Introduction
Poverty has a hefty destructive impact on modern societies. Poverty alleviation has, for a long time, been listed as an overarching objective in many countries' development plans (Iniguez-Montiel, Reference Iniguez-Montiel2014). In recent decades, the Millennium Development Goals (MDGs) and Poverty Reduction Strategies (PRSs), along with some other important documents, have aroused international interest in alleviating and eradicating poverty (Schwittay, Badiane, & Berdish, Reference Schwittay, Badiane and Berdish2011). China has also set a goal of eradicating absolute poverty by 2020 (Ye & Yin, Reference Ye and Yin2019). However, emerging evidence suggests that targeting poverty, especially relative poverty, has become tougher within the development process, as some vulnerable groups are being structurally marginalized by labour market (Li & Bai, Reference Li and Bai2010).
Poverty associated with illness and disabilities has long been a concern for the Chinese government (Liu, Reference Liu2015). Compared with other vulnerable groups, persons with severe mental illness (SMI) may be more disadvantaged and poorer because of stigma and higher social stress (Trani et al., Reference Trani, Bakhshi, Kuhlberg, Narayanan, Venkataraman, Mishra and Deshpande2015; Wilton, Reference Wilton2004). SMI is characterized as mental, behavioural or emotional disorders that can result in serious functional impairment. Commonly, it includes schizophrenia, bipolar disorders and major depressive disorders, which are the leading causes of disability (Liu et al., Reference Liu, Zhang, Pang, Li, Chen and Zheng2015). China Mental Health Survey, the first national representative community survey of mental disorders conducted during 2013 and 2015, showed that the weighted prevalence of any disorder (excluding dementia) in China was 16.6% during the participant's entire life. The lifetime prevalence of schizophrenia and any mood disorder was 0.6% and 7.4%, respectively (Huang et al., Reference Huang, Wang, Wang, Liu, Yu, Yan and Lu2019). Data with regard to poverty that is associated with SMI are scarce in China. A survey suggests that there are 11.7 million adults with a mental disability, and over 38.2% of them are living in poverty in China (Li, Pang, Du, Chen, & Zheng, Reference Li, Pang, Du, Chen and Zheng2012).
According to the social drift hypothesis (Lund et al., Reference Lund, De Silva, Plagerson, Cooper, Chisholm, Das and Patel2011), SMI may induce and exacerbate poverty by depriving one's education and employment opportunities, increasing medical costs, as well as creating isolations for persons with SMI (Breslau, Lane, Sampson, & Kessler, Reference Breslau, Lane, Sampson and Kessler2008; Corrigan, Watson, & Barr, Reference Corrigan, Watson and Barr2006). Empirical evidence strongly shows that persons with SMI are at increased risk of drifting into or remaining in poverty (Lund et al., Reference Lund, De Silva, Plagerson, Cooper, Chisholm, Das and Patel2011; Saraceno, Levav, & Kohn, Reference Saraceno, Levav and Kohn2005). By now, although the negative economic impacts of SMI were recognized as being chronic (Anakwenze & Zuberi, Reference Anakwenze and Zuberi2013), few longitudinal empirical studies exist that examine the poverty transitions associated with SMI. Targeted poverty alleviation policies may differ based on the exact poverty transition patterns, thus studies focusing on examining the factors, mechanisms and processes of poverty transitions associated with SMI are of paramount significance to guide anti-poverty practices (Trani et al., Reference Trani, Bakhshi, Kuhlberg, Narayanan, Venkataraman, Mishra and Deshpande2015).
It is widely recognized that mental illness and poverty are highly correlated and interact in a way that each can be both a cause and a consequence of the other (Lund et al., Reference Lund, De Silva, Plagerson, Cooper, Chisholm, Das and Patel2011). Apart from social drift, social causation hypothesis is also frequently mentioned, which depicted the opposite pathway from poverty to mental illness. Although previous studies suggest that social drift pathway is more applicable for SMI while social causation pathway might apply more readily to common mental illness such as anxiety (Lund et al., Reference Lund, De Silva, Plagerson, Cooper, Chisholm, Das and Patel2011), the dispute has not subsided in cross-sectional studies. Therefore, although social drift has been mentioned in many studies (Lund et al., Reference Lund, Breen, Flisher, Kakuma, Corrigall, Joska and Patel2010; Patel et al., Reference Patel, Araya, Chatterjee, Chisholm, Cohen, De Silva and Van Ommeren2007), few of them can provide long-term empirical evidence to support it. Among current existed longitudinal studies of social mobility after the onset of SMI (Harding, Brooks, Ashikaga, Strauss, & Breier, Reference Harding, Brooks, Ashikaga, Strauss and Breier1987; Ngui et al., Reference Ngui, Cohen, Courteau, Lesage, Fleury, Gregoire and Vanasse2013), few have taken a dynamic perspective to view the poverty transitions associated with SMI over time.
