Chest Email Content Delivery
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Guest Access | Sign In via User Name/Password
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF) Free
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Article Archive
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via ISI Web of Science (41)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Pappachan, J. V.
Right arrow Articles by Smith, G. B.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Pappachan, J. V.
Right arrow Articles by Smith, G. B.
(Chest. 1999;115:802-810.)
© 1999 American College of Chest Physicians

Comparison of Outcome From Intensive Care Admission After Adjustment for Case Mix by the APACHE III Prognostic System*

John V. Pappachan, FRCA; Brian Millar, BA, RGN; E. David Bennett, FRCP and Gary B. Smith, FRCA

* From the Department of Intensive Care Medicine (Drs. Pappachan and Smith), Queen Alexandra Hospital, Portsmouth, UK; Critical Audit, Ltd. (Mr. Millar), St. George's Hospital Medical School, London, UK; and Department of Intensive Care Medicine (Dr. Bennett), St. George's Hospital, London, UK.


   Abstract
TOP
 Abstract
Introduction
 Materials and Methods
 Results
 Discussion
 References
 
Study objectives: To evaluate the acute physiology, age, chronic health evaluation III (APACHE III) scoring system in the context of general adult ICUs in the United Kingdom.

Design: Prospective, noninterventional, cohort study.

Setting: Seventeen general adult ICUs in a discrete area of southwest England.

Patients: 12,793 patients admitted between April 1, 1993 and December 31, 1995.

Measurements: Sociodemographic and severity-of-illness data were collected for all patients admitted to the study units. Formal goodness-of-fit tests were applied and observed mortality was compared with that predicted by using the APACHE III system.

Results: For the group of ICUs as a whole, the risk-adjusted standardized mortality ratio (SMR) was 1.23 (95% confidence intervals, 1.12–1.25). For 11 out of 17 ICUs, the SMR was significantly greater than unity (p < 0.05). Calibration, as tested by Hosmer-Lemeshow statistics, was poor (H2 = 312.54; C2 = 332.85; df = 8; p < 0.01); however, model discrimination was good with a total correct classification rate of 82.9% and an area under the receiver operating characteristic curve of 0.89.

Conclusions: The excess mortality observed after case-mix adjustment using the APACHE III system in this study may be the result of either poor intensive care performance as compared with the United States or a failure of the APACHE III equation to fit the UK data.

Key Words: APACHE III • ICU • outcome • scoring systems • severity of illness • standardized mortality ratio


   Introduction
TOP
 Abstract
 Introduction
Materials and Methods
 Results
 Discussion
 References
 
Recently , considerable efforts have been made to develop models to permit the accurate prediction of hospital mortality for patients treated in ICUs.1 ,2 ,3 ,4 ,5 ,6 The acute physiology, age, chronic health evaluation (APACHE) III prognostic system7 was developed in the United States based on data collected from 17,440 ICU admissions (16,622 patients with complete data) to 42 ICUs between May 1988 and November 1989 and may, therefore, be representative solely of US critical care practice. There is little data regarding its use outside the United States, although one study8 has suggested that it provided good discrimination of hospital mortality in Brazilian ICUs despite a high standardized mortality ratio (SMR) for the whole group of patients studied. In the United Kingdom, APACHE III has been validated only once in a single-center comparison of the APACHE II and APACHE III systems, which again demonstrated good discrimination but an observed mortality that was higher than predicted.9 In order to evaluate the performance of the APACHE III predictive equation when applied to ICU patients in the United Kingdom, we designed a prospective study of case-mix-adjusted ICU and hospital outcome in 17 ICUs in the South of England.


   Materials and Methods
TOP
 Abstract
 Introduction
 Materials and Methods
Results
 Discussion
 References
 
Between April 1, 1993 and December 31, 1995, sociodemographic and severity-of-illness data were collected on all admissions to 17 general ICUs in the Western division of the South Thames Regional Health Authority and the Portsmouth Health District. Sixteen of the participating units were from district general hospitals and one was located in a teaching center.

Patients were excluded from the study if they were younger than 16 years; had a diagnosis of primary burn injuries; died within 4 h of ICU admission; or had been admitted to the ICU because either a separate coronary care or surgical recovery unit did not exist or was not available in the hospital concerned (ie, they did not require intensive care). For the purposes of calculating ICU and hospital mortality ratios, only data from the first admission to ICU in any one hospital admission were considered.

