• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of geronbLink to Publisher's site
J Gerontol B Psychol Sci Soc Sci. May 2010; 65B(3): 390–399.
Published online Jan 1, 2009. doi:  10.1093/geronb/gbp050
PMCID: PMC2853599

Diabetes-Related Support, Regimen Adherence, and Health Decline Among Older Adults



Social support is generally conceptualized as health promoting; however, there is little consensus regarding the mechanisms through which support is protective. Illness support has been proposed to promote regimen adherence and subsequent prevention of health decline. We hypothesize that (a) support for regimen adherence is negatively associated with self-reported health decline among older diabetic adults and that (b) regimen adherence is negatively associated with health decline among older diabetic adults.


We used the Health and Retirement Study data on individuals over the age of 60 years with type 2 diabetes mellitus (n = 1,788), examining change in self-reported health status over a 2-year period using binomial and cumulative ordinal logistic regression models.


Diabetic support is not significantly associated with health decline, but it is strongly associated with adherence to health-promoting activities consisting of a diabetic regimen. Therefore, the extent to which one receives illness support for a given regimen component is highly positively associated with adhering to that component, although this adherence does not necessarily translate into protection against perceived decline in health.


Illness-related support appears to be a mechanism through which social support matters in the diabetic population. Although this relationship did not extend to prevention of health status decline among diabetics, the relationship between support and illness management is promising.

Keywords: Chronic illness, Health determinants, Medical regimens, Patient adherence, Self-care, Self-rated health, Social support, Type 2 diabetes

SOCIOLOGICAL research has consistently emphasized the vulnerability associated with social isolation and the benefits arising from social integration (Durkheim, 1897/1951; Kohn & Clausen, 1955). However, the field has encountered more difficulty in quantifying and generalizing this relationship due to endogeneity concerns as the majority of studies are based on cross-sectional data. The past several decades of social epidemiological research on relationships has used regional prospective samples, enabling a more confident conclusion that the association is, at least in part, causally determined (Berkman & Syme, 1979; House, Landis, & Umberson, 1998; Schoenbach et al., 1986; Schulz et al., 2006; Uchino, 2004; Wills & Filer, 2001). Furthermore, great strides have been made in the conceptualization and measurement of social support and social networks in relation to the behavioral, health, and social sciences (Ajrouch, Antonucci, & Janevic, 2001; Cohen & Gottlieb, 2000; Sarason, Sarason, & Gurung, 2001). However, these relationships vary in significance by age, sex, race/ethnicity, and socioeconomic status (Bae, Hashimoto, Karlson, Liang, & Daltroy, 2001; Blazer, 1982; Everard, Lach, Fisher, & Baum, 2000; House, Robbins, & Metzner, 1982; House et al., 1998; Kaplan et al., 1988; Schoenbach et al.; Seeman, Kaplan, Knudsen, Cohen, & Guralnik, 1987). Despite the emerging empirical evidence supporting the connection between health and social relationships, there is little theoretical or empirical consensus regarding the mechanisms.

Social support and diabetes

In the examination of aging chronically ill populations, the relationship between social support and health status has also been debated in the literature. Previous studies generally suggest that social support is positively associated with health status, but research has long lacked consensus on the mechanisms through which this relationship operates (Kaplan, Cassel, & Gore, 1977). Numerous mechanisms and mediators have been suggested in the literature (Kaplan, 1989), some of which include the promotion of self-esteem and control through relationships (Krause & Borawski-Clark, 1994), the receipt of informal care (Langa et al., 2002), and illness self-management (Gallant, 2003). This literature suggests that, as the population ages, it is necessary to move beyond the individual to understand how health outcomes can be improved (Gore, 1989); however, a better understanding of the attributes of social support that contribute to the optimization of health outcomes among the chronically ill is yet to be attained (Gallant).

Older adults with type 2 diabetes mellitus are an ideal group to extend the analysis of this relationship. Diabetes mellitus is extremely costly, involving high direct medical costs (estimated $44 billion per year in 1997 dollars) and indirect costs (estimated $54 billion per year), such as lost productivity (American Diabetes Association, 1998). In addition, the indirect cost of informal care for diabetics has been estimated between $3 and $6 billion per year (Langa et al., 2002). Furthermore, older adults with diabetes mellitus of intermediate functioning have, in particular, been found to have a sharper general decline than non-diabetics (Blaum, Ofstedal, Langa, & Wray, 2003). As the proportion of Americans with diabetes grows and the population ages, it is crucial that more is known about the successful treatment of this illness. Due to the complex and rigorous regimen required for successful maintenance, diabetes mellitus has been referred to as an “exemplar” for the need to better understand the correlates of successful self-management (Hill-Briggs, 2003). The high cost, high incidence and prevalence, and complex regimen related to diabetes highlight the importance for understanding how support can improve adherence. The extent to which social support is protective for regimen adherence—and overall health—among diabetics therefore has important implications for policy and practice.

