Impact of Diabetes and Disease Duration on Work Status Among U.S. Older Adults

Abstract

Objectives: To examine the effects of diabetes and disease duration on work status over a 9-year period. Method: Multinomial logistic regression models examined the probability of retirement and disability impeding work, using data from the Health and Retirement Study (n = 5,576). Results: Among participants who had retired in 2012, almost 14% had incident diabetes (4.91 mean years with diabetes, 95% confidence interval [CI] = [4.67, 5.15]). Approximately 22% of participants who reported a disability impeded labor force participation had prevalent diabetes (17.1 mean years with diabetes, 95%CI = [16.41, 17.71]). Only prevalent diabetes that indicated longer disease duration was associated with disability (relative risk ratio [RRR] = 1.83, 95% CI = [1.30, 2.57]). There was evidence of effect modification among Hispanics only (p = .02). Discussion: Diabetes increased risk of exiting the workforce due to disability, and mean disease duration was associated with changes. Disease management and workplace interventions may enable older adults to continue being productive should they choose to remain in the workforce.

Keywords

diabetes aging workforce retirement disability

Introduction

Type 2 diabetes mellitus (referred to from this point forward as diabetes) and its complications are a major medical, psychosocial, and economic burden on societies. Diabetes is most often diagnosed in middle-aged adults, who have a substantial number of work years ahead of them (Herquelot, Guéguen, Bonenfant, & Dray-Spira, 2011). It is a leading cause of morbidity and mortality in the United States, and an enormous burden on the health care system, costing the nation over US$245 billion annually (American Diabetes Association, 2014). It is estimated that out of a total of US$245 billion in health care costs associated with diabetes and its complications, approximately US$69 billion can be attributed to lost productivity (American Diabetes Association, 2014).

Diabetes is posited to negatively impact work and economic productivity, due to work absences, disability leave, and premature departure from the workforce (Minor, 2011; Shultz & Wang, 2007; Stewart, Ricci, Chee, Hirsch, & Brandenburg, 2007; van den Berg, Elders, & Burdorf, 2010; Vijan, Hayward, & Langa, 2004). Work productivity is reportedly lower among persons with diabetes (Herquelot et al., 2011; Vijan et al., 2004). Adults with diabetes have a higher probability of retirement relative to diabetes-free adults, and the rate of self-reported disability and nonparticipation in the workforce due to health impairment among those with diabetes is twofold that of those without the disease. Furthermore, studies have also indicated a significantly increased likelihood of premature death among workers with diabetes (Herquelot et al., 2011; Vijan et al., 2004).

There are several plausible pathways through which diabetes affects workforce participation. For workers with diabetes, ability and desire to work may be influenced not only by disease symptoms and complications, but also by the side effects of treatments and medications. In persons with diabetes, hypoglycemia may affect the workforce and labor market through increased absenteeism, impairment of work, and decreased productivity (Zhang et al., 2010). These factors are associated with work discrimination (Tunceli et al., 2005), which is postulated to drive workers out of the workforce prematurely (McMahon, West, Mansouri, & Belongia, 2005). Furthermore, microvascular and macrovascular complications such as vision problems or foot disorders increase risk of disability, which may impede ability to work (Fu, Qiu, Radican, & Wells, 2009; Von Korff et al., 2005). These complications are often seen in individuals with long disease duration.

Diabetes self-management requires multiple daily activities for glycemic control, which include but may not be limited to blood glucose monitoring and insulin administration (Ruston, Smith, & Fernando, 2013; Weijman et al., 2005). These activities may be difficult to carry out in the work environment, resulting in substandard self-care, which may lead to decreased productivity and absenteeism, and diabetes complications and disability in the long run, all of which may encourage early retirement. It has been reported that newly diagnosed individuals are often overwhelmed by the diagnosis, and may lack the confidence to effectively self-manage their diabetes (Powers et al., 2016; Skovlund & Peyrot, 2005). Individuals who have had diabetes for longer periods of time may therefore have better self-management skills, and may feel more empowered to manage their disease in comparison with those who are newly diagnosed. These factors make duration with disease an important factor to consider in studies assessing effects of diabetes on work status over time.

Two previous peer-reviewed studies have used longitudinal data from the Health and Retirement Study (HRS) to examine the relationship between diabetes and workforce participation. One study focused on the contribution of prevalent diabetes to incidence of disability, retirement, and mortality among workers, as well as the economic impact and lost productivity (Vijan et al., 2004). Vijan and colleagues reported that diabetes had detrimental economic effects due to future lost income and lost productivity, attributable to incidence of early retirement, increased sick days, disability, and mortality. A second study estimated the effect of prevalent diabetes at baseline (1992) on the probability of employment, change in hours worked, work-loss days, and work limitations among those employed at 2-year follow-up (1994), and reported that diabetes was negatively associated with workforce participation (Tunceli et al., 2005).

Our study addresses several gaps in the literature. While these previous studies focused on prevalent diabetes only, our study further contributes to the literature through examination of how both diabetes incidence and prevalence might differentially impact work status. This allows us to postulate whether complications in prevalent cases or aspects of disease management and self-efficacy in newly diagnosed actually contribute to exiting the workforce. In addition, prior studies did not address how duration with diabetes might influence work status over time. Our study attempts to address diabetes duration, which has been found to significantly affect diabetes education interventions, glycemic control, and onset of disability and other complications (Ko et al., 2012; Rogus, Warram, & Krolewski, 2002), factors which are likely to impact ability to work. Furthermore, previous research did not take measures to account for mortality during the study period. Our study tries to address potential bias due to mortality during the study period.

