Value for the Future and Breast Cancer–Preventive Health Behavior

Many everyday health-related decisions involve intertemporal choices, trade-offs between immediate benefits, often pleasure or convenience, and delayed benefits, often improvements in health or survival (1). Taking antihypertensive medication, for example, may involve tolerance of side effects in the present to reduce the risk of heart disease in the future. Undergoing vaccination requires accepting the discomfort of a shot in the present to reduce the risk of an infection in the future. Individuals vary in the relative value that they assign to present versus future outcomes (i.e., their “time preference” in the psychology literature and “discounting” in the economics literature), raising the possibility that time preference is an important determinant of preventive health behavior.

In addition to the prominent role of discount rates in expected utility theory, the relationship between time preference and preventive health behavior is supported by theories of health behavior, such as the Health Belief Model and the Theory of Reasoned Action (2, 3). In these theories, the perceived benefits and risks of engaging in a health behavior influence a person's likelihood of adopting that behavior. Time preference affects these perceptions of benefits and risks. For example, if all other factors are held constant, a person who places high value on the present (compared to the future) will perceive relatively lower benefits from preventive health behavior than a person who places a relatively high value on the future. This perception of lower benefits may translate into negative attitudes about the behavior and a lower likelihood of engaging in the behavior. Thus, time preference adds greater specificity to the perceived benefits and perceived barrier components of several health behavior models.

Despite the theoretical justification for a relationship between future time preference and preventive health behavior, research examining this relationship is sparse and has yielded mixed results. Studies have found only small associations between future time preference (greater priority for future than present consequences) and health behavior, and for only some health behaviors (1, 4-6). This inconsistency may reflect, in part, the wide range of health behaviors that have been studied, including health behaviors that are less likely to be associated with perceptions about future risks. For example, studies have shown that smoking cessation is largely unassociated with perceptions about the future health risks from smoking, making it less surprising that smoking behavior has been only weakly (6) or inversely (7) associated with time preference. Additionally, most measures of future time preference in health-related studies use complex economic items that calculate numerical discount rates (e.g., the percentage increase in reward necessary to offset each year of delay) which, while appropriate for some settings, are often unreliable and difficult for subjects to understand.

Understanding the relationship between future time preference and preventive health behavior is particularly important because of the role that future time preference may play in explaining persistent socioeconomic disparities in preventive health behaviors in the U.S. (8-12). Individuals with less education or income have been found to be less willing to delay gratification, reflecting a relatively greater value placed on the present than the future (13-17). Whether socioeconomic differences in future time preference contribute to socioeconomic disparities in preventive health behavior in the U.S. is currently not known.

In this study, we examined the association between future time preference and use of genetic counseling for BRCA1/2 testing. Because BRCA1/2 testing is used to obtain information about future risk (of breast and ovarian cancer), we hypothesized that use of BRCA1/2 counseling would be associated with placing a relatively greater value on the future than the present. Furthermore, we explored the association between time preference and other breast cancer–preventive behaviors [i.e., screening mammography and self-breast examination (SBE)] that we had also assessed in our study in order to compare these behaviors with the use of BRCA1/2 counseling. Based on studies which have shown that women often see screening mammography as a method of ensuring that one does not have cancer in the present rather than as a method of preventing breast cancer death in the future (18, 19), we hypothesized that the association with time preference would be stronger for use of BRCA1/2 counseling than for adherence to breast cancer screening. To avoid the measurement error caused by complex assessments of discount rates, we used a brief, four-item measure of time preference that was easily understood and psychometrically reliable.

We conducted a health system–based case-control study to determine the association between time preference and use of genetic counseling for BRCA1/2 testing. A case-control design was used because BRCA1/2 counseling utilization is rare (<1 in 1,000 women in our population), and was provided by a single source within the University of Pennsylvania Health System, thus, allowing efficient case identification. In addition, pilot testing suggested that confidentiality concerns would interfere with accurate measurement of this outcome in general population samples. We focused on the use of BRCA1/2 counseling rather than actual use of genetic testing after counseling because genetic testing is heavily influenced by eligibility and insurance issues. At the time of this study, all interested patients had free access to genetic counseling.

