ORIGINAL ARTICLE
Attributable Length of Stay and Mortality of Major Bleeding as a Complication of Therapeutic Anticoagulation in the Intensive Care Unit Najib T. Ayas, MD,*†‡ Peter M. Dodek, MD,*† Hong Wang,† Robert Fowler, MD,§ Hubert Wong, PhD,†k¶ and Monica Norena, MSc† Objective: The aim of this study was to determine the attributable length of stay and mortality due to bleeding as a complication of therapeutic anticoagulation in intensive care unit (ICU) patients. Methods: Charts of patients from 7 ICUs in British Columbia were screened daily for the occurrence of major bleeding while receiving therapeutic heparin. To determine attributable length of stay and mortality, a matched and unmatched cohort design as well as multivariate analysis were used. We included only patients who were started on anticoagulation on or after day 2 in the ICU. Results: Between 2006 and 2009, a total of 868 patients were started on therapeutic anticoagulation and 139 bled. One hundred five patients who bled were matched to 261 controls. In the matched analysis after adjustment for potential confounders, each bleeding event was associated with an increase in ICU length of stay (hazard ratio for ICU discharge, 0.47; 95% confidence interval, 0.38–0.57; attributable ICU length of stay of 13.8 days). Hospital length of stay was also significantly increased. In the entire cohort analysis, bleeding was also associated with increased ICU length of stay (hazard ratio, 0.59; confidence interval, 0.48–0.72; attributable stay of 6.1 days) and increased hospital length of stay. In both analyses, bleeding was not associated with hospital mortality. Conclusions: Major bleeding while receiving anticoagulation is associated with a substantial increase in ICU and hospital length of stay. Key Words: intensive care unit, bleeding, adverse events, anticoagulation (J Patient Saf 2015;11: 23–27)
The most common adverse events in ICUs are related to medications such as anticoagulants, of which intravenous unfractionated heparin and subcutaneously delivered low-molecular-weight heparin are the most common agents.3 Heparin in therapeutic doses can be lifesaving when used to treat disorders caused by blood clots such as myocardial infarction and pulmonary embolism. However, the major side effect of heparin is bleeding, which can be life threatening. Because of the narrow therapeutic window of intravenous heparin, patients require frequent laboratory monitoring and adjustment of the intensity of anticoagulation. This issue is of special concern in critically ill patients who often have altered pharmacokinetics and pharmacodynamics.4 In patients who have acute coronary syndromes, for every 10-second increase in the partial thromboplastin time (PTT), the risk for bleeding increases by 7%; this finding indicates that some of these events are potentially preventable by more attention to dosing of anticoagulants.5 Of concern is that one-third of hospitalized patients who receive unfractionated heparin have a PTT in the supratherapeutic range during the first 72 hours of therapy.6 No published studies have examined the clinical consequences in terms of attributable length of stay and mortality of these bleeding events. A clearer understanding of the attributable impact would be important for priority setting and cost-effectiveness analyses of quality improvement initiatives. Therefore, the objective of this study was to estimate the attributable length of stay and mortality of bleeding as a complication of therapeutic anticoagulation in the ICU.
METHODS
P
atient harm as a consequence of medical care (adverse events) occurs in 7.5% of hospital admissions and is associated with a 20% risk for death and additional length of stay.1 Patients in the intensive care unit (ICU) have the most complex medical problems and are subjected to numerous medications and invasive medical procedures that may be associated with adverse events. It is not surprising that rates and consequences of adverse events are much greater in ICU than in non-ICU patients.2 For these reasons, the ICU is a key area for the study of patient safety.
From the *Division of Critical Care Medicine and Department of Medicine, Providence Health Care and University of British Columbia; †Center for Health Evaluation and Outcome Sciences, Providence Health Care; ‡Centre for Clinical Epidemiology and Evaluation, Vancouver Coastal Health Research Institute, Vancouver, British Columbia; §Division of Critical Care Medicine, University of Toronto, Toronto, Ontario; kSchool of Population and Public Health, University of British Columbia; and ¶CIHR Canadian HIV Trials Network, Vancouver, British Columbia, Canada. Correspondence: Najib T. Ayas, MD, Room 224 Comox Building, St. Paul's Hospital, Vancouver, British Columbia, Canada V6Z 1Y6 (e‐mail: [email protected]). The authors disclose no conflict of interest. Supported by a grant from the Michael Smith Foundation for Health Research (ICU Patient Safety Team). Dr. Ayas was supported by a Clinician Scientist Award from VCHRI. The funders had no role in the design, analysis, or interpretation of the study. Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
J Patient Saf • Volume 11, Number 1, March 2015
This was a prospective observational study based on data collected in 7 mixed medical/surgical ICUs in British Columbia, Canada, between April 1, 2006, and August 30, 2009. The purpose of this study was to assess rates and impacts of a variety of safety outcomes in the ICU including major bleeding. Three of these ICUs were in tertiary-level academic hospitals (14, 15, and 26 beds), and 4 were in community hospitals (9, 9, 10, and 15 beds). All had a closed intensivist staffing model, and none used computerized order entry for medication orders. Patient charts were screened daily for the occurrence of objectively defined major bleeding while receiving therapeutic anticoagulation. In a patient who was receiving heparin (intravenous or low molecular weight) in therapeutic doses, major bleeding was defined as either (1) clinically important bleeding at any site associated with (i) a decrease in hemoglobin of greater than 20 g/liter, (ii) transfusion of more than 2 U of red blood cells, or (iii) decrease in systolic blood pressure of more than 20 mm Hg and increase in heart rate of more than 20 per minute in the absence of other causes or (2) wound-related bleeding, requiring invasive intervention (e.g., reoperation). This definition was based on that used in a previous study of thromboprophylaxis.7 To exclude patients who were receiving anticoagulation for an unknown period before admission to ICU, we included only patients who were anticoagulated starting on or after day 2 in the ICU. For patients www.journalpatientsafety.com
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
23
J Patient Saf • Volume 11, Number 1, March 2015
Ayas et al
who had multiple ICU admissions during the study period, only the first ICU admission was used in the analyses. Other information recorded included patient characteristics (age, sex, diagnoses, Acute Physiology and Chronic Health Evaluation [APACHE] II score during the first 24 h of ICU stay), and clinical outcomes (ICU and hospital length of stay mortality).
