Hugging Face's logo

Spaces:
dep-dev
/
dataset_analysis
Running

App Files Community
dataset_analysis
File size: 51,440 Bytes
e4505d4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Clinical NLP Dataset Analysis Report</title>
<script src="https://cdnjs.cloudflare.com/ajax/libs/Chart.js/4.4.1/chart.umd.js"></script>
<style>
:root {
  --bg: #ffffff; --bg2: #f5f4f0; --bg3: #edecea;
  --text: #1a1a18; --text2: #5f5e5a; --text3: #888780;
  --border: rgba(0,0,0,0.12); --border2: rgba(0,0,0,0.08);
  --blue: #185FA5; --blue-bg: #E6F1FB; --blue-light: #B5D4F4;
  --teal: #0F6E56; --teal-bg: #E1F5EE; --teal-light: #9FE1CB;
  --coral: #993C1D; --coral-bg: #FAECE7; --coral-light: #F5C4B3;
  --amber: #854F0B; --amber-bg: #FAEEDA; --amber-light: #FAC775;
  --purple: #534AB7; --purple-bg: #EEEDFE; --purple-light: #CECBF6;
  --green: #3B6D11; --green-bg: #EAF3DE; --green-light: #C0DD97;
  --red: #A32D2D; --red-bg: #FCEBEB; --red-light: #F7C1C1;
  --radius: 12px; --radius-sm: 8px;
}
@media (prefers-color-scheme: dark) {
  :root {
    --bg: #1c1c1a; --bg2: #252523; --bg3: #2d2d2b;
    --text: #e8e6de; --text2: #b4b2a9; --text3: #888780;
    --border: rgba(255,255,255,0.12); --border2: rgba(255,255,255,0.07);
    --blue: #85B7EB; --blue-bg: #042C53; --blue-light: #0C447C;
    --teal: #5DCAA5; --teal-bg: #04342C; --teal-light: #085041;
    --coral: #F0997B; --coral-bg: #4A1B0C; --coral-light: #712B13;
    --amber: #EF9F27; --amber-bg: #412402; --amber-light: #633806;
    --purple: #AFA9EC; --purple-bg: #26215C; --purple-light: #3C3489;
    --green: #97C459; --green-bg: #173404; --green-light: #27500A;
    --red: #F09595; --red-bg: #501313; --red-light: #791F1F;
  }
}
* { box-sizing: border-box; margin: 0; padding: 0; }
body { font-family: -apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif; font-size: 15px; line-height: 1.6; color: var(--text); background: var(--bg3); padding: 2rem 1rem; }
.page { max-width: 960px; margin: 0 auto; }
h1 { font-size: 26px; font-weight: 600; margin-bottom: 0.25rem; }
h2 { font-size: 18px; font-weight: 500; color: var(--text); margin: 2rem 0 1rem; padding-bottom: 0.5rem; border-bottom: 0.5px solid var(--border); }
h3 { font-size: 15px; font-weight: 500; color: var(--text2); margin-bottom: 0.75rem; }
p { color: var(--text2); line-height: 1.7; margin-bottom: 0.75rem; }
.subtitle { font-size: 14px; color: var(--text3); margin-bottom: 2rem; }
.section { background: var(--bg); border-radius: var(--radius); border: 0.5px solid var(--border); padding: 1.5rem; margin-bottom: 1.5rem; }
.grid-2 { display: grid; grid-template-columns: 1fr 1fr; gap: 1rem; }
.grid-3 { display: grid; grid-template-columns: repeat(3, 1fr); gap: 1rem; }
.grid-4 { display: grid; grid-template-columns: repeat(4, 1fr); gap: 0.75rem; }
@media (max-width: 700px) { .grid-2, .grid-3, .grid-4 { grid-template-columns: 1fr; } }
.metric { background: var(--bg2); border-radius: var(--radius-sm); padding: 1rem 1.25rem; }
.metric .label { font-size: 12px; color: var(--text3); margin-bottom: 0.25rem; text-transform: uppercase; letter-spacing: 0.04em; }
.metric .value { font-size: 22px; font-weight: 500; color: var(--text); }
.metric .sub { font-size: 12px; color: var(--text3); margin-top: 0.25rem; }
.badge { display: inline-block; padding: 2px 10px; border-radius: 20px; font-size: 12px; font-weight: 500; }
.badge-blue { background: var(--blue-bg); color: var(--blue); }
.badge-teal { background: var(--teal-bg); color: var(--teal); }
.badge-coral { background: var(--coral-bg); color: var(--coral); }
.badge-amber { background: var(--amber-bg); color: var(--amber); }
.badge-purple { background: var(--purple-bg); color: var(--purple); }
.badge-green { background: var(--green-bg); color: var(--green); }
.badge-red { background: var(--red-bg); color: var(--red); }
.chart-wrap { position: relative; width: 100%; height: 280px; }
.chart-wrap-lg { position: relative; width: 100%; height: 340px; }
.chart-wrap-sm { position: relative; width: 100%; height: 200px; }
.legend { display: flex; flex-wrap: wrap; gap: 12px; margin-bottom: 12px; font-size: 12px; color: var(--text2); }
.legend span { display: flex; align-items: center; gap: 5px; }
.legend .dot { width: 10px; height: 10px; border-radius: 2px; flex-shrink: 0; }
.insight { background: var(--blue-bg); border-left: 3px solid var(--blue); border-radius: 0 var(--radius-sm) var(--radius-sm) 0; padding: 0.75rem 1rem; margin: 0.75rem 0; }
.insight p { color: var(--blue); margin: 0; font-size: 14px; }
.warning { background: var(--amber-bg); border-left: 3px solid var(--amber); border-radius: 0 var(--radius-sm) var(--radius-sm) 0; padding: 0.75rem 1rem; margin: 0.75rem 0; }
.warning p { color: var(--amber); margin: 0; font-size: 14px; }
.success { background: var(--teal-bg); border-left: 3px solid var(--teal); border-radius: 0 var(--radius-sm) var(--radius-sm) 0; padding: 0.75rem 1rem; margin: 0.75rem 0; }
.success p { color: var(--teal); margin: 0; font-size: 14px; }
table { width: 100%; border-collapse: collapse; font-size: 13px; }
th { text-align: left; padding: 6px 10px; font-weight: 500; color: var(--text3); font-size: 11px; text-transform: uppercase; letter-spacing: 0.04em; border-bottom: 0.5px solid var(--border); }
td { padding: 7px 10px; border-bottom: 0.5px solid var(--border2); color: var(--text); }
tr:last-child td { border-bottom: none; }
.ratio-bar { display: flex; height: 6px; border-radius: 3px; overflow: hidden; gap: 1px; margin-top: 4px; }
.ratio-bar .ip { background: #185FA5; }
.ratio-bar .op { background: #0F6E56; }
.section-header { display: flex; align-items: baseline; justify-content: space-between; margin-bottom: 1rem; }
.divider { height: 0.5px; background: var(--border); margin: 1.5rem 0; }
.tag-row { display: flex; flex-wrap: wrap; gap: 6px; margin: 0.5rem 0; }
hr { border: none; border-top: 0.5px solid var(--border); margin: 1.25rem 0; }
</style>
</head>
<body>
<div class="page">

