Healthcare resource allocation

Review Article Rationing critical care beds: A systematic review* Tasnim Sinuff, MD; Kamyar Kahnamoui, MD; Deborah J. Cook, MD; John M. Luce, MD; Mitchell M. Levy, MD; for the Values, Ethics, & Rationing in Critical Care (VERICC) Task Force Objective: Rationing critical care beds occurs daily in the hospital setting. The objective of this systematic review was to examine the impact of rationing intensive care unit beds on the process and outcomes of care. Data Source: We searched MEDLINE (1966 2003), CINAHL (1982 2003), Ovid Healthstar (1975 2003), EMBASE (1980 2003), Scisearch (1980 2003), the Cochrane Library, UBMED related articles, personal files, abstract proceedings, and reference lists. Study Selection: We included studies of seriously ill patients considered for admission to an intensive care unit bed during periods of reduced availability. We had no restriction on study design. Studies were excluded if rationing was performed using a scoring system or protocol and if cost-effectiveness was the only outcome. Data Extraction: In duplicate and independently, we performed data abstraction and quality assessment. Data Synthesis: We included ten observational studies. Hospital mortality rate was increased in patients refused intensive care unit admission vs. those admitted (odds ratio, 3.04; 95% confidence interval, 1.49 6.17). Factors associated with both intensive care unit bed refusal and increased mortality rate were increased age, severity of illness, and medical diagnosis. When intensive care unit beds were reduced, admitted patients were sicker, were less often admitted primarily for monitoring, and had a shorter intensive care unit length of stay, without other observed adverse effects. Conclusions: These studies suggest that patients who are perceived not to benefit from critical care are more often refused intensive care unit admission; refusal is associated with an increased risk of hospital death. During times of decreased critical bed availability, several factors, including age, illness severity, and medical diagnosis, are used to triage patients, although their relative importance is uncertain. Critical care bed rationing requires further investigation. (Crit Care Med 2004; 32:1588 1597) KEY WORDS: ration; triage; critical care; intensive care *See also p. 1623. From the Departments of Medicine (TS, DJC), Surgery (KK), and Clinical Epidemiology & Biostatistics (DJC), McMaster University, Hamilton, Ontario, Canada; Department of Medicine (JML), University of California, San Francisco, CA; and Department of Medicine (MML), Brown Medical School, rovidence, RI. Supported, in part, by the Canadian Institutes for Health Research (TS, DJC). Copyright 2004 by the Society of Critical Care Medicine and Lippincott Williams & Wilkins DOI: 10.1097/01.CCM.0000130175.38521.9F Healthcare resource allocation refers to the distribution of healthcare resources among individuals and populations and encompasses rationing and triage (1). Bedside rationing by a physician is the withholding of medically beneficial service because of that service s cost to someone other than the patient (2). For this systematic review, we have adopted the Values, Ethics, and Rationing in Critical Care (VERICC) Task Force definition of healthcare rationing: the allocation of potentially beneficial healthcare services to some individuals in the face of limited availability that necessarily involves the withholding of those services from other individuals. Rationing is influenced by myriad factors, including clinical judgment, patient and family preferences, and best evidence of therapeutic efficacy. Ethical approaches to rationing vary, from the model in which patient autonomy, beneficence, and distributive justice drive medical decision-making (3) to models of conspicuous paternalism (4, 5). An American Thoracic Society statement on fair intensive care unit (ICU) resource allocation (1) indicated that when demand exceeds supply, medically appropriate patients should be admitted on a first-come, first-served basis, based on an egalitarian principle of fair allocation of resources rather than on the grounds of relative benefit. The concept of resource allocation on the grounds of relative medical benefit is often referred to as triaging: the process in medicine of finding the most appropriate disposition for a patient based on an assessment of the patient s illness and its urgency (6). Triage and rationing decisions are not singular events but rather are complex, multifaceted processes with important implications for clinicians, patients, researchers, health policy makers, and society. The impact of ICU bed rationing on patient outcome remains uncertain. The objective of this systematic review was to determine the impact of bed rationing on processes of care and clinically important outcomes among patients referred for admission to an ICU. METHODS Search Strategy. We searched MEDLINE (1966 2003), CINAHL (1966 2003), Ovid Healthstar (1975 2003), EMBASE (1980 2003), Cochrane Library (1981 2003), and Scisearch (1980 2003) using these terms: acute or intensive or intermediate and care and ration or allocation or resource or triage. We searched the related articles feature of UBMED and hand-searched references and meeting abstracts from 1990 to 2003 (American Thoracic Society, American College of Chest hysicians, Society for Critical Care Medicine, and the ESICM). We reviewed personal files and wrote to experts and first authors to identify other published or unpublished studies. We had no language restrictions. Study Selection Criteria. Citations considered potentially relevant by either of two reviewers (TS or KK) were retrieved. The follow- 1588 Crit Care Med 2004 Vol. 32, No. 7

