Venous thromboembolism (VTE), which includes deep venous thrombosis (DVT) and pulmonary embolism (PE), is a leading cause of morbidity and mortality in hospitalized patients. Among more than seven million patients discharged from 944 North American acute care hospitals, postoperative VTE was the second most common medical complication, the second most common cause of excess length of stay, and the third most common cause of excess mortality and costs (Zhan 2003). Indeed, PE is the most common preventable cause of hospital death, and preventing VTE has been ranked as the number one of 79 strategies aimed to improve patient safety in hospitals (Shojania 2001).
A U.S. population-based study reported that hospital-acquired DVT and PE occurs in 1.3% and 0.4% of hospital admissions, respectively (Stein 2005), and about 60% of all VTE events occur as a result of a current or recent hospital admission (Heit 2002). The risk factors for hospital-acquired VTE are well-characterized and include surgery, acute medical illness, cancer and cancer therapy, trauma, immobilization, central venous catheters, previous history of VTE, older age and obesity (Anderson 2003). Almost all hospitalized patients have at least one risk factor for VTE, and approximately 40% have three or more risk factors (Anderson 2003; Kucher 2005; NICE 2007).
Numerous randomized clinical trials over the past 30 years have provided evidence that the use of primary thromboprophylaxis in hospitalized patients at risk for VTE is safe, effective and cost-effective in reducing DVT and PE (Geerts 2008). Clinical practice guidelines that have systematically reviewed and synthesized the evidence from these trials have strongly recommended the use of thromboprophylaxis in hospitalized patients at risk for VTE (Geerts 2008; Gould 2012; Falck-Ytter 2012; Kahn 2012; NICE 2007; Nicolaides 2006), and some have explicitly recommended that hospitals should develop a formal strategy that addresses VTE prevention, ideally in the form of a written, hospital-wide thromboprophylaxis policy (Geerts 2008). Notwithstanding the publication of more than 20 practice guidelines since 1986 recommending the use of thromboprophylaxis, audits conducted in numerous countries in various groups of hospitalized patients show that thromboprophylaxis continues to be underutilized or utilized inappropriately (Ahmad 2002; Cohen 2008; Deheinzelin 2006; Ellis 2004; Kahn 2007; Kakkar 2003; National Institute of Clinical Studies 2005; Rajaganeshan 2006; Rashid 2005; Stinnett 2005; Tapson 2005; Yu 2007). Furthermore, population-based data have not shown a reduction in either the overall incidence of VTE over time (Silverstein 1998) or the number of deaths from PE in hospitalized patients (Tsai 2012). Hence, it is clear that a gulf exists between the available evidence and the systematic implementation of this evidence into clinical practice.
In the last few years, in an effort to reduce preventable mortality and morbidity in hospital settings, there has been an increased focus at national levels on the best ways to systematically improve compliance with VTE prophylaxis recommendations (e.g. National Institute for Health and Clinical Excellence, U.K.; The Joint Commission and National Quality Forum, U.S.A.; Canadian Patient Safety Institute, Canada). Researchers have begun to address this issue as well. Examples of system-wide measures that have been evaluated to date include passive strategies such as distribution of guidelines or educational events, multi-component approaches, audit and feedback, and the use of automatic reminders such as preprinted orders and computer reminders (Kucher 2005; Schunemann 2004; Tooher 2005).
In this review, we aimed to assess the effects of various system-wide interventions designed to increase the implementation of thromboprophylaxis in hospitalized medical and surgical patients at risk for VTE, and to identify which interventions are most effective.
The objective of this review was to assess the effects of interventions designed to increase the implementation of thromboprophylaxis in hospitalized medical and surgical patients at risk for venous thromboembolism (VTE). We assessed effectiveness in terms of:
- Increase in the proportion of patients who receive prophylaxis (RP) and receive appropriate prophylaxis (RAP).
- Reduction in the proportion of symptomatic VTE (all VTE; deep vein thrombosis (DVT) [any, proximal, distal]; pulmonary embolism (PE) and fatal PE).
- Reduction in the proportion of asymptomatic VTE (detected by systematic screening of patients who do not have symptoms of DVT or PE).
- Safety of the intervention.
Criteria for considering studies for this review
Types of studies
All studies including a control group and evaluating the effects of an intervention designed to increase the implementation of thromboprophylaxis in hospitalized medical and surgical patients at risk for VTE. Thus, studies encompassed randomized controlled trials (RCTs), cluster RCTs, quasi-RCTs (for example, using pseudo-randomization such as even or odd date of birth), and non-randomized studies (NRS) with or without concurrent controls. The control group comparison could be no intervention, an existing policy, or another type of intervention. Studies could be in any language.
Types of participants
Hospitalized adult medical or surgical inpatients.
Types of interventions
Any strategies targeted to individuals or to clusters aimed to increase the use of thromboprophylaxis in hospitalized patients at risk for VTE and/or decrease the rate of symptomatic or asymptomatic VTE. Examples of interventions include electronic alerts, human alerts, sticker systems, reminders, audit and feedback, pre-printed orders or educational materials.
We excluded studies that simply distributed published guidelines and studies whose interventions were not clearly described.
Types of outcome measures
- Increase in the proportion of patients who received prophylaxis (RP) or received appropriate prophylaxis (RAP). Prophylaxis could be either pharmacologic or mechanical. The definition of appropriate prophylaxis was that used by the respective study authors.
- Reduction in the proportion of symptomatic VTE (all VTE; DVT [any, proximal, distal]; PE, fatal PE).
- Reduction in the proportion of asymptomatic VTE.
- Safety of the intervention, e.g. frequency of clinically relevant bleeding (major hemorrhage; minor hemorrhage) or other complications.
