income gradient exists in the incidence rate of first inpatient treatments for mental disorders on the national level and examined (1) how equally the income-specific trends in the incidence rates of first psychiatric hospital admissions have changed during the era of decreasing inpatient care and (2) whether the possible income-specific differences persist after adjusting for potential well-known confounding factors, including urbanicity of the living municipality,³⁶ educational attainment,¹⁰ living alone,³⁷ and income reduction during the previous 3 years before the first admission.^5,38

Methods

Data Sources and Study Population

Individual-level register data on hospital care and population registers were combined for this population-based open cohort study. We identified all first psychiatric hospital admissions and the dynamic population at risk of first admissions in Finland from 1996 through 2014. The ethical review board of the National Institute for Health and Welfare approved the study protocol. Informed consent is not required for register-based studies in Finland. We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

The population register of Statistics Finland included data on the total population on the last day of each study year, with socioeconomic variables and time of birth and death. We collected the following individual-level data: sex (man or woman), nationality (Finnish citizen or not), urbanicity of residence municipality (urban, semiurban, or rural),³⁹ household net income, size of the household dwelling unit, and for persons 20 years or older, the level of educational attainment (less than upper secondary, upper secondary, or tertiary, a national classification based on the United Nations Educational, Scientific and Cultural Organization’s International Standard Classification of Education 2011),⁴⁰ and living alone status (living alone or not).

The Finnish Hospital Discharge register (1969-1993), followed by the Care Register for Health Care, maintained by the National Institute for Health and Welfare, covers all inpatient hospital treatments in Finland, and displays good accuracy of mental health diagnoses.⁴¹ We collected all admission and discharge dates and discharge diagnoses.

Assessment of the First Psychiatric Hospital Admissions and Diagnoses

We identified all persons with first-time psychiatric hospital inpatient admissions. Outpatient visits, day-hospital treatment in psychiatric hospitals, or treatments in other general hospital wards with psychiatric diagnoses were not included. Treatments in psychiatric facilities were reliably recognized starting from the year 1976. Hence, the shortest definitive clearance period to define a first admission (ie, the time with no previous inpatient treatments before the first admission) was 20 years (January 1, 1976, through December 31, 1995). To cover the whole study period and facilitate the evaluation of temporal trends, we used a 20-year clearance period for every study year. The International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) has been used in Finland since 1996. For details, see the eMethods in the Supplement.

Assessment of Equivalized Disposable Income Deciles

We calculated the equivalized disposable income by adjusting the net income of a household dwelling unit for the size of the unit, using the Organisation for Economic Co-operation and Development–modified equivalence scale.⁴² Net income is obtained after subtracting taxes from income subject to state taxation.⁴³ We calculated 10th percentiles (deciles) of the annual Finnish income distribution to categorize the population into 10 income groups. The income deciles 1, 3, and 5 years before the first psychiatric hospital admission were also calculated, and possible decrease in income decile in the previous 3 years was recorded.

Persons who are registered as institutionalized or who are homeless, abroad, registered as unknown, or living in quarters that do not meet the definition of dwelling do not constitute household-dwelling units and are categorized as nondwelling. Therefore, no income data are available for this nondwelling population. However, instead of leaving these persons out of the analyses, we analyzed them as a separate income group.

Statistical Analysis

We computed stratum-specific incidence rates of the first psychiatric hospital admissions for every calendar year by dividing the number of first admissions by the person-years at risk in the following strata: sex, 5-year age groups, nationality, income decile, decrease in income decile, urbanicity of the residential municipality, educational level, and living alone status. Age-standardized incidence rates with 95% CIs were calculated by applying direct age standardization to the 2013 Revision of the European Standard Population.⁴⁴ Analyses were conducted separately for the main ICD-10 categories of psychiatric diagnoses.

Every person in the population register with no previous hospital admissions within the clearance period of 20 years contributed to the person-time at risk. Exact dates of immigration to and from the country, moves between municipalities, or changes in household composition were not available, and therefore changes were assumed to occur on average in midyear. Non-Finnish citizens had a high rate of missing data, for example, 13.2% of the person-years at risk had missing income data compared with 0.9% in the case of Finnish citizens. Hence, all analyses included Finnish citizens only.

