Model Driver Screening and Evaluation Program
Volume II: Maryland Pilot Older Driver Study

 

Chapter 4: 
Pilot Study Data Analysis and Results
(Continued)

Results of Crash Analyses

Motor Free Visual Perception Visual Closure Subtest (MVPT/VC). Figure 31 contains the results for the MVPT/VC. The top plot relates functional performance to crash involvement, using "all crashes" as the safety outcome measure. The middle plot relates functional status to the more restrictive outcome measure of "at-fault and unknown fault" crashes, and the bottom plot shows the distributions of License Renewal sample drivers with and without "at-fault" crashes at each level of functional ability measured by this test. In all cases, declining functional ability is indicated by an increasing number of incorrect responses, moving to the right along the x-axis.

Figure 31. 
MVPT/VC Performance Distributions and Odds Ratios for Analyses
Including All Crashes, At-Fault and Unknown-Fault Crashes,
and At-Fault Crashes Only

Graphs of Figure 31.
MVPT/VC Performance Distributions and Odds Ratios for Analyses

Inspection of this figure reveals stronger relationships moving from the top to the bottom data plot; this is associated with a progressive increase in the peak OR value from 2.21 for "all crashes" to 4.96 for "at-fault" crashes only. The peak OR (4.96), associated with a cutpoint of 5 incorrect responses, is statistically significant (c2 = 26.48, df=1, p<.001). It is also interesting to note that, in all three plots the proportion of drivers who are crash-involved begins to exceed the proportion who are crash-free at the same level of functional performance--four incorrect responses.

Finally, it may be observed that the distributions of crash-involved drivers appears bimodal, especially for at-fault crashes, while the percentages of non-crashing drivers falls off in a linear fashion with declining functional ability.

The data plotted in figure 31 are presented in tables 27, 28, and 29 of appendix F. The chi-square test results noted above, with corresponding cell counts, can be found in table 48 of appendix G.

Delayed Recall. Figure 32 shows the relationships between performance on the Delayed Recall procedure and the three indices of crash involvement analyzed here.

Figure 32. 
Delayed Recall Performance Distributions and Odds Ratios for Analyses
Including All Crashes, At-Fault and Unknown-Fault Crashes,
and At-Fault Crashes Only

 Graphs of Figure 32.
 Delayed Recall Performance Distributions and Odds Ratios for
Analyses

As shown, the association between functional performance and crash involvement, revealed through calculated OR values at each of the four possible levels for this measure, indicates elevated crash risk with a greater loss of working memory. The association is progressively stronger moving from "all crashes" to "at-fault" crashes only. In the latter case, for drivers who missed all 3 items crash risk was elevated by 2.92 times, which was statistically significant at p<.02 (c2 = 5.25, df=1). At the same time, the proportion of the sample who were crash involved began to exceed those who were crash free at the level of two incorrect responses, suggesting this as a potential cutpoint for this measure.

The data plotted in figure 32 are presented in tables 30, 31, and 32 of appendix F. The chi-square test results noted above, with corresponding cell counts, can be found in table 49 of appendix G.

Useful Field of View, Subtest 2. Figure 33 contains the results for the Useful Field of View, Subtest 2. The plots in this figure allow comparison of the distributions of crash-involved and non-crash-involved drivers at each target duration for this measure. It may be noted that poorer performance is signified when drivers need longer durations to correctly identify the target; and, each value on the x-axis is actually the midpoint of a 50 msec interval.

Figure 33.
Useful Field of View, Subset 2 Performance Distributions and Odds, 
Ratios for Analyses Including All Crashes, At-Fault and Unknown-Fault Crashes,
and At-Fault Crashes Only

Graphs of Figure 33. Useful Field of View, Subset 2 Performance Distributions and Odds,
Ratios for Analyses

While the performance level at which the proportion of crash-involved drivers exceeds non-crash-involved drivers is 250 msec, the peak OR of 2.48 for this measure obtains at a slightly longer duration, 300 milliseconds. The calculated OR is statistically significant (c2 = 6.95, df=1, p<.01) at the latter cutpoint (which is an interval with lower boundary set at 275 msec).

Though less pronounced than MVPT/VC, the plots for Subtest 2 of the Useful Field of View measure also suggest a multimodal shape for the crash-involved group, most noticeably for at-fault crashes. Interpretation is complicated by the spike at 500 msec; as noted earlier, this is an artifact of the measurement technique, inasmuch as all responses at this target duration and longer were coded with the same value.

