The use of the HIV test: a conflict choice approach.

Broyles, Robert W. ; Mwachofi, Ari ; Khaliq, Amir A. 等

INTRODUCTION AND BACKGROUND

Approximately 1.5 million Americans have been infected with HIV since 1981 and 950,000 are currently living with HIV/AIDS. CDC estimates that from 2003 to 2007 there were 1,051,875 people living with AIDS in the US. CDC also reports that from 2004-2007, HIV diagnoses increased 15% in the 34 states (Centers for Disease Control and Prevention, 2009). Although improvements in the treatment of the disease have lowered mortality, HIV remains the fifth leading cause of death among Americans between the ages of 25 and 44.

Previous research focused on variation in the quality of care provided to infected individuals, the sources of financing care, treatment protocols, patient compliance and the presence of coexisting conditions (CDC, 2004; Cohen, Berk, et al, 2001; Kilbourne, Anderson, Asch et al, 2002; Goldman, Leibowitz, Joyce et al, 2003; Dobalian, Tsao & Duncan, 2004; Landon, Wilson, Cohn et al, 2003; Landon, Wilson, Wenger et al, 2002; London, Wilson, Wenger et al, 2002; Hsiao, Wond, Kanouse et al, 2003; McInnes, Malitz et al, 2004; Doblian, Anderson & Stein, 2003; Tucker, Burnam & Sherbourne, 2003). Other studies document delays in HIV diagnoses resulting in missed opportunities for early treatment (Sherbourne, Fprge, & Kung, 2005). Although multiple dimensions of the epidemic have been explored, the distribution of diagnostic procedures or HIV tests relative to differential risk has not been explored using multivariate techniques.

The distribution of diagnostic procedures is appropriate if those who are at greatest risk make greater use of tests. Conversely, if the distribution of HIV tests is inconsistent with differential risk, utilization patterns might be improved by modifying interventions designed to prevent transmission of the virus.

Of particular importance to this study is the finding that approximately one-third of those infected in the U.S. are unaware of their HIV infection (U.S. Census Bureau, 2003). For instance revised estimate indicate that 232,700 persons in the US were living with undiagnosed HIV infection at the end of 2006 (CDC, 2008). Since no cure for HIV/AIDS exists, prevention is a critical factor in controlling the epidemic. Previous research indicates the likelihood that those who test positively modify their behavior to reduce transmitting the disease (CDC, 2008; CDC 2003).

THEORETICAL FRAMEWORK

A number of models are used to explain the demand for health services and decisions to consume preventive or early detection services. The Theory of Consumer Behavior (TCB), the Investment Theory of Demand (ITD), and the Health Belief Model (HBM) are among the most prominent of these approaches. The TCB is frequently adopted as a frame of reference to examine factors that influence the use of health services. The ITD posits that the demand for health care is derived from a demand for health. The theory employs the neo-classical framework of reference to explain consumer behavior and the health production function. The HBM focuses on perceived susceptibility, severity, benefits and barriers as the primary determinants of health care utilization.

As summarized below, the, TCB, the ITD, and the HBM are separate models that employ different conceptual frames of reference to study demand and use of health care. This paper integrates the three approaches into a single model that might be applied to a wide spectrum of health behavior and use of services.

The Basic Models

The TCB assumes that the individual is limited by a budget constraint and purchases the mix of services that maximize a defined utility function. The individual also is presumed to possess perfect information concerning available goods or services and the increment in utility associated with the consumption of an additional unit of each. In the following, let MUi represent the marginal utility or increment in satisfaction derived from consuming an additional unit of good or service i and [P.sub.i] represent the corresponding price for i = 1, ... n. Subject to the constraint imposed by a constant nominal income, M, It can be shown that utility or satisfaction is maximized when

[MU.sub.1]/[P.sub.1] = ... = MUi/[P.sub.i] = ... = [MU.sub.n]/[P.sub.n]

If markets are competitive and health care is a normal good or service, the model indicates that an increase in income or decline in price results in an increase in demand and a greater use of services. The TCB predicts that, other factors remaining constant, a decline in income or increase in price will result in a reduction in utilization. Conversely, the model suggests that an increase in income or a decline in price will stimulate demand or utilization.

