Service Utilization for Mental Problems in a Metropolitan Migrant Population in China

A multistage, stratified, randomized sample was extracted from nearly eight million citizens (age 18 or older) who had resided in Shenzhen for at least six months. Because census registration confers more social welfare benefits than does nonregistration, different data extraction methods were used for the two groups.

On the basis of previous studies (27–29), we speculated that the prevalence of mental disorders in the Shenzhen migrant population was 10%–20%. By taking 10% as the expected prevalence, allowing an error margin of 10%, and assuming a significance level of α=.05, we determined that approximately 3,600 residents would be required for each group, according to the formula of Machin and colleagues (30). Furthermore, the sample size should allow for a follow-up loss of approximately 20%. Thus the ideal sample size was determined to be approximately 9,000, with 4,500 cases per group (registered and unregistered).

Among the 561 communities in six regions of Shenzhen, 100 communities were randomly selected. A total of 45 households were sampled randomly in each community from the registered population management system database of the Shenzhen Public Security Bureau. In each household, one adult member was randomly selected via a Kish random-digits table (31). Because most unregistered residents work in factories, 52 companies were randomly sampled from the total of 3,671 companies that employed more than 500 workers and were registered with the Shenzhen Municipal Bureau of Labor and Social Security. From each sampled company, 85 unregistered workers were randomly selected.

In total, 8,920 residents (4,500 registered and 4,420 unregistered) were selected, of whom 7,134 completed the survey (response rate: 79.5% for registered and 80.4% for unregistered). According to the quality control assessment for the interview, 90 returned surveys were excluded for incompletion or inappropriate completion. Thus 7,044 returned surveys were eligible for the analysis. The Shenzhen Mental Health Survey and its administration in this study were approved by the Human Subjects Committees of Shenzhen Administration of Science and Technology. All respondents provided their written informed consent for the survey.

Using the chi square test, we compared lifetime mental health service use among participants with different socioeconomic and demographic characteristics. Moreover, we examined the prevalence of lifetime mental disorders and mental health service use according to service sector and lifetime mental disorder. The difference between use of non–health care services and use of health care services among different populations was also compared by using the chi square test. For differences in the probabilities of selection for part 2 respondents, data were weighted to estimate service use by the total sample. Logistic regression analysis with the backward stepwise method and Wald chi square statistic were used to study the predictors of lifetime mental health service use in the population. Statistical significance was evaluated by using two-sided design-based tests, and .05 was determined as the significance level. All analyses were conducted with SPSS, version 13.0.

After the data were weighted, we found that 9.0% of respondents reported using a service for mental issues in their lifetime and 4.5% had done so in the previous year (Table 1). CAM was the most often used service (5.7%), and 3.4% of those using any service had ever used a mental health specialty service. During their lifetime, 6.3% of respondents had used non–health care services, a proportion significantly higher than the 4.6% that used health care services. In most sectors, unregistered residents used services no more than did registered residents. People age 30 and over visited psychiatric professionals more often than did younger people. However, use in the non–health care service sector was significantly higher among those younger than 30. Married or cohabiting persons had lower rates of lifetime service use than those never married or previously married. Compared with nonreligious respondents, those who followed a religion used services more often in most sectors. People with more than 13 years of education and those with high per capita family income used nonpsychiatrist services, human services, and CAM significantly more often than others.

In this study, the lifetime incidence of any disorder in the population was 18.1% (Table 2). Major depressive disorder (5.7%) and obsessive-compulsive disorder (5.1%) were more prevalent than other specific disorders. Anxiety disorders were reported most frequently (9.0%), followed by mood disorders (8.8%). Psychotic disorders were the least prevalent (2.7%).

Among the respondents with a lifetime mental disorder, 18.3% used any service in their lifetime and 9.2% used services in the previous 12 months (Table 2). These proportions were 6.9% and 3.3%, respectively, in the group with no disorder. Among respondents with diagnosed disorders, 13.0% used non–health care services and 10.9% used health care services. CAM was the most often used service by all respondents. Among those with diagnosed disorders, CAM service use ranged from 9.7% of those with intermittent explosive disorder to 16.7% of those with social phobia. Mental health specialty services were more often used by respondents with bipolar disorder (13.1%) and psychotic disorder (12.9%) than others. The percentages of any service use over the lifetime by respondents with disorders ranged from 16.7% (depressive disorder not otherwise specified) to 24.5% (psychotic disorder). Service use within the previous 12 months ranged from 7.8% (intermittent explosive disorder) to 14.2% (psychotic disorder).