The data plotted in figure 33 are presented in tables 33, 34, and 35 of appendix F. The chi-square test results noted above, with corresponding cell counts, can be found in table 50 of appendix G.

Trail-Making, Part B. The results for this paper-and-pencil test of perceptual-cognitive ability are displayed in figure 34. As observed in the related, Useful Field of View (Subtest 2) plots displayed previously, the curve relating safety outcome to functional status is essentially flat using "all crashes." Also, the values on the x-axis in this figure are again actually the midpoints of intervals; each interval is 40 msec long.

Figure 34.
Trail-Making, Part B Performance Distributions and Odds Ratios 
for Analyses Including All Crashes, At-Fault and Unknown-Fault Crashes,
and At-Fault Crashes Only

Graphs of Figure 34. Trail-Making, Part B Performance Distributions and Odds Ratios for
Analyses

A strong consistency observed in these data is that the proportion of drivers in the sample who were crash-involved began to exceed those who were crash free at the 100 second performance level, across all crash categories. This suggests that 100 seconds may be the best candidate for a cutpoint on this screening measure.

The results reported in the middle plot show a somewhat stronger association overall but do not show any clear peak for the calculated OR. It isn't until the bottom plot for at-fault crashes that the OR shows a clear peak (3.50) at the 100 second level. Drivers were over 3½ times more likely to be involved in an at-fault crash if their score was 80 seconds (i.e., the lower bound of this analysis interval) or longer on this measure, a statistically significant outcome (c2 = 7.72, df=1, p<.01).

The data plotted in figure 34 are presented in tables 36, 37, and 38 of appendix F. The chi-square test results noted above, with corresponding cell counts, can be found in table 51 of appendix G.

Dynamic Trails. Figure 35 plots the results for Dynamic Trails. This automated test was related to the paper-and-pencil Trail-making (Part B) measure but was shorter, with fewer test items, and also potentially more distracting, with moving traffic in the background instead of a blank page.

Figure 35. 
Dynamic Trails Performance Distributions and Odds Ratios for Analyses
Including All Crashes, At-Fault and Unknown-Fault Crashes,
and At-Fault Crashes Only

Graphs of Figure 35. Dynamic Trails Performance Distributions and Odds Ratios for
Analyses

A peak valid OR of 1.45 was calculated for this measure at a test completion time of 25 seconds, for the "at-fault" crash category. This outcome was not statistically significant (c2 = .57, n.s.). In part, this outcome may reflect the fact that the sample size (n = 777) for this particular measure was only about half of that attained for the other procedures in the screening battery. Also, as reported anecdotally by test administrators at the MVA field data collection sites, participants had the greatest difficulty understanding the instructions on how to perform this procedure.

To the extent justified by data collection with a larger study sample, choosing a candidate cutpoint for this measure is problematic. At 20 seconds, the percentage of crash-involved drivers first exceeded crash-free drivers in the analyses of at-fault crashes; but the largest differentials between the two distributions were observed at a test completion time of 30 seconds, for all crash categories.

The data plotted in figure 35 are presented in tables 39, 40, and 41of appendix F. The chi-square test results noted above, with corresponding cell counts, can be found in table 53 of appendix G.

Scan Test. The remaining measure of perceptual ability, the Scan Test, was scored simply on a pass/fail basis. With only one criterion possible, OR calculation is irrelevant to cutpoint determination.

For this measure, 95.6 percent of all drivers in the analysis sample--whether crash-involved or not--passed. Whether this was due to insensitivity of the measurement procedure or whether these results reflect a true measurement of generally "intact" functional ability is unclear. Either way, the very small percentage of drivers failing the measure precludes reliable estimates of statistical significance. Specifically, the sample would have to be much larger, and/or the criterion to pass the test more stringent and more consistently implemented, to obtain a reliable cell count of drivers with at least one crash who failed the test (see earlier discussion of assumptions and limitations of the odds ratio technique).

Rapid Pace Walk. Figure 36 presents the plots for the Rapid Pace Walk measure. Again, a pattern of results is shown where the relationship between safety outcome and functional status appears progressively stronger moving from "all crashes" to "at-fault" crashes.

Figure 36.
Rapid Pace Walk Performance Distributions and Odds Ratios 
for Analyses Including All Crashes, At-Fault and Unknown-Fault Crashes,
and At-Fault Crashes Only

Graphs of Figure 36. Rapid Pace Walk Performance Distributions and Odds Ratios for
Analyses

A statistically-significant (c2 = 6.11, df=1, p<.01) peak OR value of 2.64 was calculated for this analysis, for the "at-fault" crash category, at the performance level designated 9.75 seconds. A second peak appears in this plot at the shorter time of 5.25 seconds, however, showing evidence of the same type of bimodal distribution of functional performance scores among crash-involved drivers that was observed earlier for MVPT/VC (while the crash-free driver distribution remains linear).

