Biomedical Research

Abstract

Objective: Vietnam’s healthcare system has undergone numerous reforms while institutionalizing a market-based mechanism. Still the poor cite medical costs and inadequate insurance as the reason to destitution risks. The reality gives rise to evidence on: 1) Effects of residency status, socioeconomic status, medical costs, and insurance on patient satisfaction towards health insurance; 2) Existence of “psychological thresholds” toward cost and insurance coverage. The article addresses these two questions.

Materials and Method: A dataset containing 900 records was collected from a 2014-15 survey on Northern Vietnamese patients, 67% of which with insurance. Attributes include: a) medical expenditures; b) (in)eligibility for insurance coverage; c) dis/satisfaction with insurance; d) length of hospitalization; e) socioeconomic/residency status. The method of analysis is logistic regression techniques involving both discrete and continuous predictor/response variables. Estimations are then used to compute empirical probabilities and threshold values.

Results: Relations are found statistically significant at 5% level, suggesting that insurance coverage and medical expenditures contribute to higher probabilities of satisfaction. Insurance has a stronger influence. Threshold coverage and expenditures show that perceptions are heterogeneous. The poor and non-resident patients show lower threshold coverage of 63.4% (mean=58%).

Conclusions: An overemphasis on raising insurance coverage is unlikely to pay off. To promote the value of health insurance, the target group should be the poor and non-resident patients. It is better for policymakers to promote a diversity of insurance schemes, including micro-health insurance, and to lower medical costs for the target group. Thresholds can serve as a reference value for policy-making.

Keywords

Health insurance, Psychological threshold, Medical expenditures, Healthcare policy, Vietnam.

Introduction

During a mid-2016 cabinet meeting, the Vietnamese Prime Minister Nguyen Xuan Phuc said: “I believe [universal insurance coverage rates] should be raised above 91%” [1] although the amended Law of Health Insurance had come into effect for about 6 months, theoretically increasing universal coverage to 100% [2]. The Vietnamese government seems adamant about covering medical expenses, but the question remains as to whether they are taking the right course of actions [3].

Vietnam’s health care system has undergone numerous reforms even before reunification in 1975, but health insurance was only introduced after 1986, when the country launched its Doi Moi reforms [3,4]. Most renovating efforts have been directed towards raising insurance coverage to 100% of expenditures for beneficiaries, and for a good cause. Financial hardships prevail where medical expenses are concerned, especially when the diseases are serious and require lengthy treatment [3-5]. In fact, a third of the poor households in debts in rural North Vietnam cite medical costs as the reason to their indebtedness [5]. There is evidence reporting the harsh reality that patients are at highest risks of destitution when being either uninsured or ineligible for adequate insurance [3,4].

There is no doubt that health insurance policies are a crucial matter, dealing with both economic and social challenges [3-6]. At present, the health insurance system in Vietnam gives patients little choice but to take part in compulsory health insurance scheme, with an option of buying commercial insurance packages, oftentimes offered by major life insurance firms licensed by the Vietnamese government [5-7]. In the current situation, the success of further health insurance reform remains to be seen as future challenges are overwhelming given the limited budget and precarious management system that frequently show weaknesses [7,8], rendering the implementation of the amended Health Insurance Law virtually impossible [2,3,9-11].

The reality of Vietnamese health and health insurance sector gives rise to the concept of “better market design” [12], which in turn would need to rely on empirical evidence and learned insights on possibilities of universal coverage [13] and willingness/capacity to pay for health insurance, especially in the rural areas and informal sector [14-16].

Objectives

The study is bound to answer two key research questions (RQ1-RQ2) as follows:

RQ1: How do residency status, socioeconomic status, total treatment cost, and actual insurance coverage impact patients’ level of satisfaction towards health insurance services?

RQ2: Does there exist some “psychological thresholds” of treatment cost and insurance coverage that would mark a turn in patients’ perception of health insurance services and if yes, what other factors come into play when calculating this threshold value?

Answering these will more likely provide some useful implications, as discussed toward the end.

Materials and Methods

Material/data

The research employs a data set containing 900 records randomly collected from a medical survey on Vietnamese patients conducted in five different provinces in Northern Vietnam–including major cities as Hanoi, Hai Phong, Quang Ninh–from August 2014 to June 2015. The survey team approached patients who actually used healthcare services, without knowing whether he/she actually held a health insurance policy. The questionnaire asks for such key information as their actual medical expenditures, (in)eligibility for insurance coverage, perceived dis/satisfaction about health insurance service, lengths of hospital stay, as well as some other standard information such as socioeconomic status (SES), and residency status (in relation to the location of the healthcare station that a corresponding patient receives treatment services), and so on.