<div class="section" style="background: linear-gradient(135deg, var(--blue-bg) 0%, var(--bg) 60%);">
  <div style="display:flex;align-items:center;gap:12px;margin-bottom:0.5rem;">
    <span class="badge badge-blue">Clinical NLP</span>
    <span class="badge badge-coral">Graph Dataset</span>
    <span class="badge badge-purple">Substance Use Disorder</span>
  </div>
  <h1>Dataset Analysis Report</h1>
  <p class="subtitle">same_graph_test_bothmasked.json &amp; same_graph_train_bothmasked.json β€” Full combined analysis</p>
  <div class="grid-4" style="margin-top:1.25rem;">
    <div class="metric"><div class="label">Total patients</div><div class="value">7,628</div><div class="sub">Test: 2,640 Β· Train: 4,988</div></div>
    <div class="metric"><div class="label">Total edges</div><div class="value">190,092</div><div class="sub">Similarity graph connections</div></div>
    <div class="metric"><div class="label">Total words</div><div class="value">2.80M</div><div class="sub">Across all clinical notes</div></div>
    <div class="metric"><div class="label">Total chars</div><div class="value">19.4M</div><div class="sub">Across all clinical notes</div></div>
  </div>
</div>

<!-- SECTION 1: Corpus Overview -->
<h2>1 Β· Corpus overview &amp; label distribution</h2>
<div class="section">
  <div class="grid-2">
    <div>
      <h3>Overall label split (combined)</h3>
      <div class="legend"><span><span class="dot" style="background:#185FA5"></span>Inpatient (IP) β€” 2,609 (34.2%)</span><span><span class="dot" style="background:#0F6E56"></span>Outpatient (OP) β€” 5,019 (65.8%)</span></div>
      <div class="chart-wrap"><canvas id="labelPie" role="img" aria-label="Pie chart showing IP 2609 vs OP 5019 patients">IP: 2609, OP: 5019</canvas></div>
    </div>
    <div>
      <h3>Split across train / test files</h3>
      <div class="legend"><span><span class="dot" style="background:#185FA5"></span>IP</span><span><span class="dot" style="background:#0F6E56"></span>OP</span></div>
      <div class="chart-wrap"><canvas id="splitBar" role="img" aria-label="Grouped bar chart of IP/OP counts in train and test sets">Train: IP 1676 OP 3312; Test: IP 933 OP 1707</canvas></div>
    </div>
  </div>
  <div class="insight"><p>The dataset is class-imbalanced at roughly 1:2 (IP:OP). This ratio is stable across both train and test splits, suggesting a stratified partition strategy was used.</p></div>
</div>

<!-- SECTION 2: Note Length Statistics -->
<h2>2 Β· Note length β€” tokens, words &amp; sentences</h2>
<div class="section">
  <h3>Combined dataset summary statistics</h3>
  <div class="grid-4" style="margin-bottom:1.25rem;">
    <div class="metric"><div class="label">Avg words / note</div><div class="value">367.2</div><div class="sub">Οƒ = 253.8</div></div>
    <div class="metric"><div class="label">Median words</div><div class="value">309</div><div class="sub">Long right tail</div></div>
    <div class="metric"><div class="label">Avg sentences</div><div class="value">48.3</div><div class="sub">Οƒ = 35.7</div></div>
    <div class="metric"><div class="label">Avg chars / note</div><div class="value">2,545</div><div class="sub">Οƒ = 1,760</div></div>
  </div>
  <div class="grid-2">
    <div>
      <h3>Word count distribution β€” all notes</h3>
      <div class="legend"><span><span class="dot" style="background:#888780"></span>All patients</span></div>
      <div class="chart-wrap-lg"><canvas id="wordDist" role="img" aria-label="Bar chart of word count distribution across all notes">Distribution of word counts in 7628 notes</canvas></div>
    </div>
    <div>
      <h3>Sentence count distribution β€” all notes</h3>
      <div class="legend"><span><span class="dot" style="background:#888780"></span>All patients</span></div>
      <div class="chart-wrap-lg"><canvas id="sentDist" role="img" aria-label="Bar chart of sentence count distribution">Distribution of sentence counts</canvas></div>
    </div>
  </div>
</div>

<!-- SECTION 3 & 4: IP vs OP length -->
<h2>3 &amp; 4 Β· IP vs OP β€” length &amp; severity</h2>
<div class="section">
  <div class="grid-3" style="margin-bottom:1.25rem;">
    <div class="metric" style="border-left:3px solid var(--blue);">
      <div class="label">IP avg words</div><div class="value" style="color:var(--blue)">438.1</div><div class="sub">Οƒ = 318.0 Β· median = 355</div>
    </div>
    <div class="metric" style="border-left:3px solid var(--teal);">
      <div class="label">OP avg words</div><div class="value" style="color:var(--teal)">330.3</div><div class="sub">Οƒ = 203.3 Β· median = 292</div>
    </div>
    <div class="metric" style="border-left:3px solid var(--amber);">
      <div class="label">Length ratio IP:OP</div><div class="value" style="color:var(--amber)">1.33Γ—</div><div class="sub">IP notes are 33% longer on avg</div>
    </div>
  </div>
  <div class="grid-2">
    <div>
      <h3>Word count by label (bucketed)</h3>
      <div class="legend"><span><span class="dot" style="background:#185FA5"></span>IP</span><span><span class="dot" style="background:#0F6E56"></span>OP</span></div>
      <div class="chart-wrap-lg"><canvas id="ipopWords" role="img" aria-label="Grouped bar chart showing word count distribution for IP and OP patients">IP: shorter notes fewer, longer more; OP: more concentrated in medium range</canvas></div>
    </div>
    <div>
      <h3>Severity score distribution</h3>
      <div class="legend"><span><span class="dot" style="background:#185FA5"></span>IP</span><span><span class="dot" style="background:#0F6E56"></span>OP</span></div>
      <div class="chart-wrap-lg"><canvas id="severityDist" role="img" aria-label="Stacked bar chart showing symptom severity scores for IP vs OP patients">IP avg severity 8.03, OP avg 6.27</canvas></div>
    </div>
  </div>
  <hr>
  <div class="grid-2">
    <div>
      <h3>Summary: IP vs OP length &amp; clinical stats</h3>
      <table>
        <thead><tr><th>Metric</th><th>IP (n=2,609)</th><th>OP (n=5,019)</th></tr></thead>
        <tbody>
          <tr><td>Mean words</td><td>438.1</td><td>330.3</td></tr>
          <tr><td>Median words</td><td>355</td><td>292</td></tr>
          <tr><td>Std words</td><td>318.0</td><td>203.3</td></tr>
          <tr><td>Mean sentences</td><td>58.5</td><td>43.0</td></tr>
          <tr><td>Mean chars</td><td>3,053</td><td>2,281</td></tr>
          <tr><td>Max words (note)</td><td>2,609</td><td>2,501</td></tr>
          <tr><td>Avg severity score</td><td>8.03</td><td>6.27</td></tr>
          <tr><td>Score β‰₯ 10 (%)</td><td>33.4%</td><td>19.1%</td></tr>
          <tr><td>Score β‰₯ 15 (%)</td><td>12.0%</td><td>4.4%</td></tr>
          <tr><td>Avg Treatment_decision tokens</td><td>5.42</td><td>3.16</td></tr>
        </tbody>
      </table>
    </div>
    <div>
      <h3>Train vs test file characteristics</h3>
      <table>
        <thead><tr><th>Metric</th><th>Train</th><th>Test</th></tr></thead>
        <tbody>
          <tr><td>Nodes</td><td>4,988</td><td>2,640</td></tr>
          <tr><td>Edges</td><td>120,492</td><td>69,600</td></tr>
          <tr><td>IP %</td><td>33.6%</td><td>35.3%</td></tr>
          <tr><td>OP %</td><td>66.4%</td><td>64.7%</td></tr>
          <tr><td>Mean words</td><td>338.5</td><td>421.3</td></tr>
          <tr><td>Median words</td><td>291</td><td>348</td></tr>
          <tr><td>Mean sentences</td><td>44.1</td><td>56.2</td></tr>
          <tr><td>Mean chars</td><td>2,325</td><td>2,961</td></tr>
        </tbody>
      </table>
      <div class="warning"><p>Test set notes are notably longer on average (421 vs 339 words). This may reflect more complex multi-visit collations, and could affect model behaviour at inference time.</p></div>
    </div>
  </div>
  <div class="insight"><p>IP notes are consistently longer, denser, and higher-severity than OP notes. This is clinically expected: inpatient cases involve more events (admission, detox, stabilisation), multiple clinicians, and higher acuity β€” all reflected in longer collated notes.</p></div>
</div>