As in the earlier timed measures, each value on the x-axis is the midpoint of an interval; in this case each interval is 1.5 seconds long. Thus the two values noted above connote analysis intervals that begin at 9.0 and 4.5 seconds, respectively. The data plotted in figure 36 are presented in tables 42, 43, and 44 of appendix F. The chi-square test results and cell counts can be found in table 53 of appendix G.

Foot Tap. Data plots of the results of the Foot Tap measure are presented in figure 37. Each value on the x-axis is actually the midpoint of a 1.5 second analysis interval.

Figure 37.
Foot Tap Performance Distributions and Odds Ratios
for Analyses Including All Crashes, At-Fault and Unknown-Fault Crashes,
and At-Fault Crashes Only

Graphs of Figure 37. Foot Tap Performance Distributions and Odds Ratios for
Analyses

As shown, there is a tendency toward higher OR's at faster times, which was somewhat unexpected. Also apparent in figure 37 is a close overlap in the distributions of crash-involved and non-crash-involved drivers, in all three plots. As a result, there are no statistically-significant differences here, even at the peak OR value of 1.50 calculated at the performance level designated 5.25 seconds in the analysis of "at-fault" crashes (c2 = 0.98, n.s.).

The data plotted in figure 37 are presented in tables 45, 46, and 47 of appendix F. The chi-square test results and cell counts can be found in table 54 of appendix G.

Head/Neck Rotation. As another binary (pass/fail) measure, no OR plots were generated for Head/Neck Rotation. Sufficient differences were found to support reliable analyses, however: 36.4 percent of drivers with 1 or more (at-fault) crashes failed this test versus only 18.2 percent of drivers in the non-crash group. The peak OR value of 2.56 for this analysis category was statistically significant (c2 = 4.69, df = 1, p<.03).

The chi-square test results and cell counts for this measure can be found in table 55 of appendix G.

Arm Reach. As with the Scan Test measure, virtually all (99.3 percent) of the drivers in the sample passed the Arm Reach test. Of those who failed, only one driver was involved in an at-fault crash. The lack of drivers failing this measure precluded reliable statistical tests, and renders this procedure of little value as a screening tool.

Relationships of Screening Measures With Conviction Data

This section quantifies and tests the significance of the statistical relationships between the functional screening measures and the conviction data extracted from the Maryland Motor Vehicle Administration files. These associations were calculated according to the conventions for measuring, sorting, and summarizing functional status and safety outcome data described previously in this report. A brief overview of the analysis technique follows.

Analysis Techniques

The strength of relationship between functional status and conviction experience was again assessed through the use of the "odds ratio" (OR) calculation. Greater detail about the nature of this calculation and the assumptions that must be met for its valid application were provided at the beginning of the preceding (crash analysis) section.

Results of the OR calculations are indicated in data plots for each functional screening measure used in the Pilot Study. Each plot shows the percentage of the distribution of drivers in the License Renewal sample who would fail a test, at each possible cutpoint, that were convicted of moving violations versus violation-free; and, it shows the calculated OR value at each possible cutpoint.

In accordance with assumptions and limitations of the OR technique explained earlier, a line representing the calculated OR value begins at the second-best level of performance, or first possible cutpoint, marked along the x-axis in each plot presented in this section. Also, in every plot a dashed line, connoting an OR of 1.0, is included for reference. At this level, a driver is as likely to be crash-involved when passing a test as he/she is when failing the test; and the OR effectively has no predictive value. Exact OR values for the data represented in the plots, including each potential cutpoint marked on the x-axis, are presented in appendix I.

Three categories of conviction data are represented in the plots presented in this section: all moving violations; all moving violations except speeding; and, all moving violations except speeding and occupant restraint citations. A variety of specific incident types are subsumed under the heading "moving violations;" these were identified earlier in the section describing the extraction of motor vehicle administration safety data.

Levels of significance of calculated OR values were assessed using chi-square (c2) tests. Test statistics were calculated by SPSS/SYSTAT for each functional performance measure where the strongest relationship with a safety outcome--indicated by the peak valid OR--was demonstrated; in all cases but one, this outcome was moving violations except speeding and restraint citations. As a general finding, it was observed that an OR value of approximately 2, or greater, was associated with a statistically significant (p<.05) chi-square test result.

