ORIGINAL ARTICLE
Effects of anticoagulant and antiplatelet drugs on the risk for hospital admission for traumatic injuries: A case-control and population-based study Stefano Di Bartolomeo, MD, PhD, Massimiliano Marino, PhD, Francesca Valent, MD, MSPH, and Rossana De Palma, MD, Bologna, Italy
The current cardiovascular literature advocates an overall beneficial balance between the advantages of oral anticoagulants and antiplatelet drugs in preventing and treating thromboembolic events and their disadvantages in promoting hemorrhage. However, traumatic injuries have usually received little attention despite several studies from the surgical literature showing worse outcomes in anticoagulated trauma registry patients. To quantify at population level too this seemingly deleterious impact, we investigated the effects of anticoagulants and antiplatelet use on the risk for hospital admission for acute traumatic causes. METHODS: A population-based, case-control study in an Italian region with 4.5 million inhabitants was conducted. Cases were all the 59,348 adult residents admitted to the hospital for traumatic injuries in the years 2010 and 2011. Controls were age- and sexmatched residents selected by incidence density sampling. By conditional logistic regression adjusted for comorbidities, we estimated the risk for traumatic hospital admission while on anticoagulant, antiplatelet, and combined medications. RESULTS: The odds ratios (ORs) for anticoagulation and combined medications were 1.21 (95% confidence interval [CI], 1.15Y1.28) and 1.39 (95% CI, 1.21Y1.62). These effects were generally consistent across subgroups of demographic and clinical characteristics and particularly important in the head injured (e.g., OR for anticoagulation, 2.00; 95% CI, 1.77Y12.27). Antiplatelets alone had no overall effect (OR, 1.02; 95% CI, 0.99Y1.05). The number-needed-to-harm of anticoagulation was 595. CONCLUSION: Oral anticoagulation increased the population risk for traumatic hospital admission, with a further increase in case of concurrent antiplatelet use. Because this effect is most likely to derive from the prohemorrhagic properties of these drugs, injured patients should be included in the future evaluations of the cost-benefit profiles of these medications. (J Trauma Acute Care Surg. 2014;76: 437Y442. Copyright * 2014 by Lippincott Williams & Wilkins) LEVEL OF EVIDENCE: Epidemiologic/prognostic study, level III. KEY WORDS: Trauma; thromboembolism/prevention and control; vitamin K/antagonists and inhibitors; risk assessment; hemorrhage/ etiology. BACKGROUND:
A
nticoagulant and antiplatelet drugs are increasingly used for the prevention and treatment of thromboembolic events, but they augment the risk for bleeding. In the current cardiovascular literature, the balance of risk and benefit is generally considered favorable,1Y3 and the common concerns regarding bleeding have been described as ‘‘overemphasized.’’4 Because bleeding is the leading cause of preventable death after trauma,5 the effects of anticoagulant or antiplatelet therapies in major trauma populations have been the focus of recent studies.6Y8 They concluded that (1) the concurrent use of these drugs among the severely injured has considerably Submitted: July 16, 2013, Revised: August 22, 2013, Accepted: August 22, 2013. Published online: January 6, 2014. From the Regional Agency for Health and Social Care of Emilia-Romagna (M.M., R.D.P.)/Azienda Ospedaliero (S.D.B.), Universitaria di Udine, Bologna; and Servizio di epidemiologia (F.V.), Direzione centrale salute, integrazione sociosanitaria e politiche sociali, Regione autonoma Friuli Venezia Giulia, Italy. Supplemental digital content is available for this article. Direct URL citations appear in the printed text, and links to the digital files are provided in the HTML text of this article on the journal’s Web site (www.jtrauma.com). Address for reprints: Stefano Di Bartolomeo, MD, PhD, Regional Agency for Health and Social Care of Emilia-Romagna/Azienda Ospedaliero-Universitaria di Udine, Viale Aldo Moro 21, 40127 Bologna, Italy; email: stefano.dibartolomeo@ uniud.it. DOI: 10.1097/TA.0b013e3182aa80f9
increased in the last years and (2) these drugs have a negative effect on survival or other outcomes. At the same time, the cardiovascular literature tends to minimize the importance of traumatic injuries when evaluating the adverse effects of anticoagulant and antiplatelet drugs or their net clinical benefit in the general population. For example, several important studies disregarded injured patients altogether (e.g., Friberg et al.,1 Singer et al.,2 Fang et al.9); others considered only head injury patients,3 while others remained ambiguous about their inclusion.10 Even when studies focused on patients with traumatic injuries, they were restricted to particular subgroups (i.e., hospitalized elderly at risk for falls11). A possible explanation for such discrepancy is a reduced consideration in the medical literature for traumatic injuries as compared with more common conditions such as thromboembolic diseases and spontaneous hemorrhages. However, traumatic injuries are known to be increasing in the elderlyVthe age group where anticoagulants and antiplatelets are mostly prescribed and that is rapidly expanding owing to population ageing.12 Therefore, the population risk for adverse traumatic events caused by these medications may be worth quantification. This study investigates the effect of anticoagulant and antiplatelet use on the risk for hospital admission for traumatic injuries in the general population of the Italian region Emilia-Romagna.
J Trauma Acute Care Surg Volume 76, Number 2
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
437
J Trauma Acute Care Surg Volume 76, Number 2
Di Bartolomeo et al.
PATIENTS AND METHODS Setting, Study Design, and Data Sources Emilia-Romagna is an Italian region with about 4.5 million inhabitants, a surface of 22 thousand square km, and approximately 80 hospitals admitting acute patients. It is one of the largest administrative divisions of the country, and its varied composition of urban and rural areas, industry, agriculture, and tourism make it well representative of the national territory. We conducted a population-based case-control study by linking data at an individual level from three regionwide longitudinal registries. The Hospital Discharge Registry contains the usual so-called administrative information recorded in the charts of all the regional hospitals. The Prescription Drug Registry collects information on dates, dosage, and quantities for every prescription dispensed within the regional boundaries, from both hospital and community pharmacies. The Demographic Registry holds information on age, sex, date of birth, residence, and vital status of all residents in Emilia-Romagna. A unique patient identifier allowed cross-linking between the databases. According to our institutional rules, neither patient consent nor ethical committee approval was necessary, given the observational, retrospective design of the study and the anonymity of the databases provided to the researchers.
