Search tips
Search criteria 


Logo of hsresearchLink to Publisher's site
Health Serv Res. 2009 June; 44(3): 1068–1087.
PMCID: PMC2699922

Preferences, Beliefs, and Self-Management of Diabetes



To assess relationships between self-assessed control over life events, subjective beliefs about longevity, time and risk preference, and other factors on use of recommended care for diabetes mellitus (DM), self-assessed control of diabetes, general health, and laboratory measures of HbA1c levels.

Data Sources

Health and Retirement Study (HRS) and 2003 HRS Diabetes Study (HRS-DS).

Study Design

We used logit and ordered logit analyses to assess use of recommended care, and subjective and objective measures of health outcomes.

Data Collection

Secondary analysis of HRS and HRS-DS data.

Principal Findings

Individuals with higher self-assessed control over life events and higher subjective probabilities of living 10 years engaged in more recommended DM care practices and had better self-assessed DM control and general health. However, these beliefs did not influence HbA1c levels. More highly educated and cognitively able persons were more likely to follow care recommendations. There were differences by race/ethnicity in health outcomes, but not in health investment among Hispanics.


Individuals' beliefs about control over life events and longevity influenced health investment and subjective health outcomes, although these beliefs did not translate into differences in HbA1c levels. Hispanics may realize lower returns on health investments, at least for diabetes care.

Keywords: Diabetes mellitus, self-management, patient preferences

Diabetes mellitus (DM) prevalence is rising (Mokdad et al. 2000; Bagust et al. 2001). Adherence of diabetes patients to recommended care can markedly reduce complication rates. In the U.K. Prospective Diabetes Study (UKPDS), tight blood glucose control and biannual monitoring lowered the microvascular complication risk by 25 percent (UKPDS Group 1998). Recommended care includes following regimens for prescribed medications, diet and exercise, and screenings, including HbA1C testing, lipid profiles, and eye exams, conducted at least annually. Diabetes outcomes appear to be more closely related to patient characteristics and decisions than to characteristics of patients' physicians or clinics (Glasgow et al. 1997; Heisler et al. 2003a; O'Connor et al. 2008).

Rates of actual use of care fall far short of recommended DM care levels. McGlynn et al. (2003) reported that 45 percent of DM-diagnosed persons in this U.S. study followed recommended care. Some other studies have found higher rates (Harris 1996; Lin et al. 2004), but overall, there is considerable variability in findings (Kell et al. 1999; Cramer 2004). Some results suggest a secular trend toward greater conformance between actual and recommended DM care (Puent, Nichols, and Scarborough 2005), but there is also contrary evidence (Sequist et al. 2006). Frequent physician visits do not assure that persons diagnosed with DM follow care recommendations. Health care providers may fail to abide by the DM care recommendations either by not providing particular services or failing to stress the importance of diabetes self-management to their patients (Sloan et al. 2004; Keating et al. 2007).

We used data from a special survey of DM-diagnosed persons, who had responded to previous waves of the Health and Retirement Study (HRS), a longitudinal database. The special cross-sectional survey, the HRS-Diabetes Study (HRS-DS), is unique in providing information on preferences and beliefs that may affect rates of utilization of care and behaviors relative to recommendations for persons diagnosed with DM and laboratory findings on respondents' HbA1c levels.

Our study assessed relationships between risk tolerance and time preference, self-assessed control over life events, subjective beliefs about longevity, and socioeconomic factors on (1) use of services and behaviors recommended for persons with a diagnosis of DM and (2) subjective and objective measures of diabetes control and health. The contribution of this study is first in its empirical analysis of the roles of preferences and beliefs in utilization decisions and on health outcomes, including a laboratory-based finding among persons with a chronic disease that is both common and becoming more so. Second, we account for differences in many factors, including insurance status, income, and preferences and beliefs, and still find disparities based on race/ethnicity in health outcomes but not health inputs.


In Grossman's (1972) health capital model, individuals demand a commodity, “good health,” both because there is positive utility associated with being healthy, and having good health determines the amount of time available for home and income production. Health capital in any period equals undepreciated health stock from the last period and gross health investment. Individuals can increase or maintain their stock of health capital through investments in medical care and time spent engaging in health-producing activities. An individual's allocation of time depends on time and money prices. One implication of the model is that the amount of time and income devoted to health investment depends on an individual's perception of the effect of current health investment on future health. A person with a high degree of self-efficacy is thus more likely to undertake health investment. People who face competing risks of death are expected to invest less in their health (Dow, Philipson, and Sala-I-Martin 1999). More time-impatient persons are expected to invest less as are more risk-tolerant persons. Higher income increases willingness to pay for good health (Viscusi and Evans 1990). Because it lowers the price of care, being insured should increase use of personal health care services.

In our empirical analysis, both inputs affecting health investment and health outcomes were considered to be endogenous. We estimated reduced form equations for health inputs and health outcomes. We did not attempt to analyze the effect of health inputs on health outcomes, that is, estimate a health production function, because each of the health inputs is endogenous to health outcomes. Individuals who invest more in controlling their diabetes may be healthier or believe they can affect outcomes better. Proper estimation of a health production function would require a panel of outcome data sufficient to estimate changes in health over time, which we do not have for most of our measured outcomes.