<!-- SECTION 5 & 6: IP vs OP patterns -->
<h2>5 &amp; 6 Β· IP vs OP β€” clinical patterns &amp; severity indicators</h2>
<div class="section">
  <h3>Pattern strength ratio (IP% Γ· OP%) β€” features that distinguish inpatient admission</h3>
  <p>Ratio > 1.0 means more prevalent in IP. Features sorted by ratio descending.</p>
  <div class="chart-wrap-lg" style="height:380px;"><canvas id="ratioChart" role="img" aria-label="Horizontal bar chart of IP-to-OP ratio for various clinical features">Features with highest IP/OP ratios: social withdrawal 2.0x, paranoia 1.78x, memory issues 1.67x</canvas></div>
  <div class="insight"><p>Paranoia (1.78Γ—), memory issues (1.67Γ—), hallucinations (1.57Γ—), nausea/vomiting (1.44Γ—), and irritability (1.30Γ—) are the strongest IP-associated symptom patterns. These reflect acute psychiatric and neurological complications requiring inpatient management.</p></div>
</div>

<!-- SECTION 7: Borderline cases -->
<h2>7 Β· Borderline &amp; ambiguous cases</h2>
<div class="section">
  <div class="grid-4" style="margin-bottom:1rem;">
    <div class="metric"><div class="label">Very short IP notes</div><div class="value">50</div><div class="sub">1.9% of IP β€” &lt;50 words</div></div>
    <div class="metric"><div class="label">Very long OP notes</div><div class="value">260</div><div class="sub">5.2% of OP β€” &gt;700 words</div></div>
    <div class="metric"><div class="label">Low-severity IP</div><div class="value">341</div><div class="sub">13.1% β€” score ≀ 2</div></div>
    <div class="metric"><div class="label">High-severity OP</div><div class="value">555</div><div class="sub">11.1% β€” score β‰₯ 12</div></div>
  </div>
  <div class="warning"><p>An estimated 10–13% of notes may constitute "borderline" cases β€” IP notes with minimal clinical documentation or OP notes with complex, high-severity presentations. These represent real-world label ambiguity and will be the hardest cases for any classifier.</p></div>
  <div class="grid-2" style="margin-top:0.75rem;">
    <div>
      <h3>Example short IP note (50 words)</h3>
      <div style="background:var(--bg2);border-radius:var(--radius-sm);padding:0.75rem 1rem;font-size:13px;color:var(--text2);line-height:1.6;font-style:italic;">
        "40-year-old married gentleman presented with history of alcohol and tobacco dependence syndrome for the past 17 years. Presented with withdrawal hallucinosis, multimodal in nature. Pulse rate is 112/min. Coarse tremors present. Plan: Treatment_decision1. Once the patient is stable, discharge..."
      </div>
    </div>
    <div>
      <h3>Severity overlap zone</h3>
      <p style="font-size:13px;">At severity score = 8 (IP median), both classes overlap heavily:</p>
      <table>
        <thead><tr><th>Score threshold</th><th>IP above (%)</th><th>OP above (%)</th></tr></thead>
        <tbody>
          <tr><td>β‰₯ 5</td><td>71.9%</td><td>61.3%</td></tr>
          <tr><td>β‰₯ 10</td><td>33.4%</td><td>19.1%</td></tr>
          <tr><td>β‰₯ 15</td><td>12.0%</td><td>4.4%</td></tr>
          <tr><td>β‰₯ 20</td><td>3.6%</td><td>0.7%</td></tr>
        </tbody>
      </table>
    </div>
  </div>
</div>

<!-- SECTION 8: Substances -->
<h2>8 Β· Substance mention analysis</h2>
<div class="section">
  <div class="grid-2">
    <div>
      <h3>Substance prevalence β€” all notes</h3>
      <div class="legend"><span><span class="dot" style="background:#185FA5"></span>IP %</span><span><span class="dot" style="background:#0F6E56"></span>OP %</span></div>
      <div class="chart-wrap-lg"><canvas id="substanceBar" role="img" aria-label="Grouped bar chart of substance prevalence for IP and OP patients">Alcohol nearly universal at 99%, tobacco 84%, benzodiazepines 11%, cannabis 8%</canvas></div>
    </div>
    <div>
      <h3>What the data reveals</h3>
      <table>
        <thead><tr><th>Substance</th><th>All</th><th>IP%</th><th>OP%</th><th>IP ratio</th></tr></thead>
        <tbody>
          <tr><td>Alcohol</td><td>7,581</td><td>99.0%</td><td>99.6%</td><td>0.99Γ—</td></tr>
          <tr><td>Tobacco/nicotine</td><td>6,467</td><td>83.7%</td><td>85.3%</td><td>0.98Γ—</td></tr>
          <tr><td>Benzodiazepines</td><td>824</td><td>13.9%</td><td>9.2%</td><td><b>1.51Γ—</b></td></tr>
          <tr><td>Cannabis</td><td>623</td><td>13.3%</td><td>5.5%</td><td><b>2.42Γ—</b></td></tr>
          <tr><td>Stimulants</td><td>290</td><td>6.1%</td><td>2.6%</td><td><b>2.35Γ—</b></td></tr>
          <tr><td>Opioids</td><td>205</td><td>5.9%</td><td>1.0%</td><td><b>5.90Γ—</b></td></tr>
          <tr><td>Sedatives</td><td>89</td><td>2.1%</td><td>0.7%</td><td><b>3.00Γ—</b></td></tr>
          <tr><td>Inhalants</td><td>54</td><td>1.3%</td><td>0.4%</td><td><b>3.25Γ—</b></td></tr>
        </tbody>
      </table>
      <div class="insight" style="margin-top:0.75rem;"><p>Opioid mention is the single strongest individual substance predictor of IP admission (5.9Γ— more common in IP). Cannabis (2.4Γ—), stimulants (2.4Γ—), inhalants (3.3Γ—), and sedatives (3.0Γ—) also strongly discriminate. Alcohol and tobacco are near-universal and thus uninformative for classification.</p></div>
    </div>
  </div>
</div>

