Abstract
Maternal psychological factors, including anxiety, depression, and substance use, may negatively affect parenting. Previous works with mothers have often assessed each of these factors in isolation despite their frequent co-occurrence. Psychological factors have been associated with neural processing of facial stimuli, specifically the amplitude (i.e., size) and latency (i.e., timing) of the face-specific N170 event-related potential. In the current study, 106 mothers completed measures assessing maternal psychological factors - anxiety, depression, and substance use. A latent profile analysis was used to identify profiles of psychological factors and assess profile associations with the N170 elicited by infant faces and with parental reflective functioning as a measure related to caregiving. Two profiles (termed high and low psychological risk) were identified, with the higher risk profile associated with delayed N170 latency responses to infant faces. An exploratory analysis evidenced an indirect effect between the higher psychological risk profile and lower parental reflective functioning through delayed N170 latency responses to infant faces. Taken together, maternal psychological risk across multiple indicators may together shape neural processing of infant faces, which may have downstream consequences for caregiving.
Keywords: Mothers, Psychological Risk, Anxiety, Depression, Substance Use, Addictive Behaviors, EEG/ERP, Latent Profile Analysis
1. Introduction
Maternal psychological factors, such as anxiety, depression, and substance use (SU), may negatively affect parenting. Maternal anxiety has been associated with low parental warmth, self-efficacy, and high parental hostility (Drake & Ginsburg, 2011; Nicol-Harper et al., 2007; Seymour et al., 2015). Previous work has also found associations between maternal depression and less efficient emotional, motivational, and technical parental scaffolding, which is additionally associated with child behavioral problems and emotion dysregulation (Hoffman et al., 2006). Further, maternal SU and addiction have been linked with compromised caregiving, decreased maternal sensitivity, and lower emotional availability (Hatzis et al., 2017; Porreca et al., 2018).
It is important to emphasize that these forms of psychopathology frequently co-occur in the perinatal period. For example, among postpartum women in a population-based study, 18% reported anxiety, of whom 35% also reported depression (Farr et al., 2014). Another population-based study reported that 18.4% of postpartum women with an anxiety disorder were also diagnosed with a depressive disorder, and 33.9% of those with a depressive disorder were also diagnosed with an anxiety disorder (Reck et al., 2008). Further, among mothers with an SU disorder (SUD), reported co-occurrences were 27% for a depressive disorder and 13% for an anxiety disorder (Hser et al., 2015). Similarly, among a population-based sample of women of child-birthing age in the United States, 27.3% of women with both an anxiety disorder and recent major depressive episode, 20.9% with a recent major depressive episode alone, and 16.4% with an anxiety disorder alone also met criteria for a SUD (Zhou et al., 2019). Given high rates of co-occurrences among anxiety, depression, and SU concerns, it is critical to consider how patterns of postpartum psychological risk may impact caregiving at multiple levels of analysis. Comprehensive assessments of maternal psychological risk impacts on maternal neural processing and parenting may better inform treatment and prevention plans and mitigate emotion dysregulation and downstream impacts on caregiving.
Prior work using electroencephalography (EEG) to examine the impact of maternal psychological factors on parenting has focused on maternal neural reactivity to infant cues, including infant facial expressions. EEG data can be time-locked to a stimulus presentation to allow for timing (latency) and magnitude (amplitude) of the response to be assessed through event-related potentials (ERPs). The N170 is a face-specific ERP that has been widely examined in mothers and has been linked to maternal behavior (Maupin et al., 2015). Prior maternal ERP research has also associated N170 amplitude and latency measures with depression (Chen et al., 2014; Noll et al., 2012) and SU (Wall et al., 2022; Lowell et al., 2020; Rutherford et al., 2017) in mothers, and with anxiety in non-parenting studies (Cui et al., 2021; Wieser & Moscovitch, 2015). However, most of these studies have assessed these psychological factors individually rather than evaluating how these symptoms may co-occur across mothers and whether different patterns of symptoms may vary in their impact on the neural processing of infant faces. Therefore, in the current study, we conducted a latent profile analysis (LPA) to examine whether groups of postpartum mothers could be identified based on multiple measures of psychological risk that have individually shown associations with the N170, including depression, anxiety, and SU. We chose to take a person-centered approach by using LPA given its potential to identify patterns of psychological factors and SU among postpartum women (e.g., high anxiety and depression but low substance use). This technique differs from variable-centered approaches, such as creating a cumulative score, which assume population homogeneity (i.e., no sub-groups). Prior work using LPA to identify groups based on symptoms of psychopathology has demonstrated that LPA can identify patterns that uniquely relate to outcomes and offer different information than variable-centered approaches (e.g., Currier et al., 2021; Gerber et al., 2014; Hasselle et al., 2019; Li et al., 2021; Nguyen et al., 2022; Reinhardt et al., 2020; Vaughn et al., 2007; Yalçın et al., 2022).
After examining whether these psychological profiles related to mothers’ neural processing of infant facial cues as indexed by the N170, we then examined whether these profiles were differentially associated with a caregiving measure, parental reflective functioning (PRF), which refers to the mothers’ abilities/tendencies to understand their infants’ behaviors in terms of mental states (Slade, 2005; Slade et al., 2005). We examined PRF given separate bodies of work showing associations between maternal psychological factors and PRF (Katznelson, 2014) and the N170 and PRF (Rutherford et al., 2017b). Finally, as an exploratory approach, we investigated the N170 as a potential mechanism in the association between maternal psychological factors and PRF. We did not specify an a priori hypothesis for how many maternal psychological profiles the LPA would yield given the absence of prior literature. Based on known associations between psychological factors and the N170, it was hypothesized that the profiles would be associated with N170 amplitude and latency responses. Finally, given past associations between the N170 and caregiving, including PRF, it was hypothesized that the N170 would be associated with PRF, and that there would be an indirect effect of psychological profile on PRF via the N170.
