Alternative approaches for analysing repeated measures data that are missing not at random

TL;DR

This paper explores two novel methods for analyzing repeated measures data with missing not at random, addressing limitations of traditional assumptions and providing practical alternatives with real-world application.

Contribution

It introduces two alternative strategies for handling non-random missing data, including a difference-in-differences approach and an instrumental variable correction, expanding analytical options.

Findings

The proposed methods outperform traditional approaches in certain missing data scenarios.

Application to Framingham Heart Study data demonstrates practical utility.

The methods effectively address violations of common missing data assumptions.

Abstract

We consider studies where multiple measures on an outcome variable are collected over time, but some subjects drop out before the end of follow up. Analyses of such data often proceed under either a 'last observation carried forward' or 'missing at random' assumption. We consider two alternative strategies for identification; the first is closely related to the difference-in-differences methodology in the causal inference literature. The second enables correction for violations of the parallel trend assumption, so long as one has access to a valid 'bespoke instrumental variable'. These are compared with existing approaches, first conceptually and then in an analysis of data from the Framingham Heart Study.