Study Participants We selected patients admitted to hospital for an injury as cases. Controls were citizens similar to the cases but who had not been admitted to hospital with an injury. Through the Hospital Discharge Registry, we identified all admissions to the hospital for a traumatic cause among adult (917 years) residents in 2010 and 2011. The DRG International Classification of DiseasesV9th Rev.VClinical Modification (ICD-9-CM) codes used to define a traumatic cause were (principal position) from 800.x to 904.x, from 910.x to 929.x, from 940.x to 957.x, and 959.x (i.e., we excluded nontrauma such as poisoning, drowning, suffocation, and nonacute diagnoses). The controls were identified from the Demographic Registry. For each case, we identified five controls, matched by year of birth and sex. We used incidence density sampling to identify controls because the matched-pairs odds ratio estimator consistently estimates the incidence density ratio when the design is of the ‘‘incidence density’’ type and the matching is random except for sampling time.13 To facilitate the matching, we used a modified automatic computer program.14
Variables and Exposure Assessment We assessed exposure to the therapies under investigation through the Prescription Drug Registry. Cases and controls were considered exposed if they had collected at least one prescription of the drugs described later from a pharmacy within 90 days before the date of the hospital admission. The choice of interval was based on an estimation of the days of supply of each prescription through the analysis of consecutive prescriptions. We also conducted sensitivity analyses, with different intervals (60 and 120 days). We investigated two groups of drugs, based on the Anatomical Therapeutic Chemical Classification System: B01AA 438
(oral anticoagulants, i.e., warfarin or acenocoumarol) and B01AC (antiplatelet agents). Because of the possible interaction between drugs of the two groups, we considered three exposure categories as follows: exposure to drugs of the first group only, exposure to drugs of the second group only, and contemporary exposure to drugs from both groups. To measure the comorbidity status, we used the Elixhauser comorbidity measure.15 This measure was developed using administrative data from a statewide California inpatient database from all nonfederal inpatient community hospital stays in California (n = 1,779,167). The Elixhauser comorbidity measure developed a list of 30 comorbidities relying on the ICD-9-CM coding manual. To identify the codes constituted in the score, we screened the hospital records of the study participants up to 5 years before the index trauma admission of the cases. To assess the severity of the injuries of the cases, we used TMPM-ICD9, a recently described anatomic injury severity score.16 This estimates empiric injury severities for each of the injuries in the ICD-9-CM lexicon based on a regression model using data from 749,374 patients admitted to 359 hospitals in the US National Trauma Data Bank. These estimates are then combined into a probability of death, ranging from 0 to 1. We defined head injury as an ICD-9-CM code 800.xx to 801.xx, 803.xx, or 850.xx to 854.xx.
Statistical Analysis We estimated the relative risk for being admitted to hospital for an injury while on anticoagulant/antiplatelet therapy by using conditional logistic regression. The relative risks were adjusted for the comorbidity status. All 30 comorbidities of the Elixhauser score were included in the model as binary variables. We conducted subgroup analyses across categories of demographic factors and of characteristics of the injured patients. As a confirmatory analysis, we compared the injury severity among the exposed (any type of antithrombotic therapy) and nonexposed cases with the Wilcoxon rank-sum test. Presenting results as relative risks is standard in observational studies but may be difficult to translate into clinical practice. The number-needed-to-harm will better reflect any adverse effect of treatment than relative risks in clinical terms.17,18 For this reason, we finally calculated the numberneeded-to-treat-for-harm (NNTH) using methods proposed by Bjerre and LeLorier19 for case-control study designs. For simplicity, we calculated it for oral anticoagulation only. Calculation of such a measure requires estimation of the incidence rate of outcome among the unexposed population. Because we were unable to directly estimate this rate from our sample, we estimated an approximate rate from the study base of residents adult citizens from which our sample arose, as done by Quilliam et al.20 This approximate background rate of traumatic hospital admissions in the general population was 0.008 per person per year.
RESULTS We identified 59,348 adults who were admitted to the hospital for traumatic injuries during the study period and 296,740 controls. The characteristics of cases and controls are shown in Table 1. Age and sex were similar in cases and controls by design. The majority of patients were older than * 2014 Lippincott Williams & Wilkins
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
J Trauma Acute Care Surg Volume 76, Number 2
Di Bartolomeo et al.
TABLE 1. Characteristics of Cases and Controls Cases All, No. Age, n (%), y 18Y44 45Y54 55Y64 65Y74 Q75 Sex, male, n (%) Comorbidity, n (%) Congestive heart failure Cardiac arrhythmias Valvular disease Pulmonary circulation disorders Peripheral vascular disorders Hypertension Paralysis Neurodegenerative disorders Chronic pulmonary disease Diabetes, uncomplicated Diabetes, complicated Hypothyroidism Renal failure Liver disease Peptic ulcer disease, no bleeding Acquired immunodeficiency syndrome/human immunodeficiency virus Lymphoma Metastatic cancer Solid tumor without metastasis Rheumatoid arthritis/collagen vascular diseases Coagulopathy Obesity Weight loss Fluid and electrolyte disorders Blood loss anemia Deficiency anemia Alcohol abuse Drug abuse Psychosis Depression Type of antithrombotic drugs, n (%) (a) None (b) Oral anticoagulants only (c) Antiplatelet agents only (d) Association of b + c
59,348 14,118 (23.79) 6,213 (10.47) 6,460 (10.88) 8,005 (13.49) 24,552 (41.37) 28,944 (48.77)
Controls 296,740 70,590 (23.79) 31,065 (10.47) 32,300 (10.88) 40,030 (13.49) 122,755 (41.37) 144,720 (48.77)
3,754 (6.33)* 5,464 (9.21)* 1,685 (2.84)* 696 (1.17)* 1,372 (2.31)* 12,558 (21.16)* 464 (0.78)* 1,834 (3.09)* 3,583 (6.04)* 3,977 (6.70)* 1,020 (1.72)* 1,265 (2.13)* 2,181 (3.67)* 1,377 (2.32)* 86 (0.14) 60 (0.10)*
12,291 (4.14)* 16,072 (5.42)* 5,674 (1.91)* 2,033 (0.69)* 5,044 (1.70)* 36,891 (12.43)* 972 (0.33)* 3,328 (1.12)* 10,347 (3.49)* 10,034 (3.38)* 2,460 (0.83)* 3,327 (1.12)* 5,819 (1.96)* 3,064 (1.03)* 322 (0.11) 80 (0.03)*
221 (0.37)* 570 (0.96)* 2,100 (3.54) 731 (1.23)*
670 (0.23)* 2,270 (0.76)* 10,198 (3.44) 1,769 (0.60)*
380 (0.64)* 636 (1.07)* 312 (0.53)* 923 (1.56)* 677 (1.14)* 825 (1.39)* 681 (1.15)* 157 (0.26)* 408 (0.69)* 1,785 (3.01)*
683 (0.23)* 2,222 (0.75)* 779 (0.26)* 2,967 (1.00)* 1,946 (0.66)* 2,429 (0.82)* 639 (0.22)* 256 (0.09)* 842 (0.28)* 3,950 (1.33)*
44,392 (74.80)* 2,684 (4.52)* 11,993 (20.21)* 279 (0.47)*
231,255 (77.93)* 9,095 (3.06)* 55,688 (18.77)* 702 (0.24)*
*p G 0.01 (W2 test).
65 years, confirming the increasing importance of traumatic injuries in the elderly. Among cases, comorbid conditions were, in general, more represented. Cases were also more often on antithrombotic therapies, for example, 25.2% as compared with 22.1% of controls.
