clinical investigations in critical care

Closure of an Intermediate Care Unit*

Impact on Critical Care Utilization
Robert J Byrick, M.D .; C. David Mazer; M .D .; and Gary M. Caskennette

We studied the effect of closing a six-bed intermediate care at the time of discharge from CCU decreased (p<O .OOOI).
area (ICA) on utilization of a multidisciplinary critical care The ICA closure altered CCU admission and discharge
unit (CCU). Data were collected on all admissions to the 7- decision-making. "Low-risk" admissions increased and pa-
bed CCU for 9 months prior to ICA closure (n=217) and tients remained in the CCU until they required less nursing
compared with 9 months after CCU expansion (7 to 9 beds) care. One factor determining utilization of a CCU is the
and ICA closure (n=407). Nonemergency CCU admissions facilities available outside the unit. A CCU management
increased from 41 to 112 after ICA closure (p<O.03). Mean system is especially important when a wide range of illness
APACHE II score within 24 h of admission decreased from severity is present. (Chest 1993; 104:876-81)
21.9±7.4 to 18.6±7.4 (p<O.OOOI). The proportion of
patients with APACHE II score < IS, increased from 301
217 to 1361407 accounting for an increase from 5.4 percent leA = intermediate care area; TISS = Therapeutic Intervention
Scoring System
to 12.7 percent ofCCU days (p<O.OOOI). Nursing workload

The high cost of providing critical care has provoked intermediate care unit that had served as a "step-
a variety of "utilization strategies" within hospitals down" facility as well as a "low-risk" monitoring unit.
to optimize use of these scarce resources. I The relative Our critical care audit system" permitted retrospective
scarcity of critical care resources and limited access analysis of the interaction of the structure and organ-
to them is a result of both increasing demand and ization of hospital facilities with utilization of a multi-
restrained supply. One strategy that many hospitals disciplinary medical-surgical CCU.
have adopted is the provision of "graded" levels of
METHODS
care in intermediate care units to apportion nursing
time and monitoring facilities for specific patient HospitallCCU
needs."? This strategy has been proposed as a cost- This study was conducted in St. Michael's Hospital, an adult,
effective alternative to critical care unit (CCU) admis- tertiary care referral center affiliated with the University ofToronto .
All medical and surgical specialtie s were available in the hospital .
sion, particularly for low-risk patients admitted for
The medical/surgical CCU was a multidisciplinary unit , supervised
monitoring," and staffed by a team of critical care physicians with residents
Most reports that evaluate the impact of an inter- available in the unit 24 hid. Th e nursing managers provided one
mediate level of care on the use of CCU resources do nurse per patient during the time period s studied .
so after the opening of an intermediate care area Indications for admission to the medical-surgical CCU included
(ICA).6.7 Few, if any, studies have evaluated the impact the requirement for mechanical ventilatory support, critical care
nursing, and/or invasive hemodynamic monitoring not available in
of closing an ICA on the utilization of a multidiscipli- other areas of the hospital . With the exception of postoperative
nary critical care facility that services both medical cardiac and neurosurgical patients and posttrauma patients who
and surgical patients. The decision to admit patients were admitted to other specialized units, all patients requiring
to the CCU for monitoring alone is controversial and critical care were admitted to the CCU during the study period .
Utilization was assessed using an ongoing CCU data collection
some tnvestigators'-" have suggested that many of these
system during two 9-month time period s separated by a 9-month
patients could be safely cared for in an ICA. transition period . During the first 9-month period , October 1, 1989
A recent budgetary crisis in our adult, tertiary care to June 30, 1990 (pre-ICA closure), the 7·bed medical-surgical CCU
referral hospital resulted in a decision to close an operated independent of a separate 6-bed ICA that admitted both
medical and surgical patients not requiring invasive hemodynamic
monitoring or mechanical ventilatory support. The medical super-
*From the Departments of Anaesthesia (Drs. Byrick and Mazer) vision of the ICA was independent of the CCU staff. Patients
and Department of Biomedical Engineering (Mr. Caskennette), requiring intensive nursing care (two nurses, one patient) but not
St. Michael's Hospital, Toronto, Ontario. Canada. continuous hemodynamic monitoring or respiratory support were
Manuscript received October 9, 1992; revision accepted January 6
Reprint requests : Dr. Byrlck, Department of Anaesthesia, St. admitted to the ICA facility based on medical and nursing needs .
Michael Hospital, 1bronto, Ontario, Canada M5B lW8 A 9-month interval between pre- and post-ICA closure data

