Self-Reported Health in Mothers: The Impact of Age, and Socioeconomic Conditions

This article was downloaded by: [University of Stockholm]On: 30 January 2012, At: 08:14Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Women & HealthPublication details, including instructions for authorsand subscription information:http://www.tandfonline.com/loi/wwah20

Self-Reported Health inMothers: The Impact of Age, andSocioeconomic ConditionsBirgitta Floderus PhD a , Maud Hagman a , GunnarAronsson PhD b , Staffan Marklund PhD c & AndersWikman PhD da Department of Public Health Sciences, NASP,Karolinska Institutet, Stockholm, SE-171 77, Swedenb Department of Psychology, Stockholm University,Stockholm, SE-106 91, Swedenc Department of Clinical Neuroscience, Section ofPersonal Injury Prevention, Karolinska Institutet,Stockholm, SE-171 77, Swedend Department of Social Science, Mälardalen University,Box 883, Västerås, SE-721 23, Sweden

Available online: 21 Oct 2008

To cite this article: Birgitta Floderus PhD, Maud Hagman, Gunnar Aronsson PhD, StaffanMarklund PhD & Anders Wikman PhD (2008): Self-Reported Health in Mothers: The Impactof Age, and Socioeconomic Conditions, Women & Health, 47:2, 63-86

To link to this article: http://dx.doi.org/10.1080/03630240802092308

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions

This article may be used for research, teaching, and private study purposes.Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expresslyforbidden.

The publisher does not give any warranty express or implied or make anyrepresentation that the contents will be complete or accurate or up to date. Theaccuracy of any instructions, formulae, and drug doses should be independentlyverified with primary sources. The publisher shall not be liable for any loss,actions, claims, proceedings, demand, or costs or damages whatsoever orhowsoever caused arising directly or indirectly in connection with or arising outof the use of this material.

Dow

nloa

ded

by [

Uni

vers

ity o

f St

ockh

olm

] at

08:

14 3

0 Ja

nuar

y 20

12

Self-Reported Health in Mothers:The Impact of Age,

and Socioeconomic Conditions

Birgitta Floderus, PhDMaud Hagman

Gunnar Aronsson, PhDStaffan Marklund, PhDAnders Wikman, PhD

ABSTRACT. The aim of the present analysis was to study health andwell-being in mothers compared to women without children, and to an-alyze potential interactions with age and socioeconomic conditions.

The study comprised 5,368 Swedish women born in 1960-1979 whowere interviewed in any of the population-based Surveys of Living

Birgitta Floderus is Professor and Maud Hagman is Biostatistician; both are affiliatedwith the Department of Public Health Sciences, NASP, Karolinska Institutet, SE-171 77Stockholm, Sweden.

Gunnar Aronsson is Professor, Department of Psychology, Stockholm University,SE-106 91 Stockholm, Sweden.

Staffan Marklund is Professor, Department of Clinical Neuroscience, Section ofPersonal Injury Prevention, Karolinska Institutet, SE-171 77 Stockholm, Sweden.

Anders Wikman is Associate Professor, Department of Social Science, MälardalenUniversity, Box 883, SE-721 23 Västerås, Sweden.

Address correspondence to: Birgitta Floderus at the above address (E-mail:[email protected]).

The study was conducted at the National Institute for Working Life (NIWL), whichwas closed down on June 30, 2007. All authors were previously affiliated with theNIWL.

The authors thank Stefan Vikenmark and Maylis Larsson at Statistics Sweden, forskillful preparation of the raw data files. Contract grant sponsor: Swedish Council forWorking Life and Social Research, grant no 2004-1101.

Women & Health, Vol. 47(2) 2008Available online at http://wh.haworthpress.com

© 2008 by The Haworth Press. All rights reserved.doi: 63 10.1080/03630240802092308

Dow

nloa

ded

by [

Uni

vers

ity o

f St

ockh

olm

] at

08:

14 3

0 Ja

nuar

y 20

12

Conditions during the period 1996-2003. Having children at home was re-lated to self-reported health symptoms and long-standing illness in multi-ple logistic regression models. The impact of age, cohabitation status,full-time or part-time work, and income level were analyzed.

While mothers were less burdened by long-standing illness, partlydue to selection mechanisms (a “healthy mother effect”), they experi-enced worse self-rated health and more fatigue than women withoutchildren, and the odds of poor self-rated health and fatigue increasedby number of children. Conditions that strengthened the association be-tween motherhood and impaired health were young maternal age, full-time employment, high income, and being alone.

The study indicates a need for improved negotiations between parentsregarding a fair share of work and family duties and extended support forlone mothers to prevent adverse health effects in women combining chil-dren and work. The results may be useful to policy-makers and employersin developing new policies. doi: [Article copies avail-able for a fee from The Haworth Document Delivery Service: 1-800-HAWORTH.E-mail address: <[email protected]> Website: <http://www.HaworthPress.com> © 2008 by The HaworthPress. All rights reserved.]

KEYWORDS. Children, effect modification, interaction, health, socio-economic status, women, work-family

INTRODUCTION

Women’s entry into the labour force is one of the most important so-cial changes of the last century. During the period 1963-2002, the num-ber of working women with paid work in Sweden increased from 53 to73%, while the proportion decreased among men from 89 to 76%. Simul-taneously, women’s mean age at first birth increased from 24 years in1970 to 29 years in 2004 (Statistics Sweden, 2007). The time trendsmay to some extent be interrelated due to difficulties for women in com-bining parenthood and vocational career.

In Sweden, the possibilities for both men and women to combine par-enthood and work are better than in most other countries, which havecontributed to a dramatic decrease in the fraction of homemakers. Ac-cording to Korpi and Stern (2006), 62% of women aged 20-50 yearswith children below seven years of age were homemakers in 1968. In2000 the percentage was nine (women with children below nine yearsof age). The accessibility to subsidized childcare, mostly high quality

64 WOMEN & HEALTH

10.1080/03630240802092308

Dow

nloa

ded

by [

Uni

vers

ity o

f St

ockh

olm

] at

08:

14 3

0 Ja

nuar

y 20

12

day-care centres is high, and about 73% of children between ages oneand five attend some form of public childcare. Furthermore, socioeco-nomic policies promote combining work and parenthood for both menand women by high acceptance of part-time work and by separate taxa-tion. In 1974, maternity leave was changed into parental leave, and to-day both parents share 12 months of paid leave for each child (and bothare subsidized for leave for sick children). Nevertheless, women takeabout 88% of the parental leave (Korpi & Stern, 2006).

