The ability to predict upper respiratory infections (URI), lower respiratory infections (LRI), and gastrointestinal tract infections (GI) in independently living older persons would greatly benefit population and individual health. Social network parameters have so far not been included in prediction models. Data were obtained from The Maastricht Study, a population-based cohort study (N = 3074, mean age (±s.d.) 59·8 ± 8·3, 48·8% women). We used multivariable logistic regression analysis to develop prediction models for self-reported symptomatic URI, LRI, and GI (past 2 months). We determined performance of the models by quantifying measures of discriminative ability and calibration. Overall, 953 individuals (31·0%) reported URI, 349 (11·4%) LRI, and 380 (12·4%) GI. The area under the curve was 64·7% (95% confidence interval (CI) 62·6–66·8%) for URI, 71·1% (95% CI 68·4–73·8) for LRI, and 64·2% (95% CI 61·3–67·1%) for GI. All models had good calibration (based on visual inspection of calibration plot, and Hosmer–Lemeshow goodness-of-fit test). Social network parameters were strong predictors for URI, LRI, and GI. Using social network parameters in prediction models for URI, LRI, and GI seems highly promising. Such parameters may be used as potential determinants that can be addressed in a practical intervention in older persons, or in a predictive tool to compute an individual's probability of infections.