Social and family factors in childhood hospital · Social and family factors in childhood hospital admission D MFERGUSSON,L J HORWOOD,ANDF T SHANNON Fromthe Christchurch ChildDevelopmentStudy,

Journal of Epidemiology and Community Health, 1986, 40, 50-58

Social and family factors in childhood hospitaladmission

D M FERGUSSON, L J HORWOOD, AND F T SHANNON

From the Christchurch Child Development Study, Department ofPaediatrics, Christchurch Clinical School ofMedicine, Christchurch Hospital, Christchurch, New Zealand

SUMMARY The relation between social, economic, and family life event measures and rates ofhospital admission during the period from birth to 5 years was studied in a birth cohort of NewZealand children. Both family social background and family life events made a significantcontribution to the variability in the risk of hospital admission. However, economic factors madeno significant contribution to rates of admission when the correlated effects of family socialbackground and life events were taken into account. In addition, the effects of family life events onrisks of admission appeared to be far more marked than the effects of family social background.Possible explanations of the consistent association between life events and rates of morbidityduring early childhood are discussed.

There have been a large number of studies of the roleof social factors in the aetiology of illness duringchildhood. At the risk of some simplification, twomajor groups of factors have emerged as beingimplicated in the risk of illness:

(a) Family social and economic situation: It hasbeen found that rates of illness tend to be higher infamilies of low socioeconomic status," single parentfamilies,"7 families of non-Caucasian ethnic origin,8"families in which parental education is poor,' 10families with young mothers,1" 12 and families withlow income or depressed living standards.7"

(b) Family life events: A number of studieshave suggested that rates of morbidity are increasedamong children in families facing various forms ofstress, adversity or changing life situation.10 12-19

However, while there have been a large number ofreports of the social correlates of illness in children,most studies appear to have been conducted oncross-sectional samples or on patient populations,and there appear to be few, if any, studies that haveexamined the way in which family life events andfamily social and economic factors influence risks ofmorbidity over a protracted time period in thegeneral child population.We report on the results of a five year longitudinal

study of rates of hospitalisation for respiratoryinfection, gastroenteritis, and accidents in a birthcohort of New Zealand children during their first fiveyears. The aims of the study were twofold:

(a) to document the way in which various familyand social factors were associated with variations inthe risk of hospitalisation during the preschool years;and

(b) to devise a proportional hazards model whichestimated the net effect of various factors on thecumulative risk of hospitalisation over time.

Method

The data were collected during the first five years ofthe Christchurch Child Development Study. In thisstudy, a birth cohort of 1265 children born in theChristchurch (NZ) urban region has been studied atbirth, at 4 months, and at annual intervals to the ageof 5 years. At each point information has beencollected by an hour-long structured interview withthe child's mother supplemented by informationfrom hospital notes, general practitioner records, andother documentary sources.7 13 20 From the datacollected in the study the following variables wereselected for analysis:

(a) HOSPITAL ADMISSION FOR INFECTION AND

ACCIDENTSInformation on hospital admission was collected inthe following ways. Each year, mothers weresupplied with a health diary record in which theywere requested to record all details of any medicalcontact the child had made during the year. The diary

50

Protected by copyright.

on January 4, 2021 by guest.http://jech.bm

j.com/

J Epidem

iol Com

munity H

ealth: first published as 10.1136/jech.40.1.50 on 1 March 1986. D

ownloaded from

http://jech.bmj.com/

Social and family factors in childhood hospital admission

record served as the primary source of informationabout health medical contacts but this wassupplemented by direct questioning of the mother ateach year. Once a child had been identified as havingbeen admitted to hospital, details of the admissionwere obtained from the case notes and relatedmaterial describing the admission. (In all cases noteswere obtained following informed maternalconsent.) It was possible to check maternal reportingaccuracy by comparing maternal reports with thecentral records held at Christchurch Hospital and thisshowed that during the five year study period onlyfive admissions were not reported and all of thesewere for relatively trivial reasons. During the fiveyear period a total of 670 admissions occurred andadmission rates show a steady tendency to declinewith age, 350/, occurring during the first year.For the purposes of the present analysis, two major

groups of admissions were considered:1 admissions for pneumonia, bronchitis,bronchiolitis, croup, other respiratory infection orgastroenteritis;2 admissions for childhood accidents includingaccidental poisoning and burns.These admissions constituted 36%/0 of the total

admissions occurring during the study period andwere selected on the basis of previous analysis whichsuggested that they were the only admissioncategories sensitive to social or familial.conditions.' 10 12 13 Consideration was given tosubdividing the data into finer diagnostic groups butthis did not give sufficient numbers of observationsper category to conduct a detailed analysis of thedata.

