Msc Economics
Faculty of Economics and Business
Specialization: Development Economics
Sustainability of Micro Health Insurance
A study on re-enrolment
Master Thesis
26
thof July, 2014
Lidwien Sol (10589031)
Abstract - The main threat to the sustainability of micro health insurance is low demand. This study goes beyond the decision of first enrolment and examines what motivates members to renew their MHI contract. In particular, the renewal decision is analyzed from a member’s perspective and five critical factors have been identified: the insurance premium, income, perceived quality of healthcare, perceived probability of illness and trust and understanding of the MHI scheme. The data used comes from the insurer of an MHI scheme in Nigeria, with a renewal rate of 46%. The findings of this study show that affordability of the premium and the correlation between perceived quality of healthcare, perceived probability of illness and trust and understanding plays a crucial role in the renewal decision process.
Primary supervisor: N. Ketel Secondary supervisor: Prof. dr. E.J.S. Plug
Table of Contents
1. Introduction ... 3
2. The concept of Micro Health Insurance ... 5
3. Literature review ... 6
3.1 Determinants of demand for health insurance ... 6
3.2 Scheme factors ... 8
3.3 Provider factors ... 10
3.4 Individual factors ... 11
4. The renewal decision model ... 13
4.1 The expected utility gain of insurance ... 13
4.2 The renewal decision ... 15
4.3 Hypotheses ... 16
5. Data and descriptive statistics ... 19
5.1 The program ... 19
5.2 Descriptive statistics ... 20
6. Methodology ... 23
6.1 The empirical model ... 23
6.2 Independent variables ... 24 7. Results ... 27 7.1 Robustness checks ... 31 8. Conclusion ... 33 9. References ... 35 2
1. Introduction
Out-of-pocket health expenses are a significant threat to the economic prosperity of the poor. In Sub-Saharan Africa, approximately 50 percent of healthcare expenses is out-of-pocket, pushing millions of people into poverty (WHO, 2005). For governments, the minimum amount needed to provide a citizen with basic healthcare is $44 annually, estimated by the World Health Organization (WHO, 2012). Developing countries do not come close to spending this amount, their mostly impoverished population pays for it out-of-pocket. In recent decades, there has been a movement towards expanded health insurance coverage for the poor in the form of micro health insurance (Onwujekwe, et al., 2009). Micro health insurance (MHI) schemes are heavily subsidized and aim at providing affordable healthcare to the poor. Despite its great potential, the success of MHI has been hampered by low take-up rates and even lower renewal rates. While there is extensive literature on take-up rates and issues related to the implementation of MHI, literature on re-enrolment and thus sustainability of MHI is scarce. This study focuses on re-enrolment and attempts to explain the puzzling problem of why individuals who were once enrolled in MHI, decide to drop out of these heavily subsidized programs. MHI schemes are now gradually maturing, thus examining re-enrolment and sustainability of MHI has become increasingly important.
The biggest threat to the sustainability of MHI is low demand, caused by low take-up rates and even lower renewal rates. Low take-up and renewal rates are inconsistent with basic economic frameworks for health insurance demand. These frameworks postulate that demand for health insurance is driven by the fact that most people are risk averse. Given the chance, people are generally willing to pay a certain amount upfront in exchange for free access to healthcare when needed (Galárraga et al.,2010). Adverse selection, working against a sustainable health insurance scheme, refers to the phenomenon that individuals with poor health status are the ones most likely to purchase health insurance (Rothschild & Stiglitz, 1976). MHI reduces financial barriers to seeking healthcare which can lead to overuse of medical care by members, a phenomenon called moral hazard. Both adverse selection and moral hazard pose a threat to the sustainability of MHI schemes. Insurance companies need a viable risk pool to overcome financial problems associated with adverse selection and moral hazard. To create these viable risk pools within MHI, demand needs to increase and thus high enrolment and renewal rates are essential. Unfortunately, micro health insurance schemes often fall short of reaching these rates, which jeopardizes its sustainability (Criel & Waelkens, 2013; Dong et al., 2009; Mladovsky, 2014).
While low renewal rates of MHI schemes have frequently been reported as a threat to the sustainability of MHI, they have rarely been analyzed in depth (De Allegri et al.,2009). A few exceptions come from two quantitative studies in Senegal and Burkina Faso and one qualitative study in Guinea. 3
The latter refers to the study of Criel and Waelkens (2003), who held focus group discussions with members and non-members to examine declining subscriptions of MHI. Population coverage of this MHI fell from the initial 8% of the target population to 6% in the following year. Dong et al. (2009) used household surveys to quantitatively examine reasons for dropping out of MHI in Burkina Faso. From the start of the program, enrolment rates had been low, 5.8% of the target population, and drop-out rates ranged from 30.9% to 45.7%. Their study focused entirely on socio-economic and health-related factors that were potentially associated with the probability of MHI renewal. Mladovsky (2014) quantitatively analyzed drop-out rates of a community based MHI in Senegal. Her study draws on data collected through household surveys on the relationship between MHI membership, social capital and active community participation. Mladovsky’s total sample size consisted of 382 households, of which 227 were currently enrolled and 155 households had dropped out. These three renewal studies concluded that MHI renewal is related to: affordability, healthcare quality, health needs, health demand, household head and household characteristics.
This study analyzes the sustainability of the Hygeia Community Health Care (HCHC) program in Kwara state, Nigeria. The HCHC program is an MHI scheme, targeting the rural poor. Consistent with other MHI schemes, the sustainability of the HCHC scheme is jeopardized by low renewal rates of approximately 46%. This study uses data from the start of the program in January 2007 until December 2013, and covers a greater time-span than previous studies that typically focused on the first two years of MHI schemes. To the best of my knowledge, this is the first time that data from an insurance company is used to analyze the renewal rates of an MHI scheme. Using data from the insurer instead of household surveys, allows for a unique and detailed look into the health seeking behavior of enrollees. Furthermore, the sample size of 103,336 individuals is significantly larger than the average sample size of 100-300 participants of previous studies. In this study, renewals are defined as individuals who were once enrolled in the scheme for one year, and decided to renew their contract for at least a second year. All renewals are included, including renewals with a long lapse time between the end of the first contract and start of the second contract.1
Whereas previous studies have mainly focused on demographics and general statistics relevant for policymakers and insurers, this study is the first to also address the MHI renewal problem from a member’s perspective. The HCHC scheme is a unique MHI scheme, with enrolment at individual level instead of the usual household level. This allows for a detailed analysis of individual behavior, instead of generalizing behavior to the household level. A ‘renewal decision model’ is created to explain the individual decision process of members at time of re-enrolment. By analyzing previous evidence and
1 In the HCHC scheme, individuals are not forced to renew at the end date of the first contract. They can be
uninsured for a while and decide to re-enrol in the scheme on a monthly basis
4
applying it to the renewal decision model, this study has identified five key determinants of the decision to renew MHI contracts. Namely: income, insurance premium, perceived quality of healthcare, perceived probability of illness and the level of trust and understanding of the MHI scheme. Given the limitations of the used dataset, the relevance of these five factors are to some extent tested with a logistic probability model on probability of renewal in the HCHC scheme. The results suggest that affordability of the insurance premium and previous healthcare utilization are highly correlated with the probability of renewal. Although the relation between healthcare utilization and MHI renewal cannot simply be attributed to adverse selection since it captures (a) the influence of perceived quality of healthcare, (b) perceived probability of illness and (c) trust and understanding. The outline of this thesis is as follows: section 2 provides background information on the concept of MHI. Section 3 discusses previous evidence on MHI demand along the lines of a conceptual model. Subsequently, the renewal decision model is presented in section 4. Section 5 describes the dataset and descriptive statistics of (ex-)members. The methodology is described in section 6, and section 7 presents the results and robustness checks. Section 8 concludes.