Studies were included if the study design, population and intervention were clearly described, if data were provided separately by intervention group and, for VTE outcomes, if VTE was diagnosed using objective, accepted criteria.
Search methods for identification of studies
The Cochrane Peripheral Vascular Diseases Group Trials Search Co-ordinator (TSC) searched the Group's Specialised Register (last searched July 2010) and the Cochrane Central Register of Controlled Trials (CENTRAL, The Cochrane Library) 2010, Issue 3. See Appendix 1 for details of the search strategy used to search CENTRAL. The Specialised Register is maintained by the TSC and is constructed from weekly electronic searches of MEDLINE, EMBASE, CINAHL, and AMED, and through handsearching relevant journals. The full list of the databases, journals and conference proceedings which have been searched, as well as the search strategies used are described in the Specialised Register section of the Cochrane Peripheral Vascular Diseases Group module in The Cochrane Library (www.thecochranelibrary.com).
In addition, the authors searched the PubMed, EMBASE, and SCOPUS databases (19 April 2010) using combinations of controlled vocabulary subject headings and free text words. See Appendix 2 and Appendix 3 for details regarding the electronic searches of PubMed and EMBASE, respectively. We used SCOPUS and Web of Science citation indexes to identify studies that cited relevant articles identified by our search.
Searching other resources
We handsearched the reference lists of relevant retrieved studies and those of narrative and systematic reviews to find additional potentially-relevant studies.
Data collection and analysis
Selection of studies
Two review authors independently reviewed full-texts of each study and indicated on a Study Eligibility Form if it should be included, excluded, or undecided. Disagreements regarding study inclusion were resolved by discussion between the two review authors and, if necessary, by the involvement of a third independent review author. All studies marked 'undecided' by one review author were discussed further between the two review authors, and then deemed included or excluded.
Data extraction and management
Two review authors independently extracted data from the included articles. The data obtained for each study were entered in duplicate into two identical databases that were designed using Microsoft Access®. The two databases were compared for inaccuracies and any data entry errors were corrected. If agreement on the data entered for a given data field could not be reached between the two extractors, a third extractor was consulted. A third, final database was populated with the final corrected data. The following information on Patients, Intervention, Comparator, Outcome, Setting (i.e. PICOS) was extracted from each study (if available) using a standardized data extraction form (one form per study) based on the Cochrane EPOC data collection template.
- Study design: randomized (cluster or non-cluster, randomization procedure, unit of randomization/analysis), non-randomized study (type of design)
- Study period
- Setting (hospital/centre characteristics): number of centres, university-affiliated hospital, community hospital, physician practice, other, type of healthcare system (public versus private), departments included
- Physician characteristics: number of physicians, physician specialties
- Patient characteristics: patient types (medical, surgical, trauma, other), inclusion/exclusion criteria, number of patients screened/included, average age, percent male, comorbidities and VTE risk profile (e.g. proportion with cancer, etc.)
- Intervention (active and control arms): type (educational, development of guidelines, distribution of guidelines, reminders, (electronic alerts, human alerts, pre-printed order, sticker system), audit and feedback, multifaceted interventions, other, and detailed description
- Control group characteristics: timing (before or concurrent with intervention group), types of patients, interventions
- VTE prophylaxis and dose used in the study (where available): pharmacologic (type, dose), mechanical
- Was appropriate prophylaxis assessed? How appropriateness of prophylaxis was defined
- Method of VTE screening and/or diagnosis
- Outcomes, raw data and/or effect estimates
- Number and/or proportion of patients who received thromboprophylaxis and appropriate thromboprophylaxis
- Number and/or proportion of patients with symptomatic VTE, DVT (any, proximal, distal), PE (any, fatal)
- Number and/or proportion of patients with asymptomatic DVT (any, proximal, distal)
- Number and/or proportion of patients with complications possibly related to the intervention: major bleeding, minor bleeding, all-cause mortality, sudden death, thrombocytopenia
- Effect estimate and variance estimates for these outcomes where raw data were unavailable
- Risk of bias
Information on methodological quality and potential biases was also extracted for each study, as described in the following section. We constructed tables of characteristics that describe study data and methodological quality (i.e. 'Risk of bias' tables) for each study.
Assessment of risk of bias in included studies
Two review authors independently assessed the quality of each study with a component approach rather than summarizing study quality in an overall score. We used the Cochrane Collaboration's tool for assessing the risk of bias (Higgins 2008). Disagreements were resolved by discussion with co-authors. We assessed the following potential sources of bias, and rated them as high risk, low risk or unclear risk of bias.
For all studies:
- Was the source population clearly defined?
- Was the study population well-described (setting, location, relevant dates)?
- Does the study population represent the source population or population of interest?
- Was the intervention well-described?
- Was the outcome defined and diagnosed appropriately?
- Were patients, physicians, assessors or others blinded to the intervention?
- Was degree of completeness of follow up described, was completeness of follow up adequate, how was loss to follow up addressed, was censoring described or addressed?
- Were incomplete outcome data (i.e. missing data) adequately addressed?
- Were statistical analyses described adequately, were analyses appropriate, did authors provide sufficient presentation of data?
- Were the authors' conclusions supported by the results of the analysis?
- Was the study report free of the suggestion of selective outcome reporting?
For randomized controlled trials:
- Was knowledge of the allocated interventions adequately prevented during the study?
- Was allocation adequately concealed?
- Was the allocation sequence adequately generated?
- Were there clinically relevant differences between the comparator groups at baseline, and were these differences accounted for in the analysis?
- For cluster design trials, was clustering accounted for in the sample size calculation and in the analysis?
