Sleep, Activity, Fatigue, and Task Effectiveness (Safte ...
January 31, 2020 The Sleep, Activity, Fatigue, and Task Effectiveness (SAFTE) Model and Fatigue Avoidance Scheduling Tool (FAST) Non-Prescriptive Tools for Effective Fatigue Management DOT Human Factors Coordinating Committee June 23, 2004 1 Steven R. Hursh, Ph.D. Program Manager, Biomedical Modeling and Analysis Science Applications International Corporation, 443-402-2701 Professor, Johns Hopkins University School of Medicine [email protected] January 31, 2020 Outline 2 FRA role and objectives in fatigue management initiatives
The SAFTE Model of fatigue The Fatigue Avoidance Scheduling Tool (FAST) Update on calibration for accidents and incidents On-the-Job Performance Study Fatigue Assessment Dashboard The Incident Fatigue Assessment Protocol (IFAP) FAST Schedule Design Wizards and XIMES Interoperability January 31, 2020 The FRA Role Partnership between Human Factors R & D Program (Volpe) and Office of Safety. Supported and coordinated with DOT human factors coordinating committee Facilitate development of fatigue management tools for the railroad environment Make available funds to advance this tool Facilitate and finance necessary studies to calibrate the tool for the railroad environment Encourage system-wide assessments combined with site specific solutions Maintain a hands-off approach to applications of the tool Ensure the confidentiality of information 3
Disseminate information about fatigue management solutions January 31, 2020 FRA Office of Safety Objectives 4 Advance a non-prescriptive approach to fatigue management. Help develop a fatigue and performance model as a nonprescriptive tool to be used at the discretion of management and labor. Help develop methods to collect and analyze information (data) that might be used to assess fatigue contributions to accidents and incidents. Integrate tools so they can be mutually supporting: FAST, Actigraphs, XIMES RAS, PERCLOS Continue to promote cooperative approaches to fatigue
mitigation solutions. January 31, 2020 Comparative Status of the SAFTE Model The DOD and DOT sponsored a comparison of six fatigue models from around the world in Seattle. All models attempted to predict the results from four standard scenarios. While all models can be improved, the SAFTE model had less error than any model tested and was combined with a convenient and logical user interface, the fatigue avoidance scheduling tool - FAST. 5 January 31, 2020 Model Comparison Independent Evaluation Scenario 2 Restricted Sleep (data from Dinges Lab U. Penn) Model Author
RRMSE (% Error) Rank Low score is better Spencer/Belyavin Moore-Ede Dawson/Fletcher Jewett/Kronauer Hursh (SAFTE 1) Hursh (SAFTE 2)* Folkard/Akerstedt 6 90.76 99.99 74.57 92.27 81.21 70.91 93.87 4 7 2 5 3 1 6 * Just prior to the Fatigue and Performance Workshop, SAFTE Model was revised and optimized using data from Scenario 5, which is similar to scenario 2. The
.RRMSE value [was] 70.91%...This constitutes a substantial improvement with respect to the earlier predictions from this and all the other models. (page 27) January 31, 2020 General Model Deficiencies Gap 7 Estimate of prediction error population variance. Representation of individual differences. Representation of countermeasures. Translation into task or job performance and risk assessment. Remedy Now included in FAST
Proposal to Army Ongoing with AF and Army Lapse Index added, other measures are part of DOT effort. January 31, 2020 Fatigue Avoidance Scheduling Tool (FAST) FAST is a fatigue assessment tool based on the SAFTE model Developed for the US Air Force and the US Army. NTI and SAIC in a Phase III SBIR program. DOT/FRA sponsored work has lead to a enhancement for transportation applications. Auto Sleep algorithm Schedule Grid data entry tool Wizards and Dashboard - funded
