Psychological theories have long emphasized the role of caregiver-child interactions in shaping attachment relationships and child psychosocial development (e.g., 1,2). Caregiving can be measured in multiple ways including self-report measures and observational measures, which each contain strengths and weaknesses. An advantage of observational measures of in-vivo interactions is the ability to account for the dynamic and reciprocal nature of caregiver-child interactions. The Observing Mediational Interactions (OMI; 10) coding scheme is one observational measure of caregiving that has the advantage of being connected with an intervention, the Mediational Intervention for Sensitizing Caregivers (MISC; 11), thereby providing immediate targets for intervention once identified. MISC focuses on how caregivers “mediate” the subjective experience of the child in common interactions with their child. The OMI captures optimal “scaffolding” that could encourage learning and psychosocial growth during conflict resolution (5). Additionally, the OMI is suggested to capture mentalizing on a granular level by operationalizing behavioral steps to achieve mentalizing during serve and return interactions (6). Mentalizing refers to the ability to reflect on one’s own and others’ mental states and to understand the relationships between mental states and behaviors and is implicated in both optimal parenting and psychosocial development (7). At present, mentalizing is considered an intrapsychic process and measures to assess it are representational in nature – this means that individuals are typically interviewed about their attachment relationships and mentalizing (or reflective functioning) is coded from these interviews. As such, these coding schemes do not provide immediate guidance on the behavioral steps needed in order to achieve a mentalizing stance (6,8). Indeed, clinicians and trainees have asked for guidance in behavioral terms of steps that can be taking intentionally to achieve mentalizing (9). The OMI offers coding of behavioral manifestations of the intrapsychic process of mentalizing (6), and is becoming more relevant as the evidence base for MISC expands (see 16 for a review).
The OMI is based on the MISC model (see Figure 1), which identifies components of caregiver-child interactions that enhance cognitive, social, and emotional development. These components include attachment-based (2) emotional components such as eye contact, synchrony, and affect. However, a principle of MISC is that “attachment is not enough” for learning to take place and therefore also includes the cognitive (learning-based) components based on Feuerstein’s cognitive modifiability theory. This theory operationalizes the behavioral components underlying individuals’ ability to learn: Focusing, Affecting (requesting or providing meaning), Expanding, Rewarding, and Regulating (see Table 1 for elaborated definitions). For instance, when resolving conflict with a child, in order to keep the child’s mind in mind (i.e. mentalizing) while also turning the conflict into a learning moment for the child, the caregiver will first have to focus the attention of the child on the topic at hand; next the caregiver may request meaning or provide meaning regarding the topic (affecting); next, the caregiver may expand on the topic by bringing previous discussion into the current discussion (expanding); the caregiver may next choose to help the child operationalize what they have learnt in an explicit way (regulating); and finally, the caregiver may give feelings of competence (rewarding). Together, the emotional and cognitive components create a “mediated learning experience,” which positively impacts the child’s needs system and approach to future experiences. Using the OMI, global ratings of emotional components are recorded. In addition, cognitive components are micro-coded as they occur in each utterance during the interaction (see Table 1). The inclusion of both attachment-based and learning-based coding variables in addition to the moment-to-moment micro-coding of cognitive components sets the OMI apart from other common observational coding schemes (e.g., 17,18) that use only macro coding systems.
OMI positive and negative cognitive components.
Cognitive component | Description | Positive example | Negative example |
---|---|---|---|
Focusing | Statement/behavior that is intentionally directed toward affecting a child’s perception or behavior. | Pointing to conflict task instructions. | |
Child response: looks to instructions, begins reading | Tone: hostile |
||
Affecting | Provision or request for meaning. Expresses significance of things. Must convey meaning, express excitement, or identify objects or people by name. | ||
Child response: |
Content: invalidating |
||
Expanding | Directed toward extending the child’s awareness beyond the immediate need or context that triggered the interaction. Must attempt to expand the child’s awareness beyond the immediate context of the interaction, attempt to connect present, past or future experiences, relate to a general, social or biological principal or process, or “tell” about things not seen or heard at the moment. | ||
Child response |
Content: invalidating [based on context] |
||
Rewarding | Expresses satisfaction with a child’s behavior and identifies a specific component that contributes to success. | ||
Child response: smiling | Content: invalidating |
||
Regulating | Behaviors or statements that model, demonstrate, and/or verbally suggest to the child regulation of behavior. Raises the child’s awareness to the possibility of “thinking” before doing and of planning steps of behavior towards a goal. | ||
Child response: |
Content: controlling |
Psychometric properties of the OMI have been investigated showing convergent validity through concurrent increases in child cognitive modifiability and performance (e.g., 9,19,20) and sociability (e.g., 21), and improved mental health outcomes (16). However, limitations remain. First, additional convergent validity will be demonstrated if relationships between the OMI and other measures of parenting or family functioning can be established. Additionally, more research is needed to fully establish the validity and reliability of the emotional components. In addition, the OMI has typically been evaluated in interactions that are not highly charged with emotion. Yet, emotionally charged situations, such as conflict discussions, may have the most clinical relevance. Meta-analytic research shows that high levels of parent-child conflict are associated with youth social and emotional maladjustment (17). However, parent-child conflict is not inherently negative as it provides opportunities to problem-solve and learn, especially in adolescence which is associated with a normative increase in parent-child conflict and increased risk of psychopathology (17). Additionally, while the OMI has been validated for use in infants (23), preschool-aged children (13,14), school-aged children (5,19), and adults with developmental disabilities (20), no studies have established its psychometric properties in adolescents.
