Evidence of individual differences in the long-term social, psychological, and cognitive consequences of child maltreatment

Study design and participants

This is a matched cohort study involving 236 Hong Kong Chinese individuals identified from an existing research database containing their contact information. The exposed cohort comprised 85 individuals who were admitted to the pediatric ward of a local teaching hospital under the age of 18 years due to one type or multiple types of maltreatment defined by the International Classification of Diseases (ICD-9-CM) codes (955.50-955.59 or E967.0-E967.9) for the first time during 2001–2010 as recorded in the Hong Kong Clinical Data Analysis and Reporting System (CDARS), which is an electronic database used by the public healthcare system in Hong Kong. All these maltreatment cases were confirmed through a detailed investigation conducted by a multidisciplinary team, which has been described in a previous study [16]. The unexposed cohort comprised 151 patients who were admitted to the same hospital under the age of 18 years due to common cold, bronchitis, influenza, or pneumonia (ICD-9-CM 460-488) in the same calendar year and month for a length of stay ≤ 14 days and without congenital anomalies (ICD-9-CM 740-759) or prior diagnosis of child maltreatment at any point in the CDARS database. All these individuals had provided consent for future research contact at the time of study enrollment.

The follow-up research was conducted during 2018–2021. We first invited the individuals in the exposed cohort to join the follow-up assessment by phone. For every exposed individual who accepted our study invitation, one unexposed individual of the same gender, similar age of onset (varying by 1 year), similar follow-up time (varying by 1 year), and similar poverty status at the time of index hospital admission (receiving government financial assistance or not) would be invited until the matching and assessment procedures were completed.

Ethical approval

All procedures involving human subjects in this study were approved by the Institutional Review Board of the University of Hong Kong/Hospital Authority Hong Kong Western Cluster (UW 18-442). In addition to the initial consent for future research contact, all individuals who attended the follow-up assessment were required to provide informed consent for participation in this follow-up research. For those under the age of 18 years at the time of data collection, the consent of their parent/guardian was also obtained.

Data collection and measures

During the assessment session, the participant underwent a blood draw, performed by a phlebotomist, and completed a comprehensive set of questionnaires with items measuring their social, cognitive, and psychological conditions. Time-varying demographic characteristics including educational attainment and employment status were also collected. For participants who had difficulty completing the questionnaires by themselves such as early adolescents (less than 10% of the sample), we asked their parent/primary caregiver to complete the questionnaires on behalf of the participants.

Social domain

The 7-item family support subscale of the Multidimensional Scale of Perceived Social Support (MSPSS) was used to assess the participant’s perceived degree of support from family members on a 7-point Likert scale, with higher scores indicating greater family support. The Chinese version of MSPSS family support subscale demonstrated excellent internal consistency (Cronbach’s alpha: 0.89) in a previous study [17].

The Parental Bonding Instrument (PBI) was used to examine the participants’ memories of their mothers and fathers during the first 16 years of life [18]. For participants under the age of 16 years, we asked them to recall the period from birth to the date of assessment. Each participant completed a scale for the mother and father, respectively. The PBI has 12 care items and 13 overprotection items on a 4-point scale, with higher scores indicating higher levels of parental care. The Chinese version of PBI has been validated in Hong Kong with good convergent validity and reliability (Cronbach’s alpha: 0.72–0.86) [19].

Cognitive domain

Executive functioning was assessed with the Adult Executive Functioning Inventory (ADEXI) [20].The ADEXI has 14 items in two subscales: working memory and inhibition. It has been shown to have high reliability (i.e. Cronbach’s alphas around 0.90 for the ADEXI full scale) and high discriminative validity with large effect sizes (sensitivity: 0.76 and specificity: 0.91) when comparing individuals with ADHD and those without [20]. The Chinese version of ADEXI has been used in previous research studies and demonstrated good internal consistency (Cronbach’s alpha 0.80 for working memory; 0.70 for inhibition) [21].

