Low flu shot rates puzzle--some plausible behavioral explanations.

Shahrabani, Shosh ; Gafni, Amiram ; Ben-Zion, Uri 等

I. Introduction

This paper describes an economic model for analyzing an individual's decision about whether or not to be vaccinated against influenza. It shows that based upon objective parameters, the vaccination rate should be high. Yet, this result is not consistent with empirical findings of low vaccination rates, for example for anti-influenza shots (MIV study group, 2005). Therefore, the second part of the paper uses the behavioral approach and subjective variables, such as perceived infection probability, time preference, subjective costs of vaccination, perceived vaccination effectiveness, and perceived severity of illness, to explain the empirical findings.

Previous economic studies that dealt with an individual's decision concerning preferred preventive behaviors (e.g., Brito et al., 1991, Francis, 1997) did not take into account subjective variables known to influence an individual's decision. For example, individuals might use the subjective probability of being infected rather than population-based epidemiological data (Mehrez and Gafni, 1987, Auld, 2003, Karni, 2003, Mullahy, 1999). Moreover, the Health Belief Model (HBM) (Rosenstock, 1988) can provide insight into why people participate in programs that prevent illness or detect disease. The HBM posits that behavior is a function of an individual's beliefs about the subjective value of an outcome and the subjective expectation that a particular behavior will achieve that outcome. Blue and Valley (2002), who employ the HBM, found that individuals who were vaccinated against influenza believed more strongly that they were susceptible to influenza and that influenza is a serious illness than did those who chose not to be vaccinated. In addition, Chapman and Coups (1999) provide some evidence that individuals' time preference patterns can explain preventive health behavior; in particular, monetary time preferences were found to predict whether people took flu shots. Since individuals' short-term discount rates are typically higher than market or social rates (West et al., 2003), the present value of the benefit from vaccination to those individuals is lower than to society.

Our paper integrates the above mentioned behavioral effects as well as known psychological biases into a theoretical economic model to explain an individual's decision regarding flu-shot vaccination as an example of preventive behavior. Because low vaccination rates have a negative externality on society, it is important to identify the decision-making process of the individual.

Our model refers to the demand side of the flu-shot, yet the supply side might also have an impact on vaccination rates. While a shortage of vaccine in certain years has reduced vaccination rates, we still observe low vaccination levels over a long period, even in years with no shortages. For example, according to the MIV study group (2005), data for the period 1997-2003 show consistently low vaccination rates, although improving over time, in developed as well as undeveloped countries. In 2002, for example, there was no shortage of flu vaccines, at least in the US (Harper et al., 2004). Yet, vaccination levels for that year (doses distributed/1000 population) in the US were 289, in Canada 328, in Japan 160 and in Germany 181, while other developed and undeveloped countries exhibited even lower vaccination levels (MIV study group, 2005).

The paper is organized as follows. Section II describes the vaccination decision model based on objective parameters. Section III uses the behavioral economic model and subjective evaluation parameters to explain empirical findings regarding vaccination. Section IV describes the policy implications, and Section V presents conclusions.

II. The Vaccination Decision-Making Model

In describing the economic model regarding an individual's decision whether or not to be vaccinated against a particular disease, we assume the use of "objective variables." We describe the case of an infectious disease such as the flu, which breaks out each year during period 1 and has an epidemiological (i.e., population-based) infection probability denoted by P. A vaccine against this disease is available to the public during the preceding period 0. During period 0, an individual chooses whether or not to be vaccinated against the disease. We also assume that the vaccination will be effective, with probability denoted by [pi].