Table 2 displays the characteristics of cases related to their traumatic injuries. More than 90% of the cases were admitted for 2 days or more, 12% of the patients experienced head injuries, and the overall mortality was 2.58%. The crude and adjusted relative risks of the investigated therapies are shown in Figure 1. Oral anticoagulation significantly increased the risk for hospital admission for trauma both in the crude and adjusted analyses (21% increase in adjusted risk). Antiplatelet therapy increased the risk only in the crude analysis, while its effect was not significant after adjustment. The combined therapy had a larger effect than oral anticoagulation alone (39% increase of adjusted risk), although with a larger confidence interval owing to the smaller number of cases (Fig. 1). Figure 2 shows the results of subgroup analyses conducted across patient characteristics. Antiplatelet agents seemed to have a significant effect in the youngest group (18Y44 years). Oral anticoagulation and combined therapy showed no special patterns across the various categories of injury severity, except in the group with missing injury severity, where their effect was stronger. In patients with very short admissions (length of hospital stay [LOS], 0Y1 day) the effect of oral anticoagulation alone was not significant, while the effect of its combination with antiplatelets was especially strong. In those with head injuries, the effect was large for all kinds of therapies; however, oral anticoagulation and combined therapy had a smaller but significant effect in patients without head injuries too. In patients admitted to the intensive care unit, the effect of oral anticoagulation and combined therapies was smaller and nonsignificant, while the effect of antiplatelets was significant.
TABLE 2. Selected Characteristics of Cases Characteristic Injury Severity,* quintiles 0.000043Y0.004514 0.004518Y0.008667 0.008672Y0.014115 0.014118Y0.019641 0.019647Y0.956034 Missing** Length of Hospital Stay, d 0Y1 2Y3 Q3 Type of injury Head injury No head injury Intensive care unit admission Yes No Mortality (30 d)
n (%) 12,907 (22.46) 11,205 (19.50) 10,912 (18.99) 11,193 (19.48) 11,248 (19.57) 1,883 (3.17) 7,321 (12.34) 12,180 (20.52) 39,847 (67.14) 6,932 (11.68) 52,416 (88.32) 2,758 (4.65) 56, 590 (95.35) 1,534 (2.58)
*TMPM-ICD9 injury severity measure, ranging from 0 to 1. See text for details. **Missing because of the fact that the TMPM-ICD9 scoring system provides for some injuries only an aspecific and adimensional severity measure (i.e., 9).
* 2014 Lippincott Williams & Wilkins
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
439
J Trauma Acute Care Surg Volume 76, Number 2
Di Bartolomeo et al.
Figure 1. Antithrombotic drugs: risk estimates of being admitted to the hospital for trauma.
Figure 2. Antithrombotic drugs: risk estimates of being admitted to the hospital for trauma. Subgroup analyses. 440
* 2014 Lippincott Williams & Wilkins
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
J Trauma Acute Care Surg Volume 76, Number 2
Di Bartolomeo et al.
The detailed effects of the 30 comorbidities of the Elixhauser classification are shown in Table 1 of the Supplemental Digital Content (http://links.lww.com/TA/A348). Figures S1 and S2 of the Supplemental Digital Content (http://links.lww.com/TA/A349 and http://links.lww.com/TA/A350) show the results of the sensitivity analyses conducted with the different definitions of exposure, that is, when the last prescriptions were collected 60 and 120 days, respectively, before hospital admission. The results are consistent throughout the various definitions but for minor changes in the effects of the combined therapy. TheTMPM-ICD9injuryseveritywassignificantlyworsein the cases exposed to antithrombotic drugs (any therapy, p G 0.01). The NNTH was 595.
DISCUSSION We found that some antithrombotic medications, namely, oral anticoagulation and its combination with antiplatelets, increased the risk for being admitted to the hospital with an injury. These results support for the first time at a population level the hypothesis that the prohemorrhagic adverse effects of these therapies may compound the physical damage after trauma. Because we could not find any previous population-based research on antithrombotic drugs and injuries, it is difficult to put our findings into perspective. However, the previously mentioned studies on cohorts of hospitalized severely injured patients indicate that these drugs do have a harmful effect, supporting the credibility of our results. The importance of our study’s findings largely depends on the actual frequency of traumatic injuries, which is probably perceived as relatively low compared with other diseases usually associated with the drugs under study. We estimated the NNTH because it is a composite measure that combines the relative risk with the underlying event rate and provides estimates of risk that better correspond with reality and with the intuitive assessment made by practicing clinicians. The NNTH of 595 estimated here for the risk for trauma admission compares well, for example, with the NNTH of 500 for spontaneous extracranial hemorrhage described in a published metanalysis.21 This indicates that the importance of traumatic injuries is probably not so negligible. Because the number needed to benefit, for stroke prevention was generally less than 100, it is unlikely that the risks of these drugs come close to outweighing their benefits, but a better overall picture of their use will be warranted if future research includes injured patients too. This study relies on some assumptions. It is an epidemiologic tenet of the case-control design that controls should be sampled from the same population that gives rise to the study cases. Ideally, all individuals who underwent traumatic events of any sort, including trivial ones causing negligible or no clinical consequences, should have composed the source population. Such a population was impossible to identify for obvious reasons. Having used instead the resident population implies that what this study truly estimated is the net combination of two distinct risks: the risk for experiencing a traumatic event and the risk for this event causing an injury worth hospital admission. However, any influence of antithrombotic
drugs on the former risk has little biologic plausibility, enabling the interpretation of their effect as if purely occurring on the latter risk. Moreover, even hypothesizing such influence, it could only be a negative one because patients aware of their anticoagulated status should augment their precautions. The resulting bias would, therefore, be toward an underestimation of the actual effects on injury severity. The other variables taken into account (i.e., age, sex, and comorbidities) are admittedly more likely to exert mixed effects on the two risks (e.g., atrial fibrillation might have increased the risk for trauma through transitory episodes of loss of consciousness and falls). However, no bias should arise as long as these variables are only used to control possible confounding, and we made no attempt to interpret their effect estimates. Further support for excess bleeding as the main explanation for our findings comes from the fact that the effect size of the three therapies matched their prohemorrhagic potential, that is, increasing from antiplatelets alone to combined therapy. That the risk estimates were highest in the group of patients known to be most sensitive to bleedingVthe head injuredVis also a strength of this study. A possible limitation is that the collection of a prescription at a pharmacy does not imply its regular consumption. This is especially important for oral anticoagulation, where previous research has shown a dose effect. It is the actual degree of anticoagulationVnot its use itselfVthat predicts adverse outcomes after traumatic injuries.22 Unfortunately, information on individual drug consumption is very difficult to obtain in a population-based research, if not impossible. However, the ensuing bias should be toward an underestimation of the real effects of the medications because of an overestimation of their actual dosage. Another limitation is that we mixed together patients with miscellaneous diseases. Antiplatelet agents and oral anticoagulation may be prescribed for a wide array of circulatory diseases, which we did not attempt to separate in the analyses. However, the main purpose of the study was to ascertain whether injured patients in general are negatively affected by concurrent antithrombotic therapies. It is in the future risk-benefit appraisals of these therapies that specific grouping by disease will be more important. One possibility that would invalidate our results is that patients on antithrombotic drugs had been admitted more often to the hospital just for precautionary reasons. Patients aware of their medication status might have sought hospital assessment more often, and/or emergency department physicians might have preferentially admitted them for observation, regardless of their actual injury severity. If it were so, however, the risk increase of being admitted should concentrate in the least injured or in those with decreased LOS. As can be seen from Figure 2, this was not the case. It is true that the combined therapy had a particularly strong effect in those with LOS of less than 2 days. However, the effect was also significant in the opposite group with the longest LOS. Furthermore, the finding that the injury severity was significantly worse in patients on antithrombotic drugs adds to the fact that the clinicians’ decisions were objective. No attention should be given to the risk increase in the subgroup where the injury severity was apparently missing (Fig. 2). The TMPM classification assigns an
* 2014 Lippincott Williams & Wilkins
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
441
J Trauma Acute Care Surg Volume 76, Number 2
Di Bartolomeo et al.