Research on the effects of social support on health status among diabetics and other chronically ill populations centers around enhancement of commitment to self-care or regimen adherence (Belgrave & Lewis, 1994; Peyrot, McMurry, & Hedges, 1987; Ruggiero, Spirito, Bond, Constan, & McGarvey, 1990). In a comprehensive meta-analysis of studies addressing the relationship between social support and chronic illness self-management, Gallant (2003, p. 170) finds that despite evidence for a “modest” positive relationship, especially among diabetics, few studies have addressed this relationship adequately, with the majority focusing on cross-sectional, relatively young, and ethnically homogenous samples. Despite previous efforts, the relationship between social support, regimen adherence, and overall health is yet to be thoroughly examined. Given the limitations of the previous studies, several factors remain ambiguous. Most are cross-sectional and lack a strong theoretical base, and are thus limited in their ability to assert the direction of causality. Furthermore, previous studies examine vastly different notions of social support (family cohesion, community involvement, social ties, etc.).

Previous research has suggested that adherence to a diabetic regimen is protective against health decline through such pathways as glycemic control and obtaining standardized tests and therapies. Through improved adherence, therefore, studies have shown that an individual will report improved outcomes not only because one’s health is better but largely because it is perceived to be better as well (Heisler, Smith, Hayward, Krein, & Kerr, 2003). Despite such findings, however, the borders between self-perceived health status, changes in health status, and regimen adherence continue to be undefined, largely due to the cross-sectional nature of most studies. Furthermore, research is needed that investigates the role regimen adherence might take as a mediator to examine either the relationship between social support and regimen adherence or the relationship between social support and health status, but not both. Therefore, such studies were unable to determine whether health benefits related to social support is largely a consequence of improved adherence. Additional research is needed that longitudinally examines the relationship between social support and health status, as well as how it relates to regimen adherence. Ideally, this research should provide insight into the mechanisms by which social support contributes to health.

Focusing on diabetic older adults, this research attempts to clarify this concern by addressing the following questions: (a) What is the association between illness-related support (in contrast to competing forms of social relationships) and health decline among this population? and (b) What is the relationship between illness-related support and regimen adherence among this population? Our approach draws upon a large nationally representative sample of older adults. Furthermore, we consider the effects of social relationships and regimen support separately. Finally, we examine sociodemographic variation in the study of social support, adherence, and health decline. Using this approach, this analysis has the potential to inform policy and practice interventions.

Conceptual/theoretical framework

We use the health decision model (HDM) as a framework to capture the relative contribution of social support in enhancing adherence and health outcomes among diabetic older adults. The HDM, from Eraker, Kirscht, and Becker (1984), builds upon Becker’s health belief model by incorporating preferences, including decision analysis and behavioral decision theory. The HDM has been used to examine patient adherence to smoking cessation interventions (Eraker, Becker, Strecher, & Kirscht, 1985) and racial differences of health-related beliefs, attitudes, and experiences of cardiac patients (Kressin et al., 2002) and can also be used to frame the analysis of the relationship between health decline among diabetics, social support, and regimen adherence. According to this model, health decisions and behavior (such as short- and long-term compliance) are influenced by sociodemographic characteristics and social interactions (social networks, support, and patient supervision). However, these factors also influence a patient’s experience with—and knowledge about—the illness (including disease, diagnostic and therapeutic interventions, and health care providers). Patient experience and knowledge independently influence outcomes (including adherence) but have additional interacting effects on patient preferences (decision-making processes) and health beliefs (specific and general).

This model is particularly useful for examining the relationship between social support and health outcomes (including adherence) among chronically ill patients with a complex regimen, as it emphasizes that patient adherence is a function of numerous complex factors, which change according to a patient’s disease attitudes and over time (Eraker et al., 1984). Furthermore, the model suggests that individual adherence does not necessarily translate into positive health outcomes, which can be moderated by other factors such as health beliefs and decisions. Finally, although few studies have comprehensively addressed the social distribution of support (House et al., 1998; Thoits, 1995; Turner &d Marino, 1994), studies have found that that sociodemographic characteristics such as race/ethnicity and sex can moderate this relationship (Connell, Fisher, & Houston, 1992; Connell, Sorandt, & Lichty, 1990; Fitzgerald et al., 1997; Gallant, 2003). Social status has been found in previous research to increase the odds of diabetic prevalence in midlife as well as health behaviors and disease management following diagnosis (Wray, Alwin, McCammon, Manning, & Best, 2006). Lower social position is a risk factor for earlier disease onset and worse management in midlife. Even holding health behaviors constant later in life, the disease prevalence and health behaviors earlier can strongly predict health outcomes among elderly diabetics.

In conclusion, the HDM suggests several mechanisms through which social support can influence health outcomes and adherence among the chronically ill, including type 2 diabetics. However, the HDM does not suggest which competing forms of social support (social relationships/networks, informal care, or illness support) would be strongest in this relationship. As argued by Coyne and DeLongis (1986), there is a need to “go beyond” social support to investigate determinants of health and well-being to further examine the relative influences of social support on health outcomes, including adherence (Gallant, 2003). Here, drawing on the HDM, we contribute to the literature on social support and health outcomes by examining the relationship between social support (focusing on illness-related support but controlling for other aspects) and disease outcomes: regimen adherence and overall health decline among diabetic older adults.


