|Home | About | Journals | Submit | Contact Us | Français|
This study examined how one of the oldest and most widely distributed child welfare practice journals addressed children's mental health issues over a 25-year period. The content of 478 articles was coded. Logistic regression findings indicate that mental health issues were discussed less frequently over the first half of the period examined, and then more frequently over the last decade. Residential treatment was discussed less frequently over time, but other community-based alternatives to residential treatment were rarely discussed at any point, so that overall the content related to treatment of mental health issues decreased as discussion of residential treatment decreased. These findings suggest that although the child welfare literature has recently focused more on children's mental health, dissemination of specific concepts from the mental health to the child welfare literature does not naturally occur over time. Efforts targeted at dissemination of effective community-based mental health interventions for foster children may be needed to support this process.
The systems of care philosophy emerged from the recognition of children and adolescents with serious emotional disturbance as an underserved population with unique service needs (Stroul & Friedman, 1996). Key features of this philosophy include the need for cross-system coordination, a focus on understanding individual child and family strengths and needs, and the importance of establishing a continuum of services such that children can be served in the least restrictive setting possible. Although effectiveness research has questioned the benefits of systems of care initiatives (Bickman, 1996), these principles have become a central aspect of the philosophy of mental health practice with children (American Academy of Child and Adolescent Psychiatry, 2007).
For children involved in child welfare systems, the philosophical shifts that have occurred in the child mental health services appear to have direct implications. Coordination across systems requires increased child welfare staff involvement in mental health services planning and possibly treatment. Increased focus on family needs requires recognition of foster parent needs for training and support. The establishment of a continuum of community-based services supports a reduced emphasis on residential treatment and a stronger focus on specialized and treatment foster care as intermediate levels of care. Some evidence that child welfare has increased its focus on mental health issues can be found. For example, the first large, nationally representative study, the National Survey of Child and Adolescent Well-being, included a focus on the mental health and service needs of children involved in child protection and foster care services (Burns, Phillips, Wagner, Barth, Kolko, & Campbell, 2004), indicating that the child welfare research community, at least, views children's mental health needs as both a child welfare and a mental health services issue. Information from nationally archived sources, however, does not support that residential services are being reduced in favor of community-based alternatives. Nationally, available data suggest that an increase rather than a decrease in the use of residential settings has occurred over the past decade (U.S. Department of Health and Human Services, 1999, 2005), with an estimated 19% of foster children residing in group homes and institutions in 2003. Although nearly half of foster children are estimated to have emotional or behavioral problems requiring treatment (Burns et al., 2004), it is unclear how shifts within children's mental health services have affected child welfare.
Understanding how the child welfare field has responded to the systems of care movement is relevant to the dissemination of practice models at two different levels. First, changes that correspond to system of care principles shifts could indicate successful implementation of a coordinated system of care across service systems, consistent with the system of care movement. Additionally, examining the extent that systems of care principles are reflected within the child welfare field provides the opportunity to understand a larger issue related to the diffusion of concepts from mental health to child welfare. Indications that systems of care principles are no more evident in how children's mental health needs are addressed in child welfare over time would point to a lack of diffusion of concepts from children's mental health services to child welfare. This larger issue was a primary focus in this study: To what extent are concepts from children's mental health services transported into the child welfare field?
This question has direct implications for how new practice models are adopted. In the mental health services field, the slow rate of dissemination of evidence-based practices from the literature to real world practices has been the focus of “translational research” that focuses on the implementation of evidence-based services (Brekke, Ell, & Palinkas, 2007; Schoenwald & Henggeler, 2004). In the child welfare field, however, an additional step in this process is needed. Since information on effective mental health treatment is primarily published in other fields, this information must first be disseminated to the child welfare literature. Otherwise, practitioners who rely on the child welfare literature to understand best practices will not have access to this information. Given the high percentage of children with child welfare involvement with significant mental health needs, information about the developments in child mental health services should be reflected in the child welfare practice literature. A lack of diffusion from children's mental health services to the child welfare field would indicate the need for mechanisms to facilitate the incorporation of concepts that are relevant to the treatment of foster children.
