Abstract
Maternal depression, substance dependence, and the comorbidity of these conditions are highly prevalent risk factors among families involved with Child Protective Services (CPS). Data from the National Survey of Child and Adolescent Well-Being I (NSCAW I) were analyzed to examine the influence of maternal substance dependence, depression, and comorbidity on parenting and child behavior over 36-months among children reported to CPS who remained in the home at all waves. Although neglect and child behavior problems were highest for mothers with comorbidity at baseline, mothers with substance dependence had the poorest self-reported parenting and child behavior problems over time. Results indicate a need for intensive targeted services to address the complex needs of CPS-involved mothers with substance dependence and their in-home children.
Keywords: Child welfare, Comorbid conditions, Maternal depression, Maternal substance dependence, Child behavior problems, Parenting, Child abuse and neglect
Introduction
Maternal depression and substance abuse are highly prevalent among mothers involved with the child welfare system (Besinger et al. 1999; Jones 2004; US DHHS 2005). Major depressive disorder in the past 12 months is more common among caregivers reported to CPS (23.3 %; US DHHS 2005) than among the general population of American adults (6.7 %; Kessler et al. 2005). Rates of caregiver substance abuse are also estimated to be higher in the child welfare system than among all American families (Besinger et al. 1999; Jones 2004). Estimates of caregiver substance abuse within CPS vary drastically depending on the level of severity used as a cut-off to indicate substance abuse and the level of CPS involvement of the sample (Jones 2004). Jones’ data suggests that substance abuse rates are progressively higher with deeper levels of involvement of the sample with CPS. Therefore, families whose children are in foster care tend to have the highest rates of substance abuse (US GAO 1998). Furthermore, depression and substance abuse are frequently comorbid, with an estimated one-third of depressed adults also having a substance use disorder (Davis et al. 2005).
Maternal depression and substance abuse result in adverse consequences that have long lasting effects on children. When caregiver substance abuse is present, children often suffer numerous behavioral and developmental consequences, including lower academic achievement, truancy, suicidal involvement, teenage pregnancy, eating disorders, and the use of tobacco, alcohol, and marijuana (Chandy et al. 1996; Hill et al. 2011). Families in which maternal substance abuse is present are more likely to experience more frequent and longer foster care placements, and are more likely to exit the foster care system through adoption (Semidei et al. 2001; Vanderploeg et al. 2007). Children of depressed parents are also at risk for adverse outcomes. Zuckerman and Beardslee (1987) highlighted findings from observational studies that found cognitive, emotional, behavioral and physical problems in children of depressed mothers. Subsequent research has identified the association between maternal depression and children’s increased risk of internalizing and externalizing problems (Kim-Cohen et al. 2005). Both maternal substance abuse and maternal depression have been strongly associated with child maltreatment (Chaffin et al. 1996). Given the high prevalence rates and negative effects, maternal substance abuse and maternal depression are pressing problems for the CPS system.
The combined impact of a mother’s substance abuse and depression may negatively influence her children’s behavior and development. Although research examining the separate impact of maternal depression and maternal substance dependence on parenting and child outcomes is valuable (Hill et al. 2011; Mustillo et al. 2011), a more comprehensive picture of the outcomes and needs when these conditions are comorbid is warranted given the frequency of comorbidity. The purpose of this study is to examine the influence of maternal substance dependence, maternal depression, and comorbid substance dependence and depression on parenting and child behavior problems in a sample of one of the country’s most vulnerable populations, children and their mothers investigated for child maltreatment by CPS. The study provides additional insight into the needs of this complex CPS population and tests the following hypotheses:
Mothers with comorbid substance dependence and depression will report higher rates of neglectful parenting over a 36-month period compared to mothers with substance dependence alone, depression alone, or neither condition.
Mothers with comorbid substance dependence and depression will report higher rates of child physical assault over a 36-month period compared to mothers with substance dependence alone, depression alone, or neither condition.
Mothers with comorbid substance dependence and depression will report higher rates of emotional maltreatment over a 36-month period compared to mothers with substance dependence alone, depression alone, or neither condition.
Children of mothers with comorbid substance dependence and depression will have poorer behavior scores over a 36-month period than children of mothers with substance dependence alone, depression alone, or neither condition.
Methods
To examine these hypotheses, data from the National Survey of Child and Adolescent Well-Being I (NSCAW I) were analyzed. NSCAW I is a longitudinal, national probability sample of children, ages 0 to 14 at the time of the investigation, and their families (n=5501) who were investigated for child maltreatment by CPS between October 1999 and December 2000. NSCAW I utilizes a two stage stratified sampling process. In the first stage, nine sampling strata were constructed to allow sampling of the entire United States. In the second stage, 81 primary sampling units (PSUs) were randomly selected from across the United States. Each PSU represents one CPS agency or a group of smaller CPS agencies and typically correspond to a county. Families were sampled from within these counties. Weighting and clustering variables were then used to provide national estimates using the PSUs. Data were collected at four time points: baseline (Wave 1), approximately 12-months post baseline (Wave 2), approximately 18-months post baseline (Wave 3), and approximately 36-months post baseline (Wave 4). Baseline data were collected between November 1999 and April 2001. This analysis utilizes data from time points when parenting measures were administered: waves 1, 3 and 4. Data used in these analyses were collected from primary caregivers, CPS caseworkers, and the child during in-person interviews. Detailed additional information about the NSCAW I study design and data collection can be found elsewhere (NDACAN 2006).
Sample
Three inclusion criteria were used to draw a subsample (n= 2204) for these analyses. First, for all hypotheses, the child must have remained in the home at waves 1, 3, and 4. The Parent Child version of the Conflict Tactics Scale (CTS-PC) was only administered when the child was in the home at the time of the interview. Therefore, all children who were placed out-of-the home at any interview were excluded, in order to examine the characteristics of a consistent sample over time. Although many studies focus on children who are removed from the home, the majority of CPS-involved children (87 %) remain in the home following a CPS investigation (Dolan et al. 2011). This analysis will focus on this understudied population. Second, the primary caregiver had to be a biological, step, or adoptive mother. Third, in order to utilize consistent measures, the child had to be age two or above at baseline. Children who were younger than two at baseline were excluded from the analyses at all time points. In NSCAW I, only one child was sampled per home so children were not nested within caregivers.
