Measurement Congruence Between Record Data and Retrospective Self-Report Measures of Child Maltreatment: Do Positive Childhood Experiences Affect Discrepancies?

Abstract

Background.

Discrepancies between retrospective self-reports and official record data of child maltreatment (CM) are well-documented, yet few studies have examined how newer self-report instruments compare with record data or what factors influence inconsistencies across methods. This study addresses two primary aims: (1) to provide the first concordance estimates between prospective child protective services (CPS) records and the Maltreatment and Abuse Chronology of Exposure (MACE), a widely used retrospective CM assessment tool; and (2) to examine the influence of positive childhood experiences on discrepancies in CM assessment.

Methods.

We utilize two maltreatment cohorts in which adults and adolescents with documented histories of CM and matched nonmaltreated controls were enrolled. Both cohorts included CM data from CPS records coded with the Maltreatment Classification System (MCS) and retrospective self-reports of CM and measures of positive childhood experiences. The cohorts vary in age at retrospective assessment (adults vs. adolescents), retrospective time lag (long vs. short), used different self-report measures (MACE vs. CTQ), and different methods for assessing positive experiences (explicit self-report vs. ratings of unconscious content). The rigorous dual-study design ensures findings are robust to study- and measurement-specific differences.

Results.

Findings revealed minimal agreement between MACE self-reports and MCS-coded CPS records for maltreatment occurring from ages 0–12. Discrepancies were primarily driven by retrospective reports of CM not documented in official records. Importantly, in both studies, individuals with more positive childhood experiences were less likely to self-report maltreatment (via MACE or CTQ) that was documented based on official records..

Conclusions.

Findings suggest that positive childhood experiences may help facilitate resilience among CM survivors by influencing memory and appraisal of childhood events. Clinical interventions that explore autobiographical memories may be particularly effective in mitigating the psychopathology sequelae of maltreatment

Introduction

Child maltreatment (CM)—including physical (PA), sexual (SA), and emotional abuse (EA), as well as neglect before age 18 (USDHHS, 2022)—is a prevalent public health issue and a major source of stress and adversity for children. CM increases risk for impaired development and maladaptation across cognitive, emotional, interpersonal, neurobiological, and physical domains (Cicchetti & Toth, 2016). Due to its far-reaching effects on physical and mental health, reliable tools to assess CM exposure are critical. However, researchers lack a consensus or gold standard for assessing an individual’s CM history (Cooley & Jackson, 2020). Studies vary in their sources (e.g., self-report vs. official records), tools (e.g., questionnaires, interviews, record coding), and timing of assessment (prospective vs. retrospective; Jackson et al., 2019; Manly, 2005). CM assessment methods often show strikingly low agreement (Cooley et al., 2022; Baldwin et al., 2019). However, few studies assess CM via multiple methods, leaving much about discrepancy in assessment unknown. This current study examines the agreement and discrepancy between two common approaches: 1) prospectively obtained official record data, and 2) retrospective self-reports in adulthood. Both methods are imperfect sources of information with different (dis)advantages in their ability to accurately detect CM (Jackson et al., 2019).

Record-based assessment of CM draws from documentation by Child Protective Service (CPS) agencies regarding CM allegations and investigation results. Researchers can rely on the formal legal substantiation status or apply coding systems like the Maltreatment Classification System (MCS; Barnett et al., 1994) to identify CM (Cooley et al., 2022; Jackson et al., 2019; Manly, 2005). These records offer detailed, contemporaneous, multi-informant, and legally standardized information that may reduce recall bias and increase confidence of true positives. However, they are costly, time-intensive, and sometimes legally prohibited to access. CPS records may also underestimate CM exposure by failing to capture less severe or less visible forms of maltreatment that are difficult to detect or are never reported (Widom, 2004). Additionally, CPS records rely on documentation and assessments made by CPS workers, who are not immune to potential human errors or biases (Narayan et al., 2024).

In contrast, retrospective self-reports are the most widely used strategy of CM assessment and capture the first-hand report by asking individuals to recount their childhood experiences with brief questionnaires or structured interviews. These reports are efficient, easily administered, and may capture a broader range of CM experiences, including those likely to be missed by official systems. Although valuable, self-reports may also introduce biased and inconsistent estimates given the fluidity of individual interpretations of autobiographical events which may be influenced by one’s appraisal, memory processes (e.g., formation, distortion, reappraisal, decay), and motivations to disclose (Coleman et al. 2024).

