Original Contribution

Using demand analysis and system status management for predicting ED attendances and rostering

Marcus Eng Hock Ong MD, MPH^a,?, Khoy Kheng Ho MD^a, Tiong Peng Tan MD^a, Seoh Kwee Koh^a, Zain Almuthar^b, Jerry Overton^c, Swee Han Lim MD^a

^aDepartment of Emergency Medicine, Singapore General Hospital, 169608 Singapore

^bService Operations Department, Singapore General Hospital, Singapore

^cRichmond Ambulance Authority, Virginia, USA

Received 12 December 2007; revised 7 January 2008; accepted 7 January 2008

Abstract

Introduction: It has been observed that emergency department (ED) attendances are not random events but rather have definite time patterns and trends that can be observed historically.

Objectives: To describe the time demand patterns at the ED and apply systems status management to tailor ED manpower demand.

Methods: Observational study of all patients presenting to the ED at the Singapore General Hospital during a 3-year period was conducted. We also conducted a Time series analysis to determine time norms regarding physician activity for various severities of patients.

Results: The yearly ED attendances increased from 113387 (2004) to 120764 (2005) and to 125773 (2006). There was a progressive increase in severity of cases, with priority 1 (most severe) increasing from 6.7% (2004) to 9.1% (2006) and priority 2 from 33.7% (2004) to 35.1% (2006). We noticed a definite time demand pattern, with seasonal peaks in June, weekly peaks on Mondays, and daily peaks at 11 to 12 AM. These patterns were consistent during the period of the study. We designed a demand-based rostering tool that matched doctor-unit-hours to patient arrivals and severity. We also noted seasonal peaks corresponding to public holidays.

Conclusion: We found definite and consistent patterns of patient demand and designed a rostering tool to match ED manpower demand.

(C) 2009

Introduction

The functions of the emergency department (ED) have evolved during the past 30 years as emergency medicine developed from one room in the hospital (the “Emergency Room”) to a new specialty with more than 100 residency training programs in the United States alone [1]. The ED is

* Corresponding author. Tel.: +65 63213590; fax: +65 63214873.

E-mail address: [email protected] (M.E.H. Ong).

unique in providing continuous, round-the-clock availability of concentrated diagnostic and therapeutic services. In many hospitals, it also serves as the portal of admission entry for most patients with acute medical and surgical conditions.

Since 1958, the number of ED visits in the USA has increased more than 600% to an estimated 108 million ED visits in 2000 [2]. In many cases, this has been accompanied by a crisis of overcrowding in the ED [3-9]. From an international perspective, the problem of ED overcrowding is not limited to the United States. It has been reported in the United Kingdom [10], Canada [11], Taiwan [12], and

0735-6757/$ - see front matter (C) 2009 doi:10.1016/j.ajem.2008.01.032

Australia [13]. Emergency department overcrowding is now recognized as an international problem [14].

Reasons for ED overcrowding vary from increased demand in terms of number and complexity of cases [15,16], insufficient resources and capacity [17], decreased inpatient capacity resulting in “bed block” [8], and suboptimum use of resources [18].

Emergency department overcrowding can compromise quality of care and may lead to Medical errors, poor outcomes, and even unnecessary patient deaths [2,19]. Lengthy Waiting times mean prolonged pain and suffering for patients and a risk that their condition may deteriorate while waiting to be seen. For ED staff, the high-stress environment contributes to staff burnout and higher turnover rates, worsens deficiencies in staffing, and impairs ongoing clinical teaching [15].

Part of the problem is that patient demand is not constant but fluctuates according to the time of day, day of week, and season of the year. Therefore, the challenge of ED rostering is to match ED staffing and resources to actual demands, according to local conditions.

It has been noticed that acute medical events and even ED health-seeking behavior are not random events, but rather have definite time-geographic distribution patterns [20-27]. Systems status management (SSM) is a technique, described by ambulance services, for matching the staffing of ambulances in anticipation of when they will be needed by using historical ambulance response data [28]. Systems status management uses flexible, real-time management of the deployment of resources to meet Ambulance demand patterns. The question is, “Are there also consistent time demand patterns to ED attendances, and how can we roster ED staffing and resources most effectively to meet these demands”

We aimed to describe the time demand patterns at our local ED during a 3-year period. This will allow us to prepare a demand analysis and, in subsequent phases, to apply SSM to tailor ED manpower to demand.

