The impact of diabetes on labour market participation: a systematic review of results and methods

Pedron, Sara; Emmert-Fees, Karl; Laxy, Michael; Schwettmann, Lars

doi:10.1186/s12889-018-6324-6

BMC Public Health

Table 2 Eligible studies evaluating the effect of diabetes on labour market outcomes

From: The impact of diabetes on labour market participation: a systematic review of results and methods

Study	Methods					Results							Other
Study	Study design^a	Outcome definition	Age group	Exposure	Statistical method	Summary measure^d	Overall		Men		Women		Confounders^e	Comorbidities/complications modelling	Endogeneity^g	Quality score
Employment
Ng et al. (2001) [27]	C	Currently working (vs. currently not working)^b	18–65	Diabetes	Probit regression	PC	−0.04	*					A, CC, E, F, G, L, MS, SH	Stratification	no	5/6
Ng et al. (2001) [27]	C	Currently working (vs. currently not working)^b		T1DM	Probit regression	PC	0.11	*					A, CC, E, F, G, L, MS, SH	Stratification	no	5/6
Bastida et al. (2002) [19]	C	Currently working (vs. currently not working)	45+	Diabetes	Probit regression	ME			− 0.08	*	− 0.07		A, E, F, H, I, MS, O	–	no	5/6
Yassin et al. (2002) [34]	C	Being employed for most of the time in the last 12 months	18–64	Diabetes	Multinomial logistic regression	OR			0.53		0.48	*	A, E, I, MC, MS, O, SM	–	no	5/6
Brown et al. (2005) [20]	C	Currently working (vs. currently not working)^b	45+	Diabetes	Probit regression	PC			−1.02	*	− 0.34	*	A, E, F, H, I, MS, O	–	yes	5/6
Brown et al. (2005) [20]	C	Currently working (vs. currently not working)^b	45+	Diabetes	Recursive bivariate probit IV	PC			−1.71	*	0.51		A, E, F, H, I, MS, O	–	yes	5/6
Klarenbach et al. (2006) [22]	C	Working at a job or business and being present at that job for the week before	20–59	T2DM	Logistic Regression	OR	0.70	*					A, CC, E, G, L, MS, O	Confounders	no	4/6
Harris (2009) [21]	C	Currently employed (vs. not working but not retired)	> 25	Diabetes	Endogenous multivariate probit model	ME			−0.07	*	− 0.09	*	A, CC, E, F, I, MS, PA, SM	Confounders	yes	4/6
Latif (2009) [23]	C	Having had a job in the last 12 months	15–64	Diabetes	Probit regression	PC			−0.65	*	− 0.44	*	A, E, H, L, MS	–	yes	5/6
Latif (2009) [23]	C	Having had a job in the last 12 months		Diabetes	Recursive bivariate probit regression IV	PC			0.96		0.19		A, E, H, L, MS	–		5/6
Zhang et al. (2009) [36]	C	Currently working (vs. currently not working)^b	18–49	Diabetes	Endogenous recursive multivariate probit model	TE (%)			−3.91	*	− 3.70		A, CC, E, MS, O, Y	Confounders	yes	4/6
Zhang et al. (2009) [36]	C	Currently working (vs. currently not working)^b	50–64	Diabetes	Endogenous recursive multivariate probit model	TE (%)			−11.47	*	− 0.20		A, CC, E, MS, O, Y	Confounders	yes	4/6
Lin (2011) [24]	C	Currently working (vs. currently not working)	45–65	Diabetes	Recursive bivariate probit model	ME	−0.24	*	− 0.19	*	− 0.15		A, E, G, I, MS	–	yes	5/6
Minor (2011) [25]	C	Worked for pay at some point during the last year	20–65	Diabetes	IV estimation (model 1)	ME	−0.42	*					A, E, F, F, J, L, MS, O, SH	–	yes	5/6
				T1DM	IV estimation (model 2)	ME	−0.06
				T2DM	IV estimation (model 2)	ME	−0.45	*
Seuring et al. (2015) [31]	C	Having worked or carried out an activity that helped with the household expenses for at least 10 h over the last week	15–44	Diabetes	Probit regression	ME			− 0.01		0.00		A, E, F, I, L, MS, O, PE	–	yes	5/6
Seuring et al. (2015) [31]	C		45–64			ME			−0.110	*	− 0.06	*	A, E, F, I, L, MS, O, PE		yes	5/6
Nielsen et al. (2016) [28]	C	Currently working (vs. currently not working)	18–103	T1DM	Linear regression	RD	−9.10	*	− 5.30	*	− 12.20	*	A, E, G, SH	–	no	4/6
Minor et al. (2016) [26]	C	Currently working (vs. currently not working)	18–65	A1c levels > 6.5%	Probit regression (model 1)	ME			−0.02		− 0.16		A, E, F, MS, O, Y	–	no	5/6
				T2DM		ME			−0.11	*	− 0.19	*
				T1DM		ME			−0.17		0.18	*
				T2DM	Probit regression (model 2)	ME			−0.09	*	− 0.19	*
				T1DM	Probit regression (model 2)	ME			−0.16		0.175	*
Tunceli et al. (2005) [46]	L	Working for pay outside the home (vs. Not working for pay outside home)	51–61	Diabetes	Probit regression (model 1)	ME			−0.09	*	− 0.06	*	A, BMI, E, F, I, J, MS, O	Confounders SA^f	no	6/8
Tunceli et al. (2005) [46]	L		51–61	Diabetes	Probit regression (model 2)	ME			−0.07	*	− 0.04		A, BMI, E, F, I, J, MS, O, CC	Confounders SA^f		6/8
Pit et al. (2012) [44]	L	Employment last week (more than one hour spent on an occupation with or without pay) (vs. less than one hour spent last week on an occupation or unemployed)	51–61	Diabetes	Robust nested multivariate longitudinal analyses (GEE)	OR	0.82	*					BMI, CC, E, F, L, MS, SM, Y	Confounders	no	4/8
Minor (2013) [43]	L	Currently working (vs. currently not working)^b	45–53	T1DM	Logistic regression	LC			0.22		−0.03		A, E, F, FH, J, L, MS, O, Y	Confounders SA^f, modelling time from diagnosis	no	6/8
				T2DM	Logistic regression	LC			−0.42	*	− 0.37	*	A, E, F, FH, J, L, MS, O, Y
				T1DM	Logistic regression	LC			0.02		0.28		A, BMI, CC, E, F, FH, J, L, MS, O, Y
				T2DM	Logistic regression	LC			−0.28		−0.36	*	A, BMI, CC, E, F, FH, J, L, MS, O, Y
Unemployment
