Journal of Critical Care 29 (2014) 472.e7–472.e12

Contents lists available at ScienceDirect

Journal of Critical Care journal homepage: www.jccjournal.org

An analysis of medicine costs of adult patients on a critical care unit☆ Sumayah Abdul-Jabbar, MPharm a,⁎, Ian Bates, BPharm, MSc, FFRPS b, Graham Davies, BPharm, MSc (Clinical Pharmacy), PhD a, Rob Shulman, MRPharmS, DipClinPharm, DHC(Pharm), FFRPS c a b c

Institute of Pharmaceutical Science, King's College London, London, UK School of Pharmacy, University College London, London, UK Department of Pharmacy, University College London Hospitals NHS Foundation Trust, London, UK

a r t i c l e Keywords: Critical care Intensive care Drug cost Severity of illness Medicine costs

i n f o

a b s t r a c t Purpose: To evaluate the costs of medicines used to treat critically ill patients in an intensive care environment and to correlate this with severity of illness and mortality. Materials and Methods: The study was conducted at a London Teaching Hospital Critical Care Unit. Data were collected for patients who were either discharged or died during September 2011 and stayed longer than 48 hours. The drug cost was related to 150 drugs that were then related to patient's acuity and outcome. Results: The median daily drug cost of the 85 patients was £26. The highest cost patients in the 85th percentile had significantly higher daily drug costs (median, £403) and higher scores for patient acuity. Patients with hematologic malignancy had a median daily drug cost (£561) more than 20 times higher than those without. A regression analysis based on patient's diversity explained 93% of the variance in the daily drug cost. Conclusions: Although the median daily drug cost for an adult critically ill patient was low, this cost significantly escalated with patient acuity and hematologic malignancy. A reference method has been designed for an indepth evaluation of daily drug cost that could be used to compare expenditure in other units. © 2014 Elsevier Inc. All rights reserved.

1. Introduction The current financial climate, together with the escalation in health care expenditure as a result of advancements in therapeutic modalities, patients' expectations, and an increasingly aging population, imposes a responsibility on health care professionals to use resources efficiently. Therefore, it is essential to optimize the use of resources in expensive specialities such as critical care because they make a significant contribution to the overall hospital costs [1]. The high costs are related to the use of sophisticated equipment, specialized pharmacotherapy, high staff-to-patient ratio, and the need for a highly trained workforce [2]. Published literature suggests that critical care costs are significantly higher than that of a general ward [1], with one study reporting that the daily cost of treating a patient in an intensive care unit (ICU) was up to 5 times higher than for treating patients in a ward setting [3]. Drug expenditure for each patient is a reflection of their chronic condition, severity of illness, and acute treatment. The cost impact of drugs used in critical care on the overall hospital drug expenditure is significant [4]. A study conducted in the United States (2003) reported ☆ Conflict of interest: None. ⁎ Corresponding author. E-mail addresses: [email protected], [email protected] (S. Abdul-Jabbar). 0883-9441/$ – see front matter © 2014 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.jcrc.2013.12.014

that ICU drugs accounted for 38% of the overall hospital drug costs. The study also reported that the annual rate of increase of ICU drugs cost was double that of non-ICU drugs (12% vs 6%) [5]. Although this is an important finding, indicating a common trend, it may not reflect typical UK cost because, in the US study, the ICU bed count accounted for one fifth of all hospital beds, higher than typically found in the UK setting. This finding is supported by a more recent study where it was reported that the United States has 7 times as many ICU beds per capita as the United Kingdom [6]. It should also be remembered that although newer, novel drugs are often used in the ICU setting, many of the branded drugs are now available as generics, with a lower associated cost than when first introduced. This is clearly important when determining the impact of earlier studies, such as that undertaken in 2003 by Weber and colleagues [5]. Therefore, a study is required to provide a more contemporary view of drug expenditure in this setting. Although there is literature focused on an aspect of critical care using a cost-effectiveness analysis model [7] (eg, drotrecogin alfa vs placebo [8] and mechanical ventilation vs nonmechanical ventilation [9]), there is less in the literature focusing on drug expenditure in general. Furthermore, those few published have been reported to have methodological bias [10] because often the use of hospital charges or bills does not reflect the actual expenses [5], or the use of average bed day price assumed constant expenditure over the entire stay [11], or using cost based on diagnosis rather than severity of