With the aging of the population, older workers are making up an increasing and vital proportion of the labor force (Schofield et al., 2013; Toossi, 2015). U.S. Bureau of Labor Statistics (BLS) data indicated a labor force participation rate of 40% among Americans aged 55 years and older in 2014, with projected increases in this rate among the 65 to 74 and 75+ age groups (Toossi, 2015). While the studies by Tunceli et al. and Vijan at al. both included adults below the traditional age of retirement, we take into account all adults above the age of 50 years who were working at baseline. By addressing these factors, we believe we are able to provide a more comprehensive picture of the impact diabetes can have on work status.

Given the gaps in literature, the objective of this study was therefore to examine the differential impact of incident and prevalent diabetes on work status over five waves of data spanning an approximately 9-year period. Due to the chronic and silent nature of diabetes, a prolonged follow-up enabled us to capture individuals who may go undiagnosed for years, or those for whom the manifestations of disease progression may take years to develop. We were interested in observing whether participants who were working in 2004, by 2012, had (a) remained in the workforce, (b) retired, or (c) reported they were no longer in the workforce due to a disability impeding their ability to work. To better understand how disease duration may influence workforce participation prospectively, we compared participants who remained free of diabetes for the study duration to those with prevalent diabetes at baseline in 2004 (17.1 mean years with diabetes), and those with incident diabetes self-reported between 2006 and 2012 (4.9 mean years with diabetes). We hypothesized that among adults aged 50 years and older, relative to participants without diabetes those with diabetes would have an increased likelihood of reporting disability or retirement by 2012, with evident differences between those with prevalent (i.e., longer mean duration with disease) and incident diabetes.

Subjects, Materials, and Methods

Data Source

HRS data from 2004 to 2012 were used for the current analysis. The HRS is a longitudinal cohort study of health and retirement among older American adults who are 50 years and older. HRS interview data are collected every 2 years, on a vast number of variables, including health conditions, demographics, socioeconomic status (SES), occupational factors, and behavioral indicators. Sampling procedures and study design for the HRS are available elsewhere in greater detail (Juster & Suzman, 1995; Sonnega et al., 2014). The HRS was approved by the Institutional Review Board at the University of Michigan, and all participants provided informed consent. HRS data are de-identified and publicly available; therefore, review and approval was not required from the Internal Review Board at the University of Texas Medical Branch.

The total HRS sample size in 2004 was 20,129. Response rates above 86% were reported in the 2004 and subsequent study waves (Sonnega et al., 2014). Relative to participants who remained in the study, those lost to follow up were more likely to be in poorer health, of lower SES (according to education proxy), employed and less likely to be retired (Michaud, Kapteyn, Smith, & Soest, 2011).

Inclusion and Exclusion Criteria

The interview question “Are you working now, temporarily laid off, unemployed and looking for work, disabled and unable to work, retired, a homemaker, or what?” was used to ascertain inclusion into the study at the 2004 Wave. Participants aged 50 years and older at baseline who reported “working now” were eligible for inclusion into the study (n = 6,333). In addition, participants had to have data on work status in 2012, unless censored at an earlier study wave due to mortality (n = 5,648). This included participants at or over the traditional retirement age (65 years) who accounted for over 24% of working adults at baseline. After exclusion of participants with missing data on the variables of interest (n = 72), the final analytic sample consisted of 5,576 participants with complete data.

Variables of Interest

The outcome of interest was work status in 2012. In each wave, participants were asked “Are you working now, temporarily laid off, unemployed and looking for work, disabled and unable to work, retired, a homemaker, or what?” Using responses from this question, a nominal variable with three discrete categories was created: (a) working, (b) retired, and (c) disabled and not-working. The BLS defines labor force participation as “the number of people working or looking for work” (Toossi, 2015). Participants in the “working” category therefore included those who were temporarily laid off, unemployed, but looking for work, and on sick leave (n = 160, 3.14%) as they indicated a potential return to the workforce. Those who were retired, had left the labor force to become homemakers since the 2004 wave, or reported “other” reasons for not working or returning to work were classified as retired. Disability impeding work was based on self-report and not on standard measures of activities of daily living (ADLs) or instrumental activities of daily living (IADLs). Participants were classified as disabled if they specifically reported that they were “disabled and unable to work.”

Self-reported diabetes diagnosis was the primary predictor of interest, which was ascertained by a confirmatory response to the question “Has a doctor ever told you that you have diabetes or high blood sugar?” For the second predictor of interest, participants were categorized into three diabetes status groups that indicated prevalence/incidence and duration with the disease: (a) no diabetes diagnosis during the study period, (b) self-reported prevalent diabetes at baseline (17.1 mean years with diabetes), and (c) self-reported incident diabetes between the 2006 and 2012 waves (4.91 mean years with diabetes).

To take into account possible mortality bias, participants who had died prior to the 2012 wave were censored at the last assessment prior to their death (Botoseneanu, Allore, Gahbauer, & Gill, 2013). For example, if a participant died in 2007, their work status from 2006 was used. The month and year of the interview in 2004 was subtracted from the reported month and year of death to create a continuous variable measuring the length of survival. Participants who were living in 2012 had their length of survival set by subtracting the interview date in the 2004 wave from the 2012 wave interview date.