Methods

Design

We conducted an observational, prospective study of all patients presenting to the ED during a 3-year period. We also conducted a time series analysis to determine time norms regarding physician activity for various severities of patients.

Setting

Prospective patients presenting to the ED of the Singapore General Hospital were enrolled in this study. The study began in January 2004 and ended in December 2006.

Singapore General Hospital is Singapore’s oldest and largest acute tertiary hospital and national referral center. It accounts for about one third of total acute Hospital beds in the public sector and about a quarter of acute beds nationwide. Annually, approximately 60000 patients are admitted to our wards and another 600000 attend to our Specialist outpatient clinics.

Subjects

Patients of all ages were included in the study. Cases were recruited using a review of ED-computerized records. The ED uses a real-time, fully computerized registration and consultation system that was developed in-house and has been in continuous use since the year 2000. All patient visits are time stamped automatically by the computerized system at the point of registration, triage, doctor’s consultation, for any procedure, and at the point of referral or discharge.

Outcomes

Data collected at the ED included demographic informa- tion, time of registration, waiting time (time of registration to time of doctor’s consultation), and processing time (time of registration to time of discharge). The processing time was entirely captured electronically from the time stamping via the computerized patient consultation system. The registra- tion time would be captured when the patient first presents at the registration counter and before triage and when his or her

Table 1 Singapore patient acuity Category (PAC) Scale

PAC These are patients who are either in a state of Scale 1 Cardiovascular collapse or in imminent danger of

collapse and would therefore be required to be attended to without delay. They would most likely require maximum allocation of staff and equipment resources for initial management.

PAC These patients are ill, nonambulant, and in various Scale 2 forms of severe distress. They would appear to be in

a stable state on initial cardiovascular examination and are not in danger of imminent collapse. The severity of their symptoms requires very early attention. Failure to do so will like result in early deterioration of their medical status. They would be trolley-based.

PAC These patients have acute symptoms but are

Scale 3 ambulant and have mild to moderate symptoms and require acute treatment, which will result in resolution of symptoms over time.

PAC These are nonemergency patients. They should not Scale 4 be presented at ED and should be managed in a

primary health care setting, such as by general practitioners or in polyclinics. They may have an old injury or condition that has been present for a long time. They do not require immediate treatment.

There is no immediate threat to their life or limb.

particulars are entered into the system. The time of discharge is captured when the patient disposition is processed, and the patient either pays his bill for discharge or is registered for inpatient admission. Triage category was also captured according to the Singapore Patient Acuity Category Scale (Table 1). In general, priority 1 (P1) patients would be seen in the resuscitation area, priority 2 (P2) would be in the nonresuscitation major emergency (Critical) area (Trolley), and priority 3 (P3) and priority 4 (P4) would be in the minor emergency (Ambulant) area.

Table 2 Yearly ED attendance by triage category

2004 (n [%])		2005 (n [%])	2006 (n [%])	Total (n [%])
Priority 1	7575	9782	11414	28771
(%)	(6.7%)	(8.1%)	(9.1%)	(8.0%)
Priority 2	38 169	42132	44101	124402
(%)	(33.7%)	(34.9%)	(35.1%)	(34.6%)
Priority 3	66 945	68090	69724	204759
(%)	(59.0%)	(56.4%)	(55.4%)	(56.9%)
Priority 4	698 (0.6%)	760 (0.6%)	534 (0.4%)	1992
(%)				(0.5%)
Total	113387	120764	125773	359924

For the time-series analysis, the timings for doctor-patient processes were analyzed for each triage category and points of contact with physicians identified. The following points of doctor contact were identified: (1) initial consultation, (2) obtaining a history from the patient or relatives, (3) interventions (eg, intravenous access), (4) minor operating theater procedures, (5) review (routine and observation), and

(6) review for admission.

The time-series study was conducted from 0000 hours on November 13, 2006, to 2359 hours on November 15, 2006. This was a longitudinal patient tracking-type survey. Survey forms were attached to the patient screening forms. Timings were separately recorded by survey staff members who were assigned to follow individual doctors stationed in specified areas during the period. They noted the start and completion times of all key activities from arrival to discharge (exit).

Nurse clinicians were briefed and trained to administer the survey, with “champions” in charge of each shift tasked with ensuring accuracy and reliability in form completion.

