Article Text

A unit-based intervention aimed at improving patient adherence to pharmacological thromboprophylaxis
  1. Charles Alexander Baillie1,
  2. James P Guevara2,
  3. Raymond C Boston3,
  4. Todd E H Hecht4
  1. 1Section of Hospital Medicine, Pennsylvania Hospital, Philadelphia, Pennsylvania, USA
  2. 2Department of General Pediatrics, Children's Hospital of Philadelphia, Philadelphia, Pennsylvania, USA
  3. 3Department of Clinical Studies, University of Pennsylvania School of Veterinary Medicine, Kennett Square, Pennsylvania, USA
  4. 4Division of General Internal Medicine, University of Pennsylvania Perelman School of Medicine, Philadelphia, Pennsylvania, USA
  1. Correspondence to Dr Charles Alexander Baillie, Section of Hospital Medicine, Pennsylvania Hospital, Philadelphia, PA 19107, USA; charles.baillie{at}


Background Pharmacological thromboprophylaxis is necessary among many hospitalised patients to prevent venous thromboembolism (VTE). However, a significant number of clinician-ordered doses are not administered with many doses refused by patients. We aimed to assess the impact and sustainability of a multifaceted intervention to improve medication adherence to pharmacological thromboprophylaxis. The intervention included a standardised nursing response to patient refusal, daily assessment of VTE prophylaxis usage and regular feedback on refusal rates.

Methods We conducted a quasi-experimental study of patients admitted between January 2010 and November 2012 to one of six hospital intervention units (three medical and three oncology units) or five control units. The primary outcome was the proportion of VTE prophylaxis doses missed for any reason.

Results A total of 20 208 admissions occurred at the six hospital units during the study period. In the pre-post analysis, the rate of missed and refused doses decreased significantly after the intervention (24.7% to 14.7% and 18.3% to 9.4%, respectively; p value <0.01 for both comparisons). In multiple regression models with interrupted time series analysis, the intervention was associated with an immediate and sustained decrease in missed (adjusted OR 0.64; 95% CI 0.55 to 0.74 and 0.98; 95%CI 0.97 to 0.99) and refused doses (adjusted OR per month 0.58; 95% CI 0.48 to 0.71 and 0.97; 95%CI 0.96 to 0.98). No immediate or sustained reduction in missed or refused doses was observed in the control units.

Conclusions Implementation of a multifaceted intervention resulted in an immediate and sustained decrease in the proportion of missed and refused doses of pharmacological thromboprophylaxis. Efforts aimed at increasing patient adherence are a promising approach to improve rates of VTE thromboprophylaxis administration.

  • Audit and feedback
  • Quality improvement
  • Qualitative research
  • Hospital medicine

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Venous thromboembolism (VTE), including deep vein thrombosis and pulmonary embolism, is a common and preventable disease.1 ,2 More than half of all VTE events are related to hospitalisation.3 VTE is described as the third most common cause of hospital-related deaths in the USA and the most common preventable cause of hospital death.4 ,5

Pharmacological thromboprophylaxis has proven to be effective, safe and cost-effective.6 ,7 VTE prophylaxis is rated by the Agency for Health Research and Quality as the single most effective patient safety practice for hospitals.8 As a consequence, significant efforts have been directed at hospital and unit-level systems to increase healthcare provider adherence with guideline-recommended pharmacological thromboprophylaxis orders.9 ,10

While improving thromboprophylaxis prescription rates are important, medications that are ordered are not always administered. Two prior studies11 ,12 suggest that the hospital-wide prevalence of non-administration of VTE thromboprophylaxis orders ranges from 5% to 13%, with patient refusal listed as the most common reason for non-administration. Significant variation in non-administration rates exist at the hospital unit level with a larger proportion of missed doses occurring on medicine units.12 Missed doses and patient refusal of VTE prophylaxis constitute a common clinical dilemma not reflected in the majority of clinical trials.13

To date, we are aware of a single study looking at the effects of a hospital-based intervention on patient adherence to recommended thromboprophylaxis regimens.14 In this study Piazza et al used one-on-one patient education sessions led by research pharmacists for all patients prescribed VTE thromboprophylaxis. While effective, the intervention was labour-intensive and required participating pharmacists to spend 1 h on each educational session. This restricts the generalisability of this intervention and made implementation at our institution infeasible. Our study attempted to address the problem of missed and refused VTE prophylaxis doses by using a unit-based intervention that we believed to be sustainable over the long term. Our objective was to assess the effectiveness and sustainability of this unit-based intervention to reduce the proportion of missed and refused doses of VTE thromboprophylaxis.