The ITD posits that the demand for health care is derived from a demand for health. Therefore the focus of the model is on disease prevention or health promotion. The ITD regards health as a capital asset that depreciates with the passage of time. As such, health promoting activities are viewed as an investment in health.

The benefits that are derived from health are of two types, namely consumption and production benefits. The consumption benefits of health consist of physical well-being and an enhanced ability to enjoy daily life. The production benefits are related to the ability to perform economic and social functions. The production benefits are measured in terms of income, suggesting that improved health enables the individual to avoid loss of income due to illness.

The ITD assumes that the health production function exhibits diminishing marginal returns of health inputs to health, suggesting that the function rises from left to right at a decreasing rate.

The HBM focuses on the influence of

1. perceived susceptibility (i.e. perceptions concerning the likelihood that the individual will contract a condition);

2. perceived severity of the condition;

3. perceived benefits of the preventive or early detection services and

4. perceived barriers to the use of early detection or preventive services

on health behavior. The model posits that the likelihood of using early detection or preventive services increases if the barriers to use decline. The barriers that are commonly cited in this context are higher prices, lower incomes and a lack of access to care including the lack of insurance. Similarly, the likelihood of using services increases as the susceptibility to the condition, perceived severity of the condition and potential benefits of use grow.

The Integrated Conflict Choice Model

As indicated, this paper relies on the integration of the TCB, the ITD and the HBM into a single analytic frame of reference. The conflict choice model employs as the unit of analysis the episode of care which consists of two phases. The first is the patient initiated phase while the second is defined as the physician or provider dominated phase.

The model proposed in this paper is limited to the patient initiated phase and is based on the proposition that the TCB, the ITD and the HBM collectively form the basis for the individual's decision to seek health care. Further, the conflict choice approach assumes that the patient initiated phase is dominated by two undesirable choices or options. Unlike the TCB, the conflict choice model contends that the consumption of medical services results in no direct utility or satisfaction. Rather, the use of health care usually results in discomfort or disutility, a feature that motivates the individual to avoid the use of service. Accordingly, it is reasonable to assume that the typical patient may choose to avoid using health care.

Similarly, the conflict choice model assumes that a rational individual also prefers to avoid the disutility of illness and the potential decline in the production function that might occur if health care is not sought. Therefore, health care is sought to avoid the disutility of illness and to derive an improvement in health and related health benefits. Consequently, the individual may choose the discomfort of illness that may accompany a decision to avoid care seeking behavior.

The integrated conflict-choice model is shown in Figure 1. The vertical axes measure the strength of the individual's tendency to avoid the negative consequences of seeking care (Choice A) and the strength of the individual's tendency to avoid the disutility of illness (Choice B) that may accompany a decision to avoid care seeking. Shown on the lower horizontal axis is the probability that the individual will select Choice B. In this case, the subject avoids care seeking, an outcome that results in the risk of illness and the potential decline in the benefits of health. The upper horizontal axis measures the probability 1 - P(B) or P(A), the likelihood that the individual will adopt Choice A, seek care and endure the disutility of use.

[FIGURE 1 OMITTED]

As indicated, the model presented in this paper assumes that the avoidance gradients are linear. However, it is also possible that the gradients are non-linear and rise exponentially as the individual approaches Choice A or Choice B. Although not estimated in this paper, it is possible to argue further that the shape or the slope of the avoidance gradient is related to the deleterious consequences that might occur if one of the choices is selected. For example, it is reasonable to suppose that the slope of the gradient increases as the adverse consequence resulting from adopting one of the choices grows.

The subject's perception of the outcome of the conflict choice situation is measured by a subjective probability. As is well known, subjective probabilities are applied to an event or outcome that is unique or occurs once. Subjective probability is directly related to the strength of the individual's belief concerning the occurrence or nonoccurrence of the outcome in the conflict choice model.

Specifically, if the individual is prepared to offer odds of H: J that an event or outcome will occur, the corresponding objective probability is expressed in the form H/(H + J). For purposes of presentation, the probability that the individual will endure the discomfort resulting from health care or suffer a loss in the benefits of health is measured objectively.