aspecific severity measure (i.e., 9) to a small subset of codes.16 If a patient has only codes of this kind, his or her injury severity ends up missing. These codes are quite various, precluding a clear interpretation of the finding. Finally, as usual in an observational research, the possibility of residual confounding must be considered. Antithrombotic drug use is associated with underlying medical conditions, which in turn are causally related to the tendency to experience more serious injury from any degree of trauma. Had the Elixhauser comorbidity measure not been able to entirely control this effect, at least some of the association we found may be caused by residual confounding. However, this bias toward overestimation of the true association would be tempered by confounding in the opposite direction because the association between antithrombotic drugs and medical conditions is not monotonic. In very frail patients, the risk for spontaneous bleeding may increase disproportionately to the risk for stroke, and antithrombotic drugs may be better avoided.23 These patients are obviously also at increased risk for unfavorable trauma outcomes. In conclusion, we found that oral anticoagulation alone and its combination with antiplatelet drugs significantly increased the population risk for being admitted to hospital after an injury, both for the head injured and for the rest of trauma admissions. Because this effect is most likely to derive from the prohemorrhagic properties of these drugs, future evaluations of their risk-benefit profiles should include injured patients. AUTHORSHIP S.D.B. conceived the study, analyzed the data, and wrote the paper. M.M. created the final database from the various databanks and contributed to the analyses. F.V. designed the study and contributed to the data analysis and writing of the paper. R.D.P. revised the paper.
DISCLOSURE The authors declare no conflicts of interest.
REFERENCES 1. Friberg L, Rosenqvist M, Lip GY. Net clinical benefit of warfarin in patients with atrial fibrillation: a report from the Swedish atrial fibrillation cohort study. Circulation. 2012;125:2298Y2307. 2. Singer DE, Chang Y, Fang MC, Borowsky LH, Pomernacki NK, Udaltsova N, Go AS. The net clinical benefit of warfarin anticoagulation in atrial fibrillation. Ann Intern Med. 2009;151(5):297Y305. 3. Olesen JB, Lip GY, Lindhardsen J, Lane DA, Ahlehoff O, Hansen ML, Raunsø J, Tolstrup JS, Hansen PR, Gislason GH, et al. Risks of thromboembolism and bleeding with thromboprophylaxis in patients with atrial fibrillation: a net clinical benefit analysis using a ‘real world’ nationwide cohort study. Thromb Haemost. 2011;106(4):739Y749. 4. Sellers MB, Newby LK. Atrial fibrillation, anticoagulation, fall risk, and outcomes in elderly patients. Am Heart J. 2011;161(2):241Y246.
442
5. Pfeifer R, Tarkin IS, Rocos B, Pape HC. Patterns of mortality and causes of death in polytrauma patientsVhas anything changed? Injury. 2009;40(9): 907Y911. 6. Dossett LA, Riesel JN, Griffin MR, Cotton BA. Prevalence and implications of preinjury warfarin use: an analysis of the National Trauma Databank. Arch Surg. 2011;146(5):565Y570. 7. Bonville DJ, Ata A, Jahraus CB, Arnold-Lloyd T, Salem L, Rosati C, Stain SC. Impact of preinjury warfarin and antiplatelet agents on outcomes of trauma patients. Surgery. 2011;150(4):861Y868. 8. Ferraris VA, Bernard AC, Hyde B. The impact of antiplatelet drugs on trauma outcomes. J Trauma Acute Care Surg. 2012;73(2):492Y497. 9. Fang MC, Go AS, Chang Y, Hylek EM, Henault LE, Jensvold NG, Singer DE. Death and disability from warfarin-associated intracranial and extracranial hemorrhages. Am J Med. 2007;120(8):700Y705. 10. Torn M, Bollen WL, van der Meer FJ, van der Wall EE, Rosendaal FR. Risks of oral anticoagulant therapy with increasing age. Arch Intern Med. 2005;165(13):1527Y1532. 11. Bond AJ, Molnar FJ, Li M, Mackey M, Man-Son-Hing M. The risk of hemorrhagic complications in hospital in-patients who fall while receiving antithrombotic therapy. Thromb J. 2005;3(1):1. 12. Victorino GP, Chong TJ, Pal JD. Trauma in the elderly patient. Arch Surg. 2003;138(10):1093Y1098. 13. Greenland S, Thomas DC. On the need for the rare disease assumption in case-control studies. Am J Epidemiol. 1982;116(3):547Y553. 14. Mounib EL, Satchi T. Matched Sampling Using SAS Software. Available at: http://www2.sas.com/proceedings/sugi25/25/po/25p230.pdf. Accessed March 26, 2013. 15. Elixhauser A, Steiner C, Harris DR, Coffey RM. Comorbidity measures for use with administrative data. Med Care. 1998;36(1):8Y27. 16. Glance LG, Osler TM, Mukamel DB, Meredith W, Wagner J, Dick AW. TMPM-ICD9: a trauma mortality prediction model based on ICD-9-CM codes. Ann Surg. 2009;249(6):1032Y1039. 17. Altman DG, Schulz KF, Moher D, Egger M, Davidoff F, Elbourne D, Gøtzsche PC, Lang T; CONSORT GROUP (Consolidated Standards of Reporting Trials). The revised CONSORT statement for reporting randomized trials: explanation and elaboration. Ann Intern Med. 2001;134: 663Y694. 18. Sierra F. Evidence-Based Medicine (EBM) in practice: applying number needed to treat and number needed to harm. Am J Gastroenterol. 2005; 100(8):1661Y1663. 19. Bjerre LM, LeLorier J. Expressing the magnitude of adverse effects in case-control studies: ‘‘the number needed to be treated for one additional patient to be harmed.’’ BMJ. 2000;320(7233):503Y506. 20. Quilliam BJ, Lapane KL, Eaton CB, Mor V. Effect of antiplatelet and anticoagulant agents on risk of hospitalization for bleeding among a population of elderly nursing home stroke survivors. Stroke. 2001; 32(10):2299Y2304. 21. Hart RG, Benavente O, McBride R, Pearce LA. Antithrombotic therapy to prevent stroke in patients with atrial fibrillation: a meta-analysis. Ann Intern Med. 1999;131(7):492Y501. 22. Pieracci FM, Eachempati SR, Shou J, Hydo LJ, Barie PS. Degree of anticoagulation, but not warfarin use itself, predicts adverse outcomes after traumatic brain injury in elderly trauma patients. J Trauma. 2007;63(3): 525Y530. 23. European Heart Rhythm Association; European Association for CardioThoracic Surgery, Camm AJ, Kirchhof P, Lip GY, Schotten U, Savelieva I, Ernst S, Van Gelder IC, Al-Attar N, Hindricks G, Prendergast B, et al. Guidelines for the management of atrial fibrillation: the Task Force for the Management of Atrial Fibrillation of the European Society of Cardiology (ESC). Eur Heart J. 2010;31:2369Y2429.
