Machine learning for patient selection in corticosteroid decision making in knee osteoarthritis: A feasibility model

Table 1 Baseline characteristics of patients included in final analysis

Characteristics	Frequency	Percentage
Age (year), mean ± SD	63.4 ± 8.3	-
Gender
Male	233	70.6%
Female	97	29.4%
Race
White/caucasian	263	79.7%
Black/afro-caribbean	59	17.9%
Other	8	2.4%
BMI (kg/m2), mean ± SD	32.3 ± 6.2	-
BMI < 30	147	44.5%
BMI ≥ 30	183	55.5%
Number of comorbidities, mean ± SD	0.7 ± 1.2	-
Comorbidities ≤ 3	317	96.1%
Comorbidities > 3	13	3.9%
KLG
KLG < 2	69	20.9%
KLG ≥ 2	261	79.1%
PASE, mean ± SD	163.1 ± 91.6	-
CESD, mean ± SD	9.7 ± 9.1	-
Past steroid use
No	118	35.8%
Yes	38	11.5%
Unknown	174	52.7%
WOMAC pain score, mean ± SD	5.2 ± 4.1	-
Non-significant change	246	74.5%
Significant change	84	25.5%

BMI: Body mass index; PASE: Physical activity scale for the elderly; CESD: Center for epidemiological studies depression; KLG: Kellgren Lawrence grade; WOMAC: Western Ontario and McMaster Universities Osteoarthritis.

Table 2 Data showing the characteristics of patients that achieved significant and non-significant change in Western Ontario and McMaster Universities Osteoarthritis pain scores, n (%)

Characteristics	Meaningful clinical significant change (MCID or MWPC)		Pearson correlation coefficient	P value
	No	Yes
Age (year), mean ± SD	63.1 ± 7.8	64.3 ± 9.5	0.062	0.259
Gender			0.005	0.932
Male	174 (70.7)	59 (70.2)
Female	72 (29.3)	25 (29.8)
Race			-0.016	0.772
White/caucasian	199 (80.9)	64 (76.2)
Black/afro-caribbean	40 (16.3)	19 (22.6)
Other	7 (2.8)	1 (1.2)
BMI (kg/m2), mean ± SD	32.2 ± 6.2	32.4 ± 6.4	0.0150	0.786
BMI < 30	114 (46.3)	36 (42.9)
BMI ≥ 30	132 (53.7)	48 (57.1)
Number of comorbidities, mean ± SD	0.7 ± 1.1	0.7 ± 1.3	0.006	0.919
Comorbidities ≤ 3	239 (97.2)	78 (92.9)
Comorbidities > 3	7 (2.8)	6 (7.1)
KLG			0.034	0.537
KLG < 2	53 (21.5)	15 (17.9)
KLG ≥ 2	193 (78.5)	69 (82.1)
PASE, mean ± SD	159.8 ± 84.7	172.8 ± 109.5	0.062	0.262
CESD, mean ± SD	9.5 ± 8.6	10.2 ± 10.2	0.034	0.543
Past steroid use			-0.013	0.820
No	90 (36.6)	28 (33.3)
Yes	25 (10.2)	13 (15.5)
Unknown	131 (53.2)	43 (51.2)
WOMAC pain score, mean ± SD	5.0 ± 4.9	5.8 ± 4.1	0.084	0.128

MCID: Minimally clinically important difference; MWPC: Meaningful within person change; BMI: Body mass index; KLG: Kellgren Lawrence grade; PASE: Physical activity scale for the elderly; CESD: Center for epidemiological studies depression; WOMAC: Western Ontario and McMaster Universities Osteoarthritis.

Table 3 Sensitivity, specificity, positive predictive value, negative predictive value, accuracy, and F-scores of the linear discriminant analysis model

Parameter	Mean	95%CI
Sensitivity	29.6%	23.6%-35.6%
Specificity	80.8%	77.2%-84.5%
Positive predictive value	36.2%	28.5%-43.9%
Negative predictive value	77.6%	74.7%-80.6%
Accuracy	67.8%	64.6%-70.9%
F1 score	30.8%	-
F2 score	29.7%	-

CI: Confidence interval.