(b) EXPLANATORY VARIABLESThe following variables were used in the analysis asexplanatory variables.1 Measures offamily social backgroundi Maternal educational level classified as: motherlacked formal educational qualifications; mother hadsecondary educational qualifications (ie, NZ SchoolCertificate, university entrance); mother had tertiaryqualifications (university degree, tertiaryprofessional qualifications).ii Child's ethnic status: Maori, PacificIsland/Caucasian, other.iii Family socioeconomic status: based on the Elleyand Irving2l scale of socioeconomic status for NewZealand. This scale divides the population into sixsocial classes on the basis of occupation.iv Child's family placement at birth: whether thechild entered an adoptive family, a (natural) twoparent family or a single parent family.v Maternal age at the birth of the child.

2 Measures offamily economic situationi Averaged family income: This was based on theaverage of the family's reported gross annual incomeduring the five years study period and was convertedto a rank measure to give an estimate of the meanincome level experienced by the child's family duringthe preschool years.ii Averaged standard of living: At each year,interviewers were asked to rate the family's materialconditions on a five point scale ranging from "verygood, obviously affluent" to "very poor, familyobviously in poverty". To obtain an estimate of thetypical material conditions of the family during thestudy period the interviewer ratings were summedover the five year study period and divided by five.iii Averaged financial difficulty: At each year,interviewers were asked to assess how much financialdifficulty the family was facing, independently of itsmaterial living standards. An averaged rating offinancial difficulty was obtained in the same way asthe averaged measure of family living standards.iv Quality of interior and exterior of theaccommodation: Each year interviewers rated thequality of the interior and exterior of the child'saccommodation on a five point scale ranging from"very good" to "very poor". Overall measures ofquality of accommodation were obtained byaveraging the ratings over the five year period.

3 Family life eventsEach year from 2 to 5 years, mothers wereinterviewed on a modified version of the Holmes andRahe Social Readjustment Rating Scale.22 Thisconsisted of 20 items which covered such areas asdeath or illness in the family, changes in parent'semployment, financial problems, and maritaldisharmony. The distributions of life events itemsover time and the characteristics of the scale havebeen described in a previous paper."3 To measure theamount of stress or adversity faced by the family, fourlife events scores were created using a simple sum ofthe number of life events reported during the periods1-2 years, 2-3 years, 3-4 years, and 4-5 years. Asimple count of the number of life events was used inpreference to a weighted score of these measuressince previous analysis"3 had shown that theunweighted sum was highly correlated (r= 0 96) withthe weighted sum. This implies that the amount ofinformation provided by weighting items in terms oftheir "importance" was negligible. The unweightedsum was used in the analysis as it was simpler tointerpret.

SAMPLE SIZEThe analysis is based on a sample of 1057 children forwhom complete data on all variables were available.

51



j.com/

J Epidem

iol Com

munity H


ownloaded from


52

This sample represents 84%/ of the initial cohort of1265 children and 94%/o of the children who remainedalive and resident in New Zealand to 5 years. Toexamine whether the sample was biased as a result oflosses to follow up, the social and demographiccharacteristics of the initial sample were comparedwith those of the sample participating at 5 years. Thisanalysis showed no significant association betweenlosses to follow up (including postnatal death) andthe child's social or perinatal characteristics.

