Analysis of health service utilization of migrants in Beijing using Anderson health service utilization model

On the basis of social systems approach, as a reliable tool, with decades of empirical support, Andersen health service utilization model is a well-validated theoretical framework aimed at understanding determinants of health services utilization, which takes into account both societal and individual determinants from the perspective of systematic analysis [19]. According to the model, health services utilization (including outpatient care, inpatient care, etc.) is determined by three dynamics: predisposing, enabling, and need variances (PEN). Predisposing factors can be social demographic characteristics such as age, race, sex, status which increase one’s need for health services. For instance, an individual who believes health service is an effective treatment for an ailment is more likely to seek care. Examples of enabling factors could be family support, access to health insurance, one’s community, etc. Enabling factors facilitate or impede use of health service. Need for care represents both self-perceived and actual need for health service.

This model is further differentiated from its predecessors by using a feedback loop to illustrate that health outcomes may affect aspects such as needs (See Fig. 1). Our outcome variables was the utilization (seek health services, and non seek health services) of health services in the past month. We hypothesize that the PEN variables will be significant predictors of health services use, and that need variables will be the strongest predictors.

The study was undertaken as a part of the National High Technology Research and Development Program-863 of the Chinese Government, which is a population-based cross sectional survey on risk factors of chronic non-communicable diseases (NCD). It was approved by the Ethical Committee of Capital Medical University, Beijing, China. Data were obtained from a 2012 cross-sectional survey in urban Beijing. Written informed consent was obtained from each participant involved in this study. For participants under the age of eighteen, written informed consents were obtained from their guardians. All participants’ information was kept confidential and tracked anonymously with identification number only.

The questionnaire was designed according to the contents of the Sixth China National Health Service Survey, and divided into four parts of contents as following: predisposing factors (demography and social structure), enabling factors (individual/family resources and community resources), need factors (general health condition), and health-seeking behavior (e.g., do nothing, self-treatment, and go to hospital). Fieldwork was performed from March to May, 2012. The quality assurance measures for this investigation included evaluating the questionnaire, training the investigators, and asking a fieldwork supervisor to monitor the investigation process. Before the investigation was implemented, it was reviewed, edited, and validated by experts from health administrative department, clinical teaching hospital and community health service institutions (CHSIs). The completed questionnaires, and validity of the collected data were independently checked. A pilot survey covering 400 persons was conducted between February 26 to 29, 2012 to assess applicability of the questionnaire and the fieldwork procedures. Beijing urban residents dwell mainly in three out of four areas were classified according to the functions of the county namely capital function core area (Dongcheng and Xicheng distircts), city functional expansion area (Chaoyang, Haidian, Fengtai and Shijingshan districts), urban development zone (Tongzhou, Shunyi, Daxing, Fangshan and Changping districts) and ecological conservation development area (Mentougou, Pinggu, Miyun, Yanqing and Huairou districts). A sample of 7000 adults was chosen from 5 of 16 counties in Beijing by using three-stage stratified and cluster random sampling. In the first stage, one county was selected from the capital functional core area (Dongcheng district), two counties were selected from the city functional expansion area (Haidian and Fengtai districts), and another two counties were selected from the urban development zone (Daxing and Shunyi districts) (See Fig. 2). In the second stage, two streets according to the economic condition and population size from each sampled county were selected. Finally, for each street selected, the total number of 7000 adults was recruited and investigated. The population was sampled in Dongcheng (1000), Haidian (2000), Fengtai (2000), Daxing (1000) and Shunyi (1000) respectively. All respondents were approached for face to face interview. Finally, 6592 out of the 7000 respondents were analyzed after excluding those whose information was missing on any of the study variables in the study. All data adopted double entry by using EpiData software (Version 3.1, EpiData association, Odense, Denmark). The two databases were compared and analyzed for discrepancies. If discrepancies exist, the original data source would be analyzed.

