J Clin Med Res

Journal of Clinical Medicine Research, ISSN 1918-3003 print, 1918-3011 online, Open Access

Article copyright, the authors; Journal compilation copyright, J Clin Med Res and Elmer Press Inc

Journal website https://jocmr.elmerjournals.com

Original Article

Volume 17, Number 12, December 2025, pages 676-687

Machine Learning-Based Model to Classify Emergency Severity Index Levels 1-3 in Febrile Patients With Tachycardia: Thailand Triage Prediction System

↓ Figure 2. Feature importance of machine learning models for predicting Emergency Severity Index (ESI) levels 1 to 3.

↓ Figure 3. Receiver operating characteristic (ROC) curves of machine learning models for predicting Emergency Severity Index (ESI) levels 1 to 3.

↓ Figure 4. Calibration plots for machine learning models predicting Emergency Severity Index (ESI) levels 1 to 3 in the test set.

↓ **Table 1.** Baseline Characteristics of Febrile Patients With Tachycardia by ESI Levels 1 to 3
Baseline characteristics	Missing	Level 1 (n = 98), n (%)	Level 2 (n = 256), n (%)	Level 3 (n = 146), n (%)	P value
bpm: beats per minute; ESI: Emergency Severity Index; DBP: diastolic blood pressure; GCS: Glasgow Coma Scale; IQR: interquartile range; MAP: mean arterial pressure; RR: respiratory rate; SBP: systolic blood pressure; SD: standard deviation; SpO₂: peripheral capillary oxygen saturation.
Male	0	61 (62.2)	122 (47.7)	74 (50.7)	0.046
Age (years), mean ± SD	0	65.1 ± 17.3	54.0 ± 21.9	37.7 ± 16.2	< 0.001
Vital signs
Temperature (°C), mean ± SD	0	38.7 ± 0.8	38.7 ± 0.8	38.6 ± 0.7	0.111
Pulse rate (bpm), mean ± SD	0	129.3 ± 21.3	118.1 ± 10.5	110.6 ± 6.1	< 0.001
SBP (mm Hg), mean ± SD	0	119.2 ± 35.1	132.6 ± 24.8	126.1 ± 18.5	< 0.001
DBP (mm Hg), mean ± SD	0	70.6 ± 23.8	78.4 ± 15.8	78.2 ± 13.3	< 0.001
MAP (mm Hg), mean ± SD	0	86.7 ± 26.5	96.4 ± 17.3	94.1 ± 13.9	< 0.001
RR (breaths/min), mean ± SD	0	29.2 ± 7.8	21.4 ± 3.2	19.6 ± 1.3	< 0.001
SpO₂ (%), mean ± SD	0	93.4 ± 8.3	96.6 ± 2.5	96.9 ± 1.7	< 0.001
GCS, median (IQR)	0	15 (14, 15)	15 (15, 15)	15 (15, 15)	< 0.001
Pain scale, median (IQR)	0	0 (0, 0)	0 (0, 0)	0 (0, 4)	< 0.001
Chief complaints
Dyspnea	0	44 (44.9)	41 (16.0)	4 (2.7)	< 0.001
Chest pain/palpitation/syncope	0	8 (8.2)	21 (8.2)	4 (2.7)	0.065
Cough	0	3 (3.1)	19 (7.4)	25 (17.1)	< 0.001
Alteration of consciousness	0	6 (6.1)	10 (3.9)	0 (0)	0.005
Seizure	0	1 (1.0)	3 (1.2)	0 (0)	0.527
Abdominal/pelvic pain	0	7 (7.1)	35 (13.7)	24 (16.4)	0.099
Vomiting/diarrhea	0	3 (3.1)	15 (5.9)	14 (9.6)	0.128
Genitourinary	0	2 (2.0)	10 (3.9)	7 (4.8)	0.550
Headache	0	2 (2.0)	8 (3.1)	14 (9.6)	0.010
Dizziness	0	0 (0)	7 (2.7)	11 (7.5)	0.005
Fatigue/poor intake	0	4 (4.1)	33 (12.9)	3 (2.1)	< 0.001
Back pain	0	0 (0)	3 (1.2)	2 (1.4)	0.712
Musculoskeletal	0	6 (6.1)	22 (8.6)	27 (18.5)	0.003
Comorbidities
Diabetic mellitus	0	21 (21.4)	64 (25.0)	12 (8.2)	< 0.001
Essential hypertension	0	39 (39.8)	99 (38.7)	22 (15.1)	< 0.001
Dyslipidemia	0	28 (28.6)	72 (28.1)	13 (8.9)	< 0.001
Chronic kidney disease	0	9 (9.2)	20 (7.8)	3 (2.1)	0.022
Cardiovascular disease	0	12 (12.2)	12 (4.7)	1 (0.7)	< 0.001
Respiratory disease	0	12 (12.2)	13 (5.1)	3 (2.1)	0.004
Neurovascular disease	0	11 (11.2)	23 (9.0)	3 (2.1)	0.004
Autoimmune disease	0	1 (1.0)	4 (1.6)	0 (0)	0.409
Hematologic disease	0	9 (9.2)	9 (3.5)	4 (2.7)	0.049
Malignancy	0	12 (12.2)	25 (9.8)	6 (4.1)	0.042
Current medications
Beta-blocker	0	11 (11.2)	19 (7.4)	3 (2.1)	0.008
Immunosuppressants	0	7 (7.1)	17 (6.6)	1 (0.7)	0.006