Using the definitions of the APACHE III prognostic system,7 where appropriate, the following data were collected for each patient: sociodemographic factors (age and sex); preexisting comorbidity; reason for admission (medical, elective surgical, or emergency surgical); diagnostic category; and severity of illness (using acute physiologic data). Outcome data (alive or dead on discharge from ICU and hospital) were also recorded. Hospital mortality was measured as the proportion of deaths during hospitalization and therefore included ICU deaths.

Personnel were identified at each unit to be responsible for data collection. Strict written guidelines were provided regarding data definition to minimize inaccuracies in data collection. Standard documentation and training were provided to all ICU personnel. Within each ICU, the collection and storage of the minimum data set was undertaken using a stand-alone computer (Macintosh LC 475; Apple Computers; Cupertino, CA) and specially designed software (Ward Watcher; Critical Audit, Ltd; London, UK). On a quarterly basis, the stored data set of a random 20% of patients in each ICU was cross-checked against the patient's medical and nursing records and other standard documentation. All data were checked for illogical, extreme, or unlikely values but no attempt was made to alter original data. Approval from local Ethics Committees for participation in the study was not required since data used for the study had already been collected for clinical purposes.

For each of 12,793 patients admitted to the study, APACHE III scores and estimates of hospital mortality were calculated using the APACHE III prognostic system in the manner described by Knaus et al.7 Individual patient APACHE III estimates of hospital mortality were summed to give an average predicted risk of mortality for all patients and for given groups of ICU admissions (ie, by age, APACHE III score, surgical status, and system diagnostic category). Mortality ratios, obtained by dividing the observed number of deaths for each group by the predicted number, were used to compare actual with estimated outcome. In addition, 95% confidence intervals (CIs) were calculated for each mortality ratio and continuity-corrected {chi}2 tests were used to analyze the difference between the observed and expected number of deaths.

The statistical analysis of the overall goodness-of-fit was undertaken using several techniques designed to test both discrimination (the ability of the model to discriminate between survivors and nonsurvivors) and calibration (the accuracy of risk predictions). Discrimination was tested using classification matrices and receiver operating characteristic (ROC) curves; 95% confidence intervals were computed for the total correct classification rate (TCCR), ie, the sum of the true-positive and true-negative rates derived from the classification tables. Calibration was tested using the Hosmer-Lemeshow H2 and C2 tests10 ,11 ; the H2 statistic divides the population into equal deciles of risk and the C2 statistic uses bands of risk with approximately equal numbers of patients. In addition, a calibration curve was constructed by plotting the predicted death rates stratified by 10% intervals of mortality (x-axis) against the observed mortality rates (y-axis).


   Results
TOP
 Abstract
 Introduction
 Materials and Methods
 Results
Discussion
 References
 
After exclusions, 12,793 patients were admitted to the study from 17 ICUs (mean per ICU, 752.5; range, 123 to 2,123). Characteristics of these patients are compared with those of the 16,622 patients in the US APACHE III database12 in Table 1 ; demographic data from our study have previously been published.13 Compared with patients in the US database, patients in the South of England were more likely to be male ({chi}2 = 23.6; p < 0.01), older than 45 years of age ({chi}2 = 22.0; p < 0.01), have greater APACHE III comorbidity ({chi}2 = 29.3; p < 0.01), and were more frequently transferred to an ICU from hospital wards and other hospitals ({chi}2 = 310.2 and {chi}2 = 246.3, respectively; p < 0.01). More American patients were admitted from an emergency room or other ICU ({chi}2 = 349.0 and {chi}2 = 116.8, respectively; p < 0.01). In the UK database, there were significantly fewer elective surgical patients ({chi}2 = 34.0; p < 0.01) and significantly more emergency surgical patients ({chi}2 = 217.3; p < 0.01) than in the US database. In the South of England, fewer ICU admissions were the result of trauma ({chi}2 = 75.7; p < 0.01), congestive heart failure ({chi}2 = 70.6; p < 0.01), or drug overdose ({chi}2 = 8.0; p < 0.01). We were unable to compare mean age, mean APACHE III scores, or mean length of ICU stay because insufficient data were available from the original APACHE III database.12