To address our research questions, we propose the following two hypotheses, which are followed by two separate analyses:

Hypothesis 1: Illness-related support is negatively associated with health status decline.

Hypothesis 2: Regimen adherence is negatively associated with health status decline.



To conduct our two analyses (a) examining the impact of social support on regimen adherence and health decline over time and (b) examining the relationship between regimen adherence and distal health outcomes, we will analyze data from the Health and Retirement Study (HRS) Waves 6 and 7 (2002 and 2004) as well as the 2003 diabetes supplement. HRS is a national population-based study that has tracked individuals and households for a 12-year period. The HRS 2003 Diabetes Study (2006) is sponsored by the National Institute on Aging (grant number NIA U01AG009740) and was conducted by the University of Michigan (Survey Research Center).

In 1992, 12,654 community-dwelling individuals born from 1931 to 1941 participated in the study (response rate 81.7%). Adjusting for respondent mortality, the response rates have remained above 84% in the six subsequent waves, with a sample size of 10,142 individuals in Wave 6 (2002) and 9,759 in Wave 7 (2004). The 2003 diabetes supplement fielded 2,381 cases (a) reporting a diagnosis of diabetes in the 2002 HRS and (b) eligible for the supplemental examination (not participating in the HRS Consumption and Activities Mail Survey). This mailed survey requested information pertinent to a variety of domains, including data on medications for diabetes, provider interactions, and comorbidities. Data from this questionnaire can be linked with the larger HRS sample by unique case identifiers. Questionnaires were returned by 1,901 respondents (a response rate of 79.8%).

Data weighting and analytic subsample.—

Of those returning questionnaires for the diabetes supplement, 1,851 reported that they had type 2 diabetes (1,603) or did not indicate which type (248). The remainder of cases reported having type 1 diabetes (50). Of the 1,851, 1,788 reported having one or more elements of the diabetes regimen (discussed subsequently) and were followed up for the seventh survey wave in 2004. Given that self-reported diabetes diagnosis has been found to be a valid and reliable indication (when compared with laboratory tests), we will restrict our study population to those indicating a positive diagnosis of type 2 diabetes mellitus or unknown diabetes types with an age of onset more than 30 years (Kaye, Folsom, Sprafka, Prineas, & Wallace, 1991; Midthjell, Holmen, Bjorrndal, & Lund-Larsen, 1992). Weights were constructed to adjust for attrition and sampling bias to generate unbiased estimates for 2003 mail-out respondents specifically.


Dependent variable.—

Decline in self-rated health from 2002 to 2004 is the outcome variable in the first analysis (the relationship between illness-related support and health decline). An indicator variable (health decline) was constructed to determine if health from the preceding wave (a) got worse or (b) maintained or improved. In 2002, respondents indicated whether they would rate their overall health as excellent, very good, good, fair, or poor. Those reporting a value upon follow-up of at least one unit of poorer overall health were labeled as having experienced health decline. Overall health status is a categorical variable in the 2002 HRS, with 15% reporting excellent health, 32% reporting that their health is “very good,” 35% reporting “good,” 15% reporting “fair,” and 2% reporting poor health.

Health change is contingent upon health status; therefore, this outcome variable is controlled for by overall self-reported health status. Self-reported and self-assessed measures have been used widely in epidemiological and social research. This global categorical measure (self-rated health on a 1–5 scale) has been found to be highly concordant with clinical assessments, as well as a reliable predictor of mortality and health care utilization (Idler & Benyamini, 1997). For the second part of the analysis (testing the relationship between self-assessed illness-related support and reported adherence), the dependent variable is self-reported adherence to the six diabetic regimen components.

Independent variables.—

A diabetic regimen is ideally customized through the patient–provider interaction to optimize adherence and successful treatment. Although the extent may vary, the prevention of diabetic complications generally calls for the following as regimen components:

  • Taking diabetes medications (pills and/or insulin);
  • exercising regularly;
  • following a recommended eating plan;
  • checking blood sugar;
  • checking feet for wounds or sores; and
  • seeing doctors or other providers.

In the 2003 diabetes supplement, participants indicated (through a 5-point Likert scale) the extent to which they can rely on family or friends to provide help and support for each regimen component (illness-related support). For each component, participants also indicated level of difficulty, or to which degree they adhered to each regimen component, ranging from so difficult that I couldn’t do it at all to not difficult; I got it exactly right (adherence). Respondents had the option of indicating if a component was not part of their regimen in which case that response is excluded from the analysis.

A global dichotomous adherence measure is used in the first analysis to test the primary hypothesis (and shown in Tables 1 and and22 and Models 1–3, discussed subsequently). Individuals are coded as relatively adherent who report relatively high levels of compliance with their overall diabetic regimen, whereas respondents are coded as relatively nonadherent who report relatively low levels of compliance. In the second analysis, to test the second hypothesis (see Table 3), health status is regressed on the six different components of the diabetic regimen discussed previously, with the Likert scale maintained.