Several researchers have suggested that multiple roles, also ex-pressed as double exposure, double shift, paid and unpaid work, may bedetrimental to the individual’s health and well-being due to time andstrain-based overload or conflicts, the role strain hypothesis (Goode,1960; Gove, & Geerken, 1977; Lundberg, 1996; Krantz, & Östergren,2001; Nordenmark, 2004; Krantz, Berntsson, & Lundberg, 2005; Chan-dola, Martikainen, Bartley, et al., 2004), while others have advocatedthat multiple roles entail beneficial effects, the role enhancement hy-pothesis, by increased stimulation, and extended access to environmen-tal resources and social affirmation (Thoits, 1983; Crossby, 1987). Itseems evident that one mechanism does not rule out the other, and thatcurrent research should aim at identifying conditions counteracting pos-itive effects of multiple roles in both women and men.

Others have studied the interaction between work and family life interms of work-to-family and family-to-work conflict, or positive andnegative “spillover” (interference) from one sphere to the other (Frone,2000; Grzywacz, 2000; Grzywacz, & Marks 2000; Hammer, Saksvik,Nytro, et al., 2004; Vaananen, Kevin, Ala-Mursula, et al., 2004; Emslie,Hunt, & Macintyre, 2004). The evidence of a link between negativework-to-family spillover and impaired health is sometimes strong(Vaananen, Kevin, Ala-Mursula, et al., 2004), but also, positive spill-over from work to family has been associated with better physical andmental health (Grzywacz, 2000). Furthermore, in the latter study, posi-tive spillover from family to work was associated with less chronic con-ditions, as well as better mental health and more positive well-being. Noconsistent evidence is available that family-to-work conflicts should bea stronger risk factor for depression or poor physical health in men thanin women, and conversely, that work-to-family conflicts should be amore important health risk factor in women than in men, due to their as-sumed priorities of the homemaker and breadwinner role, respectively(Frone, Russel, & Barnes, 1996). Cinamon and Rich (2002) have em-phasized the importance of within gender variation besides betweengender variation regarding perceptions of work and family roles, and

Floderus et al. 65

Dow

nloa

ded

by [

Uni

vers

ity o

f St

ockh

olm

] at

08:

14 3

0 Ja

nuar

y 20

12

the individual’s role priorities. They also summarize that some but notall studies, suggest work-family conflicts to be more frequent in womenthan in men.

Despite the extensive research on multiple roles, the knowledge onhealth effects is sparse. A better understanding of the lives of young cou-ples with children and the lives of lone parents based on representativesamples is warranted to narrow the scope to feasible targets for employ-ment policies and health prevention activities.

In the present study, we focused on the relation between motherhoodand health in a population-based sample of women born in the 1960s and1970s, with motherhood and number of children as measures of demandsoutside working life. Outcome variables included self-reported globalhealth, sleep problems, fatigue, anxiety symptoms, and long-standingillness–any long-standing illness and the subgroups of musculoskeletaland psychiatric disorders. The main objective was to explore health andwell-being in mothers compared to women without children and toidentify socioeconomic conditions that affected the relation of moth-erhood to health, that is, potential effect modification (interaction) withsocioeconomic conditions. The variables analyzed were: age, cohabi-tation status, full-time or part-time work, and income level. We alsocontrolled for some lifestyle factors and tried to reduce the effect of possi-ble selection factors based on health.

METHODS

Source Population and Study Group

The source population comprised all women, 20 years or older, bornin 1960 through 1979 and registered as Swedish residents (December31) at least once during the period 1990-2003 (N = 1,255,201).

The present study group included all women belonging to the sourcepopulation who, during 1996-2003, participated in any of the SwedishSurveys of Living Conditions (ULF). The surveys that are authorizedby the Swedish parliament are carried out by Statistics Sweden (Statis-tics Sweden, 2007). The data collection, covering a broad range of liv-ing conditions, is based on face-to-face interviews of population-basedrandom samples. The non-participation rate among women, aged 16-44years, during 1994-2003, varied between 15.8 and 21.9%, with anaverage of 19.4% (Statistics Sweden, 1997). If a woman was included

66 WOMEN & HEALTH

Dow

nloa

ded

by [

Uni

vers

ity o

f St

ockh

olm

] at

08:

14 3

0 Ja

nuar

y 20

12

in more than one of the eight surveys, we used data only from the firstinterview. The number of women with data from an ULF investigationwas in total 5,368 (Table 1).

Study Variables

Exposure Variables

Self-reported motherhood implied presence of residential children 18years or younger (yes/no), and number of residential children (one, two,three or more vs. no children).

Floderus et al. 67

TABLE 1. Distribution of Women by Age, Socioeconomic Conditions, andNumber of Children

Notes: *Younger = 20-29 years; Older = 30-43 years. Lone = not married and not living with partner inconsensual union;Cohabiting = married or living with partner in consensual union. Part-time: �40 hrs/week; Full-time:�40 hrs/week.High income: �upper quartile (qtl); Med income: between lower-upper qtl; Low income: �lower qtl.Metropolitan areas: Stockholm, Gothenburg, Malmö. City areas: municipalities with �90,000 inhab. within

30 km from the centre.

Dow

nloa

ded

by [

Uni

vers

ity o

f St

ockh

olm

] at

08:

14 3

0 Ja

nuar

y 20

12

Outcome Variables

Self-rated health: “How do you rate your general health condition? Isit: very good/good/in between/bad/very bad?” (e.g., Idler & Kasl, 1991). Inbetween, bad, and very bad were categorized as “poor” self- rated health,and was compared with “good” health (good/very good).

Fatigue meant remarkable tiredness during the daytime during thelast two weeks: “Have you felt remarkably tired during daytime? yes/no.” We considered tiredness in the morning, at night, and tiredness ingeneral, but discarded these variables because they were reported morefrequently and may to a larger extent have reflected the normal circa-dian rhythm. Sleep problems also referred to the last two weeks: “Haveyou had problems with sleep? yes/no.”

Anxiety symptoms were measured by the following: “I have alreadyasked about diseases but to be sure I would like to ask you about somediseases or disorders specifically. Have you possibly had any of the fol-lowing: Anxiety, restlessness or anguish? yes severe/yes light/no”; se-vere and light disorders were combined.