Developing and implementing the intervention

In late December 2010, a patient was discovered to have developed a VTE in the setting of repeated refusal of prescribed pharmacological thromboprophylaxis. The root cause analysis (RCA) of the event revealed several issues: missed doses of prophylaxis were more frequent than appreciated, patient refusal was by far the most common cause of missed doses, providers were often unaware of missed or refused doses, and the rate of missed doses and patient refusal varied significantly by hospital unit with a few units accounting for the majority of missed doses.

In response, an improvement initiative was started on the six hospital units with the highest refusal rates with the goal of reducing the rate of refused doses of VTE prophylaxis to less than 5%. The main components of this intervention were: (1) a three-step algorithm developed to standardise nurses’ response to patient refusal of pharmacological thromboprophylaxis (see online supplementary table S1 and appendix), (2) the integration of daily assessment of VTE prophylaxis into a multidisciplinary rounds checklist on the three medical units studied (see online supplementary table S2 and appendix), and (3) provision of regular audit and feedback of unit performance. Nursing leadership identified a champion on each floor who then provided a brief in-service to the nursing staff on the standardised response to patient refusal. The hospital's Anticoagulation Management Program reported the rates of missed and refused VTE prophylaxis via email to Nurse Managers and Clinical Nurse Specialists on the six hospital units involved in the intervention on a monthly basis.

When compared with unfractionated heparin (UFH) prophylaxis, low molecular weight heparin (LMWH) was judged to be better tolerated by patients and more likely to lead to higher rates of patient compliance. As a result, nurses and pharmacists were encouraged to advocate for the use of LMWH prophylaxis, particularly in cases of patient refusal.

Setting and study subjects

Our study took place at an 800-bed academic hospital comprised of 30 hospital units. The hospital is part of a healthcare system that includes three hospitals within the same metro area. Six hospital units (three medical and three oncology units) participated in the intervention. In the 12-month period prior to the intervention, these six units accounted for 39% of all missed doses and 54% of all refused doses of VTE prophylaxis in the hospital.

The study cohort included all adult patients admitted between 1 January 2010, and 31 November 2012, to one of the six hospital units participating in the intervention who received orders for one of the following pharmacological VTE prophylaxis regimens: UFH 5000 or UFH 7500 units subcutaneous injection given twice or thrice daily, enoxaparin 40 mg subcutaneous injection given once daily, or dalteparin 5000 units subcutaneous injection given once daily.

The study received expedited approval and a Health Insurance Portability and Accountability Act waiver from the Institutional Review Board.

Data source and measures

We retrieved clinical data from the inpatient electronic health record (Sunrise Clinical Manager; Allscripts, Chicago, Illinois, USA). We obtained data regarding medication administration from the hospital's electronic medication administration record. For each scheduled medication dose, a nurse records whether the dose is administered or not. When doses are late (defined as administered more than 1 h after they are ordered to be given) or missed, the nurse is prompted to enter a free text explanation for deviating from the administration schedule. Using this information, we classified all missed doses according to whether they were missed due to documented patient refusal or not.


The primary outcome is the percentage of missed doses of VTE thromboprophylaxis, defined as the number of ordered doses not administered divided by the total number of ordered doses. The secondary outcome is the percentage of doses refused. In order to ensure consistency of findings irrespective of end points chosen, we calculated a number of supplementary measures: the percentage of patients who refused any dose, the percentage of patients that missed any dose and mean percentage of doses missed per patient.

Two balancing measures15 were calculated to assess for potential negative consequences of the intervention on provider prescribing patterns, as providers could potentially discontinue VTE prophylaxis orders in order to reduce the number of missed or refused doses. We thus calculated the ‘time prophylaxis ordered’ (the percentage of the patient's hospital stay during which they were prescribed pharmacological VTE prophylaxis) and ‘time prophylaxis given’ (the percentage of the patient's hospital stay during which they were administered pharmacological VTE prophylaxis).

In order to strengthen the validity of the results, we included a non-equivalent control group that was not exposed to the intervention in our analysis. For our control group we selected hospital units from the two other hospitals within our healthcare system where the majority of patients were from the Oncology or Hospitalist services. Five hospital units met our criteria, two from Hospital A and three from Hospital B.

