Incidence of major depressive disorder: variation by age and sex in low-income individuals: a population-based 10-year follow-up study: Erratum

In the article, “Incidence of Major Depressive Disorder: Variation by Age and Sex in Low-Income Individuals: A Population-Based 10Year Follow-Up Study”, which appeared in Volume 95, Issue 15 of Medicine, the affiliation for Dr. Chun-Te Lee should appear as “From the Department of Psychiatry (C-TL), Chung Shan Medical University Hospital, Taichung; School of Medicine (C-TL), Chung Shan Medical University, Taichung.”


INTRODUCTION
D epression is an important public health issue. 1 Globally, major depressive disorder (MDD) is the most prevalent mental disorder which is often long-lived, causing substantial impairment. [2][3][4] In 1996, the prevalence of MDD in Taiwan and Korea was 0.8% and 2.3%, respectively, whereas that in western countries ranged from 3.0% in the United States to 5.8% in New Zealand. 5 A cross-national study carried out in 10 countries across America, Europe, and Asia in 2003 estimated lifetime prevalence of MDD, which ranged from 3.0% in Japan (Asia) to 16.9% in the United States (North America). 6 Several crosssectional studies have associated low income with MDD. [7][8][9] For instance, low levels of household income were associated with a number of lifetime mental disorders. 10 Furthermore, a largescale, cross-sectional epidemiological study on MDD among urban and rural residents in Beijing found low family income and female sex, among others, as independent risk factors of MDD. 11 However, large-scale cohort studies on the relationship between low income and MDD have rarely been conducted. Therefore, this study aimed at conducting a population-based 10-year follow-up study to explore the association between low income and MDD by age and sex.

MATERIALS AND METHODS
The primary data source was the National Health Insurance Research Database (NHIRD). Participants' informations were obtained from 2001 to 2010 using the Longitudinal Health Insurance Databases (LHID 2005 and 2010), each comprising 1 million individuals. The NHIRD contains detailed healthcare data of >25 million enrollees, representing >99% of Taiwan's population. Low-income individuals (specifically, Group Insurance Applicants, ie, a category ''51'' or ''52'') were identified from 2001 and 2003.
All inpatients and outpatients with depressive disorders were identified using the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes. Specifically, codes 296.2 and 296.3 were used to identify MDD. Individuals diagnosed with MDD from 2001 to 2003 were excluded from the analysis. Also excluded were 20,689 individuals whose income status changed from non-low to low from 2004 to 2010. The Institutional Review Board of Chung Shan Medical University Hospital, Taichung City, Taiwan, approved this study.
x 2 was used to analyze nominal variables and to compare the proportion of low and non-low-income individuals. Cox proportional hazard model was used to investigate the hazard ratio (HR) of MDD in low-income individuals. Adjustments were made for age, geographic location, urbanization level, and Charlson comorbidity index. The proportional assumption of Cox regression models was tested. Statistical analyses were performed using the SAS 9.3 statistical software (SAS Institute, Cary, NC).

RESULTS
The demographic characteristics of low-income and nonlow-income individuals are shown in Table 1. Eligible for the study were 15,417 low-income individuals (7247 males and 8170 females) and 1,728,531 non-low-income individuals (854,106 males and 874,425 females) from 2001 to 2003. The proportion of low-income females (52.99%) was significantly higher than their non-low-income counterparts (50.59%). However, the proportion of non-low-income males (49.41%) was significantly higher than their low-income counterparts (47.01%). Among the various age groups, the proportions of low-and high-income individuals were significantly different. The proportion of non-low-income individuals aged 18 to 44 years was significantly higher than that of other age groups. However, the proportion of low-income individuals aged 0 to 17 years (43.85%) was significantly higher when compared with other age groups. Considering the geographical location, urbanization level, and Charlson comorbidity index, the proportion of low-income individuals differed significantly when compared with their non-low-income counterparts. From 2004 to 2008, the proportion of low-income individuals diagnosed with MDD (1.73%) was significantly higher than non-lowincome individuals (0.92%). Table 2 shows the incidence rates of MDD in the non-lowincome and low-income patients categorized by age and sex. Among non-low-income individuals, the MDD incidence rates increased with age in both sexes. That is, 0.35, 0.93, 0.97, 1.40 per 10,000 person-months in males and 0.41, 1.60, 1.89, 1.95 per 10,000 person-months in females aged 0 to 17, 18 to 44, 45 to 64, and !65 years, respectively. Low-income males aged 45 to 64 years had the highest MDD incidence rate (3.04 per 10,000 person-months), followed by males aged 18 to 44 years (1.99 per 10,000 person-months), !65 years (1.59 per 10,000 personmonths), and 0 to 17 years (0.86 per 10,000 person-months). Nonetheless, low-income females aged 18 to 44 years had the highest MDD incidence rate (3.90 per 10,000 person-months) followed by those aged 45 to 64 years (3.77 per 10, 000 personmonths), !65 years (1.75 per 10,000 person-months) and 0 to 17 years (1.17 per 10,000 person-months). In both low-and nonlow-income individuals, the MDD incidence rates were higher in females than males of all the age groups. Among non-lowincome individuals, the female to male MDD incidence ratio (1.95) was highest among the age group 45 to 64 years. Among the low-income individuals, the incidence ratio was highest (1.96) among those who were 18 to 44 years old.
The proportional assumption of Cox regression models was tested. However, the output was not significant (P ¼ 0.6987). This indicates that the effect of low income on MDD was constant over time; hence, the proportional