Data were obtained from the HRS and the 2003 HRS Diabetes Study (HRS-DS). The HRS is a longitudinal national panel study beginning in 1992 with the main respondents being persons born during 1931–1941 and their spouses who could be of any age; spouses received the same interview as main respondents. The HRS over-samples blacks, Hispanics, and Florida residents, but otherwise it is representative of the U.S. population of this age. In 1998, HRS was combined with another panel survey, Aging and Health Dynamics among the Oldest Old (AHEAD), which surveyed persons born in 1923 and earlier. Also in 1998, an additional sample from the 1942–1948 birth cohort was added. Our study uses data on demographic characteristics, cognitive status, general health status, income, and beliefs about longevity from HRS interviews conducted in 2002.

The HRS-DS, conducted by mail, provides additional clinical information, including self-reports of DM care, use of prescribed medications, health behaviors, preferences, and perceived control of DM. The HRS-DS obtained the respondent's HbA1c levels, as determined by an independent laboratory.

The sample selection process was as follows: 3,194 HRS respondents had reported a DM diagnosis by the 2002 HRS interview. Of these, 680 were excluded because the respondent had participated in a prior HRS special survey. There were 129 2002 HRS respondents who died before the 2003 HRS-DS and 484 persons who did not return the HRS-DS questionnaire, leaving a sample of 1,901 (79 percent response). Valid HbA1c blood spot assays were not returned from 668 HRS-DS respondents, yielding 1,233 valid HbA1c values. Eliminating persons over age 92 reduced the analysis sample to 1,530 and to 1,034 for the HbA1c analysis.


Dependent Variables

Individual's Use of Recommended Care and Practices

We constructed a summary measure of the respondent's use of care and practices recommended by the American Diabetes Association Clinical Practice Recommendations for 2003 (American Diabetes Association 2003). Our index was based on self-reported information from the HRS-DS on whether during the last year the respondent's (1) HbA1c and/or (2) cholesterol was tested; (3) the person received an eye examination; (4) the person tried to lose weight or the person had a Body Mass Index (BMI) of 25 or less at the 2002 HRS interview; (5) s/he engaged in any type of regular exercise; (6) and whether s/he “rarely” or “never” missed taking a prescribed diabetes oral medication.

The dependent variable for an individual's use of recommended care and practices was a count variable based on the number of positive responses to these six items. We used ordered logit analysis with the count of “yes” responses as the dependent variable.

Control of Diabetes

The HRS-DS asked: “How well do you feel that your diabetes has been controlled in the past 6 months?” Responses ranged from 1 “poor” to 5 “excellent.” We analyzed the entire range of values of the dependent variable with an ordered logit analysis.

HbA1c Level

We classified the person's HbA1c level into four categories: 4–6, 7, 8, and 9–15 and used ordered logit analysis. In 2003, the American Diabetes Association regarded values of above 7 as abnormal (American Diabetes Association 2003). A value of 8 was borderline abnormal. Persons who were unwilling to make laboratory findings available to the HRS-DS or for whom values of the HbA1c were unavailable for other reasons did not differ from others on the distribution of responses to the question about self-assessed diabetes control or on age.

Health Status

We assessed determinants of self-assessed general health of persons diagnosed with DM. The health variables could take 1 of 5 values: “excellent” (5), “very good” (4), “good” (3), “fair” (2), and “poor” (1). We used a logit analysis to assess health in 2002. The dependent variable was set to 1 if the person was in good to excellent health and was 0 otherwise.

Explanatory Variables

All explanatory variables came from 2002 HRS interviews1 or from the 2003 HRS-DS. Explanatory variables were as follows:

Self-Assessed Control over Life Events

To identify the individual's perception of his or her control over life events, we used responses to the following statement in the HRS-DS: “I have little control over the things that happen to me.” Responses were reported on a five-point scale ranging from “strongly agree” to “strongly disagree,” with those strongly disagreeing having more self-assessed control. Our measure of control over life events was a binary variable set to 1 for persons who responded “disagree” or “strongly disagree” and was 0 for persons who indicated other levels of agreement with the statement.

Time Preference

The HRS-DS asked respondents to respond to this statement: “I live life one day at a time and don't think much about the future.” To measure time preference, we set responses of “strongly disagree” and “disagree” equal to 1 with the other responses, implying more time impatience for persons assigned a 0 value.

Subjective Beliefs about Longevity

In 2002, the HRS asked each respondent on a 0–100 scale (with 100 being the most likely) what the individual thought was his/her probability of living to a specific age. Persons aged 85+ were not asked this question. When there was no response, we set the missing value of the subjective probability to 0 and defined another explanatory variable for subjective beliefs missing (18 percent of respondents). Other research has used these variables from the HRS (Hurd and McGarry 2002) and from other surveys (Khwaja, Sloan, and Salm 2006). Smith, Taylor, and Sloan (2001), using data from the HRS, reported that people make quite accurate predictions of their own longevity.