Table 1. Study characteristics Study Author (Year) n/n (%) [Reference] Setting Country Years Studied Design Cohort(s) Studied Bed Shortage or Triage hysician Rationing or Triaging Beds Outcomes atients refused ICU admission Azoulay et al. (2001) 283/1,292 (21.9) atients refused admission [18] Triaging studies Joynt et al. (2001) Total n 624 [19] Admitted: 388/624 (62.2) Refused: 236/624 (37.8) Sprung et al. (1999) Total n 382 [9] Admitted: 290/382 (75.9) Later admitted: 31/382 (8.1) Never admitted: 61/382 (16.0) Metcalfe et al. (1997) Total n 651 [21] Admitted: 483/651 (74.2) Refused: 168/651 (25.8) Friso-Lima et al. (1994) Total n 127 [17] Admitted: 62/127 (48.8) Refused: 65/127 (51.2) Marshall et al. (1992) Total n 308 [14] Admitted: 295/308 (95.8) Refused: 13/308 (4.2) Rationing studies Walther and Jonasson (2001) Total n 7,479 [20] atients admitted: 1993: 2,457 1995: 2,510 1997: 2,512 Byrick et al. (1993) Total n 570 [13] atients admitted: re-ica closure: 194 ost-ica closure: 376 Strauss et al. (1986) Total n 1,151 [15] atients admitted: 1,151 Transferred out with 0 1 empty beds: 381/1,151 (33.1) Transferred out with 2 empty beds: 531/1,151 (46.1) Singer et al. (1983) Total n 1,304 [16] Admitted 1980: 734 Admitted 1981: 570 Multicenter France April 4, 1999 May 2, 1999 Hong Kong December 1997 June 1998 Israel May 1993 December 1993 Multicenter United Kingdom October 1993 January 1994 Israel January March 1992 USA April June 1988 Sweden January 1 December 31 for each of 1993, 1995, 1997 Canada October 1989 June 1990 and April 1991 January 1992 USA July 1979 April 1980 USA July December 1980 and July December 1981 R R R R Single cohort: patients refused ICU admission 2 cohorts: patients Triage: refusals admitted and refused defined as: inappropriate referrals a Triage b 3 cohorts: admitted, later admitted, never admitted Triage Intensivist Determinants of triage Futility c Triage 2 cohorts: patients Triage during: bed admitted and refused shortage Nursing shortage hysician shortage 2 cohorts: patients admitted and refused 2 cohorts: patients Triage during bed admitted and refused shortage: bed closure leading to nursing shortage Bed reduction from 16 to 10 14 beds 3 cohorts over 3 years with different bed availability (12, 10, and 8 beds available) 2 cohorts: patients admitted to ICU preand post-closure of the intermediate care area Single cohort: patients admitted 2 cohorts: patients admitted pre- and post-bed closure Intensivist Hospital mortality Standardized mortality ratios Senior resident ICU admission patterns or intensivist according to AACHE II scores and ICU bed status Hospital mortality Intensivist ICU LOS ICU mortality rate Hospital mortality rate Triage Intensivist Hospital mortality rate Bed shortage 1993: 12 beds 1995: 10 beds 1997: 8 beds Bed shortage: following closure of intermediate care area Bed shortage: periods of 0 3 beds available (comparison of 0 1 vs. 2 empty beds) Intensivist Intensivist Bed shortage and nursing shortage due to bed closure 1980: 18 beds 1982: 8 beds ICU LOS ICU mortality rate ICU LOS ICU mortality rate ost-icu mortality rate Hospital mortality rate Senior resident ICU LOS ICU mortality rate ost-icu LOS Hospital mortality rate ICU LOS Hospital mortality, prospective; ICU, intensive care unit; SCCM, Society of Critical Care Medicine; AACHE, Acute hysiology and Chronic Health Evaluation; LOS, length of stay; R, retrospective. a atients too well and expected not to derive any benefit from ICU admission; b prioritization of patients based on perceived magnitude of benefit from ICU admission and admission based on threshold of benefit to bed availability; c patient too sick to benefit from ICU regardless of bed situation. Crit Care Med 2004 Vol. 32, No. 7 1589

Table 2. Intensive care unit (ICU) characteristics and bed availability Study Author (Year) [Reference] Unit Type (Open or Closed) No. of ICU Beds Hospital: ICU Bed Ratio Nurse:atient Ratio atients refused ICU admission Azoulay et al. (2001) [18] Mixed () 14 5 UTC 1:3 Triaging studies Joynt et al. (2001) [19] Mixed (closed) 22 64:1 4.5:1 Sprung et al. (1999) [9] Mixed (closed) 8 UTC 81:1 Metcalfe et al. (1997) [21] Mixed () 7 103:1 Friso-Lima et al. (1994) [17] Mixed (closed) 20 36:1 1:2 to 2:3 Marshall et al. (1992) [14] Surgical (open) 16 dropped to 11 14 when 2 6 beds 45:1 closed Rationing studies Walther and Jonasson (2001) [20] Mixed (closed) 1993: 12 1995: 10 1997: 8 Byrick et al. (1993) [13] Mixed (closed) re-ica closure: 13 (7 ICU 6 ICA) ost-ica closure: 9 ICU beds 1993: 47:1 1995: 49:1 1997: 49:1 re-ica closure: 39:1 ost-ica closure: 52:1 Strauss (1986) [15] Mixed (closed) 18 16:1 Singer et al. (1983) [16] Mixed () 1980: 18 UTC 1981: 8 14 UTC, unable to calculate;, not reported; ICA, intermediate care area. 1993: 2.2:1 1995: 2.2:1 1997: 3.0:1 2:1 ing inclusion criteria were applied: a) adult 16 yrs, and seriously ill patients considered for admission to an ICU bed (medical, surgical, trauma, neurologic, or mixed ICU; intermediate care unit, high-dependency unit, or stepdown unit); b) retrospective or prospective cohort; c) rationing of ICU beds based on reduced bed availability or triaging of patients referred for admission; and d) any outcome including severity of illness scores, length of stay, or mortality rate. To focus on clinical decision making in ICUs, we excluded