Psychological domain

The 10-item short version of the Connor–Davidson Resilience Scale (CD-RISC10) was used to measure the participant’s level of resilience on a five-point Likert scale ranging from 0 = not true at all to 4 = true nearly all of the time [22]. The item scores are added to a total score, with a higher score reflecting higher resilience [22]. It has been widely used in local clinical and community settings and demonstrated good validity (correlation coefficients with other psychological qualities ranging from −0.26 to 0.53) and internal consistency (Cronbach’s alpha = 0.87) [23].

The subjective and objective measurements of the participant’s stress level were conducted through the analysis of questionnaire and biomarker data. The 7-item stress scale of the Depression Anxiety Stress Scale – 21 (DASS21 stress) was used to assess the participant’s perceived stress level in the preceding week on a 4-point scale, with higher scores indicating higher levels of stress. The DASS21 stress scale has demonstrated good internal consistency (Cronbach’s alpha = 0.89) and validity (intercorrelations with the other two scales of DASS-21: 0.63 for Depression and 0.67 for Anxiety scales) [24]. In addition, inflammatory markers including IL-6, IL-10, and IL-6/IL-10 ratio were used to evaluate the physiological stress responses. Specifically, serum was extracted from the peripheral blood after 15-min centrifugation at 3000 rpm and stored at −80 °C. The serum levels of IL-6 and IL-10 were determined using multiplexed bead-based immunoassays using the LEGENDplex™ Human Inflammation Panel 1 [13-plex] kit (Biolegend, San Diego, CA). Data from the flow cytometry was then acquired using the BD LSR II System (BD Biosciences, Franklin Lakes, NJ) and analyzed by the LEGENDplex™ Data Analysis Software Suit v8.0 (Biolegend).

Sample size

Among the 85 exposed individuals with a history of substantiated childhood maltreatment in the database, blood samples and survey data were collected from 63 individuals, whereas the remaining 22 individuals were not assessed due to invalid contact information (n = 7) or refusal to participate (n = 15). Based on the pre-specified matching criteria, the unexposed individuals were then approached and matched to the exposed individuals by a ratio of 1:1. Hence, this study analyzed 63 exposed individuals and 63 unexposed individuals. A meta-analysis found an odds ratio of 3.42 for the association between overall childhood maltreatment and non-suicidal self-injury [25]. Our assessed sample size was thus deemed adequate for detecting this medium effect size with 80% power at the 0.05 level of significance [26].

Data analysis

After data checking for normality of distribution by means of skewness and kurtosis, descriptive statistics were presented as means and standard deviations or frequency count as appropriate. Variables with skewed distributions including IL-6, IL-10, and monthly household income at follow-up were logarithmically transformed before data analysis. Differences in demographic characteristics between exposed and unexposed individuals were assessed using the chi-square test for categorical variables and t-test for continuous variables. In addition, we performed generalized linear regression to assess the between-group differences in all measured parameters, quantified as unstandardized regression estimate (β). Two models were constructed: first with child maltreatment status as the independent variable and second with an additional adjustment for monthly household income at follow-up. For subgroup analyses, the participants were further categorized using median splits based on their age of onset (< 8 vs ≥ 8) and follow-up duration (< 14 vs ≥ 14).

The missing data rates for the variables of interest range from 0.3% to 14.0%. Little’s MCAR X2[18] = 17.25, p = 0.506 indicated that the data were missing completely at random and appropriate for multiple imputation [27]. We imputed missing values using the R package mice [28]. Five data sets were imputed using predictive mean matching following 50 iterations of the algorithm. Since the imputed values were largely similar across the imputed data sets, we used the first imputed data set in this analysis. Sensitivity analyses also showed trivial differences between the original data and imputed data, so the imputed results are presented in the main results (Table 2 and Fig. 1) and complete cases in Additional file 1 results (Additional file 1: Tables S1 and S3a–d). All analyses were conducted using R version 4.1.1, with a p-value < 0.05 denoting statistical significance.

Fig. 1figure 1

Comparison of a log10-transformed IL-6 value, b log10-transformed IL-10 value, c log10-transformed IL-6:IL-10 value, d MSPSS family support score, e PBI care (mother) score, and f PBI care (father) score between exposed and unexposed individuals stratified by follow-up time and age of onset. *p < 0.05; **p < 0.01; ***p < 0.001. Adjusted for monthly household income at follow-up

留言 (0)

沒有登入
gif