The source population for the study was women who had a bill from a University of Pennsylvania Health System primary care provider in the 3 years before the start of the study. We excluded women with a personal breast or ovarian cancer history because BRCA1/2 counseling in these women primarily provides risk information for family members rather than information about personal future cancer risk. Cases were defined as all adult women from the source population who sought genetic counseling for BRCA1/2 testing at the University of Pennsylvania Cancer Risk Evaluation Program from December 1999 to August 2003. A total of 449 women obtained BRCA1/2 counseling; 269 did not have a breast or ovarian cancer diagnosis. Data collection was conducted prior to genetic testing. Two hundred and thirty-four of these women completed data collection for a response rate of 87.0%.

Controls were a random sample of adult women from the source population who had not sought genetic counseling at the time of enrollment and were identified through billing databases. A sample of controls was accrued every 6 months to approximate the timing of case enrollment. A total of 1,017 controls were selected, of whom five were deceased, eight were too sick to participate and one was male. Of the remaining 1,003, 47 surveys were undeliverable and 603 returned their survey for a response rate of 63.1%. Thirty-seven women were excluded because of a personal history of breast or ovarian cancer, leaving a final sample of 566.

To investigate the association between time preference and use of screening mammography and SBE, we initially restricted our analyses to the control population. Because BRCA1/2 testing was sufficiently rare that no women who was selected as a control was found to have undergone BRCA1/2 counseling, the control population represented a cross-sectional survey of a random sample of women in the University of Pennsylvania Health System primary care population. From this sample, the 398 women aged ≥50 years were included in the analyses of mammography and the 489 women aged ≥25 years were included in analyses of SBE, as recommendations are inconsistent for women below these age cutoffs. The study protocol was approved by the Institutional Review Board at the University of Pennsylvania.

Future Time Preference. Future time preference was assessed using a 4-item scale developed from the 12-item Consideration of Future Consequences Scale because of concerns about the original scale's length, redundancy, and reading level (20). To develop the new scale, the 12-item scale was tested on 10 volunteers waiting in primary care practices who were asked to comment on confusing items. The items were revised accordingly and the pilot scale was administered to 120 Philadelphia County prospective jurors (66% women; mean age, 42 years; 39% Black; mean education, 14 years). Items with the lowest inter-item correlation were removed sequentially while evaluating the content validity and internal consistency of the resultant scale. The final four-item scale had reasonable internal consistency (Cronbach's α, 0.71) and item-total correlations >0.25 (Appendix A).

Further testing of the scale was conducted in a sample drawn from adult women enrolled in a Philadelphia Medicaid Managed Care Organization. Four hundred and forty-one women completed the time preference scale as well as measures of discount rates adapted from Chapman et al. (21). In this sample, the four-item time preference scale had an internal consistency of 0.66 with all-item total correlation of >0.38. The time preference scale items had <5% missing data. In contrast, the discount rate items had high rates of missing data as well as inconsistent results. Fifty-two (32%) of the respondents to the mailed survey did not complete any of the discount items and an additional 10 respondents had inconsistent answers on at least one of the items. The four-item time preference scale was inversely correlated with the discount item scores. In addition, the four-item time preference scale was positively correlated with educational attainment and negatively correlated with cigarette smoking. Thus, based on these pilot data, we decided to use the four-item scale in the mailed survey reported in the current article.

Subjects' responses to the four time preference statements were averaged to create a future time preference score. A higher score reflects a stronger orientation towards the future (items were reverse-scored).

Screening Mammography. Mammography adherence was assessed by asking the date of last screening mammogram; having a mammogram within 12 months prior to completing the questionnaire was categorized as adherent. Women who provided only the year of last mammogram were assumed to have undergone screening in the middle of that year (June). The mammogram analyses were run with and without this assumption. When this assumption was not invoked, the data were treated as missing. The same pattern of results were seen using both approaches.

Self-Breast Examination. Frequency of SBE was assessed by asking subjects the number of times they conducted a SBE in the past year. Responses were categorized into whether each subject conducts a SBE once a month (<12 versus ≥12 times per year), although the results were similar when SBE frequency was examined as a continuous variable.