472.e8

S. Abdul-Jabbar et al. / Journal of Critical Care 29 (2014) 472.e7–472.e12

illness was misleading [10]. Any future studies should aim to minimize these types of bias in the design process. Intensive care units typically produce monthly drug expenditure reports, which are limited by their lack of specificity because they relate to all admissions from a variety of acuity and chronic conditions and a range of length of stay. There has been no previous attempt to relate the cost to patients' diversity and case mix. Consequently, our study profiles patients in terms of severity of illness, speciality, source of admission, mortality, length of stay, drugs used, and the drug cost per patient-day. The objective of this study was to evaluate the components that influence daily drug cost (DDC) per patient. A method was designed to calculate the DDC for each individual patient, focusing on the drugs that significantly contribute to expenditure. This method could serve as a way to compare drug expenditure between units, both in the United Kingdom and internationally, as well as in the non-ICU setting to compare within and between other specialities. 2. Methods 2.1. Hospital setting and study population A retrospective evaluation and analysis of drug cost per patientday within a 35-bed general adult critical care unit at a London teaching hospital housing medical admissions, transfers from other hospitals, and general surgical patients (but not postoperative cardiac or neurologic patients) was performed. The monthly trend in cost per bed day throughout 2011 was reviewed to identify any variance. September appeared to be a typical month and was selected for more in-depth analysis. The sample included patients who were either discharged or died in September 2011. Patients with a length of stay of 48 hours or less were excluded because it was agreed that patients staying in the unit for short periods were typically admitted for postanesthetic care. It was considered that patients with short-stay recovery would have diluted the data, making it less representative of the critically ill. An earlier study showed that postanesthetic care patients represented 34% of the low-cost group [12]. Those discharged from and readmitted to the unit during the study month were included in the final analysis [5]. 2.2. Cost composition and selection of drugs Because this study focused on drug costs alone, only the data relating to the hospital acquisition cost, plus value added tax, at 20%, were included in the analysis. Consequently, the cost of any reconstitution fluids, equipment, or labor used in the process of reconstituting and administering the medicines were not captured because these were not easily quantifiable and considered beyond the remit of this study. Because collecting drug use on an individual patient basis is timeconsuming, limiting the numbers included in the analysis, a streamlined method was devised to maximize the number of patients evaluated. The Pharmacy Information Management System, Proprietary System (London), reported that 475 drugs (medicines, immunoglobulin, hemofiltration fluid, parenteral nutrition, and intravenous fluids) were used within the ICU during the study month. Of these 475 drugs, 150 accounted for 97% of the total expenditure, as determined by an examination of the cumulative expenditure data. These 150 drugs were categorized into therapeutic classes and used as the basis to determine drug use and cost in our study population. 2.3. Data collection The daily drug use per patient was manually extracted from the computerized ICU information system (qs GE Medical, GE Healthcare, Anapolis, MD) by viewing retrospective administration records. Each patient-day was reviewed against the list of the 150 drugs.

