QAC380-data
team1
3/28/2022
Risk Data: Main VARS: SuicideRisk_Total / HomicideRiskTotal
latested_risk_by_client: data frame of most recently assessed risk levels
variables: StudyClientId, AssessmentYear, AssessmentType, SuicideRisk_Total, HomicideRiskTotal
Take a look at missing values of ChargeSeverity –> Non-labeled offense calss
TO DO: do we need to assign a new class to these offenses or do we just treat it as a binary variable instead (have offense history = 1, not have = 0)?
NOTE: the Offense Data DOES NOT have ClientID or any thing that could be used to reference another dataset. We might need to discard it.
Alcohol Drug Problem & Criminal History Data
Incident Data: Main VAR: IncidentScore (from Incident_byclient dataset)
IncidentScore: a composite measure based on the number of incidents corresponding to its level of severity (i.e. the sum of count * incident level)
IncidentLevel: discrete values 1,2,5 according to its type(5=most severe, i.e. [I] level incident)
All types of incidents are:
## IncidentDaTa$IncidentType
## Frequency Percent
## Aggression/Threatening Behaviors (Not to Person) [II] 24 5.7554
## Client Arrest [III] 60 14.3885
## Contraband Found [III] 89 21.3429
## Death [I] 19 4.5564
## Narcan Administration [I] 36 8.6331
## Police Action [I] 64 15.3477
## Possession of Illicit/Illegal Substances [II] 31 7.4341
## Possession of Weapons [I] 1 0.2398
## Sexual Assault Alleged [I] 35 8.3933
## Suicide Attempt [I] 4 0.9592
## Threat to Person [I] 29 6.9544
## Violence (Physical) [I] 25 5.9952
## Total 417 100.0000Trauma Data: whether or not Clients have any trauma history
Domestic Violence History
Physical Abuse Victim
Sexual Abuse Victim
Other Trauma History
Ever Diagnosed With PTSD
Has Chronic Pain Problems?
ASUS Data
AodInvolvement1: a composite measure of the lifetime involvement in the 10 major drug categories described in the literature. (QUANTITATIVE)
Disruption1: a broad measure of problems and negative consequences due to AOD use. (QUANTITATIVE)
AodLastSixMonths: measures extent of involvement and disruption from AOD use in the past six months
LegalNonConformingLastSixMonths: meaures legal problems (involvement in the adult criminal justice system) over the last six months client has been in the community.
GlobalAodPsychosocialDistruptionAndProblems: comprised of the sum of the four scales: INVOLVEMENT, DISRUPTION, MOOD, and SOCIAL NON-CONFORMING. Provides a global and overall measure of the degree to which the client is indicating life-functioning problems in the areas of substance use, mood adjustment and community compliance.
#asus: median split
#ASUS --> INVOLVEMENT1 & DISRUPTION1
#LAST SIX MONTHS vars + Scale 9 ==> 8 vars
latest_asus <- AsusData %>%
group_by(StudyClientId) %>%
arrange(AssessmentYear) %>%
summarise(AodInvolvement = last(AodInvolvement1),
Disruption = last(Disruption1),
AodLastSixMonths = last(AodLastSixMonths),
LegalNonConformingLastSixMonths = last(LegalNonConformingLastSixMonths),
GlobalAodPsychosocialDistruptionAndProblems = last(GlobalAodPsychosocialDistruptionAndProblems),
AssessmentYear = last(AssessmentYear),
AssessmentType = last(AssessmentType))ClientAllData: merged ClientData with Risk / Incident / Alcohol&Drug / Criminal History & Trauma & Address datasets
# merge all data frames together
ClientAllData <- ClientDaTa %>%
left_join(latest_criminal_by_client[,1:2], by="StudyClientId") %>%