Additional baseline covariates chosen for their association with diabetes and work status were race/ethnicity (Hispanic, non-Hispanic Black, non-Hispanic White), age, years of education (<12, 12, 12-15, ≥16), marital status (married [including cohabiting], not married[single, divorced, separated, widowed]), average hours worked per week which also indicated full- or part-time work (<35, 35-45, >45), occupational grouping (white-collar, blue-collar, service industry work), missed work for health reasons (yes/no), tenure in current type of work (≤10 years, >10 years), chronic conditions (none, one, two, or more [of lung disease, heart disease, hypertension, and cancer]), health insurance coverage (fully covered doctors’ visits, partially covered, not covered), current smoker (yes/no), physical activity (yes/no), and body mass index (BMI; Franks, 2012). Physical activity was defined as engaging in moderate or vigorous exercise two or more times a week. BMI was based on self-reported height and weight, and was treated as an ordinal variable according to the World Health Organization (WHO, 2012) classification. Due to a small number of participants with BMI below 18.5 kg/m² (<1.0%), the BMI categories underweight and normal weight were combined. The BMI variable therefore had three ordinal categories: underweight/normal (<24.9 kg/m²), overweight (25.0-29.9 kg/m²), and obese (>30.0 kg/m²). We chose not to include years since diagnosis as a measure of diabetes duration to avoid overcontrolling in the statistical models, given the differences in mean years with diabetes for participants with prevalent (17.1 years, 95% confidence interval [CI] = [16.41, 17.71]) and incident (4.9 years, 95% CI = [4.67, 5.15]) diabetes.

Analytic Strategy

Descriptive analysis of the sample was conducted using proportions for the three work status categories. Bivariate analysis between diabetes status and work variables was also conducted. The probability of retirement or disability relative to working in 2012 was estimated using multinomial logistic models, for which relative risk ratios (RRR) and 95% CIs are reported. Two models were estimated: Model 1 controlled for sociodemographic variables only, which included age, length of survival, sex, race/ethnicity, marital status, and education. Model 2 was fully adjusted for all covariates mentioned above.

To examine potential effect modification in the fully adjusted analyses, interactions between diabetes status, and other covariates (age [dichotomized at the age of retirement <65, ≥65], education, race/ethnicity, occupation group, work hours, chronic conditions, and BMI) were examined. Effect modification by gender was not examined, as gender did not show a significant association with work status. The interaction term Diabetes × Race/Ethnicity was significant for the retirement outcome only among Hispanics (p = .02). To explore this further, we conducted a logistic regression analysis for the Hispanic subpopulation, in which the odds of retiring were assessed relative to the odds of remaining in the workforce. All other tested interactions were not significant and are therefore not presented in the “Results” section. All analyses were conducted using Stata/SE 12.0 (Stata, College Station, TX).

Results

Table 1 presents the participant baseline characteristics by work status at the 9-year follow-up. The average age at baseline of participants who were working, retired, and disabled in 2012 was 57.7, 62.4, and 57.4 years, respectively. There were no gender differences in the work status rates. Married participants reported higher rates of retirement (45.9%) than nonmarried participants (33.6%). Retirement was also high among participants with 12 (50.1%) or fewer (49.6%) years of education, compared with those with 16 or more years of education (40.2%); and among those with two or more chronic conditions (53.7%) compared with those with none (39.2%). Working less than 35 hr (54.7%), having 10 years or greater of work tenure (48.2%), and working in the blue-collar (47.8%) and service (48.4%) industry at baseline were also associated with higher rates of retirement in 2012 (Table 1).

Table 1.

Baseline (2004) Characteristics by Work Status in 2012 (N = 5,576).

Baseline characteristics	Work status in 2012						p value
	Working		Retired		Disabled
	n	%	n	%	n	%
Baseline age, M (SE)	57.7 (0.13)		62.4 (0.14)		57.4 (0.24)
Length of survival, M (SE)	8.5 (0.3)		8.7 (0.02)		8.3 (0.9)
Sex							.41
Men	1,378	50.65	1,199	44.33	136	5.03
Women	1,411	49.15	1,299	45.25	161	5.61
Race/ethnicity							<.01
Hispanic	249	50.1	218	43.9	30	6.0
Non-Hispanic Black	344	45.4	349	46.1	64	8.5
Non-Hispanic White	2,103	50.5	1,872	44.9	193	4.63
Other	85	55.2	59	38.3	10	6.5
Marital status							<.01
Married	1,941	49.4	1,803	45.9	182	4.6
Not married	840	50.91	695	42.12	115	7.0
BMI							<.01
Normal/under	832	52.4	678	42.7	78	4.9
Overweight	1,101	49.4	1,025	46.0	102	4.6
Obese	848	48.2	795	45.2	117	6.7
Education							<.01
<12 years	319	40.1	398	50.1	78	9.8
12 years	781	45.2	857	49.6	91	5.3
13-15 years	732	52.8	574	41.4	80	5.8
≥16 years	949	57.0	669	40.2	48	2.8
Physical activity							<.01
Yes	1,716	52.4	1,429	43.7	127	3.9
No	1,065	46.2	1,069	46.4	170	7.4
Current smoker							<.01
No	2,301	49.9	2,104	45.6	209	4.5
Yes	480	49.9	394	41.0	88	9.2
Chronic conditions^a							.01
No chronic conditions	1,457	57.5	993	39.2	86	3.4
One chronic condition	1,014	45.6	1,067	48.0	143	6.4
Two or more chronic conditions	310	38.0	438	53.7	68	8.3
Insurance coverage							<.01
Fully covered	688	46.4	704	47.5	90	6.1
Mostly/partially	1,885	50.8	1,651	44.5	176	4.7
Not covered	208	54.5	143	37.4	31	8.1
Work hours per week							<.01
<35	614	39.8	848	54.9	82	5.3
35-45	1,404	51.0	1,202	43.6	149	5.4
≥46	763	59.8	448	35.1	66	5.2
Missed work—Health reasons							<.01
Yes	971	46.2	971	46.3	156	7.4
No	1,810	52.04	1,527	43.9	141	4.1
Tenure in type of work							<.01
≤10 years	1,569	51.7	1,274	41.9	195	6.6
>10 years	1,212	47.8	1,224	48.2	102	4.0
Occupational category							<.01
White collar	1,829	53.2	1,470	42.8	137	4.0
Service industries	382	44.2	418	48.4	64	7.4
Blue collar	570	44.7	610	47.8	96	7.5