Results of Conviction Analyses

Motor Free Visual Perception Visual Closure Subtest (MVPT/VC). Figure 38 contains the results for the MVPT/VC. The top plot relates functional performance to conviction experience using "all moving violations" as the safety outcome measure. The middle plot relates functional status to the more restrictive outcome measure of "moving violations without speeding," and the bottom plot shows the distributions of License Renewal sample drivers with and without moving violations excluding speeding and occupant restraint citations at each level of functional ability measured by this test. In all cases, declining functional ability is indicated by an increasing number of incorrect responses, moving to the right along the x-axis.

Figure 38.
MVPT/Visual Closure Subtest Distributions and Odds Ratios for Analyses 
Including All Moving Violations, Moving Violations without Speeding, and Moving Violations 
without Speeding and Occupant Restraint Citations

Graphs of Figure 38. MVPT/Visual Closure Subtest Distributions and Odds Ratios for
Analyses

Inspection of this figure reveals stronger relationships moving from the top plot, where the OR curve is virtually flat with calculated values all near 1.0, to the bottom data plot where a statistically significant (c2 = 10.83, df=1, p<.001) odds ratio of 4.53 was found. The cutpoint where this result was obtained was at a performance level of six incorrect responses. As shown in figure 38, a higher OR value was calculated for seven incorrect responses, but cell counts were too small for this calculation to be valid.

A consistent result that also is shown by this figure is the pattern in the relative percentages of drivers in the distribution with violations versus the percentage who were violation-free. In all three plots, there is a reversal at the performance level of three incorrect responses; otherwise, at every level of this measure except perfect performance (zero errors) more drivers who "failed" the test at a given cutpoint had moving violations than the number who remained violation-free.

The data plotted in figure 38 are presented in tables 56, 57, and 58 of appendix H. The chi-square test results noted above and cell counts can be found in table 77 of appendix I.

Delayed Recall. The relationships between performance on the Delayed Recall procedure and the three categories of moving violations are described by the plots shown in figure 39.

Figure 39.
Delayed Recall Distributions and Odds Ratios for Analyses Including All Moving Violations,
Moving Violations without Speeding, and Moving Violations without Speeding
and Occupant Restraint Citations

Graphs of Figure 39.
Delayed Recall Distributions and Odds Ratios for
Analyses

As shown, the association between functional status and moving violations, revealed through calculated OR values at each of the four possible levels for this measure, is generally weak. The peak valid OR, calculated for data described by the bottom plot, was 1.72. This result was obtained at the level of two incorrect responses; it approached but did not reach statistical significance (c2 = 1.58, n.s.).

The data plotted in figure 39 are presented in tables 59, 60, and 61 of appendix H. The chi-square test results noted above, with corresponding cell counts, can be found in table 78 of appendix I.

Useful Field of View, Subtest 2. Figure 40 contains the results for the Useful Field of View, Subtest 2. The plots in this figure allow comparison of the distributions of drivers with and without moving violations at each target duration characterizing different performance levels for this measure. As noted earlier in the crash analysis section, all responses at target durations longer than 500 msec were grouped together at that performance level.

Figure 40.
Useful Field of View, Subset 2 Distributions and Odd Ratios for Analyses 
Including All Moving Violations, Moving Violations without Speeding, 
and Moving Violations without Speeding and Occupant Restraint Citations

 Graphs of Figure 40. Useful Field of View, Subset 2 Distributions and Odd Ratios for
Analyses

As shown in this figure, OR values hover near 1.0 at all performance levels, for all analysis categories, with almost exactly matching distributions of drivers with and without moving violations at each cutpoint. The peak valid OR calculated for Useful Field of View, Subtest 2 was 1.67; this result obtained at the target duration designated 100 msec in the analysis of "moving violations except speeding and occupant restraint citations." This result was not statistically significant (c2 = 1.53, n.s.).

The data plotted in figure 40 are presented in tables 62, 63, and 64 of appendix H. The chi-square test results noted above, with corresponding cell counts, can be found in table 79 of appendix I.

Trail-making, Part B. The results for this paper-and-pencil test of perceptual-cognitive ability are displayed in figure 41. After MVPT/VC, this measure evidenced the strongest relationship of functional ability with moving violations found in these analyses.