A missing data analysis was conducted to compare differences in sample characteristics (children’s demographics, mothers’ demographics, family income, and the type of child maltreatment) between those reporting and those with a missing value for each dependent variable. Only one significant difference was seen across all of these comparisons. Respondents with missing data on the wave 3 dependent variables (n=364 to 370, with slight variation by dependent variable) had a slightly lower prevalence of living at or below the poverty line (approximately 48 %, with slight variation by dependent variable) than those respondents with complete data at wave 3 (approximately 54 %). Family income was controlled for in the regression analyses.
Dependent Variables
Four separate dependent variables (neglectful parenting, emotional maltreatment, physical assault, child behavior problems) were examined at waves 1, 3, and 4. The parenting variables of neglectful parenting, emotional maltreatment, and physical assault were assessed by maternal self-report on three different subscales of the CTS-PC (Straus et al. 1998). Each of the CTS-PC subscales provides a continuous range of scores with higher scores indicating higher levels of neglectful parenting, emotional maltreatment, or physical assault in the past 12 months. For this analysis, each subscale was dichotomized into those who reported one or more problematic parenting behaviors and those who did not report any of these behaviors. The continuous CTS-PC subscales were highly skewed due to a high occurrence of zero values. Dichotomization was chosen after data transformations failed to normalize these data. Behaviors on the neglectful parenting subscale include leaving the child home alone, not providing food, substance use which prevents providing care to the child, lack of medical care, or the mother not telling the child that she loves him or her. The psychological aggression subscale of the CTS-PC assessed for self-reported emotional maltreatment of the child by the mother in the past 12 months and included items to assess for the following behaviors: threatening to hurt the child, shouting/yelling/ screaming at child, swearing at child, name calling, and telling the child he or she will be sent away. Behaviors assessed on the physical assault subscale of the CTS-PC indicating concerns for physical assault of the child include hitting with a fist, kicking, beating, grabbing around the neck, burning, hitting with a hard object somewhere other than the buttocks, threatening with a knife, or throwing or knocking down the child.
Using NSCAW data, internal consistency for the caregiver report on the CTS-PC is good for the total score (Cronbach’s α=0.92) (NDACAN 2006). Internal consistency for the sub-scales range from moderate for the Psychological Aggression subscale (Cronbach’s α=0.66) to good for the very severe physical assault subscale (Cronbach’s α=0.95). Using data collected during telephone surveys of a nationally representative sample of American parents (n=1000), Straus et al. (1998) found support for construct validity of the CTS-PC in the sample by testing for correlations between the scales and demographic variables which should be present based on theoretical relationships supported by empirical literature. See Straus et al. (1998) for a full discussion of these results.
The fourth dependent variable, child behavior problems, was assessed with age-appropriate versions of the Child Behavior Checklist (CBCL): CBCL for 2–3 (Achenbach 1992) or CBCL for 4–18 (Achenbach 1991). A composite variable was created indicating behavior scores for all children ages two and above. Scores on the CBCL indicate normal (score < 60), borderline (score 60–63), or clinical (score > 63) child behavior concerns. The CBCL measures both internalizing behaviors and externalizing behaviors displayed by the child. These variables were reported separately. In NSCAW I, internal consistency of the internalizing and externalizing sub-scales of the CBCL was moderate to good (0.80 for 2 to 3 year olds; 0.90 for children 4 and older) on the internalizing sub-scale and averaging 0.91 for children 2 to 3 years old and 0.92 for children 4 and older on the externalizing subscale (NDACAN 2006). Internal consistency for the total problem behaviors scores on the CBCL are good for both 2 to 3 year olds (Cronbach’s α=0.95) and children 4 and older (Cronbach’s α=0.96). Commonly used in child welfare research, the CBCL has good reliability and validity (Achenbach 1991; Burns et al. 2004).
Independent Variables
Maternal substance dependence and maternal depression were both measured at wave 1 using the Composite International Diagnostic Interview-Short Form (CIDI-SF; Kessler et al. 1998). The CIDI-SF is a structured interview designed to screen for common psychiatric diagnoses, including substance dependence and depression, using diagnostic criteria established in the DSM-IV. Substance dependence and depression were both dichotomized for this analysis so that each mother either met clinical criteria for substance dependence or not and, separately, met the clinical threshold for depression or not. Dichotomization also allowed for the creation of the co-morbidity variable. Mothers were categorized as positive for substance dependence if their CIDI-SF scores indicated they were positive for alcohol dependence and/or drug dependence. Substance dependence denotes a severe level of substance use and likely excludes a large percentage of parents whose substance use may still negatively impact their parenting. On the CIDI-SF, responses endorsing the presence of Dysphoric or Anehedonic Depression symptoms result in additional questions being asked which determined if mothers met the diagnostic criteria for a major depressive episode or not. A four-category independent variable was constructed using the dichotomous variables for substance dependence and depression. Mothers were categorized as either having comorbid substance dependence and depression, substance dependence only, depression only, or neither condition. The reliability and validity of the CIDI has been thoroughly studied and found to be good (Andrews and Peters 1998; Ustun et al. 1997).
Control Variables
Control variables in the analyses include children’s demographics (gender, age), mothers’ demographics (race, age, educational level), family income, and the type of child maltreatment. All control variables used in this analysis were measured at wave 1. Categorical control variables include child gender (male or female), mother’s self-reported race (White or Non-white), mother’s self-report of highest degree held (no degree or high school degree/GED/higher level degree), care-giver’s report of family income (at/below poverty line or above poverty line) based on family income and the number of adults and children in the household at baseline. Child age and mother age were continuous variables based on maternal report at baseline and measured in years. Using an algorithm based on a modified version of the Maltreatment Classification System (Manly et al. 1994), the most serious type of maltreatment identified in NSCAW I was regrouped into five categories: 1) physical abuse, 2) sexual abuse, 3) failure to provide adequate food, clothing, shelter, or care, 4) failure to supervise, and 5) other types of maltreatment. Data about the types of maltreatment identified at the time of the baseline investigation were reported by the CPS caseworkers.
Data Analysis Strategy
First, univariate analyses were run to obtain descriptive characteristics of the sample (Table 1). To examine the prevalence of dependent variables at each wave, weighted frequencies of each dependent variable over time were examined (Table 2). Finally, repeated measures logistic regression with generalized estimating equations (GEE) was utilized to examine the main effects and interaction models (Diggle et al. 2002). GEE is a method of analysis that accounts for the correlated data associated with the repeated measures of our dependent variables and produces more accurate standard errors and p-values. The xtgee command in Stata was used for GEE, with the binomial specification for family to indicate the binary dependent variables represented by the dichotomized outcomes. Additionally, compound symmetry was obtained by using exchangeable for the correlation specification among the binary outcomes.