Source. Health and Retirement Study 2004-2012.

Note. n = 529 died before 2012. BMI = body mass index.

Among the following: lung disease, heart disease, hypertension, and cancer.

Approximately 9% non-Hispanic Black, 4.6% non-Hispanic White, and 6% Hispanic participants reported disability impeding ability to work. Participants with less than 12 years of education had a higher rate of disability (9.8%) than those with 16 or more years of education (2.8%). Higher rates of disability were also evident among participants reporting two or more chronic conditions (8%), smokers (9.2%), obese (6.7%), and physically inactive (7.4%) participants. Approximately 7.5% blue-collar versus 4.0% white-collar participants reported disability impeding work.

Table 2 shows the bivariate analysis of diabetes by work status and other occupational variables. Among participants who had retired by 2012, almost 13.7% had prevalent diabetes at baseline, and 13.9% had incident diabetes. Of those who reported that a disability impeded labor force participation in 2012, 22.2% had prevalent diabetes, while 14.1% self-reported diabetes after the baseline. More than 13% of participants working 45 hr a week or less at baseline reported diabetes, compared with 9.4% of those working more than 45 hr each week. Compared with blue-collar and white-collar workers, more service industry workers reported diabetes at baseline.

Table 2.

Bivariate Analysis of Diabetes Status and Occupational Variables of Interest.

	No diabetes		Prevalent diabetes		Incident diabetes		χ²	p value
	n	%	n	%	n	%	χ²	p value
2012 work status							51.02	<.01
Working	2,170	78.0	297	10.7	314	11.3
Retired	1,807	72.3	343	13.7	348	13.9
Disabled	189	63.6	66	22.2	42	14.1
Baseline characteristics:
Work hours per week							19.35	<.01
<35	1,162	75.3	208	13.5	174	11.3
35-45	2,015	73.1	378	13.7	362	13.1
≥46	989	77.5	120	9.4	168	13.1
Missed work—Health reasons							26.85	<.01
Yes	1,490	71.0	319	15.2	289	13.8
No	2,676	74.7	387	11.1	415	11.9
Tenure in type of work							0.13	.93
≤10 years	2,265	74.6	389	12.8	384	12.6
>10 years	1,901	74.9	317	12.5	320	12.6
Occupational category							16.51	<.01
White collar	2,627	76.5	397	11.6	412	12.0
Service industries	620	71.8	132	15.3	112	13.0
Blue collar	919	72.0	177	13.9	180	14.1

Table 3 shows the results of the partially and fully adjusted models for the association between self-reported diabetes diagnosis and 2012 work status. In both models, working (relative to retirement or disability) was the referent category. In Model 1, compared with being diabetes free over the study period, both prevalent (RRR = 1.30, 95% CI = [1.07, 1.55]) and incident (RRR = 1.31, 95% CI = [1.10, 1.56]) diabetes were associated with retirement in 2012. This association was attenuated for prevalent diabetes (RRR = 1.15, 95% CI = [0.95, 1.39]), and became nonsignificant after inclusion of health, behavioral, and occupational factors in Model 2. Participants with incident diabetes continued to show a significant association (RRR = 1.22, 95% CI = [1.02, 1.47]) with retirement after full adjustment of all relevant covariates. A significant association between prevalent diabetes and disability in 2012 was observed in Model 1 (RRR = 2.32, 95% CI = [1.69, 3.18]). There was no significant association between incident diabetes and disability. In Model 2, the association between disability and prevalent diabetes was decreased but remained significant (RRR = 1.83, 95% CI = [1.30, 2.57]).

Table 3.

Multinomial Logistic Regression Results for the Association Between Diabetes and Work Status in 2012.

	Retired (ref: working)						Disabled (ref: working)
	Model 1^a			Model 2^b			Model 1^a			Model 2^b
	RRR	95% CI	p ^b	RRR	95% CI	p	RRR	95% CI	p	RRR	95% CI	p
Diabetes
Prevalent diabetes at baseline	1.30	[1.07, 1.55]	<.01	1.15	[0.95, 1.39]	.15	2.32	[1.69, 3.18]	<.01	1.83	[1.30, 2.57]	<.01
Incident diabetes diagnosed after baseline	1.31	[1.10, 1.56]	<.01	1.22	[1.02, 1.47]	.03	1.48	[1.03, 2.11]	.03	1.27	[0.87, 1.86]	.21

Note. RRR refers to the relative risk ratio and the base outcome for these ratios is working. Boldface indicates statistical significance. CI = confidence interval.

