Effect of supplementary private health insurance on out-of-pocket inpatient medical expenditure: evidence from Malaysia

Abstract

The vital role of healthcare financing in achieving universal health coverage is indisputable. However, most countries, including Malaysia, face challenges in establishing an equitable and sustainable healthcare financing system due to escalating healthcare costs, an ageing population and a growing disease burden. With desirable pre-payment and risk pooling features, private health insurance (PHI) is considered an alternative financing option to reduce out-of-pocket (OOP) medical expenditure. However, ongoing theoretical and empirical debates persist regarding the adequacy of financial risk protection provided by PHI largely because it depends on its role, the benefit design and the regulations in place. Our study aimed to investigate the effect of supplementary PHI on OOP inpatient medical expenditure in Malaysia. Secondary data analysis was conducted using the Malaysian National Health and Morbidity Survey 2019 dataset. A total of 983 respondents with a history of inpatient hospitalization in the past 12 months were included in the study. Instrumental variable analysis using a two-stage residual inclusion was performed to address endogeneity bias, with wealth status and education level as the instrumental variables. Tobit regression model was used in the second stage considering the censored distribution of the outcome variable. Missing data were handled using multiple imputation. About one-fifth of the respondents had PHI. In this study, we found that having PHI significantly increased OOP inpatient medical expenditure in all three marginal effects. Additionally, age, residential location, ethnicity (citizenship), being covered by government guarantee letter, government funding and employer-sponsored health insurance were other significant factors associated with OOP inpatient medical expenditure. Our findings undermine a key justification to advocate PHI uptake among the population, with a need for the Malaysian government to reassess the role of PHI in healthcare financing and reconsider PHI subsidization policy. Regulations should also be strengthened to enhance the financial risk protection provided by PHI.

Malaysia, private health insurance, out-of-pocket expenditure, financial risk protection, instrumental variable analysis, inpatient care

Key messages

• Private health insurance (PHI), which serves a supplementary role in Malaysia, significantly increased out-of-pocket inpatient medical expenditure.

• Our findings raised fundamental questions about the quality of benefit packages and the adequacy of financial risk protection provided by PHI.

• The Malaysian government should re-examine the rationale of the ongoing subsidization of PHI purchases and the explicit encouragement of population-wide enrolment.

• The current governance and regulatory framework for PHI should be reviewed and strengthened to enhance its financial protective effect.

• Despite facing fiscal constraints, the quest to achieve universal health coverage relying on PHI remains contentious and warrants careful consideration.

Introduction

Establishing an equitable and sustainable health financing system is paramount to achieving two main goals of universal health coverage (UHC), namely: (1) providing all people with access to needed health services, including prevention, promotion, treatment and rehabilitation of sufficient quality and (2) ensuring that using these services does not put the users to financial hardship (World Health Organization, 2010). To attain UHC, all World Health Organization (WHO) member states agreed to enhance the role of pre-payment for healthcare and reduce direct payments, which were considered impediments to healthcare access (Myint et al., 2019). Based on their citizens’ values, institutional capacity and the stages of economic development, most countries have a combination of healthcare financing sources, including government revenue, private health insurance (PHI), social health insurance and private out-of-pocket (OOP) payments.

PHI has assumed a more significant role in financing healthcare over the past few decades because of several factors, including the sustainability of existing health financing arrangements, shifting consumer demand with economic growth and the increased commodification of healthcare services (Drechsler and Jutting, 2007). There has been a long debate about whether PHI brings more harm or good to the overall health system, particularly in financing, health workforce and service delivery (Mossialos and Thomson, 2002; WHO Regional Office for Europe’s Health Evidence Network, 2004; European Observatory on Health Systems and Policies, 2020). PHI proponents believe that it will empower consumers and increase efficiency in health service delivery. PHI also encourages the wealthy to pay more, reducing pressure on public finances (WHO Regional Office for Europe’s Health Evidence Network, 2004). Furthermore, establishing a pre-payment healthcare financing mechanism through PHI can serve a transitional function in low- and middle-income countries as they consider moving towards public insurance systems (Foubister et al., 2006; European Observatory on Health Systems and Policies, 2020). However, PHI has been criticized for its inability to hold down healthcare inflation due to a limited control on cost containment and higher administrative costs (Nik Rosnah and Ng, 2009). Additionally, it permits cream skimming and adverse selection, raising the possibility that the social and financial elements of health protection are neglected (Drechsler and Jutting, 2007; Wu and Ercia, 2021).

OOP payments are widely recognized as regressive and obvious obstacles to attaining UHC. Conversely, the impact of PHI is less conclusive because it is influenced by numerous factors, such as countries’ legislation, the role it plays and its design (Sekhri and Savedoff, 2004; Xu et al., 2005, 2011; World Health Organization, 2010; World Bank, 2019; European Observatory on Health Systems and Policies, 2020). Historically, PHI has been characterized as voluntary and operated for commercial gain. Nevertheless, the distinctions between public health insurance and PHI are increasingly blurred as a result of various healthcare reforms undertaken in many countries, particularly regarding the regulations on health insurance participation, the mechanism of rating insurance premium contributions and the governance of the administering entity (Organisation for Economic Co-Operation and Development, 2004). For this study, PHI is defined as the channelling of financial resources, paid voluntarily by individuals, directly to the private insuring entity to cover a defined set of health services and benefits (Organisation for Economic Co-Operation and Development, 2004; Drechsler and Jutting, 2007; European Observatory on Health Systems and Policies, 2020).

PHI mitigates the financial risk associated with individuals’ healthcare expenditures through pre-payment and pooling. Pre-payment refers to the insurance premiums paid in advance to cover any potential healthcare expenditure in the future. Pooling, on the other hand, is the process where financial resources for healthcare are accumulated and transferred to insurance entities to spread the risk of funding healthcare treatment across the entire pool rather than being borne by a single individual (Organisation for Economic Co-Operation and Development, 2004). The foundational principle of PHI is straightforward: individual healthcare needs can be unpredictable and financially burdensome, yet most individuals do not require healthcare simultaneously. As such, through the collective sharing of the financial risk associated with substantial healthcare expenditure among a diverse population, PHI effectively renders healthcare accessible and affordable to all, reducing the likelihood of OOP expenditures (Chollet Deborah and Lewis, 1997; World Bank, 2012).

