Practical Strategies for Extreme Missing Data Imputation in Dementia Diagnosis

Abstract

Accurate computational models for clinical decision support systems require clean and reliable data, but in clinical practice, data are often incomplete. Hence, missing data could arise not only from training datasets but also test datasets which could consist of a single undiagnosed case, an individual. Many popular methods of handling missing data are unsuitable for handling such missing test data. This work addresses the problem by evaluating multiple imputation and classification workflows based not only on diagnostic classification accuracy but also computational cost. In particular, we focus on dementia diagnosis due to long time delays, high variability, high attrition rates and lack of practical data imputation strategies in its diagnostic pathway. We identified and replicated the extreme missingness structure of data from a memory clinic on a larger open dataset, with the original complete data acting as ground truth. Overall, we found that computational cost, but not accuracy, varies widely for various imputation and classification approaches. Particularly, we found that iterative imputation on the training dataset combined with a reduced-feature classification model provides the best approach, compromising speed and accuracy. Taken together, this work has elucidated important factors to be considered when developing or maintaining a dementia diagnostic support system, which can be generalized to other clinical or medical domains, particularly with extreme data missingness.

Impact of different diagnostic measures on drug class association with dementia progression risk: a longitudinal prospective cohort study

“…CDRSOB scores, though comparatively less comprehensive, form a part of clinical diagnosis which leads to high correlation between the two. Due to unavailability of clinical diagnosis in some datasets, or a preference for more quantitative analysis to increase precision and reduce subjectivity, researchers often opt for cognitive scores (Bucholc et al, 2019; Ding et al, 2018; McCombe et al, 2022). However, we must be aware of the implications of choosing a particular diagnostic measure and be careful regarding how we interpret the results.…”

Distinguishing Lewy Body Dementia from Alzheimer’s Disease using Machine Learning on Heterogeneous Data: A Feasibility Study

“…This dataset had 7% missing values, which were imputed, separately from the Diagnosis column to avoid “double-dipping” [15], using the missForest algorithm [16] and R package [17]. The missForest algorithm was used as previous work by the authors found it to be an accurate and flexible imputation algorithm [18].…”

Missing Data Imputation

“…In another work on CDSS, the authors propose left-center-right method to fill unobserved biomarkers by using existing biomarker measurements from individual patient's visit records (Gupta et al, 2020). In (McCombe et al, 2021), the authors discuss CDSS for diagnosis of Dementia on extremely missing data (i.e. more than 50% of values are missing in over half of the attributes) in both training and test datasets.…”