Breast Cancer Risk Factors. The survey assessed risk factors in the Gail model, the most commonly used breast cancer risk prediction model (22): degree of family history, age at menarche, age at first live birth, and history of breast biopsy. Degree of family history was categorized as presence or absence of a first-degree relative with breast or ovarian cancer.

Sociodemographic Characteristics. Age, education, and household income were measured using items from the Behavioral Risk Factor Surveillance System 1998 questionnaire (23). Items were developed to measure race, ethnicity, and religious heritage. Race was subsequently collapsed into White and other, and education was collapsed into college degree or less than a college degree.

Pilot Testing. Study instrument drafts were reviewed with genetic counselors, physicians specializing in breast cancer and breast cancer genetics, and primary care providers. Pilot testing was conducted on 75 women who had participated in the Cancer Risk Evaluation Program prior to the start of the study and 128 potential controls who were subsequently excluded from the eligible population for the full study. Pilot subjects' comments were used to refine the final instruments.

Statistical Analyses. Descriptive statistics were used to examine the characteristics of the study samples. Mann-Whitney tests were used to compare median time preference scores between women undergoing BRCA1/2 counseling (cases) and women not undergoing BRCA1/2 counseling (controls) and, in the nested cross-sectional study, between women adherent to screening mammography or SBE guidelines and women not adherent. These analyses included age-appropriate subjects for each behavior, i.e., all subjects for use of BRCA1/2 counseling, those ≥50 years for adherence to annual mammogram, and those ≥25 years for adherence to monthly SBE.

Multivariate logistic regressions were used to assess the association between future time preference and each behavior after adjustment for sociodemographic characteristics and breast cancer risk factors. Lowess plots of the relationship between future time preference and each of the three behaviors suggested that a linear term was the appropriate time preference metric for BRCA1/2 counseling and mammography adherence (24). For SBE, the relationship between the two variables was less clear; however, because the model was not significantly improved by using a quadratic term, we present only the linear term results. As in the Mann-Whitney tests, these analyses included age-appropriate subjects for each behavior. However, because many organizations recommend annual mammography for women ≥40 years of age, the analyses of mammography adherence was repeated using an age cutoff of 40 years. The results of these secondary analyses were not substantively different.

Multinomial logistic regression was used to further explore the association of time preference with mammogram adherence and use of genetic counseling. Subjects were divided into one of four groups on the basis of their adherence to annual mammogram and use of BRCA1/2 counseling: no mammogram/no counseling, no mammogram/yes counseling, yes mammogram/no counseling, yes mammogram/yes counseling. Thus, unlike the previous analyses examining the association between time preference and adherence to annual mammogram among the controls only, these analyses included mammogram adherence status of both the controls and the cases. We fit a single model with the “baseline group” of no mammogram/no BRCA1/2 counseling. Wald tests were used to determine whether there was evidence of a relative risk (RR) that differed from the baseline for each relevant group and to test whether there were differences between specific groups. These analyses were restricted to subjects ≥50 years so that comparisons could be made between the behaviors. We also qualitatively examined whether the results of these analyses in subjects ≥50 years were supported by the patterns observed for subjects ages 40 to 49. Because of the small numbers in some cells when including covariates in the model, we present the results from the model with time preference only and from the model including adjustment for breast cancer risk factors and sociodemographic characteristics.

The associations between future time preference and sociodemographic characteristics were assessed using Kruskal-Wallis tests. These analyses were conducted both in controls and cases, and the results were substantively the same. Because the controls are a representative sample of the source population, the results for the control sample are presented.

Table 1 shows the subjects' sociodemographic characteristics and time preference scores. As described previously, cases and controls differed significantly on breast cancer risk factors and sociodemographic characteristics (25). We took account of these differences by including sociodemographic characteristics in our multivariate analyses.

Future time preference was significantly higher in women who underwent BRCA1/2 counseling (cases) than in women who did not (controls; Table 2). In analyses adjusting for sociodemographic factors and breast cancer risk factors, a one-point increase in future time preference (on a 1-5 scale) was associated with a 3-fold increase in the odds of undergoing genetic counseling for BRCA1/2 testing (Table 3).

As shown in Table 2, having had a screening mammogram in the past year was associated with future time preference among women ≥50 years of age. After adjustment for sociodemographic factors and breast cancer risk factors, the association was present but no longer statistically significant [odds ratio (OR), 1.33; P = 0.25]. In univariate or multivariate analyses, monthly SBE adherence was not associated with future time preference among women ≥25 years of age (Tables 2 and 3).

