TOWARD A FULLER UNDERSTANDING OF MATERNAL POSTPARTU M PSYCHIATRIC DISORDERS: PSYCHOMETRIC PROPERTIES AND PREDICTIVE QUALITIES OF THE POSTPARTUM DEPRESSION SCREENING SC ALE By JO M. VOGELI B.A., University of Colorado Boulder, 2010 A thesis submitted to the Faculty of the Graduate School of the University of Colorado in partial fulfillment of the requirements for the degree of Master of Arts Clinical Health Psychology Program 2014
2014 JO M. VOGELI ALL RIGHTS RESERVED
ii This thesis for the Master of Arts degree by Jo M. Vogeli Has been approved for the Clinical Health Psychology Program by Peter Kaplan, Chair Kevin Everhart Edward Dill November 19, 2014
iii Vogeli, Jo M. (M.A., Clinical Health Psychology) Toward A Fuller Understanding Of Maternal Postpartu m Psychiatric Disorders: Psychometric Properties And Predictive Qualities Of The Postpartum Depression Screening Scale Thesis directed by Professor Peter Kaplan. ABSTRACT Postpartum depression (PPD) is a mood disorder aff ecting approximately 20% of women who give birth each year in the United States The Postpartum Depression Screening Scale (PDSS) was developed to address the maternal experience after trends in qualitative research were discovered. The current r eplication study examined the reliability and validity of the PDSS as an effectiv e screener for PPD, compared the PDSS to the BDI-II with regard to sensitivity, specifici ty and predictive value, and analyzed the relationship between the Guilt/Shame subscale of th e PDSS and Structuring/NonIntrusiveness on the Emotional Availability Scale. The results largely replicated the original PDSS psychometrics. Model fit was supporte d through the Tucker-Lewis Index (0.84) with RMSEA of .07, and t-values with a minim um of 12.79 indicated that each weight was sufficient to support subscale fit. Inte rnal consistency was tested, and the PDSS Total score yielded an alpha coefficient of 0. 95, signifying subscale items were highly correlated with the overall construct of dep ression. Discriminant function analysis was significant at the p <.01 level, meaning the PDSS was able to correctly provide diagnostic classification at a rate higher than cha nce alone. The PDSSÂ’ Canonical discriminant function correlations were in the very good to excellent range for content scalesÂ’ contribution to classification results. The overall sensitivity for the PDSS for MDD diagnosis was at 76% and specificity was 83%, a nd these results were consistent
iv with results from the BDI-II. The ROC curve analysi s, which measured expected performance at various cutoff scores, was in the ex cellent range at .912. Regression analysis showed both the BDI-II and PDSS provided p ositive predictive value for the detection of a current diagnosis of clinical depres sion, with the PDSS accounting for slightly more of the variance over the BDI-II. PDSS Guilt/Shame subscale scores were highest in mothers rated as Â“over-structuringÂ” on t he EAS, consistent with the hypothesis that elevated levels on the EAS would indicate the presence of intrusive parenting. These results are an important contribution to research b ecause they essentially replicate Beck & GableÂ’s findings on the effectiveness of the PDSS for identifying mothers in need of follow-up and possibly treatment for PPD. The form and content of this abstract are approved. I recommend its publication. Approved: Peter Kaplan
v DEDICATION This thesis is dedicated to my husband, Dave, and m y son, Leo. This is our work, and I thank you for the blood, sweat, and tears we all endured during this process. Thank you both for continually reminding me of the import ant things in life. Dave, I am a better person each morning as I wake up by your side, and I look forward to continuing to grow along with you in this journey of life. Leo, you ar e my heart and soul, and I look forward to watching you grow into the fine man I know you w ill be. I love you both, 100%. And please always remember, Â“IÂ’ll be rigghhht here.Â”
vi ACKNOWLEDGEMENTS I would like to start off by acknowledging any and all of the past individuals who collected and recorded data on multiple projects, w hich eventually led to the comprehensive database from which I had to work. Ta ra Curly and Kristen Ruhl, I thank you for the tedious work of helping me with enterin g the item level data for the PDSS. To my advisors and thesis committee: Peter Kaplan, Kev in Everhart, and Ed Dill, your challenges have helped me to discover a higher leve l of researcher within myself. Although not always evident, I appreciate your supp ort and guidance throughout this process. Stephanie Hooker, you have unconditionally provided a wealth of information during this process, and I am a better researcher a s a result. And last but certainly not least, I want to acknowledge my cohort and labmates : Kaile Ross, Tattiana Romo, Shiva Fekri, Lacey Clement, and Ryan Asherin. I have a tr emendous amount of gratitude for your unwavering support during my thesis journey. Y ou are my temporary family, future colleagues, and lifelong friends. I could not ask f or a better set of people to fill these roles.
vii TABLE OF CONTENTS CHAPTER I. BACKGROUND Importance of Early Detection of PPD ................................................... ...............2 Screening for PPD...................... ................................................... ...........................3 PDSS Psychometric Properties .......... ................................................... ...................7 Maternal Experience of PPD ........... ................................................... ....................9 Cognitive Factors Involved in PPD ..... ................................................... ...............11 Emotional Availability ................ ................................................... ........................12 Emotional Availability, PPD, and Intrus iveness............................................ ........13 Purpose of Present Study .............. ................................................... ......................16 Hypotheses and Specific Aims .......... ................................................... .................17 II. METHOD Participants ........................ ................................................... ................................19 Procedures .......................... ................................................... ...............................23 Data Analyses ....................... ................................................... ............................23 Power ................................ ................................................... .................................24 Data Cleaning and Transformations .... ................................................... ..............25 Threats to Validity ................. ................................................... ...........................25 III. RESULTS Aim 1: PDSS Psychometric Property Repli cation ........................................... .....27 Aim 2: Predictive Quality of PDSS Compa red to BDI-II ..................................... 36 Aim 3: Guilt/Shame Subscale Predicting Intrusiveness on EA S ...........................38
viii IV. DISCUSSION ............................. ................................................... .......................40 REFERENCES ....................................... ................................................... .......................48
ix LIST OF TABLES TABLE 1. Demographic and Diagnostic Data ....... ................................................... .............20 2. Demographic Data for EAS/GLT Sample .. ................................................... ......22 3. PDSS, BDI-II and SCID Summary ......... ................................................... ...........27 4. PDSS Interscale Correlations .......... ................................................... ...................28 5. CFA: Maximum Likelihood Dimensions and Loadings (N=245) ........................30 6. Internal Consistency Estimates for 35-I tem PDSS .......................................... .....31 7. Discriminant Function Classification, Sensitivi ty, and Specificity of PDSS and BDI-II for MDD Detection .. ................................................... .........32 8. Correlations Between PDSS Symptom Conte nt Scales and Canonical Discriminant Function ...... ................................................... ................33 9. Hierarchical Regression of the Diagnosi s of Postpartum Depression on the BDI-II and PDSS ............... ................................................... .....................37 10. Hierarchical Regression of the Diagnosis of Postpartum Depression on the PDSS and BDI-II ............... ................................................... .....................38 11. EAS Category and GLT Scores ............ ................................................... .............39
x LIST OF FIGURES FIGURE 1. ROC Curve for PDSS: Major Depression ................................................... .......34 2. ROC curve for BDI-II: Major depression ................................................... ......... 35 3. ROC Curve for PDSS and BDI-II at Recomm ended Cutoff Scores .................... 36
xi LIST OF ABBREVIATIONS BDI-II Beck Depression Inventory, 2nd Edition EAS Emotional Availability Scales EPDS Edinburg Postpartum Depression Scale PDSS Postpartum Depression Screening Scale PPD Postpartum Depression SCID Structured Clinical Interview for Diagnosis
1 CHAPTER I BACKGROUND Considered the most common complication of birth, p ostpartum depression (PPD) is a mood disorder affecting approximately 6 millio n women annually (Beck, 2002; Freed, Chan, Boger, & Tompson, 2012; Sit & Wisner, 2009). This figure does not take into account the estimated 15-25% of cases of PPD t hat go undetected (Beck, 2002; Freeman et al., 2005). PPD can have debilitating ef fects on women, spouses or partners, and is known to have adverse effects on infant atta chment and the mother-infant dyadic relationship. These effects, in turn, can lead to a variety of social and emotional deficits due to the lack of quality interactions and sensiti vity of mothers toward their infants (Biringen, Derscheid, Vliegen, Closson, & Easterbro oks, 2014; Carter, et al., 2001; Cummings & Davies, 1994). Deficits in infant cognit ive development have also been noted (Beck, 1998; Hanson, et al., 2013; Miller, Pa llant, & Negri, 2006). Indeed, research has shown that the early infant neglect that can oc cur with depressed mothers affects the directional organization of white matter, and is as sociated with neurocognitive deficits (Hanson, et al., 2013). Neglect can result from a l ack of sensitivity on the part of the mother toward the infant (Van Doesum, Hosman, Rikse n-Walraven, & Hoefnagles, 2007), and disruption in maternal emotional availab ility due to lack of sensitivity can carry through and negatively influence children in the preschool years and beyond (Trapolini, Ungerer, & McMahon, 2008). A disruption in the dyadic relationship in depressed mother-infant dyads is not limited to a l ack of sensitivity (Biringen, et al, 2014), as caregiver intrusiveness or non-optimal st ructuring can also lead to the