876 Closure of Intermediate Care Unit (Byrick , Mazer, CBskennette)

 Table I-Demographic and AdmiAion Datafor All CCU Table 2-UtiUmtion Datafor Alllbtienta During Each
Ibtienta During Each 9-Month Time fttriod* 9-Month Time fttriod*

Pre-ICA Post-ICA Pre-lCA Post ICA

Closure Closure Significance Closure Closure Significance

No . of patients 194 376 CCU length of stay, 8.5± 15.8 6.7±22.5 NS

Age, yr 58.2± 18.6 59.9± 16.7 NS days
Sex, MIF 108186 2281148 NS Hospital length of stay, 37.3±42.1 26.5±31.3 p<O.OOl
Hospital admission days
status OutcomeCCU
Nonemergency 65 156 p<O .03 (admissions)
177
~}
Emergency 129 220 NS Survival
NS
Postopemtive admissions 105 178 Death 40
Nonemergency 46 131 p<O.05 Outcome hospital
Emergency 59 47 (patients)
Survival 133
·Data reported as mean ± standard deviation. CCU =critical care NS
Death 61
=
unit; ICA intennediate care area.
TISS score
Admission 37.1±15.9 27.9±14.9 p<O.OOOI
collection allowed an adjustment period in admission/discharge
Discharge (modified 23.8 ±16.9 13.5±13.4 p<O.OOOI
policies after closure of the ICA and redistribution of hospital beds
TISS)
and staff. During the second 9-month data collection, April 1, 1991
APACHE II score 21.9±7.4 18.6±7.4 p<O.OOOI
to January 5 , 1992 (post-ICA closure), an expanded (9-bed) medical-
(admission)
surgical CCU operated without an ICA. During this post-ICA
closure period, all patients requiring more Intensive nursing and ·Data are reported as mean ± standard deviation. CCU =critical
medical care than available on the regu\ar wards were admitted to care unit; lCA =Intennediate care area; TISS =Therapeutic Inter-
theCCU . vention Scoring System.
During both data collection periods, CCU admission and dis-
charge were determined by the same critical care team. The total (dBase Iv, Borland, Scotts Valley, Calif) was used to develop the
number of hospital beds available was reduced from a daily average management program (Carebase) for entering, maintaining, and
of503 beds (range, 354 to 560) in the pre-ICA closure period to 467 analyzing coUected data. A tmined health records technician was
(range, 348 to 497) after ICA closure. The cause of the reduced responsible for reviewing each chart and entering data. The accuracy
number of beds and reorganization of critical care facilities was an of data for nondiagnosis-related fields, such as length of stay, have
institutional (opemting) deficit of $62 million . been verified on chart reabstraction to have match rates greater
than 95 percent. U
Data Collected
Infonnation collected on every admission to the medical-surgical Sta&tic6l Analyril
CCU included the following: age, sex , primary service referring Data are reported as mean ± standard deviation. Continuous
the patient, and type of hospital admission (emergency, nonemer- variables were compared using analysis of variance for parametric
gency). As well, the CCU length of stay, bospitallength of stay, and data and the Ko\mogorov-Smlrnov test for nonparametric data (eg,
survival were noted. length of stay). A statistical pacbge (SAS, Statistical Analysis
The severity of illness during the first 24 h In the CCU was System) was used for the analysis. Frequency data were examined
assessed using the APACHE II scoring system." This methodology using the 'It where appropriate. Differences were considered
scores, on a scale of 0 to 4, each of 12 physiologic variables significant when the p value was <0.05 .
commonly measured in critically ill patients. The APACHE II score
is composed of the weighted value of these physiologic variables RESULTS
(score increases based on the deviation from normal values during
the 24 h after admission) plus points for age and chronic health Table 1 summarizes demographic and admission
status. Speci6c consideration was given to standardization of values data for patients admitted during both 9-month data
for the Glasgow Coma Score, included as one of the 12 physiologic collection periods. The number of CCU admissions
variables. Since many postopemtive surgical patients were anesthe-
increased from 217 to 407 with closure of the ICA and
tizedand partially paralyzed on admission to the CCU, the Glasgow
Coma Score was recorded as nonnal if a postoperative patient was expansion of the CCU from 7 to 9 beds. The increased
awake and following commands within 24 h of admission. throughput in CCU was from 31 patients per CCU
The nursing wnrlcload and Interventions used during the first bed per 9 months to 45.2 patients per CCU bed per
24 h In the CCU were assessed using the Therapeutic Intervention 9 months. There was no significant change in the
Scoring System (I1SS) which weights critical care nursing interven-
proportion of CCU patients who were emergency
tions (one to four points per Intervention) based on complexity of
expertise required ." A " modified TISS" scored at the time of hospital admissions (1291194 vs 2201376); however, the
discharge from the CCU was also recorded as a "one-time" proportion of nonemergency admissions increased
measurement of nursing interventions ordered for each patient, not significantly (p<O.03). After ICA closure, the number
the cumulative 24-h score for admission TISS . The purpose of of emergency postoperative admissions remained sim-
recording the interventions (modified TISS) required for patient
care on leaving the CCU was to estimate the nursing care required
ilar (59 vs 47); however, an increased number of
after discharge. nonemergency postoperative surgical admissions were
Routine data collection by nursing staff was Integrated Into a noted compared with the 9-month period before ICA
CCU audit system.· A relational database management system closure (Table 1).