Using the reference group of women who did not adhere to screening mammography or undergo BRCA1/2 counseling, future time preference was associated with adherence to screening mammography (RR, 1.5; P = 0.075), not associated with use of BRCA1/2 counseling (RR, 2.9; P = 0.314), and associated with adherence to screening mammography and use of BRCA1/2 counseling (RR, 2.8; P = 0.006; Table 4). Because so few women had undergone BRCA1/2 counseling but had not adhered to annual mammography (n = 34), the confidence intervals (CI) surrounding the point estimate for the association between time preference and the use of BRCA1/2 counseling only were relatively wide, and the difference between this estimate and the association between time preference and adherence to mammography alone was not statistically significant (P = 0.314).

This pattern of results persisted after adjustment for sociodemographic characteristics and breast cancer risk factors. For both models, the association between time preference and adherence to mammography and use of genetic counseling was greater than the association between time preference and adherence to mammography alone with borderline statistical significance in the bivariate model (χ² = 3.31, P = 0.069).

In separate models adjusting for sociodemographic characteristics and breast cancer risk factors, we found that the strength of the association between time preference and BRCA1/2 counseling after adjusting for mammography adherence (OR, 3.56; 95% CI, 1.60-7.99) was greater than the association between time preference and adherence to mammography (OR, 1.72; 95% CI, 0.96-3.11) after adjusting for use of BRCA1/2 counseling.

Among subjects ages 40 to 49, the results for the association between time preference and adherence to annual mammogram and use of BRCA1/2 counseling were qualitatively similar to the results among women 50 years of age and older. Relative to not adhering to mammography or using BRCA1/2 counseling, the RR associated with a one unit increase in time preference was 1.2 (0.58-2.4) for adherence to mammography and not using genetic counseling, and 4.7 (1.9-11.7) for both adherence to mammography and use of genetic counseling.

Future time preference was associated with increasing education (P = 0.0021) and income (P = 0.0011), but was not associated with race (P = 0.510) or age (P = 0.130; Table 5).

This study provides evidence that future time preference is strongly associated with seeking genetic counseling for BRCA1/2. A weaker association was seen between future time preference and annual mammography. No evidence was found for an association between future time preference and monthly self-breast exam.

These results extend previous work in this area. Fuchs (6) and Chapman and colleagues (4, 5) found only a small relationship between economic discount rates and health behaviors (e.g., flu vaccine, dental exams, smoking, exercise), for only some measures of time preference and for only some health behaviors. Chesson and Viscusi found a stronger future time preference (lower discount rate) among smokers than nonsmokers (7).

There are several possible explanations for the variation in these results. As suggested previously, the nature of the behavior in question may influence its relationship with future time preference. Although our study was designed to test the association between time preference and use of BRCA1/2 counseling rather than to tease out the relative strength of this association compared with other preventive behaviors, our results were consistent with our hypothesis that the behavior most explicitly related to future risk—counseling for predictive genetic testing—would be more strongly associated with future time preference than other breast cancer risk management behaviors. Despite the fact that mammography and self-breast exam have traditionally been recommended to reduce the future risk of breast cancer mortality, patient perceptions of these tests often focus on diagnosis in the present rather than their relationship to future risk (18, 19). Furthermore, the recent data that have called into question the effect of self-breast exams on breast cancer mortality may have further reduced the link between this behavior and perceptions of future risk and benefit. Although the degree of association between time preference and behavior consistently followed the pattern of BRCA1/2 counseling > mammography > self breast exam in our results, the difference between the association with BRCA1/2 counseling alone and the association with mammography alone did not meet strict statistical significance when they were combined in a single model. Although this may be attributable to a lack of statistical power because of the strong correlation among these behaviors in our sample, we cannot definitively exclude the possibility that the strength of association with time preference did not vary between these breast cancer–preventive behaviors. In addition, it is important to recognize that some findings in the literature do not fit this rule (26-28). More work is needed to refine and further test this hypothesis.