As adolescents begin to seek independence, they enter a new social-interpersonal phase, with increased time with peers, decreased time with parents, and new romantic relationships (21). These changes coincide with elevated emotionality and puberty-associated biological changes and brain development, together creating a sensitive period for cognitive development and for the onset of psychopathology (22). Caregivers play a crucial role in mediating these changes for young people to facilitate a smooth transition from adolescence to adulthood (23), thereby protecting young people from developing mental health problems and promoting resilience. In-vivo assessment of how parents manage highly charged emotional interactions such as conflict can be highly relevant for predicting the onset and course of mental health problems (5). Finally, the OMI has not been applied to clinical samples.
In adapting the OMI for parent-adolescent conflict situations appropriate for use in clinical samples, we expanded the OMI codes to include negative tone and behaviors as these are more prevalent in conflict situations and in clinical samples. Negative behaviors For example, maternal depression is associated with parental negativity, coercion, and hostility (24), and parental anxiety is often associated with greater parental control (25). Additionally, negative parenting practices have been linked to the development of child psychopathology (e.g., 28). Parenting interventions often target the reduction of maladaptive parenting behaviors, in addition to strengthening adaptive behaviors. However, the OMI manual (3) does not include codes for maladaptive or negative parenting behaviors. Thus, we developed negative versions of each cognitive component for behaviors that are similar in structure to the positive behavior but are invalidating, controlling, or coercive in content or insensitive or hostile in tone (see Measures for details and examples in Table 1). In contrast to positive components, these behaviors inhibit or discourage the child from trusting that information being shared by the caregiver is in their best interest (i.e., diminishing “epistemic trust”; 29), and impedes learning.
In summary, the current study sought to examine the validity and reliability of the OMI within a clinical sample of mother-adolescent dyads during a conflict discussion task. The first aim was to examine the reliability of the expanded/adapted version of the OMI. We specifically chose early adolescence as the targeted developmental period to capture the early stages for onset of psychopathology. We hypothesized that the measure would show adequate interrater reliability and internal structure. Second, we examined associations with other observational measures and parent-reported measures of family functioning and parenting stress as an indicators of construct validity. We hypothesized that more adaptive OMI scores would be associated with observational indicators of positive parenting, greater family functioning, and less parenting stress. For the novel negative components, we tested whether significant relationships remained after controlling for the positive cognitive components to determine the incremental utility of the negative components. Third, to capture the emotional impact of using OMI components during the interaction, we examined whether the parent’s use of the OMI components was associated with the parent’s and child’s affect following the conflict discussion. We hypothesized that more adaptive use of the OMI variables would be associated with more positive and less negative affect post-discussion, controlling for baseline affect.