2 difficulties with emotional regulation (Muscat, Obs t, Cockshaw, & Thorpe, 2014), and other health concerns as well (Field, 1981; Lumeng et al., 2012). Importance of early detection of postpartum depress ion Early detection of PPD can significantly ameliorate or even prevent for infant developmental deficits by adjusting parenting behav iors to facilitate proper social and emotional development (Field, 1998; Murray & Cooper 1997; Freeman et al., 2005; Horowitz & Goodman, 2005). Additionally, a shift in maternal mood can be met with improved interactional responsiveness on behalf of the infant (Murray & Cooper, 1997). It is important to establish a healthy dyadic relat ionship early on because infants will begin to Â“normalizeÂ” these interactional patterns r egardless of their appropriateness (Nylen, Moran, & Franklin, 2006). Because the severity and chronicity of depression c an influence learning, early detection can be equally important, particularly wh en individuals are faced with sociodemographic adversity (Kaplan, Danko, Diaz, & Kalinka, 2011). Empirically supported interventions continue to emerge to aid i n the treatment of this disability (Dennis & Chung-Lee, 2006). For example, Field (199 7) found that infants of depressed mothers experienced emotional dysregulation as evid enced by greater negative affect. This was found to be partially due to inadequate st imulation and arousal modulation; however, mood-altering interventions for both mothe r and infants produced more responsive interactions and offset negative consequ ences. As mentioned above, PPD has been linked to lower maternal sensitivity, defined as the ability to provide a warm and inviting environment for a child (Biringen, Robinso n, & Emde, 2000). Too much or too little maternal structuring have also been linked t o PPD, and the adverse effects of non-
3 optimal structuring may be compounded by overly int rusive actions by the mother toward the infant (Kim, Mayes, Feldman, Leckman, & Swain, 2012). An intrusive style of parenting is related to, but distinct from low sens itivity. A parent may provide an overstimulating environment, and not recognize cues given by a child to stop (Biringen, 2000). Van Doesum, et al. (2007) found that targete d early interventions for mothers with low maternal sensitivity resulted in an enhancement of emotional interactions and increased feelings of maternal competence. Given th at early detection and treatment is known to offset a myriad of social, emotional, and relationship impacts for the mother (Field, 1998; Freeman et al., 2005; Trapolini, et a l., 2008; Van Doesum, et al., 2007), and a host of cognitive, language, and attachment defic its for the baby (Beck, 1998; Hanson, et al., 2013; Miller, Pallant, & Negri, 2006), earl y identification is key to enable proper treatment. Screening for PPD Routine screening for PPD generally includes one of three self-report instruments used to measure symptomatology in various ways (Evi ns, Theofrastous, & Galvin, 2000). These instruments include the Edinburgh Postnatal D epression Scale (EPDS; Cox, Holden, & Sagovsky, 1987), the Beck Depression Inve ntory II (BDI-II; Beck, Steer, & Brown, 1996), and most recently, the Postpartum Dep ression Screening Scale (PDSS; Beck & Gable, 2002, 2001b). All three instruments h ave a basis in the criteria for Major Depressive Disorder as outlined in the American Psy chological Associations (APA) Diagnostic and Statistical Manual for Mental Disord ers-Fourth Edition (DSM-IV; APA, 1994), with the PDSS having additional attributes b ased on qualitative measures to enhance the detection of PPD in women. With proper identification of an effective
4 screening tool, detection of PPD will increase, and it will be possible to develop appropriate treatment through a quality case concep tualization for women with this disorder. Development of the PDSS was a result of qualitativ e research examinging the underpinnings of PPD and the difficulty of detectio n with the current gold standard scales in place (Gjerdingen & Yawn, 2007). The psychometri c problems discovered with other screening scales included using the BDI, which was developed as a general depression instrument (Floyd & Widaman, 1995; Gjerdingen & Yaw n, 2007). Symptomatology such as fatigue, sleep disturbance, and weight gain or l oss are not valid measures of mood disorder for perinatal and postpartum period becaus e they are normal sequelae of childbirth (Floyd et al, 2007). Further evidence fo r the limited sensitivity of the BDI during the postpartum period includes that it has b een shown to be ineffective at detecting minor depression (Hanna, Jarman, & Savage, 2004). T hough adequate with overall results (Hanna, Jarman, & Savage, 2004), the relati vely low sensitivity and specificity of the first version of the BDI suggested it was not a sufficient a screener for depression among women in the perinatal and postpartum period. Beck, Steer, and Brown (1996) published the second edition of the BDI (BDI-II). Similar to the BDI, the BDI-II was composed of 21 s elf-report items used to identify and measure the severity of depression in individuals a ged 13 years and older. Symptomatology was addressed in the following areas : Sadness, Past Failure, Self Dislike, Change in Sleeping Pattern, and Change in Appetite. The psychometric properties of the BDI-II were evaluated using 500 p sychiatric outpatients and 120 college students (Beck, Steer, & Brown, 1996). Though sensi tivity, specificity, positive
5 predictive value of the BDI-II for screening of dep ression was shown to be adequate with the sample populations, it was never assessed in th e postpartum population during the development of the scale. The Edinburgh Postnatal Depression Scale (Cox, Hol den & Sagovksy, 1987) was developed as a screening tool with a specific focus on PPD. This 10-item self-report instrument focuses on common depressive symptoms wi th responses in a Likert-scale format. The following questions represent symptomat ology addressed by the EPDS: Â“inability to laugh, inability to look forward to t hings with enjoyment, unnecessary blaming of oneself, anxious or worried feelings, sc ared or panicky feelings, feelings that things have been getting on top of me, sleep diffic ulties due to unhappiness, sad or miserable feelings, crying, and thoughts of self-ha rmÂ” (Cox et al., 1987). Factor analysis on the original scale led to the omission of three items after it was determined they formed a separate factor of irritability (Cox et al ., 1987). More current research indicated that irritability is a key component to PPD symptom atology, and should be included in a screening process (Beck & Indman, 2005; Miller et a l., 2006). Irritability can be directly related to PPD and the onset of obsessive thoughts and resulting compulsive behaviors distinct to the postpartum period. The EPDS has been criticized for its use of both p ositive and negative item stems (Beck & Gable, 2001b; Floyd & Widaman, 1995), as ev idence has revealed that subjectsÂ’ responses are influenced by the direction of the qu estion. Original validation of the EPDS was confounded by the fact that all participants in the nonrandom sample had already been identified as depressed (Cox et al., 1997). Va rious subsequent validation attempts put the sensitivity, specificity, and positive pred ictive value in the adequate range when
6 paired with a psychiatric interview to provide a cl inical diagnosis of depression (Hanna, Jarman, & Savage, 2004; Mancini, Carlson, & Albers, 2007). The EPDS does have the advantage of being a shorter screener, which may pr ove to be beneficial when time constraints pose an issue, but the PDSS also has a cost utility advantage of being an effective telephone screener (Mitchell, Mittelstaed t, & Schott-Baer, 2006). Additionally, physicians have shown a preference for the direct i nterview process (Sit & Wisner, 2006), which does not correspond with an abbreviate d screening measure. The qualitative approach to the development of the PDSS is what distinguishes it from the other primary screening scales. Though ade quate in validity, the EPDS maintained a focus on somatic symptoms, similar to the BDI-II. The PDSS broadened the focus to include measuring experiences of the new m other with an investigation into loss of control, loneliness, a sense of unrealness, irri tability, fear of Â“going crazy,Â” and obsessive thinking, all of which were discovered as common factors experienced by women during the postpartum period (Beck, 2002, 200 5). Emotional experiences of mothers helped to define the discordant response be tween what a mother was actually thinking and feeling versus societal expectations o f what a motherÂ’s experience should be. This disharmony in role transition was identifi ed as a continuous variable on the PDSS (Clemens, Driscoll, & Beck, 2004). This allows a practitioner to not only identify the presence of PPD, but also to understand the lev el of influence it may be having on the individual, family, and infant relationships.
7 PDSS psychometric properties The PDSS is a 35-item self-report Likert-scale-base d summative total screening tool that takes approximately five to ten minutes t o complete (Beck, 2000). Overall psychometric scale sensitivity, or identifying a tr ue positive, is rated at 91-94%; specificity, or identifying a true negative, at 7298%, and positive predictive value, or a positive outcome predicting true outcome, at 33-88% (Sharp & Lipsky, 2002), which are all at an adequate level. During initial developmen t, content validity was established through a panel of five content experts reviewing t he PDSS and a focus group composed of 15 graduate students in nursing with a specialty in obstetrics and psychiatry. Conceptual and operational definitions of seven dim ensions were evaluated on a scale of 1 (strongly disagree) to 5 (strongly agree), with a mean rating ranging from 4 to 5 for the expert panel and 3.73 to 5 for the graduate student panel. From there, a pilot version of the PDSS was developed and administered to 525 wome n (Beck & Gable, 2000). The typical participant was White, married, in her 20Â’s with some college education, primigravida 6 weeks postpartum, and no significant history of depression. The 35 items were equally divided into one of seven dimensions, including: Sleeping/Eating Disturbances; Anxiety/Insecurity; Emotional Labilit y; Cognitive Impairment; Loss of Self; Guilt/Shame; and Contemplating Harming Onesel f. Construct validity was developed through a confirma tory factor analysis. Even with the subjective basis for development, construc t validity was based on empirical data provided by respondents (Beck & Gable, 2000). Liter ature-based item-dimensions were assigned in the scale, and the analysis examined th e fit to the hypothesized model. Overall fit as well as individual item contribution was examined. Item response theory
8 was applied as an added examination of construct va lidity. The adequacy of definition for individual dimensions was analyzed, and the model f it data with regard to the fit of the 5point Likert response was examined (Beck & Gable, 2 000). Overall, pilot analysis of the PDSS supported conte nt validity, and item dimensions were empirically and critically examined (Beck & Gable, 2000). Through confirmatory factor analysis, existence of hypothes ized dimensions was empirically supported. Item response theory techniques further supported the construct validity, and analyses of the Likert 5-point response categories supported meaningful score interpretations. High levels of alpha internal cons istence reliability supported the accuracy of the assessments (Beck & Gable, 2000). Sensitivity, specificity, and positive predictive v alue of the PDSS were determined through subsequent validation analyses ( Beck & Gable, 2001a, 2001b). Using a sample of 150 women within the three month postpa rtum period, participants were asked to complete the BDI-II, the EPDS, and the PDS S in random order. Participants were then evaluated by a nurse psychotherapist util izing the Structured Clinical Interview for the DSM-IV (SCID) to assess whether criteria wa s met for major depression or depression not otherwise specified (minor depressio n). The PDSS sensitivity was at 0.94 and specificity at 0.98 for major depression diagno sis (Beck & Gable, 2001a, 2001b). Minor depression yielded sensitivity of 0.91 and sp ecificity of 0.72. A sub-analysis examined the responses given on the SCID, and found women with a major depression diagnosis had more profound descriptions of difficu lty with maternal role transitions, increased irritability with partners and friends, a nd feelings of Â“wanting to go to sleep and never wake upÂ” (Beck & Gable, 2000).