2. Methods
2.1. Participants
One-hundred-and-six mothers (Mage=29.08 years, 29.08, SD=5.90) were recruited at approximately 8 months postpartum (SD=1.98) from the New Haven, Connecticut community as part of a larger study on the maternal brain and SU. This sample overlaps with those from two other studies, which focused on adult attachment and SU (Lowell et al., 2021, Wall et al., 2022). All participants provided written informed consent, and procedures were approved by the Yale University Human Investigation Committee. Of the 106 participants, 44 identified as African American, 32 as Caucasian, 16 as Hispanic/Latina, 3 as Asian/Asian American, and 11 reported other or did not report. Fifty-one participants reported having an annual income of less than $15,000, 17 reported $15,001-$30,000, 8 reported $30,001-$45,000, 14 reported $45,001-$70,000, 4 reported $70,001-$100,000, 8 reported more than $100,001 and 8 did not report. Fifty-seven mothers were multiparous, 47 were primiparous, and 2 did not report.
2.2. Measures
The State Trait Anxiety Inventory (STAI) (Spielberger, 1983) was used to measure anxiety and is comprised of two 20-item scales. Each item is scored 1 “never” to 5 “almost all of the time.” Higher scores indicate greater levels of anxiety. On the trait scale, participants are asked to respond based on how they generally feel, and on the state scale, participants are asked to respond based on how they feel at a particular moment in time. Possible scores for each scale range from 20 to 80 with scores of 20 to 37 indicating low or no anxiety, 38 to 44 indicating moderate anxiety, and 45 to 80 indicating high anxiety (Kayikcioglu et al., 2017). Given state and trait scores were strongly correlated (r=.739, p<.001), scores were averaged to generate a single anxiety score. Good internal consistency has been reported for STAI subscales: α=0.93 (state) and α=0.92 (trait) (Fountoulakis et al., 2006).
The Beck Depression Inventory-Second Edition (BDI-II) (A. T. Beck et al., 1996) was used to assess maternal depression. The 21 items are scored 0 to 3, with 0 representing an absence of the symptom and 3 indicating maximum symptom presence experienced in the past two weeks. Items are summed for a total score with higher scores indicating greater levels of depression. Scores of 0 to 9 are indicative of limited or no depression, scores ranging from 10 to 18 are indicative of mild to moderate depression, and scores of 19 or greater indicate moderate to severe depression (Beck et al., 1988). Previous reports indicate good internal consistency for this measure, α=0.89 (Sacco et al., 2016).
The Addiction Severity Index Lite (ASI Lite) (McLellan et al., 1980) is a widely used semi-structured interview assessing use of alcohol, amphetamines, barbiturates, cannabis, cocaine, hallucinogens, heroin, hypnotics, inhalants, methadone, sedatives, tranquilizers, and non-heroin opiates/analgesics (Cacciola et al., 2007). Current SU was quantified as the number of days in the last 30 days any substance was used (0–30) consistent with prior work (Wall et al., 2022). The ASI-Lite does not measure tobacco use; participants were asked one additional binary question about their tobacco use in the last week (i.e., currently using tobacco products, n= 29, or not currently using tobacco products, n= 77).
The Parental Reflective Functioning Questionnaire (PRFQ) (Luyten et al., 2017) is an 18-item measure that assesses PRF across three subscales. The pre-mentalizing (PRFQ-PM) subscale assesses non-mentalizing modes or difficulties reflecting on the mental states of one’s infant; the certainty about mental states (PRFQ-CMS) subscale captures an almost complete absence of certainty about the infant’s internal states at the low end to over-certainty at the high end of the scale; and the interest and curiosity (PRFQ-IC) subscale evaluates interest and curiosity about the infant’s mental states. Each item is scored from 1 (strong disagreement) to 7 (strong agreement). Lower scores on PRFQ-PM, and higher scores on PRFQ-CMS and PRFQ-IC, generally indicate better PRF. Good internal consistency (α=.70-.82) has been reported for PRFQ subscales (Luyten et al., 2017).
2.3. Apparatus and Stimuli
Experimental apparatus and stimuli have been described previously (Lowell et al., 2021; Wall et al., 2022). EEG was recorded continuously using NetStation 4.2.1 with a 250 Hz sampling rate and high impedance amplifiers (0.1 Hz high pass, 100 Hz low pass) and a 128 Ag/AgCl electrode potassium chloride-soaked net (Electrical Geodesics, Inc.) with evenly and symmetrical spaced electrodes. Cz was used as the reference electrode during EEG acquisition. Electrode impedances were kept less than 40kΩ.
All stimuli were presented on a 51 cm color monitor (75 Hz, 1024 × 768 resolution) through a Pentium IV computer running E-Prime 1.2 software (Schneider et al., 2002) in an environment with sound-attenuation and dim light. Stimuli consisted of 40 infant faces presented randomly (10 happy own, 10 happy unknown, 10 sad own, 10 sad unknown) and 20 infant cries (10 own, 10 unknown). Each face and cry stimulus was presented 4 times for a total of 240 trials. Given the focus on the N170, analyses focused only on face trials. Trained staff classified infant facial expressions as happy or sad using previously validated procedures (Cole et al., 1992; Kim et al., 2017). Happy infant faces were generated from recordings of the infant playing with age-appropriate toys, and sad infant faces were generated from recordings of brief separation periods from the infant’s mother. Unfamiliar infant faces were matched to the mother’s own infant for sex, race, and age. Images were cropped to contain only the infant’s face (approximately 11cm × 11cm) and placed on black background viewed from approximately 70cm. Participants completed 12 practice trials and 4 blocks of 60 experimental trials. Each trial consisted of presentation of a fixation cross (2000ms), the face stimulus (500ms), and a blank screen (jittered 1400–2000ms).
a. Procedure
Mothers completed the BDI-II, PRFQ, and the ASI Lite and infant recordings were taken during the enrollment visit. Mothers completed the STAI and the ERP task at a second lab visit on average 34.8 days (SD=31.4) later.