Table 1.
Sample characteristics at baseline (weighted)
| Variable | Mean or percentage |
|---|---|
| Child age | 7.86 years (SD=0.15) |
| Mother age | 32.06 years (SD=0.30) |
| Child gender | |
| Male | 50.59 % |
| Female | 49.41 % |
| Mother’s race | |
| White | 54.99 % |
| Non-white | 45.01 % |
| Family’s income | |
| At or below poverty threshold | 53.46 % |
| Above poverty threshold | 46.54 % |
| Primary maltreatment type | |
| Physical abuse | 29.02 % |
| Sexual abuse | 13.99 % |
| Neglect: fail to provide food, clothing, shelter, or care | 18.55 % |
| Neglect: fail to supervise | 27.58 % |
| Other maltreatment | 10.86 % |
| Mother’s educational attainment | |
| Less than high school | 29.10 % |
| GED/High school degree or higher education | 70.90 % |
| Maternal comorbidity | |
| Neither condition | 75.11 % |
| Substance dependence only | 1.34 % |
| Depression only | 20.88 % |
| Substance dependence & depression | 2.67 % |
SD Standard deviation
Table 2.
Dependent variables by wave (weighted)
| Wave 1 | Wave 3 | Wave 4 | Across waves | |
|---|---|---|---|---|
| Physical assault | ||||
| Once or more | 12.44 % | 7.81 % | 9.28 % | 9.91 % |
| None reported | 87.56 % | 92.19 % | 90.72 % | 90.09 % |
| Neglectful parenting | ||||
| Once or more | 41.72 % | 31.75 % | 33.91 % | 35.93 % |
| None reported | 58.28 % | 68.25 % | 66.09 % | 64.07 % |
| Emotional maltreatment | ||||
| Once or more | 64.84 % | 57.59 % | 55.76 % | 59.47 % |
| None reported | 35.16 % | 42.41 % | 44.24 % | 40.53 % |
| Child behavior problems | ||||
| Total CBCL | ||||
| Clinical level | 31.30 % | 27.92 % | 25.23 % | 28.18 % |
| Borderline | 9.56 % | 8.01 % | 8.82 % | 8.82 % |
| Normal | 59.14 % | 64.08 % | 65.95 % | 63.00 % |
| Internalizing | ||||
| Clinical level | 20.23 % | 18.62 % | 15.21 % | 18.02 % |
| Borderline | 11.44 % | 7.51 % | 8.22 % | 9.11 % |
| Normal | 68.34 % | 73.87 % | 76.57 % | 72.87 % |
| Externalizing | ||||
| Clinical level | 29.44 % | 26.90 % | 25.94 % | 27.45 % |
| Borderline | 10.82 % | 10.94 % | 10.45 % | 10.73 % |
| Normal | 59.74 % | 62.17 % | 63.62 % | 61.82 % |
Six separate repeated measures logistic regression analyses were modeled. One model each was run to test each of the first three hypotheses (Table 3). The four-category independent variable, comorbidity, was regressed on each independent variable. Interactions between comorbidity and wave were analyzed to assess differences in rates of change between conditions. Next, three models were used to examine the total CBCL scores, internalizing subscale of CBCL, and externalizing subscale of CBCL (Table 5). Data management was conducted in SAS 9.3. Data were analyzed in STATA/ SE 12.1, which accounted for stratification, clustering, and weighting variables that allow national estimates to be made.
Table 3.
Logistic regression results for parenting on comorbidity
| Model | Variable | OR | SE | Z | 95 % CI | Model Wald X2 |
|---|---|---|---|---|---|---|
| Neglectful parenting on comorbidity (n=1442) | 85.24*** | |||||
| Comorbidity | ||||||
| Neither (reference) | 1.00 | |||||
| Substance dependence only | 1.99 | 1.00 | 1.37 | [0.74–5.32] | ||
| Depression only | 1.90 | 0.39 | 3.16** | [1.28–2.83] | ||
| Both | 2.08 | 0.69 | 2.23* | [1.09–3.97] | ||
| Child age | 1.17 | 0.03 | 6.42*** | [1.11–1.22] | ||
| Child gender | 0.96 | 0.15 | −0.26 | [0.71–1.30] | ||
| Mother age | 0.98 | 0.01 | −1.45 | [0.96–1.01] | ||
| Mother race | 1.27 | 0.20 | 1.46 | [0.92–1.73] | ||
| Mother education | 1.07 | 0.19 | 0.40 | [0.76–1.52] | ||
| Family income | 1.02 | 0.18 | 0.09 | [0.72–1.43] | ||
| Most serious maltreatment | ||||||
| Physical abuse (reference) | 1.00 | |||||
| Sexual abuse | 1.29 | 0.33 | 1.01 | [0.78–2.13] | ||
| Failure to provide | 1.22 | 0.33 | 0.71 | [0.71–2.08] | ||
| Failure to supervise | 1.35 | 0.26 | 1.54 | [0.92–1.97] | ||
| Other maltreatment | 1.26 | 0.36 | 0.82 | [0.72–2.19] | ||
| Wave | ||||||
| Wave 1 (reference) | 1.00 | |||||
| Wave 2 | 0.63 | 0.08 | −3.76*** | [0.50–0.80] | ||
| Wave 3 | 0.65 | 0.09 | −3.10** | [0.50–0.85] | ||
| Physical assault on comorbidity (n=1442) | 47.07*** | |||||
| Comorbidity | ||||||
| Neither (reference) | 1.00 | |||||
| Substance dependence only | 2.93 | 1.57 | 2.00* | [1.02–8.39] | ||
| Depression only | 1.23 | 0.29 | 0.87 | [0.77–1.94] | ||
| Both | 1.76 | 0.89 | 1.12 | [0.65–4.72] | ||
| Child age | 1.07 | 0.04 | 2.13* | [1.01–1.15] | ||
| Child gender | 1.51 | 0.33 | 1.87 | [0.98–2.33] | ||
| Mother age | 1.01 | 0.02 | 0.43 | [0.97–1.05] | ||
| Mother race | 1.92 | 0.41 | 3.03** | [1.26–2.92] | ||
| Mother education | 1.41 | 0.33 | 1.46 | [0.89–2.23] | ||
| Family income | 0.68 | 0.15 | −1.75 | [0.44–1.05] | ||
| Most serious maltreatment | ||||||
| Physical abuse (reference) | 1.00 | |||||
| Sexual abuse | 0.63 | 0.22 | −1.31 | [0.32–1.26] | ||