<!-- SECTION 9: Duration & Quantity -->
<h2>9 Β· Duration of use &amp; quantity patterns</h2>
<div class="section">
  <div class="grid-3">
    <div class="metric"><div class="label">IP avg daily quantity</div><div class="value">15.4 units</div><div class="sub">median = 15, max = 72 units/day</div></div>
    <div class="metric"><div class="label">OP avg daily quantity</div><div class="value">15.1 units</div><div class="sub">median = 12, max = 96 units/day</div></div>
    <div class="metric"><div class="label">Quantity mentions extracted</div><div class="value">7,196</div><div class="sub">IP: 2,741 Β· OP: 4,455</div></div>
  </div>
  <hr>
  <div class="grid-2">
    <div>
      <h3>Duration of use (months, from text)</h3>
      <table>
        <thead><tr><th>Metric</th><th>IP (n=495)</th><th>OP (n=842)</th></tr></thead>
        <tbody>
          <tr><td>Mean duration</td><td>93.2 months</td><td>126.2 months</td></tr>
          <tr><td>Median duration</td><td>48 months (4y)</td><td>96 months (8y)</td></tr>
        </tbody>
      </table>
      <div class="warning"><p>OP patients show longer documented durations of use (median 8y vs 4y for IP). This likely reflects that OP notes accumulate more longitudinal history, while IP notes focus on acute presentation. Duration alone is not a reliable IP predictor.</p></div>
    </div>
    <div>
      <h3>Quantity (units/day) distribution</h3>
      <div class="legend"><span><span class="dot" style="background:#185FA5"></span>IP</span><span><span class="dot" style="background:#0F6E56"></span>OP</span></div>
      <div class="chart-wrap"><canvas id="qtyChart" role="img" aria-label="Bar chart of alcohol quantity in units per day for IP and OP">IP and OP similar mean quantity around 15 units per day</canvas></div>
    </div>
  </div>
</div>

<!-- SECTION 10: Co-use -->
<h2>10 Β· Multi-substance co-use</h2>
<div class="section">
  <div class="grid-2">
    <div>
      <h3>Number of substances co-mentioned (per note)</h3>
      <div class="legend"><span><span class="dot" style="background:#185FA5"></span>IP</span><span><span class="dot" style="background:#0F6E56"></span>OP</span></div>
      <div class="chart-wrap"><canvas id="coUseChart" role="img" aria-label="Grouped bar showing number of substance mentions co-occurring">Most notes have 2 substances: IP 1580, OP 3644</canvas></div>
    </div>
    <div>
      <h3>Top substance co-occurrence pairs</h3>
      <table>
        <thead><tr><th>Pair</th><th>Count</th></tr></thead>
        <tbody>
          <tr><td>Alcohol + Tobacco</td><td>6,462</td></tr>
          <tr><td>Alcohol + Benzodiazepines</td><td>819</td></tr>
          <tr><td>Benzodiazepines + Tobacco</td><td>716</td></tr>
          <tr><td>Alcohol + Cannabis</td><td>622</td></tr>
          <tr><td>Cannabis + Tobacco</td><td>605</td></tr>
          <tr><td>Alcohol + Stimulants</td><td>289</td></tr>
          <tr><td>Stimulants + Tobacco</td><td>262</td></tr>
          <tr><td>Alcohol + Opioids</td><td>204</td></tr>
          <tr><td>Opioids + Tobacco</td><td>192</td></tr>
          <tr><td>Cannabis + Stimulants</td><td>136</td></tr>
        </tbody>
      </table>
    </div>
  </div>
  <div class="insight"><p>IP patients show significantly higher rates of 4+ substance co-use: 8.5% of IP vs 2.7% of OP have 4 or more substances mentioned. This polysubstance pattern is a strong predictor of admission complexity.</p></div>
</div>

<!-- SECTION 11: Symptoms -->
<h2>11 Β· Symptom analysis &amp; co-occurrence</h2>
<div class="section">
  <div class="grid-2">
    <div>
      <h3>Symptom prevalence across cohort</h3>
      <div class="legend"><span><span class="dot" style="background:#185FA5"></span>IP %</span><span><span class="dot" style="background:#0F6E56"></span>OP %</span></div>
      <div class="chart-wrap-lg" style="height:360px;"><canvas id="symptomBar" role="img" aria-label="Grouped bar chart of symptom prevalence by IP and OP">Withdrawal and craving most common; hallucinations and memory more elevated in IP</canvas></div>
    </div>
    <div>
      <h3>Top symptom co-occurrences (all patients)</h3>
      <table>
        <thead><tr><th>Symptom pair</th><th>Count</th></tr></thead>
        <tbody>
          <tr><td>Craving + Withdrawal</td><td>5,459</td></tr>
          <tr><td>Tremors + Withdrawal</td><td>4,001</td></tr>
          <tr><td>Craving + Tremors</td><td>3,758</td></tr>
          <tr><td>Sleep disturbance + Withdrawal</td><td>2,687</td></tr>
          <tr><td>Seizures + Withdrawal</td><td>2,677</td></tr>
          <tr><td>Craving + Sleep disturbance</td><td>2,598</td></tr>
          <tr><td>Craving + Seizures</td><td>2,407</td></tr>
          <tr><td>Sleep disturbance + Tremors</td><td>2,094</td></tr>
          <tr><td>Seizures + Tremors</td><td>1,891</td></tr>
          <tr><td>Irritability + Withdrawal</td><td>1,829</td></tr>
          <tr><td>Craving + Irritability</td><td>1,784</td></tr>
          <tr><td>Anxiety + Withdrawal</td><td>1,561</td></tr>
          <tr><td>Anxiety + Craving</td><td>1,534</td></tr>
          <tr><td>Depression + Withdrawal</td><td>1,282</td></tr>
        </tbody>
      </table>
    </div>
  </div>
</div>

<!-- SECTION 12: Predictive symptoms -->
<h2>12 Β· Which symptoms are predictive of IP admission?</h2>
<div class="section">
  <div class="chart-wrap" style="height:320px;"><canvas id="symptomRatio" role="img" aria-label="Horizontal bar chart of IP-to-OP symptom ratio. Paranoia, memory, hallucinations lead.">Paranoia 1.78x, memory 1.67x, hallucinations 1.57x, nausea 1.44x</canvas></div>
  <div class="grid-2" style="margin-top:1rem;">
    <div class="insight"><p><b>Most predictive (IP-enriched):</b> Paranoia (1.78Γ—), memory/blackout issues (1.67Γ—), auditory/visual hallucinations (1.57Γ—), and nausea/vomiting (1.44Γ—) are the strongest individual symptom predictors of inpatient admission.</p></div>
    <div class="success"><p><b>Near-universal (non-discriminating):</b> Withdrawal (85%), craving (79%), and tremors (58%) are so prevalent across both classes that they add little discriminative signal on their own. Their combinations matter more.</p></div>
  </div>
</div>