For non-randomized studies:
- Were confounders pre-defined and adequately measured?
- Was confounding accounted for in the design and/or analysis?
Measures of treatment effect
Studies were grouped and analyzed by type of intervention, study design and type of outcome. For our primary outcomes, we summarized the effects of the reviewed interventions using the risk difference (RD) which provides an absolute measure of effect. We examined I
Dealing with missing data
We did not contact the original investigators to request missing data. We did not use statistical methods to impute or model missing data.
Assessment of heterogeneity
We assessed and investigated clinical and methodological sources of heterogeneity across studies using the I
We tested for statistical heterogeneity by calculating the I
Assessment of reporting biases
Funnel plots centered around the pooled studies effect (separate plots for relative risks and risk differences) were graphed and examined visually to assess the potential for publication bias, keeping in mind that other factors can contribute to asymmetry of the funnel plot such as selective outcome reporting, differences in methodological quality among studies, poor methodological quality leading to spuriously inflated effects in smaller studies, true heterogeneity, artefact, and chance (Egger 1997; Higgins 2008).
Where there were sufficient data (≥ 3 studies), we calculated a summary statistic for each intervention category (alert, educational, multifaceted) and associated outcome using a random-effects model to pool risk differences (RP and RAP) or risk ratios (VTE and safety outcomes). Studies that were not pooled were tabulated and presented descriptively with the corresponding meta-analyses (RD).
Subgroup analysis and investigation of heterogeneity
To investigate sources of heterogeneity, there were only sufficient studies to perform a subgroup analysis in one grouping of studies (NRS-Multifaceted-RAP), where we investigated heterogeneity in the following subgroups: intervention component (alert versus no alert), patient type (medical, surgical, both), and hospital type (university versus non-university affiliated).
We performed a meta-regression with the subgroup variables to formally test whether there was evidence of different effects in different subgroups of studies.
As a sensitivity analysis, we performed an influence analysis, in which we assessed the degree to which excluding single studies, one by one, influenced the magnitude, precision or direction of the summary results.
Description of studies
Results of the search
Figure 1 displays the flow diagram of study selection.
|Figure 1. Flow diagram of study selection.|
We included 55 studies (57 citations) (Characteristics of included studies), including eight randomized controlled trials (Anderson 1994; Dexter 2001; Fontaine 2006; Garcia 2009; Kucher 2005a; Labarere 2007; Overhage 1996; Piazza 2009) and 47 non-randomized studies (Agu 2000; Avery 1995; Baskin 2008; Bauer 2008; Birks 2002; Boddi 2009; Brand 2009; Buhannic 1997; Bullock-Palmer 2008; Burns 2005; Byrne 1996; Cohn 2006; Dobesh 2005; Durieux 2000; Elsasser 2007; Fagot 2001; Fiumara 2010; Frankel 1999; Gladding 2007; Grupper 2006; Harinath 1998; Hawkins 2008; Huang 2000; Krimsky 2009; Labarere 2004a; Lecumberri 2008; Maynard 2010; McEleny 1998; McKenna 2009; McMullin 2006; Mosen 2004; Nendaz 2010; O'Connor 2009; Pattar 2005; Patterson 1998; Peterson 1999; Rashid 2005; Roberts 2006; Scaglione 2005; Sellier 2006; Shedd 2008; Sobieraj 2008; Stewart 2006; Stinnett 2005; Streubel 2009; Taylor 2000; Teich 2000). However, only 54 studies had data available for analysis (n = 78,343 patients), as one study (Bauer 2008) only reported data in patient-days and we were unable to extract the total number of patients included. Of the 54 studies, 50 were included in a quantitative synthesis. Four studies (Anderson 1994; Garcia 2009; Labarere 2007; Lecumberri 2008) were not included in the quantitative analysis because there were not enough other comparable studies assessing the same intervention and outcome (did not meet the minimum number of studies required to perform meta-analyses).
We excluded 21 studies which failed to meet our eligibility criteria (Characteristics of excluded studies). Eleven were excluded because the aim of the study was not to increase the use of thromboprophylaxis (Anderson 1996; Cronin 2009; Dobesh 2008; Gibbs 2009; Gidiri 2004; Labarere 2004b; Peterman 2006; Pham 2008; Samama 2006; Thomas 1983; Waltering 2007), four studies were excluded because there was no concurrent reference group (Kucher 2009; Maffei 2009; Stark 2006; Vallano 2004), two studies were not original research (Kakkar 2004; Marco 2008), two were studies whose interventions were not described (Ruttimann 2005; Schumock 2004), one study did not provide outcomes separately by intervention group (Gallagher 2009), and one study had the same active intervention in the comparison group (Baroletti 2008).
Risk of bias in included studies
The 'Risk of bias' tables for each study are given in the table Characteristics of included studies. Figure 2 shows a summary of methodological quality of the included studies, as judged by the review authors.
|Figure 2. Methodological quality graph: review authors’ judgements about each methodological quality item presented as percentages across all included studies.|
In this review, the extent of allocation concealment was unclear for all of the randomized controlled trials (RCTs) (Anderson 1994; Brand 2009; Dexter 2001; Fontaine 2006; Garcia 2009; Kucher 2005a; Labarere 2007; Overhage 1996; Piazza 2009). Among the RCTs, sequence generation was clearly reported in one trial (Labarere 2007), unclear in seven trials (Anderson 1994; Brand 2009; Dexter 2001; Fontaine 2006; Garcia 2009; Overhage 1996; Piazza 2009), and inadequate in one trial (Kucher 2005a). While sequence generation was clear in Labarere 2007, some unblinding may have occurred during the course of the study as knowledge of the allocated interventions was not adequately prevented.