left_join(latest_drug_by_client[,1:2], by="StudyClientId") %>%
left_join(latest_risk_by_client[,1:3], by="StudyClientId") %>%
left_join(Incident_by_client[,1:2], by="StudyClientId") %>%
left_join(Trauma_by_client, by="StudyClientId") %>%
left_join(AdressDaTa %>% group_by(StudyClientId) %>%
summarise(AddressType = last(AddressType),
City = last(City),
State = last(State),
Zip = last(Zip)), by="StudyClientId") %>%
left_join(latest_asus[,1:6], by='StudyClientId')
# recode NA in the IncidentScore to be 0
ClientAllData$IncidentScore[is.na(ClientAllData$IncidentScore)] <- 0
ClientAllData %>%
filter(CurrentAge >= 18) -> ClientAllDataUnivariate Plots
library(ggplot2)
ggplot(data= na.omit(subset(ClientAllData, select = c(CriminalHistory))),
aes(x=reorder(as.character(CriminalHistory), CriminalHistory), fill = as.character(CriminalHistory)))+
geom_bar(na.rm = TRUE)+
ggtitle("Clients with Criminal History")+
xlab("Criminal History")
ggplot(data= na.omit(subset(ClientAllData, select = c(AlcoholDrugProblems))),
aes(x=reorder(as.character(AlcoholDrugProblems), AlcoholDrugProblems), fill = as.character(AlcoholDrugProblems)))+
geom_bar(na.rm = TRUE)+
ggtitle("Clients with Alcohol and Drug Problems")+
xlab("Alcohol and Drug Problems")
ggplot(data= na.omit(subset(ClientAllData, select = c(SuicideRisk_Total))),
aes(x=reorder(as.character(SuicideRisk_Total), SuicideRisk_Total), fill = as.character(SuicideRisk_Total)))+
geom_bar(na.rm = TRUE)+
ggtitle("Risk of Suicide in Clients")+
xlab("Suicide Risk")
ggplot(data= na.omit(subset(ClientAllData, select = c(HomicideRisk_Total))),
aes(x=reorder(as.character(HomicideRisk_Total),HomicideRisk_Total), fill = as.character(HomicideRisk_Total)))+
geom_bar(na.rm = TRUE)+
ggtitle("Homicidal Risk of Clients")+
xlab("Homicide Risk")
ggplot(data=ClientAllData)+
geom_density(aes(x=IncidentScore))+
ggtitle("Density Plot of Client Incidents")
ggplot(data= na.omit(subset(ClientAllData, select = c(TraumaHistory))),
aes(x=reorder(as.character(TraumaHistory),TraumaHistory), fill = as.character(TraumaHistory)))+
geom_bar(na.rm = TRUE)+
ggtitle("Clients Assessed for Trauma")+
xlab("Trauma")
require(tidyverse)
top4cities <- ClientAllData %>%
group_by(City) %>%
summarise(count = n()) %>%
top_n(n = 5, wt = count)
top4cities <- na.omit(top4cities)
ggplot(data = top4cities, aes(x = City, y = count, fill= City)) +
geom_col()+
ggtitle("Clients by Top 4 Cities")
Regression Analyses
#Multiple Regression
my.lm1 <- lm(IncidentScore~CriminalHistory+AlcoholDrugProblems+SuicideRisk_Total+HomicideRisk_Total+TraumaHistory, data= ClientAllData)
summary(my.lm1)##
## Call:
## lm(formula = IncidentScore ~ CriminalHistory + AlcoholDrugProblems +
## SuicideRisk_Total + HomicideRisk_Total + TraumaHistory, data = ClientAllData)
##
## Residuals:
## Min 1Q Median 3Q Max
## -2.0496 -0.4449 -0.3659 -0.2689 15.4116
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 0.21776 0.13902 1.566 0.118
## CriminalHistory 0.03245 0.02198 1.476 0.140
## AlcoholDrugProblems -0.01581 0.02469 -0.640 0.522
## SuicideRisk_Total 0.25670 0.06162 4.166 3.42e-05 ***
## HomicideRisk_Total -0.14014 0.15168 -0.924 0.356
## TraumaHistory -0.01669 0.10096 -0.165 0.869
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 1.438 on 833 degrees of freedom
