Qual Saf Health Care 18:169-173 doi:10.1136/qshc.2008.029827
  • Original research

An exploratory study measuring verbal order content and context

  1. D S Wakefield1,
  2. J Brokel2,
  3. M M Ward3,
  4. T Schwichtenberg4,
  5. D Groath4,
  6. M Kolb5,
  7. J W Davis6,
  8. D Crandall7
  1. 1
    University of Missouri Center for Health Care Quality, Department of Health Management and Informatics, Columbia, Missouri, USA
  2. 2
    University of Iowa College of Nursing, Iowa City, Iowa, USA
  3. 3
    University of Iowa Department of Health Management and Policy, Iowa City, Iowa, USA
  4. 4
    Mercy Medical Center of North Iowa, Mason City, Iowa, USA
  5. 5
    University of Missouri Center for Health Care Quality, Columbia, Missouri, USA
  6. 6
    University of Missouri Department of Health Management and Informatics, Columbia, Missouri, USA
  7. 7
    Trinity Health, Novi, Michigan, USA
  1. Dr D S Wakefield, Center for Health Care Quality (CHCQ), MA120 Medical Sciences Bld., DC375.00, One Hospital Drive, University of Missouri, Columbia, MO 65212, USA; wakefielddo{at}
  • Accepted 1 December 2008


Background: The use of verbal orders, while essential in some healthcare settings, has been identified as a potential contributor to poor quality and less safe care. Despite the widespread use of verbal orders, little research attention has been paid to understanding and measuring the content of verbal orders or variables related to the context in which verbal orders are made.

Aim: This paper first identifies variables related to verbal order content and context, and then provides detailed analyses from two exploratory studies conducted in one community hospital.

Methods: The data presented were collected using both a paper-based manual audit, and an analysis of data generated from a computerised order entry system.

Discussion: Selected analyses focus of variations in types and timing of verbal orders hospital-wide as well as for specific inpatient units, changes in verbal order utilisation following implementation of a computerised provider order entry system, and an analysis of the presence of sound-alike and high-alert medications in verbal orders.

Patient care information and orders are commonly communicated verbally in a variety of healthcare settings including inpatient hospitals, nursing homes, ambulatory clinics and hospice.18 Verbal orders (VO) can include both face-to-face and telephone communications between a provider making the order (ie, physician, APN, PA) and a person receiving the order (ie, nurse, pharmacist). Because of the potential for information conveyed verbally to be incorrectly spoken, heard, understood or transcribed, and thus not carried out as intended, increasing emphasis has been placed on reducing the use of VO, and improving VO communication processes when VO are used.812 For example, a study of errors in outpatient computer entered laboratory orders found that about 5% contained errors, and the authors subsequently suggested substituting written and facsimile orders for VO.13

Despite the intuitive logic that VO have the potential to cause errors, to date there has been little systematic measurement or evaluation of how the content of the VO or the context in which VO are made might increase error. It is unknown to what extent VO represent a risk to patient safety.14 Inherent in understanding the potential for VO-related errors to occur is an understanding of the relationship of human factors to error occurrence. An enhanced understanding of the variability of the VO content and context is needed before interventions focusing on decreasing the risks associated with VO are implemented. Thus, it is important to examine the variability of VO content and context, such as the number of unique orders per communication episode, and the content of the orders, especially whether medication-related or not, the time of day in which a VO is given and the mode of delivery of VO.

Prior to the introduction of electronic medical records and computerised provider order entry systems, the only way to obtain VO related data was to conduct manual chart audits. Thus, there have been very few studies which have included detailed VO data.13 With the introduction of electronic data capture systems it is far easier to collect and analyse detailed VO data, with an increase in studies reporting VO data.23 With this increased ease in VO data capture, attention can be directed to more detailed evaluations of VO content and context variables. This paper provides the result of two exploratory studies developing conceptual definitions and operational measures of VO content and context characteristics that may contribute to medical error.


Two types of VO characteristics potentially contributing to errors include “VO content” and VO context. As used here, VO content refers to the numbers and types of individual orders being communicated. VO context refers to the general conditions or circumstances under which a VO is given, and the processes used to communicate, verify and transcribe the orders.

In inpatient settings, the content of VO can vary greatly in complexity, which in turn may contribute to error.3Table 1 provides examples of several different VO content characteristics including: the numbers and types of different orders being communicated, order complexity, order urgency, abbreviations and non-standard terminology, and the use of sound-alike drugs and/or doses along with high-alert medications.

