Family functioning measures: convergent and discriminant validity.

Abstract: The purpose of this study was to conduct psychometric analysis of three self-report family functioning measures. Twenty-five families were recruited for study participation. The families consisted of a very low birthweight infant, the infant's mother, and an adult partner with whom

the mother resided. Campbell and Fiske's (1959) multitrait-multimethod (MTMM) of establishing convergent and discriminant validity was employed. Traits consisted of mutuality and communication, and methods were three self-report family functioning measures: Family Dynamics Measure II (FDM II), Family Functioning Style Scale (FFSS), and the Self-Report Family Inventory (SFI). Factor analysis was employed to confirm the findings generated from the MTMM method. The mothers and adult partners were examined independently. Convergent and discriminant validity was not established.

Key words: Family Systems, Multitrait-multimethod matrix, Very low birthweight infants

**********

Despite four decades of progress in providing prenatal care to pregnant women, there has been no improvement in the preterm infant birth rate within the U.S. (American College of Obstetricians and Gynecologist, 1995). Technologic and pharmacological advancements have produced a decline in mortality for very low birthweight (VLBW) infants or infants under 1500 grams (Goldenberg, 1994; Rouse et al., 1994; Ventura, Martin, Curtin & Mathews, 1999). The rate of handicaps, however, has remained steady (Goldenberg & Rouse, 1998; Hack et al., 1991; McCormick, 1989).

Very low birthweight infants remain at risk for developmental disabilities (Thompson et al., 1994). Because economic strain, special education, and long-term care are frequently produced, preterm birth remains a significant health care concern (Morrison, 1990). Healthy developmental outcomes are attainable in VLBW infants; therefore, the recognition of influences on outcomes is critical to families delivering infants prematurely. If research agendas are to be established, interventions developed, and theoretic frameworks reassessed, revised, and extended, an early recognition of such influences is imperative (Taylor, Klein, Schatschneider & Hack, 1998).

Although biologic risk factors such as birthweight have been widely accepted as indicators of developmental outcomes, considering birthweight alone is inadequate. Exclusive attention to infants birthweight precipitated a paradigmatic shift in research agendas. A growing body of evidence, however, suggests that social influences such as the home environment, social support, and life stress serve as the best indicators of a child's long term developmental outcome (Leonard et al., 1990). As biological risks wane, the caregiving environment of family system has an increasing influence on developmental outcomes (Thompson, et al., 1998). Socioeconomic status, education, and the environment in which the child is raised appear to be the most important factors accounting for later development (Korner et al., 1993). Social influences have become increasingly important (Taylor et al., 1998).

The family system has been identified as a major environmental influence on the growth and development of the VLBW infant (Beckwith & Parmelee, 1986; Belsky, Lerner & Spanier, 1984; Minde, Perrotta & Hellman, 1988). In a system of interactions, the actions of one person (e.g., the father) can affect the responses of another person (e.g., the mother). These actions and responses can be influenced further by the presence or absence of a VLBW infant. From an ecological systems perspective, the VLBW infant is viewed as an indirect influence to which the individual and family system endeavor to adapt (Thompson et al, 1994). The infant's interactions can also affect the interactions of other people within the system. The family, therefore, is an interactional system with the potential for creating and/or modifying interactions and optimizing individual responses.

Numerous measures to examine family system interactions have been generated from differing theoretical approaches. With such a pluralistic approach to family functioning, subscales within the various measures often differ in conceptual underpinnings, item generation, and construction methods (Goldsmith, Rieser-Danner & Briggs, 1991). Differences in the types of measures, self-report versus ratings by external observers, have also made the linkages between concepts difficult (Perosa & Perosa, 1990). Psychometric issues must be addressed if systematic advances are to be achieved.

To comprehend family interactions or processes within a systems framework, research has focused on the unique contribution that each measure provides (Carlson & Grotevant, 1987; Dickerson & Coyne, 1987). Self-report construct validity studies have been discouraging (Dickerson & Coyne). Studies are therefore needed to establish such validation. The purpose of this study was to conduct psychometric analysis for three self-report measures in family systems research when one family member was a VLBW infant.

CONCEPTUAL FRAMEWORK

Bronfenbrenner's (1977) framework entitled the ecology of human development provided the foundation for this study. Although this framework has been reassessed, revised, and extended, the basic elements and imperatives of the paradigm have been strengthened through scientific evidence and argument (Bronfenbrenner, 1992). From an expansion and union of theoretical conceptions of environment, Bronfenbrenner's (1977) theoretic foundation has evolved as the basis for naturalistic and experimental approaches. Four dimensions of his ecologic framework have been identified.

Dimension 1: Bronfenbrenner's Ecological Imperative

The first dimension serves as the cornerstone of Bronfenbrenner's theoretic structure and defines his ecology: "The ecology of human development is the scientific study of the progressive, mutual accommodation, throughout the life course, between an active, growing human being, and the changing properties of the immediate setting in which the developing person lives, as this process is affected by the relations between these settings, and by the larger contexts in which the settings are embedded" (Bronfenbrenner, 1992, p. 188). The changing relation between the person and environment is conceived in systems terms (Bronfenbrenner, 1977).

Dimension 2: Bronfenbrenner's Multi-structured Environment

The second dimension within the framework and is conceived as a nested arrangement of four structures each contained within the next. Bronfenbrenner (1977) refers to these four structures as the micro, meso, exo and macrosystems. A microsystem is a "pattern of activities, roles, and interpersonal relations experienced by developing person in a given face to face setting with particular physical and material features, and containing other persons with distinctive characteristics of temperament, personality, and systems belief" (Bronfenbrenner, 1992, p. 227). Place, time, activity, physical features, participant, and role are microsystem elements. The home or family system would serve as an example of this type of environment (Bronfenbrenner, 1986).