* 2014 Lippincott Williams & Wilkins
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Effects of anticoagulant and antiplatelet drugs on the risk for hospital admission for traumatic injuries: A case-control and population-based study Stefano Di Bartolomeo, MD, PhD, Massimiliano Marino, PhD, Francesca Valent, MD, MSPH, and Rossana De Palma, MD, Bologna, Italy
The current cardiovascular literature advocates an overall beneficial balance between the advantages of oral anticoagulants and antiplatelet drugs in preventing and treating thromboembolic events and their disadvantages in promoting hemorrhage. However, traumatic injuries have usually received little attention despite several studies from the surgical literature showing worse outcomes in anticoagulated trauma registry patients. To quantify at population level too this seemingly deleterious impact, we investigated the effects of anticoagulants and antiplatelet use on the risk for hospital admission for acute traumatic causes. METHODS: A population-based, case-control study in an Italian region with 4.5 million inhabitants was conducted. Cases were all the 59,348 adult residents admitted to the hospital for traumatic injuries in the years 2010 and 2011. Controls were age- and sexmatched residents selected by incidence density sampling. By conditional logistic regression adjusted for comorbidities, we estimated the risk for traumatic hospital admission while on anticoagulant, antiplatelet, and combined medications. RESULTS: The odds ratios (ORs) for anticoagulation and combined medications were 1.21 (95% confidence interval [CI], 1.15Y1.28) and 1.39 (95% CI, 1.21Y1.62). These effects were generally consistent across subgroups of demographic and clinical characteristics and particularly important in the head injured (e.g., OR for anticoagulation, 2.00; 95% CI, 1.77Y12.27). Antiplatelets alone had no overall effect (OR, 1.02; 95% CI, 0.99Y1.05). The number-needed-to-harm of anticoagulation was 595. CONCLUSION: Oral anticoagulation increased the population risk for traumatic hospital admission, with a further increase in case of concurrent antiplatelet use. Because this effect is most likely to derive from the prohemorrhagic properties of these drugs, injured patients should be included in the future evaluations of the cost-benefit profiles of these medications. (J Trauma Acute Care Surg. 2014;76: 437Y442. Copyright * 2014 by Lippincott Williams & Wilkins) LEVEL OF EVIDENCE: Epidemiologic/prognostic study, level III. KEY WORDS: Trauma; thromboembolism/prevention and control; vitamin K/antagonists and inhibitors; risk assessment; hemorrhage/ etiology. BACKGROUND:
A
nticoagulant and antiplatelet drugs are increasingly used for the prevention and treatment of thromboembolic events, but they augment the risk for bleeding. In the current cardiovascular literature, the balance of risk and benefit is generally considered favorable,1Y3 and the common concerns regarding bleeding have been described as ‘‘overemphasized.’’4 Because bleeding is the leading cause of preventable death after trauma,5 the effects of anticoagulant or antiplatelet therapies in major trauma populations have been the focus of recent studies.6Y8 They concluded that (1) the concurrent use of these drugs among the severely injured has considerably Submitted: July 16, 2013, Revised: August 22, 2013, Accepted: August 22, 2013. Published online: January 6, 2014. From the Regional Agency for Health and Social Care of Emilia-Romagna (M.M., R.D.P.)/Azienda Ospedaliero (S.D.B.), Universitaria di Udine, Bologna; and Servizio di epidemiologia (F.V.), Direzione centrale salute, integrazione sociosanitaria e politiche sociali, Regione autonoma Friuli Venezia Giulia, Italy. Supplemental digital content is available for this article. Direct URL citations appear in the printed text, and links to the digital files are provided in the HTML text of this article on the journal’s Web site (www.jtrauma.com). Address for reprints: Stefano Di Bartolomeo, MD, PhD, Regional Agency for Health and Social Care of Emilia-Romagna/Azienda Ospedaliero-Universitaria di Udine, Viale Aldo Moro 21, 40127 Bologna, Italy; email: stefano.dibartolomeo@ uniud.it. DOI: 10.1097/TA.0b013e3182aa80f9
increased in the last years and (2) these drugs have a negative effect on survival or other outcomes. At the same time, the cardiovascular literature tends to minimize the importance of traumatic injuries when evaluating the adverse effects of anticoagulant and antiplatelet drugs or their net clinical benefit in the general population. For example, several important studies disregarded injured patients altogether (e.g., Friberg et al.,1 Singer et al.,2 Fang et al.9); others considered only head injury patients,3 while others remained ambiguous about their inclusion.10 Even when studies focused on patients with traumatic injuries, they were restricted to particular subgroups (i.e., hospitalized elderly at risk for falls11). A possible explanation for such discrepancy is a reduced consideration in the medical literature for traumatic injuries as compared with more common conditions such as thromboembolic diseases and spontaneous hemorrhages. However, traumatic injuries are known to be increasing in the elderlyVthe age group where anticoagulants and antiplatelets are mostly prescribed and that is rapidly expanding owing to population ageing.12 Therefore, the population risk for adverse traumatic events caused by these medications may be worth quantification. This study investigates the effect of anticoagulant and antiplatelet use on the risk for hospital admission for traumatic injuries in the general population of the Italian region Emilia-Romagna.
J Trauma Acute Care Surg Volume 76, Number 2
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
437
J Trauma Acute Care Surg Volume 76, Number 2
Di Bartolomeo et al.
PATIENTS AND METHODS Setting, Study Design, and Data Sources Emilia-Romagna is an Italian region with about 4.5 million inhabitants, a surface of 22 thousand square km, and approximately 80 hospitals admitting acute patients. It is one of the largest administrative divisions of the country, and its varied composition of urban and rural areas, industry, agriculture, and tourism make it well representative of the national territory. We conducted a population-based case-control study by linking data at an individual level from three regionwide longitudinal registries. The Hospital Discharge Registry contains the usual so-called administrative information recorded in the charts of all the regional hospitals. The Prescription Drug Registry collects information on dates, dosage, and quantities for every prescription dispensed within the regional boundaries, from both hospital and community pharmacies. The Demographic Registry holds information on age, sex, date of birth, residence, and vital status of all residents in Emilia-Romagna. A unique patient identifier allowed cross-linking between the databases. According to our institutional rules, neither patient consent nor ethical committee approval was necessary, given the observational, retrospective design of the study and the anonymity of the databases provided to the researchers.
