Data source

This dataset of ICE Segregation Review Management System (SRMS) records from September 1, 2018 through September 1, 2018 was released pursuant to a FOIA request by Harvard Law School and published as replication data for the report ‘“Endless Nightmare” Torture and Inhuman Treatment in Solitary Confinement in U.S. Immigration Detention’ by Harvard Immigration and Refugee Clinical Program (HIRCP) and Arevik Avedian at Harvard Law School (HLS), Peeler Immigration Lab (PIL) at Harvard Medical School (HMS), and Physicians for Human Rights (PHR).

Notebook goal

This notebook serves as an appendix to UWCHR’s NWDC Conditions Research Update: Charles Leo Daniel’s Death at NWDC in Context. Analysis focuses on ICE Segregation Review Management System (SRMS) records reported by the Northwest ICE Processing Center in Tacoma, WA; also known as the Northwest Detention Center (NWDC), and coded as “TACOMA ICE PROCESSING CENTER (NORTHWEST DET CTR) (WA)” in the below data. Given that this analysis was conducted in response to a breaking news event, it may be updated to reflect ongoing analysis.

library(pacman)
p_load(tidyverse, arrow, lubridate, zoo, digest, ggplot2, plotly, gghighlight, viridis, gridExtra)

df <- read_delim(here::here('analyze', 'input', 'SRMS_spreadsheet_9.1.2018_-_9.1.2023_Redacted.csv.gz'), delim='|',
                      col_types = cols('Placement Date' = col_date(format="%m/%d/%Y"),
                                       'Release Date' = col_date(format="%m/%d/%Y"))) %>% 
  drop_na('Tracking Number', 'Facility AOR')

names(df) %<>% stringr::str_replace_all("\\s","_") %>% tolower

nwdc_det_trinidad <- read_delim(here::here('analyze', 'frozen', 'nwdc_trinidad_and_tobago_fy11-24ytd.csv.gz'), delim = "|",
                            col_types = cols(stay_book_in_date_time = col_datetime(format=''),
                                             stay_book_out_date_time = col_datetime(format=''),
                                             detention_book_in_date_and_time = col_datetime(format=''),
                                             detention_book_out_date_time = col_datetime(format='')
                                             ))

nwdc <- "TACOMA ICE PROCESSING CENTER (NORTHWEST DET CTR) (WA)"

# Checking for any variation in SEA AOR facility names; NWDC should be only one present
stopifnot(length(unique(df[df$facility_aor == "SEA (Seattle)", "facility"])) == 1)

# Data cleaning per `dataverse_files/analysis/data_prep.do`
df <- df %>% 
  mutate(mental_illness = case_when(mental_illness == "Yes" ~ "Mental Illness",
                                    TRUE ~ mental_illness),
         facility = case_when(facility == "WEBB COUNTY DETENTION CENTER (TX)" ~ "WEBB COUNTY DETENTION CENTER (CCA) (TX)",
                              TRUE ~ facility))

# Physicians for Human Rights/Harvard Law School report omits records with ongoing solitary placements
# Uncomment this filter to reproduce PHR/HLS report results:
# df <- df %>%
# filter(!is.na(release_date))

Data overview

The HLS/PHR dataset includes a total of 14251 SMRS placement records for 125 U.S. ICE detention facilities. For a full description of the dataset, see the full report.

The dataset includes a total of 419 SRMS placement records for the Northwest Detention Center (NWDC) in Tacoma, WA; including 8 solitary placements that were ongoing at the time of release of the dataset. These ongoing solitary placements are included in this analysis.

Death of Charles Leo Daniel at NWDC

On March 7, 2024, the death of Charles Leo Daniel was reported at Northwest Detention Center (NWDC). Daniel, a citizen of Trinidad and Tobago, was reportedly in segregation at the time of his death. ICE detention records processed in a separate repository show that Daniel was booked in to the NWDC on April 2, 2020, and that he was the only citizen of Trinidad and Tobago to be detained at the facility from his arrival in 2020 until August 15, 2022 (see analyze/frozen/nwdc_trinidad_and_tobago_fy11-24ytd.csv.gz).

