Computerized provider order entry (CPOE) in health care has been introduced across North America based on a business model suggesting savings from improved patient safety.Reference Pham, Aswani and Rosen 1 This improvement is primarily achieved through forcing functions that alert health care providers to potential medication cross-reactions or allergy risks, and to evidence-based clinical decision rules that may favour particular medications or approaches. Carrying out non-advocated approaches frequently requires manual override by the provider. Pain management and encouragement to adhere to complaint-based protocols for specific conditions, such as renal colic and cerebrovascular accidents, are some examples of potential benefits of CPOE.Reference Netherton, Lonergan and Wang 2 - Reference Blankenship, Rogers and White 4 Most problems with CPOE occur with implementation rather than CPOE itself.Reference Pham, Aswani and Rosen 1
Although certain benefits have been found with CPOE, there has been little research on the impact of CPOE implementation on patient flow. Some research suggests CPOE results in prolonged length of stay (LOS) for admitted patients, reduced physician productivity, and increased time to order entry.Reference Bastani, Walch and Todd 5 - Reference Syed, Wang and Goulard 7 Other research suggests CPOE reduces LOS,Reference Innes, Grafstein and Christenson 8 - Reference Fernandes 9 though the evidence for this is either limited to a specific chief complaint or to Emergency Department Information Systems that impact emergency department (ED) patients only. As ED crowding becomes increasingly problematic, the impact of CPOE on ED flow must be examined to better evaluate whether it is beneficial (e.g., reduced time to order processing) or detrimental (e.g., slower patient disposition decisions).
The purpose of this study was to evaluate the impact of the implementation of a new hospital-wide Canadian CPOE system on three primary ED variables: LOS, wait time (WT), and the proportion of patients who left without being seen (LWBS) by a physician.
London Health Sciences Centre (LHSC), a quaternary care centre in London, Ontario, sees approximately 60,000 ED patients per year through its Victoria campus, and approximately 40,000 ED patients per year through its university campus. It is staffed by certified emergency physicians, with support from two emergency medicine residency programs, and rotating residents from various specialties.
CPOE was introduced in April 2014 under the acronym HUGO (Healthcare Undergoing Optimization). This system impacted all services and disciplines across the organization simultaneously, with education on the new system occurring in the few weeks prior to implementation. HUGO is based on a specific software solution from Cerner Corporation that has been modified for the needs of our institution.
We conducted a retrospective cohort study of all patients of 18 years and older presenting to LHSC EDs in July and August of 2013 and 2014, before and after the implementation of HUGO. We compared key throughput variables (WT, LOS, and LWBS) between groups before and after implementation. Data were extracted from the LHSC electronic database supported by the health records department. The study was approved through the Research Ethics Board of Western University.
Inclusion and exclusion criteria
We included all ED patients ages 18 years and older triaged at the study location during July and August of 2013 and 2014. We excluded patients with incomplete or incorrect ED charts. We excluded patients with negative WTs or negative lengths of stay (indicative of an erroneous triage or assessment time recorded) or extreme outliers where WTs exceeded 24 hours (presumed to represent an erroneously wrong day recorded). Any patients missing vital statistics (such as gender or CTAS) were also excluded.
Definition of variables
The following variables were calculated:
WT (minutes)=time of first physician assessment to time triaged
LOS (minutes)=time of disposition to time triaged
LWBS (%)=number of patients who LWBS/total visits for a given period
Subgroup analysis was performed on patients categorized into six a priori subgroups, which consisted of each individual CTAS level (1–5) and an additional analysis performed on admitted patients only. Each subgroup was compared before and after intervention for all three calculated throughput variables defined previously.
Descriptive statistics were calculated for age, sex, CTAS level, disposition, and total number of patients for each time-period. Variables were examined to determine whether they had a normal or non-normal distribution using the Kolmogorov-Smirnov test. Next, we compared WT and LOS between groups using the Mann-Whitney U test to assess for any overall significant change in these variables for ED patients between each time-period. We compared the LWBS proportion between groups using the chi-square test to assess for any change at each time-period.
Data were entered directly into a password-protected, study-specific Microsoft Excel database (Microsoft Corporation, Redmond, WA). All data analyses were performed using SPSS (V. 22.0, IBM Corporation).
Table 1 provides descriptive statistics of the study population before and after CPOE implementation and illustrates statistics were similar. Table 2 provides all ED throughput variables, suggesting worsening of ED flow after the intervention, though some subgroups were impacted more than others.
CPOE=computerized provider order entry; CTAS=Canadian Triage and Acuity Scale; IQR=interquartile range.
* Statistically significant at p<0.05; CPOE=computerized provider order entry; CTAS=Canadian Triage and Acuity Scale; IQR=interquartile range; LOS=length of stay; LWBS=left without being seen; WT=wait time.
We found that implementation of CPOE detrimentally affected three standard ED throughput indicators. Key findings to note were that, after CPOE, LOS for admitted patients was significantly increased, WT for CTAS 3 patients was increased, and the proportion of patients who LWBS was increased. All of these are indicators of worsening crowding and poor ED throughput.