Figure 41.
Trail-Making, Part B Distributions and Odd Ratios for Analyses 
Including All Moving Violations, Moving Violations without Speeding, 
and Moving Violations without Speeding and Occupant Restraint Citations

Graphs of Figure 41. Trail-Making, Part B Distributions and Odd Ratios for
Analyses

Inspection of the OR curves in figure 41 shows the highest values in the middle and bottom plots. The highest valid OR calculated for this measure, 1.72, was found at the performance level designated 140 seconds for the analysis of moving violations except speeding. This result was statistically significant at p<.01 (c2 = 6.70, df=1).

The 140 msec performance level was also the cutpoint at which the percentage of drivers with moving violations exceeded the percentage of violation-free drivers by the widest margins, for all three of the analysis categories.

The data plotted in figure 41 are presented in tables 65, 66, and 67 of appendix H. The chi-square test results noted above, with corresponding cell counts, can be found in table 80 of appendix I.

Dynamic Trails. Figure 42 plots the results for Dynamic Trails. This automated test was related to the paper-and-pencil Trail-making (Part B) measure but was shorter, with fewer test items, and also potentially more distracting, with moving traffic in the background instead of a blank page.

Figure 42.
Dynamic Trails Distributions and Odd Ratios for Analyses 
Including All Moving Violations, Moving Violations without Speeding, 
and Moving Violations without Speeding and Occupant Restraint Citations

Graphs of Figure 42. Dynamic Trails Distributions and Odd Ratios for
Analyses

With the exception of a spike at the 50-second performance level for the data in the bottom plot, which represented too few drivers for a valid analysis, the calculated OR value for this measure hovers near 1.0 across the board. The peak valid OR, 1.27, was found at the 25-second cutpoint in the bottom plot; this result was not statistically significant (c2 = .24, n.s.). However, there is convergence in these findings with the (at-fault) crash analysis, which also demonstrated a peak valid odds ratio at the same cutpoint.

It may again be noted that the sample size (n = 759) for this particular measure was only about half of that attained for other procedures in the screening battery.

The data plotted in figure 42 are presented in tables 68, 69, and 70 of appendix H. The chi-square test results noted above, with corresponding cell counts, can be found in table 81 of appendix I.

Scan Test. The remaining measure of perceptual ability, the Scan Test, was a binary measured scored simply on a pass/fail basis. With only one criterion possible, OR calculation is irrelevant to cutpoint determination, and no data plot was prepared.

For this measure, 95.6 percent of all drivers in the study sample--whether violation-involved or not--passed. Whether this was due to the insensitivity of the measurement procedure or whether these results reflect a true measurement of generally "intact" functional ability is unclear. Either way, the very small percentage of drivers failing the measure indicates a very limited utility for the Scan Test as a screening tool.

Of the 81 drivers who failed the Scan Test, only one was convicted of a non-speeding, non-occupant-restraint violation. This result precluded a valid OR calculation, and no chi-square test was performed for these data.

Rapid Pace Walk. Figure 43 presents the plots for the Rapid Pace Walk measure, the first of the physical screening tests for which results are reported. As shown, the calculated OR value is at or below 1.0 except for the highest test completion times in all three plots for this measure.

Figure 43.
Rapid Pace Walk Distributions and Odd Ratios for Analyses 
Including All Moving Violations, Moving Violations without Speeding, 
and Moving Violations without Speeding and Occupant Restraint Citations

Graphs of Figure 43. Rapid Pace Walk Distributions and Odd Ratios for
Analyses

The peak valid OR, 1.48, was calculated at the performance level designated 5.25 seconds in the analysis of moving violations except speeding. This result was not statistically significant (c2 = 0.96, n.s.). This same performance level was also where the percentage of drivers with moving violations exceeded the percentage without violations by the largest amount, in each analysis category.

The data plotted in figure 43 are presented in tables 71, 72, and 73 of appendix H. The chi-square test results and cell counts can be found in table 82 of appendix I.

Foot Tap. Data plots of the results of the Foot Tap measure are presented in figure 44. As shown, the odds ratio curves for the top two analysis categories are very close to the dashed horizontal line (OR = 1.0), indicating no relationship, until the poorest performance levels are reached. In fact, the peak valid OR of 2.14 is found in the top plot, at the 12.75-s level; this result approached but failed to reach statistical significance (c2 = 2.34, n.s.).