2. The concept of micro health insurance
A typical MHI scheme operates in rural areas and targets the entire population of these areas. The rural poor generally have volatile income patterns and may lack access to adequate private health insurance (Onwujekwe et al., 2009). In the absence of affordable health insurance, households generally deal with potential illness by increasing precautionary savings or avoid out-of-pocket health expenses by not seeking healthcare. The problem of unaffordable health expenses comes to some extent from the limited possibilities the poor have when coping with sudden sickness, due to imperfect capital markets (Galárraga, Sosa-Rubi, Salinas-Rodriquez, & Sesma-Vazques, 2010). MHI offers an alternative to paying for healthcare out-of-pocket by offering unlimited access to healthcare in exchange for a fixed annual insurance premium.
There are different types of MHI schemes but MHI is always non-profit by nature. Moreover, health insurance is essentially a form of risk pooling: all members pay identical premiums and the funds are redistributed to the ones with the greatest need. Members of MHI make a voluntary prepayment and have to actively renew their health insurance contract annually. Most MHI schemes are subsidized by national governments and/or local NGOs. MHI aims at increasing individual’s willingness to pre-pay for health care, so that the subsidies can phase out and MHI can eventually run on its own. Insurance premiums always have to be paid in one instalment and at the beginning of the contract. The unit of enrolment is often at household level to reduce issues associated with adverse selection. The HCHC scheme analyzed in this study, is unique in that it has individuals as the unit of
enrolment. With individual enrolment, households are no longer obliged to enroll the entire household. On the one hand this increases the possibility of adverse selection but on the other hand, it makes MHI more flexible and affordable. While the details of insurance benefit packages differ by scheme, in general benefit packages are comprised of: unlimited primary care2, maternity care,
chronic disease care, health checks, drug supply and limited use of inpatient and referral care. In principle, members have access to healthcare without any out-of-pocket expenses.
The primary goal of MHI is to reduce health related risks for the poor by alleviating them from the burden of out-of-pocket health expenses and poor health status. The risk pooling approach of MHI allows for transfers from the healthy to the sick and more equitable access to healthcare overall (Criel, Carrin, & Waelkens, 2005). Other health seeking expenses notwithstanding, the prepayment approach of MHI substantially reduces financial barriers to seeking healthcare. Once financial barriers are lowered, members are expected to frequent health facilities more regularly hereby improving their overall health status. Improved health status will in turn generate more income. Next to reduction in out-of-pocket health expenses and improvement of health status, MHI aims to strengthen developing countries’ health systems. These health systems are generally poorly functioning and provide low quality of healthcare. MHI mobilizes resources through public and private subsidies and the insurance premiums. These mobilized resources are in turn used to improve the quality and efficiency of healthcare. In sum, MHI strengthens healthcare systems by increased healthcare utilization, improved financial capacity and improved quality of care.
3. Literature review
Although MHI has come to be seen as an important means of managing health risk for the poor, demand for MHI remains relatively low (Cole et al., 2011). This section describes factors influencing MHI demand along the lines of a conceptual model. Evidence from take-up and renewal studies is used to describe the relevance of these factors for MHI renewal.
3.1
Determinants of demand for health insurance
The demand for health insurance is driven by multiple interrelated determinants. Andersen and Newman (1973) were one of the first to model these determinants. They described the determinants of health service use in the United States with their ‘Behavioral Model of Health Services Use’. According to this model, insurance demand is a function of peoples’ predisposition to use healthcare, their ability to use healthcare and their need to use healthcare. However, the micro health insurance
2 Primary care is the basic level of care that covers the most common diseases and preventative care. It is defined
by the WHO as a set of universally accessible first-level services that promote health, prevent disease and provide diagnostic, curative, rehabilitative and supportive services
6
market in developing countries is significantly different from the health insurance market in developed countries. There are differences in the quality of legal institutions, availability of informal risk sharing networks and quality of healthcare. On the individual level, there are differences in income, education, financial literacy and attitudes towards risk (Eling, Pradhan, & Schmit, 2014). Jehu-Appiah et al. (2011) modified Andersen and Newman’s model of health service use to create a model applicable to enrolment in MHI in developing countries. Where the original model by Jehu-Appiah et al. focused on enrolment in MHI, this study will use their model to analyze factors influencing the decision to re-enrol in MHI.
Following the conceptual model of Jehu-Appiah et al., in this study determinants of re-enrolment are categorized into three types of factors: insurance scheme factors, healthcare provider factors and individual factors (see figure 1). Insurance scheme factors capture the influence of the insurance premium, payment modalities and benefit packages. The original model of Jehu-Appiah et al. is expanded with the trust and understanding factor, which is grouped under insurance scheme factors. Healthcare provider factors capture the influence of the quality of care and the individual’s perception of this quality. Andersen and Newman’s behavioral model is adjusted to be grouped under individual factors, which are comprised of enabling, need and predisposing factors. Enabling factors illustrate influences on an individual’s ability to enrol in MHI. Need factors are related to individuals’ perceived health status and the amount of healthcare they are expecting to need. Predisposing factors can be seen as the determinants of enrolment. Predisposing factors are essential for policymakers and insurance companies and influence the decision to take-up MHI. However, these factors tend to stay constant over time and are therefore less relevant for a member’s renewal decision. The following
Figure 1 Modified version of Jehu-Appiah et al.'s conceptual model
The Renewal
Decision
To renew or drop out Scheme Factors Insurance premium Payment modalities Benefit package Trust and understanding Provider Factors
Quality of care Individual Factors
Pre-disposing factors Enabling factors
Need factors
section will discuss existing evidence and relevance of all determinants, along the lines of the conceptual model.