studies using a) cost-effectiveness as the only outcome; b) scoring systems or protocols to make rationing or triage decisions; or c) rationing or triaging studies of coronary care units. In duplicate and independently, two of us applied these criteria to the full articles. Disagreements were resolved by consensus. Data Abstraction and Study Quality. Independently and in duplicate, two of us abstracted data on study, patient, and ICU characteristics and on outcomes. We abstracted variables associated with refusal for ICU admission and with hospital mortality. The primary outcome was the effect of rationing or triaging on mortality rate. Differences were resolved by consensus. We developed a quality assessment form to critically appraise these studies (7, 8). A priori, we considered it important to adjust for at least two of the following: patient age, illness severity, and staff or bed shortages. We contacted authors when key data were unclear or not reported. Data Synthesis. We classified the studies as Type 1, studies comparing patients admitted to an ICU and those refused an ICU bed during a single time period (triaging studies with two group comparisons); Type 2, studies comparing patients admitted during a minimum of two different periods of time, at least one of which had reduced bed availability (rationing bed studies with two group comparisons); and Type 3, studies of patients either admitted or refused admission during a single period of bed shortage (single cohort studies). Data Analysis. We measured crude agreement between reviewers regarding study selection, data abstraction, and quality assessment. We report proportions and percentages, means and standard deviations or standard errors, medians and interquartile ranges, and odds ratios (OR), and 95% confidence intervals (CI). Because of study heterogeneity, we performed both qualitative and quantitative analyses (8). For the study by Sprung et al. (9), we combined the mortality rate of patients admitted later and those not admitted to the ICU to represent the mortality rate of patients not initially admitted to the ICU, as the authors did for their regression analysis. To evaluate the mortality rate of patients refused ICU admission, we summarized data using a random effects model (10). We tested for statistical heterogeneity using a chi-square test (11). Statistical calculations and graphical analyses were performed using SSS software, version 11.0 (SSS, Chicago, IL). Tests of significance were two-tailed, and p.05 was considered significant. A priori, we specified that three factors may influence hospital mortality rate and may explain heterogeneous results among the studies comparing patients admitted to an ICU and those refused an ICU bed (triaging studies). First, we performed a sensitivity analysis comparing studies with a total quality score greater than or equal to vs. less than the median total score. In addition, we compared studies examining confounders of mortality rate to those that did not. We hypothesized that studies of lower quality (lower total score and those not examining confounding) would Figure 1. Flow of studies considered for this systematic review as suggested in the Quality of Reporting of Meta-Analysis (QUOROM) flow diagram (12). CCU, coronary care unit; ICU, intensive care unit. find a greater difference in mortality rate between admitted and refused patients. Second, we compared studies in which 40% of the admitted cohort were surgical patients to those in which 40% were surgical. We hypothesized that the risk of mortality would be greater in studies with a greater proportion of medical patients in the refused cohort. Third, we compared studies conducted before and after 1995, hypothesizing that we would find no difference in the mortality rates between admitted and refused cohorts. RESULTS Study Selection. We identified 227 citations; 211 were excluded since they did 1590 Crit Care Med 2004 Vol. 32, No. 7

Table 2. Continued. olicy in Event of Bed Shortages or ICU Full atients refused ICU admission Azoulay et al. (2001) [18] Triaging studies Joynt et al. (2001) [19] Sprung et al. (1999) [9] Metcalfe et al. (1997) [21] Friso-Lima et al. (1994) [17] Marshall et al. (1992) [14] Rationing studies Walther and Jonasson (2001) [20] Byrick et al. (1993) [13] Strauss (1986) [15] Singer et al. (1983) [16] Society of Critical Care Medicine guidelines (20 recommendations). atients transferred to another hospital only if no ICU beds and ICU closed. Elective referrals postponed, but never refused. No specific official policy. Officially: first come first served. In practice postoperative and trauma patients received priority due to limited capacity of recovery room. When ICU full, patient not admitted regardless of condition. General indications for admission and discharge as per Critical Care Society Task Force recommendations. atients obviously moribund were not admitted. Hospital policy encouraged cooperation between ICUs for the purposes of triage so no critically ill patient would be diverted or transferred to another hospital if any ICU bed was available within the institution. Authority for interunit triage depended on the goodwill of the respective medical directors. No policy mandating cooperation between units in order to facilitate elective surgery schedule. If no bed available for elective admission, surgeon had option of postponing surgery, arranging alternative ICU care within institution, or proceeding with surgery without an available ICU bed. No specific official policy. Overflow solved by