Appl Health Econ Health Policy DOI 10.1007/s40258-014-0081-1

ORIGINAL RESEARCH ARTICLE

Healthcare Use and Voluntary Health Insurance After Retirement in Thailand Papar Kananurak

 Springer International Publishing Switzerland 2014

Abstract Background The dramatic changes occurring in the age structure of the Thai population make providing healthcare services for the elderly a major challenge for decision makers. Because the number of the elderly will be increasing, together with the number of retired workers, under the Social Health Insurance (SHI) scheme, there will be the unmet needs for healthcare use after retirement. The SHI scheme does not cover workers after retirement unless they could use free healthcare for the elderly. In addition, the government budget is tight regarding the support of universal healthcare and long-term care services for all of the elderly. Therefore, the government could support retired workers who have the ability to pay by facilitating voluntary health insurance. Objective The main objectives of the present study are to analyze the characteristics of workers that need health insurance after retirement and to identify the factors explaining healthcare use to offer healthcare services to meet the workers’ needs and expectations. Methods Four hundred insured workers under the Social Health Insurance (SHI) Scheme in Thailand were interviewed using a structured questionnaire. The Anderson– Newman model of healthcare use is the conceptual framework used in this study to understand the factors that explain healthcare use patterns of workers. Multiple

regressions are employed extensively to evaluate the variables that predict healthcare use. Results According to the survey, a person that purchases voluntary health insurance is likely to be female, have a higher personal income, and healthy. The characteristics related to healthcare use were poor health status, a high personal income, and peeople afflicted by chronic illness. Conclusions There is a gap between healthcare service use and the demand for voluntary health insurance. People that have a high income are more likely to purchase voluntary health insurance, while people in worse health and afflicted by chronic illness may have greater difficulty purchasing voluntary health insurance because they face higher premiums or are denied coverage by insurers.

Key Points for Decision Makers It was found that most variables that were defined to be important in explaining healthcare use were also important in explaining the demand for voluntary health insurance. The workers in poor health or those that anticipate future poor health are likely, but not certain, to consume more medical care in the future than those in better health, and there is high risk for this group of workers after retirement.

Electronic supplementary material The online version of this article (doi:10.1007/s40258-014-0081-1) contains supplementary material, which is available to authorized users. P. Kananurak (&) Business Economics, Martin De Tours School of Management and Economics, Assumption University, Suvarnabhumi Campus, Bangsaothong, Samuthprakarn 10540, Thailand e-mail: [email protected]

1 Introduction The rapid aging of the population, especially among the elderly population, has challenged governments to address

P. Kananurak

ways to serve the growing demand of the elderly in terms of the use of institutional healthcare services. Healthcare services for the elderly will increase markedly as the number of elderly people and the number of elderly people with disabilities and chronic illness increase. In addition, as the baby boomers enter retirement, the need for healthcare will greatly increase, turning current problems into a crisis if the governments are not well prepared [1]. Because the higher rate of chronic illness among the elderly will influence the healthcare use pattern, the understanding of healthcare use patterns relative to chronic illness such as diabetes mellitus, hypertension, high cholesterol, and cardiovascular diseases is useful for those planning and developing health service programs for older persons. Thailand has now reached a new demographic turning point with the advent of an aging society [2]. The older population or population aged 60 years and above increased from 1.5 million in 1960 to approximately 7.4 million in 2008, and it is expected to be 17.7 million in 2030 [3]. Furthermore, not only is the overall share of the aging population increasing, but also the increases are large and rapid in the extreme age group of 80 years and older [4]. From the year 2000 to 2050, the growth rate of the older adult population is projected to increase by 227 % while the elderly population (age 80 years and over) will increase by 686 % [2]. As the numbers of the older population that need healthcare and long-term care services increase, public health insurance should be adjusted and reformed to meet the demands of people after retirement. Currently, there are three public health insurance schemes in Thailand: the Social Health Insurance (SHI), the Civil Servant Medical Benefit Scheme (CSMBS), and the Universal Coverage (UC) Scheme. The UC Scheme is mandatory for all Thai citizens that are not insured by another public insurance scheme. The CSMBS covers civil servants and their immediate family members, including spouses, parents, and up to three children under the age of 20 years. It also covers retirees and their dependents. The SHI Scheme provides mandatory coverage for workers in the formal sector: since 1991, when it was applied to firms with more than 20 workers, its coverage has increased and, since 2001, it is mandatory for firms with more than one worker and for the self-employed [5]. In Thailand, the government has provided free healthcare for older persons since the introduction of universal health coverage in 2001. Under this program, all government hospitals and health centers provide free medical services to persons aged 60 years and upwards. However, only the elderly that are poor and under the UC Scheme can use the free care [6]. People under the CSMBS still receive healthcare services after retirement, whereas the SHI Scheme does not cover workers after retirement unless they can use the free healthcare for the elderly [7].

Because the number of insured workers in Thailand increased significantly, from 3.2 million in 1991 to 10.5 million in 2011, the number of retired workers under the SHI Scheme will increase accordingly to 476,000 by 2016 and 1.2 million by 2121 [8]. In general, the price of longterm care and healthcare services for the elderly is expensive. While being uninsured at any age is risky, older persons without adequate health coverage are at particular risk of experiencing adverse health events from skipping needed care, spending large shares of their income on outof-pocket costs, and accumulating medical debt. However, the Thai government budget is tight regarding the universal healthcare support and long-term care services for all of the elderly. The potential support ratio, which measures the ratio of persons aged 15–59 years to persons aged 60 years and older, is very dramatic, and will fall from 7.0 in 2000 to only 2.4 in 2030 [7]. This means that older-age Thai individuals in the future will have far fewer productive age persons per capita available to provide for their support. In addition, the healthcare for the elderly financed through taxes is not sufficient to pay for the UC Scheme. Therefore, the government could support retired workers who have the ability to pay by facilitating voluntary health insurance. However, to develop a healthcare program that meets retirees’ needs, the healthcare use of workers and the characteristics of workers that need health insurance after retirement should be understood. This understanding will help to define the unmet need for health use for the elderly. The target sample of the present study is Thai workers aged over 40 years under the SHI Scheme. Therefore, this study will analyze the characteristics of those that have private health insurance as a basis for identifying those that will probably take up voluntary health insurance after retirement. In addition, the factors explaining the healthcare use are analyzed to offer healthcare services to meet the workers’ needs and expectations. The findings are intended to be useful for those planning and developing health insurance after retirement for workers under the SHI Scheme that want better-quality healthcare services. Regarding healthcare use, one of the most frequently used frameworks is the behavioral model developed by Anderson and Newman in 1973 [9]. This model was initially developed in the late 1960s to assist in the understanding of why families use health services; to define and measure equitable access to healthcare; and to assist in developing policies to promote equitable access [10]. It assumes that a sequence of conditions contributes to the type or volume of health services a person uses. Use is dependent on: (1) predisposing factors: the predisposition of the individual to use services; (2) enabling factors: his/ her ability to secure services; and (3) need factors: illness factors [11]. Because Andersen and Newman’s model has