Figure 44.
Foot Tap Distributions and Odd Ratios for Analyses 
Including All Moving Violations, Moving Violations without Speeding, 
and Moving Violations without Speeding and Occupant Restraint Citations

Graphs of Figure 44. Foot Tap Distributions and Odd Ratios for
Analyses

In the bottom plot, higher OR values were found, but cell counts were too few for a valid analysis. Also, the OR values in this plot range from higher to lower as performance shifts from "intact" to greater and greater degrees of functional loss. This counterintuitive finding is consistent with the results observed for this measure in the earlier analysis of (at-fault) crashes.

The data plotted in figure 44 are presented in tables 74, 75, and 76 of appendix H. The chi-square test results and corresponding cell counts can be found in table 83 of appendix I.

Head/Neck Rotation. Because only pass/fail outcomes are possible for this (binary) measure, no odds ratio plot was prepared for the Head/Neck Rotation data. As noted earlier, 81.4 percent of drivers passed this test. When analyzed to examine the relationship between performance on this measure and moving violation experience, these data included only three drivers who failed the test and had at least one non-speeding, non-occupant restraint violation. This result also precluded a valid calculation of OR, and no statistical tests were performed on these data.

Arm Reach. Results for this remaining measure of physical ability, another binary (pass/fail) measure, were the most skewed among all screening activities as 99.3 percent passed, and only14 failed this test. Among those who failed, there were no drivers who received convictions for non-speeding, non-occupant-restraint violations. Accordingly, no valid OR calculations were permitted, and there are no chi-square test results to report.

Resource Requirements for Functional Screening

This section documents costs associated with the functional screening and evaluation activities undertaken in the Maryland Pilot Older Driver Study. It encompasses administrative and support activities, as well as the time actually spent by State employees performing the various testing procedures. The included cost data, as compiled by the MVA, represent the incremental costs of carrying out the Pilot Study, specifically; the costs associated with medical review of referred drivers when an activity or procedure was already a part of existing processes at the licensing agency are accounted for separately. Also, costs associated with the development of materials and procedures used during driver screening and evaluation by MVA staff are omitted from this accounting, to the extent that research team members' labor or equipment were covered under this NHTSA contract or other sources of extramural funding.

After documenting the costs experienced in a research setting to acquire the functional screening data in the Pilot Study, a projection of the cost per licensed driver interacted with by the MVA to accomplish functional screening in a production setting is presented, consistent with program parameters provided by the MVA. Supplemental costs associated with post-screening (education and counseling) activities are similarly estimated.

The cost accounting below is keyed to four categories: labor; equipment; training and quality control; and overhead. Labor costs include salary, and benefits where applicable, for the staff who conduct functional testing and who perform program administrative functions such as scheduling, customer contact, and data management. Equipment costs pertain to hardware and software resources needed to administer the functional tests. Training and quality control costs cover the time spent by MVA staff preparing to perform testing activities, and participating in periodic "refresher" sessions to maintain consistency in the administration of screening procedures. Overhead costs are limited to the space required to carry out the screening activities, apportioned according to the amount of time multi-purpose facilities at the MVA were dedicated to these activities.

Because different activities were performed in different venues, cost-per-driver-screened figures are calculated initially for screening activities performed with license renewal populations, then modified to account for differential costs in screening medical referral and residential community populations.

Beginning with functional screening for the license renewal sample in the Pilot Study. the total number of drivers who participated in screening activities was 2,381. Though only data for 1,876 were complete and valid, the costs described in this section will be based on the total number of drivers tested during the 11-month interval from the end of November to late in the following October.

To collect these data, the MVA utilized 7 line personnel who worked three days per week on this project. This translates to 4.2 full-time employees (FTE). The average hourly wage including benefits for a line employee is $15.00. Based on a work year of 2,080 hours, the cost for one FTE was $31,200; thus, the total annual cost for the 4.2 full-time employees who conducted screening may be estimated at $131,040. Adjusted for an 11-month study period, the resulting labor cost to acquire screening data for the license renewal sample was $120,120.

Administrative and logistics support for this data collection activity was provided by two research associates in the MVA Driver Safety Research Office, who devoted approximately one-third of their time each. At an hourly rate of $33.00, this resulted in an additional 0.66 FTE at an annual cost of $45,760. The adjusted figure for the 11-month duration of the Pilot Study is $41,947. Thus, total labor costs to perform functional screening for the license renewal sample in the Pilot Study may be estimated at $162,067.