Long-standing illness was based on an open-ended question: “Doyou have any longstanding illness, disorder from an accident, any dis-ability, or other ailment?” Follow-up questions were added to achieve amore accurate coding of diagnoses. For example, if the respondent gavea negative answer to this question, but later answered positively regard-ing medication, the interviewer went back to the initial question. Theanswers were coded by trained staff at Statistics Sweden according tothe WHO International Classification of Disease, 8th revision (ICD8).Any long-standing illness was compared with no such diagnosis. Themost common diagnoses were related to the musculoskeletal system,the respiratory system, injuries or accidents, the nervous system, diges-tive organs, and psychiatric disorders. Three diagnostic categories werestudied: All long-standing illness, Psychiatric disorders (ICD8 code290.0-316.9), and Musculoskeletal disorders (ICD8 code 710.0-739.9).The latter categories were chosen because today, they make up a largefraction of sick leave in Sweden.

Stratification Variables

Age was dichotomized into 30 years or younger (younger), and 31 yearsor older (older). The cut off at age 30 was chosen to obtain sufficientnumbers in the two strata. In the analyses of socioeconomic conditions, we

68 WOMEN & HEALTH

Dow

nloa

ded

by [

Uni

vers

ity o

f St

ockh

olm

] at

08:

14 3

0 Ja

nuar

y 20

12

also explored age-stratified results based on the dichotomy. Throughoutwe adjusted for age as a continuous variable (one-year intervals).

Cohabitation status was dichotomized into lone and cohabiting wo-men. Cohabiting women were included with married women. Lonewomen were those who lived without a partner. Individuals still livingwith their parents (n = 313) were excluded.

The analysis of full-time and part-time work was restricted to womenwith a contract of permanent or temporary employment, excluding, forexample, farmers and self-employed to increase the socioeconomic ho-mogeneity within and between the groups (control for potential con-founding) while studying the effect of work hours.

Data on income and residential area were added from a national de-mographic registry through linkage using personal identification num-bers of survey respondents. Income pertained to the woman’s salary,social benefits and financial compensations, and other family incomewas not included. Cut off points were set to the first and third quartile,yielding three strata: low, medium, and high income. The quartiles werebased on the year specific distribution.

Place of residence at year of interview was classified into metropoli-tan areas (Stockholm, Gothenburg, Malmö with surrounding munici-palities), city areas (municipalities with more than 90,000 inhabitantswithin 30 kilometers from the centre), and all other areas.

Place of residence was analyzed because we believed that metropoli-tan areas should be less “mother friendly” than other areas, due to moretime stress and possibly more social isolation among both lone and co-habiting mothers.

Confounding and Selection Mechanisms

All analyses were adjusted for year of interview to control for potentialconfounding from secular trends.

Some main results were analyzed with adjustment for health-relatedlifestyle factors: smoking and body mass index (BMI = weight in kilosdivided by the square of height in metres). BMI was based on self-re-ported height and weight, with overweight corresponding to a ratio of25.0 or more, underweight less than 18.5, values in between were usedas reference (normal weight). Smoking was dichotomized: current orprevious regular (daily) smoking or current occasional smoking con-trasted with never regular smoking and no current occasional smoking.

Floderus et al. 69

Dow

nloa

ded

by [

Uni

vers

ity o

f St

ockh

olm

] at

08:

14 3

0 Ja

nuar

y 20

12

At an early stage we found that long-standing illness was less frequentlyreported by women with children, and simultaneously more frequentlyreported by women with poor self-rated health. This caused us to believethat the study could be biased by a “healthy mother effect,” that womenwho became mothers, à priori, were less likely to suffer from more se-vere or long-standing illness than women with no children. This ques-tion was further explored: we restricted the study group to women withno children at the time of interview (n = 2,738) and compared womenwho subsequently had children (779) with women who still had no chil-dren at the end of 2003, with respect to their previous reports of anylong-standing illness, psychiatric diagnoses, musculoskeletal diagnoses,and self-rated health, adjusting for age and year of interview. For theseprospective analyses, data on children up to 2003 were added from thenational demographic registry.

Long-standing illness in general, as well as psychiatric diagnoses wereassociated with increased odds of not having children during follow-up(OR = 1.24, 95% CI 1.01-1.56; OR = 1.65, 95% CI 0.95-2.86), whilemusculoskeletal diagnoses showed no such effect (OR = 0.81, 95% CI0.58-1.12); neither did poor self-rated health affect the odds of havingchildren (OR = 0.91, 95% CI 0.69-1.19). The results caused us to con-trol for long-standing illness in the analyses of self-rated health, sleepproblems, fatigue, and anxiety symptoms.

Statistical Methods

We did not use data driven-model building, but made up a scheme ofanalysis à priori, according to the aim of the study. The associations be-tween exposure and outcome variables were based on pooled data fromthe ULF interviews. The analyses were conducted using multiple logisticregression (SAS statistical package (version 9.1.3. SAS Institute, Cary,North Carolina, USA 2002-2003), according to three models summa-rized as follows:

• Model 1 provided adjusted ORs with corresponding 95% confi-dence intervals (CI) for health outcome variables depending on ex-posure.

• Health outcome = Exposure � Age (continuous) � Year of inter-view � Long-standing illness.

The exposure was either having children (yes/no), or number of children(i.e., one, two, three or more children) all with no children as reference.

70 WOMEN & HEALTH

Dow

nloa

ded

by [

Uni

vers

ity o

f St

ockh

olm

] at

08:

14 3

0 Ja

nuar

y 20

12

The same covariates were used throughout, although long-standing ill-ness was excluded when this was the health outcome variable. Model 1was used for each stratum of the stratification variables: Younger andOlder, Lone and Cohabiting, Part-time and Full-time workers, womenwith High and Medium and Low income, respectively.

• Model 2 was used to obtain a statistical testing of potential effectmodification across strata (interaction). The tests answered thequestions: Is the effect of having children significantly different ifyou are young or old, lone or cohabiting, a part-time or full-timeworker, or if you have a high or moderate or low income?

• Health outcome = Exposure � Age (continuous) � Year of inter-view � Long-standing illness � Stratification variable � Expo-sure * Stratification variable.

The exposure was having children (yes/no). Model 2 was used for thestratification variables described above, and the test referred to thep-value of the cross-product of the exposure and the stratification vari-able in question.

• Model 3 yielded a statistical test for trend. It was used to estimate ifit made a difference if you had one, two, or three or more childrencompared to no children.

• Model was the same as model 1.