Applying longitudinal data from an ongoing mental health project in a less developed area of China, the objectives of this study were: (1) to identify various poverty transition patterns and the number of people who live in those patterns, and (2) to find the causes of different poverty transitions. Correspondingly, this study first provided an overview of poverty transitions for persons with SMI for over 20 years. Then, both discrete and continuous regression models were applied to search for the causes of poverty transitions and income changes for them. This study might contribute to accumulating empirical evidence on social drift, as well as providing solid evidence for further targeted poverty alleviation policies.
Methods
Data source
This study is based on the Chengdu Mental Health Project (CMHP), an ongoing longitudinal project on mental illness and mental health services in Xinjin district, Chengdu. In 1994, the first mental health survey in six townships of Xinjin was conducted, which covered a total of 123 572 people (⩾15 years old) and identified 711 persons with SMI [schizophrenia: n = 515 (72.4%); mood disorders: n = 196 (27.6%)]. In 2015, the second mental health survey in the same townships of Xinjin was conducted, in which all persons with SMI identified in 1994 were followed up. For those who had moved out to other places, phone calls were made to confirm their status. Finally, a total of 650 out of 711 persons with SMI in 1994 were followed up in 2015 (follow-up rate: 91.4%). Among them, 308 persons with SMI were still alive [schizophrenia: n = 253 (82.1%); mood disorders: n = 55 (17.9%)]. To understand the long-term impacts of SMI on poverty, all alive persons with SMI were included in the analysis in this study.
Detailed methods with regard to these two mental health surveys had been described elsewhere (Ran et al., Reference Ran, Yu, Peng, Liu, Zhang, Hu and Luo2018, Reference Ran, Xiao, Fazel, Lee, Luo, Hu and Chan2020; Yu et al., Reference Yu, Peng, Bai, Luo, Yang, Li and Ran2020a, Reference Yu, Luo, He, Yang, Liu, Guo and Ran2020b). Briefly, all persons with SMI were identified through two steps: (1) screening procedures for psychosis one household by household, and (2) general psychiatric interview. First, face-to-face interviews were conducted by trained investigators (nurses, psychologists and social workers) with each head of the household to identify potential cases of mental disorder. The Psychoses Screening Schedule (PSS) was completed by means of interviewing the heads of all the households, and discussion with village doctors and neighbours on the survey sample. Second, once a potential psychosis patient was identified, trained psychiatrists conducted a comprehensive general psychiatric interview with that subject for further diagnosis. These two mental health surveys used the same diagnostic criteria, the International Classification of Disease (ICD-10). Diagnostic reliability had been established between the trained psychiatrists with over 5 years of clinical experience. These two surveys were approved by the University Human Research Ethics Committee (HREC). All research participants provided informed consent after receiving a complete explanation of the study.
Analytic strategy
Poverty in this study refers to monetary poverty. Those with annual household income per capita lower than poverty standards were defined as poor, otherwise non-poor. Chengdu is one of the wealthiest cities in western China. Given the huge income disparity in China, applying national poverty standards may result in a higher estimation of non-poverty rate in Xinjin, Chengdu. Therefore, the Minimum Living Standard in Xinjin, defined by local government, was applied as the poverty standards in this study (Gao & Zhai, Reference Gao and Zhai2012), which were 850 CNY and 5400 CNY, respectively, for the years 1994 and 2015. The adjusted rate was similar to the growth of consumer price during this period, thus these two poverty measures were fully comparable. To better modelling the impacts of SMI-related factors on shaping poverty, income data in this paper excluded social welfare benefits allocated to households where persons with SMI reside, as those benefits were not targeted at them. Evidence showed that current social welfare benefits might reduce the poverty rate for persons with SMI (Yu et al., Reference Yu, Peng, Bai, Luo, Yang, Li and Ran2020a, Reference Yu, Luo, He, Yang, Liu, Guo and Ran2020b). Patient's mental disability level was assessed based on the Social Disability Screening Schedule (SDSS), a 10-item assessment tool to measure 10 social functions for persons with mental illness (Zhou et al., Reference Zhou, Zhou, Li, Lin, Yao, Chen and Shen2015).