Results

ASSOCIATIONS BETWEEN RATES OF HOSPITAL

ADMISSION, FAMILY SOCIAL POSITION, FAMILY

ECONOMIC CIRCUMSTANCES, AND FAMILY LIFE

EVENTSTable 1 shows rates of hospital admission (per 100children aged 0-5 years) for respiratory illness,gastroenteritis, and accidents cross-tabulated by aseries of measures of family social position, familyeconomic circumstances, and family life events. Eachcomparison is tested for statistical significance usingan F statistic based on the square root transformationof the rate data. The results show:1 There was a clear tendency for children fromsocially disadvantaged backgrounds to experiencehigher rates of admission. As a general rule, childrenof Polynesian ethnic status, children from singleparent families, children of poorly educated mothers,children of young mothers, and children fromfamilies of low socioeconomic status had the greatestrisk of admission for both accidents and infections.2 Rates of admission varied with family economicand material circumstances, children from lowincome families and from families with depressedmaterial conditions being at greater risk of hospital.admission.3 There was a consistent trend for rates of hospitaladmission to increase with the number of life eventsreported by the child's mother each year.

DATA REDUCTION AND PROPORTIONALHAZARDS ANALYSISThe results in table 1 strongly suggest that rates ofhospital admission were highest among children fromsocially disadvantaged, economically deprived andstressed families. However, the analysis does notshow the net contributions of these factors to thevariability in the rates of admission. To examine this aproportional hazards model was fitted to the data. (Adetailed account of the theoretical and mathematicalbasis of this model may be found in Cox23 andKalbfleisch and Prentice.24 However, before thisanalysis the extensive social and economic data intable 1 were reduced to two general measures:

D M Fergusson, L J Horwood, and F T ShannonTable 1 Mean number of hospital admissions per 100children (birth to S years) for infections and accidents bymeasures of family social position, family economicsituation, and family life events

Infections Accidents TotalVariable N (A) (B) (A +B)

Family social position

Maternal age (yr)15-1920-2425-2930-3435+

Maternal educationNo formal qualifications 546Secondary qualifications 306Tertiary qualifications 205

Ethnic status of childPolynesian 155Caucasian, other 902

Birth statusAdopted 36Two parents 946Single parent 75

Socioeconomic statusProfessional, executive 217Skilled, clerical 565Semiskilled, unskilled, 275unemployed

Economic and material conditions

Standard of livingAbove average 171Average 821Below average 53

Family incomeUpper quintile 20'Second quintile 211Third quintile 211Fourth quintile 214Lower quintile 21t

Financial difficultyObviously had no difficulty 244Unlikely to have difficulty 73'Probably having difficulty 7

Quality of interior of houseObviously affluent 41Very good 36:Average 57'Below average 71

Quality of exterior of houseObviously affluent 41Very good 34'Average 61(Below average 61

97 25-8 18-6 44-3323 16-7 14-2 31-0417 9-6 4-8 14-4165 8-5 11-5 20-055 7-3 1*8 9 1

p<0-01

16-99-28-3

p<0.05

5 19-42 11 9

p<0.05

6 566 11-75 32-0

p<0-001

7 6-95 12-65 18-6

p<0-05

p<0-001

9.912-85.4

NS

18-78-3

p<0-01

2-89-1

22-7

p<0-01

5.510-611-6

NS

p<0-0001

26-721-913-7

p<0.05

38-120-2

p<0-01

8-320-854-7

p<0-0001

12-423-230-2

p<0-01

8 10-7 5-6 16-36 12-1 9 0 21-13 34-0 37-7 71-7

p<0O0115 9.51 6-11 17-14 12-66 18-0

p<0.01

*6 8-19 13-72 22-2

p<0.05

*8 8-33 9-4'5 13-91 26-8

p<0.05

4 12-25 7.50 14-91 24-6

p<0-0001 p<O-OOOl

4-4 13-97-5 13-68-8 25-911-7 24-214-6 32-6

p<0.05 p<O-Ol

4-5 12-69.5 23-1

31-9 542

p<0-0001 p<O-OOOl

6-3 14-68-0 17-49-0 23-0

28-2 54-9

p<0-0001 p<O-OOOl

4-9 17-17-5 15-19-2 24-132 8 57.4

p<0-01 p<O-OOOl p<O-OOOlcondnued



j.com/

J Epidem

iol Com

munity H


ownloaded from



Table 1-continued

Infections Accidents TotalVariable N (A) (B) (A +B)