View this table:
[in this window]
[in a new window]

  		Table 1. Patient Characteristics in UK and US APACHE III Databases*

 
In the South of England, overall ICU mortality was 17.8% and varied 3.54-fold between ICUs (9.5 to 32.8%). Similarly, overall hospital mortality was 25.9% and varied 2.9-fold (15.5 to 45.0%; Table 2 ). The rank order of units for ICU mortality was similar, but not identical, to that for hospital mortality. For the group of ICUs as a whole, the risk-adjusted SMR (SMR = observed/predicted mortality) was 1.23 (95% CI, 1.12–1.25). SMRs for the 17 ICUs varied from 1.00 to 1.61. For 11 out of 17 ICUs, the SMR was significantly greater than unity (p < 0.05) (Table 2 ). The remaining six were not different in terms of sample size but, when comparing case mix, they admitted a significantly greater proportion of patients (32.6 vs 19.4%) from the accident and emergency (A&E) department than those with SMRs of > 1.0. This figure is the same as the proportion admitted from emergency rooms to ICUs in the original APACHE III database.7 The distribution of patients by severity of illness (first-day APACHE III score) is shifted to the right for UK patients as compared with those in the US APACHE III study, the cut-off beginning at a first-day APACHE III score of > 45; ie, UK patients are more seriously ill (Fig 1 ).


View this table:
[in this window]
[in a new window]

  		Table 2. ICU Admissions and Deaths and Observed and Predicted Hospital Mortality Rates by ICU

 


View larger version (21K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 1. Distribution of APACHE III scores in the first 24 h after admission to ICU for 12,793 UK patients (shaded bars) and 16,622 US patients (solid bars) showing increased severity of illness in the UK cohort of patients. * = Higher APACHE III score (p < 0.01) using {chi}2 tests. ** = Lower APACHE III score (p < 0.01) using {chi}2 tests).

 
Overall Goodness of Fit
The accuracy of risk prediction was assessed using the method suggested by Hosmer and Lemeshow.10 ,11 The goodness-of-fit tables show observed vs expected numbers of death for 10 groups (equal deciles of risk in Table 3 ) with increasing risk of mortality. Calibration was poor (H2 = 312.54; C2 = 332.85; df = 8; p < 0.01) and inspection of the tabulated values shows that APACHE III risk predictions were consistently lower than actual hospital mortality, particularly for the lower risk groups.


View this table:
[in this window]
[in a new window]

  		Table 3. Overall Goodness of Fit*

 
The calibration curve (Fig 2 ) also reflects this tendency and demonstrates the trend for observed mortality to increase as the predicted risk of hospital death increases. However, in seven of the predicted risk bands at the lower end of the risk spectrum, observed mortality was significantly higher than expected (p < 0.05). For the seven highest predicted risk bands (ie, > 70% risk of hospital death), the observed mortality was similar to that predicted. Risk prediction was most inaccurate in the strata of low predicted risks, and the curve lay closest to the diagonal (predicted = observed mortality) for higher risk groups (> 70% predicted risk). Interpretation for strata of higher predicted risks is, however, limited by the small number of cases.



View larger version (29K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 2. Calibration curve for the APACHE III equation applied to the UK data showing poor fit and apparent excess mortality in 7 of 20 risk categories in which observed risk is significantly higher than that predicted by APACHE III (* = p < 0.05).

 
The discriminative power of the system was tested by classification matrices applying 5% cut-off levels of predicted mortality. The results for decision criteria of 10, 50, and 90% predicted mortality are shown in Table 4 . The best TCCR was high at 82.9% (decision criterion, 40%), but it was lower than the best TCCR quoted in the original APACHE III paper,7 88.1%. Computation of the area under the ROC curve confirmed the good discrimination of APACHE III (area = 0.89). In the original APACHE III, the area under the ROC curve was 0.90.7


View this table:
[in this window]
[in a new window]

  		Table 4. Total Correct Classification Rate and True- and False-Positive Rates for APACHE III*

 
Uniformity of Fit
Observed and predicted mortality were similar for patients aged 16 to 35 years, but observed mortality was significantly higher than that predicted when considering all other age groups (p < 0.05; Table 5 ). In the age group 16 to 35 years, a significantly greater proportion of patients were admitted from the A&E department (38.2 vs 23.5%) than was the case for the study group as a whole. Again, this figure is the same as the proportion admitted from emergency rooms to ICU in the original APACHE III database.7