8 DOT field calibration underway. January 31, 2020 SAFTE Model and FAST Predict Performance Effectiveness Effectiveness is a measure of speed of making correct responses. Effectiveness estimates are at a 1 min. resolution. Effectiveness is highly correlated with other measures of fatigue: Lapse likelihood Reaction time Average cognitive throughput Driving simulator performance 9 Tool predicts both average person and population variance estimate. January 31, 2020 FAST Graphical Screen Options
Effectiveness Adjustable Criterion Line Lower Percentile (e.g. 20%) Sleep Periods in Blue 10 Work Periods in Red January 31, 2020 Lapses Increase with Decreasing Effectiveness from FAST (revised) Sleep Dose Response Study Experimental & Recovery Days - WRAIR Data Lapse Likelihood (Times Baseline) 12 3 hrs sleep/day 10 5 hrs sleep/day Expon. (3 hrs sleep/day) 8
6 4 y Da 2 sin a e cr n I f so ep e l gS bt e D y = 238.19e-0.0558x R2 = 0.9625 Baseline=1 (8 hrs sleep/day) 0
100 95 90 85 80 75 70 65 60 55 50 Effectiveness (FAST Prediction) 11 High Low January 31, 2020 Lapse Index
12 January 31, 2020 Commercial Applications of FAST FAST is optimized for the average person, including truck drivers, studied under a range of schedules. It has a variety of applications: Problem Definition and Assessment Work Schedule Design and Evaluation Generic Schedules (shift-schedules) Individualized Schedules (work assignments) 13 Safety and Accident Investigation Tool Training and Awareness Voluntary Self-management DOT Experience: Schedule evaluation and accident fatigue assessment underway January 31, 2020 Complements Other Existing Tools FAST provides an objective fatigue and
performance estimate that can be used in conjunction with other available tools. Other Fatigue Risk Management Tools: 14 Fatigue monitoring devices actigraphs and PERCLOS Return on Investment Models Operational/Terminal Management Models Staffing Tools Team decision-making tools Other Fatigue Estimation Tools January 31, 2020 FRA Initiatives Calibrate FAST for railroad environment Analysis of accidents and incidents with two railroad partners, pilot study Analysis of accidents and incidents with two railroad partners, large sample study Analysis of locomotive engineer performance Enhancement of tool for accident and incident investigations Fatigue indicators dashboard based on NTSB workshop Input wizards for irregular and shift schedules
XIMES interoperability Incident fatigue analysis protocol 15 January 31, 2020 Fatigue Risk Management Tools FAST Fatigue Avoidance Scheduling Tool Prospective forecasting of fatigue risk under proposed work/rest schedules. Retrospective assessment of fatigue leading up to an event. Uses the SAFTE model of fatigue developed by the DOD. IFAP Incident Fatigue Assessment Protocol Questionnaire and schedule assessment software to aid event analysis. Integrates data into FAST for fatigue assessment. 16 January 31, 2020 Validation Schematic Intrinsic Process Validation:
Primary Predictive Validation: Components of the Model Laboratory Cognitive Performance Predictions Confirming Validation: Task Performance Predictions 17 Safety & Accident Predictions Subjective Fatigue Predictions DOD & DOT Projects will contribute to these three areas of validation January 31, 2020 SAFTE MODEL Predictions and Data Total Sleep Deprivation (WRAIR 72 hr Study) Three Days of Sleep Deprivation 120 Percent of Baseline
80 January 31, 2020 Decline of Performance with Total Sleep Deprivation Sleep & Performance Model vs Angus & Heslegrave (1985) Mean of Normalized Performance Measures 120 Serial RT Decode Encode Vigilance Logical Mean SAFTE Prediction Effectiveness (Percent) 100 80 75% 60 50% 40 20
Parameters: Acrophase: 1900 hrs Awakening at 0700 hrs 0 19 0 19 10 20 30 40 Hours of Sleep Deprivation 43 50 60 January 31, 2020 SAFTE Model (Revised) Walter Reed Army Institute of Research Sleep Restriction Study
PVT Speed Chronic Restriction Adaptation 110 Mean Speed (as a % of Baseline) 95 9 Hr 80 7 Hr 5 Hr 3 Hr 65 SAFTE/FAST R2 = 0.94 50 0 20 T1 T2 B
E1 E2 E3 E4 Day E5 E6 E7 R1 R2 R3 Sleep duration is total observed according to EEG measurement. Model parameters are constant across conditions of the experiment. January 31, 2020 Congruence of SAFTE Model to Sleep Deprivation Data 21 Data Source Default Version 2 R
