Descriptive Statistics
Table S4 (China) and Table S5 (the UK) in Supplementary Materials presents the descriptive statistics and correlations between variables.
Latent Profile Solution – Chinese Samples
A series of latent profile models with one to eight were specified and estimated. The LMR test had a non-significant value for the 6-class model (p > .05), suggesting a 5-class optimal model. However, the p-value of the 5-class model LMR test was close to our p-threshold of .05 and we therefore also considered the 4-class model as a further candidate optimal model. Other than the LMR test, we also utilized criteria for AIC, BIC, and aBIC for model selection. These suggested that the five-profile solution should be preferred. However, Foti et al. (2012) highlighted that the meaning of the profile-solution picked needs to be considered when determining the number of profiles. We thus further investigated the characteristics of these solutions. We found that choosing a four-profile solution translated into the addition of a meaningful profile, while the five-profile solution shared similar characteristics with the four-profile solution (Full statistics from the latent profile analysis models see Table S6 in Supplementary Material).
Based on the aforementioned rationales, we thus explored further the four-profile model in the Chinese sample. Table 6 indicated that the entropy of the four-profile solution was 0.904, which is considered satisfactory (Morin et al., 2011). Additionally, the average latent class probabilities for the four-profile model were 0.959, 0.928, 0.937, and 0.919, above the cut-off criterion of 0.80 (Nylund, Asparouhov, & Muthén, 2007). Accordingly, we adopted the four-profile model for the Chinese samples as the best solution based on theoretical and statistical considerations.
Figure 1 displays the four-profile solution that was retained. Values on the Y-axis represented the sum scores of different types of childhood psychological maltreatment. The first latent profile was the smallest and described the 5.4% of the sample who reported the lowest level of psychological abuse and psychological neglect but the highest level of psychological non-support and was thus labelled the “Psychological Non-support” profile. The second latent profile represented 56.5% of the samples was labeled “Low-Maltreated,” given that the level of psychological abuse and psychological neglect was higher than the “Psychological Non-support” group, but with the lowest level of psychological non-support. The third latent profile, named the “High-Maltreated” profile, characterized 33.9% of the sample presenting experiences of a higher level of psychological abuse, psychological neglect, and psychological non-support. The fourth latent profile, labeled “Severe-Maltreated”, includes 4.3% of the sample and reported the highest level of psychological abuse, psychological neglect, and psychological non-support. Table 1 presents the means and standard errors of each psychological maltreatment type.
Latent Profile Solution – UK Samples
The four-class solution was also considered the best fitting model for the UK sample based on a series of fit indices (see Table S7 in supplementary Material). The LMR test had a non-significant value for the 5-class model (p > .05), suggesting a 4-class optimal model and the BIC, which has been identified as the most reliable of the available fit indices (Nylund et al., 2007), was lower than other models with significant p-values. We, therefore, adopted the four-profile model as the best solution for further analyses.
Figure 2 displays the four-profile solution for the UK sample. Values on the Y-axis represent sum scores of different types of childhood psychological maltreatment. The first latent profile was the largest and described 57.9% of the sample who reported the lowest level of childhood psychological maltreatment and were thus “Low-Maltreated”. The second latent profile represented by 25.4% of the sample was labeled “Moderate-Maltreated,” given that samples experienced a moderate level of childhood psychological maltreatment. The third latent profile, named the “High-Maltreated” profile, characterized 11.2% of samples, representing a higher level of childhood psychological maltreatment. The fourth latent profile, labeled “Severe-Maltreated”, included 5.5% of the sample who reported the highest level of childhood psychological maltreatment. Table 2 presents the means and standard error for each form of psychological maltreatment.