We used a joinpoint regression model to analyze changes in trends in age-standardized incidence rates.⁴⁵ Sex-, income decile–, and diagnosis-specific trends were analyzed separately. We used a model with a maximum of 3 joinpoints requiring at least 2 observations between joinpoints, a log-linear regression model, and the bayesian information criterion method to assess significant changes in time trends. Annual percentage changes (APCs), the estimated annual changes in rates from one joinpoint to the next in percentage, and weighted means of combined APCs were calculated. The 2-sided α level was set at .05.

To account for potential confounders, we used multivariable Poisson regression models to examine income decile-specific incidence rate ratios (IRRs) and corresponding 95% CIs. We used the incidence rates in the highest income decile as a reference. Analysis was conducted separately in 5 periods (1996-1999, 2000-2003, 2004-2007, 2008-2011, and 2012-2014) and in 3 age groups (5-19, 20-64, and ≥65 years). The division into 5 periods was used to summarize data and to make it easier to compare indicators. We adjusted the models for sex, age group, urbanicity of residential municipality, decrease in income decile, and, in the groups aged 20 to 64 years and 65 years or older, educational attainment and living alone status. We conducted separate analyses for all first admissions and for the main ICD-10 categories of psychiatric diagnoses. We replicated the analysis using the income decile 1, 3, and 5 years before the first admission, instead of the current income decile. This procedure accounted more strongly for the temporal order of having a certain level of income and the first hospital admission.

For data management and analyses, we used the following: Python, version 2.7 (Python Software Foundation); R, version 3.5.1 (R Project for Statistical Computing); Stata, version 15.1 (StataCorp LLC); and Joinpoint Regression Program, version 4.6.0.0 (Statistical Methodology and Applications Branch, Surveillance Research Program, National Cancer Institute). The analyses were conducted from August 1, 2018, through September 30, 2019.

Results

First Admissions and the First Admission Rates

A total of 6 258 033 Finnish citizens contributed 96 184 614 person-years at risk of first inpatient treatment for mental disorders during the study period. Of those at risk, 186 082 persons (93 431 men [50.2%] and 92 651 women [49.8%]) had their first admission to psychiatric inpatient care. The most commonly presented diagnostic main group consisted of mood disorders (ICD-10 codes F30-F39) in 80 548 cases (43.3%) (eTable 1 in the Supplement).

Table 1 contains annual incidence rates of first psychiatric hospital admissions in the first and last year of the study period. All years are presented in eTable 2 in the Supplement. The incidence rate per 1000 person-years varied from 1.59 (95% CI, 1.56-1.63) in 2014 to 2.11 (95% CI, 2.07-2.15) in 2008 (Figure 1A). Men had higher incidence rates in the beginning of the study period (2.15 [95% CI, 2.09-2.22]) compared with women (1.61 [95% CI, 1.56-1.66]), whereas women had higher rates in at the end (1.65 [95% CI, 1.60-1.70] vs 1.54 [95% CI, 1.49-1.59]) (Figure 1B and Table 1). Much of the variation in the overall trends occurred in the incidence rates of substance use and mood disorders, with largest percentage increase of 4.53% (95% CI, 2.27%-6.84%) in mood disorders from 1996 through 1999 and largest percentage decrease of 8.8% (95% CI, 9.69%-7.91%) in substance use disorders from 2008 through 2014 (Figure 1C and D and eFigures 1-3 and eTables 3-6 in the Supplement).

Table 1. Age-Standardized Incidence Rates of First Psychiatric Hospital Admissions in the First and Last Year of the Study Period^a.

