p u b l i c h e a l t h 1 5 1 ( 2 0 1 7 ) 8 1 e8 6

Available online at www.sciencedirect.com

Public Health

journal homepage: www.elsevier.com/puhe

Original Research

The longitudinal association between informal

caregiving and body mass index in the second half
of life: findings of the German Ageing Survey

A. Hajek*, H.-H. Ko
€nig
Department of Health Economics and Health Services Research, University Medical Center Hamburg-Eppendorf,
Hamburg, Germany

article info abstract

Article history: Objective: We aimed to investigate the relation between informal caregiving and body mass
Received 2 March 2017 index (BMI) longitudinally.
Received in revised form Study design: The data were drawn from wave 2 (2002) to wave 5 (2014) of the German
13 June 2017 Ageing Survey. This is a representative sample of the community-dwelling population aged
Accepted 24 June 2017 40 years and above in Germany.
Methods: Self-rated BMI was used. Individuals were asked whether they provide informal
care on a regular basis. Adjusting for employment status, age, marital status, morbidity
Keywords: and depressive symptoms, fixed effects regressions were used.
Body mass index Results: The fixed effects regressions showed that the onset of informal caregiving was not
Informal care associated with changes in BMI in the total sample and in women, whereas the onset of
Home care informal caregiving was associated with increasing BMI in men (b ¼ 0.15, P < 0.05). In
Aged addition, an increase in BMI was positively associated with ageing, an increase in morbidity
Longitudinal studies and a decrease in frequency of sports activities in the total sample and in both sexes.
Public health Conclusions: Our findings stress the longitudinal relation between informal caregiving and
BMI in men. Consequently, it might be helpful to generate weight management strategies
specifically designed for male informal caregivers.
© 2017 The Royal Society for Public Health. Published by Elsevier Ltd. All rights reserved.

demographic changes will lead to an increased need for long-

Introduction term care, stressing the relevance of informal care. However,
it has been shown that starting informal care is associated
Numerous studies have shown that individuals in need of care with negative health outcomes for caregivers such as
prefer to be cared for at home rather than in a nursing home increased depressive symptoms or reduced life satisfaction.4
for as long as possible.1 In Germany, care at home is mostly Thus far, there is little evidence concerning the association
provided by informal caregivers.2,3 It is expected that between informal caregiving and body mass index (BMI). Few

* Corresponding author. University Medical Center Hamburg-Eppendorf, Department of Health Economics and Health Services Research,
Hamburg Center for Health Economics, Germany. Tel. þ49 40 7410 52877; fax: þ49 40 7410 40261.
E-mail addresses: [email protected] (A. Hajek), [email protected] (H.-H. Ko€ nig).
http://dx.doi.org/10.1016/j.puhe.2017.06.022
0033-3506/© 2017 The Royal Society for Public Health. Published by Elsevier Ltd. All rights reserved.
82 p u b l i c h e a l t h 1 5 1 ( 2 0 1 7 ) 8 1 e8 6