Table 1.
Distribution of Study Variables in HRS Population, Weighted and Unweighted
Table 2.
Logistic Regression Analysis of the Probability of Health Status Decline
Table 3.
Ordinal Logit Analyses of Social Support on Regimen Adherence

As discussed previously, social support has been conceptualized and measured differently in the literature. In this study, we will focus on illness-related social support, while using support-related characteristics (provision of informal/unpaid care and social relationships) as alternative measures of support included in the analysis as covariates. Our rationale for focusing on illness- or adherence-related support in our analysis is that—as mentioned previously—previous literature has suggested that adherence is the most probable mechanism for the negative relationship between health decline and support. Including additional variables related to support as covariates will enable us to peripherally examine alternative explanations for such a relationship, provided it is found.

In the first analysis, social support for regimen components is examined with health decline, controlling for regimen adherence and health status/morbidity. For the second analysis, social support components are examined with adherence, controlling for health status, change, and additional control variables. Health status is obtained from the 2002 HRS and categorically measured through a 5-point Likert scale self-assessing overall health from poor to excellent. Morbidity is measured by an imputed variable of the Total Illness Burden Index (TIBI) score of comorbidities from the HRS 2004 tracker file. The TIBI is a composite measure of self-reported medical events and symptoms (Greenfield et al., 1995). Finally, duration of diabetes is determined by subtracting the age of diabetes diagnosis (diabetes supplement 2003) from current age (HRS tracker file 2004). The consideration of health status or morbidity in the analysis enables us to statistically isolate health decline from the potentially collinear effect of overall health.

Control variables.—

We are concerned with two relationships in this study: (a) the relationship between social support and health status change among the chronically ill and (b) the relationship between social support and regimen adherence. Although the dependent and independent variables described previously enable these analyses, social support might influence health change and adherence through other mechanisms, such as the direct provision of diabetes-related care or the health effects of companionship in old age. Marital status and informal diabetes caregiving will therefore be used to measure these competing aspects of social support. The former is obtained through the 2004 HRS respondent tracker file, with “1” indicating married and “0” a collapsed variable including divorced, widowed, separated, and never married. Informal diabetes caregiving is captured from the diabetes supplement question, “Besides your health care providers, who helps you the most in caring for your diabetes?” Respondents indicating spouse, other family members, or friends are determined to have informal diabetes caregivers. Respondents indicating “paid helper” or “nobody” are determined not to have informal diabetes caregivers.

The measured covariates (marital status and receipt of informal care) do not perfectly capture alternative forms of social support. For example, individuals who are currently married might have more interaction with a marital partner than those who are divorced, widowed, separated, and never married; however, this does not include life partners, closer personal friends, or contact with direct or extended family, community involvement, or friendships. Furthermore, the receipt of informal or unpaid care is not necessarily indicative of a higher level of support than the receipt of paid care or no care, paid or unpaid. For example, previous research has suggested that roles between formal and informal care can overlap or cross over in payment or relationship (Allen & Ciambrone, 2003; Porter, Ganong, Drew, & Lanes, 2004). That said, receipt of informal care has been raised as an alternative explanation for the relationship between social support and distal health outcomes. Including informal care as a covariate will enable illness-related support to be distinguished from the direct provision of care.

As previous research suggests (Connell et al., 1990, 1992; Fitzgerald et al., 1997; Gallant, 2003), social support and health status for the chronically ill can vary across such factors as race, class, sex, and age, which could potentially bias this analysis. Using demographic data from the 2002 HRS tracker file, sex (male or female), race/ethnicity (Black, Hispanic, and White), and age are used as controls in this analysis. In addition, educational achievement (less than high school, completion of high school, and some college or more) is used as a proxy control variable for socioeconomic status. The education variable is chosen as a proxy over other concrete statuses (such as income or attained wealth) because education is stable, whereas other indicators are highly variable at different stages of the earnings and retirement process. Measures are coded into categories to enable a more intuitive and meaningful interpretation of descriptive and analytical statistics. For the purposes of preliminary descriptive statistics, age and education levels are analyzed in groups. Four equal-range age groups are analyzed from 60–69 to 90–99 years. These groupings are intended to capture the relationship between social support, adherence, and health status change among different sociodemographic groups.

A series of statistical procedures is employed to test the hypotheses concerning social support, regimen adherence, and health status decline. As discussed subsequently, descriptive statistics on social, health, and demographic variables will provide preliminary data on sample characteristics. The hypotheses are tested through a series of binomial and ordinal logistic regression models. All analyses are weighted for nonresponse and differential subgroup sampling, unless otherwise indicated. Sampling weights were used from the 2003 diabetes mail-out study (type 2 diabetics represent more than 90% of the sample).

Descriptive statistics are calculated to examine sample characteristics among respondents experiencing health decline. The characteristics explored are demographic variables (age group, sex, years of education, and race), health and diabetes variables (self-rated health, diabetes duration, and morbidity), and social variables (marital status and informal diabetes care). Within these categorical groups, weighted proportions and standard errors are reported as well as the unweighted count of respondents in each category.

Hypothesis 1

A series of multivariate logistic regression models will enable us to test the hypothesis that social support is negatively associated with health status decline, controlling for regimen adherence, health status, and additional covariates.