For most studies, there was an unclear or high risk of bias as a result of lack of or inadequate blinding of study participants or outcome assessors (Figure 3). Only three studies (Agu 2000; Labarere 2004a; Stewart 2006) demonstrated adequate blinding of participants and assessors.
|Figure 3. Methodological quality summary: review authors’ judgements about each methodological quality item for each included study|
Incomplete outcome data
Of the 55 studies included in our review, one was judged to have a high risk of bias due to incomplete outcome reporting (Boddi 2009), 45 studies had an unclear risk of bias due to incomplete outcome reporting (Agu 2000; Anderson 1994; Avery 1995; Bauer 2008; Birks 2002; Brand 2009; Buhannic 1997; Bullock-Palmer 2008; Burns 2005; Byrne 1996; Dexter 2001; Durieux 2000; Elsasser 2007; Fagot 2001; Fiumara 2010; Fontaine 2006; Frankel 1999; Garcia 2009; Gladding 2007; Harinath 1998; Huang 2000; Krimsky 2009; Kucher 2005a; Labarere 2004a; Labarere 2007; Lecumberri 2008; Maynard 2010; McEleny 1998; McKenna 2009; McMullin 2006; Mosen 2004; Nendaz 2010; O'Connor 2009; Overhage 1996; Pattar 2005; Patterson 1998; Peterson 1999; Roberts 2006; Scaglione 2005; Shedd 2008; Sobieraj 2008; Stewart 2006; Stinnett 2005; Streubel 2009; Teich 2000) and nine had a low risk of incomplete outcome reporting (Baskin 2008; Cohn 2006; Dobesh 2005; Grupper 2006; Hawkins 2008; Piazza 2009; Rashid 2005; Sellier 2006; Taylor 2000) (Figure 3).
In 10 studies there was a low risk of bias due to selective outcome reporting (Bullock-Palmer 2008; Fiumara 2010; Hawkins 2008; Kucher 2005a; Labarere 2004a; Labarere 2007; Maynard 2010; Piazza 2009; Scaglione 2005; Sellier 2006). In 35 studies there was uncertainty regarding selective outcome reporting (Agu 2000; Avery 1995; Baskin 2008; Birks 2002; Brand 2009; Buhannic 1997; Byrne 1996; Cohn 2006; Dexter 2001; Dobesh 2005; Durieux 2000; Elsasser 2007; Fagot 2001; Fontaine 2006; Garcia 2009; Gladding 2007; Grupper 2006; Harinath 1998; Huang 2000; Krimsky 2009; McEleny 1998; McKenna 2009; Nendaz 2010; O'Connor 2009; Overhage 1996; Pattar 2005; Patterson 1998; Rashid 2005; Roberts 2006; Shedd 2008; Sobieraj 2008; Stewart 2006; Stinnett 2005; Taylor 2000; Teich 2000). Ten studies were judged to have a high risk of bias due to selective outcome reporting for the following reasons: 1) not reporting on VTE outcomes when reducing VTE was stated as a study objective; and 2) not reporting safety outcomes (bleeding, heparin induced thrombocytopenia, etc.) while reporting on VTE outcomes (Anderson 1994; Bauer 2008; Boddi 2009; Burns 2005; Frankel 1999; Lecumberri 2008; McMullin 2006; Mosen 2004; Peterson 1999; Streubel 2009).
Effects of interventions
Studies were grouped for meta-analysis based on study type (RCT, NRS), intervention type (alert, educational, multifaceted), and outcome (RP, RAP, symptomatic DVT). Table 1 summarizes the results from the meta-analyses conducted for our primary outcomes and corresponds to the forest plots described below. Additional tables 2 to 7 summarize the results for every outcome and include the results of meta-analyses in the presence of more than 2 studies or the results of individual studies where we had less than 3 studies and thus chose not to statistically pool the results.
Summary of results for effectiveness of interventions
Randomized controlled trials - Received prophylaxis
Among the RCTs, we were only able to pool studies for the 'alert' intervention category. Among the four studies pooled, 583/2893 patients randomized to the control groups and 1117/2883 patients randomized to the intervention groups received prophylaxis. The meta-analysis showed that there was a significant increase in the proportion of patients who received prophylaxis among patients in the 'alert' groups; pooled RD 0.13 (95% CI 0.01 to 0.25)). There was substantial heterogeneity in the results of the individual studies (I
|Figure 4. Forest Plot: RCT - Alerts - Received Prophylaxis|
Randomized controlled trials - Outcomes without sufficient data for meta-analysis
Two RCTs (Dexter 2001; Garcia 2009) reported the outcome 'received appropriate prophylaxis' ( Table 2). Dexter 2001 found a significant increase of 13% in the rate of appropriate prophylaxis and Garcia 2009 found a similar effect estimate, but this did not reach statistical significance. The confidence intervals for these estimates may have been wider if clustering was adequately accounted for in the analysis. Unfortunately, the studies did not provide sufficient data (ICC or adjusted confidence intervals) for us to report cluster-adjusted estimates, where appropriate.
Two RCTs (Kucher 2005a; Piazza 2009) reported venous thromboembolism outcomes. Kucher 2005a found a 41% reduction in the rate of all VTE, and a 60% reduction in the rate of PE, both statistically significant. Piazza 2009 found a 21% reduction in the rate of all VTE, 20% reduction in the rate of symptomatic DVT, and a 37% reduction in the rate of PE; however, no estimate was statistically significant.
Two RCTs (Kucher 2005a; Piazza 2009) reported safety outcomes. There were no significant effects on the rates of major bleeding (Kucher 2005a; Piazza 2009), minor bleeding (Kucher 2005a), or all cause mortality (Kucher 2005a; Piazza 2009).
