Introduction: D-dimer tests were inappropriately overused in our emergency department as a result of bloods being taken before clinical assessment to help meet the “4-hour target”. We introduced a multifaceted intervention to reduce the number of inappropriate D-dimer tests. The secondary aim was to improve the diagnostic workup of suspected pulmonary embolism (PE).
Method: Rate of D-dimer test and ventilation/perfusion scan requests were compared before, during and after a staggered intervention at two hospitals in one National Health Service Trust. Audits before and after the intervention were done to determine whether test use was appropriate and whether the diagnostic workup was complete.
Results: At hospital 1, D-dimer testing after the intervention was almost halved: ratio 0.59 (95% CI 0.55 to 0.63) (p<0.0001). There was also a small reduction at hospital 2 (control): rate 0.88 (95% CI 0.78 to 0.99) (p = 0.03). After the formal introduction of change at hospital 2, there was a further reduction in tests: ratio 0.67 (95% CI 0.58 to 0.76) (p<0.0001). In hospital 1, pretest probability assessment improved by 42% (p = 0.0004) and D-dimer test use was reduced by 12.5% (p = 0.04) between audits. Improvement in the use of D-dimer test according to the pathway was not significant (32.5%, p = 0.11), and there was no change in the proportion of patients with completion of their diagnostic workup for PE: 47.6% (95% CI 38.3% to 56%) before and 45.6% (95% CI 38.3% to 53.1%) after the intervention.
Conclusion: Implementation of a multifaceted change program reduced the number of D-dimer test requests in both hospitals and may have improved the diagnostic workup for PE at hospital 1. Processes that speed patient transit through the emergency department may impact negatively on other aspects of patient care. This should be the subject of further studies.
Statistics from Altmetric.com
D-dimer testing has been suggested as a “screening” test for pulmonary embolism (PE) without clinical assessment.1 D-dimer is produced when fibrin clots are broken down and are sensitive for detecting blood clots at standard cut-off values.2 3 However, fibrin breakdown products may be present in the blood for many reasons other than clots (trauma, infections, inflammation, cancer, pregnancy and increased age4 5 6 7). Thus, D-dimer tests have low specificity. Highly sensitive tests help rule out a diagnosis if the results are negative, but low-specificity tests do not help rule in a diagnosis if the results are positive (in other words, the likelihood of PE is not increased significantly by a positive D-dimer test result). Unfortunately, many clinicians do not appreciate this distinction.8 Once faced with a positive test result, clinicians may find it difficult to ignore the result, especially if an important and potentially fatal diagnosis such as PE is thought to be suggested by the result.9 10
Initial investigations for PE in the emergency department (ED) should be done after determining the pretest probability (PTP) of PE using either gestalt or a formal score.11 The score determines which investigations are done next.12 This approach has been prospectively validated 13 14 15 16 and has shown to be cost-effective and to reduce length of stay in hospital.17 18 19 20
However, performing blood tests in the ED leads to increased lengths of stay in the order of 2 h.21 To help speed up patient flow to meet the 4-hour h target in our ED, we had previously developed a system of blood tests performed by healthcare assistants and nurses prior to medical assessment of patients. Initially, there was no restriction on which tests were ordered. It became apparent that patients with clear alternative diagnoses were being admitted for ventilation/perfusion (VQ) scanning, on the basis of a positive D-dimer test result interpreted by the admitting senior house officer (SHO). Many of these patients had a clear alternative diagnosis on consultant review (often the day after admission). Initial investigations revealed that >600 D-dimer tests per month were being requested, representing almost one in four non-trauma presentations to the ED.
Furthermore, the cut-off point for D-dimer regarded as positive in use had been raised by our laboratory to 0.6 µg/ml in an attempt to reduce the number of false-positive test results (manufacturer’s recommendation, 0.5 µg/ml). It was also apparent that VQ scans were being reported in a unique way (the probability of PE indicated by the VQ scan was not stated on the report), and the result was being interpreted as either “positive” or “negative” by clinicians.
In the hope of improving the use of diagnostic tests for PE, we implemented a change management program in November 2005. This was a two-phase process involving separate hospitals within the same National Health Service (NHS) Trust, which involved introduction of an evidence-based pathway for PE diagnosis (fig 1).
The primary aim was to reduce the number of inappropriate D-dimer tests requested from the ED. The secondary aim was to improve the workup for patients with suspected PE.
The primary outcome was the number of D-dimer tests requested. Secondary outcomes were the number of VQ tests requested, the proportion of patients with PTP calculated, appropriate use of D-dimer testing, appropriate reporting of VQ scans, completion of the pathway for PE diagnosis and a survey of staff awareness and understanding of the pathway.
This project was approved as a clinical audit by the local regional ethics committee on the condition that the hospitals were not to be identified in the published report.