As in Brito et al. (1991), we denote the utility of a healthy individual with income (y) as u(y), where u'(y) > 0, and u"(y) < 0. [u.bar] (y) denotes the utility of the individual when ill, satisfying [u.bar]' (y) > 0, [u.bar]" (y) < 0 and u (y) > [u.bar] (y). If an individual decides to be vaccinated during period 0, it will cost him or her 0 units of utility, where [theta] < u(y). The variable [theta] reflects a vaccination service fee. In this section, we assume that the individual has no health insurance and no loss of income insurance. The market interest rate is denoted by R, while C(y) denotes the treatment costs and income loss, measured as utility loss, which the individual has to pay in case of illness. C(y) may also include treatment costs due to the externality effect of illness on family members and other relatives. We assume that the income loss costs depend on the individual's income, and that [theta](1 + R) < C (y). The utility of a person who decides to be vaccinated is:

[u(y) - [theta]] + u(y)[pi] + [u.bar](y)(1 - [pi])/1 + R (1)

Alternatively, if an individual decides not to be vaccinated, V, the expected utility is given by:

V(y, P, R, C) =

u(y) + (1 - P)u(y) + P x [u.bar](y)/1 + R - P x C(y)/1 + R (2)

From equations (1) and (2), the critical objective probability, [P.sup.*], is given by:

[P.sup.*] = [increment of u](y)(1 - [pi]) + [theta](1 + R)/[[increment of u](y) + C(y)] (3)

where [increment of u](y) = [u(y) - [u.bar](y)] reflects the perceived severity of illness to the individual when she is not ill.

Proposition (1): The objective decision rule will be: Take the vaccine if P [greater than or equal to] [P.sup.*]; otherwise, do not choose to be vaccinated.

It is reasonable to assume that for each individual, the expression [[increment of u](y) + C(y)] is by far higher than [theta], the flu shot vaccination costs. Therefore, for high values of vaccination effectiveness probability, from equation (3) we expect that [P.sup.*] will be relatively low and the likelihood of choosing to be vaccinated will be high.

Suppose we have two different groups in the adult population. The first group is the high-risk group, which includes individuals who have chronic illnesses and/or are over the age of 65, and the second group is the low-risk group (i.e., "working group"). The probability of infection for the high-risk group is higher than for the low-risk group, while the effectiveness probability of vaccination is lower for the high-risk group (Bridges et al., 2001). (1) In addition we assume that the vaccination cost, [theta], is equal to zero for individuals in the high risk group, (2) and that u(y), [u.bar](y), C(y) are functions of P. That is, the baseline utility levels are likely to be lower for a high-risk chronically ill person than for a non-chronically sick person; therefore, the two groups differ in their perceived severity of illness, [increment of u](y). Also, the treatment costs C(y) are likely to be higher for the chronically ill person. Nevertheless, income loss due to illness is lower for the high-risk group. As a result, we expect that [P.sup.*] will be lower for the high-risk group than for the working group, and the likelihood of taking the vaccine is expected to be higher for the high-risk group.

To illustrate possible value for [P.sup.*], we consider the following reasonable parameters for individual in the low-risk group:

Example 1: Let us make the following assumptions for a developed country: [theta], the vaccination costs, are $10; C(y), the costs of treatment and income loss for three days of illness, are $200; R, the market interest rate, is 4%; and [pi], the vaccination effectiveness probability, is 0.8 (Bridges et al., 2001). We also assume that [increment of u](y), the perceived severity of illness to the individual when he or she is not ill, can be measured by individual willingness to pay to eliminate the illness after it has occurred, and is, for example, $200. Substituting these illustrative numbers in equation (3) yields [P.sup.*] = 0.12, while P, the annual influenza epidemic rate, ranges from 5% to 50% as suggested by Postma et al., 2002. In this case, it is more likely that [P.sup.*] < P, so we expect the individual will decide to take the vaccine. Moreover, for individuals in the high-risk group, we expect full compliance, since [P.sup.*] for this group might be even lower while the epidemiological infection probability is much higher than that of the low-risk group. Nevertheless, vaccination levels (doses distributed/1000 population) are quite low in many developed countries (MIV study group, 2005). For example, in the U.S., the vaccination coverage in 2001 among adults with high-risk conditions aged 18-49 years and 50-64 years was 23% and 44%, respectively, substantially lower than the Healthy People 2000 and 2010 objective of 60% (Harper et al., 2004). For adults over the age of 65, the vaccination coverage was 66%, while the Healthy People 2010 objective is to achieve vaccination coverage of 90%.