CHEST I 104 I 3 I SEPTEMBER. 1993 sn

 180
170
1lSO Pre-ICA Closure (n-217)
150
140
I
Post·ICA Closure (n-407)
130
1 120
E 110
~
'S 100

! :
70
eo
50
40
30
20
10
o
7 8 9 10 11 12 13 14 15 18 17 18 19 20 > 20 Days
Length Of Stay In Unit (days)
FIGURE 1. The number of admissions to the critical care unit (CCU) during two 9-month data collection
periods and the length of CCU stay ; the first (October I , 1989 to June 30, 1990) was before closure of a
six-bed intermediate care unit (pre-ICA closure) and the second (April I, 1991 to January 5, 1992) W'LS
after ICA closure and expansion of the CCU from 7 to 9 beds (post-ICA closure). ICA = intermediate care
area.

120 -

110 -

100 · ~ Pre-ICA Closure (0-217)

!0 90 ·
1m Post ·ICA Closu re (0-407)

1 80 ·
E
'a 70 -
e
'5 80 -
j
50 -

40 -

30 -

20

10 -

0
Dol 6-10 11-15 16-20 21·215 21-30 31-35 31-40 41~

APACHE II SCore
FIGURE 2. The number of admissions to the critical care unit (CCU) with varying severity of illness
(APACHE II score") during two 9-month data collection periods before (pre-ICA closure) and after (post-
ICA closure) closure of an intermediate care area (ICA).

