In 1967, Bullard, Glaser, Heagarty, and Pivchik pointed to the difficulty that exists in establishing the precise role of the parent in failure to thrive; understanding has progressed remarkably little since then. The literature is characterised by inconsistency, and it is very difficult to be confident of the part played by maternal characteristics such as psychosocial functioning, caregiving cognitions, or childhood experiences. Current research suggests an association between parenting behaviour and infant growth retardation, but there has yet to be an adequate prospective study of the influences that guide mother and child behaviour in the interaction. This dissertation does not offer such a study: the research reported here aims to explore the long term correlates of infant growth faltering for the mother and child, by describing maternal characteristics associated with a history of failure to thrive and with the developmental consequences of non-organic failure to thrive. Such a focus is arguably justifiable, given the caveat that it does not offer a comprehensive account of potential
influences on mother or child functioning.
A descriptive analysis of maternal characteristics can be applied to examine how mothers whose children failed to thrive interpret and respond to child behaviour, and to increase
understanding of their subjective experience of past and present psychosocial stressors; this approach may be used to address the three general objectives of the present study. First, early malnourishment has previously been associated with developmental problems (e.g., Skuse et al., 1992), and so an accurate description of mother and child functioning is necessary to identify characteristics that indicate differential patterns of development among children who fail to thrive. A second issue concerns the current conceptualisation of failure to thrive as a feeding disorder (DSM-IV, American Psychiatric Association, 1996),
indicating a need to consider the mother, by virtue of her role in the feeding relationship. It should prove useful to consider current maternal functioning, through comparisons of mothers whose children failed to thrive with those who grew normally in the first 15 months of life. Finally, children who have failed to thrive may not comprise a
homogenous group (Woolston, 1985), and so analyses should serve to indicate whether it is possible to distinguish a subgroup of previously growth retarded children on the basis of mother or child characteristics at the age of six.
The present follow-up investigation of children who failed to thrive will only be useful if research addresses the methodological deficiencies of earlier investigations; as mentioned previously, the literature has been subject to criticism because of difficulties in sampling and design. The present study seeks to overcome the methodological constraints of previous research, and to examine links between current maternal functioning and the development of children who failed to thrive in infancy. Accordingly, the dissertation will adopt the following structure.
Chapter Two will outline the design of the investigation that forms the basis of this thesis, with reference to methodological issues pertinent to research in the field of failure to thrive. Subsequently, Chapter Three will explore the long term implications of early growth retardation, by comparing the current functioning of the present sample of children who failed to thrive or grew normally in infancy, in terms of their physical and psychological development. The dissertation will then go on to examine the extent to which maternal characteristics at six years are related to the occurrence, timing, and developmental
consequences of failure to thrive, by addressing areas of maternal functioning highlighted by the caregiving literature (e.g., Belsky, 1984; Mills and Rubin, 1990; Miller, 1995)
First, Chapter Four will explore the current psychological and social functioning of mothers of previously growth retarded and thriving children, in terms of maternal mental health, social support, and socio-economic circumstances. Chapter Five will then examine maternal social cognitions about events in caregiving; specifically, this section will address the mothers’ understanding of, and strategies for dealing with problem child behaviour. The third area to be examined concerns the mothers’ childhood experiences; Chapter Six will explore the extent and nature of adverse family care experienced by mothers whose children failed to thrive or grew normally in infancy. Chapter Seven aims to integrate information about each aspect of maternal functioning, to determine whether these influences inter-relate, and are associated with child functioning. Finally, Chapter Eight will critically evaluate the findings presented in this thesis, with reference to implications for research and intervention with failure to thrive.
