a b s t r a c t

Introduction: Forecasting emergency department (ED) visits is a well-studied topic. The importance of under- standing the complexity of patients along with the days and times of varying patient volumes is critical for plan- ning medical and ancillary staffing. Though multiple studies stratify their results based on Severity of disease, severity was determined by triage status. The goal of this study was to utilize a Novel method to evaluate the cor- relation between daily emergency department patient complexity, based on Current Procedure Terminology (CPT) code, and day of the week.

Methods: This was a retrospective study of subjects presenting to the ED between January 1, 2010 and December 31, 2015. We identified the correlation between subjects with each CPT code who were evaluated on a specific day of the week and evaluated the day before, the day of and the day after a legal holiday.

Results: During the study period 312,550 (48%) male and 336,348 (52%) female subjects were identified. No cor- relation between daily ED patient complexity, based on CPT code, and day of the week (p = 0.75) or any legal holidays were identified. Individual significant differences were noted among day of the week and particular CPT code as well as legal holiday and particular CPT code with no appreciable trend or pattern.

Conclusions: There was no correlation between daily ED patient complexity based on CPT code and day of the week or daily ED patient acuity and legal holiday. In light of these data, emergency department staffing and re- source allocation patterns may need to be revisited.

(C) 2017

Introduction

Forecasting emergency department (ED) visits is a well-studied topic and one with obvious practical implications. As American EDs are inundated with increasing patient volumes, the importance of un- derstanding the complexity of patients, along with the days and times of varying patient volumes, is critical for planning medical and ancillary staffing. The American College of Emergency Physicians has referred to the problem of overcrowding in the ED as a “crisis.” [1] Though the causes of this predicament are multi-factorial and include issues such as boarding inpatients in the ED, staffing may also significantly affect throughput in the ED. Multiple studies reviewing volume based on such factors as day of the week, proximity to holiday, air temperature, pollution index, and others have been published.

Many investigators have attempted to forecast ED visits. In a study in Singapore, Yan Sun et al. used Time series analysis to predict ED work- load based on temporal factors as well as weather and ambient air qual- ity. This data was stratified based on a three-pronged acuity scale

* Corresponding author at: Department of Emergency Medicine, Staten Island University Hospital, 475 Seaview Avenue, Staten Island, NY 10305, United States.

E-mail address: [email protected] (B. Hahn).

determined upon presentation to the ED. They concluded that patient volume in the ED is a predictable variable by time though not by tem- perature. They further determined different patterns based on varying levels of acuity. Specifically, P1 (most acute) arrivals were not temporal- ly related while P2 and P3 (less severe) arrivals were more highly corre- lated with day of the week, month of the year, and holidays [2]. Another Singapore study by Ong et al. similarly determined that patterns of pa- tient volume based on temporal factors are highly predictable. They have adjusted their staffing accordingly [3]. An earlier Australian study by Richardson et al. designed primarily to determine the relationship between ED activity and triage category concluded that there is no cor- relation between these two elements. However, like other studies, they did note a temporal pattern to ED volume in general [4].

A 2005 UK study by W.G. Atherton et al. studied admissions to a trauma unit based on temperatures, sunshine hours, day of the week and month of the year, and determined correlation of trauma admis- sions to both weather and temporal factors. According to this study, Monday was found to be the busiest day for all adult admissions [5].

An Israeli study by Rotstein et al. used a time series analysis over 3 years to predict patient volume based on day of the week, month, and type of day (holiday vs. regular day). They noted a lower patient volume on holidays [6]. Tandberg et al. similarly used time series

http://dx.doi.org/10.1016/j.ajem.2017.08.023

0735-6757/(C) 2017

B. Hahn et al. / American Journal of Emergency Medicine 36 (2018) 366-369 367

analysis to provide accurate ED volume forecasts in an American ED. In this study, acuity was also determined and stratified. The authors con- cluded that there are predictable temporal variations that can practical- ly affect staffing decisions. Unlike some of the above studies, they determined that patient acuity monitoring does not contribute useful information for staffing purposes [7]. McCarthy et al. employed an ex- tensive list of variables in predicting ED volume. These include ones seen frequently in other studies such as temporal, seasonal, climatic, and triage variables but also include novel variables such as patient age, gender, mode of arrival, Ambulance diversion status, and insurance status. The most important predictor of ED volume was determined to be hour of day. Additionally, there was found to be an increase of patient visits on Mondays, a trend seen in other studies as well [8].

