Table 3 Logistic regression with possible factors that help to find the people with dementia

From: How can we better use Twitter to find a person who got lost due to dementia?

  Odds ratio (95% CI) p-value
Univariate analyses with logistic regression   
 Age, (below 70 as reference)   
 70–79 6.76 (1.3135.0) 0.023
 80 or above 0.15 (0.040.60) 0.007
 Gender, male 2.22 (0.58–8.49) 0.243
 Average no. of tweet-writers 1.09 (0.91–1.30) 0.340
 Average no. of retweeters 1.01 (0.98–1.04) 0.440
 Average no. of followers (in logarithmic scale) 1.34 (1.031.74) 0.031
 Original tweet posted by police departments (Yes/No) 7.80 (0.92–65.8) 0.103
 Original tweet posted by media organisations (Yes/No) 5.33 (1.44–19.8) 0.112
 Original tweet with photo (Yes/No) 6.75 (1.5429.6) 0.011
 Original tweet with webpage links (Yes/No) 5.67 (1.4522.1) 0.012
 Original tweet with hashtags (Yes/No) 1.33 (0.39–4.55) 0.646
 Original tweet that mentioned 'Alzheimer' (Yes/No) 6.26 (0.74–53.1) 0.093
 Original tweet that mentioned 'dementia' (Yes/No) 0.16 (0.02–1.36) 0.093
 Original tweet that mentioned 'police' (Yes/No) 3.32 (0.80–13.7) 0.098
 No. of retweet / each original tweet 1.01 (0.97–1.05) 0.504
Multivariate logistic regression model (Stepwise a )   
 Age, over 80 0.08 (0.010.53) 0.008
 Original tweets posted by police departments 25.1 (1.14554) 0.041
 Original tweets with photo 34.3 (2.55462) 0.008
  1. Remarks:
  2. aVariables with p-valve <0.2 from the univariate analyses were selected into the multivariate logistic regression model. The most significant predictive factors were selected by a stepwise approach where the variables were excluded one-by-one until the best fitted model is obtained with the minimum value of Akaike information criterion (AIC). All significant results (p-value <0.05) were shown in bold