| Literature DB >> 33922735 |
Xuan Di1, Rongye Shi1, Carolyn DiGuiseppi2, David W Eby3,4, Linda L Hill5, Thelma J Mielenz6,7, Lisa J Molnar3,4, David Strogatz8, Howard F Andrews9, Terry E Goldberg9,10, Barbara H Lang10, Minjae Kim10, Guohua Li6,10.
Abstract
Emerging evidence suggests that atypical changes in driving behaviors may be early signals of mild cognitive impairment (MCI) and dementia. This study aims to assess the utility of naturalistic driving data and machine learning techniques in predicting incident MCI and dementia in older adults. Monthly driving data captured by in-vehicle recording devices for up to 45 months from 2977 participants of the Longitudinal Research on Aging Drivers study were processed to generate 29 variables measuring driving behaviors, space and performance. Incident MCI and dementia cases (n = 64) were ascertained from medical record reviews and annual interviews. Random forests were used to classify the participant MCI/dementia status during the follow-up. The F1 score of random forests in discriminating MCI/dementia status was 29% based on demographic characteristics (age, sex, race/ethnicity and education) only, 66% based on driving variables only, and 88% based on demographic characteristics and driving variables. Feature importance analysis revealed that age was most predictive of MCI and dementia, followed by the percentage of trips traveled within 15 miles of home, race/ethnicity, minutes per trip chain (i.e., length of trips starting and ending at home), minutes per trip, and number of hard braking events with deceleration rates ≥ 0.35 g. If validated, the algorithms developed in this study could provide a novel tool for early detection and management of MCI and dementia in older drivers.Entities:
Keywords: Alzheimer’s disease and related dementias; aging; artificial intelligence; dementia; driving patterns; machine learning; mild cognitive impairment; naturalistic driving study; random forests; screening
Year: 2021 PMID: 33922735 PMCID: PMC8167558 DOI: 10.3390/geriatrics6020045
Source DB: PubMed Journal: Geriatrics (Basel) ISSN: 2308-3417
Variable definitions and statistics.
| Variable | Name | Definition | Statistics | ||||
|---|---|---|---|---|---|---|---|
| Min | Max | Mean | SD | ||||
| Diagnosis_labels | MCI/Dementia | One’s disease level in a month (0-Healthy; 1-Mild Cognitive Impairment (MCI)/Dementia/Alzheimer’s | 0 | 1 | - | - | |
| Demographic characteristics | |||||||
| Age in years | Age | Age at enrollment | 65 | 79 | 71.1 | 4.1 | |
| Sex | Sex | Male; Female | NA | NA | NA | NA | |
| Race/Ethnicity | Race | Alaska Native, Native Hawaiian, Pacific Islander; American Indian, Asian; Black (non-Hispanic); White (non-Hispanic); Hispanic; Other | NA | NA | NA | NA | |
| Education | Education | Associate degree; Bachelor degree; Master, professional, or doctoral degree; Some college but no degree; Vocational, technical, business, or trade school (beyond high school level); Other | NA | NA | NA | NA | |
| Driving variables | |||||||
| 1 | Miles_n | Miles | Total number of miles driven in month | 0 | 15,783 | 762.2 | 587.8 |
| 2 | Trips | Trips | Total number of trips in month | 1 | 2341 | 115.8 | 64.6 |
| 3 | TripsLt15Miles | No. trips < 15 miles of home | Number of trips traveled in month within 15 miles of home | 0 | 1953 | 95.5 | 58.4 |
| 4 | PercentDistLt15Miles_n | % trip < 15 miles of home | Percent of trips traveled in month within 15 miles of home | 0.0 | 100.0 | 64.9 | 28.9 |
| 5 | TripsLt25Miles | No. trips < 25 miles of home | Number of trips traveled in month within 25 miles of home | 0 | 1953 | 101.7 | 59.7 |
| 6 | PercentDistLt25Miles_n | % trip < 25 miles of home | Percent of trips traveled in month within 25 miles of home | 0.0 | 100.0 | 76.5 | 26.4 |
| 7 | MilesPerTrip_n | Miles per trip | Total number of miles driven in month divided by total number of trips in month | 0.0 | 74.5 | 6.7 | 4.1 |
| 8 | MinutesPerTrip_n | Minutes per trip | Total driving minutes in month divided by total number of trips in month | 0.1 | 137.7 | 14.9 | 5.9 |
| 9 | TripMinutes_n | Total trip minutes | Total minutes of driving in month | 0.1 | 16,645.0 | 1633.4 | 1083.6 |
| 10 | TripsInDay | No. trips during day | Number of trips in month not classified as nighttime | 0 | 1279 | 107.2 | 57.4 |