3.2
Scheme factors
Traditional economic theory postulates that the demand for a product is inversely related to its price. In theory, to increase demand for MHI one should simply reduce the insurance premium. While existing evidence has shown that the insurance premium plays a significant role in the decision to re-enrol in MHI, the problem is more complex. MHI schemes are widely subsidized and members pay only a fraction of the real costs. Unlike traditional health insurance in most developed countries, MHI premiums are not based on individual risk assessment but are set at community level. In fact, subsidies allow premiums to be set below actuarially fair prices. Nevertheless, demand remains low which indicates that either other factors are causing low demand or the premium is still too high for the poor. It could be possible that the target group of MHI is too financially constrained to even have the option to choose health insurance. For example Dong et al. (2009), in their quantitative study on renewal rates of MHI in Burkina Faso, found that respondents considered the MHI insurance premium to be fair but it was unaffordable for the poorest of the poor. Mladovsky (2014) concluded that financial factors do not determine drop-out rates of an MHI scheme in Senegal, based on her quantitative analysis of household survey data. Other renewal studies identified lack of financial means, high insurance premiums and registration fees as factors for not re-enrolling in MHI in Nigeria and Guinea (Boateng & Awunyor-Vitor, 2013; Criel & Waelkens, 2003). Low demand for MHI is related to affordability of the insurance premium, although it has a bigger impact on MHI uptake than renewal. What matters for renewal rates is not the initial affordability but the change in affordability, either due to a change in the insurance premium or changes in income.
The second scheme factor, payment modalities, refers to practical issues associated with (re-) enrolling in MHI schemes. Similar to affordability of the insurance premium, payment modalities play a moderate role in the decision to renew but are of utmost importance for the take-up decision. An example of payment modalities is the unit of enrolment in combination with paying the premium in one instalment. Paying the household premium in one instalment has proven to be a barrier to (re-)enrollment, with larger households being more likely to not enroll or drop out of MHI (Criel & Waelkens, 2003; De Allegri, Sanon, Bridges, & Sauerborn, 2006; Kirigia, et al., 2005). Enrolment on individual level, as in the HCHC scheme, is more flexible and reduces issues related to this payment modality. In addition to the unit of enrolment, Dong et al. (2009) found other payment modalities to be influencing renewal rates: proximity of enrolment office, availability of enrolment officers, high registration costs, opening hours, and problems with the collection of membership cards.
The benefit package refers to the type of healthcare that is covered by the insurance, and has often been identified as a critical factor influencing MHI enrolment in SSA (Basaza, Criel, & Van der Stuyft, 2007; Jehu-Appiah, et al., 2011; Turcott-Tremblay et al., 2012). The benefit package is a typical scheme factor which influences the take-up of MHI but should not influence the decision to renew, since it stays constant over time.
The fourth scheme factor is related to the target population’s levels of trust in the MHI scheme. With MHI, much more than with micro finance products, consumers’ trust is crucial for its survival. With microcredit, consumers first receive their money and repay afterwards, while with MHI consumers bear the risks. Members first pay the insurance premium and then have to trust to be well insured in case of illness. Trust in the insurance scheme is especially relevant in developing countries with weak legal systems for enforcing payment of valid claims (Cole, et al., 2011). Schneider (2005) was one of the first to examine the role of trust in MHI demand. She examined trust-related experiences with an MHI scheme in Rwanda by using separate focus group discussions with providers, MHI managers, members and non-members. Her findings suggest that the levels of trust in MHI need to be enhanced in order to promote MHI uptake. Oriaki and Onemolease (2012) used a random sample of 360 households and analyzed factors that influenced household’s willingness to participate in MHI in Edo state, Nigeria. While this study focuses on enrolment instead of re-enrolment, the findings indicate the importance of lack of trust on MHI demand in Nigeria. Among others, Mladosky (2014) applied the issue of trust to re-enrolment in MHI and finds that 70% of currently enrolled members found the scheme trustworthy while only 33% of dropouts did so. In sum, trust between consumers, providers and insurers has proven to be a serious constraint to both uptake and renewal of MHI.
Trust is related to the extent to which people understand what MHI entails. Developing countries are not as familiar with insurance systems as developed countries. Grasping the concept of pre-paying for an event that might happen in the future can be challenging to poor people. The levels of trust and understanding are closely related to the degree of exposure to the program. Since MHI is operating in areas with illiterate and poorly educated individuals, exposure to the MHI program is the primary source of information on health insurance. Exposure is likely to continually increase understanding and trust in the MHI scheme. Members can be directly exposed to the program by frequenting a health facility themselves, or indirectly by the experience of friends and relatives. Levels of trust and understanding in the context of MHI renewal are difficult to quantify and the way of measuring these qualitative factors differs across empirical literature. Nonetheless, trust and understanding are broadly accepted to be crucial determinants of MHI uptake and renewal (Dong et al., 2009; De Allegri, Sanon & Sauerborn, 2006; Basaza, Criel & Van der Stuyft, 2007).
3.3
Provider factors
In developing countries, the quality of healthcare is a pervasive issue. Health facilities are struggling with long waiting lines, unavailability of drugs, poor attitude of staff and patient discrimination (Jehu-Appiah, et al., 2011; Dalinjong & Laar, 2012). For MHI to function properly, contracted facilities need to offer adequate quality of care. Regardless of the insurance premium, individuals are not willing to pre-pay for healthcare if they are not satisfied with the quality. Quality of care is often determined by the availability of drugs, number of doctors and nurses present, and sophistication of medical equipment. For insurance demand, it is not the objectively measured quality that matters, but the perceived quality of care. Among others, Boateng and Awunyor-Vitor (2013) examined the perceived quality of care in the Ghanian national health insurance scheme and its influence on the probability of renewal. Their renewal study was based on interviews with 300 randomly sampled households. Approximately 61% of the participants was currently enrolled, 24% had dropped out and the remaining 15% had never enrolled in the national scheme. According to Boateng and Awunyor-Vitor’s study, 58% of dropouts cited poor quality of healthcare as the reason for dropping out.