Therefore, we conclude that the two ICE SRMS records in the dataset for a male citizen of Trinidad and Tobago at NWDC represent Daniel’s segregation history while detained at the facility. The full records are presented below (note redaction of some values):

nwdc_seg_trin <- df %>% 
  filter(str_detect(facility, "TACOMA"),
         country_of_citizenship == 'TRINIDAD AND TOBAGO') %>% 
  mutate(placement_date = as.character(placement_date),
         release_date = as.character(release_date))

transpose_df <- function(df) {
  t_df <- data.table::transpose(df)
  colnames(t_df) <- rownames(df)
  rownames(t_df) <- colnames(df)
  t_df <- t_df %>%
    tibble::rownames_to_column(.data = .) %>%
    tibble::as_tibble(.)
  return(t_df)
}

knitr::kable(transpose_df(nwdc_seg_trin) %>% 
               rename(column = rowname,
                      placement_1 = "1",
                      placement_2 = "2"),
             caption = "NWDC SRMS records for citizens of Trinindad and Tobago")
NWDC SRMS records for citizens of Trinindad and Tobago
column placement_1 placement_2
tracking_number (b)(7)(E) (b)(7)(E)
alien_number (b)(6), (b)(7)(C) (b)(6), (b)(7)(C)
gender Male Male
country_of_citizenship TRINIDAD AND TOBAGO TRINIDAD AND TOBAGO
facility_aor SEA (Seattle) SEA (Seattle)
facility TACOMA ICE PROCESSING CENTER (NORTHWEST DET CTR) (WA) TACOMA ICE PROCESSING CENTER (NORTHWEST DET CTR) (WA)
report_type 14 Day 72 Hour
placement_date 2020-04-16 2021-06-25
placement_reason Detainee Request: Other Protective Custody
release_date 2021-06-23 NA
detailed_placement_reason (b)(7)(E)(b)(6)(b)(7)(C) (b)(7)(E)(b)(6)(b)(7)(C)
attorney_of_record No No
attorney_notification No No
detainee_request_segregation Detainee Request Detainee Request
compliance_with_detention_standards Yes Yes
mental_illness Serious Mental Illness Serious Mental Illness
serious_medical_illness No No
serious_disability No No
smi_spi_dis_description (b)(6), (b)(7)(C) (b)(6), (b)(7)(C)
fo_recommendation (b)(6), (b)(7)(C) (b)(6), (b)(7)(C)
created 5/5/2020 14:42 7/7/2021 17:42
length_of_stay 433 811
suicide_risk? NA NA

The first solitary placement began on 2020-04-16, shortly after Daniel was booked into the NWDC and lasted until 2021-06-23, or 433 days. The second solitary placement began two days later on 2021-06-25 and was ongoing at the time of release of the HLS/PHR data in mid-September 2023, or at least 811 days. (As noted in the replication data for the HLS/PHR report, ongoing solitary placements were dropped from analysis for the published report; we include these records here due to their relevance to issues at the NWDC.)

As noted, both stays were reported by ICE as “Detainee Request”, with the second stay categorized as “Protective Custody”. Both stays are also flagged as “Serious Mental Illness”.

As seen below, Daniel’s second placement of at least 811 days was the second longest individual solitary placement reported nationally during the time period covered by the HLS/PHR dataset. As seen below, the longest individual placement reported nationally was also at NWDC and was also ongoing at the time of release of the dataset. (Because unique personal identifiers are not included in this dataset, it is not possible to calculate cumulative stay length per person across individual solitary placements.)

t1 <- df %>% 
  mutate(ongoing_stay = is.na(release_date)) %>% 
  arrange(desc(length_of_stay)) %>% 
  head(10) %>% 
  mutate(length_of_stay = case_when(ongoing_stay == TRUE ~ paste0(as.character(length_of_stay), "+"),
                                    TRUE ~ as.character(length_of_stay))) %>% 
  dplyr::select(country_of_citizenship, facility, length_of_stay, mental_illness) %>% 
  knitr::kable(caption = "Top 10 longest solitary placements nationwide, Sept. 2018-Sept. 2023",
               row.names = TRUE)