9 Maternal experience of postpartum depression The basis for development of the PDSS includes comm on emotional experiences of women during the postpartum period. Beck (1993) developed the theory of Â“Teetering on the EdgeÂ” with a focus on loss of control a woma n may experience. According to Beck (1993), this four-stage process includes encou ntering Â“terror,Â” Â“dying of self,Â” Â“struggling to survive,Â” and Â“regaining control.Â” T he initial stage often comes as a surprise to postpartum women, and is where obsessiv e thinking is initially encountered. The depressed mother forms an appraisal of self, an d her perception of her parenting capabilities comes into question. Mothers become ex hausted from anxiety and obsessive thinking, and move to the Â“Dying of SelfÂ” stage. Th is stage is where a mother moves into feelings of isolation, unrealness, and thoughts of self-destruction. Feelings of guilt are often present due to the perception of being a fail ure as a mother. The third stage is where women seek to emerge from these debilitating feelin gs, with a regain of control in the fourth stage. Detection of women in the first and s econd stage is of primary importance. The initial stage of Â“Teetering on the EdgeÂ” stage can be related to WinnicottÂ’s theory of Â“Primary Maternal PreoccupationÂ” (1954), where a mother begins to develop a heightened level of self-preoccupation just prior t o giving birth, and this carries over to the immediate weeks of the postnatal period. In thi s heightened state of preoccupation with herself and her baby, a mother appears to be i n a withdrawn state from all others, which allows the mother to appropriately adapt to h er infantÂ’s basic needs. When a woman is unable to attend to this sensitivity to th e exclusion of all else, she may eventually recognize this as a missed opportunity a nd actively seek ways to compensate for this. This compensatory response can often be s een in the form of Â“spoilingÂ” the child
10 with inappropriate levels of attention. Conversely, when a mother becomes overly preoccupied with her infant, the potential for an o bsessive-compulsive-like state exists (Maina et al., 1999). Maternal depression is known to diminish the capacity for primary prenatal preoccupation (Field, 1992; Goodman and Go ttlieb, 1999), with depressed mothers reporting lower levels of preoccupation wit h developing a meaningful bond or relationship with their child (Feldman, et al., 199 9). A focus on experiences and emotional responses spec ific to a postpartum woman is important, as these emotions can compound causin g a Â“downward spiralÂ” (Beck, 2002), and subsequent obsessive thinking, cognitive impairment, and ego-dystonic intrusive thoughts of harming oneself or oneÂ’s chil d. Shame and guilt have been noted as reliable forecast measures for cognitively distorte d thinking, resulting in feelings of being overwhelmed (Beck, 2002). A motherÂ’s sense of selfefficacy can be negatively affected (Horowitz, Damato, Duffy, & Solon, 2005; Teti & Gel fand, 1991), and this creates a cyclical response whereby depression decreases self -efficacy, cognitions of guilt and shame ensue, and feelings of depression are exacerb ated (Fowles, 1997). Beck (2001) found a moderate effect size for the negative relat ionship between PPD and self-esteem, and consequently included maternal self-efficacy as a factor in the development the Guilt/Shame subscale of the PDSS (Beck & Gable, 200 0). Even with anticipated stress, cultural expectations for new mothers can act as a contributor to depression. Society has defined the appropriate feelings and behaviors for new mothers (Diskin, Doress, Bell & Swenson, 1976; Ditzion & Wolf, 1978; Tetoni & High, 1980), identifying the maternal experience as providing ultimate fulfillment. Additionally, there is an expectation for immediate love and a natural intuition to know
11 how to mother. When a mother is not able to meet th ese societal expectations, feelings of inadequacy, guilt, and shame ensue (Kraus & Redman, 1986). Cognitive factors involved in postpartum depression Cognitions of guilt and shame are often paired wit h anxiety (Ambrowitz et al., 2010), and postpartum anxiety can often be paired w ith obsessive thinking (Fairbrother & Woody, 2008). When women are overcome with obsessiv e thoughts, this can often result in compulsive behaviors (Paris, Bolton, & Winberg, 2008). Subsequent thoughts of selfharm and/or harming oneÂ’s baby leads to more shame, which increases the risk of suicidality. Research by Paris et al. (2008) found that suicidal thinking in postpartum women was directly correlated with low maternal sel f-esteem, negative perceptions of the mother-infant relationship, and high levels of pare nting stress resulting from incongruence in perception of expectations versus r eality. Different than postpartum psychosis, in postpartum OCD a mother recognizes th at her thoughts are her own, and that following through with them would be a poor de cision (Beck, 2006). Obsessive thoughts lead to worries about a childÂ’s health (Kim et al., 2012), and this thought process is correlated with lower mater nal sensitivity and higher levels of intrusiveness (Mercer, 1985; Nystrom & Ohrling, 200 4). Research by Zimbaldi et al. (2009) demonstrated increased risk of OCD symptoms during the postpartum period, with a relationship to emotional and cognitive diss onance. Prevalent obsessions included aggression, presenting in the form of irritability, contamination, and resulting compulsions of washing/cleaning and checking on the baby. The above referenced studies all found OCD symptoms had significant impa ct on the mother-infant relationship, as well as an increase in self-harm a nd intrusive baby-harming thoughts.
12 Depression, anxiety, faulty cognitions, and OCD are all aspects of postpartum psychiatric disorders and have immediate and long-standing impa ct on the maternal experience. Emotional Availability The Emotional Availability Scale (EAS) is a resear ch-based means of identifying and understanding the quality of the caregiver/chil d dyadic relationship (Biringen, 2000; Biringen et al., 2014). With a focus on two-way hea lthy emotional relationship, communication through verbal and nonverbal means is observed, focusing on sensitivity and responsiveness to infant cues (Biringen, 2000, Biringen et al., 2010). Based on attachment theory, the emotional connection between mother and infant was first designated as Emotional Availability (EA) by Mahler Pine and Berman (1975), and included sensitivity on behalf of the mother toward her childÂ’s emotional needs. Emde (1980) further described emotional availability as including a focus on the supportive nature that a mother should express in an optimal r elationship regardless of infant temperament. EA is a comprehensive means of approac hing a parent and childÂ’s ability to interact on an emotionally productive level by eval uating the dimensions of maternal sensitivity, structuring, non-intrusiveness, hostil ity, child responsiveness, and child involvement, with each component differentiated fro m the other (Biringen, 2000). Emotional Availability Scales (EAS) rate each of th ese dimensions based on a series of relational observation points, but from the perspec tive of the adult (Biringen, 2008). Maternal Sensitivity refers to a motherÂ’s ability to pick up on infant cues, provide warmth and soothe distress, and to be responsive in a variety of situations through quality affective interactions with her child. Structuring/ Non-Intrusiveness refers to the appropriateness of directedness on the part of the mother, providing balanced stimulation,
13 and creating learning opportunities that allow auto nomous exploration. Hostility refers to both overt and covert hostility that can be express ed in language, impatience, or anger. Child Responsiveness refers to the infantÂ’s reactio n to parental bids for engagement, and the expression of pleasure that accompanies this re action. Child Involvement refers to extent to which a child seeks to engage the mother in play as well as other means of actively seeking engagement by the mother (Biringen Robinson, & Emde, 2000). Emotional Availability, PPD, and intrusiveness Postpartum mood disorders can negatively affect EA (Easterbrooks, Bureau, & Lyons-Ruth, 2012). PPD can cause a mother to miss o r misinterpret cues from her baby (Murray, 2009). Mothers may perceive normal reactio ns in social reciprocity, such as lack of eye contact by baby to a mother with flat a ffect, as an indication that the baby does not love them (Reck et al., 2004). Maternal se lf-efficacy plays a role in EA. Teti & Gefland (1991) found that mothers perceived their b aby to less difficult when they had higher self-efficacy, but diminished self-efficacy resulted in the perception that the infant was more difficult and interactions between mother and infant were not rewarding. Research by Ambrowitz et al. (2010) showed that mor e intrusive maternal interactions cause a distress response from infants and this pro vided feedback to the mother indicating that she was not competent due to her inability to soothe baby. Beck (2006) observed that depressed mothers were more apt to lift their baby when not given the cue from the infant to do so. Overall, a mother is likely to perceive t he negative affect of her baby as reflecting on her capabilities, leaving her feeling less adequate and increasing feelings of guilt (Coleman & Karraker, 1997; Teti & Gefland, 19 91).
14 There is a large body of research looking into th e impact of PPD on maternal sensitivity (Bornstein, Suwalsky, & Breakstone, 201 2; Chaudhuri, Easterbrooks, & Davis, 2009; Geert-Jan, Stams, Juffer, & van Ijzend oorn, 2002; Kaplan, Burgess, Sliter & Moreno, 2009; Lovejoy, Gracyzk, OÂ’Hare, & Neuman, 2 000), but research is more limited when it comes to PPD and structuring and no n-intrusiveness. Guilt and shame are a crucial aspect of postpartum depression and have been shown to be associated with anxiety (Kim et al., 2012), unwanted thoughts (Fair borther & Woody, 2008; Zimbaldi et al., 2009), and subsequent intrusive parenting acti ons (Beck & Driscoll, 2006; Larsen et al., 2006). The negative impact of intrusive parent ing has emotional, physiological, and health consequences (Biringen et al., 2014; Field, 1981; Lumeng et al., 2012). Children of mothers who demonstrate heightened levels of int rusiveness exhibit more disorganized attachment (Geert-Jan et al., 2002), which can lead to higher risk for externalizing behavior in later childhood (Fearon, Bakermans-Kran enburg, van Ijzendoorn, Lapsley, & Roisman, 2010) and less emotional competence (Volli ng, McElwain, Notaro, & Herrera, 2002). Ego-resiliency, or the ability to shift beha viors and adapt emotions appropriately, is diminished when an infant is subjected to intrus ive parenting in early childhood and this subsequently alters the proper development of effortful control (Taylor, Eisenberg, Spinrad, & Widman, 2013). Infants can have a physiological response to intru sive parenting. Field (1981) found an increase in tonic heart rates in infants w ho were subjected to overstimulation by their mothers. This acceleration in heart rate crea tes activation of the HPA-axis and subsequent release of cortisol, which in turn affec ts body tissue and suppresses the immune system (Kajantee & Raikkonen, 2010). Lumeng, et al. (2012) found intrusive
15 actions by mothers led to childhood adiposity, whic h increases the risk for the development of cardiovascular disease later in life Maternal intrusiveness can have long-standing effe cts on infant development, but this behavior may be a well-intentioned result of c ompensating for guilt and shame cognitions. When a mother perceives herself as inad equate, feelings of guilt may arise (Coleman & Karraker, 1997). When she experiences eg o-dystonic thoughts of harming her baby, she feels shame and tries to avoid or dis prove these thoughts (Kim et al., 2012). Even with obsessions of harm coming to her baby, a mother may experience feelings of guilt for not doing enough to protect her child (Zi mbaldi et al, 2009). A mother may become withdrawn, but more likely she is going to g o through the process of cognitive reappraisal to remedy the unwanted thoughts and per ceptions brought on by depression (Troy, Shallcross, & Mauss, 2013). Through a process of trying to control thoughts, a mother appraises her cognition and feelings, and when depression, guilt, or shame influence the thought process, dissonance occurs (Zimbaldi et al., 2009). The moth er works to Â“put matters right,Â” by developing an alternate thought and strategy to dis prove this thought (Larsen et al., 2006). To offset obsessional thinking, individuals will often use distraction techniques (Abramowitz, Whiteside, Kalsy, & Tolin, 2003; Amir, Cashman, & Foa, 1997). Larsen et al. (2006) found that 30% of mothers experiencing p ostpartum OCD symptoms employed intrusive behavioral techniques such as checking th e baby or interacting with the baby even when not given the cues to do so by the infant The mothers endorsed the need to challenge their thoughtsÂ’ validity. Additionally, c ognitive reappraisal was highly endorsed by mothers on the Mean Thought Control Que stionnaire (TCQ; Larsen et al,
16 2006). Larsen et al. (2006) also found a correlatio n between endorsement of Reappraisal on the TCQ and elevated scores on the Yale-Brown Ob sessive Compulsive Scale (YBOCS). These results indicate that mothers who are experiencing symptoms of postpartum OCD are more likely to respond and attem pt to disprove intrusive thoughts by compensating with what they perceive to be appropri ate parenting actions, particularly when guilt and shame cognitions are involved. Taken together, this research supports the need for a focus on the maternal experience and the implications that the emotional aspects specific to the postpartum woman can have on the self, family, and the motherinfant dyadic relationship. Given the nature of the PDSSÂ’s focus on the maternal experien ce, this scale provides an appropriate measure to appraise and inform the practitioner of the presence of PPD, the potential for disruption of optimal emotional availability, and t o target the most effective and efficient intervention. Purpose of the present study The purpose of this study was to provide further v alidation of the efficacy of the PDSS as a reliable screening tool for the detection of PPD. This information may provide identification for the start of a Â“downward spiralÂ” that can result in the onset of PPD. PPD has been shown to as disruptive to the mother i nfant relationship and can have longstanding cognitive consequences on proper infant gr owth and development. Early detection and intervention is known to help offset this potential, and a valid and reliable screening tool is the first step in this identifica tion process. A focus on maternal sensitivity is important, but overlooking other aspects of EA can have detrimental effects. Optimal levels of str ucturing/non-intrusiveness can be
17 influenced by depression, low self-efficacy, and co ncurrent feelings of guilt and shame. Elevated scores on the GLT subscale of the PDSS can identify the existence of guilt and shame in a motherÂ’s cognitive processing. By noting elevated scores on the GLT subscale, practitioners can be alerted to the poten tial for this often overlooked nonoptimal parenting style. The current research sough t to find a connection between these cognitions and parenting style by looking at elevat ed scores on the GLT scale on the PDSS as a predictor of non-optimal structuring/nonintrusiveness on the EA scales. Hypothesis and specific aims Using an existing database of women who completed the PDSS, BDI-II, SCID, and EAS scored videos, the current research sought to find the predictive value of PDSS with the following aims and accordant hypotheses: Aim 1: Test the psychometric properties of the PDSS through replication with a different sample. Hypothesis 1: It was hypothesized that the psychome tric properties of the PDSS obtained by Beck & Gable would be replicated with r egard to model fit, reliability, construct validity, sensitivity, specificity, and p redictive value of a PPD diagnosis. Specifically, it was predicted that confirmatory fa ctor analysis on item-level data would produce results identifying a good model fit for th e seven subscales using the TuckerLewis Index (TLI) and root mean square error of app roximation (RMSEA). Internal consistency reliabilities of the seven dimensions o f the PDSS would be consistent with Beck & GableÂ’s (2000) findings, in the neighborhood of: Anxiety/Insecurity, 0.80; Sleeping/Eating Disturbances, 0.85; Cognitive Impai rment, 0.86; Guilt/Shame, 0.86; Emotional Lability, 0.87; Contemplating Harming One self, 0.87; and Loss of Self, 0.91.
