Using Topic Modeling to Develop Multi-level Descriptions of Naturalistic Driving Data from Drivers with and without Sleep Apnea.

One challenge in using naturalistic driving data is producing a holistic analysis of these highly variable datasets. Typical analyses focus on isolated events, such as large g-force accelerations indicating a possible near-crash. Examining isolated events is ill-suited for identifying patterns in continuous activities such as maintaining vehicle control. We present an alternative approach that converts driving data into a text representation and uses topic modeling to identify patterns across the dataset. This approach enables the discovery of non-linear patterns, reduces the dimensionality of the data, and captures subtle variations in driver behavior. In this study topic models are used to concisely described patterns in trips from drivers with and without untreated obstructive sleep apnea (OSA). The analysis included 5000 trips (50 trips from 100 drivers; 66 drivers with OSA; 34 comparison drivers). Trips were treated as documents, and speed and acceleration data from the trips were converted to "driving words." The identified patterns, called topics, were determined based on regularities in the co-occurrence of the driving words within the trips. This representation was used in random forest models to predict the driver condition (i.e., OSA or comparison) for each trip. Models with 10, 15 and 20 topics had better accuracy in predicting the driver condition, with a maximum AUC of 0.73 for a model with 20 topics. Trips from drivers with OSA were more likely to be defined by topics for smaller lateral accelerations at low speeds. The results demonstrate topic modeling as a useful tool for extracting meaningful information from naturalistic driving datasets.

Effects of Actigraphically Acquired Sleep Quality on Driving Outcomes in Obstructive Sleep Apnea Patients and Control Drivers: A Naturalistic Study.

We examined the effects of sleep quality on next day driving outcomes in a 3.5-month naturalistic driving study of 67 OSA and 47 matched control drivers. Sleep quality measures included total sleep time and sleep fragmentation from actigraphy. The driving outcomes included average speed, lateral control, longitudinal control, distraction, attention to driving- and non-driving related tasks. Sleep quality affected next day's driving performance differently for OSA and control drivers. Better sleep quality was associated with better lateral and longitudinal control during highway driving for control drivers. The reverse was true for OSA drivers. Similar effects were also seen in terms of distractions and attention to the driving task. These effects suggest improved sleep leads to greater risky driving and 'activation' among OSA drivers. Collectively, the findings suggest investment in long-term monitoring of sleep quality in commercial vehicle drivers both with and without sleep disorders may help manage safety risks.

Driving Safety and Fitness to Drive in Sleep Disorders.

Driving an automobile while sleepy increases the risk of crash-related injury and death. Neurologists see patients with sleepiness due to obstructive sleep apnea, narcolepsy, and a wide variety of neurologic disorders. When addressing fitness to drive, the physician must weigh patient and societal health risks and regional legal mandates. The Driver Fitness Medical Guidelines published by the National Highway Traffic Safety Administration (NHTSA) and the American Association of Motor Vehicle Administrators (AAMVA) provide assistance to clinicians. Drivers with obstructive sleep apnea may continue to drive if they have no excessive daytime sleepiness and their apnea-hypopnea index is less than 20 per hour. Those with excessive daytime sleepiness or an apnea-hypopnea index of 20 per hour or more may not drive until their condition is effectively treated. Drivers with sleep disorders amenable to pharmaceutical treatment (eg, narcolepsy) may resume driving as long as the therapy has eliminated excessive daytime sleepiness. Following these guidelines, documenting compliance to recommended therapy, and using the Epworth Sleepiness Scale to assess subjective sleepiness can be helpful in determining patients' fitness to drive.

CONTEXTUALIZING NATURALISTIC DRIVING DATA IN A RURAL STATE AMONG DRIVERS WITH AND WITHOUT OBSTRUCTIVE SLEEP APNEA.

In naturalistic studies, Global Positioning System (GPS) data and date/time stamps can link driver exposure to specific environments (e.g., road types, speed limits, night driving, etc.), providing valuable context for analyzing critical events, such as crashes, near crashes, and breaches of accelerometer limits. In previous work, we showed how to automate this contextualization, using GPS data obtained at 1 Hz and merging this with Geographic Information Systems (GIS) databases maintained by the Iowa Department of Transportation (DOT). Here we further demonstrate our methods by analyzing data from 80 drivers with obstructive sleep apnea (OSA) and 48 controls, and comparing the two groups with respect to several factors of interest. The majority of comparisons found no difference between groups, suggesting similar patterns of exposures to driving environments in OSA and control drivers. However, OSA drivers appeared to spend slightly more time on roads with annual traffic counts of 500-10,000 and less time driving on wider highways, during twilight, and on roads with 10,000-25,000 annual traffic counts.