In the next section, we incorporate the behavioral approach to the decision-making model in order to examine possible reasons for the discrepancy between the objective model prediction and the low vaccination rates.

III. The Behavioral Approach

The behavioral framework used here includes subjective variables as well as possible psychological estimation biases of the relevant variables to explain the low vaccination rates.

First, we assume that each individual uses a subjective infection probability, [P.sub.s], which reflects his or her own perception of the probability of being infected, rather than population-based epidemiological data. Second, we assume that each individual has a subjective probability of the perceived effectiveness of the vaccination, [[pi].sub.s]. In addition, we assume that the individual has a subjective discount rate, [R.sub.s], which is likely to differ from the market interest rate. Moreover, we denote by [[theta]sub.s] not only monetary and time costs of vaccination but also barriers to being vaccinated, such as pain, fear of injections, inconvenience of getting the vaccination, and a "status-quo" cost of change (Samuelson and Zeckhauser, 1988).

In this case the critical probability [P.sub.s.sup.*] is given by:

[P.sup.*.sub.s] [increment of u](1 - [[pi].sub.s]) + [theta](1 + [R.sub.s])/[[increment of u](y) + C(y)] (4)

Proposition (2): For a given subjective infection probability [P.sub.s], an individual will decide to be vaccinated, if [P.sub.s] [greater than or equal to] [P.sub.s.sup.*]; otherwise, the individual will prefer not to be vaccinated.

Based on empirical findings from the Health Belief Model and psychological theories, we present the following behavioral effects affecting [P.sub.s.sup.*] and [P.sub.s]. In our estimation, the subjective probability [P.sub.s] will be lower than P, and in addition, the values of the variables in equation (4) will result in a larger [P.sub.s.sup.*] than [P.sup.*], so that the individual will decide not to be vaccinated. The behavioral effects are as follows:

(a) High values of [[theta].sub.s], due to barriers such as fear of injections and estimating high monetary and time costs for receiving the flu-shot. Based on the Health Belief Model prediction, Blue and Valley (2002) showed that high barriers were one of the reasons for deciding not to be vaccinated. In addition, high values of [[theta].sub.s] can stem from high "status-quo" costs of change. The status quo effect, which describes people's reluctance to relinquish current routines, helps explain the difficulties in encouraging patients to use preventive medicine (Redelmeier et al., 1993).

(b) Low values of [increment of u](y), (3) due to underestimation of the severity of illness when someone is not ill. Hersch and Viscusi (1998) argued that if people do not fully understand how bad the outcome of ill health will be, or if they form their preferences based on erroneous information, they may attribute too little weight to a state of ill health, which, for example, will make them more likely to smoke. The same argument can hold for the low likelihood of taking a flu-shot.

(c) High values of [R.sub.s], the subjective short-run discount rate, which indicates an individual's present-oriented time preference. Since individual short-term discount rates are typically higher than market or social rates (West et al., 2003), the present value of the benefit from vaccination to those individuals will be lower than the costs and the pain incurred today. In addition, it is reasonable to believe that high-risk individuals will have even higher discount rates than low-risk individuals. Therefore, individuals will tend not to be vaccinated. Empirical measurements of a stated time preference showed that individuals' short-term discount rates range from 20%-50%, much higher than the market or social interest rate of 5%-10% (Johannesson et al. 1996, and Ben-Zion et al., 1989).

(d) Low values of C(y) due to medical insurance, where people do not pay for medical treatments when they have health insurance. In addition, lower values of C(y) can occur due to income-loss insurance, where people do not suffer loss of income when they are ill.

(e) Lower subjective infected probability [P.sub.s] than the objective probability due to the "overconfidence effect" (Griffin and Varey, 1996, Dunning et al., 2004). This effect reflects people's tendency to be unrealistically optimistic about their own health risks. As a result, for these people [P.sub.s] < [P.sub.s.sup.*], and they will prefer not to be vaccinated. This result is compatible with the findings of Blue and Valley (2002) that individuals who hold stronger beliefs that they are not susceptible to influenza tend not to be vaccinated.

