Association of Intensive Lifestyle Intervention for Type 2 Diabetes With Labor Market Outcomes

Key Points

Question

Did a 10-year intensive lifestyle intervention (ILI) promoting weight loss for individuals with type 2 diabetes improve their labor market outcomes relative to diabetes education alone during the intervention and over the longer term?

Findings

This cohort study linked clinical trial data for 3091 participants with type 2 diabetes and overweight or obesity in an ILI (aged 45-75 years at the time of randomization) with benefit and earnings data from the Social Security Administration and examined labor market outcomes for 7 years before and 15 years after randomization. The ILI was associated with higher levels of employment, but there was no change in receipt of federal disability benefits.

Meaning

These findings suggest that better chronic disease management through lifestyle modification can improve labor market outcomes among older adults with type 2 diabetes.

This cohort study assesses whether an intensive lifestyle intervention for weight loss among individuals with both type 2 diabetes and overweight or obesity was associated with changes in employment, earnings, and disability benefit receipt.

Abstract

Importance

An intensive lifestyle intervention (ILI) has been shown to improve diabetes management and physical function. These benefits could lead to better labor market outcomes, but this has not been previously studied.

Objective

To estimate the association of an ILI for weight loss in type 2 diabetes with employment, earnings, and disability benefit receipt during and after the intervention.

Design, Setting, and Participants

This cohort study included participants with type 2 diabetes and overweight or obesity and compared an ILI with a control condition of diabetes support and education. Data for the original trial were accrued from August 22, 2001, to September 14, 2012. Trial data were linked with Social Security Administration records to investigate whether, relative to the control group, the ILI was associated with improvements in labor market outcomes during and after the intervention period. Difference-in-differences models estimating relative changes in employment, earnings, and disability benefit receipt between the ILI and control groups were used, accounting for prerandomization differences in outcomes for linked participants. Outcome data were analyzed from July 13, 2020, to May 17, 2023.

Exposure

The ILI consisted of sessions with lifestyle counselors, dieticians, exercise specialists, and behavioral therapists on a weekly basis in the first 6 months, decreasing to a monthly basis by the fourth year, designed to achieve and maintain at least 7% weight loss. The control group received group-based diabetes education sessions 3 times annually during the first 4 years, with 1 annual session thereafter.

Main Outcomes and Measures

Employment and receipt of federal disability benefits (Supplemental Security Income and Social Security Disability Insurance), earnings, and disability benefit payments from 1994 through 2018.

Results

A total of 3091 trial participants were linked with Social Security Administration data (60.1% of 5145 participants initially randomized and 97.0% of 3188 of participants consenting to linkage). Among the 3091 with fully linked data, 1836 (59.4%) were women, and mean (SD) age was 58.4 (6.5) years. Baseline clinical and demographic characteristics were similar between linked participants in the ILI and control groups. Employment increased by 2.9 (95% CI, 0.3-5.5) percentage points for the ILI group relative to controls (P = .03) with no significant relative change in disability benefit receipt (−0.9 [95% CI, −2.1 to 0.3] percentage points; P = .13).

Conclusions and Relevance

The findings of this cohort study suggest that an ILI to prevent the progression and complications of type 2 diabetes was associated with higher levels of employment. Labor market productivity should be considered when evaluating interventions to manage chronic diseases.

Introduction

Type 2 diabetes is a pressing health issue for the US population, with an estimated age-standardized prevalence among adults of 14.3%.¹ Multidomain intensive lifestyle intervention (ILI; providing counseling and support for behavioral change and goal setting, healthy diet, and increased physical activity) is recommended for the prevention and management of type 2 diabetes, particularly among persons who have overweight or obesity.^2,3,4 Strong evidence suggests that multidomain ILI reduces cardiovascular risk factors among persons with type 2 diabetes across North America, Europe, and Asia, but it is less clear whether ILIs are cost-effective.^5,6,7,8,9

Assessing whether any intervention for type 2 diabetes is worth the cost requires considering the effects on labor productivity. By improving physical function and mental health, lifestyle interventions could have positive effects on labor market participation, productivity at work, and earnings.^10,11,12 Improvements in physical function could also reduce receipt of federal disability payments, which were received by 12 million adults aged 18 to 64 years and comprised $13.6 billion in monthly benefits in December 2020.¹³