Variable	1996			2014
	Admissions, No. (%)	Person-Years at Risk, Millions (%)	Incidence Rate (95% CI) per 1000 Person-Years	Admissions, No. (%)	Person-Years at Risk, Millions (%)	Incidence Rate (95% CI) per 1000 Person-Years
All populationb	9713	5.12	1.9 (1.86-1.94)	8387	5.31	1.59 (1.56-1.62)
Finnish citizens	9547 (100)	5.05	1.89 (1.85-1.93)	8056 (100)	5.10	1.59 (1.56-1.63)
Sex
Men	5423 (56.8)	2.46 (48.7)	2.15 (2.09-2.22)	3832 (47.6)	2.50 (49.0)	1.54 (1.49-1.59)
Women	4124 (43.2)	2.59 (51.3)	1.61 (1.56-1.66)	4224 (52.4)	2.60 (51.0)	1.65 (1.60-1.70)
Income decile and nondwelling population
1 (lowest)	1870 (19.6)	0.47 (9.3)	3.92 (3.72-4.14)	1804 (22.4)	0.43 (8.5)	4.02 (3.79-4.26)
2	1206 (12.6)	0.49 (9.7)	2.71 (2.55-2.89)	1116 (13.9)	0.48 (9.4)	2.62 (2.45-2.79)
3	1047 (11.0)	0.49 (9.8)	2.31 (2.16-2.47)	909 (11.3)	0.49 (9.7)	1.95 (1.82-2.10)
4	933 (9.8)	0.49 (9.8)	2.06 (1.92-2.21)	776 (9.6)	0.50 (9.8)	1.63 (1.51-1.75)
5	816 (8.5)	0.50 (9.8)	1.75 (1.62-1.90)	662 (8.2)	0.50 (9.9)	1.32 (1.22-1.43)
6	747 (7.8)	0.50 (9.8)	1.63 (1.49-1.81)	579 (7.2)	0.51 (9.9)	1.15 (1.05-1.25)
7	674 (7.1)	0.50 (9.9)	1.38 (1.25-1.57)	564 (7.0)	0.51 (10.0)	1.11 (1.02-1.22)
8	668 (7.0)	0.50 (9.9)	1.36 (1.21-1.58)	482 (6.0)	0.51 (10.0)	0.97 (0.88-1.09)
9	639 (6.7)	0.50 (9.9)	1.32 (1.18-1.53)	382 (4.7)	0.52 (10.1)	0.79 (0.70-0.91)
10 (highest)	597 (6.3)	0.50 (9.9)	1.22 (1.09-1.41)	338 (4.2)	0.52 (10.1)	0.73 (0.64-0.84)
Nondwelling	256 (2.7)	0.07 (1.3)	4.37 (3.81-5.03)	323 (4.0)	0.09 (1.7)	4.59 (4.08-5.17)
Data missing	94 (1.0)	0.05 (1.0)	5.26 (4.05-7.36)	121 (1.5)	0.04 (0.8)	6.22 (4.96-9.58)
Income decreased in previous 3 y
No	6034 (63.2)	3.26 (64.5)	1.78 (1.74-1.83)	5313 (66.0)	3.46 (67.7)	1.51 (1.47-1.56)
Yes	3309 (34.7)	1.54 (30.4)	2.07 (1.99-2.16)	2534 (31.5)	1.41 (27.7)	1.72 (1.65-1.79)
Data missing	204 (2.1)	0.26 (5.1)	5.21 (4.36-6.65)	209 (2.6)	0.24 (4.6)	4.93 (4.20-5.95)
Urbanicity of residence municipality
Urban	6687 (70.0)	3.26 (64.6)	2.05 (1.99-2.10)	5727 (71.1)	3.48 (68.3)	1.63 (1.59-1.67)
Semiurban	1460 (15.3)	0.87 (17.3)	1.70 (1.61-1.79)	1249 (15.5)	0.85 (16.6)	1.55 (1.46-1.64)
Rural	1400 (14.7)	0.91 (18.1)	1.56 (1.48-1.64)	1080 (13.4)	0.77 (15.1)	1.49 (1.40-1.59)
Educational levelc
Lower	3795 (46.5)	1.54 (40.8)	3.06 (2.95-3.17)	1974 (32.1)	0.95 (24.1)	3.10 (2.94-3.27)
Secondary	3048 (37.4)	1.35 (35.8)	2.10 (2.00-2.21)	2913 (47.4)	1.70 (43.0)	1.62 (1.56-1.69)
Tertiary	1316 (16.1)	0.88 (23.4)	1.53 (1.42-1.65)	1259 (20.5)	1.30 (32.9)	0.98 (0.92-1.05)
Living alonec
No	5105 (62.6)	2.88 (76.2)	1.73 (1.68-1.78)	3424 (55.7)	2.86 (72.2)	1.22 (1.18-1.26)
Yes	3054 (37.4)	0.90 (23.8)	3.59 (3.46-3.73)	2722 (44.3)	1.10 (27.8)	2.62 (2.51-2.73)

^aStandardized to the 2013 European Standard Population by 5-year age groups.

^bContains data for all Finland; all other rows contain Finnish citizens only.

^cIncludes those 20 years or older.

Figure 1. Trends in Age-Standardized Incidence Rates of First Psychiatric Hospital Admissions.

