Relatively little is known about the characteristics of fathers who receive social worker contact. This lack of knowledge also extends to male partners of mothers who, while not biologically related to the children, may also be significant caregivers. Increased knowledge is vital to inform the provision of services for whole families. This research note examines the characteristics of these men using a sample of individuals from a large-scale British cohort study, the Avon Longitudinal Study of Parents and Children (ALSPAC). We compare the characteristics of fathers with those of mothers who have had social work contact. Furthermore we demonstrate the use of classification tree models (Breiman, Friedman, Olshen, & Stone, 1984) as a means of predicting whether fathers or mothers are more likely to receive social work contact. Classification trees have a distinct advantage in terms of modeling complex interactions and interpretation compared with the more commonly used logistic regression model. Background There is a general consensus in the child welfare field that practice tends to focus on mothers and that fathers, including stepfathers and social fathers, tend not to be engaged (Scourfield, 2003; Strega et al., 2008). There has, however, been relatively little research on fathers in child welfare cases, and there is a dearth of high-quality evidence about the characteristics of fathers who come into contact with social workers and how they compare with mothers. Some studies have been conducted using administrative data, cross-sectional surveys, or cohort studies. For example, Strega et al. (2008), using administrative data in Canada, noted high proportions of fathers on child welfare caseloads with alcohol and drug problems and violence toward the children’s mother. Berger, Paxson, and Waldfogel (2009), using the U.S. Fragile Families and Child Wellbeing Study, found that mothers living with men who were not the biological fathers of all their children were more likely to be involved with Child Protective Services. This association persisted after controlling for characteristics of mothers and fathers. Dufour, Lavergne, Larrivée, and Trocmé’s (2008) incidence study of neglect cases in Canada found that fathers struggled less with personal problems than mothers and that fathers not biologically connected to children had greater problems than biological fathers. Malm, Murray, and Geen (2006) compared nonresident fathers of foster children with the children’s mothers, in a practitioner survey conducted in the United States. They found the demographic characteristics of fathers to be broadly similar to those of mothers, but fathers were slightly older and more likely to have been married at some point. In the United Kingdom, however, the evidence on this issue has been largely qualitative, and where there is quantitative research on social work contact, this tends not to separate out parents by gender (see, for example, Henderson, Cheung, Sharland, & Scourfield, 2015). Fathers are nontraditional clients of family services, and in this context it can be challenging to engage them. More evidence is needed if service responses to fathers are to be improved. Method Data The ALSPAC is a longitudinal study of children born in Avon, United Kingdom, between 1991 and 1992. ALSPAC recruited 14,541 pregnant women with expected dates of delivery between April 1, 1991, and December 31, 1992. These initial pregnancies resulted in 14,062 live births (roughly 80% of all live births in the Avon area during this time period) with 13,988 children alive at one year of age. Individuals in the area were slightly more affluent compared with the average residents in England at the time (Boyd et al., 2013). Information about whether the mother of an ALSPAC child and the mother’s partner had been in contact with a social worker in the past year was gathered when the study child was six years and one month old. In both cases this information was gathered from questionnaire responses from the mother and her partner, respectively. One downside to this indicator is that it contains no information as to the nature of this contact, its duration or intensity, or the reason for contact. Social work in England is free at the point of use and funded by local taxation. In the late 1990s when these ALSPAC data were collected, the term “social worker” would most commonly have been used for professional staff employed by the state, whose main priority is to fulfil a legal mandate. In relation to children this includes protection from harm and meeting need, although only a very high level of need would meet the threshold for any receipt of services. Although the ALSPAC social work question is of a very general nature, it does give an indication of routine social work contact. It is important