Model 1 controlled for age, length of survival, sex, race/ethnicity, marital status, and education.

Model 2 was fully adjusted for all covariates in Model 1 plus BMI, physical activity, smoking, insurance coverage, chronic conditions, work hours, missed work, occupation category, and tenure.

Inclusion of a Race/Ethnicity × Diabetes interaction into the multinomial model resulted in a significant effect for retirement, but only among Hispanics (p = .02). Logistic regression was therefore performed, which compared the odds of retiring with the odds of remaining in the workforce among Hispanics only (n = 467). Results indicated that prevalent (OR = 2.48, 95% CI = [1.31, 4.66]) and incident (OR = 2.14, 95% CI = [1.22, 3.74]) diabetes were strongly associated with retirement by 2012 after full adjustment for all relevant covariates (Table 4).

Table 4.

Logistic Regression Results for the Association Between Diabetes and Retirement in 2012 Among Hispanics (n = 467).

Diabetes	Retired (ref: working)^a
Diabetes	OR	95% CI	p
Prevalent diabetes at baseline	2.47	[1.31, 4.67]	<.01
Incident diabetes diagnosed after baseline	2.14	[1.22, 3.74]	.01

Note. OR = odds ratio; CI = confidence interval.

Model was fully adjusted for age, length of survival, sex, race/ethnicity, marital status, education, body mass index (BMI), physical activity, smoking, insurance coverage, chronic conditions, work hours, missed work, occupation category, and tenure.

Discussion

Previous studies (Cummings, 2007; Herquelot et al., 2011) including two that used HRS data (Tunceli et al., 2005; Vijan et al., 2004) reported that diabetes is associated with greater likelihood of exiting the workforce. Building upon these studies, we examined whether participants who were working in 2004 had remained in the workforce, retired, or had a disability that hindered participation in the workforce approximately 9 years later, using diabetes status as the primary predictor variable of interest, and taking into account disease duration.

Our findings supported our hypothesis, and were consistent with previous reports, supporting evidence of greater risk of retirement or disability impeding work in individuals with diabetes (Cummings, 2007; Herquelot et al., 2011; Tunceli et al., 2005; Vijan et al., 2004). Our findings further contribute to the literature by demonstrating how duration with diabetes in older working adults impacts decisions about workforce participation. Specifically, we found that relative to persons without diabetes, participants with prevalent diabetes at baseline (longer mean disease duration) were more likely to report that they were disabled and unable to work. The reasons for this finding are unclear; however, it is possible that participants with longer disease duration have better self-efficacy and disease management (Ruston et al., 2013; Weijman et al., 2005), which reduces the risk of diabetes-related absenteeism, impairment of work, and decreased productivity (Zhang et al., 2010), factors associated with retirement. Despite disease management and self-efficacy, however, prolonged duration with diabetes can lead to micro- and macro-vascular complications, which are known predictors of disability (Chiu & Wray, 2011; Wong et al., 2013). For participants with diabetes, these complications can develop many years after initial diagnosis (Fowler, 2011). In addition, diabetes prevalence has been associated with increased risk of comorbid chronic conditions within years of diagnosis (Huang et al., 2017), which may increase the risk of disability.

Furthermore, and somewhat surprising, participants with prevalent diabetes were more likely to remain in the workforce, while those with incident diabetes (shorter mean disease duration) were more likely to have retired by the 2012 wave. It is possible that some participants with incident diabetes may have had diabetes in 2004 that went undiagnosed and untreated for many years. Undiagnosed diabetes may progress, while remaining asymptomatic for several years, until complications arise (Beagley, Guariguata, Weil, & Motala, 2014; Serrano-Cinca, Fuertes-Callén, & Mar-Molinero, 2005), which may impact ability to work. Furthermore, blood glucose monitoring, insulin administration at work, and flexibility to take time for doctors’ visits may be challenging for persons with diabetes (Ruston et al., 2013; Weijman et al., 2005). For newly diagnosed individuals, these challenges and a lack of self-efficacy in disease self-management may lead to hypoglycemia and other complications (Ruston et al., 2013; Weijman et al., 2005), in addition to anxiety and stress due to decreased productivity. These factors may make working less desirable, particularly among adults nearing the age of retirement, who may opt to exit the workforce earlier than they would otherwise.

Our findings contribute to the literature on disparities in effects of diabetes on participation in the workforce. Race/ethnicity was an effect modifier as evidenced by the significant Hispanic × Diabetes interaction. While chance cannot be ruled out, our data suggest that relative to nondiabetic older adults, Hispanics with diabetes showed a higher propensity to retire. The reasons for this are unclear; however, research has indicated that when compared with non-Hispanic Whites, Hispanics have poorer glycemic control, and higher rates of diabetes associated complications such as amputations, nephropathy, and retinopathy (Bonds et al., 2003; Gonzalez-Zacarias, Mavarez-Martinez, Arias-Morales, Stoicea, & Rogers, 2016). In addition, Hispanics have less knowledge about diabetes self-management, inadequate access to diabetes prevention and control programs, and quality diabetes care (Bonds et al., 2003; Hausmann, Ren, & Sevick, 2010). Hispanics have high rates of uninsured persons. Lack of insurance coverage is associated with increased difficulty in obtaining diabetes testing supplies and medications, which can result in poor glycemic control and higher rates of hospitalizations (Fisher & Ma, 2015; Madden et al., 2011), factors that may affect ability to work, and increase the likelihood of exiting the workforce.