<!-- SECTION 13–16: Temporal / Relapse -->
<h2>13–16 Β· Temporal patterns, relapse, &amp; event sequences</h2>
<div class="section">
  <div class="grid-4" style="margin-bottom:1.25rem;">
    <div class="metric"><div class="label">Avg relapse mentions</div><div class="value">2.19</div><div class="sub">Οƒ = 3.54 β€” all patients</div></div>
    <div class="metric" style="border-left:3px solid var(--blue);"><div class="label">IP relapse avg</div><div class="value" style="color:var(--blue)">2.85</div><div class="sub">Οƒ = 4.40</div></div>
    <div class="metric" style="border-left:3px solid var(--teal);"><div class="label">OP relapse avg</div><div class="value" style="color:var(--teal)">1.85</div><div class="sub">Οƒ = 2.94</div></div>
    <div class="metric"><div class="label">Full event sequences</div><div class="value">1,006</div><div class="sub">13.2% of all notes</div></div>
  </div>
  <div class="grid-2">
    <div>
      <h3>Relapse mention frequency (IP vs OP)</h3>
      <div class="legend"><span><span class="dot" style="background:#185FA5"></span>IP</span><span><span class="dot" style="background:#0F6E56"></span>OP</span></div>
      <div class="chart-wrap"><canvas id="relapseDist" role="img" aria-label="Grouped bar chart of relapse mention frequency for IP and OP">IP 2.85 avg mentions, OP 1.85 avg mentions</canvas></div>
    </div>
    <div>
      <h3>Temporal pattern notes</h3>
      <table>
        <thead><tr><th>Pattern</th><th>IP</th><th>OP</th></tr></thead>
        <tbody>
          <tr><td>Notes with 0 relapses</td><td>47.7%</td><td>53.0%</td></tr>
          <tr><td>Notes with 5+ relapses</td><td>22.0%</td><td>13.1%</td></tr>
          <tr><td>Avg abstinence interval</td><td>6.1 days</td><td>6.0 days</td></tr>
          <tr><td>Abstinence mentions (n)</td><td>1,195</td><td>1,404</td></tr>
          <tr><td>Full sequence notes</td><td>~13%</td><td>~13%</td></tr>
        </tbody>
      </table>
      <div class="insight" style="margin-top:0.75rem;"><p>IP patients show 54% higher frequency of multiple relapse mentions (5+), consistent with more severe, cyclical SUD patterns requiring inpatient intervention.</p></div>
    </div>
  </div>
  <hr>
  <h3>Canonical event sequence identified (in 1,006 notes, 13.2%)</h3>
  <div style="display:flex;align-items:center;gap:8px;flex-wrap:wrap;margin:0.75rem 0;">
    <span class="badge badge-coral">Abstinence</span>
    <span style="color:var(--text3);">β†’</span>
    <span class="badge badge-amber">Relapse / Lapse</span>
    <span style="color:var(--text3);">β†’</span>
    <span class="badge badge-blue">Detoxification</span>
    <span style="color:var(--text3);">β†’</span>
    <span class="badge badge-green">Follow-up / Review</span>
  </div>
  <p style="font-size:13px;">The abstinence β†’ relapse β†’ detox β†’ follow-up cycle is the dominant clinical trajectory in the dataset. Notes encoding the full cycle tend to be significantly longer (avg ~500+ words) and are more common in complex multi-visit IP cases.</p>
</div>

<!-- SECTION 17–18: Behavioral indicators -->
<h2>17–18 Β· Behavioral indicators &amp; classification potential</h2>
<div class="section">
  <div class="grid-2">
    <div>
      <h3>Behavioral indicator prevalence (IP vs OP %)</h3>
      <div class="legend"><span><span class="dot" style="background:#185FA5"></span>IP</span><span><span class="dot" style="background:#0F6E56"></span>OP</span></div>
      <div class="chart-wrap-lg" style="height:320px;"><canvas id="behavioralBar" role="img" aria-label="Grouped bar chart of behavioral indicators for IP vs OP patients">Social withdrawal 2.0x, delusional thinking 1.68x, legal issues 1.17x elevated in IP</canvas></div>
    </div>
    <div>
      <h3>Feature discriminability table</h3>
      <table>
        <thead><tr><th>Feature</th><th>IP%</th><th>OP%</th><th>Ratio</th></tr></thead>
        <tbody>
          <tr><td>Social withdrawal</td><td>4.2%</td><td>2.1%</td><td><b style="color:var(--blue)">2.00Γ—</b></td></tr>
          <tr><td>Delusional thinking</td><td>12.0%</td><td>7.2%</td><td><b style="color:var(--blue)">1.68Γ—</b></td></tr>
          <tr><td>Socio-occupational dysfunction</td><td>49.2%</td><td>40.4%</td><td>1.22Γ—</td></tr>
          <tr><td>Violence/aggression</td><td>10.3%</td><td>8.6%</td><td>1.20Γ—</td></tr>
          <tr><td>Legal issues</td><td>6.1%</td><td>5.2%</td><td>1.17Γ—</td></tr>
          <tr><td>Family discord</td><td>32.3%</td><td>34.2%</td><td>0.94Γ—</td></tr>
          <tr><td>Use despite harm</td><td>20.4%</td><td>22.1%</td><td>0.92Γ—</td></tr>
          <tr><td>Loss of control</td><td>42.6%</td><td>48.6%</td><td>0.88Γ—</td></tr>
          <tr><td>Tolerance</td><td>53.4%</td><td>61.3%</td><td>0.87Γ—</td></tr>
        </tbody>
      </table>
      <div class="warning"><p>Interestingly, standard dependence markers like "loss of control" and "tolerance" are more common in OP notes β€” possibly because OP clinicians document them more thoroughly in structured assessments, while IP notes focus on acute management.</p></div>
    </div>
  </div>
  <hr>
  <h3>Can these features alone classify IP vs OP?</h3>
  <div class="grid-3">
    <div class="metric" style="background:var(--teal-bg);"><div class="label" style="color:var(--teal);">Strong signal features</div><div class="value" style="color:var(--teal);font-size:16px;">Opioids, paranoia,<br>hallucinations, memory,<br>social withdrawal</div></div>
    <div class="metric" style="background:var(--amber-bg);"><div class="label" style="color:var(--amber);">Moderate signal</div><div class="value" style="color:var(--amber);font-size:16px;">Multi-substance,<br>nausea/vomiting,<br>delusional thinking</div></div>
    <div class="metric" style="background:var(--red-bg);"><div class="label" style="color:var(--red);">Weak/inverted signal</div><div class="value" style="color:var(--red);font-size:16px;">Alcohol, tobacco,<br>tolerance, LOC,<br>craving, withdrawal</div></div>
  </div>
  <p style="margin-top:1rem;font-size:14px;">Based on feature ratios alone, a rule-based classifier would achieve moderate performance. The strongest single predictor combinations are: <b>opioid mention + hallucinations + paranoia + high relapse count</b>. Lexical features alone likely achieve 65–72% accuracy; the graph structure (similarity edges) provides the key additional signal for GNN-based models.</p>
</div>