Poor quality of healthcare can be caused by numerous determinants, for example discrimination against insured individuals. Discrimination is a pervasive issue and can be explained by the fact that in most MHI schemes health facilities receive a fixed capitation fee per enrolled member, regardless of the actual use of healthcare. Therefore health workers tend to prefer treating uninsured patients first, since they pay directly out-of-pocket. In the short run this generates more income for the healthcare provider but in the long run this discriminative behavior threatens the sustainability of MHI. Numerous other issues with quality of healthcare have been identified by previous uptake and renewal studies. Among others, Basaza, Criel and Van der Stuyft (2007) concluded that the subscription and supply of drugs is problematic in MHI schemes in Uganda. In Uganda, MHI schemes have been running for over ten years and the authors conclude that low demand (30,000 members) comes in part from inadequate drug supply. Their findings were based on the evaluation of two selected MHI schemes, involving a review of scheme’s records, key informant interviews and exit polls with both members and non-members. Another issue related to low quality of healthcare is that doctors often prescribe expensive drugs, not covered by the insurance or drugs that are unavailable at the hospital’s pharmacy (Criel & Waelkens, 2003). Members care about the way they are treated, the attitude of the staff, discrimination and length of waiting lines. Hence, all renewal studies show that perception of low quality of care is a significant reason for dropping out of MHI (Dong et al.,2009; Criel & Waelkens, 2003; Boateng & Awunyor-Vitor, 2013; Mladovsky, 2014).
3.4
Individual factors
3.4.1 Enabling factors
Enabling factors refer to individual determinants which influence the ability to enrol in MHI. The most significant enabling determinant is income, which captures individuals’ ability to pay the annual insurance premium. Lacking the financial means to pay the insurance premium is a critical reason for dropping out of MHI (Dong et al., 2009; Criel & Waelkens, 2013). This finding suggests that income is positively correlated with insurance demand. However, Schneider and Diop (2004) examined MHI uptake of 54 MHI schemes in Rwandaand concluded that the overall effect of income on MHI uptake is ambiguous. Their findings were based on data from 2,518 household surveys with data on enrolment, health seeking behavior and related financial implications. Where Schneider and Diop analyzed MHI enrolment, Thornton et al. (2010) examined re-enrolment in Nicaragua by implementing an evaluation design that randomly varied the costs of MHI. Data was obtained through baseline and one-year follow-up surveys on a random sample of participants and, similar to Schneider and Diop’s findings, they conclude that income is ambiguously related to MHI renewal.
Initially, a rise in income makes the insurance premium more affordable to the household and increases the probability to renew MHI. However, affordability of the premium is not the only determinant of the correlation between income and MHI renewal. For example, Dong et al. (2009) found that higher household expenditure was negatively correlated with insurance renewal. This was explained by the fact that households with higher economic status can afford to pay for better quality healthcare out-of-pocket. Implicit in this reasoning is that the perceived quality of healthcare of the MHI scheme is lower than that of other health facilities in the area. In conclusion, existing evidence on the impact of income is mixed, income is likely to play a relevant although indeterminate role in the demand for health insurance.
Distance to a contracted healthcare provider is another example of an enabling factor. MHI schemes are mainly implemented in rural areas, with low population density and large distances to health clinics. Large distances increase the costs of seeking healthcare, both by transportation costs and opportunity costs. Even though members in principle have free access to healthcare, the indirect costs of seeking healthcare can be a substantial barrier. While distance to health clinics can be an important determinant of MHI uptake, it stays constant over time and is thus less relevant for the analysis of MHI renewal.
3.4.2 Predisposing factors
Predisposing factors, ranging from culture to education and degree of risk aversion, influence attitudes about health insurance. They can be seen as general characteristics of the target population and are
essential for policymakers and insurers. However, predisposing factors are not expected to change over time and are therefore less relevant for a member’s renewal decision. Examples of individual predisposing factors are: age, gender, risk-attitude, education, marital status and characteristics of the household head. Two other relevant predisposing factors are related to community characteristics: religion and culture. Religion plays a significant role in the lives of the poor, and is often related to risk attitudes and cohesion within the community (Jehu-Appiah et al., 2011; De Allegri et al., 2009). The extent to which individuals religiously experience health shocks as outside of their control, ie. ‘an act of God’, affects enrolment in MHI. Moreover, in some cultures, it is believed that mere thinking or talking about possible illnesses will increase the likelihood of falling ill. Religion and culture can thus play a crucial role in the uptake of MHI but are less relevant for renewal analyses.
3.4.3 Need factors
Potential members are expected to weigh the benefits of having health insurance against the costs and to choose accordingly. However, people lack to a great extent control over their health status. Throughout this study, it is assumed that the decision to renew MHI is based on the ex-ante perceived probability of falling ill in the following year. Each consumer receives signals of his future health status3
before purchasing health insurance. Given these signals, individuals form a perception of their health status and assess the insurance decision (Cardon & Hendel, 2001). The interpretation of health status signals differs per person. While some people believe they will never fall ill, others prefer to have regular health checks. Health status signals can range from illness history to personal fears and beliefs that do not necessarily have a rational explanation. The interpretation of these signals might change over time, as the risk of becoming ill becomes more salient. There is a crucial difference between perceived health status and actual health status. Actual health status is the true probability of falling ill, while perceived health status is the perception of actual health status based on individual health status signals. What matters for the renewal decision is the ex-ante perceived health status, and thus the interpretation of these individual health status signals.
The renewal studies of Dong et al., Criel and Waelkens, and Boateng and Awunyor-Vitor all show a significantly positive relation between poor perceived health status and MHI renewal. The interpretation of health status signals thus appears to play a role in the renewal decision process of individuals. Age is an example of a health signal and captures to some extent the actual probability of falling ill. For example, children under five years old and the elderly in general have higher health needs than others. Gender is another example of a health signal. Compared to men, women face
3 In this study, future health status and perceived health status cover the time-span of one year, ie. the length
of the insurance contract
12
special health risks related to pregnancy, are more subject to domestic violence and have greater vulnerability to HIV/AIDS (Banthia et al., 2009). Dong et al. (2009) found no significant difference between renewals and dropouts in terms of age and sex. Past illness frequency is another example of health status signals and appears to be positively related to enrolment. For the case of MHI renewal in rural China, Zhang and Wang (2008) examined the effect of past illness frequency over three years’ time, based on panel data with more than 8000 members and non-members. The authors conclude that illness history has a significant effect on MHI enrolment. These findings point in the direction of adverse selection; sick people are the most likely to renew. However, Criel and Waelkens (2013) conclude in their qualitative renewal study that the target population was not familiar with the concept of adverse selection. Kramer (2013) examined the HCHC program in Nigeria with data from a baseline survey and follow-up survey two-years later. Since the HCHC program allows households to enrol members on individual basis, Kramer could examine adverse selection within households and concludes that there is limited evidence for this to occur. However, she concludes that past illness frequency appears to be positively related to enrolment. This is potentially due to adverse selection but potentially due to other factors, such as levels of exposure and trust.
To summarize, many factors play a role in members’ renewal decisions. Despite the fact that the decision factors are categorized into three groups, they are all interrelated. From a policymaker perspective, several scheme factors and predisposing factors are essential. Looking from a member’s perspective, changes in the individual level of information and perception of factors play a crucial role. The following section presents a renewal decision model, describing how individuals make the renewal decision based on interactions between scheme, provider and individual factors.