temporary increase in number of beds. Occasional delay of admissions to discharge ICU patients to increase bed availability. Table 3. Study Quality Study Author (Year) Reference Cohort Data Collection atient Outcomes Follow-Up 90% Confounders Total Score Azoulay et al. (2001) [18] 1 0 1 1 0 3 Joynt et al. (2001) [19] 1 0 1 1 0 3 Walther and Jonasson (2001) [20] 1 0 1 1 0 3 Sprung et al. (1999) [9] 1 0 1 1 1 4 Metcalfe et al. (1997) [21] 0 0 0 1 1 2 Friso-Lima et al. (1994) [17] 1 0 1 1 1 4 Byrick et al. (1993) [13] 1 0 1 0 0 2 Marshall et al. (1992) [14] 1 0 1 1 0 3 Strauss et al. (1986) [15] 0 1 1 1 0 3 Singer et al. (1983) [16] 0 0 1 1 0 2 not fulfill inclusion criteria (n 130) or they fulfilled exclusion criteria because they were economic evaluations (n 29), used risk factors or severity scores to allocate beds (n 39), evaluated decision tools (protocols, guidelines, algorithms) for rationing or triage (n 12), or were conducted exclusively in the coronary care unit (1). Of 16 potentially relevant citations, ten observational studies satisfied all criteria (Fig. 1) (12). Agreement for selecting abstracts was 96.5% and selecting full articles was 100%. Study Characteristics. We report study characteristics in Table 1 and ICU characteristics and bed availability in Table 2. One study was conducted in Canada (13), three in the United States (14 16), two in Israel (9, 17), one in France (18), one in Hong Kong (19), one in Sweden (20), and one in the United Kingdom (21). The funding source was reported in four studies (15, 16, 19, 20). Five studies compared cohorts of patients either admitted to ICU or refused ICU admission (triaging studies) (9, 14, 17, 19, 21), three studies compared patients admitted to an ICU (13, 16, 20) during periods of variable ICU bed availability (rationing studies), one study included patients refused ICU admission during a period of bed closure (18) (single cohort study), and in one study a cohort of patients studied was admitted to the ICU during a period of bed closure (15) (single cohort study). Four studies stated that in the event of bed shortages they encouraged the use of official policies to make decisions: a) Society of Critical Care Medicine ICU admission recommendations (18), b) out-ofhospital patient transfers (19), c) first come first served policy (17), and d) within-hospital patient transfers (14). One study had no official policy (20); five did not report a policy (9, 13, 15, 16, 21). For studies reporting the physician making the rationing or triage decisions, it was the intensivist in six studies (14, 17 21), the ICU senior resident in one (15), and either the intensivist or the senior resident in another (9). Our crude agreement was 98.2% for data abstraction. Methodological Quality. The median Crit Care Med 2004 Vol. 32, No. 7 1591

Table 4. atient characteristics and crude mortality of triaging studies Study Author (Year) n/n (%) [Reference] Age in Yrs, Mean SD or Median and IQR or Mean (SEM) Gender (% Female) Diagnosis, % AACHE II Score or Other Illness Severity Score Crude Hospital Mortality Rate or Other Classification of Mortality as Stated n/n (%) Joynt et al. (2001) Total n 624 [19] Admitted: 388/624 (62.2) Refused: 236/624 (37.8) Sprung et al. (1999) Total n 382 [9] Admitted: 290/382 (75.9) Later admitted: 31/382 (8.1) Never admitted: 61/382 (16.0) Metcalfe et al. (1997) Total N 651 [21] Admitted 483/651 (74.2) Refused: 168/651 (25.8) Friso-Lima et al. (1994) Total n 127 [17] Admitted: 62/127 (48.8) Refused: 65/127 (51.2) Marshall et al. (1992) Total n 308 [14] Admitted: 295/308 (95.8) Refused: 13/308 (4.2) Admitted: 62 (16 96) (43) Refused: 71 (20 96) (35) Admitted: 49 1 (37) Later admitted: 50 4 (39) Never admitted: 55 3 b (38) Admitted: 59 (17 91) (45) Refused: 64 (16 90) (40) Admitted: 55.9 18.6 (45) Refused: 65.4 18.6 e (60) Admitted: Refused: 53.2 12.0 (46) Admitted ARF: 24 CHF: 14 Sepsis: 13 Other medical: 36 ostoperative/emergency surgery: 13 Refused ARF: 26 CHF: 15 Sepsis: 10 Cardiac arrest: 12 Other medical: 31 ostoperative/emergency surgery: 6 Admitted Respiratory: 9.7 Sepsis: 3.4 Cardiac: 1.7 Surgical: 41.4 Trauma: 32.8 Later admitted Respiratory: 22.6 Sepsis: 9.7 Cardiac: 6.5 Surgical: 16.1 Trauma: 3.2 Never admitted Respiratory: 21.3 Sepsis: 27.9 Cardiac: 4.9 Surgical: 11.4 Trauma: 13.1 Admitted: Medical: 51 Surgical: 49 Refused: Medical: 73 Surgical: 27 Admitted Medical: 34 COD: 14.5 CHF: 3.2 Refused: Medical: 80 f COD: 32.3 g CHF: 16.9 3 Admitted: Elective surgery: 96 Emergency surgery: 94 Cardiothoracic: 59 Neurosurgical: 21 General surgery: 15 Refused: Elective surgery: 4 Emergency surgery: 7 Cardiothoracic: 2 Neurosurgical: 6 General surgery: 10 Admitted: 142/388 (36.6) Refused: 145/236 (61.4) a Admitted: 12.1 0.4 Later admitted: 5.6 1.5 Never admitted: 15.8 1.4 c Median (IQR) Admitted: 19 (1 54) Refused: Not done Admitted: 13.6 8.1 Refused: 13.9 6.6 Admitted: 16.5 6.3 Refused: 10.8 7.2 h Admitted: 40/290 (13.8) Later admitted: 11/31 (35.5) Never admitted: 28/61 (45.9) d Admitted: 168/480 (35.0) Refused: 64/165 (38.8) Admitted: 7/62 (11.3) Refused: 32/65 (49.2) 6 IQR, interquartile range; AACHE, Acute hysiology and Chronic Health Evaluation; ARF, acute respiratory failure; CHF, congestive heart failure;, not reported; COD, chronic obstructive pulmonary disease. a p.001; b p.29; c p.001; d p.001; e p.005; f p.00001; g p.05; h p.005. 1592 Crit Care Med 2004 Vol. 32, No. 7