Model 1.—

In the first model, the six social support regimen variables are regressed on the indicator variable for health decline, controlling for age and self-rated health in 2002. This analysis will provide a preliminary age- and health-adjusted indication of the relationship between support and health status decline.

Model 2.—

The second model will include the same social support regimen variables as in Model 1 but will also include core (and competing) social relationship variables of marital or coupled status and receipt of informal care. This will enable an independent evaluation of the strength of influence of different forms of support on health decline.

Model 3.—

The final model includes the core social support variables, social variables, and health- and diabetes-related variables (reported regimen adherence, duration of diabetes, and morbidity). The final model enables us to better control for illness severity and individual-level behaviors, as well as for the additional demographic characteristics sex, years of education, and race/ethnicity.

Hypothesis 2

Finally, to examine the second hypothesis that illness support is positively associated with adherence, we will conduct a series of ordinal (cumulative probability) logistic regressions. In each of the six models, illness support is regressed on its corresponding attribute of regimen adherence (e.g., adherence to a meal plan regimen would be the outcome variable and support from family or friends for meal plans is the explanatory variable). All health- and diabetes-related variables, social variables, and demographic control variables are included in the series of regressions.


Sample Characteristics

Table 1 compares weighted and unweighted samples in several factors to assess the impact of oversampling and nonresponse bias. This table demonstrates that the sample population is highly representative of the older adult diabetic population. The sample population is roughly similar in terms of age, sex, and education characteristics as the general population. Blacks and Hispanics are slightly overrepresented due to purposive oversampling. The weighted sample proportion reporting having support related to diabetes regimen components was high, ranging from nearly 72% for checking blood sugar to more than 96% for taking medications. More than 70% of the weighted sample reported following their diabetes regimen. The majority of weighted respondents were married (63%), whereas a minority received informal care from friends or family (24%). Although the unweighted sample had nearly identical proportions as the weighted sample in regimen support and adherence domains, the sample disproportionately represented those with poor and fair health, justifying the use of survey weights in this analysis due the complex design.

Preliminary Diagnostics

The proportion of the sample experiencing declining health varies by demographic, health or diabetes status, and social characteristics. Specifically, the proportion reporting declining health was positively associated with age (highest burden in age group 80–89 years), being women (37% vs. 32%), and fewer years of education. As expected, health decline was strongly related to poor health with the follow-up, with 79% of the individuals reporting poor health who had experienced health decline. Duration of diabetes and TIBI were generally positively associated with health decline. Finally, among the additional social variables examined here (marital status and provision of diabetes care by friends or family), the proportions of those reporting health decline were surprisingly similar, ranging from 34% to 35%.

Multivariate Analysis: Binomial Logistic Regression Models

Model 1.—

Among the six illness support variables tested on health decline, none was statistically significant. Self-rated health and age were highly significant as predicted (Table 2).

Model 2.—

With the inclusion of the social relationships variables, support variables continued not to be significant. Being married or coupled was not significantly associated with health status decline, whereas the receipt of informal care was strongly positively associated with decline (83%). Age and self-rated health remained statistically significant as predicted (Table 2).

Model 3.—

In the final model, diabetes support variables remained nonsignificant. Furthermore, the competing measures of social support (marital status and receipt of informal care) changed little from previous model, a reduction in the odds of decline of 8% and an increase of 97.5%. The illness support variables of interest remained stable with the inclusion of the additional health, diabetes-related, and sociodemographic variables. Self-reported regimen adherence was slightly negatively associated with decline (3.9%), whereas diabetes duration and TIBI were slightly positively associated with decline. We did not find significant differences by sex. Furthermore, education level was not statistically associated with decline when comparing respondents with education below high school with high school graduates; however, individuals with education above a high school level had 45% lower odds of health decline compared with high school graduates, which was significant. Relative to Whites, Blacks did not significantly differ in health decline; however, Hispanics had significantly higher odds of health decline relative to Whites (Table 2).

Multivariate Analysis: Ordinal Logistic Regression Models

The series of ordinal logistic regression models tested the relationship between illness support and component regimen adherence, controlling for all other factors included in Model 3. Each being highly significant, we find that a one-level increase in social illness increases the odds of adherence to the regimen by 59% (medications), 61% (appointments), 62% (checking blood sugar), 62% (checking feet), 70% (exercising), and 110% (following eating plan; Table 3).


This research examined the relationship between illness support and health status, with regimen adherence at its center. We were able to analyze—through six different regimen components—the association of support, adherence, and health status decline for a 2-year period. The finding that illness-related support was not significantly associated with distal health outcomes (but that illness-related support was significantly related to adherence) challenges previous assumptions that regimen adherence will necessarily translate into improved (or maintained) health outcomes. This is—to our knowledge—a unique finding. Further research should examine the mechanisms underlying—as well as endogeneity and temporal issues related to—the protective and risk-associated behaviors associated with illness support. Given that the period of health change is only 2 years, it is possible that illness-related support could be operating in numerous ways. Support might be increased during the time that health is in rapid decline, indicating a “need for support intervention” from friends and family. Given that these directions generally remained stable and consistent throughout the models, they should be explored in prospective quantitative and qualitative research. These relationships vary significantly by age, health status, race/ethnicity, and years of education, suggesting that illness support operates differently in sociodemographic domains. In addition, the degree to which isolation is a risk factor, or to which support has a buffering effect, warrants further investigation.