Risk Preferences

The HRS contains a question on risk preferences in the financial domain which is asked of a subsample of respondents. The HRS asked the following question: “Suppose that you are the only income earner in the family. Your doctor recommends that you move because of allergies, and you have to choose between two possible jobs. The first would guarantee your current total family income for life. The second is possibly better paying, but the income is also less certain. There is a 50–50 chance the second job would double your total lifetime income and a 50–50 chance that it would cut it by a third. Which job would you take—the first job or the second job?” The HRS then assigned a value from a four-point scale varying from low level of risk tolerance (very risk averse) to risk neutral. We defined binary variables for each point on this risk scale with risk-neutral persons (most risk tolerant) the omitted reference group. When values were missing for 2002, we used responses from 1998 and 2000 HRS interviews. Even so, we lacked values for risk tolerance for 817 respondents, which were set to 0 with another explanatory variable for missing values set to 1.

Time since Diagnosis

The HRS-DS asked respondents about the age of DM diagnosis. Given the person's age, we computed the number of years since diagnosis.

Use of Insulin and Other Diabetes Medications

Respondents were asked by HRS-DS whether they currently took insulin (25 percent of cases) and whether they took some specific oral medications to control hyperglycemia (66 percent). The HRS-DS listed 13 such oral medications. The variable for oral medications was set to 1 if the person took any of the 13 medications. We also defined a binary variable set to 1 for those who used insulin and a separate binary set to 1 for those who used oral medications.

Cognitive Status

The HRS measures cognitive performance using a modified version of the Telephone Interview for Cognitive Status and tests of immediate and delayed verbal recall (Brandt, Spencer, and Folstein 1988; Welsh, Breitner, and Magruder-Habib 1993; Sloan and Wang 2005). The cognition score had a maximum of 35. For missing values, we set the cognitive score to 0 and defined a separate binary variable for cognition missing. The cognitive score consisted of word recall, subtraction, counting backwards, and naming dates and objects tests.

Demographic Characteristics, Health Insurance, and Income

We included explanatory variables for respondent age, years of schooling, and binary variables for marital status (married=1), gender (male=1), and race–ethnicity (black; Hispanic; other race–ethnicity), family income (expressed in '0000$), and binary variables for Medicare, Medicaid, employer-based and other health insurance coverage with no health insurance coverage, the omitted reference group; 3.9 percent of persons were uninsured.


Descriptive Statistics

On average, sample persons engaged in four of six recommended diabetes management practices (Table 1). The modal response for self-assessed control of DM was 3 or “good.” The simple correlation between the four-point scale for individuals' HbA1c levels and the five-point scale for self-assessed diabetes control was −0.31 (p<0.001). Cross tabulations between the two scales showed the expected pattern, but the relationship was far from perfect (not shown). Seventy-eight percent of HRS-DS respondents had HbA1c values of 7 or lower. The modal HbA1c level for the sample was category 1, or an HbA1c level of 4–6. In 2002, 44 percent of sample persons were in fair to poor health in 2002. Nearly half of individuals (46 percent) reported having control over events in their lives and almost the same percentage (42 percent) reported that they think about the future and do not live one day at a time. On average, respondents assigned a probability of living about 10 more years at 0.36, and most individuals fell into the highest (59 percent, the omitted reference group) or lowest (30 percent) risk tolerance categories. Not reflected in the percentage distribution of persons by risk tolerance is a sizable number of missing values (53 percent).

Table 1
Descriptive Statistics

Individual's Use of Recommended Care and Practices

Persons with higher self-assessed control over life events utilized more recommended care and engaged in more recommended health practices (odds ratio [OR]=1.26, 95 percent confidence interval [CI] 1.03–1.54, Table 2) as did those with a higher subjective probability to living about another 10 years (OR=1.60, 95 percent CI 1.16–2.21). We found no statistically significant relationships between either time or risk preference and use of recommended care.

Table 2
Regression Results: Use of Recommended Care

Persons with higher cognition scores (OR=1.03, 95 percent CI 1.02–1.05), educational attainment (OR=1.06, 95 percent CI 1.02–1.09), and those taking insulin (OR=1.32, 95 percent CI 1.04–1.68) and oral medications (OR=2.50, 95 percent CI 1.83–3.40) used more recommended care.2 Men (OR=0.81, 95 percent CI: 0.66–0.99) and blacks (OR=0.74, 95 percent CI 0.56–0.96) used less recommended care, on average. We found no statistically significant differences in use of recommended care by several demographic characteristics, age, marital status, Hispanic ethnicity, income, and type of insurance.