EEG data were preprocessed using Net Station 4.5. Data were digitally filtered (30 Hz low-pass filter) and segmented into 1s epochs (100ms pre- and 900ms post-stimulus onset). Ocular Artifact Removal, with a blink slope threshold of 14 μV/ms, was used and artifact detection was: 200 μV for bad channels, 150 μV for eye-blinks, and 150 μV for eye-movements. If there were artifacts in more than 40% of trials, spline interpolation was used to replace channels. All EEG data were then re-referenced to the average reference of all electrodes and baseline corrected to 100ms pre-stimulus onset. Lastly, the EEG data were averaged across stimulus conditions for each participant. Pre-processing resulted in an average of 33.31 trials per condition across all participants (SD=6.02 trials; Happy Own M=33.56, SD=5.95; Happy Unknown M=33.71, SD=5.96; Sad Own M=33.38, SD=6.13; Sad Unknown M=33.49, SD=5.73).
N170 mean amplitude and latency (134ms–201ms) were assessed at temporal electrode sites (58,59,64,65,69,70,90,91,92,95,96,97) utilized in prior work (Lowell et al., 2021; Rutherford et al., 2017a). There were no main effects or interactions of infant familiarity, emotion, or hemisphere on N170 mean amplitude or peak latency (Wall et al., 2022). The data were therefore averaged across conditions to generate single N170 values for amplitude and latency (Figure 1a).
Figure 1.
A - Grand averaged ERP waveforms representing the average N170 response to infant faces across conditions. B - Latent profile subscale means for continuous indicator variables. C - Standardized estimates and confidence intervals from exploratory mediation model. Maternal race and household income were included as covariates. Maternal risk profile: 0, low-risk profile; 1, high-risk profile; *p < .05. Note. Tobacco use (dichotomous variable) was also an indicator of profile membership but is not depicted in the figure; 19.6% of the low-risk class and 78.6% of the high-risk class reported current tobacco use.
2.6. Data Analysis
Descriptive and bivariate statistics were first analyzed. Next, LPA was used to identify maternal psychological profiles. LPA is a person-centered analysis that identifies sub-groups within a sample based on their responses across multiple indicator variables. Comparable sample sizes to this study have been used in previous LPA research (Karlson et al., 2020; Spurk et al., 2020). Indicator variables were the STAI Anxiety composite, BDI-II depression score, ASI-Lite current SU, and current tobacco use. LPA was conducted using Mplus 8.6 (Muthén & Muthén, 2017). There were no missing data for main study variables.
The LPA was run for one to five profiles using maximum likelihood estimation (MLE) with 150 random starting values. Fit statistics examined included entropy, bootstrap likelihood ratio test (BLRT), the Lo-Mendel-Ruben likelihood ratio test (LMR), the Akaike Information Criteria (AIC), the Bayesian Information Criteria (BIC), and the adjusted Bayesian Information Criteria (aBIC) (Nylund et al., 2007).
Once an optimal number of profiles was selected based on fit statistics, profiles were evaluated as to whether they differed based on demographics, N170 latencies and amplitudes, and PRFQ subscales. The Lanza method (2013) was used for comparing profiles on categorical variables and the Bolck-Croon-Hagenaars (BCH) approach (Bakk et al., 2013) for continuous variables (Asparouhov & Muthén, 2014). For ease of interpretation, the multi-categorical variables were dichotomized after examining frequencies. Final categorical variables were race (1=Black or African American, 0=All other races), income (1=Income less than $15000, 0=Income $15000 or greater), and parity (1=multiparous, 0=primiparous). Continuous variables compared across profiles were age, N170 latency, N170 amplitude, PRFQ-PM, PRFQ-CMS, and PRFQ-IC.
2.6.2. Exploratory analysis
To evaluate the N170 as a potential mechanism in the association between maternal psychological profile and caregiving, an exploratory mediation was conducted. The model included profile membership as the independent variable (x), N170 (latency or amplitude) as the mediator variable (m), and PRFQ subscale as the dependent variable (y). Mediation analysis was conducted in MPlus using MLE.
3. Results
3.1. Preliminary Analyses
Descriptive and bivariate statistics for continuous variables are presented in Table 1. Maternal anxiety, depression, and SU were positively correlated with one another and with N170 latency. A positive correlation was observed between anxiety and PRFQ-PM. Mothers who currently smoked reported greater levels of other SU than mothers who did not (M=12.31 days and M = 2.47 days, respectively) (t(104)=−5.171, p<.001). Mothers who currently smoked also reported greater levels of depression than mothers who did not (M=13.69 and M=6.66, respectively) (t(104)=−4.285, p<.001).
Table 1.