The costs of equipment dedicated to screening activities in the Pilot Study were confined to additional computers (PC's) and peripheral devices (light pens and scanners), plus materials used for "manual" data collection (e.g., test stimuli and scoring forms). Specifically, three (3) PC's were purchased at $843 each, and subsequently were connected to a wide area network for data acquisition and data entry. Three (3) light pens were purchased at a cost of $258 each, to acquire data for measures where examinees actually needed to touch the screen to indicate their responses. And, two (2) CCD scanners used to read the bar codes on driver's licenses containing their driver identification (Soundex) numbers were purchased, at a cost of $198 each. Seven (7) test kits containing all materials and supplies used to perform the "Gross Impairment Screening" (GRIMPS) measures were also purchased, at a cost of $100 each. Total costs for equipment and supplies therefore may be estimated at $4,399.

Estimated costs associated with training and quality control may be derived based on the time that MVA staff who collected screening data and performed administrative and support functions were engaged in these activities. An initial training exercise spanning two, half-day (4-hour) sessions included ten (10) MVA line personnel and two (2) MVA research associates. For two days following initial training, ten (10) additional line personnel provided on-site supervision and observation of the individuals collecting screening data, for 6 hours each day. Through the duration of the Pilot Study, periodic visits for observation and "refresher" training to promote consistency and reduce errors in data collection and data entry procedures required a total of 12 full days of staff time at the research associate level. Together, these activities required the equivalent of 200 hours of time for line personnel, at $15/hr, plus 112 hours of research associate time at $33/hr, for a total of $6,696.

Finally, the real estate required to collect screening data for license renewal drivers in the Pilot Study consisted of a room in each of three MVA field offices. The rooms, which were used for other MVA functions when screening activities were not being performed, provided a footprint of approximately 100 square feet. At a fair market value of $12/ft²/year, the cost of this space utilized full-time, would be $3,600. Utilized three days per week, the apportioned cost of MVA office space to conduct screening was 60 percent of this amount, or $2,160.

Summing the component costs identified above associated with Pilot Study efforts to acquire the functional abilities screening data, enter and store the data, and generate raw data tables to support the project analyses, for a sample of license renewal drivers tested at MVA field offices yields an estimated total cost of $175,322.

A preliminary estimate of the cost-per-driver-screened in the research settings of the Maryland Pilot Older Driver Study is reached by dividing this amount by the number of licensed drivers tested by the MVA under this program-2,381. The result is $73.63. This estimate is termed "preliminary" because, according to an MVA research associate,3 the amount of time devoted to data collection, per se, averaged no more than 30 minutes per driver. The apparently much larger time requirement suggested by the 4.2 FTE figure cited above reflects a number of factors, most prominently challenges in recruiting the study sample: only older individuals were approached to be asked to volunteer for the license renewal study, and only about half of those approached agreed to participate.

A first step toward developing an estimate of the cost-per-driver-screened in a production setting versus the research setting is reached by limiting the time allowed per driver to only the 30 minutes (or less) that is necessary to acquire functional screening data. Because this activity would no longer be voluntary, many of the extra duties experienced by the MVA staff in the research setting would disappear. With this one adjustment, the cost element represented by the line personnel serving as data collectors in the Pilot Study is reduced to 1,191 hours (i.e., the number needed to screen the license renewal sample at one half-hour per driver) times the hourly wage of $15.00, or $17,865. Including equipment, training and quality control, and overhead costs as previously documented, the adjusted total cost is $31,120, or $13.07 per driver screened.

Next, certain cost elements were modified and others were added as data collection moved into other venues during the Pilot Study. Principal differences were the use of Driver License Examiners (DLE's) instead of line personnel to conduct screening for the medical referral sample; and, the addition of occupational therapists to provide feedback and counseling to drivers on the meaning of their screening results and changes in driving habits they should consider, with the residential community sample.

The DLE staff who performed functional screening of the medical referral sample earned a wage (salary plus benefits) of $20 per hour. The introduction of staff at this level followed observation of inconsistencies in test administration during Pilot Study data collection with the license renewal sample. The DLE staff, who were accustomed to performing a wide range of examination activities, did achieve a higher degree of consistency in administering the functional tests. In addition, because the medically-referred drivers were screened only during scheduled appointments, the test administration time was effectively limited to and consistently fell within the range of 20 to 30 minutes per driver, as stipulated above.

If all functions performed by line personnel in the cost estimate developed above--including training and quality control as well as data collection--are instead performed by DLE-level staff, the adjusted total cost for functional screening including equipment, training and quality control, and overhead increases to $38,075, or an estimate of $15.99 per driver screened.