Data markers denote the annual age-standardized incidence rates per 1000 person-years; error bars, 95% CIs. Lines represent results of the joinpoint regression model. Rates have been age standardized to the 2013 European Standard Population by 5-year age groups. ICD-10 indicates International Statistical Classification of Diseases and Related Health Problems, Tenth Revision; IR, incident rate. Codes F10 to F19 indicate mental and behavioral disorders due to psychoactive substance use; F20 to F29, schizophrenia, schizotypal, and delusional disorders; F30 to F39, mood disorders; and F40 to F48, neurotic, stress-related, and somatoform disorders.

Income-Specific Temporal Trends

In the income-specific joinpoint regression analysis, opposite trends in the age-standardized rates between different income deciles were observed in the beginning of the study period (Figure 2, Table 2, and eFigure 4 in the Supplement for the nondwelling population). Increasing trends (ie, positive APC with 95% CIs not including 0) occurred in the 7 lowest income deciles in women and in the 3 lowest deciles in men. In the highest income deciles, a continuous mean decrease of 3.71% (95% CI, 2.82%-4.59%) per year in men and 0.91% (95% CI, 0.01%-1.80%) per year in women occurred throughout the study period (Table 2). Statistically significant increasing trends did not occur in any of the 4 most common main ICD-10 categories of psychiatric diagnoses in the highest income decile in men or women, but occurred in all diagnostic categories in the lowest income decile, except in men with neurotic, stress-related and somatoform disorders (eFigures 5-8 in the Supplement). Age-specific annual incidence rates per 1000 person-years changed during the study period in girls and women aged 15 to 19 from 3.32 (95% CI, 2.62-4.15) to 7.83 (95% CI, 6.67-9.14) in the lowest income decile and 1.98 (95% CI, 1.22-3.02) to 2.74 (95% CI, 1.80-3.98) in the highest income deciles, whereas in in boys and men aged 15 to 19 years, they changed from 6.48 (95% CI, 5.39-7.71) to 5.81 (95% CI, 4.74-7.04) in the lowest income decile and 3.04 (95% CI, 2.13-4.21) to 1.42 (95% CI, 0.80-2.35) in the highest income decile (eFigures 9 and 10 in the Supplement).

Figure 2. Trends in Income Decile and Sex-Specific Age-Standardized Incidence Rates of First Psychiatric Hospital Admissions.

Data markers denote the annual age-standardized incidence rates per 1000 person-years; error bars, 95% CIs. Lines represent the results of the joinpoint regression model. Rates have been age standardized to the 2013 European Standard Population by 5-year age groups. Income is equivalized disposable income deciles, with 1 indicating the lowest income decile and 10, the highest.

Table 2. Joinpoint Analysis of Sex- and Income-Specific Age-Standardized Incidence Rates of First Psychiatric Hospital Admissions^a.