t1
Top 10 longest solitary placements nationwide, Sept. 2018-Sept. 2023
country_of_citizenship facility length_of_stay mental_illness
1 EL SALVADOR TACOMA ICE PROCESSING CENTER (NORTHWEST DET CTR) (WA) 817+ NA
2 TRINIDAD AND TOBAGO TACOMA ICE PROCESSING CENTER (NORTHWEST DET CTR) (WA) 811+ Serious Mental Illness
3 MEXICO OTAY MESA DETENTION CENTER (SAN DIEGO CDF) (CA) 759 None
4 UZBEKISTAN DENVER CONTRACT DETENTION FACILITY (CO) 727 None
5 HAITI BUFFALO (BATAVIA) SERVICE PROCESSING CENTER (NY) 652 Mental Illness
6 GUATEMALA TACOMA ICE PROCESSING CENTER (NORTHWEST DET CTR) (WA) 637 Mental Illness
7 MEXICO OTAY MESA DETENTION CENTER (SAN DIEGO CDF) (CA) 567 None
8 RUSSIA TACOMA ICE PROCESSING CENTER (NORTHWEST DET CTR) (WA) 559 None
9 MEXICO TACOMA ICE PROCESSING CENTER (NORTHWEST DET CTR) (WA) 550 Mental Illness
10 RUSSIA KROME NORTH SERVICE PROCESSING CENTER (FL) 534+ None

As reported by Physicians for Human Rights/Harvard Law School, Northwest Detention Center ranked 9th out of 125 facilities represented in the SRMS data when ranked by average solitary placement length, this figure does not change when ongoing stays are taken into account; however, note below that the number of reported solitary placements at NWDC was significantly greater than other facilities in the top 10:

t0 <- df %>% 
  group_by(facility) %>% 
  summarize(avg_placement_length = mean(length_of_stay),
            n = n()) %>% 
  arrange(desc(avg_placement_length)) %>%
  head(10) %>%
  dplyr::select(facility, avg_placement_length, n) %>% 
  knitr::kable(caption = "Avg. solitary placement length by facility, Sept. 2018-Sept. 2023",
               row.names = TRUE,
               digits=2)

t0
Avg. solitary placement length by facility, Sept. 2018-Sept. 2023
facility avg_placement_length n
1 TULSA COUNTY JAIL (DAVID L. MOSS JUSTICE CTR) (OK) 184.00 1
2 HENDERSON DETENTION CENTER (NV) 82.00 79
3 YUBA COUNTY JAIL (CA) 68.75 4
4 PIKE COUNTY CORRECTIONAL FACILITY (PA) 68.52 164
5 ORANGE COUNTY JAIL (NY) 68.25 12
6 OKMULGEE COUNTY JAIL (OK) 67.30 10
7 NYE COUNTY DETENTION CENTER, SOUTHERN (PAHRUMP) (NV) 65.75 8
8 IMPERIAL REGIONAL DETENTION FACILITY (CA) 60.62 295
9 TACOMA ICE PROCESSING CENTER (NORTHWEST DET CTR) (WA) 58.95 419
10 VIRGINIA PENINSULA REGIONAL JAIL (VA) 53.83 6

Comparison of solitary placement lengths

In fact, solitary placements reported to SRMS at the Northwest ICE Process Center were significantly longer on average than nationally, with an average stay length of 58.95 days compared to a national average of 27.03 days, and a median stay length of 27 days compared to a national median of 15 days:

df_nwdc <- df %>% 
  filter(facility == nwdc) %>% 
  mutate(group = "NWDC")

df_all <- df %>% 
  mutate(group = "Nat'l")

dat <- rbind(df_nwdc, df_all) %>% 
  mutate(group = as.factor(group))

b1 <- dat %>% 
  group_by(group) %>% 
  ggplot(aes(x = group, y = log(length_of_stay), fill=group)) +
  geom_boxplot() +
  scale_fill_viridis_d(direction=-1, breaks=rev, alpha=.5) +
  scale_x_discrete() +
  theme_minimal() +
  labs(title = "Distribution of solitary placement lengths",
       subtitle = "National versus NWDC",
       caption="Figure: UWCHR analysis\nData source: ICE SRMS data released via FOIA to PHR/HLS")

b1

A one-sample t-test shows that this difference in mean stay lengths is statistically significant:

# Data are approximately normally distributed; we expect banding around stay lengths associated with ICE reporting requirements i.e. 3, 14, 30 days
# p1 <- qqnorm(log(df$length_of_stay))
# p2 <- qqnorm(log(df[df$facility == nwdc,]$length_of_stay))

dat <- log(df[df$facility == nwdc,]$length_of_stay)
mu <- mean(log(df$length_of_stay))

t.test(x=dat, mu=mu, alternative = "greater")
## 
##  One Sample t-test
## 
## data:  dat
## t = 11.437, df = 418, p-value < 2.2e-16
## alternative hypothesis: true mean is greater than 2.672477
## 95 percent confidence interval:
##  3.245247      Inf
## sample estimates:
## mean of x 
##  3.341706

Special vulnerabilities

SRMS placements are flagged for “special vulnerability” categories, noted in the dataset as “Mental Illness”, “Serious Medical Illness”, “Serious Disability”, and “Suicide Risk?”. The following charts and plots summarize the proportion of SRMS records flagged for these categories at NWDC versus nationally; as well as distributions of placement length for each category at NWDC versus nationally. We note that in almost all categories, solitary placements at NWDC tend to be longer than for the same categories nationwide.

Mental illness

df$mental_illness <- factor(df$mental_illness, levels = c("Serious Mental Illness", "Mental Illness", "None", NA))

mi_nwdc <- df %>% 
  filter(facility == nwdc) %>% 
  count(mental_illness) %>% 
  mutate(prop = n / sum(n)) %>% 
  ungroup()

mi_natl <- df %>%
  count(mental_illness) %>% 
  mutate(prop = n / sum(n)) %>% 
  ungroup()

dat <- left_join(mi_natl, mi_nwdc, by='mental_illness', suffix=c(".natl", ".nwdc")) %>% 
  pivot_longer(cols=-c('mental_illness')) %>%
  filter(str_detect(name, "prop"))

p1 <- dat %>% 
  ggplot(aes(x = mental_illness, y = value, fill=name), color = 'black') +
    geom_col(position='dodge') +
  scale_y_continuous(labels = scales::percent) +
  scale_fill_viridis_d(limits=rev) +
  labs(title = "Proportion of solitary placements",
       subtitle = "NWDC vs. national",
       caption="Figure: UWCHR analysis\nData source: ICE SRMS data released via FOIA to PHR/HLS") +
  theme_minimal()

p1

df_nwdc <- df %>% 
  filter(facility == nwdc) %>% 
  mutate(group = "NWDC")

df_all <- df %>% 
  mutate(group = "National")

dat <- rbind(df_nwdc, df_all) %>% 
  mutate(group = as.factor(group),
         mental_illness = case_when(mental_illness %in% c("None", NA) ~ "None/NA",
         TRUE ~ mental_illness),
         mental_illness = factor(mental_illness, levels = c("Serious Mental Illness", "Mental Illness", "None/NA")))

b1 <- dat %>% 
  ggplot(aes(x = group, y = log(length_of_stay), fill = mental_illness)) +
  geom_boxplot() +
  scale_fill_viridis_d(alpha=.5) +
  theme_minimal() +
  labs(title = "Distribution of solitary placement lengths",
       subtitle = "National versus NWDC",
       caption="Figure: UWCHR analysis\nData source: ICE SRMS data released via FOIA to PHR/HLS")