Statistical Analysis We assessed attributable ICU and hospital length of stay and mortality using 2 different study designs: (1) a matched cohort design and (2) an entire cohort design that assessed the effect of the bleeding event as a time-dependent exposure. This approach allowed us to assess the consistency of the results. For the matched cohort analysis, we identified all patients who had major bleeding (the “exposed” patients) while receiving anticoagulation during the period noted above. For each exposed patient, we matched him/her to up to 3 patients who did not experience an event (the “control” patients). The matching variables were hospital site, age (within 5 y), and APACHE score (within 4 points). Control patients were included only if they had received anticoagulants for greater than or equal to the number of days of anticoagulation before the bleed in their matched exposed patient; this condition ensured that controls were at risk for the event for a comparable period as the exposed patient. For both the matched cohort and the entire cohort analyses, we used Cox proportional hazards modeling to ascertain the length of ICU and hospital stay (no censoring for death) as well as the time to death, in which discharge alive was treated as a censored observation. We did not treat death as a censored observation for length of stay because the focus was on the impact of bleeding on service use. For both analyses, we also used an adjustment term to control for ICU admitting or acquired diagnoses that may have been potential confounders: stroke, gastrointestinal bleed, cirrhosis or severe liver disease, myocardial infarction/unstable angina, peripheral vascular occlusion, and deep venous thrombosis/ pulmonary embolism. For the matched cohort analysis, the Cox model treated the matched groups as clusters and sex was included in the model as an adjustment variable. The attributable length of ICU and hospital stays was estimated as the difference in the areas under the
Kaplan-Meier curves for length of ICU and hospital stay in the exposed and control groups. For the entire cohort analysis, we used the first day of anticoagulation as time 0, and a bleeding event was treated as a timedependent exposure, which was set to the value 0 up to the time a bleeding event occurred and to the value 1 thereafter. These models were adjusted for APACHE II score, age, sex, hospital, and days in ICU before starting anticoagulation. The attributable length of ICU and hospital stay due to a bleed was obtained by averaging over all exposed patients the difference in the area under the patient-specific survival curves under the exposed and unexposed (i.e., with the value of the exposure variable set to 0 at all times) conditions. This study was approved by the Research Ethics Board of Providence Health Care.
RESULTS A total of 868 patients were started on therapeutic anticoagulation while in these ICUs. A total of 139 patients had an episode of severe bleeding while they were receiving anticoagulants (Table 1).
Matched Cohort Analysis One hundred five patients who had major bleeds during anticoagulation were matched to 261 controls who were anticoagulated but did not bleed. We excluded 34 exposed patients because no appropriate controls could be identified. Each bleeding event was associated with a significantly increased ICU length of stay after the start of anticoagulation (hazard ratio for ICU discharge, 0.47; 95% confidence interval [CI], 0.38–0.57; P < 0.0001) and an attributable ICU length of stay of 13.8 days (95% CI, 9.2–18.2 d; Fig. 1). Hospital length of stay was also significantly increased (hazard ratio for hospital discharge, 0.66; 95% CI, 0.54–0.81; P < 0.0001) and was associated with an attributable length of stay of 14.5 days (95% CI, 6.9–22.1 d; Fig. 2). Patients who bled had a higher crude hospital mortality compared with those who did not (48% versus 40%). However, after controlling for potential confounders, bleeding was not significantly associated with hospital mortality (hazard ratio for time to death, 0.84; 95% CI, 0.61–1.15; P = 0.27; Fig. 3). Interestingly, bleeding was associated with an increased time to death in the
TABLE 1. Matched Case-Control Analysis
Age, mean (SD) APACHE II score, mean (SD) Sex: female, n (%) ICU mortality rate, n (%) Hospital mortality rate, n (%) Days of anticoagulation, median (interquartile range) ICU days, median (interquartile range) ICU days from starting anticoagulation to bleed, median (interquartile range) ICU days before starting anticoagulation, median (interquartile range) ICU days after starting anticoagulation, median (interquartile range) Hospital days, median (interquartile range) Hospital days before starting anticoagulation, median (interquartile range) Hospital days after starting anticoagulation, median (interquartile range)
Patients Who Bled (n = 105)
Patients Who Did Not Bleed (n = 261)
67 (13) 25 (7) 32 (30) 28 (27) 50 (48) 6 (2–18) 23 (14–42) 3 (1–7) 3 (1–7) 17 (10–36) 53 (28–72) 7 (3–13) 39 (18–61)
66 (15) 25 (7) 90 (34) 61 (23) 105 (40) 4 (1–8) 10 (5–17) NA 2 (1–4) 6 (3–12) 27 (14–53) 4 (2–10) 20 (9–40)
NA, not applicable.
24
www.journalpatientsafety.com
© 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
J Patient Saf • Volume 11, Number 1, March 2015
FIGURE 1. ICU length of stay: matched case-control analysis. Kaplan-Meier graph demonstrating differences in ICU length of stay in the matched case-control analysis. Attributable ICU length of stay was 13.8 days (95% CI, 9.2–18.2).