Healthcare Use and Insurance in Thailand

been mainly used for explaining healthcare use [11], this study will apply their behavioral model as a framework to understand the factors explaining the healthcare use pattern of workers in Thailand. Furthermore, the market for health insurance for the elderly is still in its early stage of development. The market for voluntary health insurance after retirement, which is managed by a government authority, does not exist. Factors that determine traditional purchasing behavior of private health insurance may be strongly relevant to factors that determine purchasing behavior of voluntary health insurance after retirement. Therefore, the demand for private health insurance was applied in this study as a proxy for the demand for voluntary health insurance. Then, logistic regression analysis was employed to measure the strength of the association of various demographic and socioeconomic characteristics and health status with the demand for voluntary health insurance. These will provide insight into the future demand for voluntary health insurance after retirement in Thailand.

2 The Background of Healthcare and Public Health Insurance in Thailand The details of ‘The Background of Healthcare and Public Health Insurance in Thailand’ are given in Online Resource 1.

3 Methods 3.1 Survey Instrument The survey was conducted during June, 2011 to December, 2011. A face-to-face interview using a structured questionnaire was used in this part of the study. The questionnaire contains demographic and socioeconomic information, and information on health status, current health insurance, health risk perception, and attitude towards the future. The questionnaire is given in Online Resource 2.

Table 1 The number and proportion of insured persons under the Social Health Insurance Scheme in 2010 Number of insured persons

Population

Population (%)

Bangkok

3,203,523

33.6

Vicinity

1,944,127

20.4

Central

2,178,932

22.9

North

724,116

7.6

Northeast

795,048

8.4

South

674,438

7.1

Total

9,520,184

100.0

Source: Social Security Office, 2010 [8]

Therefore, the sample of 400 insured workers under the SHI Scheme was carried out with the assumption that there is no difference in opinion about health insurance after retirement between persons in Bangkok and vicinity, and the central, north, northeast, and southern regions of Thailand. The population was classified by area based on the size of workers, under the SHI Scheme. Based on the size of workers in proportion to the area of population, the provinces from each area were randomly selected by systematic sampling. After selecting the provinces, non-probability sampling, that is, purposive sampling, was used to determine the characteristics of the 400 samples across various provinces. In purposive sampling, the respondents should be aged between 40 and 69 years and should have or previously had SHI. In addition, they should have an income greater than 10,000 Baht per month. Therefore, a screening question in the structural questionnaire was used to ask the respondent to check his/her criteria. If the respondents passed the screening criteria, the interviewers asked them further questions about demographic and socioeconomic information, personal health status, their current health insurance contract, and their preference choice sets regarding health insurance benefits. The respondents that did not pass the criteria were screened out and the interviewer stopped the questions, and then thanked the respondents for participation.

3.2 Sampling Design and Procedures 3.3 Statistics Method The target population in this study is Thai workers that are under the SHI Scheme. Data from the Social Security Office, the Ministry of Labor and Social Welfare, showed that in 2010, Thailand had about 9.5 million insured persons under the SHI Scheme. Bangkok had the highest share of insured persons at 33.6 % of the total insured population, followed by the central region of Thailand (22.9 %) and the vicinity of Bangkok (20.4 %) [12]. The number and proportion of insured persons under the SHI Scheme are presented in Table 1.

This study examined the different characteristics between the hospitalization and non-hospitalization of respondents. Descriptive statistics were initially used to identify the various variables regarding socioeconomic and demographic information, living location, family size, number of children living with the family, health status, the acquisition of a private health insurance contract, life expectation, healthcare use (times per year), and health service expenses (Baht per time). In addition, health risks were included in this model.

P. Kananurak

The health risks were grouped into three groups: high, medium, and low. The high-risk group of respondents consisted of those that had a chronic illness and that were treated for that illness, while the medium-risk group of respondents comprised those that had a chronic illness but received no treatment or therapy. In addition, this group of respondents had infrequent health checks, rarely engaged in exercise, and often drank alcohol or smoked cigarettes. Therefore, they were at risk of having a chronic illness requiring medical treatment in old age. Those that did not have a chronic illness, regularly had health checks, engaged in exercise, and did not usually drink alcohol or smoke cigarettes were classified as having a low health risk. These significant variables provided greater understanding of the healthcare use of the workers under the SHI Scheme. Furthermore, the main objective of this study was to identify the associated factors explaining the healthcare use of workers under the SHI Scheme. To identify the predictors of healthcare use, the Andersen–Newman model was used. This model of healthcare use has been mainly used for explaining healthcare use by the elderly or by the general public [11]. On the basis of this model, the effects of predisposing, enabling, and need variables were then examined by using the multiple regression method. Theoretically, healthcare use (times per year), which indicates the number of times the event could have happened for the same length of time, is treated as count data. In addition, count data are usually analyzed using Poisson regression [13]. Nevertheless, data appropriate for a Poisson regression are rarely found because the units counted are not independent, and such a model may underestimate the variance. Healthcare use variables are usually not normally distributed, as they tend to have a mode at zero and a distribution with a long, heavily positive skew. The distribution often looks more like a lognormal than a Poisson distribution, even when the data are counts [14]. Furthermore, ordinary least squares regression assumes normality in the distribution of error terms and hence in the dependent variable; its use with this type of data is problematic if the data are not transformed to address the effects of the positive skew [13]. Therefore, this variable was transformed to a logarithmic scale, which usually shortens the log right tail, lessens heteroscedasticity, and decreases the influence of the outliers. For the independent variables such as age, education (years of schooling), and monthly personal income, these variables exhibited high variability in the data. The log transformation compressed both extreme values or spread out the small values of the independent variables. It also could be used to stabilize the variance of a sample. Therefore, a log transform for these independent variables was used to reduce the influence of extreme values and to induce symmetry [15].