Nevertheless, some studies found that PHI increased OOP spending and the occurrence of catastrophic healthcare expenditure due to several possible reasons (Ekman, 2007; Barros et al., 2011; Paccagnella et al., 2013). Individuals with PHI had a greater propensity to utilize more healthcare services, a phenomenon known as moral hazard. The practice, albeit denounceable, is an efficient consumer behaviour as individual uses more healthcare when PHI makes it more affordable. Furthermore, the increase in healthcare service utilization may also arise from the tendency of healthcare providers to offer more services than they otherwise would in the absence of PHI (Sekhri and Savedoff, 2006; World Bank, 2012). To address this, PHI plans commonly incorporate cost-sharing mechanisms and limits to encourage judicious use of healthcare. Ironically, when uniformly applied without considering the socioeconomic background, such measures may expose the insured to financial risks (Chollet Deborah and Lewis, 1997). Another contributing factor is adverse selection, wherein individuals with existing health conditions or higher risk factors are more likely to purchase PHI, given their anticipation of greater medical needs. Moreover, they commonly encounter more stringent underwriting processes due to their health status, resulting in higher premiums, more restricted benefit packages and the exclusion of pre-existing conditions from coverage (Chollet Deborah and Lewis, 1997; World Bank, 2012).

From a global perspective, the relationship between PHI and OOP health spending was weak, as observed from a correlation analysis performed across 186 countries (European Observatory on Health Systems and Policies, 2020). Despite many countries having high levels of OOP expenditure, implying significant coverage gaps, PHI only contributed a modest amount to the total healthcare expenditure, ranging from 2.4% in lower-middle-income countries to 6.3% in upper-middle-income countries. In only 23 countries worldwide, PHI accounted for more than 10% of current health spending, with the majority being middle-income countries in the Caribbean, Latin America and Africa (European Observatory on Health Systems and Policies, 2020).

Empirically, a relationship between PHI and OOP has not been clearly established. PHI has been found to reduce catastrophic health expenditure in a few countries, namely China, South Korea, Germany and Greece (Grigorakis et al., 2017; Lee et al., 2018; Blümel et al., 2020; Fu, 2021). However, in studies conducted in Brazil and Zambia, households with PHI were exposed to greater catastrophic health expenditure risk than those uninsured (Ekman, 2007; Barros et al., 2011). Paccagnella et al. (2013), who investigated the effects of PHI on healthcare expenses in 11 European countries, also obtained non-uniform findings. In the Netherlands, the researchers found that OOP medical expenditure was significantly reduced with having PHI. Conversely, owning PHI significantly increased the levels of OOP payments in Denmark, Austria, Spain and Italy. Notably, assessing the financial protective effects of PHI necessitates a nuanced understanding of its role within the healthcare system and the regulations and policies in place (European Observatory on Health Systems and Policies, 2020). The three main roles that PHI serves are as follows: (1) supplementary, which provides an alternative mechanism with faster access to healthcare services, greater choice of provider or enhanced amenities, usually covering the same range of services offered in the public healthcare sector; (2) substitutive, which covers certain populations not eligible for publicly financed healthcare services; and (3) complementary, which fills the gaps due to deficiencies in the public healthcare system’s scope and depth of coverage (World Bank, 2012; Sagan and Thomson, 2016).

PHI in Malaysia

The Malaysian healthcare system is dichotomous, comprising public and private healthcare sectors. The Ministry of Health (MOH) is the principal public healthcare provider, offering primary, secondary and tertiary healthcare services through various health facilities (Yu et al., 2008; Juni, 2014). Conversely, private healthcare providers complement public healthcare services by providing care that is typically curative. WHO has lauded Malaysia’s healthcare system as a ‘low-cost healthcare system that provides universal and comprehensive services’ (Jaafar et al., 2013). Some health analysts posit that Malaysia achieved UHC as early as the 1980s, predominantly evident in metrics such as life expectancy and infant mortality rate, of which the success was attributed to the successful control of communicable diseases and the notable improvements in maternal and child health (Savedoff William and Smith Amy, 2011).

While the public healthcare system achieved its initial objectives, it intentionally left a void in the health services sector for private facilities catering to the upper and middle classes willing to pay for expedited services and better facilities (Chee and Barraclough, 2007). This dichotomous public–private healthcare construct evolved, with the private healthcare sector now contributing half of the total healthcare expenditures (Ministry of Health Malaysia, 2020). Nevertheless, Malaysia’s earlier accomplishments are already showing signs of unsustainability due to shifting epidemiological needs and fiscal considerations. Decades of underspending in the public health system, with expenditures hovering around 2% of the gross domestic product (GDP), fall significantly below the spending benchmark of an upper-middle-income economy (4.0% to 4.5% of the GDP) (Ahmad, 2019). The low public healthcare expenditure has resulted in inadequate and ageing public facilities, poor remuneration for the health workforce and underinvestment in health information technology. Additionally, escalating medical cost inflation has contributed to a public hospital admission rate double that of the private sector, exacerbating congestion, heavy workloads and prolonged waiting time.

Supplementary PHI in Malaysia provides alternative access to healthcare treatment in private facilities, which are commonly perceived to be better equipped, deliver higher quality care and offer the freedom to choose one’s preferred healthcare providers. PHI also plays a less prominent complementary role by covering treatments not provided in public healthcare facilities, notably the costlier and latest medical treatments (Kaur et al., 2017). The purchase of PHI is voluntary, usually by individuals or in a group, as part of the job benefits provided by employers. As stipulated by the Financial Services Act 2013, PHI companies are regulated by the Central Bank of Malaysia, which is largely concerned with financial solvency and does not place significant restrictions on risk-rating, underwriting practices and benefit design (Jarrah, 2018). On the other hand, the Private Healthcare Facilities and Services Act 1998 confers power to the MOH to govern the provision of healthcare services in the private sector. The Act regulates the doctors’ and surgical fees, but it does not extend regulatory oversight to other hospital supplies and services expenses, which include laboratory fees, imaging, pharmaceuticals and nursing services (Nik Rosnah and Ng, 2009).

The Malaysian government has promoted the expansion of PHI since the 1980s to supplement public healthcare financing, with healthcare being one of the sectors targeted for privatization (Chee and Barraclough, 2007; Nik Rosnah and Ng, 2009). The government granted income tax relief for PHI purchases as a further incentive. According to the latest National Health and Morbidity Survey (NHMS) in 2019, the percentage of the population covered by PHI increased from a little over 1.5% in 1983 to 22.4% (Nik Rosnah and Ng, 2009; Institute for Health Systems Research, 2020). In recent years, the government has maintained its stance of encouraging people, especially those who can afford it, to have PHI, with public healthcare subsidies suggested to be more targeted towards poorer households (Economic Planning Unit, 2021). Currently, the government extensively subsidizes public healthcare services covering all its citizens, and patients only need to pay a nominal fee.

