Given that the overall rate of bleeding events found in patients with ACS during routine clinical practice has been reported as 30%,10 300 patients would have to be included in each group to decrease this rate by at least 30% with an α=5% and a power (1-β)=80%.

The value of 30% was set taking into account:

• •

  The decrease in bleeding events using fondaparinux and bivalirudin in clinical trials.11,12

• •

  The extrapolation of these results to the practice of routine care.

• •

  The use of other intervention measures.

All statistical analyses were performed using the SPSS software package (version 18) for Windows (SPSS, Chicago, USA) and EPIDAT V 3.1. A P value of < 0.05 was used as a cutoff for statistical significance. Qualitative variables are expressed as a frequency distribution and continuous variables are expressed as mean ± standard deviation. Numerical variables with a nonnormal distribution are expressed as median [interquartile range]. The Kolmogorov-Smirnov test was used to test for normality.

The Student t test for normally distributed variables was used to compare differences between groups. The Levene test for equal variances was used to assess the equality of variances. The nonparametric Mann-Whitney U test was used to compare the variables with a nonnormal distribution. The Pearson chi-squared test or Fisher exact test were used to study the association between categorical variables and the linear association test was used to study qualitative variables with a linear trend per category. To assess the effect of the intervention on reducing bleeding events, relative risk reduction, relative risk, odds ratio (OR), the number needed to treat, and 95% confidence intervals (95%CI) were used.

Multivariate logistic regression analysis was used to study the impact of the intervention while controlling for the following potential confounders: age > 75 years, female sex, body weight < 60kg, diabetes mellitus, chronic anemia, chronic kidney failure, a history of bleeding, percutaneous coronary intervention, and catheterization access site.2,13 The adjusted OR was obtained. The Hosmer-Lemeshow test was used to assess the goodness-of-fit and its calibration and the area under the ROC (Receiver Operating Characteristic) curve was used to assess discriminatory ability.

The cost-benefit analysis took into account the cost-savings and costs of the implementation of the program.

The impact on health outcomes was evaluated through the comparison of bleeding events and 30-day readmission between the 2 cohorts. A bleeding event was defined as any bleeding event meeting any of the criteria of the Bleeding Academic Research Consortium (BARC) classification.14 The hospital administration service provided data on the number of readmissions, which were calculated according to the patient's discharge date. Scheduled admissions were not counted.

The study was approved by the Clinical Research Ethics Committee. The hospital did not consider it necessary to obtain informed consent from the patients because the aim of the study was to improve care based on scientific and organizational evidence.

In the PRE phase, most of the patients were men (70.6%) and the mean age was 67 years. The most frequent bleeding event risk factor was percutaneous coronary intervention (79.3%), followed by the use of the femoral artery (51.2%) (Table 1). Correctable factors contributing to bleeding events were identified as:

• •

  Overdosing with an antithrombotic: in the PRE phase, 17.8% of patients were overdosed with at least 1 drug (Table 2). For this reason, the institutional protocol was revised to define in detail the standard pattern of antithrombotic use on the basis of the main clinical practice guidelines and technical specifications.15,16 In addition, it was found that the doses of antithrombotics (whose dosage depends on body weight) were frequently calculated by the approximate visual estimation of weight. However, because it has been shown that this method is imprecise,17 in the absence of an appropriate method for body weight determination, the patient or the relatives should be asked directly. Thus, weight was measured by the use of a lift coupled to a scale. In the PRE phase, weight was known in 67.4% of the patients.

  Table 2.

  Total Bleeding Events and Effectiveness of Interventions in Both Patient Cohorts

  	PRE (n=377)	POST (n=300)	P
  Total bleeding events	119 (31.6)	67 (22.3)	.010
  Bleeding events according to severity
  BARC 1	57 (15.1)	31 (10.3)	.09
  BARC 2	46 (12.2)	28 (9.3)	.29
  BARC 3a	6 (1.6)	4 (1.3)	.97
  BARC 3b	8 (2.1)	4 (1.3)	.63
  BARC 3c	1 (0.3)	0 (0.0)	.91
  BARC 5	1 (0.3)	0 (0.0)	.91
  Bleeding events according diagnosis*
  NSTEACS (n=237 and n=175)	77 (32.5)	38 (21,7)	.022
  STEACS (n=122 and n=118)	29 (23.8)	26 (22.0)	.90
  Patients overdosed with antithrombotics	67 (17.8)	18 (6.0)	< .001
  Patients receiving fondaparinux	9 (2.4)	152 (50.7)	< .001
  Patients on combination therapy with bivalirudin and abciximab	11 (2.9)	1 (0.3)	.03

  BARC, Bleeding Academic Research Consortium; NSTEACS, non—ST-segment elevation acute coronary syndrome; POST, postintervention; PRE, preintervention; STEACS, ST–segment elevation acute coronary syndrome.