Means, SDs, and Correlations for Main Study Variables
Mean | SD | Range | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | ||
---|---|---|---|---|---|---|---|---|---|---|---|---|
| ||||||||||||
1 | Anxiety (STAI) | 67.17 | 17.42 | 40–112 | 1 | |||||||
2 | Depression (BDI-II) | 8.58 | 8.13 | 0–37 | .430** | 1 | ||||||
3 | Substance Use (ASI-Lite) | 5.16 | 9.75 | 0–30 | .294** | .518** | 1 | |||||
4 | PRFQ-PM | 1.77 | 1.03 | 1–5.6 | .274** | .001 | .005 | 1 | ||||
5 | PRFQ-CMS | 4.11 | 1.10 | 1–7.0 | −.124 | −.115 | .088 | .075 | 1 | |||
6 | PRFQ-IC | 5.75 | 1.06 | 2.3–7.0 | −.031 | .181 | −0.04 | −.198* | .091 | 1 | ||
7 | N170 Latency | 164.30 | 21.38 | 65.83–216.70 | .200* | .195* | .253** | .033 | −.225* | −.087 | 1 | |
8 | N170 Amplitude | −1.28 | 1.45 | −6.34–1.74 | −.032 | −.075 | .024 | −.079 | .169 | .181 | −.042 | 1 |
Significance at the 0.01 level (2-tailed)
Significance at the 0.05 level (2-tailed)
STAI = State Trait Anxiety Index; BDI-II = Beck Depression Inventory; ASI-Lite = Addiction Severity Index Lite; PRFQ-PM = Parental Reflective Functioning Questionnaire pre-mentalizing subscale; PRFQ-CMS = Parental Reflective Functioning Questionnaire certainty about mental states subscale; PRFQ-IC = Parental Reflective Functioning Questionnaire interest and curiosity subscale
3.2. Latent Profiles
LPA fit indices are presented in Table 2. All models had an entropy above 0.97. AIC, BIC, and aBIC decreased with increasing number of profiles in each model, supporting models with more classes; however, the LMR test supported the two-profile model over models with more classes (p=.011). Additionally, the BLRT test was statistically significant at p<0.05 for all models. When reviewing models with four and five profiles, the number of participants in each class fell below 5% of the total sample. When comparing two- and three-profile models, based on the higher entropy value and the significant LMR test in the two-profile model, the two-profile model was selected (Figure 1b).
Table 2.
LPA Model Fit Indices
Parsimony Criteria | p value | |||||
---|---|---|---|---|---|---|
|
||||||
Profiles | Entropy | AIC | BIC | aBIC | BLRT | LMR |
| ||||||
1 | --- | 2570.589 | 2589.233 | 2567.118 | --- | --- |
2 | 1.000 | 2327.080 | 2359.041 | 2321.128 | <0.001 | 0.011 |
3 | 0.981 | 2291.595 | 2336.873 | 2283.164 | <0.001 | 0.508 |
4 | 0.980 | 2250.075 | 2308.671 | 2239.165 | <0.001 | 0.436 |
5 | 0.975 | 2228.823 | 2300.736 | 2215.433 | 0.013 | 0.611 |
In the two-profile model, the first profile, referred to as the low-risk profile (86.79%; n=92), participants had lower depression and anxiety scores (BDI-II M=7.09, SD=0.68, range=0–32; STAI Composite M=32.62, SD=0.83, range=20–56) and lower levels of SU (M=1.51 days, SD=0.27, range=0–15) and tobacco use (19.6% using tobacco). In the second profile, referred to as the high-risk profile (13.20%; n=14), participants had overall higher psychological risk (BDI-II M=18.43, SD=2.78, range=3–37; STAI Composite M=39.93, SD=2.80, range=22–54), with higher levels of SU (M=29.14 days, SD=0.69, range=20–30) and tobacco use (78.60% using tobacco).
3.2. Latent Profile Differences
We next evaluated whether profiles differed based on mothers’ ERP values (N170 latency, N170 amplitude), PRFQ subscale scores (PRFQ-PM, PRFQ-CMS, PRFQ-IC), and demographic factors (race and ethnicity, annual household income, parity, age). Two N170 latency scores were identified as outliers and removed from ERP analysis. There was a significant difference between profiles for N170 latency (X2(1, N=104) =11.960; p=.001), but not N170 amplitude (X2(1, N=104) =0.580; p=.446). Specifically, mothers with a high-risk profile had a delayed N170 latency (M=179.20ms, S.E=4.02) compared to mothers with a low-risk profile (M=164.16ms, S.E=1.67). Regarding parental measures, there were no significant differences between the two profiles on any of the PRFQ scales. Regarding demographic variables, profiles differed by race and annual household income, but not age or parity. The high-risk profile had less representation of mothers who identified as African American (14.3%) than the low-risk profile (47.7%) (X2(1, N=102) =9.689; p=.002), and higher representation of mothers with an annual household income of less than $15,000 (85.7%) compared to the low-risk profile (44.3%) (X2(1, N=102)=14.860; p<.001).
3.3. Exploratory Mediation Analysis
Given associations observed between risk profile membership and N170 latency, and N170 latency and PRFQ-CMS, we conducted an exploratory mediation analysis testing for an indirect association between risk profile membership and PRF, as assessed by the PRFQ-CMS, via N170 latency. Annual household income and race were included as covariates given that they differed between profiles. For mediation analysis, we hard-classified each participant based on their most probable risk profile, so that each participant was assigned 0=Low-risk class, or 1=High-risk class. Results indicated a direct effect of risk profile on N170 latency (β=.31, 95% CI[.12, .51], SE=.10, p=.002), and a direct effect of N170 latency on PRFQ-CMS (β=−.27, 95% CI[−.46, −.09], SE=.09, p=.004). There was no significant direct effect between risk profile and PRFQ-CMS scores (β=.12, 95% CI[−.09, .33], SE=.11, p=.25), but there was a significant indirect effect (β=−.09, 95% CI[−.17,−.01], SE=.04, p=.037) such that membership in the high-risk profile was indirectly associated with lower PRFQ-CMS scores via delayed N170 latencies (Figure 1c).