The subset containing 605 records for those who actually held at least a compulsory health insurance policy is used for analysis, i.e. 67.2% of the total data sample. This data subsample consists of 333 female patients and 272 male. Patients’ age spans from 1 to 92, with a majority of 67% belonging to the 40-70 age bracket. A histogram of respondents’ age is provided in Fig.1. It is also worth mentioning that a large portion of patients come from Hanoi and has a medium socioeconomic status.

The following categorical variables enter directly or indirectly into data analysis:

• Level of patient satisfaction regarding health insurance services (“SatIns”), grouped into two categories: “satis” (higher level) and “unsat” (low to medium).

• Patients’ residency status, consisting of two categories: “yes” (with residency status in the region where the healthcare unit is located) and “no” (without residency).

• Patient’s socioeconomic status (SES), divided into: “Hi” (high), “Med” (medium) and “Lo” (low).

The structured presentation of the data subset following key variables used in subsequent modeling efforts is provided in Table 1, from which it can be learned that the majority of patients are not satisfied with health insurance services (>83%).

Table 1: Empirical distribution of patients by categorical values.

Two continuous variables that also enter our modelling attempts are: i) total medical expenditures incurred to a patient (“Spent”), measured in millions of Vietnamese Dong (local currency; VND 1 million=US $47 using official exchange rate as of 2015 year-end); and, ii) the actual medical expenditures covered by the health insurance system (“Pins”), measured against “Spent”, thus with a value between [0,1] (Table 2). Empirical values of both variables fluctuate widely. It is worth observing the mean of “Pins” being <0.6, and clearly much lower than the Vietnamese government’s target 100%, not to mention 33% of respondents holding no compulsory insurance.

Table 2: Basic statistics of the continuous variables.

Methods

The subsequent analysis employs logistic regression techniques involving both discrete and continuous predictor variables while the response variable is dichotomous. The functional form of these relations drawn upon these regressions have the generic form of Eq.(1):

→(1)

where K is the largest number of predictor variables used in any modeling attempt.

In Eq.1,π(x) represents the success probability, i.e. Y_i=1; Y_i is the event to be observed from the empirical data; β₀ is the intercept; and β_i coefficients associated with the i^th predictor variable, X_i.(x) is given by:

The standard null hypothesis is β₀=0, for each i=1,…,K. In the case of X_i being a continuous variabe, if i>0 then an increase of X_i will result in the increase of π(x). The reverse is true when i<0. Therefore, when X_i increase by 1 unit, the odds of Y increase by exp (i).

The likelihood ratio test statistic is employed for hypothesis testing using:

Where l₀ is the numerical value of the likelihood function computed from the observed data using under the null hypothesis estimate (π) and l₁ under the empirical data-based estimate (π?). This G² test statistic follows a X² distribution with K degrees of freedom (df). Actual estimations and technical treatments for the analysis are provided in [17,18]. The study applies the conventional statistical significance level of 5% [19].

Results

Results for RQ1:

To examine the influence of factors: residency status, socioeconomic status, cost of treatment and insurance coverage, on the patients’ satisfaction towards heath insurance services, the study employs “SatIns” as dependent variable and “Spent”, “Pins”, “Res” and “SES” as independent variables in a logistic regression model to compute the following results:

Most coefficients in Table 3 are statistically significant, with pvalue< 0.05, suggesting that the patients’ satisfaction towards health insurance is considerably affected by their residency status, socioeconomic status, the cost of their treatment, and the actual coverage that health insurances provides them during medical care. Their empirical relations have the functional form of Eq. (RQ1):

Table 3: Estimation results for impact of “Spent”, “Pins”, “Res” and “SES” on patients’ level of satisfaction towards health insurance services.

Tables 1 and 2 provides explanations for variables in Eq(RQ1. Positive coefficients for “Spent” and “Pins” appear to indicate that patients are more likely to feel satisfied with health insurance services when costs of treatment and/or insurance coverage go up. As β1=0.010 (p<0.01) and β2=1.839 (p<0.01), increasing one unit of “Pins” will boost probabilities of satisfaction much more significantly than does the same unit of increase in “Spent”. In other words, insurance coverage is much more impactful than medical expenditures factor, in terms of improving patients satisfaction rate.

From the negative coefficient of category “yes” of variable “Res”, it can be inferred that a patient with a residency status has a lower probability of satisfaction. More precisely, the odds of satisfaction of patients with residency status in the city is estimated to be exp (-1.916)=0.147 times lower than that of patients who came from other regions. Moreover, the regression coefficients of “SES” in Eq.(RQ1.1) shows that medium to low socioeconomic status decreases the probability of satisfaction, particularly those of the medium status. These remarks all raise questions and give rise to the discussions at the end of this article.