ICU (hazard ratio, 0.51; 95% CI, 0.31–0.84; P = 0.008). The increase in time to death in the ICU (without an increase in time to hospital death) may be due to the increased ICU length of stay in the patients who bled.
Entire Cohort Analysis The entire cohort analysis included all 868 patients who received anticoagulants. Overall, the major primary admitting diagnoses included sepsis/pneumonia (26%), cardiac/vascular disease (25%), coma/encephalopathy (7%), chronic obstructive pulmonary disease (5%), and pancreatic/biliary disease (5%). Most of
Major Bleeding Due to Therapeutic Anticoagulation
FIGURE 3. Hospital mortality: matched case-control analysis. Kaplan-Meier graph demonstrating the lack of association of bleeding with hospital mortality in the matched case-control analysis.
the patients were receiving intravenous heparin infusions; only 6% of the patients received low-molecular-weight heparin for therapeutic anticoagulation. After adjusting for covariates, major bleeding was associated with a significant increase in length of ICU stay (hazard ratio for ICU discharge, 0.59; 95% CI, 0.48–0.72; P < 0.0001) and an attributable increase in ICU length of stay of 6.1 days (95% CI, 3.3–7.1 d). Hospital length of stay was also significantly increased (hazard ratio for hospital discharge, 0.59; 95% CI, 0.49–0.72; P < 0.0001). Attributable hospital length of stay was 15.4 days (95% CI, 9.9–20.9 d). Consistent with the results of the matched analysis, bleeding was not significantly associated with hospital time to death (hazard ratio for time to death, 0.94; 95% CI, 0.70–1.24; P = 0.63), but bleeding was associated with increased time to death in the ICU (hazard ratio, 0.66; 95% CI, 0.44–0.98; P = 0.04).
DISCUSSION
FIGURE 2. Hospital length of stay: matched case-control analysis. Kaplan-Meier graph demonstrating differences in hospital length of stay in the matched case-control analysis. Attributable hospital length of stay was 14.5 days (95% CI, 6.9–22.1). © 2015 Wolters Kluwer Health, Inc. All rights reserved.
In this study in a broad range of Canadian ICUs, major bleeding occurring during therapeutic anticoagulation was associated with a substantially increased ICU length of stay in both the matched cohort (13.8 d) and entire cohort (6.1 d) analyses. The reason for the difference in ICU attributable length of stay is unclear, but we believe that the entire cohort analysis may be more accurate. The matching process that we used might have resulted in 2 groups that in fact were not comparable, leading to a biased estimate of the attributable length of stay. The time-dependent exposure analysis in the entire cohort would minimize this potential problem by comparing the ICU stay trajectory of a given patient who bled with what his/her trajectory would be if he/she had not bled, rather than comparing the trajectories of patients who bled with other (potentially noncomparable) individuals. Interestingly, the impact of bleeding on hospital length of stay was similar in the matched and the cohort analysis (14.5 and 15.4 d, respectively). The patients who bled had an increased mortality rate compared with those who did not in both the matched and cohort analyses (Tables 1, 2). However, after controlling for potential confounders, bleeding was no longer associated with hospital mortality. The www.journalpatientsafety.com
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
25
J Patient Saf • Volume 11, Number 1, March 2015
Ayas et al
TABLE 2. Cohort Analysis
Age, mean (SD) APACHE II score, mean (SD) Sex: female ICU mortality rate, n (%) Hospital mortality rate, n (%) Days of anticoagulation, median (interquartile range) ICU days, median (interquartile range) ICU days from starting anticoagulation to bleed, median (interquartile range) ICU days before starting anticoagulation, median (interquartile range) ICU days after starting anticoagulation, median (interquartile range) Hospital days, median (interquartile range) Hospital days before starting anticoagulation, median (interquartile range) Hospital days after starting anticoagulation, median (interquartile range)
Patients Who Bled (n = 139)
Patients Who Did Not Bleed (n = 729)
66 (14) 25 (7) 45 (32%) 39 (28) 65 (47) 5 (2–16) 19 (10–38) 3 (0–7) 2 (1–6) 14 (7–31) 42 (21–72) 5 (2–12) 32 (14–61)
65 (15) 24 (7) 261 (36%) 155 (21) 258 (35) 3 (1–6) 8 (4–15) NA 2 (1–4) 5 (2–10) 24 (13–48) 4 (2–10) 17 (7–36)
NA, not applicable.