A mesosystem is comprised of "interrelations among major settings that contain the developing person at a particular point in their life" (Bronfenbrenner, 1977, p. 515). This structure includes church, school and daycare. Succinctly, the mesosystem is a system of microsystems (Bronfenbrenner, 1986). An exosystem consists of other structures that do not contain the developing person but encompass the immediate setting thereby influencing what transpires within the micro and mesosystems. Included are the major societal institutions such as neighborhoods, media, governmental agencies, distribution of goods, and services and transportation facilities.

A macrosystem is fundamentally different from the previous three structures. A macrosystem does not "refer to specific contexts affecting a particular person but to general prototypes that may exist within a culture" (Bronfenbrenner, 1977, p. 515). Within a certain culture, blueprints are established as to how systems should look and function. Macrosystems are viewed as ideologies of what structures should look like. Institutional patterns of a culture such, as political, economic and social systems would be an example of the macrosystem. Bronfenbrenner asserts that this system is of particular interest to child development because these systems endow meaning and motivation to particular agencies. The priority children are provided will eventually dictate their treatment.

Dimensions 3 & 4: Ecologic Validity and Experiment

The third and fourth dimensions in Bronfenbrenner's ecology of human development are ecologic validity and the ecologic experiment. Bronfenbrenner (1977) delineated a series of propositions that outline requirements of an ecological model as stipulated in his theoretical framework of environment. The particular proposition of interest within the microsystem and of interest in this study states, "in contrast to the conventional dyadic research model, which is limited to assessing the direct effect of two agents on each other, the design of an ecological approach must take into account the existence in the setting of systems that include more than two persons (p. 520). System analysis must consider all possible subsystems (triads, dyads) and the potential second or higher order effects associated with them (Bronfenbrenner). The indirect influence on subsystems is defined by Bronfenbrenner as a second order effect. For example, a father's action and the mother's responses may be influenced by the presence or absence of a VLBW infant and the effect of the infant's behavior on the interaction of the father with the mother.

Beyond Bronfenbrenner: Second-order Effect

The examination of a second order effect is important to consider when examining systems beyond dyads because dyadic interactions might be influenced by an infant. Als' (1982) theory of synactive development, is used to examine the infant's second order effect within the family system. Als (1982) posits that at each developmental stage and at each moment of functioning, various systems interactionally co-exist. Frequently, however, the systems are in a holding pattern as if to define and differentiate the foundation of each system. These systems include the autonomic, motor, state-organizational, attention and interactive systems and a self-regulating balancing system. Not all newborns are "equally able to increasingly build up the interactive attentional capacity" (Als, 1982, p. 232). The pre-term infant is not an inadequate full-term infant; rather, a pre-term infant is an appropriately functioning person in a particular environment. When the infant is suddenly placed in a vastly different environment, system functioning is irreversibly altered. Thus, the infant is in an environment poorly matched to infant expectations. The result is a behavior pattern that is quite different from the full-term infant.

LITERATURE REVIEW

Few studies have examined microsystem triads when one family member is a VLBW infant. Little attention has been focused on the home versus laboratory environment. The application of a triad model to a developmental context outside the home is rare. Bronfenbrenner (1977) suggest that as one progresses beyond the dyad, the resulting structures may offer possibilities for greater stability and mutual assistance. Triadic combinations, therefore, constitute a promising ecological domain for developmental research.

Mutuality and communication are two concepts or traits within the microsystem subjected to review. These two traits have been selected for several reasons. There is a growing consensus among family theorists that core traits of family systems exist (Barnhill, 1979; Dickerson & Coyne, 1987; Perosa & Perosa, 1990). Sawin and Harrigan (1994), however, question whether mutuality and communication are distinct. Although agreement exists that there are traits that cut across major models of family functioning, linkages between the traits have been disappointing. Construct validity needs to be established.

A literature review was conducted to determine the construct validity of three self-report family system measures; Family Dynamics Measure II (FDM II), Family Functioning Style Scale (FFSS) and Self-Report Family Inventory (SFI). No published studies using the FDM II were found; however, Lasky et al. (1985) reported high correlation coefficients in the original Family Dynamics Measure. Sawin and Harrigan contend that communication may be an "index" subscale because of the high correlations with all other subscales; therefore, factor analysis was suggested. The family functioning style scale (FFSS) was employed by McGrath (1997) to examine medical and environmental risk. The sample consisted of high-risk children. Alpha coefficients were not reported. The SFI literature was reviewed and Cronbach alphas between .84 and .93 were reported (Beavers & Hampson, 2000). The cohesion subscale of the SFI correlated with the cohesion scale from the FACES III (r = -.67), a self-report scale for the Circumplex Model of Marital and Family Functioning (Beavers & Hampson, 1990; Olson, 1986). Clinical validity of the SFI was demonstrated by its capacity to discriminate groups of psychiatric patients with differing diagnoses (Beavers & Hampson, 1990).

METHOD

A descriptive correlational design was selected to determine construct validity through convergence and discrimination. Convergent and discriminant validity was examined employing the Campbell and Fiske (1959) multitrait-multimethod (MTMM) model utilizing a matrix of correlations. Unlike other approaches to establish construct validity, the MTMM approach provides for the ability to separate trait and method variance (Polit & Hungler, 1999). The criterion for the establishing convergence is that correlations of the same trait by different methods should be high in relation to reliability estimates. Goldsmith et al. (1991) contends that three criteria for discriminant validity exist. The least stringent criterion for correlations between the same trait measured by different methods occurs when convergence exceeds the correlations between different traits measured by different methods (heterotrait-heteromethod). The second criterion of discrimination occurs when convergent validity coefficients exceed correlations of different traits measured by the same method or trait variance surpasses method variance. The third and most stringent criterion for discrimination occurs when the pattern of inter-trait correlations is similar across different methods of measurement.

