Back

Reliability of self-reported risk factors for age-related brain disease

Tack, R. W.; Senff, J. R.; Kimball, T. N.; Prapiadou, S.; Yechoor, N.; Rosand, J.; Singh, S. S.; Anderson, C. D.

2024-04-26 neurology
10.1101/2024.04.25.24306370 medRxiv
Show abstract

Importancemodifiable risk factors, including hypertension, hypercholesterolemia, diabetes, kidney disease, hearing impairment, and being overweight, increase risk of stroke, dementia, and late life depression (LLD), but may be poorly understood by patients. ObjectiveTo aid in the design of brain health risk stratification tools, we aimed to determine whether self-reported modifiable risk factors can identify at-risk patients on a population level. Designcohort study using cross-sectional data of ten iterations of the National Health and Examination Surveys (NHANES) from 1999 to 2018. Analyses were conducted between February and April of 2024. SettingUS population-based cohorts Participantsall participants who had both questionnaire data and objective measurements available. Exposurewe collected data on answers to common questions and compared these to objectively measured risk factors for age-related brain disease. Main outcomes and measureswe compared answers to simple questions to objectively measured risk factors. We reported means and 95% confidence intervals of objective measurements, created confusion matrices to determine common metrics, and compared performance of individual questions. We defined a question reliable if performance, measured through F1 scores, was > 0.7, as moderately reliable if F1 score was 0.5-0.7 and as unreliable if F1 score was < 0.5. ResultsParticipants with both objective measurements and questionnaire data ranged from 16,966 (median age 32, 9,113 [20%] black, 19,993 [45%] white) in hearing impairment to 63,834 (median age 39, 14,156 [22%] black, 25,754 [40%] white) in diabetes. Mean values of objective measurements were significantly increased across all risk factors in patients that responded "Yes" to common questions on their presence compared to those who answered "No". Performance of questions in identifying at-risk patients measured by F1 score was 0.25 in kidney disease, 0.44 in hypercholesterolemia, 0.56 in hypertension, 0.59 in hearing impairment, 0.71 in diabetes and 0.81 in being overweight. Conclusion and relevancein this study of multiple cohorts of up to 63,834 Americans, self-reported awareness of meeting clinical criteria for diabetes or being overweight were reliable, hypertension or hearing impairment were moderately reliable, and hypercholesterolemia or kidney disease were unreliable. These results could be used to guide construction of risk factor screening tools.

Matching journals

The top 8 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.