lack of a significant effect of bleeding on hospital mortality was somewhat surprising, and the reasons are open to speculation. It is possible that patients in the ICU who bleed are closely monitored, and bleeding may be rapidly recognized leading to prompt institution of supportive care (e.g., blood products, fluids) that might have reduced any potential mortality impact of bleeding. Another possibility is that there may be unrecognized confounders that may have reduced the association between bleeding and mortality. Our results build on those published by Cook and colleagues,8 who assessed attributable mortality and ICU length of stay of clinically important gastrointestinal bleeding using an observational design. These investigators used 3 statistical analyses (matched cohort method, model-based matched cohort method, and regression). In their study, although risk for death was increased in some of the analyses, in the fully adjusted regression analysis, bleeding was not significantly associated with an increased risk for mortality. In contrast, bleeding was significantly associated with increased ICU length of stay regardless of which method was used. Our study extends these findings to patients who had major bleeds from any source in the setting of anticoagulation. The most striking finding in this study is the magnitude of excess ICU stay associated with this adverse event. Given that costs in the ICU in the United States range from $3968 to $10,794 per patient day, the economic implications of bleeding are likely substantial.9 This study provides the first step toward understanding the cost implications of these events. In the future, cost models can be developed to assess the cost-effectiveness of interventions designed to prevent these events. For example, a recent study showed that lack of pharmacist participation in anticoagulation therapy was associated with a 49% increased risk for bleeding complications and 39% more patients requiring transfusions in ICU patients treated for thromboembolic disease.10 Broader use of specific weight-based heparin protocols11 and information technology12 may provide opportunities to reduce the risk for bleeding. The magnitude of attributable length of stay of bleeding events during anticoagulation is similar to that of other important safety outcomes in ICUs. For instance, a recent meta-analysis of 8 studies found an attributable ICU length of stay of 14.3 days in patients who experienced a catheter-related blood stream infection.13 Similarly, a 2005 systematic review demonstrated that ventilatorassociated pneumonia is associated with an attributable ICU length of stay of 6.1 days.14
26
www.journalpatientsafety.com
Strengths and Limitations One strength of our study is the broad range of the ICUs represented; this feature increases the generalizability of our study. In addition, we used a prospective study design using a previously validated definition of severe bleeding to minimize risk for bias from a retrospective design, such as imperfect case finding, lack of documentation of bleeding, temporal relations with anticoagulant use, and missing data. Finally, we used 2 different statistical approaches to assess the attributable length of stay and mortality related to bleeding—both approaches led to similar findings. There are several limitations to our study. A primary limitation of this study is the possibility of inadequately controlled confounding variables between those who bled and those who did not. That is, the increased length of stay may not have been due to the bleed per se but instead may have been due to the confounding effect of comorbidities, disease severity, or the indication for anticoagulation. However, we matched and/or adjusted for diagnoses that may be associated with both bleeding and increased morbidity/ mortality, severity of illness scores, duration in ICU before the bleeding event, and patient age. Second, these data were collected solely in Canada, and the applicability of our findings outside a Canadian context could be questioned. However, we believe that the care of patients in Canada is likely similar to that in other highly resourced countries. Third, we included only severe bleeding episodes; our results may not be applicable to patients who have less severe bleeding episodes. Fourth, the proportion of bleeds that were potentially preventable was not known. Specifically, we did not measure adherence to anticoagulation protocols or errors in management of the heparin infusions that may have contributed to bleeding. We also have no information on the intensity of anticoagulation (e.g., PTT values). Recording of times when the PTT was supratherapeutic would have provided interesting data. Fifth, we did not have detailed economic information relating to cost of care, and thus, the true costs of these events are unclear. However, given the impact on length of stay, this would likely be substantial. Major bleeding while receiving therapeutic doses of anticoagulants is associated with a substantially increased ICU and hospital length of stay. Quality improvement measures focused on preventing these events could result in substantial cost savings. Similar methodology can be used to assess the impact of other important safety outcomes. © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
J Patient Saf • Volume 11, Number 1, March 2015
REFERENCES 1. Baker GR, Norton PG, Flintoft V, et al. The Canadian Adverse Events Study: the incidence of adverse events among hospital patients in Canada. CMAJ. 2004;170:1678–1686. 2. Cullen DJ, Sweitzer BJ, Bates DW, et al. Preventable adverse drug events in hospitalized patients: a comparative study of intensive care and general care units. Crit Care Med. 1997;25:1289–1297. 3. Rothschild JM, Landrigan CP, Cronin JW, et al. The Critical Care Safety Study: the incidence and nature of adverse events and serious medical errors in intensive care. Crit Care Med. 2005;33:1694–1700. 4. Barletta JF, Cooper B, Ohlinger MJ. Adverse drug events associated with disorders of coagulation. Crit Care Med. 2010;38:S198–S218. 5. Anand SS, Yusuf S, Pogue J, et al. Relationship of activated partial thromboplastin time to coronary events and bleeding in patients with acute coronary syndromes who receive heparin. Circulation. 2003;107:2884–2888. 6. Valenstein PN, Walsh MK, Meier F. Heparin monitoring and patient safety: a College of American Pathologists Q-Probes study of 3431 patients at 140 institutions. Arch Pathol Lab Med. 2004;128:397–402. 7. Arnold DM, Rabbat C, Lauzier F, et al. Adjudicating bleeding outcomes in a large thromboprophylaxix trial in critical illness. Blood. 2009;114:A2471.
© 2015 Wolters Kluwer Health, Inc. All rights reserved.
Major Bleeding Due to Therapeutic Anticoagulation
8. Cook DJ, Griffith LE, Walter SD, et al. The attributable mortality and length of intensive care unit stay of clinically important gastrointestinal bleeding in critically ill patients. Crit Care. 2001;5:368–375. 9. Dasta JF, McLaughlin TP, Mody SH, et al. Daily cost of an intensive care unit day: the contribution of mechanical ventilation. Crit Care Med. 2005;33:1266–1271. 10. MacLaren R, Bond CA. Effects of pharmacist participation in intensive care units on clinical and economic outcomes of critically ill patients with thromboembolic or infarction-related events. Pharmacotherapy. 2009;29: 761–768. 11. Brown G, Dodek P. An evaluation of empiric vs. nomogram-based dosing of heparin in an intensive care unit. Crit Care Med. 1997;25:1534–1538. 12. Kershaw B, White RH, Mungall D, et al. Computer-assisted dosing of heparin. Management with a pharmacy-based anticoagulation service. Arch Intern Med. 1994;154:1005–1011. 13. Siempos II, Kopterides P, Tsangaris I, et al. Impact of catheter-related bloodstream infections on the mortality of critically ill patients: a meta-analysis. Crit Care Med. 2009;37:2283–2289. 14. Safdar N, Dezfulian C, Collard HR, et al. Clinical and economic consequences of ventilator-associated pneumonia: a systematic review. Crit Care Med. 2005;33:2184–2193.
