To the Editor:

We read with great interest the paper by Bergman, Mackay, and Pell (Reference Bergman, Mackay and Pell2021) about dementia in veterans and non-veterans. Authors found that there was no evidence in the subsequent risk of dementia between veterans and non-veterans (Bergman et al., Reference Bergman, Mackay and Pell2021). This study has provided an excellent basis for a follow-up study of the association of veterans with dementia. However, there are some concerns we would like to discuss further.

First, we note that there's possibility of important residual confounders. We suggest further match or adjust known risk factors of dementia to rule out residual confounding bias, such as hypertension, diabetes, hypercholesterolemia, obesity, and cardiovascular diseases. Further subgroup analysis on these important covariates for clinical applications would be appreciated (Baumgart et al., Reference Baumgart, Snyder, Carrillo, Fazio, Kim and Johns2015; Shang et al., Reference Shang, Fratiglioni, Marseglia, Plym, Welmer, Wang and Xu2020). Additionally, it would be necessary for further matching or stratified analysis of periodontal disease and chronic inflammatory diseases (Lee et al., Reference Lee, Lee, Hu, Huang, Chao, Lin and Chen2016; Ma et al., Reference Ma, Hasturk, Carreras, Dedeoglu, Veeravalli, Huang and Wei2021).

Second, we also concern about the coding accuracy of ICD-9 and -10 for dementia, post-traumatic stress disorder, and mood disorder that may be underdiagnosed in many claim-based databases. We suggest that the authors to provide validation data or references for the definition of outcome to avoid information bias.

We are grateful for the work done by the authors and looking forward to their response.