Limitations

There are several limitations to this study. The diabetes variable was based on self-report, which may affect results due to bias, or misinterpretation (Zandwijk et al., 2015) and makes it difficult to ascertain the prevalence of undiagnosed diabetes. Self-reported diagnosis of chronic conditions such as diabetes is affected by awareness and recall bias in older participants (Nam, Al Snih, & Markides, 2015). However, self-report of physician-diagnosed chronic conditions, including diabetes and medical records have been previously reported to be highly correlated (Simpson et al., 2004). Inability to account for undiagnosed diabetes may have resulted in conservative estimates of the relationship between diabetes and work status change. Factors such as flexibility, supportive work environment, and discrimination that may affect ability to effectively manage diabetes and motivation to remain in the workforce were not available in the HRS.

We readily acknowledge that the method we have used to account for mortality bias presents some challenges, in particular, if there is a long lag time between last employment observation and date of death. While a model that accounts for competing risks and complex disease pathways would have addressed this drawback (Joly, Commenges, Helmer, & Letenneur, 2002; Leffondré, Touraine, Helmer, & Joly, 2013), our objective of assessing both prevalent and incident cases restricted our ability to use a time to event model, which would focus on incident cases only. Furthermore, our study definition of disability did not include ADLs and IADLs, which have been found to greatly impact ability to work. We however wanted to focus on the data that were collected in the HRS employment questionnaire, and allow for comparability with previous studies that have used this data. Despite these shortcomings, this study has several strengths that lend weight to our conclusions, including the use of prospective data from a large cohort, with diverse sociodemographic and work factors.

Implication for Policy and Public Health

Diabetes is a highly complex disease to manage, and employers and supervisors reportedly have limited understanding of the implications of diabetes management, or how the work environment may impact an employee’s ability to successfully manage the disease (Ruston et al., 2013). Employers may therefore be unable to adequately provide support for their diabetic workforce to ensure appropriate self-management (Ruston et al., 2013). Workers who are unable to properly manage their diabetes at work place both their health and ability to continue working at great risk (Ruston et al., 2013). In turn, employers who lack knowledge or fail to support the needs of diabetic workers are likely to experience lost productivity due to absenteeism and from employees who prematurely exit the workforce.

Policies and interventions on diabetes management at work would be beneficial to both employees and employers, and would need to target both (Munir, Randall, Yarker, & Nielsen, 2009; Ruston et al., 2013). Diabetes self-management education and diabetes self-management support are viewed as important components of diabetes control by the American Diabetes Association, the American Association of Diabetes Educators, and the Academy of Nutrition and Dietetics (Powers et al., 2016) and should be incorporated into workplace wellness programs. Workplace interventions for chronic disease prevention and management, and programs to support and educate the workforce have been proven to be effective in reducing chronic disease progression and absenteeism, as well as hospital admissions, readmissions, and overall health care costs (Meng et al., 2017; Powers et al., 2016). Multilayer interventions would need to address education on self-management of diabetes, individual health behaviors, available personal and community resources, and the work social environment. Such interventions would need to provide information and support in different easily accessible formats, which can be understood by diverse populations including those for whom English is not a first language and those with low literacy levels or SES. Investing in workplace health programs that improve the health and quality of life of older adults from diverse backgrounds would enable them to continue being productive should they choose to remain in the workforce.

Conclusion

While the incidence and prevalence of diabetes and its risk factors continue to rise in the United States (Menke, Casagrande, Geiss, & Cowie, 2015), so too is the percentage of older adults who work for pay (Toossi, 2015). More research examining the effects of diabetes on work in this subpopulation is therefore warranted. Future research needs to take into account the work environment, undiagnosed and untreated diabetes, and disease self-management, as these are likely to impact an individual’s ability to remain healthy and continue working. Furthermore, race/ethnic disparities in diagnosis and disease management particularly in the work context, and the effects of diabetes on employment should continue to be explored. As the number of older adults in the workforce continues to increase, the rates of employees requiring support in diabetes management at work will also increase, and a focus on workplace health management strategies will become increasingly important.

Footnotes

Authors’ Note

Miriam Mutambudzi is now affiliated with University of Glasgow, Scotland, as a research associate.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Miriam Mutambudzi

References

American Diabetes Association. (2014). Statistics about diabetes. Retrieved from http://www.diabetes.org/diabetes-basics/statistics/#sthash

Beagley

Guariguata

Weil

Motala

A. A.

(2014). Global estimates of undiagnosed diabetes in adults. Diabetes Research and Clinical Practice, 103, 150-160. doi:10.1016/j.diabres.2013.11.001

Bonds

D. E.

Zaccaro

D. J.

Karter

A. J.

Selby

J. V.

Saad

Goff

D. C.

(2003). Ethnic and racial differences in diabetes care: The insulin resistance atherosclerosis study. Diabetes Care, 26, 1040-1046. doi:10.2337/DIACARE.26.4.1040

Botoseneanu

Allore

H. G.

Gahbauer

E. A.

Gill

T. M.

(2013). Long-term trajectories of lower extremity function in older adults: Estimating gender differences while accounting for potential mortality bias. Journals of Gerontology, Series A: Biological Sciences & Medical Sciences, 68, 861-868. doi:10.1093/gerona/gls228

Chiu

C.-J.

