The Structure of Women’s Mood in the Early Postpartum

Abstract

The “postpartum blues” is a mild, predictable mood disturbance occurring within the first several days following childbirth. Previous analyses of the “blues” symptom structure yielded inconclusive findings, making reliable assessment a significant methodological limitation. The current study aimed to explicate the symptom structure of women’s mood following childbirth, and to examine psychometric properties of the Daily Experiences Questionnaire (DEQ), an adapted version of the Kennerley Blues Questionnaire that included additional items from the Positive and Negative Affect Schedule. Mothers who recently delivered (N = 216) were recruited from a university hospital and asked to complete mood ratings on six consecutive days using the DEQ. Exploratory factor analysis yielded an interpretable two-factor solution identified as Negative Affect and Positive Affect. Scale reliability indices were excellent, with a high level of agreement in factor structure over time. This two-factor model provides reliable assessment of women’s mood in the early postpartum, informing the study of reproductive-related mood disorders.

Keywords

postpartum blues factor analysis positive and negative affect perinatal mental health

The postpartum blues is often characterized as a mild mood disturbance linked to childbirth that occurs within the first 7 to 10 days after delivery (Harris et al., 1994; Iles, Gath, & Kennerley, 1989). Because there are no agreed on criteria for the postpartum blues, prevalence rates range from 26% to 84%, reflecting more (O’Hara, Schlechte, Lewis, & Wright, 1991) or less (Oakley & Chamberlain, 1981) stringent criteria. The etiology of the postpartum blues is not well-understood, but it is commonly believed to have a biological basis due to the abrupt withdrawal of ovarian hormones following childbirth, in particular estradiol and progesterone (e.g., Bloch, Daly, & Rubinow, 2003; Glover & Kammerer, 2004). This very abrupt change in the reproductive hormonal milieu may produce perturbations in mood that are relatively unique compared with other periods of hormonal changes, such as menstruation (Glover & Kammerer, 2004). More specifically, estradiol levels by the end of pregnancy are as much as 50 times the maximum menstrual cycle level and return to early follicular levels by Day 3 postpartum. Furthermore, progesterone levels increase by 10 times the highest menstrual cycle levels by the third trimester and typically normalize to follicular levels between postpartum Days 3 to 7 (Bloch et al., 2003; Nappi et al., 2001). Sacher et al. (2010) found a link between lowered estrogen levels during the first 3 to 4 days postpartum and MAO-A (monoamine oxidase A) synthesis, identifying elevated MAO-A levels in the early postpartum period as a marker for a monoamine-lowering process that may contribute to the mood dysregulation characteristic of the postpartum blues. Moreover, these researchers showed that MAO-A binding was highest on Day 5 postpartum, the day most strongly associated with a peaking phenomenon across studies of postpartum blues (e.g., Harris et al., 1994).

Emotional lability appears to be a hallmark of the postpartum blues and is characterized by symptoms reflecting abrupt changes in mood and emotion as well as characteristics such as having up and down mood and being overly sensitive (Ballinger, Buckley, Naylor, & Stansfield, 1979; Kendell, McGuire, Connor, & Cox1981; O’Hara, 1995). Timing of symptoms is critical, potentially informing the etiologic biological underpinnings of the postpartum blues. A “peaking phenomenon” has been identified as highly characteristic of the blues syndrome, with most evidence to suggest that negative mood symptoms are at their most intense between 3 and 5 days after delivery (e.g., Gonidakis, Rabavilas, Varsou, Kreatsas, & Christodoulou, 2007; Harris et al., 1994; O’Hara, 1995); other researchers have shown a peak in mood between 4 and 6 days postpartum (e.g., Stein, 1980).

Despite the relatively mild clinical course of the postpartum blues, it may reflect a vulnerability to reproductive-related mood disorders. For example, the postpartum blues has been linked to later postpartum depression (e.g., Henshaw, Foreman, & Cox, 2004; O’Hara et al., 1991; Reck, Stehle, Reinig, & Mundt, 2009; Watanabe et al., 2008). Additionally, women who experience premenstrual symptoms are at increased risk for the postpartum blues (Bloch, Rotenberg, Koren, & Klein, 2005; Henshaw et al., 2004; O’Hara et al., 1991). These findings underscore the importance of understanding the etiology and characteristics of symptom expression in the postpartum blues as well as the possibility that there is a unique structure to mood in the early postpartum period.

Although it may potentially inform both research and clinical work in the area of women’s mental health, existing research on the postpartum blues has important methodological limitations. These limitations include the use of nonspecific measures to assess the blues syndrome. More specifically, research has relied on a variety of measures to assess the blues syndrome, including visual analogue scales (Kendell et al., 1981), adjective checklists, and rating scales designed to measure symptoms of depression, such as the Beck Depression Inventory–I (Handley, Dunn, Waldron, & Baker, 1980) and the Edinburgh Postnatal Depression Scale, a specific measure of postpartum depression (Hannah, Adams, Lee, Glover, & Sandler, 1992). These measures have significant limitations, including (a) failure to address the full range of women’s affective symptoms, (b) poor design for daily administration, and (c) inadequate psychometric validation in the context of the postpartum blues (Henshaw, 2003).

Early investigators of the postpartum blues developed scales by selecting items based on clinical descriptions of women’s emotional experiences following childbirth, a rational—rather than a factor analytic—approach. In general, this strategy yielded one-factor solutions (e.g., Pitt, 1973; Yalom, Lunde, Moos, & Hamburg, 1968). Whiffen (1988) conducted a principal components analysis to evaluate the coherence of blues symptoms identified in previous research. In this study, participants (N = 124) were asked to rate the severity of 15 symptoms retrospectively on a 7-point scale on Day 11 postpartum. Through this process, a four-component solution emerged, accounting for 62.1% of the variance. The components were labeled (a) Maternity Blues, (b) Tension, (c) Cognitive Impairment, and (d) Blues Attributions. However, these findings should be interpreted with caution, given the use of a small item pool and retrospective mood ratings in scale development.