Self-Assessed Control of Diabetes

Greater self-assessed control over life events led to higher self-assessed DM control (OR=1.37, 95 percent CI 1.12–1.68, Table 3). Persons who were relatively optimistic about living another 10 years reported better DM control (OR=2.37, 95 percent CI 1.70–3.32). Relatively risk-averse individuals (lowest risk tolerance) were much more likely to report having good DM control than were risk-neutral individuals, the omitted reference group (OR=1.52, 95 percent CI 0.99–2.33), although the result was not quite statistically significant at conventional levels (p=.057). Time preference was unrelated to self-assessed diabetes control.

Table 3
Regression Results: Perceived DM Control, HbA1c, Self-Reported Health 2002 and 2006

Insulin-dependent persons were much less likely to report that their DM was well controlled than were others (OR=0.52, 95 percent CI 0.41–0.67) and males reported better DM control than did females (OR=1.34, 95 percent CI 1.09–1.64). Older persons also reported better DM control (OR=1.02, 95 percent CI 1.01–1.04). Hispanics were much less likely to report good diabetes control than were whites, the omitted reference group (OR=0.56, 95 percent CI 0.39–0.80).

HbA1c Level

In contrast to the results for use of recommended care, most of the statistically significant covariates in the analysis of HbA1c levels related to measures of DM severity, demographic characteristics, and Medicaid coverage status rather than to beliefs. Neither self-assessed control over life events nor the subjective probability of living another 10 years was associated with the person's HbA1c level. Time preference was not statistically significant. Among the risk tolerance variables, only the low-medium risk tolerance variable was statistically significant, predicting higher HbA1c levels relative to persons who were risk neutral (OR=1.81, 95 percent CI 1.03–3.18). Other statistically significant findings were as follows: use of insulin (OR=2.06, 95 percent CI 1.60–2.65); use of oral medication (OR=1.71, 95 percent CI 1.27–2.31); years since DM diagnosis (OR=1.02, 95 percent CI 1.01–1.02); black race (OR=1.89, 95 percent CI 1.41–2.52) and Hispanic ethnicity (OR=1.61, 95 percent 1.11–2.34); and enrollment in Medicaid (OR=1.51, 95 percent CI 1.07–2.13).

Health Status

Persons with better self-assessed control over life events (OR=1.39, 95 percent CI 1.09–1.77) and who attached a higher subjective probability to living about another 10 years (OR=5.24, 95 percent CI 3.46–7.93) were more likely to be in good to excellent health in 2002. Neither time nor risk preference was related to self-reported health status. Persons with higher cognitive ability (OR=1.03, 95 percent CI 1.01–1.05) and higher family income (OR=1.06, 95 percent CI 1.03–1.10) were more likely to be in good-excellent health; individuals taking insulin (OR=0.55, 95 percent CI 0.41–0.74), blacks (OR=0.59, 95 percent CI 0.43–0.83) and persons covered by Medicare (OR=0.69, 95 percent CI 0.48–0.99) and Medicaid (OR=0.39, 95 percent CI 0.26–0.59) were more likely to report being in fair-poor health.


Beliefs about one's ability to control life events influenced both use of recommended care, and self-assessed control of diabetes, and health for persons diagnosed with DM. However, such beliefs did not translate into our objective measure of diabetes control, the individual's HbA1c level. There is empirical evidence that among persons with DM, beliefs about the effectiveness of activities to control glucose levels predict better DM self-management (Hampson 1997). However, the vast majority of studies of psychosocial predictors of diabetes self-management and objective control have used health- or diabetes-specific measures of control and self-efficacy (Glasgow, Toobert, and Gillette 2001) rather than generalized measures which may better capture underlying personality traits.

There are possible reasons for the discrepancy in findings between the subjective and objective measures of health investment and outcomes. First, diabetes control and general health are more all-encompassing concepts than the single, albeit widely used and objective measure of health of persons with diabetes, the HbA1c. When people think about control of diabetes, they may think more broadly than just about their HbA1c level. Included in the broader definition is control of hypertension and blood cholesterol levels, weight, and general fitness in addition to the HbA1c. A second, more negative interpretation is that persons diagnosed with DM are too optimistic on average about how well their diabetes is controlled, that is, they may underestimate their HbA1c levels and therefore have unrealistic beliefs about their health and DM control.

We found a positive association between the person's subjective probability of living 10 more years and use of recommended care, and between self-assessed control of diabetes and health. However, there was no relationship between subjective beliefs about longevity and the HbA1c level. As the health capital model would predict, persons who believe that they will live many more years invested more in their health as represented in our analysis by use of care recommended to persons with diagnosis of DM. The benefit from investments in health that may improve health outcomes of persons with DM often takes many years to be realized. The UKPDS Group (1998) found that lower glucose levels may take 7+ years to yield health benefits. Thus, persons who perceive that they have few years to live, often because of a competing health risk, have a lower incentive to invest in behaviors that would affect the course of the disease if their longevity prospects were greater (Dow, Philipson, and Sala-I-Martin 1999).

Conceptually, the link between time preference and health investment is clear. More time-impatient individuals should invest less and hence experience poorer health outcomes. Our results provide no support for this view, but, to place our findings on time preference in perspective, it is noteworthy that the literature on the relationship between time preference and health behaviors is mixed.3 In addition, our study used only one question to gauge time preference rather than a validated scale; thus, our lack of findings may be due to the inability of our measure to capture the dimension of time preference that affects health decisions.