Based on one's poor or non-poor status in 1994 and 2015, four poverty transition patterns were identified: (1) persistent poverty, (2) fell into poverty, (3) escaped poverty and (4) never experienced poverty. Descriptive analysis was applied to show the number of people who live in those patterns. Then, both discrete and continuous regression models were applied to explore the causes of different poverty transitions. The discrete modelling approach estimated poverty transitions, while the continuous approach was based on income and income growth to provide complementary interpretations on poverty transitions (Lawson, McKay, & Okidi, Reference Lawson, McKay and Okidi2006).
For the discrete modelling part, first, a multinomial logistic model was applied to predict the likelihood that a given person with SMI would belong to any of these four poverty transition patterns based on person's initial status in 1994 and later changes of some key variables. The group of persons who had never experienced poverty was taken as a base. Then, poverty transitions were treated as a sequential result that can be affected by two sets of factors: those that made people more likely to become poor or non-poor in 1994, and those that enabled them to change their poverty status by 2015. Binary logistic models were applied, which consisted of three different regression models. The first model addressed the issue of one's poverty status in 1994, the second and third models addressed the issue of poverty status in 2015.
For the continuous modelling part, the first model took income in 2015 as the dependent variable and estimated how SMI-related factors in 1994 and changes on some key factors might have contributed to the reported income in 2015. The second model took the income growth between the period of 1994 and 2015 as the dependent variable. Both discrete and continuous models applied similar explanatory variables. Key change variables in this study included the marital status, work ability, physical illness, educational attainment, disability level of persons with SMI and the household size. Data analysis was finished using SPSS 23.0.
Results
Characteristics of participants with SMI
Table 1 shows the sociodemographic and clinical characteristics of participants (n = 308) in 1994 and 2015. There were more females (54.9%) and persons with schizophrenia (82.1%) in this data set. The rate of married of these participants decreased from 69.6% in 1994 to 61.7% in 2015. In the mental health survey in 1994, they had an average of 10.5 years of mental illness duration. There were substantial changes for people with SMI, as they were found to have significantly higher rates of being unable to work (21.8%, p < 0.001), having comorbidity of physical illness (32.1%, p < 0.001), being severely mentally disabled (26.9%, p < 0.01) and medication (73.0%, p < 0.001) in 2015. The poverty rate and reported household income per capita also increased significantly in 2015 (p < 0.001).
Sociodemographic and clinical characteristics for the followed-up and unfollowed-up cases
×p < 0.05, ××p < 0.01, ×××p < 0.001 (differences for the followed-up panel in 1994 and 2015); *p < 0.05, **p < 0.01, ***p < 0.001 (differences between the followed-up and unfollowed-up cases in 1994).
In this study, 56.7% (n = 403) of the original 1994 samples were not included. Table 1 also shows the characteristics of those unfollowed-up cases in 1994. Compared with followed-up cases, those unfollowed-up cases had significantly higher rates of being unable to work (26.2%, p < 0.001), having physical illness (15.8%, p < 0.001) and being severely mentally disabled (28.3%, p < 0.01), as well as had lower marriage rate (59.3%, p < 0.05) in 1994. The mean age and illness duration of unfollowed-up cases were 57.2 and 14.8 years, respectively, which were significantly higher than followed-up cases. The family size for unfollowed-up cases was significantly smaller. Although the unfollowed-up cases had a higher poverty rate and lower annual income per capita than the followed-up cases, the differences were not statistically significant.
Figure 1 shows the poverty incidences of the participants in 1994 and 2015, as well as the percentages of four poverty transition types. The poverty rate for the participants increased from 39.9% to 49.4% (increase rate: 23.8%). In 1994, 185 persons with SMI (60.1%) were not in poverty, but after 21 years, 27.3% of these participants (n = 84) had fallen into poverty. There were 68 participants (22.1%) who were poor at both time points, whereas 55 participants (17.8%) were able to escape poverty. In general, a larger proportion of participants had fallen into poverty (27.3%) rather than moved out of it (17.8%).