Using kinematic driving data to detect sleep apnea treatment adherence.

People spend a significant amount of time behind the wheel of a car. Recent advances in data collection facilitate continuously monitoring this behavior. Previous work demonstrates the importance of this data in driving safety but does not extended beyond the driving domain. One potential extension of this data is to identify driver states related to health conditions such as obstructive sleep apnea (OSA). We collected driving data and medication adherence from a sample of 75 OSA patients over 3.5 months. We converted speed and acceleration behaviors to symbols using symbolic aggregate approximation and converted these symbols to pattern frequencies using a sliding window. The resulting frequency data was matched with treatment adherence information. A random forest model was trained on the data and evaluated using a held-aside test dataset. The random forest model detects lapses in treatment adherence. An assessment of variable importance suggests that the important patterns of driving in classification correspond to route decisions and patterns that may be associated with drowsy driving. The success of this approach suggests driving data may be valuable for evaluating new treatments, analyzing side effects of medications, and that the approach may benefit other drowsiness detection algorithms.

DRIVING PERFORMANCE AND DRIVER STATE IN OBSTRUCTIVE SLEEP APNEA: WHAT CHANGES WITH POSITIVE AIRWAY PRESSURE?

We evaluated naturalistic driving in 65 drivers with obstructive sleep apnea (OSA) before and after positive airway pressure (PAP) therapy and in 43 comparison drivers. Driving performance metrics included speed (mean, variability), and lateral, and longitudinal acceleration (g's). Driver state measures included sleepiness and attention to the driving task based on sampled trigger and baseline video clips. OSA drivers showed less variability in speed and lateral g's compared to control drivers before and after PAP treatment when vehicle speed was <45mph. There were no driving performance differences when vehicle speed exceeded 45 mph. OSA drivers remained less alert than comparison drivers before and after PAP. Average hours of nightly PAP-use predicted improved alertness and lower levels of sleepiness among OSA drivers. The findings suggest increased crash risk among OSA drivers may result from lower levels of attention to the driving task that result in performance lapses that may lead to crashes, rather than to clear and specific patterns of performance deficits in vehicle control.

Sleep remains disturbed in patients with obstructive sleep apnea treated with positive airway pressure: a three-month cohort study using continuous actigraphy.

OBJECTIVE: Some patients with obstructive sleep apnea (OSA) remain sleepy despite positive airway pressure (PAP) therapy. The mechanisms by which this occurs are unclear but could include persistently disturbed sleep. The goal of this study was to explore the relationships between subjective sleepiness and actigraphic measures of sleep during the first three months of PAP treatment. METHODS: We enrolled 80 patients with OSA and 50 comparison subjects prior to treatment and observed them through three months of PAP therapy. PAP adherence and presence of residual respiratory events were determined from PAP machine downloads. Epworth Sleepiness Scale (ESS), Functional Outcomes of Sleep Questionnaire (FOSQ), and actigraphic data were collected before and at monthly intervals after starting PAP. RESULTS: Patients with OSA were sleepier and showed a greater degree of sleep disruption by actigraphy at the baseline. After three months of PAP, only ESS and number of awakenings (AWAKE#) normalized, while wake after sleep onset and sleep efficiency remained worse in patients with OSA. FOSQ was improved in patients with OSA but never reached the same level as that of comparison subjects. ESS and FOSQ improved slowly over the study period. CONCLUSIONS: As a group, patients with OSA show actigraphic evidence of persistently disturbed sleep and sleepiness-related impairments in day-to-day function after three months of PAP therapy. Improvements in sleepiness evolve over months with more severely affected patients responding quicker. Persistent sleep disruption may partially explain residual sleepiness in some PAP-adherent OSA patients.

LINKING GPS DATA TO GIS DATABASES IN NATURALISTIC STUDIES: EXAMPLES FROM DRIVERS WITH OBSTRUCTIVE SLEEP APNEA.