Most prior studies on the association between chronic disease management and labor market outcomes have relied on observational data or instrumental variables estimation. For example, research has found that body weight is negatively associated with wages (for White women) and that diabetes is negatively associated with employment.^14,15 In particular, estimates from observational studies may be biased by unobserved individual attributes that are correlated with clinical status and labor market outcomes. Two randomized clinical trials of workplace wellness programs (which include components similar to ILIs and seek to improve chronic disease management)^16,17,18,19 found no effect on workplace productivity or earnings. However, these programs were more widely targeted across employees and had no effect on body mass index or cardiovascular risk factors (cholesterol levels, hypertension, or blood glucose level) and thus may not represent the effect of a more targeted and clinically effective lifestyle intervention on employment outcomes. One prominent study, the Multiple Risk Factor Intervention Trial (MRFIT),²⁰ randomized approximately 13 000 middle-aged men to receive a bundle of interventions to reduce mortality due to coronary heart disease. MRFIT had beneficial effects on cholesterol level, smoking, and blood pressure and increased earnings and family income, suggesting the potential economic benefits of improving chronic disease management. Whether these labor market benefits accrue in patients with better managed type 2 diabetes remains unknown. Given the numerous microvascular and macrovascular complications of type 2 diabetes and their associated effects on functional status, it is important to address this evidence gap. Our objective was to assess whether the Look AHEAD (Action for Health in Diabetes) ILI for weight loss among individuals with both type 2 diabetes and overweight or obesity was associated with changes in employment, earnings, and disability benefit receipt.

Methods

Study Overview and Design

This cohort study was conducted as an ancillary study to the Look AHEAD randomized clinical trial.^21,22 Participants aged 45 to 75 years with both type 2 diabetes and overweight or obesity were randomized to an ILI group or a control group receiving diabetes support and education between August 22, 2001, and April 30, 2004, within 16 sites around the US. Participants in the ILI group received one-on-one sessions with counselors, dieticians, exercise specialists, and behavioral health specialists that occurred on a weekly basis in the first 6 months, decreasing in frequency to a monthly basis by the fourth year. The control group participated in 3 group sessions in the first 4 years and annual sessions thereafter. Participants in the ILI group had reduced weight and hemoglobin A_1c (HbA_1c) levels and increased fitness and disability-free life-years relative to the control group (among other benefits), but the ILI did not reduce the rate of cardiovascular events.^5,23 The intervention period ended September 14, 2012, but the study continued to follow up participants. As a part of 2 cohort studies that followed the Look AHEAD cohort after the intervention period (Look AHEAD Continuation and Extension), Look AHEAD obtained informed consent to perform linkages with administrative data. We linked study participants who provided written informed consent with data from the Social Security Administration (SSA) from 1994 through 2018. The Institutional Review Boards of the University of Southern California and the University of Minnesota approved the linkages. Wake Forest University Health Sciences and the SSA deferred their reviews to the University of Southern California Institutional Review Board. Study outcomes were prespecified on ClinicalTrials.gov^21,22 and analyses were prespecified in the study proposals to the Look AHEAD study (eAppendix in Supplement 1). The trial protocol for the Look AHEAD ILI has been previously published.⁵ We followed the Consolidated Standards of Reporting Trials (CONSORT) reporting guideline as applied to a secondary analysis of clinical trial data.

Data Linkage and Study Population

Of the original 5145 Look AHEAD participants, 3753 (1907 ILI and 1846 control participants) participated in either the Continuation or Extension studies. Of these, 3188 (1618 ILI and 1570 control participants) consented to administrative data linkages and provided Social Security Numbers. We successfully linked 3091 participants (1559 ILI and 1532 control participants) to Social Security databases, comprising 60.1% of those initially randomized and 97.0% of those consenting to linkages (Figure 1). One of the 16 Look AHEAD sites (the Southwestern American Indian Center) did not participate in administrative data linkages. Across the remaining 15 sites, the linkage rate ranged from 44% to 77%. The linkage rate exceeded 50% for 14 of 15 sites and exceeded 60% for 11 of 15 sites. The eMethods in Supplement 1 includes details on the data linkage.

Figure 1. Study Flow Diagram.

ILI indicates intensive lifestyle intervention.