for service planning to establish the social patterning of routine use of social work. For the purposes of gender comparison, we have restricted mothers’ partners to male partners only, ignoring the 22 who were female partners (0.3% of partners in the study). In the majority of cases these men were the biological father of the ALSPAC study child (94.0%) and were living with the mother (95.5%). As such, we refer to mothers’ male partners as “fathers,” using that term inclusively, in the rest of the article. The ALSPAC also includes a variety of other information about adversities and life circumstance that are of interest. In this article we look at these circumstances or events that have occurred in the previous year. The choice of variables was influenced by previous research using ALSPAC, which looked at factors predicting the likelihood of children being placed on the child protection register before the age of seven (Sidebotham, Heron, & ALSPAC Study Team, 2006). The number of variables was also somewhat limited by the need to have same set of information for mothers and fathers in the analysis. The variables selected were whether the respondent has been depressed, got married, got divorced, has been hospitalized, has taken cannabis, has an alcohol problem, lost their job, and has gained a new job; respondents were also asked whether their partner has been emotionally cruel to them. In the first part of this research note we describe and compare the profiles of mothers and fathers who received social work contact and those who did not, to compare the demographic characteristics of both groups. In the second part of this text, we examine which events are associated with social work contact for both mothers and fathers. To do this we make use of classification tree models (Breiman et al., 1984). This is a technique more frequently used in data mining, as it approaches the problem of making predictions in ways very different from logistic regression. Classification Trees Many researchers are familiar with interpreting odds ratios from logistic regression models, but these models have a number of drawbacks. First, despite their familiarity, researchers often misunderstand what odds ratios are, commonly confusing them with relative risks (Davies, Crombie, & Tavakoli, 1998). Second, logistic regression models cannot account for complex interactions between different predictors unless these are specified by the analyst. For instance, imagine that having an alcohol problem increases the odds of an individual receiving social work, but only for men and not for women. In this case, the logistic regression model would not take this interaction into account unless it was specified in the model. There is an impossibly large number of potential interactions to include in a model—especially when we are dealing with very complex phenomena. Some of these issues with logistic regression models can be resolved by using classification trees. Like logistic regression models, classification tree models also use data to produce models that can predict the probability of whether a case falls into a certain group. Classification tree models use decision trees as their basis for making predictions. These models can be easier to interpret than logistic regression models. Figure 1 shows a sample classification tree predicting whether a passenger on the HMS Titanic survived. The model was made using a subset of the total passenger data, and it uses passenger age and gender data to create the predictive model. To understand how a prediction is made, we simply follow the branches of the decision tree. For instance, the first thing we need to know is whether a passenger was male or female. If they were female then the model predicts that they have a 75.5% chance of surviving. If the passenger was male, we need to consider the age of the passenger: those age six or younger had a 66.7% chance of survival, whereas only 17.9% of older male passengers actually survived the incident. Figure 1: View largeDownload slide Titanic Survival Model Figure 1: View largeDownload slide Titanic Survival Model The decision tree captures an interesting interaction between age and sex: If a passenger was female, then her age adds little or no extra information regarding her chances of survival, whereas the survival of male passengers is highly dependent on their age. It is possible to specify this sort of interaction between gender and age in a logistic regression model, but this interaction has to be specified in advance by the analyst. This is clearly unfeasible when there is even a modest amount of information in the data set because there are so many interactions to consider. Furthermore, prior theories or previous studies may not offer any guidance as to which interactions are important