Mutuality and communication were the multitraits used in the Campbell and Fiske (1959) model and the multimethods included three self-report family system measures, the FDM II (White, Wu & Grzankowski, 1997), the FFSS (Trivette, et al, 1990), and the SFI (Beavers & Hampson, 1990). The trait, mutuality, was measured by three subscales: FDM II mutuality, FFSS family commitment, and SFI cohesion. The second trait, communication, also consisted of three traits; FDM II clear communication, FFSS interactional patterns and SFI expressiveness (See figure 1).

Problems have been identified when the traditional Campbell and Fiske (1959) MTMM model is used to establish construct validity. These problems include a) zero order correlations in which correlations between scales are affected by the reliability of the scales, b) failure to provide a formal statistic for evaluating construct validity, c) degree to which operationalizations measure the traits, and d) the inability to discriminate trait and method variance (Bagozzi, 1978; Perosa & Perosa, 1990; Ten Haaf, Janssens & Gerris, 1994). Decisions regarding the MTMM model have been made on simple visual inspection. For convergence to exist, correlations from various measures of the same trait should be high enough to warrant further examination of the model. The criterion for establishing discriminant validity is more subjective and less clear (Ten Haaf et al.). For purposes of this study, adaptations to the MTMM model were made (see Procedure section).

Sample

A convenience sample consisted of twenty-five family systems with a VLBW infant as the product of the mother's first pregnancy. Statistical power analysis for a planned multiple regression dictated a sample size of 25, alpha 0.05, and R square of 0.61 for thirteen independent variables. A power of 0.85 was established for this procedure. The family system consisted of two or more adults who resided together and who indicated a commitment to each other. The VLBW infant was between six and twelve months of adjusted age at the time of the interview. The adjusted age was calculated by subtracting the weeks of prematurity from the infant's chronological age. The VLBW infant was a singleton delivery without congenital anomalies.

Instrumentation

Three self-report family functioning measures were employed. These measures were selected for several reasons, including: a) early development of the instrument, b) insufficient psychometric testing, and c) use of a systems framework to study families (Sawin & Harrigan, 1994).

Family Dynamics Measure II

Barnhill's (1979) conceptualization of family functioning provided the foundation for the formulation of one of the three self-report measures, the FDM II. Four themes of healthy family functioning were identified by Barnhill and extended across eight bipolar subscales. The FDM II consists of six subscales: individuation, clear communication, stability, flexibility, mutuality, and role reciprocity. The FDM II is a 62 item measure with a 6 point Likert-type format (from strongly agree to strongly disagree). The lower the score the healthier the family. Positively and negatively stated items are randomly ordered. The subscales of mutuality and clear communication are each comprised of eleven items. Each subscale has reverse-scored items. Subscale scores are derived by summing items. Preliminary internal consistency estimates varied from .68 to .92 for the 66-item measure (White et al., 1997).

Family Functioning Style Scale

The FFSS was created to measure the concepts aimed at enabling and empowering families. A family's style is defined as the family's strengths, capabilities, or uniqueness in coping with life events and promoting growth (Sawin & Harrigan, 1994). A five-point Likert scale includes these choices: Not at all, a little, sometimes, generally, or always "like my family" (Sawin & Harrigan). The FFSS consists of 26 items across five subscales: interactional patterns, family values, coping strategies, family commitment, and resource mobilization. Internal consistency estimates for the FFSS range from .77 to .85. The family commitment subscale consists of three items, and the interactional patterns scale consists of ten items. Individual scores are derived by item summation.

Self-Report Family Inventory

The SFI consists of 36 items. Concepts central to the model are competence and style. Competence refers to the family's ability to organize and manage. Egalitarian leadership, strong parental coalition, and clear generational boundaries characterize competence. Competent family members are autonomous and their behavior reflects trust, clear, and direct communication and the ability to solve conflict. Style is the second level of family process, and it occurs on a continuum. At one end, family members seek gratification primarily from within the family and at the other end, family members seek gratification outside the family (Sawin & Harrigan, 1994). A Likert-like scale ranging from 1 to 5 is used. A one is "fits our family very well , 3 is fits our family some" and 5 is "does not fit our family." The SFI has five subscales: health/competence, cohesion, conflict, leadership, and expressiveness. Internal consistency estimates for SFI subscales range from .84 to .88. The cohesion and expressiveness subscales are each comprised of five items. Reverse-scored items are present within each subscale. Items are summed for each of the five subscale scores.

Mutuality focuses on the emotional closeness and unity between family members. The ability to share and clarify messages expressed by family members represents communication. (See figure 2).

[FIGURE 2 OMITTED]

Procedure

The procedure for data collection was adapted from the research of Goldsmith et al. (1991) and designed to contain few impediments to convergence. Therefore, the three family functioning measures were administered at a single session to minimize variations due to situational factors. Such an approach was appropriate for assembling the content common to similarly named scales from different measures. Additionally, measures were administered with six possible random combinations. To assure MTMM is appropriately applied, the researcher assumed that performance on each measure was independent and not influenced by performance on any other measure.

Approval for the study was received from the human subjects committees at each of the two participating facilities. A list of VLBW infants was provided to the researcher in addition to the mother's name, her gravid status, and the infant's date of birth. Data collection began in 1995. Once infants were identified, the researcher contacted each mother of a VLBW infant by telephone to determine their willingness to participate. Nine family systems consisting of the mother, adult partner, and VLBW infant were recruited in 1995.