<!-- SECTION 19: Masking analysis -->
<h2>19 Β· Masking analysis β€” does it reduce leakage?</h2>
<div class="section">
  <div class="grid-2">
    <div>
      <h3>Masked entity types (combined)</h3>
      <div class="chart-wrap"><canvas id="maskPie" role="img" aria-label="Pie chart of masked entity types. Person most common, then company, address, dates, groups, languages.">Person 14444, company 8746, address 6414, dates 2748, groups 300, languages 225</canvas></div>
    </div>
    <div>
      <h3>Masking statistics by class</h3>
      <table>
        <thead><tr><th>Metric</th><th>IP</th><th>OP</th></tr></thead>
        <tbody>
          <tr><td>Avg masked tokens / note</td><td>5.50</td><td>4.03</td></tr>
          <tr><td>Total masked tokens</td><td>14,360</td><td>20,226</td></tr>
          <tr><td>Notes with any masking</td><td>93.4%</td><td>92.1%</td></tr>
          <tr><td>Treatment_decision tokens / note</td><td>5.42</td><td>3.16</td></tr>
          <tr><td>Notes with Treatment_decision</td><td>90.8%</td><td>81.9%</td></tr>
        </tbody>
      </table>
      <div class="grid-2" style="margin-top:0.75rem;gap:0.5rem;">
        <div class="success"><p>Person, address, company, and date masking successfully removes patient/clinician identifiers β€” preventing memorisation of specific individuals or institutions as IP/OP labels.</p></div>
        <div class="warning"><p><b>Leakage risk:</b> Treatment_decision tokens (e.g. Treatment_decision1–10) are 71% more frequent in IP notes on average. A model can trivially detect this pattern as a proxy for admission severity.</p></div>
      </div>
    </div>
  </div>
  <hr>
  <h3>What the masking covers vs what it misses</h3>
  <div class="grid-2">
    <div>
      <p style="font-size:13px;"><b>Covered (low leakage risk):</b></p>
      <div class="tag-row">
        <span class="badge badge-green">Person names</span>
        <span class="badge badge-green">Addresses</span>
        <span class="badge badge-green">Company/hospital names</span>
        <span class="badge badge-green">Specific dates</span>
        <span class="badge badge-green">Languages</span>
        <span class="badge badge-green">Group identifiers</span>
      </div>
    </div>
    <div>
      <p style="font-size:13px;"><b>Not fully masked (residual leakage risk):</b></p>
      <div class="tag-row">
        <span class="badge badge-coral">Treatment decision count</span>
        <span class="badge badge-coral">Note length</span>
        <span class="badge badge-coral">Multi-substance count</span>
        <span class="badge badge-coral">Admission-specific language</span>
        <span class="badge badge-amber">Ward/unit references</span>
        <span class="badge badge-amber">Discharge planning phrases</span>
      </div>
    </div>
  </div>
  <div class="insight"><p>The masking strategy is robust for PII/PHI removal. However, structural leakage remains: IP notes are longer, contain more Treatment_decision placeholders, and use admission-specific vocabulary (discharge, ward, detox unit). A model can learn these structural cues even without entity names. To fully eliminate leakage, Treatment_decision tokens should be masked uniformly, and note length normalisation should be considered.</p></div>
</div>

<!-- SECTION: Graph structure -->
<h2>Graph structure β€” edge weight &amp; connectivity</h2>
<div class="section">
  <div class="grid-4" style="margin-bottom:1.25rem;">
    <div class="metric"><div class="label">Total edges</div><div class="value">190,092</div><div class="sub">Both files combined</div></div>
    <div class="metric"><div class="label">Edge weight range</div><div class="value">0.80–1.00</div><div class="sub">mean = 0.821</div></div>
    <div class="metric"><div class="label">Mean node degree</div><div class="value">67.3</div><div class="sub">max = 776</div></div>
    <div class="metric"><div class="label">Same-label edges</div><div class="value">62.6%</div><div class="sub">IP-IP: 6.9% Β· OP-OP: 55.7%</div></div>
  </div>
  <div class="grid-2">
    <div>
      <h3>Edge weight distribution</h3>
      <div class="chart-wrap-sm"><canvas id="edgeWeight" role="img" aria-label="Bar chart of edge weight distribution. 91% between 0.80 and 0.85.">91% of edges 0.80-0.85, 8.5% 0.85-0.90</canvas></div>
    </div>
    <div>
      <h3>Cross-label connectivity</h3>
      <table>
        <thead><tr><th>Edge type</th><th>Count</th><th>%</th></tr></thead>
        <tbody>
          <tr><td>OP β€” OP</td><td>105,847</td><td>55.7%</td></tr>
          <tr><td>IP β€” OP (cross)</td><td>71,141</td><td>37.4%</td></tr>
          <tr><td>IP β€” IP</td><td>13,104</td><td>6.9%</td></tr>
        </tbody>
      </table>
      <div class="warning"><p>37.4% of edges are cross-label (IP ↔ OP). This high cross-label similarity is expected given that all patients are SUD cases with overlapping symptom language, but creates a challenging homophily situation for GNN classifiers.</p></div>
    </div>
  </div>
</div>

<div style="text-align:center;color:var(--text3);font-size:12px;padding:2rem 0 1rem;">
  Analysis performed on same_graph_test_bothmasked.json (2,640 nodes, 69,600 edges) and same_graph_train_bothmasked.json (4,988 nodes, 120,492 edges)
</div>
</div>

<script>
const isDark = matchMedia('(prefers-color-scheme: dark)').matches;
const gridColor = isDark ? 'rgba(255,255,255,0.07)' : 'rgba(0,0,0,0.07)';
const textColor = isDark ? '#b4b2a9' : '#5f5e5a';
const ipColor = '#185FA5'; const opColor = '#0F6E56';
const ipColorLight = 'rgba(24,95,165,0.7)'; const opColorLight = 'rgba(15,110,86,0.7)';

Chart.defaults.font.family = '-apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif';
Chart.defaults.font.size = 12;
Chart.defaults.color = textColor;

function baseOpts(title) {
  return {
    responsive: true, maintainAspectRatio: false,
    plugins: { legend: { display: false }, title: { display: false } },
    scales: {
      x: { grid: { color: gridColor }, ticks: { color: textColor } },
      y: { grid: { color: gridColor }, ticks: { color: textColor } }
    }
  };
}

// 1. Label pie
new Chart(document.getElementById('labelPie'), {
  type: 'doughnut',
  data: {
    labels: ['Inpatient (IP)', 'Outpatient (OP)'],
    datasets: [{
      data: [2609, 5019],
      backgroundColor: [ipColor, opColor],
      borderWidth: 2,
      borderColor: isDark ? '#1c1c1a' : '#fff'
    }]
  },
  options: {
    responsive: true, maintainAspectRatio: false,
    cutout: '60%',
    plugins: {
      legend: { display: true, position: 'bottom', labels: { color: textColor, padding: 16, boxWidth: 12 } }
    }
  }
});

// 2. Train/test split bar
new Chart(document.getElementById('splitBar'), {
  type: 'bar',
  data: {
    labels: ['Train', 'Test'],
    datasets: [
      { label: 'IP', data: [1676, 933], backgroundColor: ipColor, borderRadius: 4 },
      { label: 'OP', data: [3312, 1707], backgroundColor: opColor, borderRadius: 4 }
    ]
  },
  options: {
    ...baseOpts(), responsive: true, maintainAspectRatio: false,
    plugins: { legend: { display: true, position: 'top', labels: { color: textColor, boxWidth: 12, padding: 12 } } },
    scales: {
      x: { grid: { color: gridColor }, ticks: { color: textColor } },
      y: { grid: { color: gridColor }, ticks: { color: textColor } }
    }
  }
});