Data Sources

We used SSA databases providing annual earnings from employment and self-employment (derived from Internal Revenue Service data), participation in SSA disability programs, and detailed information on disability payments. Further details on SSA databases are included in the eMethods in Supplement 1. We linked the SSA data with trial data for each participant, including information on baseline characteristics, intervention assignment, clinic site, and clinical outcomes during the study.

Study Measures

We constructed labor market outcomes, including employment, earnings, receipt of SSA disability payments, and amount of disability payments. We defined study years as 12-month increments starting with the month of randomization. We defined employed as having any earnings in a given study year from employment or self-employment, and we defined earnings as the sum of earnings across all employers in a given study year. Because study years do not usually align with our calendar-year Internal Revenue Service data, participants were identified as employed if they worked in either calendar year overlapping with a given study year. Similarly, we constructed person-year earnings by taking the weighted mean of earnings across the calendar years overlapping with a study year. In the analysis of earnings, we included nonemployed participants as earning $0 in a given year. Earnings were adjusted for inflation and expressed in 2018 US dollars.

The SSA administers 2 disability programs. Social Security Disability Insurance (SSDI) provides payments to individuals who have a disability and have accrued sufficient work credits to be eligible for benefits. Supplemental Security Income (SSI) provides payments to individuals who have disabilities and low incomes who have not accrued a sufficient work history to receive SSDI. Individuals with SSDI transition to SSA Retirement Benefits when they reach full retirement age (currently 66 years and 10 months in 2021), while those receiving SSI transition to aged-based SSI when they reach 65 years of age. We hypothesized that improved health from the ILI could lead to a reduction in receipt of disability benefits for the ILI vs the control group, which could occur due either to fewer applications or to program participation falling as participants’ health status improved. We observed monthly enrollment and payments for both programs. We constructed a binary outcome measure for enrollment in the 2 federal disability programs, indicating whether participants received benefits from either disability program in a given study year. We also constructed annual payments across the 2 programs for each study year. In our analysis of payments, we included nonrecipients as receiving $0 in benefits. Disability benefits were adjusted for inflation and expressed in 2018 US dollars. Further details of data construction are described in the eMethods in Supplement 1.

Statistical Analysis

First, we examined whether observed characteristics of trial participants linked to SSA databases were similar between the ILI and control groups and to the full set of participants initially randomized in the trial. We also investigated whether the ILI had a similar effect on weight and HbA_1c levels in the linked sample compared with those initially randomized.

In our main analysis, we used difference-in-differences regression models that estimated relative changes in outcomes for the ILI vs the control groups. Notably, we compared postintervention changes in outcomes between the ILI and control groups rather than postintervention levels (as indicated in our prespecified analyses) because we observed baseline differences in labor market outcomes between the ILI and the control groups (ie, prior to the intervention) (eTable 1 in Supplement 1). However, in supplemental analyses described in the eMethods and eFigures 3A and 4A of Supplement 1, we found that the ILI and control groups exhibited parallel outcome trends prior to the intervention, which supported the linked control group being used as a counterfactual for the ILI group.

For employment and earnings, we were able to link participants from 7 years prior to randomization through 15 years after randomization. For disability benefit receipt and payments, we assessed participants 7 years prior to randomization through 13 years after randomization. To estimate a mean postintervention effect, we used difference-in-differences models to estimate the mean postintervention change in outcomes for the ILI relative to the control groups, using the 3 years prior to randomization as the baseline period. All analyses controlled for baseline demographic and economic characteristics (age, sex, self-reported race and ethnicity), clinical attributes at randomization (body mass index [calculated as weight in kilograms divided by height in meters squared], HbA_1c level, and history of cardiovascular disease, hypertension, and baseline cardiovascular fitness), socioeconomic status (educational attainment), and initial study clinic site. We estimated linear regressions in our main difference-in-differences regressions, for both continuous and binary outcomes. In eTable 7 in Supplement 1, we report marginal effects using logit models for binary outcomes.

