Our study of internal medicine inpatients demonstrates that the volume of blood taken for diagnostic testing strongly predicts hemoglobin and hematocrit changes during hospitalization. For every 1 mL of phlebotomy, mean (SD) decreases in hemoglobin and hematocrit values were 0.070 (0.011) g/L and 0.019% (0.003%), respectively. Accordingly, for 100 mL, hemoglobin and hematocrit levels would be expected to change by 7.0 g/L and 1.9%, respectively.
While small changes in hemoglobin may be clinically inconsequential, a clinically significant change has been reported to be between 6.6 and 10 g/L.19,20
In our population, the mean drop in hemoglobin during admission was 7.9 g/L, and larger volumes of phlebotomy resulted in larger falls. Expected changes in hemoglobin and hematocrit corresponding to volumes of phlebotomy due to hypothetical clinical settings are presented in . Changes larger than predicted warrant further investigation.
Volumes of Blood Draw and Predicted Drops in Hemoglobin and Hematocrit Based on Clinical Scenarios
Our findings are consistent with prior studies that have assessed the impact of diagnostic phlebotomy on hemoglobin/hematocrit changes in internal medicine patients. Studies by Colimon et al. and Joosten et al. found changes similar to ours,2,4
although Colimon et al., in contrast to our study, did not find a significant decrease in hemoglobin until the length of hospitalization was greater than 4 weeks and the phlebotomy volume was greater than 100 mL.2
Our study was larger and therefore had greater power to detect changes in hemoglobin/hematocrit levels with smaller phlebotomy volumes. In addition, we better controlled for other factors that may cause changes in hemoglobin/hematocrit, such as intravascular volume status, and comorbid and chronic illnesses.
In our study, patients with higher admission hemoglobin/hematocrit values had greater falls in hemoglobin/hematocrit during hospitalization. There are two potential explanations for this finding. First, patients with higher hemoglobin/hematocrit levels will lose more red blood cells per mL of phlebotomy than those with lower levels, hence accounting for the larger drop. Second, patients with higher hemoglobin/hematocrit values on admission may have been “hemoconcentrated” and a drop on the basis of rehydration, in addition to phlebotomy, would be expected. While we adjusted for intravascular volume status in our multivariable analysis, it is possible that our results regarding admission hemoglobin/hematocrit levels represent residual confounding for this variable, that is, volume contraction not captured by our BUN/Cr ratio and osmolality definition. We also found that younger patients had larger hemoglobin/hematocrit changes during hospitalization. We do not have a clear explanation for this effect. It is possible that younger patients received more volume resuscitation because physicians are generally less concerned about fluid administration to younger patients.
Interestingly, post hoc analysis of our cohort revealed that investigations for anemia, including iron indices and fecal occult blood, were performed during 56 (13.9%) hospitalizations. During 11 of these 56 (19.6%) hospitalizations, patients were neither anemic on admission nor admitted for conditions that may have caused anemia. These investigations were likely, at least in part, attributable to changes in hemoglobin values caused by phlebotomy.
Our study has several limitations. First, our results were based on medical record reviews and therefore information pertinent to the inclusion or exclusion criteria may have been incompletely recorded, thereby resulting in misclassification. Second, because our study was retrospective, establishing causation is challenging. It is possible that large falls in hemoglobin/hematocrit, for other reasons, may have triggered investigations requiring large volumes of phlebotomy. We think this is unlikely because anemia-driven blood testing would have been expected to increase daily phlebotomy volumes during the later stages of admission. Our results (not shown), however, indicate that the average daily volume of phlebotomy per patient decreases over time. Third, we only included approximately 38.4% of the hospitalizations due to stringent exclusion criteria. While this may limit the generalizability of our results, we believe that our explicit inclusion and exclusion criteria allow us to make a more accurate assessment of the impact of phlebotomy on hemoglobin/hematocrit levels. Fourth, we were unable to control for the use of medications that may idiosyncratically affect red cell homeostasis (e.g., antibiotics) or cause subclinical blood loss (e.g., Nonsteroidal Antiinflammatory Drugs [NSAIDSs]). We believe any unmeasured effect is likely small and would not significantly affect our results. Finally, accurately controlling for intravascular volume status was difficult and, as mentioned above, our results regarding admission hemoglobin/hematocrit may reflect residual confounding. Also, volume status is likely not a dichotomous variable as it pertains to changes in hemoglobin/hematocrit levels, that is, the relationship is likely more continuous. Our definition is based on those used by others in the literature17,18
and is easy to interpret, in contrast to small changes in BUN/Cr ratios or osmolalities.
Our results highlight the need to perform blood testing judiciously. Physicians do not always check the blood work ordered for the management of their patients.21
Therefore, education of physicians and nurses about the potential risk of anemia as a consequence of phlebotomy should be the first step.16
In addition, other techniques such as reporting of cumulative volumes of phlebotomy to physicians and the use of pediatric-sized blood collection tubes22–25
that have been shown to decrease volumes of phlebotomy in the critical care setting could easily be adapted to general medical wards.