## (2745 observations deleted due to missingness)
## Multiple R-squared: 0.02313, Adjusted R-squared: 0.01727
## F-statistic: 3.945 on 5 and 833 DF, p-value: 0.001529my.lm2 <- lm(IncidentScore~CriminalHistory+AlcoholDrugProblems+SuicideRisk_Total+HomicideRisk_Total+TraumaHistory+CurrentAge+as.factor(MaritalStatus)+as.factor(Race), data= ClientAllData)
summary(my.lm2)##
## Call:
## lm(formula = IncidentScore ~ CriminalHistory + AlcoholDrugProblems +
## SuicideRisk_Total + HomicideRisk_Total + TraumaHistory +
## CurrentAge + as.factor(MaritalStatus) + as.factor(Race),
## data = ClientAllData)
##
## Residuals:
## Min 1Q Median 3Q Max
## -2.2387 -0.4964 -0.3571 -0.2083 15.2343
##
## Coefficients:
## Estimate Std. Error t value
## (Intercept) -0.333225 1.191227 -0.280
## CriminalHistory 0.033498 0.022654 1.479
## AlcoholDrugProblems -0.022824 0.025781 -0.885
## SuicideRisk_Total 0.260025 0.063007 4.127
## HomicideRisk_Total -0.163307 0.153426 -1.064
## TraumaHistory -0.016929 0.102674 -0.165
## CurrentAge 0.011144 0.004973 2.241
## as.factor(MaritalStatus)Civil Union -0.342551 1.562909 -0.219
## as.factor(MaritalStatus)Divorced/Annulled -0.107220 1.197983 -0.090
## as.factor(MaritalStatus)Legally separated 0.083069 1.251447 0.066
## as.factor(MaritalStatus)Married 0.018202 1.198301 0.015
## as.factor(MaritalStatus)Not Specified 0.037473 1.204298 0.031
## as.factor(MaritalStatus)Single/Never Married 0.091131 1.182061 0.077
## as.factor(MaritalStatus)Widow/widower 0.629448 1.329528 0.473
## as.factor(Race)American Indian or Alaskan Native 0.410987 0.552294 0.744
## as.factor(Race)Asian -0.516238 1.044583 -0.494
## as.factor(Race)Caucasian or White 0.063367 0.129017 0.491
## as.factor(Race)Multi-Racial -0.012667 0.221008 -0.057
## as.factor(Race)Not on file -0.272445 1.185920 -0.230
## as.factor(Race)Some other race 0.002278 0.132248 0.017
## as.factor(Race)Undisclosed -0.715118 0.843170 -0.848
## Pr(>|t|)
## (Intercept) 0.7798
## CriminalHistory 0.1396
## AlcoholDrugProblems 0.3763
## SuicideRisk_Total 4.05e-05 ***
## HomicideRisk_Total 0.2875
## TraumaHistory 0.8691
## CurrentAge 0.0253 *
## as.factor(MaritalStatus)Civil Union 0.8266
## as.factor(MaritalStatus)Divorced/Annulled 0.9287
## as.factor(MaritalStatus)Legally separated 0.9471
## as.factor(MaritalStatus)Married 0.9879
## as.factor(MaritalStatus)Not Specified 0.9752
## as.factor(MaritalStatus)Single/Never Married 0.9386
## as.factor(MaritalStatus)Widow/widower 0.6360
## as.factor(Race)American Indian or Alaskan Native 0.4570
## as.factor(Race)Asian 0.6213
## as.factor(Race)Caucasian or White 0.6235
## as.factor(Race)Multi-Racial 0.9543
## as.factor(Race)Not on file 0.8184
## as.factor(Race)Some other race 0.9863
## as.factor(Race)Undisclosed 0.3966
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 1.443 on 818 degrees of freedom
## (2745 observations deleted due to missingness)
## Multiple R-squared: 0.03366, Adjusted R-squared: 0.01004
## F-statistic: 1.425 on 20 and 818 DF, p-value: 0.1021my.lm3 <- lm(IncidentScore ~ SuicideRisk_Total + CriminalHistory +
SuicideRisk_Total*AlcoholDrugProblems, data = ClientAllData)
summary(my.lm3)##
## Call:
## lm(formula = IncidentScore ~ SuicideRisk_Total + CriminalHistory +
## SuicideRisk_Total * AlcoholDrugProblems, data = ClientAllData)
##
## Residuals:
## Min 1Q Median 3Q Max
## -3.3771 -0.5345 -0.4358 -0.2884 15.2003
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 0.18975 0.12130 1.564 0.1179