In naturalistic studies, it is vital to give appropriate context when analyzing driving behaviors. Such contextualization can help address the hypotheses that explore a) how drivers perform within specific types of environment (e.g., road types, speed limits, etc.), and b) how often drivers are exposed to such specific environments. In order to perform this contextualization in an automated fashion, we are using Global Positioning System (GPS) data obtained at 1 Hz and merging this with Geographic Information Systems (GIS) databases maintained by the Iowa Department of Transportation (DOT). In this paper, we demonstrate our methods of doing this based on data from 43 drivers with obstructive sleep apnea (OSA). We also use maps from GIS software to illustrate how information can be displayed at the individual drive or day level, and we provide examples of some of the challenges that still need to be addressed.

EFFECTS OF FATIGUE ON REAL-WORLD DRIVING IN DISEASED AND CONTROL PARTICIPANTS.

This study evaluated real world driver errors and sleepiness in 66 drivers with Obstructive Sleep Apnea (OSA) and 34 matched controls (24 younger and 22 older). Driving errors and driver state were derived from analyses of video data from "black-box" event recorders. Sleep fragmentation data in OSA was derived from actigraphy for 15 days prior to beginning standard treatment (positive airway pressure, PAP) and 15 days after beginning PAP treatment. Prior to starting PAP, OSAs appeared sleepier than controls in general and particularly at intersections, while making safety errors following nights with high levels of fragmented sleep compared to matched controls. Adverse effects of sleep fragmentation during the pre-PAP phase were reduced post-PAP. Greater hours of PAP-use were associated with lower sleepiness and errors on the road. PAP-use was associated with a decrease in high sleep fragmented nights. Findings suggest reduction in acute sleepiness is unlikely to be the only mediating factor that explains the driving safety benefits of PAP in OSA.

FEEDBACK FROM NATURALISTIC DRIVING IMPROVES TREATMENT COMPLIANCE IN DRIVERS WITH OBSTRUCTIVE SLEEP APNEA.

As part of a study in drivers with obstructive sleep apnea (OSA), we conducted a randomized clinical trial to assess whether individualized feedback can increase compliance with continuous positive airway pressure (CPAP) therapy. After completing 3.5 months of naturalistic driving monitoring, OSA drivers were randomized either to receive an intervention, which was feedback regarding their own naturalistic driving record and CPAP compliance, or to receive no such intervention. In the week immediately after the intervention date, drivers receiving feedback (n=30) improved their CPAP usage by an average of 35.8 minutes per night (p=0.008; 95% CI=9.6, 62.0) to a mean level of 296 minutes. By contrast, CPAP usage in the non-feedback group (n=36) decreased an average of 27.5 minutes per night (p=0.022; 95% CI=4.0, 51.0) to a mean level of 236 minutes. The mean group-specific changes were higher (better) in the feedback group than in the non-feedback group during the first, second, and third weeks of follow-up (p<0.001, p=0.001, and p=0.027, respectively). By weeks 4 through 10, the effect of the feedback had lost its significance (p>0.25 in all cases). Our study suggests that CPAP compliance can be increased using individualized feedback, but that follow-up feedback sessions or reminders may be necessary for sustained improvement.

Variations on a theme: Topic modeling of naturalistic driving data.

This paper introduces Probabilistic Topic Modeling (PTM) as a promising approach to naturalistic driving data analyses. Naturalistic driving data present an unprecedented opportunity to understand driver behavior. Novel strategies are needed to achieve a more complete picture of these datasets than is provided by the local event-based analytic strategy that currently dominates the field. PTM is a text analysis method for uncovering word-based themes across documents. In this application, documents were represented by drives and words were created from speed and acceleration data using Symbolic Aggregate approximation (SAX). A twenty-topic Latent Dirichlet Allocation (LDA) topic model was developed using words from 10,705 documents (real-world drives) by 26 drivers. The resulting LDA model clustered the drives into meaningful topics. Topic membership probabilities were successfully used as features in subsequent analyses to differentiate between healthy drivers and those suffering from Obstructive Sleep Apnea.

USING FEEDBACK FROM NATURALISTIC DRIVING TO IMPROVE TREATMENT ADHERENCE IN DRIVERS WITH OBSTRUCTIVE SLEEP APNEA.

We are studying the effects of individualized feedback upon adherence with therapy (CPAP) in ongoing research aimed at improving driving safety in at-risk individuals with obstructive sleep apnea (OSA). The feedback includes specific samples of the individual's own naturalistic driving record, both alert and drowsy, and record of CPAP adherence. We report on this methodology, provide data examples of CPAP usage, and show preliminary data on the results in the first eleven drivers who received this intervention.