Variable	Segment 1		Segment 2		Segment 3		Segment 4		Total Follow-up, MAAPC (95% CI)
	Segment Start	AAPC (95% CI)	Segment Start	AAPC (95% CI)	Segment Start	AAPC (95% CI)	Segment Start	AAPC (95% CI)
Overall Trends
Total	1996	2.26 (0.71 to 3.83)	2000	0.01 (−0.61 to 0.64)	2008	−3.96 (−4.78 to −3.15)	NA	NA	−0.84 (−1.03 to −0.38)
Men	1996	−0.52 (−1.00 to −0.04)	2008	−4.93 (−6.33 to −3.52)	NA	NA	NA	NA	−2.01 (−2.54 to −1.49)
Women	1996	4.69 (2.18 to 7.25)	2000	0.95 (−0.00 to 1.91)	2008	−3.25 (−4.46 to −2.02)	NA	NA	0.34 (−0.38 to 1.06)
Income-Specific Trends
Men
Nondwelling population	1996	1.12 (−0.25 to 2.50)	2005	7.20 (−6.17 to 22.48)	2008	−7.41 (−9.58 to −5.19)	NA	NA	−0.85 (−2.99 to 1.34)
1 (lowest decile)	1996	1.31 (0.62 to 2.01)	2008	−7.98 (−18.03 to 3.29)	2011	−0.47 (−6.61 to 6.08)	NA	NA	−0.60 (−2.57 to 1.41)
2	1996	1.29 (0.43 to 2.16)	2008	−5.54 (−8.12 to −2.89)	NA	NA	NA	NA	−1.04 (−2.02 to −0.05)
3	1996	0.72 (0.05 to 1.39)	2008	−6.09 (−8.12 to −4.01)	NA	NA	NA	NA	−1.60 (−2.37 to −0.83)
4	1996	−0.23 (−1.50 to 1.05)	2007	−5.18 (−7.79 to −2.49)	NA	NA	NA	NA	−2.19 (−3.38 to −0.98)
5	1996	−1.02 (−2.05 to 0.01)	2008	−4.82 (−7.82 to −1.72)	NA	NA	NA	NA	−2.30 (−3.43 to −1.16)
6	1996	−2.13 (−2.88 to −1.37)	2011	−8.40 (−16.94 to 1.02)	NA	NA	NA	NA	−3.20 (−4.74 to −1.64)
7	1996	−2.26 (−2.93 to −1.60)	NA	NA	NA	NA	NA	NA	−2.26 (−2.93 to −1.60)
8	1996	−1.06 (−2.71 to 0.61)	2002	−7.16 (−15.66 to 2.20)	2005	2.26 (−8.68 to 14.50)	2008	−5.26 (−6.93 to −3.55)	−2.98 (−5.10 to −0.81)
9	1996	−3.91 (−4.44 to −3.38)	NA	NA	NA	NA	NA	NA	−3.91 (−4.44 to −3.38)
10 (highest decile)	1996	−3.71 (−4.59 to −2.82)	NA	NA	NA	NA	NA	NA	−3.71 (−4.59 to −2.82)
Women
Nondwelling population	1996	13.39 (0.89 to 27.44)	1999	−7.94 (−24.85 to 12.76)	2002	8.66 (2.07 to 15.68)	2007	−2.73 (−5.08 to −0.33)	1.97 (−1.76 to 5.83)
1 (lowest decile)	1996	5.61 (2.36 to 8.96)	2002	1.44 (−2.27 to 5.30)	2008	−4.04 (−6.88 to −1.12)	NA	NA	0.93 (−0.77 to 2.65)
2	1996	3.02 (2.18 to 3.87)	2007	−3.43 (−4.97 to −1.88)	NA	NA	NA	NA	0.46 (−0.27 to 1.20)
3	1996	2.68 (1.43 to 3.96)	2005	−2.19 (−3.36 to −1.02)	NA	NA	NA	NA	0.22 (−0.57 to 1.01)
4	1996	1.31 (0.40 to 2.23)	2008	−3.37 (−5.83 to −0.84)	NA	NA	NA	NA	−0.27 (−1.23 to 0.69)
5	1996	1.58 (0.29 to 2.88)	2007	−3.97 (−6.34 to −1.53)	NA	NA	NA	NA	−0.62 (−1.74 to 0.52)
6	1996	3.77 (1.74 to 5.85)	2002	−0.46 (−1.41 to 0.49)	2012	−8.99 (−18.33 to 1.42)	NA	NA	−0.07 (−1.37 to 1.25)
7	1996	2.49 (0.38 to 4.65)	2005	−2.12 (−4.00 to −0.21)	NA	NA	NA	NA	0.16 (−1.13 to 1.47)
8	1996	7.27 (−1.86 to 17.24)	2000	−1.19 (−2.40 to 0.03)	NA	NA	NA	NA	0.63 (−1.37 to 2.67)
9	1996	0.06 (−0.81 to 0.93)	2010	−5.35 (−10.38 to −0.03)	NA	NA	NA	NA	−1.17 (−2.42 to 0.09)
10 (highest decile)	1996	−0.91 (−1.80 to −0.01)	NA	NA	NA	NA	NA	NA	−0.91 (−1.80 to −0.01)

Abbreviations: AAPC, average annual percentage change; MAAPC, mean AAPC; NA, not applicable.

^aStandardized to the 2013 European Standard Population by 5-year age groups.