Non-randomized studies - Received prophylaxis
Among the five NRS of alerts pooled to assess the proportion of patients who received prophylaxis, 3055/6845 patients in the control groups and 3833/6608 in the intervention groups received prophylaxis. The pooled effect estimate suggests a positive effect of alerts, however this did not reach statistical significance; RD 0.09 (95% CI -0.00 to 0.19). There was substantial heterogeneity in the results of the individual studies (I
|Figure 6. Forest Plot: NRS - Alerts - Received Prophylaxis|
Non-randomized studies - Received appropriate prophylaxis
Among the ten NRS of alerts pooled to assess the proportion of patients who received appropriate prophylaxis, 2330/3721 patients in the control groups and 2746/3664 patients in the intervention groups received appropriate prophylaxis. The meta-analysis showed that alerts were associated with a significant increase in the proportion of patients receiving appropriate prophylaxis; RD 0.18 (95% CI 0.12 to 0.24). There was substantial heterogeneity in the results of the individual studies (I
|Figure 7. Forest Plot: NRS - Alerts - Received Appropriate Prophylaxis|
Non-randomized studies - All venous thromboembolism
Among the three NRS of alerts pooled to assess VTE, there were 53/8943 patients in the control groups and 47/8860 in the intervention groups who presented with a VTE. There was no significant reduction in DVT associated with an alert, RR 0.85 (95% CI 0.49 to 1.46). There was moderate heterogeneity of effects among the individual studies (I
|Figure 8. Forest Plot: NRS - Alerts - All VTE|
Non-randomized studies - Outcomes without sufficient data for meta-analysis
Among the NRS of alerts, one study (Lecumberri 2008) reported other venous thromboembolism outcomes ( Table 3). Lecumberri 2008 found no significant differences in the rates of symptomatic DVT or PE. Two studies (Fiumara 2010; Lecumberri 2008) reported safety outcomes. Lecumberri 2008 found no difference in the rate of major bleeding. Fiumara 2010 found no difference in the rate of minor bleeding but noted a significant 52% increase in the rate of all cause mortality.
There were no RCTs of education interventions ( Table 4).
Non-randomized studies - Received prophylaxis
Among the four NRS of education pooled to assess the proportion of patients who received prophylaxis, there were 426/718 patients in the control groups and 547/734 in the intervention groups who received prophylaxis. The pooled effect estimate suggests a positive effect of education, however this did not reach statistical significance; RD 0.08 (95% CI -0.02 to 0.18). There was substantial heterogeneity of effects among the individual studies (I
|Figure 9. Forest Plot: NRS - Education - Received Prophylaxis|
Non-randomized studies - Received appropriate prophylaxis
Among the six NRS of education pooled to assess the proportion of patients who received appropriate prophylaxis, there were 734/1399 patients in the control groups and 671/1078 in the intervention groups who received appropriate prophylaxis. The meta-analysis showed that educational interventions were associated with a significant increase in the proportion of patients receiving appropriate prophylaxis; RD 0.11 (95% CI 0.06 to 0.17). There was moderate heterogeneity of effects among the individual studies (I
|Figure 10. Forest Plot: NRS - Education - Received Appropriate Prophylaxis|
Non-randomized studies - Outcomes without sufficient data for meta-analysis
Three NRS of educational interventions (Boddi 2009; Frankel 1999; Streubel 2009) reported venous thromboembolism outcomes ( Table 5). Streubel 2009 found a reduction in the rate of all VTE and Frankel 1999 found a reduction in the rate of symptomatic DVT, but neither was statistically significant. Boddi 2009 found a significant 60% reduction in the rate of symptomatic DVT but no effect for PE.
Randomized controlled trials - Outcomes without sufficient data for meta-analysis
Two RCTs of multifaceted interventions (Anderson 1994; Labarere 2007) reported the outcome 'received prophylaxis' ( Table 6). Anderson 1994 found a 28% increase in the rate of prophylaxis, which was statistically significant. Labarere 2007 found a 7% increase, but this did not reach statistical significance.
One RCT (Labarere 2007) reported venous thromboembolism outcomes. Labarere 2007 noted an increase in the rate of asymptomatic DVT, but this was not statistically significant. Two RCTs (Anderson 1994; Labarere 2007) reported safety outcomes. Anderson 1994 reported that there was no difference in the rate of all cause mortality, and Labarere 2007 reported no difference in the rate of thrombocytopenia.