These models also enabled us to test the extent to which illness support might be confounded by social relationships or receipt of informal care. The latter was consistently significant. The extent to which illness support is associated with regimen adherence is of particular concern. From the series of ordinal logistic regressions, we were able to assert that illness support is significantly associated with adherence for each regimen component tested here, controlling for other factors. These findings are consistent with previous literature examining the relationship between illness support and regimen adherence among chronically ill populations—particularly diabetics (Connell et al., 1990, 1992; Gallant, 2003; Sherbourne et al., 1992; Wilson et al., 1986).

This relationship between illness support, adherence, and health outcomes does not conflict with Eraker and colleagues’ (1984) HDM, which suggests that health determinants might relate to outcomes in a nonlinear fashion. This study marks a stepping-stone in the elucidation of the “black box” of social support theory into tangible mechanisms of its relation to health status and how it varies by sociodemographic attributes. We learned that illness support (and reported adherence) might not translate into improved health outcomes, as indicated in this study of diabetics for a period of 2 years.

Despite the advantages of this study, there are several noteworthy limitations. First, a period of 2 years might not adequately capture the translation of regimen adherence into prevention of health decline. Second, we rely heavily on subjective measures of health, which—although commonly used in health and social research—measure only a limited aspect of health status. Furthermore, subjective health status—and self-reported health decline—could further be impacted by psychosocial factors. For example, research has suggested that chronically ill older adults who rate their health more optimistically will have relatively greater perceived control over their illnesses (Hong, Oddone, Dudley, & Bosworth, 2005). As a result, we reran our analyses with variables measuring aspects of disease outcome optimism and disease self-efficacy. Although there were minor changes in the coefficients we reported, our key findings remained consistent.

In addition, our analysis of health decline as a dependent variable relies not only on an appropriate measurement of health status but also on the appropriate measure of change. By assessing decline, we introduce some error into our analysis in the form of a “floor effect,” whereby those reporting poor health at baseline cannot report relatively worse health. However, given that the number of respondents reporting poor health at baseline is very small (less than 3%), it is unlikely that this imposes biases on our overall findings. We suggest that future research should examine the complex pathways of health status and functional limitations (maintenance, decline, and improvement) as health trajectories, with multiple periods of observation and over an extended period of time.

We were unable to examine additional measures of social support found to be significant in previous health research (social networks and ties, community involvement and participation), as well as other important predictors in the HDM (such as health beliefs and knowledge). Furthermore, the availability of self-reported adherence measures (support for diabetes activities, regimen adherence variables) only with the 2003 diabetes supplement poses limitations on analyzing the full relationship over time. In addition, the analysis was restricted to individuals reporting a full diabetic regimen (consisting of the six regimen components). Missing data on the outcome variable, health and diabetes variables, social variables, and demographic variables were negligible and apparently missing at random (or addressed through weighting). An imputation was conducted on item missing data and did not significantly change the outcomes, so the original data (with sampling weights from the 2003 mail-out study) were kept. Sampling weights adjusted for nonresponse (including participant mortality). Finally, all measures used in this analysis were based on self-report, which, as mentioned previously, might impose some systematic bias on the results of this study. However, given the relative stability of coefficients and standard errors across different models as well as previous studies on the validity and reliability of psychometric measures used in this analysis, it is unlikely these findings are heavily impacted by a large degree of measurement error (see Alwin, 2007).


The relationship between illness support, adherence, and health is nuanced and multifaceted. Diabetic support appears to be associated as protective for some regimen components but a susceptibility factor for others. Diabetic support is, however, highly associated with adherence for each regimen attribute, controlling for other factors. Future research should examine the strength of these relationships in different subgroups of the population, particularly by race/ethnicity, by social class, and by sex. Finally, as this study seeks to understand the role of illness support within the context of health and illness, much of the texture of these disease trajectories are lost. Subsequent analyses—qualitative and quantitative—are necessary to better understand how support influences health trajectories and disease pathways for different groups. Such research can shed light on potential protective factors of social support for individuals and communities.


J.L.’s effort was supported by the National Institute on Aging via Grants R01-AG154124 and R01-AG028116 and the Michigan Claude D. Pepper Older Americans Independence Center (P60-AG08808).


The authors are grateful to the Agency for Healthcare Research and Quality and to the John A. Hartford Foundation for supporting the research of the first author. In addition, the authors thank Steven Heeringa, Jason Owen-Smith, and Caroline Blaum for their helpful comments. Finally, they express their gratitude to the editor and to the three anonymous reviewers for their thoughtful feedback and helpful suggestions. E.J.N. planned the study, conducted data analysis, and wrote the article. J.L. contributed to the revision of the article.


