| JOURNAL HOME | CME HOME | THIS MONTH | PAST ISSUES | ETOC | COLLECTIONS |
| AUTHORS | REVIEWERS | EDITORIAL BOARD | FEEDBACK | RSS | HELP |
 
|
|
|
||
|
|
© 2000 International Anesthesia Research Society
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionReferences |
|---|
Previous publications suggest that handwritten anesthesia records are less accurate when compared with computer-generated records, but these studies were limited by small sample size, unblinded study design, and unpaired statistical comparisons. Eighty-one pairs of handwritten and computer-generated neurosurgical anesthesia records were retrospectively compared by using a matched sample design. Systolic arterial pressure (SAP), diastolic arterial pressure (DAP), and heart rate (HR) data for each 5-min interval were transcribed from handwritten records. In computerized records, the median of up to 20 values was calculated for SAP, DAP, and HR for each consecutive 5-min epoch. The peak, trough, standard deviation, median, and absolute value of the fractional rate of change between adjacent 5-min epochs were calculated for each case. Pairwise comparisons were performed by using Wilcoxon tests. For SAP, DAP, and HR, the handwritten record peak, standard deviation, and fractional rate of change were less than, and the trough and median were larger than, those in corresponding computer records (all with P < 0.05, except DAP median and HR peak). Considering together all the recorded measurements from all cases, extreme values were recorded more frequently in computerized records than in the handwritten records.
Implications: The discrepancies between handwritten and computerized anesthesia records suggest that some of the data in handwritten records are inaccurate. The potential for inaccuracy should be considered when handwritten records are used as source material for research, quality assurance, and medicolegal purposes.
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
As computerized anesthesia record keeping has become more common, it is clear that the automated means of data collection introduces the potential for more accurate and objective reporting of patient hemodynamic data. In contrast, the hemodynamic data in the handwritten records are subject to "the phenomenon of smoothing" that most likely stems from resistance to reporting extremes and rapid changes in values (1). We quantified the extent of this smoothing phenomenon in the anesthesia records of two cohorts of patients who underwent neurosurgical anesthesia before and after the introduction of a computerized anesthesia record keeping system.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
The study protocol was approved as a retrospective investigation by the Institutional Review Board of the Mount Sinai School of Medicine and was exempt from a requirement for patient consent. The study was designed as a comparison of matched pairs of computerized and handwritten anesthesia records.
All anesthesia records studied were of patients who had undergone neurosurgical procedures conducted under general anesthesia in one tertiary care teaching institution. The attending anesthesiologists were familiar with the anesthesia record keeping software (CompuRecordTM; Anesthesia Recording, Inc., Pittsburgh, PA) after several years of experience with this software in a suite of 22 operating rooms outside of the neurosurgical area. Anesthesiology housestaff had used the computerized system since the beginning of the residency program. Beginning in February 1999, the record keeping of all neurosurgical anesthetics became computerized.
The matching was performed by using a pool of approximately 200 handwritten anesthesia records. The records were drawn from 2 mo in 1998 and approximately the same number of computerized records from 2 mo in 1999 (more than 1 mo after the introduction of the computerized system). The matching was accomplished by one experienced neurosurgical anesthesiologist according to the following criteria: surgical procedure type, nearest available anesthetic duration, and patient age. Three physiologic variables were compared between the matched pairs of computerized and handwritten records: systolic arterial pressure (SAP), diastolic arterial pressure (DAP), and heart rate (HR).
The data in the handwritten records had been documented at 5-min intervals throughout the anesthetic. All legible recordings were estimated to the nearest 5 mm Hg for pressures or 5 (bpm) for HR, transcribed into a computer spreadsheet by two observers, and cross-checked to eliminate errors. We assumed that the handwritten record data were devoid of artifact.
The computerized anesthesia record files registered all available data every 15 s throughout the anesthetic. As a safeguard against artifact, we performed a comprehensive review of the physiologic data and eliminated any obvious errors. As a result of this review, the following criteria were used to identify and eliminate artifactual values:
- SAP < 50 mm Hg
- SAP > 250 mm Hg
- DAP < 20 mm Hg
- DAP > 150 mm Hg
- HR < 20 bpm
- HR > 175 bpm
- DAP
SAP
- SAP - DAP < 10 mm Hg
- SAP > 250 mm Hg
To equalize the frequency of data points (between computerized and handwritten records) and further screen out artifacts, we calculated the median value for each 5-min interval (up to 20 acquired data points) for each variable. These 5-min medians constituted the data from the computerized records that were compared with the handwritten 5-min recordings.