To afford increased privacy to mothers, the procedure was revised in 1996. Mothers were subsequently recruited from a local newspaper that provided names of the mothers, addresses, the hospital of infant birth in addition to infant birthweight. If infants were VLBW, the mother was contacted by letter. The study purpose was explained; the criteria for inclusion and a request to participate were provided. Mothers were given a telephone number to call to answer any questions and to enroll in the study. Those who agreed to participate were asked to establish an interview time when the researcher could visit them and administer the three family functioning measures.

If a suitable time could not be established, the measures were mailed to the families. Two hundred and forty four letters were sent to families with VLBW infants. Separate packets were provided for each of the two adult family members; the three measures were randomly combined. Families were instructed by telephone to complete the measures independently, and without discussion of their individual responses. The couples marital status and socioeconomic status was ascertained by telephone. If the mailed measures were not returned within the time allotted, the couple received a follow-up telephone call. Sixteen letters were returned with insufficient address, and fifteen mothers called to inform the researcher that they did not qualify for study participation. From the remaining 213 families contacted by letter, sixteen families participated.

Demographic data were collected on the family systems. Demographics included: a) hospital of the infant's birth, b) maternal position within the family system, c) infant age at interview, d) location of the interview, e) infant gestational age at birth, f) mother's education, g) mother's occupation, h) mother's hollingshead score, i) adult partner's education, j) adult partner's occupation, k) adult partner's hollingshead score, l) order of the family functioning measures, and m) the infant's birthweight in grams.

At the completion of the interview session, the social status of the family was determined by the Hollingshead Four Factor index of Social Status (Hollingshead, 1975). Maternal position within the family was categorized as spouse, friend, or child. These categories were based on the following criteria. If the mother was married and living in a family with the husband, the mother's position was categorized as spouse. If the adult partner with whom the mother lived with was not the marriage partner, the maternal position was categorized as friend. When the mother was living in a family headed by the mother's parents, the maternal position was categorized as child. Data collection concluded in 2001.

RESULTS

An alpha of 0.05 was used for all tests of significance. Data were analyzed by descriptive and inferential statistics including Cronbach alpha, t-test, multiple regression, Pearson product correlation, and factor analysis.

Descriptive Analysis

Means and standard deviations for seven of the thirteen demographic variables were obtained. Shapiro-Wilk tests for normalcy was performed revealing a lack of normalcy on five of the seven demographic variables. To induce normalcy, the five variables were logarithmically transformed (See table 1). Table 2 presents the remaining six demographic variables: maternal education, maternal occupation and maternal Hollingshead score, adult partner education, adult partner occupation, and adult partner Hollingshead score. Table 2 also includes means, standard deviations, and logarithmic transformation for the variables lacking normalcy.

Table 3 presents the six maternal dependent variables: FDM II mutuality, FFSS family commitment, SFI cohesion, FDM II communication, FFSS interactional patterns, and SFI expressiveness. Four of the six dependent variables lacked normalcy and were logarithmically transformed. Likewise, the six adult partner dependent variables are presented in Table 4 with the appropriate transformations.

Cronbach alpha coefficients were calculated for the six dependent variables of the mother and the six adult partner dependent variables (See Tables 5 and 6). The three socioeconomic independent variables were also subjected to t-test analyses. Statistically significant differences were determined on the transformed occupation (t = -2.51, p = 0.02) and Hollingshead score (t = -2.57, p = 0.02). Further analyses were conducted separately for the mother and the adult partner.

Regression Analysis

Six hierarchical regression equations were executed, one for each maternal dependent variable. The thirteen demographic variables were employed as independent variables. R square values for maternal dependent variables are: FDM II log mutuality (.70), FFSS log family commitment (.74), SFI cohesion (.66), FDM II communication (.59), FFSS log interactional patterns (.60), and SFI log expressiveness (.70). With the same thirteen independent variables, six hierarchical regression equations were performed for the adult partner dependent variables. R square values for these dependent variables are: FDM II mutuality (.45), FFSS log family commitment (.51), SFI cohesion (.57), FDM II communication (.54), FFSS interactional patterns (.75) and SFI log expressiveness (.60). Graphic and residual analyses were used to test for violations of causal modeling and statistical assumptions underlying the use of multiple regression. Multiple regression residuals, or the unexplained variance from the regression equations, were employed in the generation of the MTMM matrix of Pearson correlation coefficients (See Tables 5 and 6). Appropriate residual plots demonstrated horizontal scatter with no systematic trends.

Convergent and Discriminant Validity

MTMM matrices were created for mothers and their adult partners as recommended by Campbell and Fiske (1959). Correlation coefficients were computed between each trait and method (See tables 5 & 6). The most direct evidence for determining construct validity comes when correlations are generated between two different methods for measuring the same trait. Two methods, FDM II log mutuality and FFSS log family commitment, examined the maternal trait: mutuality. A coefficient of 0.31 was calculated, indicating poor convergence. Convergent validity entries should be higher than those correlations between measures that have neither method nor trait in common. For instance, a mutuality coefficient of 0.31 (FDM II log mutuality and FFSS log family commitment) should surpass coefficients of 0.39 (FDM II communication and FFSS log family commitment) and 0.35 (FDM II log mutuality and FFSS log interactional patterns). See Table 5. Clearly, a coefficient of 0.31 does not surpass the "heterotrait-heteromethod" values of 0.39 and 0.35.