Fig. 2 Average patient arrivals by day of the week: 2004-2006.

Data were entered into Microsoft Excel using third-party entry staff.

Based on survey results, computer system data, and doctor’s feedback, the frequency and duration of interven- tions were determined. Subsequently, the average interven- tion time per patient was worked out. All timings were then added up to calculate the average time doctors spent on patients in P1, P2, and P3 categories.

Data analysis

Data entry was carried out using Microsoft Excel 2003 (Microsoft Corporation). All data analyses were performed using SPSS version 14.0 (SPSS, Chicago, Ill), presenting descriptive statistics and frequencies.

Results

From January 2004 through December 2006, a total of 359924 patients were enrolled in the study. Table 2 shows the breakdown of patients by year and triage category in the study. There was a yearly increase in attendance from 113387 in 2004, to 120764 in 2005, and to 125773 in 2006.

Fig. 1 shows the average daily patient arrivals by month from 2004 to 2006. There seems to be higher arrivals during the months of January/February and June/July. Fig. 2 shows the average patient arrivals by day of the week. The highest attendances were on Mondays and Tuesdays.

Fig. 1 Average daily patient arrivals by month: 2004-2006.

Fig. 3 Comparing average patient arrivals by hour of the day on a quarterly basis: 2004-2006.

Fig. 3 shows the ED attendances by hour of the day on a quarterly basis for Fridays. This is compared with the average patient arrivals by hour of the day for all days. The daily peak in arrivals occurred between 1000 hours and 1100 hours, with a smaller peak in the evenings from 2100 hours

to 2200 hours. Fig. 4 shows the distribution of patient arrivals by hour of the day according to triage category. The patterns were consistent with P2 and P3 arrivals; however, P1 arrivals were more constant.

For the time-series analysis, 1014 forms were issued, of which 957 (94.4%) were usable for analysis. There was no systematic error noted, with good quality of data entry. The mean doctor process time for P1 was 60 minutes, P2 was 40 minutes, and P3 was 30 minutes.

Fig. 5 shows the average doctor unit hours by hour of the day. The 50th, 80th, 95th, and 100th percentiles are shown, respectively. One doctor unit hour was defined as the patient

Fig. 4 Distribution of patient arrivals by hour of the day

according to triage category. Fig. 5 Doctor unit hours by hour of the day: 2004-2006.

Fig. 6 Rostering tool showing current mismatch of demand and ED manpower: 2004-2006.

demand requiring 1 doctor, 1 hour of his or her time (work output). This definition is irrespective of the severity of the patients seen. For example, 1 doctor unit hour might equate to a doctor seeing 1 P1 patient in an hour, or 2 P3 patients instead. This unit hour allows patient demand to be roughly translated into the actual number of medical staffing required for a shift.

Fig. 6 shows the rostering tool that we developed as a result of this study. This tool allows comparison of patient demand and actual ED manpower. On top of this, we added actual waiting times for the period as a reflection of actual performance of the system. There is currently a mismatch in doctor supply and patient demand, with supply lagging behind demand in the mornings. This results in prolonged waiting times between 1000 hours and 1300 hours.

Fig. 7 shows the seasonal peaks after public holidays, which was consistent on a yearly basis.

Fig. 7 Post-holiday demand: 2004-2006.

Discussion

In this study, we noticed a definite time demand pattern in our ED, with seasonal peaks in June, weekly peaks on Mondays, and daily peaks at 11 to 12 AM. These patterns were consistent during the period of the study. We designed a demand-based rostering tool that matched doctor unit hours to patient arrivals and severity. We also noted seasonal peaks corresponding to public holidays. Our study demonstrates the utility of using a demand analysis for planning ED manpower deployment.

There has been a growing understanding that Medical emergencies and ED attendances are not random events but occur in patterns and trends that can be observed historically. This is related to how Circadian rhythms affect disease processes, the underlying population demographics, and health-seeking behaviors.

For example, it has been observed that sudden cardiac arrests follow a circadian pattern [29-32], with an increasing incidence in the mornings until noon [25,33]. weekly patterns have also been noticed, with increased frequency of cardiac events on Mondays [34-37]. seasonal variations have also been described [34]. Similar patterns have been observed in acute myocardial infarction [38-44], stroke [32,45], and pulmonary embolism [46]. Various explana- tions for this effect that have been proposed include Circadian variations in the autonomic nervous system [47], variations in electrical activity [48-50], vascular changes [51,52], and hormonal/metabolic fluctuations [53]. It is interesting that our ED attendances seem to correspond to these circadian rhythms.

