Iatrogenic dependence linked to hospitalisation DILH is very frequent, severe but avoidable in 80% of cases. Extensive research has identified 6 main causes of DILH in the elderly, which interact and have common risk factors: immobilisation, falls, undernutrition, de novo urinary incontinence, confusion and drug side-effects. Our main hypothesis is that promoting access to, knowledge of and implementation of good practice recommendations by healthcare staff (medical and paramedical) concerning DILH in the elderly, based on graded interventions and targeting the 6 main causes of DILH, via the TAKE CARE intervention could reduce the incidence of DILH in the population of patients aged 75 or over hospitalised in acute geriatric units excluding post-operative and post-resuscitation, in the context of unscheduled hospitalizations. A Multicenter, randomized, cluster trial evaluating the effect of the TAKE CARE intervention compared with usual care in patients will be conducted across 11 acute geriatric services in France. A total of 2200 patients will be needed for a total period of 6 months with a 6 months followup prior inclusion. Data will be collected directly from the patients' medical records and entered by an investigator or a clinical study technician in an electronic CRF (CleanWeb) managed by the URC PSL. The trial will also include a qualitative component to assess, among other things, the implementation aspects of this complex intervention (individual interviews, focus group). A medico-economic component with a cost-utility analysis to assess the incremental cost-utility ratio, and a hospital microcosting study to assess the costs associated with the TAKE CARE intervention. The impact of the TAKE CARE intervention will be studied using a win-ratio approach based on a hierarchical composite criterion that takes into account vital status at D7 and the change in ADL score between D15 before admission and D7. The difference in ADL score between D-15 and D7 will be compared between the 2 randomisation groups using a mixed linear model with the randomisation arm as a fixed effect and a random effect on the cluster. A win-ratio significantly greater than 1 will allow us to conclude that the TAKE CARE intervention is effective.