chapter two
methodological Issues
page2.1
introduction ... 48
2 .1 .a
methodological issues... ... 48
2 .1.a.i sampling ...48 2 .1 .a.II methods of assessment... 502 .1.a.ii.a questionnaire and interview m ethods...50
2.1. a. a. b the research setting... 54
2 . 1.a.ii.c methods of assessment: conclusions... 54
2 .2
the method of the present study... 55
2 .2 .a
s a m p le ... 55
2.2.b
assessment of maternal functioning...55
2.2.b.i the mother’s interview ...552.2.b.i.a psychosocial functioning ...56
2.2.b.i.b cognitions about caregiving ...56
2.2.b.i.c childhood experiences... 57
2.2.b.ii the pilot s tu d y ... 57 2.2.b.iii p ro c e d u re ...58
2.2.C
assessment of child functioning ... 59
2.2.0.i the school visit ... 59 2.2.C.Ü the pilot study ... 60 2.2.c.iii p ro c e d u re ...612.3
ethical issues ... 61
2 .3 .a
sensitive information ... 62
2.3.b
intervention ... 63
2.3.C
confidentiality ...63
2.4
methodological limitations ... 64
2 .4 .a
the retrospective design... 64
2 .4 .b
matched pair com parisons...65
2.4.C
reliability ... 66
2.4.d
practical constraints ...67
2.5
statistical issues ...68
2 .5 .a
a conservative approach to analyses ...69
2.5.b
analyses of non-parametric data ... 70
2.5.Cmatched pairs analyses ... 70
methodological issues
2.1
introduction
Chapter One highlighted several theoretical inadequacies in the failure to thrive literature, including lack of reference to current perspectives in caregiving research. Chapter Two will consider the methodological problems of previous research; these difficulties have limited any conclusions that could be drawn from the literature about the psychological characteristics of mothers whose infants fail to thrive, and so they have implications for the design and method of the present follow-up study of mothers and their previously growth retarded children. Accordingly, this chapter will also include a general discussion of the methodological approach of the research reported here, with specific reference to sampling and the style of measurement. The present study has sought to address a range of
problems encountered in previous investigations, but it is nonetheless subject to a number of constraints, and so finally this chapter will consider ethical concerns, potential
methodological difficulties, and statistical issues.
2 .1 .a
methodological Issues
Methodological problems in failure to thrive research are widespread and take many forms. The need for empiricism has been considered in many previous reviews (notably, Skuse,
1985; Lachenmeyer and Davidovicz, 1987; Drotar, 1990), and a comprehensive discussion of methodology is not necessary here. Nevertheless, because sampling
problems are so widespread (Drotar, 1990) they warrant particular attention in this context. In addition, variability in approaches to the study of parenting has led to inconsistency in the literature, and so it is important to account for problems with measures of maternal functioning used in previous investigations.
2 .1 .a.I sampling
Theoretical progress concerning non-organic failure to thrive has been restricted by persistent sampling problems, despite a move towards controlled studies (Drotar, 1990). Research has been considerably constrained by definitional inconsistency, which limits the
comparability of results of different studies; these concerns are discussed in more detail in Chapter Three. Notably, there is disparity among researchers regarding the severity and duration of growth faltering that is defined as failure to thrive, and this variation is
aggravated by diverse approaches to the measurement and analysis of anthropometric data.
Sampling problems have also been exacerbated because investigators still tend to study hospital referred cases, although research using hospitalised infants is prone to a number of confounding variables. Foremost, it is doubtful that referred cases are representative of infants who fail to thrive. Skuse, Wolke, and Reilly (1992) provided evidence of the need for community based research from a whole population survey; only 28% of children identified as growth retarded had been referred to a paediatrician about their developmental delay, implying that children who are referred are not typical of those who fail to thrive. In addition, Mitchell, Gorrell, and Greenberg (1980) noted considerable differences
between failure to thrive in hospital referred cases and in the primary care setting, indicating that findings based on hospitalised samples cannot be generalised to a wider setting.
The use of hospital samples fails to consider the impact of referral and intervention on the family. Clinical intervention tends to emphasise the role of non-medical factors, and this emphasis probably contradicts the parents’ understanding of the syndrome; Sturm and Drotar (1991) found that mothers were most likely to attribute their child’s growth failure to medical problems, so they were probably unhappy with non-medical intervention. In addition, a non-medical approach may be seen as threatening by parents because it implies their failure to care for their child (Kotelchuk and Newberger, 1983; Miguel and Burton, 1990). Problems of sample attrition have affected a number of investigations (e.g., Drotar and Eckerle, 1989; Sturm and Drotar, 1991); these difficulties might occur because the assessment procedure is threatening or invasive in some way.