Previous studies have described variations and trends in ED presentation and how population demographics and health-seeking behavior have affected ED demand [24,54,55]. Attempts have also been made to forecast ED attendances using Mathematical models [20,22,23]. Our

study is unique in attempting to translate patient demand into “Doctor-Unit-Hours.” This is a simple tool that can be used to match physician rostering to actual patient demand. We have also designed a Microsoft Excel-based rostering tool that is able to graphically display expected demand with rostered manpower (Fig. 6). We found that currently, there is a mismatch in physician supply, which lags behind patient demand in the mornings. This has been addressed in subsequent rostering of manpower, and we are monitoring the effect on waiting times.

Limitations

Limitations of our study include that fact that demand prediction assumes that there are no sudden changes in the population demographics, disease patterns, and health- seeking behavior. Any rostering system should have a built-in overcapacity to cope with daily surges, and we recommend staffing to the 95th percentile of expected demand, not the median. Nevertheless, experience from emergency medical services agencies that use demand analysis suggests that this is a reliable tool [56]. Another limitation is that using a historical demand analysis also does not account for large-scale surges [57], and local provisions still have to be made for disaster and mass casualty situations [58]. Systems status management is a dynamic process, which requires real-time monitoring of ED demand [59] and on-the-spot adjustments to adapt to changing circumstances. Another limitation to consider is that demand analysis cannot address all the causes for ED overcrowding. Other factors including availability of ED and Hospital resources, hospital bed block, and systems workflow issues must all be considered when addressing this issue. In particular, hospital bed block can have a significant impact on ED workflows. boarding of patients can affect waiting time, nursing resources, radiology availability, and physician time. Emergency department resources are spent on continuing care of boarders, including the level of care in intensive care unit, when they should be seeing new patients coming in. Thus, Patient throughput and output will be

definitely impacted.

Nevertheless, we have found our demand model and rostering tool to be useful for planning manpower in the ED. We intend to follow up this study by looking into a real-time demand monitoring system that can provide instant feedback on patient demand and waiting times. We are also adapting our tool to reflect nursing and ancillary manpower unit hours in the model.

Conclusions

We found definite and consistent patterns of patient demand in our ED and designed a rostering tool to match

ED manpower demand. Our study shows a potential for demand analysis and SSM theories to be applied in the ED.

Acknowledgments

The authors thank the following: Derek Andresen, Richmond Ambulance Authority, Va, and Nur Shahidah Ahmad, Department of Emergency Medicine, Singapore General Hospital.

References

1. Young GP, Sklar D. Health care reform and emergency medicine. Ann

Emerg Med 1995;25:666-74.

1. Committee on Pediatric Emergency Medicine. Overcrowding crisis in our nation’s emergency departments: is our safety net unraveling? Pediatrics 2004;114:878-88.
2. Bradley VM. Placing emergency department crowding on the decision agenda. J Emerg Nurs 2005;31:247-58.
3. Dickinson G. Emergency department overcrowding. CMAJ 1989;140:

270-1.