Kennerley and Gath (1989) employed a cluster analytic technique to develop the Kennerley Blues Questionnaire (KBQ). The original item pool in this scale development process consisted of 47 different adjectives describing women’s emotional experiences during the first few days following childbirth. Findings from the cluster analysis yielded seven clusters: Primary blues, Decreased alertness, Hypersensitivity, Decreased self-confidence, Depression, Despondency, and Reservation (Kennerley & Gath, 1989).

Research findings on mood in the early postpartum period suggest that its structure may be more complex than mood at other times in a woman’s life (e.g., Kennerley & Gath, 1989; Whiffen, 1988). This is significant because it has implications for the assessment of mood symptoms in the early postpartum period and suggests that there is something unique about a woman’s experience during this time that goes beyond the adjustments of being a new mother. Within the general mood literature, structural analyses consistently have identified two varimax-rotated factors defined as Positive Affect (PA) and Negative Affect (NA; Watson & Tellegen, 1985). PA is defined by mood items such as enthusiastic, active, and alert, whereas NA is characterized by descriptors such as angry, disgusted, guilty, and afraid (Watson, Clark, & Tellegen, 1988). These two symptom dimensions are widely used in the self-report mood literature to model affect (Watson et al., 1988) and may extend to women’s affective experiences in the early postpartum as well.

In the present study, exploratory factor analysis (EFA) was used to examine the underlying symptom structure of the blues and to assess the psychometric properties of the Daily Experiences Questionnaire (DEQ), a modified and expanded version of the KBQ. To determine the stability of the most interpretable and meaningful factor solution, factor comparability analyses were conducted by comparing models derived on the first 6 days of the study assessment period; we also examined the replicability of the factor loadings via congruence coefficients. Although we collected data on 10 consecutive evenings, postpartum Day 6 was selected as a cutoff based on previous research suggesting that the “peaking phenomenon,” highly characteristic of the blues, occurs between postpartum Days 3 and 5 (e.g., Gonidakis et al., 2007; Harris et al., 1994; Kendell, Mackenzie, West, McGuire, & Cox, 1984), and that the blues remit around Day 10 (Stein, 1980).

Method

Participants

Eligible mothers (N = 216) spoke English and were between the ages of 18 and 45 years. Mothers were included in the study regardless of birth complications, number of days spent on the maternity unit, and method of delivery to ensure a wide range of symptoms. The majority of mothers did not endorse any birth complications (76%); reported birth complications were blood loss requiring medication/transfusion (4%), significant tearing or stitches (10%), infection (2%), and other (8%). Vaginal delivery was most common (70%), with 30% delivering via cesarean section/other method. Participants were, on average, 28 years old (SD = 5.6) and college educated (M = 15.0 years, SD = 2.9). Approximately 80% of the participants were Caucasian, 75% were breastfeeding at time of entry into the study, and approximately 70% were married. More than half of the participants were working (61.1%) and roughly half (49.9%) had an annual income equal to or greater than $40,000. The number of participants completing mood assessments for each postpartum day reported in these analyses ranged from N = 123 to N = 172 (Day 1, N = 172; Day 2, N = 130; Day 3, N = 123; Day 4, N = 130; Day 5, N = 127; and Day 6, N = 126); 75 participants (35%) completed ratings on all 6 days. Ten participants (7%) were experiencing a major depressive episode at time of completing the Day 1 mood assessment.

Procedures

On receiving approval from the University of Iowa’s Institutional Review Board, mothers who recently delivered were recruited from the maternity unit at the University of Iowa Hospitals and Clinics on postpartum Day 0, 1, or 2. Postpartum Day 1 was defined as beginning at 12:01 a.m. on the first day after childbirth. Recruiters approached mothers meeting inclusion criteria (English speaking and 18 years or older) and asked for their participation in a research study that examined women’s emotional experiences following childbirth. Recruiters explained the study protocol and reviewed the consent form with interested participants; on receiving informed consent, recruiters asked participants to complete three brief questionnaires: (a) Demographic Questionnaire, (b) Mood Variability Questionnaire, and (c) the DEQ. Participants were given the option to complete the daily mood ratings either online (60% of participants) or over the telephone (40%) with a researcher between the hours of 5 p.m. and 8 p.m. If this was not possible, participants were asked to complete the questionnaire before bedtime that evening or by noon the following day. If 1 day was missed, participants were instructed to complete two questionnaires (one for the missed day and one for the current day) in the evening of the following day. Online questionnaires were completed using a web-based survey application offered to researchers by The University of Iowa.

Measures

Demographic Questionnaire

Information was collected on participant’s age, education, income, marital status, and breastfeeding status, as well as parity and method of delivery.

Daily Experiences Questionnaire

The DEQ (see the appendix) is a 38-item scale adapted from the KBQ (Kennerley & Gath, 1989). We expanded the original 28-item Kennerley scale (one item, depression, was inadvertently left off the original Kennerley scale; thus, 27 items from the scale were included in our item pool) to include 10 items from the PANAS (Positive and Negative Affect Schedule; Watson et al., 1988), and one item to assess loss of appetite. Because of minimal representation of positive mood descriptors in previous measures of the postpartum blues (e.g., Kennerley & Gath, 1989; Stein, 1980), we included six positive mood items (active, determined, enthusiastic, inspired, interested, and proud) from the PANAS in our item pool. To maximize representation of NA-related mood descriptors, we also included four negative mood descriptors (afraid, ashamed, guilty, and hostile) from the PANAS. Finally, we included the item loss of appetite because of its relation to depression and variability in the postpartum period (Stein, 1980). Each item was rated (similar to the PANAS) on a 1 to 5 Likert-type scale ranging from 1 (very slightly or not at all) to 5 (extremely), reflecting the mother’s affective experience on that day.