The hypothesis was tested by using multiple regression analysis. The main objective was to identify the associated factors explaining the healthcare use of workers under the SHI Scheme. The hypothesis and alternative hypothesis are as follows: Ho: The predisposing, enabling, and need variables such as gender, age, education, marital status, employment, family size, health behaviors, personal income, the acquisition of a private health insurances contract, living location, health status, health risk perception, and the number of chronic illnesses do not explain a significant portion of the variance in healthcare use in Thailand. Ha: The predisposing, enabling, and need variables such as gender, age, education, marital status, employment, family size, health behaviors, personal income, the acquisition of a private health insurances contract, living location, health status, health risk perception, and the number of chronic illnesses explain a significant portion of the variance in healthcare use in Thailand. The multiple regression model is specified as: yi ¼ b0 þ b1 male þ b2 logðageÞ þ b3 logðeducationÞ þ b4 single þ b5 employment + b6 family size þ b7 health behaviour + b8 logðpersonalin comeÞ þ b9 private health insurance þ b10 BKK + b11 health status þ b12 healthrisk perception þ b13 chronicill,

where yi is log of the number of hospitalizations (times per year). The independent variables are as follows: gender (0 = female, 1 = male), log of age (continuous, years), log of education (continuous, years of schooling), marital status (0 = not single, 1 = single), employment (1 = employment with fixed income, 0 = others), family size (continuous, number of persons in household), health behavior (average rating from 1 to 5 of health behavior, 1 being the poorest health behaviors and 5 being the most excellent health behaviors), log of personal income (continuous, Baht per month), private health insurance contract (0 = no, 1 = yes), living location (BKK (Bangkok) = 1, 0 = others), selfhealth assessment (1 = excellent or good, 0 = average or below), health risk perception (1 = high, 2 = medium, 3 = low), and chronic illness (continuous, the number of chronic illnesses). When the regression model was analyzed, statistical tools based on the principle of correlation were used to assess the regression results. The coefficient of determination was particularly important because it represents the amount of variability of a dependent variable explained by an independent variable. The R-squared coefficient provides useful information for researchers in assessing the overall ‘‘goodness of fit’’ of a given regression equation—a higher value indicates a better fit. The adjusted R-squared coefficient is also reported and corrects for the number of

Healthcare Use and Insurance in Thailand

independent variables that are specified in a regression equation when more than one predictor variable is being analyzed. However, the regression analyzes of the use data, the values of R-squared, are usually in the order of less than 20 % [14]. This should not be surprising considering how difficult it would be to predict one’s own use. The low values of R-squared indicate that we cannot predict well for an individual; however, more often we are trying to predict the average use for a group of individuals, and regression equations can often do this quite satisfactorily [14]. In addition, an analysis of the overall F statistic from regression analyses is explored to assess the statistical significance of the regression equation. Subsequently, residual analysis and a multicollinearity test were carried out to test the assumption of the regression model for the least squares analysis. The presence of an interaction implies that the effect of one independent variable on the outcome of interest is influenced by the levels of another independent variable. Based on literature reviews on health use, age and sex are the most common covariates because they are reasonable proxies for a person’s need for services [14]. Verbrugge [16] has noted that sex differences in health and health reporting are principally due to differentials in roles, life chances and styles, and stress and prevention practices. In addition, Galvin and Fan (1975) found that age and chronic illness increase hospitalization and the need for healthcare [17]. Education has an important relationship with health service use; individuals, regardless of sex, report higher levels of health status with higher levels of education [17]. Additionally, better education may lead to better income status, nutrition, and access to healthcare [18]. Marital status affects health services use as well. Married persons feel better, function better, and have fewer illnesses, fewer serious illnesses, better illness outcomes, and less mortality [19]. In addition, families with children benefit more than couples without children, and those with two children benefit more than those with large families or with adult children in the home [20]. To incorporate interaction into the multiple regression model, the product terms of the independent variables (X1  X2 ) were converted into the multiple regression model. There is evidence of an interaction between independent variables (X1 and X1 ) if the coefficient on the product terms of the independent variables (X1  X2 ) is significant. The following interactions were tested: • • • •

Age 9 gender Age 9 chronic illness Education 9 personal income Marital status 9 family size

To evaluate the determinants of the demand for voluntary health insurance after retirement, the demand for private health insurance was used as a proxy of demand for health insurance after retirement. Then, logistic regression analysis was employed to measure the strength of the association between various demographic and socioeconomic characteristics and health status with the demand for voluntary health insurance. The hypothesis and alternative hypothesis are as follows: Ho: The variables of demographic, socioeconomics, and health status, including gender, age, living location, marital status, education, employment, family size, number of children in family, personal income, self-health assessment, and the number of chronic illnesses, do not explain a significant portion of the variance in the demand for voluntary health insurance. Ha: The variables of demographic, socioeconomics, and health status, including gender, age, living location, marital status, education, employment, family size, number of children in family, personal income, self-health assessment, and the number of chronic illnesses, explain a significant portion of the variance in the demand for voluntary health insurance. The logistic regressions is specified as yi ¼ b0 þ b1 male þ b2 logðageÞ þ b3 BKK + b4 single þ b5 logðeduÞ þ b6 employment þ b7 logðincomeÞ þ b8 fmsize þ b9 children þ b10 health þ b11 chronicill,

where yi is the respondents that have a private health insurance contract. The independent variables are as follows: gender (0 = female, 1 = male), log of age (continuous, years), living location (BKK = 1, 0 = others), marital status (0 = not single, 1 = single), log of education (continuous, years of schooling), employment (1 = employment with fixed income, 0 = others), log of personal income (continuous, Baht per month), family size (continuous, number of persons in household), number of children in family (continuous, number of children in family), self-health assessment (1 = excellent or good, 0 = average or below and chronic illness (continuous, the number of chronic illnesses). The effectiveness of the logistic regression model was evaluated from (a) overall model evaluation, (b) statistical tests of individual predictors, (c) goodness-of-fit statistics, and (d) validation of predicted probabilities [21]. The overall significance was tested using what SPSS calls the Model Chi-square, which is derived from the likelihood of observing the actual data under the assumption that the model that has been fitted is accurate. The difference between -2LL for the best-fitted model and -2LL for the intercept-only model is that it is distributed like a chi-

P. Kananurak

squared one; this difference is namely the Model Chisquare. The statistical significance of the individual regression coefficients (Bs) was tested using the Wald Chisquare statistic. The goodness-of-fit statistic assesses the fit of the logistic model against actual outcomes. The inferential goodness-of-fit test is the Hosmer–Lemeshow (H–L) test. If the H–L goodness-of-fit statistic is greater than 0.05, it implies that the model’s estimates fit the data at an acceptable level. The descriptive measures of goodness-offit are the R2 indices, defined by Cox and Snell (1989) and Nagelkerke (1991) [21].