A person-month was the unit of indicated receipt of services in a health setting at least once in a month, not the total number of times care received. Associations between categorical variables were tested using chi-square independence and binary logistic regression analysis. First, for the dataset, descriptive statistics and chi-square independence test stratified by nationality were calculated. Second, the binary logistic regression was performed in full model (considering all of the predisposing characteristics, enabling resources and the need factors were entered into the model, resulting in model I for migrants and model II for residents with “Hukou”), to estimate the potential major determinants that affect the patterns of health services utilization among two groups population, controlling for possible confounding associations on health services utilization. The dependent variable has two categories, including seek health services and non seek health services (self-medication or do nothing). The list of independent variables is summarized in the Table 1. All analyses were conducted using the IBM Statistical Package for Social Science software program (IBM-SPSS) Version 20.0 for Windows and all the tests are two sided. P < 0.05 was considered statistically significant, and the results were expressed as odds ratios (OR) with 95% confidence intervals (CI).

The study revealed that 503 migrants (25.0%) and 1441 (31.5%) residents with “Hukou” reported at least one episode of discomfort in the past month. In the group of 25 to 34 years old, the rate of having symptoms of migrants was lower than that of residents with “Hukou” in the past month, but difference was not identified in other age groups (See Table 2).

Table 3 indicates the differences of therapy measures selected between two groups population when they have symptoms in the past month. The rate of seeking health services one or more times was 46.7% for migrants as compared to 62.7% for residents with “Hukou”. The proportion of taking self-medication as well as taking nothing therapy measures were higher in migrants (24.3 and 29.0%) than residents with “Hukou” (15.2 and 22.1%).

Descriptive statistics and chi-square independence test were used to analyze the difference in two groups of population by socio-economic and demographic factors. Chi-square independence test showed that age, ethnicity, employment status, having chronic disease and the degree of symptom were the major determinants affecting migrants to receive health services. For the residents with “Hukou”, the main influence factors included gender, age, ethnicity, marital status, employment status, education level, incoming, having chronic disease, self-evaluation of health and the degree of symptom in the past month (See Table 4).

The finding reveals that the full models for migrants and residents with “Hukou” containing all predictors were statistically significant, (model I, χ2 = 41.059, p = 0.000; model II, χ2 = 190.112, p = 0.000), indicating that the models was able to distinguish between respondents who seek health services and those who do not to seek health services after their having symptoms.

In the model summary, the model I as a whole explained between 7.8% (Cox and Snell R square) and 10.5% (Nagelkerke R square) of the variance in health services utilization of migrants was explained by all the predictor variables entered into the logistic regression model. The model II as a whole explained between 12.4% (Cox and Snell R square) and 16.9% (Nagelkerke R square) of the variance in health services utilization of residents with “Hukou” was explained by all the predictor variables entered into the logistic regression model (See Table 5).

Tables 6 and 7 predict the determinants of health services utilization for migrants (model I) and residents with “Hukou” (Model II) by binary logistic regression. ModelIin Table 4 shows the migrants with moderate degree and severe degree of symptom in the past month were at 1.623-times (OR = 1.623) and 5.035-times (OR = 5.035) higher chances of seeking health services respectively, comparing to the mild degree of symptom. Minority migrants were less likely to seek health services than Han migrants (OR = 0.282, 95%-CI 0.128, 0.621, P < =0.002). Model II in Table 5 indicates that for the residents with “Hukou”, female (OR = 1.314), age above 55 years old (OR = 2.253), having chronic disease (OR = 1849), with moderate and severe degree of symptom in the past month (OR = 2.032 and OR = 5.786), were more likely to seek health services. Additionally, the chances of seeking health services were decreased by 33.3% in moderate health status (OR = 0.667) and 38.2% in poor health status (OR = 0.618), comparing to self-evaluated good health status.

Consistent with the theory of “healthy migrant effect” that first-generation migrants are often healthier with lower overall morbidity and mortality than local-born populations [20], our study also find that the recent one-month prevalence rate of illness among migrants was lower than residents with “Hukou” without classified by age group. Subsequently, a large number of studies find that the migrant health advantage diminishes dramatically especially in middle age [21, 22]. In our study, migrants had lower prevalence rate of illness than that of residents with “Hukou” in the group of 25 to 34 years old, but the distinction was not revealed in other age groups. Meanwhile, the recent one-month physician visit rate in the past month among migrants was lower than residents with “Hukou”. It remains a concern that migrants are often over-optimistic about their health status due to the poor understanding of health which is induced by the socioeconomic environment they live and work in [23].