Multivariable Analysis

In the Poisson regression model adjusted for potential confounders, a negative income gradient was observed in the IRRs of the first hospital admissions for mental disorders in adults aged 20 to 64 years: the lower the income decile, the higher the IRR compared with the highest decile. In adults, the IRR of the lowest income decile in different periods varied from 2.94 (95% CI, 2.78-3.11) to 4.46 (95% CI, 4.17-4.76) (Figure 3B). The negative income gradient was observed in the 4 most common ICD-10 main groups, and the gradient was steepest in schizophrenia and related psychotic disorders, followed by substance use disorders (ICD-10 diagnoses F20-F29 and F10-F19, respectively). The gradients were less steep in mood disorders and neurotic disorders (ICD-10 diagnoses F30-F39 and F40-F48, respectively). Estimated IRRs of the lowest income decile were 5.89 (95% CI, 5.77-6.02) for schizophrenia and psychotic disorders, 5.21 (95% CI, 5.08-5.33) for substance use disorders, 2.81 (95% CI, 2.68-2.94) for mood disorders, and 2.89 (95% CI, 2.76-3.02) for neurotic disorders (eFigures 11-12 and eTable 7 in the Supplement). In children and adolescents, a constant gradient in the IRRs was observed in the 5 lowest deciles (Figure 3A). In nondwelling children and adolescents, the IRRs compared with the highest-income decile varied from 6.76 (95% CI 5.75-7.94) to 8.83 (95% CI 7.56-10.32) (eFigure 13 in the Supplement). In persons 65 years and older, no clear gradient was observed, but the IRRs were highest in the lowest income decile, varying from 1.93 (95% CI, 1.55-2.41) to 3.56 (95% CI, 2.83-4.47) (Figure 3C). Corresponding analyses were reproduced with income decile statuses 1, 3, and 5 years before the first admissions. The observed income gradients decreased but did not disappear. In adults, the highest IRRs in the lowest income decile were 3.68 (95% CI, 3.45-3.92) 1 year before first admissions, 2.97 (95% CI, 2.79-3.15) 3 years before first admissions, and 2.71 (95% CI, 2.56-2.87) 5 years before first admissions (eFigures 14-16 in the Supplement).

Figure 3. Incidence Rate Ratios for the First Psychiatric Hospital Admissions by Income.

Data markers denote the incidence rate ratio estimates compared with the highest income decile; error bars, 95% CIs. The model is adjusted for age group, sex, urbanicity, and decrease in income decile among those aged 5 to 19 years and for age group, sex, urbanicity, decrease in income decile, educational level, and living alone in those aged 20 to 64 years and 65 years or older. Income is equivalized in disposable income deciles, with 1 indicating the lowest income decile and 10, the highest (reference group).

Discussion

In this comprehensive, nationwide cohort study of register data, we found a clear negative income gradient in the incidence rates of first hospital admissions for mental disorders in the adult population, even after adjusting for potential confounders, including the level of education, urbanicity, living alone, and decrease in income within the previous 3 years. Low household income was associated with higher incidence during the whole study period among different age groups and between men and women, although in those 20 years or younger and 65 years or older, differences between the highest deciles diminished. Trends in incidence varied between income groups. The trends decreased throughout the whole study period only in the high-income groups. To our knowledge, this is the first national-level study showing that a robust income gradient is also present in the incidence rates of first hospitalizations due to mental disorders.

The income gradient in the adult population was evident at all levels of household income in overall and ICD-10 diagnostic main category-specific incidence rates. This finding is in line with the psychosocial theory of health inequalities, which states that adversity and stress associated with lower income increase the risk of a variety of illnesses,^46,47 although the mechanisms linking income and mental health vary between disorders.²² Mental disorders have been associated with low future income,⁴⁸ but changes in income after the first admissions were not evaluated in the present study. The disorder resulting in the first hospital treatment may also have influenced the person’s ability to earn or maintain their level of income already before the first admission.⁴⁹ However, the income gradient was smaller, although still clearly observed, when income 1, 3, or 5 years before the first admissions was used for grouping the income deciles.

Opposite trends in the beginning of the study period suggested an increase in disparity between the highest and lowest income groups, which is another main finding of this study. Individuals with higher income might be in more stable and secure positions that make them more willing to undergo or more capable of receiving more intense outpatient care and avoiding first hospitalizations. This possibility is in convergence with the diffusion of innovations and cultural capital explanations of health inequalities, which state that adoption of new behaviors and the earlier uptake of new interventions, in this case outpatient care, occur earlier in higher socioeconomic positions.⁵⁰ This also supports the view that some of the mechanisms linking income and health may be located within the health care system itself.²⁵

The current approach facilitates comparisons in equality of the trends between income groups but offers limited means to interpret the population’s mental health in general. However, the overall rate of first admissions decreased during the study period in men and first increased and then decreased after 2008 in women. This finding is convergent with the decreasing total number of individuals in hospital care in Finland during the same time.³² Some evidence suggests that the true rate of mental health problems has increased in adolescents,^51,52 and increasing disparity in mental health has previously been observed in the United States, United Kingdom, and Finland.^53,54,55 The present results showed the increasing disparity between income groups also in first hospital-treated mental disorders; this seems to be the case especially in girls and women aged 15 to 19 years.