## SuicideRisk_Total 0.48201 0.08178 5.894 4.59e-09 ***
## CriminalHistory 0.04922 0.01988 2.476 0.0134 *
## AlcoholDrugProblems -0.02451 0.02343 -1.046 0.2958
## SuicideRisk_Total:AlcoholDrugProblems -0.03455 0.02075 -1.665 0.0961 .
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 1.763 on 1606 degrees of freedom
## (1973 observations deleted due to missingness)
## Multiple R-squared: 0.04228, Adjusted R-squared: 0.0399
## F-statistic: 17.73 on 4 and 1606 DF, p-value: 2.997e-14Latent Class Analysis
Categorize variables for latent class analysis
RISK data: Suicide_class / Homicide_class ( low - medium - high)
SuicideRisk –> scale 0-8, <3 is low, 3-5 is medium, 6-8 is high
HomicideRisk -> scale 0-7, <2 is low, 2-3 is medium, > 3 is high
ASUS data: Aod1_class / Aod2_class (last 6 mos) / Disruption_class / LegalNonConform_class / LifeFunctionality_class ( low - medium - high)
Other: Trauma_class (yes - no/unknown) / age_group (level: 18-30 / 31-50 / >50)
# age groups: 18-30, 31-50, and > 50
ClientAllData <- ClientAllData %>%
mutate(age_group = case_when(CurrentAge <= 30 ~ '18-30',
CurrentAge %in% c(31:50) ~ '31-50',
CurrentAge > 50 ~ '>50'),
# split AodInvolvement based on 0-0.35, 0.25-0.75, 0.75 quantiles
Aod1_class = cut(AodInvolvement,
c(-Inf, quantile(AodInvolvement, c(.25, .75),na.rm=TRUE), Inf),
labels=c('low','medium','high')),
# split AodLastSixMonths based on 0-0.35, 0.25-0.75, 0.75 quantiles
Aod2_class = cut(AodLastSixMonths,
c(-Inf, quantile(AodLastSixMonths, c(.25, .75),na.rm=TRUE), Inf),
labels=c('low','medium','high')),
# split Disruption based on 0-0.35, 0.25-0.75, 0.75 quantiles
Disruption_class = cut(Disruption,
c(-Inf, quantile(Disruption, c(.25, .75),na.rm=TRUE), Inf),
labels=c('low','medium','high')),
# split LegalNonConformingLastSixMonths based on 0-0.35, 0.25-0.75, 0.75 quantiles
LegalNonConform_class = cut(LegalNonConformingLastSixMonths,
c(-Inf, quantile(LegalNonConformingLastSixMonths, c(.25, .75),na.rm=TRUE), Inf),
labels=c('low','medium','high')),
# split GlobalPsychological problems based on 0-0.35, 0.25-0.75, 0.75 quantiles
LifeFunctionality_class = cut(GlobalAodPsychosocialDistruptionAndProblems,
c(-Inf, quantile(GlobalAodPsychosocialDistruptionAndProblems, c(.25, .75),na.rm=TRUE), Inf),
labels=c('low','medium','high')),
Suicide_class = case_when(SuicideRisk_Total < 3 ~ 'low',
SuicideRisk_Total %in% 3:5 ~ 'medium',
SuicideRisk_Total > 5 ~ 'high'),
Homicide_class = case_when(SuicideRisk_Total < 2 ~ 'low',
SuicideRisk_Total %in% 2:3 ~ 'medium',
SuicideRisk_Total > 3 ~ 'high'),
Trauma_class = ifelse(TraumaHistory == 1, "yes", "no/unknown")
)
ClientAllData$Trauma_class[is.na(ClientAllData$Trauma_class)] <- "no/unknown"