Wray

L. A.

(2011). Physical disability trajectories in older Americans with and without diabetes: The role of age, gender, race or ethnicity, and education. The Gerontologist, 51, 51-63. doi:10.1093/geront/gnq069

Cummings

(2007). Diabetes and employment prospects. Practical Diabetes International, 24(2), 67. doi:10.1002/pdi.1055

Fisher

M. A.

(2015). Medicaid-insured and uninsured were more likely to have diabetes emergency/urgent admissions. The American Journal of Managed Care, 21(5), e312-e319. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/26167779

Fowler

M. J.

(2011). Microvascular and macrovascular complications of diabetes. Clinical Diabetes, 29, 116-122. doi:10.2337/diaclin.29.3.116

Franks

P. W.

(2012). The complex interplay of genetic and lifestyle risk factors in type 2 diabetes: An overview. Scientifica, 2012, Article 482186. doi:10.6064/2012/482186

10.

A. Z.

Qiu

Radican

Wells

B. J.

(2009). Health care and productivity costs associated with diabetic patients with macrovascular comorbid conditions. Diabetes Care, 32, 2187-2192. doi:10.2337/dc09-1128

11.

Gonzalez-Zacarias

A. A.

Mavarez-Martinez

Arias-Morales

C. E.

Stoicea

Rogers

(2016). Impact of demographic, socioeconomic, and psychological factors on glycemic self-management in adults with type 2 diabetes mellitus. Frontiers in Public Health, 4, Article 195. doi:10.3389/fpubh.2016.00195

12.

Hausmann

L. R.

Ren

Sevick

M. A.

(2010). Racial differences in diabetes-related psychosocial factors and glycemic control in patients with type 2 diabetes. Patient Preference and Adherence, 4, 291-299. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/20859456

13.

Herquelot

Guéguen

Bonenfant

Dray-Spira

(2011). Impact of diabetes on work cessation: Data from the GAZEL cohort study. Diabetes Care, 34, 1344-1349. doi:10.2337/dc10-2225

14.

Huang

Refaat

Mohammedi

Jayyousi

Al Suwaidi

Abi Khalil

(2017). Macrovascular complications in patients with diabetes and prediabetes. Biomed Research International, 2017, 1-9. doi:10.1155/2017/7839101

15.

Joly

Commenges

Helmer

Letenneur

(2002). A penalized likelihood approach for an illness-death model with interval-censored data: Application to age-specific incidence of dementia. Biostatistics, 3, 433-443. doi:10.1093/biostatistics/3.3.433

16.

Juster

F. T.

Suzman

(1995). An overview of the health and retirement study. The Journal of Human Resources, 30, S7-S56. doi:10.2307/146277

17.

S. H.

Park

S. A.

Cho

J. H.

S. H.

Shin

K. M.

Lee

S. H.

. . . Ahn

Y. B.

(2012). Influence of the duration of diabetes on the outcome of a diabetes self-management education program. Diabetes & Metabolism Journal, 36, 222-229. doi:10.4093/dmj.2012.36.3.222

18.

Leffondré

Touraine

Helmer

Joly

(2013). Interval-censored time-to-event and competing risk with death: Is the illness-death model more accurate than the Cox model? International Journal of Epidemiology, 42, 1177-1186. doi:10.1093/ije/dyt126

19.

Madden

M. H.

Tomsik

Terchek

Navracruz

Reichsman

Clark

T. D.

. . . Werner

J. J.

(2011). Keys to successful diabetes self-management for uninsured patients: Social support, observational learning, and turning points: A safety net providers’ strategic alliance study. Journal of the National Medical Association, 103, 257-264. doi:10.1016/S0027-9684(15)30292-3

20.

McMahon

B. T.

West

S. L.

Mansouri

Belongia

(2005). Workplace discrimination and diabetes: The EEOC Americans With Disabilities Act research project. Work, 25(1), 9-18. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/16006671

21.

Meng

Wolff

M. B.

Mattick

K. A.

DeJoy

D. M.

Wilson

M. G.

Smith

M. L.

(2017). Strategies for worksite health interventions to employees with elevated risk of chronic diseases. Safety and Health at Work, 8, 117-129. doi:10.1016/j.shaw.2016.11.004

22.

Menke

Casagrande

Geiss

Cowie

C. C.

(2015). Prevalence of and trends in diabetes among adults in the United States, 1988-2012. Journal of the American Medical Association, 314, 1021-1029. doi:10.1001/jama.2015.10029

23.

Michaud

P.-C.

Kapteyn

Smith

J. P.

van Soest

(2011). Temporary and permanent unit non-response in follow-up interviews of the Health and Retirement Study. Longitudinal and Life Course Studies, 2, 145-169. doi:10.14301/llcs.v2i2.114

24.

Minor

(2011). The effect of diabetes on female labor force decisions: New evidence from the National Health Interview Survey. Health Economics, 20, 1468-1486. doi:10.1002/hec.1685

25.

Munir

Randall

Yarker

Nielsen

(2009). The influence of employer support on employee management of chronic health conditions at work. Journal of Occupational Rehabilitation, 19, 333-344. doi:10.1007/s10926-009-9199-7

26.

Nam

Al Snih

Markides

K. S.

(2015). Sex, nativity, and disability in older Mexican Americans. Journal of the American Geriatrics Society, 63, 2596-2600. doi:10.1111/jgs.13827

27.

Powers

M. A.

Bardsley

Cypress

Duker

Funnell

M. M.

Fischl

A. H.

. . . Vivian

(2016). Diabetes self-management education and support in type 2 diabetes: A joint position statement of the American Diabetes Association, the American Association of Diabetes Educators, and the Academy of Nutrition and Dietetics. Clinical Diabetes, 34, 70-80. doi:10.2337/diaclin.34.2.70

28.