Family life events

Life events 1-2 yr0 143 105 4-2 14-71-2 484 8 5 6-2 14 73-4 291 13-8 14-8 28 55+ 139 295 18-0 47-5

p<O-OO1 p<O-OOl p<O-OOOlLife events 2-3 yr

0 187 9 1 6-4 15-51-2 545 12-8 8-1 20-93-4 235 8 5 10-2 18-75+ 90 33 3 26-7 60-0

p<O-OOl p<0-OOl p<O-000lLife events 3-4 yr

0 221 8-6 5 4 14-01-2 500 110 8-2 19-23-4 248 15-3 13-7 29-05+ 88 28-4 19-3 47-7

p<0-OOl p<O-OOl p<O-OOOlLife events 4-5 yr

0 242 4-6 6-6 11-21-2 520 13-1 7-1 20-23-4 216 15-7 17-6 33-35+ 79 30 4 16-5 46-8

p<0-001 p<O-OOl p<O-OOOl

1 Family social position This was a weightedlinear combination of the measures of maternal age,maternal education, ethnicity, the child's birthplacement, and family socioeconomic status. Thisweighted sum defined a general dimension whichranged from extremely socially disadvantagedfamilies (ie, those with a young mother of pooreducation, of Polynesian ethnicity, in which the childentered a single parent, low socioeconomic statusfamily at birth) to extremely highly advantagedfamilies (ie, those with an older mother with tertiaryeducation, of Caucasian ethnicity, where the childentered a two parent family of professionalsocioeconomic status).

2 Family economic background This was a

weighted linear combination ofthe measures offamilyincome, family living standards, and family housingconditions. The measures defined a generaldimension of economic material well-being whichranged from families with low income and poor livingstandards and housing conditions to families withhigh income and superior housing and livingconditions.These indices were obtained by factor analysing

the variables and using the least squares estimates ofthe factor score coefficients to obtain an estimate ofthe underlying dimension. A detailed account of themethod of scale construction has been given in a

previous paper.25 There were three major reasoningsfor reducing the data in this way:(a) Since the emphasis of the study was on theeffects of broad social and related differencesbetween families, it was felt that composite measuresof social and economic conditions would providemore robust and interpretable measures of thegeneral aspects of family functioning.(b) A number of the explanatory variables in table1 were highly intercorrelated, and this could raiseserious problems of multicollinearity in an analysis inwhich the original raw data were entered. Byreducing the data into general indices based on sets ofintercorrelated variables many of these problemswere reduced.(c) The proportional hazards model describedbelow includes a dynamic time dependent covariate,and preliminary investigation suggested that withsuch a variable it was not possible to enter allexplanatory variables into the model if the originalraw data were analysed. By combining variables intomore general indices, the demands on the availablecomputing space were reduced to manageabledimensions.However, while the reduction of the social and

economic data to broad indices has a number ofobvious theoretical and practical advantages, therewas a possibility that such classification could distortthe results. To examine this a number of exploratorymultiple regression analyses were conducted tocompare the predictive power of the combinedmeasures with that of the original raw data. Thisanalysis suggested that combining social andeconomic variables into more general indices had anegligible effect on the predictive power of the data.Table 2 shows the results of fitting a proportional

hazards model to the risk of at least one hospitaladmission for accident or infection during the fiveyear study period. In this model the explanatoryvariables were the fixed covariates of family socialposition and economic situation and the timedependent covariate of family life events. (In a timedependent analysis the values of the covariates arepermitted to change throughout the period beingstudied. A more detailed discussion of the propertiesof such covariates is provided by Kalbfleisch andPrentice'). The model fitted was:

Ak(z;t)=X.(t) ez',O+z-tPwhere X(z;t) denotes the hazard or instantaneous rateof hospitalisation at time t for a subject with covariatevector z. In this instance the vector z comprises the setof fixed covariates zi and the set of time dependentcovariates z2 measured at time t. XO(t) denotes thehazard which exists for the base line population withcovariate values of 0 at time t, and 13s, /2 are a set ofregression like coefficients which show the extent to