b1

dat %>% 
  group_by(group, mental_illness) %>% 
  summarize("average placement length" = mean(length_of_stay, na.rm=TRUE),
            "median placement length" = median(length_of_stay, na.rm=TRUE),
            "total placements" = n()) %>% 
  pivot_wider(names_from = c('group'), names_glue = "{group} {.value}", values_from = c('average placement length', 'median placement length', "total placements")) %>% 
  mutate("NWDC % longer than national average" = (`NWDC average placement length` / `National average placement length` - 1) * 100,
         "NWDC % longer than national median" = (`NWDC median placement length` / `National median placement length` - 1) * 100
         ) %>% 
  # dplyr::select(-c("NWDC % longer than national average",
  #                   "NWDC % longer than national median")) %>% 
  knitr::kable(digits=2)
mental_illness National average placement length NWDC average placement length National median placement length NWDC median placement length National total placements NWDC total placements NWDC % longer than national average NWDC % longer than national median
Serious Mental Illness 34.45 89.52 13 28 642 23 159.87 115.38
Mental Illness 23.38 66.42 10 29 2876 90 184.10 190.00
None/NA 27.56 54.45 17 25 10733 306 97.55 47.06 
p1 <- df %>% 
  filter(facility == nwdc,
         placement_date >= "2018-09-01") %>% 
  mutate(qtr = as.factor(quarter(placement_date, fiscal_start=10, type="year.quarter"))) %>% 
  count(qtr, mental_illness) %>% 
  ggplot(aes(x = qtr, y = n, fill = mental_illness)) +
  geom_col(position='stack') +
  labs(title="NWDC solitary placements, Sept. 2018-Sept. 2023",
       subtitle="`Mental Illness` categories",
       caption="Figure: UWCHR analysis\nData source: ICE SRMS data released via FOIA to PHR/HLS") +
  ylab("Total Placements") +
  xlab("Placement date (FY quarter)") +
  scale_fill_viridis_d(na.value='grey') +
  theme_minimal() +
  theme(axis.text.x=element_text(angle=45,hjust=1))
  
p1

Serious medical illness

df_nwdc <- df %>% 
  filter(facility == nwdc) %>% 
  mutate(group = "NWDC")

df_all <- df %>% 
  mutate(group = "Nat'l")

dat <- rbind(df_nwdc, df_all) %>% 
  mutate(group = as.factor(group))

# dat %>% 
#   count(group, serious_medical_illness)

b1 <- dat %>% 
  ggplot(aes(x = group, y = log(length_of_stay), fill = serious_medical_illness) )+
  geom_boxplot() +
  scale_fill_viridis_d(alpha=.5) +
  theme_minimal() +
  labs(title = "Distribution of solitary placement lengths",
       subtitle = "National versus NWDC",
       caption="Figure: UWCHR analysis\nData source: ICE SRMS data released via FOIA to PHR/HLS")

b1

Serious disability

df_nwdc <- df %>% 
  filter(facility == nwdc) %>% 
  mutate(group = "NWDC")

df_all <- df %>% 
  mutate(group = "Nat'l")

dat <- rbind(df_nwdc, df_all) %>% 
  mutate(group = as.factor(group))

# dat %>% 
#   count(group, serious_disability)

b1 <- dat %>% 
  ggplot(aes(x = group, y = log(length_of_stay), fill = serious_disability) )+
  geom_boxplot() +
  scale_fill_viridis_d(alpha=.5) +
  theme_minimal() +
  labs(title = "Distribution of solitary placement lengths",
       subtitle = "National versus NWDC",
       caption="Figure: UWCHR analysis\nData source: ICE SRMS data released via FOIA to PHR/HLS")

b1

Detainee-requested versus Facility-initiated solitary placements

Note larger proportion of “detainee-requested” solitary placements reported at NWDC compared to national; similar distribution of placement lengths between these categories comparing NWDC to national stay length.

dr_nwdc <- df %>% 
  filter(facility == nwdc) %>% 
  count(detainee_request_segregation) %>% 
  mutate(prop = n / sum(n)) %>% 
  ungroup()

dr_natl <- df %>%
  count(detainee_request_segregation) %>% 
  mutate(prop = n / sum(n)) %>% 
  ungroup()

dat <- left_join(dr_natl, dr_nwdc, by='detainee_request_segregation', suffix=c(".natl", ".nwdc")) %>% 
  pivot_longer(cols=-c('detainee_request_segregation')) %>%
  filter(str_detect(name, "prop"))

p1 <- dat %>% 
  ggplot(aes(x = detainee_request_segregation, y = value, fill=name), color = 'black') +
    geom_col(position='dodge') +
  scale_y_continuous(labels = scales::percent) +
  scale_fill_viridis_d(limits=rev) +
  labs(title = "Proportion of solitary placements",
       subtitle = "NWDC vs. national",
       caption="Figure: UWCHR analysis\nData source: ICE SRMS data released via FOIA to PHR/HLS") +
  theme_minimal()

p1

df_nwdc <- df %>% 
  filter(facility == nwdc) %>% 
  mutate(group = "NWDC")

df_all <- df %>% 
  mutate(group = "Nat'l")

dat <- rbind(df_nwdc, df_all) %>% 
  mutate(group = as.factor(group))

b1 <- dat %>% 
  ggplot(aes(x = group, y = log(length_of_stay), fill = detainee_request_segregation)) +
  geom_boxplot() +
  scale_fill_viridis_d(alpha=.5) +
  theme_minimal() +
  labs(title = "Distribution of solitary placement lengths",
       subtitle = "National versus NWDC",
       caption="Figure: UWCHR analysis\nData source: ICE SRMS data released via FOIA to PHR/HLS")

b1

Placement reason

Note no placements categorized as “Hunger Strike”, “Mental Illness” (some “Medical/Mental”), or “Suicide Risk” at NWDC. Note several of these categories are not mutually-exclusive.

df <- df %>% 
  mutate(placement_reason_short = case_when(
    str_detect(placement_reason, "Detainee Request: Other") ~ "Detainee Request",
    str_detect(placement_reason, "Disciplinary") ~ "Disciplinary",
    str_detect(placement_reason, "Facility Initiated: Other") ~ "Facility Initiated",
    str_detect(placement_reason, "Facility Security Threat") ~ "Facility Security Threat",
    str_detect(placement_reason, "Facility Security Threat: Due to Seriousness of Criminal Conviction") ~ "Facility Security Threat",
    str_detect(placement_reason, "Facility Security Threat: Gang Member Status (Not Protective Custody)") ~ "Facility Security Threat",
    str_detect(placement_reason, "Facility Security Threat: Other") ~ "Facility Security Threat",
    str_detect(placement_reason, "Facility Security Threat: Violent or Disruptive Behavior") ~ "Facility Security Threat",
    str_detect(placement_reason, "Hunger Strike") ~ "Hunger Strike",
    str_detect(placement_reason, "Medical/Mental") ~ "Medical/Mental",
    str_detect(placement_reason, "Medical: Detox/Withdrawal Observation") ~ "Medical",
    str_detect(placement_reason, "Medical: Disabled or Infirm") ~ "Medical",
    str_detect(placement_reason, "Medical: Observation") ~ "Medical",
    str_detect(placement_reason, "Medical: Other Infectious Disease") ~ "Medical",
    str_detect(placement_reason, "Medical: Other") ~ "Medical",
    str_detect(placement_reason, "Medical: Segregation Unit") ~ "Medical",
    str_detect(placement_reason, "Medical: Tuberculosis") ~ "Medical",
    str_detect(placement_reason, "Mental Illness") ~ "Mental Illness",
    str_detect(placement_reason, "Mental Illness: Observation") ~ "Mental Illness",
    str_detect(placement_reason, "Other: Pending General Population Bedspace Availability") ~ "Other",
    str_detect(placement_reason, "Other: Unique Reason Not Listed (Explain in Comments)") ~ "Other",
    str_detect(placement_reason, "Pending Investigation of Disciplinary Violation") ~ "Pending Disciplinary",
    str_detect(placement_reason, "Protective Custody") ~ "Protective Custody",
    str_detect(placement_reason, "Protective Custody: Criminal Offense (i.e. Sex Offender)") ~ "Protective Custody",
    str_detect(placement_reason, "Protective Custody: Gang Status (Protective Custody Only)") ~ "Protective Custody",
    str_detect(placement_reason, "Protective Custody: Lesbian, Gay, Bisexual, Transgender (LGBT)") ~ "Protective Custody",
    str_detect(placement_reason, "Protective Custody: Other Detainee Safety") ~ "Protective Custody",
    str_detect(placement_reason, "Protective Custody: Special Vulnerability Other") ~ "Protective Custody",
    str_detect(placement_reason, "Protective Custody: Victim of Sexual Assault") ~ "Protective Custody",
    str_detect(placement_reason, "Suicide Risk Placement") ~ "Suicide Risk",
    str_detect(placement_reason, "Other") ~ "Other",
  ))

df_nwdc <- df %>% 
  filter(facility == nwdc) %>% 
  mutate(group = "NWDC")

df_all <- df %>% 
  mutate(group = "Nat'l")

dat <- rbind(df_nwdc, df_all)

b1 <- dat %>% 
  ggplot(aes(y = placement_reason_short, x = log(length_of_stay), fill = group)) +
  geom_boxplot() +
  scale_fill_viridis_d(limits=rev, alpha=.5) +
  theme_minimal() +
  labs(title = "Distribution of solitary placement lengths",
       subtitle = "National versus NWDC",
       caption="Figure: UWCHR analysis\nData source: ICE SRMS data released via FOIA to PHR/HLS")

b1

dr_nwdc <- df %>% 
  filter(facility == nwdc) %>% 
  count(placement_reason_short) %>% 
  mutate(prop = n / sum(n)) %>% 
  ungroup()

dr_natl <- df %>%
  count(placement_reason_short) %>% 
  mutate(prop = n / sum(n)) %>% 
  ungroup()

dat <- left_join(dr_natl, dr_nwdc, by='placement_reason_short', suffix=c(".natl", ".nwdc")) %>% 
  pivot_longer(cols=-c('placement_reason_short')) %>%
  filter(str_detect(name, "prop"))

p1 <- dat %>% 
  ggplot(aes(y = placement_reason_short, x = value, fill=name), color = 'black') +
    geom_col(position='dodge') +
  scale_x_continuous(labels = scales::percent) +
  scale_fill_viridis_d(limits=rev) +
  labs(title = "Proportion of solitary placements",
       subtitle = "NWDC vs. national",
       caption="Figure: UWCHR analysis\nData source: ICE SRMS data released via FOIA to PHR/HLS") +
  theme_minimal()

p1

Interactive charts

The following interactive charts offer an overview of detailed “Placement Reason” values over time at NWDC and nationally:

NWDC solitary placements by “Placement Reason”

p1 <- df %>% 
  filter(facility == nwdc,
         placement_date >= "2018-09-01") %>% 
  mutate(qtr = as.factor(quarter(placement_date, fiscal_start=10, type="year.quarter")),
         placement_reason = str_wrap(placement_reason, width=25)) %>% 
  count(qtr, placement_reason) %>% 
  ggplot(aes(x = qtr, y = n, fill = placement_reason)) +
  geom_col(position='stack') +
  labs(title="NWDC solitary placements, Sept. 2018-Sept. 2023",
       subtitle="`Placement Reason` categories",
       caption="Data source: ICE SRMS data released via FOIA to PHR/HLS\nFigure: UWCHR analysis") +
  ylab("Placements") +
  xlab("Placement date (FY quarter)") +
  scale_fill_viridis_d(na.value='grey') +
  theme_minimal() +
  theme(axis.text.x=element_text(angle=45,hjust=1))
  
ggplotly(p1)

National solitary placements by “Placement Reason”

p1 <- df %>% 
  filter(placement_date >= "2018-09-01") %>% 
  mutate(qtr = as.factor(quarter(placement_date, fiscal_start=10, type="year.quarter")),
         placement_reason = str_wrap(placement_reason, width=25)) %>% 
  count(qtr, placement_reason) %>% 
  ggplot(aes(x = qtr, y = n, fill = placement_reason)) +
  geom_col(position='stack') +
  labs(title="National ICE solitary placements, Sept. 2018-Sept. 2023",
       subtitle="`Placement Reason` categories",
       caption="Data source: ICE SRMS data released via FOIA to PHR/HLS\nFigure: UWCHR analysis") +
  ylab("Placements") +
  xlab("Placement date (FY quarter)") +
  scale_fill_viridis_d(na.value='grey') +
  theme_minimal() +
  theme(axis.text.x=element_text(angle=45,hjust=1))
  
ggplotly(p1)