Version June 2024
| Data source | Pros | Cons |
| Health insurance claims data | Capture information from large populations | Long-term follow-up may be limited, if patients change insurance providers |
| Provide longitudinal information | Information may be incomplete or unreliably recorded, depending on how clinicians have assigned billing codes | |
| Electronic health record (EHR) data | Can provide detailed information that can sometimes be missing from insurance claims | Data can be incomplete or inaccurate, depending on clinician documentation |
| If a patient receives care at multiple institutions, data can be incomplete | ||
| Patient registries | Can provide specific and detailed information about patients with a given condition or receiving a given treatment | Can be subject to recall bias, selective reporting, and missing data |
| May only represent a highly select patient population | ||
| Patient generated data | Can be continuously measured, for example with wearable devices | Can be subject to recall bias, selective reporting, and missing data |
| Can be a good source of quality of life and adverse effects data | ||
| Vital statistics (ie, from death certificates) | Are a good source for date of death | Individual causes of death are often inaccurately recorded (instead categories of death, eg, death from cardiovascular causes, are more accurate and such data can be collected) |
| Can be linked to other data sources such as claims data | ||
| Linkage across data types | Multiple sources of data can be linked together, addressing gaps in any individual source | Linking can be costly and difficult to perform |
| Data are often collected for different purposes from different sources, making it difficult to consolidate in a reliable way |
آیا می خواهید مدیلیب را به صفحه اصلی خود اضافه کنید؟