WRAIR 72 hr sleep deprivation data 0.89 Angus-Heslegrave 54 hr sleep deprivation data 0.98 Mean 0.94 WRAIR Restricted Sleep Study (Revised Model-All Groups) 0.94 January 31, 2020 Missed Horns Analysis ITRI Locomotive Simulator Missed Horns for Effectiveness Quintiles (Excess of misses summed over group) Day 4 Day 5 Day 6
0% -20% -26.6% 22 Chart 3 -40% Top Quintile 2nd Quintile Middle Quintile 4th Quintile Bottom Quintile January 31, 2020 SAFTE/FAST Analysis of Incident Data Pilot Study Pilot study of 50 events from two railroads. Report here data for only one railroad. For each event, there were usually two operators: engineer and conductor. Results are tentative because of small sample size; large study of 500 events is
pending railroad approval. Results illustrates the analysis approach. 23 January 31, 2020 Important Data for FAST 24 Work starts and stops Call times Commute times Particular sleep habits Actual sleep times Quality of sleep environment Schedule predictability
Considered Considered Considered Not considered Not considered Not considered Not considered January 31, 2020 Important Data Not Considered Schedule delays or misinformation Training and experience Medical conditions and sleep disorders Medications and/or drug use Observations of operator performance and 25 appearance Concurrent stress, family issues, or work demands Crew resource management (communication problems) January 31, 2020 Fatigue Analysis Overview
FAST is a tool for evaluating and combining fatigue factors in a schedule and providing an objective and impartial performance prediction. The model is designed to predict fatigue, not incidents and accidents. Many human factors accidents occur without fatigue. Some non-human factors accidents may have a fatigue component that was not recorded. 26 However, our hypothesis was that the model should indicate an increased probability of events at the extremes of predicted fatigue (low effectiveness). January 31, 2020 RR-X Incident Effectiveness (Least Effective Crewmember) Incident Crew Effectiveness (Worst Member) versus Overall Crew Effectiveness Time 30.00 Non-Human Factors Human Factors Time 25.00 Percent
January 31, 2020 RR-X Incident Effectiveness Probabilities (Least Effective Crewmember) Incident Crew Effectiveness (Worst Crew Members) - Probability by Effectiveness Category 4.5 Non-Human Factors 4 Human Factors 3.5 Probability 3 2.5 2 1.5 Random 1 0.5 0 100-95
28 95-90 90-85 85-80 80-75 75-70 Effectiveness Categories 70-65 65-60 60-55 55-50 January 31, 2020 FAST Can Discriminate Human Factors vs Nonhuman Factors Events Receiver Operator Curve FAST with effectiveness criterion at 67% Percent correct is 68% A = .77 Comparisons:
Materials Testing (detection of cracks in airplane wings) Ultrasound - 0.68 Eddy Current - 0.93 Medical Imaging (detecting tumors) Ultrasound, adrenal gland - 0.83 Weather Forecasting Extreme cold - 0.89 Rain - 0.82 Fog - 0.76 Storms - 0.74 Temperature Intervals - 0.71 Polygraph Lie Detection Real crimes - 0.70 to 0.92 Promising but more data needed. 29 January 31, 2020 On-the-Job Performance Study Purpose and Process To determine if the FAST software can make valid predictions of operator performance changes that result from fatigue and circadian variations. Analysis Steps:
Collaborate with two or more Class 1 railroads to collect event recorder data. Correlate performance measures from engineers along with work schedule information and actigraph measures of sleep. Determine if the FAST predictions are consistent with general variations in performance resulting from sleep patterns and time-of-day effects. Determine if interventions reduce performance changes. Status: Looking for collaborators could use historical data, if available. FAA-CAMI (Tom Nesthus) is planning a study of fatigue in coordination with the International Flight Inspection Office to test the predictions of FAST. 30 Actigraphs, Logbooks, ARES Performance Battery, and perhaps oral temperature.