18 We predicted that ROC curve analysis would identify a PDSS total score to be 80 as a positive screen for major PPD. ROC curve analysis w ould also indicate that the PDSS has an adequate level of psychometric sensitivity, defined as true positive hits and specificity, defined as true negative hits (.080), representing a balanced and reliable screener for PPD. Aim 2: Evaluate the predictive quality of the PDSS as a PPD screen and compare its value to the BDI-II. Hypothesis 2: It was hypothesized that the PDSS wil l provide added predictive value when compared to the BDI-II as an effective screeni ng tool for PPD. Specifically, hierarchical regression should indica te that both the BDI-II and PDSS are effective screeners for a diagnosis of PPD, but the proportion of variance accounted for in SCID diagnosis should be greater for the PDSS than the BDI-II. Aim 3: Investigate the factor of Guilt/Shame of the maternal experience for negative impacts on the mother-infant dyadic relati onship. Hypothesis 3: Elevated the Guilt/Shame subscale of the PDSS will predict elevated scores on the Structuring/Non-Intrusiveness scale o f the EAS. Specifically, it was predicted that regression anal ysis should indicate that the Guilt & Shame subscale of the PDSS would predict non-optima l scores (below 4.5 or above 6) on the Structuring/Non-Intrusiveness scale of the EAS. Additionally, a relationship between categorical non-optimal structuring and GLT scores should be identified.
19 CHAPTER II METHOD Participants An existing dataset was used for the current study Participants were women recruited from Colorado Parent Magazine and Early Head Start centers. Of 245 participants with completed PDSS data, 238 particip ants completed demographic data, BDI-II total score, PDSS total score, and SCID resu lts, while 114 completed the PDSS and had EAS coded recordings available. A confirmat ory factor analysis (CFA) was conducted on all participants with completed PDSS d ata ( N =245). Further analyses were conducted on the subsample ( N =238) that contained PDSS, BDI-II measures and SCID outcomes. Table 1 presents demographic and PDSS and BDI-II d iagnostic information for the N =238 sample. MotherÂ’s ages ranged from 16 to 41 yea rs, with a mean age of 30.24 ( SD =5.34). Infant age ranged from 138 days to 459 days with a mean age of 348.67 days ( SD =51.61). One hundred sixty-one of the infants were identified as White (67.8%), 41 as Latino (17.3%), 22 as African American (9.3%), 1 0 as Asian (4.2%), and 3 infants were identified as Native American (1.3%). The moda l education level was 6.0 (earned bachelorÂ’s degree), with 5.0 = earned an associateÂ’ s degree and 4.0 = earned high school diploma). One hundred eighty mothers reported being currently married (75.6%) and 58 reported not being currently married (24.4%). Media n household income was in the $31,000 $40,000 range ( n =32, 13.4%), with reported household income ranging from $0-$6000 ( n =23, 9.7%) to over $50,000 annually ( n =109, 45.8%).
21 There were no differences as a function of PDSS or BDI-II category and depression diagnosis for infant age, maternal age, education a nd number of children. Those with lower household income (<$20,000 annually) (p < .05 ), minorities (p < .01), and those who were not married (p < .001) were more likely to be depressed. Women with lower household income (p < .01) and unmarried mothers (p < .001) more likely to be in Positive Screen category of the PDSS and BDI-II. Table 2 presents the demographic data for the subs et of participants ( n =114) that provided data for the EAS. MotherÂ’s ages ranged fro m 16 to 40 years, with a mean age of 30.70 ( SD =5.14). Infant age ranged from 180 days to 452 days with a mean age of 357.50 days ( SD =43.38). Eighty-five of the infants were identified as White (75.6%), 12 as Latino (10.1%), 10 as African American (10.84%), 4 as Asian (3.4%), and 3 infants were identified as Native American (2.5%). The moda l education level was 6.0 (earned bachelorÂ’s degree), with 5.0 = earned an associateÂ’ s degree and 7.0 = earned masterÂ’s degree). Ninety-eight mothers reported being curren tly married (84.9%) and 16 reported not being currently married (15.1%). Median househo ld income was in the $41,000 $50,000 range ( n =14, 12.1%), and modal household income was over $5 0,000 annually ( n =61, 53.4%). There were no differences as a functio n of PDSS or BDI-II category and depression diagnosis for infant age, maternal age, minority status, household income, education and number of children. Those who were no t married were more likely to be depressed in addition to being more likely to be in Positive Screen category of the PDSS and BDI-II (p < .001).
23 Procedure Participants in the dataset completed the 35-item PDSS, the BDI-II, and a SCID administered by a trained graduate student. Item an alysis of PDSS, subscale scores of PDSS, and overall scores of PDSS and BDI-II were an alyzed for specificity, sensitivity, and predictive quality of MDD as determined by SCID results. Blind raters viewing a 10-minute, free play, dyadi c interaction coded the Emotional Availability Scales (EAS). Coders had no prior knowledge of psychological assessment results or maternal diagnosis. Participa nts were given an appropriate score for each of the five scales. The Structuring/Non-Intrus iveness 9-pt. scale assessed the level of directed involvement seeking a demonstration of app ropriate measures to assure success without becoming overly involved and a balanced lev el of stimulation. Lower scores are indicative of low or non-existent structuring, an o ptimal score of 5 indicates appropriate structuring, and higher scores represent the presen ce of high levels of stimulation, too much structuring, and intrusive behaviors by the pa rent. Data analysis Data were analyzed using IBM SPSS Statistics versi on 21 (SPSS Inc., 2012) and Mplus version 7.1 (Muthn & Muthn, 2013). Descript ive statistics (e.g., means, standard deviations, frequency distributions) were calculate d separately to describe both samples (PDSS replication N=238 and PDSS Guilt/Shame & EAS Structuring/Intrusivene ss N= 114). Internal consistency of the PDSS was measured by c oefficient alpha to determine reliability for total score and content scales. Mod el fit for the PDSS was determined
24 through confirmatory factor analysis with standardi zed weights for five items assigned to each of the seven content scales. Construct validit y for the PDSS was determined through correlation with the BDI-II, which measures aspects of depression, and hierarchical regression analysis, to explain variance in diagnos tic classification. Sensitivity, Specificity, Positive Predictive Value and Negative Predictive Value for both the PDSS and BDI-II were determined. ROC curves were constru cted to determine the sensitivity, specificity, and predictive value of the PDSS and B DI-II over a range of cut-off scores using DSM-IV major depression criteria. The relationship between the Guilt & Shame (GLT) s ubscale of the PDSS and the Structuring/Non-Intrusiveness scale of EAS was meas ured through multiple regression with both the GLT and EAS variables remaining conti nuous. A one-way within subject ANOVA was also conducted, with the EAS Structuring/ Non-Intrusiveness scale categorized into 3 levels: 1 = 0 Â– 4.5 (Low in Stru cturing), 5 Â– 6 (Optimal), and 6.5 Â– 9 (High in Intrusiveness). The GLT subscale was inclu ded as a continuous variable for this analysis. Power Statistical power is the probability of detecting a significant effect when a true effect exists and operates on the function of sample size, effect size, and alpha level. In order to determine if the existing database provided the min imal number of subjects required to achieve the desired power for regression analyses, a power analysis was computed using suggested values with a medium effect size of 2 = .15, alpha of .05, and desired power of .95 (Cohen, 1988). Using these criteria, it was det ermined a sample size of N=89 was needed in order to adequately decrease the likeliho od of a Type II error.
25 Data cleaning and transformation The full dataset of 245 mothers for item level ana lyses of the PDSS was entered into SPSS. Assumptions for CFA were assessed, and d ata with missing values were coded Â“99999Â” in order to distinctly identify these items. The dataset for 238 mothers for all other analyses were also entered into SPSS, wit h missing values also coded as Â“99999Â”. Descriptive statistics including mean, med ian, mode, standard deviation, and frequency distribution charts were analyzed to iden tify any outliers or data entry errors. Normality and linearity were assessed to assure the re were no violations in the assumptions of regression. Both the total scores fo r the PDSS and BDI-II violated the assumption of normal distribution for skewness, and appropriate data transformation was performed to correct this violation. The goal of a transformation was to achieve a variable with a more normal distribution to maximize the ass ociation (r) and statistical power (Tabachnick & Fidell, 2013). After a logarithm tran sformation of the BDI-II total score and a square root transformation of the PDSS total score, data approximated a range of normal distribution. Threats to validity All efforts were made to assure that any threats to validity are minimal at most. Consideration was made for any inconsistencies with the current sample and the original sample and the impact this may have had, including working with a reduced predictor for hierarchical regression. Additionally, the original PDSS testing distinguished between Major and Minor depression, and the current data di d not account for this distinction. Finally, though adequate in empirical support, the current study used the 2nd Edition of
26 Emotional Availability Scales. Currently, EAS is on its 4th Edition. One factor to consider is that in the newest version of the EAS, Structuri ng/Non-Intrusiveness scale is no longer one scale, and is now two distinct scales. This lim its the opportunity to distinguish between structuring and non-intrusiveness, but thro ugh communication with the developer of the EAS, it was established that for t he purposes of the present study, working with the 2nd Edition provided adequate and valid results (Z. Bi ringen, personal communication, July 11, 2014).