| Failure to provide | 0.91 | 0.33 | −0.25 | [0.45–1.85] | ||
| Failure to supervise | 0.88 | 0.24 | −0.48 | [0.51–1.49] | ||
| Other maltreatment | 0.50 | 0.17 | −2.01* | [0.24–0.98] | ||
| Wave | ||||||
| Wave 1 (reference) | 1.00 | |||||
| Wave 2 | 0.54 | 0.11 | −2.94** | [0.36–0.81] | ||
| Wave 3 | 0.55 | 0.14 | −2.32* | [0.34–0.91] | ||
| Emotional maltreatment on comorbidity (n=1442) | 60.38*** | |||||
| Comorbidity | ||||||
| Neither (reference) | 1.00 | |||||
| Substance dependence only | 3.16 | 1.41 | 2.57* | [1.31–7.59] | ||
| Depression only | 1.53 | 0.30 | 2.15* | [1.04–2.26] | ||
| Both | 2.01 | 1.09 | 1.30 | [0.70–5.80] | ||
| Child age | 1.16 | 0.03 | 5.43*** | [1.10–1.22] | ||
| Child gender | 0.99 | 0.15 | −0.09 | [0.72–1.34] | ||
| Mother age | 0.97 | 0.01 | −2.56* | [0.94–0.99] | ||
| Mother race | 1.05 | 0.17 | 0.32 | [0.78–1.45] | ||
| Mother education | 1.18 | 0.22 | 0.87 | [0.82–1.70] | ||
| Family income | 1.03 | 0.17 | 0.20 | [0.74–1.44] | ||
| Most serious maltreatment | ||||||
| Physical abuse (reference) | 1.00 | |||||
| Sexual abuse | 0.59 | 0.15 | −2.02* | [0.35–0.98] | ||
| Failure to provide | 0.79 | 0.20 | −0.91 | [0.48–1.30] | ||
| Failure to supervise | 0.90 | 0.18 | −0.55 | [0.61–1.32] | ||
| Other maltreatment | 1.18 | 0.30 | 0.67 | [0.72–1.93] | ||
| Wave | ||||||
| Wave 1 (reference) | 1.00 | |||||
| Wave 2 | 0.71 | 0.09 | −2.71** | [0.56–0.91] | ||
| Wave 3 | 0.69 | 0.08 | −3.44** | [0.56–0.85] |
Note.
p<.05,
p<.01,
p<.001
Table 5.
Logistic regression results for child behavior problems on comorbidity
| Model | Variable | Exp(b) | SE | Z | 95 % CI | Model Wald X2 |
|---|---|---|---|---|---|---|
| Total CBCL scale on comorbidity (n=1701) | 202.75*** | |||||
| Comorbidity | ||||||
| Neither (reference) | 1.00 | |||||
| Substance dependence only | 1.68 | 0.24 | 3.67*** | [1.27–2.22] | ||
| Depression only | 1.47 | 0.06 | 9.31*** | [1.36–1.60] | ||
| Both | 1.36 | 0.16 | 2.67** | [1.09–1.70] | ||
| Child age | 1.03 | 0.01 | 4.69*** | [1.02–1.04] | ||
| Child gender | 0.91 | 0.03 | −2.45* | [0.85–0.98] | ||
| Mother age | 1.00 | 0.00 | −1.03 | [0.99–1.00] | ||
| Mother race | 0.85 | 0.03 | −4.29*** | [0.79–0.92] | ||
| Mother education | 0.87 | 0.04 | −3.28** | [0.81–0.95] | ||
| Family income | 0.92 | 0.03 | −2.16* | [0.86–0.99] | ||
| Most serious maltreatment | ||||||
| Physical abuse (reference) | 1.00 | |||||
| Sexual abuse | 0.94 | 0.05 | −1.06 | [0.85–1.05] | ||
| Failure to provide | 0.89 | 0.05 | −2.03* | [0.80–0.996] | ||
| Failure to supervise | 0.85 | 0.04 | −3.37** | [0.77–0.93] | ||
| Other maltreatment | 0.86 | 0.06 | −2.27* | [0.76–0.98] | ||
| Wave | ||||||
| Wave 1 (reference) | 1.00 | |||||
| Wave 2 | 0.93 | 0.02 | −3.17** | [0.89–0.97] | ||
| Wave 3 | 0.90 | 0.02 | −4.83*** | [0.86–0.94] | ||
| Internalizing subscale on comorbidity (n=1701) | 205.70*** | |||||
| Comorbidity | ||||||
| Neither (reference) | 1.00 | |||||
| Substance dependence only | 1.49 | 0.19 | 3.19** | [1.17–1.91] | ||
| Depression only | 1.36 | 0.05 | 8.28*** | [1.26–1.46] | ||
| Both | 1.44 | 0.15 | 3.62*** | [1.18–1.76] | ||
| Child age | 1.02 | 0.01 | 3.62*** | [1.01–1.03] | ||
| Child gender | 0.83 | 0.03 | −5.71*** | [0.78–0.89] | ||
| Mother age | 1.00 | 0.00 | 0.46 | [1.00–1.01] | ||
| Mother race | 0.97 | 0.03 | −1.00 | [0.91–1.03] | ||
| Mother education | 0.89 | 0.03 | −3.28** | [0.83–0.95] | ||
| Family income | 0.97 | 0.03 | −0.78 | [0.91–1.04] | ||
| Most serious maltreatment | ||||||
| Physical abuse (reference) | 1.00 | |||||
| Sexual abuse | 0.97 | 0.05 | −0.64 | [0.88–1.07] | ||
| Failure to provide | 0.98 | 0.05 | −0.47 | [0.88–1.08] | ||
| Failure to supervise | 0.88 | 0.04 | −2.90** | [0.81–0.96] | ||
| Other maltreatment | 0.96 | 0.06 | −0.73 | [0.86–1.07] | ||
| Wave | ||||||
| Wave 1 (reference) | 1.00 | |||||
| Wave 2 | 0.91 | 0.02 | −4.37*** | [0.88–0.95] | ||
| Wave 3 | 0.87 | 0.02 | −6.79*** | [0.83–0.91] | ||
| Externalizing subscale on comorbidity (n=1701) | 200.04*** | |||||
| Comorbidity | ||||||
| Neither (reference) | 1.00 | |||||
| Substance dependence only | 1.78 | 0.25 | 4.15*** | [1.35–2.33] | ||
| Depression only | 1.41 | 0.06 | 8.52*** | [1.30–1.53] | ||
| Both | 1.24 | 0.14 | 1.89 | [0.99–1.54] | ||
| Child age | 1.04 | 0.01 | 6.23*** | [1.02–1.05] | ||
| Child gender | 1.00 | 0.04 | −0.13 | [0.93–1.07] | ||
| Mother age | 0.99 | 0.00 | −1.87 | [0.99–1.00] | ||
| Mother race | 0.85 | 0.03 | −4.54*** | [0.79–0.91] | ||
| Mother education | 0.87 | 0.03 | −3.46** | [0.81–0.94] | ||
| Family income | 0.90 | 0.03 | −2.99** | [0.83–0.96] | ||
| Most serious maltreatment | ||||||
| Physical abuse (reference) | 1.00 | |||||
| Sexual abuse | 0.92 | 0.05 | −1.49 | [0.83–1.03] | ||
| Failure to provide | 0.86 | 0.05 | −2.72** | [0.77–0.96] | ||