Table 1 Examples of verbal order (VO) content and context variables

In a similar manner, VO may be used differently across different inpatient and outpatient care settings units, vary by time of the day, involve a variety of personnel and be conveyed by a number of communication modes. In addition, characteristics of both the personnel and environment may influence VO use, transmission and accuracy. Characteristics of the personnel involved include how well the provider giving the VO and the person receiving it know each other and the patient involved. Environmental characteristics that can play a key role in VO include ambient noise levels, background activity levels and the degree to which the VO is an interruption. Thus, the context in which VO are given can be highly variable. Table 1 identifies five types of VO contextual variables potentially contributing to medical error: type of care setting; timing communication processes; personnel characteristics; and environment. Asterisked items represent those which may lend themselves to automatic electronic data capture.

To stimulate understanding of the nature and some of the underlying variability of VO content and context characteristics, we report here the results of operational measures for two exploratory studies: one using manual medical record abstracting and one using a computer-based approach.


Study site

The results of both exploratory studies presented here are drawn from one Midwestern, large not-for-profit rural referral hospital. The hospital has approximately: 13 000 inpatient discharges and 420 000 outpatient visits annually, 2750 employees, 260 staffed beds and 293 physicians on the medical staff. Nearly all physicians are in private practice and are not employed by the hospital. While there are both hospitalists and intensivists in the hospital, they are not universally used by all physicians to manage inpatient admissions. There is a Family Medicine Residency programme, and the hospital also serves as a clinical training site for cardiology residents from other institutions, as well as nursing and other allied health students. With the exception of the Family Medicine Residency programme, on-call coverage is handled by the attending physicians through traditional on-call scheduling. Because of a transition from a paper-based to computerised provider order entry (CPOE) system, this hospital provided a unique opportunity to study VO utilisation and characteristics in both paper-based and computer-based environments.

Data collection and analysis

The study received IRB approval and the data were collected for the first exploratory study using manual paper-based medical record abstracting, and for the second study analysing an electronic computerised database which pulled data from a CPOE system. Data collected in both exploratory studies included selected VO content and context characteristics included in fig 1. We limited the analysis to VO given by physicians and received by nurses in the manually abstracted data collection. Limiting the focus to only physician-nurse VO does not include VO given directly by physicians to pharmacists, and thus will understate VO utilisation. In using the computerised database, this limitation has been eliminated because the VO taker is automatically identified for each order.

Figure 1

Summary of mean, upper and lower confidence limits of the probability of orders being entered using verbal orders by specialty.

VO were operationally defined to include all telephone and face-to-face patient care orders that: (a) the ordering physician communicated verbally (eg, by telephone or face-to-face); (b) required transcription to the patient’s medical record and to the hospital’s order forms by the nurse receiving the order; and (c) required a physician signature at a subsequent time to validate the order. Thus, for demonstration purposes, we have combined telephone and face-to-face VO into a single category. Because many different individual VO may be communicated for one patient during the same verbal communication event (eg, start drug A, stop drug B, order lab tests C–E), we analysed data at the VO even, and for unique VO levels of analysis.

The paper-based audit tool development included several iterations in content specification, coding, pilot testing and training the three RN abstracters and has been described in detail elsewhere3 (supplemental tables). During May, 2005, a team of researchers and quality improvement practitioners developed and used the paper-based tool to audit 1 week’s (7 days) worth of VO data,3 prior to a planned CPOE implementation.15 Because of the labor intensity of the data collection we did not collect data related to all orders (eg, both written and verbal communicated orders). Double data entry was used to verify data entry accuracy.

Following the CPOE implementation the computerised database was used to collect data for 12 months (August 2005 to September 2006) worth of total orders (ie, computer entered, VO and written). Per IRB approval, all reported data were deidentified and aggregated.


Exploratory study 1: paper-based audit tool of VO applications

Using the manual chart review, we conducted exploratory studies of the number of VO use, medication content of VO, non-medication content of VO, VO use by time of day and VO use across inpatient care units.

Number and communication mode of VO use

For the 7-day audit period, we identified 1522 VO events containing 8171 unique VO for specific patient care services (table 2). The number of unique VO per VO event ranged between 1 and 28, with a mean of 5.4 (SD = 6.1). Approximately 43% of VO events involved face-to-face communications, 33% used telephone communication, and for 24% the communication method was not available.