Highway Healthcare: How Naturalistic Driving Data Index Adherence to CPAP Therapy in Obstructive Sleep Apnea.

Drowsy driving is a major factor in many vehicle crashes around the world. Sleep disorders, such as obstructive sleep apnea (OSA), underpin many of these crashes. Continuous positive airway pressure (CPAP) therapy is an effective treatment for sleep apnea but it requires consistent use and is often rejected by OSA patients. Rejection of CPAP treatment creates a dangerous on-road environment for both OSA sufferers and the general public. Algorithms capable of detecting CPAP use and its effects on driving are integral to identifying and mitigating this danger. This work uses naturalistic kinematic driving data to develop an algorithm which can detect nightly CPAP abstinence and adequate CPAP use. Speed and lateral acceleration data were collected using a data recorder in participant's primary vehicle and CPAP data were collected by downloading adherence data from participant CPAP machines. The speed and acceleration data were reduced to a set of symbols using Symbolic Aggregate approximation (SAX) time-series analysis. The symbols were converted into a sequence frequency dataset using sliding windows of size 1 to 10 s with a 1 Hz sampling rate. A Random Forest classifier was trained on the data to create a classification algorithm. On a held aside testing set, the Random Forest algorithm correctly identified 71% of the instances and had an area under the receiver operating characteristic curve of 0.76. The variable importance of the algorithm suggested that kinematic patterns associated with common drowsy driver crash types were key features in the algorithm's prediction performance.

The Language of Driving: Advantages and Applications of Symbolic Data Reduction for Analysis of Naturalistic Driving Data.

Recent advances in onboard vehicle data recording devices have created an abundance of naturalistic driving data. The amount of data exceeds the resources available for analysis; this situation forces researchers to focus on analyses of critical events and to use simple heuristics to identify those events. Critical event analysis eliminates the context that can be critical in understanding driver behavior and can reduce the generalizability of the analysis. This work introduced a method of naturalistic driving data analysis that would allow researchers to examine entire data sets by reducing the sets by more than 90%. The method utilized a symbolic data reduction algorithm, symbolic aggregate approximation (SAX), which reduced time series data to a string of letters. SAX can be applied to any continuous measurement, and SAX output can be reintegrated into a data set to preserve categorical information. This work explored the application of SAX to speed and acceleration data from a naturalistic driving data set and demonstrated SAX's integration with other methods that could begin to tame the complexity of naturalistic data.

CAN INTERMITTENT VIDEO SAMPLING CAPTURE INDIVIDUAL DIFFERENCES IN NATURALISTIC DRIVING?

We examined the utility and validity of intermittent video samples from black box devices for capturing individual difference variability in real-world driving performance in an ongoing study of obstructive sleep apnea (OSA) and community controls. Three types of video clips were coded for several dimensions of interest to driving research including safety, exposure, and driver state. The preliminary findings indicated that clip types successfully captured variability along targeted dimensions such as highway vs. city driving, driver state such as distraction and sleepiness, and safety. Sleepiness metrics were meaningfully associated with adherence to PAP (positive airway pressure) therapy. OSA patients who were PAP adherent showed less sleepiness and less non-driving related gaze movements than nonadherent patients. Simple differences in sleepiness did not readily translate to improvements in driver safety, consistent with epidemiologic evidence to date.

EFFECTS OF ENVIRONMENTAL FACTORS ON NATURALISTIC DRIVING IN OBSTRUCTIVE SLEEP APNEA.

Reduced visibility and other environmental factors can impair driver ability to respond to roadway hazards. We examined the effects of reduced visibility on naturalistic driving in 66 drivers, including 45 at-risk drivers with obstructive sleep apnea (OSA) and 21 controls. We analyzed three months of electronic data using "black box" recorder technology and assessed the extent to which driver speed, longitudinal acceleration, and lateral acceleration metrics depend on ambient visibility from web-based environmental data archives. We calculated summary driving metrics within 10-second intervals, and reduced these to within-subject means and tested for associations of interest. OSA drivers did not differ from controls with respect to electronic measures or visibility conditions in which they drove. On average, drivers drove slower when visibility was reduced. After controlling for speed, variations in lateral and longitudinal acceleration were positively associated with high-visibility conditions. These findings suggest that drivers exert greater vehicular control when visibility is limited, and that this association is not just due to slower speeds. Weaker relationships between visibility and driving measures in OSA suggest reduced adaptive strategies. Our methods provide a framework for analyzing the effects of other environmental factors on driving, and we provide an additional example using wind speed.