Convergent validity coefficients should exceed the correlations between measures of different traits by a single method. The MTMM matrix reveals convergence coefficients of 0.31 (FDM II log mutuality and FFSS log family commitment) and--0.03 (FFSS log family commitment and SFI cohesion). These coefficients do not exceed 0.65 (FDM II log mutuality and FDM II communication) and 0.74 (FFSS log family commitment and FFSS log interactional patterns). SFI cohesion and SFI log expressiveness yield a convergence coefficient of 0.02. Because the last two requirements were not fulfilled, insufficient evidence exists for discriminant validity.

Validity is predicated on adequate internal consistency or reliability. When adult partner data were analyzed, the alpha for the FFSS log family commitment was -0.07. Further examination of the MTMM matrix was aborted.

Factor Analysis

To examine the underlying structure of the MTMM matrix, factor analysis was employed as an alternative method of analysis (Ten Haaf et al., 1994). Factor analysis was selected because the analysis is unaffected by the reliability of the scales and unrelated to zero order correlations (Cole, 1987). The analysis was conducted in two phases. During the first phase, factor extraction was conducted using the mother's six dependent variables; two factors were extracted--both with Eigenvalues of one or greater (Kristjanson, Sloan, Dudgeon & Adaskin, 1996). Two factors were also extracted when the adult partner's six dependent measures were applied. Orthogonal factor rotations were performed. During the second phase, factors loadings equal to or exceeding 0.4 were accepted. Five dependent variables loaded on factor one and two dependent variables on factor two (see Table 7). The FDM II log mutuality that remained skewed after logarithmic transformation, loaded on factors one and two, suggesting a lack of discrimination. The two remaining skewed dependent variables, FFSS log family commitment and FFSS log interactional patterns, loaded on factor one.

For the adult partners, four dependent variables loaded on factor one and two dependent variables on factor two after orthogonal rotation. One of the six dependent variables, FFSS log family commitment, remained skewed after logarithmic transformation and loaded on factor two.

DISCUSSION

Establishing construct validity is a difficult and challenging task. To assist in defining connections between concepts or traits, three self-report family functioning measures, with different conceptual underpinnings, were employed. Based on results of this study, construct validity was not established. Contrarily, this study confirmed Perosa and Perosa's (1990) assertion that efforts to reduce the salient features of family functioning to basic traits, particularly mutuality, are premature. When the mothers of VLBW infants were examined, a reliability of 0.4 was obtained on the FFSS family commitment subscale that may have been attributed to skewness, low-test variance on a three-item subscale, and homogeneity within the family system of mothers. Two of the three maternal dependent measures, FDM II mutuality and FFSS family commitment, also remained skewed after logarithmic transformation, limiting the ability to draw inferences. When factor analysis was executed on FDM II log mutuality, this subscale loaded on two factors suggesting a lack of clarity of the underlying attribute and a lack of discrimination.

Confusion in trait definitions must be considered; otherwise, validation studies are unlikely to confirm expected predictions. When examining the trait, communication, reliability coefficients exceeded 0.76 on FDM II communication, FFSS interactional patterns, and SFI expressiveness. Convergence was evidenced on FDM II communication and FFSS log interactional patterns (0.35, p = 0.10), FDM II communication and SFI log expressiveness (-0.56, p = 0.005), and FFSS log interactional patterns and SFI log expressiveness (-0.66, p = 0.0006). The FFSS log interactional patterns remained skewed post-logarithmic transformation. FDM II communication, FFSS log interactional patterns, and SFI log expressiveness loaded on one factor suggesting convergence and discrimination.

The adult partners within the family system of VLBW infants were more heterogeneous consisting of spouses and the mother's parents. An alpha of -0.07 was obtained, however, on the FFSS family commitment; an alpha of 0.54 was obtained for SFI cohesion. Factor analysis revealed FDM II mutuality and SFI cohesion loading on the first factor and FFSS log family commitment loading on the second factor. The FFSS log family commitment subscale remained skewed post-logarithmic transformation. Again, inferences are questionable; convergence is suspect. On the trait, communication, alpha coefficients ranged from 0.58 to 0.86. All three subscales: FDM II communication, FFSS interactional patterns, and SFI log expressiveness loaded on one factor.

RECOMMENDATIONS

Recommendations for future research include performing comparative studies with VLBW infant and full-term infant family systems, obtaining a larger and more heterogeneous sample, increasing the number of items on the FFSS subscale of family commitment, and reverse-scoring the entire FFSS measure. Based upon factor analysis, careful scrutiny of emergent factors, with explicit factor naming, may contribute to clearer trait definitions and more discriminating methods once related measures are combined.

Figure 1. Trait definitions and sample items

MUTUALITY

FDM II Mutuality = "emotional closeness; intimacy, that is possible when individuals have clearly defined boundaries" (Barnhill, 1979, p. 95) Sample item: "I look out for others"

FFSS Family commitment = making decisions that benefit the entire family depending on other family members, and trying to solve problems within the family unit before asking for help (Dunst, Trivette & Deal, (1994) Sample item: "We make decisions like moving or changing jobs for the good of all family members"

SFI Cohesion = "family togetherness, satisfaction received from inside the family versus outside, and spending time together" (Beavers & Hampson, 2000, p. 136) Sample item: "Our happiest times are at home"

COMMUNICATION

FDM II Clear communication = "successful exchange of information; check out communication to clarify meaning" (Barnhill, 1979, p. 96) Sample item: "We ask questions when we don't understand each other"

FFSS Interactional patterns = not taking each other for granted, spending time together, listening to all points of view, and sharing concerns and feelings (Dunst et al., 1994) Sample item: "We are always willing to pitch in and help each other"