Effect of PHI on OOP medical expenditure

Despite the increase in PHI coverage over the years, PHI still only funded 8% of the total health expenditure in Malaysia in 2019, with OOP spending contributing 35% (Malaysia National Health Accounts Section, 2021). To alleviate the effects of adverse selection, PHI premiums in Malaysia are allowed to be risk-rated, taking into account the individuals’ health status (Nik Rosnah and Ng, 2009; Kefeli@Zulkefli and Jones, 2012). Individuals with pre-existing comorbidities are deemed higher risk and will have to pay more expensive insurance premiums. Depending on their health conditions, some people may also have additional exclusions from their PHI policy, or they may not even be allowed to purchase PHI at all, such as those with cancer or living with human immunodeficiency virus (European Observatory on Health Systems and Policies, 2020; Kong et al., 2020; The Malaysian AIDS Foundation, 2020). Benefits for PHI are frequently capped at a certain amount for each individual service and, in many cases, an annual limit. In addition, PHI, especially individual insurance, typically only cover hospital inpatient care, and patients are expected to pay the hospital bill first before submitting a claim for reimbursement. The reimbursement process may take a long time, coupled with the uncertainty of how much of the healthcare expenses paid are eventually covered (Nik Rosnah and Ng, 2009; Kefeli@Zulkefli and Jones, 2012; Kong et al., 2020).

Given these limitations, it was unsurprising that the OOP payments for healthcare expenses remained high. There were people insured with PHI who still reported catastrophic health expenditure (Bhoo-Pathy et al., 2019; Kadravello et al., 2021). This gives rise to concerns whether PHI provides adequate financial risk protection with sufficient benefit packages, potentially leading to patients experiencing financial distress while suffering from their medical conditions. Upon exhaustion of their insurance coverage, some may even forgo recommended treatment due to unaffordability or continue treatment at public healthcare facilities which are often congested and less well equipped (Atun et al., 2016; Yap et al., 2019; Kong et al., 2020; Kadravello et al., 2021). Having said that, for some patients, PHI still plays a vital role in mitigating financial risk from their experience, particularly for catastrophic illness (Kong et al., 2020; Kadravello et al., 2021). According to a study conducted among cancer patients in Malaysia, those without health insurance were twice as likely to suffer from catastrophic health expenditures (Bhoo-Pathy et al., 2019).

Currently, there is scant evidence about the financial risk protection provided by PHI in Malaysia, which is often limited to qualitative studies conducted among cancer patients. Existing studies mainly investigate the factors associated with the purchase of PHI and factors associated with catastrophic health expenditure (Abu Bakar et al., 2012; Sharifa Ezat et al., 2012; Ellyana et al., 2020; Sukeri and Sayuti, 2020; Balqis-Ali et al., 2021). This evaluation is timely and much needed as the government continues encouraging the population to purchase PHI, while efforts are made to revamp the healthcare subsidies to be more targeted. In the hope of achieving UHC while facing fiscal constraints, this study could also guide policymakers in other countries in deciding their health financing arrangements, whether supplementary PHI is a viable option. Thus, this study aimed to assess the adequacy of financial risk protection provided by supplementary PHI in Malaysia by examining its effect on OOP inpatient medical expenditure.

Methods

Study population and design

We used data from the NHMS 2019. This nationwide community-based cross-sectional survey collected information on socio-demographics, non-communicable disease status and healthcare utilization (Institute for Health Systems Research, 2020). The NHMS 2019 applied a two-stage stratified cluster sampling design based on the Department of Statistics Malaysia’s sampling frame, from which over 75 000 enumeration blocks (EBs) with each containing 80–120 living quarters (LQs) were considered. An average of 500 to 600 people resided in each LQ. In total, 350 and 113 EBs were selected from urban and rural areas across all states in Malaysia, yielding 6482 eligible LQs for the study. From each selected EB, 14 LQs were randomly chosen. This study included all eligible household members in the selected LQs. The survey was conducted via face-to-face interviews at respondents’ places of residence between July and October 2019. Each participant or guardian provided informed written consent, and participants aged 7 to 18 years signed an additional assent form. For the present analysis, 983 out of the total 16 688 respondents with a history of inpatient admission in the last 12 months before the survey were included. The analysis was performed at the individual level.

Measurement

Dependent variable

The outcome variable was the respondents’ total OOP inpatient medical expenditure for all visits, which comprised payments for consultation, medications, procedures, tests and appliances, including implants, intra-ocular lenses and wheelchairs. Our study focused on inpatient medical expenditure, as an individual PHI usually covers treatments requiring hospitalization. Only payments made by own money/family/household members were recorded, excluding reimbursements made by the insurance/employer. The OOP inpatient medical expenditures did not account for indirect costs, such as lost wages and transportation expenses.

Independent variables

The explanatory variable of main interest was whether the respondent had supplementary PHI. Respondents with PHI were defined as individuals who owned either a stand-alone or combined PHI with other payer options, taking the value of 1 if the respondents had PHI and 0 otherwise. Ten respondents had a missing value for this variable.

The decision of whether to purchase supplementary PHI and the incurrence of OOP inpatient medical expenditure may be associated with socio-demographic, economic and health status. The following covariates were included in the analysis: age, sex (male and female), ethnicity (Malay, Chinese, Indian, aborigines/natives of Sabah and Sarawak and others), marital status (single/widow/widower and married), residential location (urban and rural), household size, occupation (government employee, private employee, self-employed and unpaid worker/housewife/not working), perceived health status (excellent, good, fairand poor/very poor), having diabetes mellitus, hypertension or hypercholesterolemia and whether they were covered by government guarantee letter (GL), government health funding programme, employer-sponsored health insurance and social security organization (SOCSO). Government GL refers to the assurance that the government bears all medical expenses incurred by a government employee and his or her immediate dependents in public healthcare facilities. The government health funding programme is part of the government’s financial assistance initiatives targeted at helping poor households receive healthcare services. On the other hand, SOCSO is a social security protection that provides employees with free medical treatment, rehabilitation and financial assistance if accidents or diseases have reduced their ability to work or rendered them incapacitated.

There was a large amount of missing data (170 respondents) for perceived health status, as the survey question was to be answered by respondents aged ≥13 years, leading to missing completely at random. Multiple imputation was performed to account for the missing observations rather than dropping the variable, which was an important confounder.