Rates of lifetime service use were significantly associated with various factors: unmarried status, high family income, coming from a metropolitan area, mother deceased, history of mental problems of female caregiver, history of homelessness, history of attempted suicide, and diagnosis of psychotic disorder or anxiety disorder (Table 3).

No significant correlations were found between lifetime service use and any sociodemographic variables except marital status, family income, and hometown type. Registration status, age, gender, years of education, years since migrating to Shenzhen, employment, and religion did not affect the results of the logistic regression analysis.

Sampling bias is a possibility in this study because much of the sample was drawn from census registration data, which exclude residents who are unemployed or unregistered. In addition, since 20% of the sample were nonrespondents, there is another possibility of bias because nonrespondents may also use services to treat mental health problems. The next limitation concerns the CIDI-3.0. Because of the possible differences in expressing mental health symptoms in the Chinese population compared with other populations, we cannot assess the validity of part 2 of the CIDI-3.0 for our sample without verification data on service use, although the CIDI-3.0 has been used in more than 20 countries (32). Also, the CIDI-3.0 is based on self-report, which may reduce the accuracy of the diagnosis of mental disorders, particularly psychotic disorders. Potentially faulty recall of mental health service use may have occurred (35,36), and some of the participants could have used mental health services before migration. Psychiatric problems at the time of the interview may also have created difficulties during the interview process. In addition, the Chinese translation of the CIDI-3.0 excluded certain DSM-IV mental disorders, such as schizophrenia, substance use disorders, posttraumatic stress disorder, and eating disorders. Therefore, some respondents classified as having no disorder may actually have had one. All of these factors may have resulted in an underestimation or overestimation of service use by this population.

On the basis of our results and facts concerning the census composition of the Shenzhen migrant population in 2005, we speculated that 10.9% of this population had used mental health services within their lifetime and 6.1% of them had done so within the previous year. Between 2001 and 2004, a series of similar surveys using the CIDI-3.0 was initiated by WHO in 17 countries (24). Persons in the Chinese sample were from the registered population of Beijing (N=914) and Shanghai (N=714). Their results indicated that 3.4%±.6% (mean±SE) of Chinese respondents had used services within the previous year (23). Our figure was 5.1% (Table 1). Another large survey in four Chinese provinces suggested that 7.3% of the registered urban population had used services within their lifetime (22). In the study we report here, the figure was 12.2%±.9%. These results may imply a greater mental health care need in the migrant population than in the native population.

Previous studies of various populations have reported that the health care service sector, compared with non–health care services, is being used most often for mental health issues (24). In this study, however, services from the non–health care sector, especially CAM, were more popular. The high proportion of CAM use in Chinese populations has been verified in previous research (15). The lack of mental health professionals and small number of mental health care settings in Shenzhen may have influenced this result.

Of those with mental disorders as identified by the DSM-IV, 18.3% had sought services for a mental health problem. Of the service users in this population, approximately 33% (3.3% of 9%) had some mental disorder. Similar to the results of Phillips and colleagues (22), our results showed that respondents with psychotic disorders were the most frequent users of any services among the diagnosis groups.

As noted above, in Shenzhen, registration brings certain social benefits such as medical insurance, compulsory education for children, and disability insurance. However, registration was not significantly correlated with service use within either the previous year or the respondent’s lifetime. A possible explanation is that the social environment and poor health service system have had little influence on migrants’ personal service use behavior for the short history of this migrant population.

Sociodemographic variables, such as age, gender, education, and employment status, have been found to be associated with mental health service use (37–42). However, we could not verify this among Shenzhen migrants. Religious beliefs may have encouraged greater care seeking (43), whereas marriage had a significant dampening effect on service use, which has been verified by Western surveys (37,38). High family income was associated with increased service use, whereas low family income limited the ability to pay for such services. Migrants coming from metropolitan areas were more likely than those from rural areas to use mental health services. Homelessness, attempted suicide, and diagnosis of a psychotic disorder or an anxiety disorder may have increased the respondents’ need for mental health services.

Caregiver’s gender was also associated with service use in this population. Respondents with deceased mothers or who had female caregivers with troubled mental histories had an increased possibility of lifetime service use. Previous studies found that having parents with chronic disabilities increased service use (44), and better relationship support was associated with decreased use of specialty mental health care services (45). In most Chinese families, mothers perform the bulk of caregiving (46). Therefore, mothers may have greater influence than fathers on their offspring’s mental health. Migrants to Shenzhen had poor relationship support, and the absence of the mother-child relationship may have additional impact on their mental health. However, gender differences observed in the responses require further investigation in order to clarify the associated factors.

Among the variables above, personal characteristics seemed to play a leading role in service utilization. Future studies of service utilization patterns in migrant populations should give special attention to personal characteristics, such as family support.