We estimated 3 sets of subgroup analyses based on baseline characteristics: (1) comparing individuals with at least a college degree vs less than a college degree, (2) individuals with above or below median income in the sample, and (3) older vs younger participants at baseline. We hypothesized that the effects of the intervention differed according to job characteristics. For instance, it is likely that reduced weight and improved health has more impact in physically demanding jobs. Because we did not observe occupation in the data, we also estimated separate effects of the ILI on study outcomes for participants with a higher level of education at baseline (at least a college degree) vs less education (less than a college degree). This was motivated by external data showing that individuals with less education are considerably more likely to be employed in physically demanding jobs.²⁴ We estimated separate estimates for participants who were older at baseline (≥60 years) vs younger (<59 years), reflecting that eligibility for Social Security disability programs transition to age-based eligibility starting at 65 years of age. For all analyses, we calculated clustered-robust SEs at the participant level.

Data were analyzed using Stata, version 16.0 (StataCorp LLC). Two-sided P < .05 indicated statistical significance. Analysis of disability outcomes began July 13, 2020, and analysis of employment and earnings began September 11, 2020, in each case after submitting outcome preregistration, and was completed May 17, 2023. We did not include adjustments for multiple comparisons for our main analysis, as we prespecified our primary outcomes on ClinicalTrials.gov separately for the employment-related analysis (employment and earnings) and the disability program analysis (disability benefit receipt and disability benefit application, the latter outcome was not feasible to consistently measure), alleviating concerns of type I errors.²⁵ For subgroup analysis, we report whether the results would be significant after applying a Bonferroni correction for the 6 subgroups we examined, which results in a threshold for statistical significance of 2-sided P = .008. We estimated ex post minimum detectable effects and found that we were powered to detect economically meaningful effect sizes (eTable 2 in Supplement 1). Additional details on the data sources, outcome definitions, and statistical analyses are provided in the eMethods in Supplement 1.

Results

A total of 3091 trial participants were linked with SSA data, including 1836 (59.4%) women and 1255 (40.6%) men with a mean (SD) age of 58.4 (6.5) years. In terms of race and ethnicity, 525 (17.0%) were non-Hispanic Black, 422 (13.7%) were Hispanic, 2033 (65.8%) were non-Hispanic White, and 111 (3.6%) were other race or ethnicity (including Asian or Pacific Islander, American Indian or Alaska Native, and other, pooled due to small sample sizes for some groups). The linked ILI and control group participants were similar across most baseline characteristics (Table 1). The 1559 ILI and 1532 control group participants were of similar age (mean [SD], 58.3 [6.4] vs 58.5 [6.5] years, respectively) and had similar baseline clinical characteristics such as body mass index (35.9 [5.9] vs 36.0 [5.8], respectively) and HbA_1c levels (7.2% [1.1%] vs 7.2% [1.2%], respectively). However, a higher percentage of the ILI group were college graduates (703 [45.1%] vs 636 [41.5%]). We also compared the characteristics of the 3091 linked participants relative to the 2054 participants who we were unable to link to SSA data (eTable 3 in Supplement 1). Overall, linked and nonlinked participants were similar across many baseline characteristics, but we found moderate differences in educational attainment (post-secondary without college degree, 1103 of 3091 [35.7%] vs 813 of 2054 [39.6%]), prior cardiovascular disease (368 of 3091 [11.9%] vs 344 of 2054 [16.7%]), duration of diabetes (mean [SD], 6.5 [6.2] vs 7.3 [7.0] years), and lower HbA_1c levels (mean [SD], 7.2% [1.1%] vs 7.4%[1.2%]). While linkage rates varied across study sites (44% to 77%), we found similar patterns comparing overall differences, mean within-site differences, and the distribution of differences across sites. One important difference was that that the linked sample had a lower percentage of participants who were not Black, Hispanic, or White (111 [3.6%] vs 298 [14.5%]). This finding reflects that the Southwestern American Indian Center Look AHEAD site did not participate in administrative data linkages and thus was excluded from this study; however, the mean within-site difference for this measure was only 0.4 (95% CI, −0.6 to 1.5) percentage points. Of note, there were no statistically significant within-site differences in baseline characteristics between the ILI and control groups within the linked sample, indicating that attrition was unrelated to treatment status, alleviating concerns of bias (eTable 4 in Supplement 1). Moreover, relative reductions in weight and HbA_1c level for linked ILI participants were similar to those of the full cohort, suggesting that the effects of the ILI on labor market outcomes for linked ILI participants may have broader generalizability (eFigure 1 in Supplement 1). Relative weight reductions occurred for ILI participants across educational attainment groups (eFigure 2 in Supplement 1).