cross-sectional studies compared the BMI of informal care- In the waves 1, 2, 3 and 5, a national representative base-
givers and non-caregivers. In a young and rather specific line sample was drawn and followed-up afterwards, whereas
sample (n ¼ 57 black female recipients; 37 caregivers and 20 wave 4 was a pure panel survey, i.e. including solely in-
non-caregivers; median age was 28.2 years), no significant dividuals who had already been interviewed before. 1526 in-
differences in BMI between caregivers and non-caregivers dividuals from wave 1 were re-interviewed in wave 2. In wave
were observed.5 Contrarily, large, representative cross- 3, 1995 individuals were reinterviewed and 6205 individuals
sectional studies found significant differences in BMI be- were first time participants. In wave 5, about 6000 individuals
tween informal caregivers and individuals not providing were first time participants, whereas over 4000 individuals
informal care. For example, a cross-sectional American study were reinterviewed. Register sampling of the community-
(Behavioral Risk Factor Surveillance System; n ¼ 292,813; dwelling individuals was used, disproportionally stratified by
74,135 were self-identified as caregivers and 216,652 were age, gender and geographical location. The panel data struc-
non-caregivers; telephone interview in the year 2009) ture allows investigation of intraindividual changes over time.
demonstrated that caregivers (mean BMI: 27.8 kg/m2) had a A total of 5194 participants were interviewed in wave 2
slightly higher BMI than non-caregivers (mean BMI: 27.4 kg/ (response rate: 38%). 8200 individuals were interviewed in
m2).6 This finding was also supported by a cross-sectional wave 3 (response rate: 38%) and 4855 participants were inter-
study conducted in Thailand in 2009 (n ¼ 60,569) where care- viewed in wave 4 (response rate: 56%). In wave 5, 10,355 in-
givers reported higher BMI compared to non-caregivers.7 dividuals took part (response rate: 33%). The response rates are
Drawing data from the Nurses' Health Study (n ¼ 54,411, an similar to other large survey studies carried out in Germany.12
ongoing cohort of the US female registered nurses aged 46e71 The main reasons for the lack of follow-up data were refusal to
years; questionnaire), another study has shown that women participate and bad health condition. Concerning the DEAS
providing care to an ill/disabled spouse had a higher BMI study (e.g. sample composition, panel attrition or sample se-
(mean BMI: 27.0 kg/m2) compared to those not providing care lection), further details were provided elsewhere.13,14
at all (mean BMI: 25.6 kg/m2).8 Informal caregiving, BMI and various other variables were
However, longitudinal studies are missing investigating included from wave 2 onwards. Consequently and worth
the relation between informal care and BMI. Longitudinal repeating, data from wave 2 to wave 5 were used in the pre-
studies are needed to get insights into the causal relationship sent study. Written informed consent was given prior to the
between these factors. Moreover, using panel econometric interview.
techniques can mitigate the problem of unobserved hetero-
geneity.9 Therefore, based on a representative sample of in- Dependent variable
dividuals in the second half of life (40 years and above), the
aim of the present study was to investigate whether informal Self-reported weight (kg) divided by self-reported height
caregiving and BMI were associated longitudinally. We (metre) squared was used to calculate the BMI. The BMI is a
hypothesise that informal caregiving is positively related to widely used body weight classification system (e.g. to quantify
BMI longitudinally in the total sample and in both sexes. excess weight) in adults.15
The knowledge about a longitudinal association between
informal caregiving and BMI is meaningful because this might Independent variables
stress consequences of informal caregiving. This is important
as an increased BMI is a risk factor for various physical ill- Several potential confounders were included in the current
nesses, including diabetes, or cardiovascular diseases.10 study. Age, marital status (married and living together with
Furthermore, an increased BMI is associated with reduced spouse and others [married and living separately, divorced,
quality of life and social exclusion.11 In total, this might help widowed and single]) and occupational status (employed,
to create interventional strategies for informal caregivers. retired and others [not employed]). Based on the International
Physical Activity Questionnaire,16 the frequency of sports ac-
tivities was quantified by asking ‘How often do you do sports
Methods such as hiking, soccer, gymnastics, or swimming?’ (daily;
several times a week; once a week; 1e3 times a month; less
Sample often and never). This item was validated elsewhere.17 The
widely used and well-validated Center for Epidemiologic
Data were drawn from the wave 2 (2002), wave 3 (2008), wave 4 Studies Depression Scale18 was used to measure depressive
(2011) and wave 5 (2014) of the German Ageing Survey (DEAS, symptoms. In addition, the sum of physical illnesses (e.g.
‘Deutscher Alterssurvey’). It is a nationally representative bladder problems; stomach and intestinal problems; cardiac
cross-sectional and longitudinal survey of the community- and circulatory disorders; joint, bone, spinal or back problems;
dwelling population in the second half of life (40 years and cancer and diabetes, ranging from 0 to 11) was used as an
above) in Germany organised by the German Centre of explanatory variable. The presence of providing informal care
Gerontology (DZA, ‘Deutsches Zentrum für Altersfragen’) and was assessed with the item ‘Are there people you look after or
funded by the Federal Ministry for Family Affairs, Senior Cit- care for regularly due to their poor state of health, either on a
izens, Women and Youth (BMFSFJ). The first wave took place private or volunteer basis?’ (no; yes).
in 1996. The DEAS study has a cohort-sequential longitudinal Afterwards, respondents were asked for whom they pro-
design with a wide thematic spectrum (e.g. social exclusion, vide support (e.g. father-in-law/partner's father, mother, fa-
occupation, psychological factors or meaning of ageing). ther, mother-in-law/partner's mother, spouse/partner and
 p u b l i c h e a l t h 1 5 1 ( 2 0 1 7 ) 8 1 e8 6 83