| Failure to supervise | 0.84 | 0.04 | −3.65*** | [0.76–0.92] | ||
| Other maltreatment | 0.81 | 0.05 | −3.31** | [0.71–0.92] | ||
| Wave | ||||||
| Wave 1 (reference) | 1.00 | |||||
| Wave 2 | 0.96 | 0.02 | −2.10* | [0.92–0.997] | ||
| Wave 3 | 0.94 | 0.02 | −2.83** | [0.90–0.98] |
Note.
p<.05,
p<.01,
p<.001
Results
Sample Characteristics
Sample characteristics at baseline are reported in Table 1 and account for weighting. Next, weighted frequencies of the prevalence of each dependent variable by wave were run (Table 2). Across waves, self-reported emotional maltreatment (59.47 %) was more prevalent than neglectful parenting (35.93 %), or physical assault (9.91 %). The prevalence of each self-reported negative parenting behavior was highest at baseline. Total scores on the CBCL across waves indicated that over one-third of the children in the sample scored in the clinical (28.18 %) or borderline range (8.82 %) (Table 2). For the internalizing subscale of the CBCL, slightly over one-quarter of children in the sample scored in the clinical (18.02 %) or borderline (9.11 %) range across waves. Across waves, 27.45 % of children scored in the clinical range on the externalizing subscale with another 10.73 % in the borderline range. Clinical levels, indicating a need for mental health treatment, fell at each wave for all measures of child behavior problems.
Bivariate Analyses
Chi-square tests were run to examine the relationship between comorbidity and the categorical sample characteristic variables, parenting variables across waves, and child behavior scores across waves. Only statistically significant associations are presented. Comorbidity was associated with family income (Wald χ2(3)=12.30, p<.05), maternal report of neglectful parenting across waves (Wald χ2(3)=15.87, p<.01), maternal report of emotional maltreatment (Wald χ2(3)=12.07, p<.05), total scores on the CBCL (Wald χ2(6)=31.20, p<.001), child internalizing behavior scores (Wald χ2(6)= 20.61, p<.01), and child externalizing behavior scores (Wald χ2(6)=33.27, p<.001).
Logistic Regression Analyses
Parenting Behavior Analyses
Controlling for child age and gender, maternal age, race, education, family income, most serious type of maltreatment, and wave, the models regressing maternal self-report of neglectful parenting on comorbidity (Wald χ2(15)=85.24, p<.0001), maternal self-report of physical assault on comorbidity (Wald χ2(15)=47.07, p<.0001), and maternal self-report of emotional maltreatment on comorbidity (Wald χ2(15)= 60.38, p<.0001) were statistically significant (Table 3). Significant relationships were seen between substance dependence and depression with each parenting model. The odds of a mother with depression only (OR=1.90) or a mother with comorbidity (OR=2.08) self-reporting neglectful parenting were around two times the odds of a mother with neither condition. The odds of a mother with substance dependence only self-reporting physical assault was almost three times (OR=2.93) the odds of a mother with neither condition. Finally, the odds of a mother with substance dependence only (OR=3.16) or a mother with depression only (OR=1.53) self-reporting emotional maltreatment were higher than for mothers with neither condition. For all three models, a significant relationship was seen between self-reported negative parenting behaviors and wave with the odds of reporting negative parenting behaviors decreasing from wave 1 to wave 2 and wave 1 to wave 3.
In order to examine differences in parenting behaviors by both wave and comorbidity while controlling for other variables in the model, predicted probabilities were run. An example of how to interpret the predicted probabilities is presented for neglect (Table 4). Predicted probabilities for self-reported neglect by wave and comorbidity indicated statistically significant differences in the 12 probabilies, F(11, 72)= 5.66, p<.0001, and an interaction of comorbidity by wave, F(6, 77)=2.68, p<.05. At baseline, predicted probabilities for self-reported neglect were highest for mothers with comorbidity (0.76) compared to substance dependent (0.52), depressed (0.58), and neither condition (0.36). At wave 3, predicted probabilities for all groups decreased with the highest rates for mothers with substance dependence only (0.46) followed by depression only (0.42), neither condition (0.29), and comorbidity (0.26). At wave 4, predicted probabilities of self-reporting neglect increased slightly for the group with neither condition (0.30), decreased for mothers with depression only (0.38), and increased but to levels lower than baseline for mothers with substance dependence only (0.50) and comorbidity (0.47). Overall, neglect declined between waves 1 and 4, but at wave 4 it remained highest among mothers with substance dependence only at baseline.
Table 4.