An Australian study by Champion et al., compares statistical models for forecasting number of patients by month in a regional ED. They de- termined little variation by month and found that the various Statistical models employed yielded similar results [9]. Jones et al. also compares multiple statistical models for best accuracy in determining ED visits based on seasonal and Weekly variation. Much like the above studies cited, they found seasonal and weekly patterns of patient volume. For example, there was a direct correlation between maximum daily tem- peratures and daily ED volume. They also determined that more ad- vanced statistical analysis tools did not improve prediction of this pattern as compared to standard statistical models [10]. Schweiger et al. reports a diurnal variation in busiest times in the ED during the day with the highest volume, predictably, later in the day. Both statisti- cal models that they employed served to forecast these variations well. Multiple studies have attempted to identify severity of disease in ad- dition to hourly and daily patient volumes. The thought process behind this approach is that variation in volumes may be meaningless if the complexity is all very high or low. For example, a day with lower patient volumes but much higher complexities may require more staffing due to greater time demand and resource utilization, as opposed to a day with higher patient volumes but all with much lower patient complex- ities. Although these studies stratified their results based on severity of disease, severity was determined by triage status. There are no studies to date that correlate ED resource utilization and actual severity of dis- ease determined after the patient has been evaluated and has had a dis- position established. The Current Procedure Terminology (CPT) is a code set used to standardize the description of medical services ren- dered based on complexity. Meaning, a low 99281 CPT code would cor- respond to a visit for an uncomplicated insect bite while a more complex 99285 corresponds to visits for such entities as severe burns or chest pain, requiring more extensive diagnostic and therapeutic in- terventions. Upon disposition, CPT codes are assigned to each patient’s chart and designating the complexity. Therefore, CPT codes are a precise

and standardized way to classify patient acuity.

The goal of this study was to evaluate the correlation between daily ED patient complexity, based on CPT code, and day of the week. Second- ary goals include determining the correlation between daily emergency department patient acuity and legal holidays.

Methods

Study design

This was a retrospective, observational study of all patients present- ing to the ED between January 1, 2010 and December 31, 2015. This study was approved by the Institutional Review Board at our institution.

Study setting and population

The study was conducted at Staten Island University Hospital, a 700- bed, tertiary-care teaching hospital in Staten Island, NY. The ED has sep- arate adult and pediatric departments.

Information was extracted from a computer database. All adult pa- tients over 18 years of age were included. Patients were de-identified and assigned a study number. The CPT code for each patient visit, along with demographic data, including age and sex, was recorded. ED CPT codes, in order of increasing patient complexity include 99281, 99282, 99283, 99284, 99285. Critical care CPT codes are time-based and include 99291 and 99292. At our institution, CPT coding is per- formed entirely by LogixHealth (Bedford, MA), an independent coding and billing service. LogixHealth utilizes proprietary software to ensure accuracy and consistency in coding charts.

Outcome measures

The database was evaluated to determine the number of subjects with each CPT code who were;

[1] evaluated on a specific day of the week and [2] evaluated the day before, the day of and the day after a legal holiday. Labor Day, Memorial Day and Thanksgiving all occur on the same day of the week each year. Therefore, the average number of visits over the entire study period was analyzed for these holidays. Since Thanksgiving is the only holiday which routinely occurs on a Thursday, the Monday after Thanksgiving was also analyzed. As Christmas, New Years and Independence Day all occur on different days of the week, each year was analyzed individually.

Data analysis

Summary statistics for demographic characteristics are presented for all patients. Categorical data is summarized using frequency counts and percentages. Continuous variables are summarized by descriptive statistics, including mean and standard deviation.

The primary objective of the study was to evaluate the association between day of the week and ED patient complexity. The association of ED patient complexity with legal holidays will be also explored. The Chi-square test was used to evaluate the general association of each of the independent variables and ED patient complexity and as a test of heterogeneity among multiple proportions.

All statistical tests are two-sided and conducted at the 0.05 level of significance. Data analyses was conducted using SAS (Statistical Analy- sis System) software Version 9.3.

Results

During the study period between January 1, 2010 and December 31, 2015, 644,532 adult patient visits over 18 years of age were identified. All subjects were included in the final analysis. 310,209 (48%) were male and 334,323 (52%) were female. The mean age was 49.40 (stan- dard deviation, 19.80).

Table 1 displays the total patient volumes per day over the study pe- riod as well as the distribution of CPT codes on each day.