The exposure was number of children on a scale from 0 through 3. Thetest for trend referred to the p-value of this variable. Model 3 was usedfor each stratum of the stratification variables.

Thus, we explored the associations between motherhood and healthwithin each stratum to get a high degree of homogeneity within groups,probably yielding additional control of unmeasured confounding (apartfrom the covariates) in the stratum specific results (Model 1 and 3) andfurther, by Model 2 we assessed potential effect modification (inter-action) across strata with control for age, year of interview and long-standing illness.

Ethical Approval

Approval to use the data for this study was obtained by Statistics Swe-den, working in compliance to the principles of the Helsinki Declaration.The present study was further approved by the Research ethics committeelocated at the Karolinska Institutet, Stockholm (www.epn.se/eng/start/),

Floderus et al. 71

Dow

nloa

ded

by [

Uni

vers

ity o

f St

ockh

olm

] at

08:

14 3

0 Ja

nuar

y 20

12

(Dnr 2005/32-31). The data obtained from Statistics Sweden did notcontain any personal identifiers.

RESULTS

A total of 1,726 (32%) women had no contract of employment, in-cluding students, involuntarily unemployed, self-employed, farmers, andhomemakers (Table 1). The mean number of work hours reported byfull-time employees was 40 (SD 3.3) hours, and in part-time workers25 hours (SD 10.9).

Out of the 2,915 (54%) mothers, 30% were 20-29 years of age, 15%were lone, 35% were part-time workers, and 33% had an income abovethe upper quartile. Women without children were younger than moth-ers, 78% were between 20-29 years of age, more often lone (78%), lessoften part-time workers (21%), and less often in the high-income cate-gory (16%); 34% of mothers lived in metropolitan areas versus 43%among women without children (Table 1).

Self-Rated Health

The prevalence of poor self-rated health was 16.1%. Mothers per-ceived their health as poor more often than women without children.The effect modification by age was statistically significant, that is, theassociation between children and poor self-rated health was stronger inyounger than in older women. The OR for poor self-rated health increasedby number of children overall, and in both age strata (Tables 2 and 3).

Lone mothers had higher odds of poor health compared to lone womenwho were not mothers, and the OR increased with number of children.Cohabiting women had a similar result, but the effect of motherhoodamong cohabiting women was less pronounced than among lone women(Tables 2 and 3).

In part-time employees, no difference was observed in self-rated health(Table 2). Among full-time employees, mothers had higher odds of poorself-rated health compared to women without children (Table 2). TheOR increased with increasing number of children, overall, and in bothyounger and older full-time workers (Table 3). The effect of motherhoodamong full-time workers differed significantly from the effect amongpart-time workers (Table 2).

72 WOMEN & HEALTH

Dow

nloa

ded

by [

Uni

vers

ity o

f St

ockh

olm

] at

08:

14 3

0 Ja

nuar

y 20

12

TA

BLE

2.A

djus

ted

Odd

sR

atio

s(O

R)

with

95%

Con

fiden

ceIn

terv

als

for

Sel

f-R

ated

Hea

lth,S

leep

Pro

blem

s,F

atig

uean

dA

nxie

tyin

Wom

enw

ithan

dW

ithou

tChi

ldre

n,by

Age

and

Soc

ioec

onom

icC

ondi

tions

*

Not

es:*

n-va

lues

,see

Tab

le1.

**O

Rad

just

edfo

rye

arof

inte

rvie

w,a

ge(c

ontin

uous

),an

dlo

ng-s

tand

ing

illne

ss;p

-val

ues

refe

rto

effe

ctm

odifi

catio

n .

73

Dow

nloa

ded

by [

Uni

vers

ity o

f St

ockh

olm

] at

08:

14 3

0 Ja

nuar

y 20

12

Income was also a clear effect modifier. Motherhood was not associatedwith poor self-rated health among women in the low-income category. Thepoint estimate was increased in the medium-income category, and amongmedium-income earners the OR tended to increase with number of chil-dren. In the highest income group, poor self-rated health was clearly in-creased in mothers compared to women without children, and the ORincreased with number of children (Tables 2 and 3).

Other Self-Reported Symptoms

The prevalence of sleep problems was 21.7%. Motherhood was notassociated with sleep problems. However, the OR was slightly higher inyounger compared to older women. Lone mothers, and mothers working

74 WOMEN & HEALTH

OR** OR** OR** p**

All women 1.40 1.11 - 1.76 1.46 1.16 - 1.83 1.92 1.45 - 2.53 .0001Younger 1.44 1.04 - 1.98 1.87 1.28 - 2.72 2.91 1.66 - 5.09 .0001Older 1.23 0.88 - 1.73 1.17 0.87 - 1.59 1.53 1.09 - 2.15 .03

Lone 1.88 1.28 - 2.77 2.23 1.44 - 3.47 2.58 1.42 - 4.67 .0001Younger 2.15 1.24 - 3.74 3.52 1.62 - 7.67 1.96 0.48 - 8.06 .0004Older 1.73 1.00 - 3.00 1.87 1.07 - 3.27 2.65 1.32 - 5.33 .002Cohabiting 1.38 1.02 - 1.87 1.65 1.23 - 2.22 2.29 1.62 - 3.24 .0001Younger 1.40 0.92 - 2.13 2.01 1.26 - 3.20 3.88 2.02 - 7.47 .0001Older 1.16 0.73 - 1.84 1.26 0.84 - 1.90 1.66 1.06 - 2.58 .02

Part-time 1.05 0.67 - 1.66 0.81 0.52 - 1.27 0.86 0.49 - 1.50Younger 1.02 0.55 - 1.91 0.51 0.22 - 1.21 0.53 0.11 - 2.52Older 1.16 0.56 - 2.38 0.97 0.52 - 1.80 1.01 0.50 - 2.03Full-time 1.13 0.75 - 1.70 1.58 1.07 - 2.33 2.20 1.32 - 3.65 .001Younger 0.80 0.37 - 1.71 1.99 0.92 - 4.30 4.79 1.31 - 17.56 .03Older 1.26 0.74 - 2.13 1.49 0.92 - 2.39 1.94 1.10 - 3.44 .02