Consider first a situation of disequilibrium and assume the individual assigns a probability of P(B') to the likelihood of adopting option B and correspondingly a probability of 1- P(B') or P(A') of selecting choice A. As indicated, the strength of the individual's tendency to avoid choice A is measured by line segment ac while the strength of the tendency to avoid B is represented by line segment bc. Since line segment ac exceeds line segment bc, the individual is influenced by a net tendency to increase the likelihood of selecting choice B and to reduce the probability of adopting option A, an outcome that results in a behavioral equilibrium characterized by the probabilities P(B) and 1-P(B).

As indicated in Figure 2, a change in the position of the avoidance gradient A alters the location of behavioral equilibrium and related probabilities. Specifically, a shift in the avoidance gradient A from A to A' reduces the tendency of the individual to avoid the selection of Choice A, thereby increasing the probability that the subject will seek service and endure the disutility of use.

[FIGURE 2 OMITTED]

Consistent with the HBM and the TCB, a decline in the barriers to use such as a reduction in effective price, distance traveled to the source of care or an increase in income is expected to lower the avoidance gradient from A to A'. Similarly, consistent with the ITD, an improvement in medical technology and related change in the health production function is expected to increase the perceived benefits of health and thereby lower the avoidance gradient from A to A'. Hence, the analysis suggests that a decline in the effective price, improved access to care, advances in medical technology and the resulting improvement in perceived benefits increase the likelihood that the individual will select option A, elect to use early detection or preventive service and endure the disutility of use.

[FIGURE 3 OMITTED]

As indicated in Figure 3 a change in the position of avoidance gradient B from B to B' also increases the likelihood that the individual will select Choice A. It is possible to argue that perceived susceptibility and severity co-vary positively with the disutility of illness and the expected magnitude of foregone benefits derived from health. Hence, the model suggests that an increase in the perceived severity of illness or susceptibility results in an increase in the individual's propensity to avoid the disutility of illness and the foregone benefits of health. For example, if the a priori likelihood of illness increases from 2:1 to 4:1 and the perception of severity grows from a significant disability to potential death, the avoidance gradient B shifts upward and the likelihood the individual will use a preventive or early detection service will increase.

[FIGURE 4 OMITTED]

Assume that the objective of policy deliberations is to develop and implement options that stimulate the use of preventive or early detection services. Referring to Figure 4, suppose further that the elimination of barriers to care or an improvement in the benefits of service lowers avoidance gradient A from A to A'. Similarly, assume that perceived susceptibility and severity increases, resulting in a shift of avoidance gradient B from B to B'. In this case, gradient B' lies above and to the right of gradient A', implying that the strength of the individual's tendency to avoid Choice B exceeds the strength of the tendency to avoid Choice A for all probabilities. Hence, the model suggests that an upward shift in avoidance gradient B and a downward shift in avoidance gradient A will induce the individual to use service and endure the direct disutility or discomfort that accompanies the procedure.

Conversely, suppose the policy analyst introduces options that restrict or limit the use of preventive or early detection services. In this case, an increase in the effective price, a decline in income, a more stringent set of other barriers to care or a decline in the perceived benefits of care shifts avoidance gradient A upward and to the right. Similarly, a decline in perceived susceptibility and severity lowers avoidance gradient B in a downward direction. When combined, the policy options reduce the probability of use and increase the likelihood that the individual will adopt option B (i.e. avoid the use of service and endure the disutility of illness).

THE DATA AND METHODS

As mentioned earlier, the theoretical framework of the conflict-choice model was applied to the data derived from the responses of 196,081 individuals who participated in the Behavioral Risk Factor Surveillance System (BRFSS) survey in 2003. BRFSS is a well known population survey conducted annually on risk factors and health conditions in the general population in the U. S. Data on variables that may influence the use of HIV test are available in the 2003 survey but unavailable in more recent years. To test the validity of the conflict-choice model that integrates three separate theoretical perspectives into a single modality, a logistic regression analysis was carried out using HIV tests as the dependent variable. As such, the use or non-use of HIV tests is treated as a function of the integrated conflict-choice model.

The definitions, means and standard deviations of the dependent variable and the set of related covariates are summarized in Table 1. Respondents who were 65 years of age or more and those who were unwilling or unable to report their use or non-use of the HIV test were eliminated from the study. As indicated in the table, the focus of the logistic regression analysis is on the binary variable HIV which identifies 86,527 individuals who reported that they used the HIV test. For purposes of presentation, the set of covariates represent barriers to use, perceived risk or susceptibility to contracting the disease and the perceived benefits of early detection or prevention. In terms of the HBM, it is assumed that all respondents were familiar with the severity of the condition and its related health outcomes.