Table 1. Baseline Characteristics of Linked Participants.

Characteristic	Participant groupa
	Control (n = 1532)	ILI (n = 1559)
Age, mean (SD), y	58.5 (6.5)	58.3 (6.4)
Sex
Men	611 (39.9)	644 (41.3)
Women	921 (60.1%)	915 (58.7)
Race and ethnicity
Hispanic	211 (13.8)	211 (13.5)
Non-Hispanic Black	263 (17.2)	262 (16.8)
Non-Hispanic White	1007 (65.7)	1026 (65.8)
Otherb	51 (3.3)	60 (3.8)
Educational attainment
High school or less	322 (21.0)	327 (21.0)
Some college	574 (37.5)	529 (33.9)
College graduate	636 (41.5)	703 (45.1)
Family incomec
<$20 000	146 (10.3)	149 (10.5)
$20 000 to <$40 000	303 (21.4)	290 (20.5)
$40 000 to <$60 000	291 (20.5)	280 (19.8)
$60 000 to <$80 000	235 (16.6)	254 (18.0)
≥$80 000	442 (31.2)	440 (31.1)
Baseline clinical status
Weight, mean (SD), kg	101.0 (18.7)	101.2 (19.6)
BMI, mean (SD)	36.0 (5.8)	35.9 (5.9)
Hemoglobin A1c level, mean (SD), % of total hemoglobin	7.2 (1.2)	7.2 (1.1)
History of cardiovascular disease	173 (11.3)	195 (12.5)
Hypertension	1258 (82.1)	1305 (83.7)
Duration of type 2 diabetes, mean (SD), y	6.4 (5.8)	6.6 (6.6)
Metabolic equivalents, mean (SD)	7.3 (2.0)	7.3 (2.0)

Abbreviations: BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); ILI, intensive lifestyle intervention.

^a Unless otherwise indicated, data are expressed as No. (%) of participants. Percentages have been rounded and may not total 100.

^b Includes Asian or Pacific Islander, American Indian or Alaska Native, and other race or ethnicity (pooled due to small sample sizes for some groups).

^c Baseline family income information available for only 1413 ILI and 1417 control participants.

Association of ILI With Labor Market Outcomes

We plotted employment rates overall and by level of baseline educational attainment (adjusting for the baseline demographic, economic, and clinical characteristics described above) and compared the employment rate for the control group with that for the ILI group (Figure 2). Among control participants, employment fell from 89% 7 years prior to randomization to 32% 15 years after randomization, reflecting that the cohort was aging and leaving employment during this period (Figure 2A). A nominally lower percentage of ILI participants were employed in the years prior to randomization relative to the control group (87% vs 89%, 7 years prior to randomization). However, employment rates converged between the ILI and control groups soon after randomization and were eventually higher for ILI participants, implying a relative increase in employment for the ILI group that extended beyond the intervention period. A similar and more pronounced pattern existed for ILI participants with less than a college degree (Figure 2B). In contrast, ILI and control participants with at least a college degree had similar employment rates before and after the intervention (Figure 2C).

Figure 2. Percentage of Employed Look AHEAD (Action for Health in Diabetes) Participants.

The first vertical dashed line indicates the year prior to randomization and the second vertical dashed line indicates the median intervention length (10 years). ILI indicates intensive lifestyle intervention.

The percentage of control group participants receiving disability benefits increased from 1.9% (7 years prior to randomization) to 6.5% (4 years after randomization), before decreasing over the rest of the study period as individuals transitioned from federal disability programs to either SSA retirement benefits or aged-based SSI when they turned 65 years of age (Figure 3A). Participants in the ILI group were more likely to receive disability benefits prior to randomization (3.0% 7 years prior to randomization), but the percentage receiving disability benefits converged with and subsequently fell below that of the control group after randomization; this suggests a relative reduction in disability benefits received for the ILI group. A similar but more pronounced pattern existed for ILI and control participants with less than a college degree (Figure 3B). In contrast, there was a small nominal relative increase in disability benefits received for the ILI group among participants with at least a college degree (Figure 3C). Notably, Figures 2 and 3 suggest that trends in employment and disability benefit receipt were parallel for the ILI vs the control groups prior to randomization. Using an event study analysis (discussed in eMethods in Supplement 1), we could not reject that outcome trends were parallel between the ILI and control groups in the prestudy period (eFigures 3A and 4A in Supplement 1). This finding supports the assumption that the difference in outcomes between the ILI and control groups would have been stable in the absence of the intervention, which is an important assumption for the internal validity of the difference-in-differences analysis.