  • Ajrouch KJ, Antonucci TC, Janevic MR. Social networks among blacks and whites: The interaction between race and age. Journal of Gerontology: Psychological Science and Social Sciences. 2001;56:S112–S118. [PubMed]
  • Allen SM, Ciambrone D. Community care for people with disability: Blurring boundaries between formal and informal caregivers. Qualitative Health Research. 2003;13:207–226. [PubMed]
  • Alwin DF. Margins of error: A study of reliability in survey measurement. New York: Wiley; 2007.
  • American Diabetes Association. Economic consequences of diabetes mellitus in the U.S. in 1997. Diabetes Care. 1998;21:296–309. [PubMed]
  • Bae S, Hashimoto H, Karlson EW, Liang MH, Daltroy LH. Variable effects of social support by race, economic status, and disease activity in systemic lupus erythematosus. Journal of Rheumatology. 2001;28:1413–1422. [PubMed]
  • Belgrave FZ, Lewis DM. The role of social support in compliance and other health behaviors for African Americans with chronic illnesses. Journal of Health and Social Policy. 1994;5:55–68. [PubMed]
  • Berkman LF, Syme SL. Social networks, host resistance, and mortality: A nine-year follow-up of Alameda County residents. American Journal of Epidemiology. 1979;109:186–204. [PubMed]
  • Blaum CS, Ofstedal MB, Langa KM, Wray LA. Functional status and health outcomes in older Americans with diabetes mellitus. Journal of the American Geriatrics Society. 2003;51:735–897. [PubMed]
  • Blazer D. Social support and mortality in an elderly community population. American Journal of Epidemiology. 1982;115:684–694. [PubMed]
  • Cohen S, Gottlieb BH. Social support measurement and intervention: A guide for health and social scientists. New York: Oxford University Press; 2000.
  • Connell CM, Fisher EB, Houston CA. Relationships among social support, diabetes outcomes, and morale for older men and women. Journal of Aging and Health. 1992;4:77–100.
  • Connell CM, Sorandt M, Lichty W. Impact of health belief and diabetes-specific psychosocial context variables on self-care behavior, metabolic control, and depression of older adults with diabetes. Behavior Health and Aging. 1990;1:183–196.
  • Coyne JC, DeLongis A. Going beyond social support: The role of social relationships in adaptation. Journal of Consulting and Clinical Psychology. 1986;54:454–460. [PubMed]
  • Durkheim E. Suicide. New York: Free Press; 1951. (Original work published 1897)
  • Eraker SA, Becker MH, Strecher VJ, Kirscht JP. Smoking behavior, cessation techniques, and the Health Decision Model. The American Journal of Medicine. 1985;78:817–825. [PubMed]
  • Eraker SA, Kirscht JP, Becker MH. Understanding and improving patient compliance. Annals of Internal Medicine. 1984;100:258–68. [PubMed]
  • Everard KM, Lach HW, Fisher EB, Baum MC. Relationship of activity and social support to the functional health of older adults. Journal of Gerontology: Psychological Science and Social Sciences. 2000;55:S208–S212. [PubMed]
  • Fitzgerald JT, Anderson RM, Funnell MM, Arnold MS, Davis WK, Amam LC, Jacober SG, Grunberger G. Differences in the impact of dietary restrictions on African Americans and Caucasians with NIDDM. Diabetes Educator. 1997;23:41–47. [PubMed]
  • Gallant MP. The influence of social support on chronic illness self-management: A review and directions for research. Health Education and Behavior. 2003;30:170–195. [PubMed]
  • Gore S. Social networks and social supports in health care. In: Freeman HE, Levine S, editors. Handbook of medical sociology. 4th ed. Englewood Cliffs, NJ: Prentice Hall; 1989. pp. 306–331.
  • Greenfield S, Sullivan L, Dukes KA, Silliman R, D’Agostino R, Kaplan S. Development and testing of a new measure case mix for use in office practice. Medical Care. 1995;33(Suppl.):AS47–AS55. [PubMed]
  • Health and Retirement Study, 2003 Diabetes Study. Produced and distributed by the University of Michigan with funding from the National Institute on Aging. Ann Arbor, MI: 2006. (grant number NIA UG01AG009740)
  • 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. 2003;26:738–745. [PubMed]
  • Hill-Briggs F. Problem solving in diabetes self-management: A model of chronic illness self-management behavior. Annals of Behavioral Medicine. 2003;25:182–193. [PubMed]
  • Hong TB, Oddone EZ, Dudley TK, Bosworth HB. Subjective and objective evaluations of health among middle-aged and older veterans. Journal of Aging and Health. 2005;17:592–608. [PubMed]
  • House JS, Landis K, Umberson D. Social relationships and health. Science. 1998;241:540–45. [PubMed]
  • House J, Robbins C, Metzner H. The association of social relationships and activities with mortality: Prospective evidence from the Tecumseh Community Health Study. American Journal of Epidemiology. 1982;116:123–140. [PubMed]
  • Idler EL, Benyamini Y. Self-rated health and mortality: A review of twenty-seven community studies. Journal of Health and Social Behavior. 1997;38:21–37. [PubMed]
