Ageing is a physiological phenomenon associated with cognitive and functional decline which, in the long term, could hamper the execution of daily life activities and threaten both social and independent life. The onset of chronic diseases can intensify this process, increasing the risk of hospitalisation and admission to long term care. This represents a significant burden on public health and reduces the quality of life of those affected. Early detection of unhealthy decline is therefore key, but the similarity to normal ageing hinders its prompt screening. This study presents a first step towards the early screening of unhealthy ageing, based on an innovative instrumented ink pen to ecologically assess handwriting performance in different age groups: 40-59 (Group 1), 60-69 (Group 2) and 70+ (Group 3) years old. Raw handwriting data were collected from 60 healthy subjects and used to extract fourteen indicators related to gesture and tremor. The indicators were then used to discriminate between subjects of different age groups in three binary classification tasks, using a selection of machine learning algorithms. This approach produced remarkable results, particularly in the task of greatest interest, identifying subjects at the very beginning of the ageing process (Group 2) from elderly subjects (Group 3), achieving an accuracy of 97.5%, an F1 score of 97.44% and a ROC-AUC of 95%. Explainability of the model, facilitated by the analysis of the Shapley values of the learned indicators, revealed age-dependent sensitivity of handwriting and tremor-related indicators. The proposed method represents a promising solution for the early detection of abnormal signs of ageing, and is designed for the remote, non-invasive, unsupervised home monitoring, to improve the care of older adults.