For every case, the peak, trough, median, and standard deviation of all the 5-min values were calculated for SAP, DAP, and HR. For every case, the absolute fractional change (2) from each 5-min value to the next one was calculated for SAP, DAP, and HR by using the following formula:
|
|
Pairwise comparisons between computerized and handwritten records were performed by using the Sign Test for binary comparisons and Wilcoxon’s signed rank test for continuous data. For all statistical analyses, a two-tailed P value < 0.05 was considered significant.
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
The pairing of handwritten and computerized anesthesia records produced two samples that had very similar demographics with regard to duration of anesthesia and patient age (Table 1). The matched pairs included 50 of the same sex (28 female-female and 22 male-male) and 31 of the opposite sex. The percentage of female patients was nearly equal in both sets of records (53% for handwritten and 54% for computerized). The distribution of surgical procedures among the matched pairs is described in Table 2.
|
|
Systolic Arterial Pressure
We analyzed 4044 SAP data points from handwritten records and 4481 5-min epochs from computerized records. The SAP peak was 6% higher (P = 0.012) and the SAP trough was 8% lower (P < 0.001) in the computerized records. The SAP median was 4% lower (P = 0.0097) in the computerized records. The mean |FCM| was 60% larger (P < 0.001) in the computerized records, and the standard deviation was 22% greater (P < 0.001) (Table 3).
|
Diastolic Arterial Pressure
We analyzed 4021 DAP data points from handwritten records and 4481 5-min epochs from computerized records. The DAP peak was 4% higher (P < 0.001) and the DAP trough was 11% lower (P < 0.001) in the computerized records. The DAP median was not significantly different between computerized and handwritten records. The mean |FCM| was 59% greater (P < 0.001) in the computerized records, and the standard deviation was 30% greater (P < 0.001) (Table 3).
Heart Rate
We analyzed 4600 HR data points from handwritten records and 4544 5-min epochs from computerized records. The HR peaks did not differ significantly, but the HR trough was 9% lower (P = 0.007) in the computerized records. The HR median was 7% lower (P = 0.034) in the computerized records. The mean |FCM| was 44% greater (P < 0.001) in the computerized records, and the standard deviation was 14% greater (P = 0.037) (Table 3).
The number and percentage of matched pairs in which the handwritten record value was greater than, less than, or equal to that in the computerized record for the peak, trough, median, and standard deviation of the SAP, DAP, and HR are shown in Table 4.
|
To gain further insight as to the nature of the differences between the two methods of record keeping, extreme values of SAP, DAP, and HR in any handwritten or computerized record at any point during the anesthetic were explored. There were fewer extreme values reported in the handwritten records (Table 5). One exception to this finding is for HR data < 40 bpm, in which the handwritten record from one case accounts for all 19 data points. We further confirmed this finding by stratifying the data into three groups according to procedure duration. The data (not shown) in the three groups are consistent with the result reported in Table 5.
|
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Handwritten records demonstrated a "smoothing effect" that consisted of lower peak and higher trough values for both SAP and DAP and higher trough values for HR. Additionally, there was markedly less variation between adjacent HR, SAP, and DAP values as reflected in the absolute value of fractional changes between temporally adjacent values. Moreover, the differences between handwritten and computerized records were found despite significant mathematical smoothing of the data from the computerized records for each five-minute epoch.
Devitt et al. (3) conducted a study of the completeness and accuracy of 124 handwritten anesthetic records created in a simulator laboratory. The individual subjects were blinded to the purpose of the study and had various levels of experience. In comparing the handwritten records with electronically stored monitored values, the degree of completeness and accuracy of the handwritten record was low, regardless of the anesthesiologist’s age, level of training, or number of years in practice. Subjects at all levels of experience failed to chart significantly abnormal hemodynamic variables; the ranges of discrepancy for HR and systolic BP were 11.2%–20.6% and 20.6%–34.2%, respectively.
Cook et al. (4) compared 46 handwritten and automated anesthesia records for up to one hour per record. The highest automated SAP was at least 20 mm Hg greater than the highest handwritten SAP in 19 cases and at least 40 mm Hg greater in 4 cases. In total, 9.4% of the automated SAP readings were greater than the highest handwritten SAP at any time in the corresponding handwritten record. The pattern was similar for maximum diastolic pressures, whereas the discrepancies between handwritten and automated records for minimum systolic and diastolic pressures were not as great. The causes of the discrepancies were attributed to unobserved data, faulty memory, and observer bias in favor of less controversial values.