Visual vigilance in drivers with obstructive sleep apnea.

OBJECTIVE: To determine the effects of obstructive sleep apnea (OSA) on visual vigilance during simulated automobile driving. METHODS: Twenty-five drivers with OSA and 41 comparison drivers participated in an hour-long drive in a high-fidelity driving simulator. Drivers responded to light targets flashed at seven locations across the forward horizon. Dependent measures were percent correct [hit rate (HR)] and reaction time (RT). Self-assessment of sleepiness used the Stanford Sleepiness Scale (SSS) before and after the drive and the Epworth Sleepiness Scale (ESS). RESULTS: OSA drivers showed reduced vigilance based on lower HR than comparison drivers, especially for peripheral targets (80.7+/-14.8% vs. 86.7+/-8.8%, P=.03). OSA drivers were sleepier at the end of the drive than comparison drivers (SSS=4.2+/-1.2 vs. 3.6+/-1.2, P=.03), and increased sleepiness correlated with decreased HR only in those with OSA (r=-0.49, P=.01). Lower HR and higher post-drive SSS predicted greater numbers of driving errors in all subjects. Yet, ESS, predrive SSS, and most objective measures of disease severity failed to predict driving and vigilance performance in OSA. CONCLUSIONS: Reduced vigilance for peripheral visual targets indicates that OSA drivers have restriction of their effective field of view, which may partly explain their increased crash risk. This fatigue-related decline in attention is predicted by increased subjective sleepiness during driving. These findings may suggest a means of identifying and counseling high-risk drivers and aid in the development of in-vehicle alerting and warning devices.

USEFUL FIELD OF VIEW IMPAIRMENTS IN DRIVERS WITH OBSTRUCTIVE SLEEP APNEA.

As a group, drivers with obstructive sleep apnea (OSA) have an increased risk for motor vehicle crashes, but determining individual crash risk is difficult. We tested the hypothesis that drivers with OSA have impaired visual attention, as indexed by reduced useful field of view (UFOV), a predictor of high-risk driving. Forty-one drivers with untreated OSA and 50 comparison drivers were assessed by UFOV. OSA drivers performed significantly worse than controls on all UFOV subtests and had reduced UFOV as indicated by a higher mean total UFOV score (p = 0.0017). However, only 4 OSA and 2 control drivers had values indicative of high crash risk (UFOV reduction >23%). Drivers with OSA have reduced UFOV compared to drivers without neurological or sleep disorders. However, as UFOV identifies few high-risk drivers, its role in assessing crash risk in an unselected population of drivers with OSA appears to be limited.

Driver Performance in the Moments Surrounding a Microsleep.

This study examined if individuals who are at increased risk for drowsy-driving because of obstructive sleep apnea syndrome (OSAS), have impairments in driving performance in the moments during microsleep episodes as opposed to during periods of wakefulness. Twenty-four licensed drivers diagnosed with OSAS based on standard clinical and polysomnographic criteria, participated in an hour-long drive in a high-fidelity driving simulator with synchronous electroencephalographic (EEG) recordings for identification of microsleeps. The drivers showed significant deterioration in vehicle control during the microsleep episodes compared to driving performance in the absence of microsleeps on equivalent segments of roadway. The degree of performance decrement correlated with microsleep duration, particularly on curved roads. Results indicate that driving performance deteriorates during microsleep episodes. Detecting microsleeps in real-time and identifying how these episodes of transition between wakefulness and sleep impair driver performance is relevant to the design and implementation of countermeasures such as drowsy driver detection and alerting systems that use EEG technology.

Driving impairment in patients with obstructive sleep apnea syndrome.

Drivers with obstructive sleep apnea syndrome (OSAS) have an increased risk of motor vehicle crashes. Unfortunately, neither clinical nor polysomnographic features allow clinicians to reliably identify high-risk drivers. One potential means of identifying these drivers is with the use of driving simulators. Several investigators have shown that OSAS patients perform worse than healthy control drivers and results from our studies have demonstrated declines in driving performance during EEG-defined "microsleeps." The use of simulators, and in-vehicle detection and alerting devices may mitigate some of the suffering caused by these crashes.