High income 2.00 1.13 - 3.52 2.68 1.65 - 4.34 3.54 2.07 - 6.06 .0001Younger 3.58 1.36 - 9.41 4.92 2.00 - 12.09 4.37 1.34 - 14.18 .0004Older 1.27 0.62 - 2.59 1.81 1.01 - 3.25 2.57 1.38 - 4.81 .001Med income 1.55 1.13 - 2.13 1.37 0.99 - 1.89 1.75 1.15 - 2.65 .01Younger 1.53 0.98 - 2.38 1.61 0.95 - 2.74 2.60 1.12 - 6.06 .007Older 1.51 0.95 - 2.40 1.20 0.78 - 1.85 1.49 0.89 - 2.48Low income 0.90 0.58 - 1.41 1.03 0.62 - 1.71 1.47 0.79 - 2.74Younger 0.94 0.55 - 1.63 1.32 0.63 - 2.78 3.20 1.08 - 9.48Older 0.74 0.33 - 1.64 0.75 0.37 - 1.53 0.89 0.39 - 2.01

95% CI 95% CI95% CI

Poor self-rated health

One vs no children Two vs no children > Three vs no children

TABLE 3. Adjusted Odds Ratios (OR) with 95% Confidence Intervals for Self-Rated Health in Women, According to Number of Children (No Children as Refer-ence), by Age and Socioeconomic Conditions*

Notes: *n-values, see Table 1.

**OR adjusted for year of interview, age (continuous), and long-standing illness; p-values refer to test for trend.

Dow

nloa

ded

by [

Uni

vers

ity o

f St

ockh

olm

] at

08:

14 3

0 Ja

nuar

y 20

12

full-time showed increased ORs of sleep problems. Significant differencesbetween strata were only observed for younger/older women and full-time/part-time workers (Table 2).

Fatigue (prevalence 34.3%) showed a pattern somewhat similar toself-rated health, but the associations were weaker. Overall, more moth-ers reported fatigue than women without children, and this was attrib-uted to younger women. Furthermore, an association was found amonglone women and (to a lesser degree) in cohabiting women, among womenwith full-time employment and in high-income earners. The differencesbetween strata were only significant for age and cohabitation status. Atrend of increasing OR for fatigue with increasing number of chil-dren was apparent overall, among younger women, among lone and co-habiting women, full-time workers, and high-income earners. (Tables 2and 4).

Anxiety symptoms were reported by 23.4% of women. Overall, theodds of anxiety symptoms were lower in mothers compared to womenwithout children, and the ORs decreased by number of children in olderwomen, part-time workers, low- and medium-income earners. Contraryto this pattern, associations between children and anxiety symptomswere apparent in lone mothers, and high-income earners, supported byincreasing ORs with number of children (Tables 2 and 5).

Long-Standing Illness

Long-standing illness was reported by 34.3%; the corresponding fig-ures for musculoskeletal and psychiatric disorders were 11.3 and 4.3%,respectively.

The odds of long-standing illness were lower among mothers com-pared to women without children; this pertained to both age groups.Cohabitation status was an effect modifier, mainly due to an OR mark-edly below unity among cohabiting women. ORs were also reduced forlong-standing illness for both part-time and full-time workers, with nomajor difference. The OR was below unity in the low- and medium-in-come categories, and close to unity in high-income earners, and the het-erogeneity across level of income was significant (Table 6).

We found no consistent evidence of an association between moth-erhood and musculoskeletal disorders. Only among younger lone wom-en, were children related to increased odds of musculoskeletal disorders(Table 6).

Mothers were less burdened by psychiatric disorders compared toother women. In older women, older cohabiting women, older part-time

Floderus et al. 75

Dow

nloa

ded

by [

Uni

vers

ity o

f St

ockh

olm

] at

08:

14 3

0 Ja

nuar

y 20

12

workers, and low-income earners (both younger and older), mothershad lower odds of psychiatric disorders compared to women withoutchildren, whereas higher odds were found among younger lone women,and high-income earners (based on small numbers). Statistically signifi-cant differences between the strata were found for all stratification vari-ables (Table 6).

Residential Area

The results for residential area were heterogeneous with few consis-tent patterns, and the results were omitted from the tables due to spaceconstraints. The highest OR was found for poor self-rated health amongyounger women in city areas, that is, not metropolitan areas (OR = 2.17;

76 WOMEN & HEALTH

TABLE 4. Adjusted Odds Ratios (OR) with 95% Confidence Intervals for Fatiguein Women, According to Number of Children (No Children as Reference), by Ageand Socioeconomic Conditions*

Notes: *n-values, see Table 1.

**OR adjusted for year of interview, age (continuous), and long-standing illness; p-values refer to test for trend.

Two

Dow

nloa

ded

by [

Uni

vers

ity o

f St

ockh

olm

] at

08:

14 3

0 Ja

nuar

y 20

12

95% CI = 1.42-3.31), and the OR increased by number of children,OR = 1.75 (95% CI = 1.05-2.91), 2.46 (95% CI = 1.35-4.46) and 4.10(95% CI = 1.84-9.15).

Health-Related Behavior

To test further the validity of the main results, we controlled for thetwo lifestyle factors considered: smoking and BMI. The results wereomitted from the tables because there were no major changes. For ex-ample, the ORs adjusted for smoking and BMI were for women withchildren compared to those with no children 1.48 (95% CI 1.22-1.79)and 0.66 (0.57-0.76) for poor self-rated health and long-standing illness,respectively.

Floderus et al. 77

TABLE 5. Adjusted Odds Ratios (OR) with 95% Confidence Intervals for AnxietySymptoms in Women, According to Number of Children (No Children as Refer-ence), by Age and Socioeconomic Conditions*

Notes: *n-values, see Table 1.

**OR adjusted for year of interview, age (continuous), and long-standing illness; p-values refer to test for trend.

Dow

nloa

ded

by [

Uni

vers

ity o

f St

ockh

olm

] at

08:

14 3

0 Ja

nuar

y 20

12

DISCUSSION

While mothers were less burdened by long-standing illness, partlydue to selection mechanisms (a “healthy mother effect”), they experi-enced poor global health, and fatigue more often than women withoutchildren. The results were supported by increasing odds with increasingnumber of children. Conditions that strengthened the association be-tween motherhood and impaired health were young age, full-time em-ployment, high income, and being lone.

We have not found any study within work-family research focusingpotential health effects of having children. The unpaid work, tradi-tionally and still, mainly shouldered by women (also in Sweden), com-prises a large variety of tasks. We choose to focus on children, becausewe believe that parenthood is the most demanding role outside work.

