How to Better Estimate Bedtime in Actigraphic Records
When a patient forgets to hit the event mark button on an actigraph, there are still ways that you can determine an accurate “lights out” point.
Defining the sleep periods within actigraphic records can be the toughest part of scoring data from these popular devices. Sure, the proprietary algorithms of each actigraph provider can estimate when the patient fell asleep. But when did they go to bed?
Home actigraphic recordings are burdened by the very thing that makes them convenient and popular. Actigraphs collect data in an uncontrolled environment. So, did the patient fall asleep in front of the TV for an hour and then go to bed? Or did they go to bed and then read for a time before turning the light off and trying to sleep? The accurate definition of the sleep period (lights off to lights on) affects many of the actigraphically derived variables, not the least of which is latency to persistent sleep (LPS).
The following are factors I take into consideration when estimating the “lights-off” point in an actigraph recording.
- Light signals: Not every actigraph has an ambient light sensor. If yours does, use it to see if you can detect the literal “lights-off” point just prior to primary block of sleep. You’re looking for a precipitous and sustained drop-off in ambient light.
- Event marks: When you have a compliant patient, event marks can be extremely valuable and your best indication of “going to bed with the intention of sleeping.” But since event buttons can sometimes be marked accidentally, or the patient may decide to press the button to indicate something other than originally intended, make sure the event mark location makes sense with regard to the activity pattern.
- Sleep diaries: This information can be very accurate or totally inaccurate, and there’s little way of gauging the credibility of diary data unless entries are made in real time and time-stamped electronically (as in an electronic diary). Otherwise, the accuracy of a paper diary can be measured against the activity record for each night.
- Context from within the data set: In most applications, there should be more than 24 hours of data. Multinight recording is one of the strengths of actigraphy. So make use of it. Does the patient appear to have a habitual bedtime? This is a good starting point.
- Activity patterns and threshold: Another clue that a patient has gone to bed is usually a precipitous drop in activity levels. Sometimes, with no other input, this is the only information with which you will be able to estimate the sleep period. This technique assumes that the patient is fairly active just prior to the sleep period. However, if the patient reads in bed, the earlier drop in high activity levels could be a false indicator, appearing to be an extended period of latency to persistent sleep. Light levels, event marks, and diaries would be important to clarify this. Some actigraph scoring programs have an algorithm included that provides a first-pass estimate of a probable sleep period based on activity levels. Check that you are always using consistent settings (if there are any) and, if possible, modify this initial sleep period estimate if you have credible information from another source. Also note that daytime activity levels vary between populations and individuals. Choose a threshold that makes sense given your patient’s behavior.
Consistency counts! Set up rules for scoring within your facility, including what to do when a lights-off point is not clear. In general, the point of activity drop-off has the final say unless the light value drop-off, event mark, or diary indication occurs after this point and before the first block of persistent sleep (as indicated by the sleep scoring algorithm). When a “lights off” point cannot be confidently determined, the start of the sleep period is set at the minute before the first block of persistent sleep and latency is not reported. When in doubt about setting a lights-out point, consult with a knowledgeable coworker to achieve a consensus and document your reasoning in the patient’s actigraphic report. Make sure the actigraph scorers are trained in your facility’s methods and set up a schedule for rescoring a percentage of actigraph recordings by a second scorer to establish reliability within your facility.
Karina Palafox is clinical data manager at Clinilabs Inc, where she is the lead data analyzer for actigraphy.