↓ Table 1. Baseline Characteristics of Febrile Patients With Tachycardia by ESI Levels 1 to 3

Baseline characteristics

Missing

Level 1 (n = 98), n (%)

Level 2 (n = 256), n (%)

Level 3 (n = 146), n (%)

P value

bpm: beats per minute; ESI: Emergency Severity Index; DBP: diastolic blood pressure; GCS: Glasgow Coma Scale; IQR: interquartile range; MAP: mean arterial pressure; RR: respiratory rate; SBP: systolic blood pressure; SD: standard deviation; SpO₂: peripheral capillary oxygen saturation.

Male

61 (62.2)

122 (47.7)

74 (50.7)

0.046

Age (years), mean ± SD

65.1 ± 17.3

54.0 ± 21.9

37.7 ± 16.2

< 0.001

Vital signs

Temperature (°C), mean ± SD

38.7 ± 0.8

38.6 ± 0.7

0.111

Pulse rate (bpm), mean ± SD

129.3 ± 21.3

118.1 ± 10.5

110.6 ± 6.1

< 0.001

SBP (mm Hg), mean ± SD

119.2 ± 35.1

132.6 ± 24.8

126.1 ± 18.5

< 0.001

DBP (mm Hg), mean ± SD

70.6 ± 23.8

78.4 ± 15.8

78.2 ± 13.3

< 0.001

MAP (mm Hg), mean ± SD

86.7 ± 26.5

96.4 ± 17.3

94.1 ± 13.9

< 0.001

RR (breaths/min), mean ± SD

29.2 ± 7.8

21.4 ± 3.2

19.6 ± 1.3

< 0.001

SpO₂ (%), mean ± SD

93.4 ± 8.3

96.6 ± 2.5

96.9 ± 1.7

< 0.001

GCS, median (IQR)

15 (14, 15)

15 (15, 15)

< 0.001

Pain scale, median (IQR)

0 (0, 0)

0 (0, 4)

< 0.001

Chief complaints

Dyspnea

44 (44.9)

41 (16.0)

4 (2.7)

< 0.001

Chest pain/palpitation/syncope

8 (8.2)

21 (8.2)

4 (2.7)

0.065

Cough

3 (3.1)

19 (7.4)

25 (17.1)

< 0.001

Alteration of consciousness

6 (6.1)

10 (3.9)

0 (0)

0.005

Seizure

1 (1.0)

3 (1.2)

0 (0)

0.527

Abdominal/pelvic pain

7 (7.1)

35 (13.7)

24 (16.4)

0.099

Vomiting/diarrhea

3 (3.1)

15 (5.9)

14 (9.6)

0.128

Genitourinary

2 (2.0)

10 (3.9)

7 (4.8)

0.550

Headache

2 (2.0)

8 (3.1)

14 (9.6)

0.010

Dizziness

0 (0)

7 (2.7)

11 (7.5)

0.005

Fatigue/poor intake

4 (4.1)

33 (12.9)

3 (2.1)

< 0.001

Back pain

0 (0)

3 (1.2)

2 (1.4)

0.712

Musculoskeletal

6 (6.1)

22 (8.6)

27 (18.5)

0.003

Comorbidities

Diabetic mellitus

21 (21.4)

64 (25.0)

12 (8.2)

< 0.001

Essential hypertension

39 (39.8)

99 (38.7)

22 (15.1)

< 0.001

Dyslipidemia

28 (28.6)

72 (28.1)

13 (8.9)

< 0.001

Chronic kidney disease

9 (9.2)

