To our knowledge, this is the first systematic review to document in a quantified way the time differences between computer- and paper-based documentation among studies that assessed the documentation activities of nurses and physicians. Time efficiency is only one possible outcome for which the success of EHR integration can be assessed, and studies in this review also reported on direct patient care time,13,15,16,37,42,47,48,50
accuracy of the information,13,15,40
completeness of data entered,15,38,41,43,49,50
and the overall impact on workflow.16,22,44,45
However, time efficiency is recognized as an important facilitator or barrier of EHR implementation,3,10,11,18,21,22,23
and consequently needs to be assessed with rigorous methodologies. Only 23 studies (36% of total retrieved) involved a quantitative examination of the integration of EHR into clinical workflow. One possible explanation of the paucity of research may be the limitations associated with the methods available to accurately document the impact of EHR on time. Continuous observation of work processes as captured by time and motion or video-recording methods are seen as the most accurate data collection techniques to monitor clinical activities33
as they provide precise estimates of time spent in each activity. Fifty-eight percent of the reviewed studies used these methods despite the higher costs of one-to-one direct observations. Under the work-sampling technique, data are collected at predefined intervals of time, which allows the observation of multiple individuals by a single observer, which is seen as a major advantage over the time and motion technique. Activities are captured as “snapshots” of professional processes. Single counts of categorized activities do not provide any information on the real time spent performing the activity.54
The overall proportion of time must be estimated using the number of snapshots in one category over the total number of snapshots that were recorded during the work-sampling time period. Recognized as a valid approach to evaluate work patterns,54,55
a major disadvantage of the work-sampling technique lies in its need for very large sample sizes for the time estimates to have an acceptable level of precision, a criterion not often met.33
This limitation of work sampling methods is not likely to influence the conclusions of our review since the errors in estimating time related to the chart documentation would occur in both the computer- and paper-based groups.
Only a few study results44,48
reported in our review had large confidence intervals ( and ), and although sample size may influence the width of confidence intervals, only one had a fairly small sample size (pre/post, 14/39 patients).48
Population characteristics also play an important role in the variability of the data, and, hence, on the width of confidence intervals. For example, the two studies44,48
were conducted in environments (general internal medicine and community clinics) where care delivery is highly variable because of the population's heterogeneity. In comparison, studies conducted in highly specialized settings such as the one by Weinger et al.48
are more likely to have uniform care delivery patterns, less variability across patients and physicians, and therefore narrower confidence intervals, despite a smaller sample size.
Results of this review suggest that nurses are more likely than physicians to gain time efficiencies by using a computer system to document patient information. Several reasons may explain the difference between nurses and physicians. First, nurses and physicians document different types of information. Nurses often document using standardized forms or care plans,56
while physicians rarely use standardized templates to write their clinical notes.
Retrieval or viewing of information is part of the work processes of both nurses and physicians. However, it is much more intricately related to the documentation process of physicians. This may have played an important role in time efficiencies of CPOE systems that combine retrieval, viewing of information, data entry, and, in many cases, responses to alerts and reminders. These additional factors are difficult to capture by time and motion or work-sampling methods as both have limited capacity in capturing simultaneous activities,57
and these may have accounted for the extra time that physicians take to document or enter orders on a computer. Several studies have shown that computers increase the completeness of information being documented.15,38,41,43,49,50,58
This additional information available to physicians will influence the time required to retrieve information,59
and their motivation to use EHRs if part of that information is perceived as unnecessary to their clinical activities.60,61
While both nurses and physicians see the added value of integrating EHR into their daily practice,17,59,62
physicians and nurses differ in their incentives to use the EHR56
and in their speed of adoption.63
These can be influenced by the fact that nurses tend to work in a single location and will therefore be more frequently exposed to the EHR in contrast to physicians who tend to work in several locations, both inside and outside the hospital. The degree of exposure to a newly implemented EHR may influence the learning curve and ability to become an efficient user more rapidly. As employees of a health care organization, nurses may be more likely to receive support from clinical leaders and paid training sessions, both of which have been identified as essential requirements for EHR adoption.64
The autonomy and accountability of nurses and physicians are different and may influence their performance.65
Those may explain why nurses tend to be more time efficient than physicians. Both groups also differ in their work processes. For example, nurses are part of a care team and need to verbally transmit information to their colleagues at the end of their working shifts. The use of computers has been shown to reduce the time devoted to the end-of-shift report,13
and this change in workflow may have been a strong incentive for nurses to become efficient users of the system. Our results support this assumption, with all studies examining the impact of EHR over working shift periods, reporting favorable time efficiencies compared to those with patients or patient encounters as the sampling units. In our review, all studies on physicians, except for CPOE studies, used patients as their unit of analysis and most reported an unfavorable impact of the EHR. Time gains, at the patient level, may be difficult to achieve and examining the impact of EHR time on the overall clinic or hospital day may have yielded different results for physicians.