When researchers include both methods, they tend to analyze only one due to the complications of discrepant reports between sources (Cooley & Jackson, 2020). This assumes the two measures identify similar groups of individuals and can be used interchangeably in clinical and research practices (Baldwin et al, 2019), but evidence suggests otherwise (Cooley et al., 2022). Baldwin et al. (2019) found poor agreement between record and questionnaire sources (Cohen’s kappa = .19) and others have shown the two assessment methods identify distinct groups with differing outcomes (Danese & Widom, 2020). The field has moved from litigating the validity of each method to understanding the causes of discrepancy (Cooley et al., 2022). This study advances that effort through two aims.

Aim 1: Concordance Between Prospective Record Data and the MACE Instrument

Most studies on concordance between record data and retrospective CM measures use a narrow set of self-report tools, such as the Childhood Trauma Questionnaire (CTQ), Conflict Tactics Scale (CTS), or study-specific instruments (Cooley et al., 2022; Baldwin et al., 2019). Although measures such as the CTQ are widely used, we have limited knowledge on the concordance of other instruments designed to assess CM retrospectively from adulthood. This is largely due to the rarity of longitudinal cohorts with linked official records, which limits the breadth of self-report tools that have been evaluated for concordance. This study provides the first concordance estimates between prospective CPS record data and the Maltreatment and Abuse Chronology of Exposure (MACE). Uniquely, the MACE was explicitly designed to better capture the developmental timing of CM exposure (Teicher & Parigger, 2015) and it has received some of the strongest psychometric ratings among adult CM self-report instruments (Georgieva et al., 2023). The MACE is a highly useful tool for CM scientists invested in understanding how variation in the developmental timing and chronicity affect sequelae. Despite these beneficial elements and psychometric strengths, unlike other prominent CM self-report instruments, the MACE’s concordance with prospective records has not yet been evaluated. We hypothesized that concordance between the MACE and MCS-coded records would be in line with estimates for other self-report measures (Baldwin et al., 2019).

Aim 2: The Role of Positive Childhood Experiences in CM Reporting Discrepancies

If prospective record data and retrospective self-report CM measures identify mostly different individuals, it is important to understand why. Why do some with documented CM not report it in adulthood? What characterizes this subgroup? Researchers point to several possible explanations including reluctance to disclose due to shame, stigma, avoidance of distress, discomfort with researchers, fear of mandated reporting and distrust of systems (Baldwin et al., 2019; Desir & Karatekin, 2019; Herman, 2003; Narayan et al., 2017), disagreement with the accuracy of the records (Narayan et al., 2024), and reappraisal processes shaped by personality or competing memories (Baldwin et al., 2019; Coleman et al., 2024; Francis et al., 2023; Najman et al., 2020; Reuben et al., 2016). Coleman et al. (2024) extensively describe how processes involved in the formation, consolidation, and retrieval of memories may contribute to poor recall and influence discrepancies. These include memory fallibilities such as poor retrieval of early childhood events (e.g., infantile amnesia), lack of memory consolidation due to low emotional valence, or protective mechanisms (e.g., denial or suppression of traumatic memories).

Few studies have empirically tested these explanations, with very little exploration of positive childhood experiences as a factor. One exception is Coleman et al. (2024), who utilized qualitative data from interviewer notes to investigate potential sources of discrepancies between prospective parent-reported CM and retrospective self-reported CM at age 18. They found that individuals often reinterpret childhood events as non-abusive, even when prospectively reported by parents as otherwise. They identified several possible explanations for incongruence, including difficulties in rapport, memory difficulties, and reappraisal processes related to intention, responsibility and acceptability of behaviors.

The current study investigates the hypothesis that positive childhood experiences reduce the likelihood that individuals with documented CM report it retrospectively. Positive memories may interfere with encoding or retrieval of abusive experiences, biasing recall and reappraisal toward more favorable narratives that supersede negative memories (Coleman et al., 2024). To test this, we applied a rigorous dual-cohort design with two separate longitudinal CM study samples: adults and adolescents with prospectively documented CM and matched nonmaltreated controls. Both studies assessed CM through coded CPS record data and retrospective self-report measures and included retrospective assessments of positive childhood experiences. The studies differed in important ways that allow us to test whether findings are consistent across samples and methods, including differences in age range (adults v. adolescents), retrospective time lag (short v. long), CM self-report instrument (MACE v. CTQ), and method for assessing positive childhood experiences—the Adult Attachment Interview (AAI; George et al., 1996) and the Benevolent Childhood Experiences Scale (BCEs; Narayan et al., 2018). The AAI and the BCEs are important complements in this replication study. Both address early life experiences but differ in focus and methodology. The AAI relies on inferred ratings of interview content to assess an individual’s internal representations of attachment relationships and reflects how one interprets or makes meaning of early experiences (Roisman, 2009). In contrast, the BCEs reflects concrete, observable positive experiences and external protective resources (e.g., having a supportive adult or safe home; Narayan et al., 2018).