The research reported in this article used content coding of articles from a major child welfare practice journal, Child Welfare, to explore the extent to which children's mental health issues and services were addressed over a 25-year period from 1980-2004. The primary goal of this study was to examine the extent that concepts from the children's mental health field were disseminated into the child welfare practice literature over time. We chose 1980-2004 because this period encompasses the start of the community-based systems of care orientation and the publication of Knitzer's (1982) study in Unclaimed Children, which documented the enormity of children's unmet mental health needs and provided the impetus for reform in children's mental health services. Overall, we expected that the literature would reflect child welfare's increased role in recognizing and addressing child mental health issues over time. Additionally, we hypothesized that when mental health issues were discussed, authors would be more likely to also discuss treatment, indicating greater child welfare involvement as a collaborator in treatment planning and implementation. We also expected that the child welfare literature would reflect a greater emphasis on community-based care and the role of other child-serving systems, reflecting the philosophical shifts that occurred in children's mental health services in this period.
Specifically, from 1980-2004, we expected to find the following trends:
This project involved content coding of articles from the journal Child Welfare between 1980-2004. We chose Child Welfare because of its long history of publication, its broad distribution, and the relevance of this journal to the child welfare field. As most widely distributed, oldest child welfare journal during this period, it is most likely to reflect the perspectives and concerns of the child welfare practice and research communities. Due to the large number of articles published over the last 25 years, only every other article was coded, beginning with the second article, which was randomly determined by a coin flip. Research, practice, conceptual, and historical articles were included in the analysis. However, articles were excluded if they appeared as follows: book reviews, nota bene, classified advertisements, book briefs, letters from the editor, readers forum, or folio. Furthermore, articles that addressed international issues were selected for coding but are not included in this analysis due to the variations in country systems and service delivery. Overall, 544 articles were selected for coding. A total of 66 articles were eliminated due to their international status, resulting in a sample size of 478 articles.
The development of the coding form occurred through a series of pre-tests using child welfare articles randomly selected from journals other than Child Welfare. Journals used in the construction of the coding form included Children and Youth Services Review, Child and Adolescent Social Work, Child Maltreatment, Child and Youth Care Forum, Journal of Child and Family Studies and Child Welfare Review. Approximately 30 articles were used in developing the coding sheet utilized in the analysis.
Upon completion of the coding form, the first two authors completed a pre-test with 50 articles from Child Welfare. Articles were chosen using the same criteria as described above from the years 1978 and 1979. Twenty articles were initially selected, but because the kappa coefficients were below .80 for some questions, additional testing was completed. Reliability reached .8 for all main questions of interest in the second set of 30 articles. After completing the pretest, we both coded the first 30 articles in 1980 and again inter-rater reliability was very good. The questions “was children's mental health discussed”, “were mental health services discussed”, “was residential treatment discussed”, “were juvenile justice services discussed”, and “were educational services a primary focus of the article” resulted in kappa coefficients of .86, .87, .78, 1.0, and 1.0 respectively.
Reliability checks were conducted on a regular basis after coding began. A total of 38 articles were randomly selected by SPSS in order to be coded by the other rater for reliability purposes. Inter-rater reliability revealed that again, agreement between the two coders was adequate. For the questions, “was mental health discussed”, “were mental health services discussed”, and “was residential treatment discussed”, kappa coefficients were .84, .73, and 1.0, respectively. We were unable to examine the kappa coefficients for the questions “were juvenile justice services discussed” and “was specialized or treatment foster care discussed” due to the low frequency of positive responses for these items.
The first two authors initially began coding the articles. However, due to both time constraints and the large sample size, an additional coder, the third author, was added to the project and assisted in coding articles from 1990-2004. Training consisted of four weekly meetings between the three coders in which the third coder was provided definitions used during the coding process and was given articles to code each week. The third coder was given the first 30 articles from 1980 that had been previously coded by the first two coders and used to determine inter-rater reliability. The third coder's ratings were then compared to the consensus that had been reached by the first two coders. A total of 100 articles were used for training and the pre-test for the third coder. Inter-rater reliability proved to be good, with each primary question resulting in a kappa coefficient of at least .75. Once inter-rater reliability was established, the third coder began coding articles from 1991. Periodic checks remained in place to reduce drift of ratings over time.