Non-randomized studies - Received prophylaxis
Among the 15 NRS of multifaceted interventions pooled to assess the proportion of patients who received prophylaxis, there were 5081/9939 patients in the control groups and 8327/10212 in the intervention groups who received prophylaxis. The meta-analysis showed that multifaceted interventions were associated with a significant increase in the proportion of patients receiving prophylaxis; RD 0.17 (95% CI 0.09 to 0.25). There was substantial heterogeneity of effects among the individual studies (I
|Figure 11. Forest Plot: NRS - Multifaceted - Received Prophylaxis|
|Figure 12. Cumulative meta-analysis by precision for NRS - Multifaceted - Received Prophylaxis to investigate publication bias.|
Non-randomized studies - Received appropriate prophylaxis
Among the 14 NRS of multifaceted interventions pooled to assess the proportion of patients who received appropriate prophylaxis, there were 1562/4628 patients in the control groups and 1808/3538 patients in the intervention groups who received appropriate prophylaxis. The meta-analysis showed that multifaceted interventions were associated with a significant increase in the proportion of patients receiving appropriate prophylaxis; RD 0.19 (95% CI 0.11 to 0.26). There was substantial heterogeneity of effects among the individual studies (I
|Figure 13. Forest Plot: NRS - Multifaceted - Received Appropriate Prophylaxis|
|Figure 14. Cumulative meta-analysis by precision for NRS - Multifaceted - Received Appropriate Prophylaxis to investigate publication bias.|
Non-randomized studies - All venous thromboembolism
Among the four NRS of multifaceted interventions pooled to assess VTE, there were 147/11070 patients in the control groups and 117/12921 patients in the intervention groups who presented with a VTE. The meta-analysis showed that multifaceted interventions were not associated with a significant increase in the risk of VTE; RR 1.01 (95% CI 0.51 to 1.98). Important heterogeneity was detected (I
|Figure 15. Forest Plot: NRS - Multifaceted - all VTE|
Non-randomized studies - Symptomatic deep vein thrombosis
Among the three NRS of multifaceted interventions pooled to assess symptomatic DVT, there were 132/15275 patients in the control groups and 94/17700 patients in the intervention groups who developed symptomatic DVT. The meta-analysis showed that multifaceted interventions were not associated with a difference in the risk of symptomatic DVT; RR 0.59 (95% CI 0.18 to 1.98). There was substantial heterogeneity in the results (I
|Figure 16. Forest Plot: NRS - Multifaceted - symptomatic DVT|
Non-randomized studies - Outcomes without sufficient data for meta-analysis
Among the NRS of multifaceted interventions, three studies (Labarere 2004a; Maynard 2010; Sellier 2006) reported other venous thromboembolism outcomes ( Table 7). Maynard 2010 reported no effect on the rate of PE. Labarere 2004a found a 66% reduction in the rate of asymptomatic DVT and Sellier 2006 found a 39% reduction; both were statistically significant. One study (Sellier 2006) reported safety outcomes, finding no significant differences in the rates of major bleeding, minor bleeding, or thrombocytopenia.
Subgroup analyses for non-randomized studies - multifaceted interventions - received appropriate prophylaxis outcome
We chose to conduct our subgroup analyses among the intervention group that had the largest number of studies and for our most precise outcome of 'received appropriate prophylaxis'.
Subgroup A - Intervention components: In this analysis, we divided our studies into those with multifaceted interventions that included an alert/reminder and those that did not. This subgroup analysis explained some of the heterogeneity, as indicated by the decreased I
Subgroup B - Patient types: In this analysis, we divided our studies into those that included medical patients only, surgical only, or both. This subgroup analysis explained some of the heterogeneity, as the I
Subgroup C - Type of hospital: In this analysis, we divided our studies into those that were conducted in university-affiliated hospitals and those that were not. This subgroup analysis explained some of the heterogeneity, as indicated by the decreased I
Subgroup D – Control group prophylaxis rate: In this analysis, we divided our studies into those which had a low baseline/control rate of prophylaxis (< 50%) versus those with a high baseline rate (≥ 50%). This subgroup analysis did not explain heterogeneity and pooled effects were similar in both groups ( Table 8 ) (P = 0.796) ( Table 9).
Subgroup E – Region: In this analysis, we divided our studies into those conducted in North America, Europe, and elsewhere (in this case, Australia). This subgroup analysis did not explain heterogeneity and pooled effects were similar in all three groups ( Table 8) ( Table 9).
Summary of main results
VTE is a leading, potentially preventable cause of morbidity and mortality in hospitalized patients. We systematically reviewed 55 studies which implemented a variety of system-wide strategies aimed to improve thromboprophylaxis rates in many settings and patient populations. Among 55 studies eligible for review (that included a total of 78,343 patients), we found statistically significant improvements in prescription of prophylaxis associated with alerts (randomized controlled trials (RCTs)) and multifaceted interventions (RCTs and NRS), and improvements in prescription of appropriate prophylaxis in non-randomized studies (NRS) with the use of education, alerts or multifaceted interventions. However, in NRS, multifaceted interventions with an alert component seem to be more effective than alerts alone or education alone.
Strengths of our review are that we included studies of various designs, interventions, settings and patient populations, our review protocol was peer reviewed, we reviewed the reference lists of all included studies and reviews in an attempt to identify additional studies, screening of titles/abstracts and full texts was done by two reviewers, we included studies published in any language, and we performed double data extraction and assessment of risk of bias.
Overall completeness and applicability of evidence
Many studies did not report on appropriateness of prophylaxis. Most studies did not assess clinical outcomes such VTE and bleeding. Overall, there were few RCTs (8 out of 55 studies reviewed).
Quality of the evidence
Our systematic approach to searching, study selection, and data extraction followed that of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2008). The methodological quality of the included studies was variable (Figure 3). For the RCTs, incomplete reporting did not allow proper scoring of relevant study design features such as sequence generation and allocation concealment in the majority of included studies.
Potential biases in the review process
Our search was last performed on 19 April 2010. After completion of this review, a search of PubMed in February 2012 using our search strategy (Appendix 2) identified 226 new articles published since April 2010, of which 22 may be relevant to our review. These and future studies will be considered in future updates of this review. As only titles and abstracts were included in our search strategy, there is the possibility that studies that did not have an abstract or that did not contain our designated search terms in their abstract or title were missed. We did not specifically search any foreign language databases, such as LILACS, and as such there may be a bias towards studies published in North America, Australia and Europe.
We believe that there is a low potential for bias due to our screening process. We were very inclusive, eliminating only those studies with interventions that did not aim to increase the use of thromboprophylaxis, appropriate thromboprophylaxis, or decrease the incidence of VTE. Furthermore, record screening was done in duplicate and any study identified as potentially relevant by either review author was included in the full-text stage of study selection.