A spreadsheet was used to enter the unit cost of each drug within its therapeutic class for the corresponding day of stay. The unit cost for each drug was calculated by dividing pack price by size to determine cost per tablet, capsule, infusion bag or injection or where vials and ampoules were used, and the number required to administer the dose. The cost of any oral liquids administered was based on the dose volume rather than the whole bottle. Inhaler cost was taken as that for the whole pack because the device could not be reused by another patient. The list did not include any topical creams or eye drops because these were in the low cost (3%) and thus excluded from the data collection. 2.4. Measurement of acute illness morbidity and outcome Data routinely recorded by the unit to reflect both morbidity and mortality were collected to represent the patients' health acuity at admission, during their stay, and on discharge. These reflect the mandatory requirement under the Critical Care Minimum Dataset used by the National Health Service for remuneration. Thus, the Critical Care Minimum Dataset is a useful indicator of workload in ICUs within the United Kingdom [13]. On admission, the Acute Physiology and Chronic Health Evaluation II (APACHE II) score was used to reflect acuity because it has been shown to be a good predictor of outcome and to significantly correlate with overall resource use for the entire ICU stay [14,15]. During the patients' stay, factors known to influence expenditure were recorded, including organ system supports and levels of care. Assessment of changes in organ function plays a role in describing the magnitude of a patient's acute illness. Therefore, the type and number of organ supported reflect resource use and, in turn, the extent of the patient's morbidity. Organ support was described in 2 ways. The first was based on the use of mechanical ventilation during the stay (respiratory support). The second was a cardiovascular-renal index. This index was readily available because it could be determined from the database of drug usage itself and was based on an aggregate established from both cardiovascular and renal organ support. Level of care (LoC) classifies patients based on their needs using definitions provided by the UK Department of Health. Level 2 corresponds to patients requiring more detailed observation and intervention or have a single-organ support (“high dependency” care), and level 3 describes those requiring mechanical ventilation and also where there is a multiorgan support (“intensive” care). The unit has combined level 2 and 3 beds, and patients often fluctuate between the 2 LoCs during their stay, so it is not possible to provide a definitive number within each level for the study month. Consequently, to overcome this problem, patients were allocated to an overall LoC based on a simple frequency calculation (aggregate) and on their length of stay. In addition, any patients who deteriorated during their stay, that is, move up from level 2 to 3, were identified for further analysis, as it was hypothesized that such patients could add to the overall costs, whereas the outcome of ICU stay was based on patients' mortality. 2.5. Data analysis and presentation Descriptive statistics were used to summarize patients' characteristics and costs. Data were expressed as median and interquartile range (IQR), whereas categorical data were presented as number and percentage. Average drug cost per day was calculated for each patient as the total cost of stay divided by the length of stay. This was referred to as the DDC for each patient. The trend in the DDC was presented using the percentage cost per day ([total day cost/total cost of stay] × 100). Because DDC was not normally distributed, nonparametric statistical tests were used. Comparison between groups was carried out by Mann-Whitney (2 groups) and Kruskal-Wallis tests (N2 groups). Spearman correlation (r values) coefficient was used to

S. Abdul-Jabbar et al. / Journal of Critical Care 29 (2014) 472.e7–472.e12

sample was 63 years (IQR, 47-73). The demographic details are shown in Table 1.

Table 1 Demographic characteristics of the patients (n = 85) Characteristics Sex Male Female Source of admission Ward Accident and emergency Another hospital ward Another hospital critical care unit Speciality Hematology Oncology Medicine and infectious disease Surgery Cardiothoracic and vascular General, gastrointestinal, urology, and colorectal Other surgery

472.e9

No. (%)

3.2. Drug expenditure 44 (51.8) 41 (48.2) 60 (70.6) 16 (18.8) 6 (7.1) 3 (3.5) 9 (10.6) 3 (3.5) 10 (11.8) 63 (74.1) 21 (24.7) 33 (38.8) 9 (10.6)

investigate relationships. Statistical significance was considered when a P value was less than .05. To allow comparison of patients by expenditure, an “expensive” category had to be created. Patients in the 85th percentile of DDC were considered to be expensive because this “cutoff,” determined by visual inspection of the total drug expenditure data for each patient, appeared to reflect a difference between 2 groups of patients (low and high expense). The characteristics of each group were determined and compared with each other. Multiple regression was carried out following an exploratory simple regression to determine the variables that best predicted the outcome (DDC per patient). For the model to be significant, both F ratio and t statistics for each variable should have a P value less than .05. Statistical analysis was performed using Statistical Package for the Social Sciences (SPSS), version 17 (Chicago, Ill).

3. Results 3.1. Patients' characteristics A total of 203 patients either were discharged from the unit or died during the study period. Of these, 87 had an ICU stay of longer than 48 hours and were included in the analysis. Two patients were readmitted to ICU during the study period, resulting in a final sample of 85 patients, representing 87 discharges. The median age of the