or—more likely—they may recommend far too many interactions. In the latter case, when we are fitting complex statistical models with too little data, we run the risk of creating models with results that are not generalizable to a wider population (Babyak, 2004). In contrast, classification tree models will inductively only include interactions that improve predictions. This sort of flexibility makes classification tree models particularly useful if we wish to see whether different factors are important for predicting whether a father or mother has social work contact. Classification trees models have been used in a range of fields from genetics to ecology (De’ath & Fabricus, 2000; Dudoit, Fridlyand, & Speed, 2002), but rarely in welfare-related fields such as social work (see M. A. Johnson, Brown, & Wells, 2002, for an exception). Results In ALSPAC around 4.2% of mothers with male partners had contact with a social worker. In comparison only 2.8% of fathers had contact with a social worker. There were 67 cases in which both fathers and mothers had social work contact. However, there were 95 mothers who reported social work contact where the father did not, and 50 fathers who reported social work contact where the mother did not. There were 3,954 couples where neither partner reported social work contact. Table 1 shows the proportion of individuals receiving social work contact who had experienced depression in the last year, had been hospitalized, and so forth. It also shows the proportions for those who did not have social work contact. For both mothers and fathers, those who had contact with social workers were far more likely to have had depression in the last year, to have been the recipient of emotional cruelty from his or her partner, and to have an alcohol problem. Mothers with social work contact were more likely to have been hospitalized in the last year: 23.5% compared with 13.5% for those without social work contact. In contrast, fathers with social work contact were not more likely to have been hospitalized than those without social work contact. However, fathers who had social work contact were more likely to have lost their jobs in the past year (10% compared with 5.5% of those without social work contact). Mothers who had social work contact were not more likely to have experienced job loss, possibly because fewer mothers had been working to begin with. Mothers who had social work contact were more likely to have taken cannabis in the last year (8% compared with 3.9%). Men who had social work contact were less likely to be the biological father of the ALSPAC study children than men who did not have social work contact (81.7% compared with 94.5%). However, men in both groups were equally likely to be living with the mother. Table 1: Descriptive Summary of Sample Variable Fathers Mothers No Social Work Contact (%) Social Work Contact (%) p No Social Work Contact (%) Social Work Contact (%) p Had depression in past year 25.30 53.90 <.001 21.60 40.50 <.001 Married in past year 1.00 4.20 .009 1.10 2.20 .103 Divorced in past year 0.60 0.80 .497 1.50 2.80 .057 Hospitalised in past year 7.20 10.10 .366 13.50 23.50 <.001 Taken cannabis in past year 6.60 10.00 .135 3.90 8.00 .001 Partner has been emotionally cruel 6.60 14.20 .003 6.50 14.20 <.001 Has an alcohol problem 1.70 7.60 .002 1.00 3.10 .003 Lost job in past year 5.50 10.00 .039 3.20 2.50 .624 Gained new job in past year 19.80 22.50 .413 21.50 23.60 .364 Biological father 94.50 81.70 <.001 Lives with mother 98.80 98.30 .654 Variable Fathers Mothers No Social Work Contact (%) Social Work Contact (%) p No Social Work Contact (%) Social Work Contact (%) p Had depression in past year 25.30 53.90 <.001 21.60 40.50 <.001 Married in past year 1.00 4.20 .009 1.10 2.20 .103 Divorced in past year 0.60 0.80 .497 1.50 2.80 .057 Hospitalised in past year 7.20 10.10 .366 13.50 23.50 <.001 Taken cannabis in past year 6.60 10.00 .135 3.90 8.00 .001 Partner has been emotionally cruel 6.60 14.20 .003 6.50 14.20 <.001 Has an alcohol problem 1.70 7.60 .002 1.00 3.10 .003 Lost job in past year 5.50 10.00 .039 3.20 2.50 .624 Gained new job in past year 19.80 22.50 .413 21.50 23.60 .364 Biological father 94.50 81.70 <.001 Lives with mother 98.80 98.30 .654 Table 1: Descriptive Summary of Sample Variable Fathers Mothers No Social Work Contact (%) Social Work Contact (%) p No Social Work Contact (%) Social Work Contact (%) p Had depression in past year 25.30 53.90 <.001 21.60 40.50 <.001 Married in past year 1.00 4.20 .009 1.10 2.20 .103 Divorced in past year 0.60 0.80 .497 1.50 2.80 .057 Hospitalised in past year 7.20 10.10 .366 13.50 23.50 <.001 Taken cannabis in past year 6.60 10.00 .135 3.90 8.00 .001 Partner has been emotionally cruel 6.60 14.20 .003 6.50 14.20 <.001 Has an alcohol problem 1.70 7.60 .002 1.00 3.10 .003 Lost job in past year 5.50 10.00 .039 3.20 2.50 .624 Gained new job in past year 19.80 22.50 .413 21.50 23.60 .364 Biological father 94.50 81.70 <.001 Lives with mother 98.80 98.30 .654 Variable Fathers Mothers No Social Work Contact (%) Social Work Contact (%) p No Social Work Contact (%) Social Work Contact (%) p Had depression in past year 25.30 53.90 <.001 21.60 40.50 <.001 Married in past year 1.00 4.20 .009 1.10 2.20 .103 Divorced in past year 0.60 0.80 .497 1.50 2.80 .057 Hospitalised in past year 7.20 10.10 .366 13.50 23.50 <.001 Taken cannabis in past year 6.60 10.00 .135 3.90 8.00 .001 Partner has been emotionally cruel 6.60 14.20 .003 6.50 14.20 <.001 Has an alcohol problem 1.70 7.60 .002 1.00 3.10 .003 Lost job in past year 5.50 10.00 .039 3.20 2.50 .624 Gained new job in past year 19.80 22.50 .413 21.50 23.60 .364 Biological father 94.50 81.70 <.001 Lives with mother 98.80 98.30 .654 Figure 2 displays the results of the classification tree model for predicting social work contact in the past year. The model technique can inductively fit interactions between gender and other predictors associated with the individual’s life situation in the past year, allowing us to further understand whether different factors predict social worker contact for fathers and mothers. However, the results of the final classification tree model show that it needs to use at most only four pieces of information (out of the nine variables entered) to predict social work contact: whether the participant was depressed, had an alcohol problem, experienced emotional cruelty from his or her partner, or lost his or her job in the past year. Given these four pieces of information, the respondent’s status as father or mother does not actually help us predict whether they received social work contact or not. The most important factor in making predictions was whether a respondent experienced depression in the past year; only 2.4% of those who did not experience depression had social work contact. Those who had experienced both depression and alcohol problems were almost eight times more likely to have social work contact (19% based on 58 cases, p < .001). A higher proportion of those who had been depressed, had alcohol problems, experienced cruelty, and lost a job had social work contact. However, only 10 individuals had this unique combination of life circumstances so there is a considerable margin of error around what the true proportion would be in the general population. As explained, the absence of gender as a predictor in the model also means that the same factors are used to predict social worker contact for both fathers and mothers. Figure 2: View largeDownload slide Classified Tree Model Predicting Social Work Contact Figure 2: View largeDownload slide Classified Tree Model Predicting Social Work Contact Conclusion The different profiles of mothers and fathers with social workers suggested by the bivariate analyses (Table 1) are in keeping with a gendered picture of family life, with lack of work seen as more problematic for men. The finding that nonbiological fathers were more likely to have had social work contact than biological fathers is consistent with Strega et al.’s (2008) Canadian child welfare research. However, the finding from the multivariate classification tree that there was no interaction with gender would suggest that the service needs of fathers and mothers are in fact more similar than different, namely the need for help with employment, alcohol, depression, and damaging relationships. An overall similarity of fathers’ and mothers’ problems was also noted by Malm et al. (2006) in the United States, when looking at the parents of children in foster care. Given the similarity of mothers’ and fathers’ problems, it may be that the challenge of engaging fathers in social work services lies less with the nature of their difficulties than the way of approaching them. This would imply the need for practitioners to develop novel techniques for engaging with fathers. However, two notes of caution must be sounded about the finding of similar problems for both parents. The first is in relation to the gender symmetry in reporting cruelty from a partner, because most evidence on domestic abuse from agency samples shows a highly gendered picture of men abusing women (M. P. Johnson, 2006). In light of this evidence it is possible that male perpetrators are deflecting responsibility onto their partners. Second, an acknowledged limitation of the study is that the social work variable relied on self-report and—as with all UK cohort studies—was not sufficiently detailed to capture nature, intensity, or focus of the social work interventions themselves. Meng Le Zhang, PhD, is research associate, University of Sheffield, Floor 3, ICOSS, 219 