GEOspatial FRAILty and Recovery in Independent Living (GEOFRAIL)

The GEOspatial FRAILty and Recovery in Independent Living (GEOFRAIL) dataset consists of the following interconnected tables linked by shared primary and foreign keys, which are denoted by underlining below.To protect participant privacy, the dataset has been de-identified. Raw latitude, longitude, and postal code information are not shared. Locations where participants resided or visited are represented exclusively using coded location identifiers (L0001–L4096). In addition, small random perturbations (3–5%) were applied to non-zero amenities, crime rate, and census-derived variables to reduce the risk of re-identification while preserving aggregate distributions. The dates on which participant data were collected were also temporally perturbed to further limit the risk of re-identification. 1) Demographics. The first table (18 × 11) contains participant-level demographic and study information, where 18 represents the number of participants and 11 represents the number of variables, including:• participant ID (1 variable);• monitoring period start and end dates (2 variables);• demographic characteristics: age, sex, ethnicity, education level, employment status, relationship status, and fracture type (7 variables); and• location ID of participant's home (1 variable). 2) Sensor features. The second table (18 × 56 × 48) contains features extracted on a daily basis from multiple continuous multimodal sensor streams collected from participants, as described in Abedi et al.24. Each feature represents a specific characteristic of a sensor modality on a day; for example, one step-derived feature reports the total number of steps per day. In total, 46 sensor-derived features are provided per participant per day, with 56 days of monitoring per participant. Detailed information on the features extracted from different data modalities is provided in Supplementary Information 2.• participant ID (1 variable);• timestamp date (1 variable);• acceleration (16 variables); heart rate (6 variables); motion (5 variables); location (3 variables); sleep (11 variables); and step features (5 variables). 3) Clinical. The third table (18 × 4 × 63) contains clinical questionnaire responses and physical performance test results collected biweekly or every 8 weeks.• participant ID (1 variable);• timestamp bi-weekly or every 8 weeks date (1 variable);• SIS (6 variables); OHS (12 variables); OKS (12 variables); TUG test (1 variable); 30-second Chair Stand test (1 variable); RAPA (9 variables); CFS-HQ (16 variables); grip strength (2 variables); and body weight and height (2 variables) 4) Temporal location. The fourth table (18,843 × 3) contains deidentified temporal location identifiers representing locations where participants stayed or visited.• participant ID (1 variable);• timestamp second level (1 variable); and• location ID of locations participants stayed in or visited (1 variable). 5) Amenities. The fifth table (2,881 × 6) contains deidentified unique location IDs linked to amenity proximity data, including counts of selected amenities within a 1 km radius of locations where participants stayed or visited. These 2,881 unique location IDs account for 18,843 temporal location records, reflecting repeated visits or stays at the same locations across participants and time. • location ID (1 variable);• number of community centers; libraries; parks; food establishments; and places of worship (5 variables). 6) Crime rate. The sixth table (2,881 × 10) contains deidentified unique location IDs linked to crime rates in the neighborhoods where participants stayed or visited.• location ID (1 variable);• assault; auto theft; bike theft; break and enter; homicide; robbery; shooting; theft from motor vehicle; and high value theft rate (9 variables). 7) Census. The seventh table (2,881 × 1,624 × 6) contains deidentified unique location IDs linked to 1,624 census variables describing neighborhood level characteristics of the areas where participants stayed or visited, reported across six population measures, total count, men count, women count, total rate, men rate, and women rate. Detailed information on the census variables is provided in Supplementary Information 3.• location ID (1 variable);• census data (1,624 × 6 variables).