Twenty-one-year trend of poverty for persons with SMI (n = 308).
Discrete models on poverty transitions
Table 2 shows the results of the multinomial logistic model, with the base being the group which had never experienced poverty. Participant's work ability, physical health condition, education level, level of disability and treatment status in 1994 were significantly attributed to their later poverty transitions. Specifically, those who were unable to work, had physical illness or were severely mentally disabled in 1994 were more likely to live in persistent poverty or fall into poverty in 2015. Compared to illiterate ones, persons with SMI who achieved high school and above education were less likely to live in persistent poverty. Medication also mattered, as persons without medication had a higher probability of living in persistent poverty or falling into poverty. If they received medication, persons with SMI were significantly less likely to live in persistent poverty or fall into poverty (p < 0.01). Persons who resided in larger households were less likely to live in persistent poverty. Losing work ability during the period of 1994 and 2015 was significantly related to a participant's status of falling into poverty (p < 0.01). The comorbidity of physical illness, as well as the increase of disability levels, was also related to maintain them in persistent poverty or move them into poverty.
Results of the multinomial logistic model, with base equal to never experienced poverty
*p < 0.05, **p < 0.01, ***p < 0.001.
Table 3 shows the results of binary logistic models. Factors associated with whether or not persons with SMI were poor in the initial year, 1994, were examined separately from the factors associated with changes in poverty status over the 21-year period. In the first model, the likelihood of a person with SMI being poor in 1994 was significantly associated with one's gender (p < 0.01), marital status (p < 0.01), disability level (p < 0.001), treatment status (p < 0.05) and household size (p < 0.05). Specifically, females, married, with medication and residing in larger households were significantly less likely to be poor in 1994 (p < 0.01). Having severe or moderate disability increased the likelihood of a person being poor in 1994.
Results of the binary logistic models, 1994 and 2015 (poor = 1, non-poor = 0)
*p < 0.05, **p < 0.01, ***p < 0.001.
In the second model, for those poor persons in 1994 (n = 74), their work ability, physical health status, educational level, duration of SMI and treatment status were all significant predictors of later poverty transitions. Those who had no physical illness, had the ability to work, high school and above education level, shorter duration of mental illness and received medication in 1994 were more likely to escape poverty. During the 21-year period (1994–2015), those who lost their work ability and suffered from physical illness were more likely to remain in poverty. Receiving medication was a positive predictor of escaping poverty. In the third model, for those non-poor persons in 1994 (n = 234), severe mental disability and insufficient medication, as well as negative changes on their work ability and physical health status, could have driven them into poverty status by 2015. Those who turned to medication were more likely to remain non-poor status by 2015 (p < 0.01).
Continuous models on income and income growth
Table 4 shows the model of the predictive factors for the participant's income in 2015. Persons with SMI who lost work ability, had physical illness, severe mental disability and received no medication in 1994 had significantly lower income than their counterparts in 2015. During the period of 1994–2015, negative changes on participant's work ability, physical health status and disability levels were also significantly associated with lower income in 2015, whereas receiving medication was a predictive factor of higher income. Educational level was also significant in predicting income for persons with SMI, as those who received high school and above education in 1994 had significantly higher income (p < 0.01).
Results of linear regressions on income in 2015 and income growth
*p < 0.05, **p < 0.01, ***p < 0.001.
Table 4 also shows the model on income growth. The growth of income for persons with SMI was significantly lower for those who were unable to work (p < 0.001), severely mentally disabled (p < 0.01), had physical illness (p < 0.001) or never received treatment (p < 0.01) in 1994. Negative changes on their work ability, comorbidity of physical illness and mental disability level could predict one's lower income growth in 2015 (p < 0.05). Turning to medication was positively associated with higher income growth (p < 0.01). The growth of income was significantly higher for those who had a higher income in 1994 (p < 0.001).
