Large population studies are important sources for medical research. These studies are well planned, well organized, and costly. However, people record health data themselves using different sensors, which are mostly unplanned, unorganized and inexpensive. Nevertheless, self-recorded data might be an important supplement to population studies. The question is how to access and use this data. In the seventh survey of the Tromsøcohort study, questionnaires and accelerometers were used to collect data on physical activity (PA). We now plan to collect historical PA data from these participants, using mobile sensor data already stored in the cloud. We will examine the feasibility of this approach and the quality of this data. Objectively measured historical data will provide valuable insights in the potential and limitations of mobile sensors as new data collection tools in medical research.