variousdoccasionally reporteddother options). Intragenera- The statistical significance was determined with P < 0.05.
tional care was defined as providing care for spouse/partner. Stata 14.0 (StataCorp, College Station, Texas, USA) was used to
Intergenerational caregiving was defined as follows: providing perform analysis.
care for father, father-in-law/partner's father, mother and
mother-in-law/partner's mother. These two variables were
used for sensitivity analysis (instead of informal caregiving in Results
general).
Because there is an evidence that BMI is also associated Sample characteristics
with self-esteem,19 the main model was extended by adding
this variable in sensitivity analysis. This factor was not Descriptive characteristics (pooled) of the observations
included in the main analysis, as endogeneity bias is likely to included in FE regression analysis (n ¼ 20,850) are depicted in
occur (e.g. BMI affecting self-esteem).20,21 Self-esteem was Table 1. Among others, nearly half of the observations were
quantified using the well-validated and widely used Rosen- female (49.0%) and mean age was 63.2 years (±11.4 years;
berg scale.22 We also tested whether the effect of informal 40e95 years). 72.3% were married and living together with
caregiving on BMI is moderated by age in the total sample and spouse. About half of observations were retired (52.0%) and
in both sexes. For descriptive purposes, education23 (Interna- 14.5% provided informal care. The average BMI was 26.7 kg/m2
tional Standard Classification of Education [ISCED-97], dis- (±4.4 kg/m2; 14.8e69.1 kg/m2).
tinguishing between low [0e2], medium [3e4] and high [5e6]
education) was also used. Regression analysis