Past research also suggests that time discounting in the health domain may strongly depend on the length of the delay in the payoff of a health behavior with time discounting increasing sharply as health payoffs are pushed farther into the future (Asenso-Boadi, Peters, and Coast 2008). Because health benefits from investments in improving the course of one's diabetes may take years to be realized, persons who only expect to live a few years may not consider returns from very long-term investments in their decision making. Benefits of use of recommended care and control of HbAlc may be outside of a reasonable time horizon for an older population of individuals who face other, more immediate, competing health risks. Controversy surrounds the use of the discounted utility model, which is embedded in the health capital model, although it continues to be used, in part for analytic convenience (Frederick, Loewenstein, and O'Donoghue 2002).

Risk tolerance did not predict use of recommended care. The most risk-averse persons were more likely to report having good DM control (although this result was not quite statistically significant at p=.05), and for HbA1c levels, the relationship for risk tolerance did not show a monotonically declining relationship as one would have expected from the results for DM control.

As with time preference, results from the literature on the role of risk preference and health behaviors are mixed. While studies find that risk health behaviors such as smoking, drinking, and seatbelt use are more prevalent among risk-tolerant individuals (see, e.g., Barsky et al. 1997 who used the same HRS measure of risk tolerance we did), other findings are contradictory.4 Rather than suggest that risk preferences are not important in decisions about diabetes care, it is more likely that the measures employed to date have not really captured how risk preferences enter an individual's health decision making process.

We found persons with a higher cognitive score tended to use more recommended care and had better self-reported health, but no relationship was observed for diabetes control or HbA1c. Higher educational attainment was also positively associated with use of recommended care, but there were not statistically significant relationships with the health outcome measures. Previous research has linked educational attainment and socioeconomic status more generally to adherence to DM guidelines for care and for other chronic diseases (Goldman and Smith 2002). In our study, income was only associated with better self-reported health, but not with the diabetes-specific measures.

Successful DM self-management imposes substantial demands on the individual's cognitive skills. Low health literacy is likely to result in poor comprehension of written health care instructions, poor understanding of numerical instructions, for example, number of doses per day, interpreting blood glucose values, reading labels on pill bottles, and making and keeping scheduled appointments (Schillinger et al. 2002). Low health literacy is more common among persons with low educational attainment, older persons, racial and ethnic minorities, and immigrants (Ad Hoc Committee on Health Literacy for the Council on Scientific Affairs 1999). Lower health literacy is associated with poorer self-management of chronic diseases and higher use of services (Baker et al. 1997; Baker et al. 1998). An assessment by physicians of patient recall and comprehension revealed that patients with better recall and comprehension were more likely to have HbA1c levels below the sample mean (Schillinger et al. 2002). Brown et al. (2008) found that higher risk elderly DM patients had significantly more concerns about remembering to take their medication, needing help taking medications, and feeling overwhelmed after doctor's visits.

We found no differences by Hispanic ethnicity in use of recommended care, but worse self-reported control of diabetes by Hispanics, worse self-reported health by blacks, and higher HbA1c levels among both Hispanics and blacks. The pattern of results implies that Hispanics, on average, invest about the same amounts in diabetes control but have a lower return on their investments. For blacks the results imply lower rates of adherence but no difference in perceived control of diabetes. These results are broadly consistent with much of the literature on racial disparities in diabetes care and outcomes. However, our study shows a clearer discrepancy between health investment and outcomes than does past research. For example, Chin, Zhang, and Merrell (1998) reported disparities in access to care, poorer quality of care, and difficulty in getting doctor's visits, suggesting barriers to high-quality care that make our finding of about equal investment even more impressive. De Rekeneire et al. (2003) reported higher levels of HbA1c for blacks than whites even after controlling for socioeconomic characteristics, including educational attainment and insurance coverage. Heisler et al. (2003b) measured five indicators of use of recommended care and found statistically differences in use among two of the five. In a meta-analysis, Kirk et al. (2005) studied DM control among blacks, Hispanics, and whites finding that disparities by race-ethnicity were greatest for glycemic control and least for LDL-C.

Our study has several strengths. Most important are (1) use of measures of self-control, preferences, and beliefs in an assessment of use of recommended care and practices and health outcomes; (2) use of outcomes as perceived by respondents and laboratory-based outcomes; and (3) incorporation of clinical variables as explanatory variables, for example, insulin dependence. Most frequently, studies have included clinical and demographic variables, but they have lacked data on how people think about their care and their health. Harwell et al. (2002) found that only 24 percent of persons could recall their HbA1c levels. However, the HRS-DS obtained the laboratory results on respondents' HbA1c. Social science–oriented studies have generally lacked a clinical component.