Alavinia et al. (2008) [18]	C	Currently unemployed (vs. Having done any kind of paid work in the last four weeks)	50–65	Diabetes	Logistic regression	OR	1.38						A, AL, BMI, CC, E, G, MS, PA, SM	Confounders	no	4/6
Smith et al. (2014) [32]	C	Currently not employed due to health reasons (vs. currently employed)	25–74	Diabetes	Logistic regression	OR	2.22	*					A, BMI, CC, E, F, G, I, L, MS, Y	Confounders	no	3/6
Van Der Zee-Neuen et al. (2017) [33]	C	Currently unemployed (vs. currently employed)	18–65	Diabetes	Multinomial logistic regression	OR	1.88						A, BMI, E, G, SM	–	no	4/6
Yassin et al. (2002) [34]	C	Transition from employment to no employment due to health reasons	18–64	Diabetes	Logistic regression	OR			3.1	*	2.9		A, E, I, MC, MS, O, SM	–	no	5/6
Rumball-Smith et al. (2014) [30]	C	More than one year of absence from the labour force or retirement (vs. Currently employed)	> 50	Diabetes	Cox proportional hazards models (matching diabetes subject with seven non-diabetes matches)	HR	1.30	*	1.26	*	1.34	*	A, E, G, L	–	no	5/6
Kraut et al. (2001) [41]	L	Not in the labour force (not employed and not seeking job) vs. in the labour force	18–64	Diabetes (w comp)	Logistic regression	OR	2.07	*					A, G, L, MS, O	Exposure	no	6/8
			18–64	Diabetes (w/o comp)	Logistic regression	OR	1.20
		Unemployed (no job but actively looking for it) vs. employed (with job)	18–64	Diabetes	Logistic regression	OR	1.45
				Diabetes (w comp)	Logistic regression	OR	1.69
				Diabetes (w/o comp)	Logistic regression	OR	1.35
Kouwenhoven-Pasmooij et al. (2016) [40]	L	Transition from employment to unemployment	> 50	Diabetes	Multinomial logistic regression	OR	1.17						A, CC, E, G, L, MS,	Confounders	no	6/8
Majeed et al. (2015) [42]	L	“Early paid work” (vs. “mostly in the labour force”)^c	45–50	Diabetes	Multinomial logistic regression	OR	1.44	*					BMI, E, F, I, MS, SM	–	no	4/8
Early retirement
Vijan et al. (2004) [47]	C	Currently retired (vs. currently working)	51–61	Diabetes	Logistic regression	OR	1.3						A, E, F, G, MS, O	–	no	4/6
Alavinia et al. (2008) [18]	C	Currently retired (vs. Having done any kind of paid work in the last four weeks)	50–65	Diabetes	Logistic regression	OR	1.33	*					A, AL, BMI, CC, E, G, MS, PA, SM	Confounders	no	4/6
Pit et al. (2013) [29]	C	Retirement due to health reasons (vs. Working)	45–65	Diabetes	Multinomial logistic regression	OR			1.44	*	1.30		A, CC, E, MS	Confounders	no	3/6
Pit et al. (2013) [29]		Retirement for other reasons (vs. Working)		Diabetes	Multinomial logistic regression	OR			1.16		1.07		A, CC, E, MS			3/6
Yen et al. (2011) [35]	C	Age at retirement	50–75	Diabetes at age 50	OLS regression	OLS	−1.39	*					CC, E, G, I, J, L, O	Confounders	no	3/6
Vijan et al. (2004) [47]	L	Incremental duration of retirement over the 8 years follow-up	51–61	Diabetes at baseline	Two-part multivariable model (logistic regression + OLS)	OLS	0.14	*					A, E, F, G, MS, O	–	no	6/8
Shultz et al. (2007) [45]	L	Transition from employment to retirement	47–64	Diabetes at baseline	Multinomial logistic regression	OR	3.37	*					A, CC, G, I, O	Confounders	no	4/8
Herquelot et al. (2011) [38]	L	Transition from employment to retirement	35–60	Diabetes (in at least three consecutive yearly questionnaire)	Cox proportional-hazard regression	HR	1.6	*					A, BMI, G, J	–	no	6/8
Kang et al. (2015) [39]	L	Transition from employment to early retirement due to health problems	45–70	Diabetes at baseline	Cox proportional hazard model	HR	1.47	*	1.52		1.40		A, AL, BMI, CC, G, I, J, PA, SH, SM	Confounders	no	5/8
Kouwenhoven-Pasmooij et al. (2016) [40]	L	Transition from employment to retirement	> 50	Diabetes	Multinomial logistic regression	OR	1.06						A, CC, E, G, L, MS,	Confounders	no	6/8
Disability pension
Vijan et al. (2004) [47]	C	Currently receiving a disability pension (vs. currently working)^a	51–61	Diabetes	Logistic regression	OR	3.1	*					A, E, F, G, MS, O	–	no	4/6
Van Der Zee-Neuen et al. (2017) [33]	C	Currently receiving a disability pension (vs. Currently employed)	18–65	Diabetes	Multinomial logistic regression	OR	2.32	*					A, BMI, E, G, SM	–	no	3/6
Vijan et al. (2004) [47]	L	Incremental duration of disability pension over the 8 years follow-up	51–61	Diabetes at baseline	Two-part multivariable model (logistic regression + OLS estimation)	Cumulative impact of diabetes (years)	0.79	*					A, E, F, G, MS, O	–	no	6/8
Herquelot et al. (2011) [38]	L	Transition from employment to disability pension	35–60	Diabetes (in at least three consecutive years)	Cox proportional-hazard regression	HR	1.4						A, BMI, G, J	–	no	6/8
Ervasti et al. (2016) [37]	L	Transition from employment to disability pension	30–65	Diabetes at baseline (vs. No metabolic condition)	Cox proportional-hazard regression (model 1)	HR	1.84	*					A, G, SES	Confounders SA^f	no	7/8
Ervasti et al. (2016) [37]	L	Transition from employment to disability pension	30–65	Diabetes at baseline (vs. No metabolic condition)	Cox proportional-hazard regression (model 2)	HR	1.56	*					A, AL, BMI, CC, G, J, PA, SES, SM	Confounders SA^f	no	7/8
Kouwenhoven-Pasmooij et al. (2016) [40]	L	Transition from employment to disability pension	> 50	Diabetes or high blood glucose levels	Multinomial logistic regression	OR	2.37	*					A, CC, E, G, L, MS,	Confounders	no	6/8