View this table:
[in this window]
[in a new window]

  		Table 5. Age Range of Admissions, Observed and Actual Deaths, and SMRs

 
Observed mortality was higher than that predicted for both surgical and medical groups of patients (p < 0.05; Table 6 ). Similarly, when patients were grouped by major system diagnostic category, the observed mortality was significantly higher than that predicted for medical patients in the cardiovascular, respiratory, GI, neurologic, and trauma categories, and for surgical patients in the cardiovascular, GI, and hematologic categories.


View this table:
[in this window]
[in a new window]

  		Table 6. Observed and Predicted Deaths and SMRs for Surgical and Nonoperative Patients Within Diagnostic Categories

 
Table 7 demonstrates that, when patients were grouped by the acute physiologic component of the APACHE III score, SMRs were significantly higher in the acute physiologic component range between 10 and 89. Below and above this range (ie, for the 350 patients with the least and the 980 patients with the most severely deranged physiology), the ICUs performed as predicted. Similarly, when patients were grouped by total APACHE III score, SMRs were significantly higher than predicted for 12,076 patients who scored between 0 and 109 APACHE III points. For the most severely ill patients, with the exception of 48 patients whose APACHE III score was 150 to 159, the observed number of deaths was similar to the predicted number.


View this table:
[in this window]
[in a new window]

  		Table 7. SMR in Patients Grouped by Acute Physiology Scores (Acute Physiologic Component of APACHE III)

 

   Discussion
TOP
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
References
 
Our study represents the first reported use of the APACHE III prognostic system in a group of English ICUs and, to date, the largest assessment of the APACHE III system outside the United States. We discovered significant differences in the case mix of patients in our study compared with the US APACHE III database.7 ,12 Notably, patients in the South of England were sicker, had greater comorbidity, and more frequently entered an ICU from the general hospital wards. Their stay in an ICU appears to have been shorter (4.03 vs 4.73 days, but lack of sufficient data makes statistical comparison impossible), and both their ICU mortality (17.8 vs 10%) and hospital mortality (25.9 vs 17%) were significantly higher. Using the APACHE III system to adjust for measured differences in patient case mix and to estimate hospital mortality, the English patients had a significantly higher hospital mortality than predicted (p < 0.05) after correction for age, physiologic derangement, and comorbidity. Overall, hospital mortality for English ICU patients was 25% higher than predicted (SMR, 1.25); this varied among ICUs, with one unit having a hospital death rate 61% higher than predicted. For only 6 of the 17 ICUs studied did the SMR suggest performance was as good as expected based on the APACHE III equation. It is interesting to note that in these units, a significantly higher proportion of patients were admitted directly from A&E departments.

If it is assumed that the APACHE III system is accurate and is sufficiently robust to control for international variations in ICU admission practice, it must be accepted that the performance of 64% (11 of 17) of ICUs in the South of England is poor compared with ICUs in the United States. Indeed, this finding is particularly worrying, as it suggests that the admission of identical groups of patients to units in the United States and the South of England would result in 25% greater hospital mortality in the latter group. Such an excess might be explained by the well-recognized differences between the two countries in the structure and organization of intensive care, availability of technology, and the training of medical staff. For instance, in the United States, resource allocation for intensive care is considerably greater,14 ,15 ,16 intensive care is a recognized specialty,17 ICUs are more likely to have a full-time clinical director, and dedicated critical care training programs have existed for some time.18 Resource shortages in the United Kingdom often lead to refusal of ICU admission,19 ,20 delay in ICU admission,20 a high requirement for interhospital transfers,21 ,22 early ICU discharge,23 and increased ICU re-admission rates,24 all of which have the potential to worsen patient outcome. Many patients are referred late for intensive care25 ; this is reflected in our study by the greater number of ICU patients who were admitted from general wards, rather than from the emergency department as often happens in the United States. Results similar to ours have been reported by Bastos et al,8 who recorded an SMR of 1.67 (range, 1.01 to 2.31) for 10 Brazilian ICUs using APACHE III. Subsequently, the performance of individual units was attributed to the availability of ICU technology.26