The perceived barriers to use are measured by a set of binary variables that identify individuals who reported that they earned a high income, encountered no situation in which cost prevented the use of service and identified a usual source of care. In each case, the set of variables represent a relatively low barrier to use and it is expected that respondents who were well educated, earned a high income, reported that cost did not prevent the use of medical care and identified a usual source of service were more likely to use the HIV test than their counterparts in the corresponding reference group.

The perceived benefits that might be derived from early detection also are measured by a set of binary variables. As indicated in the table, the binary variables identify those who correctly noted that treatment may prevent the transmission of the disease to the baby during delivery and that medical care may enable infected individuals to live longer. In addition, it is assumed that those who attained a college degree, those who reported that it was important or very important to know their HIV status and those who reported a consultation with a health professional regarding preventive measures are more likely to perceive the benefits of early detection or prevention than their counterparts in the corresponding reference group. Accordingly, in each case, the set of variables represent, in varying degrees of precision, the individual's perception of the benefits derived from the HIV test, suggesting that the coefficients derived for this set of factors will be positive and significant.

The final set of covariates serves as surrogates for the individual's risk or susceptibility of contracting HIV/AIDS. As indicated, the most direct measure of susceptibility, HIV RISK, identifies individuals who reported the use of intravenous drugs, treatment for a sexually transmitted or venereal disease engaged in prostitution or participated in anal sex without the use of a condom. Assuming that individuals are risk averse, the study examines the proposition that the probability of using the HIV test is greater among those who are at greatest direct risk than among their less susceptible counterparts.

Less direct indicators of a greater comparative risk include the individual's racial background, represented by the binary variables BLACK and NATIVE AMERICAN. The differential risk associated with age is measured by the binary variable that identifies those who are between 18 and 44 years of age while gender specific susceptibility is captured by the binary variable, FEMALE. Given the differential rates of infection cited previously, it is expected that the probability of use is greater among members of these groups than their counterparts in the corresponding reference category.

RESULTS

The results of the logistic regression analysis are presented in Table 2. As indicated, the analysis correctly identified 64.2 percent of the respondents. It should also be noted that all coefficients are statistically significant (P < .01).

The analysis uniformly and without exception supports the expectations derived from the conflict choice model and, in particular the predicted influence of barriers to care, reported susceptibility and the perceived benefits of early detection on the use of the HIV test. The results also suggest that indicators represented by AGE RISK,

ATTITUDE and PROVIDER CONSULTATION were the

most important predictors of use. In this regard, respondents who reported that it was important to know their health status and those who consulted with a provider of care regarding preventive measures were among the most likely to use the HIV test. In short, the results of the analysis are fully consistent with the expectations derived from the conflict choice model.

CONCLUSIONS AND QUALIFICATIONS

The results that are presented in this paper are limited by several considerations. First, the analysis is based on cross sectional data, suggesting that causality is neither possible nor intended. Second, data limitations prevented an assessment of differential perceptions of severity, a feature that exerts an unknown influence on reported results. In addition, several of the variables represent gross surrogates for their theoretical counter parts. In particular, the analysis of the barriers to use would benefit from a more precise indicator of the differential prices imposed on potential users while the educational attainment of the individual is a proxy for the perceived benefits of early detection and perhaps perceptions concerning the severity of the health outcomes that accompany infection. Data limitations prevented the assessment of other factors such as the distance from the potential user's residence to the source of service. Finally, the study was not able to measure the slope of the avoidance gradient or the magnitude change in the position of avoidance gradients. Rather the analysis was limited to an assessment of the influence of each variable on the change in the likelihood that an individual used the HIV test. However, given the uniformity of results it is possible to argue that the results reported here reflect differences in the position of the avoidance gradients.

