In this paper, authors describe a novel method to infer mental and physical state of individuals based on their digital behavioral changes, mainly through the data collected from smart phone usage patterns. About 200 students in a dorm were given smartphones which collected usage patterns such as number of daily contacts and the diversity of daily connections and movements. Smartphones were also programmed to collect the results of a daily survey on mental and physical state of the student. The survey results along with the daily variations in digital behavioral metrics were fed into a classification model, the result of which could be used to predict individual's health status from daily monitored digital behavior signals.
Th authors mentioned two important applications for this work: 1- Preventive healthcare and 2- Upgrade into current epidemiological models which assume individual behavior in the society stays the same after infection. A possible future work is to extend this classification model so that it uses other forms of digital data such as social networking usage patterns.
Strengths:
1- The general idea along with the methodology are novel.2- Medical research community has often discussed the possibility of a link between physical health deterioration and stress/depression. The temporal (causal) analysis between digital behavioral changes, depression/stress and high-intensity symptoms (such as fever and influenza) verifies the link between the mental and physical state.
Weaknesses:
1- Authors fail to mention ethical and privacy concerns regarding constant monitoring of an individual's digital behavior and lifestyle.2- Individual outliers with exceptionally high or low social interaction/exposure, might have skewed the classification results. Such individuals will exhibit large or no entropy decrease when infected. Authors fail to discuss how personal traits affect their methodology. Perhaps, the classification model and the final trained model should account for personal traits as well as digital behavioral changes.
3- Temporary events such as traveling, excessive mobility due to temporary work requirements and other similar lifestyle changes can produce the same effect in monitored digital behavior as does sickness or infection.
