Srinivasan Rajaraman

Predictive Monitoring for Improved Management of Glucose Levels

Background: Recent developments and expected near-future improvements in continuous glucose monitoring (CGM) devices provide opportunities to couple them with mathematical forecasting models to produce predictive monitoring systems for early, proactive glycemia management of diabetes mellitus patients before glucose levels drift to undesirable levels. This article assesses the feasibility of data-driven models to serve as the forecasting engine of predictive monitoring systems. Methods: We investigated the capabilities of data-driven autoregressive (AR) models to (1) capture the correlations in glucose time-series data, (2) make accurate predictions as a function of prediction horizon, and (3) be made portable from individual to individual without any need for model tuning. The investigation is performed by employing CGM data from nine type 1 diabetic subjects collected over a continuous 5-day period. Results: With CGM data serving as the gold standard, AR model-based predictions of glucose levels assessed over nine subjects with Clarke error grid analysis indicated that, for a 30-minute prediction horizon, individually tuned models yield 97.6 to 100.0% of data in the clinically acceptable zones A and B, whereas cross-subject, portable models yield 95.8 to 99.7% of data in zones A and B. Conclusions: This study shows that, for a 30-minute prediction horizon, data-driven AR models provide sufficiently-accurate and clinically-acceptable estimates of glucose levels for timely, proactive therapy and should be considered as the modeling engine for predictive monitoring of patients with type 1 diabetes mellitus. It also suggests that AR models can be made portable from individual to individual with minor performance penalties, while greatly reducing the burden associated with model tuning and data collection for model development.

A unified mathematical model to quantify performance impairment for both chronic sleep restriction and total sleep deprivation.

Performance prediction models based on the classical two-process model of sleep regulation are reasonably effective at predicting alertness and neurocognitive performance during total sleep deprivation (TSD). However, during sleep restriction (partial sleep loss) performance predictions based on such models have been found to be less accurate. Because most modern operational environments are predominantly characterized by chronic sleep restriction (CSR) rather than by episodic TSD, the practical utility of this class of models has been limited. To better quantify performance during both CSR and TSD, we developed a unified mathematical model that incorporates extant sleep debt as a function of a known sleep/wake history, with recent history exerting greater influence. This incorporation of sleep/wake history into the classical two-process model captures an individuals capacity to recover during sleep as a function of sleep debt and naturally bridges the continuum from CSR to TSD by reducing to the classical two-process model in the case of TSD. We validated the proposed unified model using psychomotor vigilance task data from three prior studies involving TSD, CSR, and sleep extension. We compared and contrasted the fits, within-study predictions, and across-study predictions from the unified model against predictions generated by two previously published models, and found that the unified model more accurately represented multiple experimental studies and consistently predicted sleep restriction scenarios better than the existing models. In addition, we found that the model parameters obtained by fitting TSD data could be used to predict performance in other sleep restriction scenarios for the same study populations, and vice versa. Furthermore, this model better accounted for the relatively slow recovery process that is known to characterize CSR, as well as the enhanced performance that has been shown to result from sleep banking.

A new metric for quantifying performance impairment on the psychomotor vigilance test

We have developed a new psychomotor vigilance test (PVT) metric for quantifying the effects of sleep loss on performance impairment. The new metric quantifies performance impairment by estimating the probability density of response times (RTs) in a PVT session, and then considering deviations of the density relative to that of a baseline‐session density. Results from a controlled laboratory study involving 12 healthy adults subjected to 85 h of extended wakefulness, followed by 12 h of recovery sleep, revealed that the group performance variability based on the new metric remained relatively uniform throughout wakefulness. In contrast, the variability of PVT lapses, mean RT, median RT and (to a lesser extent) mean speed showed strong time‐of‐day effects, with the PVT lapse variability changing with time of day depending on the selected threshold. Our analysis suggests that the new metric captures more effectively the homeostatic and circadian process underlying sleep regulation than the other metrics, both directly in terms of larger effect sizes (4–61% larger) and indirectly through improved fits to the two‐process model (9–67% larger coefficient of determination). Although the trend of the mean speed results followed those of the new metric, we found that mean speed yields significantly smaller (∼50%) intersubject performance variance than the other metrics. Based on these findings, and that the new metric considers performance changes based on the entire set of responses relative to a baseline, we conclude that it provides a number of potential advantages over the traditional PVT metrics.

A biomathematical model of the restoring effects of caffeine on cognitive performance during sleep deprivation.

RATIONALE While caffeine is widely used as a countermeasure to sleep loss, mathematical models are lacking. OBJECTIVE Develop a biomathematical model for the performance-restoring effects of caffeine in sleep-deprived subjects. METHODS We hypothesized that caffeine has a multiplicative effect on performance during sleep loss. Accordingly, we first used a phenomenological two-process model of sleep regulation to estimate performance in the absence of caffeine, and then multiplied a caffeine-effect factor, which relates the pharmacokinetic-pharmacodynamic effects through the Hill equation, to estimate the performance-restoring effects of caffeine. RESULTS We validated the model on psychomotor vigilance test data from two studies involving 12 subjects each: (1) single caffeine dose of 600mg after 64.5h of wakefulness and (2) repeated doses of 200mg after 20, 22, and 24h of wakefulness. Individualized caffeine models produced overall errors that were 19% and 42% lower than their population-average counterparts for the two studies. Had we not accounted for the effects of caffeine, the individualized model errors would have been 117% and 201% larger, respectively. CONCLUSIONS The presented model captured the performance-enhancing effects of caffeine for most subjects in the single- and repeated-dose studies, suggesting that the proposed multiplicative factor is a feasible solution.

Predicting human subcutaneous glucose concentration in real time: A universal data-driven approach

Continuous glucose monitoring (CGM) devices measure and record a patients subcutaneous glucose concentration as frequently as every minute for up to several days. When coupled with data-driven mathematical models, CGM data can be used for short-term prediction of glucose concentrations in diabetic patients. In this study, we present a real-time implementation of a previously developed offline data-driven algorithm. The implementation consists of a Kalman filter for real-time filtering of CGM data and a data-driven autoregressive model for prediction. Results based on CGM data from 3 different studies involving 34 type 1 and 2 diabetic patients suggest that the proposed real-time approach can yield ∼10-min-ahead predictions with clinically acceptable accuracy and, hence, could be useful as a tool for warning against impending glucose deregulation episodes. The results further support the feasibility of “universal” glucose prediction models, where an offline-developed model based on one individuals data can be used to predict the glucose levels of any other individual in real time.