methodological quality score of these studies was 3 (interquartile range, 2 4; Table 3). Five studies identified a priori potential confounders in their studies (15, 17 19, 21). Three performed adjusted analyses for mortality (9, 17, 21). Of the ten studies, seven scored 3 (9, 14, 15, 17 20). Our crude agreement on study quality was 92.4%. Triaging Studies In Table 4, we present patient characteristics for the five triaging studies (9, 14, 17, 19, 21), which compared patients admitted to ICU and those refused an ICU bed. There was variability in the case mix, severity of illness, and proportion of patients refused ICU admission. The proportion of patients refused ICU admission was 16 51%. None of the studies reported the do-not-resuscitate status of patients or indicated when triage decisions were made (day or night, or day of the week). Three of the studies determined factors associated with refusal to admit patients. These included age (9, 17, 19), illness severity (19), and medical diagnosis (9, 17). Four of the five studies reported unadjusted hospital mortality rate (9, 17, 19, 21) (Table 5) and were included in the quantitative analysis of unadjusted mortality rate comparing patients admitted to ICU and those refused ICU admission (total n 1,778). For patients refused ICU admission, compared with those admitted, the pooled OR for mortality was 3.04 (95% CI, 1.49, 6.17), although results were heterogeneous (p.001, Table 5,Fig. 2). Following sensitivity analysis, three studies with quality scores 3.0 (9, 17, 19) (total n 1,133) demonstrated an increased risk of mortality in patients refused ICU admission (pooled OR, 4.07; 95% CI, 2.38, 6.97; Fig. 2). Differences in methodological quality, population (surgical vs. medical), and year of the study did not explain the heterogeneity. Variables associated with increased mortality rate included greater age, higher Acute hysiology and Chronic Health Evaluation II score, and medical status (9, 17, 21). Study summaries are available upon request. Table 5. Hospital mortality rate of triaging studies Author (Year of ublication) [Reference] Admitted to ICU No./ Total (%) of Deaths Refused Admission to ICU No./Total (%) of Deaths Odds Ratio [95% CI] Joynt et al. (2001) [19] 142/388 (37) 145/236 (61) 2.76 [ 1.98, 3.85] Sprung et al. (1999) [9] 40/290 (14) 39/92 (42) 4.60 [ 2.70, 7.82] Metcalfe et al. (1997) [21] 168/480 (35) 64/165 (39) 1.18 [ 0.82, 1.70] Friso-Lima et al. (1994) [17] 32/65 (49) 7/62 (11) 7.62 [3.02, 19.21] ooled odds ratio [95% CI] 3.04 [ 1.49, 6.17] ICU, intensive care unit; CI, confidence interval. Rationing Beds Studies Three studies compared patients admitted to the ICU (15 16, 20) during periods of differing bed availability, whereas one study (13) compared patients admitted to the ICU before and after closure of the intermediate care unit (Table 6, rationing studies). None of the studies collected data on patients not admitted to the ICU or reported do-not-resuscitate status. Strauss et al. (15) demonstrated Figure 2. Odds of hospital mortality in patients refused intensive care unit (ICU) admission. Unit of expression equates an odds ratio (OR) with the risk of mortality in patients refused ICU admission. CI, confidence interval. 1 Test for homogeneity: 2 27.03, p.001; 2 test for homogeneity: 2 5.7, p.06. Crit Care Med 2004 Vol. 32, No. 7 1593

Table 6. atient characteristics of rationing studies Study Author (Year) n/n (%) [Reference] Age in Yrs, Mean SD or Median and IQR or Mean (SEM) Gender (% Female) Diagnosis, % AACHE II Score or Other Illness Severity Score Walther and Jonasson (2001) Total n 7,479 [20] atients admitted: 1993: 2,457 1995: 2,510 1997: 2,512 Byrick et al. (1993) Total n 570 [13] atients Admitted: re-ica closure: 194 ost-ica closure: 376 Strauss et al. (1986) Total n 1,151 [15] Admitted: 1,151 Transferred out with 0 1 empty beds: 381/1,151 (33.1) Transferred out with 2 empty beds: 531/1,151 (46.1) Singer et al. (1983) Total n 1,304 [16] Admitted 1980: 734 Admitted 1981: 570 1993: 53.3 (0.5) 1995: 1997: 54.2 (0.5) re-ica closure: 58.2 18.6 (44) ost-ica closure: 59.9 16.7 (39) 1993: Nonsurgical: 76.3 Surgical: 23.7 1995: Nonsurgical: 79.5 Surgical: 20.5 1997: Nonsurgical: 87.7 Surgical: 12.3 re-ica closure: Nonemergent: 33.5 Emergent: 66.5 ostoperative nonemergent: 43.8 ostoperative emergent: 56.2 ost-ica closure: Nonemergent: 41.5 b Emergent: 58.5 ostoperative nonemergent: 73.6 a ostoperative emergent: 26.4 50 21 1980: 62 yrs 1981: 61 yrs 1980: recordial pain: 34 Respiratory distress: 17 Coma: 6.3 GI bleed: 3.0 1981: recordial pain: 36 Respiratory distress: 18 Coma: 5.9 GI bleed: 3.5 1993: 12.6 7.9 1995: 13.5 8.0 1997: 12.8 8.2 re-ica Closure: 21.9 7.4 ost-ica Closure: 18.6 7.4 a No. of major interventions per patient: 1980: 0.49 1981: 0.69 c IQR, interquartile range; AACHE, Acute hysiology and Chronic Health Evaluation;, not reported; ICA, intermediate care area; GI, gastrointestinal. a p.001; b p.03; c p.01. that during bed shortages, significantly more patients with acute myocardial infarction were admitted, with an inverse relationship between severity of illness and bed availability. In the study by Singer et al. (16), bed shortages were associated with an increased proportion of patients admitted with acute myocardial infarction, an increase in major interventions per patient, and a decreased proportion admitted for monitoring only. The mean bed occupancy increased by 13%, whereas the proportion of ICU beds available at the time of referral decreased by 40%. Byrick et al. (13) compared patients admitted to the ICU before and after closure of the intermediate care unit. The proportion of nonemergent medical and surgical patients