4 Results 4.1 Healthcare Use Tables 2 and 3 present the profiles of hospitalization (including outpatient care [OPD] and inpatient care [IPD]) and non-hospitalization (including OPD and IPD). For those that were hospitalized, the healthcare use was around 2.86 times per year and the health service expenses were around 2,377.23 Baht per time. It was found that workers under the SHI Scheme that were hospitalized tended to be female rather than male, lived in Bangkok, had more children in the family, and had a higher monthly personal income (17,500.50 Baht per person) and also a higher monthly household income (35,000.50 Baht). Interestingly, the number of hospitalizations was more likely to increase with age (age around 48.58 years). The number of hospitalizations seemed to increase with the respondents that exhibited good health behaviors more than average to poor health behaviors because of concern for self-healthcare. Those that engaged in self-healthcare wanted to protect themselves against health risks by frequently going for a check-up at the hospital at the beginning stage of the disease. Furthermore, those that were hospitalized had a high health risk; they expected their life to be shorter than those that were not hospitalized.

the predisposing and enabling factors. It was also found that the log of personal income was statistically significant with a positive sign. Nevertheless, gender, age, marital status, education, employment, family size, living location, private health insurance contract, and health risk were not statistically significant. As a result, the characteristics related to healthcare use were poor health condition, high personal income, and the increasing number of chronic illnesses. As the number of chronic illnesses increases with age, it is important to explore the relation of healthcare use and the most common chronic illnesses of the elderly in Thailand. The top three chronic illnesses of the elderly in Thailand in 2007 were hypertension (31.7 %), followed by diabetes (13.3 %) and cardiovascular disease (7 %) [7]. The most common chronic illnesses, including diabetes, hypertension, high cholesterol, and cardiovascular disease, were substituted in the model. Regarding Table 5, the F statistic shows that at least one of the explanatory variables was not statistically equal to zero—meaning that these variables could explain a significant portion of the variance in healthcare use in Thailand. It was found that health status was statistically significant, with a negative sign, while the log of personal income and the most common chronic illnesses of the elderly, including diabetes, hypertension, cardiovascular disease, and high cholesterol, were all statistically significant with a positive sign. Income and health status still had a strong impact on healthcare use. In addition, those respondents that were afflicted by chronic illness such as diabetes, hypertension, cardiovascular disease, and high cholesterol had to be admitted to the hospital and had to go for a checkup at the OPD regularly; therefore, healthcare use can be seen to be significantly increasing. Nevertheless, gender, age, marital status, education, employment, family size, living location, private health insurance contract, and health risk were not statistically significant. As a result, the characteristics related to healthcare use were poor health status, high personal income, and those afflicted by a chronic illness.

4.2 The Determinants of Healthcare Use 4.2.1 Healthcare Use of Outpatient and Inpatient Care Regarding Table 4, the F statistic shows that at least one of the explanatory variables was not statistically equal to zero, meaning that they can explain a significant portion of the variance in healthcare use in Thailand. It was found that health status was statistically significant with a negative sign and chronic illness was statistically significant with a positive sign. Consequently, those with the worst health status using more healthcare services may come from having a chronic illness. In addition, the need variables seemed to have a greater impact for healthcare use than for

The hospitalization between outpatients and inpatients was different. OPD is the care given to the patient that is not hospitalized for 24 h or more but that visits a hospital for diagnosis or treatment, whereas IPD is the care given to the patient that is admitted to the hospital and stays overnight. Therefore, the healthcare use of the OPD and IPD was analyzed separately to gain more insight into the predictor variables of healthcare use among workers under the SHI scheme in these two types of service.

Healthcare Use and Insurance in Thailand Table 2 Profile of respondents hospitalized and non-hospitalized Profile

Nonhospitalization (%)

Hospitalization (%)

Gender Female

54.4

53.6

46.4

Province

53.1

46.9

Bangkok

40.2

59.8

Employment with fixed income

49.6

50.4

Others

49.5

50.5

Not single

50.0

50.0

Single

47.8

52.2

34.0 58.5

66.0 41.5

No

58.7

41.3

Yes

25.9

74.1

5.407*

Employment

0.001

Marital status

0.134

Self-assessed health Average or lower Good or excellent

v -test

2.548 45.6

Male Living location

2

22.298**

Have chronic illness

34.680**

Private health insurance contract

1.300

Yes

51.2

48.8

No

44.7

55.3

High

29.7

70.3

Medium

Health risk

32.972** 67.5

32.5

Low Number of responses

55.2 202

44.8 198

Number of respondents

400

* Significant at the 10 % level ** Significant at the 5 % level

4.2.1.1 The Healthcare Use of the OPD In general, the patient goes to the OPD to check up on his/her illness with the physician, and receives medical treatment and medicine. Therefore, if a person has a chronic illness, especially diabetes, hypertension, and high cholesterol, it has a greater effect on the healthcare use of the OPD than if the person does not have a chronic illness. According to Table 6, the F statistic shows that at least one of the explanatory variables was not statistically equal to zero—meaning that these variables could explain a significant portion of the variance in healthcare use in Thailand. It was found that health status was statistically significant with a negative sign, while the log of personal income and chronic illness, especially diabetes, hypertension, and high cholesterol,

were statistically significant with a positive sign. Nevertheless, whether the respondents were afflicted with cardiovascular disease had an insignificant effect on the healthcare use of the OPD. Additionally, gender, age, marital status, education, employment, family size, living location, private health insurance contract, and health risk were not statistically significant. As a result, the characteristics related to healthcare use of the OPD were poor health condition, high personal income, and those that were afflicted by diabetes, hypertension, and high cholesterol. 4.2.1.2 The Impact of Healthcare Use of the IPD Inpatient care refers to medical treatment that is provided in a hospital or other facility, and requires at least one overnight stay. The patient admitted to the IPD normally shows more severe symptoms and acute chronic illness than the patient that goes to the OPD. Consequently, if people are afflicted by a chronic illness, particularly cardiovascular disease and high cholesterol, it has a greater effect on the healthcare use of the IPD than if people are not afflicted by a chronic illness. Regarding Table 7, the F statistic shows that at least one of the explanatory variables was not statistically equal to zero—meaning that these variables could explain a significant portion of the variance in healthcare use in Thailand. It was found that chronic illness, particularly cardiovascular disease and high cholesterol, was statistically significant with a positive sign. Although education level had an effect on health service use for Thai elders [22], the acquisition of a private insurance contact had a stronger effect on the healthcare use of the IPD. Persons that had private insurance usually want to be admitted to the hospital because they could claim health insurance benefits, as health insurance plans cover treatment expenses when the insured is hospitalized for more than 24 h. In addition, they can receive work compensation when admitted to the hospital. However, if the respondents had diabetes and hypertension, it had an insignificant effect on the healthcare use of the IPD. Interestingly, the log of education was significant with a negative sign, while the acquisition of a private insurance contact was significant with a positive sign. Nevertheless, respondents’ characteristics such as gender, age, marital status, employment, family size, living location, and health risk were not statistically significant. As a result, the characteristics related to healthcare use of the IPD were education level, the acquisition of private health insurance, and those that were afflicted by cardiovascular disease and high cholesterol. 4.2.2 Healthcare Use of Chronically Ill Respondents Because of the increasing number of elderly persons in Thailand, the patients that are afflicted by a chronic disease