Predicted probabilities (weighted)
| Wave 1 | Wave 3 | Wave 4 | Difference in 12 means or probabilities | Interaction of comorbid* wave | |
|---|---|---|---|---|---|
| Neglectful parenting | F(11, 72)=5.66*** | F(6, 77)=2.68* | |||
| Neither condition | 0.36 | 0.29 | 0.30 | ||
| Substance dep. only | 0.52 | 0.46 | 0.50 | ||
| Depression only | 0.58 | 0.42 | 0.38 | ||
| Comorbidity | 0.76 | 0.26 | 0.47 | ||
| Physical assault | F(11, 72)=3.38*** | F(6, 77)=2.44* | |||
| Neither condition | 0.10 | 0.05 | 0.08 | ||
| Substance dep. only | 0.36 | 0.22 | 0.03 | ||
| Depression only | 0.13 | 0.10 | 0.04 | ||
| Comorbidity | 0.29 | 0.05 | 0.03 | ||
| Emotional maltreatment | F(11, 72)=3.13** | F(6, 77)=1.07 | |||
| Neither condition | 0.62 | 0.54 | 0.53 | ||
| Substance dep. only | 0.90 | 0.61 | 0.89 | ||
| Depression only | 0.72 | 0.64 | 0.61 | ||
| Comorbidity | 0.78 | 0.57 | 0.75 | ||
| Total CBCL | F(11, 72)=8.14*** | F(6, 77)=1.02 | |||
| Neither condition | 54.6 | 54.0 | 52.8 | ||
| Substance dep. only | 62.6 | 65.5 | 62.7 | ||
| Depression only | 61.7 | 59.4 | 59.5 | ||
| Comorbidity | 63.7 | 61.9 | 58.3 | ||
| Internalizing | F(11, 72)=7.11*** | F(6, 77)=0.56 | |||
| Neither condition | 51.8 | 50.4 | 49.9 | ||
| Substance dep. only | 58.9 | 61.0 | 60.0 | ||
| Depression only | 58.4 | 56.4 | 55.4 | ||
| Comorbidity | 58.6 | 59.1 | 54.8 | ||
| Externalizing | F(11, 72)=4.45*** | F(6, 77)=1.07 | |||
| Neither condition | 54.8 | 54.5 | 53.6 | ||
| Substance dep. only | 62.4 | 64.8 | 62.3 | ||
| Depression only | 61.1 | 59.1 | 59.9 | ||
| Comorbidity | 63.5 | 61.0 | 58.5 |
Note.
p<.05,
p<.01,
p<.001
Child Behavior Problem Analyses
After controlling for all control variables and wave, the models child behavior scores parenting on comorbidity (Wald χ2(15)=202.75, p<.0001), internalizing behaviors on comorbidity (Wald χ2(15)=205.70, p<.0001), and externalizing behaviors on comorbidity (Wald χ2(15)= 200.04, p<.0001) were statistically significant (Table 5). Again, significant relationships were seen between substance dependence and depression with each parenting model. The odds of a mother with substance dependence only (OR=1.68), depression only (OR=1.47), or comorbidity (OR=1.36) having a child with problematic behavior on the total CBCL scale were higher than those of mothers with neither condition. For the internalizing subscale, again the odds of a mother with either condition or the co-morbidity of both having a child with an internalizing behavior problem were higher than mothers with neither substance dependence nor depression. For the externalizing subscale, the odds of a mother with substance dependence only (OR=1.78) or depression only (OR= 1.41) having a child with an externalizing behavior problem were higher than mothers with neither condition. For all three models, a significant relationship was seen between child behavior problems and wave with the odds of reporting behaviors meeting the clinical cut-off decreasing from wave 1 to wave 2 and wave 1 to wave 3.
Discussion
This analysis of a large, national probability sample of children and families in child welfare confirms the adverse consequences associated with maternal substance dependence and depression, and demonstrates that the relationship of co-morbidity with parenting and child behavior is complex. Outcomes changed over the three waves of data used in this analysis: baseline (Wave 1), approximately 18-months post baseline (Wave 3), and approximately 36-months post baseline (Wave 4). At baseline, parenting and child behavior problems were at concerning levels when maternal substance dependence only, depression only, or both conditions were present. However, by wave 4, mothers with substance dependence only had the poorest parenting and child behavior outcomes. In this group, almost no improvement was seen in parenting or child behavior from wave 1. Children of mothers with substance dependence only fared worse across all time points. They had the highest predicted probabilities of mean total CBCL scores, internalizing scores, and externalizing scores at both 18- and 36-month follow up. Furthermore, the total and externalizing scores remained in the clinically significant range across all time points. Child behavior problems in this group were worse at 18-months than at baseline and showed very little to no improvement from baseline to 36-months later. Hence, the children who remained in the home with a substance dependent mother did not improve over time. These results further emphasize the need for treatment specific to mothers with substance dependence only. Child behavior scores may remain at such high levels due to the continuation of neglect and emotional maltreatment in this group.
Study findings indicate a need to further examine the role of maternal substance dependence without depression on negative parenting practices. Predicted probabilities indicate concerning patterns of self-reported neglect, physical assault, and emotional maltreatment by mothers with substance dependence only. Unlike the other groups, mothers with substance dependence only maintained similar levels of both neglectful parenting and emotional maltreatment at 36-months compared to baseline. This pattern may indicate a need for services targeted towards mothers with substance dependence only and could be a reflection of a lack of service receipt or the receipt of ineffective services. These mothers meet criteria for substance dependence, indicating a high level of reliance on alcohol and/or drugs. This is particularly relevant since substance abuse is a strong predictor of children being placed into foster care (Berger et al. 2010). Our findings suggest the possibility that these high levels of self-reported substance use went undetected by CPS in some cases and, therefore, services were not provided to this group of mothers. This relationship could be examined in future studies. Although prior research by Chedgzsey Smith-McKeever et al. (2011) indicated that maternal depression is a stronger predictor of child behavior problems than maternal substance abuse, the authors did not conduct separate analyses to compare comorbidity with each separate condition. Therefore, it is difficult to compare implications from their study to the results of this study, but it does highlight the need for further research to disentangle these complicated relationships.
Clearly, adverse parenting remains an ongoing problem in many homes, which is likely related to the high prevalence of externalizing and internalizing problems in these children. This was demonstrated by the finding that over one-quarter of children two and older who remained in the home at all waves met clinical criteria for a mental health disorder, indicating a need for treatment. While rates of child behavior problems and negative parenting remain concerning in this vulnerable population, there were improvements in both outcomes over time. Decreases in child behavior problems across the entire sample may reflect a number of factors including: changes in the home, an increasing length of time since a negative experience, or the receipt of services addressing negative child behaviors. While outside the scope of this analysis, reasons for improvements among the entire sample should be explored in future studies.