Instrumental variables

In the analysis of the effect of PHI, the validity of estimates is often questioned because of potential endogeneity bias caused by unmeasured confounders and reverse causality between explanatory and outcome variables. Adverse selection is a case in point, whereby sick people with high OOP expenditures may have a greater tendency to purchase PHI. This scenario may lead to a biased estimation of the true effects of insurance on OOP medical expenditures, as some of these confounding factors are unobservable to researchers.

Instrumental variable (IV) analysis is commonly used to address endogeneity bias (Aso and Yasunaga, 2020; Wooldridge, 2013). IV analysis mimics the treatment assignment process in a randomized controlled study. Several essential criteria need to be fulfilled to be a valid instrument, which are as follows: (1) the IV should be strongly associated with the explanatory variable of main interest, in this case, the PHI status; (2) the IV is not associated with unmeasured confounders after conditioning on measured confounders; and (3) the IV affects the outcome variable only through the explanatory variable, which is known as ‘exclusion restriction’ (Ertefaie et al., 2017; Aso and Yasunaga, 2020). After performing instrument relevance tests among variables potentially fulfilling the criteria, two IVs were identified and employed in our regression model. The first IV was education level, categorized as no formal education, primary, secondary or tertiary education. The second IV was household income, categorized into the Top 20%, Middle 40% and Bottom 40%. Household income was calculated based on each individual’s self-reported income and categorized using state-specific cut-off values obtained from the Department of Statistics Malaysia.

Statistical analysis

We described the characteristics of the study population, categorized according to PHI status. We used a Tobit model implemented with a two-stage residual inclusion (2SRI) framework to analyse the effect of supplementary PHI on OOP inpatient medical expenditure.

The outcome variable, OOP inpatient medical expenditure, was a continuous variable with substantial observations having a value of 0. Models with such dependent variables were censored at 0 and characterized by nonlinearity. To account for the censored distributions of the outcome, Tobit model was employed for the study. Tobit model utilizes the maximum likelihood estimation method based on the censored density functions of the outcome variable with limiting values (Mcdonald and Moffitt, 1980; Wooldridge, 2013). The Tobit model in our study was:

$${y^*} = {\beta _0} + x\beta + u,u|x\,\sim\,N\left( {0,{\sigma ^2}} \right)$$

(1)

$$y = \,\left\{ {\begin{array}{*{20}{c}} {\,{y^*}\,if\,{y^*}\, \gt 0}\\ {0\,\,if\,{y^*}\,\,\, \le 0} \end{array}} \right.$$

(2)

where |${y^*}$| represents the latent variable of OOP inpatient medical expenditure and |$y$| is the observed variable.

The 2SRI approach, proposed by Terza and widely adopted in health economics, was used to implement IV estimation within the framework of the Tobit model (Terza et al., 2008; Terza, 2018). Simulations have demonstrated that the 2SRI estimator can generate consistent estimates in nonlinear models. The 2SRI-Tobit framework took the following structure:

$$y = M\left( {{\beta _e}\,PHI + \,{\beta _0}{x_0} + \,{\beta _u}{x_u}} \right) + e$$

(3)

$$PHI = N\left( {Z\,\alpha } \right) + \,{x_u}\,,\,Z = \,\left[ {{x_0}\,,\,\,IVs} \right]$$

(4)

where |$PHI$| represents private health insurance status, which is the endogenous variable, |${x_0}$| represents a matrix of exogenous covariates, |${x_{\rm{u}}}$| represents the unobserved confounders, |$e$| represents the regression error term, and |$\alpha $| is a column vector of regression parameters. In the first stage of 2SRI, a Probit regression was performed to estimate the probability of having PHI on the exogenous covariates and IVs and generated the residuals |$\widehat {{x_u}}$|⁠. By substituting |$\widehat {{x_u}}$| from Equation (4) for the unobserved confounders |${x_u}$| in Equation (3), consistent estimates of PHI could be generated if the IVs are valid. The estimation of standard errors should account for uncertainty in both stages of the 2SRI approach. To that end, we bootstrapped the two-stage regressions with 500 repetitions to obtain more accurate standard errors (Palmer et al., 2017). To demonstrate the consequence of addressing the endogeneity effect, we compared the results from the Tobit regression with and without applying the 2SRI approach.

We conducted multiple imputation to address missing data for the perceived health status variable. The combined application of bootstrap inference and multiple imputation has garnered attention and discussions in recent years (Schomaker and Heumann, 2018; Brand et al., 2019; Bartlett and Hughes, 2020). We first applied multiple imputation using Rubin’s rules, followed by bootstrapping for each imputed dataset to estimate the within-imputation complete data variance (Schomaker and Heumann, 2018; Bartlett and Hughes, 2020). Ten imputations were performed in line with the claim that 5 to 10 imputed datasets are sufficient to attain adequate efficiency (Rubin, 1987; Siddique and Belin, 2008).

Marginal effects on the expected value of the latent variable, expected value condition on non-zero OOP inpatient medical expenditure and censored expected value were calculated to achieve meaningful interpretations, denoted as ME1, ME2 and ME3, respectively. ME1 refers to the average marginal effect of the independent variables on the expectation of the latent variable y*. In the present analysis, it represents the influence of the independent variables on OOP inpatient medical expenditure. ME2 represents the average marginal effects of independent variables on the expectation of the outcome variable y conditioning on y>0, which can be alternatively interpreted as the effect of the independent variables on the OOP inpatient medical expenditure in the sample with observed non-zero OOP inpatient medical expenditure. Moreover, ME3 indicates the average marginal effects of the independent variables on the expectation of the actual outcome variable y, unconditionally. It explains the effect of the independent variables on the observed OOP inpatient medical expenditure. The three marginal effects were calculated as follows:

$${\rm{ME}}1 = \,\frac{{\partial E\left[ {{y^*}|x} \right]}}{{\partial x}} = \,\beta $$

(5)

$${\rm{ME}}2 = {{\,}}\frac{{\partial E[y\,,\,y \gt 0|x]}}{{\partial x}} = \,\left\{ {1 - \,\lambda \left( \omega \right)\left[ {\omega + \,\lambda \left( \omega \right)} \right]} \right\}\beta $$

(6)

$${\rm{ME}}3 = \,\frac{{\partial E[y|x]}}{{\partial x}} = \,{{\Phi }}\left( \omega \right)\beta $$

(7)

where |$\omega = \,\frac{{x{^{^{\prime}}}\,\beta }}{\sigma }$|⁠.