27 CHAPTER III RESULTS Analyses were conducted to address the three Specif ic Aims and test the above hypotheses. Table 3 presents the PDSS, BDI-II and S CID diagnostic results or the main sample ( N = 238). Table 3. PDSS, BDI-II and SCID summary No Depression MDD FR MDD PR MDD Other Current Dx PDSS 58.77 74.78 89.95 114.70 82.29 N = 120 32 39 30 17 50.4% 13.4% 16.4% 12.6% 7.1% BDI-II 9.03 9.16 16.31 26.00 16.24 N = 120 32 39 30 17 50.4% 13.4% 16.4% 12.6% 7.1% PDSS Total mean = 74.76 BDI-II Total mean = 10.88 N = 238 Columns are defined based on PDSS and BDI-II mean s core MDD-FR = Full Remission; MDD-PR = Partial Remission Other Current Dx includes bipolar disorder, panic d isorder, GAD, and anxiety NOS Aim 1: PDSS psychometric properties replication The validity of a test can be defined as its accur ateness in measuring the psychological characteristics intended to be measur ed (Anastasi, 1988). Construct validity specifically addresses whether a test capt ures and quantifies a theoretical psychological characteristic. The seven dimensions of the PDSS are theoretically defined dimensions of PPD, and the overall measure posits t o capture interrelated aspects of different dimensions for defining PPD. Interscale c orrelations of these seven dimensions
28 are presented in Table 4. Most scales were correlat ed at the moderate to high level, indicating that content scales represent interrelat ed constructs of PPD. Individual content scales were more highly correlated with PDSS Total score than with another content scale, indicating that each content scale were more directly related to PPD symptoms than to any other individual symptom cluster measured. Table 4. PDSS interscale correlations Diagnostic Sample ( N = 254) SLP ANX ELB MNT LOS GLT SUI Sleeping/Eating Disturbance (SLP) Anxiety/Insecurity (ANX) .681 Emotional Lability (ELB) .655 .868 Mental Confusion (MNT) .649 .819 .846 Loss of Self (LOS) .642 .806 .826 .824 Guilt/Shame (GLT) .525 .729 .705 .694 .767 Suicidal Thoughts (SUI) .363 .540 .527 .539 .608 .649 PDSS Total Score (TOT) .777 .917 .924 .910 .915 .823 .636 Confirmatory factor analysis was also used to eval uate construct validity and allow researchers to estimate the fit of actual dat a to hypothesized dimensions (Joreskog, 1969). Table 5 presents the results of the CFA of t he PDSS. The table includes standardized factor loadings for the five items ass igned to each of the seven content scales. Results are nearly identical to those of Be ck & Gable (2000), who reported ranges of .62 to .78 on the Sleep/Eat Disturbance (SLP) sc ale, .66 to .73 on the Anxiety/Insecurity (ANX) scale, .74 to .84 on the E motional Lability (ELB) scale, .79 to .85 on the Mental Confusion (MNT) scale, .85 to .90 on the Loss of Self (LOS) scale, .72
29 to .87 on the Guilt/Shame (GLT) scale, and .75 to 92 on the Suicidal Thoughts (SUI) scale. Each of the weights had a minimum t score of 12.79, indicating that all of the items fit the model as predicted. Goodness-of-fit i ndices were also calculated. The Tucker-Lewis index (TLI) of 0.84 indicated each wei ght was sufficient to support subscale fit. RMSEA of .072 was larger than the com monly recommended cutoff score of .06, but still provided moderate support for model fit between observed and model data (Hu and Bentler, 1999). All factor loadings were si gnificant ( p < .001). This compares with Beck & Gable (2000), who reported a TLI of 0.8 7, RMSEA of .05, and t-values with a minimum of 14.79. However, the results of the cu rrent CFA were Â“non-positive definite.Â” This indicates the possibility of a high er order latent variable, and although valid, results should be viewed with caution.
30 Table 5. CFA: Maximum-likelihood dimensions and loa dings (N=245) Item I II III IV V VI VII Sleep/Eat Disturbance (SLP) 1. I had trouble sleeping even when my baby was asl eep. .74 8. I lost my appetite. .65 15. I woke up on my own in the middle of the night and had trouble getting back to sleep. .83 22. I tossed and turned for a long time at night tr ying to fall asleep. .83 29. I knew I should eat but I could not. .61 Anxiety/Insecurity (ANX) 2. I got anxious over even the littlest things that concerned my baby. .61 9. I felt really overwhelmed. .73 16. I felt like I was jumping out of my skin. .71 23. I felt all alone. .74 30. I felt like I had to keep moving or pacing. .66 Emotional Lability (ELB) 3. I felt like my emotions were on a roller coaster .80 10. I was scared that I would never be happy again. .77 17. I cried a lot for no real reason. .72 24. I have been very irritable. .75 31. I felt full of anger ready to explode. .70 Mental Confusion (MNT) 4. I felt like I was losing my mind. .84 11. I could not concentrate on anything. .77 18. I thought I was going crazy. .81 25. I had a difficult time making even a simple dec ision. .73 32. I had difficulty focusing on a task. .75 Loss of Self (LOS) 5. I was afraid that I would never be my normal sel f again. .82 12. I felt as thought I had become a stranger to my self. .84 19. I did not know who I was anymore. .83 26. I felt like I was not normal. .84 33. I did not feel real. .77 Guilt/Shame (GLT) 6. I felt like I was not the mother I wanted to be. .81 13. I felt like so many mothers were better than me .81 20. I felt guilty because I could not feel as much love for my baby as I should. .61 27. I felt like I had to hide what I was thinking o r feeling toward the baby. .58 34. I felt like a failure as a mother. .87 Suicidal Thoughts (SUI) 7. I have thought that death seemed like the only w ay out of this living nightmare. .95 14. I started thinking that I would be better off d ead. .93 21. I wanted to hurt myself. .77 28. I felt that my baby would be better off without me. .64 35. I just wanted to leave this world. .93
31 To test for reliability, internal consistency esti mates were calculated using CronbachÂ’s coefficient alpha (Cronbach, 1988). Reli ability refers to a testÂ’s ability to measure what it purports to measure, and internal c onsistency reports the intercorrelations among items in a scale. Nunnally & Bernstein (1994) define a minimum coefficient alpha of .70 for measures defining emotional constructs. Table 6 presents internal consistency estimates for PDSS Total score and seven symptom co ntent scales. PDSS Total score yielded an alpha coefficient of .95 and coefficient s ranged from .68 to .93 for content scales. At .68, the SUI scale fell below the .70 re commended cutoff, indicating that this subscale may not be as reliable at effectively meas uring suicidal thoughts as possible. These results are consistent with Beck & Gable (200 0) who reported a PDSS Total score alpha coefficient of .96, and content scales rangin g from .80 (ANX) to .91 (LOS). Concurrent validity refers to a testÂ’s ability to predict an outcome given a criterion measured. Using the seven PDSS dimensions, a discr iminant function analysis was performed to determine the accurate predictive valu e of the PDSS for group membership to either depressed or non-depressed groups. As det ermined by SCID results, 175 women Table 6. Internal consistency estimates for 35-Item PDSS Coefficient Alpha Sleeping/Eating Disturbance (SLP) .78 Anxiety/Insecurity (ANX) .92 Emotional Lability (ELB) .93 Mental Confusion (MNT) .91 Loss of Self (LOS) .91 Guilt/Shame (GLT) .84 Suicidal Thoughts (SUI) .68 PDSS Total Score .95
32 were classified as non-depressed and 36 women were classified as depressed The discriminant function calculation was significant a t the p < .01 level ( X2 = 100.36), and WilksÂ’ Lambda = .653, which indicates the discrimin ant function accounted for 55.3% of the variability between the depressed and non-depre ssed groups. The classification of groups for the PDSS and BDIII is presented in Table 7. Of the total sample of 238 women, 168 mothers were cor rectly classified in the nondepressed group and 29 of the 36 mothers were corre ctly classified in the depressed group for the PDSS. For the BDI-II, 179 mothers wer e correctly classified in the nondepressed group, and for the depressed group, 27 of the 36 mothers were correctly classified. Table 7. Discriminant function classification, sens itivity and specificity of PDSS and BDI-II for MDD detection PDSS BDI-II SCID Diagnosis SCID Diagnosis Depressed Not Depressed Total Depressed Not Depressed Total Pos. Screen 29 (True +) 34 (True -) 63 Pos. Screen 27 (True +) 23 (True -) 50 Neg. Screen 7 (False -) 168 (True -) 175 Neg. Screen 9 (False -) 179 (True -) 188 Total 36 202 238 Total 36 202 238 Sensitivity = 29/38 = 76% Sensitivity = 27/36 = 75% Specificity = 168/202 = 83% Positive PV = 29/63 = 46% Negative PV = 168/175 = 96% Specificity = 179/202 = 89% Positive PV = 27/50 = 54% Negative PV = 179/189 = 95% (N = 238 ) Positive screen for PDSS is considered Total score 80 Positive screen for BDI-II is considered Total scor e 14 Sensitivity, specificity, and predictive value ref er to the ability of an instrument to correctly identify women who are depressed (i.e., t rue-positive rate) or not depressed (i.e., true-negative rate), and to differentiate be tween positive predictive value (i.e., screened positive when positive) and negative predi ctive value (i.e., screened positive PDSS Outcome BDI-II Outcome
33 when negative). The PDSS overall Sensitivity for MD D diagnosis was determined to be .76 and Specificity was .83. These are lower than t he original findings by Beck & Gable (2001a, 2001b), who reported Sensitivity at .94 and Specificity at .98 for MDD diagnosis. Table 7 presents a comparative evaluation of Sensit ivity and Specificity for the PDSS and BDI-II for MDD diagnosis. The PDSS and BDI-II produ ced comparable results, with the PDSS being marginally higher in Sensitivity and Neg ative Predictive Value and the BDIII being marginally higher in Specificity and Posit ive Predictive Value. Table 8 shows the seven dimensions of the PDSS in relation to the discriminant function of each as a predictor between depressed a nd non-depressed mothers. According to Tabachnick & Fidell (2013), correlations above 0 .33 are considered interpretable. The correlation values suggested that all of these scal es qualified as interpretable predictors, with Emotional Lability (ELB) being the strongest p redictor of group membership, followed closely by Guilt and Shame (GLT). Loadings greater than 0.71 are considered excellent, 0.63 is very good, 0.55 is good, 0.45 is fair, and 0.33 and below are poor (Comrey & Lee, 1992). All dimensions were in the ve ry good to excellent range. Table 8. Correlations between PDSS symptom content scales an d Canonical discriminant function PDSS Content Scale Correlation With First Discriminant Function Sleeping/Eating Disturbances (SLP) .658 vg Anxiety/Insecurity (ANX) .725 e Emotional Lability (ELB) .816 e Mental Confusion (MNT) .810 e Loss of Self (LOS) .751 e Guilt/Shame (GLT) .814 e Suicidal Thoughts (SUI) .707 vg ( N = 238)
34 Receiver Operating Characteristic (ROC) is a metho d used to determine the optimum cut-off point, appropriately weighing sensi tivity, specificity, and predictive value (Jekel, Elmore, & Katz, 1996). A ROC curve w as constructed to determine the optimal cutoff point and the overall predictive val ue of the PDSS as determined by the area under the curve. Figure 1 presents the ROC cur ve for the PDSS. The curve for Figure 1 represents the PDSSÂ’ performance in screen ing 238 mothers for major depression. Area under the curve is considered exce llent at .90 Â– 1, good at .80 .90, and fair at .70 .80. The PDSSÂ’ area under the curve w as .91, putting this in the excellent range for balance between sensitivity and specifici ty, with a significance value of p < .001. The optimal cutoff score was determined to be 80 for major depression diagnosis, consistent with Beck & Gable, with Sensitivity = .8 9 and Specificity = .71 when using this score. FIGURE 1 ROC curve for PDSS: Major depression. Area = 0.91 2 ( SD = .02)
35 ROC curve was also developed to determine the optim al cutoff score and overall predictive value for the BDI-II based on MDD diagno sis. Figure 2 presents the ROC curve for the BDI-II. The area under the curve for the BDI-II was .93 (p<.001), which is in the excellent for specificity and sensitivity. T he optimal cutoff score was determined to be 14.5 with sensitivity and specificity for this s core of .94 and .73 respectively. FIGURE 2 ROC curve for BDI-II: Major depression. Area = 0. 931 ( SD = .02) Additionally, a ROC curve was constructed to provi de a visual comparative analysis of the sensitivity and specificity of the PDSS against the BDI-II. Figure 3 presents the results. Consistent with results prese nted in Table 7, the BDI-II outperforms the PDSS with regard to specificity, but also has m ore sensitivity when the cutoff score is set to 14.5. If the BDI-II cutoff score was lowered to 14, sensitivity would remain at .94, but the specificity would be reduced to .69.