78 WOMEN & HEALTH

TABLE 6. Adjusted Odds Ratios (OR) with 95% Confidence Intervals for Long-Standing Illness, Musculoskeletal and Psychiatric Disorders in Women withand Without Children, by Age and Socioeconomic Conditions*

Notes: *n-values, see Table 1.

**OR adjusted for year of interview, and age (continuous); p-values refer to effect modification.

Long-standing

Dow

nloa

ded

by [

Uni

vers

ity o

f St

ockh

olm

] at

08:

14 3

0 Ja

nuar

y 20

12

Children need time and energy-consuming attention and care, and house-hold work increases with number of children, entailing psychologicaland physical demands with a low degree of flexibility. Emslie and co-workers (2004) showed that having children was an important predictorof work-family conflict in women (but not in men) (Emslie, Hunt, &Macintyre, 2004). Motherhood and number of children can be mea-sured with high precision, and the variables are easy to comprehendcompared with other concepts used in work-family research, like work-family conflicts or spillover. Another advantage of the present study wasthe large population-based sample of women that enabled us to studyeffect modification (interaction) with a comparatively high statisticalpower. Potential limitations of the study that are discussed below concernthe cross-sectional setting, the health parameters that were based on sub-jective reports, and uncontrolled confounding and selection mechanisms.

The cross-sectional nature of the study means, for example, that theresults may be biased by pre-existing illness, even if we controlled forlong-standing illness. It is noteworthy, however, that residual bias fromhealth selection most likely would have led to underestimated ORs sincepre-existing illness was negatively associated with having children,while at the same time positively associated with poor health at inter-view. We interpreted the decreased odds of long-standing illness inmothers as a sign of pre-selection–a healthy mother effect–which wassupported by the side-analyses described in the Methods section, andwe believe that an increased awareness of selection mechanisms infuture studies on parenthood and health is warranted. Previously, Khlat,Sermet and Le Pape (2000) have discussed the “healthy mothers” effect.

A cross-sectional design is not ideal for studying the effect of moth-erhood on chronic diseases, while outcome variables of a more “acute”or reversible nature can be studied more accurately, such as self-ratedhealth, fatigue, sleep problems and possibly also anxiety symptoms.Our aim was not to study long-term effects of having children, but ratherto study the current health status among mothers compared to womenwithout children, considering contemporaneous socioeconomic condi-tions. A review including prospective studies on different health mea-sures related to previous employment and sometimes involving thepresence of children has been provided by Klumb and Lampert (2004).

The use of subjective reports on health status may introduce mis-classification. Self-rated health is a frequently used measure, assumedto capture the individual’s underlying disease burden or vulnerability,and has been shown to be a solid predictor of morbidity and mortality(Swedish Council for Planning and Coordination of Research, 1996).

Floderus et al. 79

Dow

nloa

ded

by [

Uni

vers

ity o

f St

ockh

olm

] at

08:

14 3

0 Ja

nuar

y 20

12

Most studies on the validity of self-rated health were based on elderlypeople, but studies incorporating young individuals as well, do alsoshow convincing predictions (Sundquist, & Johansson, 1997; McGee,Liao, Cao, & Cooper, 1999). In a recent meta analysis, including studieswith a comparative wording of the question (Compared to other peopleyour age . . .), self-rated health was a strong predictor of mortality, evenafter adjustment for various morbidity measures (11 studies), depres-sion (5 studies), functional status (10 studies), cognitive function (4 stud-ies), and socioeconomic status (9 studies) (DeSalvo, Bloser, Reynolds,He, & Muntner, 2005). Probably, the findings of higher risks of poorself-rated health in mothers compared to women without children indifferent socioeconomic situations may be attributed to different types ofillness and background mechanisms. For example, the excess of poorhealth in lone mothers may be due to depressive illness caused by maritaldisruption, while the excess in mothers with high income should involveother pathways. Among lone women, the presence of children may alsoreinforce the financial pressure, which is an important risk factor forthis group (Hope, Power, & Rodgers, 1999; Fritzell, & Burstrom, 2006).

Sleep problems may differ and have different causal background, inthat psychological strain may be a risk factor, while physical exhaustionmay rather be protective, “. . . higher physical strain tends to ‘drive’ thebodily systems toward recuperative sleep spontaneously, psychologicalstrain tends to do the reverse, making sleep more difficult to initiate andmaintain” (Winwood, & Lushington, 2006). Both physical exhaustionand psychological strain may be linked to motherhood and may thencounteract each other. This could partly explain the lack of a consistentassociation between children and sleep problems. In lone mothers, inmothers working full-time, or being in the highest income category,psychological strain may come across, yielding the increased odds inthese subgroups. Deprivation of sleep in particular, may also be coveredby the question on pronounced tiredness during daytime (fatigue), forwhich the results were more in line with expectations of increased oddsin mothers.

We have no immediate explanation for the “protective effect” ofmotherhood for psychiatric disorders. Mothers may underreport suchillness to a higher extent than non-mothers due to feelings of guilt of notbeing a “good” mother. Further, considering the low prevalence, thequestionnaire may have captured only the most severe mental illnessthat may be incompatible with having children or having residentialchildren. But, probably, the most likely explanation is that childrenbring meaning to life and prevent mental illness, and the similarity

80 WOMEN & HEALTH

Dow

nloa

ded

by [

Uni

vers

ity o

f St

ockh

olm

] at

08:

14 3

0 Ja

nuar

y 20

12

between the results of psychiatric disorders and anxiety symptoms isnoteworthy.

Evidence is consistent that employed women are healthier than un-employed women, and that married/cohabiting women are better offthan unmarried/lone women (Burstrom, Diderichsen, Shouls, & White-head, 1999; Khlat, Sermet, & Le Pape, 2000; Vaananen, Kevin, Ala-Mursula, et al., 2004; Emslie, Hunt, & Macintyre, 2004; Klumb, &Lampert 2004; Roos, Lahelma, Saastamoinen, & Elstad, 2005; Krantz,Berntsson, & Lundberg, 2005). These associations were apparent in ourdata as well, also in models controlling for prevalence of children (datanot shown). Known and unknown factors contribute to the differencesin health status between socioeconomic groups, factors that might con-vey confounding. We therefore analyzed motherhood and health bystratification, not only to explore effect modification but also to reduceconfounding. A similar approach was used by Khlat, Sermet, and LePape (2000). They analyzed, for example, the odds of poor self-ratedhealth in women 30 years of age or older, for different role profiles(marriage, work, children) within income strata. Their results were notconsistent with ours except for an increased OR for having children innon-married women. The lack of consistency could partly be explainedby different age distributions.