Discussion
Chronic poverty is at the forefront of targeting poverty policy debates (Dercon, Hoddinott, & Woldehanna, Reference Dercon, Hoddinott and Woldehanna2012; Kimberlin & Berrick, Reference Kimberlin, Berrick, Fernandez, Zeira, Vecchiato and Canali2015). In order to develop appropriate anti-poverty interventions, it is of paramount significance to understand the factors that had trapped people in poverty or had pulled them out. Social drift theory suggests that people with SMI are more likely to live in a status of chronic poverty (Lund et al., Reference Lund, De Silva, Plagerson, Cooper, Chisholm, Das and Patel2011; Mitra, Posarac, & Vick, Reference Mitra, Posarac and Vick2013), the results of this study provide further evidence of the long-term poverty transitions associated with SMI. Applying a longitudinal dataset from an ongoing mental health project in China, this study provides an overview of 21-year poverty trajectories for persons with SMI and explores the causes of different poverty transition patterns. To the best of our knowledge, this is the first study to do so in the context of a Chinese rural community.
Originally, in 1994, there were 711 persons with SMI. After 21 years, although the follow-up rate was high (91.4%), only 308 persons with SMI (43.3%) were alive and eligible for this long-term poverty transition study. Compared with unfollowed-up cases, they were significantly younger and had a better social function in 1994. During this long period of time, they had experienced overall significant negative changes in their work ability, comorbidity and mental disability. A larger proportion of persons with SMI had been impacted by physical illness and severe mental disability, which may contribute to their lack of work ability. Recovery from SMI is a hard task (Perese, Reference Perese2007), particularly in places where community mental health services had not been strengthened (Tse, Ran, Huang, & Zhu, Reference Tse, Ran, Huang and Zhu2013). The overall rates of medication in persons with SMI increased, which may be directly benefited from China's rapid social development and increased awareness of mental health problems (Yu et al., Reference Yu, Peng, Bai, Luo, Yang, Li and Ran2020a, Reference Yu, Luo, He, Yang, Liu, Guo and Ran2020b).
In this study, although participants' overall income increased significantly from 1994 to 2015, a larger proportion of them were found to fall in poverty (27.3%) rather than move out of it (17.8%). The increasing rate of poverty was 9.5% in persons with SMI during the 21-year follow-up. This could be a strong empirical evidence to support social drift theory (Lund et al., Reference Lund, De Silva, Plagerson, Cooper, Chisholm, Das and Patel2011). Although economic deprivation associated with ageing was common (Mezuk, Edwards, Lohman, Choi, & Lapane, Reference Mezuk, Edwards, Lohman, Choi and Lapane2012), adverse poverty transitions may be more marked among persons with SMI. Our previous studies based on the same datasets found that, compared with the annual net income per person in the general population in Xinjin district, income for persons with SMI was significantly lower in both 1994 and 2015. Meanwhile, the income disparity between the general population and persons with SMI increased significantly from 1994 to 2015 (Ran et al., Reference Ran, Weng, Liu, Zhang, Yu, Peng and Xiang2019; Yu et al., Reference Yu, Peng, Bai, Luo, Yang, Li and Ran2020a, Reference Yu, Luo, He, Yang, Liu, Guo and Ran2020b).
It should be noted that this study on poverty transitions is rooted in a special period of fast socioeconomic development in China. Guided by the strategy of opening and reforming, based on a strong belief in the trickle-down effect (Norton, Reference Norton2002), the Chinese government has launched a package of programmes to stimulate development (Wang, Reference Wang2008). One positive outcome of this is that the population living in extreme poverty had decreased sharply from nearly 240 million in 1978 to 16.6 million in 2018, and the extreme poverty is expected to be eradicated by the end of 2020 (Ye & Yin, Reference Ye and Yin2019). In this study, although a larger proportion of persons with SMI were found to fall into poverty rather than move out it, there were still around 50% persons with SMI not living in poverty at the follow-up. For persons with a long history of SMI, this result may not be so frustrating while comparing it with studies in other countries (Harding et al., Reference Harding, Brooks, Ashikaga, Strauss and Breier1987; Lund et al., Reference Lund, Breen, Flisher, Kakuma, Corrigall, Joska and Patel2010). Albeit so, social drift phenomenon in China is still noteworthy, and persons with SMI are relatively harder to benefit from social development and escape poverty than others in the same society (Ran et al., Reference Ran, Weng, Liu, Zhang, Yu, Peng and Xiang2019; Yu & Ran, Reference Yu and Ran2019).