SFI Expressiveness = "verbal and nonverbal expression of warmth, caring and closeness" (Beavers & Hampson, 2000, p. 136) Sample item:" Our family members touch and hug each other"

Table 1. Demographics of Families with Very Low Birthweight Infants
(N=25)

                            MEAN    STAN. DEV.   SHAPIRO-WILK   P VALUE

Hospital of Birth            1.6        1.04         0.62       <.0001
1=public, level 3
  nursery (N= 18)
2=private, level 3
  nursery (N= 1)
3=private (N=4)
4=private (N=2)
5=private (N=0)
Log transformation           0.31       0.53         0.61       <.0001

Maternal Position
  within Family              1.48       0.87         0.53       <.0001
1=married (n= 19)
2=friend (N=0)
3=child (N=6)
Log transformation           0.26       0.48         0.53       <.0001

Infant Adjusted Age at
  Interview (months)         8.16       2.03         0.85       0.002
Log transformation           2.07       0.24         0.87       0.004

Location of Interview        2.36       0.95         0.95       <.0001
1=home, by mail (N=8)
2=health care office
  (N=0)
3=home, interview
  (N=17)
Log transformation           0.75       0.52         0.59       <.0001

Infant Gestational Age
  at Birth (weeks)          29.28       1.86         0.95       0.29

Family Functioning
  Measures Order             3.44       1.71         0.90       0.02
I=FDM II, SFI, FFSS
  (N=3)
2=SFI, FFSS, FDM II
  (N=6)
3=FFSS, SFI, FDM II
  (N=5)
4=FDM II, FFSS, SFI
  (N=4)
5=SFI, FDM II, FFSS
  (N=2)
6=FFSS, FDM II, SFI
  (N=5)
Log transformation           1.09       0.57         0.9        0.02

Birthweight (grams)       1184.96     261.77         0.93       0.07

Table 2. Demographics of mother and Adult Partner (N=25)

                             MEAN   STAN. DEV.   SHAPIRO-WILK   P VALUE

Mother's Education          13.32      4.16          0.93       0.1
1=<7th grade
2=8th-9th grade
3=10th-11 th grade
4=12th grade
5=partial college
6=college or university
  education
7=graduate degree (times
  factor of three)

Mother's Occupation         15.6      13.41          0.76       <.0001
1=farm laborers, no
  occupation
2=unskilled workers
3=machine operators
4=small business
5=clerical and sales
  workers
6=technicians
7=small business owners
8=administrators,
  lesser professions
9=major professions, high
  executives (times
  factor of five)
Logarithmically
  transformation             2.39      0.86          0.78       <.0001

Mother's Hollingshead
  Score                     28.92     15.71          0.87       0.004
Logarithmically
  transformation             3.32      0.54          0.93       0.10

Adult Partner's Education   13.8       3.87          0.86       0.003
Logarithmically
  transformation             2.58      0.33          0.082      0.0005

Adult Partner's
  Occupation                23.8      13.3           0.9        0.02
Logarithmically
  transformation             2.96      0.74          0.85       0.002

Adult Partner's
  Hollingshead Score        37.6      15.7           0.94       0.12

Table 3. Mother's dependent variables

                                     MEAN    STAN.   SHAPIRO-   P VALUE
                                              DEV.     WILK

MUTUALITY SUBSCALES
FDM II Mutuality (N=24)              57.38   6.86      0.8       0.0003
Logarithimically transformed          4.04   0.13      0.76      0.0001
FFSS Family Commitment (N=25)        13.64   1.52      0.83      0.0007
Logarithimically transformed          2.61   0.12      0.82      0.0005
SFI Cohesion (N=25)                  11.68   3.15      0.94      0.12

COMMUNICATION SUBSCALES
FDM II Communication (N=25)          52.12   6.86      0.92      0.05
FFSS Interactional Patterns (N=23)   43      6.41      0.86      0.004
Logarithimically transformed          3.75   0.16      0.82      0.0009
SFI Expressiveness (N=25)             9      3.9       0.78      0.0001
Logarithimically transformed          2.12   0.38      0.83      0.0008

Table 4. Adult Partner's dependent variables

                                     MEAN    STAN.   SHAPIRO-   P VALUE
                                              DEV.     WILK

MUTUALITY SUBSCALES
FDM II Mutuality (N=25)              55.84   6.07      0.97      0.54
FFSS Family Commitment (N=25)        13.52   1.53      0.84      0.001
Logarithimically transformed          2.6    0.12      0.83      0.007
SFI Cohesion (N=25)                  11.42   2.87      0.93      0.1

COMMUNICATION SUBSCALES
FDM II Communication (N=25)          51.44   6.55      0.98      0.8
FFSS Interactional Patterns (N=24)   41.79   5.49      0.93      0.08
SFI Expressiveness (N=24)             9.35   2.84      0.89      0.02
Logarithimically transformed          2.2    0.28      0.94      0.2

Table 5. Multitrait-multimethod correlation matrix of multiple
regression residuals for mutuality (trait) and communication (trait)
subscales from three family self-report measures (methods) of mothers
with very low birthweight infants (N=25)

                              FDM II         FFSS             SFI
                            Mutuality      Log Family        Cohesion
                                           Commitment

MUTUALITY SUBSCALES
FDM II Log Mutuality     (.82)
FFSS Log Family           .31 (p=.14)     (.40)
  Commitment
SFI Cohesion              .29 (p=.17)     -.03 (p=.89)     (.64)