(f) Lower values of [[pi].sub.s], the subjective probability of the perceived effectiveness of the vaccination, in comparison to the objective effectiveness probability, [pi]. Many individuals fail to get vaccinated against the flu because they believe the vaccine will be ineffective against multiple strains. Chapman and Coups (1999) found a positive relationship between vaccination acceptance and the perceived effectiveness of the vaccine.

Example 2: Let us assume that [[theta].sub.s], representing vaccination costs, inconvenience and status-quo costs, is $60; C(y), representing the treatment costs and income loss of an insured individual, is $70; and [R.sub.s], the subjective discount rate, is 20%. We also assume that [[pi].sub.s], the effectiveness probability, is 0.2, and that [increment of u](y), the perceived severity of illness to the individual when he or she is not ill, measured by willingness to pay to eliminate the illness when it does occur, is $70. Substituting these illustrative numbers in equation (4) yields [P.sub.s.sup.*] = 0.91, while P, the annual influenza epidemic rate, ranges from 5% to 50%. In this case [P.sub.s.sup.*] > P and we expect that the individual will decide not to be vaccinated.

IV. Policy implications

Anti-influenza vaccination has been shown to be a cost-effective policy from a societal perspective, both for targeted high-risk groups and for healthy adult workers (Nichol, 2001; Dille, 1999; Lee et al. 2002). Therefore, we assume that the costs of implementing the policy are much smaller than the full costs to society in case of an outbreak of the disease. Taking into consideration the positive externality of the vaccine to society, it is in the mutual interest of insurance companies, employers and health care providers that more individuals be vaccinated. Following our model, we suggest potential actions for influencing the subjective parameters, which, in turn, will increase individuals' motivation to be vaccinated. The suggested policies involve actions taken by insurance companies, employers and health care providers, which affect the following subjective parameters:

(a) Increase the values of [increment of u](y), the expected severity of illness in terms of utility, by providing easily accessible and understandable information regarding the seriousness of the illness, especially for those groups for whom the severity of illness is higher than for other groups.

(b) Increase the value of C(y), the costs of illness, by offering insurance with differential premiums, meaning that individuals will have to face extra premiums for their health insurance if they decide not to take the vaccine. Or, to put it differently, individuals who take the vaccine will be entitled to a reduction in their health insurance premiums. An increase in C(y) can also stem from information regarding the negative externality of the illness on relatives and family members.

(c) Decrease the value of [[theta].sub.s], by reducing the travel and time costs to get the vaccine. This can be accomplished if free vaccinations are offered not only in medical centers, but also at more work places, shopping centers, etc.

(d) Increase the value of [P.sub.s], the subjective infection probabilities, by publishing public information explaining how the disease spreads. Individuals are more likely to get vaccinated against influenza as the perceived disease threat increases (Li et al., 2004).

The suggested actions in Steps (a)-(c) may increase [P.sub.s.sup.*], while action (d) decreases [P.sub.s]. Therefore, according to our model an individual's incentive to be vaccinated will be higher.

Some of the steps suggested above might help physicians guide the health decisions of their patients. In our model, physicians use population probabilities, while the patient is motivated by "subjective probabilities." In addition, it is reasonable to believe that physicians, due to their medical knowledge and their objective attitude as advisors, are not subject to the "overconfidence effect" regarding the probability of infection. (4) Moreover, physicians are not subject to the underestimation bias of the severity of illness that some individuals might have. Nevertheless, doctors do not know their patients' utility functions. In addition, physicians cannot force patients to take an influenza shot (as reflected by the low compliance rates). Therefore, a potential mechanism by means of which physicians may influence patients' decision is by engaging in a type of encounter known as shared treatment decision-making (Charles et al., 1997, 1999).

V. Summary and Conclusions

In this paper, we used a behavioral economic approach to show why a pure objective model does not predict the low rates of vaccination against influenza in the population. We showed that lower subjective variables, such as infection probability, effectiveness probability of vaccination and perceived severity of illness, as well as higher subjective variables, such as infection probability and costs of vaccination, including pain and inconvenience, may explain the low rates of vaccination.