Thrush (5) compared 13 handwritten and computerized records to determine if values for physiological variables, including SAP, DAP, and HR, lay outside predetermined ranges. Both records were completed for the same case with the automated record created by a second anesthesiologist who verified the readings and noted artifacts on printout. The early portion (
1.5 hours) of the records was compared. The handwritten records had a significantly lower frequency of low SAP, DAP, and HR than the automated records. These findings were attributed to observer bias, missed readings, and errors in recall.
Lerou et al. (6) analyzed 30 elective ophthalmologic procedures that were dually documented by a standard handwritten record and an automated recording device in an unblinded fashion. The authors computed fractions of time during which there were missing or erroneous data in the handwritten record. The variables SAP, DAP, and HR had relatively low frequencies of missing data. The error fractions for SAP and DAP were 7% and 11%, respectively. HR had a small error fraction rate. Most errors were identified as occurring during the induction period and at the end of the case.
Block (1) retrospectively analyzed 118 cases that were recorded by two automated devices to determine the prevalence of physiologic data outside of arbitrarily set ranges of normal. For SAP, DAP, and HR, 3.6%, 13.3%, and 4.3% of data was outside the normal range, respectively. These data were not compared with the corresponding handwritten records.
In 1977, Zollinger et al. (7) found that 43 of 100 cases contained "major discrepancies" between computerized and handwritten records. They attributed these findings to missed readings during critical periods of the anesthetic during the induction and emergence. This is a plausible observation, because patient monitors in that era did not have recall functions for retrieving physiologic data.
Our findings largely concur with those of others, but are based on a large group of anesthetic records of long duration that were created by practitioners who were unaware that their records would be studied. We also carefully analyzed the variability of the whole sets of data and of changes between temporally adjacent measurements, and we used pairwise statistical comparisons to control for case duration and procedure type. Finally, we believe that using the median values of five-minute epochs from the computer-acquired data is the fairest method of comparing computer-acquired with handwritten data. In all likelihood, these procedures eliminated artifactual data points caused by arterial line insertion and flushing, physical manipulation of the sensors, and similar causes. There is a small chance that some of the extreme values could reflect artifacts that were the result of factors present at the time of data collection, but that were not obvious upon retrospective review (e.g., incorrect transducer height that was not noted in the anesthesia record).
A limitation of the current study is that matched pairs of cases were analyzed, rather than a simultaneous evaluation of each case using both methods. Although the samples are comparable in overall anesthetic duration, the individual pairs had case-length discrepancies that may have influenced the data. Other factors that may have varied among the matched pairs include preoperative hypertension, anesthetic technique, and concurrent medical therapy. Advantages of the design in the current investigation included the lack of awareness of the anesthesiologists that their records would be studied and the consistent nature of the anesthetics in a subspecialized anesthesiology discipline (neurosurgical anesthesia). The generalizability of the results of the study, however, may be limited by the choice of a specialized group of long-duration anesthetics in one institution.
The findings in our handwritten records may reflect less data smoothing than occurs in other institutions. The anesthesiology staff at our institution was experienced with computerized record keeping because it has been in routine use outside of the neurosurgical operating suite for eight years. The staff may therefore have been less concerned by "unsmooth" records. Additionally, it is our impression that the majority of extreme values are concentrated during the induction and emergence portions of the anesthetic, which represent a smaller proportion in anesthetic records of long duration.
Our results raise concern regarding the wider implications of the data smoothing phenomenon in handwritten anesthesia records. Although we cannot determine with certainty the importance of the data that may be missing from handwritten records, preliminary data from our institution (8) suggest that extreme hemodynamic values have prognostic significance over and above those of preoperative risk factors. The opportunity to identify high risk surgical populations and to propose, test, and monitor the effects of therapeutic interventions (e.g., sympatholytic therapy) that may reduce adverse outcomes is diminished in the absence of accurate physiologic data recordings.
In conclusion, this study has demonstrated a considerable "data smoothing" phenomenon in handwritten anesthesia records for the physiological variables SAP, DAP, and HR. This consisted of the truncation of extreme values and the reduction in the fractional change between adjacent measurements. The discrepancies between handwritten and computerized anesthesia records suggest that some of the physiologic data in handwritten records are inaccurate. The potential for inaccuracy should be considered when handwritten records are used as source material for research, quality assurance, and medicolegal purposes.
|
Acknowledgments |
|---|
The authors gratefully acknowledge James B. Eisenkraft, MD, for his critical review of the manuscript.