We acknowledge these study limitations. First, we relied on individual self-reports of use of care and health practices. Discrepancies between patient and physician reports of the extent to which patients follow diabetes care guidelines have been reported (Lutfey and Ketcham 2005). However, physician reports in this domain may not be the “gold standard” because physicians cannot monitor actual behavior of their patients in their daily lives. A second relates to endogeneity of some covariates. To mitigate the problem of endogeneity, we used questions from the HRS-DS that related to the individual in general rather to the individual's diabetes, even to the point that variables may have been insufficiently domain specific as for risk preference. Nevertheless, some problems of endogeneity may remain. In particular, subjective beliefs about longevity are likely to reflect current health and as the reverse. People knowing that they are unlikely to adhere to recommended care may report lower expected longevity. However, in this context, finding an instrumental variable that is sufficiently well correlated with subjective beliefs about longevity and not with health and use of recommended care is a difficult task.

In conclusion, individuals' beliefs about control over life events and longevity matter to both investments in care and subjective measures of health outcomes. While such beliefs were associated with self-assessed outcomes, we were not able to establish a link between these beliefs and the most widely used single objective measure of diabetes outcomes. Even after controlling for differences in preferences and beliefs, clinical, cognitive, and demographic factors, substantial differences in health outcomes, both subjectively and objectively measured, but not in health investment, by race and ethnicity remained. This implies that blacks and Hispanics may realize lower returns on health investments, at least for diabetes care, than do non-Hispanic whites, even after controlling for socioeconomic factors, cognitive status, and preferences and beliefs. Identifying reasons for these disparities should be a priority for future research.


Joint Acknowledgment/Disclosure Statement: This research was supported in part by a grant from the National Institute on Aging (2R3737-AG-17473-05A1).

Disclosures: There are no conflicts of interest associated with the publication of this paper.

Disclaimers: None.


1We used data from prior waves to fill in missing values.

2The binary variables denoting missing values for use of insulin (OR=0.52, 95 percent CI 0.30–0.53) and oral medications (OR=0.62, 95 percent CI 0.45–0.86) were statistically significant, predicting less use of recommended care.

3Fuchs (1982) found very little evidence that time preference was related to preventive health behaviors such as exercise, dental visits, and seatbelt use. Gurmankin Levy et al. (2006) found weak or nonexistent associations between future time preference and adherence to recommended screening exams. However, Picone, Sloan, and Taylor (2004) found that lower rates of time preference, measured by a person's financial planning horizon, were associated with more demand for early diagnostic screenings and exams in the HRS. Khwaja, Silverman, and Sloan (2007) found that time preference did not vary by smoking status. Bickel, Odum, and Madden (1999), however, using a much smaller sample of younger individuals, reported that smokers do have higher discount rates than nonsmokers.

4Using the HRS risk tolerance measure, Picone, Sloan, and Taylor (2004) found that risk-tolerant women were more likely to have regular recommended cancer screenings, which runs counter to the view that such individuals would be the ones least likely to undergo cancer screening. However, Stiggelbout et al. (1994) found that more risk-averse persons with cancer were more likely to undergo chemotherapy rather than only undergo a surveillance protocol. van der Pol and Ruggeri (2008) found that individuals are risk averse with respect to monetary gambles and gambles between immediate death and 5 life years, while individuals were risk seeking with respect to longevity and quality-of-life tradeoffs, implying that risk tolerance varies in effect, depending on the outcome being considered.

Supporting Information

Additional supporting information may be found in the online version of this article:

Appendix SA1: Author Matrix.

Please note: Wiley-Blackwell is not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.


