A generalized interrupted time series model for assessing complex health care interventions

Area of Science:

• Health Services Research
• Biostatistics
• Epidemiology

Background:

• Assessing complex interventions' impact on health outcomes is crucial for healthcare and policy.
• Interrupted time series (ITS) designs are quasi-experimental methods for retrospective intervention impact analysis.
• Existing ITS statistical models predominantly handle continuous outcomes, limiting analysis of discrete health data.

Purpose of the Study:

• To propose the Generalized Robust ITS (GRITS) model for analyzing discrete health outcomes (binary, count).
• To expand ITS methodology for outcomes following exponential family distributions.
• To formally implement change point testing for discrete ITS.

Main Methods:

• Developed the Generalized Robust ITS (GRITS) model for discrete outcomes.
• The GRITS model accommodates outcomes from the exponential family of distributions.
• Methodology allows for change point estimation, multi-unit data pooling, and pre/post-intervention comparisons.

Main Results:

• The GRITS model effectively models binary and count health outcomes in ITS.
• The methodology successfully tested for and estimated change points in discrete ITS data.
• Demonstrated ability to borrow information across multiple units and assess intervention effects.

Conclusions:

• The GRITS model significantly advances ITS methodology for discrete health outcomes.
• This approach provides a robust framework for evaluating complex interventions using non-continuous data.
• The GRITS model offers enhanced capabilities for analyzing patient falls and similar health events.