53



j.com/

J Epidem

iol Com

munity H


ownloaded from


54

which the hazard for a subject with a non zerocovariate vector is increased when compared to thehazard for the base populations.The table is divided into two major sections.The first section shows maximum likelihood

estimates of the regression coefficients of the modeland their standard errors. This information makes itpossible to test the significance of the net effects ofeach of the explanatory variables on the variation inrates of hospital admission. The results show thatfamily social position (t=3.07; p<001) and familylife events (t=3-88; p<0001) were significantlyrelated to rates of hospital admission. However,family economic situation was not a significantpredictor (t=057; p>7050). This implies that theapparent correlations between family economicconditions and rates of admission arose largely fromthe correlations of this measure with family socialposition and family life events.The second part of the table shows the estimated

proportional hazards coefficients. These may beinterpreted in a way that is analogous to the morefamiliar notion of relative risk: the increase in thehazard or instantaneous rate of hospital admissionwhich is associated with a given level of anexplanatory variable when compared to the risk thatexists for a defined base line population. In thisinstance the base line population is the group ofchildren in the top quartile of the social positionmeasure whose families reported one or fewer lifeevents in any given year. The findings show a cleartendency for decreasing social position to beassociated with increasing risks of admission, andchildren who were in the lowest quartile of thepopulation had nearly twice the rate of admission aschildren in the base line population. Similarly, thereis a marked tendency for risks of admission toincrease with increasing reports of life events, and forany year children whose families reported five ormore life events were 2 7 times more likely to beadmitted to hospital.

While the results in table 2 show the relative risksof admission for various groups of subjects, they donot indicate the overall levels of admission for groupswith varying combinations of covariate values. To dothis requires estimation of the survival curves (ie, theproportions of subjects not admitted to hospital overtime) for various covariate values. However,estimation of survival curves for the life eventsmeasure is complex, since this measure was treated asa time dependent covariate with the result that ratesof survival are conditional on the pattern of lifeevents reporting over the five year period. Toproduce estimates of the survival curves anapproximate method was therefore used. Thisinvolved computing, for each child, the average

D M Fergusson, L J Horwood, and F T Shannon

Table 2 Results ofproportional hazards model

(a) Fitted regression coefficients

Variable Coefficient SE t p

Family life events 0 330 0-086 3-88 <0-001Family social position 0-229 0-075 3 07 <0 01Family economic situation 0-051 0-089 0 57 >0 50

(b) Proportional hazards coefficients for levels of significant variables

Variable ezglFamily life eventsNo events 11-2 events 1-3973-4 events 1-9525+ events 2-726

Family social positionTop quartile 1Second quartile 1-257Third quartile 1 580Bottom quartile 1-986

hazard score for the life events history he or sheexperienced throughout the five year study period.The averaged hazard was defined as:

A(t;z)= Xo(t)lezf/4where ezP was evaluated for the life events measure ateach of the four time periods studied. It will be notedthat since XO(t), ,f, and e are known constants for anytime t, the averaged hazard is proportional to theaverage number of life events reported per year or,alternatively, to the accumulated number of lifeevents that the family reported.

Figures 1 and 2 show the proportion of children notadmitted to hospital for infection or accidents overthe five year study period subdivided by both thesocial position score and the averaged hazard score.The comparisons are shown in two forms:(i) The unadjusted results are the observedsurvivorship curves for each of the groups.(ii) The adjusted curves are the estimatedsurvivorship curves, taking into account theintercorrelations of family life events and the familysocial position. The adjusted curves were obtainedby: (a) stratifying the sample on the basis of eachfactor; (b) solving the theoretical survival curves foreach stratum; and (c) combining the resultingtheoretical curves to produce overall adjusted curvesusing a weighted average of the estimates.

Inspection of figs 1 and 2 suggests two majorconclusions:(i) The effects of adjusting the curves for theintercorrelation of life events and the family socialposition is to reduce the overall effects of each factoron rates of admission. Before adjustment, childrenwhose families reported an average of five or morelife events per year had rates of admission by five



j.com/

J Epidem

iol Com

munity H


ownloaded from


55Social and family factors in childhood hospital admission

0)

90-

I

ED

aa

c 7

0 60

1 2 3 4 5Age

Fig 1 Cumulative proportions ofchildren not admitted to hospital by a given age for varying levels ofthe mean number oflife events per annum (a) unadjusted for the effects offamily social position and (b) adjusted for the effects offamily socialposition:* <1 life event per annum; 0 1-2 life events per annum.