COMMUNICATION
SUBSCALES
FDM II Communication      .65 (p=.0001)    .39 (p=.05)     -.16 (p=.44)
FFSS Log Interactional    .35 (p=.10).      74 (p=.0001)   -.23 (p=.30)
  Patterns
SFI Log Expressiveness   -.56 (p=.005)    -.51 (p=.01)      .02 (p=.92)

                             FDM II            FFSS          SFI
                         Communication     Interactional     Log
                                              Patterns     Express-
                                                           iveness

MUTUALITY SUBSCALES
FDM II Log Mutuality
FFSS Log Family
  Commitment
SFI Cohesion

COMMUNICATION
SUBSCALES
FDM II Communication     (.76)
FFSS Log Interactional    .73 (p=.0001)   (.87)
  Patterns
SFI Log Expressiveness   -.88 (p=.0001)   -.66 (p=.0006)    (.81)

Note: Diagonal elements in parentheses are the Cronbach alpha
coefficients of reliability
FDM II (Family Dynamics Measure II)
FFSS (Family Functioning Style Scale)
SFI (Self-report Family Inventory)

Table 6. Multitrait-multimethod correlation matrix of multiple
regression residuals for mutuality (trait) and communication
(trait) subscales from three family self-report measures (methods) of
adult partners with very low birthweight infants (N=25)

                            FDM II             FFSS            SFI
                           Mutuality        Log Family       Cohesion
                                                            Commitment

MUTUALITY SUBSCALES
FDM II Mutuality         (.78)
FFSS Log Family           .05 (p=.82)     (-.07)
  Commitment
SFI Cohesion              .05 (p=.79)       .10 (p=.65)    (.54)

COMMUNICATION
SUBSCALES
FDM II Communication      .49 (p=.01)       .14 (p=.50)     .57 (p=.79)
FFSS Interactional        .54 (p=.006)      .23 (p=..29)   -.04 (p=.84)
  Patterns
SFI Log Expressiveness   -.59 (p=.002)     -.13 (p=.53)     .12 (p=.59)

                            FDM II           FFSS          SFI
                         Communication   Interactional     Log
                                           Patterns      Express-
                                                         iveness

MUTUALITY SUBSCALES
FDM II Mutuality
FFSS Log Family
  Commitment
SFI Cohesion

COMMUNICATION
SUBSCALES
FDM II Communication     (.75)
FFSS Interactional        .59 (p=.002)   (.86)
  Patterns
SFI Log Expressiveness   -.36 (p=.09)    -.40 (p=.06)     (.58)

Note: Diagonal elements in parentheses are the Cronbach alpha
coefficients of reliability
FDM II (Family Dynamics Measure II)
FFSS (Family Functioning Style Scale)
SFI (Self-report Family Inventory)

Table 7. Factor Analysis Orthogonal Transformed Matrix

                        MOTHERS
Mutuality Subscales               Factor 1   Factor 2

FDM II Mutuality                    0.63       0.61
FFSS Log Family Commitment          0.77      -0.11
SFI Cohesion                       -0.12       0.91

Communication Subscales           Factor 1   Factor 2

DM II Communication                 0.92       0.08
FSS Interactional Patterns          0.88      -0.24
FI Log Expressiveness              -0.86      -0.13

                    ADULT PARTNERS
Mutuality Subscales               Factor 1   Factor 2

FDM II Log Mutuality                0.82      -0.16
FFSS Log Family Commitment          0.23       0.7
SFI Cohesion                       -0.2        0.72

Communication Subscales           Factor 1   Factor 2

FDM II Communication                0.74       0.16
FFSS Log Interactional Patterns     0.81       0.19
SFI Log Expressiveness             -0.74       0.17

Note: Factor loadings [greater than or equal to] 0.4 are italicized

REFERENCES

Als, H. (1982). Toward a synactive theory of development: Promise for the assessment and support of infant individuality. Infant Mental Health Journal, 3(4), 229-243.

American College of Obstetricians and Gynecologists (1995). ACOG technical bulletin: Preterm labor. International Journal of Gynecology and Obstetrics, 50, 303-313.

Bagozzi, R. P. (1978). The construct validity of the affective, behavioral, and cognitive components of attitude by analysis of covariance structures. Multivariate Behavioral Research, 13, 9-31.

Barnhill, L R. (1979). Healthy family systems. The Family Coordinator, 28, 94-100.

Beavers, W. R. & Hampson, R. B. (1990). Successful families: Assessment and intervention. New York: W. W. Norton.

Beavers, R. & Hampson, R. B. (2000). The Beavers systems model of family functioning. Journal of Family Therapy, 22(2), 128-143.

Beckwith, L & Parmelee, A. (1986). EEG patterns of preterm infants, home and later IQ. Child Development, 57, 777-789.

Belsky, J., Lerner, R. & Spanier, G. (1984). The child in the family. Reading, MA: Addison-Wesly.

Bronfenbrenner, U. (1977). Toward an experimental ecology of human development. American Psychologist, 32, 513-531.

Bronfenbrenner, U. (1986). Ecology of the family as a context for human development: Research perspectives. Developmental Psychology, 22(6), 723-742.

Bronfenbrenner, U. (1992). Ecological Systems Theory. Six Theories of Child Development: Revised Formulations and Current Issues. Annals of Child Development, 6, 187-249.

Campbell, D.T. & Fiske, D.W. (1959). Convergent and discriminant validation by the multitrait-multimethod matrix. Psychological Bulletin, 56(2), 81-105.

Carlson, C. & Grotevant, H. (1987). Rejoinder: The challenges of reconciling family theory and method. Journal of Family Psychology, 3(1), 62-65.

Cole, D. (1987). Utility of confirmatory factor analysis in test validation research. Journal of Consulting and Clinical Psychology, 55(4), 584-594.