The sample included 56 youth aged 10-15 years and their primary female caregivers (54 biological, 2 adoptive mothers) drawn from a larger longitudinal study (citation removed for blind review) with three waves of assessments spaced 9 months apart. The longitudinal study was ongoing at the time of data selection and all participants who completed assessments were included in our sample. Forty-three dyads were from the first wave (wave 2 n = 10, wave 3 n = 3). Youth were recruited from pediatric primary care and psychiatric clinics in an American metropolitan city. Study eligibility required that youth were currently receiving psychotherapy and/or psychiatric medication management for a mood or behavioral problem. Youth were also phone screened to oversample for clinically significant emotional instability (scores of ≥ 12 on the Emotional Instability subscale of the Personality Assessment Inventory-Adolescent version (28) and ≥ 7 on the McClean Screening Instrument for Borderline Personality Disorder (29)). Exclusion criteria were an Intelligence Quotient estimate < 70 (assessed using the Peabody Picture Vocabulary Test-4th edition; 30), parent-reported neurological medical condition, parent-reported autism spectrum disorder, or current manic or psychotic episode (assessed using the Kiddie Schedule for Affective Disorders and Schizophrenia (K-SADS; 31). All parents had legal and primary physical custody (> 50% of the time). Table 2 summarizes demographic, diagnostic (KSADS and Childhood Interview for Borderline Personality Disorder; 31,32), and psychopathology (Child Behavior Checklist; 32) characteristics. Power analyses for bivariate correlation models in G*Power determined that a sample size of at least N = 46 (power = .80,
Sample characteristics (N = 56 mother – child dyads).
Child | n or M | SD or % |
---|---|---|
Age | 12.29 | .83 |
Gender (female) | 30 | 53.6% |
Race\Ethnicity* | ||
Black | 21 | 37.5% |
White | 23 | 41.1% |
Hispanic | 2 | 3.6% |
Multiracial | 12 | 21.4% |
CBCL psychopathology (borderline; clinical range, n=50) | ||
Total problems | 12; 7 | 21.4%; 12% |
Internalizing problems | 8; 8 | 14.3%; 14.3% |
Externalizing problems | 15; 5 | 26.8%; 8.9% |
K-SADS DSM-5 diagnoses | ||
Attention deficit hyperactivity disorder | 35 | 62.5% |
Oppositional defiant disorder | 26 | 46.4% |
Major depressive disorder | 20 | 35.7% |
Generalized anxiety disorder | 15 | 26.8% |
Separation anxiety disorder | 13 | 23.3% |
Conduct disorder | 12 | 21.4% |
Disruptive mood dysregulation disorder | 4 | 7.1% |
Social anxiety disorder | 6 | 10.7% |
Post-traumatic stress disorder | 3 | 5.4% |
Panic Disorder | 2 | 3.6% |
Childhood Interview for Borderline Personality Disorder diagnosis | 18 | 32.1% |
Mother | SD or % | |
Age | 40.30 | 7.28 |
Race\Ethnicity* | ||
Black | 19 | 33.9% |
White | 33 | 58.9% |
Hispanic | 2 | 3.6% |
Multiracial | 3 | 5.4% |
Marital status | ||
Married/living as married | 20 | 37.5% |
Never married | 20 | 37.5% |
Divorced/separated | 14 | 25% |
Highest level of education | ||
< High school | 7 | 12.5% |
High school/GED | 20 | 36.4% |
Associate’s | 10 | 18.2% |
Bachelor’s/graduate | 18 | 32.14% |
Household income | .50 | |
< $20,000 | 17 | 31.5% |
$20,000 - $39,000 | 10 | 18.5% |
$40,000 - $59,000 | 12 | 22.2% |
> $60,000 | 15 | 27.8% |
Note. Race/Ethnicity were each assessed as binary yes-no variables
Parents and youth completed semi-structured interviews, self-report measures, and the widely used “Hot Topics” conflict paradigm (34). Before the conflict discussion, caregivers and youth each completed a 25-item survey on parent-child conflict Participants endorsed common conflict areas (e.g., behavior toward siblings, internet/cell phone usage, behavior in school), and indicated its frequency (1 =
Two coders were trained by the OMI co-developers and brought to coding reliability through standard reliability procedures (35). These procedures included 1) the use of a predetermined coding or rating scheme (OMI), which allows for manualized procedures and decision rules to assign codes to emotional and cognitive components, 2) 3-day training, 3) practice until deemed competent, 4) competency test and approval as trained coder by the senior author (CS), and 5) quarterly supervision and consensus meetings with the senior author (CS) to maintain fidelity in coding, including calibration reliability checks evaluating whether the coder was still reliable based on supervisor assessment.
Following training, the two trained raters followed standard coding procedures as outlined by the OMI manual (3). First, they viewed the video and transcription without scoring. Then, raters rewatched the interaction and rated the emotional components on a macro level (i.e., one global score). Raters finally watched the interaction for a third time scoring the cognitive components on a micro level (i.e., utterance-by-utterance tally as they occur during the interaction).
