Personal Informatics

Measuring Strategy for Healthy Eating & Sleep

Up until I took the Personal Informatics class, I had never really taken my eating habits and their impact on my life seriously. I have always eaten without considering the effects of food on my body. There are some effects which show up instantly like drowsiness after overeating or a sense of anxiety and restlessness after drinking coffee. But the long term and slow effects often go unnoticed. As someone who erratically suffers from insomnia and has irregular eating habits, I decided to focus on the challenge of measuring the impact of blood sugar levels on sleep. This will help me understand how my body reacts to different foods and establish relationships between sleep and glucose intake.

In order to gather information corresponding to this issue, I decided to break down the measurement strategy into two separate units without wanting to lose the relationship between them. The first part will cover the measurement of blood sugar levels in the body prior to sleeping and after waking up. This can be done relatively accurately using the medical equipment available today which promises acceptable and accurate results. FDA approved Glucometers are required to have a tolerance of ±20 points for glucose levels above 75 mg/dl and ±15 points below that level (“Blood Glucose Meters 2014,” n.d.). The second part is more subjective and depends on a combination of self-reporting techniques and quantitative sleep tracking devices. The sleep tracking device will keep track of sleep duration, sleep phases and quality of sleep and provide quantitative data over a period of time. As part of the self-reporting regime, the user will have to keep a record of his/her emotional state before going to sleep and after waking up. This will provide us with a qualitative aspect of the user’s perspective regarding his/her sleeping activity. This will entail recording the user’s emotional state whether he feels fresh or drowsy and his satisfaction level with his sleep. I have not included the type of food intake as a parameter as it diverges into a broad array of other topics which are entirely different metrics on their own. Also, the blood sugar levels provide a fair correlation with varying glucose levels in the body and the hormonal interruptions to sleeping cycles. A combination of measuring these 3 attributes will help us gather data around glucose, sleep and emotional feedback of a user which we will be able to analyze and draw conclusions from. Hence, a typical use case scenario represents a user recording his blood glucose level before he sleeps, during which he is wearing the sleep tracking device amongst the likes of Misfits Shine or Rem-Fit which will actively track the duration of sleep and phase cycles corresponding to it. When the user wakes up, the user will make a log entry onto a pre-formatted report which would ask him/her to rate the quality of sleep on a semantic differential interval scale and circle an emotion from the emotions listed on the log report. This data will have a timestamp on it which can later be collected and synthesized to gain insights about the possible interconnection between glucose level and its impact on sleep.

This approach collects three distinct types of user data from the user, two of which are quantitative in nature and the last one is qualitative. In terms of usability, the glucose testing methods of today can seem daunting due to the fact that user has to prick a finger and lay down a drop of blood on the strip and has been prevalent in the society. Although prick-less methods have been devised, they are expensive and not very widespread. This might cause concerns amongst the users and needs to be addressed using less invasive and painless methods. The sleep monitoring devices today are non-invasive and don’t need any input from the user except for having it strapped to their hand or placed below their bed. They do not entail any burden on the user and provide a non-intrusive experience to the user. In terms of the self-reporting of the quality of sleep, I intend on keeping the log to a minimal by just including quality of sleep and the emotional response of the user. Although this might seem to be a chore, it is a small one and does not seem excessive. I believe that a combination of these techniques will provide us with a rich data and help us establish relationships between sugar levels and quality of sleep. There is also a relationship between glucose levels with sleep where blood sugar levels rise during sleep(“The Somogyi Effect & The Dawn Phenomenon | Cleveland Clinic,” n.d.). This can help us manage irregularities in data by correlating glucose levels before and after sleep. The self-reporting technique will help the user analyze how his mood is affected by the duration of sleep and will be better able to understand the quality of sleep and how that is affected by the food he consumes. The validity of these methods can be visualized in the validity matrix as shown below. The major concern which stands out is that of content stability over time as sleeping times and eating habits can vary greatly from a weekday and a weekend. There might also be circumstances where the user has more work to finish and has to stay up late or wake up early in order to complete it. This might skew our results as it won’t be considered a normal scenario. Another concern which stands out is the capability and accuracy of sleep monitoring devices. The devices available today mainly depend on motion sensors and heart rate to track of sleep cycles and phases like REM. These are not completely accurate and do bring a degree of error with them(“How Do Sleep Trackers Work And Are They Reliable? ” 2015). These devices are accurate with respect to the duration of sleep but not how they keep track of sleep cycles. We can have a tolerance for random error within this data and address it statistically. The self-reporting of emotion proves to be an effective qualitative assessment of sleep but due to its subjective nature can not escape from Role selection and response bias as it might be affected by events in the life of the user.


Although there are the mentioned concerns with the validity of these methods, no experiment is free from error. Some of these issues can be addressed statistically as random error over a period of time by collecting a considerable number of results and aggregating deviations out of the findings. Some of the devices considered while designing this strategy were the Misfit Shine 2, Dexcom G5, Rem-fit which are all FDA approved now and can be reliably used to get plausible data from the users which can be used for our synthesis.


Glucose Level Sleep Tracking Emotional Response TOTAL
Examination 0 0 1 1
Role Selection 0 0 1 1
Measurement Effects 1 0 0 1
Response Bias 0 0 1 1
Interviewer Effects 0 0 0 0
Change in Instrument 0 0 1 1
Population Restrictions 1 0 0 1
Pop. Variance over Time 0 0 0 0
Pop. Change over Area 1 0 0 1
Measureability 0 0 0 0
Content Stability over Time 1 1 0 2
Content Stability over Area 0 0 0 0
Dross Rate 0 1 0 1
Access to Desc. Cues 0 0 0 0
Replicability 0 0 0 0
User Burden 0 0 1 1



Blood Glucose Meters 2014. (n.d.). Retrieved October 26, 2016, from

Dexcom G5 Mobile CGM System | Glucose on your phone. (2016, March 3). Retrieved October 26, 2016, from

How Do Sleep Trackers Work And Are They Reliable? (2015, December 15). Retrieved October 26, 2016, from

Misfit Shine 2 Advanced Fitness + Sleep Tracker – Misfit. (n.d.). Retrieved October 26, 2016, from

REM-Fit Active 100 Bundle for Halloween 2016 | (n.d.). Retrieved October 26, 2016, from

The Somogyi Effect & The Dawn Phenomenon | Cleveland Clinic. (n.d.). Retrieved October 26, 2016, from

Leave a Reply

Your email address will not be published. Required fields are marked *