We focused on all first inpatient treatments rather than on diagnosis-specific rates so that possible sex-, income-, or race/ethnicity-related differences or variation in the temporal stability of the recorded diagnoses did not confound the analysis of the income gradient.^56,57,58,59 Diagnosis-specific analysis, on the other hand, offers insight into temporal variations in the overall trends. In addition to changes in population mental health, these variations may reflect changes in the health care system and diagnostic practices. For example, during the deinstitutionalization process, inpatient treatment shifted to other facilities in Finland, which may at least partly explain the reduction in the rate of substance use–related first admission in men.⁶⁰ In women, an increase in depression severity is a possible explanation for the increased rate of first hospitalizations.⁵⁵ Decreasing trends in first admission rates of individual disorders, such as schizophrenia, have been associated with increased outpatient care and variations in diagnostic practices.^61,62,63

Diagnosis-specific income gradient was steepest in schizophrenia and related psychotic disorders. Interestingly, social selection may be a more important mechanism in this diagnostic group.²³ Differences in gradients between disorders need to be interpreted with caution, however, because help-seeking patterns and the proportion of outpatient care may have varied among income groups, disorders, and time. In substance use disorders, social consequences, rather than severity of dependence, may be associated with treatment entry and may partly explain the gradient.⁶⁴ Mood disorders presented as the most common diagnostic group, with relatively smaller IRRs between the highest and lowest income deciles. This finding may reflect the relatively endogenous and evenly devastating nature of depression severe enough to require hospital treatment.

In terms of increased life expectancy, the deinstitutionalization of mental health care in Finland was successful for people with previous hospitalizations from 1981 to 2003, except for those with alcohol and substance abuse.⁶⁵ Major legislative changes in 1991 transferred the responsibility to provide mental health services to the municipalities. This change coincided with an economic recession period that lasted until 1993 and led to significant regional differences in the capacity of psychiatric outpatient care. The general increase in the first-time hospitalizations in that era may reflect occasional failures in outpatient management. Nearer the global economic crisis in 2008, service organizing was partly recentralized as larger hospital districts started to take over the managing of psychiatric services. These systemic changes may moderate the general trends in hospitalizations but do not explain the income-related trends. The finding that individuals with low income are overrepresented in first hospitalizations, regardless of their diagnoses or the dominant service provider institutions at the time, should be taken into account even in countries with relatively low income inequality and universal welfare policies, such as Finland.⁶⁶

Individual household members of different ages contribute differently to their household income. Children and adolescents in Finland have few possibilities of contributing to their families’ income, but family income is associated with many aspects of health.^67,68,69 Nondwelling children and adolescents had high incidences of hospitalizations throughout the study period, and being institutionalized under foster care is probably the most common reason for nondwelling. By Nordic standards, foster care is relatively common in Finland,⁷⁰ and children and adolescents in foster care have a substantial frequency of psychiatric hospitalizations and mental health problems.⁷¹

In persons 65 years and older, differences diminished between the highest income deciles, but those in the lowest income decile had the highest IRRs compared with the highest income decile. This finding is in line with some previous research,⁷² but in these age groups, persons’ economic circumstances are also affected by pension systems, household composition, and wealth and savings. Comorbid medical conditions are associated with mental disorders and low income and hence may partly explain the observed association. Whether the reasons for having low or reduced income are associated with the income gradient in mental health is an interesting question for further study.

Limitations

This study had some important limitations. First, no comprehensive data on outpatient treatments before the first inpatient treatments or population mental health in general were available for the study period. Change in income 3 years before the first admission was controlled instead. Second, not all monetary income and no wealth are captured with the national statistics of equivalized disposable income. Third, owing to the period covered by the registers, we used a 20-year clearance period to define a first hospital admission, which is a relatively short period of time in the group 65 years and older. Finally, the nature of causality and causal inference in epidemiology and health inequality are matters of debate. On its own, the present observational, register-based cohort study can describe the income gradient existing in the incidence of first psychiatric inpatient treatments.