Tests for endogeneity for regressor, PHI status in our analysis, were performed to see if instrumentation was warranted. The assessment was achieved using the Durbin and Wu–Hausman endogeneity tests, with a finding of non-significance (P > 0.05) indicating that the regressor was exogenous and not needing instrumentation (Babington and Cano-Urbina, 2016; Garson, 2018). In addition, over-identification and weak IV tests were also conducted to test the validity of the IVs. Over-identification test was done to check if the instruments were valid and not correlated with the regression error term of the outcome variable, with the Sargan test being the most common (Baum et al., 2003; Garson, 2018). The null hypothesis of the test was that both IVs in the models were exogenous. On the other hand, two tests were performed to rule out weak instrumentation. The first test was performed by examining the F-statistic of the linear probability model obtained from the first-stage regression, where a value of <10 was suggested as the rule of thumb, indicating a weak instrument (Baum et al., 2007; Garson, 2018, Stock et al., 2002). The second test, also known as the Stock–Yogo relative bias test, uses the Cragg–Donald Wald F-statistic, where a minimum eigenvalue of >15% rejection level cut-off indicating that instrumentation is not weak (Stock and Yogo, 2005; Baum et al., 2007; Garson, 2018).

All analyses were performed with Stata/SE software version 16 (Stata Corp, College Station, TX, USA).

Results

Descriptive results

Table 1 shows the baseline characteristics of respondents based on their PHI ownership status. Among the 973 respondents, 21.48% reported having PHI. Compared to those without PHI, PHI owners were significantly younger (mean: 32.73 vs 40.69 years, P < 0.001) and healthier without comorbidities, for instance, diabetes mellitus (89.76% vs 78.84%, P < 0.001), hypertension (81.07% vs 70.47%, P = 0.002) and hyperlipidaemia (86.83% vs 74.93%, P < 0.001). A larger proportion of PHI owners reside in urban areas (79.43% vs 54.97%, P < 0.001), had other health payers options, such as government GL (29.81% vs 22.73%, P = 0.04), employer-sponsored health insurance (31.73% vs 13.80%, P < 0.001) and SOCSO (23.56% vs 9.45%, P < 0.001), received tertiary education (39.71% vs 15.22%, P < 0.001), belonged to the Top 20% wealth status (23.44% vs 5.37%, P < 0.001) and were Chinese or Indian (38.75% vs 11.78%, P < 0.001). For occupation, a larger percentage of those who were unpaid workers, housewives and not in active employment were without PHI (69.90% vs 47.37%, P < 0.001).

Table 1.

Respondents’ characteristics based on PHI status (N = 973)

	Have PHI (n = 209, 21.48%)		No PHI (n = 764, 78.52%)
Variables	n	%	N	%	P-value
Sex					0.813
Male	84	40.19	314	41.10
Female	125	59.81	450	58.90
Age, in years					<0.001
Mean (SD)	32.73 (19.09)	40.69 (23.69)
Ethnicity					<0.001
Malay	119	56.94	553	72.38
Chinese	44	21.05	40	5.24
Indian	37	17.70	50	6.54
Aborigines/Natives of Sabah and Sarawak	8	3.83	107	14.01
Others	1	0.48	14	1.83
Marital status					0.758
Single/widow/widower	49	23.44	187	24.48
Married	160	76.56	577	75.52
Education level (n = 971)					<0.001
No formal education	33	15.79	141	18.50
Primary	29	13.88	215	28.22
Secondary	64	30.62	290	38.06
Tertiary	83	39.71	116	15.22
Occupation					<0.001
Government employee	31	14.83	54	7.07
Private employee	55	26.32	104	13.61
Self-employed	24	11.48	72	9.42
Unpaid worker/housewife/not active	99	47.37	534	69.90
Household income					<0.001
Bottom 40%	78	37.32	569	74.48
Middle 40%	82	39.23	154	20.16
Top 20%	49	23.44	41	5.37
Strata					<0.001
Urban	166	79.43	420	54.97
Rural	43	20.57	344	45.03
Perceived health status (n = 805)					<0.001
Excellent	30	17.96	70	10.97
Good	90	53.89	272	42.63
Fair	39	23.35	233	36.52
Poor/very poor	8	4.79	63	9.87
Diabetes mellitus (n = 966)					<0.001
No	184	89.76	600	78.84
Yes	21	10.24	161	21.16
Hypertension (n = 968)					0.002
No	167	81.07	537	70.47
Yes	39	18.93	225	29.53
Hyperlipidaemia (n = 963)					<0.001
No	178	86.83	568	74.93
Yes	27	13.17	190	25.07
Household size			0.097
Mean (SD)	3.83 (1.94)	4.10 (2.14)
Government GL (n = 969)					0.035
No	146	70.19	588	77.27
Yes	62	29.81	173	22.73
Government health funding programme (n = 971)					0.099
No	202	97.58	725	94.90
Yes	5	2.42	39	5.10
Employer-sponsored health insurance (n = 969)					<0.001
No	142	68.27	656	86.20
Yes	66	31.73	105	13.80
SOCSO (n = 970)					<0.001
No	159	76.44	690	90.55
Yes	49	23.56	72	9.45

Abbreviation: SD, standard deviation.

Determinants of PHI ownership status

Based on the first-stage Probit regression predicting PHI ownership status (Table 2), we found that age, ethnicity, education level, occupation, wealth status, household size and being covered by SOCSO were statistically significant determinants of PHI ownership. The Probit regression model showed a good fit to the data, as demonstrated by the Hosmer–Lemeshow test (P = 0.27), classification table which correctly classified 84.09% and area under ROC curve of 0.85. The mean of the variance inflation factor (VIF) was 2.22, with the VIF for all independent variables <10, ranging from 1.04 to 5.55, suggesting that multicollinearity was unlikely.

Table 2.