From Eq.(RQ1), a conditional probability of a patient’s satisfaction follows:

For a patient with residency status in the city, bearing a low SES, paying VND 50 million for medical treatment, 60% of which being covered by health insurance, there is a 11.2% probability that he/she finds his/her health insurance to be satisfactory. This is noteworthy as mean of insurance coverage is <60% (see Table 2); and yet, the patient is still much more inclined to feel unsatisfied. Perhaps, in the case of insurance coverage, more is not always better as other factors come into play.

With other residency and socioeconomic statuses, specific computations from Eq.(RQ1) are presented in Table 4.

“SES”|“RES”	“YES”	“NO”
“Lo”
“MED”
“Hi”

Table 4: Calculations of probability of satisfaction by patients’ residency status and SES.

Results for RQ2:

The (psychological) “thresholds” of treatment cost (“Spent”) and insurance coverage (“Pins”) is defined as the point where π_satis=π_unsat=50% in each case, meaning that the patients are equally as probable to be satisfied or unsatisfied with health insurance services. All values other than the threshold will show π_satis ≠ π_unsat. More precisely, at points where π_satis<π_unsat, patients are less likely to be satisfied with health insurance services. Likewise, when π_satis<π_unsat, patients tend to perceive insurance benefits more positively.

Possible threshold medical expenditures: The next logistic regression is performed with “Spent” and “Res” being predictor variables, and “SatIns” response. The actual estimations, provided in Appendix A, lead to relationship Equation (RQ2.1).

From Eq.(RQ2.1), the formula of probability of patients’ satisfaction, taking into account their residency status and cost of treatment, is as follows:

The likelihood of satisfaction among patients with residency status is thus determined by:

In the same vein, the probability of satisfaction for patients without residency status is given by:

The threshold that affects patients’ perception of insurance services is determined as “Spent” value where π_satis=50%. Substituting this value of π_satis into Eq. (RQ2.1.1):

Therefore, the threshold treatment cost for patients with residency status is approximately VND 245.8 million. Similarly, by substituting π_satis=50% into Eq.(RQ2.1.2), the threshold value of “Spent” for patients without residency status is obtained.

Figure 1 shows the probabilistic trends for dis/satisfaction of patients towards health insurance services. The two lines intersect at the threshold point of treatment cost as a factor affecting patients’ dis/satisfaction towards insurance services. As surprising as it may sound, the higher the expenditures are, the more likely the patients will find their insurance benefits to be satisfactory. However, in terms of absolute value, patients with a residency have a numerical threshold nearly 2.5 times as those without residency, specifically VND 255 million against VND 104 million.

Figure 1: Probability of patients’ dis/satisfaction towards health insurance conditional on residency status (data from Appendix B).

Possible threshold insurance coverage rate: This investigation uses logistic regression, provided in Appendix C, with response variable “SatIns” and predictors “Pins”, “Res” and “SES”. In an effort to quantify the influence of insurance coverage as well as residency and SES on patients’ perception, Eq.(RQ2.2) become useful.

To compute the probabilities of satisfaction conditional on residency and SES, substituting “Pins” values corresponding to different categories of “Res” and “SES” yields the following formula:

A set of computed values against different conditions of residency and SES is given in Appendix D. Figures 2-4, using Appendix E data, unveil some insights regarding patients’ perception of health insurance services in different cases. In these figures, the lines representing the probability of satisfaction (“satis”) tend to go upwards as insurance rates rise. The reverse trend can be observed in “unsat” lines (dissatisfaction). Where “satis” and “unsat” cross paths on each graph is the point of threshold, at which π_satis=π_unsat=50%. Figure 2 represents the perception of patients with a residency bearing a low SES. “Satis" and “unsatis” do not intersect at any value of “Pins”. Remarkably, the same phenomenon occurs among other patients of medium or high SES (Figures 3 and 4).

Figure 2: Probabilities of dis/satisfaction by residency status among patients with low SES.

Figure 3: Probabilities of dis/satisfaction by residency status among medium SES patients.

Figure 4: Probabilities of dis/satisfaction by residency status among higher SES patients.

On the right hand sides of Figures 2-4, “satis” and “unsat” intersect at the threshold insurance coverage affecting patients’ perception of health insurance. Empirically, “Pins” thresholds for patients with low, medium and high SES are 63.4%, 86.6% and 17.6%, respectively. This means that without a residency, a modest 17.6% insurance coverage is already enough to content a higher SES patient. But, medium-SES patients show a much higher threshold insurance rate: 86.6%.