The IFIO missions are typically 3-wks out (Europe, Far East, or wherever). January 31, 2020 Supported by Transportation Research Board Research Recommendations (1) Measurements of effectiveness of safety interventions of any sort. Relationship between fatigue and operator performance. (7) Sensitivity of fatigue models to performance-related measures? Relationship between fatigue and operator performance. (8) Incidence of fatigue in train crews based on work schedule data. 31 January 31, 2020 NTSB Workshop on Fatigue Analysis in Transportation Accidents 32 Rosekind and Dinges Expert Consultants
Recommended in-depth assessment of operator schedules, sleep need and habits, medical history, and medications. Recommended calculation of fatigue indicators. FRA now sponsoring work to add fatigue indicators to FAST, along with performance prediction. Feature will provide immediate fatigue explanation of low predicted performance to aid the accident investigator. January 31, 2020 Other Initiatives Fatigue Indicators Dashboard Incident Fatigue Assessment Protocol - Update Schedule Design Wizard FAST XIMES RAS interoperability (Details follow) 33 Performance Indicators Effectiveness (PVT speed) (% normal) Mean Cognitive (% normal)
Lapse Probability (times normal) Reaction Time (% normal) <90%=Y <75%=R <90%=Y <75%=R >2=Y >4=R >110%=Y >133%=R Fatigue Factors Acute Sleep Debt (Hrs in last 24 hrs) 8 hrs=R
Cumulative Sleep Debt (Total Hours prior to current day) 8 hrs=R Continuous Hours Awake Time of Day Circadian Disruption (Hrs out of phase) 17 hrs=R 0000-0600, 1500-1700 =R 3 hrs=R All of the indicators can be computed from variables available within FAST. All the performance indicators are computed as transforms from Effectiveness shown on the graph. Fatigue factors within a dangerous range are indicated by a red flag based on criteria established at NTSB workshop. The performance indicator will change color from green to yellow to red. When performance is less than green, the flagged fatigue factors
will provide an explanation. All these values would be continuously updated as the cursor is moved along the time line of the schedule, giving an instantaneous assessment at any moment or at some critical event. January 31, 2020 Schedule Name: Date:Time Performance Indicators 35 Fatigue Factors Effectiveness (PVT speed) (% normal) Mean Cognitive (% normal) Lapse Probability (times normal) Reaction Time (% normal) Acute Sleep Debt (Hrs in last
24 hrs) Cumulative Sleep Debt (Total Hours prior to current day) Continuous Hours Awake Time of Day Circadian Disruption (Hrs out of phase) 67% 71% 6 140% 7 12 20 0630
.6 When activated, the dashboard will appear as a window within the schedule window and update with the movement of the mouse. A left mouse button depression within the schedule area will clear the table; release the mouse button and it will rewrite with new values appropriate for the new cursor position. The dashboard can be dragged to any convenient place within the graphical window. A right mouse click will allow the user to print or copy the dashboard to the clipboard for paste into another application. January 31, 2020 Incident Fatigue Assessment Protocol Computer-based investigation questionnaire of operators and witnesses Computerized version of Technical Bulletin (proof of principal) Keppen Associates initiative to test and improve questionnaire Computer-based schedule recording and analysis Access database creation Automatic porting to FAST Potential to be ported to PDA 36 January 31, 2020 Incident Fatigue Assessment Protocol (IFAP) Schedule Entry Screen
37 January 31, 2020 Schedule Design Wizards Funded (Volpe Project) Shift-schedule wizard Irregular schedule wizard Dialog data entry with options to branch to tabular or grid entry (TurboTax metaphor) Descriptive results output Graphical Tabular Fatigue factors Narrative 38 January 31, 2020 FAST and XIMES RAS Interoperability Funded (Volpe Project) XIMES RAS is an Austrian developed shift schedule design tool. Assists in design and descriptive analysis of shift schedules Contains rule of thumb schedule evaluation Does not assess fatigue potential
39 Initiative will create standard schedule file format that can be shared by these two tools and an other schedule analysis software. Schedules created with XIMES RAS can be imported into FAST for sleep assessment and fatigue analysis. January 31, 2020 Recommendations Continue to advance our knowledge and understanding of software modeling of fatigue as one tool to address fatigue. Continue to seek cooperation of railroad partners in collecting essential data. Continue to seek performance indices to quantify the role of fatigue in railroad operations. Continue to coordinate and strengthen transfer of our knowledge to other modes.