Laboratory and Nuclear Medicine databases were queried to extract the numbers of tests done from September 2005 to May 2006. Rates of D-dimer and VQ tests were analysed. Electronic medical records and VQ scan request forms were reviewed to determine follow-up investigations and outcomes. The first audit of investigation of PE in the ED at hospital 1 was conducted from July to December 2004. This audit was undertaken by the Audit and Effectiveness Department on behalf of one of the authors (BE). A random sample of patients undergoing VQ testing over the 6-month period was analysed (audit 1, baseline). The second audit corresponded to phase 2 of the intervention (conducted from March to May 2006 by PJ and BE). All VQ scan requests in this time period were analysed (audit 2, postintervention in hospital 1).
Appropriate D-dimer test use: PTP calculated and D-dimer test requested only when PTP is low (audit 1) or unlikely (audit 2)12
Appropriate VQ scan reporting: standard reporting (PIOPED II) nomenclature22
Completion of pathway: PE confirmed with positive imaging test: high-probability VQ; positive computerised tomography pulmonary angiography or PE ruled out (post-test probability, <2%): PTP low (audit 1) or unlikely (audit 2) with D-dimer test result negative, low probability VQ with either PTP low or D-dimer test result negative and/or normal VQ scan or computerised tomography pulmonary angiograph result.
Periods used for interventions and investigation request rate calculations:
Baseline: before intervention—September to November 2005
Phase 1: intervention at hospital 1/hospital 2 as control—December 2005 to February 2006
Phase 2: intervention at hospital 2—March to May 2006
Hospital 1 has an annual ED census of ≈80 000, with ≈25 000 non-trauma presentations. There were 12 ED consultants, six staff grades and 22 SHOs. Hospital 2 is located 20 min from hospital 1 and has an annual ED census of ≈30 000. ED consultant staff rotate between the hospitals, providing 4 hours of cover per day at hospital 2. Otherwise, five staff grades cover ED at this hospital. All patients potentially requiring VQ scans at hospital 2 are admitted by an SHO under a medical assessment unit consultant. SHOs rotate between the hospitals. All VQ scans for the Trust are done at hospital 1. There is a separate laboratory at hospital 2, with oversight from the laboratory at hospital 1.
Change management process
A detailed description of the process can be found in Appendix 1.
Graphs were used to display the numbers of tests done by month. Daily rates of testing and 95% confidence intervals (CIs) were calculated using Openepi (http://www.openepi.com). Categorical variables were compared using Fisher exact test, and CIs for proportions were calculated using (Graphpad Software, La Jolla, California, USA). Statistical significance for differences between groups was set at p<0.05 (two-tailed).
There was a reduction in the number of D-dimer test requests at hospital 1 after the institution of the change process (fig 2). At baseline, daily rate of testing was 20.5 (95% CI 19.6 to 21.5). After the intervention, the rate was 12.1 (95% CI 11.4 to 12.8), a ratio of 0.59 (95% CI 0.55 to 0.63) (p<0.0001). During the same periods, at hospital 2, the rates were 6.3 (95% CI 5.9 to 6.9) and 5.5 (95% CI 5.1 to 6.1), respectively, with a ratio of 0.88 (95% CI 0.78 to 0.99) (p = 0.03). During phase 2 of the intervention, there was no further change in the rate at hospital 1: rate was 11.6 per day, a ratio of 0.96 (95% CI 0.88 to 1.05) (p = 0.3); however, there was a further and more marked drop in the rate at hospital 2 (during intervention): rate 3.7 (95% CI 3.3 to 4.1), a ratio of 0.67 (95% CI 0.58 to 0.76) (p<0.0001), compared with that before the intervention.
There was also a reduction in the number of VQ scan requests at hospital 1 after the institution of the change process (fig 3). At hospital 1, the baseline daily rate of testing was 3.2 (95% CI 2.8 to 3.6). After phase 1 of the intervention, the rate was 2.1 (95% CI 1.8 to 2.4), a ratio of 0.7 (95% CI 0.6 to 0.8) (p<0.0001). During the same periods, at hospital 2, the rates per 10 days were 5.6 (95% CI 4.2 to 7.4) and 5.8 (95% CI 4.3 to 7.6), respectively, a ratio of 1.0 (95% CI 0.7 to 1.5) (p = 0.87). During phase 2 of the intervention, there was no further change in the rate at hospital 1: rate was 2.3 (95% CI 2.0 to 2.7) per day (p = 0.33). There was also no change in the rate at hospital 2: rate per 10 days was 6 (95% CI 4.5 to 7.8), a ratio of 1.04 (95% CI 0.71 to 1.5) (p = 0.86), compared with that before the intervention.