Although limited by several considerations, the analysis supports several tentative conclusions. First, the results are uniformly consistent with the expectations derived from the conflict choice model and each coefficient was statistically significant (P< .01). Consistent with the TCB and the HBM, the analysis clearly supports the proposition that a decline in barriers to use tends to increase the probability the individual used the test. Also consistent with the TCB, the paper indicates that policies designed to stimulate the use of early detection or preventive services should consider a reduction in price or the elimination of the price of service. The results also indicate that a usual source of care reduces the barrier to use. When combined with findings concerning a consultation with a provider of care and the individual's attitude toward the importance of securing service, the results indicate a need to ensure that the members of the population have access to a personal source of service.

The analysis also indicates that, if the individual seeks to avoid the risk of an adverse outcome, the indicators of susceptibility were among the most important predictors of the use of the HIV test. These results indicate a need to communicate accurate information that depicts the interrelation among risky behaviors, contracting the disease and the potential deleterious effects on the individual's health status. Given the importance of racial and gender effects, the analysis also suggests that whites, males and those occupying a low socio-economic status may represent appropriate targets for interventions designed to ensure appropriate use of the test.

Finally, the study supports the proposition that early detection requires the implementation of a coordinated policy option that not only reduces barriers to service, to include price, but also emphasizes the behaviors that contribute to susceptibility and the potential benefits that are derived from a timely use of the test. In short, the results of the paper clearly support the adoption of policy options that lower the individual's tendency to avoid care seeking behavior while simultaneously increasing the avoidance of undesirable health outcomes, thereby improving the likelihood that the individual benefit from an improved health status.

REFERENCES

Allen S, Serufilira A, Bogaerts J, et al. Confidential HIV testing and condom promotion in Africa. Impact on HIV and gonorrhea rates. Journal of the American Medical Association. 1992;268:3338-3343.

CDC. (2003). MMWR 52(15).

CDC. (2004). HIV/AIDS Surveillance Report Vol 15.

CDC. (2008). HIV Prevalence Estimates--United States, 2006. MMWR, 57(39); 1073-1076

Centers for Disease Control and Prevention. (2009). Cases of HIV infection and AIDS in the United States and dependent areas, by race/ethnicity, 2003-2007. HIV/AIDS Surveillance Supplemental Report 14 (No. 2)

Cohen, S. E., M.L. Berk, et al. (2001). The care of HIV-infected adults in rural areas of the United States. Journal of Acquired Immune Deficiency Syndrome, 28, 385-392.

Doblian, A. R. Andersen, J. Stein. (2003). The impact of HIV on oral health and subsequent use of dental services. Journal of Public Health Dentistry 63(2) 78-85.

Dobalian, A. J. Tsao and R. Duncan. (2004). Pain and the use of outpatient services among persons with HIV. Medical Care 42(2) 129-138 2004

Goldman, G.P., A. Leibowitz, G. Joyce et al. (2003). Insurance status of HIV-infected adults in the post HAART era: Evidence from the United States. Applied Economics and Health Policy 2(2) 85-91.

Higgins D, Galavotti C, O'Reilly KR et al. (1991). Evidence for the effects of HIV antibody counseling and testing on risk behaviors. Journal of the American Medical Association. 266:2419-2429.

Hsiao, A. M. Wong D. Kanouse et al. (2003). Complementary and alternative medicine use and substitution for conventional therapy by HIV-infected patients. Journal of Acquired Immune Deficiency Syndromes 33, 157-165 2003

Kaiser Family Foundation. (June 2004) Survey of Americans on HIV/AIDS.

Kaiser Family Foundation. (June, 2004). HIV/AIDS Policy Fact Sheet.

Kilbourne, A. M., R. M. Anderson, S. Asch et al. (2002). Response to symptoms among a U.S. national probability sample of adults infected with human immunodeficiency virus. Medical Care Research and Review 59 (1) 36-58.

Kuo AM, Hauloos JS, Witt MD, et al. (2005). Recognition of Undiagnosed HIV Infection: an Evaluation of Missed Opportunities in Predominantly Urban Minority Population. Aids Patient Care and STDs Vol 19 (4).

Landon, B., I. Wilson, S. Cohn et al.(2003). Physician specialization and antiretroviral therapy for HIV. Journal of General Internal Medicine 18, 233-241.

Landon, B., I. Wilson, N. Wenger et al. (2002). Specialty training and specialization among physicians who treat HIV/AIDS in the United States. Journal of General Internal Medicine 17, 12-22.