Hospital-acquired VTE (HA VTE) rates were obtained for the intervention units. Cases are actively identified and reviewed independently by the hospital's department of Clinical Effectiveness and Quality Improvement (CEQI). The CEQI staff systematically review all discharge International Classification of Diseases Ninth Edition (ICD-9) codes, the reports of all duplex scans of the upper and lower extremities, as well as all contrast enhanced CT scans of the chest. Tests for VTE were generally performed based on clinical signs and symptoms, rather than routine screening. The department of CEQI exerts no influence over decisions about whether or not testing is done.

Sample size estimates performed as part of our prespecified data analysis plan suggested that our study period would be underpowered to detect a difference in HA VTE rates. HA VTE rates were thus considered exploratory end points for our analysis

Statistical analysis

Descriptive analysis

We assessed the baseline clinical and demographic characteristics of the patients in the entire study cohort, the preintervention group, and the postintervention group. We analysed categorical data using the χ2 statistic. For continuous variables whose distributions approximated normality, a t test was used for comparisons. When normality assumptions were not satisfied, the Mann-Whitney U test was used.

Pre-post and interrupted time series analysis

We used pre-post and interrupted time series (ITS) analytical approaches to assess the impact of the intervention on the primary and secondary outcomes. For the pre-post analyses, to determine whether differences in outcome measures between the two time periods were statistically significant, the outcomes were modelled using a dichotomous ‘intervention’ predictor variable in multiple logistic regression models. Regression models to adjust for potential confounders were built using manual backward-stepwise variable selection. Variables that were considered included: age, gender, race, body mass index, marital status, type of VTE prophylaxis ordered (LMWH or UFH), and length of stay. The ‘intervention’ variable was locked in the model a priori, while all other variables were retained at p<0.20 to reduce the risk of overfitting.

In addition, we used an ITS design to evaluate the longitudinal effects of the intervention on the outcomes of interest.16 While pre-post analyses effectively compare average rates before and after the intervention, the ITS approach allows more formal conclusions about the impact of an intervention on the outcome and are able to control for secular trends. We used ITS logistic modelling for the rate of missed and refused doss and adjusted for the variables found to be associated with the outcomes in the pre-post analysis. We used the Huber-White-sandwich method for robust SEs17 ,18 to adjust for patient level clustering.

We analysed HA VTE rates using Poisson regression models. During the last 3 months of the study time period the department of CEQI changed its definition of HA VTE. Analysis for the HA VTE outcome was thus performed with the full study period and repeated using a restricted study period that excluded the last 3 months to provide conservative estimates of change.


Table 1 describes the patient characteristics for the 20 208 admissions during the study period to the hospital units participating in the intervention; 8293 (41%) admissions occurred prior to the intervention and 11 915 (59%) afterwards. There was a significant change in prescribing patterns and median doses of pharmacological thromboprophylaxis. This change was expected as the intervention explicitly encouraged the use of LMWH with its lower frequency of administration.

Table 1

Patient characteristics

In the pre-post analysis (table 2), we adjusted for age, body mass index, length of hospital stay and type of VTE prophylaxis (UFH or LMWH). These four covariates were found after backward elimination to be independently associated with missed doses of VTE prophylaxis. The rate of missed doses of pharmacological thromboprophylaxis decreased significantly after the intervention from 24.7% to 14.7% (p value <0.01). The decrease in missed doses was driven by a decrease in patient refusal from 18.3% to 9.4% (p value <0.01). These findings were consistent irrespective of outcome chosen. The balancing measures calculated did not provide any evidence for a deleterious effect of the intervention on prescribing patterns. The small decline in the time prophylaxis ordered in patients who refused any dose was offset by the rise in the time prophylaxis given.

Table 2

Pre-post analysis

For the ITS models (table 3 and figure 1A), we adjusted for the same four variables noted above. In the intervention cohort, the proportion of missed doses was stable prior to the intervention (adjusted monthly OR 1.01; p value 0.07), decreased significantly with the intervention (adjusted OR 0.64; p value <0.001) and continued to decrease in the months following the intervention (adjusted monthly OR 0.98; p value <0.001). The proportion of refused doses was found to be increasing prior to the intervention (adjusted monthly OR 1.03; p value 0.007), decreased significantly with the intervention (adjusted OR 0.58; p value <0.001) and continued to decrease in the months following the intervention (adjusted monthly OR 0.97; p value <0.001). These results were unchanged when outcomes were stratified by type of VTE prophylaxis ordered (see online supplementary figure S1 and supplementary appendix).

Table 3

Interrupted time series analysis of missed and refused doses

Figure 1

(A) Missed and refused doses over time for intervention and control groups. (B) Venous thromboembolism (VTE) rates in the intervention units over time with fitted values from an interrupted time series (ITS) model.