A far more likely explanation of the disparity between actual and predicted outcome for our group of patients may be that the APACHE III equation does not fit the South of England data. This hypothesis is supported by the poor calibration and uniformity of fit seen in our study. Although there is good discrimination, some would argue that this has nothing to do with model fit and only with the ordering of the predicted chance of the patients dying. In such circumstances, the performance of UK and US ICUs could be similar, but masked by the inability of the APACHE III system to control for international differences in case-mix factors, ICU admission practices, or lead-time bias.27 Indeed, considerable differences in case-mix factors have been demonstrated between our data and that of the APACHE III database.7 ,12 The APACHE III equation was formulated from a population that was more likely to be female, younger, less sick, with less comorbidity and a greater chance of being admitted directly from the emergency room. Each of these factors might "bias" the APACHE III equation to outcome such that it does not function accurately when applied to a population with a different composition. Similar disparity between the case-mix of test and reference databases were also reported by Bastos et al,8 and they too discovered that the APACHE III showed good discrimination but poor calibration and uniformity of fit.

To our knowledge, there have only been two other large, published multicenter analyses of case-mix-adjusted outcome undertaken in the United Kingdom, both using the APACHE II system.28 ,29 One, undertaken by the UK Intensive Care Society, demonstrated that the goodness of fit of the APACHE II equation to data from 26 ICUs was good.28 However, the Intensive Care Society found poor uniformity of fit when patients were grouped by age, diagnosis, or APACHE II score.28 Nevertheless, the overall SMR was 1.02 (95% CI, 0.98–1.06) and this has been taken to imply that the performance of these 26 ICUs was as good as similar units in the United States. In contrast, the work of Goldhill and Withington29 demonstrated a significantly higher SMR (1.11; 95% CI, 1.06–1.16) for the group as a whole and for certain subroups of patients. They also demonstrated "... substantial differences in case mix... " between their data and the original APACHE II study.29 This pattern of apparent excess mortality and a disparity of case-mix factors between study and reference populations mirrors the results of both our study and that from Brazil.8 The fact that the APACHE III equation does not fit our study data might thus be explained by differences in case mix between the two countries.

Other reasons why there may be a poor fit of the APACHE III equation to the South of England data might include systematic differences in medical definitions or diagnostic labeling, in the effectiveness of therapy, and in data collection (automated data collection systems record measurements more frequently than manual chart systems and are more common in the United States), or a failure of the APACHE III system to weight certain comorbidities sufficiently for their impact on UK mortality. In addition, the effect of lead-time bias may also have a greater impact in the United Kingdom, where Tunnell et al27 have demonstrated significant increases in APACHE III severity-of-illness scores and estimated risk of hospital mortality, if data collected before definitive intervention outside the ICU were included. This might suggest that the factor within APACHE III that corrects for lead-time bias may not necessarily adjust across all patient subgroups. Finally, the validity of using the SMR as an indicator of ICU performance has been questioned previously.27 ,30 ,31 It must be accepted that, although the SMR has been proposed as a measure of ICU performance, there is no alternative independent standard by which SMRs themselves have been or can be compared or validated. Additionally, the SMR reflects the ratio of observed to predicted hospital mortality and looks at the overall system of patient management, not just ICU activity.

In summary, the excess mortality observed in 17 ICUs in the South of England after adjustment for case mix using APACHE III is almost certainly a reflection of a failure of the APACHE III equation to fit the UK data. However, the lack of a universally recognized and verified marker of ICU performance makes it impossible to exclude poor UK intensive care performance as the cause of this excess. Two potential methods of determining whether clinical underperformance is responsible exist. The first involves a change in the method of delivery of health care in the South of England with subsequent reassessment of outcome using the APACHE III system. The second would be to match patients in our study with those of the APACHE III reference database on the basis of case-mix profile and risk predictions and then compare outcome in these groups.


   Acknowledgements
 
We acknowledge the directors and staff of the following ICUs for their cooperation and enthusiasm in the collection of data for this study: Ashford Hospital, Middlesex; Crawley Hospital; East Surrey Hospital, Redhill; Epsom General Hospital; Frimley Park Hospital; Haywards Heath Hospital; Kingston Hospital; Mayday Hospital, Croydon; Queen Mary's Hospital, Roehampton; Royal Surrey County Hospital, Guildford; St. Helier Hospital, Carshalton; St. Peter's Hospital, Chertsey; St. Richard's Hospital, Chichester; Worthing Hospital; General ICU, St. George's Hospital, Tooting; Queen Alexandra Hospital, Portsmouth; and St. Mary's Hospital, Portsmouth.