Days of the week with the highest patient volumes were Friday then Saturday. Days with the lowest patient volumes were Monday then Sunday. No statistically significant difference was noted between pa- tient volume and day of the week. The trends in CPT codes based on day of the week can also be seen in Fig. 1.

Days of the week that demonstrated a significant difference in CPT codes are listed in Table 2. Legal holidays that demonstrated a signifi- cant difference in CPT codes are listed in Table 3. In these tables, CPT codes corresponding to each individual day are compared to the same CPT codes corresponding to a composite of all other days of the week. P values in each table are only listed for significant differences. All other comparisons were not statistically significant. Since documenta- tion of CPT codes 99281 and 99292 were so small as to approach zero, these codes were not significant.

368 B. Hahn et al. / American Journal of Emergency Medicine 36 (2018) 366-369

Table 1

ED CPT code frequency by day of week.

Discussion

This 6-year retrospective study evaluated the relationship between patient volume, complexity, and the temporal factors of days of the week and relationship to holiday. Our study of 644,532 patients did not find a statistically significant pattern between these factors. There were individual significant differences noted among day of the week and particular CPT code, though they appeared to be arbitrary and with no appreciable trend. Similarly, while there were some significant differences noted with respect to relationship to legal holiday, there did not appear to be any trend, or consistency among the various holidays. As mentioned previously, the causes of overcrowding are multi- factorial and include issues such as boarding inpatients in the ED. This

Fig. 1. CPT trend by day of week.

study focuses solely on patient volume and complexity, which together appear to significantly impact ED throughput. Earlier studies similarly attempted to evaluate patient complexity, and presumably resource uti- lization, as it relates to various ED characteristics. As opposed to these studies which determined patient complexity based on triage category, we used CPT code to make this designation. The thought process for evaluating patient complexity based on CPT code is that variations in volumes may not be as meaningful if the complexity of the patients is all very high or low. For example, a day with lower patient volumes but much higher complexities may require more staffing due to greater time and resource utilization, as opposed to a day with higher patient volumes but all with much lower patient complexities. Our approach may be more precise, because unlike other studies, our complexity was determined after complete physician evaluation and intervention. We believe that this is a more accurate method of assessing resource utilization required for an individual patient. A patient may seem com- plex upon arrival but this may not ultimately be the case. Triage catego- ry is an excellent method in attempting to predict what resources will be required, however CPT coding can accurately quantify the degree of resource utilization that was actually required. Furthermore, in the case of critical care patients, CPT coding can even identify the amount of time was required by a patient. To our knowledge this is a novel and more refined approach which has not previously been utilized.

This set of data has obvious practical implications, made even more applicable by the long study period and large volume of patients which were included. Prior studies and general ED sentiment is that Monday is the busiest day of the week. An interesting aspect of the results of this study is that they do not tend to mirror previous studies in relation to daily volumes. These findings, along with the lack of trend in terms of patient acuity and temporal factors, may have important ramifications for staffing and resource allocation decisions. These decisions do not re- late only to staffing the ED in general, they are also applicable to staffing in particular sections of the ED. For example, if higher acuity volume is noted on a particular day, an ED could more heavily staff the “critical care” area on that day.

B. Hahn et al. / American Journal of Emergency Medicine 36 (2018) 366-369 369

Table 2

Day of week and average percent of emergency department CPT code.

	99282				99283				99284				99285				99291
	Day	Ref	p	Change	Day	Ref	p	Change	Day	Ref	p	Change	Day	Ref	p	Change	Day	Ref p Change
Monday	3.17	3.17			24.23	24.12			30.54	30.63			37.86	38.05			3.46	3.35
Tuesday	3.01	3.19	0.004	-0.18	24.47	24.08	0.01	+0.39	30.60	30.62			37.77	38.07			3.42	3.36
Wednesday	3.22	3.16			24.25	24.12			30.84	30.58			37.84	38.06			3.23	3.39 0.01 -0.16
Thursday	3.25	3.15			24.28	24.11			30.48	30.64			37.76	38.07			3.48	3.35
Friday	3.13	3.17			24.10	24.14			30.36	30.66			38.44	37.96	0.00	+0.48	3.28	3.38
Saturday	3.20	3.16			23.82	24.19	0.01	-0.37	30.36	30.66			38.58	37.93	b0.001	+0.15	3.41	3.36
Sunday	3.19	3.16			23.80	24.19	0.01	-0.39	31.15	30.53	b 0.001	+0.62	37.90	38.05			3.31	3.38

Day = day under evaluation.

