An application of ordinal regression to extract social dysfunction levels through behavioral problems

Area of Science:

• Psychology
• Psychometrics
• Mental Health Assessment

Background:

• Accurate identification of psychological problems is crucial for intervention.
• Questionnaires like the Strengths and Difficulties Questionnaire (SDQ) are vital tools for assessing behavioral issues and distress.
• Social dysfunction significantly impacts daily life, necessitating reliable measurement methods.

Purpose of the Study:

• To investigate if social dysfunction, measured by SDQ impact scores, can be predicted from SDQ behavioral problems in Indian university students.
• To analyze the relationship between specific SDQ difficulty components and levels of psychological distress during the COVID-19 pandemic.

Main Methods:

• Two surveys were conducted on distressed college/university students in India during the COVID-19 pandemic.
• Ordinal regression (OR) models with a negative log-log link function were used to estimate distress levels (Normal, Borderline, Abnormal) from SDQ behavioral problems.
• Model fit was assessed using Cox and Snell, Nagelkerke, and McFadden tests.

Main Results:

• Hyperactivity-inattention and emotional symptoms were significant predictors of distress in the first survey.
• Peer problems emerged as a significant predictor in the second survey, alongside hyperactivity-inattention and emotional symptoms.
• Ordinal regression models effectively estimated the 'Normal' and 'Abnormal' distress categories but showed limitations in accurately estimating the 'Borderline' category.

Conclusions:

• SDQ behavioral problems, particularly hyperactivity-inattention, emotional symptoms, and peer problems, are valuable for estimating psychological distress levels in university students.
• The study highlights the utility of SDQ for large-scale mental health screening but points to challenges in precisely categorizing borderline distress.
• Further research may be needed to refine models for better estimation of the borderline distress category using complex SDQ data.