www.journalpatientsafety.com
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
27
Attributable Length of Stay and Mortality of Major Bleeding as a Complication of Therapeutic Anticoagulation in the Intensive Care Unit Najib T. Ayas, MD,*†‡ Peter M. Dodek, MD,*† Hong Wang,† Robert Fowler, MD,§ Hubert Wong, PhD,†k¶ and Monica Norena, MSc† Objective: The aim of this study was to determine the attributable length of stay and mortality due to bleeding as a complication of therapeutic anticoagulation in intensive care unit (ICU) patients. Methods: Charts of patients from 7 ICUs in British Columbia were screened daily for the occurrence of major bleeding while receiving therapeutic heparin. To determine attributable length of stay and mortality, a matched and unmatched cohort design as well as multivariate analysis were used. We included only patients who were started on anticoagulation on or after day 2 in the ICU. Results: Between 2006 and 2009, a total of 868 patients were started on therapeutic anticoagulation and 139 bled. One hundred five patients who bled were matched to 261 controls. In the matched analysis after adjustment for potential confounders, each bleeding event was associated with an increase in ICU length of stay (hazard ratio for ICU discharge, 0.47; 95% confidence interval, 0.38–0.57; attributable ICU length of stay of 13.8 days). Hospital length of stay was also significantly increased. In the entire cohort analysis, bleeding was also associated with increased ICU length of stay (hazard ratio, 0.59; confidence interval, 0.48–0.72; attributable stay of 6.1 days) and increased hospital length of stay. In both analyses, bleeding was not associated with hospital mortality. Conclusions: Major bleeding while receiving anticoagulation is associated with a substantial increase in ICU and hospital length of stay. Key Words: intensive care unit, bleeding, adverse events, anticoagulation (J Patient Saf 2015;11: 23–27)
The most common adverse events in ICUs are related to medications such as anticoagulants, of which intravenous unfractionated heparin and subcutaneously delivered low-molecular-weight heparin are the most common agents.3 Heparin in therapeutic doses can be lifesaving when used to treat disorders caused by blood clots such as myocardial infarction and pulmonary embolism. However, the major side effect of heparin is bleeding, which can be life threatening. Because of the narrow therapeutic window of intravenous heparin, patients require frequent laboratory monitoring and adjustment of the intensity of anticoagulation. This issue is of special concern in critically ill patients who often have altered pharmacokinetics and pharmacodynamics.4 In patients who have acute coronary syndromes, for every 10-second increase in the partial thromboplastin time (PTT), the risk for bleeding increases by 7%; this finding indicates that some of these events are potentially preventable by more attention to dosing of anticoagulants.5 Of concern is that one-third of hospitalized patients who receive unfractionated heparin have a PTT in the supratherapeutic range during the first 72 hours of therapy.6 No published studies have examined the clinical consequences in terms of attributable length of stay and mortality of these bleeding events. A clearer understanding of the attributable impact would be important for priority setting and cost-effectiveness analyses of quality improvement initiatives. Therefore, the objective of this study was to estimate the attributable length of stay and mortality of bleeding as a complication of therapeutic anticoagulation in the ICU.
METHODS
P
atient harm as a consequence of medical care (adverse events) occurs in 7.5% of hospital admissions and is associated with a 20% risk for death and additional length of stay.1 Patients in the intensive care unit (ICU) have the most complex medical problems and are subjected to numerous medications and invasive medical procedures that may be associated with adverse events. It is not surprising that rates and consequences of adverse events are much greater in ICU than in non-ICU patients.2 For these reasons, the ICU is a key area for the study of patient safety.
From the *Division of Critical Care Medicine and Department of Medicine, Providence Health Care and University of British Columbia; †Center for Health Evaluation and Outcome Sciences, Providence Health Care; ‡Centre for Clinical Epidemiology and Evaluation, Vancouver Coastal Health Research Institute, Vancouver, British Columbia; §Division of Critical Care Medicine, University of Toronto, Toronto, Ontario; kSchool of Population and Public Health, University of British Columbia; and ¶CIHR Canadian HIV Trials Network, Vancouver, British Columbia, Canada. Correspondence: Najib T. Ayas, MD, Room 224 Comox Building, St. Paul's Hospital, Vancouver, British Columbia, Canada V6Z 1Y6 (e‐mail: [email protected]). The authors disclose no conflict of interest. Supported by a grant from the Michael Smith Foundation for Health Research (ICU Patient Safety Team). Dr. Ayas was supported by a Clinician Scientist Award from VCHRI. The funders had no role in the design, analysis, or interpretation of the study. Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
J Patient Saf • Volume 11, Number 1, March 2015
This was a prospective observational study based on data collected in 7 mixed medical/surgical ICUs in British Columbia, Canada, between April 1, 2006, and August 30, 2009. The purpose of this study was to assess rates and impacts of a variety of safety outcomes in the ICU including major bleeding. Three of these ICUs were in tertiary-level academic hospitals (14, 15, and 26 beds), and 4 were in community hospitals (9, 9, 10, and 15 beds). All had a closed intensivist staffing model, and none used computerized order entry for medication orders. Patient charts were screened daily for the occurrence of objectively defined major bleeding while receiving therapeutic anticoagulation. In a patient who was receiving heparin (intravenous or low molecular weight) in therapeutic doses, major bleeding was defined as either (1) clinically important bleeding at any site associated with (i) a decrease in hemoglobin of greater than 20 g/liter, (ii) transfusion of more than 2 U of red blood cells, or (iii) decrease in systolic blood pressure of more than 20 mm Hg and increase in heart rate of more than 20 per minute in the absence of other causes or (2) wound-related bleeding, requiring invasive intervention (e.g., reoperation). This definition was based on that used in a previous study of thromboprophylaxis.7 To exclude patients who were receiving anticoagulation for an unknown period before admission to ICU, we included only patients who were anticoagulated starting on or after day 2 in the ICU. For patients www.journalpatientsafety.com
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
23
J Patient Saf • Volume 11, Number 1, March 2015
Ayas et al
who had multiple ICU admissions during the study period, only the first ICU admission was used in the analyses. Other information recorded included patient characteristics (age, sex, diagnoses, Acute Physiology and Chronic Health Evaluation [APACHE] II score during the first 24 h of ICU stay), and clinical outcomes (ICU and hospital length of stay mortality).