The overall drug expenditure for the study sample was £114 083, with a median DDC of £26 (IQR, 15-46). Fig. 1 illustrates the drug expenditure with respect to speciality. Because hematology patients accounted for the greatest portion of the total cost (80%), comparison was carried out vs nonhematology patients. The median DDC for a hematology patient (£561) was significantly more than that for nonhematology patient (£23; Mann-Whitney U = 38, P b .001). The nonhematology specialities did not have any significant intergroup difference in their DDC (Kruskal-Wallis H(4) = 3.91, P = .08. 3.3. Length of stay and allocation of drug cost Data were collected for a total of 775 patient-days (number of patients multiplied by the cumulative length of the stay), with a median length of stay of 4 days (IQR, 3-8).Consequently the cost of the 4 days was taken as 100% of the stay to represent the percentage drug cost on each day (Fig. 2). The drug expenditure for day 2 was significantly higher than for any other, progressively declining thereafter. When the median length of stay was further explored, only 4 patients had died, whereas 40 were discharged. 3.4. Acute illness morbidity and outcome The relationship between acuity variables and DDC was examined using correlation. The cardiovascular-renal index had the strongest association with DDC (Spearman r = 0.458, P ≤ .001) followed by aggregated LoC (r = 0.417, P ≤ .001). The APACHE II score had a nonsignificant trend toward correlation (r = 0.206, P = .055). Thirty-seven (43%) patients were mechanically ventilated at some point during their stay. The DDC of these patients (median, £27; IQR, 19-97) was significantly higher than that of those who did not receive mechanical ventilation (median, £22; IQR, 13-41; Mann-Whitney U = 689, P = .03). The median DDC of the “deteriorating” patients (defined as increased their LoC during their stay) was £269, 11 times the expense of those who did not. Within the data collection period, the overall sample mortality was 13%. Nonsurvivors were significantly more expensive (DDC: median, £136 [IQR, 46-795]) than survivors (DDC: median, £23 [IQR, 15-41]; Mann-Whitney U = 122, P b .001).

Fig. 1. Drug expenditure by speciality.

472.e10

S. Abdul-Jabbar et al. / Journal of Critical Care 29 (2014) 472.e7–472.e12

patients for Pneumocystis jirovecii and accounted for 50% of their total antibacterial expenditure. Similarly, rasburicase and terlipressin were only used for hematology patients and together accounted for 97% of their miscellaneous drug expenditure. Although drug expenditure in the cardiovascular, immune response, antibacterial and miscellaneous category was significantly different between hematology and nonhematology patients, antifungals and parenteral nutrition were the major contributors in each patient group. Antifungal drugs contributed 35% of the total cost of hematology patients, and parenteral nutrition and intravenous fluids accounted for 23% of the total cost of nonhematology patients. The pattern of anti-infective (antibacterial, antiviral, antifungal, and antiprotozoal) use within the ICU is of a particular interest and can reflect the use of multiple agents for empiric therapy or may be a result of positive cultures. Both situations could be associated with higher costs. To confirm this, the number of broad-spectrum antiinfectives prescribed during their stay was compared with DDC. The DDC significantly increased as the number of anti-infective classes prescribed increased (Kurskal-Wallis H(4) = 33, P b .001). Fig. 2. The allocation of drug cost for 4 days.

3.6. Expensive patients 3.5. Drug use The total drug expenditure for each therapeutic class and the median DDC per speciality were determined and are shown in Table 2. Antifungal and miscellaneous drugs (acetylcysteine, insulin, rasburicase, terlipressin) accounted for 50% of the total drug expenditure (£114 083). Liposomal amphotericin (Ambisome, Gilead Sciences Ltd., Cambridge, UK) costs were 51% of the antifungal expenditure and 12% of the total drug expenditure. Rasburicase accounted for 87% of the miscellaneous drug expenditure and 13% of the total drug expenditure. The “Other surgery (maxillofacial, orthopaedics, and obstetrics)” patients used only 9 of the 14 therapeutic classes and had the least total drug expenditure (see Fig. 1). An analysis was conducted to establish any association between therapeutic class and DDC for both hematology and nonhematology patients. The DDC for certain categories (cardiovascular, immune response, antibacterial and miscellaneous drugs) was significantly different when hematology and nonhematology patients were considered (P b .05). Drugs in the cardiovascular, immune response, and antibacterial categories were used more extensively by hematology patients. Co-trimoxazole was only used to treat hematology

Percentile analysis showed the patients in the 85th percentile of DDC accounted for 77 % of the total DDC. Patients in the 85th percentile (“expensive patients”) tended to be male (69% vs 47%) and hematology patients (54% vs 3%) when compared with patients who were relatively less expensive. The median DDC and length of stay for expensive patients were significantly higher (P b .001) than those for nonexpensive patients (£403 [135-809], 10 days [7-19], and £22 [15-39], 4 days [3-7], respectively). For morbidity, patients classed as expensive had a higher severity of illness. This was reflected by a significantly higher percentage of expensive patients using mechanical ventilation and deteriorating in terms of LoC. In addition, 46% of the expensive patients died compared with 7% for those who were less expensive. Those who were not in the 85th percentile were distinctively different in that they received no antifungal agents, which were the major contributor toward expenditure. 3.7. Multiple regression analysis Following an exploratory regression, some variables were rejected as predictors for DDC (dependent variable). Those rejected in a simple