*p-value< 0.05
^aC: cross-sectional study; L: longitudinal study;
^bNot clearly stated but understood from context, interpretation, questions asked in survey
^cOther outcomes considered (“increasingly paid work”, “gradually not in paid work”, “mostly not in paid work”) are not reported here
^dOR: Odds Ratio HR: Hazard Ratio ME: Marginal Effect PC: Probit Coefficient LC: Logit Coefficient TE: Treatment Effect RD: Risk Differences OLS: OLS-Coefficient
^eA Age, AL Alcohol use, BMI Body-Mass-Index, CC Comorbidities/complications, E Education/Years of schooling, F Family related features (Number of children; Family size; People living in houehold; Household size; Living with someone who needs care; Competing activities;); FH Family health, G Gender, H Owns home, I Income/Wealth, J Employment characteristics, (Self-employment; Job tenure; Work experience; Part time; Occupational status;) L Region, Area of living/residence, MC Medical cost, MS Marital status, O Origin (Race, Australian born, Immigrant status) 0 PA Physical activity, PE Parental education, SH Subjective health/health related quality of life, SM Tobacco use/Smoking, Y Year
^fComplications were used in the sensitivity analysis as confounders;
^gPresence of endogeneity: yes = endogeneity of diabetes was detected; no = endogeneity of diabetes was not detected
Other information (e.g. sample size, country, method of data collection, results of IV tests) are not included in the table due to space limitations and are available from the corresponding author upon request

Back to article page

ISSN: 1471-2458

Contact us

Submission enquiries: bmcpublichealth@biomedcentral.com
General enquiries: ORSupport@springernature.com