Welcome partnerships with other modes to fill data gaps. 40 Continue to strengthen our working relationships with AARs Work Rest Task Force and NARAP. January 31, 2020 Non-Prescriptive Tools for Effective Fatigue Management End of Presentation Steven R. Hursh, Ph.D. Science Applications International Corporation, 443-402-2701 Professor, Johns Hopkins University School of Medicine [email protected] 41 January 31, 2020 Modeling Approaches Fatigue Audit InterDyne (FAID) Model: Drew Dawson
42 (University of South Australia) - fatigue model currently being applied as a fatigue management tool in Australia and by the Union Pacific. Sleepwake Predictor: Torbjorn kerstedt (Karolinska Institute) & Prof. Simon Folkard (University of Wales) Interactive Neurobehavioral Model: Megan Jewett, (Harvard University) System for Aircrew Fatigue Evaluation (SAFE): Spencer and Belyavin (QinetiQ, Inc, UK) Circadian Alertness Simulator (CAS): Martin Moore-Ede (Circadian Technologies, Inc., USA) Sleep, Activity, Fatigue and Task Effectiveness (SAFTE) Model: Steven Hursh (SAIC and Johns Hopkins University, USA).
By using parallelism, authors or speakers implicitly invite their readers or audiences to compare and contrast parallel elements. "It was the best of times, it was the worst of times." - Dickens "I have a dream…I have a dream…I have...
Pre-competition anxiety Exercise and stress reduction PE and self-esteem Personality and sports success Exercise adherence Aggression in sport Overtraining Substance abuse in athletes Team cohesion K-12 Physical Education Teach physical education in public & private schools Additional endorsements include special...
Arial Garamond Times New Roman Wingdings Symbol Borda Microsoft Equation 3.0 Centro Universitário da Fundação Educacional de Barretos Propriedades Mecânicas dos Materiais Slide 3 Slide 4 Slide 5 Slide 6 Slide 7 Slide 8 Slide 9 Slide 10 Slide 11...
The hypothesis that if I pour 100ml of coffee on four pea plants and pour 100ml of water on another four pea plants, then the plants with coffee will grow taller because the caffeine will stimulate the plants was supported....
Kneehigh Theatre Company. It's fun, it's physical, its musical and it breaks the fourth wall. That's not a quote, its just me summing them up. features- of their practice. Audience interaction. ... Puppetry. Chorus/ ensemble. Cross gender casting.
Preparation of the Engineering Body of Knowledge. NSPE Licensure and Qualifications for Practice Committee. 27 Members. 8 Disciplines Represented. Licensure Oriented. ... Higher value-added, leading-edge services and products from the U.S. needed.
A few pharmacodynamic interactions have also been described. Pharmacodynamicinteractions may be synergetic(e.g. interaction between the anticoagulant warfarinwith antiplatelet herbs), or antagonistic (e.g. kava possesses dopaminergic antagonistic propertiesand hence might reduce the pharmacological activity ofthe anti-parkinson drug levodopa) .
Ready to download the document? Go ahead and hit continue!