Controlled quality improvement studies—different types of controls take care of different types of confounding
| Type of control | Description | What it controls for (and what not) |
| Location-based controls | The control is another location that is similar to the intervention location but has not received the intervention, for example, another ward in the same hospital or another similar hospital. | Controls for confounding factors that will affect both locations, but not factors uniquely tied to the intervention location. |
| Characteristic-based controls | If the intervention targets specific patient groups defined by their characteristics, for example, elderly, patients with a specific diagnosis, then non-targeted groups may act as controls. | Controls for confounding factors that will affect both patient groups, but not factors specific for the intervention group. If the characteristic defining who is intervention/control (eg, age) is a confounder, then this cannot be controlled for. |
| Historical controls | This type of control can only be used if such historical data are available (and with sufficient completeness). | Potentially, it might help to control for seasonal effects, but cannot control for confounding factors that are specific for the period when the intervention was implemented (eg, changes in case mix). |
| Outcome controls | The control is another outcome in the same population not targeted by the intervention (or expected to be not very sensitive for the intervention). | Controls for those confounding factors that would affect both the intervention and control outcome, but not those unique for the intervention outcome. |