The emotional components were each scored on a 5-point scale: smiling, eye contact, turn taking, verbal expressions, mutual attention, mutual engagement, synchrony, adult affect, child affect, and dyad affect. Physical closeness and touch were not rated because dyads were instructed to sit across from each other. As instructed by the manual (10), emotional component ratings were summed to yield a total emotional component score.
Cognitive components were micro-coded utterance-by-utterance using the video and a transcription of the dialogue: Focusing, Affecting, Expanding, Regulating, Rewarding. Negative versions of cognitive components were applied to statements that were similar in structure to the respective positive cognitive behaviors but were invalidating, controlling, or coercive in content or insensitive or hostile in tone. Negative cognitive components were often followed by the child responding negatively (e.g., hostile tone, disagreement) or disengaging (e.g., looking down, not responding; see Table 1.)
Each positive and negative cognitive component was tallied to yield its frequency. While most studies have used the cognitive components separately, some have created composite scores by summing across components (e.g., 20,35). In this study, we calculated individual component scores, positive cognitive composites (sum of positive components), and negative cognitive composites (sum of negative components). Given that this is a novel population and context for the OMI, scoring was guided by the frequency of each component and only components that occurred in at least 20 of the interactions were included in the analyses.
The conflict discussions were also coded by independent raters using an adapted version of the System for Coding Interactions & Family Functioning (SCIFF; 17). The SCIFF is a behavioral assessment of family functioning focused on how families handle conflict, disagreement, and problem solving. The SCIFF has shown adequate reliability and validity in samples similar to the current sample (37). We used two dyadic codes (cohesiveness, focus on problem) and seven parent-focused codes (positive affect, anger and frustration, rejection and invalidation, withdrawal, coerciveness, emotional support, respect for autonomy). Each was rated on a 5-point scale indicating how characteristic the behavior was of whole interaction. Raters completed two training assignments and achieved agreement with a “master coder.” Interrater reliability was demonstrated in the larger dataset from which our sample was drawn (38). In the current sample, interrater reliability was evaluated using a two-way random effects model to calculate intraclass coefficients based on absolute agreement across five randomly selected interactions rated by three independent coders (poor ICC < .5, moderate ICC = .5 - .75, good ICC = .75 - .9, excellent ICC ≥ .9; 39) Both dyadic codes exhibited strong interrater reliability (ICC’s=.92-.93). Regarding parent codes, positive affect, emotional support, and respect for autonomy exhibited good to excellent reliability (ICCs =.86 -.98), and anger (ICC= .70), rejection/invalidation (ICC=.67), withdrawal (ICC=.56), and coerciveness exhibited moderate reliability (ICC=.53).
The FACES-IV (40) is a 42-item self-report inventory based on the Circumplex Model of Marital and Family Systems. The model hypothesizes that balanced or mid-range levels of cohesion (i.e., “the emotional bonding that family members have toward one another”) and flexibility (i.e., “the quality and expression of leadership and organization, role relationship, and relationship rules and negotiations”) are most conducive to healthy family functioning (38, p. 65). Items are rated on a five-point Likert scale, yielding two balanced subscales (balanced cohesion and balanced flexibility) for which higher scores indicate better family functioning as well as four unbalanced subscales for which higher scores indicate poorer functioning. Parents completed the FACES-IV and we used the balanced cohesion and balanced flexibility percentile scores (higher percentile indicates better functioning), which have shown adequate reliability and construct validity in terms of strong correlations with other family functioning measures (41). In the current sample (n = 42), internal consistency was adequate for the balanced cohesion scale (
The PSS (42) is an 18-item self-report survey assessing parenting stress and three components: positive emotional benefits, sense of enrichment and personal fulfillment, and negative components of parenting. Parents report on their relationship with their children using a Likert scale from “strongly disagree” to “strongly agree.” The PSS has shown adequate reliability and construct validity with correlations with similar measures of parenting stress (40). In the current sample (n = 48), internal consistency was good (
Positive affect (PA) and negative affect (NA) were measured via self-report. Directly before and after the conflict discussion task, subjects were asked to indicate the extent to which they were presently feeling ten specific emotions (PA: alert, calm, proud, happy, accepted; NA: sad, nervous, upset, mad, ashamed) using a five-point Likert scale (1= Very Slightly or Not at all, 5= Extremely). Ratings were summed for each domain to yield total PA and NA scores. Internal consistency for the child-report just below adequate pre-conflict (PA: α= .66, NA:
Interrater reliability of the emotional component sum, positive cognitive composite, negative cognitive composite, and each cognitive component was evaluated in a subset (16.1%) of double-coded interactions. Reliability was calculated with a two-way random effects model to produce intraclass correlation coefficients. A threshold of .70 was used to determine adequate interrater reliability based on Stemler (2004). Coefficient alpha was used to examine internal consistency of the emotional component sum score. The internal structure of the OMI was evaluated by inspecting Pearson’s correlations between the emotional component sum, positive cognitive composite, negative cognitive composite, and each cognitive component.