Discussion

Attempts to model the probability of satisfaction among patients and determine a threshold value of medical expenditures and insurance coverage have shown that the key variables are all significant to patients’ perception. Further insights and policy implications are offered below. Firstly, a rise in insurance coverage will prompt an increase in probability of the patients’ satisfaction towards health insurance services. In the first place, this might sound obvious, and Vietnamese policy-makers seemed to be thinking the same when setting the idealistic 100% insurance rate [20]. But reality shows that it is not possible for any citizen to exclusively rely on insurance to pay for their entire treatment [21], this can be seen at first glance in Table 2, where the maximum insurance coverage in real life cases only attains 90%, minimum coverage is null, and average coverage stays merely 60%. Due to possible frictions in the administrative system and insurance eligibility based on residency status, patients are often stuck with inadequate insurance coverage. In the end, they do not receive what they are promised to be given, and remain frustrated with the system [22].

Second, a possible explanation for patients to find higher medical expenditures to “help” on improving satisfaction rate is needed. This is probably is the psychology of the majority of patients: “It is still good to have money to spend on treatments” [23]. Notably, the threshold of treatment cost for patients with residency status to perceive health insurance services as more positive is VND 246 million (~US$12,000); this amount is huge as Vietnam’s per capita GDP is only US $2,300 in 2016. Meanwhile, at less than half that cost, patients from other regions who come to the city/province for medical treatment already start to content themselves with health insurance services. This shows a large gap in standards between patients of different SES, similar to the situation analyzed in [24]. Third, for patients with residency status, there is no insurance coverage threshold for their satisfaction, even if the insurance rate is highest possible (90%). On the other end of the spectrum, high SES non-resident patients who show exceptional leniency towards health insurance services, tend to be content with an insurance coverage as low as 17.6%. The vast difference between the various groups of patients again reinforces the population’s heterogeneity, and plays key evidence to our later concluding remarks [25].

Regarding SES, those belonging in high classes tend to feel better about health insurance services, while the middle class is the hardest to please. In fact, patients with higher SES already have little to worry about financial burdens, so insurance coverage plays a lesser role in determining their level of satisfaction. Figure 4 show that high-SES patients have the lowest average insurance coverage among all groups [26]. It is speculated that these people might care more about the actual medical services provided during treatment as well as how convenient the administrative procedure is [27]. In stark contrast with high-class patients, the middle class is extremely critical towards health insurance services, as the threshold is an impossible 85%.

Conclusion

This concluding section provides some useful policy implications drawn upon the results and discussion:

• The above remarks suggest that efforts to raise satisfaction for both of these classes by raising insurance coverage would be unlikely to pay off. This leaves us with the most vulnerable group, i.e. poor, non-resident patients, as a target group if the government and health system want to promote the value of compulsory insurance policies. As the threshold rate is a little >63%, reducing vulnerability of this group of patients is most feasible and least costly, while having far-reaching effect as far as health policies are concerned.

• The current health insurance system lacks options and becomes rigid given the observed heterogeneity of the population. Therefore, it would be more beneficial if policy makers, instead of promising an unattainable universal coverage, target the vulnerable and build more diverse insurance schemes to accommodate their needs, such as micro-health insurance [28]. This can be done by lowering general costs – which would be a long way to go-and modifying health insurance policies – which is a more feasible measure, in shorter terms.

• As insurance policies need to stay balanced between covering patients’ expenses and keeping the fund stable, higher coverage does not necessarily mean better policies; rather, the amount should be optimal [29,30]. The threshold values of insurance coverage that are determined by the above analysis are expected to serve as a reference point for future policy making [31].

Acknowledgement

This work was supported in part by Hanoi-based research philanthropy Quang-Vu-Hung Hoi (Foundation) (QVHH.BR16-01).