OL 3-4 life events per annum; * 5 + life events per annum.

100-

Soo.. .7 90-

-9 8

so

01 2 3 4 5

Age

c 70-

60 .4

0

Ce--_ ..-----7

I I I

1 2 3Age

Fig 2 Cumulative proportions ofchildren not admitted to hospital by a given age for quartiles ofthe family social positionindex (a) unadjusted for the effects offamily life events and (b) adjusted for the effects offamily life events:* <top quartile; 0 second quartile.O third quartile; * S + bottom quartile.

0100

90'la

l 80'a0C_ 70

L.

A!60-

Age

0100.

0$

c

60

I 1

4. 5

b



j.com/

J Epidem

iol Com

munity H


ownloaded from


56

years which were nearly five times higher than thosewhose families reported an average of less than oneevent per year (39% versus 8%/, respectively). Afteradjustment for the common effects of family socialposition this difference reduced to 3 to 1 (29% versus9% respectively). Similar trends are present for thesocial position measure: before adjustment for familylife events children in the lowest quartile of thedistribution had five year admission rates which werenearly 224 times those in the top quartile (24% versus10%/6 respectively); after adjustment this differencereduced to 1-8 to 1 (21% versus 12%/o respectively).(ii) In terms of the contribution to the variability inrates of admission it is apparent from both theunadjusted and adjusted curves that the family lifeevents variable makes a greater contribution than thesocial position measure.

Discussion

The findings of this research confirm the results ofprevious studies of the social aetiology of childhoodmorbidity:1-19 rates of hospitalisation were highestamong children from socially disadvantaged families,economically depressd families, and families facingvarious forms of stress and adversity. However, thelongitudinal nature of the data and associatedanalysis serve to place a number of important issuesinto perspective.

Firstly, it is often implicitly assumed that"poverty" is one of the major social determinants ofhospital admission and ill health in children.However, the findings here indicate that familyeconomic and material conditions made nosignificant contribution to the variability in rates ofhospital admission when the correlated factors offamily social background and family life events weretaken into account. This suggests that the apparentcorrelations between family material and economicconditions and hospital admission arise because bothhigher rates of morbidity and depressed economiccircumstances are symptomatic of a sociallydisadvantaged or stressed family environment. Ofcourse, this conclusion holds true only for the rangeof social conditions in the population under study,and the effects of poverty on child health inunderdeveloped Third World countries are wellknown. What the findings indicate is that in arelatively economically homogeneous society, such asNew Zealand, economic factors play little role in theaetiology of hospital admission once the effects offamily social background and family life events aretaken into account.

It is also often assumed that social class orsocioeconomic status is the major social determinantof hospital admission. However, the findings here

D M Fergusson, L J Horwood, and F T Shannon

show very clearly that family life events made aconsistently larger contribution to the variability inrates of illness than family social background factors.This observation has been noted in previous studiesof this cohort.10 12 13 This perhaps suggests that in asociety such as New Zealand variations inintra-family dynamics make a greater contribution tothe risk of childhood illness than broad differences interms of family social or demographic factors.The persistent correlations between life events

measures and child health seen in this research andother studies10 119 raise a number of importantquestions about the mechanisms by which life eventsare related to childhood illness. Broadly speaking,there appear to be three major explanatoryhypotheses which could account for the associations.

(a) MEASUREMENT ERROR AND RESPONDENTBIASIt is possible that the apparent correlations betweenlife event reports and rates of illness arise from aseries of measurement errors or biases. In particular,it is possible that while life events do not alter illnessthey may alter health care seeking behaviour so thatfamilies under stress or making social readjustmentsmay seek medical attention more often than otherfamilies. This is a difficult issue to resolve sincealmost invariably measures of morbidity are based onmedical attendance. However, it is perhaps worthnoting that if such bias does exist it appears to applynot only to hospital admission but to other types ofmedical attendance since a previous study of thiscohort has found similar correlations betweengeneral practitioner attendances, hospital outpatientattendance, and life events.13 It is also possible thatmothers whose children have been admitted tohospital may invent or over-report life events in anattempt to explain the reason for the hospitalisation.For the present data this seems unlikely since the lifeevents chosen for study related to common and majorproblems-health problems in the family, financialproblems, marital problems, bereavement, familyconflicts (see Beautrais, Fergusson and Shannon13)which would have been hard to invent or over-report.

