Under-diagnosing and under-treating iron deficiency in hospitalized patients with gastrointestinal bleeding

Table 1 Percentage of patients on iron supplements n (%)

	GI bleeders(n = 307)	GI bleeders with proven iron deficiency1(n = 45)
Prior to hospitalization	44 (14.3)	7 (15.6)
During hospitalization	71 (23.1)	29 (64.4)
After hospitalization (discharge instructions)	68 (22.1)	29 (64.4)

¹Iron deficiency by laboratory results during the current hospitalization defined as either iron saturation < 15% or ferritin < 45 μg/L.

Table 2 Predictors of iron deficiency anemia investigation n (%)

Variable		Univariate analysis1		Multivariate analysis
		No iron studies(n = 188)	Iron studies(n = 94)	P value	OR (95%CI)
Anemia documentation2	None	35 (92.1)	3 (7.9)		Reference
	Before discharge summary	153 (62.7)	91 (37.3)	0.001	8.45 (2.35-30.33)
History of GI bleed	None	150 (63.0)	88 (37.0)		Reference
	Yes	38 (86.4)	6 (13.6)	0.001	0.19 (0.07-0.50)
Hematemesis	None	130 (61.6)	81 (38.4)		Reference
	Yes	58 (81.7)	13 (18.3)	0.003	0.32 (0.15-0.68)
FOBT	Negative	56 (62.9)	33 (37.1)	< 0.0013	Reference
	Positive	59 (55.7)	47 (44.3)	0.07	1.87 (0.95-3.69)
	Not performed	73 (83.9)	14 (16.1)	0.005	0.32 (0.14-0.71)
MCV4		90.5 ± 9.4 fL	82.9 ± 19.2 fL	< 0.001	0.95 (0.93-0.98)

¹All variables included in multivariable analysis were statistically significant on univariate analysis, including all the variables shown in this table;

²Anemia documentation on any note including admission, consultation or progress notes but excluding discharge summary;

³P value for trend;

⁴MCV is a continuous variable, so a OR < 1 means that higher MCV was inversely associated with iron deficiency investigation (with every 1 fL increase in MCV there was a 5% decrease in probability of getting iron studies). GI: Gastrointestinal; IDA: Iron deficiency anemia; FOBT: Fecal occult blood testing; MCV: Mean corpuscular volume. Variables included in the multivariable logistic regression but excluded from the table above for being not statistically significant are: Iron supplements on admission, occult bleeding, and lowest hemoglobin. All patients’ characteristics that were not included in the multivariate regression were not significantly different on univariate analysis.