Outcomes of Latent Profile – Chinese Samples
The relationships between four latent profiles and mental health outcomes for the Chinese sample are displayed in Table 3. The results indicated that the sample in the “Psychological Non-support” profile exhibit the best mental health outcomes with the lowest depression, anxiety, and aggression levels and the highest self-esteem levels. In addition, those assigned to the “Severe-Maltreated” profile show poorer mental health outcomes than other profiles, with the highest levels of depression, anxiety, and aggression, and the lowest levels of self-esteem. The “Psychological Non-support” profile was categorized as having minimal depression. The other three profiles were associated with mild or moderate depression. The differences in levels of depression between all profiles were significant, except for the difference between “Moderate-Maltreated” and “Severe-Maltreated” profiles. When compared on anxiety levels, the “Low-Maltreated” profile displayed the lowest level of anxiety, while other profiles were all above the clinical cut-off point. There was no difference between the “High-Maltreated” and “Severe-Maltreated” profiles but differences between other profiles were significant.
Only the “Psychological Non-support” profile was below the norm mean score when compared on anger. The difference between all profiles was significant. When compared on physical aggression scores, “Psychological Non-support” and “Low-Maltreated” profiles were below the norm mean score, while “High-Maltreated” and “Severe-Maltreated” profiles were above the norm mean scores. The differences between all profiles were significant, except for the difference between “Moderate-Maltreated” and “Severe-Maltreated”. When compared on verbal aggression scores, only the “Psychological Non-support” profile was below the norm mean score. The differences between the “Low-Maltreated”. “Moderate-Maltreated”, and “Severe-Maltreated” were non-significant, while the differences between “Psychological Non-support” with other profiles were significant. Only the “Psychological Non-support” profile was below the norm mean score when compared on the hostility scores. Other profiles were all above the mean scores. All the profiles showed significant differences compared with other profiles.
When comparing on the self-esteem levels, samples in the “Psychological Non-support” profiles showed a higher level of self-esteem, while other profiles were within the normal ranges. There was no difference between the “Psychological Non-support” and “Low-Maltreated” profiles; however, the differences between other profiles were significant.
Outcomes of Latent Profile – UK Samples
The relationships between four latent profiles of mental health outcomes for the UK samples are displayed in Table 4. The results indicated that the sample in the “Low-Maltreated” profile exhibited better mental health outcomes such as higher levels of self-esteem and lower levels of anxiety, depression, and aggression. Moreover, the “Severe-Maltreated” profile displayed poorer mental health outcomes with the highest level of depression, anxiety, aggression, and the lowest level of self-esteem.
When compared on depression levels, the “Low-Maltreated” profile was categorized as showing minimal depression, “Moderate-Maltreated” and “High-Maltreated” were categorized as showing mild and moderate depression, while the “Severe-Maltreated” profile was categorized as showing moderately severe depression. The difference between profiles was significant, except for the “High-Maltreated” and “Severe-Maltreated” profiles. Only the “Low-Maltreated” profile was below the clinical cut-off point when compared on anxiety levels. Other profiles were all above the clinical cut-off point. Except for the difference between the “Moderate-Maltreated” and “Severe-Maltreated” profiles being non-significant, differences between other profiles were all significant.
Regarding anger scores, the “Low-Maltreated” profile was below the norms mean score, while other profiles were all above the norms mean score. Differences between “Moderate-Maltreated”, “High-Maltreated,” and “Severe-Maltreated” were non-significant, while the difference between the “Low-Maltreated” profile and other profiles was significant. When compared on physical aggression, all the profiles were below the norms mean score. Like anger, the difference between the “Low-Maltreated” and other profiles was significant, while the differences between the other three profiles were non-significant. When compared on verbal aggression scores, all profiles were above the norms mean scores except the “Low-Maltreated” profile. Differences between all profiles were non-significant except for the difference between the “Low-Maltreated” and “Moderate-Maltreated” profiles. When compared on hostility scores, all the profiles were above the norms mean scores. Differences between all profiles were significant, except for the difference between the “Moderate-Maltreated” and “High-Maltreated” profiles.
When compared on self-esteem scores, all the profiles were within the normal range. The differences between “Moderate-Maltreated” and “High-Maltreated” and “High-Maltreated” and “Severe-Maltreated” profiles were non-significant. Other profiles showed significant differences between each other.