Helping Researchers Use Statistical Methods for Analyzing Data When Treatments Change Over Time

This implementation project is complete.

PCORI implementation projects promote the use of findings from PCORI-funded studies in real-world healthcare and other settings. These projects build toward broad use of evidence to inform healthcare decisions.

	This PCORI-funded implementation project helped researchers use statistical methods to analyze data on the effectiveness of treatments that change over time.
In randomized controlled trials, or RCTs, researchers assign patients by chance to different treatments to compare their benefits and harms. RCTs are the most reliable way for researchers to learn how well treatments work for patients. But RCTs aren’t well suited for comparing treatments that change over time.

What was the goal of this implementation project?

Dynamic treatment is when doctors change a patient’s treatment over time based on how the patient responds. But it can be hard for researchers to analyze data about how effective dynamic treatments are. A PCORI-funded project found that two statistical methods worked equally well to analyze data on dynamic treatments. The two methods were inverse probability weighting and the parametric g-formula.

This project helped researchers use the two methods, collectively called g-methods, to study dynamic treatments using data from patients’ health records.

What did this project do?

The project team worked with eight groups of researchers and statisticians at four hospital systems. They used the g methods to compare types of care or treatments for topics such as:

• How to assess the severity of traumatic injuries in children
• Use of medicines for pain relief
• Pneumonia
• Arthritis
• Differences in health outcomes by race and ethnicity for mothers and children

The project team provided training on use of the g-methods through webinars. The team also offered coaching and consulting sessions for the researchers.

At the end of the project, researchers taking part completed a survey about their use of the g-methods. More than three-fourths said that g-methods worked well for their research topics. Half indicated that they intend to use the g-methods in future research projects.

For more information about this project, please contact the Project Team at [email protected].