4. The renewal decision model
4.1
The expected utility gain of insurance
In this section, a renewal decision model is presented based on the assumption that individuals are expected utility maximizers. This assumption implies that all individuals4 weigh their options and
choose the one with the highest expected utility. However, MHI often deals with the poorest of the poor who face severe financial constraints. Prepayment for a future event, even though it has an expected utility gain, might be unattainable for the poor and could be a severe limitation to the maximization of expected utility assumption. Following Costa & García (2003), individuals are assumed to be risk averse and the chance of falling ill is ex-ante a random event that influences individual utility
4 Except for young children, whose parents are expected to make the decision
13
functions. The original model of Costa and Garcia is based on the Spanish Catalan health system, and focuses on the initial take-up decision. Consumers could choose between private health insurance, national health insurance and no health insurance at all. The model presented in this section is to some extent based on Costa & García’s model, but focuses on the renewal decision instead of the take-up decision. Moreover, members of MHI only have two consumption options instead of three since they generally have no access to national health insurance schemes.
For simplicity, it is assumed that there are only two types of health status: healthy and sick. All utility functions are concave and increasing in their arguments5. The concavity of utility functions
depends on individuals’ level of risk aversion. The decision to take up insurance depends on four utility functions in total. The determinants of the utility functions differ per health status and insurance status (see table 1). Scheme factors only play a role in the expected utility functions of being insured, since uninsured individuals are not exposed to the scheme. For the insured, the utility function when sick (
)
consists of all three types of factors: scheme, provider and individual factors. The utility function when healthy ()
is determined by scheme factors and individual factors. The utility function of uninsured healthy individuals is solely based on individual factors. For the uninsured, the utility function when sick consists of provider factors and individual factors. The uninsured can frequent healthcare providers outside the scheme. Their provider factors may therefore differ from provider factors of the insured. Moreover, healthy individuals do not frequent healthcare providers, thus provider factors do not enter the utility functions of healthy individuals.At time of take-up, individuals face a trade-off between the uncertain utility gain resulting from one year free access to healthcare and the immediate disutility of paying the insurance premium. Individuals have two consumption options, the first one is to enrol in the scheme, thereby securing free access to healthcare when needed for one year but reducing disposable income by the fixed insurance premium. Alternatively, individuals may choose not to enrol. In this second option, the individual does not pay an insurance premium but bears the risk of paying for healthcare out-of-pocket
5 0, U < 0 , apostrophes denote first and second derivatives
Table 1: Determinants of the utility functions
Sick Healthy
Insured Scheme, provider and individual factors Scheme and individual factors
Uninsured Provider and individual factors Individual factors
14
in case of illness ex-post. Ultimately, the decision depends on expected utilities. The expected probability of falling ill ex-post is given by . It ranges from 0 to 1, with 1 representing 100% chance of falling ill. Naturally, the expected probability of being healthy is (1 ). Given the perceived health status, in other words the expected probability of illness, individuals weigh their options according to the following expected utility functions:
=
(
,
.,
) + [1
]
(
,
)
(1)
=
(
,
) + [1
]
(
)
(2)
The functions above show the expected utility when an individual is insured (1) versus uninsured (2). Where Provcontr. refers to provider factors of contracted healthcare providers and Provall also refers to
providers outside the scheme. An individual will decide to take up health insurance if ex-ante the expected utility gain ( ) of being insured for one year is positive:
=
0
(3)
4.2
The renewal decision
From a member’s perspective, the initial take-up decision is similar to the renewal decision, but taken at a different point in time. Since this study focuses on the renewal decision, all individuals have already once made the take-up decision and decided to enrol in the scheme. Therefore it is certain that at time of take-up, the individual’s expected utility gain of enrolling in MHI was positive:
> 0
In other words, the individual initially expected to gain utility from having health insurance versus paying out-of-pocket. The renewal decision is taken at least one year after the take-up decision and many factors may have changed within a year. The expected utility gain at time of take-up is therefore not necessarily equal to the expected utility gain at time of renewal: .
Dropouts have decided to discontinue their insurance contracts, implying that at time of renewal the expected utility from being uninsured must have exceeded the expected utility from being insured: . This implies that the expected utility gain at time of renewal is negative while it was initially positive, thus for dropouts it is certain that . For renewals, both expected utility gains are positive but not necessarily equal:
, 0 .
To answer the puzzling question of why individuals who were once enrolled in MHI decide to drop out, factors that might have changed over time need to be identified. The change in these factors should explain why the is no longer the same as . The renewal decision model is created to identify these factors. Of all factors described in the conceptual model of section 3, only five factors may have changed between time of take-up and renewal. All other factors are general characteristics that are not expected to change within one year. The five relevant factors are: the insurance premium ( ), income ( ), perceived quality of healthcare ( ), the perceived probability of illness ( ) 6 and the level of trust and understanding of the scheme ( ).
. refers to the quality
of healthcare providers within the scheme, refers to the quality of all healthcare providers. The out-of-pocket expenses for healthcare when uninsured are represented by variable c. The relevance of this variable will be explained in the following section.
The change in the five key factors could potentially explain the change in expected utilities and ultimately the change in the expected utility gain. The following functions represent the renewal decision model, they show the change in expected utility functions between time of take-up and time of renewal:
=
, ,(4)
With , = , ., , + [1 ] ( )
With = take-up, renewal
=
, ,(5)
With
, = ( , ) + [1 ] ( )
With = take-up, renewal
=
=
(6)
4.3
Hypotheses
The insurance premium ( ) and the level of trust and understanding of the insurance scheme ( ) are scheme factors. Changes in these scheme factors affect the expected utility of being insured (4) while the expected utility of being uninsured (5) remains constant. An increase of the insurance premium will reduce disposable income by . Hence it reduces and and ultimately the
6 The perceived probability of illness is likely to be correlated with income. In the long run, health issues indirectly
affect income through reduced labor supply and productivity. Changes in perceived probability of illness thus affect income with a lag. Since the renewal decision model only covers the average time-span of one year, this indirect effect is likely to be outside the scope of this study and therefore not added to the model
16
probability of renewal. Furthermore, the literature review has shown that increased levels of trust and understanding of MHI increase the probability of renewal. This leads to the following hypotheses regarding scheme factors:
Hypothesis 1: Higher levels of trust and understanding increase the probability of renewal
Hypothesis 2: An increase in the insurance premium decreases the probability of renewal
Regarding provider factors, the general costs of frequenting a health facility when uninsured ( ) are assumed to stay constant while perceived quality of contracted healthcare providers ( .,
)
can be subject to change. As previously mentioned, the determinants of provider factors differ per healthcare provider. Once an individual drops out of the scheme, he faces no restrictions regarding choice of healthcare provider (Provall). If he remains enrolled, he is still restricted to frequenting
contracted healthcare providers only (Provcontr.). For simplicity, it is assumed that between time of
take-up and renewal, the perceived quality of healthcare providers outside the scheme does not change, . This does not necessarily mean that the quality of all other providers remains constant, but members are assumed to have a constant perception of this quality. This assumption is based on the notion that enrolled members are expected to frequent contracted providers only. Existing literature has identified low perceived quality of care as a main determinant of dropping out of MHI schemes. Therefore, the renewal decision model allows for changes in the perceived quality of contracted healthcare providers, ., . The actual quality of healthcare is not expected to change,
only the extent to which individuals are aware of this quality level and thus their perception. People base their quality expectations on the amount of information available to them ex-ante. In rural areas where the majority of people is illiterate, the primary source of information is exposure to these health facilities, either by themselves or friends and family. The level of exposure and thus information is likely to change during the course of the insurance contract. The effect of being exposed to health facilities can work both ways. A positive experience which exceeds quality expectations will likely increase the probability of renewal, while the opposite occurs after a negative experience.