Knowing the factors affecting the decision not to be vaccinated can help in designing a suitable policy to prevent the burden of treatment costs and output loss caused by the disease in society. For example, health care providers can encourage individuals with large families, or individuals with high-risk members of their families, to get vaccinated in order to prevent the negative externality of illness. Moreover, since it was found that higher education has a positive effect on the decision to be vaccinated (Wu, 2003), we assume that supplying clearer information to the public regarding the severity of illness may have an impact on the subjective infection probability and on the perceived seriousness of illness.

Yet, supplying free vaccination is not sufficient to convince individuals to be vaccinated if the other vaccination barriers, such as inconvenience and status quo costs of change, are high. Efforts to reduce those barriers should be implemented.

Finally, the framework analysis of this paper can be applied to understand other individual decisions regarding health-related issues, such as the decision to stop smoking, to have regular medical check-ups and to control weight.

References

Auld, M. C., 2003. Choices, Beliefs, and Infectious Disease Dynamics. Journal of Health Economics 22, 361-377.

Ben-Zion, U., Rapoport A., Yagil J., 1989. Discount Rates Inferred from Decisions: An experimental Study. Management Science 35, 270-284.

Blue, C. L., Valley J. M., 2002. Predictors of Influenza Vaccine: Acceptance among Healthy Adult Workers. AAOHN Journal Thorofare 50(5), 227-235.

Bridges, C.B., Cox N.J., Fukuda S.K., Singleton J.A., Iwane M., 2001. Prevention and control of influenza: Recommendations of the Advisory Committee on Immunization Practices (ACIP). The MMWR series, CDC, U.S. Department of Health and Human Services; 50(RR04), 1-46.

Brito, L.D., Sheshinski E., Intriligator M.D., 1991. Externalities and Compulsory Vaccinations. Journal of Public Economics 45(1), 69-90.

Centers for Disease Control and Prevention. Influenza Vaccine information. Retrieved from http://www.cdc.gov/ncidod/diseases/flu/fluvac. htm April 8, 2002, 1-8.

Chapman, B.G., Coups E.J., 1999. Time Preferences and Preventive Health Behavior: Acceptance of the Influenza Vaccine. Medical Decision Making 19(3), 307-314.

Charles, C., Gafni A., Whelan T., 1997. Shared Decision-Making in the Medical Encounter: What Does It Mean? (or It Takes at Least Two to Tango). Social Science and Medicine 44(5), 681-92.

Charles, C., Gafni A., Whelan T., 1999. Decision Making in the Physician-Patient Encounter: Revisiting the Shared Treatment Decision Making Model. Social Science and Medicine 49, 651-661.

Dille, J.H., 1999. A Worksite Influenza Immunization Program: Impact on Lost Work Days, Healthcare Utilization, and Healthcare Spending. AAOHN Journal 47(7), 301-309.

Dunning, D., Heath C., Suls J.M., 2004. Flawed Self-Assessment: Implications for Health, Education, and Business. Psychological Science in the Public Interest 5(3), 69-83.

Francis, J.P., 1997. Dynamic Epidemiology and the Market for Vaccinations. Journal of Public Economics 63(3), 383-406.

Griffin, W.D., Varey C.A., 1996. Towards a Consensus on Overconfidence. Organizational Behavior and Human Decision Processes 65(3), 227-231.

Harper, S.A., Fukuda S.K., Uyeki T.M., Cox N.J., Bridges C.B., 2004. Prevention and control of influenza: Recommendations of the Advisory Committee on Immunization Practices (ACIP). The MMWR series, CDC, U.S. Department of Health and Human Services; 53(RR06), 1-40.

Hersch, J., Viscusi W.K., 1998. Smoking and Other Risky Behaviors. Journal of Drug Issues 28(3), 645-661.

Johannesson, M., Johannesson P.O., 1996. The Discounting of Life-Years Saved in Future Generation Some Empirical Results. Health Economics 5, 329-332.

Karni E., 2003. On the Representation of Beliefs by Probabilities. The Journal of Risk and Uncertainty 26, 17-38.