Finally, when older drivers in the residential community sample were screened in the Pilot Study, an occupational therapist (OT) was available to provide feedback and counseling services. By design, these interactions were to be tailored as follows: functionally intact drivers would receive educational information about the relationship between functional ability and driving risk, advice on self-testing and what to do when abilities begin to decline in the future; while persons "failing" one or more screening measures, in addition to receiving educational information, would be counseled on specific risks posed by their functional impairments and/or what actions were needed vis-à-vis changes in driving habits, where they should go for more in-depth assessment, and what options might be explored to remediate their functional loss. As a practical matter, however, the OT's time was limited to interactions with drivers for whom the screening activities indicated the most pronounced functional deficits. The occupational therapists participating in the Pilot Study were outside consultants, i.e., not MVA staff personnel, who were paid $45 per hour.

If OT's, nurses, or similarly-qualified professionals were engaged to provide counseling services on a broader scale, the incremental cost associated with this service would be driven by the percentage of drivers screened who would "fail" the functional ability screening, and the fraction of this group who would require one-on-one attention from a medical professional to have their questions answered or to receive the necessary referrals for further evaluation and/or to identify remediation options.

It is the perspective of MVA officials4 that not more than 25 percent of the population of renewing drivers in the 55+ cohort would "fail" functional screening using a to-be-selected subset of the measures examined in the Pilot Study, and applying the cutpoints that are best supported by available data relating functional status to safety outcomes as per the analyses reported herein; and further, that a majority of even the "failing" drivers could have their needs for feedback and counseling effectively met by properly trained DLE-level staff. Only those individuals whose questions could not be answered adequately or whose need for an immediate referral required the attention or action of a medical professional would interact with an OT or nurse after completing screening. Accordingly, incremental cost estimates for the provision of post-screening services to the license renewal sample, in a production setting, are based on the following assumptions:

  1. Post-screening feedback for all of the "functionally intact" drivers (75 percent of the total number screened) would be accommodated through interactions with the DLE that focus on education and promote awareness of the functional abilities needed for safe driving, at 5 minutes per interaction;
  2. Eighty percent of drivers with significant functional loss (20 percent of the total number screened) would be accommodated through more extensive interactions with the DLE, at 10 minutes per interaction; and
  3. Twenty percent of drivers with significant functional loss (5 percent of the total number screened) would receive initial feedback from the DLE, lasting up to 10 minutes, then would require additional consultation with a medical professional, at 20 minutes per interaction.

Based on the $20/hr and $45/hr costs experienced in the Pilot Study for DLE and OT labor, respectively, these assumptions yield an incremental cost of $6,745, raising the total cost for screening and evaluation activities in a license renewal context to $44,819 and the cost-per-driver interacted with by the MVA to $18.82.

It deserves mention that no costs have been included in these estimates for Pilot Study involvement by the Chief or the Daily Duty Doctors serving on the Medical Advisory Board at the MVA. While these individuals played key roles in the early planning and later evaluation of screening activities, an ongoing screening program is viewed as but one additional source of information complementing other data currently considered in medical reviews for fitness-to-drive determinations. Since fitness-to-drive determinations are a defining characteristic of the MAB, the only incremental cost in this process is represented by the acquisition of screening data plus whatever post-screening educational and counseling services, if any, are provided to drivers. The consideration of screening outcomes within the context of responsibilities normally discharged by the MAB, by comparison, does not represent an incremental cost.

Perhaps more importantly, it must be emphasized that the cost analysis in this section reflects screening activities (including data entry) that were performed mostly on a manual and labor-intensive basis--only two of the measures were automated--and by MVA staff for whom this was a completely novel assignment. As with any procedure, staff became more efficient and skilled in administering the functional tests with experience, especially the Driver License Examiners.

Most important from a cost standpoint is the potential to automate the majority of the most-promising measures emerging from the Pilot Study. Automation of data entry as well as data collection functions could enable one staff member to direct and monitor the screening of two or perhaps three drivers, and still provide feedback within the parameters outlined above. Under this scenario, the cost-per-driver-screened could be reduced to the range of $5 to $10.

Further discussion relating the cost estimates developed above to the anticipated benefits of a functional capacity screening program to identify persons at high risk of driving impairment is presented in Volume 1 of this report. 


3 pers. comm., Mr. Jack Joyce, Senior Research Associate, Maryland Motor Vehicle Administration Office of Driver Safety Research, 8/9/02.
4 pers. comm.., Dr. Robert Raleigh, Chief, Maryland Medical Advisory Board, telephone conversation on 8/08/02.

BACK  |  CONTENTS  |  NEXT