Statistical Analysis We assessed attributable ICU and hospital length of stay and mortality using 2 different study designs: (1) a matched cohort design and (2) an entire cohort design that assessed the effect of the bleeding event as a time-dependent exposure. This approach allowed us to assess the consistency of the results. For the matched cohort analysis, we identified all patients who had major bleeding (the “exposed” patients) while receiving anticoagulation during the period noted above. For each exposed patient, we matched him/her to up to 3 patients who did not experience an event (the “control” patients). The matching variables were hospital site, age (within 5 y), and APACHE score (within 4 points). Control patients were included only if they had received anticoagulants for greater than or equal to the number of days of anticoagulation before the bleed in their matched exposed patient; this condition ensured that controls were at risk for the event for a comparable period as the exposed patient. For both the matched cohort and the entire cohort analyses, we used Cox proportional hazards modeling to ascertain the length of ICU and hospital stay (no censoring for death) as well as the time to death, in which discharge alive was treated as a censored observation. We did not treat death as a censored observation for length of stay because the focus was on the impact of bleeding on service use. For both analyses, we also used an adjustment term to control for ICU admitting or acquired diagnoses that may have been potential confounders: stroke, gastrointestinal bleed, cirrhosis or severe liver disease, myocardial infarction/unstable angina, peripheral vascular occlusion, and deep venous thrombosis/ pulmonary embolism. For the matched cohort analysis, the Cox model treated the matched groups as clusters and sex was included in the model as an adjustment variable. The attributable length of ICU and hospital stays was estimated as the difference in the areas under the
Kaplan-Meier curves for length of ICU and hospital stay in the exposed and control groups. For the entire cohort analysis, we used the first day of anticoagulation as time 0, and a bleeding event was treated as a timedependent exposure, which was set to the value 0 up to the time a bleeding event occurred and to the value 1 thereafter. These models were adjusted for APACHE II score, age, sex, hospital, and days in ICU before starting anticoagulation. The attributable length of ICU and hospital stay due to a bleed was obtained by averaging over all exposed patients the difference in the area under the patient-specific survival curves under the exposed and unexposed (i.e., with the value of the exposure variable set to 0 at all times) conditions. This study was approved by the Research Ethics Board of Providence Health Care.
RESULTS A total of 868 patients were started on therapeutic anticoagulation while in these ICUs. A total of 139 patients had an episode of severe bleeding while they were receiving anticoagulants (Table 1).
Matched Cohort Analysis One hundred five patients who had major bleeds during anticoagulation were matched to 261 controls who were anticoagulated but did not bleed. We excluded 34 exposed patients because no appropriate controls could be identified. Each bleeding event was associated with a significantly increased ICU length of stay after the start of anticoagulation (hazard ratio for ICU discharge, 0.47; 95% confidence interval [CI], 0.38–0.57; P < 0.0001) and an attributable ICU length of stay of 13.8 days (95% CI, 9.2–18.2 d; Fig. 1). Hospital length of stay was also significantly increased (hazard ratio for hospital discharge, 0.66; 95% CI, 0.54–0.81; P < 0.0001) and was associated with an attributable length of stay of 14.5 days (95% CI, 6.9–22.1 d; Fig. 2). Patients who bled had a higher crude hospital mortality compared with those who did not (48% versus 40%). However, after controlling for potential confounders, bleeding was not significantly associated with hospital mortality (hazard ratio for time to death, 0.84; 95% CI, 0.61–1.15; P = 0.27; Fig. 3). Interestingly, bleeding was associated with an increased time to death in the
TABLE 1. Matched Case-Control Analysis
Age, mean (SD) APACHE II score, mean (SD) Sex: female, n (%) ICU mortality rate, n (%) Hospital mortality rate, n (%) Days of anticoagulation, median (interquartile range) ICU days, median (interquartile range) ICU days from starting anticoagulation to bleed, median (interquartile range) ICU days before starting anticoagulation, median (interquartile range) ICU days after starting anticoagulation, median (interquartile range) Hospital days, median (interquartile range) Hospital days before starting anticoagulation, median (interquartile range) Hospital days after starting anticoagulation, median (interquartile range)
Patients Who Bled (n = 105)
Patients Who Did Not Bleed (n = 261)
67 (13) 25 (7) 32 (30) 28 (27) 50 (48) 6 (2–18) 23 (14–42) 3 (1–7) 3 (1–7) 17 (10–36) 53 (28–72) 7 (3–13) 39 (18–61)
66 (15) 25 (7) 90 (34) 61 (23) 105 (40) 4 (1–8) 10 (5–17) NA 2 (1–4) 6 (3–12) 27 (14–53) 4 (2–10) 20 (9–40)
NA, not applicable.
24
www.journalpatientsafety.com
© 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
J Patient Saf • Volume 11, Number 1, March 2015
FIGURE 1. ICU length of stay: matched case-control analysis. Kaplan-Meier graph demonstrating differences in ICU length of stay in the matched case-control analysis. Attributable ICU length of stay was 13.8 days (95% CI, 9.2–18.2).