Table 2 Total drug expenditure for each therapeutic class and median DDC for each class within the specialities Therapeutic class

Median DDC £ (IQR) Hematology

Total drug expenditure (£)

Nonhematology Oncology Medicine and Cardiothoracic General, gastrointestinal, Other surgery infectious disease and vascular urology, colorectal

Parenteral anticoagulant Cardiovasculara Respiratory Immune responsea (including immunoglobulin) Antibacteriala Antifungal Antiviral Antiprotozoal (antibacterial indication) Analgesia, sedation and anesthesia Gastrointestinal Nervous system Parental nutrition and fluids Hemofiltration fluids Miscellaneousa Total (£)

6 (1-51) 12 (9-141) 104 (0-247) 7 (2-66) 17 (11-23) 359 (218-518) 7 (4-45) 7 (0) 14 (2-23) 1 (0-2) 206 (0) 11 (5-20) 14 (3-19) 114 (6-434) 91 448

– 4 (0) – 3 (0) 5 (0) 42 (0) – – 4 (0) 1 (0) – 5 (0) 5 (0) 6 (0) 2388

1 (1-2) 2 (1-3) – – 3 (1-7) – 24 (0) – 5 (1-9) 1 (0-2) 60 (0) 5 (2-12) 6 (0) 5 (2-7) 2471

Blank cells indicate that speciality did not use the therapeutic class. a DDC was significantly different between hematology and nonhematology patients.

1 (1-3) 3 (1-9) 21 (0) 2 (0-4) 5 (2-7) – 7 (0) – 2 (1-4) 1 (0-2) 44 (0) 5 (2-13) 10 (0) 4 (3-7) 6624

1 (0-1) 3 (1-8) 2 (0-2) 1 (1-2) 5 (2-10) 344 (0) – – 3 (1-10) 1 (1-2) 3 (0) 5 (2-16) 12 (0) 3 (2-7) 10 189

1 (0-1) 3 (0) – 2 (1-3) 2 (1-5) – – – 6 (4-14) 2 (0) 3 (0) 4 (2-12) – 2 (2-3) 962

3598 9593 8534 4991 7158 34 619 4029 205 3452 349 7329 6842 1718 21 667 114 083

S. Abdul-Jabbar et al. / Journal of Critical Care 29 (2014) 472.e7–472.e12

regression were as follows: sex, age, source of admission, and use of mechanical ventilation (P N .05). Aggregated LoC, APACHE II score, cardiovascular-renal index, and mortality were significant predictors in a simple regression but were insignificant in a multiple regression analysis (P N .05). However, hematologic malignancy, deterioration, number of antiinfectives (ie, antifungal, antibiotic, antiviral, or antiprotazoal), and day 2 of admission drug cost were significant predictors in multiple regression and were used to produce a model for predicting DDC per patient. The model generated from those 4 variables had an R 2 of 0.929 and a probability of F test less than .001.