admitted to ICU increased after intermediate care unit closure, whereas the Acute hysiology and Chronic Health Evaluation II score decreased. In Table 7, we report study outcomes. During bed shortages, two studies (15, 20) documented a reduced ICU length of stay, whereas one (13) documented reduced hospital length of stay. The percentage of patients with ICU length of stay 3 days was significantly higher during the year of reduced bed availability in the study by Singer et al. (16). Studies of atients Refused ICU Admission In this multiple-center cohort study (18), only 25% of patients for whom admission was requested were considered futile. Refusals were made either when the ICU was full or during phone-triage for 71% of patients. Society for Critical Care Medicine recommendations were more likely adhered to following bedside patient examination by the intensivist than during phone-triage and when the ICU was not full compared with full. Factors associated with ICU refusal were age 65 yrs, poor performance status, underlying malignancy (associated with multiple-system organ failure or terminal metastatic disease), and chronic respiratory or cardiac failure. DISCUSSION In this systematic review, we summarized studies assessing the effects of patient triage and ICU bed rationing on processes of care and patient outcomes. 1594 Crit Care Med 2004 Vol. 32, No. 7

Table 7. atient outcomes of rationing studies Study Author (Year) No. [Reference] ICU LOS, Median (IQR), Hrs ICU Mortality Rate (%) Hospital LOS Walther and Jonasson (2001) Total n 7,479 [20] atients admitted: 1993: 2,457 1995: 2,510 1997: 2,512 Byrick et al. (1993) Total n 570 [13] atients admitted: re-ica closure: 194 ost-ica closure: 376 Strauss et al. (1986) Total n 1,151 [15] Admitted: 1,151 Transferred out with 0 1 empty beds: 381/1,151 (33.1%) Transferred out with 2 empty beds: 531/1,151 (46.1%) Singer et al. (1983) Total n 1,304 [16] Admitted 1980: 734 Admitted 1981: 570 1993: 6.6 (19.2) 1995: 1997: 4.8 (16.9) a Mean SD days re-ica closure: 8.5 15.8 ost-ica closure: 6.7 22.5 Entire admitted cohort, mean SD days: 3.3 5.8 Mean days: 1980: 3.5 1981: 3.3 1993: 11.7 1995: 9.2 1997: 10.6 re-ica closure: 40/217 (18.4) ost-ica closure: 65/407 (16.0) Entire admitted cohort 146/1,151 (12.7) 1980 (10) 1981 (10) re-ica closure: 37.3 42.1 days ost-ica closure: 26.5 31.3 days a ost-icu LOS (mean days) 0 1 empty beds: 18.0 2 empty beds: 18.3 ICU, intensive care unit; LOS, length of stay; IQR, interquartile range;, not reported; OR, odds ratio; CI, confidence interval; ICA, intermediate care area. a p.0001; b p.05. Factors associated with refused admission included advanced age, high illness severity, medical diagnosis, poor performance status, and bed shortages. These studies indicate that patients triaged and refused ICU admission largely because of a perceived minimum potential to benefit from critical care had a three-fold higher risk of hospital mortality compared with those admitted. In contrast, studies comparing patients admitted to the ICU (13, 15, 16, 20) during periods of reduced compared with usual bed availability showed that during periods of bed shortage, more seriously ill patients were admitted overall, and fewer patients were admitted for monitoring. In two studies (15, 16), when fewer beds were available, acuity of illness was higher at the time of discharge, with ICU length of stay shorter. Bed shortages did not appear to influence ICU readmission rates and mortality rates. These studies suggest that physicians ration ICU beds during times of decreased availability without increasing readmission or mortality rate by using different admission and discharge thresholds. However, these observational studies preclude conclusions about whether patients who were refused admission to an ICU bed were actually denied treatment that would have been beneficial. This review suggests that several factors are used to triage patients when the ICU is full, such as age, illness severity and complexity, and admitting diagnosis; these factors are consistent with stated attitudes, although their relative importance is unclear. In a survey of 600 clinicians about the distribution of ICU resources, 12% stated that age should limit admission; most indicated that quality of life, probability of hospital survival, acute illness reversibility, and comorbidities were important considerations when triaging (22). rofessional society ICU admission guidelines (1, 6, 23 25) have not been formally evaluated for their effect on outcomes and may not be used in practice. For example, during bed shortages and phone triage, intensivists followed fewer recommendations than other times (18). Further research is warranted to assess the merit of telephone triage, outof-icu consults, and ICU admission guidelines. Reduced bed availability may result in premature, untimely discharge. For example, one study found a significant increase in ICU night discharges in 1995 98 compared with 1988 90 (26). Night discharges were associated with a greater case-mix adjusted hospital mortality rate compared with daytime discharges (OR, 1.33; 95% CI, 1.06 1.65). remature initially unplanned discharge due to bed shortages was more common nocturnally and associated with an increased mortality rate (OR, 1.35; 95% CI, 1.10 1.65). Mortality rate was not significantly increased after adjustment for premature discharge. Insufficient healthcare resources require that some patients who may potentially benefit from ICU cannot receive it. Alternatives to ICU admission are needed for patients who need stabilization or whose illnesses are too complex for general wards. rovision of graded levels of care may also help with the optimal utilization of critical care resources (27). Although their cost-effectiveness has not been demonstrated (28), intermediate care areas offer theoretical advantages for lower risk patients needing monitoring and less intense nursing care; these areas may also improve triaging, result in more appropriate use of ICU beds, and help to avoid ICU readmission (29). The recently developed medical emergency team could serve a similar function (30 31). This systematic review has several limitations. The observational design of the Crit Care Med 2004 Vol. 32, No. 7 1595