We calculated Pearson’s correlations between the emotional component sum, and positive and negative cognitive composite scores with the adapted SCIFF, parent-reported family functioning (FACES-IV cohesion and flexibility), and parenting stress (PSS total). Based on bivariate relationships, we ran hierarchical linear regression models testing the incremental utility of the negative over the positive composite to predict construct validity variables. Child age and gender would be controlled for in these models if they showed significant relationships with OMI scores. Tolerance greater than .40 was considered acceptable.
Pearson’s correlations were conducted between OMI variables and self-reported post-conflict PA and NA. For significant findings, we conducted hierarchical linear regressions controlling for pre-conflict affect. Post-discussion PA or NA was entered as the dependent variable, pre-discussion PA or NA was entered at step 1, and an OMI variable (i.e., emotional component sum, positive composite, or negative composite) was entered at step 2.
OMI component coding frequency and interrater reliability are displayed in Table 3. Summing all codes across participants showed that positive and negative Affecting and Expanding each accounted for less than 1% of the total positive/negative codes. Therefore, only Affecting, Expanding, and Regulating were included in each composite variable and analyzed. Interrater reliability was adequate for all composite variables and components except for negative Expanding. The emotional component sum showed good internal consistency (
Frequency of OMI components and Interrater Reliability.
n with 1+ code | ICC | |
---|---|---|
1 Emotional sum | NA | .76* |
2 Pos. composite | NA | .92** |
3 Neg. composite | NA | .94** |
4 Focusing | 13 | NA |
5 Affecting | 56 | .92** |
6 Expanding | 54 | .93** |
7 Rewarding | 6 | NA |
8 Regulating | 51 | .81* |
4 Focusing | 1 | NA |
5 Affecting | 46 | .92** |
6 Expanding | 20 | .63 |
7 Rewarding | 0 | NA |
8 Regulating | 27 | .98** |
p < .05,
p <.0;
Pearson’s correlations among OMI variables and construct validity variables.
1. | 2. | 3. | 4. | 5. | 6. | 7. | 8. | 9. | 10. | 11. | 12. | 13. | 14. | 15. | 16. | 17. | 18. | 19. | 20. | 21. | |||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
1 Emotional sum | |||||||||||||||||||||||
2 Pos. composite | .50** | ||||||||||||||||||||||
3 Neg. composite | -.51** | -.44** | |||||||||||||||||||||
4 Focusing | -.01 | -.02 | .06 | ||||||||||||||||||||
5 Affecting | .49** | .94** | -.43** | .01 | |||||||||||||||||||
6 Expanding | .03 | .07 | -.12 | -.16 | -.16 | ||||||||||||||||||
7 Regulating | .16 | .48** | -.08 | .02 | .22 | .03 | |||||||||||||||||
8 Affecting | -.48** | -.41** | .97** | .04 | -.39** | -.12 | -.12 | ||||||||||||||||
9 Expanding | -.39** | -.31** | .55** | -.14 | -.34* | .09 | -.10 | .43** | |||||||||||||||
10 Regulating | -.25 | -.23 | .54** | .23 | -.26 | -.16 | .15 | .36** | .23 | ||||||||||||||
11 Cohesiveness | .59** | .34* | -.23 | -.27* | .36** | -.12 | .14 | -.20 | -.20 | -.17 | |||||||||||||
12 Focus of Problem | .28* | .22 | -.25 | -.10 | .27* | .02 | -.11 | -.23 | -.14 | -.18 | .51** | ||||||||||||
13 Positive Affect | .57** | .35** | -.20 | .20 | .39** | -.17 | .10 | -.20 | -.23 | -.04 | .58** | .36** | |||||||||||