   Footnotes
 
For editorial comment see page 614.

Correspondence to: G.B. Smith, FRCA, Department of Intensive Care Medicine, Queen Alexandra Hospital, Portsmouth PO6 3LY, UK

Abbreviations: A&E = accident and emergency; APACHE = acute physiology, age, chronic health evaluation; CI = confidence interval; ROC = receiver operating characteristic; SMR = standardized mortality ratio; TCCR = total correct classification rate

Received for publication March 31, 1998. Accepted for publication October 21, 1998.


   References
TOP
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
 References
 

  1. Knaus, WA, Zimmerman, JE, Wagner, DP, et al (1981) APACHE—acute physiology and chronic health evaluation: a physiologically based classification system. Crit Care Med 9,591-597[ISI][Medline]
  2. Knaus, WA, Draper, EA, Wagner, DP, et al (1985) APACHE II: a severity of illness classification sysytem. Crit Care Med 13,818-829[ISI][Medline]
  3. Le Gall, JR, Loirat, P, Alperovitch, A, et al (1984) A simplified acute physiology score for ICU patients. Crit Care Med 12,975-977[ISI][Medline]
  4. Le Gall, JR, Lemeshow, S, Saulnier, F (1993) A new simplified acute physiology score (SAPS II) based on a European/North American Multicenter Study. JAMA 270,2957-2963[Abstract]
  5. Lemeshow, S, Teres, D, Pastides, H, et al (1985) A method for predicting survival and mortality of ICU patients using objectively derived weights. Crit Care Med 13,519-525[ISI][Medline]
  6. Lemeshow, S, Teres, D, Klar, J, et al (1993) Mortality probability models (MPM II) based on an international cohort of intensive care unit patients. JAMA 270,2478-2486[Abstract]
  7. Knaus, WA, Wagner, DP, Draper, EA, et al (1991) The APACHE III prognostic system: risk prediction of hospital mortality for critically ill hospitalized adults. Chest 100,1619-1639[Abstract/Free Full Text]
  8. Bastos, PG, Sun, X, Wagner, DP, et al (1996) Application of the APACHE III prognostic system in Brazilian intensive care units: a prospective multicenter study. Intensive Care Med 22,564-570[ISI][Medline]
  9. Beck, DH, Taylor, BL, Millar, B, et al (1997) Prediction of outcome from intensive care: a prospective cohort study comparing Acute Physiology and Chronic Health Evaluation II and III prognostic systems in a United Kingdom intensive care unit. Crit Care Med 25,9-15[CrossRef][ISI][Medline]
  10. Lemeshow, S, Hosmer, DW (1982) A review of goodness of fit statistics for use in the development of logistic regression models. Am J Epidemiol 115,92-106[Abstract/Free Full Text]
  11. Hosmer, DW, Lemeshow, S (1989) Applied logistic regression. ,135-149 John Wiley & Sons New York.
  12. Knaus, WA, Wagner, DP, Draper, EA, et al (1993) Variations in mortality and length of stay in intensive care units. Ann Intern Med 118,753-761[Abstract/Free Full Text]
  13. Pappachan, JV, Millar, BW, Bennett, ED, et al (1997) A study of the APACHE III prognostic system in 17 UK intensive care units: casemix and outcome. Clin Intensive Care 8,101
  14. Chalfin, DB, Fein, AM (1994) Cost-containment issues in the United States: critical care medicine in managed competition and a managed care environment. New Horiz 2,275-282[Medline]
  15. Bion J. Cost containment: Europe—The United Kingdom New Horiz 1994; 2:341–344
  16. National ITU audit 1992/1993. London, UK: Royal College of Anaesthetists, 1994
  17. Grenvik, A, Leonard, JJ, Arens, JF, et al (1981) Critical care medicine: certification as a multidisciplinary subspecialty. Crit Care Med 9,117-125[ISI][Medline]
  18. Task Force on Guidelines, Society of Critical Care Medicine. Recommendations for program content for fellowship training in critical care medicine. Crit Care Med 1987; 15:971–977
  19. Mitchell, I, Grounds, M, Bennett, D (1995) Intensive care in the ailing UK health care system. Lancet 345,652
  20. Metcalfe, MA, Sloggett, A, McPherson, K (1997) Mortality among appropriately referred patients refused admission to intensive care units. Lancet 350,7-11[CrossRef][ISI][Medline]
  21. Mackenzie, PA, Smith, EA, Wallace, PGM (1997) Transfer of adults between intensive care units in the United Kingdom: postal survey. BMJ 314,1455-1456[Free Full Text]
  22. Smith, GB, Taylor, BL, McQuillan, PJ, et al (1995) Rationing intensive care: intensive care provision varies widely in Britain. BMJ 310,1412-1413[Free Full Text]
  23. Bull, AR (1995) Rationing intensive care: crisis reactions to shortages should be replaced by analysis of strategic services. BMJ 310,1010[Free Full Text]
  24. Crighton, IM, Winter, RJ (1997) Failure to recognise the need for readmission to an intensive care or high dependency unit. Br J Intensive Care 7,46-48
  25. Pilkington, S, McQuillan, PJ, Allan, AJ, et al (1998) Confidential inquiry into quality of care before admission to intensive care. BMJ 316,1853-1858[Abstract/Free Full Text]
  26. Bastos, PG, Knaus, WA, Zimmerman, JE, et al (1996) The importance of technology for achieving superior outcomes from intensive care. Intensive Care Med 22,664-669[ISI][Medline]
  27. Tunnell, RD, Millar, BW, Smith, GB (1998) The effect of lead time bias on severity of illness scoring, mortality prediction and standardized mortality ratio in intensive care: a pilot study. Anaesthesia 53,1045-1053[CrossRef][ISI][Medline]
  28. Rowan, KM, Kerr, JH, Major, E, et al (1993) Intensive Care Society's APACHE II study in Britain, and Ireland: II. outcome comparisons of intensive care units after adjustment for casemix by the American APACHE II method. BMJ 307,977-981
  29. Goldhill, DR, Withington, PS (1996) The effect of casemix adjustment on mortality as predicted by APACHE II. Intensive Care Med 22,415-419[CrossRef][ISI][Medline]
  30. Boyd, O, Grounds, RM (1993) Physiological scoring systems and audit. Lancet 341,1573-1574[CrossRef][ISI][Medline]
  31. Goldhill, DR, Withington, PS (1996) Mortality predicted by APACHE II: the effects of changes in physiological values and post-ICU hospital mortality. Anaesthesia 51,719-723[ISI][Medline]



