Premium
Machine learning predictions of depression from time spent in cadence bands
Author(s) -
Forrester Jenna,
Thomas Diana,
Adams Bryan,
Stice Eric,
Arigo Danielle,
Scioletti Michael,
TudorLocke Catrine
Publication year - 2019
Publication title -
the faseb journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.709
H-Index - 277
eISSN - 1530-6860
pISSN - 0892-6638
DOI - 10.1096/fasebj.2019.33.1_supplement.693.18
Subject(s) - cadence , depression (economics) , logistic regression , cluster (spacecraft) , linear regression , regression analysis , statistics , national health and nutrition examination survey , psychology , regression , machine learning , artificial intelligence , computer science , physical therapy , medicine , mathematics , physical medicine and rehabilitation , population , environmental health , economics , macroeconomics , programming language
Background Depression impairs normal human functionality and can cause severe interruptions of normal everyday activities affecting productivity at work and at home. Physical activity is often proposed as an effective treatment for depression, however the relationship between physical activity and depression has not been evaluated with unsupervised machine learning algorithms. Unsupervised machine learning algorithms determine relationships between all variables without requiring user input information on which variables should be explanatory. Methods The National Health and Nutrition Examination Survey 2005–2006 (NHANES) dataset, which includes demographic characteristics, depression scores from the Patient Health Questionnaire (PHQ‐9), and 2‐weeks of accelerometer‐determined step data collected from 4,000 participants. Movement intensity using peak 30‐minute cadence (an index of best natural effort). Total movement volume was retained as steps/day. Linear and logistic regression techniques were employed to predict depression. We also performed a k‐means cluster and principal component analysis to identify those predictors that contributed most to depression variance. Results The linear regression model revealed that an increase of one unit of peak 30‐min cadence predicts a decrease in sum of the depression scores by 0.03 units. Similarly, logistic regression indicated that an increase in peak 30‐min cadence by 20 units predicted a 1.5 times decrease in depression. The k‐means cluster analysis resulted in 4 distinct clusters. In clusters 1,2,3 and 4, 5.0%, 7.9%, 6.0% and 6.5% were categorized as depressed respectively. Cluster 2 represented the most sedentary cluster. Conclusions Machine learning models combined with rich accelerometer data can be used to rigorously quantify the relationship between step‐defined physical activity and depression. These models can personalize physical activity treatment recommendations f for depression in terms of steps/d and peak 30‐minute cadence. Support or Funding Information None This abstract is from the Experimental Biology 2019 Meeting. There is no full text article associated with this abstract published in The FASEB Journal .