878 Closure of Intermediate Care Un" (Byrick. Mazer, C8skennelte)

 A reduction in the mean length of hospital stay the utilization of key service components, such as
(Table 2) was found after leA closure; however, no critical care, also changes. The high costs of critical
difference in CCU length of stay was noted between care and the limited acces s to these services necessi-
the two time periods. The range of ceu length of stay tates an organized approach to resource management
varied from a minimum of 1 day to a maximum of 156 during periods of adjustment. Reports of CCU utili-
days before ICA closure and from 1 to 105 days after zation often study altered patterns of use when a
ICA closure. These data are highly skewed to the left "step-down unit" or ICA is opened . Our data dem-
(Fig 1). The median eeu length of stay was 3 days onstrated the extent of altered utilization of a multi-
pre-ICA closure and 2 days after ICA closure. disciplinary ceu when an ICA was closed and em-
A significant decrease in the APACHE II score on phasized the need for a utilization management
admission to the ecu was noted from 21.9±7.4 pre- system. The two key changes in utilization pattern
ICAclosure to 18.6± 7.4 after ICAclosure (p<O.OOOI). were the increased number of patients admitted with
The number of APACHE II points for chronic health a low severity of illness and an increased number of
disability decreased from 4.52 ± 1.10 to 3.82 ± 1.47 patients discharged with significantly lower nursing
(p<O.OOOl). The proportion of CCU admissions with workload requirements. The magnitude of these two
APACHE II scores greater than or equal to 15 did not changes in utilization pattern was large (Table 2), yet
increase (187/217 vs 271/407) signtficantly after closure commonly used measures, such as length of stay, did
of the ICA (Ftg 2). However, a larger number of ceu not detect them. If physicians are to participate
admissions with APACHE II scores less than 15 were effectively in resource management with nursing and
admitted to the CCU after closure of the ICA (1361 administrative colleagues, we must develop an organ -
407) than prior to closure (301217). The total number ized approach to utilization analysis to understand the
of CCU days attributed to patients with APACHE II factors that influence and measure use of resources.
scores less than 15 increased from 5.4 percent (100/ This study demonstrates the value of the APACHE II
1850) to 12.7 percent (329/2702). No death occurred methodology as a validated tool for utilization assess-
in the 29 patients admitted prior to ICA closure with ment and emphasizes the need to develop more
APACHE II scores less than 15. After ICA closure, validated measures of nursing workload . Our experi-
there were only two deaths among patients admitted ence also suggests that the beneficial effects reported
to ceu with APACHE II scores less than 15 (one in with creation of an ICA37 are reversed if the unit is
CCU and one after CCU discharge). Both patients closed and reinforces the concept that the presence of
died suddenly; one unexpectedly 26 days after CCU an ICA altered decision-making in a ceu . The
discharge of pulmonary embolus and the other of strategy of identifying "low-risk" groups requiring
irreversible central nervous system damage within 2 fewer nursing resources and allocating these patients
days of admission to the CCU. to less intensively monitored or staffed locations was
The average TISS on admission to the CCU de - inhihited by closing the ICA.
creased (p<O .OOOl) after ICA closure (Table 2). In Critical care physicians and nurses have two poten-
addition, the TISS (modified) at the time of discharge tial options to manage access to critical care facilities.
decreased from 23.5 ± 16.9 to 13.5 ± 13.4 (p<O.OOOI) The first strategy is to establish independent ICAs
after ICA closure . that provide limited, well-defined intermediate levels
After ICA closure, the number of short-stay patients of care to decrease demand for critical care beds, The
(1- and 2-day CCU length of stay) increased signifi- alternate strategy would he the development offlexihle
cantly from 92 to 243. The hospital mortality rate of staffing patterns and resource use within a multidis-
short-stay patients was 20.5 percent and 15.9 percent, ciplinary CCU. For efficient use of personnel and
respectively, in each time period and the percentage resources, this option would necessitate stratification
of total CCU days attributed to short-stay patients of critical care delivery according to patient need
increased from 6.64 percent to 11.59 percent after within a diverse unit. Altering the supply of and
ICA closure. Short-stay (1 to 2 days) patients had demand for critical cart' resources could theoretically
admission TISS, after ICA closure, that was signifi- he accomplished hy either strategy, and either appears
cantly lower (22.8 ± 14.0) than for patients staying preferable to formal explicit rationing. I
longer than 2 days (34.7 ± 3.3) (p<O .OOO6). The mod - Cost considerations and increased throughput have
ified TISS at the time of discharge was decreased been the prime impetus encouraging alternate sites of
significantly in the short-stay group from 21.3 ± 17.2 care for critically ill patients. Spivack" has estimated
to 13.0± 13.9 (p<O .OOO6) between the two time pe - that approximately $100,000 could he saved annually
riods. per hospital if 10 percent of patients were diverted
from a 15-lw.d ceu with a 95 percent occupancy rate
DISCUSSION
to a noninvasive unit . He estimated that implementing
When hospital structure and organization change, this strategy nationally in the United States could