Another possible reason for the difference in our results is the approach to measuring future time preference. Most prior studies of time preference and health behavior used numeric discount rates measured by asking people to quantify the difference in value between a benefit in the present and a benefit at a specific time in the future. These measures often lead to considerable missing data and wide ranges of estimates, some of which seem implausible (e.g., equating $100,000 in a year to $1,000 now; ref. 29). We attempted to address this limitation by using a scale of time preference, which improved the consistency and reliability of responses. However, measuring time preference is challenging and our scale may also capture other constructs, such as risk preference, raising the possibility that the difference in our findings relates to the inclusion of risk preference rather than the use of a more reliable measure of time preference alone (30). Although this possibility cannot be completely excluded, the exclusion of the two items (items 1 and 4) that are potentially linked to risk preference does not alter the association between the items clearly assessing time preference (items 2 and 3) and use of BRCA1/2 counseling. Furthermore, because most people are risk-averse (31, 32), to the extent that the time preference measure overlaps with risk preference, the variation in the measure is mostly due to variation in people's time preference. Thus, the time preference component of the measure is likely to be the predominant determinant of the association with behavior.

Empirical evidence demonstrating a relationship between future time preference and use of BRCA1/2 counseling is important for three reasons. First, it adds to current knowledge about determinants of health behavior by giving credence to the theoretical argument that time preference should influence behaviors involving intertemporal choice. Second, as future time preference is associated with higher levels of education and income, our results suggest that differences in time preference may contribute to the lower rates of some preventive health behaviors among individuals of low socioeconomic status in the U.S., including use of genetic susceptibility testing (25). Third, this evidence raises the possibility that time preference may be a target for behavioral change interventions. These strategies may be most effective if they emphasize the future benefits of preventive health behaviors and the importance of a future orientation.

Although the association seen in this study between lower education and income levels and weaker future orientation is supported by several prior studies (13-17, 35), the direction of the relationship remains unclear. It seems equally plausible that the decision to continue in school beyond the mandatory requirements is driven by future time preference as it does that education makes individuals more able to delay gratification and more oriented towards the future. Alternatively, weaker future orientation among low income individuals may simply reflect perceptions of a relatively low likelihood of long-term survival compared with high-income individuals, a form of lay epidemiology (34). However, the differences in future time preference between individuals of low and high socioeconomic status were very small, and more work is needed to determine if time preference contributes to lower rates of preventive health behavior among low socioeconomic status populations in the U.S.

The results of this study must be considered within its limitations. We studied women who had seen a primary care provider in one health system in the Mid-Atlantic region of the U.S. who may be more likely than the general population to adhere to preventive care guidelines. However, there is no reason to believe that these women would have a different future time preference than the general population, suggesting that the association found in this study may be unbiased. This study was correlational in nature, and we therefore cannot draw conclusions about a causal link between time preferences and behavior. Although it is difficult to imagine how engaging in breast cancer–preventive behaviors could change one's future time preference, this pathway cannot be ruled out. Our measure of future time preference, although reasonably reliable, was brief and has not been widely used. In addition, self-report of health behavior has been shown to be <100% accurate (35-37), which may have led to misclassification of the behavior of some subjects. However, this type of nondifferential measurement error would have biased our results towards the null, making it more difficult to find an association between future time preference and behavior. Finally, as with any survey research, response bias is a concern. Although the response rates of 87% and 63% are respectable, it is possible that nonresponders differed from responders. However, Health Insurance Portability and Accountability Act regulations forbid us from collecting data on those who did not consent to participation, which leaves us unable to assess the existence of a nonresponse bias.

Nevertheless, the results of this study provide evidence of an association between future time preference and preventive health behavior against breast cancer risk. The results suggest that the mixed findings regarding the relationship between future time preference and preventive health behavior may be attributable to the extent to which a health behavior is perceived as protecting against future risk compared with whether it leads to detection or reduction of disease in the present. More work is needed to test this hypothesis and to examine whether a weak future orientation among lower socioeconomic status individuals can explain the lower rates of preventive health behavior that they exhibit.

Responses to each of the four items were scored on a four-point Likert scale from “not at all like me” to “extremely like me.”

The authors are grateful to Barbara Weber, Jill Stopfer, Susan Domchek, Amy Badler, and the women who participated in the study.