Table 7. Continued. Crude Hospital Mortality Rate or Other Classification of Mortality as Stated Within 30 Days of Discharge from ICU (%) 1993: 511/2,457 (20.8) 1995: 404/2,510 (16.1) 1997: 457/2,512 (18.2) b re-ica closure: 61/194 (31.4) ost-ica closure: 94/376 (25.0) Unadjusted OR [95% CI] for Mortality as Defined in Adjacent Column for eriod of Bed Shortage/Reduced Bed Availability 0.85[0.74, 0.98] 0.73[0.50, 1.07] These studies suggest that patients who are perceived not to benefit from critical care are more often refused intensive care unit admission; refusal is associated with an increased risk of hospital death. ost-icu discharge mortality rate: Admitted: 90/1,151 (7.8) 0 1 empty beds: 42/381 (11.0) 2 empty beds: 48/531 (9.0) 1980: 37/734 (5.0) 1981: 29/570 (5.1) studies means that the increased mortality rate among patients refused admission may not hold after adjusting for other confounders. The four studies that performed risk adjustment (9, 17, 19, 21) determined that patients refused ICU admission were more likely to die either in hospital or at 90 days. These adjusted mortality estimates differ only in degree, not direction, compared with the unadjusted rates. In other words, vital status cannot be attributed to critical care bed refusal in these studies. Second, although the patients and their illness severity were similar across studies, geographic differences in the indications for ICU admission and refusal may exist. Since the triaging studies were predominantly from Europe and Asia, and studies of reduced bed availability were predominantly North American, their results may not be generalizable elsewhere. A third limitation is the potential for temporal trends; from 1983 to 2001 when these studies were published, technology utilization, ICU workload, and societal values such as patient preferences for life support may have changed. Furthermore, in recent studies, influences of families and clinicians such as social workers were not considered and may be important. Nevertheless, we did not find any variability in the outcomes across studies due to publication date. Fourth, we did not incorporate studies of scoring systems and protocols, since these are uncommonly used in practice. Strengths of this review include rigorous methods and transparent reporting (8). We searched multiple data sources and sought non-english literature. We evaluated validity of the primary studies and conducted both qualitative and quantitative syntheses when appropriate. Because of the potential for spurious results of meta-analyses of observational studies (32) and since there is no consensus on how to proceed if qualitative but not quantitative heterogeneity exists among studies (8, 33 35), we presented a qualitative summary of the studies, in addition to a summary estimate of hospital mortality rate for patients refused ICU admission. Due to this heterogeneity, results should be interpreted cautiously (32 33). CONCLUSION 1.25[0.81, 1.93] 1.00 Although our systematic review provides some insights into the effects of bed rationing and patient triage, questions remain. In practice, intensivists make the majority of triaging and rationing decisions a priori, acting as gatekeepers. Rationing decisions can result in the admission or discharge of a group of patients with a narrowing spectrum of illness severity, favoring patients of higher acuity. Although these decisions have important ramifications, they are not always objective (14). It remains unclear how clinicians reason and ration ethically. Moreover, as the health of our aging population deteriorates, and as critical care becomes more expensive, the need to optimize triage and rationing decisions will intensify. Our scarce critical care resources underscore the need to examine triaging and rationing using policy analysis and qualitative research (36). Although the studies in this systematic review represent the totality of the literature in this area, the observational nature, heterogeneity, moderate to poor methodological quality, and lack of succinct conclusions of the individual studies preclude strong conclusions. Additionally, the synthesis of these studies into this systematic review does not provide clear recommendations or guidance for rationing critical care resources. The medical community would ideally work toward consistent definitions for rationing and triage (37). Higher quality studies are needed that address a) how rationing and triaging decisions are made; b) the morbidity and mortality impact on patients; and c) the public s perspective on rationing critical care beds. ACKNOWLEDGMENTS We thank Bruce Weaver, who developed the computer program for performing the meta-analyses. We are especially grateful to Bob Truog, Marion Danis, and Dan Brock of the Values, Ethics, & Rationing in Critical Care Task Force for their thoughtful suggestions. 1596 Crit Care Med 2004 Vol. 32, No. 7