Statistical analysis The results of the FE regression analysis with the BMI as
outcome measure are depicted in Table 2 (first column: total
Generally, a potential threat of cross-sectional regression sample; second column: men and third column: women).
analysis is self-selection, i.e. individuals decide according to
their characteristics (e.g. optimism) whether they provide
informal care or not. Most often, individuals differ in omitted
variables that affect explanatory variables as well as the Table 1 e Characteristics of observations included in
outcome variable. As a consequence, regression techniques fixed effects regression analysis (wave 2 to wave 5,
that do not take these unobserved factors into account are pooled; n ¼ 20,850).
biased.24 Variables N (%)/mean (SD)
For example, more optimistic individuals might select into Female: N (%) 10,219 (49.0)
informal caregiving and have higher satisfaction with life. Education (ISCED-97): N (%)
When optimism is not observed and between-comparisons Low education 1355 (9.3)
are conducted, the resulting association between informal Medium education 7679 (52.6)
High education 5566 (38.1)
caregiving and higher life satisfaction is biased by optimism
Age (in years): mean (SD); range 63.2 (±11.4); 40e95
(and other unobserved factors).
Married, living together with 15,068 (72.3)
When within-estimators are used (such as the fixed effects spouse: N (%)
[FE] estimator), solely intraindividual changes over time are Employment status: N (%)
exploited. Therefore, self-selection does not bias the FE-esti- Working 7611 (36.5)
mates.25 For example, let us assume that satisfaction with life of Retired 10,851 (52.0)
an individual is higher after starting informal caregiving and Other (not employed) 2388 (11.5)
Frequency of sports activities: N (%)
that optimism is constant within individuals over time. Because
Daily 1626 (7.8)
only changes within individuals over time are used in FE Several times a week 4956 (23.8)
regression, changes in satisfaction with life can be ascribed to Once a week 3674 (17.6)
changes in informal caregiving, even though optimism is un- 1e3 times a month 1607 (7.7)
observed. Under the assumption of strict exogeneity, FE re- Less than 1e3 times a month 2513 (12.1)
gressions lead to unbiased estimates even if time-invariant Never 6474 (31.0)
Total number of physical diseases: 2.4 (±1.8); 0e11
factorsdunobserved and observeddexist and are systemati-
mean (SD); range
cally associated with explanatory variables.24 Consequently, FE
Depressive symptoms (CES-D): 6.5 (±5.9); 0e45
regressions were used in the current study. The SarganeHansen mean (SD); range
test (Hausman test with cluster-robust standard errors) sup- Providing informal care: N (%) 3026 (14.5)
ported the choice of using FE regression analysis. Please refer Self-esteem (Rosenberg scale): 3.4 (±0.4); 1.2e4
the study by Cameron and Trivedi for technical details.24 mean (SD); range
It is worth noting that individuals were only included in FE BMI (kg/m2): mean (SD); range 26.7 (±4.4); 14.8e69.1

regressions when they had changes in informal caregiving BMI, body mass index; CES-D, Center for Epidemiologic Studies
between 2002 and 2014. For example, if an individual had Depression Scale; SD, standard deviation.
already been an informal caregiver prior to 2002 and his or her Note that the variables, sex and education were not included in FE
regressions as explanatory variables since these variables are time-
status remained unchanged afterwards, this individual did
invariant (i.e. they did not vary within individuals in old age). These
not contribute to our FE estimates (because of the absence of variables are only used to describe the sample.
within information).
84 p u b l i c h e a l t h 1 5 1 ( 2 0 1 7 ) 8 1 e8 6

Table 2 e Predictors of BMIdresults of fixed effects regressions (total sample; men; women).
Independent variables (1) BMIdtotal sample (2) BMIdmen (3) BMIdwomen
Age 0.06*** (0.01) 0.04*** (0.01) 0.08*** (0.01)
Other marital statuses (ref.: married, living together with spouse) 0.32** (0.12) 0.26þ (0.15) 0.46** (0.17)
Employment status: retired (ref.: working) 0.17þ (0.09) 0.10 (0.12) 0.26þ (0.14)
Other: not employed 0.10 (0.09) 0.17 (0.13) 0.09 (0.13)
Frequency of sports activities: several times a week (ref.: daily) 0.28*** (0.08) 0.35* (0.14) 0.21* (0.10)
Once a week 0.36*** (0.08) 0.42*** (0.12) 0.31** (0.10)
1e3 times a month 0.33*** (0.10) 0.55*** (0.14) 0.12 (0.13)
Less than 1e3 times a month 0.32*** (0.09) 0.30* (0.13) 0.38** (0.13)
Never 0.31*** (0.08) 0.34** (0.13) 0.31** (0.11)
Total number of physical diseases 0.05** (0.02) 0.06* (0.02) 0.05* (0.03)
Depressive symptoms 0.01** (0.00) 0.02* (0.01) 0.01 (0.01)
Informal care (ref.: not providing informal care) 0.09þ (0.05) 0.15* (0.07) 0.06 (0.07)
Constant 23.08*** (0.36) 24.54*** (0.53) 21.65*** (0.50)
Observations 20,850 10,631 10,219
Number of individuals 13,421 6861 6560
R2 0.04 0.03 0.06

BMI, body mass index.