CHEST I 104 13 I SEPTEMBER , 1993 879

result in a $500 million saving. We found that dismant- overriding determinant in allocating access to critical
ling the ICA resulted in 136 CCU patient admissions care.
with APACHE II scores less than 15 as compared with Admission and discharge decision-making regarding
30 in the pre-ICA closure period. This increase "low-risk" monitored admissions will depend on what
represented approximately 25 percent of total CCU alternatives are available outside the CCU. Closure of
admissions, but accounted for less than 10 percent of the ICA reduced flexibility in discharging patients and
CCU bed days in the 9 months after ICA closure. thus, the capacity to adjust discharge decision-making
We cannot comment on the cost-effectiveness to the was lost. This lost flexibility resulted in no change in
hospital of the decision to close the ICA because we mean CCU length of stay, in spite of the increased
do not have accurate cost per patient data and because number of "low-risk" admissions with low APACHE
the case-mix distribution clearly changed. The patient II scores; these patients stayed only 1 or 2 days.
population served by the ICA was admitted to either Medical suitability remains the basis for CCU admis-
the wards or to the CCU in the post-ICA closure sion and discharge decision-making. Little attention
period. After ICA closure, we found an increased has been paid to establishing priorities for discharge,
proportion of CCU admissions and patient days that although physicians do this daily.
were accounted for by patients requiring less nursing In this study, important changes in CCU utilization
care with lower APACHE II scores. The potential for occurred without any significant differences in overall
inefficient use of critical care personnel and resources CCU length of stay. This happened because the effect
increased when the ICA with its "graded" level of care of an increased number of short-stay patients with low
was withdrawn without a change in CCU management. APACHE and 11SS scores was offset by the necessity
Since personnel costs account for approximately 80 of keeping "sicker" patients longer in the CCU because
percent of critical care expenditures, the need for a of the lack of stepdown facilities to discharge them to .
nursing workload management system during periods Similarly, the reduction in calculated mortality rate
of changing utilization is particularly important. In the for patients staying in the CCU only 1 or 2 days
absence of an ICA, increased cost-effectiveness might probably does not reflect a change in the quality of
be gained by providing graded levels of care within a care provided because the severity of illness and
CCU to patients with possible acute myocardial in- preadmission chronic health disability decreased .
farction.v' respiratory failure," and medical surgical Thus, the CCU length of stay and mortality rates,
patients," Henning et al l 3 proposed using severity of taken in isolation, are not sufficient measures of CCU
illness measures (APACHE II system) to identify utilization, especially during times of administrative
patients eligible for intermediate levels of care. How- change.P
ever, this index alone may not be appropriate for all In conclusion, we found that the structure and
groups and must be investigated further,14 because availability of non-CCU facilities, such as the ICA, to
nursing workload and level of care do not always a significant extent determined the utilization of a
correlate directly with APACHE II scores," Thus, a critical care facility. Our data emphasize the need to
utilization analysis process that includes nursing work- take a "system" approach to utilization analysis in
load assessment is most important when the hetero- critical care. Utilization analysis must therefore con-
geneity of severity of illness increases within a CCU. sider both the need for CCU and the facilities available
A successful ICA strategy should emphasize that in the CCU, as well as in the hospital and the health
unit managers focus on discharge policies and priori- care system. These facilities vary with time and place
ties, and design ICA units to meet specific hospital and must be defined in order to properly interpret
needs. Our data suggest that discharge criteria (eg, CCU utilization data. The need for a continuous audit
modified 11SS) are as important as admission criteria ofCCU utilization is most marked when organizational
for efficient utilization. Physicians can alter priorities strategies are changing. The result of our analysis was
for admission and discharge when confronted with a the reestablishment of a unit (ICA) to provide specified
resource (bed or nursing) shortage.w'" Strauss et al l7 monitoring (ECG and pulse oximetry) to "low-risk"
have shown that the relationship between bed availa- monitored patients in a less intensively staffed area.
bility on a given day and the severity of illness of As well, the need for a CCU audit became the focus
patients admitted and discharged was inversely pro- of a utilization management system with an emphasis
portional. Factors other than bed availability that can on nursing workload assessment and severity of illness.
influence admission and discharge criteria are prog- It is only by taking advantage of these "natural
nostic uncertainty, patient age,18 expected social ca- experiments" which occur in all organizations that we
pacity;" quality of life,oo family preferences, and the will be able to study factors that determine the proper
organization of critical care delivery," Kalb and Miller! allocation and utilization of critical care resources.
suggested that severity of illness and medical suitabil- ACKNOWLEDGMENTS: The authors wish to express appreciation
ity rather than "marginal benefit" should be the to Ms. Cathy Mechetuk and David Leung of the Medical Records