References

1. Thoa D. PM Nguyen Xuan Phuc: By 2020, more than 90% of the population will participate in health insurance. CPV Online Newspaper, 2016.
2. Vietnamese National Assembly. Law No. 46/2014/QH13 - Amendments to the Law on Health Insurance, 2014.
3. Vuong QH. Be rich or don't be sick: estimating Vietnamese patients' risk of falling into destitution. Springerplus 2015; 4: 529.
4. Ensor T. Introducing health insurance in Vietnam. Health Policy Plan 1995; 10: 154-163.
5. Whitehead M, Dahlgren G, Evans T. Equity and health sector reforms: can low-income countries escape the medical poverty trap?. The Lancet 2001; 358: 833-836.
6. Tran VT, Hoang TP, Mathauer I, Nguyen TKP. A health financing review of Vietnam with a focus on social health insurance. World Health Organization, Geneva, 2011.
7. Ekman B, Nguyen TL, Ha AD, Axelson H. Health insurance reform in Vietnam: a review of recent developments and future challenges. Health Policy Plan 2008; 23: 252-263.
8. Sepehri A, Sarma S, Simpson W. Does non-profit health insurance reduce financial burden? Evidence from the Vietnam living standards survey panel. Health Economics 2006; 15: 603-616.
9. Lieberman S, Wagstaff A. Health financing and delivery in Vietnam. World Bank: Washington D.C, 2009, 64-65.
10. Jowett M, Contoyannis P, Nguyen DV. The impact of public voluntary health insurance on private health expenditures in Vietnam. Social Sci Med 2003; 56: 333-342.
11. Somanathan A, Tandon A, Dao HL, Hurt KL, Fuenzalida-Puelma HL. Moving toward universal coverage of social health insurance in Vietnam: assessment and options. World Bank Publications, 2014.
12. Cramton P, Katzman B. Reducing healthcare costs requires good market design. The Economists’ Voice 2010; 7: 1-5.
13. Lagomarsino G, Garabrant A, Adyas A, Muga R, Otoo N. Moving towards universal health coverage: health insurance reforms in nine developing countries in Africa and Asia. The Lancet 2012; 380: 933-943.
14. Lofgren C, Thanh NX, Chuc NT, Emmelin A, Lindholm L. People's willingness to pay for health insurance in rural Vietnam. Cost Eff Resour Alloc 2008; 6: 16.
15. Wagstaff A, Nguyen HTH, Dao H, Bales S. Encouraging health insurance for the informal sector: a cluster randomized experiment in Vietnam. Health Economics 2016; 25: 663-674.
16. Mitra S, Palmer M, Mont D, Groce N. Can Households Cope with Health Shocks in Vietnam? Health Econ 2016; 25: 888-907.
17. Vuong QH, Napier NK. Resource curse or destructive creation in transition: evidence from Vietnam’s corporate sector.Management Res Rev 2014; 37: 642-657.
18. Agresti A. Categorical Data Analysis. John Wiley & Sons, Hoboken, New Jersey.
19. Penn State Science. Analysis of discrete data: Further topic on logistic regression, 2016.
20. Vuong QH, Nguyen TK. Data on Vietnamese patients’ financial burdens and risk of destitution. Data Brief 2016.
21. Lu JF, Hsiao WC. Does universal health insurance make health care unaffordable? Lessons from Taiwan. Health Aff (Millwood) 2003; 22: 77-88.
22. Rodwin VG. The health care system under French national health insurance: lessons for health reform in the United States. Am J Public Health 2003; 93: 31-37.
23. Vuong QH; Nguyen TK. Vietnamese patients’ choice of healthcare provider: in search of quality information. Int J Behav Healthcare Res 2015; 5: 184-212.
24. Mohammed S, Souares A, Bermejo JL, Babale SM, Sauerborn R, Dong H. Satisfaction with the level and type of resource use of a health insurance scheme in Nigeria: health management organizations’ perspectives. Int J Health Plan Manage2014; 29: e309-e328.
25. Zhou X, Mao Z, Rechel B, Liu C, Jiang J, Zhang Y. Public vs. private administration of rural health insurance schemes: a comparative study in Zhejiang of China. Health Economics Policy Law 2013; 8: 269-288.
26. Robyn PJ; Sauerborn R, Barnighausen T. Provider payment in community-based health insurance schemes in developing countries: a systematic review. Health Policy Plan 2013; 28: 111-122.
27. Jerant A, Fenton JJ, Bertakis KD, Franks P. Satisfaction with health care providers and preventive care adherence: a national study. Med Care 2014; 52: 78-85.
28. Habib SS, Perveen S, Khuwaja HMA. The role of micro health insurance in providing financial risk protection in developing countries: a systematic review. BMC Public Health 2016; 16: 281.
29. Hoang VM, Oh J, Tran TA, Tran TG, Ha AD. Patterns of Health Expenditures and Financial Protections in Vietnam 1992-2012. J Korean Med Sci 2015; 30 Suppl 2: S134-138.
30. Nguyen C. The impact of health insurance programs for children: evidence from Vietnam. Health Econ Rev 2016; 6: 34.
31. Drummond MF, Sculpher MJ, Claxton K, Stoddart GL, Torrance GW. Methods for the economic evaluation of health care programmes (4th Ed). Oxford University Press, New York, 2015.