Large decreases in self-reported negative parenting practices were seen over the course of the 36-month study window among all children who remained in the home. These decreases may be reflective of attempts by mothers to stop using negative parenting practices following a CPS investigation. Another possible explanation is that the mothers became less likely to report negative parenting behaviors by wave 3. However, increases in self-reported physical assault and neglectful parenting from waves 3 to 4 do not provide support for this reporting bias. If mothers participating in the study simply learned to stop reporting their abusive behaviors, one would expect to continue to see this decline at wave 4. Although increases in negative parenting behaviors from wave 3 to 4 are concerning, results indicating that the prevalence of neglect and physical assault at wave 4 are still much lower than at wave 1 are promising. That emotional maltreatment rates are similar at baseline and wave 4 does suggest the need for interventions that not only target physical discipline and neglect, but that also seek to reduce emotionally abusive parenting behaviors.
While overall our findings result in the rejection of study hypotheses, these analyses indicate that comorbidity plays a role in both parenting and child behavior outcomes, but the relationship is complicated. Though the hypothesized relationship (highest rates when comorbidity was present) was not always supported, comorbidity did have a significant relationship with neglectful parenting, physical assault, emotional maltreatment, and all child behavior indicators, even after controlling for demographics and most serious maltreatment type. These results are consistent with prior research indicating that having a depressed parent or a parent who misuses substances is associated with being maltreated (Chaffin et al. 1996) and negative child behavior (Kim-Cohen et al. 2005; Osborne and Berger 2009).
Limitations
When comparing these results to national and child welfare estimates, it is important to consider that our sample does not contain families in which the children were placed in out-of-home care. Given that children of substance abusing parents are more likely to be placed into out-of-home care following a maltreatment investigation, it is likely that substance dependence is even more prevalent in that sample. Also, NSCAW I assessed only for substance dependence, rather than lower levels of substance use disorders, limiting this sample to only those with the most severe levels of substance problems. This may exclude families in which maternal substance use was present at a level which affects parenting but does not meet the criteria for dependence. The parenting behavior variables were each based on mother’s self-report of her parenting behaviors. All mothers were involved with CPS and some may have chosen to not report some negative parenting behaviors. Although all caregivers were informed that their responses would be kept confidential this may have been a concern for some mothers. However, it is important to note that far over half of the sample reported negative parenting behaviors and many continued to report negative parenting behaviors over time. The relationship between maternal depression and substance dependence and the parenting and child behavior outcomes reported at waves 3 and 4 may be impacted by services received by the families. While some service data is available in NSCAW I, key aspects, such as dosage and the quality of services, are limited or unavailable. Therefore, service receipt was not examined in the analyses but should reflect a national estimate of treatment as usual for CPS-involved families in the United States.
Conclusions
Findings suggest that substance dependence is more highly associated with self-reported neglect, physical assault, emotional maltreatment, and poorer child behavior over time than either comorbidity, depression only, or neither condition. These negative outcomes persist over time when maternal substance dependence but not depression is present. These concerning findings indicate a need to know more about family dynamics and parenting in families where maternal substance dependence is present. When children remain in the home, intensive targeted services are needed for mothers with substance dependence only and their children to address severe needs in parenting and child behavior problems following entry into the child welfare system. The persistence of negative outcomes in this population indicates the need for follow-up beyond 18-months to determine if improvements persist over time. Additional longitudinal data are needed to determine the trajectory of parenting behavior and children’s behavior problems beyond 3 years post CPS investigation. Although some interventions have been developed or adapted to address maternal substance abuse in CPS-involved families (Osterling and Austin 2008), little is known about how these interventions impact parenting behavior or children’s behavior and well-being over time (Oliveros and Kaufman 2011). Prenatal substance exposure, a factor associated with child behavior problems (Sood et al. 2006), may be impacting behavior outcomes in this sample of children. Interventions developed for families where maternal substance dependence is present must evaluate for the presence of prenatal substance exposure and target services to address negative parenting practices, the use of substances, and child behaviors. Further innovation is necessary to evaluate and develop interventions that are effective at addressing the complex interplay between maternal substance use, parenting knowledge, and children’s well-being for CPS-involved families whose children remain in the home.
Acknowledgments
Research reported in this publication was supported by the National Institute on Drug Abuse under Award Numbers F31DA034442 (K. Seay, PI) and 5T32DA015035 and by the National Institute of Mental Health (RO3MH082203; P. Kohl, PI). Dr. Seay has support from a Doris Duke Fellowship and a Fahs-Beck Doctoral Dissertation Grant. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
References
- Achenbach TM. Manual for the child behavior checklist: 4–18 and 1991 profile. Burlington: University of Vermont, Department of Psychiatry; 1991. [Google Scholar]
- Achenbach TM. Manual for the child behavior checklist: 2–3 and 1992 profile. Burlington: University of Vermont, Department of Psychiatry; 1992. [Google Scholar]
- Andrews G, Peters L. The psychometric properties of the composite international diagnostic interview. Social Psychiatry and Psychiatric Epidemiology. 1998;33(2):80–88. doi: 10.1007/s001270050026. [DOI] [PubMed] [Google Scholar]