Beta-coefficients are reported; cluster-robust standard errors in parentheses. ***P < 0.001, **P < 0.01, *P < 0.05, þP < 0.10.
Observations with missing values were dropped (listwise deletion). It is worth noting that the Stata command for FE regression analysis (‘xtreg,
fe’) include individuals with only one observation in calculating the number of observations because these individuals deliver information
about the variance components, the constant, the between R2 and so on. Nevertheless, it does not affect the beta-coefficients as well as the
standard errors.

Adjusting for potential confounders, FE regressions

showed that the onset of informal caregiving was not asso- Discussion
ciated with changes in BMI in the total sample and in women,
whereas the onset of informal caregiving was associated with Based on a representative sample of community-dwelling
an increase in BMI in men (b ¼ 0.15, P < 0.05). individuals aged 40 years and above, the aim of the present
Besides, an increase in BMI was positively associated with study was to provide insights into the longitudinal relation-
ageing (total sample, b ¼ 0.06, P < 0.001; men, b ¼ 0.04, P < 0.001 ship between transitions in informal caregiving and changes
and women, b ¼ 0.08, P < 0.001), an increase in morbidity (total in BMI. Extending previous cross-sectional studies, the asso-
sample, b ¼ 0.05, P < 0.01; men, b ¼ 0.06, P < 0.05 and women, ciation between informal caregiving and BMI was examined
b ¼ 0.05, P < 0.05) and a decrease in the frequency of sports using a longitudinal approach. Previous large cross-sectional
activities in the total sample and in both sexes (e.g. from studies have shown that informal caregiving is positively
‘daily’ to ‘several times a week’: total sample, b ¼ 0.28, associated with BMI.6e8 However, self-selection is likely to
P < 0.001; men, b ¼ 0.35, P < 0.05 and women, b ¼ 0.21, P < 0.05). occur in cross-sectional studies.25 Time-constant, unobserved
While changes from ‘married, living together with spouse’ factors might bias cross-sectional regression analysis.25
to other marital statuses were associated with a decrease in Overcoming these shortcomings, FE regressions used in our
BMI in the total sample (b ¼ 0.32, P < 0.01) and in women study demonstrated that transitions to informal caregiving
(b ¼ 0.46, P < 0.01), and an increase in depressive symptoms were associated with an increase in BMI in men. We assume
was associated with a decrease in BMI in the total sample that starting informal caregiving is associated with increasing
(b ¼ 0.01, P < 0.01) and in men (b ¼ 0.02, P < 0.05). stress in the caregivers, which is not fully captured by con-
In sensitivity analysis (results not shown, but available trolling for depressive symptoms and comorbidity. Stress in
upon request), the main model was extended by adding self- turn is positively associated with alcohol intake and eating in
esteem. In terms of effect sizes and significance, findings the absence of hunger which are factors that are positively
remained virtually the same. For example, the onset of associated with BMI.26 Thus, it appears plausible that informal
informal caregiving was associated with an increase in BMI in caregiving is associated with increasing BMI longitudinally.
men (b ¼ 0.15, P < 0.05). Changes in BMI were not significantly However, it remains an open question why this association
associated with changes in self-esteem in the total sample was only significant in men. A possible explanation might be
and in both sexes. Furthermore, it is worth noting that the that people vary in their response to stress. Some increase
interaction term ‘informal caregiving  age’ did not reach energy intake but others start fasting.27,28 For example,
Kivima € ki et al.29 have found that high job strain and low job
statistical significance (neither in the total sample nor after
stratification by sex). control were associated with weight gain among overweight
The FE regressions showed that starting intergenerational men (BMI >27 kg/m2), whereas this association was not
caregiving was associated with an increase in BMI in men observed among women.
(b ¼ 0.26, P < 0.05), whereas it was not associated with changes The present study has several strengths. Extending previ-
in BMI in the total sample and in women. In addition, starting ous knowledge about a cross-sectional association between
intragenerational caregiving was not associated with changes BMI and informal caregiving, we provide first evidence that
in BMI in the total sample and in both sexes. changes in informal caregiving and changes in BMI are
 p u b l i c h e a l t h 1 5 1 ( 2 0 1 7 ) 8 1 e8 6 85

