Development, validation, and proof-of-concept implementation of a two-year risk prediction model for undiagnosed atrial fibrillation using common electronic health data (UNAFIED)

Table 2. Logistic regression variable parameters

Parameter^a,b	Description	Estimate^c	Odds ratio (95% confidence interval)
Intercept		0.4063
Age (years)	40 ≤ Age < 56	− 0.9741	0.38 (0.36, 0.4)
	67 ≤ Age < 77	0.7216	2.06 (1.94, 2.19)
	Age ≥ 77	1.5844	4.88 (4.54, 5.23)
Heart disease (derived)	Present	0.5053	1.66 (1.55, 1.77)
Albumin (g/dL)	Albumin < 3.5	0.7438	2.1 (1.93, 2.3)
BMI (kg/m²)	Missing	0.0884	1.09 (0.99, 1.21)
	BMI < 18.5	0.6086	1.84 (1.26, 2.69)
	24.9 ≤ BMI ≤ 29.9	0.0384	1.04 (0.92, 1.17)
	BMI > 29.9	0.3723	1.45 (1.3, 1.62)
COPD diagnosis	Present	0.5259	1.69 (1.57, 1.82)
Gender	Female	− 0.6226	0.54 (0.51, 0.56)
Heart failure diagnosis	Present	1.0609	2.89 (2.53, 3.31)
Insurance	Commercial	− 0.4111	0.66 (0.62, 0.71)
	Medicaid	0.0378	1.04 (0.95, 1.13)
	Other/unknown	− 0.8584	0.42 (0.39, 0.46)
Kidney disease (derived)	Present	0.58	1.79 (1.59, 2.01)
Shock diagnosis	Present	0.6219	1.86 (1.67, 2.08)

^aSee Additional file 1, for codes and logic used for extraction and derivation of parameters
^bReference parameters: Age 56–66 years (inclusive), no calculated heart disease, albumin ≥ 3.5 (or missing value), normal BMI (19.5–24.9), no COPD, male, no heart failure, Medicare insurance, no calculated kidney disease, and no shock.
^cEach eligible patient had a risk score of exp(raw score)/(1 + exp(raw score)), where the raw score was the sum of the intercept and parameter estimates corresponding to the patients characteristics in each parameter.

ISSN: 1472-6947