Methods

Participants and Procedures

Recruitment Procedures for Studies 1 and 2

For both study samples, the CM-exposed youth were recruited by a Department of Human Services (DHS) liaison, who identified a random set of families with a child in the targeted age range who had experienced CM based on DHS CPS records. The DHS liaison approached the families, explained the study, and if interested, the parent signed a release form to share contact information with the project staff. Because the vast majority of families with documented CM were low-income, the DHS liaison identified a random set of families 1) with a child in the targeted age range, 2) who were receiving Temporary Assistance for Needy Families (TANF), and 3) who had no DHS records of CM. As with the maltreatment sample, the DHS liaison approached the families, explained the study, and if interested, the parent signed a release form to share contact information with project staff. These youth comprised the counterfactual group for comparison purposes. All families (CM and comparison) were living at <200% of the poverty rate.

For all families, a project staff member met with the mother and provided more details about the study. Mothers provided informed consent and permission for the youth’s participation. Youth participants provided assent. Additionally, all mothers provided consent for the research team to have access to all DHS records pertaining to the family for subsequent coding of maltreatment histories of the youth (for the maltreatment group) and confirmation of absence of records for youth in the comparison group. Absence of maltreatment was further confirmed by maternal interview (Maternal Child Maltreatment Interview; Cicchetti et al., 2003). Families were excluded from the comparison group if information was provided suggesting that the children may have experienced maltreatment. Study procedures were approved by Institutional Review Board at all waves and participants provided written, informed consent.

Study 1.

Once enrolled, youth in Study 1 participated in research summer camp between 1986–2004 (Wave 1). The week-long camp included recreational activities with groups of 8–10 children (same-sex) and reports of child functioning both self-reported by the children and independently by the camp counselors at the end of the week. Further details regarding camp procedures can be found in previous publications (Cicchetti & Manly, 1990).

At Wave 2, participants completed two research lab visits when they were between 24 – 45 years old (M = 30.90, SD = 5.41). Adult participants (n = 346) were racially and ethnically diverse (57.6% Black, 20.6% White, 14.8% Hispanic, 7% another race or ethnicity) 50.9% female, and 46.8% CM exposed based on records.

Study 2.

Conversely, youth in Study 2 participated in a series of individual research sessions conducted in private interview rooms by trained research staff. The same interviewer met with each adolescent for all research sessions to enhance rapport. The participants included 303 adolescents, approximately 15-years-old (M=15.03, SD = 1.20); 58.1% were male, 57.1% had documented CM exposure. Participants were racially and ethnically diverse (54.1% Black, 29% White, 10.9% Hispanic, 6% another race or ethnicity).

Measures

Prospective, Record-Based Maltreatment

For both studies, record-based maltreatment exposure was assessed via Maltreatment Classification System (MCS; Barnett et al., 1993) coding of CPS records from birth to age at initial study phase. The MCS is a comprehensive coding system of maltreatment subtype, severity, chronicity, and perpetrator—independent of CPS designations, which results in greater measurement sensitivity. Clinical psychologists completed MCS coding, reaching acceptable reliability; weighted κ with the criterion standard ranged from .86 to .98. Both studies utilize binary exposure to overall maltreatment and each subtype (physical, sexual, and emotional abuse and neglect), as well as the number of subtypes of exposure. Record data spanned birth to the participants’ age at entry into each Study (Study 1 = ~age 12, Study 2 = ~age 15). Exposure was indexed by age and developmental period.

Self-Report Measures

Study 1.

Retrospective, Self-Reported CM.

Retrospective self-reports of CM were collected at Time 2 in adulthood using the 75-question MACE (Teicher & Parigger, 2015). The MACE is designed to assess several CM subtypes from ages 1–18. It evaluated the severity of the following ten maltreatment subtypes at each age: sexual abuse, parental verbal abuse, parental non-verbal emotional abuse, parental physical abuse, witnessing abuse of siblings, witnessing interparental violence, peer verbal abuse, peer physical abuse, emotional neglect, and physical neglect. Severity scores for each subtype corresponded with the number of subtype questions positively endorsed. Each subtype has a unique severity threshold at or above which an individual is considered to have reported clinically-significant exposure to that subtype. See supplemental material for further description.