14 Anger/Frustration | -.23 | -.05 | .49** | .39** | -.01 | -.15 | -.01 | .43** | .47** | .32 | -.40** | -.29** | -.15 | ||||||||||
15 Rejection/ Invalidation | -.24 | -.11 | .25 | .27* | -.13 | .08 | -.05 | .20 | .26 | .25 | -.57** | -.47** | -.29 | .58** | |||||||||
16 Withdrawal | -.42** | -.33* | -.03 | -.005 | -.25 | -.23 | -.19 | -.07 | -.001 | .12 | -.27** | -.18 | -.15 | -.006 | .06 | ||||||||
17 Coerciveness | -.08 | -.004 | .24 | .35** | .001 | -.07 | .04 | .21 | .09 | .25 | -.25 | -.26 | -.14 | .48** | .39** | .03 | |||||||
18 Emotional Support | .40** | .35** | -.36** | -.20 | .39** | -.06 | .04 | .35** | -.29* | -.13 | .38** | .55** | .44** | -.51** | .66** | -.22 | -.45** | ||||||
19 Respect for Autonomy | .48** | .39** | -.47** | -.09 | .44** | -.08 | .04 | -.46** | -.42** | -.13 | .53** | .48** | .37** | -.39** | -.54** | -.09 | -36** | .81** | |||||
20 Cohesion | .31* | .30* | -.59** | -.20 | .32* | .11 | -.01 | -.46** | -.53** | -.43** | .13 | .19 | .06 | -.51** | -.50** | -.08 | -.26 | .45** | .53** | ||||
21 Flexibility | .03 | .33* | -.40** | -.15 | .27 | .12 | .24 | -.29 | -.43** | -.33** | .05 | .05 | -.11 | -.43** | -.34* | -.11 | -.22 | .34* | .35* | .67** | |||
22 |
-.03 | -.29 | -.10 | .11 | -.19 | -.16 | -.31* | -.08 | -.12 | -.09 | -.08 | .19 | -.06 | .03 | .25 | .07 | -.08 | -.15 | -.06 | -.07 | .19 |
Note, p < .05,
p < .01
Pearson’s correlations among OMI variables and construct validity variables are displayed in Table 4. The emotional component sum showed positive correlations with SCIFF cohesiveness, focus on problem, emotional support, respect for autonomy, and FACES-IV cohesion, and a negative correlation with SCIFF withdrawal. The positive cognitive composite showed a similar pattern of correlations, except that it was not significantly correlated with SCIFF focus on problem and was positively correlated with FACES-IV flexibility and negatively correlated with parenting stress. positive affect. The negative cognitive composite was positively correlated with SCIFF anger/frustration and negatively correlated with SCIFF emotional support and respect for autonomy and FACES-IV cohesion and flexibility. Pearson’s correlations showed no significant relationships between child age and OMI scores. T-tests revealed no significant differences in OMI scores between male and female youth. Thus, demographic variables were not controlled for in analyses.
Five hierarchical linear regression models tested the incremental utility of the negative composite over the positive composite in predicting construct validity variables (Table 5). For each model, the positive cognitive composite was entered at step 1, the negative cognitive composite was entered at step 2, and the construct validity measure was entered as the dependent variable. Tolerance (.81 - .92) was within acceptable limits. As displayed in Table 5, for models predicting SCIFF anger/frustration, SCIFF respect for autonomy, FACES-IV cohesion, and FACES-IV flexibility, adjusted R2 values indicated a significant change in explained variance due to the addition of the negative cognitive composite to the model. The model predicting SCIFF emotional support showed no significant change in explained variance due to the addition of the negative cognitive composite.