Figure 3. Percentage of Look AHEAD (Action for Health in Diabetes) Participants Receiving Disability.

The first vertical dashed line indicates the year prior to randomization and the second vertical dashed line indicates the median intervention length (10 years). ILI indicates intensive lifestyle intervention.

Difference-in-Differences Analysis

Using difference-in-differences regressions, we estimated baseline differences and postintervention changes in each outcome for the ILI relative to the control group (Table 2). For the full sample, the prerandomization difference in employment between the ILI and the control group was statistically nonsignificant (−2.1 [95% CI, −4.6 to 0.3] percentage points; P = .08). Relative to the control group, employment in the ILI group increased by 2.9 (95% CI, 0.3-5.5) percentage points (P = .03). Among participants with less than a college degree, prerandomization employment was 4.0 percentage points lower for the ILI vs the control group (95% CI, −7.5 to −0.5 percentage points; P = .02) and employment in the ILI group increased by 5.2 percentage points relative to the control group (95% CI, 1.6-8.8 percentage points; P = .005); this estimate is still significant in that it accounts for the 6 subgroup analyses using a Bonferroni correction with an effective significance threshold of .008. The event study results also show that relative increases in employment extended beyond the intervention period for the full linked sample and participants with less than a college degree (eFigure 3A and B in Supplement 1). There was no statistically significant change for participants with at least a college degree (−0.2 [95% CI, −3.9 to 3.5] percentage points; P = .97). There were no statistically significant baseline differences or differential changes in earnings for any group.

Table 2. Estimated Association of ILI With Economic Outcomes, Full Matched Sample and by Baseline Educational Attainment^a.

Outcome	Control baseline mean	Baseline ILI differenceb		Difference-in-differences estimatec
		Absolute (95% CI)	P value	Absolute (95% CI)	Percentage (95% CI)	P value
Full linked sample (n = 3091)
Employment, %	79.2	−2.1 (−4.6 to 0.3)	.08	2.9 (0.3 to 5.5)	3.7 (0.4 to 6.9)	.03
Earnings, $	45 043	−107 (−4290 to 4076)	.96	309 (−3270 to 3888)	0.7 (−7.3 to 8.6)	.87
Disability benefit receipt, %	5.4	1.2 (−0.3 to 2.7)	.11	−0.9 (−2.1 to 0.3)	−16.7 (−38.9 to 5.6)	.13
Disability benefit payments, $	759	237 (−5 to 479)	.06	−137 (−349 to 75)	−18.1 (−46.0 to 9.9)	.21
Less than college degree (n = 1752)
Employment, %	76.7	−4.0 (−7.5 to −0.5)	.02	5.2 (1.6 to 8.8)	6.8 (2.1 to 11.5)	.005
Earnings, $	31 972	−750 (−4275 to 2775)	.68	781 (−2303 to 3865)	2.4 (−7.2 to 12.1)	.62
Disability benefit receipt, %	7.6	2.1 (−0.3 to 4.4)	.08	−2.2 (−4.2 to −0.3)	−28.9 (−55.3 to −3.9)	.03
Disability benefit payments, $	995	402 (42 to 762)	.03	−365 (−695 to −35)	−36.7 (−68.2 to −3.5)	.03
At least a college degree (n = 1339)
Employment, %	82.9	0.1 (−3.3 to 3.4)	.97	−0.2 (−3.9 to 3.5)	−0.2 (−4.7 to 4.2)	.91
Earnings, $	63 458	−419 (−8976 to 8138)	.92	942 (−6206 to 8090)	1.5 (−9.8 to 12.7)	.80
Disability benefit receipt, %	2.2	−0.0 (−1.7 to 1.6)	.96	0.7 (−0.4 to 1.9)	31.8 (−18.2 to 86.4)	.22
Disability benefit payments, $	426	−12 (−337 to 313)	.94	171 (−61 to 403)	40.1 (−14.3 to 94.6)	.15

Abbreviation: ILI, intensive lifestyle intervention.