This variable was coded with a 1 for articles that discussed children's mental health problems and with a 0 for articles that had very minimal or no discussion of mental health problems. Discussion of children's mental health or behavioral problems was rated as present in all articles that included at least 1-2 sentences focused on children's mental health or behavioral needs, disorders, issues, or problems. Articles that mentioned mental health problems more briefly than this (i.e, in a word or phrase) were not coded as positive for discussion of mental health problems. Similarly, articles that focused on children' adjustment problems without specifying emotional or behavioral adjustment problems were not coded as positive.
This variable was also dichotomous. Children's mental health services were coded as present whenever an article discussed services focused on lessening children's mental health or behavioral problems. Mental health services could include outpatient, inpatient, or other residential services as well as non-traditional services such as foster parent training and support that specifically focused on lessening mental health problems. As with mental health problems, only articles that discussed mental health services in at least 1-2 sentences were coded as positive for child mental health services.
Treatment in restrictive settings was indicated by discussion of residential treatment in 1-2 sentences or more. Whether the authors of the article were apparent proponents of residential treatment was not coded; therefore, an article that criticized the use of residential treatment would be coded as positive for discussion of residential treatment in addition to an article focused on describing outcomes from a residential treatment center. Discussion of treatment or specialized foster care and foster parent support and training services were also coded in two separate dichotomous variables to understand whether alternatives to the treatment of children with significant needs were more likely to be discussed over time.
Coding of discussion of the juvenile justice system or services followed the same format as coding of other variables; any discussion in 1-2 sentences or more led to a positive rating for juvenile justice. Coding of educational services was more restrictive. While testing the coding form, we found that detecting when educational services were mentioned in a sentence or two was more difficult than coding juvenile justice services, since juvenile justice services were nearly always discussed in the context of delinquency, which was also included in coding as a serious behavior problem. In contrast, educational services could be mentioned in an article that did not mention any of the other variables coded for the study. To increase the accuracy of coding, educational services were coded as positive only when educational services were a major focus of the article. Thus, the indicator of educational services was a less sensitive indicator than the other indicators. Both of the resulting variables were dichotomous.
We used logistic regression analyses to test the first three hypotheses. Probability of discussion of a topic in an article was estimated in three different models. In the first model, the dependent variable was discussion of child mental health; in the second model, discussion of child mental health services; and in the third, discussion of residential treatment. The mental health services model included only the subset of articles that included a discussion of children's mental health issues, since discussion of child mental health services could only occur in articles that discussed mental health issues. Discussion of treatment foster care, juvenile justice systems, and education occurred too infrequently to use logistic regression, and so we used chi-square analyses and descriptive statistics for these analyses. Time was included in all models as a continuous independent variable that ranged from 0 (for 1980) to 25 (for 2004). Nonlinear time (quadratic and cubic) effects were tested in all models that had a significant coefficient for time. Selection of nonlinear effects in the final model was based on likelihood ratio tests.
Because testing the effect of time on the dependent variables was the central focus of the analyses, all models included control variables that varied over time that might account for a significant time effect. For example, over time, Child Welfare articles were more likely to specifically focus on child welfare services and issues (as opposed to, for example, research with a non-child welfare population), and so this variable was included as a control variable. Other control variables included a focus on families or children and discussion of abuse and neglect. Missing data were minimal (less than 2% for all analyses) and so missing data were deleted listwise from all analyses.
To assess the adequacy of the logistic regression models, the model chi-square and the Hosmer and Lemeshow goodness-of-fit test statistic were examined. A probability level of .05, assuming a two-tailed test, was used to identify statistically significant relationships in all analyses.
A total of 478 articles were coded and included in the analyses. From 1980-2004, 29.5% of the articles included content on children's mental health needs, and 11.9% included content on children's mental health services. Placement in residential treatment was discussed in just 6.3% of the articles, and treatment or specialized foster care was discussed even less (<1%). Discussion of other child-serving systems or services (i.e., juvenile justice and education) was also rare (4.2% and 1.3%, respectively).
Contrary to our hypothesis, the probability of discussion of child mental health needs decreased over time from 1980-1991. For several years after this point, the probability of discussion was essentially level, until it began to increase in 1994. In 2001, discussion of mental health issues was at the same level as in 1980. In the last three years of the observation period, probability of discussion was finally higher than in 1980. This nonlinear trend is shown in Graph 1. It should be noted that the probabilities shown in the graph cannot be generalized to the sampled articles overall, since these probabilities only indicate the effect of an article's year of publication. To calculate the probability for a particular article (Pi), the odds ratios for the other significant variables (i.e, child, family, and parent focus and child in placement) and the model constant would need to be multiplied by the odds ratio for year to calculate the article's odds ratio (Oi), which can then be used to calculate the probability for the article [Pi = Oi/(1+Oi)].