Conclusions

This study observed a robust but temporally varying negative income gradient in first hospital-treated mental disorders. It appears that reduction in the use of inpatient care has not taken place equally between persons at different income levels. Many individuals admitted to psychiatric inpatient care for the first time are from low-income households. Hence, meeting the needs of these individuals in vulnerable positions in society appears to be an ongoing challenge for mental health services.

Supplement.

eMethods. Main ICD-10 Categories of Psychiatric Diagnoses

eTable 1. Recorded Main ICD-10 Categories of Psychiatric Discharge Diagnoses of First Psychiatric Hospital Treatments in Finland 1996-2014

eTable 2. Annual Age-Standardized Incidence Rates (IRs) of First Psychiatric Hospital Admissions

eTable 3. JoinPoint Analysis of Sex- and Income-Specific Age-Standardized Incidence Rates of First Psychiatric Hospital Admissions With Diagnosis of Mental and Behavioral Disorders Due to Psychoactive Substance Use

eTable 4. JoinPoint Analysis of Sex- and Income-Specific Age-Standardized Incidence Rates of First Psychiatric Hospital Admissions With Diagnosis of Schizophrenia, Schizotypal and Delusional Disorders

eTable 5. JoinPoint Analysis of Sex- and Income-Specific Age-Standardized Incidence Rates of First Psychiatric Hospital Admissions With Diagnosis of Mood Disorders

eTable 6. JoinPoint Analysis of Sex- and Income-Specific Age-Standardized Incidence Rates of First Psychiatric Hospital Admissions With Diagnosis of Neurotic, Stress-Related and Somatoform Disorders

eTable 7. Contrasts of the Predicted Incidence Rate Ratios Between ICD-10 Diagnostic Main Groups in Different Income Deciles

eFigure 1. Trends in Age-Standardized Incidence Rates of First Psychiatric Hospital Admissions With Organic Mental Disorders, Behavioral Syndromes Associated With Physiological Disturbances, and Personality Disorders

eFigure 2. Trends in Age-Standardized Incidence Rates of First Psychiatric Hospital Admissions With Disorders With Onset Usually Occurring in Childhood and Adolescence

eFigure 3. Trends in Age-Standardized Incidence Rates of First Psychiatric Hospital Admissions With Disorders With Major Psychotic Features

eFigure 4. Trends in Age-Standardized Annual Incidence Rates of First Psychiatric Hospital Admissions in the Nondwelling Finnish Population

eFigure 5. Trends in Income Decile- and Sex-Specific, Age-Standardized Incidence Rates of First Psychiatric Hospital Admissions With Substance Use Disorders (ICD-10 Codes F10-F19)

eFigure 6. Trends in Income Decile- and Sex-Specific, Age-Standardized Incidence Rates of First Psychiatric Hospital Admissions With Schizophrenia, Schizotypal and Delusional Disorders (ICD-10 Codes F20-F29)

eFigure 7. Trends in Income Decile- and Sex-Specific, Age-Standardized Incidence Rates of First Psychiatric Hospital Admissions With Mood Disorders (ICD-10 Codes F30-F39)

eFigure 8. Trends in Income Decile- And Sex-Specific, Age-Standardized Incidence Rates of First Psychiatric Hospital Admissions With Neurotic, Stress-Related and Somatoform Disorders (ICD-10 Codes F40-F48)

eFigure 9. Annual Sex- and Age Group–Specific Incidence Rates of First Psychiatric Hospital Admissions in the Lowest and Highest Income Deciles

eFigure 10. Annual Numbers of Person-Years at Risk of First Psychiatric Inpatient Treatments in the Lowest and Highest Income Deciles

eFigure 11. Incidence Rate Ratios for the First Psychiatric Hospital Admissions by Income With Different Diagnoses in Persons Aged 20-64 Years

eFigure 12. Predicted Incidence Rate Ratios for the First Psychiatric Hospital Admissions in Different ICD-10 Diagnostic Groups in Persons Aged 20-64 Years

eFigure 13. Incidence Rate Ratios for First Psychiatric Hospital Admissions in the Nondwelling Population

eFigure 14. Incidence Rate Ratios for the First Psychiatric Hospital Admissions by Income 1 Year Before First Admission

eFigure 15. Incidence Rate Ratios for the First Psychiatric Hospital Admissions by Income 3 Years Before First Admission

eFigure 16. Incidence Rate Ratios for the First Psychiatric Hospital Admissions by Income 5 Years Before First Admission