Results for the control group are presented in table 3 and figure 1A. In the ITS analysis, there was no immediate effect during the intervention time period. In the time period after the intervention, the proportion of refused doses, but not missed doses, was seen to increase in the control group (adjusted OR per month 1.01, 95% CI 1.00 to 1.02; p=0.03).

In the pre-post analysis of HA VTE rates (table 2), there was a significant decrease in the rate of HA VTE in the intervention units. This decrease was not significant when the restricted study period was used. ITS models (table 4 and figure 1B) were developed using the full study period as well as the restricted study period. Neither model revealed an immediate or sustained postintervention change in HA VTE rates.

Table 4

Interrupted time series analysis of VTE rates


We developed a multifaceted quality improvement intervention aimed at increasing patient adherence with clinician-ordered VTE thromboprophylaxis. We focused our efforts on six units with the highest historical rates of non-administration and patient refusal. Average non-administration and patient refusal rates for clinician-ordered injectable VTE thromboprophylaxis on these hospital units prior to the intervention were 25% and 18%, respectively. These rates of non-administration and patient refusal are similar to those seen among low-performing hospital units in a prior study.12 Non-administration rates declined significantly at the time of the intervention and continued to decline in the 21-month follow-up period after the intervention. Average rates of non-administration and patient refusal in the last 3 months of the study period were 11% and 6%, respectively (figure 1). Rates of non-administration and patient refusal in the non-equivalent control group did not decline during the study period, further supporting the effect of the intervention on reducing non-administration rates.16

Many reported improvement efforts have relied on brief study periods. Sustainability is an important component of the external validity of any intervention.19 The 21-month follow-up period of this study strengthens the external validity of our findings and sustainability of the intervention. Features of the intervention that contributed to sustainability were likely the explicit attempt at integrating the intervention within existing nursing workflow, as well as using audit and feedback.20 Both aspects of the intervention are associated with low maintenance costs and facilitated acceptance of the intervention by hospital staff members.

Our study has important limitations. First, we used a multicomponent intervention in this study. As with any such intervention, the complexity of multicomponent interventions may limit generalisability and makes it difficult to assess which parts were essential and which were unnecessary. However, the three principal components of our intervention did not require significant resources, were easily integrated into daily clinical practice and were seen to be sustainable over the long term. Second, the sudden decrease in missed and refused doses of VTE prophylaxis may be partly attributed to the impact of the sentinel event on clinical practice and attitudes towards VTE prophylaxis. The event was a focal point of unit-based council meetings at that time, as well as the focus of a RCA involving many staff members from the unit. As such a significant proportion of the unit staff was aware of the event. However, it is worth noting that the intervention also impacted study outcomes on units other than the one where the sentinel event occurred and was associated with a sustained effect that is less likely to be a result of a single clinical event. Third, our study was limited to a single academic medical centre, and may not be generalisable to all medical centres and patient populations. In particular, we purposefully chose to study units with a high baseline rate of missed and refused doses. Our RCA revealed that rates of refused prophylaxis doses vary dramatically by unit type and that efforts at improving patient adherence may be best served if focused on the few units with the highest rates of refused doses. Fourth, the study period was not long enough to enable us to provide robust estimates of the impact of the intervention on rates of hospital-associated VTE. Given the existing evidence supporting the use of pharmacological prophylaxis,7 and the adoption of administration rates of VTE prophylaxis as a core quality measure,21 we felt that administration rate of clinician-ordered VTE prophylaxis was an appropriate primary outcome and was sufficient to support the beneficial impact of the intervention.

In summary, implementation of a multifaceted intervention aimed at improving patient adherence to VTE thromboprophylaxis orders resulted in an immediate and a sustained decrease in the proportion of missed and refused doses of pharmacological thromboprophylaxis. Patient refusal represents an important barrier to VTE prevention and efforts aimed at improving patient education and adherence can significantly improve administration rates.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.


  • Contributors All authors contributed to this work. CAB participated in study development, analysed the data, and prepared the manuscript. TEHH executed the design of interventions, participated in study development, data analysis and preparation of the manuscript. RCB and JPG participated in study development, analysis activities and preparation of the manuscript. All authors contributed to and approved the final manuscript.

  • Competing interests None declared.

  • Ethics approval University of Pennsylvania IRB.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement The current data set has not been fully deidentified. Prior to any data sharing, the data would be fully deidentified and a deidentified data use agreement would be signed.

Linked Articles