First-stage Probit regression on PHI status

	Univariable analysis		Multivariable analysis
Variables	Coefficient	SE	Coefficient	SE
Ethnicity
Malay (ref)
Chinese	0.986***	0.148	1.386***	0.187
Indian	0.738***	0.147	0.898***	0.176
Aborigines/Natives of Sabah and Sarawak	−0.552***	0.186	−0.636***	0.215
Others	−0.575	0.501	−0.373	0.538
Sex
Male (ref)
Female	0.022	0.092
Age	−0.009***	0.002	−0.021***	0.003
Marital status
Single/widow/widower (ref)
Married	0.033	0.106
Education level
No formal education (ref)
Primary	−0.302**	0.151	−0.085	0.209
Secondary	−0.033	0.134	0.126	0.192
Tertiary	0.670***	0.142	0.617***	0.204
Occupation
Government employee (ref)
Private employee	−0.050	0.173	−0.535**	0.225
Self-employed	−0.329*	0.197	0.008	0.239
Unpaid worker/housewife/ not active	−0.663***	0.151	−0.303	0.191
Household income
Bottom 40% (ref)
Middle 40%	0.780***	0.105	0.797***	0.128
Top 20%	1.284***	0.147	1.422***	0.184
Residential location
Urban (ref)
Rural	−0.648***	0.101
Household size	−0.037*	0.022	−0.154***	0.032
Perceived health status
Excellent (ref)
Good	−0.151	0.173
Fair	−0.513***	0.174
Poor/very poor	−0.682***	0.253
Government GL
No (ref)
Yes	0.214**	0.103
Government funding programme
No (ref)
Yes	−0.428*	0.253
Employer-sponsored insurance
No (ref)
Yes	0.633***	0.110
SOCSO
No (ref)
Yes	0.647***	0.125	0.690***	0.192
Diabetes mellitus
No (ref)
Yes	−0.475***	0.131
Hypertension
No (ref)
Yes	−0.331***	0.108
Hyperlipidaemia
No (ref)
Yes	−0.442***	0.120

***

P < 0.01,

**

P < 0.05,

*

P < 0.10.

Effect of supplementary PHI on OOP inpatient medical expenditure

Tobit model was used to determine the effect of supplementary PHI on OOP inpatient medical expenditure due to the censored outcome distributions, with three distinct sets of marginal effects calculated (Table 3). The marginal effect of PHI on the latent variable (ME1) was 3415.61 Malaysian Ringgit (MYR)¹ (95% CI 170.58, 6660.64; P = 0.04), which indicated that supplementary PHI had a positive correlation with OOP inpatient medical expenditure. Additionally, the marginal effect of PHI on the expected OOP inpatient medical expenditure conditioned on non-zero expenses (ME2) was 1399.22 MYR (95% CI −161.00, 2959.43; P = 0.08). Lastly, the marginal effect of PHI on the expected value of the observed OOP inpatient medical expenditure (ME3) was 1101.69 MYR (95% CI −103.14, 2306.32; P = 0.07). The statistical significance of all marginal effects confirmed the effect of supplementary PHI on OOP inpatient medical expenditure, whereby having PHI increased OOP inpatient medical expenditure. In addition, other variables, such as residing in rural areas and being covered by government GL, government funding and employer-sponsored insurance, had significantly lower OOP inpatient medical expenditure. On the other hand, increasing age and ethnicity, specifically, the ‘others’ subcategory comprised mainly non-citizens compared to the ‘Malay’ subcategory, had significantly higher OOP inpatient medical expenditure.

Table 3.

Marginal effects of Tobit model on OOP inpatient medical expenditure

	ME 1		ME 2		ME 3
Variables	ME	SE	ME	SE	ME	SE
PHI
No (ref)
Yes	3415.61**	1655.66	1399.22*	796.04	1101.69*	614.67
Ethnicity
Malay (ref)
Chinese	427.93	985.28	147.70	358.40	123.01	293.32
Indian	−946.38	1021.59	−283.43	290.70	−250.79	263.32
Aborigines/Natives of Sabah and Sarawak	693.06	437.13	245.58	161.57	202.41	131.59
Others	3480.28***	1295.00	1583.61**	761.11	1205.67**	561.56
Sex
Male (ref)
Female	567.41	439.16	189.11	145.07	159.90	123.51
Age	23.02*	13.75	7.76*	4.68	6.53*	3.94
Marital status
Single/widow/widower (ref)
Married	−84.83	482.65	−28.708	164.49	−24.13	138.28
Occupation
Government employee (ref)
Private employee	336.98	1136.84	102.02	337.29	89.91	300.27
Self-employed	520.78	1130.80	160.78	339.68	140.46	300.69
Unpaid worker/housewife/ not active	870.44	1076.88	278.90	319.54	239.80	284.51
Residential location
Urban (ref)
Rural	−841.56**	−420.37	−278.70**	137.46	−236.15**	117.25
Household size	153.25	96.57	51.64	32.62	43.48	27.55
Perceived health status
Excellent (ref)
Good	−352.01	510.34	−119.33	174.99	−100.23	146.60
Fair	210.59	626.24	74.29	222.32	61.41	183.85
Poor/very poor	−1057.69	1325.40	−332.13	401.57	−288.11	354.23
Government GL
No (ref)
Yes	−3547.10***	1057.79	−967.77***	245.40	−895.84***	245.28
Government funding programme
No (ref)
Yes	−1851.03	1280.88	−522.73*	299.37	−475.57	296.72
Employer-sponsored insurance
No (ref)
Yes	−1389.70**	658.58	−427.41**	185.18	−374.13**	168.83
SOCSO
No (ref)
Yes	355.02	709.45	122.83	252.35	102.32	208.78
Diabetes mellitus
No (ref)
Yes	157.89	819.35	53.73	285.45	45.06	238.02
Hypertension
No (ref)
Yes	537.41	795.76	185.55	281.16	154.70	233.29
Hyperlipidaemia
No (ref)
Yes	−709.93	839.73	−229.92	262.76	−196.79	228.83

***

P < 0.01,

**

P < 0.05,

*

P < 0.10.

Without addressing the endogeneity effect using the 2SRI approach, all three marginal effects obtained from the Tobit regression showed a similar direction of association (Table 4). However, they were not statistically significant.

Table 4.

Comparison of marginal effects of PHI on OOP inpatient medical expenditure with and without 2SRI

	ME 1		ME 2		ME 3
Analysis	ME	SE	ME	SE	ME	SE
With 2SRI	3415.61*	1655.66	1399.22†	796.04	1101.69†	614.67
Without 2SRI	732.26	599.02	257.45	219.44	212.99	178.72

*

P < 0.05,

†

P < 0.10.

Further tests

Test for endogeneity of regressor

Durbin and Wu–Hausman endogeneity tests were performed with a null hypothesis that the regressor was exogenous and did not need instrumentation. Both Durbin |$\left( {{x^2} = 12.08} \right)$| and Wu–Hausman (F = 11.92) tests were statistically significant (P < 0.001), thus rejecting the null hypothesis and confirming an endogenous relationship between PHI ownership status and OOP inpatient medical expenditure.