36 FIGURE 3 ROC curve for PDSS and BDI-II at recommended cuto ff scores Aim 2: Predictive quality of the PDSS as compared t o BDI-II Correlation and multiple regression analyses were conducted to examine the relationship between SCID diagnosis of PPD and elev ated PDSS and BDI-II scores as potential predictors. Table 9 summarizes the descri ptive statistics and analysis results. To examine the construct validity of the PDSS, correla tional data were analyzed with the BDI-II, a known valid and reliable screener. Correl ation results are listed in the left-side triangle of Table 9. The PDSS was strongly correlat ed with the BDI-II ( r = .675 p < .01), evidencing that the two instruments measure similar aspects of depression. In order to determine the ability of the PDSS to e xplain variance of diagnostic classification of PPD that provides added value ove r the BDI-II, incremental validity of the PDSS was assessed through hierarchical regressi on analysis. A new screening
37 instrument should be judged on the basis of the con tribution over and above the bestknown existing instrument (Cronbach & Gleser, 1957) The BDI-II variable was entered first into the regression equation, followed by the PDSS. The right side of Table 9 contains the regression results. The BDI-II account ed for 22.5% of the variance explained in SCID depression diagnosis. The PDSS increased th e variance of depression diagnosis by 10.0% ( p = .001). This increase in prediction of group classif ication supports the construct validity of the PDSS. Table 9. Hierarchical regression of the diagnosis of postpar tum depression on the BDI-II and PDSS. Variables PPD Diagnosis BDI-II PDSS R R 2 Adjusted R2 eta BDI-II .601* .475 .225* .222 .167** PDSS .599* .675* .570 .325 .319 .441*** Means -.69 12.86 75.03 Standard deviation .729 10.86 29.54 *p = <.01 **p = <.02 ***p = <.001 When the PDSS was entered first, followed by the BD I-II, the regression model produced R2 = .325, F (2, 235) = 56.63, p =.027, meaning the BDI-II added 1.4% predictive value when holding the PDSS constant (r2 change = .014). Table 10 presents the results when the PDSS is entered first, followe d by the BDI-II. Overall, both the PDSS and BDI-II provide significant and unique pred ictive value toward depression diagnosis.
38 Table 10. Hierarchical regression of the diagnosis of postpar tum depression on the PDSS and BDI-II. Variables PPD Diagnosis BDI-II PDSS R R 2 Adjusted R2 eta BDI-II .601* .558 .311* .308 .441*** PDSS .599* .675* .570 .325** .319 .167** Means -.69 12.86 75.03 Standard deviation .729 10.86 29.54 *p = <.01 **p = <.05 ***p = <.001 Aim 3: Guilt/Shame subscale predicting intrusivenes s on EAS A regression analysis was conducted to test the th ird aim looking at the five-item response total of the PDSSÂ’ Guilt/Shame (GLT) subsc ale as a predictor of an elevated score on the Structuring/Non-Intrusive scale of the EAS. This sample included 114 women who had completed the PDSS and had EAS coding results. The regression model with GLT as a predictor produced R2 = .024, F (1, 112) = 2.72, p = .102, ns, meaning that GLT did not contribute significant predictive value Further analysis was conducted to explore the rela tionship between elevated scores on the GLT subscale and non-optimal EAS scor es. Analysis of variance (ANOVA) was conducted to compare the effect of cate gorical structuring/nonintrusiveness level on GLT scores, with low structu ring defined as (< 4.5), optimal structuring as (4.5 Â– 6), and high structuring (Int rusiveness) as (> 6). There was a significant effect of EAS Structuring level on GLT scores, F (2, 111) = 3.11, p = .05, 2 = .053. These results suggest that EAS levels do have an effect on GLT scores. Specifically, when mothers were more intrusive in p arenting style they are more likely to
39 be experiencing more guilt and shame symptoms as de termined by the PDSS. Table 11 presents the information regarding GLT scores and E AS category. Table 11. EAS category and GLT scores EAS Category No. in Category GLT Mean Score 1 Low in Structuring 12 6.67 ( SD = 2.39) 2 Optimal Style 65 9.43 ( SD = 3.84) 3 High in Intrusiveness 37 10.19 ( SD = 5.27) Total 114 9.38 ( SD = 4.33)
40 CHAPTER IV DISCUSSION The current study proposed and tested the psychome tric properties of the PDSS as a valid and reliable screener for detection of PPD in women. These factors were investigated within the items and scales of the PDS S, as well as in comparison with the BDI-II. Additionally, further predictive value of t he PDSS was investigated to ascertain if the dimension of PDSS guilt and shame correlated wi th EAS-rated intrusive parenting style. With regard to psychometric properties of the PDSS the current study looked at construct validity, reliability, concurrent validit y, sensitivity, specificity, and predictive validity. Our findings revealed that the construct validity of the PDSS for our sample was good as indicated by interscale correlations and co nfirmatory factor analysis. Adequate reliability was established through internal consis tency estimates. Concurrent validity was established through discriminant function class ification and Canonical discriminant function. In comparison to the BDI-II, results indi cate the PDSS had adequate levels of sensitivity, specificity, positive predictive value and negative predictive value, though slightly less robust than the BDI-II. ROC Curve ana lyses were performed to determine the optimal cutoff score for the PDSS and BDI-II fo r MDD diagnosis, with both performing at the excellent level. The predictive p ower of the PDSS and BDI-II for MDD diagnosis was further tested through regression ana lysis, and it was established that both account for a significant level of variance for MDD diagnosis. The predictive power for maternal EA, specifically the hypothesized guilt an d shame effect, was not demonstrated
41 through regression analysis, but subsequent ANOVA d id support a linear relationship between categorical ratings of the EAS Structuring/ Non-Intrusiveness scale and elevated GLT scores. The need for an effective screener for PPD that ta kes the Â“maternal experienceÂ” into account has been addressed in the literature a nd prompted the development of the PDSS. The PDSS contributes to screening for PPD in ways similar to the EPDS and clarifies symptomatology addressed by the BDI-II, m aking it potentially more appropriate for the postpartum population. The PDSS also has the added contribution of addressing experiences a mother may have that contr adict the societal expectation for a new mothersÂ’ emotional and relationship experience. The current study began with a pre-existing sample of women from the Denver, Colorado area. When comparing demographic informati on to the original sample used by Beck & Gable (2000), there are some notable differe nces. The original development sample included women with a similar mean age (31.2 years vs. 30.2 years for the current sample). Mean level of education was also similar w ith most participants completing 4 years of college, but the current sample had 31% of participants with a high school diploma as their highest level of education attainm ent, compared to 8% for the original sample. The current sample ( N =238) had differences with regard to infant age, in fant ethnicity, and maternal marital status. Specificall y, the average infant age at time of testing was 5.63 weeks for the Beck & GableÂ’s devel opment sample, compared to 348.67 days (49.81 weeks) for the current sample. The curr ent sample had an ethnic makeup that included 67.8% identifying as white whereas the dev elopment sample reported 87% white. The current sample also had 17% Hispanic and 9% African American participants
42 compared to 3% Hispanic and 8% African Americans re spectively. Twenty-four percent of the current studyÂ’s participants reported not be ing married compared to 11% in the development study. The original findings did not re port annual household income, so no comparison could be made. However, this is an impor tant aspect, as the current research found that lower annual household income was relate d to higher levels of MDD diagnosis and higher scores on the PDSS. Overall, the demogra phic makeup of the current study differs from the development sample, making the cur rent research more generalizable to a diverse population. Aim 1 of the current study proposed and tested the construct validity, concurrent validity, discriminant function, specificity, sensi tivity, and predictive value of the PDSS. The current research found results to be overall co nsistent with the original findings. A CFA was conducted for the PDSS using 35-item re sponses by 245 mothers. Findings from the CFA support good model fit, and t he identification of the seven dimensions of: Sleeping/Eating Disturbances; Anxiet y/Insecurity; Emotional Lability; Mental Confusion; Loss of Self; Guilt/Shame; and Su icidal Thoughts. However, it was noted that the possibility of a higher order latent variable exists due to a Â“non-positive definiteÂ” matrix. This most likely means that there are several factors that are essentially measuring the same thing. The PDSS, like most psych ological screening measures, is based on a pre-determined theory of emotional state (i.e., depression). It is presumed that aspects of depression are interrelated (e.g., loss of sleep can cause mental confusion). Therefore, it is not unexpected that factors on mos t subscales would be related and would have some dependency on each other. It is not known if the original developers encountered such an error, as this is not always re ported. This does not discount the
43 PDSS as a valuable screener; rather, it opens the p otential for future research to gain a more in-depth understanding of the intricacies of t he maternal experience and the factors that drive PPD and other mood an anxiety disorders in the perinatal and postpartum period. In most cases, the analyses conducted with the pr esent sample were able to replicate the initial development findings. One are a where there were differences was with regard to sensitivity and specificity. Althoug h results put overall sensitivity and specificity in the excellent range, when taking cut off scores into account, the PDSS missed almost 30% of true depression cases. One not able observation from the findings includes the breakdown of levels of depression and other psychiatric disorders during the postpartum period as detailed on Table 3. Although the current study identified 15.2% of the population as having MDD during the postpartum period, which is consistent with reported prevalence of PPD in the United States (Be ck, 2002; Freed, Chan, Boger, & Tompson, 2012; Sit & Wisner, 2009), the other facto rs such as minor depression, major depression in partial remission, and other common m ood and anxiety disorders were not addressed here. When taking partial remission or ot her psychiatric conditions into account, the PDSS and BDI-II may be able to identif y women in need of intervention at a much higher rate. One possible reasons for the relatively lower rat es of psychometric sensitivity and specificity here is that these women were gener ally 12 months into the postpartum period and depressed mothers may have been experien cing symptoms for an extended period of time. What may be true reportable symptom s of depression may be perceived by mothers as the norm of motherhood, and their dis tress may be expected and hence
44 downplayed in an interview style format. The curren t studyÂ’s sample had a significant proportion of women in the Â“partial remissionÂ” cate gory, meaning they were possibly still experiencing identifiable symptoms but not at the l evel or duration that would warrant a MDD diagnosis, which would reduce psychometric sens itivity for the identification of a mood disturbance. Mothers in Beck & GableÂ’s (2000) developmental sample were all within the 12-week postpartum period, and may not h ave had a chance to remit or partially remit, thereby increasing the numbers of identifiable SCID diagnoses for MDD. Another consideration is that SCID diagnosis for MD D is not specific enough for the postpartum woman. Future research would benefit fro m taking a broader perspective that is more inclusive of other psychiatric disorders du ring the perinatal and postpartum period and determinations made as to how valid and reliable the current screeners are at detecting such disorders. Aim 2 of the current study proposed correlational and hierarchal regression analysis to identify the PDSS as a valid and reliab le screener based on the standard in the field, the BDI-II. Results support that both are va lid and reliable screeners, but the PDSS accounted for a greater proportion of variance depr ession diagnosis. However, there was minimal discrepancy between R2 and adjusted R2, suggesting that each contributed unique variance to PPD diagnosis. These differences may be due to nature of the population for which they were developed (i.e., gen eral depression vs. postpartum depression). Having a more symptom-based focus, the BDI-II may be a more appropriate screener when trying to assess somatic distress. Al though the BDI-II does measure certain cognitive aspects of depression, it is not specific to the maternal experience and
45 may not capture symptoms that are causing distress that are unique to this subset of depressed individuals. When seeking to identify ways in which a mother ma y be experiencing distress in various biological, psychological, and social aspec ts, the PDSS may be a more appropriate choice the BDI-II or EPDS. One area add ressed by the PDSS that is not considered on the BDI-II or the EPDS is the Â“Loss o f Self.Â” This subscale on the PDSS asks about feelings of separation from oneÂ’s person al identity, which can elicit feelings of worry about feeling Â“normalÂ” again. Overall, the PD SS was developed to be a more comprehensive tool for the conceptualization of a m otherÂ’s postnatal experience, and should be considered for use when trying to identif y the full scope of distress a mother may be experiencing. Aim 3 of the study proposed that the GLT subscale of the PDSS would be a positive predictor of intrusive parenting style as indicated by elevated scores on the EAS Structuring/Non-Intrusive subscale. Although analys es examining EAS-based structuring/non-intrusiveness as a continuous measu re did not yield significant results, subsequent analyses using a categorical measure of this construct were able to show that mothers classified as overly structuring/intrusive in parenting style also had elevated selfreports of guilt and shame on the PDSS. This is con sistent with the hypothesis reviewed above that guilt and shame about inappropriate nega tive feelings about the infant and low parenting competence may stimulate mothers into com pensatory intrusive behaviors. Future research is needed to go beyond the correlat ional data presented here, to further investigate this causal hypothesis.