- Berger LM, Slack KS, Waldfogel J, Bruch SK. Caseworker-perceived caregiver substance abuse and child protective services outcomes. Child Maltreatment. 2010;15(3):199–210. doi: 10.1177/1077559510368305. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Besinger BA, Garland AF, Litrownik AJ, Landsverk JA. Caregiver substance abuse among maltreated children placed in out-of-home care. Child Welfare. 1999;78(2):221–239. [PubMed] [Google Scholar]
- Burns BJ, Phillips SD, Wagner HR, Barth RP, Kolko DJ, Campbell Y, Landsverk J. Mental health need and access to mental health services by youths involved with child welfare: a national survey. Journal of the American Academy of Child and Adolescent Psychiatry. 2004;43(8):960–970. doi: 10.1097/01.chi.0000127590.95585.65. [DOI] [PubMed] [Google Scholar]
- Chaffin M, Kelleher K, Hollenberg J. Onset of physical abuse and neglect: psychiatric, substance abuse, and social risk factors from prospective community data. Child Abuse & Neglect. 1996;20:191–203. doi: 10.1016/s0145-2134(95)00144-1. [DOI] [PubMed] [Google Scholar]
- Chandy JM, Blum RW, Resnick MD. History of sexual abuse and parental alcohol misuse: risk, outcomes and protective factors in adolescents. Child and Adolescent Social Work Journal. 1996;13(5):411–432. [Google Scholar]
- Chedgzsey Smith-McKeever T, Falconnier L, Gao W. African American and white mothers’ substance abuse, depression, and criminality as risk factors for child behavior problems. Families in Society. 2011;92(1):63–68. [Google Scholar]
- Davis LL, Rush JA, Wisniewski SR, Rice K, Cassano P, Jewell ME, McGrath PJ. Substance use disorder comorbidity in major depressive disorder: an exploratory analysis of the sequenced treatment alternatives to relieve depression cohort. Comprehensive Psychiatry. 2005;46(2):81–89. doi: 10.1016/j.comppsych.2004.07.025. [DOI] [PubMed] [Google Scholar]
- Diggle PJ, Heagerty P, Liang KY, Zeger SL. Analysis of longitudinal data. 2. New York: Oxford University Press; 2002. [Google Scholar]
- Dolan M, Smith K, Casanueva C, Ringeisen H. NSCAW II baseline report: Introduction to NSCAW II. OPRE Report #2011-27a. Washington: Office of Planning, Research and Evaluation, Administration for Children and Families, U.S. Department of Health and Human Services; 2011. [Google Scholar]
- Hill SY, Tessner KD, McDermott MD. Psychopathology in offspring from families of alcohol dependent female probands: a prospective study. Journal of Psychiatric Research. 2011;45(3):285–294. doi: 10.1016/j.jpsychires.2010.08.005. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Jones L. The prevalence and characteristics of substance abusers in a child protective service sample. Journal of Social Work Practice in the Addictions. 2004;4(2):33–50. [Google Scholar]
- Kessler RC, Andrews G, Mroczek D, Ustun TB, Wittchen HU. The world health organization Composite International Diagnostic Interview Short Form (CIDI-SF) International Journal of Methods in Psychiatric Research. 1998;7(4):171–185. doi: 10.1002/mpr.168. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Kessler RC, Chiu WT, Demler O, Walters EE. Prevalence, severity, and comorbidity of twelve-month DSM-IV disorders in the National Comorbidity Survey Replication (NCS-R) Archives of General Psychiatry. 2005;62:617–627. doi: 10.1001/archpsyc.62.6.617. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Kim-Cohen J, Moffitt TE, Taylor A, Pawlby SJ, Caspi A. Maternal depression and children’s antisocial behavior. Archives of General Psychiatry. 2005;62:173–181. doi: 10.1001/archpsyc.62.2.173. [DOI] [PubMed] [Google Scholar]
- Manly JT, Cicchetti D, Barnett D. The impact of subtype, frequency, chronicity, and severity of child maltreatment on social competence and behavior problems. Development and Psychopathology. 1994;6:121–143. [Google Scholar]
- Mustillo SA, Dorsey S, Conover K, Burns BJ. Parental depression and child outcomes: the mediating effects of abuse and neglect. Journal of Marriage and Family. 2011;73:164–180. [Google Scholar]
- National Data Archive on Child Abuse and Neglect. National Survey of Child and Adolescent Well-Being (NSCAW): Combined waves 1–4 data file user’s manual restricted release version. Ithaca: Author; 2006. [Google Scholar]
- Oliveros A, Kaufman J. Addressing substance abuse treatment needs of parents involved with the child welfare system. Child Welfare. 2011;90(1):25–41. [PMC free article] [PubMed] [Google Scholar]
- Osborne C, Berger LM. Parental substance abuse and child well-being: a consideration of parents’ gender and co-residence. Journal of Family Issues. 2009;30(3):341–370. [Google Scholar]
- Osterling KL, Austin MJ. Substance abuse interventions for parents involved in the child welfare system: evidence and implications. Journal of Evidence-Based Social Work. 2008;5(1/2):157–189. doi: 10.1300/J394v05n01_07. [DOI] [PubMed] [Google Scholar]
- Semidei J, Radel LF, Nolan C. Child Welfare. Substance abuse and child welfare: clear linkages and promising responses. 2001;80(2):109–128. [PubMed] [Google Scholar]
- Sood B, Delaney-Black V, Covington C, Nordstrom-Klee B, Ager J, Templin T, Sokol RJ. Prenatal alcohol exposure and childhood behavior at age 6 to 7 years: I. dose–response effect. Pediatrics. 2006;108(2):e34. doi: 10.1542/peds.108.2.e34. [DOI] [PubMed] [Google Scholar]
- Straus MA, Hamby SL, Finkelhor D, Moore DW, Runyan D. Identification of child maltreatment with the parent–child conflict tactics scale: development and psychometric data for a national sample of American parents. Child Abuse & Neglect. 1998;22(4):249–270. doi: 10.1016/s0145-2134(97)00174-9. [DOI] [PubMed] [Google Scholar]
- U.S. Department of Health and Human Services, Administration on Children, Youth, and Families 1 (US DHHS ACF) National Survey of Child and Adolescent Well-being (NSCAW): CPS sample component wave 1 data analysis report. Washington: U.S. Government Printing Office; 2005. [Google Scholar]
- U.S. Government Accounting Office. Foster care: Agencies face challenges securing stable homes for children of substance abusers (GAO/HEHS-98-182) Washington: Author; 1998. [Google Scholar]
- Ustun B, Compton W, Mager D, Babor T, Baiyewu O, Chatterji S, Sartorius N. WHO study on the reliability and validity of the alcohol and drug use disorder instruments: overview of methods and results. Drug and Alcohol Dependence. 1997;47:161–169. doi: 10.1016/s0376-8716(97)00087-2. [DOI] [PubMed] [Google Scholar]
- Vanderploeg JJ, Connell CM, Caron C, Saunders L, Katz KH, Tebes JK. The impact of parental alcohol or drug removals on foster care placement experiences: a matched comparison group study. Child Maltreatment. 2007;12(2):125–136. doi: 10.1177/1077559507299292. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Zuckerman BS, Beardslee WR. Maternal depression: a concern for pediatricians. Pediatrics. 1987;79(1):110–117. [PubMed] [Google Scholar]