(b) CHANGES IN SUSCEPTIBILITY TO ILLNESS

The higher rate of hospitalisation for infection amongchildren from families reporting high levels of lifeevents leads to the speculation that perhaps childrenin these families have a greater susceptibility toillness due to a transient immunodeficiency caused bystress and readjustment in the family. There havebeen a number of tantalising hints that this may bethe case. Meyer and Haggerty,"7 in a study ofstreptococcal sore throats in children, found that theonset of illness was often preceded by a significant



j.com/

J Epidem

iol Com

munity H


ownloaded from



family life event. Further, a series of studies havesuggested changes in catecholamines,corticosteroids, and immune function in subjectsfacing various forms of stress and difficulty.26 31 Itmay be hypothesised that stress leads to an increasedcatecholamine and corticosteroid production whichinhibits immune function27"29 which in turn leads togreater rates of morbidity.26 30 31 While the availableevidence tends to be consistent with this hypothesis,critical evidence linking life events to changes inimmune function and thence to greater risks ofinfection is still lacking.

(C) CHANGES IN PARENTING BEHAVIOURWhile it is possible that the greater rates of infectionin children in families reporting significant life eventsmay be explained by a change in immune function,such an explanation cannot be applied to thepersistent correlations that have been found betweenfamily life events and childhood accidents." 12 14 16 19It seems more likely that this correlation arisesbecause of changes in parenting behaviour whichplace the child at greater risk from variousenvironmental hazards. In particular, it seems likelythat in families facing readjustment, parentalvigilance and care of children may be reduced,leading to a greater risk of accidents.12 However,again critical evidence linking life events to a changein parenting behaviour and thence to a change in thehazards faced by the child is not yet available.

Life events research has been an area that has beenextremely controversial, and opinions have tended topolarise into an uncritical acceptance of the view thatlife events cause morbidity to an over criticalrejection of this view.32 A more realistic assessmentof the field is that there have been a large number ofcorrelations reported between life event measuresand various measures of morbidity. Thesecorrelations have not been explained adequatelyeither by the critics who espouse the view that theyarise from measurement error or the enthusiasts whoassume that they reflect causal associations. Clearlythere is still much to be learned about the way inwhich adversity and readjustment in the lives ofpeople influence their health and well-being. Amajor aim of this paper has been to show that, of thesocial factors implicated in hospital admission duringthe preschool years, family life events appear to exertby far the most persistent and consistent influence onadmission rates.

This research was funded by grants from the MedicalResearch Council of New Zealand and the NationalChildren's Health Research Foundation.

References

'Taylor B, Wadsworth J, Golding J, Butler N.Breastfeeding, bronchitis and admissions to hospital forlower respiratory illness and gastroenteritis during thefirst five years. Lancet 1982; i: 1227-9.

2Scott HD, Mackie A. Decisions to hospitalise and tooperate: a socioeconomic perspective in an urban state.Surgery 1975; 77: 311-7.

'Colley JRT, Reid DD. Urban and social origins ofchildhood bronchitis in England and Wales. Br Med J1970; 2: 213-7.

'Miller FJW, Court SDM, Walton WS, Knox EG. Growingup in Newcastle upon Tyne: a continuing study ofhealthand illness in young children within their families.London: Oxford University Press, 1960.

'Wadsworth J, Burnell I, Taylor B, Butler N. Family typeand accidents in preschool children. J EpidemiolCommunity Health 1983; 37: 100-4.

6Fergusson DM, Horwood EJ, Shannon FT. Birthplacement and child health. NZ Med J 1981; 93: 37-41.

7Fergusson DM, Horwood LJ, Shannon FT. Birthplacement and childhood disadvantage. Soc Sci Med1981; 15E: 315-26.

8Fergusson DM, Horwood LJ, Shannon FT. Family ethniccomposition, socioeconomic factors and childhooddisadvantage. NZ J Educ Studies 1982; 17: 171-9.