  Data are expressed as no. (%).

  *

  The sum does not match the total number of patients because there were 25 for whom the type of acute coronary syndrome was not included in the report.

• •

  The use of drugs with a more unfavorable safety profile regarding bleeding events and alternatives with a lower risk of bleeding events.

  • –

    Enoxaparin in patients with non—ST-segment elevation acute coronary syndrome and high risk of bleeding according to the CRUSADE scale. In the PRE phase, 97.6% of the patients received enoxaparin treatment. However, the use of fondaparinux in this subgroup of patients has been demonstrated to have a reduced risk of bleeding.11 Only 2.4% of patients were treated with fondaparinux in the PRE phase.

  • –

    Unfractionated heparin in combination with abciximab. The alternative to this combination (bivalirudin) has shown a decreased risk of bleeding without reduced effectiveness.12 In the PRE phase, 17.0% of the patients received treatment with unfractionated heparin and abciximab and 8.7% with bivalirudin.

• •

  The use of combinations of antithrombotics with a higher risk of bleeding, such as the combination of abciximab and bivalirudin. In the PRE phase, 11 patients received treatment with this combination. However, the clinical practice guidelines recommend that its use should be restricted and provisional because the systematic use of bivalirudin and a glycoprotein IIb/IIIa inhibitor has a demonstrated increased risk of bleeding, but without greater efficacy compared with bivalirudin alone.12

The baseline characteristics of the patients and the presence of bleeding risk factors in POST were similar to those in the PRE phase (Table 1). Regarding the degree of acceptance of the proposed intervention measures, it was found that:

• •

  The percentage of overdosed patients decreased by 66.3% in POST (17.8% vs 6.0%; P < .001). The percentage of patients with known weight increased from 67.4% to 88.7% (P<.001) (Table 2).

• •

  Regarding the use of drugs with a more favorable safety profile:

  • –

    The percentage of patients treated with fondaparinux increased in POST (2.4% vs 50.7%; P<.001), while the use of enoxaparin decreased (97.6% vs 49.3%; P<.001) (Table 2).

  • –

    Only 3.3% of patients were treated with bivalirudin (P=.007), while the combination of unfractionated heparin and unfractionated abciximab was 21.3% (P=.019) in POST.

• •

  Regarding the use of combinations of drugs with an increased risk of bleeding, only 1 patient (0.3%) was treated with the combination of bivalirudin and abciximab (P=.03) (Table 2) in POST.

The overall percentage of in-hospital bleeding events in POST was reduced by 29.2% (31.6% PRE vs 22.3% POST; OR, 0.62; 95%CI, 0.44-0.88) (Table 2). The adjusted OR was 0.58 (95%CI, 0.40-0.85). The result of the Hosmer-Lemeshow test was not statistically significant (P=.36) and the area under the curve was 0.66.

Regarding the number needed to treat, 11 patients needed intervention measures to prevent a bleeding event (95%CI, 7-39). The readmission rate was 7.7% in the PRE phase and 5% in the POST phase (P=.20) (14 readmissions in the PRE phase and 8 readmissions in the POST phase were for cardiovascular reasons).

The present study is limited by its retrospective design. Its most important bias lies in the difficulty of validating the information obtained from medical records. However, the reliability of the study is increased by the comparison of 2 patient groups given that the magnitude of bias is similar in both groups. Another aspect associated with the limitations of retrospective studies was the inability to record ischemic and bleeding risk a posteriori, which would undoubtedly have been very useful information.23 However, it should be clarified that the assessment of GRACE and CRUSADE scores is considered in the Cardiology Service Protocols of our hospital for the indication for antiplatelet therapy in patients with ACS.9 Secondly, savings were estimated using a partial extrapolation of the data published by Soldevila et al.5 rather than by using data from our own hospital. The estimation was done in this way because the aim of the study was to obtain an approximate cost-benefit analysis rather than an exhaustive pharmacoeconomic evaluation. Furthermore, we consider it appropriate to use data from the aforementioned study because it was performed under similar conditions (reperfusion strategies used, antithrombotic therapy used, national study). Third, the frequent current use of prasugrel and ticagrelor may suggest that the bleeding event rate could have been higher than that reported in the study. However, this eventuality would not diminish the effectiveness of the intervention in reducing bleeding events. Finally, it was difficult to demonstrate a statistical difference in less prevalent events, such as readmissions, because a very large sample size would have been needed in each group.