Introduction

Depression, characterized by symptoms of sadness, depressed mood, and loss of interest, is one of the most prevalent disorders worldwide and accounts for 5.9% of total all-cause disability-adjusted life years (DALYs)1. Depression is associated with poor quality of life, prevalence of cancer, chronic diseases and suicide, and thus may contribute to an increase in mortality2. Depression casts a heavy burden on families, communities and health services in both developed and develo** countries3,4.

The prevalence of depressive symptoms (DS) varies widely, from 1% to 16% among middle-aged and elderly people in studies conducted in Western countries5,6,7. One study reported DS prevalence of elderly Chinese people being 3.9% during 1980s and 1990s8. Two recent meta-analysis studies reported the prevalence has been increasing rapidly because of urbanization and environmental changes in China, up to 17% among people aged 55 years and above using the Center for Epidemiologic Studies Depression Scale (CES-D)9,10.

Several socio-demographic characteristics, such as age, gender, marital status, education level, and income have been associated with DS11, although the evidence is not entirely consistent7,10. Age is associated with DS in many studies12, and interestingly, this association is inverse or U-shaped in several studies13,14,43 Other factors, including geography and urbanization, are found to be important factors in DS development, especially in China because from 2012 more than half of its population live and work in urban areas and there has been massive rural-to-urban migration to Chinese cities44. In a 2011 Migration and Quality of Life Survey, Chen et al. reported highly populated cities along the eastern coast such as Shanghai illustrated potential effects of urbanization as a stressor for mental health45.

In a Hong Kong study, Li et al. found that underweight individuals had a high prevalence of DS and overweight ones a low prevalence, compared to those in the normal weight range46. Yu et al. found that thin men had a higher prevalence of DS than normal weight men in Taiwan47. In rural China, obese men were found to be less likely to suffer from DS, compared with normal weight men48. Consistent with these Chinese studies, we found that BMI was negatively associated with prevalence of DS in men and in women. However, in two meta-analysis studies, one for cohort studies and another for cross-sectional studies, mainly conducted in Western populations, reported a significantly positive association between DS and obesity49,50. A United States-based study reported that major depressive symptoms were associated with a low BMI in men but a high BMI in women22. A study in the Netherlands observed a significant U-shaped association between BMI (underweight, normal, overweight and obesity) and DS, i.e., both obesity and underweight were associated with an increased risk of DS51. The underlying mechanism for this discrepancy is not entirely clear. Among middle-aged and elderly Chinese people, the prevalence of obesity is low (3.7% in this study) and severe obesity is rare; the association pattern might be explained by the ‘happy mind and fat body’ culture because many overweight or modestly obese people tend to be more optimistic and happier than those who are thinner in China21,46. But in Western countries, obesity, particularly morbid obesity, could be related to greater stress, stigma, chronic diseases and devaluation, causing obese people to suffer lower self-esteem and have a more negative self-image, resulting in more DS49. Although several studies reported that BMI or obesity is inversely correlated with DS, some of them only observed the impact of DS on BMI and did not observe an inverse effect of BMI on DS as we did in our study. This directionality could lead to a different conclusion. Forman-Hoffman et al. reported that the effects of DS on weight differ by gender, and concluded that DS at baseline may predict both weight loss and weight gain52. However, Konttinen et al. reported that men with depressive symptoms were likely to have a higher BMI, while women with a higher BMI were likely to develop depressive symptoms53. In middle aged and elderly Asian populations, depression can lead to weight loss rather than obesity, and being underweight may elicit DS unlike among Westerners who show a positive relationship between depression and obesity. These findings have implications for public health interventions by gender in China54.