In terms of specific mechanisms that trapped persons with SMI into poverty, the results of this study indicated that reduced work ability could significantly predict chronic poverty status for persons with SMI. Persons without adequate work ability are excluded by labour market, thus result in a decrease in income (She & Livermore, Reference She and Livermore2009). Without appropriate social support policy, a person with SMI may be constantly marginalized in the society (Richter & Hoffmann, Reference Richter and Hoffmann2019). This study also showed that physical illness and level of mental disability were significant predictors of chronic poverty, as well as reasons that trapped non-poor persons into poverty. Studies in many countries have indicated that persons with mental illness report high rates of physical illness (De Hert et al., Reference De Hert, Correll, Bobes, Cetkovich-Bakmas, Cohen, Asai and Ndetei2011; Ehrlich et al., Reference Ehrlich, Kendall, Frey, Kisely, Crowe and Crompton2014). Meanwhile, without sufficient mental health service resources, one's mental disability status may be further deteriorated (Barr et al., Reference Barr, Taylor-Robinson, Stuckler, Loopstra, Reeves and Whitehead2016). Physical illness and deteriorated mental disability further reduce the work ability for persons with SMI, as well as increase their treatment burden, thus further aggravate poverty (Lund et al., Reference Lund, Breen, Flisher, Kakuma, Corrigall, Joska and Patel2010). For participants who were poor in 1994 in this study, those who had a shorter duration of mental illness and received medication were more likely to escape poverty.
The results of this study suggest the following measurements and interventions for reducing poverty among persons with SMI. First, early diagnosis and treatment should be crucial for reducing long-term poverty status of persons with SMI. Although China had lunched free medication programmes for persons with SMI (Patel et al., Reference Patel, Xiao, Chen, Hanna, Jotheeswaran, Luo and Yu2016), there were still several barriers for them to approach timely medication. Insufficient awareness of SMI, stigma towards mental illness, fear on side effects of medication were common, especially in less-developed areas (Yu & Ran, Reference Yu and Ran2019; Zhang et al., Reference Zhang, Wong, Yu, Ni, He, Bacon-Shone and Ran, M.S.2018). Health policy should pay efforts on public education, thus to improve public awareness of SMI and reduce stigma towards it. Meanwhile, special education towards persons with SMI and their caregivers are also needed to enhance medication adherence (Ran, Chan, Ng, Guo, & Xiang, Reference Ran, Chan, Ng, Guo and Xiang2015). Second, poor work ability and severe mental disability were significant factors that had kept persons with SMI in poverty status or driven them into it, further policies should be developed to strengthen community mental health services and improve the social function of persons with SMI (Ran et al., Reference Ran, Xiao, Fazel, Lee, Luo, Hu and Chan2020; Yu & Ran, Reference Yu and Ran2019). For those who are fully or partly capable to work, work training and anti-stigma interventions should be conducted to enable them to work as others (Yu et al., Reference Yu, Peng, Bai, Luo, Yang, Li and Ran2020a, Reference Yu, Luo, He, Yang, Liu, Guo and Ran2020b; Zhang et al., Reference Zhang, Wong, Yu, Ni, He, Bacon-Shone and Ran, M.S.2018). In the Chinese context, at least, social welfare enterprises should first lower the threshold for persons with a mental disability, thus allowing them to earn incomes and live a valuable life within the social protection system (Song & Liu, Reference Song and Liu2015). Third, physical health condition and its deterioration were also predictive for chronic poverty. Although the treatment of SMI has been covered by specialized programmes (Patel et al., Reference Patel, Xiao, Chen, Hanna, Jotheeswaran, Luo and Yu2016), persons with SMI may still face high costs for treatment of comorbidities (e.g. hypertension, diabetes, etc.), thus health insurance should be further refined to lower their treatment burdens on physical illness. Last, education was critical to shape one's poverty transition status, even for those with SMI. Although specialized educational programmes have been established for disabled persons in China, the current one-size-fits-all framework is more applicable for physical disabilities rather than mental disability (Xiong et al., Reference Xiong, Yang, Yu, Hou, Li, Li and Jiao2011). The particularity of SMI and mental disability should be seriously considered in further promoting special education for those with SMI.