P. Kananurak Table 3 Descriptive statistics of respondents that were hospitalized and nonhospitalized

Profile

Statistic

Age (years) Education level (years of schooling) Personal income (Baht per month) Household income (Baht per month)

Non-hospitalization

Hospitalization

P value 0.001**

Mean

46.32

48.58

Median

42.00

47.00

Mean

13.01

13.32

Median

12.00

14.00

Mean

21,894.44

32,983.17

Median

12,500.50

17,500.50

Mean

41,566.16

56,881.69

Median

25,000.50

35,000.50

0.325 0.001** 0.005**

Family size

Mean

1.66

1.77

0.300

Number of children

Mean

1.13

1.33

0.089*

Rating of health behavior

Mean

3.56

3.66

0.079*

Number of chronic illnesses

Mean

0.17

0.60

0.000**

Life expectation (years)

Mean

75.05

73.02

0.031**

Healthcare use (times per year)

Mean

0.00

2.86

0.000**

Mean

0.00 202

2,377.23 198

0.000**

* Significant at the 10 % level

Health service expenses (Baht per time) Number of responses

** Significant at the 5 % level

Number of respondents

400

Table 4 The determinants of healthcare use (model 1) Model

Unstandardized coefficient

Standardized coefficient

B

B

Standard error

P value

Collinearity statistics Tolerance

VIF

1.147

Constant

-3.316

1.116

Male

-0.099

0.067

-0.070

0.143

0.872

0.431

0.274

0.082

0.117

0.716

1.397

-0.052

0.166

-0.019

0.752

0.571

1.751

0.113

0.082

0.067

0.166

0.836

1.196

Log of age Log of education Single

0.003**

Employment with fixed income

0.055

0.077

0.036

0.480

0.777

1.287

Family size

0.021

0.034

0.030

0.534

0.829

1.206

Rating of health behavior

0.042

0.069

0.033

0.546

0.639

1.564

Private health policy

0.059

0.076

0.037

0.436

0.890

1.124

Log of income

0.213

0.063

0.200

0.001**

0.561

1.781

0.081 -0.261

0.081 0.074

0.051 -0.178

0.317 0.000**

0.748 0.773

1.337 1.294

Bangkok Health status Number of chronic illnesses

0.304

0.076

0.308

0.000**

0.332

3.016

Health risk

0.041

0.062

0.051

0.511

0.328

3.051

Test

Statistics value

Significance

Analysis of variance

F test = 9.454

0.000

R squared

0.242

Adjusted R squared

0.216

Model summary

* Significant at 10 % level ** Significant at 5 % level

will increase accordingly. Normally, chronically ill patients are consistently high users of healthcare. The higher rate of chronic illness on the part of the elderly will influence the healthcare use pattern. It is also shown in this study that the

healthcare use pattern among the chronically ill elderly increases with age. According to Table 8, the F statistic shows that at least one of the explanatory variables was not statistically equal to zero—meaning that these variables

Healthcare Use and Insurance in Thailand Table 5 The determinants of the healthcare use (model 2) Model

Unstandardized coefficient

Standardized coefficient

B

B

Standard error

P value

Collinearity statistics Tolerance

VIF

Constant

-3.236

1.126

Male

-0.092

0.068

-0.065

0.173

0.869

1.151

0.410

0.276

0.078

0.138

0.710

1.409

-0.068

0.169

-0.024

0.688

0.554

1.804

0.118

0.082

0.070

0.151

0.834

1.199

Log of age Log of education Single

0.004**

Employment with fixed income

0.048

0.079

0.031

0.541

0.754

1.326

Family size

0.022

0.034

0.032

0.515

0.817

1.223

Rating of health behavior

0.074

0.068

0.059

0.275

0.667

1.498

Private health policy

0.045

0.077

0.028

0.554

0.879

1.138

Log of income

0.216

0.063

0.202

0.001**

0.561

1.783

0.084 -0.268

0.081 0.075

0.053 -0.182

0.304 0.000**

0.743 0.755

1.346 1.325

Diabetes mellitus

0.293

0.133

0.111

0.028**

0.787

1.271

Hypertension

0.273

0.125

0.125

0.029**

0.603

1.657

Cardiovascular disease

0.575

0.254

0.106

0.024**

0.896

1.116

High cholesterol

0.304

0.122

0.135

0.013**

0.667

1.500

Health-risk

0.005

0.057

0.006

0.932

0.401

2.496

Bangkok Health status

Test Model summary Analysis of variance

Statistics value

Significance

F test = 7.764

0.000

R squared

0.245

Adjusted R squared

0.213

* Significant at 10 % level ** Significant at 5 % level

could explain a significant portion of the variance in healthcare use in Thailand. It was found that health status was statistically significant with a negative sign, while log of age was statistically significant with a positive sign. Nevertheless, gender, marital status, education, employment, family size, living location, having a private health insurance contract, and health risk were not statistically significant. As a result, of all the explanatory variables, the characteristics related to the healthcare use of the chronically ill elderly were older-age workers and poor health status.

older-age female persons and the older-age in the chronically ill elderly, and these interactions increased healthcare use. Interestingly, the interaction between a lower level of education and higher income also had an effect on the healthcare use of workers under the SHI Scheme. As a result, the main determinants of healthcare use for workers under the SHI Scheme in the model with interaction were single workers, higher income, older age of female workers, older age in the chronically ill elderly, and lowereducated but high-income workers (Table 9).

4.2.3 Model with Interaction

4.3 The Main Determinants of Voluntary Health Insurance

As the interaction terms in the healthcare use model were added and the variables in the regression model were adjusted, only one interaction variable, marital status 9 family size, was not statistically significant. The three interaction terms of age 9 gender, age 9 chronic illness, and education 9 personal income were statistically significant. Therefore, there were interactions between the

According to Table 10, a test of Model Chi-square (-2LL) shows that the model is a good-fitting one. The Wald criterion also demonstrated that the factors influencing the probability of having insurance were gender, personal income, and the number of chronic illnesses. There was a positive relation of personal income with the demand for voluntary health insurance, whereas the relationship of