The data extraction form used in this review was designed to capture data on a variety of outcomes and the data were extracted by two different review authors. It is therefore unlikely that any data were missed if they were reported within the article itself. However, we did not contact the authors of the various studies to ask if there were omitted or missing data, and the 'Risk of bias' assessment presented in this review was based solely on what was presented in the articles. In addition, one study (Bauer 2008) reported data in patient-days, which our extraction form was not designed to capture.
For our meta-analysis, we used unadjusted risk differences as very few studies reported adjusted effect estimates. As a result, there is a potential for confounding. The 'received prophylaxis' group may be an imperfect surrogate for 'received appropriate prophylaxis'. There were only two non-randomized studies that were true, controlled, before and after studies with a no intervention control group in both the pre and post time periods, permitting adjustment for temporal changes in prophylaxis rates.
Agreements and disagreements with other studies or reviews
There have been two previous systematic reviews of this topic which were less methodologically rigorous than our review and are now outdated. In Tooher’s review (Tooher 2005) of 30 studies published between 1996 and 2003, it was reported that passive dissemination of guidelines is unlikely to improve VTE prophylaxis practice and that multiple active strategies used together that incorporate methods to remind clinicians to assess patients for DVT risk and aid with the selection of appropriate prophylaxis are likely to achieve better outcomes (Tooher 2005). A subsequent, less comprehensive review that searched the literature to 2006 drew similar conclusions (Michota 2007) to the Tooher 2005 article. More recently, there was a literature review of methods to improve prophylaxis and decrease VTE events in the hospitalized patient (Mahan 2010). This review referenced articles published between 1996 and 2008 and came to similar conclusions as Tooher 2005; namely, that interventions that are active, rather than passive, appear to be more effective (Mahan 2010).
Implications for practice
Among 55 studies with data available for meta-analysis, we found statistically significant improvements in rates of prescription of prophylaxis and prescription of appropriate prophylaxis associated with each of education, alerts, and multifaceted intervention strategies. Multifaceted interventions that included an alert component appeared to be more effective (as suggested by greater pooled risk difference) than alerts alone or education alone.
Having any one type of intervention can be effective, but having a multifaceted approach that combines various interventions, including education (consistently effective) and an alert (greater pooled risk difference) appears to have the greatest effect. An educational intervention is expected to be less costly than an alert intervention, although a recent evaluation of the economic impact of an e-alert to prevent venous thromboembolism (VTE) at a single hospital in Spain described reduced incidence of VTE and a net cost saving with use of the e-alert (Lecumberri 2011). We hope that the results of our review will help physicians, hospital administrators and policy makers make practical decisions tailored to their own settings about adoption of specific system-wide measures to improve prevention of VTE in hospitalized patients. VTE prevention quality initiative programs could also benefit from our findings.
Implications for research
Several questions remain unsolved about the effectiveness of system-wide interventions to increase the implementation of thromboprophylaxis in hospitalized medical and surgical patients at risk for VTE. While most of the interventions we reviewed were effective at increasing rates of prophylaxis or appropriate prophylaxis, absolute differences tended to be modest (in the 10 to 20% range). Research is needed to understand why such interventions do not have a more pronounced effect on prescribing of prophylaxis, and why effects of interventions were greater in studies of non university-affiliated than university-affiliated hospitals and in studies of surgical patients than medical patients. Most studies were performed at single centers; future multicenter studies should evaluate the effectiveness and generalizability of applying a given intervention at multiple centers, ideally including university-affiliated and community hospitals of various sizes. Study of the comparative cost-effectiveness of various system-wide interventions is also required.
Dr Kahn is a recipient of a National Research Scientist Award from the Fonds de la Recherche en Santé du Québec.
This review was funded in part by a Canadian Institutes for Health Research Knowledge Synthesis Grant (# KRS-103271).
Data and analyses
This review has no analyses.
Appendix 1. CENTRAL Search Strategy
Appendix 2. Authors' Pubmed Search Strategy
[19 April 2010] (n = 1163)
(((medical[tiab] OR hospitalized[tiab] OR high-risk[tiab] OR “high risk”[tiab] OR surgical[tiab] OR older[tiab] OR “at risk”[tiab]) AND patients[tiab]) OR inpatient[tiab] OR inpatients[tiab]) AND (“Electronic alerts”[tiab] OR Intervention[tiab] OR strategy[tiab] OR “preprinted order”[tiab] OR educational[tiab] OR education[tiab] OR audit[tiab] OR feedback[tiab] OR “preprinted sticker”[tiab] OR order[tiab] OR “electronic tool”[tiab] OR “computerized alerts”[tiab] OR “computer reminders”[tiab] OR(prescription[tiab] AND aids[tiab])) AND (("Venous Thrombosis"[Mesh] OR “vein thrombosis”[tiab] OR DVT[tiab] OR “deep vein thrombosis”[tiab] OR “deep venous thrombosis”[tiab] OR VTE[tiab] OR “venous thromboembolism”[tiab] OR PE[tiab] OR “pulmonary embolism”[tiab] OR “blood clot”[tiab] OR “phlebitis”[tiab] OR “clot”[tiab] OR “thrombosis”[tiab] OR “thrombus”[tiab] OR phlebothrombosis[tiab] OR emboli[tiab] OR embolism[tiab] OR anticoagulant[tiab]) AND (Prophylaxis[tiab] OR Prevention[tiab] OR reduction[tiab] OR decrease[tiab] OR diminish[tiab] OR prophylactic[tiab] OR preventative[tiab] OR prevent [tiab] OR ((adherence[tiab] OR compliance[tiab]) AND guidelines[tiab]) OR Thromboprophylaxis[tiab]))
Appendix 3. Authors' EMBASE Search Strategy
[1980-April 2010] (n = 1285)
- “high risk”.ti,ab.