The strongest support in our study for adverse health effects in moth-ers was the consistent associations between number of children andpoor self-rated health. An exposure-response pattern was not only foundfor lone, but also for cohabiting women. The results are consistent withstudies from Spain, based on number of household members (Artazcoz,Borell, & Benach, 2001; Artazcoz, Artieda, Borell, et al., 2004), anda Swedish study showing that women with three or more children had atotal workload of 90 hours per week (compared with about 70 hours inmen with corresponding number of children) (Lundberg, 1996).

Mothers working part-time did not have increased odds of ill health,possibly because they can benefit from the advantages of active em-ployment, which brings about social integration, support, and develop-ment of skills, while at the same time avoiding an overload becauseof the restricted hours of work. Several mechanisms may be involved inthe increased odds of poor health in mothers working full-time. Full-timework can be a question of necessary breadwinning with demands of timeand energy that remains unbalanced by potential payback from enhance-ment mechanisms, particularly in lone mothers (Figure 1). In a Swedishstudy based on 743 women from the white-collar sector, women with atleast 50 hours of paid work and household work more than 20 hours per

Floderus et al. 81

Dow

nloa

ded

by [

Uni

vers

ity o

f St

ockh

olm

] at

08:

14 3

0 Ja

nuar

y 20

12

week had doubled odds of “a high level of symptoms” compared to thosewith less than 50 hours paid work and less than 20 hours of householdwork per week (Krantz, Berntsson, & Lundberg, 2005). Hewitt, Baxter,and Western (2006) reported results from Australia that are coherentwith those of the present study–combining full-time employment andchildren had a negative impact on health in women, while combiningchildren with part-time work or no employment was beneficial.

The consistent effect modification from income was not anticipated.Additional analyses on self-rated health, controlling for education didnot change the pattern (primary school, upper secondary school, highereducation), neither did adjustment for part-time/full-time employment.High-income jobs may nowadays entail a very high pressure which candiminish the time and energy left for children. High-income earnersshould also hold more qualified and challenging jobs, and children mightbe perceived as a threat to their occupational identity (Frone 2000), withincreased psychological strain and negative effects on health and well-be-ing. This would be in line with the findings for women in high-rankingpositions, showing higher norepinephrine levels after work in mothers

82 WOMEN & HEALTH

FIGURE 1. Adjusted Odds Ratios (dots) for Poor Self-Rated Health in Womenwith Children Compared to Women Without Children, by Cohabitation Statusand Full-Time/Part-Time Employment

Note: Vertical lines represent 95% confidence intervals.

Cohabiting, Full-timeSingle, Full-time Single, Part-time Cohabiting, Part-time

Dow

nloa

ded

by [

Uni

vers

ity o

f St

ockh

olm

] at

08:

14 3

0 Ja

nuar

y 20

12

compared to women without children (Lundberg & Frankenhaeuser,1999). In Sweden, it is unusual to hire a person for domestic services,also among economically privileged women.

Poor self-rated health in mothers may in the long run lead to more se-vere health conditions, a question that requires a follow-up into higherages (Klumb & Lampert, 2004). Most types of chronic disease are ex-tremely rare in the age range represented in the present study. However,in a U.S. longitudinal study, for example, the incidence of coronaryheart disease was approximately twice as high in employed women withthree or more children as among employed women without children(Haynes & Feinleib, 1980). The choice of health variables is a delicateissue and different measures cannot be seen as interchangeable markersof health (Wikman, Marklund, & Alexanderson, 2005).

The present study suggests a need for improved negotiations betweenthe parents regarding a fair share of work and family duties, and extendedsupport for lone mothers. The results may be useful to policy-makers andemployers in developing new policies. The research agenda within thework-family field is dynamic and shifts across time and nations (Lewis &Cooper, 1999), with different welfare systems (Allen, 2003).

REFERENCES

Allen, S. (2003). Working parents with young children: cross-national comparisons ofpolicies and programmes in three countries. International Journal of Social Wel-fare, 12, 261-273.

Artazcoz, L., Borrell, C., & Benach, J. (2001). Gender inequalities in health amongworkers: the relation with family demands. Journal of Epidemiology and Commu-nity Health, 55, 639-647.

Artazcoz, L., Artieda, L., Borell, C., Cortès, C., Benach, J., & Garcìa, V. (2004). Com-bining job and family demands and being healthy. What are the differences betweenmen and women?. European Journal of Public Health, 14, 43-48.

Burstrom, B., Diderichsen, F., Shouls, S., & Whitehead M. (1999). Lone mothers inSweden: trends in health and socioeconomic circumstances, 1979-1995. Journal ofEpidemiology and Community Health, 53, 750-756.

Cinamon, R.G., Rich, Y. (2002). Gender differences in the importance of work andfamily roles: implications for work-family conflict. Sex Roles: A Journal of Re-search, Dec issue.

Chandola, T., Martikainen, P., Bartley, M., Lahelma, E., Marmot, M., Michikazu, S.,Nasermoaddeli, A., Kagamimori, S. (2004). Does conflict between home and workexplain the effect of multiple roles on mental health? A comparative study ofFinland, Japan, and the UK. International Journal of Epidemiology, 33, 884-893.

Floderus et al. 83

Dow

nloa

ded

by [

Uni

vers

ity o

f St

ockh

olm

] at

08:

14 3

0 Ja

nuar

y 20

12

Crosby, F.J. (1987). Spouse, parent, worker: on gender and multiple roles. NewHaven: London: Yale University Press.

Desalvo, K.B., Bloser, N., Reynolds, K. He, J., & Muntner, P. (2005). Mortality predic-tion with a single general self-rated health question. Journal of Genetic and InternalMedicine, 21, 267-275.

Emslie, C., Hunt, K., & Macintyre, S. (2004). Gender, work-home conflict, andmorbidity amongst white-collar bank employees in the United Kingdom. Interna-tional Journal of Behavioral Medicine, 11, 127-134.

Fritzell, S., & Burstrom, B. (2006). Economic strain and self-rated health among loneand couple mothers in Sweden during the 1990s compared to the 1980s. Health Pol-icy, 79, 253-264.

Frone, M.R., Russel, M. & Barnes, J.M. (1996). Work-family conflict, gender, andhealth-related outcomes: a study of employed parents in two community samples.Journal of Occupational Health Psychology, 1, 57-69.