To maximize congruence between the MACE and MCS operationalizations of each maltreatment subtype, only those MACE subtypes corresponding most closely to MCS operationalizations (neglect, emotional abuse, physical abuse, and sexual abuse) were used for the purpose of this study. Presence of neglect was assigned if the individual met or exceeded the severity threshold for either physical or emotional neglect. Presence of emotional abuse was assigned if the individual met severity threshold for either parental verbal abuse, parental non-verbal emotional abuse, or witnessing interpersonal violence. Given these four maltreatment categories, multiplicity of maltreatment on the MACE could range from 0 to 4 clinically-significant subtypes. Clinically-significant exposure to these four maltreatment subtypes were evaluated in three segmented developmental periods (0–4 years, 5–7 years, and 8–12 years) as well as the full age range of interest (0–12 years) to map onto the developmental periods captured by prospective MCS information in Study 1.

Positive Childhood Experiences.

Also at Time 2, adult participants retrospectively reported their childhood experiences via the Benevolent Childhood Experiences Scale (BCEs; Narayan et al., 2018), a reliable and valid measure for adversity-exposed samples. This self-report measure assesses for 10 positive childhood experiences in a binary manner. Responses were summed to create a total BCEs score for the current study (M = 8.44, SD = 1.71).

Study 2.

Retrospective, Self-Reported CM.

Retrospective maltreatment was assessed via the Childhood Trauma Questionnaire – Short Form (CTQ; Bernstein et al., 2003; Bernstein & Fink, 1998), a 28-item self-report measure of traumatic experiences. In addition to overall maltreatment, the CTQ assesses for emotional, physical, and sexual abuse, and emotional and physical neglect. Using established cutoff scores, presence or absence of each type was coded according to standard cutoff scores (Bernstein et al., 2003; Bernstein & Fink, 1998). The current study combined emotional and physical neglect into a general neglect subtype.

Positive Childhood Experiences.

The Adult Attachment Interview (AAI; George et al., 1996) is a semi-structured, hour-long interview in which participants are asked to describe memories related to their early childhood relationships with their primary caregivers and the effects they perceived those experiences to have on their development. AAI coding focuses on the internal consistency of participants’ narratives as well as the extent to which participants become emotionally activated while discussing childhood memories. In the current study, we used a slightly modified AAI (Ward & Carlson, 1995).

The AAI narratives were subsequently transcribed and coded, using the AAI Q-Sort (Kobak, 1993), by raters centrally trained and certified as reliable on the AAI categorical coding system (Main et al., 1984–2003). To ensure reliability, 20% of the AAI transcripts were sorted by two coders, who were reliable (≥0.6 using the Person-Brown prophecy formula) on 88% of cases. Where two coders were discrepant, a third coder sorted the case. The final sort in these instances was computed by averaging the two sorts (reliable ≥0.6) that were the most highly correlated. Coders had no knowledge of maltreatment status.

To operationalize positive childhood experiences, we relied on one of the inferred parental experience subscales (“mother loving”). This scale reflects how much love, warmth, nurturance, and overall affection was present in the caregiving relationship and is scored on a scale of 1 (none; no loving qualities) to 7 (very strong; exceptionally strong loving attachment). M = 3.34 (SD = 1.14). Research with the AAI has documented the phenomenon by which an adult’s explicit report of their childhood experiences is not consistent with the narrative they generate via the AAI (Roisman, 2009), suggesting this measure taps unconscious information.

Data Analytic Plan

Aim 1: Record Data v. Self-Reported Maltreatment Concordance

In Study 1, the magnitude of agreement between MCS and MACE on presence/absence of CM (and subtypes) was determined via kappa coefficients, which were calculated at the full assessed age range (0–12) and at each of the three respective overlapping developmental epochs where possible (i.e., ages 0–4, 5–7, and 8–12). Guidelines for interpreting kappa values across studies are as follows: .20 or less (no agreement), .21-.39 (minimal), .40-.59 (weak), .60-.79 (moderate), .80 or greater (strong; McHugh, 2012). Intra-class coefficients were generated as estimates of concordance between record-based and self-reported assessment of continuous operationalizations of CM exposure, such as the number of subtypes experienced (ranging from 0–4) and the number of developmental periods in which CM occurred (ranging from 0–3). ICC values < .5 were interpreted as weak agreement. To determine the directionality of (in)congruence, we conducted 2 (record-based exposure) x 2 (self-reported exposure) crosstabulations.

In Study 2, magnitude and direction of agreement between MCS and CTQ were estimated for maltreatment status and each subtype in the same manner described above.