Study Participants We selected patients admitted to hospital for an injury as cases. Controls were citizens similar to the cases but who had not been admitted to hospital with an injury. Through the Hospital Discharge Registry, we identified all admissions to the hospital for a traumatic cause among adult (917 years) residents in 2010 and 2011. The DRG International Classification of DiseasesV9th Rev.VClinical Modification (ICD-9-CM) codes used to define a traumatic cause were (principal position) from 800.x to 904.x, from 910.x to 929.x, from 940.x to 957.x, and 959.x (i.e., we excluded nontrauma such as poisoning, drowning, suffocation, and nonacute diagnoses). The controls were identified from the Demographic Registry. For each case, we identified five controls, matched by year of birth and sex. We used incidence density sampling to identify controls because the matched-pairs odds ratio estimator consistently estimates the incidence density ratio when the design is of the ‘‘incidence density’’ type and the matching is random except for sampling time.13 To facilitate the matching, we used a modified automatic computer program.14
Variables and Exposure Assessment We assessed exposure to the therapies under investigation through the Prescription Drug Registry. Cases and controls were considered exposed if they had collected at least one prescription of the drugs described later from a pharmacy within 90 days before the date of the hospital admission. The choice of interval was based on an estimation of the days of supply of each prescription through the analysis of consecutive prescriptions. We also conducted sensitivity analyses, with different intervals (60 and 120 days). We investigated two groups of drugs, based on the Anatomical Therapeutic Chemical Classification System: B01AA 438
(oral anticoagulants, i.e., warfarin or acenocoumarol) and B01AC (antiplatelet agents). Because of the possible interaction between drugs of the two groups, we considered three exposure categories as follows: exposure to drugs of the first group only, exposure to drugs of the second group only, and contemporary exposure to drugs from both groups. To measure the comorbidity status, we used the Elixhauser comorbidity measure.15 This measure was developed using administrative data from a statewide California inpatient database from all nonfederal inpatient community hospital stays in California (n = 1,779,167). The Elixhauser comorbidity measure developed a list of 30 comorbidities relying on the ICD-9-CM coding manual. To identify the codes constituted in the score, we screened the hospital records of the study participants up to 5 years before the index trauma admission of the cases. To assess the severity of the injuries of the cases, we used TMPM-ICD9, a recently described anatomic injury severity score.16 This estimates empiric injury severities for each of the injuries in the ICD-9-CM lexicon based on a regression model using data from 749,374 patients admitted to 359 hospitals in the US National Trauma Data Bank. These estimates are then combined into a probability of death, ranging from 0 to 1. We defined head injury as an ICD-9-CM code 800.xx to 801.xx, 803.xx, or 850.xx to 854.xx.
Statistical Analysis We estimated the relative risk for being admitted to hospital for an injury while on anticoagulant/antiplatelet therapy by using conditional logistic regression. The relative risks were adjusted for the comorbidity status. All 30 comorbidities of the Elixhauser score were included in the model as binary variables. We conducted subgroup analyses across categories of demographic factors and of characteristics of the injured patients. As a confirmatory analysis, we compared the injury severity among the exposed (any type of antithrombotic therapy) and nonexposed cases with the Wilcoxon rank-sum test. Presenting results as relative risks is standard in observational studies but may be difficult to translate into clinical practice. The number-needed-to-harm will better reflect any adverse effect of treatment than relative risks in clinical terms.17,18 For this reason, we finally calculated the numberneeded-to-treat-for-harm (NNTH) using methods proposed by Bjerre and LeLorier19 for case-control study designs. For simplicity, we calculated it for oral anticoagulation only. Calculation of such a measure requires estimation of the incidence rate of outcome among the unexposed population. Because we were unable to directly estimate this rate from our sample, we estimated an approximate rate from the study base of residents adult citizens from which our sample arose, as done by Quilliam et al.20 This approximate background rate of traumatic hospital admissions in the general population was 0.008 per person per year.
RESULTS We identified 59,348 adults who were admitted to the hospital for traumatic injuries during the study period and 296,740 controls. The characteristics of cases and controls are shown in Table 1. Age and sex were similar in cases and controls by design. The majority of patients were older than * 2014 Lippincott Williams & Wilkins
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
J Trauma Acute Care Surg Volume 76, Number 2
Di Bartolomeo et al.
TABLE 1. Characteristics of Cases and Controls Cases All, No. Age, n (%), y 18Y44 45Y54 55Y64 65Y74 Q75 Sex, male, n (%) Comorbidity, n (%) Congestive heart failure Cardiac arrhythmias Valvular disease Pulmonary circulation disorders Peripheral vascular disorders Hypertension Paralysis Neurodegenerative disorders Chronic pulmonary disease Diabetes, uncomplicated Diabetes, complicated Hypothyroidism Renal failure Liver disease Peptic ulcer disease, no bleeding Acquired immunodeficiency syndrome/human immunodeficiency virus Lymphoma Metastatic cancer Solid tumor without metastasis Rheumatoid arthritis/collagen vascular diseases Coagulopathy Obesity Weight loss Fluid and electrolyte disorders Blood loss anemia Deficiency anemia Alcohol abuse Drug abuse Psychosis Depression Type of antithrombotic drugs, n (%) (a) None (b) Oral anticoagulants only (c) Antiplatelet agents only (d) Association of b + c
59,348 14,118 (23.79) 6,213 (10.47) 6,460 (10.88) 8,005 (13.49) 24,552 (41.37) 28,944 (48.77)
Controls 296,740 70,590 (23.79) 31,065 (10.47) 32,300 (10.88) 40,030 (13.49) 122,755 (41.37) 144,720 (48.77)
3,754 (6.33)* 5,464 (9.21)* 1,685 (2.84)* 696 (1.17)* 1,372 (2.31)* 12,558 (21.16)* 464 (0.78)* 1,834 (3.09)* 3,583 (6.04)* 3,977 (6.70)* 1,020 (1.72)* 1,265 (2.13)* 2,181 (3.67)* 1,377 (2.32)* 86 (0.14) 60 (0.10)*
12,291 (4.14)* 16,072 (5.42)* 5,674 (1.91)* 2,033 (0.69)* 5,044 (1.70)* 36,891 (12.43)* 972 (0.33)* 3,328 (1.12)* 10,347 (3.49)* 10,034 (3.38)* 2,460 (0.83)* 3,327 (1.12)* 5,819 (1.96)* 3,064 (1.03)* 322 (0.11) 80 (0.03)*
221 (0.37)* 570 (0.96)* 2,100 (3.54) 731 (1.23)*
670 (0.23)* 2,270 (0.76)* 10,198 (3.44) 1,769 (0.60)*
380 (0.64)* 636 (1.07)* 312 (0.53)* 923 (1.56)* 677 (1.14)* 825 (1.39)* 681 (1.15)* 157 (0.26)* 408 (0.69)* 1,785 (3.01)*
683 (0.23)* 2,222 (0.75)* 779 (0.26)* 2,967 (1.00)* 1,946 (0.66)* 2,429 (0.82)* 639 (0.22)* 256 (0.09)* 842 (0.28)* 3,950 (1.33)*
44,392 (74.80)* 2,684 (4.52)* 11,993 (20.21)* 279 (0.47)*
231,255 (77.93)* 9,095 (3.06)* 55,688 (18.77)* 702 (0.24)*
*p G 0.01 (W2 test).
65 years, confirming the increasing importance of traumatic injuries in the elderly. Among cases, comorbid conditions were, in general, more represented. Cases were also more often on antithrombotic therapies, for example, 25.2% as compared with 22.1% of controls.