Ref = reference (percent of same CPT code for other days of week). P= p value.

It is unlikely that “downcoding” (assigning a lower value code to a higher acuity patient visit) contributed in any meaningful way to the re- sults of this study. The charts in this study were, as per routine, reviewed for this with no resultant significant findings.

There are limitations to this study which must be noted. First, this is

a retrospective study which, like all such studies, is inherently less valu- able than a similar prospective trial would be. Also, this is a single center study which contributes to decreased applicability of data to other sites. Another limitation is that there are many potential factors that con- tribute to ED overcrowding, resulting in staffing and resource needs. Various studies have attempted to identify which elements likely con- tribute to these factors. However, this study endeavors to evaluate just

one additional element which is that of patient complexity.

It should be noted that although CPT coding is a more accurate mea- sure of patient complexity than the initial nurse triage, there are limita- tions to this method. CPT coding is dependent on documentation and may not always accurately reflect the resource utilization required by the patient. Furthermore, charts may be coded to maximize reimburse- ment based on documentation.

Furthermore, due to limitations of the database from where data was extracted, individual shifts or times of day were not able to be

Table 3

Legal holiday and significantly different emergency department CPT code.

	99283					99285
	Day	Ref	p	Change		Day	Ref	p	Change

Labor day (all years)

Day before 33.56 37.91 0.008 -4.35

Day of 28.86 24.2 0.002 4.66

Memorial day (all years)

Day before 27.16	23.79	0.022	3.37	33.45	37.91	0.006	-4.46
Day of 28.95	24.2	0.001	4.75	33.5	37.84	0.008	-4.34
Christmas (2013)
Day before				28.4	36.86	0.043	-8.46

New years (2010)

Day of 39.03 28.21 0.007 10.82

Independence (2014)

Day of 32.03 22.04 0.01 9.99

Day = holiday under evaluation. Ref = reference day for rest of year. P= p value.

analyzed independently. It is possible that significant differences may have been identified if this analysis was possible and may be a source for future investigation.

Finally, there is a lack of corroboration between our study and simi- lar studies which may place the results into question. For example, Mondays as the busiest day was found to be a true phenomenon in other studies while not in ours. Similarly, in other studies holidays were noted to be busier in general while in our study, no such trend was seen. There are likely numerous reasons for these discrepant re- sults, and those would be worthwhile to explore in future studies.

Conclusions

In summary, this study did not find a correlation between daily ED patient complexity based on CPT code, and day of the week. Further- more, no correlation was determined for the secondary endpoint of daily ED patient acuity and legal holidays. In light of these data, the prac- tice of emergency department staffing and resource allocation patterns geared towards individual days may need to be revisited.

References

1. ACEP. Emergency Medicine Crowding and Boarding//ACEP. https://www.acep.org/ content.aspx?id=32050; 2016. [Accessed March 28, 2017].
2. Sun Y, Heng BH, Seow YT, et al. Forecasting daily attendances at an emergency de- partment to aid resource planning. BMC Emerg Med 2009;9(1).
3. Ong MEH, Ho KK, Tan TP, et al. Using demand analysis and system status manage- ment for predicting ED attendances and rostering. Am J Emerg Med 2009;27(1): 16-22.
4. Richardson D. No relationship between emergency department activity and triage categorization. Acad Emerg Med 1998;5(2):141-5.
5. Atherton W, Harper W, Abrams K. A year’s trauma admissions and the effect of the weather. Injury 2005;36(1):40-6.
6. Rotstein Z, Wilf-Miron R, Lavi B, et al. The dynamics of patient visits to a public hos- pital ED: a statistical model. Am J Emerg Med 1997;15(6):596-9.
7. Tandberg D, Qualls C. Time series forecasts of emergency department patient vol- ume, length of stay, and acuity. Ann Emerg Med 1994;23(2):299-306.
8. Mccarthy ML, Zeger SL, Ding R, et al. The challenge of predicting demand for emer- gency department services. Acad Emerg Med 2008;15(4):337-46.
9. Champion R, Kinsman LD, Lee GA, et al. Forecasting emergency department presen- tations. Aust Health Rev 2007;31(1):83-90.
10. Schweigler LM, Desmond JS, Mccarthy ML, et al. Forecasting models of emergency department crowding. Acad Emerg Med 2009;16(4):301-8.