4. Discussion Our Unit's annual drug expenditure was £1.4 million (equivalent to $2.2 million) at the time of this study, a figure significantly lower than the range reported of between $6 and 12 million during an earlier study conducted in the United States in 2002 and may reflect important differences between patients treated in the ICU setting in these two countries [5,6]. A study involving a number of ICUs in the United Kingdom found a median drug cost per day of £111.74 (IQR, 91.28-143.10) [16], considerably higher than reported in our study. The authors acknowledged that the cost block methodology used was not designed to study cost difference between patients and that the results may not be representative of all ICUs in the United Kingdom [16]. This may, in part, explain the differences seen between the two studies because we believe that our method has the advantage of analyzing the cost with respect to patient's case mix and can determine a more accurate picture of the actual costs incurred. Furthermore, our study was able to explore the variation in DDC with respect to patients' morbidity and mortality. It is no surprise to find that the DDC was significantly higher in those patients who died compared with those who were discharged. This probably reflects that death is often preceded by a worsening of organ function and the use of maximal treatment, including drug therapy, with a subsequent associated cost. Although this is supported by the work of others [17,18], care should be taken before concluding that “nonsurvivors” always cost more. This is because patients who die rapidly result in shorter durations of stay, which will have an effect on overall cost. In addition, the APACHE II score has been shown previously to correlate well with mortality [12] but not with drug expenditure. Our study found a trend in this relationship (P = .055) suggesting that its impact on cost needs further evaluation. This is the first study to explore the value of the cardiovascularrenal index in explaining the DDC. The correlation reported suggests that it is a useful measure for use in future studies, especially because it is readily available from routine records. Daily scoring systems (eg, therapeutic intervention scoring system) were not used to evaluate the severity of illness because these data were not available in the study setting. We confirmed the findings of others that respiratory support is associated with a higher cost within intensive care [19]. The DDC was low on day 1 of admission, significantly highest on day 2, and declined thereafter (Fig. 2). A possible explanation to this is that day 1 could be a fraction of a day or a diagnosis day, followed by day 2 being the first comprehensive therapy day. Because only 4 patients died by the fourth day, the progressive decline in expenditure thereafter could be attributed to patient stabilisation. The designed method had the advantage of also explaining DDC in terms of drug use pattern. Costly antifungal treatment is used typically for treatment of suspected or confirmed invasive fungal infection in hematology patients [20]. In this study, antifungals were used to treat 67% of hematology patients. Multiagent antibacterial therapy is also typical within hematology; co-trimoxazole is the first-line treatment of P jirovecii [21].

472.e11

Inotropes have previously been identified to represent an insignificant drug cost in intensive care but a significant cost in patients who use the highest percentage of resource [5]. In this study, expensive patients (including hematology and nonsurvivors) were high users of inotropes. Antibacterial DDC was significantly different between survivors and nonsurvivors. Previously, it was found that the number of antibacterial agents used correlated with intensive care mortality [22]. The treatment of infections within intensive care has been studied widely and was always associated with high drug expenditure [4,23]. In this study, the number of broad anti-infectives used directly affected the DDC. The extensive diversity of patients treated within intensive care [24] was reflected on the multiple regression model built as a predictive tool for DDC. It took into account case mix (speciality and LoC), drug use in terms of anti-infectives, and day 2 drug cost. It explained 93% of the variance in the DDC. The present study designed a method for evaluating the DDC for patients within intensive care by calculating a value rather than one averaged across all patients. This is very time-consuming to calculate, hence the use of our pragmatic method of capturing the top 150 drugs, accounting for 97% of expenditure. We suggest that this method is more appropriate and sensitive to use because it reflects the type of patient and acuity rather than the traditional, simpler method of dividing total drug cost by number of bed days occupied [17]. Therefore, a method with minimum bias was successfully designed [11]. Furthermore, by excluding the lowest 3% of drug cost, the DDC integrity has not been significantly compromised because this range of drugs has a minimum impact on overall drug cost. The ICU hematology patients had a higher DDC compared with all other specialities, often protocol driven, suggesting that in future studies, their costs should be reported separately. The main limitation of the study was that excluded patients were not analyzed to confirm that they were actually postoperative recovery patients rather than patients dying within 2 days. In conclusion, it is crucial to appreciate the diversity and heterogeneity of the ICU patients to maintain the accuracy of cost evaluations. Taking all these variables mutually would allow for a better understanding of resource allocation and, hence, resource optimization within this high-demand speciality. This study identified the variables that influenced drug cost and also designed a reference method to evaluate the DDC per patient. This new patient-focused approach is unique and warrants further use. We invite others to compare their Unit's cost by following the method we have described, which would allow comparisons of clinical efficiency in the treatment for similar patients. This method has a potential in the development of electronic prescribing systems to provide direct download of this type of data. References [1] Edbrooke D, Hibbert C, Ridley S, Long T, Dickie H. The development of a method for comparative costing of individual intensive care units. The Intensive Care Working Group on Costing. Anaesthesia 1999;54(2):110–2. [2] Kahn JM. Understanding economic outcomes in critical care. Curr Opin Crit Care 2006;12(5):399–404. [3] Shorr AF. An update on cost-effectiveness analysis in critical care. Curr Opin Crit Care 2002;8(4):337–43. [4] Biswal S, Mishra P, Malhotra S, Puri GD, Pandhi P. Drug utilization pattern in the intensive care unit of a tertiary care hospital. J Clin Pharmacol 2006;46(8):945–51. [5] Weber RJ, Kane SL, Oriolo VA, Saul M, Skledar SJ, Dasta JF. Impact of intensive care unit (ICU) drug use on hospital costs: a descriptive analysis, with recommendations for optimizing ICU pharmacotherapy. Crit Care Med 2003;31(1 Suppl): S17–24. [6] Wunsch H, Angus DC, Harrison DA, Linde-Zwirble WT, Rowan KM. Comparison of medical admissions to intensive care units in the United States and United Kingdom. Am J Respir Crit Care Med 2011;183(12):1666–73. [7] Pines JM, Fager SS, Milzman DP. A review of costing methodologies in critical care studies. J Crit Care 2002;17(3):181–6. [8] Neilson AR, Burchardi H, Chinn C, Clouth J, Schneider H, Angus D. Costeffectiveness of drotrecogin alfa (activated) for the treatment of severe sepsis in Germany. J Crit Care 2003;18(4):217-2.