Discussion
Maternal psychological factors including anxiety, depression, and SU are associated with parenting, including decreased sensitivity, warmth, emotional availability, and PRF, as well as increased parental hostility (Drake & Ginsburg, 2011; Hatzis et al., 2017; Hoffman et al., 2006; Katznelson, 2014; Nicol-Harper et al., 2007; Seymour et al., 2015). However, previous assessments of maternal risk have often focused on one aspect of risk in isolation, despite high rates of co-occurrence between anxiety, depression, and SU (Back & Brady, 2008; Davis et al., 2008; Farr et al., 2014; Field et al., 2010; Hser et al., 2015; Lamers et al., 2011; Reck et al., 2008; Zhou et al., 2019). Changes in the N170 ERP response to caregiving cues have also been associated with maternal (and non-maternal) psychological risk (Chen et al., 2014; Cui et al., 2021; Lowell et al., 2020; Noll et al., 2012; Rutherford et al., 2017a; Wieser & Moscovitch, 2015). We evaluated whether distinct profiles of postpartum mothers could be identified based on their anxiety, depression, and SU, and compared N170 response, parental reflective functioning, and demographic factors between profiles. The LPA identified two profiles, one with overall higher levels (i.e., elevated levels of anxiety, depression, and SU) and the other with overall lower levels (i.e., lower levels of anxiety, depression, and SU) of psychological factors. The low-risk profile had an average BDI score of 7.09, indicative of limited to no depression whereas the high-risk profile had a higher mean BDI score of 18.43, indicative of moderate to severe depression (Beck et al., 1988). Likewise, participants in the low-risk profile had an average STAI score of 32.62 suggesting low or no anxiety whereas the high-risk profile had a higher average STAI score of 39.93, indicating moderate anxiety (Kayikcioglu et al., 2017). Further, the participants in the high-risk profile reported SU an average of 29 days per month, compared to participants in the low-risk group reporting SU an average of 1.51 days per month. Moreover, mothers in the high-risk group were more likely to report using tobacco (78.6% of group) compared to mothers in the low-risk group (19.6%). In sum, mothers in the high-risk group were more likely to score in the clinical range for measures of anxiety and depression and to use substances on a more frequent basis than mothers in the low-risk profile. There was also a significant difference between profiles for N170 latency such that mothers in the high-risk class had a delayed N170 response compared to mothers in the low-risk class, suggesting less efficient neural processing of infant faces in mothers belonging to the high-risk group. Therefore, having high levels of psychological risk may be associated with a delay in the speed of processing infant cues during the postpartum period, which may have downstream consequences for caregiving.
When comparing psychological profiles for annual household income, mothers with an annual household income of less than $15,000 were more likely to be in the high-risk profile. Previous work has also identified low socio-economic status as a potential risk factor for anxiety, depression, and SU and addiction (Bassuk et al., 1998; Sareen et al., 2011). Results additionally suggested higher likelihood of membership in the high-risk profile among mothers identifying as a race other than African American. This result may be related to heterogeneity among non-African American participants, making the result difficult to interpret.
While the two profiles did not differ in levels of PRF, the PRFQ-CMS and N170 latency were correlated. We further explored the N170 as a mediator between psychological profile and this measure of parenting in a mediation analysis. We found that higher-risk profile membership was indirectly associated with less optimal mentalizing, as indicated by lower PRFQ-CMS scores, via delayed N170 latency, suggesting N170 latency as a potential link between postpartum psychological risk and caregiving behavior. This finding provides preliminary evidence for neural processing of caregiving cues as one mechanism that may explain the link between maternal psychological risk and caregiving outcomes.
It is important to consider these findings in the context of their limitations and directions for future work. One limitation to consider is that the model of best fit in the latent profile analysis contained two profiles, one in which participants scored highly across all measures and one in which participants scored low on all measures. Relatedly, the high-risk profile contained a minority of participants (13.2%). Prior research using LPA to identify subgroups based on psychopathology has also yielded two-class solutions similar in terms of containing “high” and “low” groups (Currier et al., 2021). While a three-profile class with high anxiety and low depression and substance/tobacco use was observed, the two-profile solution was better supported by fit statistics, which suggested very strong separation (entropy = 1) between the two classes. Nevertheless, it is possible that this lack of emergence of a more nuanced third profile was due to our sample size, which is on the lower end for LPA studies (Wurpts & Geiser, 2014). It is important for future studies with larger sample sizes to evaluate whether additional, more nuanced profiles of psychopathology (i.e., high scores in some symptoms domains and low in others) emerge when examining psychopathology and SU in postpartum populations, or whether only a high-risk and low-risk group again emerge, to better understand potential patterns of psychological risk during this period and their effects on ERPs and caregiving.
Further, given that the two-profile solution supported by our data was comprised of a low-risk and high-risk group, it is possible that a continuous or composite measure of overall psychopathology would have yielded similar findings. We chose to use LPA as a person-centered approach to understand whether there were risk subgroups based on scores across multiple clinical measures – in contrast to a continuous or composite measure that would represent a variable-centered approach which does not offer an opportunity to examine risk subgroups. By using LPA, we were able to identify a group in our sample that reported high levels of SU, depression, and anxiety, and find that this group was more likely to have a delayed N170 in response to caregiving cues. The presence of a group with high SU and psychiatric comorbidity, alongside their differential neural response to caregiving cues, is aligned with prior literature (Cui et al., 2021; Hser et al., 2015; Noll et al., 2012; Wall et al., 2022), and underlines the need for support for this high-risk population.
Also worth noting as a limitation is that our psychological factors were all self-report assessments, which may be subject to recall error and/or responder bias given the stigma associated with SU and mental health concerns during the postpartum period. Future work should also incorporate forms of measurement other than self-report such as behavioral assessments of parenting and interview measures of psychopathology and SU. An additional limitation of this work is the duration between the enrollment visit, where two of the psychological factors were assessed, and the EEG/ERP visit, as symptoms may have changed during this period. Concurrent assessment of self-report and neural measures are warranted as this work continues. Moreover, a limitation in regard to our post-hoc mediation analysis is that we did not have data from multiple time points to evaluate the mediation effects over time (i.e., with psychological risk collected prior to EEG/ERP, followed by PRF). Future work should examine neural responses as a mediator of the association between psychological concerns and caregiving using longitudinal data. It is also important to consider that our stimuli conditions were limited to infant faces. While the results suggest high levels of maternal psychological risk may be associated with delayed processing of infant cues and downstream consequences for caregiving, it is also possible this result extrapolates to faces more generally (including adult faces). Therefore, future work should consider additional face stimuli (e.g., own and unknown adult faces) as well as neutral stimuli (e.g., houses) to further discriminate if maternal psychological risk in the postpartum period impacts infant face processing specifically or face processing more generally. Additionally, of note, we did not identify associations between infant emotional expression or familiarity and the N170. Previous literature assessing associations between familiarity and emotion content and the N170 response are mixed. However, variability in experimental approaches may contribute to reported differences. For example, studies involving passive viewing frequently no emotion effects on the N170 in mothers (Lowell et al., 2020, 2021; Malak et al., 2015; Maupin et al., 2019; Rutherford, Byrne, et al., 2017; Rutherford et al., 2019; Rutherford, Maupin, et al., 2017; Wall et al., 2022), whereas emotion recognition tasks in which participants categorize infant emotions during each experimental trial have frequently reported associations between infant facial expression and maternal N170 responses (Bernard et al., 2018; Ma et al., 2017; Márquez et al., 2019; Proverbio et al., 2006). Lowell et al., (2021) posits that these findings may reflect intentional directing of attention to the infant face stimuli required by emotion recognition tasks, and future work directly comparing different experimental tasks is required to support this proposal.