Data Analyses

EFA was conducted using the principal axis factoring extraction method, with varimax (orthogonal) rotation, to factor analyze the daily mood ratings from the DEQ. Separate analyses were conducted on each of the first 6 days of the postpartum period. EFA was selected for factor analysis and for the examination of the basic psychometric properties of the DEQ because it is designed for exploring a data set with limited theoretical basis for specifying a priori the number and patterns of common factors (Costello & Osborne, 2005; Floyd & Widaman, 1995). A parallel analysis (Horn, 1965; O’Connor, 2000) was performed to identify the appropriate number of factors for extraction. The initial solution derived on postpartum Day 1 was then validated using factor comparison analyses based on two different approaches: coefficients of congruence (Tucker, 1971) and comparability coefficients derived from factor scores (Gorsuch, 1983). Postpartum Day 1 data were used to run the EFA because it represented the largest sample size (N = 172) over the 6-day assessment period.

Results

Exploratory Factor Analysis

Factor retention

Parallel analysis (O’Connor, 2000) for the 95th percentile in 1,000 random data sets identified 4 factors as the maximum number of meaningful dimensions. At root 5, the eigenvalue (1.58) fell below the mean random data eigenvalue (1.64), indicating that the fifth factor was modeling error variance. Thus, we considered four factors a maximum and examined two-, three-, and four-factor solutions. In examining the loadings and the number of items with high cross-loadings in the two-, three-, and four-factor solutions, it appeared that the two-factor solution yielded the most distinguishable set of factors with meaningful content; these dimensions could be identified as NA and PA, respectively. In the two-factor solution, there were relatively few items with cross-loadings greater than |.40|, compared with the three- and four-factor solutions, suggesting a clear delineation between these two dimensions. The three- and four-factor solutions yielded additional redundant factors, tapping variations of NA and PA. Additionally, we reran the analyses using an oblique promax rotation and obtained results very similar to those generated using varimax rotation.

Factor similarity

Two sets of analyses were conducted to provide a quantitative comparison of the factor structures of NA and PA across the first 6 days of the assessment period. Additionally, we conducted analyses of factor similarity for the three- and four-factor solutions discussed previously to ensure that these models did not enhance the overall quality of the structure beyond two factors. A well-established method of factor comparison is computing the coefficient of congruence (Tucker, 1971; Wrigley & Neuhaus, 1955). Congruence coefficients are based on a comparison of factor loadings between solutions and provide a measure of the similarity of the factor-loading matrices in these solutions. Factor comparisons based on the coefficient of congruence method are common in the literature on scale development and appropriate for determining the replicability of a factor solution (Cordano, Scherer, & Owen, 2003).

According to Tucker (1971), a coefficient of 1.0 suggests perfect agreement between two factors being compared, whereas a coefficient of 0 represents no agreement. More specifically, a coefficient greater than .90 is indicative of congruent factors, and less than .45 suggests no agreement (Tucker, 1971). These coefficients ranged from .98 to .99 for each of the factors in the two-factor solution (i.e., NA and PA), suggesting a high level of agreement in the replicability of the factor structure over the assessment period (see Table 1).

Table 1.

Congruence Coefficients across Postpartum Days 1 to 6 for Negative Affect and Positive Affect Factors

Postpartum day	1	2	3	4	5	6
1. Day 1		.99	.99	.99	.99	.99
2. Day 2	.99		.98	.99	.99	.99
3. Day 3	.99	.99		.99	.99	.99
4. Day 4	.99	.99	.99		.99	.99
5. Day 5	.99	.99	.99	.99		.98
6. Day 6	.99	.99	.99	.99	.99

Note. N = 38. Coefficients for negative affect are presented above the diagonal. Coefficients for positive affect are presented below the diagonal.

Comparability coefficients also were computed. In these analyses, regression-based factor scoring weights are calculated separately for each factor on each postpartum day. These weights are then multiplied by each participant’s standardized responses to produce an overall score for each person on each factor on each day. The weights derived from one solution then can be applied to others, yielding multiple scores for each factor (one based on the weights from each of the solutions) that are then correlated within the same data set; in this case, we computed these comparability correlations using the Day 1 data, which had the largest sample size (N = 172). The comparability correlations based on the two-factor model ranged from .98 to .99 for each of the factors (NA and PA) across the 6 postpartum days, a pattern similar to that found in the coefficient of congruence analyses (see Table 2).

Table 2.

Comparability Correlations across Postpartum Days 1 to 6 for the Negative Affect and Positive Affect Factors

Postpartum day	1	2	3	4	5	6
1. Day 1		.98	.99	.98	.99	.99
2. Day 2	.98		.99	.98	.98	.98
3. Day 3	.98	.98		.98	.98	.98
4. Day 4	.99	.99	.99		.99	.99
5. Day 5	.99	.98	.99	.99		.99
6. Day 6	.99	.99	.99	.99	.99

Note. These correlations were computed using the Day 1 responses (N = 172); see text for details. Correlations for negative affect are presented above the diagonal. Correlations for positive affect are presented below the diagonal.

Additionally, the three- and four-factor solutions derived on postpartum Day 1 were compared with solutions found on postpartum Days 2 through 6. The third factor in the three-factor solution (not shown) accounted for an additional 10.2% of the variance; in this solution, PA dimension split apart with markers of concentration and energy forming a third factor. Thus, these findings yielded a modified version of the two-factor solution, with a dimension identified as Energy emerging as a third factor. Congruence coefficients for the three-factor solution suggested a high degree of similarity for the NA (.93 to .97) and PA (.99) factors; however, the coefficients for the Energy factor (−.01 to .94) indicated that it was not strongly replicable across days. Similarly, comparability correlations for NA (.69 to .97) and PA (.95 to .99) showed good to excellent factor similarity over time, whereas correlations for the Energy factor (−.05 to .94) indicated that it did not replicate over time, consistent with findings based on congruence coefficients.

The fourth factor in the four-factor solution (not shown) accounted for an additional 6.1% of the variance; in this solution, markers of both NA and PA split off to form two hybrid factors. Some markers of NA (e.g., up and down in mood, irritable, crying) split to form a third dimension, which we identified as a Blues factor. Similarly, some markers of PA (e.g., alert, active, able to concentrate) split to form a fourth dimension identified as Energy, similar to the Energy dimension identified in the three-factor solution. The congruence coefficients based on the four-factor solution suggested that the factors representing NA (.60 to .97), PA (.96 to .99), and the Blues (.91 to .97) were congruent and replicated over time, whereas coefficients for the Energy factor (.12 to .90) indicated that it was not consistently replicable across the study period. Findings based on comparability correlations were somewhat consistent, with correlations for PA (.95 to .99) and the Blues (.63 to .97) suggesting good agreement in factor similarity over time. In contrast, correlations for NA (−.02 to .90) and Energy (.01 to .92) showed little to no support for replication of the four-factor structure over time.

The factor similarity data for the two-factor solution provided a clear contrast to the three- and four-factor solutions. The two-factor solution yielded factors that were clearly delineated and highly replicable, a sharp contrast to the three- and four-factor solutions that yielded smaller dimensions (Energy and the Blues) that were relatively unstable over time. Thus, we concluded that the three- and four-factor solutions did not enhance the overall structure beyond the first two dimensions, and adopted the two-factor solution as the final model.

Polychoric analyses

Because items do not have normal distributions, an alternative approach to item-level factor analysis is to use polychoric correlation matrices; polychoric correlations are estimates of what the Pearson correlations would be had the items been normally distributed (Holgado-Tello, Chacón-Moscoso, Barbero-García, & Vila-Abad, 2010). To examine this as a potential consideration in the current data analyses, a polychoric EFA was run using Mplus (Version 6.1). The results were highly consistent with the original two-factor solution identified using Pearson correlations (see Table 3). More specifically, the EFA based on polychoric correlations supported a clear two-factor solution identified as NA and PA (see Table 4). The root mean square error of approximation (RMSEA) estimate was .05 and the value for the root mean square residual (RMSR) was .09. Suggested cutoff values for incremental fit indices that represent acceptable lower bounds of good model fit are .08 for RMSEA and .10 for the RMSR (Vandenberg & Lance, 2000).

Table 3.

Pearson Correlation–Based Factor Loadings of Items from the Daily Experiences Questionnaire

Item	Negative Affect	Positive Affect
Up and down in mood	.82	−.13
Mentally tense	.72	−.27
Overemotional	.70	−.04
Oversensitive	.68	−.10
Brooding	.67	−.17
Low spirited	.62	−.27
Tearful	.62	−.17
Irritable	.61	−.21
Helpless	.58	−.32
Guilty	.58	−.24
Crying without being able to stop	.58	−.16
Feeling sorry for yourself	.55	−.23
Ashamed	.55	−.18
Afraid	.53	−.18
Changeable in your spirits	.53	−.13
Restless	.48	−.05
Anxious	.47	−.06
Loss of appetite	.46	−.09
Forgetful, muddled	.43	−.27
Hostile	.43	.02
Finding it difficult to show feelings	.40	−.11
Tired	.34	−.19
Emotionally numb	.34	−.19
Wishing you were alone	.27	−.17
Inspired	−.08	.72
Confident	−.31	.72
Lively	−.07	.71
Enthusiastic	−.20	.67
Interested	−.17	.65
Alert	−.15	.64
Determined	.02	.63
Calm, tranquil	−.29	.63
Active	−.02	.60
Proud	−.23	.59
Happy	−.42	.59
Able to concentrate	−.22	.59
Elated	−.22	.47
Mentally relaxed	−.41	.43

Note. N = 172 (Postpartum Day 1). Item loadings meeting criteria as a marker for each factor are in boldface. Criteria for Pearson correlations are |.40| on the primary factor and |.30| on the secondary factor.

Table 4.

Polychoric Correlation–Based Factor Loadings of Items from the Daily Experiences Questionnaire

Item	Negative Affect	Positive Affect
Up and down in mood	.90	.12
Mentally tense	.76	−.37
Overemotional	.83	−.01
Oversensitive	.79	−.06
Brooding	.71	−.29
Low spirited	.72	−.43
Tearful	.74	−.13
Irritable	.69	−.20
Helpless	.65	−.47
Guilty	.72	−.45
Crying without being able to stop	.79	−.29
Feeling sorry for yourself	.69	−.38
Ashamed	.77	−.39
Afraid	.59	−.20
Changeable in your spirits	.63	−.11
Restless	.52	−.07
Anxious	.55	−.08
Loss of appetite	.50	−.11
Forgetful, muddled	.48	−.29
Hostile	.65	−.05
Finding it difficult to show feelings	.56	−.25
Tired	.40	−.19
Emotionally numb	.54	−.34
Wishing you were alone	.33	−.20
Inspired	−.08	.77
Confident	−.34	.73
Lively	−.12	.76
Enthusiastic	−.23	.71
Interested	−.17	.73
Alert	−.18	.68
Determined	.01	.72
Calm, tranquil	−.33	.64
Active	−.03	.66
Proud	−.25	.69
Happy	−.46	.68
Able to concentrate	−.23	.58
Elated	−.26	.52
Mentally relaxed	−.46	.46

Note. N = 172 (Postpartum Day 1). Item loadings meeting criteria as a marker for each factor are in boldface. In moving to polychoric interitem correlation matrices, we adjusted our original criteria upward to determine markers on each of the factors—using |.50| or greater on the primary factor and less than |.40| on the secondary factor (Note: In our original analyses, we used |.40| on the primary factor and |.30| on the secondary factor)—because polychoric correlations and factor loadings tend to be higher, relative to Pearson correlations and factor loadings (Holgado-Tello et al., 2010).

Additionally, the initial solution derived on postpartum Day 1 using polychoric interitem correlation matrices was validated by computing coefficients of congruence with the two Pearson correlation-based factors; as previously discussed, congruence coefficients are a well-established method of factor comparison (Tucker, 1971; Wrigley & Neuhaus, 1955). These coefficients were .98 or greater for both factors in the two-factor solution (NA and PA). Thus, the same two factors emerged, regardless of whether Pearson correlations or polychoric correlations were used in the EFAs. These findings also are consistent with our original factor similarity analyses using Pearson correlation matrices and suggest a very high level of agreement in the replicability of the factor structure over the 6-day assessment period.

Scale creation

Given that these two approaches clearly yielded very similar results, we used the original Pearson-based factors to create scales. In creating scales based on the two-factor solution, an item was considered a marker for that particular scale if it had a loading of |.40| or greater on the primary factor and a loading less than |.30| on the secondary factor in the Day 1 ratings. Using these criteria, items that did not clearly load on either factor were dropped from the analyses. A total of seven items were subsequently dropped from the scales. The following four items were removed from Factor 1: helpless, emotionally numb, tired, and wishing you were alone; the following three items were removed from Factor 2: confident, happy, and mentally relaxed. After removing the seven items from the original item pool, a total of 31 items were included in the two scales (see Table 5 for item content and loadings). The first scale was labeled NA, reflecting symptoms of low mood, crying, and mood lability, and was composed of 20 items. The second scale was labeled PA, reflecting a sense of enthusiasm and feeling lively and engaged in life and consisted of 11 items. The mood ratings on this reduced set of 31 items were again subjected to a principal factor analysis with varimax rotation to ensure that all items on the two scales had factor loadings of |.40| or greater on the primary factor, and no cross loadings greater than |.30| on the secondary factor (see Table 5); these two factors accounted for 40.8% of the variance.

Table 5.

Varimax-Rotated Factor Loadings of Items in Final Scales from the Daily Experiences Questionnaire

Item	Negative Affect	Positive Affect
Up and down in mood	.82	−.13
Mentally tense	.72	−.27
Overemotional	.70	−.04
Oversensitive	.68	−.10
Brooding	.67	−.17
Low spirited	.62	−.27
Tearful	.62	−.17
Irritable	.61	−.21
Guilty^a	.58	−.24
Crying without being able to stop	.58	−.16
Feeling sorry for yourself	.55	−.23
Ashamed^a	.55	−.18
Afraid^a	.53	−.18
Changeable in your spirits	.53	−.13
Restless	.48	−.05
Anxious	.47	−.06
Loss of appetite	.46	−.09
Forgetful, muddled	.43	−.27
Hostile^a	.43	.02
Finding it difficult to show feelings	.40	−.11
Inspired^a	−.08	.72
Lively	−.07	.71
Enthusiastic^a	−.20	.67
Interested^a	−.17	.65
Alert	−.15	.64
Determined^a	.02	.63
Calm, tranquil	−.29	.63
Active^a	−.02	.60
Proud^a	−.23	.59
Able to concentrate	−.22	.59
Elated	−.22	.47

Note. N = 172 (Day 1). Items loading on each factor are in boldface. Ratings were made on a 5-point scale (1 = very slightly or not at all, 5 = extremely).

Items from the Positive and Negative Affect Schedule.

Scale characteristics

Internal consistency reliabilities (Cronbach’s coefficient alpha and average interitem correlations [AICs]), and scale intercorrelations were computed for NA and PA scales on postpartum Days 1 to 6, and are presented in Table 6. The alpha reliabilities are consistently strong for both scales, with coefficients ranging from .91 to .94 for NA and from .88 to .92 for PA. According to Clark and Watson (1995), an AIC for a scale in the moderate range (.15 to .50) reflects an appropriate level of internal consistency within the items. The AICs for both scales fell within this range. The AICs across days for NA ranged from r = .35 to r = .44; those for PA ranged from r = .39 to r = .41, which again illustrates the stability of the two-factor solution across days. The correlations between the NA and PA scales across days ranged from r = −.31 to r = −.49, which was in the expected range given the item content of these scales (Watson & Clark, 1997; Watson & Tellegen, 1985).

Table 6.

Basic Scale Characteristics for the Daily Experiences Questionnaire

		Negative affect (NA)		Positive affect (PA)
Day	N	M	SD	M	SD	NA α	PA α	NA AIC	PA AIC	NA–PA intercorrelation
1	172	30.63	10.45	34.87	8.48	.91	.88	.35	.41	−.42
2	130	32.84	11.95	34.11	9.06	.92	.89	.37	.43	−.37
3	123	33.29	12.08	34.40	9.46	.92	.91	.38	.48	−.49
4	130	33.88	12.24	32.85	8.70	.92	.90	.38	.44	−.40
5	127	40.72	13.19	30.11	8.55	.92	.91	.39	.49	−.39
6	126	33.16	13.60	32.79	9.70	.94	.92	.44	.39	−.31

Note. α = Cronbach’s alpha; AIC = average interitem correlation.

Discussion

Main Findings

This study addressed methodological issues with using current nonspecific measures to assess the blues, including (a) failure to capture the full range of affect-related mood descriptors, both negative and positive, experienced by recently delivered mothers; (b) poor design for daily administration; and (c) inadequate psychometric validation in the context of the blues. Thus, we developed a more comprehensive measure of the blues for daily administration, with additional items tapping PA and NA. EFA was employed in this study and yielded a two-factor solution that was identified as having the best fit for modeling the final 31 mood items on the DEQ, with a clean split between two symptom dimensions, NA and PA. Notably, our findings showed that the strongest loadings on the PA scale were associated with positive mood items extracted from the PANAS, indicating that positive mood descriptors (e.g., inspired, enthusiastic) may play an important role in the characterization of women’s mood following childbirth. Finally, factor similarity analyses confirmed that the same two-factor structure emerged across 6 consecutive days following delivery, supporting the stability of the two-factor model over time.

Comparison With Previous Studies

Previous studies examining the symptom structure of the blues have relied on rational or structural analytic approaches to scale development. Using a rational approach, scales were developed using affect-related mood descriptors elicited by newly delivered mothers. This approach generally yielded one-factor solutions (e.g., Handley et al., 1980). Previous investigators have also relied on structural approaches, including principal components (Whiffen, 1988) and cluster analytic techniques (Kennerley & Gath, 1989). Although these techniques are considered acceptable for scale development, methodological issues, such as small sample size and retrospective mood ratings (Whiffen, 1988) and lack of reliability indices and replication data (Kennerley & Gath, 1989) raise concerns about the stability, comprehensiveness, and accuracy of these models. Therefore, it is difficult to generate conclusions on how symptoms relate to each other and the extent to which the blues serve as a reliable predictor of clinically significant disorders.

In contrast to previous studies on scale development, the current study used EFA with principal axis factoring, a robust and well-established data reduction method (Floyd & Widaman, 1995), to establish a two-factor solution of NA and PA as having the best fit for modeling the final 31 items on the DEQ. Multiple mood assessments in the present study also facilitated a rigorous examination of the structure of women’s emotional experiences (Hedges, Jandorf, & Stone, 1985) during the early postpartum, and allowed for factor comparisons between different days of sampling using the coefficient of congruence and comparability coefficients methods (Teel & Verran, 1991; Tucker, 1971; Wrigley & Neuhaus, 1955). Earlier structural-analytic studies did not include multiple assessments and did not examine the replicability of the factor structure of postpartum mood over time (Kennerley & Gath, 1989; Whiffen, 1988).

Previous research findings suggest that there is a distinct cluster (Kennerley & Gath, 1989) or component (Whiffen, 1988) representing the blues syndrome. In the current study, the classic symptoms (e.g., up and down in mood, crying, low spirited) forming the “Primary blues” cluster identified by Kennerley and Gath (1989) showed moderate to strong loadings on the NA factor, suggesting these items are (a) cohesive and (b) substantially linked to other types of negative mood. Moreover, factor-similarity analyses based on three- and four-factor solutions (which included a Blues factor) showed a poor level of replicability in the factor structure over time. In sum, there was little support for a separate Blues factor in the present study.

Limitations

Our study had some limitations. We relied on relatively small sample sizes (e.g., postpartum Day 1, N = 172; Day 2, N = 130) for conducting EFA, which may potentially affect the generalizability of the results. However, the very strong level of factor similarity we observed across the 6 days mitigates this concern to a considerable extent (Hogarty, Hines, Kromrey, Ferron, & Mumford, 2005). Furthermore, our sample was primarily Caucasian (80%) and roughly half (49.9%) reported an annual income of greater than or equal to $40,000. Thus, replication of the current findings is needed in more diverse samples. There also was a high rate of attrition in this study. The total sample size for each postpartum day reported on in this study ranged from N = 123 to N = 172, with a total of 216 women enrolled in the study. It was revealed through discussions with participants that remembering to complete questionnaires daily was burdensome, especially while caring for a newborn. However, only a small proportion of mood ratings (7%) were completed at times outside of protocol (e.g., after 8 p.m. or on the following day). Thus, adherence to study protocol was minimally compromised.

Women were asked to complete the DEQ on a daily basis between the hours of 5 p.m. to 8 p.m., reflecting on their mood over the entire day when responding to each of the items on the questionnaire. It is possible that assessing mood at a different time of the day or asking about mood at the time of the rating specifically would have yielded different results. However, the approach we used represents standard protocol in the mood literature (e.g., Watson et al., 1988) and provides an overall representation of women’s mood throughout the day. Another limitation is that 7% of the women (N = 10/148) were experiencing a major depressive episode at time of completing the Day 1 mood assessment. The presence of depression at time of mood assessment may have potentially inflated the report of negative mood and/or diminished report of positive mood. Finally, as noted earlier, the depression item was inadvertently left off the DEQ. Nevertheless, even without this item, the NA factor was clearly defined by 20 negative mood descriptors with strong loadings and weak cross- loadings.

Implications and Future Research

This is the first study to employ a well-established and robust data reduction method, EFA with principal axis factoring, to examine the psychometric properties and underlying factor structure of a postpartum blues measure, the DEQ. The DEQ is brief and suitable for daily administration, with strong psychometric properties that support its use as a reliable measure to assess women’s emotional experiences following delivery. A two-factor model identified as NA and PA was determined to have the best fit for modeling the 31 items on the DEQ, consistent with the two-factor model of affect found in the general mood literature (Watson et al., 1988; Watson & Tellegen, 1985). This suggests that the structure of women’s mood during the first week after delivery is no different than at other times outside the early postpartum, and that women’s mood is not affected by the process of hormonal normalization suggested in previous research (Glover & Kammerer, 2004). Additionally, our findings lend support for the role of positive mood descriptors in the characterization of women’s mood following delivery. Fluctuations in positive mood potentially are important and should not be overlooked; existing blues measures do not provide adequate representation of positive mood items (Handley et al., 1980; Kennerley & Gath, 1989; Stein, 1980).

Future research should include replication studies using the DEQ to model NA and PA and to examine the invariance of the two-factor model before, during, and after approximately 2 weeks postpartum to determine whether this widely used model of general affect generalizes to the postpartum period as well. Because it is important to improve understanding into the etiologic biological correlates of the postpartum blues, it is worthwhile to examine potential correlates using the DEQ. In addition, it may be worthwhile to use the DEQ to assess mood in a recently delivered population with complex mood descriptors removed from the scale (e.g., ashamed, enthusiastic, inspired), to determine if the two-factor structure replicates. Finally, mood monitoring throughout the day using the DEQ rather than simply at the end of the day has the potential to capture a great deal more mood variability in recently delivered women (Hedges et al., 1985). A sampling technique, such as ecological momentary assessment (Stone & Shiffman, 1994) may be beneficial in future research as it would allow for reporting of symptoms in close proximity to the actual experience.

Footnotes

Appendix

Original Items Comprising the Daily Experiences Questionnaire (DEQ)

1.	Crying without being able to stop
2.	Mentally tense
3.	Able to concentrate
4.	Low spirited
5.	Elated (extremely happy and excited)
6.	Helpless^a
7.	Alert
8.	Finding it difficult to show your feelings
9.	Forgetful, muddled
10.	Mentally relaxed^a
11.	Wishing you were alone^a
12.	Proud^b
13.	Emotionally numb, without feelings^a
14.	Feeling sorry for yourself
15.	Inspired^b
16.	Brooding on things (thinking a long time about things that make you sad, worried, or angry)
17.	Active^b
18.	Overemotional
19.	Happy^a
20.	Changeable in your spirits
21.	Anxious
22.	Tired^a
23.	Confident^a
24.	Irritable
25.	Lively
26.	Oversensitive
27.	Tearful
28.	Up and down in mood
29.	Restless
30.	Calm, tranquil
31.	Guilty^b
32.	Interested^b
33.	Hostile^b
34.	Determined^b
35.	Ashamed^b
36.	Enthusiastic^b
37.	Afraid^b
38.	Loss of appetite

Items removed in subscale development.

Items from the Positive and Negative Affect Schedule (PANAS).

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

The authors received no financial support for the research, authorship, and/or publication of this article.

References

Ballinger

C. B.

Buckley

D. E.

Naylor

G. J.

Stansfield

D. A.

(1979). Emotional disturbance following childbirth: Clinical findings and urinary excretion of cyclic AMP. Psychological Medicine, 9, 293-300.

Bloch

Daly

R. C.

Rubinow

D. R.

(2003). Endocrine factors in the etiology of postpartum depression. Comprehensive Psychiatry, 44, 234-246.

Bloch

Rotenberg

Koren

Klein

(2005). Risk factors associated with the development of postpartum mood disorders. Journal of Affective Disorders, 88, 9-18.

Clark

L. A.

Watson

(1995). Constructing validity: Basic issues in objective scale development. Psychological Assessment, 7, 309-319.

Cordano

Scherer

R. F.

Owen

C. L.

(2003). Dimensionality of the Women as Managers scale: Factor congruency among three samples. Journal of Social Psychology, 143, 141-143.

Costello

A. B.

Osborne

J. W.

(2005). Best practices in exploratory factor analysis: Four recommendations for getting the most from your analysis. Practical Assessment, Research & Evaluation, 10, 1-9.

Floyd

Widaman

(1995). Factor analysis in the development and refinement of clinical assessment instruments. Psychological Assessment, 7, 286-299.

Glover

Kammerer

(2004). The biology and pathophysiology of peripartum psychiatric disorders. Primary Psychiatry, 11, 37-41.

Gonidakis

Rabavilas

A. D.

Varsou

Kreatsas

Christodoulou

G. N.

(2007). Maternity blues in Athens, Greece: A study during the first 3 days after delivery. Journal of Affective Disorders, 99, 107-115.

10.

Gorsuch

R. L.

(1983). Factor analysis (2nd ed.). Hillsdale, NJ: Erlbaum.

11.

Grice

J. W.

(2001). Computing and evaluating factor scores. Psychological Methods, 6, 430-450.

12.

Handley

S. L.

Dunn

T. L.

Waldron

Baker

J. M.

(1980). Tryptophan, cortisol and puerperal mood. British Journal of Psychiatry, 136, 498-508.

13.

Hannah

Adams

Lee

Glover

Sandler

(1992). Links between early postpartum mood and postnatal depression. British Journal of Psychiatry, 160, 777-780.

14.

Harris

Lovett

Newcombe

R. G.

Read

G. F.

Walker

Riad-Fahmy

(1994). Maternity blues and major endocrine changes: Cardiff puerperal mood and hormone study II. British Medical Journal, 308, 949-953.

15.

Hedges

Jandorf

Stone

(1985). Meaning of daily mood assessments. Journal of Personality and Social Psychology, 48, 428-434.

16.

Henshaw

(2003). Mood disturbance in the early puerperium: A review. Archives of Women’s Mental Health, 6, 33-42.

17.

Henshaw

Foreman

Cox

(2004). Postnatal blues: A risk factor for postnatal depression. Journal of Psychosomatic Obstetrics & Gynecology, 25, 267-272.

18.

Hogarty

K. Y.

Hines

C. V.

Kromrey

J. D.

Ferron

J. M.

Mumford

K. R.

(2005). The quality of factor solutions in exploratory factor analysis: The influence of sample size, communality, and overdetermination. Educational and Psychological Measurement, 65, 202-226.

19.

Holgado-Tello

F. P.

Chacón-Moscoso

Barbero-García

Vila-Abad

(2010). Polychoric versus Pearson correlations in exploratory and confirmatory factor analysis of ordinal variables. Quality & Quantity, 44, 153-166.

20.

Horn

J. L.

(1965). A rationale and test for the number of factors in factor analysis. Psychometrika, 30, 179-185.

21.

Iles

Gath

Kennerley

(1989). Maternity blues. II. A comparison between post-operative women and post-natal women. British Journal of Psychiatry, 155, 363-366.

22.

Kendell

R. E.

Mackenzie

W. E.

West

McGuire

R. J.

Cox

J. L.

(1984). Day-to-day mood changes after childbirth: Further data. British Journal of Psychiatry, 145, 620-625.

23.

Kendell

R. E.

McGuire

R. J.

Connor

Cox

J. L.

(1981). Mood changes in the first three weeks after childbirth. Journal of Affective Disorders, 3, 317-326.

24.

Kennerley

Gath

(1989). Maternity blues. I. Detection and measurement by questionnaire. British Journal of Psychiatry, 155, 356-362.

25.

Nappi

R. E.

Petraglia

Luisi

Polatti

Farina

Genazzani

A. R.

(2001). Serum allopregnanolone in women with postpartum “blues.” Obstetrics and Gynecology, 97, 77-80.

26.

Oakley

Chamberlain

(1981). Medical and social factors in post-partum depression. Journal of Obstetrics and Gynaecology, 1, 182-187.

27.

O’Connor

B. P.

(2000). SPSS and SAS programs for determining the number of components using parallel analysis and Velicer’s MAP test. Behavior Research Methods, Instruments, & Computers, 32, 396-402.

28.

O’Hara

M. W.

(1995). Postpartum depression. Causes and consequences. New York, NY: Springer.

29.

O’Hara

M. W.

Schlechte

J. A.

Lewis

D. A.

Wright

M. A.

(1991). Prospective study of postpartum blues: Biologic and psychosocial factors. Archives of General Psychiatry, 48, 801-806.

30.

Pitt

(1973). Maternity blues. British Journal of Psychiatry, 122, 431-433.

31.

Reck

Stehle

Reinig

Mundt

(2009). Maternity blues as a predictor of DSM-IV depression and anxiety disorders in the first three months postpartum. Journal of Affective Disorders, 113, 77-87.

32.

Sacher

Wilson

A. A.

Houle

Rusjan

Hassan

Bloomfield

P. M.

. . . Meyer

J. H.

(2010). Elevated brain Monoamine Oxidase A binding in the early postpartum period. Archives of General Psychiatry, 67, 468-474.

33.

Stein

G. S.

(1980). The pattern of mental change and body weight change in the first postpartum week. Journal of Psychosomatic Research, 24, 165-171.

34.

Stone

A. A.

Shiffman

(1994). Ecological momentary assessment (EMA) in behavorial medicine. Annals of Behavioral Medicine, 16, 199-202.

35.

Teel

Verran

J. A.

(1991). Focus on psychometrics. Factor comparison across studies. Research in Nursing and Health, 14, 67-72.

36.

Tucker

L. R.

(1971). Relations of factor score estimates to their use. Psychometrika, 36, 427-436.

37.

Vandenberg

R. J.

Lance

C. E.

(2000). A review and synthesis of the measurement invariance literature: Suggestions, practices and recommendations for organizational research. Organizational Research Methods, 3, 4-70.

38.

Watanabe

Wada

Sakata

Aratake

Kato

Ohta

Tanaka

(2008). Maternity blues as a predictor of postpartum depression: A prospective cohort study among Japanese women. Journal of Psychosomatic Obstetrics & Gynecology, 29, 206-212.

39.

Watson

Clark

L. A.

(1997). Measurement and mismeasurement of mood: Recurrent and emergent issues. Journal of Personality Assessment, 68, 267-296.

40.

Watson

Clark

L. A.

Tellegen

(1988). Development and validation of brief measures of positive and negative affect: The PANAS scales. Journal of Personality and Social Psychology, 54, 1063-1070.

41.

Watson

Tellegen

(1985). Toward a consensual structure of mood. Psychological Bulletin, 98, 219-235.

42.

Whiffen

V. E.

(1988). Vulnerability to postpartum depression: A prospective multivariate study. Journal of Abnormal Psychology, 97, 467-474.

43.

Wrigley

C. S.

Neuhaus

J. O.

(1955). The matching of two sets of factors. American Psychologist, 10, 418-419.

44.

Yalom

I. D.

Lunde

D. T.

Moos

R. H.

Hamburg

D. A.

(1968). Postpartum blues syndrome. A description and related variables. Archives of General Psychiatry, 18, 16-27.