Over-identification test of IVs

Over-identification test using Sargan statistics was performed to determine if the IVs employed were valid and exogenous. The P-value for the Sargan test conducted |$\left( {{x^2} = 0.41} \right)$| was 0.52, confirming that the IVs were exogenous and not correlated with the regression error term.

Test of weak instruments

F-statistic of the linear probability model obtained from the first-stage regression was 36.06, more than the critical value of 10, thus ruling out weak instrumentation. The Stock–Yogo relative bias test conducted further confirmed that the IVs were not weak, where the minimum eigenvalue of 44.93 was >15% rejection level cut-off of 11.59.

Discussion

To our knowledge, our study is the first to examine the effect of supplementary PHI on OOP inpatient medical expenditure in Malaysia. Using a Tobit model implemented with the 2SRI approach to control for endogeneity bias, we found that supplementary PHI significantly increased OOP inpatient medical expenditure, as evidenced by all three marginal effects. This finding contrasts with another study conducted among cancer patients in Malaysia which found that health insurance (including PHI and employer-sponsored insurance) reduced OOP spending, specifically lowering the risk of catastrophic health expenditure by half (Bhoo-Pathy et al., 2019). However, the same study also pointed out that 20% of the cancer survivors reported experiencing catastrophic health expenditure despite having PHI. It is important to note that the study focused on cancer treatment which typically involves substantial costs, while our study investigated all inpatient admissions regardless of diagnosis. Similarly, Lee et al. (2018) found that households with PHI had a lower incidence of catastrophic health expenditure but again focused on cancer patients. Another study found that PHI significantly reduced OOP spending in the Netherlands. However, it served a complementary role in covering healthcare services excluded from the publicly financed benefits package, in contrast to the supplementary role it plays in Malaysia (Paccagnella et al., 2013). Other studies, which found that PHI provided a significant financial risk protective effect, investigated it in combination with other payer options (Grigorakis et al., 2017; Fu, 2021).

We postulated several potential reasons why supplementary PHI significantly increased OOP inpatient medical expenditure observed in our study. Owning PHI has been found to increase the likelihood of admission to private hospitals rather than public hospitals. This may also be the case in our study because of the perceived better quality of care, well-equipped facilities, shorter waiting time and less congestion (Eldridge et al., 2016; Doiron and Kettlewell, 2018; Rana et al., 2020; Balqis-Ali et al., 2023). A recent study conducted in Malaysia found that having PHI significantly increased private inpatient utilization (Balqis-Ali et al., 2023). However, admissions to private hospitals may lead to medical expenses not covered by PHI and unanticipated by the patients. These may include upgrade of room and board and non-medical miscellaneous expenses, such as entertainment systems and toiletries (Nik Rosnah and Ng, 2009). Treatments and medications deemed either not indicated or not cost-effective compared to the standard treatments by the insurers, or admissions that are not medically necessary are also excluded from PHI coverage. These exclusions are usually not clearly spelt out in the PHI policies and may differ under different scenarios leading to confusion. A systematic review conducted in China on the effects of PHI towards achieving UHC also found that most OOP expenses were due to these exclusions (Wu et al., 2020). Due to poor health insurance literacy, patients may also not fully comprehend their PHI policies, especially concerning the waiting period and coverage exclusions, such as pre-existing illness, congenital conditions and mental disorders and may end up paying OOP (Nik Rosnah and Ng, 2009; Kong et al., 2020; Call et al., 2021; Edward et al., 2021).

To reduce moral hazards associated with PHI, private health insurers may apply some cost sharing to individuals, either through co-payments, deductibles or co-insurance (Organisation for Economic Co-Operation and Development, 2004; Paccagnella et al., 2013; European Observatory on Health Systems and Policies, 2020). Depending on the level of cost sharing, this may represent a significant amount of OOP inpatient medical expenditure. The same reason has been postulated to explain why having PHI significantly increased OOP spending in Denmark, Austria, Italy and Spain (Paccagnella et al., 2013). Additionally, private health insurers typically impose ceilings on the amounts reimbursed to policyholders to limit their cost exposure. These measures are commonly implemented through annual and lifetime reimbursement limits, coverage amount restriction for specific services and maximum days of inpatient admission (Organisation for Economic Co-Operation and Development, 2004). Once the ceilings are reached, patients are left with no alternative but to cover their medical expenditures through OOP. With healthcare cost inflation and medical technology advancement, the probability of exceeding these limits increases, particularly for older PHI policies with lower ceilings and catastrophic illnesses requiring substantial spending (Kefeli@Zulkefli and Jones, 2012; Kong et al., 2020).

In addition, our study found several other variables significantly associated with OOP inpatient medical expenditure. Having coverage through employer-sponsored insurance was significantly associated with lower OOP expenditure, as group insurance with better risk spreading usually provides more comprehensive coverage and benefit packages with fewer exclusions. On the other hand, respondents covered by government GL and funding were also found to spend less on OOP, as they were more likely to be admitted to subsidized public hospitals. The same applies to respondents residing in rural areas who were inclined to visit public hospitals due to the distribution of hospitals and accessibility (Yap et al., 2019). Conversely, increasing age was significantly associated with greater OOP expenditure as advanced age was commonly related to higher healthcare utilization and the need for longer hospitalization due to their health conditions (Lisk et al., 2019). Besides that, non-citizens in Malaysia experienced significantly greater OOP inpatient medical expenditure because they were charged higher unsubsidized fees in public hospitals as per government policies. These individuals may also face greater challenges in navigating the healthcare system (Straiton and Myhre, 2017; Loganathan et al., 2019; 2020). Nevertheless, the sample size for non-citizens in our study was small. Our study also observed that health status generally was not significantly associated with OOP expenditure, perhaps indicating that Malaysia’s heavily subsidized healthcare system did confer adequate financial risk protection to the population for inpatient care (Chua and Cheah, 2012). In addition, the determinants of PHI ownership status found in our study, which was not our main interest, were comparable to other studies conducted in Malaysia (Balqis-Ali et al., 2021; Al-Sanaani et al., 2022; Abd Khalim et al., 2023).