This article has been cited by other articles:


Home page
 ChestHome page
A. Garland
Improving the ICU: Part 1
Chest, June 1, 2005; 127(6): 2151 - 2164.
[Abstract] [Full Text] [PDF]

Home page
 JAMAHome page
T. Stefanec
Diuretics in Critically Ill Patients With Acute Renal Failure
JAMA, March 19, 2003; 289(11): 1380 - 1380.
[Full Text] [PDF]

Home page
 ChestHome page
C. A. Sirio, K. Tajimi, N. Taenaka, Y. Ujike, K. Okamoto, and H. Katsuya
A Cross-Cultural Comparison of Critical Care Delivery : Japan and the United States
Chest, February 1, 2002; 121(2): 539 - 548.
[Abstract] [Full Text] [PDF]

Home page
 ChestHome page
D. A. Cook
Performance of APACHE III Models in an Australian ICU
Chest, December 1, 2000; 118(6): 1732 - 1738.
[Abstract] [Full Text] [PDF]

Home page
 ANN INTERN MEDHome page
R. M. Poses, D. K. McClish, W. R. Smith, E. C. Huber, F. L. W. Clemo, B. P. Schmitt, D. Alexander, E. M. Racht, and C. C. Colenda III
Results of Report Cards for Patients with Congestive Heart Failure Depend on the Method Used To Adjust for Severity
Ann Intern Med, July 4, 2000; 133(1): 10 - 20.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF) Free
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Article Archive
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via ISI Web of Science (41)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Pappachan, J. V.
Right arrow Articles by Smith, G. B.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Pappachan, J. V.
Right arrow Articles by Smith, G. B.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS