Plots and tables with reskit

The aims of this vignette are to:

  • illustrate usage patterns of some of the functions in {reskit}
  • provide a set of test/example outputs of tables and charts derived from NHP demand model results data.

The latter may be visually compared to / validated against visualisations currently produced by the NHP Outputs app.

Firstly we load the functions in the {reskit} package.

Code 
# library(reskit)
purrr::walk(dir(here::here("R"), full.names = TRUE), source)
seed <- 1871

Then we create a fake results object which consists of synthetic data created by some internal reskit functions. It ought to go without saying that the numbers in these synthetic tables are entirely fictitious and unrealistic. They exist purely to provide a test basis for the data processing and visualisation functions in this package.

This is so we don’t have to go and read in actual data for the sake of this vignette.

In real life however you would do something a bit like the steps in the next little section…

Example using azkit and reskit to download results

Here is an example of some code that you might use to access results data from Azure storage. (We’re not running it here, it’s just provided as an example.)

token <- azkit::get_auth_token()
results_container <- azkit::get_container(
  Sys.getenv("AZ_RESULTS_CONTAINER"),
  token = token
)

# use various filters to get the right dataset (PartitionKey), model version and
# scenario from a runs lookup table
results_path <- azkit::read_azure_table(
  Sys.getenv("AZ_RUNS_TABLE"),
  token = token,
  filter = "PartitionKey eq 'national' and app_version eq 'v5.1'",
  select = "aggregated_results_path"
) |>
  dplyr::pull("aggregated_results_path") |>
  grepv(pattern = "NDG2\\-zero") |>
  dplyr::first() # ensure we only get 1 path!

results <- read_results_parquet_files(results_container, results_path)

Back to our synthetic results data

OK so back in vignette world (worst theme park ever), let’s use reskit’s experimental create_demo* functions to create some fake results data.

Code 
default_tbl <- create_demo_default_tbl(seed)
tretspef_losgroup_tbl <- create_demo_tretspef_losgroup_tbl(seed)
sex_agegroup_tbl <- create_demo_sex_agegroup_tbl(seed)
sex_tretspef_tbl <- create_demo_sex_tretspef_tbl(seed)
stepcounts_tbl <- create_demo_stepcounts_tbl(seed)

# fmt: skip
tbl_names <- c(
  "default", "tretspef+los_group", "sex+age_group",
  "sex+tretspef_grouped", "step_counts"
)

results <- list(
  default_tbl,
  tretspef_losgroup_tbl,
  sex_agegroup_tbl,
  sex_tretspef_tbl,
  stepcounts_tbl
) |>
  rlang::set_names(tbl_names)

We now have a real results object (an R list) to work with for the remainder of the vignette.

Moving on to the outputs (tables and charts)

Having shown one way of accessing the results parquet files, this vignette now runs through the suggested pipelines of functions that generate the NHP Outputs app tables and charts, which usually comprise in their simplest form:

  • a data preparation function which creates a data frame ready to be used by:
  • a visualisation function that creates a table or chart.

Some data preparation functions have sub-functions that may occsaionally be useful; there are also export* functions designed to create csv versions of prepared results data (in the same form as the data used by the visualisation function, but with all sites, measures, activity types etc included.)

Outputs: tables

“Main summary table” - overall change by Point of Delivery

This uses the “default” table (from default.parquet), passing it to the compile_principal_pod_data function which prepares the data for the table.

The only relevant filter here is for site selection.

Then the make_principal_pod_table function generates the table.

Code 
results |>
  compile_principal_pod_data() |>
  make_principal_pod_table()
Point of Delivery Baseline Principal Change Percent Change
Inpatient Admissions
Non-Elective Admission 109,000
   77,166
   -31,834
   -29%
Regular Day Attender Admission 109,000
   86,141
   -22,859
   -21%
Regular Night Attender Admission 100,000
   89,678
   -10,322
   -10%
Elective Admission 69,000
   56,009
   -12,991
   -19%
Maternity Admission 45,000
   41,994
   -3,006
   -7%
Inpatient Bed Days
Regular Day Attender Bed Days 105,000
   109,058
   4,058
   4%
Elective Bed Days 99,000
   99,257
   257
   0%
Non-Elective Bed Days 95,000
   85,483
   -9,517
   -10%
Regular Night Attender Bed Days 79,000
   74,273
   -4,727
   -6%
Maternity Bed Days 49,000
   40,230
   -8,770
   -18%
Outpatient
Follow-up Outpatient Attendance 103,000
   100,458
   -2,542
   -2%
Outpatient Procedure 68,000
   51,664
   -16,336
   -24%
First Outpatient Attendance 66,000
   63,935
   -2,065
   -3%
First Outpatient Tele-attendance 66,000
   57,270
   -8,730
   -13%
Follow-up Outpatient Tele-attendance 50,000
   47,624
   -2,376
   -5%
A&E
A&E Arrivals 897,000
   729,576
   -167,424
   -19%

Now the same thing again but testing the site selection feature. We should get a similar style table but with smaller numbers as we have filtered down to just one site.

Code 
results |>
  compile_principal_pod_data(sites = "site1") |>
  make_principal_pod_table()
Point of Delivery Baseline Principal Change Percent Change
Inpatient Admissions
Non-Elective Admission 78,000
   55,220
   -22,780
   -29%
Regular Day Attender Admission 75,000
   59,271
   -15,729
   -21%
Regular Night Attender Admission 66,000
   59,187
   -6,813
   -10%
Elective Admission 39,000
   31,658
   -7,342
   -19%
Maternity Admission 8,000
   7,466
   -534
   -7%
Inpatient Bed Days
Elective Bed Days 82,000
   82,213
   213
   0%
Non-Elective Bed Days 67,000
   60,288
   -6,712
   -10%
Regular Night Attender Bed Days 67,000
   62,991
   -4,009
   -6%
Regular Day Attender Bed Days 64,000
   66,474
   2,474
   4%
Maternity Bed Days 33,000
   27,093
   -5,907
   -18%
Outpatient
Follow-up Outpatient Attendance 64,000
   62,421
   -1,579
   -2%
Outpatient Procedure 54,000
   41,027
   -12,973
   -24%
First Outpatient Tele-attendance 41,000
   35,577
   -5,423
   -13%
First Outpatient Attendance 27,000
   26,155
   -845
   -3%
Follow-up Outpatient Tele-attendance 16,000
   15,240
   -760
   -5%
A&E
A&E Arrivals 583,000
   472,298
   -110,702
   -19%

The principal LoS table

This uses data from the “tretspef+los_group” results table.

With measure set to “beddays”:

Code 
results |>
  compile_principal_los_data(measure = "beddays") |>
  make_principal_los_table()
Length of Stay Baseline Principal Change Percent Change
Regular Day Attender Bed Days
0 days 100,300
   83,226
   -17,074
   -17%
1 day 79,300
   65,754
   -13,546
   -17%
2 days 80,800
   67,261
   -13,539
   -17%
3 days 90,100
   75,723
   -14,377
   -16%
4-7 days 101,400
   85,436
   -15,964
   -16%
8-14 days 95,900
   80,534
   -15,366
   -16%
15-21 days 97,800
   82,951
   -14,849
   -15%
22+ days 90,300
   76,552
   -13,748
   -15%
Elective Bed Days
0 days 97,500
   81,124
   -16,376
   -17%
1 day 100,400
   83,780
   -16,620
   -17%
2 days 87,400
   72,162
   -15,238
   -17%
3 days 78,300
   66,917
   -11,383
   -15%
4-7 days 104,700
   85,246
   -19,454
   -19%
8-14 days 92,300
   76,541
   -15,759
   -17%
15-21 days 90,200
   73,555
   -16,645
   -18%
22+ days 81,100
   67,897
   -13,203
   -16%
Non-Elective Bed Days
0 days 94,100
   78,204
   -15,896
   -17%
1 day 95,500
   79,374
   -16,126
   -17%
2 days 82,200
   71,609
   -10,591
   -13%
3 days 93,300
   82,960
   -10,340
   -11%
4-7 days 81,500
   67,480
   -14,020
   -17%
8-14 days 97,100
   84,537
   -12,563
   -13%
15-21 days 91,600
   77,736
   -13,864
   -15%
22+ days 85,300
   73,484
   -11,816
   -14%
Maternity Bed Days
0 days 86,200
   70,993
   -15,207
   -18%
1 day 83,200
   66,968
   -16,232
   -20%
2 days 80,900
   66,755
   -14,145
   -17%
3 days 81,500
   68,332
   -13,168
   -16%
4-7 days 92,300
   76,311
   -15,989
   -17%
8-14 days 84,600
   72,475
   -12,125
   -14%
15-21 days 104,400
   88,441
   -15,959
   -15%
22+ days 95,800
   81,114
   -14,686
   -15%
Regular Night Attender Bed Days
0 days 82,500
   66,331
   -16,169
   -20%
1 day 89,000
   76,440
   -12,560
   -14%
2 days 89,900
   74,860
   -15,040
   -17%
3 days 96,100
   81,177
   -14,923
   -16%
4-7 days 88,700
   76,878
   -11,822
   -13%
8-14 days 86,600
   72,239
   -14,361
   -17%
15-21 days 88,900
   72,293
   -16,607
   -19%
22+ days 80,900
   69,209
   -11,691
   -14%

With measure set to “admissions”:

Code 
results |>
  compile_principal_los_data(measure = "admissions", sites = "site2") |>
  make_principal_los_table()
Length of Stay Baseline Principal Change Percent Change
Elective Admission
0 days 22,200
   18,731
   -3,469
   -16%
1 day 27,200
   22,651
   -4,549
   -17%
2 days 28,800
   24,063
   -4,737
   -16%
3 days 27,900
   23,977
   -3,923
   -14%
4-7 days 27,900
   23,658
   -4,242
   -15%
8-14 days 28,700
   24,375
   -4,325
   -15%
15-21 days 30,000
   25,127
   -4,873
   -16%
22+ days 27,600
   23,685
   -3,915
   -14%
Non-Elective Admission
0 days 25,100
   21,905
   -3,195
   -13%
1 day 26,200
   22,915
   -3,285
   -13%
2 days 26,700
   22,474
   -4,226
   -16%
3 days 26,400
   22,887
   -3,513
   -13%
4-7 days 28,000
   24,231
   -3,769
   -13%
8-14 days 28,400
   24,125
   -4,275
   -15%
15-21 days 25,500
   22,683
   -2,817
   -11%
22+ days 21,600
   17,621
   -3,979
   -18%
Regular Day Attender Admission
0 days 22,900
   19,884
   -3,016
   -13%
1 day 30,800
   25,951
   -4,849
   -16%
2 days 26,900
   23,071
   -3,829
   -14%
3 days 23,700
   20,617
   -3,083
   -13%
4-7 days 25,100
   21,269
   -3,831
   -15%
8-14 days 24,700
   20,532
   -4,168
   -17%
15-21 days 26,800
   23,333
   -3,467
   -13%
22+ days 24,000
   20,301
   -3,699
   -15%
Regular Night Attender Admission
0 days 25,000
   22,184
   -2,816
   -11%
1 day 27,300
   23,134
   -4,166
   -15%
2 days 24,100
   20,623
   -3,477
   -14%
3 days 26,700
   22,092
   -4,608
   -17%
4-7 days 25,100
   20,825
   -4,275
   -17%
8-14 days 28,900
   24,373
   -4,527
   -16%
15-21 days 23,800
   20,489
   -3,311
   -14%
22+ days 21,500
   18,268
   -3,232
   -15%
Maternity Admission
0 days 22,600
   19,400
   -3,200
   -14%
1 day 26,000
   21,924
   -4,076
   -16%
2 days 25,300
   21,385
   -3,915
   -15%
3 days 25,500
   21,903
   -3,597
   -14%
4-7 days 25,300
   21,904
   -3,396
   -13%
8-14 days 19,800
   16,717
   -3,083
   -16%
15-21 days 22,200
   18,954
   -3,246
   -15%
22+ days 22,200
   18,744
   -3,456
   -16%

Outputs: charts

Impact of changes charts

For these we need to use carefully chosen combinations of measure, activity type, and pod.

Overall (“waterfall”) chart with baseline

As this is for admissions, “efficiencies” should not be shown on the y-axis.

Code 
results |>
  compile_grouped_impact_data(
    measure = "admissions",
    activity_type = "ip",
    pods = c(
      "ip_elective_admission",
      "ip_maternity_admission",
      "ip_non-elective_admission"
    ),
    include_baseline = TRUE
  ) |>
  make_grouped_impact_plot()

Code 
results |>
  compile_grouped_impact_data(
    measure = "beddays",
    activity_type = "ip",
    pods = c(
      "ip_elective_admission",
      "ip_maternity_admission",
      "ip_non-elective_admission"
    ),
    sites = "site1",
    include_baseline = TRUE
  ) |>
  make_grouped_impact_plot()

Let’s look at an outpatients chart…

Code 
results |>
  compile_grouped_impact_data(
    measure = "attendances",
    activity_type = "op",
    pods = c("op_first", "op_follow-up")
  ) |>
  make_grouped_impact_plot()

…and one without the baseline:

Code 
results |>
  compile_grouped_impact_data(
    measure = "attendances",
    activity_type = "op",
    pods = c("op_first", "op_follow-up"),
    include_baseline = FALSE
  ) |>
  make_grouped_impact_plot()

Individual change factor (TPMA) impact charts

And a couple of individual change factor charts:

Code 
results |>
  compile_tpma_impact_data(
    measure = "attendances",
    activity_type = "op",
    pods = NULL
  ) |>
  make_tpma_impact_plot()

Code 
results |>
  compile_tpma_impact_data(
    measure = "arrivals",
    activity_type = "aae",
    pods = NULL,
    sites = "site2"
  ) |>
  make_tpma_impact_plot()

Activity in detail charts

These are even more tricky to specify as they require several arguments to be provided.

A chart by sex and age_group

Code 
results |>
  compile_detailed_activity_data(
    measure = "walk-in",
    activity_type = "aae",
    aggregation = "age_group"
  ) |>
  make_detailed_activity_table(final_year = "2029/30")
Age Group Baseline Final (2029/30) Change Percent Change
Female
0 199,400
   178,963
   -20,437
   -10%
1-4 229,500
   207,405
   -22,095
   -10%
5-9 156,200
   128,376
   -27,824
   -18%
10-15 142,000
   123,529
   -18,471
   -13%
16-17 180,900
   157,305
   -23,595
   -13%
18-34 199,200
   187,863
   -11,337
   -6%
35-49 191,400
   186,426
   -4,974
   -3%
50-64 188,900
   160,225
   -28,675
   -15%
65-74 199,700
   174,407
   -25,293
   -13%
75-84 205,100
   177,127
   -27,973
   -14%
85+ 230,100
   216,937
   -13,163
   -6%
Male
0 220,100
   180,740
   -39,360
   -18%
1-4 161,500
   148,783
   -12,717
   -8%
5-9 167,100
   155,750
   -11,350
   -7%
10-15 202,100
   195,830
   -6,270
   -3%
16-17 209,800
   177,665
   -32,135
   -15%
18-34 277,800
   236,942
   -40,858
   -15%
35-49 182,200
   168,698
   -13,502
   -7%
50-64 203,200
   167,055
   -36,145
   -18%
65-74 254,400
   228,743
   -25,657
   -10%
75-84 216,600
   195,876
   -20,724
   -10%
85+ 162,500
   151,378
   -11,122
   -7%

A chart by sex and treatment speciality

Code 
results |>
  compile_detailed_activity_data(
    measure = "beddays",
    activity_type = "ip",
    aggregation = "tretspef_grouped"
  ) |>
  make_detailed_activity_table(final_year = "2029/30")
Treatment Specialty Baseline Final (2029/30) Change Percent Change
Female
100: General Surgery 22,700
   17,854
   -4,846
   -21%
101: Urology 21,700
   18,076
   -3,624
   -17%
110: Trauma and Orthopaedic 15,200
   12,041
   -3,159
   -21%
120: Ear Nose and Throat 23,200
   18,538
   -4,662
   -20%
130: Ophthalmology 17,300
   13,198
   -4,102
   -24%
140: Oral Surgery 21,300
   16,352
   -4,948
   -23%
150: Neurosurgical 22,400
   17,406
   -4,994
   -22%
160: Plastic Surgery 17,500
   13,467
   -4,033
   -23%
170: Cardiothoracic Surgery 21,900
   17,551
   -4,349
   -20%
300: General Internal Medicine 17,200
   15,210
   -1,990
   -12%
301: Gastroenterology 13,400
   10,607
   -2,793
   -21%
320: Cardiology 23,000
   18,037
   -4,963
   -22%
330: Dermatology 21,400
   16,440
   -4,960
   -23%
340: Respiratory Medicine 19,400
   14,869
   -4,531
   -23%
400: Neurology 17,900
   13,995
   -3,905
   -22%
410: Rheumatology 15,500
   12,451
   -3,049
   -20%
430: Elderly Medicine 11,600
   9,084
   -2,516
   -22%
Other 19,000
   15,854
   -3,146
   -17%
Male
100: General Surgery 20,800
   16,329
   -4,471
   -21%
101: Urology 15,100
   12,398
   -2,702
   -18%
110: Trauma and Orthopaedic 22,400
   18,312
   -4,088
   -18%
120: Ear Nose and Throat 14,900
   12,064
   -2,836
   -19%
130: Ophthalmology 12,500
   9,739
   -2,761
   -22%
140: Oral Surgery 13,200
   9,742
   -3,458
   -26%
150: Neurosurgical 18,100
   13,807
   -4,293
   -24%
160: Plastic Surgery 14,100
   11,313
   -2,787
   -20%
170: Cardiothoracic Surgery 12,700
   10,324
   -2,376
   -19%
300: General Internal Medicine 15,400
   12,518
   -2,882
   -19%
301: Gastroenterology 20,500
   15,665
   -4,835
   -24%
320: Cardiology 17,000
   13,348
   -3,652
   -21%
330: Dermatology 21,500
   16,838
   -4,662
   -22%
340: Respiratory Medicine 20,200
   16,334
   -3,866
   -19%
400: Neurology 17,600
   13,145
   -4,455
   -25%
410: Rheumatology 18,100
   14,124
   -3,976
   -22%
430: Elderly Medicine 24,600
   19,470
   -5,130
   -21%
Other 13,000
   10,423
   -2,577
   -20%

Distribution tables and charts

These show the distribution across all model runs rather than just the calculated mean or principal values. For the synthetic results data we are using here, there are only 64 runs - the real model would usually use 512 or 256 runs.

Create a summary table:

Code 
results |>
  compile_distribution_summary_data("median") |> # median is the default value
  make_distribution_summary_table()
Measure Baseline Median Change Percent
Change
80% prediction interval
Lower Upper
Inpatient Non-Elective Admission
Admissions 109,000 76,488 −32,512 −30% 68,358 87,368
Bed days 95,000 84,731 −10,269 −11% 75,726 96,784
Inpatient Regular Day Attender Admission
Admissions 109,000 85,382 −23,618 −22% 76,309 97,529
Bed days 105,000 108,098 3,098 3% 96,611 123,476
Inpatient Regular Night Attender Admission
Admissions 100,000 88,888 −11,112 −11% 79,442 101,533
Bed days 79,000 73,619 −5,381 −7% 65,795 84,091
Inpatient Elective Admission
Admissions 69,000 55,516 −13,484 −20% 49,616 63,414
Bed days 99,000 98,384 −616 −1% 87,928 112,379
Inpatient Maternity Admission
Admissions 45,000 41,624 −3,376 −8% 37,201 47,545
Bed days 49,000 39,876 −9,124 −19% 35,638 45,548
Outpatient Follow-up Outpatient Attendance
Attendances 103,000 99,574 −3,426 −3% 88,993 113,739
Tele-attendances 50,000 47,205 −2,795 −6% 42,188 53,919
Outpatient Outpatient Procedure
Attendances 68,000 51,209 −16,791 −25% 45,767 58,493
Outpatient First Outpatient Attendance
Attendances 66,000 63,372 −2,628 −4% 56,637 72,388
Tele-attendances 66,000 56,766 −9,234 −14% 50,733 64,840
A&E Type 5 Department (SDEC)
Ambulance 122,000 89,518 −32,482 −27% 80,005 102,253
Walk-in 73,000 58,526 −14,474 −20% 52,307 66,852
A&E Type 2 Department
Ambulance 108,000 81,069 −26,931 −25% 72,454 92,602
Walk-in 64,000 57,064 −6,936 −11% 51,000 65,182
A&E Type 4 Department
Ambulance 94,000 68,603 −25,397 −27% 61,313 78,362
Walk-in 110,000 99,174 −10,826 −10% 88,634 113,282
A&E Type 3 Department
Ambulance 92,000 72,060 −19,940 −22% 64,402 82,311
Walk-in 83,000 85,038 2,038 2% 76,000 97,135
A&E Type 1 Department
Ambulance 88,000 68,480 −19,520 −22% 61,203 78,222
Walk-in 63,000 43,621 −19,379 −31% 38,985 49,827

An alternative:

Code 
results |>
  compile_distribution_summary_data("principal") |>
  make_distribution_summary_table()
Measure Baseline Principal Change Percent
Change
80% prediction interval
Lower Upper
Inpatient Non-Elective Admission
Admissions 109,000 77,166 −31,834 −29% 68,358 87,368
Bed days 95,000 85,483 −9,517 −10% 75,726 96,784
Inpatient Regular Day Attender Admission
Admissions 109,000 86,141 −22,859 −21% 76,309 97,529
Bed days 105,000 109,058 4,058 4% 96,611 123,476
Inpatient Regular Night Attender Admission
Admissions 100,000 89,678 −10,322 −10% 79,442 101,533
Bed days 79,000 74,273 −4,727 −6% 65,795 84,091
Inpatient Elective Admission
Admissions 69,000 56,009 −12,991 −19% 49,616 63,414
Bed days 99,000 99,257 257 0% 87,928 112,379
Inpatient Maternity Admission
Admissions 45,000 41,994 −3,006 −7% 37,201 47,545
Bed days 49,000 40,230 −8,770 −18% 35,638 45,548
Outpatient Follow-up Outpatient Attendance
Attendances 103,000 100,458 −2,542 −2% 88,993 113,739
Tele-attendances 50,000 47,624 −2,376 −5% 42,188 53,919
Outpatient Outpatient Procedure
Attendances 68,000 51,664 −16,336 −24% 45,767 58,493
Outpatient First Outpatient Attendance
Attendances 66,000 63,935 −2,065 −3% 56,637 72,388
Tele-attendances 66,000 57,270 −8,730 −13% 50,733 64,840
A&E Type 5 Department (SDEC)
Ambulance 122,000 90,313 −31,687 −26% 80,005 102,253
Walk-in 73,000 59,046 −13,954 −19% 52,307 66,852
A&E Type 2 Department
Ambulance 108,000 81,789 −26,211 −24% 72,454 92,602
Walk-in 64,000 57,571 −6,429 −10% 51,000 65,182
A&E Type 4 Department
Ambulance 94,000 69,212 −24,788 −26% 61,313 78,362
Walk-in 110,000 100,055 −9,945 −9% 88,634 113,282
A&E Type 3 Department
Ambulance 92,000 72,700 −19,300 −21% 64,402 82,311
Walk-in 83,000 85,793 2,793 3% 76,000 97,135
A&E Type 1 Department
Ambulance 88,000 69,088 −18,912 −21% 61,203 78,222
Walk-in 63,000 44,009 −18,991 −30% 38,985 49,827

Create a beeswarm chart:

Code 
results |>
  compile_distribution_plot_data(measure = "walk-in", activity_type = "aae") |>
  make_beeswarm_distrib_plot()

The same beeswarm chart but with the zero point shown:

Code 
results |>
  compile_distribution_plot_data(measure = "walk-in", activity_type = "aae") |>
  make_beeswarm_distrib_plot(show_zero = TRUE)

Create a cumulative distribution (ecdf) chart:

Code 
results |>
  compile_distribution_plot_data(measure = "walk-in", activity_type = "aae") |>
  make_cumulative_distrib_plot()

The same cumulative distribution chart but with the zero point shown:

Code 
results |>
  compile_distribution_plot_data(measure = "walk-in", activity_type = "aae") |>
  make_cumulative_distrib_plot(show_zero = TRUE)

Conclusion

This vignette has included a summary of the main methods to process results data using functions from {reskit}, and to produce tables and charts.

There are many more variations on the charts shown, depending on the selection of activity type, PoD, measure and site. But this vignette shows the main processes you might need to follow - amend as necessary!