Table 2 Paper-based audit of verbal orders (VO)

Medication content in VO

Medication-related orders (eg, starting new medications, changing medication doses, and stopping medications) were by far the most common single type of order accounting for 51% of all unique VO (N = 4199) and were found in about 80% (1226) of the VO events (table 3). The mean number of individual medication-related orders per VO event was 3.4 (SD 3.7), with 28% of VO events containing four or more individual medication-related orders. Approximately 29% of VO events with unique medication-related orders also contained orders for range or conditional medication administration, which increases the complexity of the order.3

Table 3 Paper-based audit of medication-related verbal orders (VO)

Non-medication content in VO

Non-medication-related orders such as orders for diagnostic testing, treatments or therapies, dietary changes, or intravenous and blood products accounted for half of the total number of individual VO, and were found in 62% of all VO events (table 4). Over half were for non-medication treatments (eg, physical therapy, respiratory therapy), and 30% of non-medication related orders were for diagnostic services (eg, lab, radiology).

Table 4 Paper-based audit of non-medication-related verbal orders (VO)

VO use by time of day

To analyse VO patterns by time of day, we divided the 24 h in a day into six 4 h time intervals (eg, 0:00–03:59, 04:00–7:59, etc). The total numbers of VO events and unique VO were the lowest during the 0:00–03:59 and 04:00–07:59 time periods (table 5). For example, in the 0:00–03:59 time period, there was an average of 7.8 unique orders per VO event, and of the 91% of VO events containing medication-related orders there was an average of about 4.4 medication related orders per VO event. Surprisingly, the two time periods (08:00–11:59 and 12:00 to 13:59) when one might expect it is more likely for physicians to be physically present and write orders accounted for 53% of all VO events and 51% of unique verbal orders. The high number of VO, combined with the high percentage of face-to-face VO made during this time period (data not reported) suggests that VO may be used more for convenience than for medical necessity.

Table 5 Verbal order (VO) audits analyses by time of day

VO use in different care units

For the 7-day period, we also found great variation in the use of VO by different care units. The number of VO events ranged from a minimum of 19 on Cardio Vascular Recovery to a maximum of 344 on the Cardiac Step Down Unit; total unique VO ranged from 135 on Acute Rehab to 2002 on Cardiac Step Down Unit. Likewise, the mean number of individual orders per VO event ranged from 2.8 for the ICU up to 8.7 in the Birth Center. The average number of VO events per staffed bed ranged between 0.8 for Acute Rehab and 9.8 for Ortho/Neuro, with the mean number of unique VO per Staffed Bed ranging from 3.6 for Acute Rehab to 6.5 for the Ortho/Neuro. There was also significant variation among patient care units in terms of VO being communicated by telephone (9% for the ICU vs 58% for Cardio-Vascular Recovery), face to face (5% for the ICU vs 80% for the Birth Center) and not being documented (2% for the Birth Center vs 86% for the ICU).

Exploratory study 2: CPOE-based audit of VO applications

We used the CPOE system to examine VO use over time by different inpatient units and use of sound-alike and high-alert medications in VO.

VO use over time

For each physician, we analysed monthly totals and percentages of orders that were electronically entered using the CPOE system, given verbally or by telephone, or handwritten on the inpatient care units. For the 12-month period there were a total of 972 560 unique physician-generated orders with about 20% (n = 200 394) being VO. There was marked variation in the use of VO by physician specialty and resident status. To analyse changes in VO patterns over time, we grouped physicians by medical or surgical specialty and resident status. Figure 1 provides a graphical presentation of the mean probability of different groups of physicians using VO, along with upper and lower confidence limits (alpha = 0.05). The results clearly suggest great variability in VO use rates and that changes in VO use following implementation of CPOE also varied over time among different physician groupings. Surprisingly, with one exception (Urology), physicians in this hospital showed little change in the use of VO over time. That is, the introduction of CPOE did not appear to have a widespread influence on VO use.