Hierarchical Linear Regressions Testing Utility of Negative Composite over Positive Composite in Predicting Construct Validity Variables
b | SE | β | t | p | Adj. R2 | ΔAdj. R2 | |
---|---|---|---|---|---|---|---|
DV = SCIFF Anger/frustration | |||||||
Step 1a | -.016 | ||||||
Pos. cognitive composite | -.003 | .01 | -.05 | -.36 | .72 | ||
Step 2b | .246 | .262*** | |||||
Pos. cognitive composite | .01 | .01 | .20 | 1.56 | .12 | ||
Neg. cognitive composite | .06 | .01 | .58 | 4.44 | < .001 | ||
DV = SCIFF Emotional support | |||||||
Step 1c | .105 | ||||||
Pos. cognitive composite | .02 | .01 | .35 | 2.70 | .008 | ||
Step 2d | .146 | .041 | |||||
Pos. cognitive composite | .01 | .01 | .24 | 1.70 | .096 | ||
Neg. cognitive composite | -.03 | .02 | -.26 | -1.88 | .066 | ||
DV = SCIFF Respect for autonomy | |||||||
Step 1e | .137 | ||||||
Pos. cognitive composite | .02 | .01 | .39 | 3.10 | .003 | ||
Step 2f | .236 | .099** | |||||
Pos. cognitive composite | .01 | .01 | .23 | 1.76 | .085 | ||
Neg. cognitive composite | -.03 | .01 | -.37 | -2.82 | .007 | ||
DV = FACES-IV Cohesion | |||||||
Step 1g | .070 | ||||||
Pos. cognitive composite | .21 | .10 | .30 | 2.0 | .048 | ||
Step 2h | .333 | .263*** | |||||
Pos. cognitive composite | .10 | .09 | .15 | 1.10 | .278 | ||
Neg. cognitive composite | -.87 | .21 | -.55 | -4.14 | < .001 | ||
DV = FACES-IV Flexibility | |||||||
Step 1i | .088 | ||||||
Pos. cognitive composite | .23 | .10 | .33 | 2.25 | .03 | ||
Step 2j | .174 | 8.6* | |||||
Pos. cognitive composite | .16 | .10 | .23 | 1.6 | .118 | ||
Neg. cognitive composite | -.52 | .23 | -.34 | -2.3 | .028 |
Note: *p<.05,
p<.01,
p<.001;
model not significant, F(1, 54) = .13, p = .72;
model significant, F(2,53) = 9.96, p < .001;
model significant, F(1, 54) = 7.5, p < .01;
model significant, F(2,53) = 5.7, p < .01;
model significant, F(1, 54) = 9.8, p <.01;
model significant, F(2,53) = 9.49, p < .001;
model significant, F(1, 54) = 4.2, p < .05;
model significant, F(2,53) = 11.5, p < .001;
model significant, F(1, 54) = 5.1, p < .05;
model significant, F(2,53) = 5.4, p < .001.
The mother’s post-discussion PA was positively associated with the positive cognitive composite (
Hierarchical Linear Regressions Testing the Effect of Baseline Affect.
b | SE | β | t | p | Adj. R2 | ΔAdj. R2 | |
---|---|---|---|---|---|---|---|
DV = Mother’s post-discussion PA | |||||||
Step 1a | .505 | ||||||
Pre-discussion PA | .81 | .11 | .72 | 7.6 | < .001 | ||
Step 2b | .519 | .014 | |||||
Pre-discussion PA | .77 | .11 | .68 | 7.12 | < .001 | ||
Pos. mediation composite | .04 | .02 | .15 | 1.58 | .120 | ||
DV = Mother’s post-discussion PA | |||||||
Step 1c | .505 | ||||||
Pre-discussion PA | .81 | .11 | .72 | 7.6 | < .001 | ||
Step 2d | .541 | .036* | |||||
Pre-discussion PA | .74 | .11 | .65 | 6.8 | < .001 | ||
Neg. mediation composite | -.10 | .04 | -.22 | -2.3 | .027 |
Note: *p<.05,
p<.01,
p<.001;
model significant, F(1, 54) = .57.2, p < .001;
model significant, F(2,53) = 30.6, p < .001;
model significant, F(1, 54) = 57.2, p < .001;
model significant, F(2,53) = 33.4, p < .001
The overall aim of the current study was to establish the psychometric properties of the OMI for coding parenting during a conflict interaction. We hypothesized that the expanded OMI would demonstrate adequate reliability and construct validity and predict affect following the conflict discussion.
Results firstly indicated that positive and negative Focusing and Rewarding were rarely used. The lack of Focusing suggests that parents of adolescents may be more likely to use statements that contain instructions or sequencing of behaviors, which are coded as Regulation (e.g., “First, tell me about the arguments with your sister”) rather than simple calls for attention. In addition, the conflict task may preclude the need to focus the adolescent, as participation necessitates shared focus. Rewarding may be more common when the child is working on a concrete task such as in a teaching or play interaction. Future studies with similar samples and contexts could consider only coding positive and negative Affecting, Expanding, and Regulating. However, other laboratory tasks such as a teaching interaction may require coding of focusing and/or rewarding.
Regarding Aim 1 (Reliability), interrater reliability was adequate for the emotional component sum, positive and negative composites, and individual components except for negative Expanding. The emotional component sum exhibited good internal consistency. The emotional component sum and composite scores were highly associated with each other; however, positive and negative Expanding and Regulating were not significantly related to most other OMI variables. Therefore, the internal structure remains a question for future research using factor analysis with larger sample sizes. The evidence of reliability of the emotional component sum supports its use and is a particularly critical addition to the literature considering the scarcity of previous research. However, the inadequate reliability of negative Expanding suggests that further elaboration in the coding instructions is required.