- “at risk”.ti,ab.
- 1 OR 2 OR 3 OR 4 OR 5 OR 6 OR 7
- 8 AND 9
- 10 OR 11
- “electronic alerts”.ti,ab.
- “preprinted order”.ti,ab.
- “preprinted sticker”.ti,ab.
- “electronic tool”.ti,ab.
- “computerized alerts”.ti,ab.
- “computer reminders”.ti,ab.
- “prescription aids”.ti,ab.
- 13 OR 14 OR 15 OR 16 OR 17 OR 18 OR 19 OR 20 OR 21 OR 22 OR 23 OR 24 OR 25 OR 26
- exp vein thrombosis/
- vein thrombosis.ti,ab.
- venous thrombosis.ti,ab.
- "deep vein thrombosis".ti,ab.
- "deep venous thrombosis".ti,ab.
- venous thromboembolism.ti,ab.
- pulmonary embolism.ti,ab.
- blood clot.ti,ab.
- 28 OR 29 OR 30 OR 31 OR 32 OR 33 OR 34 OR 35 OR 36 OR 37 OR 38 OR 39 OR 40 OR 41 OR 42 OR 43 OR 44 OR 45 OR 46
- 48 OR 49 OR 50 OR 51 OR 52 OR 53
- 55 OR 56
- 57 AND 58
- 54 OR 59
- 47 AND 60
- 61 OR 62
- 10 AND 23 AND 63
Appendix 4. Influence Analysis
Contributions of authors
1. Article reviewers: Dave Morrison (DM), Dr Susan Kahn (SK), Jacqueline Cohen (JC), Dr Vicky Tagalakis (VT), and Jessica Emed (JE)
2. Resolving disputes: SK, JC
3. Statistical expertise: JC
4. Content expertise: SK, VT, JE, Dr Andre Roussin (AR), Dr William Geerts (WG)
5. Administrative coordination: DM
6. Literature searches: DM, JC
7. Drafting the manuscript: SK, DM, JC
8. Revising the manuscript: SK, DM, JC, JE, VT, AR WG.
Declarations of interest
The authors of this review have not received any funding to undertake this review other than the peer-reviewed grant noted above. The authors report the following declarations of interest:
JE received an honorarium for participation in a single meeting (focus group) with LEO Pharma for work unrelated to the submitted review.
WG reports board membership (Canadian Patient Safety Institute (Safer Health Care Now) National lead for venous thromboembolism prevention), consultancy (Bayer Healthcare, Boehringer Ingelheim, Leo Pharma, Pfizer, Sanofi) and payment for lectures (Bayer Healthcare, Boehringer Ingelheim, Leo Pharma, Pfizer, Sanofi) and development of educational presentations (Bayer Healthcare, Leo Pharma). Other support has been received by his institution from Boehringer Ingelheim and Bayer Healthcare for clinical and quality of care outreach programs. WG reports that these relationships in no way impact on his involvement with this Cochrane review.
SK has received grant support from public granting agencies (CIHR) for research on the treatment of venous thrombosis. She participated in one industry-sponsored advisory board (Boehringer-Ingelheim) on the treatment of venous thrombosis and provided expert testimony for the Canadian Medical Protective Association. SK also reports that Sanofi Aventis has partnered with her institution to help create a center of excellence in thrombosis and anti coagulation. The funds will go to renovate space for the center.
VT has received and currently holds grant support from the CIHR for research in venous thrombosis. She also reports consultancy activities for Sanofi Aventis, Bayer, and Pfizer and development of educational materials on venous thrombosis for Sanofi Aventis. She has received support for investigator-initiated research projects from Pfizer and Sanofi Aventis.
AR reports board membership and consultancy activities for BMS, BI, Pfizer, Bayer and received payment for lectures from BMS, BI, Bayer, Pfizer, Astra, and Merck not related to this review. AR also reports that his institution has received a CIHR grant for Aids vascular research and payment for development of educational presentations from BI, Bayer, BMS, Pfizer for the preparation of university accredited symposiums and slide kits.
Sources of support
- National Research Scientist Award from the Fonds de la Recherche en Santé du Québec, Canada.
- Canadian Institutes for Health Research Knowledge Synthesis Grant (# KRS-103271), Canada.
- Chief Scientist Office, Scottish Government Health Directorates, The Scottish Government, UK.The PVD Group editorial base is supported by the Chief Scientist Office.
Differences between protocol and review
There are no substantial differences between our protocol (Kahn 2010) and our final review. Minor changes include that were unable to perform sensitivity analyses based on methodological quality to assess the robustness of our summary results, and we decided not to combine randomized and non-randomized studies.
In our protocol we specified we would include only prospective studies. We decided not to exclude studies based on this criterion because it was difficult to determine when the data were collected relative to when the study was conducted. Further, because many of these studies used medical record review to assess prophylaxis use, even if the study was conducted retrospectively, the data were collected in a prospective manner. Therefore, we decided to include studies regardless of their prospective and retrospective nature which was often either unclear or unimportant.
Medical Subject Headings (MeSH)
*Hospitalization; Anticoagulants [therapeutic use]; Australia; Europe; Hospitals; North America; Postoperative Complications [epidemiology; prevention & control]; Pulmonary Embolism [epidemiology; prevention & control]; Randomized Controlled Trials as Topic; Venous Thromboembolism [epidemiology; *prevention & control]; Venous Thrombosis [epidemiology; prevention & control]
MeSH check words
* Indicates the major publication for the study