ICM-2009-01177.R3

Campion et al

Online Resource 4

Campion TR Jr., May AK, Waitman LR, Ozdas O, Gadd CS. Effects of blood glucose transcription mismatches on a computer-based intensive insulin therapy protocol. Intensive Care Medicine.

Corresponding author:

Thomas R. Campion, Jr., MS

Department of Biomedical Informatics

Vanderbilt University School of Medicine

400 Eskind Biomedical Library

2209 Garland Avenue

Nashville TN 37232

+1 (615) 936-5092

Fax: +1 (615) 936-1427


Online Resource 4

Potential reasons for missing data and mismatched data observed in study.

Missing Data / Mismatched Data
·  Testing device malfunction
·  Testing device clock set incorrectly
·  Incorrect medical record number used on testing device
·  BG value obtained from testing device not integrated with laboratory information system
·  Failure/forgetfulness to check a BG value
·  Deliberate falsification of BG value
·  Device to laboratory information system transfer failure
·  Laboratory information system to clinical data repository transfer failure
·  Other clinical information system failure (e.g. data backup) / ·  Chance
·  Nursing workload and human factors [13]
o  Two hour vs. one hour monitoring interval
·  Units of measure (mg/dL vs. mmol/L)
o  A device value of 138 mg/dL may be transcribed incorrectly as 238 mg/dL, but it is less likely that 7.6 mmol/L (138 mg/dL) would be entered as 13.2 mmol/L (238 mg/dL).
·  Deliberate action
o  Entering lower BG value in order to administer less insulin due to “fear of hypoglycemia” [14]
o  Entering different BG value to avoid triggering alert (e.g. 61 mg/dL vs. 60 mg/dL)
o  Entering different BG value to avoid changing multiplier (e.g. 110 mg/dL vs. 111 mg/dL)
o  Entering different BG value to avoid administering D50 (e.g. 80 mg/dL vs. 79 mg/dL)

Our time-based approach to linking manually transcribed and device captured BG values for data comparison—one hour before to one hour after— might have also produced missing and mismatched data. However, the approach is justified given the challenge of examining voluminous retrospective data that is only identified by medical record number and timestamp. Additionally, missing data suggests that additional mismatches could have occurred. See [9] for more detail regarding BG device data transmission issues