DISCUSSION AND CONCLUSIONS
This is the first large-scale longitudinal follow-up study to evaluate the association between low income and MDD by age and sex. In this study, income was significantly associated with MDD as opposed to some previous studies. A cross-sectional study of relatively smaller sample size (clinical depression n ¼ 330) reported higher MDD prevalence in individuals with lower monthly income. However, income was not found to serve as a significant risk factor. 12 Moreover, in this study, the incidence but not prevalence rate was determined. In another cross-sectional study on major depression during the economic crisis in Greece, higher numbers of depressive symptoms were found in more financially strained than less-strained individuals. 7 Nonetheless, our study found an association between low income and MDD by sex and age.
Liu et al 11 found that female sex, age above 45 years, low family income, and poor family environment were independent risk factors of MDD. However, the study was not stratified by income. Interestingly, this study had more findings. For example, among low and non-low-income individuals, the MDD incidence rates were higher in females than males in all age groups. Furthermore, low income was not found to serve as a risk factor of MDD in both males and females aged !65 years. A previous study found age !64 years and low income among others as independent risk factors of MDD in diabetes patients. 13 The finding that older individuals (!65 years' old) are not associated with depression is not surprising. This is because older adults have been shown to be more protective against depression, 14 as they seem to react less to stressors than younger adults. 15 Another finding of this study was that lowincome females aged 18 to 44 years had the highest incidence of MDD, whereas those aged 0 to 17 years had the highest HR for MDD. However, in low-income males, the highest incidence rate and HR of MDD were found in individuals aged 45 to 64 years. A study conducted on Chinese adolescents showed that girls had more depressive symptoms than boys. 16 These results align with our findings wherein low-income females between the ages of 0 and 17 years had higher incidence rates and HR of MDD than males. Some possible explanations for these findings are girls have been found to experience higher levels of interpersonal stress than boys, and tend to show more anxiety and depression when confronted with interpersonal stress in the family and among peers. 17,18 The differences in depression between males and females escalate from early to late adolescence probably because girls are more likely to blame themselves for relationship problems and are more concerned about negative evaluations by peers than boys. Besides, once distressed, adolescent girls may even continue to create a more stressful environment. 17 A study conducted on adolescent girls associated body dissatisfaction and body mass index with depression. 19,20 Adolescence is a time of increased vulnerability to depression. Females who may be aware of their low-income status at that stage are likely to be more predisposed to depression. Some studies showed that women were roughly twice as likely as men to be classified as having major depression. [21][22][23][24] In such studies, however, subjects were not grouped by low income and age as in the present study. The female to male MDD incidence ratios among non-low-income individuals aged 0 to 17, 18 to 44, 45 to 64 and !65 years were, respectively, 1.17, 1.72, 1.95, and 1.39. Among low-income individuals, the female to male MDD incidence ratios were 1.36, 1.96, 1.24, and 1.10 for the age groups 0 to 17, 18 to 44, 45 to 64, and !65 years, respectively. These variations observed between females and males may be related to income and age.
In this study, we assessed the proportional assumption of the Cox regression models. The output was not significant (P ¼ 0.6987). This is an indication that the effect of low income on MDD was constant over time; hence, the proportional assumption was met. Additionally, the KM-plots stratified by age and sex showed that the curves did not cross or intersect each other (figures not shown).
It is possible that the effects of income on MDD may vary over time. However, there were no data on the time-varying income variable; hence, the effects of varying income on MDD over time could not be determined. Individuals were simply divided into low and non-low-income groups. Those whose status changed from non-low to low income from 2004 to 2010 were excluded to minimize potential bias. That is, participants in the control group were those with stable income.
There were some major confounding factors that were not considered during analysis. The association between low income and MDD may have been confounded by environmental (including policy or social support system) and physiological problems (such as depression, anxiety disorders, schizophrenia, personality disorders, adjustment disorders, and family problems). However, these variables were not available and thus could not be adjusted for. This was our study limitation.
In conclusion, MDD incidence rates for the non-lowincome men and women were found to increase with age. Furthermore, low-income females aged 18 to 44 years and males aged 45 to 64 years had the highest incidence of MDD. Finally, low income was found to serve as a significant risk factor of MDD only in individuals younger than 65 years. Each model was adjusted for geographical location, urbanization level, and Charlson comorbidity index. CI ¼ confidence interval, HR ¼ hazard ratio, MDD ¼ major depressive disorder.