In the study, the finding validated our hypothesis that need factors (having chronic disease, self-evaluation general health status, and the degree of symptom) were the most dominant predictors of health services utilization especially for residents with “Hukou”, even when adjusting for predisposing and enabling variables. Migrants in the moderate and severe status of symptom have higher likelihood of seeking healthcare services. There might be some factors account for the differences. Firstly, these results might be due to the fact that the migrants are relatively young, i.e. nearly half of respondents were aged between 15 and 35 in this and other studies [8, 24]. Selection bias in migration indicated that those who voluntarily migrate tend to have healthier lifestyles and lower prevalence rate of illness than either their urban counterparts born in host countries or their non-migrating peers in the original places [20]. Secondly, under the pressures of heavy working, younger migrants always have a satisfactory level of self-reported health risk awareness, and were more inclined to ignore the health problems than their urban counterparts [11]. Furthermore, considering high health expenses, along with deficiency in health insurance for lowering health risks, they might be more likely to take regular medication or palliative care instead of seeking healthcare services in medical institutions when they only have minor illness. Finally, most of migrants would return to their hometown to treat severe illness or enjoy old age, from an economical perspective [25]. For the above reasons, most migrants made use of health services only when they were in serious health status.

For migrants, we observed that ethnicity of predisposing variables contributed significantly to the variance in health services utilization. Ethnic minority migrants are less likely to use health services than Han migrants, and the result was consistent with the previous research of Guangxi [26]. The possible explanations could be divided into two aspects, one is that ethnic minority migrants mainly rely on relatives, friends, or fellow villagers to acquire resources in urban communities. Social assistance networks were rarely provided by organizations, such as grassroots government, community neighborhood committees, enterprises, civil organizations [27]. On the other hand, special religious beliefs might affect their health service seeking behaviors. They might pray to god or spirit for relieving their sufferings, or take special herb medicine to cure or alleviate their illness instead of seeking health services [28]. Taking into account the sample size of minority selected is small in this study, in the future, ethnic minority migrants should be studied in terms of their health service utilization behaviors and influencing factors as a unique group.

In this research, the result was different from some previous China-based studies which indicated that insurance participation increases the likelihood of seeking health services in the general population [29]. Comparing to residents with “Hukou”, less migrants had insurance in Beijing. The root cause of this phenomenon is the household registration system (Hukou) in China, which divides the population into rural and urban citizens [30]. Although the “Hukou” system is no longer used to prevent the flow of migrants, the obstacles for the majority of interprovincial migrants acquired permanent urban residency rights and many of the associated social welfare still existed. China’s government launched the policy of “New Rural Cooperative Medical System” (NRCMS) for rural “Hukou” residents to provide basic healthcare in 2003. By the end of 2013, 98.7% (802 million) of the rural population were covered by the NRCMS. There is no doubt that NRCMS improved both hospitalization and outpatient services utilization to prevent catastrophic health payment in rural China, yet it was useless for migrants with rural “Hukou” which account for 80% of the total migrants in immigration cities [31, 32]. Many migrants had insurance in their hometown, and it is nearly impossible to transfer insurance relations between different provinces in China. Thus, they need to return to hometown to reimburse their medical expense. Although “migrant worker insurance” was launch in 2006, it covered the migrant workers with stable employment relation with urban employers, and was not applicable for the self-employed or the unregistered migrants [33]. Working without a contract is common to labor migrants in most industries in majority immigrant countries, which often can prevent privilege to the typical national or local health insurance, support services or social assistance [34‐36]. Lower health insurance coverage and complicated reimbursement procedure increased the barriers to access health services for migrants. For the reasons mentioned above, health insurance is not a determinant to health services utilization for migrants.