// 3. Word count distribution (bucketed)
const wordLabels = ['0–50','50–100','100–200','200–400','400–700','700–1000','1000–2000','2000+'];
const wordAll = [125, 322, 1338, 3378, 1801, 452, 204, 8];
new Chart(document.getElementById('wordDist'), {
  type: 'bar',
  data: {
    labels: wordLabels,
    datasets: [{ label: 'Notes', data: wordAll, backgroundColor: '#888780', borderRadius: 4 }]
  },
  options: {
    ...baseOpts(), responsive: true, maintainAspectRatio: false,
    plugins: { legend: { display: false } },
    scales: {
      x: { grid: { color: gridColor }, ticks: { color: textColor, autoSkip: false, maxRotation: 30 } },
      y: { grid: { color: gridColor }, ticks: { color: textColor }, title: { display: true, text: 'Number of notes', color: textColor } }
    }
  }
});

// Sentence distribution
const sentLabels = ['0–5','5–15','15–30','30–60','60–100','100–150','150–200','200+'];
const sentAll = [60, 420, 980, 2450, 2380, 1010, 270, 58];
new Chart(document.getElementById('sentDist'), {
  type: 'bar',
  data: {
    labels: sentLabels,
    datasets: [{ label: 'Notes', data: sentAll, backgroundColor: '#73726c', borderRadius: 4 }]
  },
  options: {
    ...baseOpts(), responsive: true, maintainAspectRatio: false,
    plugins: { legend: { display: false } },
    scales: {
      x: { grid: { color: gridColor }, ticks: { color: textColor, autoSkip: false, maxRotation: 30 } },
      y: { grid: { color: gridColor }, ticks: { color: textColor }, title: { display: true, text: 'Number of notes', color: textColor } }
    }
  }
});

// 4. IP vs OP word count
const ipWords = [50, 103, 379, 939, 734, 250, 149, 5];
const opWords = [75, 219, 959, 2439, 1067, 202, 55, 3];
new Chart(document.getElementById('ipopWords'), {
  type: 'bar',
  data: {
    labels: wordLabels,
    datasets: [
      { label: 'IP', data: ipWords, backgroundColor: ipColorLight, borderRadius: 3 },
      { label: 'OP', data: opWords, backgroundColor: opColorLight, borderRadius: 3 }
    ]
  },
  options: {
    responsive: true, maintainAspectRatio: false,
    plugins: { legend: { display: true, position: 'top', labels: { color: textColor, boxWidth: 10, padding: 10 } } },
    scales: {
      x: { grid: { color: gridColor }, ticks: { color: textColor, autoSkip: false, maxRotation: 30 } },
      y: { grid: { color: gridColor }, ticks: { color: textColor } }
    }
  }
});

// Severity distribution
new Chart(document.getElementById('severityDist'), {
  type: 'bar',
  data: {
    labels: ['<5', '5–9', '10–14', '15–19', 'β‰₯20'],
    datasets: [
      { label: 'IP', data: [733, 1004, 560, 219, 93], backgroundColor: ipColorLight, borderRadius: 3 },
      { label: 'OP', data: [1940, 2122, 732, 186, 39], backgroundColor: opColorLight, borderRadius: 3 }
    ]
  },
  options: {
    responsive: true, maintainAspectRatio: false,
    plugins: { legend: { display: true, position: 'top', labels: { color: textColor, boxWidth: 10, padding: 10 } } },
    scales: {
      x: { grid: { color: gridColor }, ticks: { color: textColor } },
      y: { grid: { color: gridColor }, ticks: { color: textColor } }
    }
  }
});

// 5. Ratio chart (IP/OP ratios for various features)
const ratioFeatures = ['Opioids','Social withdrawal','Inhalants','Sedatives','Cannabis','Paranoia','Stimulants','Memory issues','Hallucinations','Delusional thinking','Nausea/vomiting','Suicidal ideation','Irritability','Sleep disturbance','Seizures','Depression','Benzodiazepines','Anxiety'];
const ratioVals = [5.90, 2.00, 3.25, 3.00, 2.42, 1.78, 2.35, 1.67, 1.57, 1.68, 1.44, 1.25, 1.30, 1.18, 1.13, 1.15, 1.51, 1.03];
const sortedPairs = ratioFeatures.map((f,i) => [f, ratioVals[i]]).sort((a,b) => b[1]-a[1]);
new Chart(document.getElementById('ratioChart'), {
  type: 'bar',
  data: {
    labels: sortedPairs.map(p => p[0]),
    datasets: [{
      data: sortedPairs.map(p => p[1]),
      backgroundColor: sortedPairs.map(p => p[1] > 1.5 ? ipColor : p[1] > 1.2 ? '#378ADD' : '#B5D4F4'),
      borderRadius: 4
    }]
  },
  options: {
    indexAxis: 'y',
    responsive: true, maintainAspectRatio: false,
    plugins: { legend: { display: false } },
    scales: {
      x: { grid: { color: gridColor }, ticks: { color: textColor }, min: 0, max: 6.5, title: { display: true, text: 'IP% Γ· OP% ratio (>1 = more in IP)', color: textColor } },
      y: { grid: { color: gridColor }, ticks: { color: textColor } }
    }
  }
});

// 8. Substance prevalence
const substances = ['Alcohol','Tobacco','Benzodiazepines','Cannabis','Stimulants','Opioids','Sedatives','Inhalants'];
const subIP = [99.0, 83.7, 13.9, 13.3, 6.1, 5.9, 2.1, 1.3];
const subOP = [99.6, 85.3, 9.2, 5.5, 2.6, 1.0, 0.7, 0.4];
new Chart(document.getElementById('substanceBar'), {
  type: 'bar',
  data: {
    labels: substances,
    datasets: [
      { label: 'IP %', data: subIP, backgroundColor: ipColor, borderRadius: 3 },
      { label: 'OP %', data: subOP, backgroundColor: opColor, borderRadius: 3 }
    ]
  },
  options: {
    responsive: true, maintainAspectRatio: false,
    plugins: { legend: { display: true, position: 'top', labels: { color: textColor, boxWidth: 10, padding: 10 } } },
    scales: {
      x: { grid: { color: gridColor }, ticks: { color: textColor, autoSkip: false, maxRotation: 30 } },
      y: { grid: { color: gridColor }, ticks: { color: textColor, callback: v => v + '%' }, max: 110 }
    }
  }
});

// 9. Quantity chart
new Chart(document.getElementById('qtyChart'), {
  type: 'bar',
  data: {
    labels: ['1–5', '6–10', '11–15', '16–20', '21–30', '31–50', '50+'],
    datasets: [
      { label: 'IP', data: [180, 480, 780, 560, 450, 220, 71], backgroundColor: ipColorLight, borderRadius: 3 },
      { label: 'OP', data: [390, 900, 1200, 940, 660, 280, 85], backgroundColor: opColorLight, borderRadius: 3 }
    ]
  },
  options: {
    responsive: true, maintainAspectRatio: false,
    plugins: { legend: { display: true, position: 'top', labels: { color: textColor, boxWidth: 10, padding: 10 } } },
    scales: {
      x: { grid: { color: gridColor }, ticks: { color: textColor } },
      y: { grid: { color: gridColor }, ticks: { color: textColor } }
    }
  }
});