9O'Donnell JL, Fergusson DM, Horwood LJ, Shannon FT.Health care in early infancy.NZ MedJ 1978; 88: 315-7.

°Beautrais AL, Fergusson DM, Shannon FT. Childhoodaccidents in a New Zealand birth cohort. Aust Paed J1982; 18: 238-42.

"Taylor B, Wadsworth J, Butler NR. Teenage mothering:hospitalisation and accidents during the first five years.Arch Dis Child 1983; 58: 6-11.

1Beautrais AL, Fergusson DM, Shannon FT. Accidentalpoisoning in the first three years of life. Aust Paed J1981; 17: 104-9.

1Beautrais AL, Fergusson DM, Shannon FT. Life eventsand childhood morbidity: a prospective study. Pediatrics1982; 70: 935-40.

14Padilla ER, Rohsenow DJ, Bergman AB. Predictingaccident frequency in children. Pediatrics 1976; 58:223-6.

15Heisel JS, Ream S, Raitz R, et al. The significance of lifeevents as contributing factors in diseases of children. JPediatr 1973; 88: 119-23.

Meyer RJ, Roelefs HA, Bluestone J, et al. Accidentalinjury to the preschool child.JPediatr 1963; 63: 95-105.

7 Meyer RJ, Haggerty RJ. Streptococcal infections infamilies: factors altering individual susceptibility.Pediatrics 1962; 29: 539-49.

18Boyce WT, Jensen W, Cassel JC, Collier AM, Smith AH,Ramey CI. Influence of life events and family routineson childhood respiratory tract illness. Pediatrics 1977;60: 609-15.

9Sibert R. Stress in families of children who have ingestedpoisons. Br Med J 1975; 3: 87-9.

0Beautrais AL, Fergusson DM, Shannon FT. Family lifeevents and behavioural problems in preschool-agedchildren. Pediatrics 1982; 70: 774-9.

2Elley WB, Irving JC. Revised socioeconomic index forNew Zealand. NZ J Educ Studies 1976; 1: 25-36.

2Holmes TH, Rahe RH. The social readjustment ratingscale. J Psychosom Res 1967; 11: 213-8.

23Cox DR. Regression models and life tables (withdiscussion). J Royal Stat Soc B, 1972; 34: 187-220.

4Kalbfleisch JD, Prentice RL. The statistical analysis offailure time data. New York: John Wiley, 1980.

57



j.com/

J Epidem

iol Com

munity H


ownloaded from


58

25Fergusson DM, Dimond ME, Horwood LJ, Shannon FT.The utilisation of preschool health and educationservices. Soc Sci Med 1984; 19: 1173-80.

26Jemmott JB, Borysenko JZ, Borysenko M, et al.Academic stress, power motivation, and decrease insecretion rates of salivary secretion immunoglobulin A.Lancet 1983; i: 1400-3.

27Rogers MR, Dubey D, Reich P. The influence of thepsyche and the brain on immunity and diseasesusceptibility: a critical review. Psychosom Med 1979;41: 147-64.

Borysenko M, Borysenko J. Stress, behaviour andimmunity: animal models and mediating mechanisms.Gen Hosp Psychiatry 1982; 4: 59-67.

D M Fergusson, L J Horwood, and F T Shannon29Ader R, ed. Psychoneuroimmunology. New York:

Academic Press, 1981.30Bartrop RW, Lazarus L, Lockhurst E, et al. Depressed

lymphocyte function after bereavement. Lancet 1977; i:834-36.

31 Dorian BJ, Keystone W, Garfinkel PE, Brown CM.Immune mechanisms in acute psychological stress.Psychosom Med 1981; 43: 84 (abstr).

32Dohrenwend BS, Dohrenwend BP, eds. Stressful lifeevents: their nature and effects. New York: John Wiley,1974.



j.com/

J Epidem

iol Com

munity H


ownloaded from


Documents

Social and family factors in childhood hospital · Social and family factors in childhood hospital admission D MFERGUSSON,L J HORWOOD,ANDF T SHANNON Fromthe Christchurch ChildDevelopmentStudy,