Table 1 shows the results of the two audits of VQ scan requests. PTP use improved by 42% (p = 0.0004) and D-dimer test use was reduced by 12.5% (p = 0.04) between audits. There was no statistically significant difference in the use of D-dimer test according to the pathway (32.5% improvement, p = 0.11). All VQ scans were reported in a standard manner after the intervention compared with none before. There was no difference in the number of patients with a definitive diagnosis (PE ruled either in or out) after the intervention.
The last century has seen the development of many investigations to help clinicians, the “science” of diagnosis.23 Investigations should come after interpretation of clinical findings, to complement the “art” of diagnosis.24 25
With the ease of access to many tests and busy clinical workloads, this traditional model of patient assessment is not always applied. Recently, there are concerns about the lack of thought that goes into diagnostic testing in the NHS.26 27 Such concerns about the resource implications of inappropriately used tests have been expressed for >30 years,28 and attention has focused on clinician’s traits as a cause of differential test ordering.29 30
In the modern NHS, other factors have come into play. As a response to unacceptable delays in assessing patients, a target of >97% of patients leaving the ED within 4 h of arrival was set.31 Financial incentives drive this,32 and work practices have been changed to speed up patient flow. One change that was made in our hospital is that patients have blood tests on arrival, so that results are available sooner. The suggested benefit is to reduce short-stay admissions wherein patients wait for test results.33 Great caution needs to be used with this approach. If tests irrelevant to the patient’s presenting problem are done, an abnormal test result may allow patients to be labelled with a potential (but incorrect) diagnosis, which may inadvertently “facilitate” moving them to a ward for further assessment within the target time. There is some evidence that investigations in the ED have increased since the 4-hour target was introduced.34 We believed that, in our department, patients with false-positive D-dimer test results were admitted to have further investigations, delaying the correct diagnosis and leading to inappropriate treatment.
Our intervention appeared to reduce the number of D-dimer and VQ tests requested and increase the use of a formal PTP score at hospital 1. However, a small improvement in how the D-dimer test was used did not reach statistical significance, and there was no improvement in the proportion of patients with a definite diagnosis after the intervention.
A systematic search of major medical databases (Cochrane, Medline and Embase) found three previous studies of PE pathways, all of which found low compliance, in the order of 40%–60%. One of these was a baseline assessment,35 another was the rate during a prospective clinical trial36 and the last was a postintervention rate of compliance.37 No study has explored the reasons for low compliance with the pathways. A qualitative research methodology would be required to do this.
The “study” and “control” hospitals were dissimilar in size and levels of staffing. However, the culture of inappropriate D-dimer testing and interpretation was similar between them.
Because of resource constraints, we were limited in the design of this study to an interrupted time series to assess the primary outcome and to a before-and-after method to assess secondary outcomes. These are accepted methods to assess change management programs in healthcare settings.38 Ideally, the important patient-centred outcomes (number of patients with PE treated appropriately and number without PE not treated with anticoagulants) would have been assessed prospectively with formal follow-up in a blinded manner.39 Using number of test requests as the primary outcome measure and inferring appropriateness of diagnostic workup by electronic records review is a less robust but practical choice. It is possible that as a result of the change process, some patients who should have been investigated for PE were not, and we were unable to determine this.
There was also a difference in the methods of sampling between the two audits, the first of which was a random sample of VQ scan requests during summer and autumn months, the second was a consecutive sample of VQ scan requests during winter and spring. Ideally, similar months across several years would have been compared, but because of changes in record keeping, changes to service provision across the Trust and the time pressure to introduce change, this was not possible. The seasonal increase in PE investigation and diagnosis in winter months would be likely to bias towards more use of D-dimer and VQ tests in audit 2 compared with audit 1.40 In an attempt to confirm that the changes observed were not caused by a Trust-wide improvement in PE workup not related to the change process, we used hospital 2 as a control site. Our control site was not completely masked from the study site because of staff movement between sites and shared intranet and imaging resources. Because of the pressure to introduce the change process in hospital 2 after apparent success in hospital 1 (partly driven by estimated cost savings of £2600 per month in D-dimer test and £4200 per month in VQ scan requests), the second audit occurred during the introduction of the process in hospital 2. Both of these factors would be likely to bias results against showing a difference in favour of the change process between hospitals.
Implementation of a multifaceted change program reduced the number of D-dimer tests in both hospitals studied and may have improved the diagnostic workup for PE. Processes that speed patient transit through the ED may impact negatively on other aspects of patient care. This should be the subject of further studies.
This project formed the basis for a postgraduate certificate in evidence-based healthcare at Oxford University, Department of Continuing Education. Statistical advice was provided by Greg Gamble, MSc. Lucy Roche and Mahima contributed to data collection for audit 2.
Funding The University Hospitals of Coventry and Warwickshire NHS Trust provided part financial support to PJ to undertake the postgraduate certificate in evidence-based healthcare that this article was based on.
Competing interests None.
Provenance and Peer review Not commissioned; externally peer reviewed.
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.