Hypothesis 3: An increase in perceived quality of care, increases the probability of renewal
Regarding individual factors, the individual predisposing factors are assumed to stay constant over time7. Lives of the poor are characterized by many risks, therefore the enabling factor income is
7 With the exception of age, which increases by at least one year
17
expected to fluctuate. Consistent with the findings of previous studies, the renewal decision model predicts an ambiguous effect of income on the renewal decision. Namely, an increase in income affects both and positively but the overall effect on is ambiguous. Given the assumption that all utility functions are concave and increasing in their arguments, higher income will increase when insured by more than when uninsuredsince disposable income is reduced by the insurance premium8. The opposite will occur for when insured and
when uninsured, assuming the insurance premium is smaller than out-of-pocket expenditure for healthcare9. Therefore the overall effect on is ambiguous, leading to the following hypothesis:
Hypothesis 4: Changes in income have an ambiguous effect on the renewal decision
The individual need factor perceived probability of illness ( ) is expected to change over time. This factor consists of two components: the actual probability of falling ill, and the interpretation of health status signals. The perception of health status might change as a result of a change in the actual probability of falling ill. For example pregnant women have increased needs for healthcare. It might also be the case that the underlying true probability of falling ill stays constant, while health status signals and their interpretations have changed. For example, being diagnosed with a chronic disease gives a strong signal about the true health status. Even though the actual health status of a member has not changed, at time of take-up the member was already chronically ill, the perceived health status will worsen as a result of this health status signal. Another example is the effect of an ill household member. Even though this does not affect a member’s own health status, the risk of falling ill becomes more salient and the member might start interpreting health status signals in a different manner. All these cases lead to a change in perceived health status and affect the renewal decision. From an economic point of view, by only looking at disposable income, the change in due to an increase in the perceived probability of illness, is greater than the change in .10 Ceteris
paribus, an increase in the perceived probability of illness has a positive effect on the renewal decision since the expected utility gain is positive:
8
, ( ) > , ( ) since < 0
9
, ( ) > , ( ) since < 0 and >
where subscript denotes first and second derivatives wrt income
10 = , ., , + ( 1) ( ) ( , ) + ( 1) ( ) = , ., , ( , ) + ( ) ( ) (+) (+) Thus
> 0
18Hypothesis 5: The renewal decision is positively associated with an individual’s perceived probability of illness
5. Data and descriptive statistics
5.1
The program
This study analyzes renewal rates of the Hygeia Community Health Care (HCHC) program in Kwara state, Nigeria. HCHC is the only health insurance program operating in Kwara state (Kramer, 2013). Based on GDP per capita, Nigeria is ranked as a middle income country though its health indicators tell a different story. Each year, one million children die before reaching the age of five, over three million individuals live with HIV/AIDS and life expectancy at birth is only 52 years (Gustafsson-Wright, Van der Gaag, & Tanovic, 2013; The World Bank, 2012). Kwara state is the typical MHI target area: a rural state with a poor functioning health system, limited awareness, high out-of-pocket health expenses and low healthcare utilization (Kramer, 2013). Consistent with other MHI schemes, membership has to be actively renewed yearly. In the insurance contract, members state their preferred healthcare clinic and are expected to frequent this clinic or the referral clinic. Due to large distances between contracted healthcare clinics, cross-overs between clinics are unlikely to occur.
The HCHC program was implemented in 2007 by the local HMO Hygeia Ltd., in collaboration with the Dutch PharmAccess Foundation. Consistent with other MHI schemes, the eligible target population consists of all inhabitants of the region and is approximately 201,642. The annual insurance premium of 300 Naira (1.8 USD) 11 has to be paid in one installment and is heavily subsidized12. The
insurance premium amounts to 23.1% of yearly per capita health expenses before the program was implemented (Kramer, 2013). The benefit package covers primary and maternity care and to some extent referral care. Enrolment in the program is voluntary and the unit of enrolment is at individual level, instead of the usual household level. This is partly due to the fact that individual enrolment makes MHI more affordable for potential members. Another reason why HCHC has individual enrolment, is that the definition of households can be indistinct in this particular setting. In Kwara, men are allowed to have multiple wives and often a household consists of many generations. Therefore household sizes might be larger than conventional western standards and problematic to define.
11 On 01-02-2009 the insurance premium was increased from 200 Naira to 300 Naira 12 The premium paid by members is only 7.5% of the actual premium costs
19
5.2
Descriptive statistics
This study uses a rich dataset with information on the insured people of HCHC from the start of the program in February 2007 until December 201313. The dataset allows for detailed information on
enrollee characteristics, insurance contracts and utilization data per enrollee. Data on the three largest14 healthcare facilities in the program was used, which resulted in a sample size of 103,336
ever-enrolees. Dropouts are defined as individuals who have reached the end of their health insurance contract and never renewed for the next contract. Individuals who have renewed their contract, regardless of the lapse time between end of contract and date of renewal, are defined as renewals. Members who have been enrolled in the MHI scheme from the start can be at most on their 7th
contract, since the program started in 2007. Of all members who dropped out over the years (87,036), 64.5% dropped out after the end of their first contract (see table A2). The other 35.5% renewed their first contract but dropped out in the following years. These numbers indicate that the dropout problem is the most severe after one year of enrolment. Therefore, this study will focus on the renewal decision at the end of the first insurance contract and will not go into depth on the decision to renew the second to seventh contract.
