Can a mathematical model predict an individual's trait‐like response to both total and partial sleep loss?

Abstract

SUMMARYHumans display a trait-like response to sleep loss. However, it is not known whether this trait-like response can be captured by a mathematical model from only one sleep-loss condition to facilitate neurobehavioural performance prediction of the same individual during a different sleep-loss condition. In this paper, we investigated the extent to which the recently developed unified mathematical model of performance (UMP) captured such trait-like features for different sleep-loss conditions. We used the UMP to develop two sets of individual-specific models for 15 healthy adults who underwent two different sleep-loss challenges (order counterbalanced; separated by 2-4 weeks): (i) 64 h of total sleep deprivation (TSD) and (ii) chronic sleep restriction (CSR) of 7 days of 3 h nightly time in bed. We then quantified the extent to which models developed using psychomotor vigilance task data under TSD predicted performance data under CSR, and vice versa. The results showed that the models customized to an individual under one sleep-loss condition accurately predicted performance of the same individual under the other condition, yielding, on average, up to 50% improvement over nonindividualized, group-average model predictions. This finding supports the notion that the UMP captures an individual's trait-like response to different sleep-loss conditions. IN TROD UCTI ONInsufficient sleep impairs alertness and neurobehavioural performance. Results from Van Dongen et al. (2004) and Rupp et al. (2012) showed that substantial interindividual variability exists with regard to response to sleep loss. More importantly, both studies also showed that this response was trait-like. Van Dongen et al. (2004) evaluated the same 21 individuals under three separate total sleep deprivation (TSD) challenges of 36 h each and observed that the neurobehavioural deficits resulting from sleep loss were stable within individuals across the three challenges. Rupp et al. (2012) showed that an individual's neurobehavioural response to 64 h of TSD was correlated positively with that individual's response to chronic sleep restriction (CSR) [seven consecutive days of 3 h nightly time in bed (TIB)]. In both studies, they quantified individual response to sleep loss by averaging neurobehavioural performance [i.e. psychomotor vigilance task (PVT) data] over the last 12-24 h of the sleep-loss condition, and used the intraclass correlation coefficient (ICC) to assess the extent to which an individual's vulnerability rank among a group of individuals was preserved across sleep-loss conditions. Large ICC values observed in the two studies suggest a high degree of trait preservation. In other words, performance on the last day of a particular sleep-loss challenge can predict accurately the relative rank of an individual for a subsequent sleep-loss challenge. However, such analyses provide little or no information regarding the temporal dynamics of an individual's performance, which is critical in operational settings.In the past, many biomathematical m...

2B‐Alert App: A mobile application for real‐time individualized prediction of alertness

“…The 2B‐Alert App generates individualized and group‐average alertness predictions for mean response time (RT) as the PVT statistic. We chose mean RT because it is one of the most frequently used PVT statistics (Basner & Dinges, ), and because it allows for the most accurate estimation of individual‐specific predictions of alertness under both TSD and CSR (Liu et al, ; Ramakrishnan et al, ). Nevertheless, the app also stores raw RT data.…”

“…We chose mean RT because it is one of the most frequently used PVT statistics (Basner & Dinges, 2011), and because it allows for the most accurate estimation of individual-specific predictions of alertness under both TSD and CSR (Liu et al, 2017;Ramakrishnan et al, 2015).…”

“…The UMP is well validated: we have assessed its predictions on 442 subjects from 14 different studies, including 24 different sleep/wake conditions (from 3 to 10 hr of sleep per night to 88 hr of TSD) and nine different caffeine conditions during sleep deprivation (from single/repeated doses of 100–600 mg). We have also shown that, when individually customized, the UMP captures the individual’s trait‐like response to sleep deprivation under different sleep‐deprivation challenges (Ramakrishnan et al, ).…”

Real‐time individualization of the unified model of performance

“…To date, such models have been used primarily as off‐line planning tools to predict the performance of an ‘average’ individual (Dawson et al ., ; Hursh et al ., ). However, given the large intersubject variability in the response to sleep loss (Van Dongen et al ., ), implementing these models in mobile computing smartphone devices would allow for individualized model customization and more accurate predictions (Ramakrishnan et al ., ).…”

“…Recently, our group developed the UMP, which predicts individual performance on PVT data accurately under conditions ranging from CSR to TSD and ‘learns’ an individual's trait‐like response to sleep loss (Ramakrishnan et al ., ). However, individualization of such models has thus far been achieved in a post‐hoc manner by fitting the model parameters to the complete set of PVT data collected during the course of an entire sleep‐loss challenge.…”