^a Outcomes include annual employment, earnings, receipt of disability benefits (Supplemental Security Income or Social Security Disability Insurance), and disability benefit payments. The disability analysis included 49 456 person-year observations for the full sample (3091 participants across 16 years), 28 032 person-year observations for the participants with less than a college degree (1752 participants across 16 years), and 21 424 person-year observations for the participants with a college degree or higher (1339 participants across 16 years). The employment and earnings sample included 52 547 person-year observations for the full sample (3091 participants across 17 years), 29 784 person-year observations for the participants with less than a college degree (1752 observations across 17 years), and 22 763 person-year observations for the participants with a college degree or higher (1339 observations across 17 years).

^b Baseline differences indicate the outcome differences between the ILI and control groups in the 3 years prior to randomization.

^c Indicates relative changes in outcome from 3 years prior to randomization to postrandomization for ILI vs control groups. Absolute changes are in percentage point units (for employment and disability benefit receipt) and dollar units (for earnings and disability benefit payments). Percentage changes are relative to the control group mean in the year prior to randomization.

For the full sample, the prerandomization difference in disability benefit receipt for the ILI vs the control groups was statistically nonsignificant (1.2 [95% CI, −0.3 to 2.7] percentage points; P = .11) and there was no statistically significant relative change in disability benefit receipt for the ILI group (−0.9 [95% CI, −2.1 to 0.3] percentage points; P = .13). Among participants without a college degree, there was no significant difference in prerandomization disability benefit receipt between the ILI and the control groups (2.1 [95% CI, −0.3 to 4.4] percentage points; P = .08). There was a statistically significant relative reduction in disability benefit receipt for ILI participants with less than a college degree (−2.2 [95% CI, −4.2 to −0.3] percentage points; P = .03), but this estimate was not robust to a Bonferroni correction. There was no statistically significant prerandomization difference or differential change in disability benefit receipt for ILI participants with a college degree or in disability benefit payments for the full sample or college graduates. Among ILI participants with less than a college degree, prerandomization mean disability benefit payments were $402 (95% CI, $42-$762; P = .03) higher than the control group, and payments decreased by $365 (95% CI, −$695 to −$35; P = .03) after the intervention, but this estimate was not robust to a Bonferroni correction.

We conducted a number of sensitivity analyses. Briefly, we estimated the main analysis excluding a small number of participants who died late in the sample period (2014-2018) (eTable 5 in Supplement 1), stratified the analysis by baseline income (eTable 6 in Supplement 1), estimated logit models rather than linear probability models (eTable 7 in Supplement 1), stratified the estimates by baseline age and focused on younger participants prior to aging out of disability benefit eligibility (eTables 8 and 9 in Supplement 1), and estimated a range of alternative difference-in-differences specifications (eTable 10 in Supplement 1). These estimates were all consistent with our primary findings.

Discussion

In this cohort study, we linked participants in the Look AHEAD trial with SSA data to investigate the association of an ILI targeting weight loss among individuals with type 2 diabetes and overweight or obesity with labor market outcomes. We found that the ILI was associated with increased employment but not with receipt of disability benefits. In a subgroup analysis focusing on participants with less than a college degree, we found larger associations between the ILI and employment and statistically significant reductions in disability benefit receipt and payments, but the disability estimates could reflect random chance due to multiplicity. The magnitudes of the estimates were sizable, representing a 4% increase in employment overall and a 7% increase in employment for the group with lower educational attainment. These effects are commensurate to the ILI’s effects on weight (8% reduction in the first year of the trial), physical fitness (10% increase in metabolic equivalents in the first 4 years of the trial), and mobility loss (48% reduction in the first 4 years of the trial).^5,26

A previous cost-effectiveness analysis of Look AHEAD (focused on health utility and health care spending)⁷ found mixed evidence for whether the benefits of the ILI justified the costs. Our findings imply that the ILI had benefits in terms of employment and disability benefit receipt that are not captured by typical cost-effectiveness frameworks, although overall effects on earnings and disability payments were statistically insignificant. This finding aligns with the Second Panel on Cost-Effectiveness in Health and Medicine recommendation that evaluations of clinical interventions should consider broader societal effects on non–health care outcomes, including labor market productivity, earnings, and the use of social services, and shows that this is particularly important for participants with less educational attainment.²⁷