In conclusion, we employed LPA to understand patterns of psychological risk during the perinatal period and whether different patterns of risk may demonstrate differences in maternal neural processing of infant facial cues and in one aspect of caregiving, parental reflective functioning. Mothers in the profile represented by high levels of SU, depression, and anxiety, and by higher likelihood of tobacco use, were those who had a delayed N170 response to infant faces and lower levels of certainty about their infant’s mental states. These findings suggest the importance of considering the effects of multiple forms of psychopathology when understanding the effects of maternal psychological risk on neural response to infant cues and on caregiving.
Acknowledgements:
We would like to thank all the mothers that participated.
Funding sources:
This work was supported by grants from the National Institutes of Health [R01 DA026437-08, T32 MH018268, F32 DA055389, T32 NS041228]. The views presented in this manuscript are those of the authors and do not necessarily reflect those of the funding agencies.
Footnotes
Conflict of Interest Disclosure: Kathryn M. Wall, Francesca Penner, Jaclyn Dell, Amanda Lowell, Linda C. Mayes, and Helena J. V. Rutherford declare no conflicts of interest. Marc N. Potenza has consulted for and advised Opiant Pharmaceuticals, Idorsia Pharmaceuticals, AXA, Game Day Data and the Addiction Policy Forum; has been involved in a patent application with Yale University and Novartis; has received research support from the Mohegan Sun Casino and Connecticut Council on Problem Gambling; has participated in surveys, mailings or telephone consultations related to drug addiction, impulse control disorders or other health topics; and has consulted for law offices and gambling entities on issues related to impulse control or addictive disorders.
Ethics approval: All participants provided written informed consent, and all procedures were approved by the Yale University Human Investigation Committee.
Data availability statement:
Research data are not shared.
References
- Asparouhov T, & Muthén B (2014). Auxiliary Variables in Mixture Modeling: Three-Step Approaches Using Mplus. Structural Equation Modeling: A Multidisciplinary Journal, 21(3), 329–341. 10.1080/10705511.2014.915181 [DOI] [Google Scholar]
- Back SE, & Brady KT (2008). Anxiety Disorders with Comorbid Substance Use Disorders: Diagnostic and Treatment Considerations. Psychiatric Annals, 38(11), 724–729. 10.3928/00485713-20081101-01 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Bakk Z, Tekle FB, & Vermunt JK (2013). Estimating the Association between Latent Class Membership and External Variables Using Bias-adjusted Three-step Approaches. Sociological Methodology, 43(1), 272–311. 10.1177/0081175012470644 [DOI] [Google Scholar]
- Bassuk EL, Buckner JC, Perloff JN, & Bassuk SS (1998). Prevalence of Mental Health and Substance Use Disorders Among Homeless and Low-Income Housed Mothers. American Journal of Psychiatry, 155(11), 1561–1564. 10.1176/ajp.155.11.1561 [DOI] [PubMed] [Google Scholar]
- Beck A, Steer R, & Carbin. (1988). Psychometric properties of the Beck Depression Inventory: Twenty-five years of evaluation. Clinical Psychology Review, 8(1), 77–100. 10.1016/0272-7358(88)90050-5 [DOI] [Google Scholar]
- Beck AT, Steer RA, Ball R, & Ranieri WF (1996). Comparison of Beck Depression Inventories-IA and-II in Psychiatric Outpatients. Journal of Personality Assessment, 67(3), 588–597. 10.1207/s15327752jpa6703_13 [DOI] [PubMed] [Google Scholar]
- Cacciola JS, Alterman AI, McLellan AT, Lin Y-T, & Lynch KG (2007). Initial evidence for the reliability and validity of a “Lite” version of the Addiction Severity Index. Drug and Alcohol Dependence, 87(2), 297–302. 10.1016/j.drugalcdep.2006.09.002 [DOI] [PubMed] [Google Scholar]
- Chen J, Ma W, Zhang Y, Wu X, Wei D, Liu G, Deng Z, Yang L, & Zhang Z (2014). Distinct Facial Processing Related Negative Cognitive Bias in First-Episode and Recurrent Major Depression: Evidence from the N170 ERP Component. PLOS ONE, 9(10), e109176. 10.1371/journal.pone.0109176 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Cole PM, Barrett KC, & Zahn-Waxler C (1992). Emotion displays in two-year-olds during mishaps. Child Development, 63(2), 314–324. 10.2307/1131481 [DOI] [PubMed] [Google Scholar]
- Cui L, Dong X, & Zhang S (2021). ERP evidence for emotional sensitivity in social anxiety. Journal of Affective Disorders, 279, 361–367. 10.1016/j.jad.2020.09.111 [DOI] [PubMed] [Google Scholar]
- Currier JM, Foster JD, Karatzias T, & Murphy D (2021). Moral injury and ICD-11 complex PTSD (CPTSD) symptoms among treatment-seeking veterans in the United Kingdom. Psychological Trauma: Theory, Research, Practice and Policy, 13(4), 417–421. 10.1037/tra0000921 [DOI] [PubMed] [Google Scholar]
- Davis L, Uezato A, Newell JM, & Frazier E (2008). Major depression and comorbid substance use disorders. Current Opinion in Psychiatry, 21(1), 14–18. 10.1097/YCO.0b013e3282f32408 [DOI] [PubMed] [Google Scholar]