For Aim 2 (Relationships with Parenting and Family Functioning Variables), we examined the construct validity of the emotional component sum and composite variables given their demonstrated reliability. Consistent with our hypotheses, relationships with a variety of dyadic and parent-focused variables on the SCIFF observational coding scheme and self-report family functioning and parenting stress measures emerged. Additionally, the novel negative cognitive component showed incremental utility over the positive cognitive composite in predicting parent anger/frustration, emotional support, respect for autonomy, and family cohesion and flexibility. This finding justifies the need for negative cognitive scores in the context of observing conflict interactions. This the first study to demonstrate construct validity of the OMI variables through relationships with observational or selfreport variables, rather than concurrent increases in variables following MISC. It is notable that the indicators of construct validity included both observational coding by independent researchers using a measure based in a distinct theoretical orientation and global parent-ratings related to general family functioning and overall parenting stress. These findings add to existing support for the validity of the positive cognitive components, provide some of the first evidence of the validity of the emotional components, and support the validity of the novel negative cognitive scores, supporting their use in future studies.
Aim 3 (Emotional Impact of OMI Components Associations with Parent and Child Affect Post-Discussion) findings were in partial support of our hypotheses. Analyses revealed that mothers with a higher negative cognitive composite reported less positive affect post-discussion, controlling for pre-discussion positive affect. This again underscores the utility of the novel negative components as neither the emotional nor positive cognitive components predicted affect. This suggests that engaging in fewer negative parenting behaviors could positively impact caregiver mental health, consistent with MISC research showing increased caregiver sense of competence and self-efficacy following intervention (4). The finding also underscores the promise of recent MISC adaptations for mothers with psychopathology (10). In contrast with our hypotheses, OMI coders were not associated with child affect. It is possible that power was too weak to detect an effect or the self-report inadequately captured child affect. Future studies should examine if the OMI predicts short-term (e.g., affect) and long-term (e.g., mental health) child outcomes.
This study has several limitations. The sample size may have had inadequate power to detect smaller effects and did not allow for full testing of differences based on demographic characteristics. Future studies could examine measurement invariance based on age, gender, race, or ethnicity. However, the study has strengths including a racially diverse sample and detailed data achieved through time-intensive micro-coding. Sample sizes were particularly low for the FACES-IV and PSS because these measures were not administered throughout the entire larger study. We also acknowledge that multiple testing of variables could have led to false positive results. Moreover, we limited our study to female caregiver because the small number of fathers in the larger study sample precluded us from examining group differences. Future research should extend to fathers. While we deliberately chose to examine early adolescence, studies could also examine older adolescents as the OMI is not age restricted. Additionally, our sample was selected based on clinical features of the child and results may differ based on the parent’s clinical status, which is important to explore for adapting MISC adaptations for parents with psychopathology. In addition, in examining construct validity, we used an adapted version of the SCIFF that showed only moderate inter-rater reliability for some variables. While we examined both an external observational measure and global self-reports of family functioning and parenting stress, other indicators of construct validity (e.g., child reports of parenting) should be used in future studies. Future research could also compare the OMI to other coding systems (e.g., (e.g., 17,18) or self-report measures of parenting in the same model predicting outcomes to determine the unique utility of the OMI. Finally, we acknowledge that while observational measures go beyond self-report in capturing the quality of parent-child interactions, the fact that they are observational may introduce observational bias.
Withstanding limitations, this study is an important addition to the literature. The expanded OMI demonstrated acceptable reliability and validity, although the negative Expanding component should be elaborated in the coding instructions. Findings support the use of the emotional components, and positive and negative cognitive composites in studies examining parent-adolescent conflict in clinical samples. While the original OMI contains five cognitive components, only three (Affecting, Expanding, and Regulating) were relevant to the present interactions. Future studies focused on similar populations and contexts may consider only coding these types of components. Additionally, findings suggest that the novel negative cognitive composite may add clinically useful information beyond the existing OMI. As this is the first examination of the expanded OMI, future work should attempt to extend findings to other populations and settings and further elaborate the coding instructions as needed. Additionally, now that the OMI has shown good psychometric properties in a clinical sample of mother-adolescent dyads, researchers should consider examining relationships between the OMI and caregiver/adolescent mental health problems to inform ongoing adaptations of MISC (10) and to identify targets for prevention and intervention more broadly.