Studies of referred samples have often been additionally constrained by inadequate use of comparisons. For example, Dubowitz, Zuckerman, Bithoney and Newberger (1989) compared growth retarded and physically abused children, but failed to consider families
who were not experiencing problems, and Ramsay, Gisel and Boutry (1993) studied feeding skills among children with organic and non-organic growth retardation, but did not include observations of thriving children. Findings of other researchers (e.g., Ayoub and Milner, 1985; Drotar, Malone, Devost, Brickell et al., 1985) are considerably limited by the omission of any sort of control group. Unless growth retarded and thriving children are studied, it is impossible to determine whether findings are characteristic of failure to thrive, or whether they describe some more general trait. Furthermore, several studies that have included thriving comparisons are subject to potential bias because the investigators were not blind to case-control status (e.g., Pollitt, Eichler and Chan, 1975; Gorman, Leifer, and Grossman, 1993); this situation may not be surprising if research is carried out by the clinicians responsible for intervention.
In conclusion, studies of hospital samples may have some value in informing clinicians about the cases they are likely to see, but until improvements are made in identification and referral, hospital based research cannot provide a representative picture of failure to thrive, and so can only make a limited contribution to theoretical development. To understand the long term correlates of failure to thrive, there is a need for community-based research that samples an unreferred population of previously growth retarded children and includes a comparison group of children who thrived in infancy.
2.1 a .ii methods of assessment
2.1 .a .ii.a questionnaire and interview methods
Dowdney, Skuse, Heptinstall, Puckering and Zur-Szpiro (1987) drew attention to the variability of assessment measures in the failure to thrive literature, which may explain the diversity of research findings. The extensive use of questionnaires has probably further impeded theoretical development. Reviews of the literature (e.g., Lachenmeyer and Davidovicz, 1987; Drotar, 1991) have concluded that theory and research must account for the complexity of the syndrome; this requirement has been acknowledged in behavioural studies, which have adopted increasingly detailed empirical assessments, with, for example, the use of structured schemes for coding specific aspects of mother and infant
behaviour (e.g., Heptinstall, Puckering, Skuse, Start, Zur-Szpiro, and Dowdney, 1987; Mathisen, Skuse, Wolke and Reilly, 1989; Drotar, Eckerle, Satola, Pallotta, and Wyatt,
1990; Wolke, Skuse, and Mathisen, 1990). Less stringent requirements have been accorded to the study of parental psychological characteristics, which continues to rely on questionnaires and global assessments.
Several studies reported problems of “retardation” or variable literacy among mothers whose children do not thrive (e.g., Jacobs and Kent, 1977; Singer, Song, Hill, and Jaffe,
1990); these individuals will experience difficulty with paper and pencil measures. Singer et al. ( 1990) excluded an unspecified number of mothers from their sample because they were “intellectually unable”, “mentally impaired”, or “psychotic”, and so, presumably, they were unable to respond to the questionnaire measures used in the study. The occurrence of these characteristics in mothers of children who are not thriving surely warrants
consideration; their exclusion creates biased sampling and offers an inadequate picture of maternal characteristics associated with failure to thrive. Many studies report a low level of formal maternal education; questionnaire measures may be threatening to individuals who have limited verbal skills or who have had little experience of written tests. These concerns seem particularly apposite since earlier research with the mothers who took part in the present study found that they had diverse cognitive abilities (see Skuse, Wolke, Reilly, and Chan, 1995).
Questionnaire methods may also be criticised because they generally demand that the parent thinks in fairly abstract terms - such as rating her degree of “contentment with child”
(Casey, Bradley, and Wortham, 1984). Most people do not customarily think in abstract global terms, so maternal responses may be an artifact of the test, because the mother generates a series of transient beliefs in response to the items in the questionnaire (see Stratton, Heard, Hanks, Munton, Brewin, and Davidson, 1986). Holden and Edwards ( 1989) conducted a wide ranging review of studies of parental attitudes, which argued for a reappraisal of questionnaire methods. They also conducted an interview study o f mothers’ interpretations of a parental attitude questionnaire, and observed “an alarming amount of
error” (p 46), stemming from a range of problems, including vague wording, a mismatch between questions and experience, lack of situational specificity, and difficulties with the Likert rating scale. Holden and Edwards concluded that the standard questionnaire format can never adequately assess parental thinking, and suggested that investigations of maternal social cognition should be based on realistic stimuli reflecting commonly occurring child- rearing situations or problems.