1. Kunz Howard P. Overcrowding: not just an emergency department issue. J Emerg Nurs 2005;31:227-8.
2. Laskowski-Jones L. Starling’s curve: a way to conceptualize emergency department overcrowding. J Emerg Nurs 2005;31:229-30.
3. Lynn SG, Kellermann AL. Critical decision making: managing the emergency department in an overcrowded hospital. Ann Emerg Med 1991;20:287-92.
4. Trzeciak S, Rivers EP. Emergency department overcrowding in the United States: an emerging threat to patient safety and public health. Emerg Med J 2003;20:402-5.
5. Walters SD. Is booming demand dooming emergency department? Mich Hosp 1990;26:4-9.
6. Bagust A, Place M, Posnett JW. Dynamics of bed use in accommodating emergency admissions: stochastic Simulation model. BMJ 1999;319:155-8.
7. Schull MJ, Szalai JP, Schwartz B, et al. Emergency department overcrowding following systematic hospital restructuring: trends at twenty hospitals over ten years. Acad Emerg Med 2001;8:1037-43.
8. Shih FY, Ma MH, Chen SC, et al. ED overcrowding in Taiwan: facts and strategies. Am J Emerg Med 1999;17:198-202.
9. Acute Health Division. Emergency demand management. Melbourne: Victorian Government, Department of Human Services; 2001.
10. Graff L. Overcrowding in the ED: an international symptom of health care system failure. Am J Emerg Med 1999;17:208-9.
11. Derlet RW. Overcrowding in emergency departments: increased demand and decreasED capacity. Ann Emerg Med 2002;39:430-2.
12. Lambe S, Washington DL, Fink A, et al. Trends in the use and capacity of California’s emergency departments, 1990-1999. Ann Emerg Med 2002;39:389-96.
13. Eastaugh SR. Overcrowding and fiscal pressures in emergency medicine. Hosp Top 2002;80:7-11.
14. Siegel B. Triage for overcrowding. Hospitals should fix the emergency department problems they can control. Mod Healthc 2003;33:24.
15. Gordon JA, Billings J, Asplin BR, et al. Safety net research in emergency medicine: proceedings of the Academic Emergency Medicine Consensus Conference on “The Unraveling Safety Net”. Acad Emerg Med 2001;8:1024-9.
16. Champion R, Kinsman LD, Lee GA, et al. Forecasting emergency department presentations. Aust Health Rev 2007;31:83-90.
17. Clark MJ, Purdie J, FitzGerald GJ, et al. Predictors of demand for emergency prehospital care: an Australian study. Prehosp Disaster Med 1999;14:167-73.
18. Jones SA, Joy MP, Pearson J. Forecasting demand of emergency care. Health Care Manag Sci 2002;5:297-305.
19. Milner PC. Forecasting the demand on accident and emergency depart- ments in health districts in the Trent region. Stat Med 1988;7:1061-72.
20. Milner PC, Nicholl JP, Williams BT. Variation in demand for accident and emergency departments in England from 1974 to 1985. J Epidemiol Community Health 1988;42:274-8.
21. Levine RL, Pepe PE, Fromm Jr RE, et al. Prospective evidence of a circadian rhythm for out-of-hospital cardiac arrests. JAMA 1992;267: 2935-7.
22. Warden CR, Daya M, Legrady LA. Using geographic information systems to evaluate Cardiac arrest survival. Prehosp Emerg Care 2007; 11:19-24.
23. Lerner EB, Fairbanks RJ, Shah MN. Identification of out-of-hospital cardiac arrest clusters using a geographic information system. Acad Emerg Med 2005;12:81-4.
24. Overton J, Stout J, Kuehl A. System design, preHospital systems and medical oversight, National Association of EMS Physicians. 3rd ed. Kendall/Hunt Publishing; 2002. p. 114-31.
25. Willich SN, Levy D, Rocco MB, et al. Circadian variation in the incidence of sudden cardiac death in the Framingham Heart Study population. Am J Cardiol 1987;60:801-6.
26. Savopoulos C, Ziakas A, Hatzitolios A, et al. Circadian rhythm in sudden cardiac death: a retrospective study of 2,665 cases. Angiology 2006;57:197-204.
27. Jones-Crawford JL, Parish DC, Smith BE, et al. Resuscitation in the hospital: circadian variation of cardiopulmonary arrest. Am J Med 2007;120:158-64.
28. Hayashi S, Toyoshima H, Tanabe N, et al. Daily peaks in the incidence of sudden cardiac death and fatal stroke in Niigata Prefecture. Jpn Circ J 1996;60:193-200.
29. Muller JE, Ludmer PL, Willich SN, et al. Circadian variation in the frequency of sudden cardiac death. Circulation 1987;75:131-8.