46 Particularly important would be to use the current version of the EAS, which not only separates the Structuring and Non-Intrusive su bscales, but also provides a clearer and more distinct definition of what intrusive pare nting looks like in the dyadic interaction. This area is important to investigate, as known interventions are available to help correct these non-optimal parenting styles. When taken together, Aims 1, 2, and 3 of the curre nt study provide important contributions toward identifying and understanding the mechanisms involved in PPD. With early identification as the objective to help offset the myriad of potential negative effects of PPD, the PDSS should be considered as th e screener that will provide this detection and will also help to conceptualize and f ormulate a proper treatment plan. Women who are endorsing high levels on any or all s cales of the PDSS should be identified as potentially needing intervention, not just if depression is indicated by a positive screen. As we are now learning that elevat ed guilt and shame scores carry concern for non-optimal parenting style, thereby cr eating more potential for future developmental issues for a child, it is important t o view the perinatal and postpartum period through the lens of the maternal experience. The current research also provides valuable considerations for further investigation, including investigating the potential for an overarching mechanism in a negative maternal exp erience. Biological considerations of PPD should also be considered, as these have bee n shown to play a crucial role for the development of postpartum psychiatric disorders and may negatively influence the maternal-infant relationship. As science and societ y are moving toward a fuller understanding of the broad psychiatric reactions bo th during and after pregnancy, it is
47 important to have a screener that is more encompass ing of this dynamic to help support with early intervention.
48 REFERENCES Abramowitz, J.S., Meltzer-Broday, S., Leserman, J., Killenberg, S., Rinaldi, K., Mahaffey, B.L., & Pedersen, C. (2010). Obsessional thoughts and compulsive behaviors in a sample of women with postpartum moo d symptoms. Archive of WomenÂ’s Mental Health, 13, 523-530. Abramowitz, J. S, Whiteside, S.P. Kalsy, S.A. & Tol in, D.F. (2003). Thought control strategies in obsessive-compulsive disorder: A rep lication and extension. Behavior Research and Therapy, 14, 529-540. Amir, N., Cashman, L., & Foa, E.B. (1997). Strategi es of thought control in obsessivecompulsive disorder. Behavior Research and Therapy, 35, 775-777. Anastasi, A. (1988). Psychological testing New York: Macmillan. Arnold, L. M. (1999). A case series of women with p ostpartum-onset obsessivecompulsive disorder. Primary Care Companion Journal of Clinical Psychiat ry, 1 (4), 103-108. Beck, C. T., (1993). Teetering on the edge: A subst antive theory of postpartum depression. Nursing Research, 42, 42-48. Beck, A. T., Steer, R., & Brown, G. (1996). BDI-II manual. San Antonio, TX: Psychological Corporation. Beck, C. T. (1998). The effects of postpartum depre ssion on child development: a metaanalysis. Archives of Psychiatric Nursing 12 (1), 12Â–20. Beck, C. T., & Gable, R. K. (2000). Postpartum depr ession screening scale: Development and psychometric testing. Nursing Research 49 (5), 272-282.
49 Beck, C.T. (2001). Predictors of Postpartum Depress ion. Nursing Research 50 (5), 275Â– 285. Beck, C. T., & Gable, R.K. (2001a). Further valida tion of the postpartum depression screening scale. Nursing Research 50 (7),155-164. Beck, C.T., & Gable, R.K. (2001b). Comparative anal ysis of the performance of the Postpartum Depression Screening Scale with two othe r depression instruments. Nursing Research 50 (5), 242-250. Beck, C. T. (2002). Postpartum Depression: A Metasy nthesis. Qualitative Health Research 12 (4), 453Â–472. Beck, C. T., & Indman, P. (2005). The many faces of postpartum depression. Journal of Obstectric, Gynecologic, and Neonatal Nursing, 34 (5), 569Â–76. Beck, C. T. (2006). Postpartum depression: It isnÂ’t just the blues. Associated Journal of Nursing, 106 (5), 40-50. Beck, C.T., & Driscol, J.W. (2006). Postpartum mood and anxiety disorders: A clinicianÂ’s guide. Sandbury, MA: Jones and Bartlett Publishers. Biringen, Z., & Robinson, J. (1991). Emotional avai lability in mother-child interactions: A reconceptualization for research. American Journal of Orthopsychiatry, 61 (2), 258-271. Biringen, Z. (2000). Emotional availability: Concep tualization and research findings. American Journal of Orthopsychiatry, 70 (1), 104-114. Biringen, Z., Robinson, J., & Emde, R. (2000). Appe ndix B: The Emotional Availability Scales. Attachment & Human Development (3rd Ed.); 256-270.
50 Biringen, Z. (2008). The emotional availability (EA) scales, 4th edition Boulder, CO: Emotional Availability.com. Biringen, Z. & Easterbrooks, M.A. (2012). Emotional availability: Concept, research, and window on developmental psychopathology. Development and Psychopathology, 24, 1-8. Biringen, Z., Altenhofen, S., Aberle, J., Baker, M. Brosal, A., Bennett, S, Â… Swaim, R. (2012). Emotional availability, attachment, and int ervention in center-based child care for infants and toddlers. Development and Psychopathology, 24, 23-34. Biringen, Z., Derscheid, D., Vliegen, N., Closson, L., Easterbrooks, M.A. (2014). Emotional availability (EA): Theoretical background empirical research using the EA scales, and clinical applications. Developmental Review, 34, 114-167. Bornstein, M. H., Suwalsky, J. T., & Breakstone, D. A. (2012). Emotional relationships between mothers and infants: Knowns, unknowns, and unknown unknowns. Developmental and Psychopathology, 24, 113-123. Carter, S., Garrity-Rokous, F. E., Chazan-Cohen, R. Little, C., & Briggs-Gowan, M. J. (2001). Maternal depression and comorbidity: predic ting early parenting, attachment security, and toddler social-emotional problems and competencies. Journal of the American Academy of Child and Adolescent Psychiatry 40 (1), 18Â–26. Chaudhuri, J.H., Easterbrooks, M.A., & Davis, C.R. (2009). The relation between emotional availability and parenting style: Cultura l and economic factors in a diverse sample of young mothers. Parenting Science and Practice, 9, 277-299. Coleman, P.K. & Karraker, K.H. (1997). Self-efficac y and parenting quality: Findings and future applications. Developmental Review, 18, 47-85.
51 Comrey, A. L., & Lee, H. B. (1992). A first course in factor analysis Hillsdale, NJ: Erlbaum. Cox, J. L., Holden, J. M., & Sagovsky, R. (1987). D etection of postnatal depression. Development of the 10-item Endinburgh postnatal de pression scale. British Journal of Psychiatry, 150, 782-786. Cronbach L. J. Gleser G. C. (1957). Psychological tests and personnel decisions Urbana: University of Illinois Press, 1957. Cummings, E.M., & Davies, P.T. (1994). Maternal dep ression and child development. Journal of Child Psychology & Psychiatry, 35, 73-112. Dennis, C.L. and Chung-Lee, L. (2006), Postpartum d epression help-seeking barriers and maternal treatment preferences: A qualitative syste matic review. Birth, 33 : 323Â–331. Easterbrooks, M.A., Bureau, J.F., & Lyons-Ruth, K. (2012). Developmental correlates and predictors of emotional availability in motherchild interaction: A longitudinal study from infancy to middle childhood. Development and Psychopathology, 24, 65-78. Emde, R. N. (1980). Emotional availability: A recip rocal reward system for infants and parents with implications for prevention of psychos ocial disorders. In P. Taylor (Ed.), Parent-infant relationships (pp. 87-115). Orlando, FL: Grune & Stratton. Evins, G. G., Theofrastous, J. P., & Galvin, S. L. (2000). Postpartum depression: A comparison of screening and routine clinical evalua tion. American Journal of Obstetrics and Gynecology 182 (5), 1080Â–1082. Fairbrother, N., & Woody, S. R. (2008). New mothers Â’ thoughts of harm related to the newborn. Archive of WomenÂ’s Mental Health, 11, 221-229.