In our study, we applied the 2SRI with instrumentation to address the endogeneity of PHI towards OOP medical expenditure, as demonstrated by the IV analysis. Despite some limitations and the need to fulfil key IV criteria, this approach possessed greater advantages than other alternatives, such as propensity score matching, which could only match the control group to the treatment group based on the observable variables and may result in a smaller sample size and reduced statistical power (Laborde-Casterot et al., 2015; Ertefaie et al., 2017; Aso and Yasunaga, 2020). The marginal effects for OOP inpatient medical expenditure were greater with instrumentation than with conventional Tobit regression, indicating that the true effect was more significant when endogeneity bias was addressed. In addition, we used multiple imputation to handle missing data in our study. While the perceived health status variable was missing completely at random, performing a complete case analysis would reduce the statistical power. Our study also involved bootstrapping to estimate standard errors considering the uncertainty in both stages of the 2SRI. Hence, a combination of bootstrap inference and multiple imputation was performed. Recent literature discussing the implications of applying bootstrapping in the context of missing data has shown that our analysis method could generate valid confidence intervals (Schomaker and Heumann, 2018; Bartlett and Hughes, 2020).

Our findings have important implications for healthcare financing planning and PHI regulation in Malaysia as efforts are made to reform its healthcare system inherited from the British since its independence (Ahmad, 2019; Ismail, 2023). The public healthcare system is widely believed to be chronically underfunded, with public healthcare expenditure far below the proposed spending for an upper-middle-income country (Ahmad, 2019). Considering the fiscal constraint, the government has stated its intention for PHI to complement and supplement public health expenditure (Economic Planning Unit, 2021). Nevertheless, urging the uninsured to purchase PHI without first ascertaining its financial protective effect would do injustice to the population. The rationale of subsidizing PHI purchase, which has already been criticized for benefiting the more affluent population, could be even more contentious if it fails to offer adequate financial risk protection (Chee and Barraclough, 2007).

Our study underscores the existence of notable areas requiring improvement concerning PHI in Malaysia. First, the governance structure overseeing the regulation and supervision of PHI may require strengthening, given its current sole oversight by the Central Bank of Malaysia. Acknowledging the intricate nature of PHI governance and its increasingly significant interaction with the public healthcare system, there is a compelling argument for the MOH to assume a more pivotal role (World Bank, 2012). Second, health insurance literacy, a relatively new concept involving the knowledge, ability and confidence in seeking, understanding, evaluating and using information related to health insurance, should be enhanced and assessed periodically (Bardy, 2023). Studies have shown that poor health insurance literacy hinders informed decision-making and effective use of health insurance, increasing the likelihood of forgoing necessary care and financial burdens (Adepoju et al., 2019). Third, the introduction of standardized benefit packages should be considered to facilitate easier comparison among PHI products and to safeguard consumers’ interest, akin to the essential health benefits in the USA (World Bank, 2012; Sagan and Thomson, 2016). In addition, options should be made available for policyholders to increase their insurance coverage levels in alignment with healthcare cost inflation, medical advancement and evolving personal circumstances, without the encumbrance of extensive underwriting scrutiny (World Bank, 2012). Finally, premium affordability should be closely monitored as it is a primary factor contributing to underinsurance (Lavarreda et al., 2011; Institute for Health Systems Research, 2020).

Our study could provide guidance to other countries searching for the optimum balance between public and private healthcare financing. Moreover, our findings regarding the effect of supplementary PHI on OOP medical expenditure contribute to the current body of knowledge about the adequacy of its financial protective effect, which remains controversial and inconclusive (World Bank, 2007; Paccagnella et al., 2013). This study also highlights areas in governing PHI which may require greater government intervention through policies, regulations and incentives. Nevertheless, we reiterate that the interpretation of the findings needs to consider the country’s context and the interaction of PHI with the overall healthcare system.

Our study has several limitations. First, there might be a possibility of information bias as the data were self-reported, especially for numerical variables such as OOP expenditure and income. Second, our study performed a secondary data analysis on a nationwide health survey, whose original interests were disease prevalence and healthcare demand. Thus, greater details about PHI, for example, insurance benefits packages, financial coverage and respondents’ health insurance literacy, were unavailable. Future studies capturing these details are warranted to provide a more comprehensive understanding of the financial protective effects of PHI. More granular data on OOP inpatient medical expenditure breakdown may also be helpful, in differentiating medical and non-medical expenses. Lastly, it may be argued that the IVs applied in our study may still correlate with unmeasured and residual confounders, such as health insurance literacy. However, the relationship between health insurance literacy and the IVs, which is highly context-specific, has not been extensively studied in Malaysia, with existing literature failing to show significant associations consistently (Adepoju et al., 2019; Feinburg et al., 2019; Mamun et al., 2021). Furthermore, the over-identification test we performed found that the IVs were exogenous. The IVs were also not significantly associated with the outcome variable.

Conclusions

Supplementary PHI in Malaysia significantly increased OOP inpatient medical expenditures. Our findings raised fundamental questions about the quality of benefit packages and the adequacy of financial risk protection provided by PHI. Moreover, the Malaysian government should re-examine the rationale of the ongoing subsidization of PHI purchases and the explicit encouragement of population-wide enrolment. The current governance and regulatory framework for PHI should also be reviewed and strengthened to enhance its financial protective effect. Moving forward, further exploration of this topic is imperative to provide crucial input for the Malaysian government in assessing the role that PHI should play in the country’s healthcare financing landscape, aligning with the objectives of UHC.

Data availability

The data underlying this article were provided by the MOH Malaysia by permission. Data will be shared on request to the corresponding author with permission of the MOH Malaysia.

Funding

This study was self-funded.

Acknowledgements

The authors would like to thank the MOH Malaysia for their permission to utilize the NHMS 2019 data for this study.

Author contributions

All authors contributed to the conception or design of the work. R.J.N. assisted with the data collection. All authors assisted with the data analysis and interpretation. R.J.N. assisted with drafting the article. W.Y.C., C.-W.N. and N.N.H. assisted with the critical revision of the article. All authors provided final approval of the version to be submitted.

Reflexivity statement

The authors comprised of one male and three females. Concerning seniority, R.J.N. is a doctoral student under the scholarship from the MOH Malaysia, supervised by W.Y.C., C.-W.N. and N.N.H. who are professors in the University of Malaya, Malaysia. All authors are based in Malaysia which is a middle-income country.

Ethical approval.

This study performed secondary data analysis using the NHMS 2019. The NHMS 2019 survey received ethical approval from the Medical Research and Ethics Committee, MOH Malaysia, bearing registration number NMRR-18-3085-44207.

Conflict of interest.

The authors have no conflict of interest to declare.

Footnotes

1.

1 USD = 4.14 MYR in 2019; median household income = 6338 MYR/month.

Author notes

‡

Both authors contributed equally to this paper.