There are some limitations to this study. First, this study applied a follow-up dataset with a large time span of over 20 years, the sample size (n = 308) was limited to perform more sophisticated statistical analysis. Meanwhile, even if it was known that a person was poor at both time points, this did not mean that they were poor in the years between. This was a common limitation in examining poverty dynamics (Addison, Hulme, & Kanbur, Reference Addison, T., Hulme, D., & Kanbur2008). Second, there may be considerable variations due to measurement error between two rounds of surveys. The survey context varies and the data quality may improve or deteriorate (Grosh & Glewwe, Reference Grosh and Glewwe2000). Third, poverty measurement in this study was based on self-reporting of annual household income per capita. Further refined study should add asset measurement to increase the precision of poverty measurement. Besides, apart from the characteristics of persons with SMI, other factors may also contribute on shaping the poverty measurement, such as the male ratio of a household, educational attainment of other family members and the family structure (Rahman, Reference Rahman2013). Household-level data were limited, thus it may potentially magnify the role of SMI in shaping long-term poverty transitions. Fourth, although the follow-up rate was high, the mortality rate of persons with SMI was also very high in this study which may limit the inferences that can be made. Fifth, this study followed persons with SMI only. Without follow-up data on the general population, it was limited to show whether getting poorer alongside the ageing process was more marked among persons with SMI.
Overall, in China, as well as many other developing countries, mental health problems have been largely ignored or marginalized in social and health policy frameworks. Both fiscal investments and the treatment rate for physical health problems are much higher than for mental illness (Yu & Ran, Reference Yu and Ran2019). Even in China's targeted anti-poverty frameworks for the disabled, physical disabilities are emphasized more than mental disability (Xiong et al., Reference Xiong, Yang, Yu, Hou, Li, Li and Jiao2011). This study reveals how SMI has resulted, at least deteriorated poverty for persons with SMI within a period of social prosperity. It should be noted that, in collective societies which highlight the role of family on patient care, the negative impacts of SMI may also occur at the household level (Hulme, Moore, & Shepherd, Reference Hulme, Moore and Shepherd2001). Having a family member with SMI is associated with enormous costs of treatment, caregiving and income loss (Patel et al., Reference Patel, Araya, Chatterjee, Chisholm, Cohen, De Silva and Van Ommeren2007). Because of this, households with persons with SMI are less likely to benefit from the trickle-down effect of external social development (Ran et al., Reference Ran, Weng, Liu, Zhang, Yu, Peng and Xiang2019; Yu et al., Reference Yu, Peng, Bai, Luo, Yang, Li and Ran2020a, Reference Yu, Luo, He, Yang, Liu, Guo and Ran2020b) and are more likely to face chronic economic challenges (Hudson, Reference Hudson2005). Currently, China has a national strategic aim of targeted poverty alleviation (Li, Su, & Liu, Reference Li, Su and Liu2016). However, no targeted anti-poverty plans have been formulated for persons with SMI, as well as for families that provide long-term supports for them. Based on the long-term evidence of social drift revealed in this study and current research on SMI prevalence (Huang et al., Reference Huang, Wang, Wang, Liu, Yu, Yan and Lu2019), formulating intervention plans and policies to alleviate poverty, especially relative poverty, for this unique group may be crucial for achieving China's overall goal of poverty alleviation.
Data
The data are not publicly available due to the information that could compromise the privacy of research participants.
Acknowledgements
The authors thank all the collaborative institutes (e.g. Chengdu Xinjin Second People's Hospital, Chengdu Mental Health Center, Sichuan Veterans Hospital, Jingzhou Mental Health Center, Guangyuan Mental Health Center, Santai Mental Health Center, Sichuan College of Traditional Chinese Medicine, Southwest Medical University, Chongqing Medical University, Chengdu Mental Hospital, Panzhihua Mental Health Center) and the Chengdu Mental Health Project (CMHP) study group for collaboration, support and data collection.
Author contributions
Ran designed and conducted the study. Ran, Xiang, Luo, Liu and Kuang participated in data collection. Yu conducted the major analysis and wrote the first draft of the paper. All authors participated in the discussions and revisions of the paper.
Financial support
The mental health survey in 1994 was supported in part by a grant from the China Medical Board in New York (CMB, 92–557). The mental health survey in 2015 was supported in part by Seed Funding Programme for Basic Research (HKU, 2014–2016), Seed Funding Programme for Applied Research (HKU, 2014–2016), Strategic Research Theme (SRT): Contemporary China Seed Funding (HKU, 2014–2016), and Mental Health Research in Chengdu, China (Dept. Matching Fund, 2015–2017).
Conflict of interest
None.