  • Kaplan G, Salonen J, Cohen R, Brand R, Syme S, Puska P. Social connections and mortality from all causes and cardiovascular disease: Prospective evidence from Eastern Finland. American Journal of Epidemiology. 1988;128:370–380. [PubMed]
  • Kaplan H. Social psychology of disease. In: Freeman HE, Levine S, editors. Handbook of medical sociology. 4th ed. Englewood Cliffs, NJ: Prentice Hall; 1989. pp. 53–70.
  • Kaplan H, Cassel JC, Gore S. Social support and health. Medical Care. 1977;15:47–58. [PubMed]
  • Kaye SA, Folsom AR, Sprafka JM, Prineas RJ, Wallace RB. Increased incidence of diabetes mellitus in relation to abdominal adiposity in older women. Journal of Clinical Epidemiology. 1991;44:329–334. [PubMed]
  • Kohn ML, Clausen JA. Social isolation and schizophrenia. American Sociological Review. 1955;20:265–273.
  • Krause N, Borawski-Clark E. Clarifying the functions of social support in later life. Research on Aging. 1994;16:251–279.
  • Kressin NR, Clark JA, Whittle J, East M, Peterson ED, Chang B, Rosen AK, Ren X, Alley LG, Kroupa L, et al. Racial differences in health-related beliefs, attitudes, and experiences of VA cardiac patients: Scale development and application. Medical Care. 2002;41(Suppl.):I72–I85. [PubMed]
  • Langa KM, Vijan S, Hayward RA, Chernew ME, Blaum CS, Kabeto MU, Weir DR, Katz SJ, Willis RJ, Fendrick AM. Informal caregiving for diabetes and diabetic complications among elderly Americans. Journal of Gerontology: Psychological Science and Social Sciences. 2002;57:S177–S186. [PubMed]
  • Midthjell K, Holmen J, Bjorrndal A, Lund-Larsen F. Is questionnaire information valid in the study of chronic disease such as diabetes? The Nord-Trondelt Diabetes Study. Journal of Epidemiology and Community Health. 1992;46:537–542. [PMC free article] [PubMed]
  • Peyrot M, McMurry JF, Hedges R. Living with diabetes: The role of personal and professional knowledge in symptom and regimen management. Research in the Sociology of Health Care. 1987;6:107–146.
  • Porter EJ, Ganong LH, Drew N, Lanes TI. A new typology of home-care helpers. The Gerontologist. 2004;44:750–759. [PubMed]
  • Ruggiero LA, Spirito A, Bond A, Constan D, McGarvey S. Impact of social support and stress on compliance in women with gestational diabetes. Diabetes Care. 1990;13:441–443. [PubMed]
  • Sarason BR, Sarason IG, Gurung RA. Close personal relationships and health outcomes: A key to the role of social support. In: Sarason BR, Duck S, editors. Personal relationships: Implications for clinical and community psychology. New York: Wiley; 2001. pp. 15–41.
  • Schoenbach V, Kaplan B, Freedman L, Kleinbaum D. Social ties and mortality in Evans County, Georgia. American Journal of Epidemiology. 1986;123:577–591. [PubMed]
  • Schulz AJ, Israel BA, Zenk SN, Parker EA, Lichtenstein R, Shellman-Weir S, Klem L. Psychosocial stress and social support as mediators of relationships between income, length of residence and depressive symptoms among African American women on Detroit’s eastside. Social Science & Medicine. 2006;62:510–522. [PubMed]
  • Seeman TE, Kaplan GA, Knudsen L, Cohen R, Guralnik J. Social network ties and mortality among the elderly in the Alameda County Study. American Journal of Epidemiology. 1987;126:714–723. [PubMed]
  • Sherbourne CD, Hays RD, Ordway L, DiMatteo MR, Kravitz RL. Antecedents of adherence to medical recommendations: Results from the medical outcomes study. Journal of Behavioral Medicine. 1992;15:447–468. [PubMed]
  • Thoits PA. Stress, coping, and social support processes: Where are we? What next? Journal of Health and Social Behavior. 1995;35:53–79. [PubMed]
  • Turner RJ, Marino F. Social support and social structure: A descriptive epidemiology. Journal of Health and Social Behavior. 1994;35:193–212. [PubMed]
  • Uchino BN. Social support and physical health outcomes: Understanding the health consequences of our relationships. New Haven, CT: Yale University Press; 2004.
  • Wills TA, Filer M. Social networks and social support. In: Baum A, Revenson TA, Singer JE, editors. Handbook of health psychology. Mahwah, NH: Erlbaum; 2001. pp. 209–234.
  • Wilson W, Ary DV, Biglan A, Glasgow RE, Toobert DJ, Campbell DR. Psychosocial predictors of self-care behaviors and glycemic control in NIDDM. Diabetes Care. 1986;9:614–622. [PubMed]
  • Wray LA, Alwin DF, McCammon RJ, Manning T, Best LE. Social status, risky health behaviors, and diabetes in middle-aged and older adults. Journal of Gerontology: Psychological Science and Social Sciences. 2006;61:S290–S298. [PubMed]

Articles from The Journals of Gerontology Series B: Psychological Sciences and Social Sciences are provided here courtesy of Oxford University Press
PubReader format: click here to try


Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...


  • MedGen
    Related information in MedGen
  • PubMed
    PubMed citations for these articles

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...