Planned missing data in early literacy interventions: A replication study with an additional gold standard

Introduction In a digital early literacy intervention RCT, children born late preterm fell behind peers when in a control condition, but outperformed them when assigned to the intervention. Results did however not replicate previous findings. Replication is often complicated by resource quality. Gold Standard measures are generally time-intensive and costly, while they closely align with, and are more sensitive to changes in, early literacy and language performance. A planned missing data approach, leaving these gold standard measures incomplete, might aid in addressing the origin(s) of non-replication. Methods Participants after consent were 695 p Dutch primary school pupils of normal and late preterm birth. The high-quality measures, in additional to simpler but complete measures, were intentionally administered to a random subsample of children. Five definitions of gold standard alignment were evaluated. Results Two out of five gold standard levels improved precision compared to the original results. The lowest gold standard level did not lead to improvement: precision was actually diminished. In two gold standard definitions, an alphabetical factor and a writing-only factor the model estimates were comparable to the original results. Only the most precise definition of the gold standard level replicated the original results. Conclusion Gold standard measures could only be used to improve model efficiency in RCT-designs under sufficiently high convergent validity.

Three-Form Split Questionnaire Design for Panel Surveys

Longitudinal or panel surveys are effective tools for measuring individual level changes in the outcome variables and their correlates. One drawback of these studies is dropout or nonresponse, potentially leading to biased results. One of the main reasons for dropout is the burden of repeatedly responding to long questionnaires. Advancements in survey administration methodology and multiple imputation software now make it possible for planned missing data designs to be implemented for improving the data quality through a reduction in survey length. Many papers have discussed implementing a planned missing data study using a split questionnaire design in the cross-sectional setting, but development of these designs in a longitudinal study has been limited. Using simulations and data from the Health and Retirement Study (HRS), we compare the performance of several methods for administering a split questionnaire design in the longitudinal setting. The results suggest that the optimal design depends on the data structure and estimand of interest. These factors must be taken into account when designing a longitudinal study with planned missing data.

Reflection on modern methods: planned missing data designs for epidemiological research

Taking advantage of the ability of modern missing data treatments in epidemiological research (e.g. multiple imputation) to recover power while avoiding bias in the presence of data that is missing completely at random, planned missing data designs allow researchers to deliberately incorporate missing data into a research design. A planned missing data design may be done by randomly assigning participants to have missing items in a questionnaire (multiform design) or missing occasions of measurement in a longitudinal study (wave-missing design), or by administering an expensive gold-standard measure to a random subset of participants while the whole sample is administered a cheaper measure (two-method design). Although not common in epidemiology, these designs have been recommended for decades by methodologists for their benefits—notably that data collection costs are minimized and participant burden is reduced, which can increase validity. This paper describes the multiform, wave-missing and two-method designs, including their benefits, their impact on bias and power, and other factors that must be taken into consideration when implementing them in an epidemiological study design.