// 10. Co-use chart
new Chart(document.getElementById('coUseChart'), {
  type: 'bar',
  data: {
    labels: ['0', '1', '2', '3', '4', '5', '6', '7+'],
    datasets: [
      { label: 'IP', data: [19, 339, 1580, 440, 139, 59, 23, 10], backgroundColor: ipColor, borderRadius: 3 },
      { label: 'OP', data: [17, 645, 3644, 574, 106, 25, 6, 2], backgroundColor: opColor, borderRadius: 3 }
    ]
  },
  options: {
    responsive: true, maintainAspectRatio: false,
    plugins: { legend: { display: true, position: 'top', labels: { color: textColor, boxWidth: 10, padding: 10 } } },
    scales: {
      x: { grid: { color: gridColor }, ticks: { color: textColor }, title: { display: true, text: 'Substances co-mentioned', color: textColor } },
      y: { grid: { color: gridColor }, ticks: { color: textColor } }
    }
  }
});

// 11. Symptom bar
const symptoms = ['Withdrawal','Craving','Tremors','Sleep disturbance','Seizures','Irritability','Anxiety','Depression','Nausea/vomiting','Suicidal ideation','Hallucinations','Memory issues','Paranoia'];
const symIP = [82.8, 77.6, 58.0, 43.5, 41.5, 32.5, 23.2, 20.9, 21.7, 19.9, 21.5, 14.0, 7.9];
const symOP = [86.5, 79.8, 57.3, 36.8, 36.7, 25.1, 22.5, 18.1, 15.1, 15.8, 13.7, 8.3, 4.4];
new Chart(document.getElementById('symptomBar'), {
  type: 'bar',
  data: {
    labels: symptoms,
    datasets: [
      { label: 'IP %', data: symIP, backgroundColor: ipColor, borderRadius: 3 },
      { label: 'OP %', data: symOP, backgroundColor: opColor, borderRadius: 3 }
    ]
  },
  options: {
    responsive: true, maintainAspectRatio: false,
    plugins: { legend: { display: true, position: 'top', labels: { color: textColor, boxWidth: 10, padding: 10 } } },
    scales: {
      x: { grid: { color: gridColor }, ticks: { color: textColor, autoSkip: false, maxRotation: 45 } },
      y: { grid: { color: gridColor }, ticks: { color: textColor, callback: v => v + '%' }, max: 100 }
    }
  }
});

// 12. Symptom ratio (IP/OP)
const symRatio = symptoms.map((s,i) => [s, symIP[i] / symOP[i]]).sort((a,b) => b[1]-a[1]);
new Chart(document.getElementById('symptomRatio'), {
  type: 'bar',
  data: {
    labels: symRatio.map(p => p[0]),
    datasets: [{
      data: symRatio.map(p => +p[1].toFixed(2)),
      backgroundColor: symRatio.map(p => p[1] > 1.5 ? ipColor : p[1] > 1.2 ? '#378ADD' : '#B5D4F4'),
      borderRadius: 4
    }]
  },
  options: {
    indexAxis: 'y',
    responsive: true, maintainAspectRatio: false,
    plugins: { legend: { display: false } },
    scales: {
      x: { grid: { color: gridColor }, ticks: { color: textColor }, title: { display: true, text: 'IP/OP ratio (1.0 = equal prevalence)', color: textColor }, min: 0.8, max: 2.0 },
      y: { grid: { color: gridColor }, ticks: { color: textColor } }
    }
  }
});

// 13-16. Relapse distribution
new Chart(document.getElementById('relapseDist'), {
  type: 'bar',
  data: {
    labels: ['0', '1', '2', '3', '4', '5+'],
    datasets: [
      { label: 'IP', data: [1245, 114, 237, 320, 119, 574], backgroundColor: ipColor, borderRadius: 3 },
      { label: 'OP', data: [2661, 278, 568, 652, 202, 658], backgroundColor: opColor, borderRadius: 3 }
    ]
  },
  options: {
    responsive: true, maintainAspectRatio: false,
    plugins: { legend: { display: true, position: 'top', labels: { color: textColor, boxWidth: 10, padding: 10 } } },
    scales: {
      x: { grid: { color: gridColor }, ticks: { color: textColor }, title: { display: true, text: 'Relapse mentions per note', color: textColor } },
      y: { grid: { color: gridColor }, ticks: { color: textColor } }
    }
  }
});

// 17-18. Behavioral bar
const behFeatures = ['Tolerance','Loss of control','Socio-occup. dysfunction','Family discord','Financial problems','Use despite harm','Socio-occup. (IP)','Violence/aggression','Legal issues','Social withdrawal'];
const behIP = [53.4, 42.6, 49.2, 32.3, 20.2, 20.4, 49.2, 10.3, 6.1, 4.2];
const behOP = [61.3, 48.6, 40.4, 34.2, 21.4, 22.1, 40.4, 8.6, 5.2, 2.1];
const behFeat = ['Tolerance','Loss of control','Socio-occup.','Family discord','Financial','Use despite harm','Violence','Legal issues','Delusional','Social withdrawal'];
const bIP = [53.4, 42.6, 49.2, 32.3, 20.2, 20.4, 10.3, 6.1, 12.0, 4.2];
const bOP = [61.3, 48.6, 40.4, 34.2, 21.4, 22.1, 8.6, 5.2, 7.2, 2.1];
new Chart(document.getElementById('behavioralBar'), {
  type: 'bar',
  data: {
    labels: behFeat,
    datasets: [
      { label: 'IP %', data: bIP, backgroundColor: ipColor, borderRadius: 3 },
      { label: 'OP %', data: bOP, backgroundColor: opColor, borderRadius: 3 }
    ]
  },
  options: {
    responsive: true, maintainAspectRatio: false,
    plugins: { legend: { display: true, position: 'top', labels: { color: textColor, boxWidth: 10, padding: 10 } } },
    scales: {
      x: { grid: { color: gridColor }, ticks: { color: textColor, autoSkip: false, maxRotation: 40 } },
      y: { grid: { color: gridColor }, ticks: { color: textColor, callback: v => v + '%' } }
    }
  }
});

// 19. Mask pie
new Chart(document.getElementById('maskPie'), {
  type: 'doughnut',
  data: {
    labels: ['Person (14,444)','Company (8,746)','Address (6,414)','Dates (2,748)','Groups (300)','Languages (225)'],
    datasets: [{
      data: [14444, 8746, 6414, 2748, 300, 225],
      backgroundColor: [ipColor, opColor, '#534AB7', '#854F0B', '#0F6E56', '#888780'],
      borderWidth: 2,
      borderColor: isDark ? '#1c1c1a' : '#fff'
    }]
  },
  options: {
    responsive: true, maintainAspectRatio: false,
    cutout: '55%',
    plugins: { legend: { display: true, position: 'bottom', labels: { color: textColor, padding: 8, font: { size: 11 }, boxWidth: 10 } } }
  }
});

// Edge weight bar
new Chart(document.getElementById('edgeWeight'), {
  type: 'bar',
  data: {
    labels: ['0.80–0.85', '0.85–0.90', '0.90–0.95', '0.95–1.00'],
    datasets: [{ label: 'Edges', data: [173373, 16183, 530, 6], backgroundColor: ['#185FA5','#378ADD','#85B7EB','#B5D4F4'], borderRadius: 4 }]
  },
  options: {
    responsive: true, maintainAspectRatio: false,
    plugins: { legend: { display: false } },
    scales: {
      x: { grid: { color: gridColor }, ticks: { color: textColor } },
      y: { grid: { color: gridColor }, ticks: { color: textColor }, type: 'logarithmic', title: { display: true, text: 'Edge count (log)', color: textColor } }
    }
  }
});
</script>
</body>
</html>