associated in men longitudinally. Data were drawn from a 4. Hajek A, Ko € nig H-H. Informal caregiving and subjective well-
large representative sample of individuals in the second half being: evidence of a population-based longitudinal study of
of life (2002e2014). Various potential confounders such as the older adults in Germany. J Am Med Dir Assoc 2016;17:300e5.
5. Rice MC, Wicks MN, Martin JC. Health risk characteristics of
frequency of sports activities or depressive symptoms were
black female informal caregivers on welfare. Clin Nurs Res
included in regression analysis. The FE regressions were used 2008;17:20e31.
to remedy the problem of unobserved heterogeneity which is 6. Do EK, Cohen SA, Brown MJ. Socioeconomic and demographic
a main challenge in non-experimental research. factors modify the association between informal caregiving
However, the current study has also several limitations. and health in the Sandwich Generation. BMC Public Health
Self-rated health was used, suggesting that BMI is biased 2014;14:1.
downwards.30 Moreover, body fat percentage was not 7. Yiengprugsawan V, Harley D, Seubsman S-a, Sleigh AC.
Physical and mental health among caregivers: findings from a
assessed in the DEAS study. Also for reasons of data avail-
cross-sectional study of Open University students in
ability, frequency of alcohol consumption and energy intake Thailand. BMC Public Health 2012;12:1.
were not included in FE regression analysis. Even though self- 8. Lee S, Colditz GA, Berkman LF, Kawachi I. Caregiving and risk
selection bias is likely to occur in the DEAS study, it has been of coronary heart disease in US women: a prospective study.
demonstrated that this bias might be rather small.13 In addi- Am J Prev Med 2003;24:113e9.
tion, panel attrition cannot be ruled out. However, it has been 9. Brüderl J, Ludwig V. Fixed-effects panel regression. In: Wolf C,
editor. The Sage handbook of regression analysis and causal
shown that this is not a main problem in the German Ageing
inference. Los Angeles: SAGE; 2015. p. 327e57.
Survey.14 In addition, the type of disease of the care recipient
10. Must A, Spadano J, Coakley EH, Field AE, Colditz G, Dietz WH.
is unknown. Furthermore, we did not know about the exact The disease burden associated with overweight and obesity.
timing of starting informal caregiving. Further analysis is JAMA 1999;282:1523e9.
necessary to clarify adaptation in BMI following informal 11. Sutin AR, Stephan Y, Carretta H, Terracciano A. Perceived
caregiving. Adaptation effects have, e.g. been shown in sub- discrimination and physical, cognitive, and emotional health
jective well-being research.31 in older adulthood. Am J Geriatric Psychiatry 2015;23:171e9.
12. Neller K. Kooperation und Verweigerung. Eine Non-Response-
The present study demonstrated that the onset of informal
Studie. ZUMA Nachrichten 2005;29:9e36.
caregiving is associated with an increase in BMI in men, but 13. Klaus D, Engstler H. Daten und Methoden des Deutschen
not women. Thus, it might be fruitful to generate weight Alterssurveys. In: Mahne K, Wolff JK, Simonson J, Tesch-
management strategies specifically designed for male Ro€ mer C, editors. Altern im Wandel: Zwei Jahrzehnte Deutscher
informal caregivers, e.g. respite care might be helpful in Alterssurvey (DEAS). Berlin; 2016. p. 25e42.
reducing stress symptoms of private caregivers. 14. Schiel S, Kerr P, Dickmann C, Aust F. Methodenbericht zur
Durchführung der fünften Erhebungswelle 2014. Bonn: infas
