The Effect of Multiple Contact Patterns During the COVID-19 Lockdown on Older People: A Doubly Robust Approach to Propensity Score Weighting with Gradient Boosted Regression

Loneliness is a significant public health concern due to its effects on health, longevity, and well-being. During the COVID-19 pandemic, restrictions on physical contact and lockdowns drastically reduced in-person relations, thereby increasing feelings of loneliness among older adults. Conversely, remote contact has increased, helping to counteract the negative effects of reduced face-to-face interactions. This study explores the causal impact of these altered contact patterns on perceived loneliness among older Italians. We employ gradient boosted regression for propensity score weighting to balance key pre-treatment differences. The Average Treatment Effect on the Treated (ATT) associated with changes in non-physical and in-person interactions is estimated using a doubly robust approach. Our findings suggest that increased non-physical contacts significantly mitigated the effects of reduced in-person interactions during the lockdown.