- Drake KL, & Ginsburg GS (2011). Parenting Practices of Anxious and Nonanxious Mothers: A Multi-Method, Multi-Informant Approach. Child & Family Behavior Therapy, 33(4), 299–321. 10.1080/07317107.2011.623101 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Farr SL, Dietz PM, O’Hara MW, Burley K, & Ko JY (2014). Postpartum Anxiety and Comorbid Depression in a Population-Based Sample of Women. Journal of Women’s Health, 23(2), 120–128. 10.1089/jwh.2013.4438 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Field T, Diego M, Hernandez-Reif M, Figueiredo B, Deeds O, Ascencio A, Schanberg S, & Kuhn C (2010). Comorbid depression and anxiety effects on pregnancy and neonatal outcome. Infant Behavior and Development, 33(1), 23–29. 10.1016/j.infbeh.2009.10.004 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Fountoulakis KN, Papadopoulou M, Kleanthous S, Papadopoulou A, Bizeli V, Nimatoudis I, Iacovides A, & Kaprinis GS (2006). Reliability and psychometric properties of the Greek translation of the State-Trait Anxiety Inventory form Y: Preliminary data. Annals of General Psychiatry, 5(1), 2. 10.1186/1744-859X-5-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Hatzis D, Dawe S, Harnett P, & Barlow J (2017). Quality of Caregiving in Mothers With Illicit Substance Use: A Systematic Review and Meta-analysis. Substance Abuse: Research and Treatment, 11, 1178221817694038. 10.1177/1178221817694038 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Hoffman C, Crnic KA, & Baker JK (2006). Maternal Depression and Parenting: Implications for Children’s Emergent Emotion Regulation and Behavioral Functioning. Parenting, 6(4), 271–295. 10.1207/s15327922par0604_1 [DOI] [Google Scholar]
- Hser Y-I, Lanza HI, Li L, Kahn E, Evans E, & Schulte M (2015). Maternal Mental Health and Children’s Internalizing and Externalizing Behaviors: Beyond Maternal Substance Use Disorders. Journal of Child and Family Studies, 24(3), 638–648. 10.1007/s10826-013-9874-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Karlson CW, Sarver DE, Raiker JS, Espil FM, Cox AD, Elkin TD, & Annett RD (2020). The contribution of neurocognitive functions to academic and psychological outcomes in pediatric cancer: A latent profile analysis. Child Neuropsychology: A Journal on Normal and Abnormal Development in Childhood and Adolescence, 26(7), 881–899. 10.1080/09297049.2020.1734553 [DOI] [PubMed] [Google Scholar]
- Katznelson H (2014). Reflective functioning: A review. Clinical Psychology Review, 34(2), 107–117. 10.1016/j.cpr.2013.12.003 [DOI] [PubMed] [Google Scholar]
- Kayikcioglu O, Bilgin S, Seymenoglu G, & Deveci A (2017). State and Trait Anxiety Scores of Patients Receiving Intravitreal Injections. Biomedicine Hub, 2(2), 1–5. 10.1159/000478993 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Kim S, Iyengar U, Mayes LC, Potenza MN, Rutherford HJV, & Strathearn L (2017). Mothers with substance addictions show reduced reward responses when viewing their own infant’s face. 10.1002/hbm.23731 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Lamers F, van Oppen P, Comijs HC, Smit JH, Spinhoven P, van Balkom AJLM, Nolen WA, Zitman FG, Beekman ATF, & Penninx BWJH (2011). Comorbidity patterns of anxiety and depressive disorders in a large cohort study: The Netherlands Study of Depression and Anxiety (NESDA). The Journal of Clinical Psychiatry, 72(3), 341–348. 10.4088/JCP.10m06176blu [DOI] [PubMed] [Google Scholar]
- Lanza ST, Tan X, & Bray BC (2013). Latent Class Analysis With Distal Outcomes: A Flexible Model-Based Approach. Structural Equation Modeling: A Multidisciplinary Journal, 20(1), 1–26. 10.1080/10705511.2013.742377 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Lowell AF, Dell J, Potenza MN, Strathearn L, Mayes LC, & Rutherford HJV (2021). Adult attachment is related to maternal neural response to infant cues: An ERP study. Attachment & Human Development, 0(0), 1–18. 10.1080/14616734.2021.1880057 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Lowell AF, Maupin AN, Landi N, Potenza MN, Mayes LC, & Rutherford HJV (2020). Substance use and mothers’ neural responses to infant cues. Infant Mental Health Journal, 41(2), 264–277. 10.1002/imhj.21835 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Luyten P, Mayes LC, Nijssens L, & Fonagy P (2017). The parental reflective functioning questionnaire: Development and preliminary validation. PLOS ONE, 12(5), e0176218. 10.1371/journal.pone.0176218 [DOI] [PMC free article] [PubMed] [Google Scholar]
- McLellan AT, Luborsky L, Woody GE, & O’Brien CP (1980). An improved diagnostic evaluation instrument for substance abuse patients. The Addiction Severity Index. The Journal of Nervous and Mental Disease, 168(1), 26–33. 10.1097/00005053-198001000-00006 [DOI] [PubMed] [Google Scholar]
- Muthén LK, & Muthén B, O. (2017). Mplus: Statistical Analysis with Latent Variables: User’s Guide (Version 8). https://www.scirp.org/(S(czeh2tfqyw2orz553k1w0r45))/reference/ReferencesPapers.aspx?ReferenceID=2123077