Frone, M.R. (2000). Work-family conflict and employee psychiatric disorders: the Na-tional Comorbidity Survey. Journal of Applied Psychology, 85, 888-895.

Goode, W. (1960). A theory of role strain. American Sociological Review, 25, 483-496.Gove, W.R., & Geerken, M.R. (1977). The effect of children and employment on the

mental health of married men and women. Social Forces, 56, 66-76.Grzywacz, J. (2000). Work-family spillover and health during midlife: is managing

conflict everything? American Journal of Health Promotion, 14, 236-243.Grzywacz, J., & Marks, N.F. (2000). Reconceptualizing the work-family interface: an

ecological perspective on the correlates of positive and negative spillover betweenwork and family. Journal of Occupational Health Psychology, 5, 111-126.

Hammer, T., Saksvik, P.O., Nytro, K., Torvatn, H., & Bayasit, M. (2004). Expandingthe psychosocial work environment: workplace norms and work-family conflict ascorrelates of stress and health. Journal of Occupational Health Psychology, 9,83-97.

Haynes, S.G., & Feinleib, M. (1980). Women, work and coronary heart disease: Pro-spective findings from the Framingham heart study. American Journal of PublicHealth, 70, 133-1412.

Hewitt, B., Baxter, J., & Western, M. (2006). Family, work and health – The impact ofmarriage, parenthood and employment on self-reported health of Australian menand women. Journal of Sociology, 42, 61-78.

Hope, S., Power, C., & Rogers, B. (1999). Does financial hardship account for elevatedpsychological distress in lone mothers? Social Science & Medicine, 49, 1637-1649.

Idler, E.L. & Kasl, S. (1991). Health perceptions and survival: do global evaluations ofhealth status really predict mortality? Journal of Gerontology, 46, 55-65.

Khlat, M., Sermet, C., Le Pape, A. (2000). Women´s health in relation with their familyand work roles: France in the early 1990s. Social Science & Medicine, 50, 1807-1825.

Klumb, P.L., & Lampert, T. (2004). Women, work, and well-being 1950-2000: a re-view and methodological critique. Social Science & Medicine, 58, 1007-1024.

Korpi, T., & Stern, C. (2006). Women’s Employment in Sweden: Globalization, Dein-dustrialization, and the Labor Market Experiences of Swedish Women 1950 – 2000.

84 WOMEN & HEALTH

Dow

nloa

ded

by [

Uni

vers

ity o

f St

ockh

olm

] at

08:

14 3

0 Ja

nuar

y 20

12

In Blossfeld, H-P. and H. Hofmeister (ed.) Globalization, Uncertainty, and Wom-en’s Careers. An International Comparison. Edward Elgar.

Krantz, G., Berntsson, L., & Lundberg, U. (2005). Total workload, work stress and per-ceived symptoms in Swedish male and female white-collar employees. EuropeanJournal of Public Health, 15, 209-214.

Krantz, G., & Ostergren, P.O. (2001). Double exposure, the combined impact of do-mestic responsibilities and job strain on common symptoms in employed Swedishwomen. European Journal of Public Health, 11, 413-419.

Lewis, S., & Cooper, C.L.. (1999). The work-family research agenda in changing con-texts. Journal of Occupational Health Psychology, 4, 382-393.

Lundberg, U. (1996). Influence of paid and unpaid work on psychophysiological stressresponses of men and women. Journal of Occupational Health Psychology, 1,117-130.

Lundberg, U., & Frankenhaeuser, M. (1999). Stress and workload of men andwomen in high-ranking positions. Journal of Occupational Health Psychology, 4,142-151.

McGee, D.L., Liao, Y., Cao, G., & Cooper, R.S. (1999). Self-reported health status andmortality in a multiethnic US cohort. American Journal of Epidemiology, 149,41-46.

Nordenmark, M. (2004). Balancing work and family demands. Do increasing demandsincrease strain? A longitudinal study. ScandinavianJournal of Public Health, 32,450-455.

The Research Ethics Committee (2007, April 19). www.epn.se/eng/start/background.aspx

Roos, E., Lahelma, E., Saastamoinen, P. & Elstad, J.-I. (2005). The association of em-ployment status and family status with health among women and men in fourNordic countries. Scandinavian Journal of Public Health, 53, 250-260.

Statistics Sweden (SCB) (1997). Living conditions and inequality in Sweden - a20-years perspective 1975-1995. Population and welfare statistics, the nationalSwedish survey of living conditions (ULF). Örebro: Sweden: SCB-Tryck,.

Statistics Sweden (SCB) (2007, April 19). http://www.scb.seSwedish Council for Planning and Coordination of Research (FRN) (1996). Bjorner,

J.B., Kristensen, T.S., Orth-Gomér, K., Tibblin, G., Sullivan, M., & Westerholm, P.eds. Report 96:9: Self-rated health. A useful concept in research, prevention andclinical medicine. Baarn: Holland: M.C. Escher & Cordon Art.

Sundquist, J., & Johansson, S.E. (1997). Self reported poor health and low educationallevel predictors for mortality: a population based follow up study of 39,156 peoplein Sweden. Journal of Epidemiology and Community Health, 51, 35-40.

Thoits, P.A. (1983). Multiple identities and psychological well-being: a reformulationand test of the social isolation hypothesis. American Sociological Review, 48,174-187.

Vaananen, A., Kevin, M.V., Ala-Mursula, L., Pentti, J., Kivimaki, M., & Vahtera, J.(2004). The double burden of and negative spillover between paid and domesticwork: associations with health among men and women. Women & Health, 40,1-18.

Floderus et al. 85

Dow

nloa

ded

by [

Uni

vers

ity o

f St

ockh

olm

] at

08:

14 3

0 Ja

nuar

y 20

12

Wikman, A., Marklund, S., & Alexanderson, K. (2005). Illness, disease, and sicknessabsence: an empirical test of differences between concepts of ill health. Journal ofEpidemiology and Community Health, 59, 450-454.

Winwood, P.C., & Lushington, K. (2006). Disentangling the effect of psychologicaland physical work demands on sleep. Recovery and maladaptive chronic stress out-comes within a large sample of Australian nurses. Journal of Advanced Nursing, 56,679-689.

86 WOMEN & HEALTH

Dow

nloa

ded

by [

Uni

vers

ity o

f St

ockh

olm

] at

08:

14 3

0 Ja

nuar

y 20

12