Table 2 displays the characteristics of cases related to their traumatic injuries. More than 90% of the cases were admitted for 2 days or more, 12% of the patients experienced head injuries, and the overall mortality was 2.58%. The crude and adjusted relative risks of the investigated therapies are shown in Figure 1. Oral anticoagulation significantly increased the risk for hospital admission for trauma both in the crude and adjusted analyses (21% increase in adjusted risk). Antiplatelet therapy increased the risk only in the crude analysis, while its effect was not significant after adjustment. The combined therapy had a larger effect than oral anticoagulation alone (39% increase of adjusted risk), although with a larger confidence interval owing to the smaller number of cases (Fig. 1). Figure 2 shows the results of subgroup analyses conducted across patient characteristics. Antiplatelet agents seemed to have a significant effect in the youngest group (18Y44 years). Oral anticoagulation and combined therapy showed no special patterns across the various categories of injury severity, except in the group with missing injury severity, where their effect was stronger. In patients with very short admissions (length of hospital stay [LOS], 0Y1 day) the effect of oral anticoagulation alone was not significant, while the effect of its combination with antiplatelets was especially strong. In those with head injuries, the effect was large for all kinds of therapies; however, oral anticoagulation and combined therapy had a smaller but significant effect in patients without head injuries too. In patients admitted to the intensive care unit, the effect of oral anticoagulation and combined therapies was smaller and nonsignificant, while the effect of antiplatelets was significant.
TABLE 2. Selected Characteristics of Cases Characteristic Injury Severity,* quintiles 0.000043Y0.004514 0.004518Y0.008667 0.008672Y0.014115 0.014118Y0.019641 0.019647Y0.956034 Missing** Length of Hospital Stay, d 0Y1 2Y3 Q3 Type of injury Head injury No head injury Intensive care unit admission Yes No Mortality (30 d)
n (%) 12,907 (22.46) 11,205 (19.50) 10,912 (18.99) 11,193 (19.48) 11,248 (19.57) 1,883 (3.17) 7,321 (12.34) 12,180 (20.52) 39,847 (67.14) 6,932 (11.68) 52,416 (88.32) 2,758 (4.65) 56, 590 (95.35) 1,534 (2.58)
*TMPM-ICD9 injury severity measure, ranging from 0 to 1. See text for details. **Missing because of the fact that the TMPM-ICD9 scoring system provides for some injuries only an aspecific and adimensional severity measure (i.e., 9).
* 2014 Lippincott Williams & Wilkins
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
439
J Trauma Acute Care Surg Volume 76, Number 2
Di Bartolomeo et al.
Figure 1. Antithrombotic drugs: risk estimates of being admitted to the hospital for trauma.
Figure 2. Antithrombotic drugs: risk estimates of being admitted to the hospital for trauma. Subgroup analyses. 440
* 2014 Lippincott Williams & Wilkins
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
J Trauma Acute Care Surg Volume 76, Number 2
Di Bartolomeo et al.
The detailed effects of the 30 comorbidities of the Elixhauser classification are shown in Table 1 of the Supplemental Digital Content (http://links.lww.com/TA/A348). Figures S1 and S2 of the Supplemental Digital Content (http://links.lww.com/TA/A349 and http://links.lww.com/TA/A350) show the results of the sensitivity analyses conducted with the different definitions of exposure, that is, when the last prescriptions were collected 60 and 120 days, respectively, before hospital admission. The results are consistent throughout the various definitions but for minor changes in the effects of the combined therapy. TheTMPM-ICD9injuryseveritywassignificantlyworsein the cases exposed to antithrombotic drugs (any therapy, p G 0.01). The NNTH was 595.
DISCUSSION We found that some antithrombotic medications, namely, oral anticoagulation and its combination with antiplatelets, increased the risk for being admitted to the hospital with an injury. These results support for the first time at a population level the hypothesis that the prohemorrhagic adverse effects of these therapies may compound the physical damage after trauma. Because we could not find any previous population-based research on antithrombotic drugs and injuries, it is difficult to put our findings into perspective. However, the previously mentioned studies on cohorts of hospitalized severely injured patients indicate that these drugs do have a harmful effect, supporting the credibility of our results. The importance of our study’s findings largely depends on the actual frequency of traumatic injuries, which is probably perceived as relatively low compared with other diseases usually associated with the drugs under study. We estimated the NNTH because it is a composite measure that combines the relative risk with the underlying event rate and provides estimates of risk that better correspond with reality and with the intuitive assessment made by practicing clinicians. The NNTH of 595 estimated here for the risk for trauma admission compares well, for example, with the NNTH of 500 for spontaneous extracranial hemorrhage described in a published metanalysis.21 This indicates that the importance of traumatic injuries is probably not so negligible. Because the number needed to benefit, for stroke prevention was generally less than 100, it is unlikely that the risks of these drugs come close to outweighing their benefits, but a better overall picture of their use will be warranted if future research includes injured patients too. This study relies on some assumptions. It is an epidemiologic tenet of the case-control design that controls should be sampled from the same population that gives rise to the study cases. Ideally, all individuals who underwent traumatic events of any sort, including trivial ones causing negligible or no clinical consequences, should have composed the source population. Such a population was impossible to identify for obvious reasons. Having used instead the resident population implies that what this study truly estimated is the net combination of two distinct risks: the risk for experiencing a traumatic event and the risk for this event causing an injury worth hospital admission. However, any influence of antithrombotic
drugs on the former risk has little biologic plausibility, enabling the interpretation of their effect as if purely occurring on the latter risk. Moreover, even hypothesizing such influence, it could only be a negative one because patients aware of their anticoagulated status should augment their precautions. The resulting bias would, therefore, be toward an underestimation of the actual effects on injury severity. The other variables taken into account (i.e., age, sex, and comorbidities) are admittedly more likely to exert mixed effects on the two risks (e.g., atrial fibrillation might have increased the risk for trauma through transitory episodes of loss of consciousness and falls). However, no bias should arise as long as these variables are only used to control possible confounding, and we made no attempt to interpret their effect estimates. Further support for excess bleeding as the main explanation for our findings comes from the fact that the effect size of the three therapies matched their prohemorrhagic potential, that is, increasing from antiplatelets alone to combined therapy. That the risk estimates were highest in the group of patients known to be most sensitive to bleedingVthe head injuredVis also a strength of this study. A possible limitation is that the collection of a prescription at a pharmacy does not imply its regular consumption. This is especially important for oral anticoagulation, where previous research has shown a dose effect. It is the actual degree of anticoagulationVnot its use itselfVthat predicts adverse outcomes after traumatic injuries.22 Unfortunately, information on individual drug consumption is very difficult to obtain in a population-based research, if not impossible. However, the ensuing bias should be toward an underestimation of the real effects of the medications because of an overestimation of their actual dosage. Another limitation is that we mixed together patients with miscellaneous diseases. Antiplatelet agents and oral anticoagulation may be prescribed for a wide array of circulatory diseases, which we did not attempt to separate in the analyses. However, the main purpose of the study was to ascertain whether injured patients in general are negatively affected by concurrent antithrombotic therapies. It is in the future risk-benefit appraisals of these therapies that specific grouping by disease will be more important. One possibility that would invalidate our results is that patients on antithrombotic drugs had been admitted more often to the hospital just for precautionary reasons. Patients aware of their medication status might have sought hospital assessment more often, and/or emergency department physicians might have preferentially admitted them for observation, regardless of their actual injury severity. If it were so, however, the risk increase of being admitted should concentrate in the least injured or in those with decreased LOS. As can be seen from Figure 2, this was not the case. It is true that the combined therapy had a particularly strong effect in those with LOS of less than 2 days. However, the effect was also significant in the opposite group with the longest LOS. Furthermore, the finding that the injury severity was significantly worse in patients on antithrombotic drugs adds to the fact that the clinicians’ decisions were objective. No attention should be given to the risk increase in the subgroup where the injury severity was apparently missing (Fig. 2). The TMPM classification assigns an
* 2014 Lippincott Williams & Wilkins
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
441
J Trauma Acute Care Surg Volume 76, Number 2
Di Bartolomeo et al.