472.e12

S. Abdul-Jabbar et al. / Journal of Critical Care 29 (2014) 472.e7–472.e12

[9] Mayer SA, Copeland D, Bernardini GL, Boden-Albala B, Lennihan L, Kossoff S, et al. Cost and outcome of mechanical ventilation for life-threatening stroke. Stroke 2000;31(10):2346–53. [10] Gyldmark M. A review of cost studies of intensive care units: problems with the cost concept. Crit Care Med 1995;23(5):964–72. [11] Bams JL, Miranda DR. Outcome and costs of intensive care. A follow-up study on 238 ICU-patients. Intensive Care Med 1985;11(5):234–41. [12] Noseworthy TW, Konopad E, Shustack A, Johnston R, Grace M. Cost accounting of adult intensive care: methods and human and capital inputs. Crit Care Med 1996;24(7):1168–72. [13] Kinsella G, Thomas AN, Taylor RJ. Electronic surveillance of wall-mounted soap and alcohol gel dispensers in an intensive care unit. J Hosp Infect 2007;66(1):34–9. [14] Sage WM, Rosenthal MH, Silverman JF. Is intensive care worth it? An assessment of input and outcome for the critically ill. Crit Care Med 1986;14(9):777–82. [15] Knaus WA, Draper EA, Wagner DP, Zimmerman JE. APACHE II: a severity of disease classification system. Crit Care Med 1985;13(10):818–29. [16] Edbrooke DL, Ridley SA, Hibbert CL, Corcoran M. Variations in expenditure between general intensive care units in the UK. Anaesthesia 2001;56:208–16. [17] Jacobs P, Edbrooke D, Hibbert C, Fassbender K, Corcoran M. Descriptive patient data as an explanation for the variation in average daily costs in intensive care. Anaesthesia 2001;56(7):643–7.

[18] Girotti MJ, Brown SJ. Reducing the costs of ICU admission in Canada without diagnosis related case-mix groupings. Can Anaesth Soc J 1986;33(6):765–72. [19] Dasta JF, McLaughlin TP, Mody SH, Piech CT. Daily cost of an intensive care unit day: the contribution of mechanical ventilation. Crit Care Med 2005;33(6): 1266–71. [20] Ananda-Rajah MR, Cheng A, Morrissey CO, Spelman C, Dooley M, Neville AM, et al. Attributable hospital cost and antifungal treatment of invasive fungal diseases in high-risk hematology patients: an economic modelling approach. Antimicrob Agents Chemother 2011;55(5):1953–60. [21] Lang A, De Fina G, Meyer R, Aschbacher R, Rizza F, Mayr O, et al. Comparison of antimicrobial use and resistance of bacterial isolates in a haematology ward and an intensive care unit. Eur J Clin Microbiol Infect Dis 2001;20(9):657–60. [22] Bernieh B, Al Hakim M, Boobes Y, Siemkovics E, El Jack H. Outcome and predictive factors of acute renal failure in the intensive care unit. Transplant Proc 2004;36(6): 1784–7. [23] Hartmann B, Junger A, Brammen D, Röhrig R, Klasen J, Quinzio L, et al. Review of antibiotic drug use in a surgical ICU: management with a patient data management system for additional outcome analysis inpatients staying more than 24 hours. Clin Ther 2004;26(6):915–24. [24] Ridley S, Burchett K, Gunning K, Burns A, Kong A, Wright M, et al. Heterogeneity in intensive care units: fact or fiction? Anaesthesia 1997;52(6):531–7.