At the end of the first contract, members have three options: renew immediately, renew with lag time, or drop out. In the HCHC program, it takes on average 123 days before members renew their first year of membership. 14,763 members were at the end of their first contract in 2013, these members have had considerably less time to renew their contract than members who were at the end of their contract in 2008. It can be seen from table 2 that both take-up and renewal is declining over time, with the renewal rate dropping from the initial 65.8% to 23.6%. To reduce issues associated with
13PharmAccess Foundation checks the quality of data (consistency, validity and accuracy), however some
limitations to its quality may still apply
14 Data on the other two contracted healthcare providers was omitted (29,326 observations) from the dataset
because these facilities were significantly different in terms of size and characteristics of the target population
Table 3 General scheme statistics
General statistics Count
Scheme total 103,335
Renewals 46%
Female 54%
Average Age 24.4
Visits per member 1.45
Hospitalized members 45%
Chronic diagnosis 9.8%
Maternity diagnosis 18.9%
Table 2 Enrolment and renewal per year
New enrollees Renewal rate
2007* 22,981 65.8% 2008 19,713 51.2% 2009 19,037 46.5% 2010 12,322 42.9% 2011 14,520 30.2% 2012 14,763 23.6%
*year of contract start
20
censoring15, renewals and dropouts are defined on the basis of contract data until April 2014, which
captures the average of 123 days lag time for renewals. In this analysis, a person is considered to be a dropout if he failed to renew his health insurance contract after the end of the first contract. Renewals are individuals who renewed their first contract, regardless of the time between contract end date and date of renewal. Of the 103,336 members in total, 47,215 members renewed their first contract, resulting in a renewal rate of 46% (see table 3). The program enrolled more women than men and members are on average 24.4 years old. On average, members visit the healthcare provider 1.45 times during their first contract. 45% of members have visited a clinic, of which 9.8% have been diagnosed with a chronic disease16 and 18.9% received maternity care.
The dataset contains information on enrolled members of three preferred healthcare providers, see table 4. Clinic 1 is the largest clinic, with 37,305 members ever enrolled, and has the highest renewal rate (50%). Clinic 3 has the lowest renewal rate, 37.9%. The final two columns show scores on the quality of care at clinic level. The SafeCare score is obtained from data of the SafeCare Initiative, which measures objective quality of health clinics based on internationally recognized standards. All three clinics have been assessed by the SafeCare Initiative, clinic 1 performed the best with a score of 61 on a scale from 0-100. MyCare is a monitoring instrument of the PharmAccess Foundation by which 20 enrolled patients are interviewed per health clinic on ten key quality dimensions. Health clinics receive a MyCare score, based on the experience of these patients. On a scale from 0-100, clinic 2 scored remarkably well, with a score of 95 while clinic 1 has the lowest score of 57. Although SafeCare and MyCare scores are revealed to the clinics, members are likely to be unaware of these scores and their meaning.
15 In section 7 a robustness check for censoring is performed by omitting all data from 2013
16 Diagnoses are classified as ‘chronic, maternity or other’ using the International Classification of Primary Care,
2nd edition (ICPC-2) system
Table 4 General characteristics of contracted healthcare clinics
Total ever-members Renewal rate SafeCare score MyCare score
Clinic 1* 37,505 50.0% 61 57
Clinic 2 33,744 48.3% 40 95
Clinic 3 32,087 37.9% 42 82
21
Predisposing factors do not enter the renewal decision model17 but are essential to
policymakers and insurance providers. Table 5 shows the general characteristics of dropouts and renewals. The first four columns summarize the predisposing and need factors of dropouts and renewals, respectively. The final column shows the p-values of chi-square tests. Most members are between 15 and 49 years of age, no significant difference is detected between dropouts and renewals in terms of this age category. The other three age categories show significant differences between the groups, and renewals are on average slightly older than dropouts. Moreover, renewals are more likely to be female, to have a male household head and to have more household members enrolled in the scheme than dropouts. In terms of healthcare utilization, there are significant differences between renewals and dropouts. Renewals are more likely than dropouts to have frequented a health facility during their first contract. 41.6% of dropouts versus 33.6% of renewals did not frequent a health clinic themselves but a household member did. In 23.5% of all dropout cases and 10.5% of renewal cases, no one from the household frequented a health clinic.
Table 6 shows healthcare utilization data of members who have visited a health clinic at least once, ie hospitalized people. In total, 45% (45,957) of all members visited a health clinic during their first contract. 43% (19,566) of hospitalized individuals decided to drop out and the other 57% (26,391) decided to renew their first contract. There is a significant difference in average number of health
17 Except for age and gender, which are health status signals and thus related to need factors
Table 5 Descriptive statistics of dropouts and renewals
Total members (N=103,336) Dropout Renewal N= 56,121 SD N= 47,215 SD 2 p-value Age- category Age <5 17.7% 1.18 19.4% 1.18 0.000 Age 5-14 20.7% 2.79 18.1% 2.71 0.000 Age 15-49 50.4% 8.71 50.8% 8.91 0.205 Age 50+ 11.2% 7.55 11.7% 7.25 0.007 Average age 24.2 18.3 24.7 18.6 0.000 Gender (Female) 51.3% 0.50 56.5% 0.50 0.000
Household head gender (Female) 42.2% 0.49 37.2% 0.48 0.000
Average hh size enrolled 8.41 7.73 9.88 7.50 0.000
Healthcare utilization Enrollee visit 34.9% 55.9% 0.000 Household visit 41.6% 33.6% 0.000 No visit 23.5% 10.5% 0.000 22
visits, with renewals frequenting health clinics more often than dropouts. Moreover, renewals were more likely to be diagnosed with a chronic disease or to use maternity care18. The final four rows of
table 6 show data on the number of days between the last health visit of a member and the end date of the insurance contract. 44.8% of renewals have visited a health clinic within the final 100 days of their contracts, against only 24.5% of dropouts. Dropouts were significantly more likely to have more than 300 days between their last visit and the end of their contract. The association between these discounted health visits, an indicator of exposure, and the probability of MHI renewal will be further examined in the empirical analysis below.