The results from the logistic regression also indicated that articles that focused on families or children and that included children who were placed in substitute care were more likely to discuss children's mental health issues than other types of articles (Table 1). A focus on child welfare issues and discussion of maltreatment were not associated with discussion of children's mental health. The Hosmer and Lemeshow statistical test for this model was nonsignificant (p=.508), indicating adequate model fit.
We did not find support for our hypothesis that discussion of children's mental health services would increase over time. In fact, in the subsample of articles that discussed child mental health issues, mental health services were discussed less over time, as indicated by the negative coefficient for year in Table 2. Child welfare content, the discussion of maltreatment, and the discussion of children in placement were nonsignificant, as shown in Table 2. Non-linear effects for time were marginally significant (p = .07) The Hosmer and Lemeshow statistic was again nonsignificant (p=.83), indicating adequate model fit.
We hypothesized that the services typically provided for children with serious mental health disorders would focus less on restrictive settings (e.g., residential treatment) and more on community-based treatment (e.g., treatment or specialized foster care) over time. Consistent with this hypothesis, the odds that an article discussed residential treatment decreased over time (Table 3). Again, Hosmer and Lemeshow statistic was nonsignificant (p = .68), indicating adequate model fit.
In contrast, no changes occurred in the probability of discussion related to less restrictive forms of care for children with mental health issues. Throughout the entire period, little discussion of specialized or treatment foster care occurred. Specialized or treatment foster care was discussed in just .6% of the articles. When we examined specialized or treatment foster care in five-year increments, we found no significant difference in the frequency of discussion across the periods [χ2 (4,478) = 1.933, p = .748]. Foster parent training and support was discussed in 1.9% of the articles and was also non-significant [χ2 (4,478) = 3.877, p = .423].
We also examined specialized or treatment foster care in five-year increments for the subsample of articles that addressed children's mental health issues. Again, we found these placements were rarely discussed (1.4%) and no significant differences occurred across the time periods [χ2 (4, 141) = 3.931, p = .415]. Foster parent training and support was also rarely discussed throughout the 25-year period (1.4%). A marginally significant difference occurred over time [χ2 (4,141) = 8.973, p = .062]. This finding appeared to be due to a higher level of discussion of foster parent training and support in the time period spanning from 1985-1989 as 7.7% of articles discussed these services in 1985-1989 in comparison to <1% in all other time periods.
We also found no support for our hypothesis that other child serving systems (e.g., juvenile justice and educational services) would be discussed more frequently over time. Juvenile justice services were discussed in 4.2% of all articles and 10.6% of the articles that discussed children's mental health issues. When we examined the incidence of discussion over five-year time periods, we found no differences across time in the entire sample [χ2 (4,478) = .631, p = .960] or in the subset of articles that included a discussion of child mental health [χ2 (4,141) = .591, p = .964].
Educational services were rarely the major focus of the articles, occurring only in 1.3% of the entire sample of articles, and in 2.8% of articles also addressing children's mental health issues. The frequency of focusing on educational services did not change over five-year time periods either in the entire sample [χ2 (4,478) = 4.905, p = .297] or in the subsample of articles discussing child mental health issues [χ2 (4,141) = 4.285, p = .369].
Foster children's unique emotional and behavioral needs have been recognized from almost the beginning of child welfare practice in the United States. As early as the 1920s and 30s, investigators attempted to identify the factors contributing to foster children's risk for emotional and behavioral problems and how to best treat these problems (Cowan & Stout, 1939; Healy, Bronner, Baylor, & Murphy, 1929). This early focus on understanding and treating children's mental health problems would suggest that current child welfare practitioners and researchers would be cognizant of trends in the children's mental health services field. Accordingly, in this study we expected to find evidence of the dissemination of systems of care principles into the child welfare practice literature over the past 25 years. Our findings, however, provide only mixed support for this hypothesis.