The manner in which ED crowding contributes to the risk of spread for potentially lethal disease has been noted previously.Reference Crane and Noon 10 In the years since, efforts to better control hospital flow have yielded mixed results. A number of centres have reported dramatic improvements in ED patient flow,Reference Fernandes 9 , Reference Watson and Leeson 11 but this has been predicated on improved inpatient access. As LitvakReference Litvak, Lachman and Leitch 12 has shown, efforts such as Lean ThinkingReference George 13 or Toyota Production SystemReference Liker and Gardner 14 require approximately 80% inpatient bed occupancy to be successful. Many Canadian hospitals, including ours, function with far higher occupancy, sometimes up to 125%, thus any gains from implementation of these processes may be negligible in such settings. As well, increased boarding time in the ED has been found to correlate with increased mortality.Reference Singer, Thode and Viccellio 15 Thus, dramatically worsening bed occupancy within the ED as may be the case from HUGO implementation, without addressing inpatient occupancy, could contribute to further stretching of limited resources, such as nursing and physician time, and have the potential to worsen patient outcomes.
Our results suggest that HUGO increases ED LOS for admitted patients by a median of 63 minutes. This shows the dramatic effect of poor implementation on the process of patient care. It is difficult to argue that gains through efficiencies would override this inefficiency, particularly when WTs for non-admitted patients have also increased. One concern is that resilient emergency physicians could attempt to overcome this lack of capacity by examining patients in nonstandard locations, such as hallways and chairs. This could lead to more risk to the patient because the physical exam likely becomes less reliable in a disadvantaged area.
Application of queuing theory demonstrates the negative effect of HUGO on non-admitted patients. We analysed CTAS 3 patients at Victoria campus, which has approximately 40 ED beds, using the M/M/s model for multiple servers with a single queue,Reference Litvak, Lachman and Leitch 12 where s is the number of servers (beds available), λ is the average number of arrivals per unit of time (approximately four CTAS 3 patients per hour), and μ is the average number of requests served per unit of time (1/WT, in our case). Most of the time, s is 15 at our Victoria campus (the rest of the beds are taken up by admitted patients or CTAS 1 or 2 cases). In this case, using a queuing theory calculator (e.g., supositorio.com), Wq goes from 0.18 minutes pre-HUGO to 0.31 minutes post-HUGO, where Wq is the average waiting time for a CTAS 3 patient to access the next available bed (a 70% increase in inefficiency). When we have only 10 available beds, as may be the case when many admitted patients are occupying beds due to increased LOS, HUGO effects escalate for Wq (14.33 minutes to 22.72 minutes). It is arguable that the primary reason that the system is able to function is emergency physician resiliency efforts, as mentioned previously.
It has previously been demonstrated that it is possible to improve LWBS proportion through use of a Fast-track process.Reference Fernandes, Daya and Barry 16 - Reference Al Darrab, Fan and Fernandes 20 Such a process focuses on CTAS 4 and 5 patients, who otherwise are a very low priority and sometimes even overlooked. The LWBS proportion is primarily a measure of efficiency in the care of CTAS 4 and 5 patients. The Fast-track process is in keeping with modern flow theories that look at streaming of patients to appropriate resources (fewer resources for CTAS 4/5, more for higher acuity patients). In our case, we see that the impact of HUGO implementation on LWBS has been most dramatic for the lower acuity patients. Again, any gains from such processes as a Toyota Production System or Lean Thinking will be lessened through poor CPOE implementation. Another possible concern is that patient satisfaction may have worsened as a result of HUGO. It has been shown that the LWBS proportion is an indicator of patient satisfactionReference Fernandes, Daya and Barry 16 - Reference Fernandes, Price and Christenson 19 ; an increasing proportion of LWBS may be an early signal that a department may need to urgently address patient dissatisfaction before institutional reputation is permanently harmed.
There are a number of limitations to keep in mind when interpreting our results. Our study was not randomized and was conducted at a single academic health centre, using a unique software package. We did not examine various flow issues through specific days and times, and assessed patients during only 2 specific months in 2 consecutive years. This study did not examine solutions that may have been in the process of implementation by LHSC to solve resultant flow issues. Further, we did not look at the impact of HUGO on specific services and disciplines other than emergency medicine.
An important aspect for future research will be examining whether the throughput variables we assess improved over time. It is possible that we are now further down the “learning curve,” and difficulties with efficiency after HUGO have been resolved. Maybe the real lesson here is that CPOE implementation needs to be more thoroughly tested for inefficiencies, and training needs to be more extensive prior to implementation.
We found that CPOE implementation at our health care organization detrimentally impacted patient flow in the ED. All throughput variables were involved, some with greater significance than others. The most striking clinically relevant result we found was an increase in LOS of 63 minutes for admitted patients. Our results suggest that the potential patient safety risks may outweigh the benefits when considering CPOE implementation.
Competing interests: None declared.