Aim 2: Testing the Influence of Positive Childhood Experiences on Congruence

In both studies, we identified the following four congruence groups for exposure to any CM and each subtype: 1) CM not present on either official record data nor self-report assessments (“Neither Record Data or Self-Report”), 2) official record of CM exposure without self-reported exposure (“Record Data Only”), 3) retrospective self-report of exposure without official record indication (“Self-Report Only”), and 4) official record data indicating exposure with corresponding self-report (“Record Data & Self-Report”).

To test whether positive childhood experiences predict CM concordance groups, we conducted multinomial logistic regressions with 95% Confidence Intervals (CI). The contrast between the “Record Data & Self-Report” group (reference) and the “Record Data Only” group was of central interest.

Sensitivity Analyses.

To test the robustness of the findings in Study 1, we re-estimated the multinomial regressions with the following covariates which are known to potentially bias retrospective reporting: observed childhood internalizing symptoms, current adult-reported depressive symptoms, and current adult-reported neuroticism(Reuben et al., 2016; Goltermann et al., 2023). See Supplemental materials for measurement description.

Results

Aim 1: Record Data v. Self-Reported Maltreatment Concordance

Study 1.

Magnitude of Agreement Between MCS and MACE.

Regarding agreement between MCS-coded official records and the MACE, there was minimal agreement for presence/absence of any CM, SA, and EA, and no agreement for neglect and PA. CM prevalence estimates, kappa coefficients, and ICCs are reported in Table 1.

Table 1.

Study 1 Rates of Maltreatment and MACE-MCS Agreement

Subtype Rates and Agreement by Developmental Period:
	Ages 0–12			Ages 0–4			Ages 5–7			Ages 8–12
	Kappa	N (%) Maltreated		Kappa	N (%) Maltreated		Kappa	N (%) Maltreated		Kappa	N (%) Maltreated
		MCS	MACE		MCS	MACE		MCS	MACE		MCS	MACE
Any CM	.268	162 (46.8%)	179 (51.7%)	.162	128 (37%)	109 (31.5%)	.107	34 (9.8%)	136 (39.3%)	.080	40 (11.6%)	167 (48.3%)
SA	.280	23 (6.6%)	52 (15%)	.303	16 (4.6%)	14 (4%)	.120	3 (0.9%)	27 (7.8%)	.118	4 (1.2%)	40 (11.6%)
PA	.194	61 (17.6%)	84 (24.3%)	.200	41 (11.8%)	20 (5.8%)	.128	12 (3.5%)	45 (13%)	.013	15 (4.3%)	79 (22.8%)
EA	.246	103 (29.8%)	119 (34.4%)	.086	81 (23.4%)	41 (11.8%)	.081	12 (3.5%)	76 (22%)	.062	18 (5.2%)	106 (30.6%)
N	.152	126 (36.4%)	118 (34.1%)	.106	96 (27.7%)	86 (24.9%)	.119	24 (6.9%)	96 (27.7%)	.039	31 (9%)	108 (31.2%)
Number of Subtypes and Number of Developmental Periods:
Number of Subtypes			Number of Dev Periods
	MCS	MACE		MCS	MACE
0	184 (53.2%)	167 (48.3%)	0	184 (53.2%)	167 (48.3%)
1	63 (18.2%)	66 (19.1%)	1	126 (36.4%)	44 (12.7%)
2	54 (15.6%)	51 (14.7%)	2	32 (9.2%)	38 (11.0%)
3	38 (11.0%)	43 (12.4%)	3	4 (1.2%)	97 (28.0%)
4	7 (2.0%)	19 (5.5%)	ICC:	.336
ICC:	.525

Notes: MACE: Maltreatment and Abuse Chronology of Exposure Scale; MCS: Maltreatment Classification System; CM: Child Maltreatment; SA: Sexual Abuse; PA: Physical Abuse; EA: Emotional Abuse; N: Neglect; ICC: intraclass correlation

Directionality of Agreement Between MCS and MACE.

Visualization of the cell proportions and (in)congruence percentages for the crosstabulations are shown in Figure 1. For any CM, SA, PA, and EA, most disagreement was due to adult self-report in the absence of MCS-coded indication of maltreatment. Specifically, results showed that 34% of CM (any type) captured by official records did not have corresponding adult self-reports, whereas 40.2% of self-reported maltreatment was not indicated by official records. For SA, 43.48% of SA captured by official records did not have corresponding adult self-reports, whereas 75% of self-reported SA was not indicated by official records. For PA, 57.38% of PA captured by official records did not have corresponding adult self-reports, whereas 69.05% of self-reported PA was not indicated by official records. For EA, 47.57% of EA captured by official records did not have corresponding adult self-reports, whereas 54.62% of self-reported EA was not indicated by official records. For neglect, discrepancy was driven by record data indication in the absence of self-report; 56.35% of neglect captured by official records did not have corresponding adult self-reports, whereas 53.4% of self-reported neglect was not indicated by official records.

