05_conclusion.qmd

# Conclusions

The purpose of this research was to evaluate the benefits of UDOT's expanded IMT program by evaluating and comparing IMT performance measures for all crashes in 2018 and 2022 that were responded to by IMTs as well as the user impacts for those corresponding crashes. Regression models provided better understanding of the incident characteristics and crash variables that have the greatest impact on performance measures and user impacts.

The $Year 2018$ variable represents the difference in crash data of the IMT program between years 2018 and 2022 between which time the program expanded from 13 units to 25 units. The results of the analysis of this variable show that there was a statistically significant difference between IMT RT in 2018 and 2022 of approximately 2.7 minutes between the two years. This demonstrates that the expansion allows IMTs to respond faster to crashes presumably due to there being more of them covering the study area. The variables with the greatest effect on $RT$ were the number of IMTs ($N. IMTs$) that responded to an incident and $TWNLC$, where each additional IMT that responded was correlated with approximately 2.0 minutes less RT, and each added lane that was closed for the duration of the incident was correlated with an added 1.9 minutes of RT; this also includes the effect of the number of IMTs within the same model.

RCT and ICT did not have any statistically significant improvements due to the program expansion, likely due to the IMTs emphasizing safety as part of post-COVID-19 pandemic protocols and an increase in the proportion of PI crashes to PDO crashes in 2022. The variables that were shown to best explain these performance measures were the $N. IMTs$ and $N. UHP Teams$ that responded to an incident as well as $RT$. While each of the variables mentioned previously were continuous variables, the $Time Range$ variable (categorical) provided added meaning to the model by eliminating outliers in the morning off peak period.

Each user impact had a statistically significant and practically significant $Year 2018$ variable coefficient that showed that AV decreased by approximately 11 percent between 2018 and 2022 and that both ETT and EUC decreased by approximately 50 percent due to the added number of IMTs in the program. This shows that more IMTs responding to more crashes made a moderate reduction to the number of vehicles who experienced delay for a given crash but made a major reduction in the total delay and time cost experienced per by all roadway users who experienced delay in a given crash.

One of the limitations of the $Year 2018$ variable is that this accounts for all differences between 2018 and 2022 variables not accounted for by other variables, and while it is assumed to primarily account for the difference in IMT program size and the number of units on the road, it may not account for the effect of post-pandemic traffic conditions even though traffic flow rates have largely returned to normal.

The key independent variables that explain user impacts of incidents responded to by IMTs are shown in @tbl-conclusions . The impact of each variable includes the effects of T~7~-T~0~ (which had a natural log transformation applied to the variable due to the non-linear trends observed in the data). The independent variables which had a natural log transformation taken for them are interpreted relatively in percent changes rather than in units of time due to their effect on user impacts being non-linear. For the purpose of summarizing the impacts of each variable, a 10 percent increase of the independent variable was used to make a standard comparison.

For T~7~-T~0~, a 10 percent increase resulted in a 9 percent increase in AV; the $LnT_7-T_0*Year 2018$ variable indicated a difference of only 1 percent the effect of T~7~-T~0~ on AV in 2018 and 2022. T~7~-T~0~ had a greater impact on ETT and EUC of a 20 percent increase per 10 percent increase in T~7~-T~0~ when analyzed individually and a 17 percent increase when the effects of the variables *Ln RCT* and either $TWNLC$ or $N. Lanes Closed$ were included. This demonstrates that the time for which speed of traffic is reduced significantly below normal is the primary independent variable influencing delay. The $Ln T_7-T_0*Year 2018$ variable was not statistically significant for ETT and EUC, showing that the effect of the IMT program expansion on The T~7~-T~0~ did not have a significant impact on ETT and EUC.

RCT did not have a statistically significant impact on AV when analyzed in the same model as T~7~-T~0~; while it was a statistically significant variable for ETT and EUC, it only increased these by 2 percent for each 10 percent increase in RCT. The number of lanes closed and TWNLC were not found to have a statistically significant effect on AV. However, the impact of 1 additional lane closed at any point during an incident increased ETT and EUC by 45 percent, and the impact of 1 additional lane closed for the duration of an incident increased ETT and EUC by 62 percent. The variable other than T~7~-T~0~ that was found to have the greatest impact on AV was the total number of lanes at the bottleneck of a crash, which caused a 19 percent increase in AV and 21 percent increase in ETT and EUC for each added lane. This demonstrates that the number of lanes closed at any point during an incident and the time for which they are closed have a large impact on the delay experienced by roadway users while the total number of lanes is better correlated with the number of roadway users affected by a crash.

```{r conclusions, echo=FALSE}
#| label: tbl-conclusions
#| tbl-cap: Impact of Independent Variables on User Impacts
library(kableExtra)
tibble::tribble(
 ~ `Independent Variable`, ~ `Affected Volume`, ~ `Excess Travel Time and Cost`,
 "$T_7-T_0$", "9% increase per 10% increase in $T_7-T_0$"   , "17-20% increase per 10% increase in $T_7-T_0$ (see regression models)",
"RCT" , NA , "2% increase per 10% increase in RCT",
 "N. Lanes Closed", NA , "45% increase per added lane closed at any point during the crash",
"TWNLC"     , NA , "62% increase per added lane closed for the duration of the crash",
"Total Lanes" , "19% increase per added lane" , "21% increase per added lane"
) |> 
  kbl(booktabs = TRUE, escape = TRUE) |> 
  kable_styling()
```

This study is the first of its kind to analyze both IMT performance measures and user impacts for detailed crash data to evaluate the expansion of an existing IMT program. The findings on the expansion of UDOT's IMT program provides a valuable reference point for other transportation agencies considering expanding their own program due to the significant benefits of reduced user impacts associated with additional IMTs responding to more crashes within a given coverage area.

This study also highlights the crash data variables that have the greatest impacts on IMT performance measures and user impacts to help provide a better understanding to the field of TIM researchers and professionals. While other studies have performed detailed analysis on performance measures given incident characteristics, few studies have incorporated RT to model RCT and ICT. Even fewer studies have been conducted on delay and user impacts associated with crashes responded to by IMTs using detailed incident data instead of simulation models, and the findings from this study provide well-fitted, simplistic models reflecting the impacts of incident characteristics and IMT performance measures on user impacts.