Dickerson, V. C. & Coyne, J. C. (1987). Family cohesion and control: A multitrait-multimethod study. Journal of Marital and Family Therapy, 13(3), 275-285.

Dunst, C., Trivette, C. & Deal, A. (1994). Supporting and Strengthening Families: Methods, Strategies and Practices, 1, Brookline Books.

Goldenberg, R. L. (1994). The prevention of low birthweight and its sequelae. Preventative Medicine, 23(5), 627-31.

Goldenberg, R. L & Rouse, D. J. (1998). Prevention of premature birth. The New England Journal of Medicine, 339(5), 313-320.

Goldsmith, H. H., Rieser-Danner, L. A. & Briggs, S. (1991). Evaluating convergent and discriminant validity of temperament questionnaires for preschoolers, toddlers, and infants. Developmental Psychology, 27(4), 566-579.

Hack, M., Horbar, J. D., Malloy, M. H., Tyson, J. E., Wright, E. & Wright, L. (1991). Very low birthweight outcomes of the National Institute of Child Health and Human Development Neonatal Network. Pediatrics, 87(5), 587-97.

Hollingshead, A. (1975). Four Factor Index of Social Status. New Haven: Yale University.

Korner, A. E, Stevenson, D. K., Kraemer, H. C., Spiker, D., Scott, D. T., Constantinou, J. & Dimiceli, S. (1993). Prediction of the development of low birth weight preterm infants by a new neonatal medical index. Developmental and Behavioral Pediatrics, 14(2), 106-111.

Kristjanson, L. J., Sloan, J. A., Dudgeon, D. & Adaskin, E. (1996). Family members' perceptions of palliative cancer care: Predictors of family functioning and family members' health. Journal of Palliative Care, 12(4), 10-20.

Lasky, P., Buckwalter, K. C., Whall, A., Lederman, R., Speer, J., McLane, A., King, J. M. & White, M. A. (1985). Developing an instrument for the assessment of family dynamics. Western Journal of Nursing Research, 7(1), 40-57.

Leonard, C. H., Clyman, R. I., Piecuch, R. E., Juster, R. P., Ballard, R. A. & Behle, M. B. (1990). Effect of medical and social risk factors on outcome of prematurity and very low birth weight. The Journal of Pediatrics, 116(4), 620-626.

McCormick, M. C. (1989). Long-term follow-up of infants discharged from neonatal intensive care units. JAMA, 261(12), 1767-1772.

McGrath, M. M. (1997). Estimating risk and protective indexes in high-risk children. Clinical Effectiveness in Nursing, 1, 92-104.

Minde, K., Perrotta, M. & Hellmann, J. (1988). Impact of delayed development in premature infants on mother-infant interaction: A prospective investigation. Journal of Pediatrics, 112(1), 136-142.

Morrison, J. (1990). Preterm birth: A puzzle worth solving. Obstetrics and Gynecology, 76(1 Suppl), 5S-12S.

Olson, D. (1986). Circumplex model VII validation studies and FACES III. Family Process, 25, 337-351.

Perosa, L. M. & Perosa, S. L. (1990). Convergent and discriminant validity for family self-report measures. Educational and Psychological Measurement, 50, 855-868.

Polit, D. F. & Hungler, B. P. (1999). Nursing Research Principles and Methods. Philadelphia: Lippincott.

Rouse, D. J., Goldenberg, R. I., Oliver, S. P., Cutter, G. R., Mennemeyer, S. T. & Fargason, C.A. (1994). Strategies for the prevention of early-onset neonatal group B streptococcal sepsis: A decision analysis. Obstetrics and Gynecology, 83, 483-94.

Sawin, K.J. & Harrigan, M. (1994). Measures of family functioning for research and practice. Scholarly Inquiry for Nursing Practice, 8(1), 92-100.

Taylor, H. G., Klein, N., Schatschneider, C. & Hack, M. (1998). Predictors of early school age outcomes in very low birth weight children. Developmental and Behavioral Pediatrics, 19(4), 235-243.

Ten Haaf, P. G. J., Janssens, J. M. A. M. & Gerris, J. R. M. (1994). Child-rearing measures: Convergent and discriminant validity. European Journal of Psychological Assessment, 10(2), 111-128.

Thompson, R. J., Catlett, A. T., Oehler, J. M., Gustafson, K. E., Goldstein, R. F & Prochaska, J. J. (1998). Home environment and developmental outcomes of african american and white infants with low birthweight. Children's Health Care, 27(1), 1-14.

Thompson, R. J., Goldstein, R. E, Oehler, J. M., Gustafson, K. E., Catlett, A. T. & Brazy, J. E. (1994). Developmental outcome of very low birth weight infants as a function of biological risk and psychosocial risk. Developmental and Behavioral Pediatrics, 15(4), 232-238.

Trivette, C. M., Dunst, C. J., Deal, A. G., Hammer, A. W. & Propst, S. (1990). Assessing family strengths and family functioning style, Topics in Early Childhood Special Education, 10, 16-35.

Ventura, S, Martin, J, Curtin, S. & Mathews, T. (1999). April. Births: Final data for 1997. National Vital Statistics Reports, 47(18), 1-74.

White, M. A., Wu, Y. B. & Grzankowski, J. (1997, November). Using informatics to study family dynamics cross culturally. Poster session presented at the annual meeting of the American Academy of Nursing, Washington, D. C.

Deborah B. Nelson, DSN, ARNP, is an Adjunct Professor at Pensacola Junior College, Warrington Campus in Pensacola, Florida. She may be contacted at dlbn5@bellsouth.net. The author wishes to thank the University of Mobile in Mobile, Alabama.