Figure 1.

Directionality of MCS-MACE Discrepancies

Note: Graphs on the left shows rates of congruence (green) and discrepancy (red) within MCS-defined maltreatment groups; graphs on the right show rates of congruence (green) and discrepancy (red) within MACE-defined maltreatment groups.

Study 2.

Given the existence of meta-analyses testing agreement between official records of CM and the CTQ (Baldwin et al., 2019), we did not extensively detail the magnitude and directionality of measurement agreement in Study 2. Estimates were in line with meta-analytic results and similar to results of Study 1 (i.e., no agreement for neglect and EA; minimal agreement for any CM, PA, and SA).

Aim 2: Testing the Influence of Positive Childhood Experiences on Congruence

Multinomial Regression Results.

Study 1.

See Table 2 and Figure 2 for full results. Among adults with record-indicated CM histories, those who retrospectively reported fewer positive childhood experiences were more likely to be in the “Record Data & Self-Report” group (i.e., individual-confirmed CM exposure) vs. the “Record Data Only” group (i.e., no corresponding self-report of CM). For each additional BCE, the likelihood of being in the “Record Data Only” group (v. “Record Data & Self-Report”) increased by 60%. In other words, for each additional positive childhood experience, individuals with record-documented maltreatment were less likely to self-report maltreatment. This effect held after adjusting for childhood internalizing symptoms, and adult depressive symptoms and neuroticism. Further, this pattern of results was consistent for each distinct subtype of CM. See Supplemental Materials for mean level BCEs reported across groups.

Table 2.

Results of multinomial logistic regression.

Comparison		Predictor	β	p-value	Exp(β)	Confidence Intv
						Lower Bound	Upper Bound
Record Data-Only CM (v. Record & Self-Report)
	Study 1	BCE sum	.465	<.001	1.593	1.249	2.030
		BCE sum (adj. model)	.322	.014	1.380	1.066	1.786
	Study 2	AAI-Mother Loving	.698	<.001	2.009	1.449	2.787
Record Data-Only SA (v. Record & Self-Report )
	Study 1	BCE sum	.417	.067	1.518	.971	2.372
		BCE sum (adj. model)	.303	.255	1.354	.803	2.283
	Study 2	AAI-Mother Loving	.780	.021	2.181	1.124	4.232
Record Data-Only PA (v. Record & Self-Report )
	Study 1	BCE sum	.358	.013	1.431	1.078	1.90
		BCE sum (adj. model)	.321	.062	1.379	.984	1.934
	Study 2	AAI-Mother Loving	.774	<.001	2.168	1.350	3.481
Record Data-Only EA (v. Record & Self-Report )
	Study 1	BCE sum	.205	.066	1.228	.987	1.527
		BCE sum (adj. model)	.127	.326	1.135	.881	1.463
	Study 2	AAI-Mother Loving	1.353	<.001	3.867	2.178	6.865
Record Data-Only Neglect (v. Record & Self-Report )
	Study 1	BCE sum	.418	<.001	1.518	1.206	1.912
		BCE sum (adj. model)	.349	.009	1.417	1.089	1.844
	Study 2	AAI-Mother Loving	.711	<.001	2.035	1.139	2.979

Note: “Adj. Model” refers to model that includes the following control variables: Adult-reported depressive symptoms, adult-reported neuroticism, observed childhood internalizing symptoms. Bold font indicates significant 95% CI.

Figure 2.

Forest plot of odds ratios predicting membership in the “Record Data Only” group relative to membership in “Record Data & Self-Report” group.

Study 2.

Among adolescents with record-indicated CM histories, those who were rated as having fewer loving interactions with their mother were more likely to be in the “Record Data & Self-Report” group (i.e., individual-confirmed CM) vs the “Record Data Only” group (i.e., no corresponding self-report of CM). For a one-unit increase in the mother loving AAI scale, the likelihood of being in the “Record Data Only” group (v. “Record Data & Self-Report”) increased two-fold. This pattern was consistent for each distinct form of CM.