Institut für angewandte Sozialwissenschaft GmbH; 2015.
15. Flegal KM, Shepherd JA, Looker AC, Graubard BI, Borrud LG,
Author statements Ogden CL, et al. Comparisons of percentage body fat, body
mass index, waist circumference, and waist-stature ratio in
Ethical approval adults. Am J Clin Nutr 2009;89:500e8.
16. Craig CL, Marshall AL, Sjorstrom M, Bauman AE, Booth ML,
None sought. Ainsworth BE, et al. International physical activity
questionnaire: 12-country reliability and validity. Med Sci
Sports Exerc 2003;35:1381e95.
Funding
17. Wurm S, Tomasik MJ, Tesch-Ro € mer C. On the importance of a
positive view on ageing for physical exercise among middle-
None declared. aged and older adults: cross-sectional and longitudinal
findings. Psychol Health 2010;25:25e42.
Competing interests 18. Radloff LS. The CES-D scale a self-report depression scale for
research in the general population. Appl Psychol Meas
None declared. 1977;1:385e401.
19. Yusufov M, Dalrymple K, Bernstein MH, Walsh E,
Rosenstein L, Chelminski I, et al. Body mass index,
references depression, and suicidality: the role of self-esteem in bariatric
surgery candidates. J Affect Disord 2017;208:238e47.
20. Huang C-C, Yabiku ST, Kronenfeld JJ. The effects of household
technology on body mass index among Chinese adults. Popul
1. Hajek A, Lehnert T, Wegener A, Riedel-Heller SG, Ko € nig H-H.
Res Pol Rev 2015;34:877e99.
Factors associated with preferences for long-term care 21. Kinge JM. Waist circumference, body mass index, and
settings in old age: evidence from a population-based survey employment outcomes. Eur J Health Econ 2016:1e13.
in Germany. BMC Health Serv Res 2017;17:156. 22. Rosenberg M. Society and the adolescent self-image. Princeton:
2. Hajek A, Lehnert T, Wegener A, Riedel-Heller SG, Ko € nig H-H.
Princeton University Press; 1965.
€
Informelles Pflegepotenzial bei Alteren in Deutschland. Z 23. UNESCO. International Standard Classification of Education. ISCED
Gerontol Geriatr 2017:1e7. 1997. Re-edition ed. Paris: UNESCO; 2006.
3. Roloff J. Pflege ist weiblichegeschlechterdifferenzierte 24. Cameron AC, Trivedi PK. Microeconometrics: methods and
Betrachtung des Pflegebedarfs und Pflegepotenzials. applications. New York: Cambridge University Press; 2005.
Bev€
olkerungsforsch Mittl aus dem Bundesinstitut für 25. Brüderl J. Applied panel data analysis using Stata. Munich. 2015.
Bev€
olkerungsforsch 2009;30:6e10.
86 p u b l i c h e a l t h 1 5 1 ( 2 0 1 7 ) 8 1 e8 6

26. Payette H, Gray-Donald K, Cyr R, Boutier V. Predictors of bidirectional effects of job strain on body mass index in the
dietary intake in a functionally dependent elderly population Whitehall II study. Int J Obes 2006;30:982e7.
in the community. Am J Public Health 1995;85:677e83. 30. Gorber SC, Tremblay M, Moher D, Gorber B. A comparison of
27. Hickson M. Malnutrition and ageing. Postgrad Med J direct vs. self-report measures for assessing height, weight
2006;82:2e8. and body mass index: a systematic review. Obes Rev
28. Rosenbloom CA, Whittington FJ. The effects of bereavement 2007;8:307e26.
on eating behaviors and nutrient intakes in elderly widowed 31. Powdthavee N. What happens to people before and after
persons. J Gerontol 1993;48:S223e9. disability? Focusing effects, lead effects, and adaptation in
29. Kivima € ki M, Head J, Ferrie J, Shipley M, Brunner E, Vahtera J, different areas of life. Soc Sci Med 2009;69:1834e44.
et al. Work stress, weight gain and weight loss: evidence for