ICU (hazard ratio, 0.51; 95% CI, 0.31–0.84; P = 0.008). The increase in time to death in the ICU (without an increase in time to hospital death) may be due to the increased ICU length of stay in the patients who bled.
Entire Cohort Analysis The entire cohort analysis included all 868 patients who received anticoagulants. Overall, the major primary admitting diagnoses included sepsis/pneumonia (26%), cardiac/vascular disease (25%), coma/encephalopathy (7%), chronic obstructive pulmonary disease (5%), and pancreatic/biliary disease (5%). Most of
Major Bleeding Due to Therapeutic Anticoagulation
FIGURE 3. Hospital mortality: matched case-control analysis. Kaplan-Meier graph demonstrating the lack of association of bleeding with hospital mortality in the matched case-control analysis.
the patients were receiving intravenous heparin infusions; only 6% of the patients received low-molecular-weight heparin for therapeutic anticoagulation. After adjusting for covariates, major bleeding was associated with a significant increase in length of ICU stay (hazard ratio for ICU discharge, 0.59; 95% CI, 0.48–0.72; P < 0.0001) and an attributable increase in ICU length of stay of 6.1 days (95% CI, 3.3–7.1 d). Hospital length of stay was also significantly increased (hazard ratio for hospital discharge, 0.59; 95% CI, 0.49–0.72; P < 0.0001). Attributable hospital length of stay was 15.4 days (95% CI, 9.9–20.9 d). Consistent with the results of the matched analysis, bleeding was not significantly associated with hospital time to death (hazard ratio for time to death, 0.94; 95% CI, 0.70–1.24; P = 0.63), but bleeding was associated with increased time to death in the ICU (hazard ratio, 0.66; 95% CI, 0.44–0.98; P = 0.04).
DISCUSSION
FIGURE 2. Hospital length of stay: matched case-control analysis. Kaplan-Meier graph demonstrating differences in hospital length of stay in the matched case-control analysis. Attributable hospital length of stay was 14.5 days (95% CI, 6.9–22.1). © 2015 Wolters Kluwer Health, Inc. All rights reserved.
In this study in a broad range of Canadian ICUs, major bleeding occurring during therapeutic anticoagulation was associated with a substantially increased ICU length of stay in both the matched cohort (13.8 d) and entire cohort (6.1 d) analyses. The reason for the difference in ICU attributable length of stay is unclear, but we believe that the entire cohort analysis may be more accurate. The matching process that we used might have resulted in 2 groups that in fact were not comparable, leading to a biased estimate of the attributable length of stay. The time-dependent exposure analysis in the entire cohort would minimize this potential problem by comparing the ICU stay trajectory of a given patient who bled with what his/her trajectory would be if he/she had not bled, rather than comparing the trajectories of patients who bled with other (potentially noncomparable) individuals. Interestingly, the impact of bleeding on hospital length of stay was similar in the matched and the cohort analysis (14.5 and 15.4 d, respectively). The patients who bled had an increased mortality rate compared with those who did not in both the matched and cohort analyses (Tables 1, 2). However, after controlling for potential confounders, bleeding was no longer associated with hospital mortality. The www.journalpatientsafety.com
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
25
J Patient Saf • Volume 11, Number 1, March 2015
Ayas et al
TABLE 2. Cohort Analysis
Age, mean (SD) APACHE II score, mean (SD) Sex: female ICU mortality rate, n (%) Hospital mortality rate, n (%) Days of anticoagulation, median (interquartile range) ICU days, median (interquartile range) ICU days from starting anticoagulation to bleed, median (interquartile range) ICU days before starting anticoagulation, median (interquartile range) ICU days after starting anticoagulation, median (interquartile range) Hospital days, median (interquartile range) Hospital days before starting anticoagulation, median (interquartile range) Hospital days after starting anticoagulation, median (interquartile range)
Patients Who Bled (n = 139)
Patients Who Did Not Bleed (n = 729)
66 (14) 25 (7) 45 (32%) 39 (28) 65 (47) 5 (2–16) 19 (10–38) 3 (0–7) 2 (1–6) 14 (7–31) 42 (21–72) 5 (2–12) 32 (14–61)
65 (15) 24 (7) 261 (36%) 155 (21) 258 (35) 3 (1–6) 8 (4–15) NA 2 (1–4) 5 (2–10) 24 (13–48) 4 (2–10) 17 (7–36)
NA, not applicable.