Discussion

Overview and Key Contributions

This study is the first to estimate concordance between MCS-coded records of childhood maltreatment (CM) and the MACE—a robust retrospective CM tool not yet systematically evaluated in this context. This study also stands among the first to show, using a rigorous dual-study design, that among individuals with record-documented maltreatment, those reporting more positive childhood experiences were less likely to retrospectively report maltreatment.

MACE Concordance Estimates

In line with established reviews (Baldwin et al., 2019; Cooley & Jackson, 2022), we found minimal to poor agreement between retrospective MACE reports and prospective MCS-coded records of CM from ages 0–12. However, the MACE showed slightly better agreement than other self-report tools across all CM subtypes. For any CM, prior meta-analyses found “poor agreement” (k = .19; Baldwin et al., 2019) between retrospective questionnaires (not including the MACE) and official records. We observed “minimal agreement” (k = .27) between the MACE and MCS-coded records. This may be due to MACE’s chronological structure, which could enhance recall, or to both the MACE and MCS-coded records being highly sensitive, non-CPS-dependent designations.

As in previous studies, the magnitude and directionality of concordance varied by CM subtype. Agreement was stronger for emotional and sexual abuse (“minimal”) than for physical abuse and neglect (“none”), mostly consistent with meta-analytic results (Baldwin et al., 2019) and systematic reviews (Cooley & Jackson, 2022). Baldwin et al (2019) found that agreement for emotional abuse was similarly lower as neglect. Most discrepancies reflected self-reported CM that was absent from official records, except for neglect, where records more commonly indicated CM that participants did not report. This is consistent with the directionality of disagreement between CM record data and other self-report questionnaires (Baldwin et al., 2019; Cooley & Jackson, 2020).

We offer a brief explanation for these differences, which have been extensively discussed elsewhere (see Cooley & Jackson, 2020; Cooley et al., 2022; Font & Maguire-Jack, 2020). Official records are likely to under-capture EA given the highly subjective nature—what CPS deems to cause “observable injury to psychological safety and emotional stability” (Font & Maguire-Jack, 2020) may not match an individuals’ interpretation (Font & Maguire-Jack, 2020; McCrory et al., 2017). Likewise, official records may fail to capture PA and SA perpetrated by non-parental caregivers (Font & Maguire-Jack, 2020). Conversely, individuals likely underreport neglect because it is a subtype that lacks definitional consensus/standards, involves complex issues such as parental capacity, intent, and culpability, and refers to acts of omission, all of which make it difficult to appraise (Cooley & Jackson, 2020; Font & Maguire-Jack, 2020).

Do Positive Childhood Experiences Affect Discrepancies?

Research suggests that prospective record data and retrospective self-report CM measures identify largely different groups of individuals (Baldwin et al., 2019). Our findings support this and point to positive childhood experiences as a contributing factor to discordance (Coleman et al., 2024). In Study 1, each additional benevolent childhood experience decreased the likelihood of individuals self-reporting CM that was indicated by record data by 60%, even when accounting for confounders like current depression, neuroticism, and childhood internalizing symptoms. Study 2 replicated the effect—a one-unit increase in inferred maternal lovingness decreased the likelihood of measurement concordance by a factor of 2.

This dual-study design—incorporating different ages, time lags, and measurement types—strengthens the case that these effects are not sample- or method-specific. Study 1 used adult retrospective self-reports of childhood experiences (MACE and BCE) with a long recall gap. Conversely, Study 2 relied on adolescent CTQ responses to assess CM exposure with short recall lag and coded AAI ratings tapping unconscious material pertaining to maternal caregiving. The findings suggest positive experiences, whether consciously remembered or implicitly encoded, influence agreement between record data and self-report of CM. These results held across CM types and samples, with minor variations. Study 1 showed statistically significant effects for CM, neglect, and PA; effects for SA and EA approached significance. In Study 2, all were significant, with larger effect sizes likely due to shorter recall lag, greater use of coded data, and emphasis on maternal relationships.

Finally, retrospective reporting and/or reappraisals of adverse childhood experiences may be positively biased by personality features (e.g., high optimism and agreeableness, low neuroticism; Reuben et al., 2016) and negatively biased by past or present psychopathology symptoms (e.g., negative bias is a core feature of depression: Goltermann et al., 2023). Thus, it is noteworthy that the influence of positive childhood experiences in Study 1 was robust to the effects of personality traits and both past and current depressive symptoms. This increases confidence that the unique influence of positive childhood experiences on retrospective recall was isolated from the relative influence of individual characteristics or mood states. Future investigations into the factors that influence congruence between prospective maltreatment records and retrospective recall should account for possible confounding factors.