Lee P.Y., Matchar D.B., Clements D.A., Huber J., Hamilton J.D., Peterson E.D., 2002. Economic Analysis of Influenza Vaccination and Antiviral Treatment for Healthy Working Adults. Annals of Internal Medicine 137(4), 225-31.

Li, Y.C., Norton E.C., Dow W.H., 2004. Influenza and Pneumococcal Vaccination Demand Responses to Changes in Infectious Disease Mortality, Health Services Research 39(4), 905-925.

Macroepidemiology of Influenza Vaccination (MIV) study group, 2005. The Macroepidemiology of Influenza Vaccination in 56 Countries, 1997-2003. Vaccine 23(44), 5133-5143.

Mehrez, A., Gafni A., 1987. The Optimal Treatment Strategy from a Patient Perspective. Management Science, 33, 1602-1612.

Mullahy, J., 1999. It'll only Hurt a second? Microeconomic Determinants of Who Gets Flu Shots. Health Economics 8(1), 9-24.

Nichol, L.K., 2001. Cost-Benefit Analysis of a Strategy to Vaccinate Healthy Working Adults against Influenza. Archives of Internal Medicine 161,749-59.

Postma, M.J., Jansema P., Van Genugten M.L., Heijnen M.L., Jager J.C., De Jong-van den Berg L.T., 2002. Pharmacoeconomics of Influenza Vaccination for Healthy Working Adults. Drugs 62(7), 1013-1024.

Redelmeier, D.A., Rozin P., Kahneman D., 1993. Understanding Patients' Decisions. The Journal of American Medical Association 270(1), 72-77.

Rosenstock, L.M., Strecher, V.J., Becker, M.H., 1988. Social Learning Theory and the Health Belief Model. Health Education Quarterly 15(2), 175-183.

Samuelson, W., Zeckhauser R., 1988. Status Quo Bias in Decision Making. Journal of Risk and Uncertainty 1, 7-59.

West, R.R., McNabb R., Thompson A.G., Sheldon T.A., Grimley E.J., 2003. Estimating Implied Rates of Discount in Healthcare Decision-Making. Health Technology Assessment 7(38), 1-60.

Wu, S., 2003. Sickness and Preventive Medical Behavior. Journal of Health Economics 22, 675-689.

Notes

(1.) In a report for the Centers for Disease Control and Prevention (CDC), Bridges et al. (2001) claimed that when the vaccine and circulating viruses are well-matched, the vaccine prevents influenza illness among approximately 70%-90% of healthy adults aged less than 65 years. In addition, a randomized trial among non-institutionalized persons over the age of 60 reported a vaccine efficacy of 58% against influenza respiratory illness (Bridges et al., 2001).

(2.) Many countries provided public reimbursement for vaccination of recommended target groups through either national or social health insurance (MIV group 2005, page 5134).

(3.) In equation (4), ([partial derivative][P.sup.*]/[partial derivative][increment of u]) < 0, if C < [[theta](1 + R)/(1 - [pi])]. For an insured individual, this condition is reasonable since C(y), treatment costs and loss of income are very low.

(4.) Nevertheless, in a personal decision, physicians might be subject to this effect. For example, MIV study group (2005) argued that vaccination rates in the group of healthcare workers are widely regarded as inadequate.

by Shosh Shahrabani, * Amiram Gafni, ** and Uri Ben-Zion ***

* Shosh Shahrabani, Economics and Management Department, The Emek Yezreel College, Emek Yezreel 19300, Israel. shoshs@yvc.ac.il.

** Amiram Gafni, Centre for Health Economics and Policy Analysis, Department of Clinical Epidemiology and Biostatistics, McMaster University, Ontario, Canada. gafni@mcmaster.ca.

*** Uri Ben-Zion, Economics Department, Ben-Gurion University, Beer-Sheva, Israel. benzionu@mail .bgu.ac.il.

**** A pervious version of this paper was presented at the SABE conference, Philadelphia July 2004. We would like to thank the participants of the seminars, as well as Gregory Yom-Din, for their useful comments and suggestions.