It was surprising to see that studies that observed clinicians relatively soon after implementation time (three months or less) showed a slight reduction in documentation time, while those that waited longer tended to show increases. It is possible that once clinicians become familiar with the system, they begin to take advantage of its other functionalities and thus may appear to be less efficient. Another reason may be that most projects have intensive support in the early implementation phase and that support may decrease over time. The optimal time period for assessment of time efficiencies post-implementation of EHRs remains a challenge and will require further research.
To understand the role that system use may play in time efficiencies, standardized audit trail information needs to be collected that would allow assessment of the extent to which individual components of a system are used. This review clearly highlighted the absence of any consistency or agreement on a standard time period after which a system should be tested. In fact, 25% of the studies in our review neglected to mention the time period in which the evaluation was performed despite the importance of this time period on adoption, use, and efficiency rates.15,16,46
We attempted to characterize the different EHR systems reviewed in this paper in a systematic way and reported for each system the location (bedside or central station), data entry format (structured, free text, keyboard, touch screen), and the main functionalities (POE, complete clinical notes). Obviously, other characteristics such as the number of available fields in the EHR that one must navigate to enter data and the speed of the computer were not systematically reported and would likely play a major role in the clinician's time efficiency. For informed and valid comparisons of time efficiency within and across studies, timed standardized tasks would be helpful in establishing baseline expected efficiencies as some EHRs may not have the capacity to be time efficient in comparison to paper charting, regardless of the user or the environment. Knowing this information prior to EHR implementation will influence the deployment and training strategies. The focus on time efficiency should then be oriented toward the overall processes of care delivery rather than toward the potential time gains in performing specific activities, like documenting or ordering tests.
Limitations of the Study
Only 23 papers met our selection criteria despite the fact that we examined all papers published since 1966 and did not limit our search to RCTs. The concept of time efficiency is reported in the health informatics literature through quantitative or qualitative results or anecdotal evidence, but our focus was on quantitative results only. The inclusion of the numerous qualitative or anecdotal evidence studies may have provided valuable information to this area of knowledge but would have prevented summarizing the results in a quantitative way, which we thought was highly informative. A wide range of EHR systems were covered in this review (from POE to full clinical notes system). We grouped time differences on the basis of users, systems, and sampling unit as we could not assume that, for example, a 10% increase in time efficiency per patient would be the equivalent of a 10% increase for the total working shift. Different grouping approaches may yield different time averages, but the overall direction of results, time efficiency versus time inefficiency, should remain the same.
We recognize that papers included in this review cover a ten-year time period during which technology was rapidly evolving. Combining results of studies conducted in the 1990s with studies from the early 2000s may be debatable. However, our results did not identify a clear trend toward enhanced time efficiency despite the increased speed of computers, the availability of customized software, and the large array of user interfaces and input devices. The role of factors that are external to the information systems in contributing to the time efficiency of clinicians needs to be better understood. The methods used in the selection and review of these papers did not allow us to examine the impact of these factors. Further studies are required to examine the role of clinicians, professional practice, and organizational environment in facilitating or not the efficient use of EHRs.