P. Kananurak Table 6 The determinants of healthcare use of the outpatient care Model

Unstandardized coefficient

Standardized coefficient

B

B

Standard error

P value

Collinearity statistics Tolerance

VIF

Constant

-3.612

1.132

Male

-0.101

0.068

-0.072

0.139

0.869

1.151

0.417

0.277

0.080

0.133

0.710

1.409

Log of age

0.002**

Log of education

0.033

0.169

0.012

0.844

0.554

1.804

Single

0.113

0.082

0.067

0.172

0.834

1.199

Employment with fixed income

0.058

0.079

0.038

0.461

0.754

1.326

Family size

0.030

0.034

0.044

0.381

0.817

1.223

Rating of health behavior

0.074

0.068

0.060

0.276

0.667

1.498

-0.036

0.077

-0.023

0.639

0.879

1.138

0.216

0.064

0.204

0.001**

0.561

1.783

0.090 -0.221

0.082 0.076

0.057 -0.151

0.274 0.004**

0.743 0.755

1.346 1.325

Diabetes mellitus

0.307

0.133

0.117

0.022**

0.787

1.271

Hypertension

0.241

0.125

0.111

0.056*

0.603

1.657

Cardiovascular disease

0.389

0.255

0.072

0.128

0.896

1.116

High cholesterol

0.289

0.123

0.130

0.019**

0.667

1.500

Health risk

0.008

0.057

0.010

0.883

0.401

2.496

Private health policy Log of income Bangkok Health status

Test Model summary Analysis of variance

Statistics value

Significance

F test = 6.962

0.000

R squared

0.225

Adjusted R squared

0.193

* Significant at 10 % level ** Significant at 5 % level

gender and the number of chronic illnesses was negative. Those who are likely to purchase voluntary health insurance are female, have a higher personal income, and are healthy. Consequently, if they have chronic conditions, they are less likely to purchase private health insurance.

5 Discussion According to the survey, the factors influencing the probability of having or purchasing insurance were gender, personal income, and the number of chronic illnesses. A person that purchases voluntary health insurance is likely to be female, have a higher personal income, and be healthy. Much of the literature indicates that women use more healthcare services than men. Several explanations have been offered. These differences may be associated with reproductive biology and conditions specific to gender, higher rates of morbidity in women than in men, differences in health perceptions, and the reporting of symptoms and illnesses, or a greater likelihood that women

seek help for prevention and illness [23]. In addition, the higher the income the more likely individuals are to purchase voluntary health insurance [24]. An increase in income has been found to lead to an increase in the probability of having insurance as well as the quantity of insurance purchases [2]. Additionally, there was a negative relationship between health insurance and the presence of chronic conditions; those that exhibited fair or poor health condition are actually less likely to purchase private health insurance, consistent with the findings review from the health insurance literature. Health status is relevant to the decision to obtain health insurance because it affects an individual’s likely future consumption of healthcare, and individuals vary in their expectations of healthcare use. However, people that have voluntary health insurance are significantly more likely to insure themselves against other risks that are not likely to be correlated with health risks [25]. Conversely, people in worse health may have greater difficulty than those in better health in obtaining health insurance coverage, either because they face higher premiums that make health insurance unaffordable or because

Healthcare Use and Insurance in Thailand Table 7 The determinants of healthcare use of the inpatient care Model

Unstandardized coefficient

Standardized coefficient

B

B

Standard error

P value

Collinearity statistics Tolerance

VIF

Constant

0.472

0.713

Male

0.010

0.043

0.013

0.807

0.869

1.151

Log of age

-0.019

0.175

-0.006

0.913

0.710

1.409

Log of education

-0.246

0.107

-0.151

0.013

0.052

0.013

Single Employment with fixed income

0.508

0.021**

0.554

1.804

0.803

0.834

1.199

0.001

0.050

0.001

0.991

0.754

1.326

Family size

-0.030

0.022

-0.074

0.169

0.817

1.223

Rating of health behavior

-0.008

0.043

-0.011

0.856

0.667

1.498

Private health policy

0.191

0.049

0.204

0.000**

0.879

1.138

Log of income

0.039

0.040

0.063

0.332

0.561

1.783

-0.040 -0.102

0.051 0.048

-0.044 -0.121

0.439 0.032

0.743 0.755

1.346 1.325

Diabetes mellitus

0.022

0.084

0.015

0.790

0.787

1.271

Hypertension

0.050

0.079

0.040

0.525

0.603

1.657

Cardiovascular disease

0.307

0.161

0.098

0.057*

0.896

1.116

High cholesterol

0.134

0.077

0.104

0.083*

0.667

1.500

Health risk

0.022

0.036

0.048

0.536

0.401

2.496

Bangkok Health status

Test Model summary Analysis of variance

Statistics value

Significance

F test = 2.316

0.003

R squared

0.088

Adjusted R squared

0.050

* Significant at 10 % level ** Significant at 5 % level

they are often denied coverage by insurers [26]. Assuming that people that buy health insurance are more risk averse, risk-averse people will be more willing to pay a premium that exceeds their expected benefits, reflecting the value they obtain from protection from the financial risks associated with medical spending. Therefore, healthy people that buy health insurance may want to avoid financial risks as well as gain access to healthcare [12]. The survey of healthcare use of workers under the SHI Scheme revealed that health status was statistically significant with a negative sign, and chronic illness was statistically significant with a positive sign. Nonetheless, gender, age, marital status, education, employment, family size, living location, private health insurance contract, and health risk were not statistically significant. Although the majority of studies have investigated the effect of age on hospitalization, only one-third reported that older people had higher hospital use than younger individuals [11]. However, as age increases, the number of chronically ill patients will increase accordingly. In addition, those with

the worst health status use more healthcare services and this may come from having a chronic illness, which requires more treatment and therapy. Consequently, the characteristics related to healthcare use were poor health status, high personal income, and those that are afflicted by chronic illness. The result from this study is also consistent with the empirical literature on healthcare use based on the behavioral model. Much of the literature saw a positive correlation between income and health [27], which exists for psychological as well as physical health. However, the relationship between income and healthcare use varies across countries. In addition, health conditions seem to be a more important factor in explaining healthcare use [28]. When need or health status is taken into consideration, several studies suggest that these structural, social, and psychological factors have comparatively little effect on patient behavior [28]. Furthermore, the number of chronic illnesses is one of the clearest determinants of health status in almost all studies conducted among the elderly in terms of healthcare use [29]. Poor health conditions with the

P. Kananurak Table 8 Healthcare use of chronically ill patients Model

Constant

Unstandardized coefficient

Standardized coefficient

B

B

Standard error

P value

Collinearity statistics Tolerance

VIF

-4.945

2.459

Male

0.048

0.155

0.032

0.760

0.878

1.139

Log of age

1.144

0.596

0.228

0.058*

0.668

1.496

-0.033

0.350

-0.012

0.925

0.573

1.746

0.097

0.200

0.051

0.631

0.848

1.179

Log of education Single

0.047

Employment with fixed income

0.005

0.190

0.003

0.981

0.630

1.588

Family size

0.042

0.074

0.059

0.577

0.863

1.159

-0.053

0.123

-0.047

0.669

0.785

1.274

Rating of health behavior Private health policy

0.082

0.179

0.048

0.648

0.874

1.145

Log of income

0.186

0.128

0.183

0.149

0.600

1.666

-0.214 -0.453

0.188 0.158

-0.133 -0.290

0.260 0.005**

0.690 0.920

1.448 1.088

Bangkok Health status Test

Statistics value

Significance

0.084

Model summary Analysis of variance

F test = 1.710

R squared

0.178

Adjusted R squared

0.074

* Significant at 10 % level ** Significant at 5 % level Table 9 Interaction healthcare use model Model