Rogus

J. J.

Warram

J. H.

Krolewski

A. S.

(2002). Genetic studies of late diabetic complications: The overlooked importance of diabetes duration before complication onset. Diabetes, 51, 1655-1662. doi:10.2337/diabetes.51.6.1655

29.

Ruston

Smith

Fernando

(2013). Diabetes in the workplace—Diabetic’s perceptions and experiences of managing their disease at work: A qualitative study. BMC Public Health, 13(1), Article 386. doi:10.1186/1471-2458-13-386

30.

Schofield

D. J.

Callander

E. J.

Shrestha

R. N.

Passey

M. E.

Percival

Kelly

S. J.

(2013). Multiple chronic health conditions and their link with labour force participation and economic status. PLoS ONE, 8(11), e79108. doi:10.1371/journal.pone.0079108

31.

Serrano-Cinca

Fuertes-Callén

Mar-Molinero

(2005). Measuring DEA efficiency in Internet companies. Decision Support Systems, 38, 557-573. Retrieved from http://www.who.int/about/licensing/%5Cnhttp://apps.who.int/iris/bitstream/10665/204871/1/9789241565257_eng.pdf

32.

Shultz

K. S.

Wang

(2007). The influence of specific physical health conditions on retirement decisions. The International Journal of Aging & Human Development, 65, 149-161. doi:10.2190/AG.65.2.c

33.

Simpson

C. F.

Boyd

C. M.

Carlson

M. C.

Griswold

M. E.

Guralnik

J. M.

Fried

L. P.

(2004). Agreement between self-report of disease diagnoses and medical record validation in disabled older women: Factors that modify agreement. Journal of the American Geriatrics Society, 52, 123-127. doi:10.1111/j.1532-5415.2004.52021.x

34.

Skovlund

S. E.

Peyrot

(2005). The Diabetes Attitudes, Wishes, and Needs (DAWN) program: A new approach to improving outcomes of diabetes care. Diabetes Spectrum, 18, 136-142. doi:10.2337/diaspect.18.3.136

35.

Sonnega

Faul

J. D.

Ofstedal

M. B.

Langa

K. M.

Phillips

J. W. R.

Weir

D. R.

(2014). Cohort profile: The Health and Retirement Study (HRS). International Journal of Epidemiology, 43, 576-585. doi:10.1093/ije/dyu067

36.

Stewart

W. F.

Ricci

J. A.

Chee

Hirsch

A. G.

Brandenburg

N. A.

(2007). Lost productive time and costs due to diabetes and diabetic neuropathic pain in the US workforce. Journal of Occupational and Environmental Medicine, 49, 672-679. doi:10.1097/JOM.0b013e318065b83a

37.

Toossi

(2015). Labor force projections to 2024: The labor force is growing, but slowly. Monthly Labor Review. Retrieved from https://www-jstor-org-443.web.bisu.edu.cn/stable/monthlylaborrev.2015.12.004

38.

Tunceli

Bradley

C. J.

Nerenz

Williams

L. K.

Pladevall

Lafata

J. E.

(2005). The impact of diabetes on employment and work productivity. Diabetes Care, 28, 2662-2667. doi:10.2337/diacare.28.11.2662

39.

van den Berg

T. I. J.

Elders

L. A. M.

Burdorf

. (2010). Influence of health and work on early retirement. Journal of Occupational and Environmental Medicine, 52, 576-583. doi:10.1097/JOM.0b013e3181de8133

40.

Vijan

Hayward

R. A.

Langa

K. M.

(2004). The impact of diabetes on workforce participation: Results from a national household sample. Health Services Research, 39(6p1), 1653-1670. doi:10.1111/j.1475-6773.2004.00311.x

41.

Von Korff

Katon

Lin

E. H. B.

Simon

Ciechanowski

Ludman

Young

. (2005). Work disability among individuals with diabetes. Diabetes Care, 28, 1326-1332. doi:10.2337/diacare.28.6.1326

42.

Weijman

Ros

W. J. G.

Rutten

G. E. H. M.

Schaufeli

W. B.

Schabracq

M. J.

Winnubst

J. A. M.

(2005). The role of work-related and personal factors in diabetes self-management. Patient Education and Counseling, 59, 87-96. doi:10.1016/j.pec.2004.10.004

43.

Wong

Backholer

Gearon

Harding

Freak-Poli

Stevenson

Peeters

(2013). Diabetes and risk of physical disability in adults: A systematic review and meta-analysis. The Lancet Diabetes & Endocrinology, 1, 106-114. doi:10.1016/S2213-8587(13)70046-9

44.

World Health Organization. (2012). Obesity and overweight. Retrieved from http://www.who.int/mediacentre/factsheets/fs311/en/#.U2gDIH5zIZ4.mendeley

45.

Zandwijk

Van Koppen

Van Mameren

Mesters

Winkens

De Bie

(2015). The accuracy of self-reported adherence to an activity advice. European Journal of Physiotherapy, 17, 183-191. doi:10.3109/21679169.2015.1075588

46.

Zhang

Wieffer

Modha

Balar

Pollack

Krishnarajah

(2010). The burden of hypoglycemia in type 2 diabetes: A systematic review of patient and economic perspectives. Journal of Clinical Outcomes Management, 17, 547-557. Available from www.jcomjournal.com