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
- Block FE. Normal fluctuation of physiologic cardiovascular variables during anesthesia and the phenomenon of "smoothing." J Clin Monit 1991; 7: 141–5.[Web of Science][Medline]
- Reich DL, Osinski TK, Bodian C, et al. An algorithm for assessing intraoperative mean arterial pressure lability. Anesthesiology 1997; 87: 156–61.[Web of Science][Medline]
- Devitt JH, Rapanos T, Kurrek M, et al. The anaesthetic record: accuracy and completeness. Can J Anaesth 1999; 46: 122–8.[Web of Science][Medline]
- Cook R, McDonald JS, Nunziata E. Differences between handwritten and automatic blood pressure records. Anesthesiology 1989; 71: 385–90.[Web of Science][Medline]
- Thrush D. Are automated anesthesia records better? J Clin Anesth 1992; 4: 386–9.[Web of Science][Medline]
- Lerou JGC, Dirksen R, van Daele M, et al. Automated charting of physiological variables in anesthesia: a quantitative comparison of automated versus handwritten anesthesia records. J Clin Monit 1988; 4: 37–47.[Web of Science][Medline]
- Zollinger RM Jr, Kreul JF, Schneider AJL. Man-made versus computer-generated anesthesia records. J Surg Res 1977; 22: 419–24.[Web of Science][Medline]
- Reich DL, Bodian CA, Krol M, et al. Intraoperative hemodynamic predictors of mortality, stroke, and myocardial infarction following coronary artery bypass surgery. Anesth Analg 1999; 88: 814–22.
This article has been cited by other articles:
|
|
 |
|
  S. E. Regenbogen, J. M. Ehrenfeld, S. R. Lipsitz, C. C. Greenberg, M. M. Hutter, and A. A. Gawande Utility of the Surgical Apgar Score: Validation in 4119 Patients Arch Surg, January 1, 2009; 144(1): 30 - 36. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. M. Ehrenfeld, J. L. Walsh, and W. S. Sandberg Right- and Left-Sided Mallinckrodt Double-Lumen Tubes Have Identical Clinical Performance Anesth. Analg., June 1, 2008; 106(6): 1847 - 1852. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
W. D. Driscoll, M. A. Columbia, and R. A. Peterfreund An Observational Study of Anesthesia Record Completeness Using an Anesthesia Information Management System Anesth. Analg., June 1, 2007; 104(6): 1454 - 1461. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Kheterpal, R. Gupta, J. M. Blum, K. K. Tremper, M. O'Reilly, and P. E. Kazanjian Electronic Reminders Improve Procedure Documentation Compliance and Professional Fee Reimbursement Anesth. Analg., March 1, 2007; 104(3): 592 - 597. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D. L. Reich, S. Hossain, and C. A. Bodian What Good are Large Databases of Intraoperative Data? Anesth. Analg., July 1, 2006; 103(1): 252 - 252. [Full Text] [PDF]
|
|
|
|
|
|
D. L. Reich, S. Hossain, M. Krol, B. Baez, P. Patel, A. Bernstein, and C. A. Bodian Predictors of Hypotension After Induction of General Anesthesia Anesth. Analg., September 1, 2005; 101(3): 622 - 628. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. Klasen, A. Junger, B. Hartmann, M. Benson, A. Jost, A. Banzhaf, M. Kwapisz, and G. Hempelmann Differing Incidences of Relevant Hypotension with Combined Spinal-Epidural Anesthesia and Spinal Anesthesia Anesth. Analg., May 1, 2003; 96(5): 1491 - 1495. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D. L. Reich, E. Bennett-Guerrero, C. A. Bodian, S. Hossain, W. Winfree, and M. Krol Intraoperative Tachycardia and Hypertension Are Independently Associated with Adverse Outcome in Noncardiac Surgery of Long Duration Anesth. Analg., August 1, 2002; 95(2): 273 - 277. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
B. Hartmann, A. Junger, J. Klasen, M. Benson, A. Jost, A. Banzhaf, and G. Hempelmann The Incidence and Risk Factors for Hypotension After Spinal Anesthesia Induction: An Analysis with Automated Data Collection Anesth. Analg., June 1, 2002; 94(6): 1521 - 1529. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Mouren, G. De Winter, S. P. Guerrero, C. Baillard, M. Bertrand, and P. Coriat The Continuous Recording of Blood Pressure in Patients Undergoing Carotid Surgery Under Remifentanil Versus Sufentanil Analgesia Anesth. Analg., December 1, 2001; 93(6): 1402 - 1409. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Junger, B. Hartmann, M. Benson, E. Schindler, G. Dietrich, A. Jost, A. Beye-Basse, and G. Hempelmannn The Use of an Anesthesia Information Management System for Prediction of Antiemetic Rescue Treatment at the Postanesthesia Care Unit Anesth. Analg., May 1, 2001; 92(5): 1203 - 1209. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