London A., C. Foote-Ardah and M. Shapiro. (2003). Use of alternative therapists among people in care for HIV in the United States. American Journal of Public Health 93(6) 980-987.

McCaig, L. A. Hardy and D. Winn. (1991). Knowledge about AIDS and HIV in the US population: Influence of the local incidence of AIDS. American Journal of Public Health 81(12) 1591-1595.

McInnes, M. F. Malitz et al.(2004). Differences in patient and clinic characteristics at CARE Act funded HIV clinics. AIDS Care 16(7) 851-857.

Pergua, A. and M. Rivo. (1998). Racial differences in AIDS knowledge among adults. AIDS Education Preview 4(1) 52-60.

Samet, J. M. Winter L. Grant and R. Hingson. (1997). Factors associated with HIV testing among sexually active adolescents: a Massachusetts survey. Pediatrics 100 (3) 371-377.

Sherbourne, C. M Fprge F-Y Kung (2003). Personal and psychosocial characteristics associated with psychiatric conditions among women with human immunodeficiency virus. Women's Health Issues 13, 104-110.

Tucker J., M. Burnam C. Sherbourne Substance use and mental health correlates of nonadherence to antiretroviral medications in a sample of human immunodeficiency virus infection. American Journal of Medicine 114, 573-580.

U.S. Census Bureau. (2003). Annual Estimates of the Population by Sex, Race and Hispanic or Latino Origin for the United States.

ROBERT W. BROYLES

University of Oklahoma

ARI MWACHOFI

East Carolina University

AMIR A. KHALIQ

University of Oklahoma

Table 1

Definitions, Means and Standard Deviations of the
Dependent Variable and the Covariates

Variable    Definition                                   Mean (SD)

DEPENDENT VARIABLE

TEST        If respondent was tested for HIV = 1; 0      .44 (.49)
            otherwise

SUSCEPTIBILITY

AGERISK     If respondent was between 18 and 44 =1;      .54(.49)
            0 otherwise

BEHAVIOR    If respondent engaged in risky behavior      .03(.18)
            during previous year = 1; 0 otherwise

FEMALE      If respondent was female = 1; 0              .59(.49)
            otherwise

INDIAN      If respondent was Native American = 1; 0     .01(.05)
            otherwise

BLACK       If respondent was African American = 1;      .01(.04)
            0 otherwise

SINGLE      If respondent was single = 1; 0 otherwise    .43(.49)

BENEFITS

IMPORTANT   If respondent believed it is important       .73 (.37)
            or very important to know his/her HIV
            status, =1; 0 otherwise

BIRTH       If respondent knew that HIV may be           .52 (.49)
            prevented during child birth, = 1; 0
            otherwise

LIFE        If respondent knew that treatment for        .93(.25)
            HIV might prolong life, = 1; 0 otherwise

CONTACT     If respondent talked to a health care        .10(.30)
            provider about the use of condoms to
            prevent STDs, = 1;

COLLEGE     0 otherwise If respondent was college        .34(.47)
            educated = 1; 0 otherwise
(.30)

BARRIERS

INCOME      If respondent earned an annual income of     .37(48)
            $50,000 or more year = 1; 0 otherwise

USUAL       If respondent reported a usual source of     .81(.39)
            care= 1; 0 otherwise

COST        If respondent reported that cost never       .86(.35)
            prevented use of service = 1; 0
            otherwise

Table 2

Logistic Regression Analysis of the Use or Non-use
of the HIV Test

Co-variate          Coefficient   .95 OR CI

RISK
AGERISK             .89 (a)       2.40, 2.50
BEHAVIOR            .50 (a)       1.57,
1.75
  FEMALE            .02 (a)       1.01,1.05
  INDIAN            .58 (a)       1.44,2.19
  BLACK             .64 (a)       1.53,2.34
SINGLE
  BENEFITS
  IMPORTANT         .83 (a)       2.22, 2.39
  BIRTH             .22 (a)       1.22, 1.28
  LIFE              .30 (a)       1.30, 1.39
  CONTACT COLLEGE   .78 (a)       2.10, 2.25
BARRIERS
  INCOME            .15 (a)       1.14, 1.89
  USUAL             .15 (a)       1.14, 1.20
  COST              .42           1.48, 1.56