20 (7.8)

3 (2.1)

0.022

Cardiovascular disease

12 (12.2)

12 (4.7)

1 (0.7)

< 0.001

Respiratory disease

12 (12.2)

13 (5.1)

3 (2.1)

0.004

Neurovascular disease

11 (11.2)

23 (9.0)

3 (2.1)

0.004

Autoimmune disease

1 (1.0)

4 (1.6)

0 (0)

0.409

Hematologic disease

9 (9.2)

9 (3.5)

4 (2.7)

0.049

Malignancy

12 (12.2)

25 (9.8)

6 (4.1)

0.042

Current medications

Beta-blocker

11 (11.2)

19 (7.4)

3 (2.1)

0.008

Immunosuppressants

7 (7.1)

17 (6.6)

1 (0.7)

0.006

↓ **Table 2.** Performance Metrics of Machine Learning Models for Predicting ESI Levels 1 to 3
Models	AuROC	95% CI	Accuracy	Recall	Precision	F1 score
AuROC: area under the receiver operating characteristic curve; CI: confidence interval; ESI: Emergency Severity Index; F1 score: harmonic mean of precision and recall; XGBoost: Extreme Gradient Boosting.
Random forest
ESI level 1	1.00	0.99, 1.00	1.00	0.70	1.00	0.82
ESI level 2	0.93	0.88, 0.98	0.82	0.92	0.82	0.87
ESI level 3	0.96	0.93, 1.00	0.86	0.86	0.86	0.86
XGBoost
ESI level 1	1.00	0.99, 1.00	1.00	0.70	1.00	0.82
ESI level 2	0.94	0.89, 0.98	0.83	0.88	0.83	0.86
ESI level 3	0.97	0.93, 1.00	0.81	0.90	0.81	0.85
Gradient boosting machine
ESI level 1	1.00	0.99, 1.00	1.00	0.70	1.00	0.82
ESI level 2	0.93	0.88, 0.98	0.83	0.86	0.83	0.85
ESI level 3	0.96	0.92, 1.00	0.79	0.90	0.79	0.84

↓ Table 2. Performance Metrics of Machine Learning Models for Predicting ESI Levels 1 to 3

Models

AuROC

95% CI

Accuracy

Recall

Precision

F1 score

AuROC: area under the receiver operating characteristic curve; CI: confidence interval; ESI: Emergency Severity Index; F1 score: harmonic mean of precision and recall; XGBoost: Extreme Gradient Boosting.

Random forest

ESI level 1

1.00

0.99, 1.00

1.00

0.70

1.00

0.82

ESI level 2

0.93

0.88, 0.98

0.82

0.92

0.82

0.87

ESI level 3

0.96

0.93, 1.00

0.86

XGBoost

ESI level 1

1.00

0.99, 1.00

1.00

0.70

1.00

0.82

ESI level 2

0.94

0.89, 0.98

0.83

0.88

0.83

0.86

ESI level 3

0.97

0.93, 1.00

0.81

0.90

0.81

0.85

Gradient boosting machine

ESI level 1

1.00

0.99, 1.00

1.00

0.70

1.00

0.82

ESI level 2

0.93

0.88, 0.98

0.83

0.86

0.83

0.85

ESI level 3

0.96

0.92, 1.00

0.79

0.90

0.79

0.84

↓ **Table 3.** Confusion Matrix for the Random Forest Model
Predicted ESI (n, %)	Observed ESI (n, %)	Total
Observed ESI: Emergency Severity Index level assigned by triage nurses; Predicted ESI: Emergency Severity Index level assigned by the random forest model.
1	14 (14.0)	6 (6.0)	0 (0)	20 (20.0)
2	0 (0)	47 (47.0)	4 (4.0)	51 (51.0)
3	0 (0)	4 (4.0)	25 (25.0)	29 (29.0)
Total	14 (14.0)	57 (57.0)	29 (29.0)	100 (100)

↓ Table 3. Confusion Matrix for the Random Forest Model

Predicted ESI (n, %)

Observed ESI (n, %)

Total

Observed ESI: Emergency Severity Index level assigned by triage nurses; Predicted ESI: Emergency Severity Index level assigned by the random forest model.

14 (14.0)

6 (6.0)

0 (0)

20 (20.0)

0 (0)

47 (47.0)

4 (4.0)

51 (51.0)

0 (0)

4 (4.0)

25 (25.0)

29 (29.0)

Total

14 (14.0)

57 (57.0)

29 (29.0)

100 (100)

↓ **Table 4.** Confusion Matrix for the Extreme Gradient Boosting Model
Predicted ESI (n, %)	Observed ESI (n, %)	Total
Observed ESI: Emergency Severity Index level assigned by triage nurses; Predicted ESI: Emergency Severity Index level assigned by the extreme gradient boosting model.
1	14 (14.0)	6 (6.0)	0 (0)	20 (20.0)
2	0 (0)	45 (45.0)	6 (6.0)	51 (51.0)
3	0 (0)	3 (3.0)	26 (26.0)	29 (29.0)
Total	14 (14.0)	54 (54.0)	32 (32.0)	100 (100)

↓ Table 4. Confusion Matrix for the Extreme Gradient Boosting Model

Predicted ESI (n, %)

Observed ESI (n, %)

Total

Observed ESI: Emergency Severity Index level assigned by triage nurses; Predicted ESI: Emergency Severity Index level assigned by the extreme gradient boosting model.

14 (14.0)

6 (6.0)

0 (0)

20 (20.0)

0 (0)

45 (45.0)

6 (6.0)

51 (51.0)

0 (0)

3 (3.0)

26 (26.0)

29 (29.0)

Total

14 (14.0)

54 (54.0)

32 (32.0)

100 (100)

↓ **Table 5.** Confusion Matrix for the Gradient Boosting Machine Model
Predicted ESI (n, %)	Observed ESI (n, %)	Total
Observed ESI: Emergency Severity Index level assigned by triage nurses; Predicted ESI: Emergency Severity Index level assigned by gradient boosting machine model.
1	14 (14.0)	6 (6.0)	0 (0)	20 (20.0)
2	0 (0)	44 (44.0)	7 (7.0)	51 (51.0)
3	0 (0)	3 (3.0)	26 (26.0)	29 (29.0)
Total	14 (14.0)	53 (53.0)	33 (33.0)	100 (100)

↓ Table 5. Confusion Matrix for the Gradient Boosting Machine Model

Predicted ESI (n, %)

Observed ESI (n, %)

Total

Observed ESI: Emergency Severity Index level assigned by triage nurses; Predicted ESI: Emergency Severity Index level assigned by gradient boosting machine model.

14 (14.0)

6 (6.0)

0 (0)

20 (20.0)

0 (0)

44 (44.0)

7 (7.0)

51 (51.0)

0 (0)

3 (3.0)

26 (26.0)

29 (29.0)

Total

14 (14.0)

53 (53.0)

33 (33.0)

100 (100)

↓ **Table 6.** Predicted Triage Misclassification and Penalty Scores Across Machine Learning Models
Model	Under-triage (two steps) (n, %)	Under-triage (one step) (n, %)	Over-triage (two steps) (n, %)	Over-triage (one step) (n, %)	Mean penalty
Under-triage (two steps): predicted ESI two levels lower than observed (e.g., observed level 1 predicted as level 3); Under-triage (one step): predicted one level lower (e.g., observed level 2 predicted as level 3); Over-triage (two steps): predicted two levels higher (e.g., observed level 3 predicted as level 1); Over-triage (one step): predicted one level higher (e.g., observed level 2 predicted as level 1); XGBoost: extreme gradient boosting.
Random forest	0 (0)	4 (4.0)	0 (0)	10 (10.0)	0.22
XGBoost	0 (0)	3 (3.0)	0 (0)	12 (12.0)	0.21
Gradient boosting machine	0 (0)	3 (3.0)	0 (0)	13 (13.0)	0.22

↓ Table 6. Predicted Triage Misclassification and Penalty Scores Across Machine Learning Models

Model

Under-triage (two steps) (n, %)

Under-triage (one step) (n, %)

Over-triage (two steps) (n, %)

Over-triage (one step) (n, %)

Mean penalty

Under-triage (two steps): predicted ESI two levels lower than observed (e.g., observed level 1 predicted as level 3); Under-triage (one step): predicted one level lower (e.g., observed level 2 predicted as level 3); Over-triage (two steps): predicted two levels higher (e.g., observed level 3 predicted as level 1); Over-triage (one step): predicted one level higher (e.g., observed level 2 predicted as level 1); XGBoost: extreme gradient boosting.

Random forest

0 (0)

4 (4.0)

0 (0)

10 (10.0)

0.22

XGBoost

0 (0)

3 (3.0)

0 (0)

12 (12.0)

0.21

Gradient boosting machine

0 (0)

3 (3.0)

0 (0)

13 (13.0)

0.22