30. Arntz HR, Willich SN, Schreiber C, et al. Diurnal, weekly and seasonal variation of sudden death. Population-based analysis of 24,061 consecutive cases. Eur Heart J 2000;21:315-20.
31. Arntz HR, Muller-Nordhorn J, Willich SN. Cold Monday mornings prove dangerous: epidemiology of sudden cardiac death. Curr Opin Crit Care 2001;7:139-44.
32. Gruska M, Gaul GB, Winkler M, et al. Increased occurrence of out-of- hospital cardiac arrest on Mondays in a community-based study. Chronobiol Int 2005;22:107-20.
33. Willich SN, Lowel H, Lewis M, et al. Weekly variation of acute myocardial infarction. Increased Monday risk in the working population. Circulation 1994;90:87-93.
34. Muller JE, Tofler GH, Willich SN, et al. Circadian variation of cardiovascular disease and Sympathetic activity. J Cardiovasc Pharmacol 1987;10(Suppl 2):S104-9 [discussion S110-1].
35. Lucente M, Rebuzzi AG, Lanza GA, et al. Circadian variation of ventricular tachycardia in acute myocardial infarction. Am J Cardiol 1988;62:670-4.
36. Muller JE, Tofler GH, Stone PH. Circadian variation and triggers of onset of acute cardiovascular disease. Circulation 1989;79:733-43.
37. Goldberg RJ, Brady P, Muller JE, et al. Time of onset of symptoms of acute myocardial infarction. Am J Cardiol 1990;66:140-4.
38. Hausmann D, Lichtlen PR, Nikutta P, et al. Circadian variation of myocardial ischemia in patients with stable coronary artery disease. Chronobiol Int 1991;8:385-98.
39. van der Palen J, Doggen CJ, Beaglehole R. Variation in the time and day of onset of myocardial infarction and sudden death. N Z Med J 1995;108:332-4.
40. Cannon CP, McCabe CH, Stone PH, et al. Circadian variation in the onset of unstable angina and non-Q-wave acute myocardial infarction (the TIMI III Registry and TIMI IIIB). Am J Cardiol 1997;79:253-8.
41. Manfredini R, Boari B, Smolensky MH, et al. Circadian variation in stroke onset: identical temporal pattern in ischemic and hemorrhagic events. Chronobiol Int 2005;22:417-53.
42. Gallerani M, Manfredini R, Portaluppi F, et al. Circadian variation in the occurrence of fatal pulmonary embolism. Differences depending on sex and age. Jpn Heart J 1994;35:765-70.
43. Piepoli MF, Capucci A. Autonomic nervous system in the genesis of arrhythmias in chronic heart failure: implication for risk stratification. Minerva Cardioangiol 2007;55:325-33.
44. Wennerblom B, Lurje L, Karlsson T, et al. Circadian variation of Heart rate variability and the rate of autonomic change in the morning hours in healthy subjects and angina patients. Int J Cardiol 2001;79: 61-9.
45. Beyerbach DM, Kovacs RJ, Dmitrienko AA, et al. Heart rate-Corrected QT interval in men increases during winter months. Heart Rhythm 2007;4:277-81.
46. Oren H, Cosgun A. Weekly variation of the QT dispersion in healthy subjects and in patients with coronary heart disease. Cardiology 2007; 108:55-61.
47. Pringle E, Phillips C, Thijs L, et al. Systolic Blood pressure variability as a risk factor for stroke and cardiovascular mortality in the elderly hypertensive population. J Hypertens 2003;21:2251-7.
48. Otto ME, Svatikova A, Barretto RB, et al. Early morning attenuation of endothelial function in healthy humans. Circulation 2004;109: 2507-10.
49. Li JJ. Circadian variation in myocardial ischemia: the possible mecha- nisms involving in this phenomenon. Med Hypotheses 2003;61: 240-3.
50. Lee AH, Meuleners LB, Zhao Y, et al. emergency presentations to Northern Territory public hospitals: demand and access analysis. Aust Health Rev 2003;26:43-8.
51. Svenson JE. Emergency department visits in Wisconsin 1998-2002: trends in usage and accuracy of reported data. WMJ 2005;104:56-8.
52. Brown LH, Lerner EB, Larmon B, et al. Are EMS call volume predictions based on demand pattern analysis accurate? Prehosp Emerg Care 2007;11:199-203.
53. Jenkins JL, O’Connor RE, Cone DC. Differentiating large-scale surge versus daily surge. Acad Emerg Med 2006;13:1169-72.
54. Flowers LK, Mothershead JL, Blackwell TH. Bioterrorism prepared- ness. II: the community and emergency medical services systems. Emerg Med Clin North Am 2002;20:457-76.
55. Hoot NR, Zhou C, Jones I, et al. Measuring and forecasting emergency department crowding in real time. Ann Emerg Med 2007; 49:747-55.