Constant

Unstandardized coefficient

Standardized coefficient

B

B

Standard error

P value

Collinearity statistics Tolerance

VIF

-2.289

0.562

Single

0.196

0.102

0.116

0.055*

0.540

1.853

Employment

0.022

0.074

0.014

0.771

0.838

1.194

Family size

0.042

0.035

0.060

0.240

0.759

1.318

Rating of health behavior

0.092

0.058

0.074

0.113

0.915

1.093

Private health policy

0.042

0.076

0.026

0.580

0.894

1.118

Log of income

0.248

0.057

0.232

0.000**

0.692

1.445

Living location

0.084

0.080

0.053

0.289

0.774

1.291

-0.003 0.005

0.002 0.001

-0.090 0.297

0.084* 0.000**

0.729 0.805

1.372 1.242

Age 9 gender Age 9 chronic illness

0.000**

Education 9 personal income

-0.001

0.001

-0.119

0.049**

0.541

1.849

Marital status 9 family size

-0.108

0.077

-0.084

0.162

0.543

1.841

Test

Statistics value

Significance

Analysis of variance

F test = 11.065

0.000

R squared

0.239

Adjusted R squared

0.217

Model summary

* Significant at 10 % level ** Significant at 5 % level

Healthcare Use and Insurance in Thailand Table 10 The main determinants of voluntary health insurance Predictor

B

Constant Male Bangkok Age

Standard error

Wald

df

Significance

Odd ratios

-18.36612

5.39955

11.56962

1

0.0007**

0.00000

-0.52992

0.31581

2.81565

1

0.0933*

0.58865

0.11146

0.34181

0.10633

1

0.7444

1.11791

1.04046

1.21939

0.72805

1

0.3935

2.83051

-0.65000

0.74140

0.76865

1

0.3806

0.52204

Employment

0.67523

0.42660

2.50532

1

0.1135

1.96448

Personal income

1.50283

0.27004

30.97253

1

0.0000**

4.49438

Single

-0.02241

0.43747

0.00263

1

0.9591

0.97783

Number of children

-0.08940

0.20026

0.19928

1

0.6553

0.91448

Family size

0.02638

0.11502

0.05260

1

0.8186

1.02673

Self-health assessment

0.45951

0.34560

1.76789

1

0.1836

1.58330

-0.60368

0.28209

4.57961

1

0.0324**

0.54679

Education

Number of chronic illnesses Test

Statistics value

df

Significance

8

0.923

Model summary -2 Log likelihood

284.023

Cox and Snell R square

0.223

Nagelkerke R square

0.320

Goodness-of-fit Hosmer and Lemeshow

3.169

* Significant at 10 % level ** Significant at 5 % level

presence of a number of chronic illnesses will lead to an increase in the amount of health service use accordingly (Fig. 1). It was found that most variables that are defined to be important in explaining healthcare use are also important in explaining the demand for voluntary health insurance. However, there is an unmet need for coverage between healthcare service use and demand for voluntary health insurance. People in worse health may have greater difficulty in obtaining health insurance coverage. Furthermore, the SHI scheme does not cover workers after retirement. As a result, the workers in poor health or those that anticipate future poor health are likely, but not certain, to consume more medical care in the future than those in better health, and there is a high catastrophic risk for this group of workers after retirement. In Thailand, the market for healthcare and long-term care insurance for the elderly is still under early development. If the government does not manage the health and long-term care insurance for the elderly, the private insurers will fail to meet some targets for expanding coverage and lowering healthcare costs. Therefore, the government should develop voluntary health insurance for these workers. Based on the analysis, the study results suggest some key policy recommendation for voluntary health insurance after retirement. Recently, the Thai government has

guaranteed healthcare coverage for all Thai elderly people in a program funded from general tax revenues [7]. Most low-income people, in urban and rural areas, use public facilities and are covered by a public insurance scheme; namely, the UC Scheme. In contrast, medium- to highincome people usually use private facilities, and they are the group that can buy private health insurance. This private insurance gives them access to ‘‘premium’’ services provided by private facilities while being insured. The privately insured are decidedly better off than others with respect to both service accessibility and financial hedges against costs. However, most people cannot afford private health insurance. In addition, the voluntary health insurance schemes in Thailand are still at an early stage of development and have yet to seriously address the questions of equity and efficiency, while private health insurance is limited to people that can afford the premium [7]. Given the increases in healthcare costs at the older ages, and the association of retirement with low income and chronic illness, the retired workers typically face the greatest obstacles to obtaining private health insurance [30]. While the decision to buy private health insurance is likely to be made before 50 years of age during one’s earlier working life, government authorities should provide support for both people that want to buy health insurance and the insurance providers that are issuing the insurance

P. Kananurak Fig. 1 The gap between demand for voluntary health insurance and health use

Health Utilization

Demand for Voluntary Health Insurance

Female

High monthly personal income

High monthly personal income

More presence of chronic conditions

Less presence of chronic conditions

Average or Poor health status Gap of uninsured person

policy. Then, the workers after retirement with or without chronic illness can buy the health insurance at an affordable premium rate. This study proposes health insurance after retirement for workers under the SHI Scheme organized by government authorities but provided by private issuers. However, it is conceivable that difficulties will appear in meeting the expectations of heterogeneous groups demanding ‘‘personalized’’ healthcare, more choices, and promptness of delivery. In addition, the various types of health service benefits associated with the premium will be segmented by target customers, such as gender and the individuals’ characteristics according to their needs and expectations. This study includes only the urban area where the dependency ratio is higher than the dependency ratio in the rural area. This study may not represent Thai workers in rural areas and nationwide, under the SHI Scheme. The Anderson–Newman model of healthcare use was used as a conceptual framework to understand the factors explaining the healthcare use pattern of workers in Thailand. Therefore, there might be specific determinants of health status in the Thai population that were not covered by the sociodemographic and health-related variables study such as the psychosocial variable. Presently, most of the health research studies aim to focus on long-term care insurance for the elderly. As long-term care expenditures represent one of the largest uninsured financial risks faced by the elderly, the limited insurance coverage for long-term care expenditures has important implications for the welfare

loss of the elderly in most countries of the world. In addition, the features of health insurance and long-term care insurance are different from the view point of insurers and consumers, which is not included in this study. Nevertheless, the demand on health insurance varies with age cohort, therefore this study focused only on workers under the SHI Scheme aged between 40 and 69 years. Future research should investigate workers’ preferences, especially regarding long-term insurance for the elderly. Because long-term care expenditures represent one of the largest uninsured financial risks facing the elderly, limited insurance coverage for long-term care expenditures has important implications in terms of welfare loss for the elderly in most countries in the world. It is thus hoped that the lessons drawn from this study will provide a useful guide for other countries in developing health insurance after retirement or in reforming health insurance benefit schemes for the elderly to meet their needs and preferences. Acknowledgments Thanks to Associate Professor Dr. Udomsak Seenprachawong, Assistant Professor Dr. Dararatt Anantanasuwong, and Assistant Professor Dr. Amornrat Apinunmahakul for their useful comments and suggestions. Author Contributors Conception and design: Kananurak Papar, Administrative support: Kananurak Papar, Financial support: Kananurak Papar, Collection and assembly of data: Kananurak Papar, Data analysis and interpretation: Kananurak Papar, Manuscript writing: Kananurak Papar.