We found partial support for our hypothesis that there would be an increase in discussion of child mental health problems and children's mental health services from 1980-2004. Although nearly 30% of all articles had some discussion of children's mental health issues, we found that the chances that an article would discuss children's mental health problems fell during 1980-1992. After the mid-1990s, however, discussion of mental health issues was more likely over time, so that the chances that an article would include content about child mental health issues was slightly higher at the very end of the period than in the early 1980s. Although the increase in focus on behavioral issues occurred later than we expected, we may have expected too immediate a response within the child welfare field to the shifts within children's mental health services. Cross-system coordination was only occasionally mentioned in the children's mental health services literature up until the early 1990s, when results from demonstration projects funded by the Child and Adolescent Service System Program initiated by the National Institute of Mental Health began to be published (Day & Roberts, 1991). The increase in chances of discussion of mental health issues that occurred in Child Welfare during the 1990s might correspond to these developments.
Influential publications unrelated to the systems of care movement might have also contributed to the increase in discussion of child mental health issues. In particular, the Surgeon General's report on children's mental health needs might have affected perceptions about the role of the child welfare system in assessing and treating children's mental health needs, since this report emphasized integrating child mental health services into all systems that serve children and youth (U.S. Public Health Service, 2000, p. 3).
Unexpectedly, however, we did not find that the chances of discussing mental health services increased over time when discussion of child mental health issues occurred. This hypothesis was based on the expectation that as the children's mental health services field attempted to increase cross-system coordination, the child welfare practice literature would reflect a more prominent role in child treatment. In contrast, the chances of discussion of mental health services actually decreased over time in Child Welfare articles. This suggests that even though child mental health issues were addressed more often from the mid 1990s through the end of the observation period, content focused on how to provide mental health services did not follow the same trend. This might sound counter intuitive to mental health practitioners and researchers, who might assume that any focus on child mental health needs would include discussion of services as well. However, child mental health issues were frequently discussed in a more general way, or solely in relation to a particular experience, such as abuse, without any discussion of treatment.
A finding that was consistent with our hypotheses related to discussion of residential treatment over time. As we hypothesized, residential treatment was addressed significantly less often over time in the literature, despite the fact that actual use of residential treatment did not decrease. However, no changes occurred in the amount of discussion related to specialized or treatment foster care, a community-based alternative to residential treatment that is consistent with system of care principles regarding community treatment in the least restrictive setting possible. Specialized and treatment foster care were rarely topics in the articles we coded, with discussion occurring in only 1.6% of the articles that included content on child mental health issues. Articles also rarely had a major focus on educational needs (1.3% of all articles) or discussed the juvenile justice system (4.2% of all articles), providing no evidence of any increase in a cross-system orientation to addressing the needs of children involved with the child welfare system.
Overall, our findings indicate that Child Welfare has addressed residential treatment less frequently over the past 25 years and has increased its focus on child mental health issues in the past decade. However, support for the infusion of systems of care concepts from the mental health services literature to the child welfare literature is mixed, given the lack of support for our other hypotheses. Instead, the decrease in discussion of residential treatment is likely to be related to other changes related to child welfare initiatives focused on providing care in less restrictive community-based settings (Naylor, Anderson, & Morris, 2003) and the wider deinstitutionalization movement in child and adult mental health services (Neirman & Lyons, 2001). Residential treatment is the most expensive form of out of home care and has been found to have no benefits relative to treatment foster care (Chamberlain & Reid, 1998). Additionally, specialized and treatment foster care are clearly preferred placement types given that they provide care in a family setting rather than an institution. As residential treatment has fallen out of favor and states and counties have attempted to limit children's entry into restrictive settings (Naylor et al., 2003), the child welfare literature appears to have followed this trend and focused less often on residential treatment as well. Why the literature did not increase its attention on research and practice guidelines related to less restrictive care settings as this has occurred is unclear.
It is also unclear why mental health services in general were less likely to be discussed over time in articles that discussed child mental health issues. One explanation for the decrease in discussion of children's mental health services might be that as residential treatment has been addressed less frequently, the discussion of mental health services might have declined since residential treatment might have been a primary service focused on in these earlier articles. In our original hypotheses, we did not consider this possibility, since we expected the increase in focus on community-based services to be greater than the decrease in residential services. To test our alternative hypothesis, we conducted a post hoc logistic regression analysis predicting children's mental health services that included discussion of residential treatment as a predictor. In this analysis, residential treatment was a significant predictor of discussion of mental health services, as expected (OR = 4.31, p < .01), and the effect of time was nonsignificant. These findings support the explanation that the decrease mental health services content over time was accounted for by the decrease in residential treatment content.