Interpretation and Mechanisms

Our results align with prior findings that CM assessment discrepancies are predictable (Kobulsky et al., 2018; Najman et al., 2020; Reuben et al., 2016). While Coleman et al. (2024) used qualitative data to explore mismatch between parent and child CM reports, we are the first to statistically test positive experiences as a source of discrepancy. Their findings—suggesting discrepancies stem from differing appraisals—support ours.

The current findings may be interpreted through memory and appraisal theory (see Coleman et al., 2024 for excellent review). Positive experiences may interfere with the encoding, storage, or retrieval of negative events. Memories of CM may dissipate as contemporaneous positive experiences overwrite, supersede, and/or dilute negative memories, affecting retrospective recall. Likewise, the formation and retention of CM memories can be subject to interference that may color interpretations of events (Coleman et al., 2024). Prior positive childhood experiences may compete with CM experiences, shaping the interpretation of childhood in favor of a positive or neutral appraisal, which then reduces the likelihood negative memories are encoded, consolidated, and retained for recall (Coleman et al., 2024). Retrospective reports are also filtered through individual (re)appraisals at the time of recall, which may bias, distort, or even reshape existing memories (Coleman et al., 2024). Individuals may struggle to simultaneously identify as both a victim of CM and the beneficiary of positive care (Wekerle et al., 2001). This is consistent with research showing that individuals with CM histories can experience overgeneral memory (see Valentino, 2011; Valentino et al., 2009), which results in more generalized (vs specific) autobiographical memory, and greater likelihood of a categorical framing of childhood. Thus, it may be that individuals with both CM exposure and positive childhood experiences generally recall positive memories rather than specific CM exposures. Finally, positive memories may facilitate forgiveness or trauma resolution, reducing recall of CM (Hardt & Rutter, 2004; Reuben et al., 2016; Robins et al., 1985).

An alternative explanation is that, in some cases, incongruence may stem from records that overestimate or underestimate the presence of maltreatment. CPS records and documentation inevitably involve a degree of individual decision-making and qualitative judgement, leaving the potential for human error and biases (e.g., Drake et al., 2023; Kim & Drake, 2018; Narayan et al., 2024). In the context of our findings, positive childhood experiences may increase the likelihood of being in the “Record-Only” group because, in some instances, these experiences more accurately represent the individuals’ lived experience.

Strengths and Limitations

Results are contextualized in the presence of study limitations. First, we interpret findings via theory on memory processes and appraisal and cannot rule out motivational factors (e.g., shame/guilt, interviewer rapport/trust; Coleman et al., 2024) as the cause of discrepant reports. Similarly, misclassification into one of the four (in)congruence groups is possible as neither measurement is infallible. However, the strengths are substantial: a dual-study, multi-method, multi-informant, multi-population design with high-quality, prospectively collected data; demographically-matched non-maltreated comparison groups; and valid instruments. Cohort studies such as these are incredibly rare—we capitalize on two. Finally, using MCS-coded records—not just CPS substantiations—improves accuracy of maltreatment designations and reduces many biases through standardization and inter-rater reliability.

Implications for Research and Practice

Positive childhood experiences help facilitate resilience among CM survivors (Cicchetti, 2013), and our findings may elucidate one possible mechanism. Danese and Widom (2020) have demonstrated that self-reported CM is more predictive of mental health outcomes than record data (Danese & Widom, 2020), implying that memory/appraisal of CM is the key factor in determining mental health risk. Our results suggest that positive memories may modulate these appraisals in a way that could mitigate psychopathology sequelae. Clinicians who work with CM survivors should consider exploring positive memories with clients, as in the “angels in the nursery” approach from Child-Parent Psychotherapy (Lieberman et al., 2005).

Conclusion

This study underscores the minimal concordance between record data and retrospective self-report CM assessments and highlights how positive childhood experiences may reduce recall of, or agreement with, documented CM. Future research should consider both positive and negative childhood experiences to better understand memory, measurement, and clinical outcomes in maltreatment research. Likewise, CM investigations should consider assessing CM exposure with both self-report and record-based assessments.

Abbreviations:

CM

Child Maltreatment

CPS

Child Protective Services

MACE

Maltreatment and Abuse Chronology of Exposure

CTQ

Childhood Trauma Questionnaire

MCS

Maltreatment Classification System

EA

Emotional Abuse

PA

Physical Abuse

SA

Sexual Abuse

Data Availability.

Data is not available due to the inclusion of information obtained from Child Protective Service (CPS) records and data sharing restrictions that were part of the data agreement with governing bodies.