Healthcare Use and Insurance in Thailand Conflicts of interest interest exist.

The author has declared that no conflicts of

References 1. Martins JO, de la Maisonneuve C. The drivers of public expenditure on health and long-term care: an integrated approach, OECD economic studies no. 43. OECD Publishing, European Union, 2006. 2. John K, Napaporn C. Gender and ageing in Thailand: a situation analysis of older women and men. Michigan: Population Studies Center; 2008. 3. Banloo S. Situation of the Thai elderly 2008. Bangkok: Foundation of Thai Gerontology Research and Development Institute; 2009. 4. John K, Napaporn C. Older persons in Thailand: a demographic, social and economic profile. Ageing Int. 2009;33:3–14. 5. Yot T, Viroj T. Designing a reproductive health services package in the universal health insurance scheme in Thailand: match and mismatch of need, demand and supply. Health Policy Plan. 2004;19(Suppl 1):i31–9. 6. Preker AS, Zweifel P, Schellekens OP (eds). Global marketplace for private health insurance: strength in numbers. World Bankfree PDF; World Bank Publications, Washington, D.C., 2009. http://www.google.co.th/books. Accessed May 2011 7. Banloo S. Situation of the Thai elderly 2008 (in Thai). Bangkok: Foundation of Thai Gerontology Research and Development Institute; 2009. 8. Social Security Office. 2010. The number of insured persons under the Social Health Insurance Scheme in 2010. Retrieved 13 November, 2010 from http://www.sso.go.th. 9. Kathryn P, Kathleen M, Ronald A, Aday L. Assessing variables in healthcare utilization. Health Serv Res. 1998;33:571–91. 10. Ronald A. Revisiting the behavioral model and access to medical care: does it matter? J Health Soc Behav. 1995;36:1–10. 11. Angela B, Wouter W, Hanneke H. Predictors of health care utilization in the chronically ill: a review of the literature. Health Policy. 1997;42:101–15. 12. John N. The value of health insurance: the access motive. J Health Econ. 1999;18:141–52. 13. Hutchinson KM, Holtman MC. Analysis of count data using poisson regression. Res Nurs Health. 2005;28:408–18. 14. Paula D. Methods for analyzing health care utilization and costs. Annu Rev Public Health. 1999;20:125–44. 15. Box GEP, Cox DR. An analysis of transformations. J R Stat Soc Ser B (Methodological). 1964;26:211–52. 16. Verbrugge LM. Gender and health: an update on hypotheses and evidence. J Health Social Behav. 1985;1985(26):156–82. 17. Young JT. Illness behavior: a selective review and synthesis. Sociol Health Illness. 2004;26:1–31. 18. Barbara B. Diversities in conceptions of health and physical fitness. J Health Human Behav. 1961;2:39–46.

19. Williams DR. Socioeconomic differentials in health: a review and redirection. Social Psychol Q. 1990;53:81–99. 20. Rogers RG, Hummer RA, Nam CB. Living and dying in the USA: behavioral, health, and social differentials of adult mortality. USA: Academic Press; 2000. 21. Peng CYJ, Lee KL, Ingersoll GM. An introduction to logistic regression analysis and reporting. J Educ Res. 2002;96:3–14. 22. Jitapunkul S, Na SM, Chayovan N, Choprapawon C, Kachondham Y, Buasai S. A national survey of health-services uses in Thai elders. Age Ageing. 1999;28:67–71. 23. Bertakis KD, Azari R, Helms LJ, Callahan EJ, Robbins JA. Gender differences in the utilization of health care services. J Fam Pract. 2000;49:147–52. 24. Joan C, Jaume G. Demand for private health insurance: how important is the quality gap? Health Econ. 2003;12:587–99. 25. Buchmueller TC, Alan CM. Employer-sponsored health insurance and the promise of health insurance reform. Inquiry. 2009;46:187–202. 26. Elisabeth S, Cathy S, Stephanie B. Market failure? Individual insurance markets for older Americans. Health Aff. 2001;20: 139–49. 27. Susan E. New evidence on the relationship between income and health. J Health Econ. 1996;15:67–85. 28. Coulton C, Frost AK. Use of social and health services by the elderly. J Health Social Behav. 1982;23:330–9. 29. Haseen F, Adhikari R, Soonthorndhada K. Self-assessed health among Thai elderly. BMC Geriatr. 2010;10:30. 30. McWilliams JM, Zaslavsky AM, Meara E, Ayanian JZ. Health insurance coverage and mortality among the near-elderly. Health Aff. 2004;23(4):223–33. 31. Viroj T, Walaiporn P, Chitpranee V, Phusit P, Pongpisut J, Samrit S, Jadej T. Thailand health financial review 2010. Bangkok: WHO SEARO; 2010. 32. Jitapunkul S, Chayovan N, Kespichayawattana J. Chapter 6: national policies and long term care of elderly in Thailand. In: David RP, Alfred CMC, editors. Ageing and long-term care: national policies in the Asia Pacific. Singapore: Institute of Southeast Asian Studies; 2002. p. 181–213. 33. Worawet S. Poverty and financial security of the elderly in Thailand. Ageing Int. 2009;2009(33):50–61. 34. Banloo S. Situation of the Thai elderly 2008(in Thai). Bangkok: Foundation of Thai Gerontology Research and Development Institute; 2009. 35. National Statistic Office. Report on the 2007 Survey of the Older Persons in Thailand. Bangkok: National Statistic Office (in Thai); 2007. 36. Joseph A. Health care financing in Thailand: modeling and sustainability. Bangkok: World Bank; 2007. 37. Viroj T, Anuwat S, Jongkul L. The social security scheme in Thailand: what lessons can be drawn? Social Sci Med. 1999;48(7):913–23. 38. Somsak C, Kanittha B. Long-term care in developing countries: ten case-studies. Geneva: World Health Organization; 2003. 39. Viroj T, Walaiporn P, Chitpranee V, Phusit P, Pongpisut J, Samrit S, Jadej T. Thailand Health Financial Review. Bangkok: WHO SEARO; 2010.