52 Fearon, R.P., Lapsley, A., Bakersmans-Kraneburg, M. J., Ijzendoorn, M. H., & Roisman, G. (2010). The significance of insecure attachment and disorganization in the development of childrenÂ’s externalizing behavior: A meta-analytic study. Child Development, 81 (2), 435-456. Field, T. (1992). Infants of depressed mothers. Developmental Psychology, 4, 49-66. Field, T. (1998). Maternal depression effects on in fants and early interventions. Preventative Medicine, 27 200-203. Field, T. (1981). Infant gaze aversion and heart ra te during face-to-face interactions. Infant Behavior and Development, 4, 307-315. Floyd, F. J., & Widaman, K. F. (1995). Factor analy sis in the development and refinement of clinical assessment instruments. Psychological Assessment 7 (3), 286Â– 299. Feldman R., Weller, A., Leckman, J.F., Kvint, J., & Eidelman, A.I., (1999). The nature of the motherÂ’s tie to her infant: The formation of pa rent-infant bonding in healthy and at-risk dyads. Journal of Child Psychology and Psychiatry, 40, 929-939. Freed, R. D., Chan, P. T., Boger, K. D., & Tompson, M. C. (2012). Enhancing maternal depression recognition in health care settings: a r eview of strategies to improve detection, reduce barriers, and reach mothers in ne ed. Families, Systems & Health:The Journal of Collaborative Family Healthcare 30 (1), 1Â–18. Freeman, M. P., Wright, R. O., Watchman, M., Wahl, R. A., Sisk, D. J., Fraleigh, L., et al. (2005). Screening feasibility, prevalence, and risk factors. Journal of Women's Health, 14 (10), 929Â–936.
53 Fowles, E.R. (1998). The relationship between mater nal role attainment and postpartum depression. Healthcare for Women International, 19, 83-94. Geert-Jan, J. M., Stams, F. J., & van IJzendoorn, M H. (2002). Maternal sensitivity, infant attachment, and temperament in early childho od predict adjustment in middle childhood: The case of adopted children and their b iologically unrelated parents. Developmental Psychology, 38 (5), 806-821. Gjerdingen, D. K., & Yawn, B. P. (2007). Postpartum depression screening: importance, methods, barriers, and recommendations for practice Journal of the American Board of Family Medicine: JABFM 20 (3), 280Â–8. Goodman, S.H., & Gottlieb, I.H. (1999). Risk for ps ychopathology in the children of depressed mothers: a developmental model for under standing mechanisms of transmission. Psychological Review 106, 458-490. Hanna, B. (2004). The Early Detection of Postpartum Depression: Midwives and Nurses Trial a Checklist. Journal of Obstetric, Gynecologic, and Neonatal Nur sing 33 (2), 191Â–197. Hanna, B., Jarman, H., & Savage, S. (2004). The cli nical application of three screening tools for recognizing post-partum depression. International journal of nursing practice 10 (2), 72Â–9. Hanson, J. L., Adluru, N., Chung, M. K., Alexander, A. L., Davidson, R. J., & Pollak, S. D. (2013). Early Neglect Is Associated With Alterat ions in White Matter Integrity and Cognitive Functioning. Child Development 00 (0), 1-13.
54 Horowitz, J. A., Damato, E. G., Duffy, M. E., & Sol on, L. (2005). The relationship of maternal attributes, resources, and perceptions of postpartum experiences to depression. Research in Nursing & Health, 28, 159-171. Horowitz, J. A., & Goodman, J. H. (2005). Identifyi ng and treating postpartum depression. Journal of obstetric, gynecologic, and neonatal nur sing: JOGNN / NAACOG 34 (2), 264Â–73. Jekel, J.F., Elmore, J.G., Katz, D.L. (1996). Epidemiology, biostatistics and preventive medicine Philadelphia, PA: WB Saunders Company. Jreskog, K. G. (1969). A general approach to confi rmatory maximum likelihood factor analysis. Psychometrika 34 (2), 183-202. Kaitz, M., & Maytal, H. (2005). Interactions betwee n anxious mothers and their infants: An integration of theory and research findings. Infant Mental Health Journal, 26, 570-597. Kajantee, E., & Raikkonen, K. (2010). Early life pr edictors of the physiological stress response later in life. Neuroscience & Biobehavioral Review, 35 (11), 23-32. Kaplan, P.S., Burgess, A.P., Sliter, J.K., & Moreno A.J. (2009). Maternal sensitivity and the learning-promoting effects of depressed and no n-depressed mothersÂ’ infantdirected speech. Infancy, 14 (2), 143-161. Kaplan, P. S., Danko, C. M., Diaz, A., Kalinka, C. J. (2011). An associative learning deficit in 1-year-old infants of depressed mothers: role of depression duration. Infant Behavior and Development, 34 (1) 35-44.
55 Kim, P., Mayes, L., Feldman, R., Leckman, J.F., & S wain, J.E. (2012). Early postpartum parental preoccupation and positive parenting thoug hts: Relationship with parentinfant interaction. Infant Mental Health Journal, 1-13. Larsen, K. E., Schwarts, S.A., Whiteside, S. P., Kh andker, M., Moore, K. M., & Ambrowitz, J. S., Thought control strategies used b y parents reporting postpartum obsessions. Journal of Cognitive Psychotherapy: An Internationa l Quarterly, 20 (4), 435-445. Litze H., & Bentler, P.M. (1999) Cutoff criteria fo r fit indexes in covariance structure analysis: Conventional criteria versus new alterna tives. Structural Equation Modeling: A Multidisciplinary Journal, 6(1), 1-55. Lovejoy, M.C., Graczyk, P.A., OÂ’Hare, E., & Neuman, G. (2000). Maternal depression and parenting behavior: A meta-analytic review. Clinical Psychology Review, 20 (5), 561-592. Lumeng, J.C., Ozbeki, T.N., Appugliesse, D.P., Raci roti, N., Corwyn, R.F., & Bradley, R.H. (2012). Observed assertive and intrusive mater nal feeding behaviors increase child adiposity. American Journal of Clinical Nutrition, 3, 640-647. Mahler, M.S., Pine, F., & Bergman, A. (1975). The psychological birth of the human infant: Symbiosis and individuation. New York: Basic Books. Maina, G., Albert, U., Bogetto, F., Vaschetto, P., & Ravizza, L. (1999). Recent life events and obsessive-compulsive disorder (OCD): Th e role of pregnancy/delivery. Psychiatry Review, 89 (1), 49-58.
56 Mancini, F., Carlson, C., & Albers, L. (2007). Use of the Postpartum Depression Screening Scale in a collaborative obstetric practi ce. Journal of midwifery & womenÂ’s health 52 (5), 429Â–34. Mercer, R.T. (1985). The process of maternal role a ttainment over the first year. Nursing Research, 34, 198-204. Miller, R. L., Pallant, J. F., & Negri, L. M. (2006 ). Anxiety and stress in the postpartum: is there more to postnatal distress than depression ? BMC psychiatry 6 12. Mitchell, A. M., Mittelstaedt, M.E., & Schott-Baer D. (2006). Postpartum depression: the reliability of telephone screening. American Journal of Maternal Child Nursing 31 (6):382Â–387. Murray, L. & Cooper, P.J. (1997). Effects of postna tal depression on infant development. Archive of Disease in Childhood, 77, 99-101. Murray, L. (2009). The development of children of p ostnatally depressed mothers: Evidence from the Cambridge longitudinal study. Psychoanalytic Psychotherapy, 23 (3), 185-199. Muscat, T., Obst, P., Cockshaw, W., & Thorpe, K. (2 014). Beliefs about infant regulation, early infant behaviors and maternal pos tnatal depressive symptoms. Birth, 41 (2), 206-213. Muthn, B. & Muthn, L. (2013). Mplus version 7.1 [ Computer software]. Nunnally, J.C., & Bernstein, I.H. (1999). Psychometric Theory (3rd ed.). New York: McGraw-Hill. Nystrom, K., & Ohrling, K. (2004). Parenthood exper iences during the childÂ’s first year: Literature review. Journal of Advanced Nursing, 46, 319-330.
57 Paris, R., Bolton, R. E., & Weinberg, M. K. (2009). Postpartum depression, suicidality, and mother-infant interactions. Archives of womenÂ’s mental health 12 (5), 309Â–321. Reck, C., Hunt, A., Fuchs, T., Weiss, R., Noon, A., Moehler, E., Â… Mundt, C. (2004). Interactive regulation of affect in postpartum depr essed others and their infants: An overview. Psychopathology, 37, 272-280. Righetti-Veltema, M., Conne-Perrard, E., Bousquet, A., & Manzano, J. (1998). Risk factors and predictive signs of postpartum depressi on. Journal of affective disorders 49 (3), 167Â–80. Sharp, L. K., & Lipsky, M. S. (2002). Screening for depression across the lifespan: A review of measures for use in primary care settings American Family Physician, 66 (6), 1001-1009. Sit, D.K., & Wisner, K.L. (2009). The identificatio n of postpartum depression. Clinical Obstetrics & Gynecology, 53 (3), 456-468. SPSS Inc. (2012). Statistical package for the socia l sciences v. 21 [Computer software]. Tabachnick, B. G., and Fidell, L. S. (2013). Using multivariate statistics, (6th ed.) Boston: Pearson. Taylor, Z. E., Eisenberg, N., Spinrad, T.L., & Widm an, K.F. (2013). Longitudinal relations of intrusive parenting and effortful con trol to ego-resiliency during childhood. Child Development, 84 (4), 1145-1151. Teti, D. M., & Gelfand, D. M. (1991). Behavioral co mpetence among mothers of infants in the first year: The meditational role of materna l self-efficacy. Child Development, 62, 918-929.
58 Trapolini, T., Ungerer, J., & McMahon, C. (2008). M aternal depression: Relations with maternal caregiving representations and emotional a vailability during the preschool years. Attachment & Human Development, 10 (1), 73-90. Troy, A. S., Shallcross, A.J., & Mauiss, I.B. (2013 ). A person-by-situation approach to emotion regulation: Cognitive reappraisal can eithe r help or hurt, depending on the context. Psychological Science, 24 (12), 2505-2514. Van Doesum, K., Hosman, M., Riksen-Walraven, J., & Hoefnagels, C. (2007). Correlates of depressed mothersÂ’ sensitivity toward their infa nts: The role of maternal, child, and contextual characteristics. Journal of the American Academy of Child and Adolescent Psychiatry, 46 (6), 747-756. Volling, B. L., McElwain, N.L., Notaro, P.C., & Her rera, C. (2002). ParentsÂ’ emotional availability and infant emotional competence: Predi ctors of parent-infant attachment and emerging self-regulation. Journal of Family Psychology, 16 (4), 447-465. Winnicott, D.W. (1975 ). Primary maternal pre occupation. In: Collected Papers: Through paediatrics to psycho-analysis. Basic Books : New York, 300-305. Wisner, K. L., Parry, B. L., & Piontek, C. M. (2002 ). Postpartum depression. New England Journal of Medicine, 347 (3), 194-199. Zimbaldi, C. F., Cantillno, A., Montenegro, A. C., Paes, J. A., Thiago, C. A., & Sougey, E. B. (2009). Postpartum onset obsessive-compulsive disorder: Prevalence and clinical characteristics. Journal of Comprehensive Psychiatry, 50 (6), 503-509. Ziv, Y., Aviezer, O., Gini, M., Sagi, A., & Koren-K arie, N. (2000). Emotional availability in the mother-infant dyad as related t o the quality of infant-mother attachment relationship. Attachment & Human Development, 2 (2), 149-169.