REFERENCES 1. American Thoracic Society Statement. Fair allocation of intensive care unit resources. Am J Respir Crit Care Med 1997; 156:1282 1301 2. Ubel A, Goold S: Recognizing bedside rationing: Clear cases and tough calls. Ann Intern Med 1997; 126:74 80 3. Osborne M, atterson J: Ethical allocation of ICU resources: A view from the USA. Intensive Care Med 1996; 22:1009 1014 4. Osborne M, Evans TW: Allocation of resources in intensive care: A transatlantic perspective. Lancet 1994; 343:778 780 5. Lemaire F, Ferrand E: Refusal of care in the ICU: Any meaning for doctors? Intensive Care Med 1998; 24:204 205 6. Society of Critical Care Medicine Ethics Committee. Consensus statement on the triage of critically ill patients. JAMA 1994; 271:1200 1203 7. Moher D, ham B, Jones A, et al: Does quality of reports of randomized trials affect estimates of intervention efficacy reported in meta-analyses? Lancet 1998; 352:609 613 8. Stroup DF, Berlin JA, Morton SC, et al: For the Meta-analysis of Observational Studies in Epidemiology (MOOSE) Group. Meta-analysis of observational studies in epidemiology: A proposal for reporting. JAMA 2000; 283: 2008 2012 9. Sprung CL, Geber D, Eidelman LA, et al: Evaluation of triage decisions for intensive care admission. Crit Care Med 1999; 27:1073 1079 10. Fleiss JL: The statistical basis of meta-analysis. Stat Methods Med Res 1993; 1:121 145 11. DerSimonian R, Laird N: Meta-analysis in clinical trials. Control Clin Trials 1986; 7:177 188 12. Moher D, Cook DJC, Eastwood S, for the QUOROM group: Improving the quality of reports of meta-analyses of randomized controlled trials: QUOROM statement. Lancet 1999; 354:1896 1900 13. Byrick RJ, Mazer CD, Caskennette GM: Closure of an intermediate care unit. Impact on critical care utilization. Chest 1993; 104:876 881 14. Marshall MF, Schwenzer KJ, Orsina M, et al: Influence of political power, medical provincialism, and economic incentives on the rationing of surgical intensive care unit beds. Crit Care Med 1992; 20:387 394 15. Strauss MJ, LoGerfo J, Yeltatzie JA, et al: Rationing of intensive care unit services. An everyday occurrence. JAMA 1986; 255: 1143 1146 16. Singer DE, Carr L, Mulley AG, et al: Rationing intensive care physician responses to a resource shortage. N Engl J Med 1983; 309:1155 1160 17. Friso-Lima, Gurman G, Schapira A, et al: Rationing critical care what happens to patients who are not admitted? Theoretical Surgery 1994; 39:208 211 18. Azoulay E, ochard F, Chevret S, et al: For the ROTOCETIC Group. Compliance with triage to intensive care recommendations. Crit Care Med 2001; 29:2132 2136 19. Joynt GM, Gomersall CD, Tan, et al: rospective evaluation of patients refused admission to an intensive care unit: Triage, futility, and outcome. Intensive Care Med 2001; 27:1459 1465 20. Walther SM, Jonasson U: A prospective cohort study of 6-month mortality in a community hospital experiencing a gradual reduction in critical care services. Intensive Care Med 2001; 27:700 705 21. Metcalfe MA, Sloggett A, Mcherson K: Mortality among appropriately referred patients refused admission to intensive care units. Lancet 1997; 350:7 12 22. Society of Critical Care Medicine: Attitudes of critical care medicine professionals concerning distribution of intensive care resources. Crit Care Med 1994; 22:358 362 23. European Society of Intensive Care Medicine. Guidelines for the utilization of intensive care units. Intensive Care Med 1994; 20:163 164 24. American College of Critical Care Medicine of the Society of Critical Care Medicine: Guidelines on admission and discharge for adult intermediate care units. Crit Care Med 1998; 26:607 610 25. Ethics Committee of the Society of Critical Care Medicine: Consensus statement of the Society of Critical Care Medicine s Ethics Committee regarding futile and other possibly inadvisable treatments. Crit Care Med 1997; 25:887 891 26. Goldfrad C, Rowman K: Consequences of discharges from intensive care at night. Lancet 2000; 355:1138 1142 27. Kalb E, Miller DH: Utilization strategies for intensive care units. JAMA 1989; 261: 2389 2395 28. Keenan S, Massel D, Inman KJ, et al: A systematic review of the cost-effectiveness of noncardiac transitional care units. Chest 1998; 113:172 177 29. Cheng DCH, Byrick RJ, Knobel E: Structural models for intermediate care areas. Crit Care Med 1999; 27:2266 2271 30. Lee A, Bishop G, Hillman KM, et al: The medical emergency team. Anaesth Intens Care 1995; 23:183 186 31. Bristow J, Hillman KM, Tien C, et al: Rates of in-hospital arrests, deaths and intensive care admissions: The effect of a medical emergency team. Medical Journal Australia 2000; 173:236 240 32. Egger M, Schneider M, Davey Smith G: Spurious precision? Meta-analysis of observational studies. BMJ 1998; 316:140 145 33. Davey Smith G, Egger M: Meta-analysis of observational data should be done with due care. BMJ 1999; 318:56 34. Beral V: The practice of meta-analysis : Discussion. Meta-analysis of observational studies: A case study of work in progress. J Clin Epidemiol 1995; 48:165 166 35. Egger M, Davey Smith G, hillips AN: Metaanalysis: rinciples and procedures. BMJ 1997; 315:1533 1537 36. Rocker GM, Cook DJ, Martin DK, et al: Seasonal bed closures in an intensive care unit: A qualitative study. J Crit Care 2003; 18:25 30 37. Strosberg MA, Teres D: Gatekeeping in the intensive care unit. Chicago, Health Administration ress, 1997. Foundation of the American College of Healthcare Executives. Crit Care Med 2004 Vol. 32, No. 7 1597