Our study is among the first, to our knowledge, to link clinical trial data with administrative records to investigate the association between effective chronic disease management and labor market outcomes. The MRFIT successfully reduced cholesterol levels, smoking, and blood pressure and decreased the percentage of participants who were unable to work due to disability, which is consistent with our findings for Look AHEAD participants with less than a college degree.²⁰ In contrast to Look AHEAD, MRFIT increased earnings and family income, but not employment, and MRFIT also increased earnings among white collar workers, with less effect for blue collar workers. The Look AHEAD ILI’s positive association with employment, in contrast to MRFIT, may be explained by reductions in weight and improvements in physical fitness (primary intervention outcomes for the trial) allowing participants to work longer. Notably, Look AHEAD participants were older at baseline than MRFIT participants (aged 45 to 75 years vs 35 to 57 years, respectively), and employment rates were falling prior to the intervention as participants left the labor market. Moreover, improvements in physical function may have been more important for participants with more physically demanding jobs, who are less likely to have a college degree.²⁴

Our findings are consistent with past work identifying educational gradients in outcomes for persons with diabetes.²⁸ In particular, for persons with type 1 diabetes, intensive diabetes management in the seminal Diabetes Control and Complications trial was shown to be more valuable for those who are less educated.²⁸ This result implied that those with more education were better at self-managing their disease, and hence there was less for the care team to improve upon clinically. The present study finds a similar pattern for type 2 diabetes and extends those benefits to labor market outcomes.

Our results also speak to the long-standing scientific inquiry into how health affects socioeconomic status. Prior work has established the importance of in utero and childhood health on the formation of cognitive and noncognitive abilities and economic outcomes during adulthood.^29,30 Our study provides evidence that improving health later in life (in this case, by improving type 2 diabetes management) may also have positive economic benefits by extending the ability to work and reducing participation in SSA disability programs.

Limitations

The primary limitation of our study is that we were only able to link Look AHEAD participants to SSA data if they had consented to linkages as a part of observational follow-up after the trial intervention was stopped. In addition, the linkage rate varied across study sites. However, this included 60.1% of the initially randomized sample, the ILI and control groups continued to be balanced on observable characteristics, and we found similar associations of the ILI with weight loss and HbA_1c levels relative to the full sample.

Conclusions

The findings of this cohort study suggest that an ILI to prevent the progression and complications of type 2 diabetes was associated with higher employment. This result suggests the potential of better chronic disease management for improving labor market outcomes. Labor market productivity should be considered when evaluating interventions to control chronic diseases.

Supplement 1.

eMethods. Data Accrual and Analyses

eFigure 1. Adjusted Weight and HbA_1c Level by Treatment Group for Linked Sample

eFigure 2. Adjusted Weight by Treatment Group for Linked Sample, by College Degree

eFigure 3. Event Study Estimates: Employment

eFigure 4. Event Study Estimates: Disability Benefit Receipt

eTable 1. Baseline Labor Market Outcomes for ILI and Control Group, Year Prior to Randomization

eTable 2. Post Hoc Minimum Detectable Effects (MDE)

eTable 3. Baseline Characteristics of Linked and Nonlinked Look AHEAD Participants

eTable 4. Baseline Characteristics of ILI vs Control Group Within Linked Sample

eTable 5. Comparison of Difference-in-Differences Estimates Between Main Sample (Overall and by Educational Subgroup) vs the Survivor Sample of Participants Who Did Not Die During the Sample Period

eTable 6. Estimated Association of ILI With Labor Market Outcomes, by Above and Below Median Baseline Income

eTable 7. Comparison of Marginal Effects From Difference-in-Differences Model Estimated With Linear Probability Model vs Logit Model

eTable 8. Estimated Association of ILI With Economic Outcomes by Baseline Age

eTable 9. Disability Estimates for Participants 58 Years or Younger at Baseline With Less Than College Degree, First 6 Years of Intervention (911 Participants)

eTable 10. Alternative Difference-in-Differences Regression Specifications for Primary Outcomes

eAppendix. Manuscript Proposals to Look AHEAD Trial

Supplement 2.

Data Sharing Statement