880 CIosuI8 oIlntermedlate Care Unll (ByrIck. M8Z~ ClJskenneltfJ)

Department for assistance with data entry, Mr. Colin Kay for update 1983. Crit Care Med 1983: 11:1-3
statistical analysis, and Ms. Kerry McClenaghan for preparation of 12 Report of the Ontario Data Quality Reahstracting Study. Ontario
the manuscript. Hospital Association, Ontario Ministry of Health, Hospital
MedkaI Records Institute, April 1991
REFERENCES 13 Henning RJ, McClish 0 , Daly B, Nearman H, Franklin C,
1 Kalb PE, Miller OH. Utilization strategies for intensive care Jackson O . Clinical characteristics and resource utilization of
units. ]AMA 1989: 261:2389-95 ICU patients: implications for organization of intensive care.
2 Fineberg Hv, Scadden 0, Goldman L. Care of patients with Crit Care Med 1987: 15:~
low probability of acute myocardial infarction. N EngI J Med 14 Clvetta JM , Hudson-Civetta ]A, Nelson LO. Evaluation of
1984: 310:1301-07 APACHE II for cost containment and quality assurance. Ann
3 Mulley AG, Thibault GE. Hughes RA, Barnett GO, lieder VA, Surg 1990; 212:266-74
Sherman EL. The course of patients with suspected myocardial 15 Singer OE, Carr PL, Mulley AG, Thibault GE. Rationing
infarction: the identification ofJow.risk patients for early transfer intensive care: physician responses to resource shortage. N Engl
from intensive care. N Engl J Med 1980; 302:943-48 J Med 1983; 309:1155-60
4 Brush JE, Brand OA, Acompora 0, Chalmers B, ~rs FJ. 16 Selker Hf, Griggltb JL, Dorey F], O'Agostlno RB. How do
Use of the initial electrocardiogram to predict in-hospital physicians adapt when the coronary care unit is full? ]AMA
complications of acute myocardial infarction. N EngI J Med 1987; 257:1181-85
1985:312:1137~1 17 Strauss MJ, LoGeno Jf, Yeltlatzie Jf, Temkins JS, Hudson CO.
5 Wagner Of, Knaus WA, Draper EA, Zimmerman JE. Identifi- Rationing of intensive care unit services: an everyday occur-
cation of low-risk monitor patients within a medical-surgical rence. ]AMA 1986; 255:11~
intensive care unit. Med Care 1983: 21:425-34 18 Campion Evv, Mulley AG, Goldstein RL, Barnett GO, Thibault
6 Byrick RJ, Power JO, Yeas JO, Brown KA. Impact of an GE. Medkalintensive care lOr the elderly: a study of current
intermediate care area on ICU utilization after cardiac surgery. use, costs and outcomes. ]AMA 1981: 246:2052-56
Crit Care Med 1986: 14:869-72 19 Crane O. The sanctity of soda1 life: pbyslclans treatment of
7 Bone RC, Balk RA. Noninvasive respiratory care unit: a cost - critically ill patients. New York: Russell Soge FOundation, 1975
effective solution for the future. Chest 1988: 93:390-94 20 Pearlman RA, Invl 1'5, Canter WB . Variability in physician
8 Spivack O. The high cost of acute health care: a review of bloethical decision-making. Ann Intern Med 1982: 97:420-25
escalating costs and limitations of such exposure in intensive 21 Marshall MF, Scbwenzer KJ, Orslna M, Fletcher JC, Durbin
care units. Am Rev Respir Dis 1987: 136:1007-11 CG. ln8uence of political power, medkaI provlnciallsm and
9 Byriclc RJ, Caskennette GM. Audit of critical care: alms, uses, economic incentives on the rationing of surglca\ intensive care
costs and limitations of a Canadian system. Can J Anaesth 1992; unit beds. Crit Care Med 1992; 20:387-94
39:260-$ 22 Mazer OM, ByricIcRJ, SibbaJd WJ, Cbavaz PM, Goodman SJL,
10 Knaus WA, Draper EA, Wagner Of, Zimmerman JE. APACHE Girotti MJ, et aI. I\:lst-operative utilization of critical care
II: a severity of disease classification system. Crit Care Med. services by cardiac surgery: a multi-centre studyin the Canadian
1985: 13:818-29 health care system. Crlt Care Med (in press)
11 Keene AE , Cullen OJ. Therapeutic intervention scoring system:

CHEST I 104 I 3 I SEPTBoeER, 1883 881