Environmental research has been transformed by big data technologies, such as electronic health records linkage, wearables and remote sensing. These tools provide the opportunity to perform large population-based studies, with longitudinal collection of individual-level variables and finely reconstructed spatio-temporal exposure maps. However, these studies require innovative analytical methods. Here we present a new study design called case time series for investigating transient health risks in epidemiological analyses. The design is based on the reconstruction of longitudinal profiles of health outcomes and time-varying predictors in subject-specific series. This adaptable framework combines the individual-level setting and confounding control of case-only methods, such as case-crossover and self-controlled case series, with the flexibility and temporal structure of time series methods to model trends and lagged effects. It is applicable with either continuous or (multiple) event outcomes. Estimation exploits the computational efficiency of conditional regression models. Applications are demonstrated through three case studies that illustrate flexibility and wide applicability of the design: 1) analysis of associations between air pollution and asthma recurrence using a general practitioners patients cohort; 2) country-wide study of temperature-mortality associations using small-area data and high-resolution exposure maps; 3) analysis of effects of weather on musculoskeletal pain in a cohort of patients, with daily questionnaires and geo-located exposures collected through a smartphone app. The case time series design combines several advantages, such as an individual-level setting, strict confounding control, modelling flexibility, and computational efficiency. Its longitudinal structure allows investigation of complex temporal dependencies and the definition of individual risk profiles, making full use of new big data resources for environmental studies.
展开▼