6. Methodology
6.1
The empirical model
According to the renewal decision model, individuals base their choice of renewal on the expected utility gain. Given that the population of interest was already once enrolled in MHI, the difference in and is determined by the potential change in five factors: the insurance premium, income, perceived quality of healthcare, perceived probability of illness and trust and understanding. Given the limitations of the available dataset, the association between these five factors and the probability of renewal is to some extent tested with the use of proxy variables. The dependent variable (R) represents the decision to renew the insurance contract and will take on the value of 1 if the member renewed and 0 if he dropped out. Given the binary nature of the dependent
18 Hospitalized individuals can be diagnosed with several diseases, therefore the percentages on chronic,
maternity and other do not necessarily add up to 100%
Table 6 Healthcare utilization of hospitalized people
Hospitalized Individuals (N= 45,957)
Dropout (N= 19,566) Renewal (N= 26,391) 2
p-value
Average no. of health visits 2.66 3.69 0.000
Chronic 7.5% 11.6% 0.000
Maternity 18.1% 19.5% 0.000
Other 87.6% 90.6% 0.000
Days between last visit and contract end date
0- 100 days 24.5% 44.8% 0.000 100- 200 days 23.3% 23.8% 0.000 200- 300 days 28.4% 19.9% 0.000 300- 400 days 23.8% 11.4% 0.000 23
variable, a linear probability model (LPM) is used to determine the extent to which different variables affect the probability of renewal. All variables are described in the section below. Three regressions are run, the first one is the main LPM model and will take the following form:
( = 1) =
+
+
+
+
+
+
+
+
+
+
+
1 +
2 +
3 +
+
(7)
Where
,
and are insurance premium dummies,,
and are diagnosis dummies for chronic, maternity and other diseases.,
and are dummies representing exposure to the scheme.is a vector of personal characteristics.
1, 2
and3
are clinic dummies and represents year fixed effects.The random term ( ) captures unobserved heterogeneity and is assumed to be binomially distributed and heteroskedastic. The other two regressions are modifications of equation (7) with extra independent variables and interaction terms on exposure to health clinics and discounted exposure.6.2
Independent variables
Control variables - is a set of the following control variables: age, gender, household size19 and
gender of the household head. The continuous variable age is divided into four categories; 0-4, 5-14, 15-49 and 50+. These four categories are often used in medical literature to allow for differences in health needs between the groups. A gender dummy is included in the model which will take on the value of 1 if the individual is female. An interaction term between gender and age category 5-49 is added to control for the distinctive health needs of women in the maternity age versus men of the same age. Unfortunately, the dataset lacks information on income of enrolees. Therefore, in this analysis gender of the household head is used as a proxy for income. Given the traditional background of most enrollees, it is expected that households with female heads have on average lower income and are more vulnerable to income shocks than households with male heads20. The gender of the
household head is a dummy which takes on the value of 1 if the household head is female. Year fixed effects ( ) are added to control for shocks that might differ per year but affect all individuals equally.
Exposure to premium increase - , and are dummies capturing members’ level of exposure to the premium increase. In February 2009, the insurance premium was raised from 200 Naira (1.2 USD) to 300 Naira (1.8USD). Since not all members enrolled at the same time, there are three different
19 This is a proxy for the real household size. Since there is no direct information on household size, household
size refers to the total amount of household members ever-enrolled in the scheme.
20 Households with female heads will likely lack a male breadwinner and thus have lower average income
24
levels of exposure to the premium increase and members are grouped accordingly. The first group faced the lowest price of 200 Naira both at time of take-up and renewal. The second group explicitly experienced the premium increase, they initially paid 200 Naira but faced the price of 300 Naira at renewal. The third group faced the high premium at enrolment and re-enrolment. , and take on the value of 1 if the member belongs to the first, second or third group, respectively.
Diagnoses - , and are dummy variables that take on the value of 1 if the individual, during the first year of membership, has been diagnosed with a chronic, maternity or other disease. Being diagnosed with a disease is a strong health signal that alters one’s perception on the probability of illness. Therefore, diagnosis dummies are used as proxies for the perceived probability of illness.
Healthcare providers - 1, 2 and 3 are dummy variables that take on the value of 1 if the individual is contracted with health clinic 1, clinic 2 or clinic 3, respectively. As explained in section 5 the clinics may differ in quality, therefore these dummies are added to control for differences at the clinic level. In a modified version of the LPM, interaction terms between exposure and health clinics are added. Unfortunately there is no direct measure of individual perceptions of healthcare quality, therefore these interaction terms are used as proxies for perceived quality of care.
Exposure to the scheme - The variables , and are dummies that represent different levels of exposure to the HCHC scheme.There are three levels of exposure: never exposed, exposed through healthcare utilization of a household member and exposed through own healthcare utilization. takes on the value of 1 if a member and his household members have never been exposed to the scheme. takes on the value of 1 if a household member has been exposed to the scheme, and is 1 if the member himself has used healthcare. The coefficients on the exposure dummies capture the effect of three factors: perceived probability of illness, the perceived quality of healthcare and the levels of trust and understanding of the scheme.
Exposure to the scheme influences perceived probability of illness through changes in (the interpretation) of health signals. This can either be due to higher risk saliency or adverse selection. To isolate the effect of risk saliency, health visits of household members are added to the empirical model. Health visits by a household member do not change one’s own health signals but potentially change the interpretation of these signals as the risk of falling ill becomes more salient. Higher risk saliency might in turn lead to a different way of interpreting health status signals, ultimately leading to a change in member’s perceived probability of illness. The effect of healthcare utilization of the member himself will be discounted by the number of days between the last health visit and the end
date of the insurance contract. Namely, the original member exposure variable ( ) is broken down into four dummies on discounted health visits. There will be four categories of the number of days between last health visit and end of contract: 0-100 days, 101-200 days, 201-300 days and 300-400 days. The dummy variables take on the value of 1 if the number of days falls within the respective range. It is expected that members who visited the clinic more recently, have a higher perceived probability of illness. This can be due to adverse selection reasons or higher risk saliency. In other words, this can be due to changes in health signals or changes in their interpretation.
The second factor that is expected to change with the level of exposure to healthcare providers, is the perceived quality of care. Information on the quality of the healthcare providers, including MyCare and SafeCare scores, is not widely available to members. The primary source of information is exposure to these health facilities, either direct or indirect. As previously mentioned, interaction terms between exposure and health clinics are used as proxies for the perceived quality of care. The third and final factor that is affected by different levels of exposure, is the scheme factor trust and understanding. It is assumed that the scheme is in principle trustworthy and that exposure continually increases trust and understanding of the scheme. The effect of changes in trust and understanding are hard to capture in one variable, especially given the limitations of the used dataset. Therefore in the empirical analysis, variables on exposure to the program will serve as proxies.
In sum, the coefficients on the exposure dummies should be interpreted with care since they capture three different factors of the renewal decision. Table 7 presents an overview of the five key factors of the renewal decision with their associated variables and hypothesized signs.
Table 7 Overview determinants, variables and their hypothesized signs
Determinants Hypothesized
sign
Variables Hypothesized sign*
Insurance premium -
Low insurance premium +
Increased insurance premium -
High insurance premium -
Income +/-
Female household head -
Perceived Quality + Clinic*Exposure +/- Perceived probability of illness + Age + Female + Chronic diagnosis Maternity diagnosis Other diagnosis + + + Exposure +
Trust and understanding +
Exposure +
*The hypothesized sign of the variables refers to the effect on the determinant, not the probability of renewal 26