lack of a significant effect of bleeding on hospital mortality was somewhat surprising, and the reasons are open to speculation. It is possible that patients in the ICU who bleed are closely monitored, and bleeding may be rapidly recognized leading to prompt institution of supportive care (e.g., blood products, fluids) that might have reduced any potential mortality impact of bleeding. Another possibility is that there may be unrecognized confounders that may have reduced the association between bleeding and mortality. Our results build on those published by Cook and colleagues,8 who assessed attributable mortality and ICU length of stay of clinically important gastrointestinal bleeding using an observational design. These investigators used 3 statistical analyses (matched cohort method, model-based matched cohort method, and regression). In their study, although risk for death was increased in some of the analyses, in the fully adjusted regression analysis, bleeding was not significantly associated with an increased risk for mortality. In contrast, bleeding was significantly associated with increased ICU length of stay regardless of which method was used. Our study extends these findings to patients who had major bleeds from any source in the setting of anticoagulation. The most striking finding in this study is the magnitude of excess ICU stay associated with this adverse event. Given that costs in the ICU in the United States range from $3968 to $10,794 per patient day, the economic implications of bleeding are likely substantial.9 This study provides the first step toward understanding the cost implications of these events. In the future, cost models can be developed to assess the cost-effectiveness of interventions designed to prevent these events. For example, a recent study showed that lack of pharmacist participation in anticoagulation therapy was associated with a 49% increased risk for bleeding complications and 39% more patients requiring transfusions in ICU patients treated for thromboembolic disease.10 Broader use of specific weight-based heparin protocols11 and information technology12 may provide opportunities to reduce the risk for bleeding. The magnitude of attributable length of stay of bleeding events during anticoagulation is similar to that of other important safety outcomes in ICUs. For instance, a recent meta-analysis of 8 studies found an attributable ICU length of stay of 14.3 days in patients who experienced a catheter-related blood stream infection.13 Similarly, a 2005 systematic review demonstrated that ventilatorassociated pneumonia is associated with an attributable ICU length of stay of 6.1 days.14
26
www.journalpatientsafety.com
Strengths and Limitations One strength of our study is the broad range of the ICUs represented; this feature increases the generalizability of our study. In addition, we used a prospective study design using a previously validated definition of severe bleeding to minimize risk for bias from a retrospective design, such as imperfect case finding, lack of documentation of bleeding, temporal relations with anticoagulant use, and missing data. Finally, we used 2 different statistical approaches to assess the attributable length of stay and mortality related to bleeding—both approaches led to similar findings. There are several limitations to our study. A primary limitation of this study is the possibility of inadequately controlled confounding variables between those who bled and those who did not. That is, the increased length of stay may not have been due to the bleed per se but instead may have been due to the confounding effect of comorbidities, disease severity, or the indication for anticoagulation. However, we matched and/or adjusted for diagnoses that may be associated with both bleeding and increased morbidity/ mortality, severity of illness scores, duration in ICU before the bleeding event, and patient age. Second, these data were collected solely in Canada, and the applicability of our findings outside a Canadian context could be questioned. However, we believe that the care of patients in Canada is likely similar to that in other highly resourced countries. Third, we included only severe bleeding episodes; our results may not be applicable to patients who have less severe bleeding episodes. Fourth, the proportion of bleeds that were potentially preventable was not known. Specifically, we did not measure adherence to anticoagulation protocols or errors in management of the heparin infusions that may have contributed to bleeding. We also have no information on the intensity of anticoagulation (e.g., PTT values). Recording of times when the PTT was supratherapeutic would have provided interesting data. Fifth, we did not have detailed economic information relating to cost of care, and thus, the true costs of these events are unclear. However, given the impact on length of stay, this would likely be substantial. Major bleeding while receiving therapeutic doses of anticoagulants is associated with a substantially increased ICU and hospital length of stay. Quality improvement measures focused on preventing these events could result in substantial cost savings. Similar methodology can be used to assess the impact of other important safety outcomes. © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
J Patient Saf • Volume 11, Number 1, March 2015
REFERENCES 1. Baker GR, Norton PG, Flintoft V, et al. The Canadian Adverse Events Study: the incidence of adverse events among hospital patients in Canada. CMAJ. 2004;170:1678–1686. 2. Cullen DJ, Sweitzer BJ, Bates DW, et al. Preventable adverse drug events in hospitalized patients: a comparative study of intensive care and general care units. Crit Care Med. 1997;25:1289–1297. 3. Rothschild JM, Landrigan CP, Cronin JW, et al. The Critical Care Safety Study: the incidence and nature of adverse events and serious medical errors in intensive care. Crit Care Med. 2005;33:1694–1700. 4. Barletta JF, Cooper B, Ohlinger MJ. Adverse drug events associated with disorders of coagulation. Crit Care Med. 2010;38:S198–S218. 5. Anand SS, Yusuf S, Pogue J, et al. Relationship of activated partial thromboplastin time to coronary events and bleeding in patients with acute coronary syndromes who receive heparin. Circulation. 2003;107:2884–2888. 6. Valenstein PN, Walsh MK, Meier F. Heparin monitoring and patient safety: a College of American Pathologists Q-Probes study of 3431 patients at 140 institutions. Arch Pathol Lab Med. 2004;128:397–402. 7. Arnold DM, Rabbat C, Lauzier F, et al. Adjudicating bleeding outcomes in a large thromboprophylaxix trial in critical illness. Blood. 2009;114:A2471.
© 2015 Wolters Kluwer Health, Inc. All rights reserved.
Major Bleeding Due to Therapeutic Anticoagulation
8. Cook DJ, Griffith LE, Walter SD, et al. The attributable mortality and length of intensive care unit stay of clinically important gastrointestinal bleeding in critically ill patients. Crit Care. 2001;5:368–375. 9. Dasta JF, McLaughlin TP, Mody SH, et al. Daily cost of an intensive care unit day: the contribution of mechanical ventilation. Crit Care Med. 2005;33:1266–1271. 10. MacLaren R, Bond CA. Effects of pharmacist participation in intensive care units on clinical and economic outcomes of critically ill patients with thromboembolic or infarction-related events. Pharmacotherapy. 2009;29: 761–768. 11. Brown G, Dodek P. An evaluation of empiric vs. nomogram-based dosing of heparin in an intensive care unit. Crit Care Med. 1997;25:1534–1538. 12. Kershaw B, White RH, Mungall D, et al. Computer-assisted dosing of heparin. Management with a pharmacy-based anticoagulation service. Arch Intern Med. 1994;154:1005–1011. 13. Siempos II, Kopterides P, Tsangaris I, et al. Impact of catheter-related bloodstream infections on the mortality of critically ill patients: a meta-analysis. Crit Care Med. 2009;37:2283–2289. 14. Safdar N, Dezfulian C, Collard HR, et al. Clinical and economic consequences of ventilator-associated pneumonia: a systematic review. Crit Care Med. 2005;33:2184–2193.
www.journalpatientsafety.com
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
27