Sound-alike and high-alert medications in VO

Because medications constituted such a large portion of the total VO, we were interested in identifying VO used in the ordering of sound-alike or commonly confused drug name, and high-alert medications. The Institute for Safe Medication Practices (ISMP), in recognising that sound-alike or commonly confused drug names and high-alert medications represent important medication safety concerns, has published lists for each.1617 For a pilot examination, we looked at the VO during a 1-week period for a group of non-invasive and invasive cardiologists. For that 1 week period, there were 5276 individual orders, including 2251 (43%) medication-related orders. Among medication-related orders, 454 (21%) were VO, 1273 (56%) were entered by use of the CPOE system, 452 (20%) were handwritten, 64 (3%) were generated by protocol, and eight (1%) were nurse-entered orders. Of the 2251 medication-related orders, 442 (20%) matched medications listed on the ISMP’s list of commonly confused medications, with 93 (21%) of these being VO, 300 (68%) made by the physicians using the CPOE system and 49 (11%) being handwritten. The five most frequently occurring commonly confused medication names that were ordered by VO were Fentanyl, Lasix, Lidocaine 1% Inj, Versed and Dextrose 50% Syringe.

High-alert medications included those identified by ISMP18 and additional high-alert medications of interest in our own facility. Using this combined high-alert list we found that 919 of the 2251 total medication orders (40%) matched the high-alert list. Of the high-alert medications, 189 (20%) were VO, with the most frequent including orders for narcotic and opiate medicines (17%), sedating agents (7%), anticoagulants (7%) and low-molecular-weight heparin (6%).


For the foreseeable future, VO will continue to be used in most care settings because of geographic separation of the order prescribers from where orders can be directly written or electronically entered, clinical necessity and/or prescriber preference and convenience. Because of both their frequency and potential of error and harm, VO merit more systematic and intensive examination.

The VO content and context framework used in this study provides a useful means to gain new insights about the nature, extent and variation of VO use. Using the framework, we collected revealing data on a variety of content and context issues for VO. As shown here, both a paper-based audit tool and computerised data collection yielded important information.

For example, the magnitude of VO use in the study hospital was surprising in terms of VO totals, and the numbers and types of orders being communicated at different times of day. If we assume that 99.4% (ie, 4-Sigma level of performance) of the 200 000 plus VO were each correctly communicated, understood and transcribed, about 1200 of these orders would have had some type of error. Assuming further that about half of all VO are medication-related, at the 4-Sigma performance level there would have been about 600 defective VO medication-related orders made, and potentially carried out. Unknown is whether the VO processes at this hospital actually perform at a 4-Sigma performance level. Clearly, systematic examination of the reliability of the various processes used to make VO, and the effect of the physical environments in which VO are used on the reliability of VO-related communication processes, represents a promising area of future research.

Because most hospitals do not yet have fully functioning CPOE systems, manual data collection of VO content and context data is the near-term option. Manual systems have the advantage of being highly flexible in that there is no reliance on data collected previously for other purposes. Its primary disadvantage is the labour intensiveness of data abstraction, coding and entry, as well as interpretation of handwritten records. The high degree of labour intensity limits expanded and repeated data-collection efforts. For hospitals implementing CPOE systems, computerised databases have a number of advantages including the potential to conduct secondary data analyses on data already collected. Depending on the level of data already being captured, there may be an opportunity to analyse large numbers of individual VO and order types by individual or groups of prescribers, and/or specific types of patients using both cross-sectional and longitudinal analytical techniques. CPOE-driven electronic databases’ primary disadvantages include being limited to previously collected data that were not operationally defined to address the specific questions of interest, difficulty in getting new data elements included in future data-capture efforts, mastering data dictionaries and data-mapping strategies to determine what data are available, and developing new ways of thinking about the specific questions of interest and how to optimally analyse the available data. Finally, to convert the wealth of potential VO-related data into useful information, additional resources will be needed for more advanced statistical analyses. Despite these limitations, the ability to better understand VO-related processes and outcomes will be greatly increased in the CPOE environment.

To advance the study, understanding and management of VO, it is important to use a framework to better understand how both the content of the VO and the context in which VO are made may contribute to both quality of care and reduced patient care safety. Future research is needed to further test, modify and develop frameworks allowing a greater understanding of VO, a commonly used communication approach in a variety of healthcare settings.


The authors thank Mercy Medical Center-North Iowa pharmacists and Practice Quality & Informatics Nurses and B Wakefield for assistance with survey development, administration, and data entry and interpretation.


  • See Commentary, p 164

  • Additional tables are published online only at

  • Funding: This work is supported by funding from AHRQ-THQIT Implementation #1 UC1HS015196 and the University of Missouri Center for Health Care Quality.

  • Competing interests: None.