Our study has several limitations that should be considered before discussing the implications of our findings. We chose to analyze articles from Child Welfare because of its history of publication and its unique role as the primary publication of the Child Welfare League of America. As the dominant child welfare practice journal, shifts in its content potentially affect a broad range of policy makers and practitioners. Arguably, however, our study design might have been stronger if we had sampled from a wider range of child welfare journals. A related issue is that several new child welfare journals have been started in the past decade (e.g., Child Maltreatment, and Journal of Public Child Welfare). These relatively new journals might have received a disproportionate number of articles related to residential treatment or mental health services, leading to a reduction of these types of submissions to Child Welfare over time, which would reduce the generalizability of our findings to the child welfare literature as a whole. However, our analyses did include some control variables that might capture this type of bias (i.e., whether an article focused on families and children).
Our findings also do not address whether articles focused on mental health services for child welfare involved children were more likely to be published in mental health rather than child welfare journals over time. If this did occur, however, the problem of poor dissemination of mental health practice concepts to the child welfare field remains, since mental health journals are not generally accessible to child welfare professionals.
Another limitation of our study is that we only coded for some of the key concepts found in system of care principles and the programs that implemented these principles. We chose central concepts that were relatively easy to identify, but did not code principles such as whether understanding individual child and family strengths and needs and creating individualized service plans was a central part of service planning. Including additional variables might have provided more evidence of the dissemination of systems of care concepts to the child welfare literature.
Findings from this study suggest that the child welfare practice literature was slow in disseminating new trends in child mental health to the child welfare practice community. Recognition of children's mental health issues has increased in the past decade, but content on mental health services did not follow the same trend through 2004. Given that nearly half of foster children have clinically significant mental health symptoms (Burns et al., 2003), this finding is of concern. It is essential that the child welfare field take an active role in the treatment of foster children with mental health problems. Studies have called into question the effectiveness of children's mental health services as generally practiced in the community, despite the fact that many effective treatments have been developed in controlled clinic settings (Kolko, 1996; Weiss, Catron, Harris, & Phung, 1999; Weisz, 2000). Without active child welfare involvement in identifying and disseminating effective mental health interventions, it is unlikely that foster children will receive appropriate treatment.
Child welfare has traditionally taken a central role in providing mental health treatment to emotionally and behaviorally disturbed foster children, but this treatment has primarily involved residential care. Our findings indicate that as support for residential treatment waned, the child welfare practice and research community did not maintain its focus on child treatment by focusing on another community-based option for the treatment of foster children's mental health problems. In fact, in the past decade residential treatment has been provided to more rather than fewer children and adolescents over time. Particularly since effective treatment foster care models are becoming available (Chamberlain, 2002; Fisher, Burraston, & Pears, 2005), treatment foster care is an important topic that receives little attention in the child welfare practice literature. Other community-based models, such as school-based interventions, might also be critically important for children who have child welfare involvement, given the number of moves they may experience and their poor academic outcomes (, under review). These types of service models and clinic-based models that have been shown to be effective with foster children (see Chaffin & Friedrich, 2004) should receive more attention in the literature.
We examined system of care principles to assess the extent that concepts from the child mental health literature cross over into child welfare literature to understand whether dissemination of key concepts naturally occurs over time, creating a conduit of information on appropriate treatment to the child welfare community. Our findings suggest that specific mechanisms might be needed to facilitate communication between child welfare and mental health service systems to support dissemination of information about mental health services. These mechanisms could include special issues of child welfare journals focused on evidence-based child mental health services, and similarly, special issues of child mental health journals focused on child welfare populations.
Findings from this study suggest that although the child welfare literature has focused more on children's mental health issues over the past decade, dissemination of specific concepts from the mental health to the child welfare literature does not naturally occur over time. To support this process, efforts targeted at dissemination of highly relevant interventions for foster children, such as effective treatment foster care programs, might be needed. Discussion of these types of services in the child welfare literature is only the first step in establishing effective services for children with child welfare involvement. This step, however, may be a key part of the process through which effective interventions are ultimately adopted.
This study was supported by NIMH K01 070580.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.