  • Ad Hoc Committee on Health Literacy for the Council on Scientific Affairs, AMA. Health Literacy: Report of the Council on Scientific Affairs. Journal of the American Medical Association. 1999;281(6):552–7. [PubMed]
  • American Diabetes Association. Standards of Medical Care for Patients with Diabetes Mellitus. (Position Statement) Diabetes Care. 2003;26(1):S33–49. [PubMed]
  • Asenso-Boadi F, Peters TJ, Coast J. Exploring Differences in Empirical Time Preference Rates for Health: An Application of Meta-Regression. Health Economics. 2008;17(2):235–48. [PubMed]
  • Bagust A, Hopkinson PK, Maier W, Currie CJ. An Economic Model of the Long-Term Health Care Burden of Type II Diabetes. Diabetologia. 2001;44(12):2140–55. [PubMed]
  • Baker DW, Parker RM, Williams MV, Clark WS. Health Literacy and the Risk of Hospital Admission. Journal of General Internal Medicine. 1998;13(12):791–8. [PMC free article] [PubMed]
  • Baker DW, Parker RM, Williams MV, Clark WS, Nurss J. The Relationship of Patient Reading Ability to Self-Reported Health and Use of Health Services. American Journal of Public Health. 1997;87(6):1027–30. [PubMed]
  • Barsky RB, Juster FT, Kimball MS, Shapiro MD. Preference Parameters and Behavioral Heterogeneity: An Experimental Approach in the Health and Retirement Study. Quarterly Journal of Economics. 1997;112(2):537–79.
  • Bickel WK, Odum AL, Madden GJ. Impulsivity and Cigarette Smoking: Delay Discounting in Current, Never, and Ex-Smokers. Psychopharmacology. 1999;146(4):447–54. [PubMed]
  • Brandt J, Spencer M, Folstein M. The Telephone Interview for Cognitive Status. Neuropsychiatry, Neuropsychology, and Behavioral Neurology. 1988;1:111–7.
  • Brown S, Meltzer D, Chin M, Huang E. Perceptions of Quality-of-Life Effects of Treatments for Diabetes Mellitus in Vulnerable and Nonvulnerable Older Patients. Journal of the American Geriatrics Society. 2008;56(7):1183–90. [PMC free article] [PubMed]
  • Chin MH, Zhang JX, Merrell K. Diabetes in the African-American Medicare Population. Morbidity, Quality of Care, and Resource Utilization. Diabetes Care. 1998;21(7):1090–5. [PubMed]
  • Cramer JA. A Systematic Review of Adherence with Medications for Diabetes. Diabetes Care. 2004;27(5):1218–24. [PubMed]
  • de Rekeneire N, Rooks RN, Simonsick EM, Shorr RI, Kuller LH, Schwartz AV, Harris TB. Racial Differences in Glycemic Control in a Well-Functioning Older Diabetic Population: Findings from the Health, Aging and Body Composition Study. Diabetes Care. 2003;26(7):1986–92. [PubMed]
  • Dow WH, Philipson TJ, Sala-I-Martin X. Longevity Complementarities under Competing Risks. American Economic Review. 1999;89(5):1358–71.
  • Frederick S, Loewenstein G, O'Donoghue T. Time Discounting and Time Preference: A Critical Review. Journal of Economic Literature. 2002;40(2):351–401.
  • Fuchs V. Time Preference and Health: An Exploratory Study. In: Fuchs V, editor. Economic Aspects of Health. Chicago: University of Chicago Press; 1982. pp. 93–120.
  • Glasgow REP, Hampson SEP, Strycker L A B A, Ruggiero LP. Personal-Model Beliefs and Social-Environmental Barriers Related to Diabetes Self-Management. Diabetes Care. 1997;20(4):556–61. [PubMed]
  • Glasgow RE, Toobert DJ, Gillette CD. Psychosocial Barriers to Diabetes Self-Management and Quality of Life. Diabetes Spectrum. 2001;14(1):33–41.
  • Goldman DP, Smith JP. Can Patient Self-Management Help Explain the SES Health Gradient? Proceedings of the National Academy of Sciences. 2002;99(16):10929–34. [PubMed]
  • Grossman M. On the Concept of Health Capital and the Demand for Health. Journal of Political Economy. 1972;80(2):223–55.
  • Hampson SE. Personal Models and the Management of Chronic Illness: A Comparison of Diabetes and Osteoarthritis. European Journal of Personality. 1997;11(5):401–14.
  • Harris MI. Medical Care for Patients with Diabetes. Annals of Internal Medicine. 1996;124(1):117–22. [PubMed]
  • Harwell TS, Dettori N, McDowall JM, Quesenberry K, Priest L, Butcher MK, Flook BN, Helgerson SD, Gohdes D. Do Persons with Diabetes Know Their (AIC) Number? The Diabetes Educator. 2002;28(1):99–105. [PubMed]
  • Heisler M, Smith DM, Hayward RA, Krein SL, Kerr EA. How Well Do Patients' Assessments of Their Diabetes Self-Management Correlate with Actual Glycemic Control and Receipt of Recommended Diabetes Services? Diabetes Care. 2003a;26(3):738–43. [PubMed]
  • Heisler M. Racial Disparities in Diabetes Care Processes, Outcomes, and Treatment Intensity. Medical Care. 2003b;41(11):1221–32. [PubMed]
  • Hurd MD, McGarry K. The Predictive Validity of Subjective Probabilities of Survival. Economic Journal. 2002;112(482):966–85.
  • Keating NL, Zaslavsky AM, Herrinton LJ, Selby JV, Wolf RE, Ayanian JZ. Quality of Diabetes Care among Cancer Survivors with Diabetes. Medical Care. 2007;45(9):869–75. [PubMed]