| 11 | PercentTripsInDay_n | % trips during day | Percent of trips in month not classified as nighttime | 0.0 | 100.0 | 93.1 | 8.0 |
| 12 | TripsAMPeak | No. trips in AM peak | Number of trips in month during 7–9 AM on weekdays | 0 | 167 | 8.6 | 9.4 |
| 13 | PercentTripsAMPeak_n | % trips in AM peak | Percent of trips in month during 7–9 AM on weekdays | 0.0 | 100.0 | 7.3 | 6.9 |
| 14 | TripsAtNight | No. trips at night | Number of trips during which at least 80% of a trip was during nightime in month (Nightime was defined as civil twilight or a solar angle greater than 96 deg) | 0 | 1143 | 8.7 | 16.6 |
| 15 | PercentTripsAtNight_n | % trips at night | Percent of trips during which at least 80% of a trip was during nightime in month (Nightime was defined as civil twilight or a solar angle greater than 96 deg) | 0.0 | 100.0 | 6.9 | 8.0 |
| 16 | TripsPMPeak | No. trips in PM peak | Number of trips in month during 4–6PM on weekdays | 0 | 150 | 10.9 | 9.5 |
| 17 | PercentTripsPMPeak_n | % trips in PM peak | Percent of trips in month during 4–6PM on weekdays | 0.0 | 100.0 | 9.3 | 6.7 |
| 18 | LeftTurnCount | No. left turns | Number of left turns made in month | 0 | 2592 | 261.6 | 159.9 |
| 19 | RightTurnCount | No.right turns | Number of right turns made in month | 0 | 2751 | 242.6 | 150.1 |
| 20 | RightToLeftTurnRatio_n | Right to left turn ratio | Ratio of all right-hand to left-hand turning events for a driver in a month | 0.0 | 7.0 | 0.9 | 0.2 |
| 21 | TripsVgt60 | No. trips on high speed roads | Number of trips in month where 20% of distance travelled was at a speed of 60 MPH or greater | 0 | 226 | 13.9 | 15.0 |
| 22 | PercentTripsVgt60_n | % trip on high speed roads | Percent of trips in month where 20% of distance travelled was at a speed of 60 MPH or greater | 0.0 | 12.6 | 100.0 | 12.3 |
| 23 | SpeedGt80mphCount | No. speeding events | Number speeding events in month (speed > 80 MPH sustained for at least 8 s) | 0 | 3300 | 7.3 | 31.3 |
| 24 | DecelCntLtN3pt5Mps2 | No. hard braking events with deceleration rates ≥ 0.35 g | Number of events with a deceleration rate ≥ 0.35 g in a month | 0 | 1112 | 3.8 | 8.2 |
| 25 | DecelCntLtN4pt0Mps2 | No. hard braking events with deceleration rates ≥ 0.40 g | Number of events with a deceleration rate ≥ 0.4 g in a month | 0 | 734 | 0.9 | 4.2 |
| 26 | DecelCntLtN7pt5Mps2 | No. hard braking events ≥ 0.75 g | Number of events with a deceleration rate ≥ 0.75 g in a month | 0 | 24 | 0.005 | 0.2 |
| 27 | TripChains | Trip chains | Number of trip chains in month (Note: chain is a series of trips starting and ending at home) | 0 | 180 | 8.2 | 7.8 |
| 28 | MilesPerChain_n | Miles per chain | Total miles of chains in month divided by total number of trip chains in month | 0.00 | 4273.3 | 100.7 | 121.1 |
| 29 | MinutesPerChain_n | Minutes per chain | Total driving minutes for chains divided by total number of trip chains in month | 0.0 | 6606.3 | 222.0 | 219.8 |
NA, not applicable.
Performance of random forests models with different covariates in predicting incident mild cognitive impairment or dementia.
| Model | Covariates | Accuracy | Precision or PPV | Recall or Sensitivity | Specificity | NPV | F1 Score | AUC | Out-of-Bag Error Rate | Confusion Matrix | |||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Predicted | Observed | ||||||||||||
| % | SD | 0 | 1 | ||||||||||
| 1 | Age only | 0.46 | 1.00 | 0.06 | 1.00 | 0.45 | 0.11 | 0.56 | 6.00 | 0.01 | 0 | 77 | 96 |
| 1 | 0 | 6 | |||||||||||
| 2 | Age, sex, race/ethnicity, and education | 0.53 | 1.00 | 0.17 | 1.00 | 0.48 | 0.29 | 0.64 | 0 | 77 | 85 | ||
| 4.27 | 0.01 | 1 | 0 | 17 | |||||||||
| 3 | Driving variables only | 0.66 | 0.79 | 0.56 | 0.81 | 0.58 | 0.66 | 0.76 | 2.60 | 0.75 | 0 | 62 | 45 |
| 1 | 12 | 57 | |||||||||||
| 4 | Age and driving variables | 0.80 | 0.89 | 0.74 | 0.88 | 0.72 | 0.81 | 0.91 | 2.14 | 0.50 | 0 | 68 | 27 |
| 1 | 9 | 75 | |||||||||||
| 5 | Age, sex, race/ethnicity, education and driving variables | 0.86 | 0.86 | 0.90 | 0.81 | 0.86 | 0.88 | 0.90 | 2.07 | 0.57 | 0 | 62 | 10 |
| 1 | 15 | 92 | |||||||||||
PPV, positive predictive value; NPV, negative predictive value; AUC, area under the receiver operating characteristic curve.
Figure 1Feature importance ranking.