aspecific severity measure (i.e., 9) to a small subset of codes.16 If a patient has only codes of this kind, his or her injury severity ends up missing. These codes are quite various, precluding a clear interpretation of the finding. Finally, as usual in an observational research, the possibility of residual confounding must be considered. Antithrombotic drug use is associated with underlying medical conditions, which in turn are causally related to the tendency to experience more serious injury from any degree of trauma. Had the Elixhauser comorbidity measure not been able to entirely control this effect, at least some of the association we found may be caused by residual confounding. However, this bias toward overestimation of the true association would be tempered by confounding in the opposite direction because the association between antithrombotic drugs and medical conditions is not monotonic. In very frail patients, the risk for spontaneous bleeding may increase disproportionately to the risk for stroke, and antithrombotic drugs may be better avoided.23 These patients are obviously also at increased risk for unfavorable trauma outcomes. In conclusion, we found that oral anticoagulation alone and its combination with antiplatelet drugs significantly increased the population risk for being admitted to hospital after an injury, both for the head injured and for the rest of trauma admissions. Because this effect is most likely to derive from the prohemorrhagic properties of these drugs, future evaluations of their risk-benefit profiles should include injured patients. AUTHORSHIP S.D.B. conceived the study, analyzed the data, and wrote the paper. M.M. created the final database from the various databanks and contributed to the analyses. F.V. designed the study and contributed to the data analysis and writing of the paper. R.D.P. revised the paper.
DISCLOSURE The authors declare no conflicts of interest.
REFERENCES 1. Friberg L, Rosenqvist M, Lip GY. Net clinical benefit of warfarin in patients with atrial fibrillation: a report from the Swedish atrial fibrillation cohort study. Circulation. 2012;125:2298Y2307. 2. Singer DE, Chang Y, Fang MC, Borowsky LH, Pomernacki NK, Udaltsova N, Go AS. The net clinical benefit of warfarin anticoagulation in atrial fibrillation. Ann Intern Med. 2009;151(5):297Y305. 3. Olesen JB, Lip GY, Lindhardsen J, Lane DA, Ahlehoff O, Hansen ML, Raunsø J, Tolstrup JS, Hansen PR, Gislason GH, et al. Risks of thromboembolism and bleeding with thromboprophylaxis in patients with atrial fibrillation: a net clinical benefit analysis using a ‘real world’ nationwide cohort study. Thromb Haemost. 2011;106(4):739Y749. 4. Sellers MB, Newby LK. Atrial fibrillation, anticoagulation, fall risk, and outcomes in elderly patients. Am Heart J. 2011;161(2):241Y246.
442
5. Pfeifer R, Tarkin IS, Rocos B, Pape HC. Patterns of mortality and causes of death in polytrauma patientsVhas anything changed? Injury. 2009;40(9): 907Y911. 6. Dossett LA, Riesel JN, Griffin MR, Cotton BA. Prevalence and implications of preinjury warfarin use: an analysis of the National Trauma Databank. Arch Surg. 2011;146(5):565Y570. 7. Bonville DJ, Ata A, Jahraus CB, Arnold-Lloyd T, Salem L, Rosati C, Stain SC. Impact of preinjury warfarin and antiplatelet agents on outcomes of trauma patients. Surgery. 2011;150(4):861Y868. 8. Ferraris VA, Bernard AC, Hyde B. The impact of antiplatelet drugs on trauma outcomes. J Trauma Acute Care Surg. 2012;73(2):492Y497. 9. Fang MC, Go AS, Chang Y, Hylek EM, Henault LE, Jensvold NG, Singer DE. Death and disability from warfarin-associated intracranial and extracranial hemorrhages. Am J Med. 2007;120(8):700Y705. 10. Torn M, Bollen WL, van der Meer FJ, van der Wall EE, Rosendaal FR. Risks of oral anticoagulant therapy with increasing age. Arch Intern Med. 2005;165(13):1527Y1532. 11. Bond AJ, Molnar FJ, Li M, Mackey M, Man-Son-Hing M. The risk of hemorrhagic complications in hospital in-patients who fall while receiving antithrombotic therapy. Thromb J. 2005;3(1):1. 12. Victorino GP, Chong TJ, Pal JD. Trauma in the elderly patient. Arch Surg. 2003;138(10):1093Y1098. 13. Greenland S, Thomas DC. On the need for the rare disease assumption in case-control studies. Am J Epidemiol. 1982;116(3):547Y553. 14. Mounib EL, Satchi T. Matched Sampling Using SAS Software. Available at: http://www2.sas.com/proceedings/sugi25/25/po/25p230.pdf. Accessed March 26, 2013. 15. Elixhauser A, Steiner C, Harris DR, Coffey RM. Comorbidity measures for use with administrative data. Med Care. 1998;36(1):8Y27. 16. Glance LG, Osler TM, Mukamel DB, Meredith W, Wagner J, Dick AW. TMPM-ICD9: a trauma mortality prediction model based on ICD-9-CM codes. Ann Surg. 2009;249(6):1032Y1039. 17. Altman DG, Schulz KF, Moher D, Egger M, Davidoff F, Elbourne D, Gøtzsche PC, Lang T; CONSORT GROUP (Consolidated Standards of Reporting Trials). The revised CONSORT statement for reporting randomized trials: explanation and elaboration. Ann Intern Med. 2001;134: 663Y694. 18. Sierra F. Evidence-Based Medicine (EBM) in practice: applying number needed to treat and number needed to harm. Am J Gastroenterol. 2005; 100(8):1661Y1663. 19. Bjerre LM, LeLorier J. Expressing the magnitude of adverse effects in case-control studies: ‘‘the number needed to be treated for one additional patient to be harmed.’’ BMJ. 2000;320(7233):503Y506. 20. Quilliam BJ, Lapane KL, Eaton CB, Mor V. Effect of antiplatelet and anticoagulant agents on risk of hospitalization for bleeding among a population of elderly nursing home stroke survivors. Stroke. 2001; 32(10):2299Y2304. 21. Hart RG, Benavente O, McBride R, Pearce LA. Antithrombotic therapy to prevent stroke in patients with atrial fibrillation: a meta-analysis. Ann Intern Med. 1999;131(7):492Y501. 22. Pieracci FM, Eachempati SR, Shou J, Hydo LJ, Barie PS. Degree of anticoagulation, but not warfarin use itself, predicts adverse outcomes after traumatic brain injury in elderly trauma patients. J Trauma. 2007;63(3): 525Y530. 23. European Heart Rhythm Association; European Association for CardioThoracic Surgery, Camm AJ, Kirchhof P, Lip GY, Schotten U, Savelieva I, Ernst S, Van Gelder IC, Al-Attar N, Hindricks G, Prendergast B, et al. Guidelines for the management of atrial fibrillation: the Task Force for the Management of Atrial Fibrillation of the European Society of Cardiology (ESC). Eur Heart J. 2010;31:2369Y2429.
* 2014 Lippincott Williams & Wilkins
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