That contention corresponds with Andrews and Brown’s (1993) observations about the efficacy of interview and questionnaire measures in a large (N = 146) sample of women from an inner-city population similar to that used in the present study. They found that an interview measure provided a more sensitive predictor of respondents’ psychological functioning, relative to more global questionnaire assessments, and concluded that
interview methods are more likely to reflect real-life situations, because they are less limited by abstract and generic descriptions. In addition, Brewin, Andrews and Gotlib’s (1993) discussion of retrospective measures of childhood experience suggested that criticisms of the reliability of retrospective recall may be overcome by use of semi-structured interview techniques, and investigator-based rating. Global judgements, of the kind required by questionnaire methods such as the Parental Bonding Instrument (Parker, Tupling and Brown, 1979), may be more subject to distortion, and the authors noted that interview methods “seem to be most suited for research assessments of specific childhood experiences” (p 93).
Interviewing may also improve sample co-operation and response rates. Cartwright’s ( 1988) comparison of the efficacy of interview and postal questionnaire methods in a study of maternity services reported a 92% response rate for interviewing, compared with 75% using questionnaires. This suggests that interviewing may be advantageous for the present research, given the concerns about sample attrition noted above. It may be concluded that semi-structured interviewing offers a useful approach for studying caregiving cognitions, psychosocial functioning and childhood care experiences among mothers in the present study. Nonetheless, it cannot be assumed that interview methods are inevitably reliable or
valid, since the information provided may be prone to bias or socially desirable responding (see e.g., Cannel and Kahn, 1968).
Interviews may be subject to bias stemming from a number of sources; Andrews (1991) noted that information may not be reported if it has been forgotten, or is considered too insignificant to be reported. The perceived threat of material for the respondent may also hinder recall and reporting, or information may be lost simply because the respondent did not know about an event. There is evidence of these difficulties from interview studies in a variety of settings. For example, Fairbum and Beglin ( 1994) compared interview and questionnaire assessments of disordered eating in a sample of young women and observed that the interview method was less successful in assessing complex symptoms such as concerns about body shape. In addition, Martin, Anderson, Romans, Mullen et al. (1993) carried out a large scale survey of retrospective accounts of childhood sexual abuse (CSA), and found higher rates of disclosure of abuse within the family using a postal questionnaire method, compared with interviewing.
Together, these observations imply that reports of personally threatening experiences such as abuse or disordered eating may be limited by an interview method, but it is noteworthy that Martin and colleagues discussed a number o f women who initially did not report CSA, but went on to give accounts of abuse in response to detailed descriptive questions during interview. This finding supports the contention that the efficacy of interviewing may depend on the nature of questioning. Andrews commented that non-judgmental wording of questions should increase the accuracy and fullness of participants’ responses, since it is less likely to cause embarrassment or threaten self-image, and Brewin et al. highlighted the importance of phrasing questions to elicit accounts of specific experiences (which are less subject to distortion than more global measures); these descriptions can then form the basis of independent investigator ratings that avoid the potential biases of participant based judgements. With care to address these requirements, it seems probable that an interview
based approach can offer valid and reliable assessment of maternal functioning in the present study.
2.1 a .ii.b the research setting
The environment in which families are studied also influences the reliability and validity of research findings, and hospital based assessments may provide an inadequate picture of family relationships. Skuse (1985) observes that mother-infant dyads are seldom “their usual selves” when interacting in the clinic; in addition, a clinical assessment of failure to thrive precludes consideration of the family context of the syndrome (Drotar, 1991) and of situation specific stressors that influence child rearing (Roberts and Maddux, 1982). Studies of parent-child interaction in a clinical setting will offer a less representative picture of family life than research carried out in the home.
An additional, pragmatic advantage for home-based research is based on the argument that visiting families at home may ensure greater co-operation, because participation requires less time and effort from respondents than a clinic-based study, in which the family has to travel to the researcher. Home-based assessment may therefore provide a valuable means of addressing the problems of sample attrition that have constrained several previous studies of failure to thrive; for example, almost 50% of Drotar and Eckerle’s (1989)