  • Kell SH, Drass J, Bausell RB, Thomas KA, Osborn MA, Gohdes D. Measure of Disease Control in Medicare Beneficiaries with Diabetes Mellitus. Journal of the American Geriatrics Society. 1999;47(4):417–22. [PubMed]
  • Khwaja A, Silverman D, Sloan F. Time Preference, Time Discounting, and Smoking Decisions. Journal of Health Economics. 2007;26(5):927–49. [PubMed]
  • Khwaja A, Sloan F, Salm M. Evidence on Preferences and Subjective Beliefs of Risk Takers: The Case of Smokers. International Journal of Industrial Organization. 2006;24(4):667–82.
  • Kirk JK, Bell RA, Bertoni AG, Arcury TA, Quandt SA, Goff DC, Jr., Narayan KMV. Ethnic Disparities: Control of Glycemia, Blood Pressure, and LDL Cholesterol among US Adults with Type 2 Diabetes. Annals in Pharmacotherapy. 2005;39(9):1489–501. [PubMed]
  • Levy A, Micco E, Putt M, Armstrong K. Value for the Future and Breast Cancer-Preventive Health Behavior. Cancer Epidemiology Biomarkers and Prevention. 2006;15(5):955–60. [PubMed]
  • Lin EHB, Katon W, Von Korff M, Rutter C, Simon GE, Oliver M, Ciechanowski P, Ludman EJ, Bush T, Young B. Relationship of Depression and Diabetes Self-Care, Medication Adherence, and Preventive Care. Diabetes Care. 2004;27(9):2154–60. [PubMed]
  • Lutfey KE, Ketcham JD. Patient and Provider Assessments of Adherence and the Sources of Disparities: Evidence from Diabetes Care. Health Services Research. 2005;40(6):1803–17. [PMC free article] [PubMed]
  • McGlynn EA, Asch SM, Adams J, Keesey J, Hicks J, DeCristofaro A, Kerr EA. The Quality of Health Care Delivered to Adults in the United States. New England Journal of Medicine. 2003;348(26):2635–45. [PubMed]
  • Mokdad AH, Ford ES, Bowman BA, Nelson DE, Engelgau MM, Vinicor F, Marks JS. Diabetes Trends in the U.S.: 1990–1998. Diabetes Care. 2000;23(9):1278–83. [PubMed]
  • O'Connor P, Rush W, Davidson G, Louis T, Solberg L, Crain L, Johnson P, Whitebird R. Variation in Quality of Diabetes Care at the Levels of Patient, Physician, and Clinic. Preventing Chronic Disease: Public Health Research, Practice, and Policy. 2008;5(1):1–9. [PMC free article] [PubMed]
  • Picone G, Sloan F, Taylor D., Jr. Effects of Risk and Time Preference and Expected Longevity on Demand for Medical Tests. Journal of Risk and Uncertainty. 2004;28(1):39–53.
  • Puent BD, Nichols KK, Scarborough R. The Frequency of Diliated Diabetic Eye Examinations by Optometrist in the State of Ohio. Optometry and Vision Science. 2005;82(11):959–63. [PubMed]
  • Schillinger D, Grumbach K, Piette J, Wang F, Osmond D, Daher C, Palacios J, Sullivan GD, Bindman AB. Association of Health Literacy with Diabetes Outcomes. Journal of the American Medical Association. 2002;288(4):475–82. [PubMed]
  • Sequist TD, Adams A, Zhang F, Ross-Degnan D, Ayanian JZ. Effect of Quality Improvement on Racial Disparities in Diabetes Care. Archives of Internal Medicine. 2006;166(6):675–81. [PubMed]
  • Sloan FA, Bethel MA, Lee PP, Brown DS, Feinglos MN. Adherence to Guidelines and Its Effects on Hospitalizations with Complications of Type 2 Diabetes. Review of Diabetic Studies. 2004;1(1):29–38. [PubMed]
  • Sloan F, Wang J. Disparities among Older Adults in Measures of Cognitive Function by Race or Ethnicity. Journals of Gerontology. 2005;60B(5):242–50. [PubMed]
  • Smith VK, Taylor D, Sloan F. Longevity Expectations and Death: Can People Predict Their Own Demise? American Economic Review. 2001;91(4):1126–34.
  • Stiggelbout AM, Kiebert GM, Kievit J, Leer JWH, Stoter G, De Haes JCJM. Utility Assessment in Cancer Patients: Adjustment of Time Tradeoff Scores for the Utility of Life Years and Comparison with Standard Gamble Scores. Medical Decision Making. 1994;14(1):82–90. [PubMed]
  • U.K. Prospective Diabetes Study (UKPDS) Group. Intensive Blood-Glucose Control with Sulphonylureas or Insulin Compared with Conventional Treatment and Risk of Complications in Patients with Type 2 Diabetes (UKPDS 33), Part 1. Lancet. 1998;352(9131):837–53. [PubMed]
  • U.K. Prospective Study Group. Effect of Intensive Blood-Glucose Control with Metformin on Complications in Overweight Patients with Type 2 Diabetes (ukpds 34) Lancet. 1998;352:854–65. [PubMed]
  • van der Pol M, Ruggeri M. Is Risk Attitude Outcome Specific within the Health Domain? Journal of Health Economics. 2008;27(3):706–17. [PubMed]
  • Viscusi WK, Evans WN. Utility Functions That Depend on Health Status: Estimates and Economic Implications. American Economic Review. 1990;80(3):353–74.
  • Welsh K, Breitner J, Magruder-Habib K. Properties of the Telephone Interview for Cognitive Status: Application in Epidemiological and Longitudinal Studies. Neuropsychiatry Neuropsychology and Behavioral Neurology. 1993;6(2):103–10.

Articles from Health Services Research are provided here courtesy of Health Research & Educational Trust