Pitfalls of automated clinical data entry

Neither a borrower nor a lender be.

The number of healthcare institutions installing clinical information systems (CIS) is growing steadily, with providers of these systems often spruiking their time-saving features.

One feature that is of interest to critical care is the ability to automate charting of patients’ physiological observations in hospital records through an interface with patient monitors.

Before installing such systems, their features require careful evaluation and the potential pitfalls need to be understood. One example is our study of automated charting of central venous pressure (CVP) in the intensive care unit using a clinical information system and data quality.

Background

In 1994, the intensive care unit of St Vincent’s Hospital Melbourne introduced the CareVue clinical information system from Philips Healthcare to replace traditional paper and pen charts. In March 2009, CareVue was upgraded to the current Philips product, IntelliSpace Critical Care and Anaesthesia.

It is common for a patient in intensive care to have more than 75 different observations entered into his or her hospital record every hour. To record this information more efficiently, the ‘auto-chart’ functionality of the CIS was activated for some physiological variables. Auto-charting extracts data from the bedside patient monitor and automatically populates the patient’s CIS record.

Auto-charting is a two-step procedure. First, the nurse selects the appropriate chart time on the CIS flow sheet. At this point data are extracted from the bedside patient monitor and the relevant fields populated. Second, the nurse verifies, with a mouse click, the auto-charted values.

During routine data audits, it was noted that some auto-charted values were not physiologically plausible. Auto-charted observations of CVP were of particular concern.

CVP is the blood pressure in the large veins close to the heart. It is typically measured using a transduced catheter placed in the superior vena cava and the measurement is displayed on the patient’s bedside monitor in real time. The unit of measurement is millimetres of mercury (mmHg).

The CVP value reflects the volume of blood in the circulation and the pumping efficiency of the heart. CVP values are used to help guide the treating clinicians in the administration of intravenous fluid and other therapies that support the circulation.

To obtain an accurate measurement of CVP, several factors need to be considered: the calibration of the transducer to atmospheric pressure; the position of the transducer relative to the heart; the position of the patient; use of the CVP catheter for drug and fluid infusions; and the patient’s breathing rate and effort.

The many variables that can affect the accuracy of CVP measurement make it susceptible to measurement error: CVP values displayed on bedside monitors cannot always be taken at face value.

We hypothesised that a CVP measurement read from the monitor by the nurse and manually entered into the CIS would be more accurate than values entered via auto-charting.

Methods

The setting for this study was the 15-bed ICU of St Vincent’s Hospital Melbourne, a university-affiliated tertiary referral hospital. Consecutive patients admitted to intensive care during 2010-2012 after open heart surgery (coronary artery bypass grafts) were identified from a prospective database.

From the start of 2010 until mid-March 2011, the auto-chart function for CVP was active, and from mid-March 2011 to the end of 2012 the auto-chart function was inactive, allowing comparison of CVP measurements in cohorts with and without auto-charting. CVP values documented in the CIS during each patient’s first 24 hours in intensive care were analysed.

Results

The analysis involved 461 patients, of whom, 235 (51 per cent) were exposed to auto-charting of CVP. Several statistically significant differences in the two cohorts were observed.

Compared to CVP measurements documented manually, auto-charted documentation occurred more frequently (median 23 v 22). Also, documented CVP values tended to be higher when auto-charting was used (mean 13.2 v 12.3) and the range of values (the maximum value minus the minimum value) charted for each patient was greater when auto-charting was employed (12 v 10).

Changes in the CVP from one measurement to the next that exceeded 10 mmHg were considered to be potentially the result of measurement error. Such swings occurred for more patients in the auto-chart group (20.4 per cent v 9.3 per cent). Similarly, CVP measurements over 25mmHg were considered to be potentially erroneous, and we found that CVP measurements of over 25mmHg were recorded for more patients in the auto-chart group (21.3 per cent v 10.2 per cent).

Conclusions

Our analysis of CVP data indicates that the method by which the CIS was populated affected the values entered into the patients’ clinical records. Automated CVP charting was associated with greater frequency of documentation, higher documented values, greater clinically significant variation in the documented values and more frequent documentation of potentially spurious values.

Compared to manual charting, auto-charting of CVP could adversely affect clinical decision-making and this could ultimately have an adverse impact on patient care.

A clinical information system provides many benefits, including more legible clinical records, integration with bedside devices and information systems, databases of clinical data, automated reporting, clinical alerts, and the automation of physiological data collection.

Our natural instinct may be to adopt all the features a CIS has to offer, particularly those that save time. However, not everything that can be downloaded digitally is accurate, so be careful what you borrow.

About the authors:

  • David Reid, clinical data analyst, Department of Critical Care Medicine, St Vincent’s Hospital Melbourne
  • Roger Smith, research coordinator, Department of Critical Care Medicine, St Vincent’s Hospital Melbourne
  • Dr Antony Tobin, deputy director, Department of Critical Care Medicine, St Vincent’s Hospital Melbourne
  • Associate Professor John Santamaria, director, Department of Critical Care Medicine, St Vincent’s Hospital Melbourne

Posted in Australian eHealth

Comments   

# Philip Massarelli 2015-02-25 08:30
Imagine that - garbage in - garbage out phenomenon. I believe a lot of technology is also taken for granted. There is much truth to "old school" methods for certain measurements. The Laws of physics do apply!


Thanks for the great read.
# Terry Hannan 2015-02-25 22:20
It should be of interest to the authors in terms of system design in CRITICALLY ILL patients they should read documentation from the HELP system in Utah in 1990 and other aspects of this system.

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2245546/

You need to log in to post comments. If you don't have a Pulse+IT website account, click here to subscribe.

Sign up for Pulse+IT eNewsletters

Sign up for Pulse+IT website access

For more information, click here.

Copyright © 2017 Pulse+IT Magazine
No content published on this website can be reproduced by any person for any reason without the prior written permission of the publisher.