- Nicol-Harper R, Harvey AG, & Stein A (2007). Interactions between mothers and infants: Impact of maternal anxiety. Infant Behavior and Development, 30(1), 161–167. 10.1016/j.infbeh.2006.08.005 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Noll LK, Mayes LC, & Rutherford HJV (2012). Investigating the impact of parental status and depression symptoms on the early perceptual coding of infant faces: An event-related potential study. Social Neuroscience, 7(5), 525–536. 10.1080/17470919.2012.672457 [DOI] [PubMed] [Google Scholar]
- Nylund KL, Asparouhov T, & Muthén BO (2007). Deciding on the Number of Classes in Latent Class Analysis and Growth Mixture Modeling: A Monte Carlo Simulation Study. Structural Equation Modeling: A Multidisciplinary Journal, 14(4), 535–569. 10.1080/10705510701575396 [DOI] [Google Scholar]
- Porreca A, Biringen Z, Parolin M, Saunders H, Ballarotto G, & Simonelli A (2018). Emotional Availability, Neuropsychological Functioning, and Psychopathology: The Context of Parental Substance Use Disorder. BioMed Research International, 2018, e5359037. 10.1155/2018/5359037 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Reck C, Struben K, Backenstrass M, Stefenelli U, Reinig K, Fuchs T, Sohn C, & Mundt C (2008). Prevalence, onset and comorbidity of postpartum anxiety and depressive disorders. Acta Psychiatrica Scandinavica, 118(6), 459–468. 10.1111/j.1600-0447.2008.01264.x [DOI] [PubMed] [Google Scholar]
- Rutherford HJV, Maupin AN, Landi N, Potenza MN, & Mayes LC (2017a). Current Tobacco-Smoking and Neural Responses to Infant Cues in Mothers. Parenting, 17(1), 1–10. 10.1080/15295192.2017.1262176 [DOI] [Google Scholar]
- Rutherford HJV, Maupin AN, Landi N, Potenza MN, & Mayes LC (2017b). Parental reflective functioning and the neural correlates of processing infant affective cues. Social Neuroscience, 12(5), 519–529. 10.1080/17470919.2016.1193559 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Sacco R, Santangelo G, Stamenova S, Bisecco A, Bonavita S, Lavorgna L, Trojano L, D’Ambrosio A, Tedeschi G, & Gallo A (2016). Psychometric properties and validity of Beck Depression Inventory II in multiple sclerosis. European Journal of Neurology, 23(4), 744–750. 10.1111/ene.12932 [DOI] [PubMed] [Google Scholar]
- Sareen J, Afifi TO, McMillan KA, & Asmundson GJG (2011). Relationship Between Household Income and Mental Disorders: Findings From a Population-Based Longitudinal Study. Archives of General Psychiatry, 68(4), 419–427. 10.1001/archgenpsychiatry.2011.15 [DOI] [PubMed] [Google Scholar]
- Schneider W, Eschman A, & Zuccolotto A (2002). E-Prime User’s Guide. Psychology Software Incorporated. [Google Scholar]
- Seymour M, Giallo R, Cooklin A, & Dunning M (2015). Maternal anxiety, risk factors and parenting in the first post-natal year. Child: Care, Health and Development, 41(2), 314–323. 10.1111/cch.12178 [DOI] [PubMed] [Google Scholar]
- Slade A (2005). Parental reflective functioning: An introduction. Attachment & Human Development, 7(3), 269–281. 10.1080/14616730500245906 [DOI] [PubMed] [Google Scholar]
- Slade A, Grienenberger J, Bernbach E, Levy D, & Locker A (2005). Maternal reflective functioning, attachment, and the transmission gap: A preliminary study. Attachment & Human Development, 7(3), 283–298. 10.1080/14616730500245880 [DOI] [PubMed] [Google Scholar]
- Spielberger CD (1983). State-Trait Anxiety Inventory for Adults—PsycNET. APA PsycTests. https://psycnet.apa.org/doiLanding?doi=10.1037%2Ft06496-000 [Google Scholar]
- Spurk D, Hirschi A, Wang M, Valero D, & Kauffeld S (2020). Latent profile analysis: A review and “how to” guide of its application within vocational behavior research. Journal of Vocational Behavior, 120, 103445. 10.1016/j.jvb.2020.103445 [DOI] [Google Scholar]
- Wall Dell J., Lowell A., Potenza M., Mayes L., & Rutherford HJ. (2022). Current substance use and maternal neural responses to infant faces and cries. Journal of Mental Health and Addiction. [Google Scholar]
- Wang Y-P, & Gorenstein C (2021). Chapter 16 - The Beck depression inventory: Uses and applications. In Martin CR, Hunter L-A, Patel VB, Preedy VR, & Rajendram R (Eds.), The Neuroscience of Depression (pp. 165–174). Academic Press. 10.1016/B978-0-12-817933-8.00020-7 [DOI] [Google Scholar]
- Wieser MJ, & Moscovitch DA (2015). The Effect of Affective Context on Visuocortical Processing of Neutral Faces in Social Anxiety. Frontiers in Psychology, 6. https://www.frontiersin.org/articles/10.3389/fpsyg.2015.01824 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Wurpts IC, & Geiser C (2014). Is adding more indicators to a latent class analysis beneficial or detrimental? Results of a Monte-Carlo study. Frontiers in Psychology, 5. https://www.frontiersin.org/articles/10.3389/fpsyg.2014.00920 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Zhou J, Ko JY, Haight SC, & Tong VT (2019). Treatment of Substance Use Disorders Among Women of Reproductive Age by Depression and Anxiety Disorder Status, 2008–2014. Journal of Women’s Health, 28(8), 1068–1076. 10.1089/jwh.2018.7597 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
Research data are not shared.