Groffen et al. reported that only current smoking, not past smoking, was associated with DS in white women in US55. Current smoking was associated with increased prevalence of DS in the Korean and Japanese female populations56,57. On the other hand, a study in Eastern Europe found that current and past smoking was associated with DS in men but not in women58. Effects of smoking on DS are complex: Nicotine has antidepressant properties, releasing dopamine in the mesolimbic reward pathway, thereby elevating mood and relieving stress59. However, evidence has also suggested that smoking increases a person’s risk of DS as a result of changes in neurophysiology60. In our study, ever smoking rate in men was 71% (50.3% for current smoking, 20.7% for past smoking), much higher than in many western countries, but the prevalence of DS was low (2.4%). We found no statistical difference in DS between smokers and non-smokers among men in our study. However, smoking rate was low in women (2.0%) in our study, and female current smokers had a high prevalence DS compared with non-smokers. It is unclear whether such association was caused by women with DS seeking stress relief via smoking. An alternative explanation is that smoking was not socially acceptable for women, especially in urban China. Thus, female smokers could have low self-esteem which led to an increase of DS57,61.

Another contribution was that our study assessed whether DS was more likely to occur in an early vs. late stage of chronic disease (i.e., within 5 years of diagnosis, or after this period) to identify when risk for depression is most elevated. Depression is a significant problem for cancer patients in particular, either as a pre-existing or co-existing condition, or as a consequence of cancer diagnosis and treatment. Several studies have reported that the associations of cancer with DS lasted for varying lengths of time depending on the type of cancer23,62. However, few studies have focused on assessing DS among long-term cancer survivors in a non-medical setting62,63. In our population-based study, we found an increased DS prevalence primarily within 5 years post-cancer diagnosis. This association persisted but was attenuated when the time interval since diagnosis increased to beyond 5 years in women. These findings suggest a need of proper intervention and care for mental health for cancer survivors, particularly during the period soon after cancer diagnosis. Consistent with previous findings24,25, we found that patients with Parkinson’s disease or stoke had an increased prevalence of DS that appeared relatively consistent both before and after 5 years post-diagnosis. These associations call for a long term management/treatment approach.

Our study has many strengths, including a large sample size, a population-based design, and adjustment for a wide range of socioeconomic characteristics, lifestyle factors, chronic diseases, and cancer. Another noticeable strength is that the measure of all physical illnesses took place prior to the CES-D measurement, minimizing the risk of reverse causation. However, several limitations of our study should also be acknowledged. (1) The relatively low participation rate of 74.0% among men in the SMHS, compared with that of women in the SWHS, could lead to selection bias in parameter estimation. Reasons for non-participation included refusal, being out of the area during enrollment and other miscellaneous reasons, including health and hearing problems. However, lower participation of men is consistent with other large psychiatric epidemiology studies64, including those conducted in China65. A study from the Danish National Birth Cohort reported that in a cohort study based on prospective data, the decision to participate cannot be based on future outcomes; this suggests a minimal risk of bias in our study given that DS outcomes were collected prospectively66. (2) We used data from two large cohort studies: the SWHS and the SMHS. However, these datasets did not allow examination of some key risk factors, including but not limited to early childhood adversity and psychosocial stress at work. (3) We only included 6 CES-D questions instead of the full 20- or 26-item measure in the study, which may not have captured all possible DS and therefore could have resulted in an underestimate of the prevalence of DS in our study population. Further, the CES-D indicates the likelihood of clinical depression, not clinical depression itself. Our study would have been strengthened if our shortened DS measure was validated by a sub-sample where clinical interviews were conducted to evaluate our shortened measure’s ability to accurately detect clinical depression. However, these concerns are at least somewhat mitigated by the psychometric analyses that we conducted above on our 6-item measure. (4) Our study was conducted among long-term urban residents in Shanghai, one of the most developed cities in China. Thus, the results of our study may not be generalizable to all Chinese, particularly those living in rural areas or those who migrated to urban areas.

In conclusion, in this large population-based study of 103,595 Chinese women and men, we found that 2.4% and 5.6% of middle-aged and elderly Chinese men and women living in Shanghai experienced DS. Age and several socio-demographic characteristics and lifestyle factors, such as education, income, smoking, cancer and several chronic diseases were associated with DS. In future studies, qualitative approaches would be quite valuable in further elucidating the nature of DS among elderly, and whether these are tied to perceived changes in aging, health status, socioeconomic status, or cultural standing in relation to modernization. Findings of our study are value for development of prevention programs in identifying elderly individuals with DS for early intervention.