Portobello, Sheffield, UK S1 4DP; e-mail: email@example.com. Jonathan Scourfield, PhD, and Sin Yi Cheung, PhD, are professors, School of Social Science, Cardiff University, Cardiff, Wales, UK. Elaine Sharland, PhD, is professor, Social Work Research, Sussex University, Brighton, UK. The authors are extremely grateful to all the families who took part in this study, the midwives for their help in recruiting them, and the entire Avon Longitudinal Study of Parents and Children (ALSPAC) team, which includes interviewers, computer and laboratory technicians, clerical workers, research scientists, volunteers, managers, receptionists, and nurses. The UK Medical Research Council and the Wellcome Trust (grant ref: 102215/2/13/2) and the University of Bristol provided core support for ALSPAC. This research was funded by the Nuffield Foundation, grant number CPF/41218. References Babyak, M. A. ( 2004). What you see may not be what you get: A brief, nontechnical introduction to overfitting in regression-type models. Psychosomatic Medicine, 66, 411– 421. Berger, L. M., Paxson, C., & Waldfogel, J. ( 2009). Mothers, men, and Child Protective Services involvement. Child Maltreatment, 14, 263– 276. Google Scholar CrossRef Search ADS Boyd, A., Golding, J., Macleod, J., Lawlor, D. A., Fraser, A., Henderson, J., et al. . ( 2013). Cohort profile: The “Children of the 90s”—The index offspring of the Avon Longitudinal Study of Parents and Children. International Journal of Epidemiology, 42( 1), 111– 127. Google Scholar CrossRef Search ADS Breiman, L., Friedman, J. H., Olshen, R. A., & Stone, C. G. ( 1984). Classification and regression trees . Belmont, CA: Wadsworth. Davies, H. T., Crombie, I. K., & Tavakoli, M. ( 1998). When can odds ratios mislead? British Medical Journal, 316( 7136), 989– 991. Google Scholar CrossRef Search ADS De’ath, G., & Fabricus, K. E. ( 2000). Classification and regression trees: A powerful yet simple technique for ecological data analysis. Ecology, 81, 3178– 3192. Google Scholar CrossRef Search ADS Dudoit, S., Fridlyand, J., & Speed, T. ( 2002). Comparison of discrimination methods for the classification of tumors using gene expression data. Journal of the American Statistical Association, 97( 457), 37– 41. Google Scholar CrossRef Search ADS Dufour, S., Lavergne, C., Larrivée, M.-C., & Trocmé, N. ( 2008). Who are these parents involved in child neglect? A differential analysis by parent gender and family structure. Children and Youth Services Review, 30, 141– 156. Google Scholar CrossRef Search ADS Henderson, M., Cheung, S. Y., Sharland, E., & Scourfield, J. ( 2015). The effect of social work use on the mental health outcomes of parents and the life satisfaction of children in Britain. Children and Youth Services Review, 58, 71– 81. Google Scholar CrossRef Search ADS Johnson, M. A., Brown, C., & Wells, S. J. ( 2002). Using classification and regression trees (CART) to support worker decision making. Social Work Research, 26, 19– 29. Google Scholar CrossRef Search ADS Johnson, M. P. ( 2006). Conflict and control: Gender symmetry and asymmetry in domestic violence. Violence Against Women, 12, 1003– 1008. Google Scholar CrossRef Search ADS Malm, K., Murray, J., & Geen, R. ( 2006). What about the dads? Child welfare agencies’ efforts to identify, locate and involve nonresident fathers . Washington, DC: U.S. Department of Health and Human Services. Scourfield, J. ( 2003). Gender and child protection . Basingstoke, UK: Palgrave Macmillan. Google Scholar CrossRef Search ADS Sidebotham, P., Heron, J., & ALSPAC Study Team. ( 2006). Child maltreatment in the “children of the nineties”: A cohort study of risk factors. Child Abuse & Neglect, 30, 497– 522. Google Scholar CrossRef Search ADS Strega, S., Fleet, C., Brown, L., Dominelli, L., Callahan, M., & Walmsley, C. ( 2008). Connecting father absence and mother blame in child welfare policies and practice. Children and Youth Services Review, 30, 705– 716. Google Scholar CrossRef Search ADS © 2018 National Association of Social Workers This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)
Social Work Research – Oxford University Press
Published: Jan 4, 2018
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.
Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.
All the latest content is available, no embargo periods.
“Hi guys, I cannot tell you how much I love this resource. Incredible. I really believe you've hit the nail on the head with this site in regards to solving the research-purchase issue.”Daniel C.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud
“I must say, @deepdyve is a fabulous solution to the independent researcher's problem of #access to #information.”@deepthiw
“My last article couldn't be possible without the platform @deepdyve that makes journal papers cheaper.”@JoseServera