Rule-based and machine learning algorithms identify patients with systemic sclerosis accurately in the electronic health record

Table 2 Performance of electronic health record algorithms for systemic sclerosis

Algorithm¹	PPV (%)	Sensitivity (%)	F-score (%)
ICD-9 codes only
≥ 1 count of the ICD-9 code (710.1)	52	100	81
≥ 2 counts	63	88	74
≥ 3 counts	79	72	75
≥ 4 counts	86	67	75
ICD-10 codes only
≥ 1 count of the ICD-10 codes (M34*)	82	94	88
≥ 2 counts	84	91	87
≥ 3 counts	88	85	87
≥ 4 counts	91	85	88
ICD-9 or ICD-10 codes
≥ 1 count	52	98	68
≥ 2 counts	70	97	81
≥ 3 counts	86	94	90
≥ 4 counts	91	91	91
ICD-9 code AND ANA positive²
≥ 1 count of the ICD-9 codes AND ANA	53	81	64
≥ 2 counts of the ICD-9 codes AND ANA	68	81	74
≥ 3 counts of the ICD-9 codes AND ANA	84	70	77
≥ 4 counts of the ICD-9 codes AND ANA	93	64	76
ICD-10 codes AND ANA positive
≥ 1 count of the ICD-10 codes AND ANA	95	53	68
≥ 2 counts AND ANA	95	53	68
≥ 3 counts AND ANA	100	50	67
≥ 4 counts AND ANA	100	50	67
ICD-9 code AND Raynaud’s (RP) keyword
≥ 1 count of the ICD-9 code AND RP	78	90	84
≥ 2 counts AND RP	86	80	83
≥ 3 counts AND RP	92	66	77
≥ 4 counts AND RP	91	60	73
ICD-9 code, RP, ANA positive
≥ 1 count of the ICD-9 code AND ANA OR RP	55	95	70
≥ 2 counts AND ANA OR RP	67	89	76
≥ 3 counts AND ANA OR RP	85	77	81
≥ 4 counts AND ANA OR RP	94	70	81
≥ 1 count AND ANA AND RP	75	75	75
≥ 2 counts AND ANA AND RP	87	75	80
≥ 3 counts AND ANA AND RP	94	66	77
≥ 4 counts AND ANA AND RP	93	59	72

¹All algorithms included at least one or more counts of the SSc ICD-9 (710.1) or ICD-10-CM (M34*) codes for SSc
²ANA positive (titer ≥ 1:80)

ISSN: 1478-6362