a b s t r a c t

Background: Several previous studies have investigated the clinical utility of age-adjusted D-dimer cutoffs for diagnosing pulmonary embolism (PE).

Objectives: We performed a pre/post implementation study, using data from a mid-Atlantic healthcare system comprising 6 hospitals and 400,000 ED visits to determine whether implementing age adjusted D-dimer cutoffs reduced the number of imaging tests performed.

Methods: Retrospective study of all patients who had a D-dimer performed during ED visits between September 2015 to September 2018. On March 21, 2017, the D-dimer Upper limit of normal system-wide was increased for patients over 50 to: Age (years) x 0.01ug/mL. D-dimer results were displayed as normal or high based on auto- mated age adjustment. EHR Chart review was performed 1.5 years prior to implementation of age-adjusted D-dimer cutoffs, as well as 1.5 years after to evaluate mortality and test accuracy characteristics such as False negative rates. Comparisons were made using chi-square testing.

Results: 22,302 D-dimers were performed pre-implementation of which 10,837 (48.6%) were positive resulting in 7218 (32.3%) imaging studies. After implementation of age-adjusted d-dimer, 25,082 were performed of which 10,851 (43.2%) were positive resulting in 7017 (28.0%) imaging studies. (pre: 48.6%, post: 43.2%; p < 0.01). A significantly lower proportion of patients had a positive d-dimer (pre: 48.6%, post: 43.2%; p < 0.01) and underwent imaging post-implementation (pre: 32.3%, post: 28.0%; p < 0.05) a relative risk reduc- tion of 13.3. This absolute risk reduction of 4.4% is associated with 1104 less scans in the post-implementation group while still increasing test accuracy from 53.7% to 59.2% (p < 0.05).

Conclusion: Implementation of an automated age-adjusted D-dimer positive reference value reduced CT and V/Q imaging in this population by 4.4% while increasing test accuracy in a regional, heterogeneous six-hospital system.

(C) 2020 Published by Elsevier Inc.

1. Introduction

Pulmonary embolism (PE) is a potentially life-threatening condition with a relatively rare incidence (69 per 100,000 persons annually) whose diagnosis proves challenging given the associated non-specific complaints of dyspnea and chest pain [1]. The diagnosis of PE requires diagnostic imaging, most commonly computerized tomographic pul- monary angiography (CTPA) or ventilation perfusion (V/Q) scanning which exposes patients to ionizing radiation and IV contrast. Clinical de- cision tools to reduce the reliance on diagnostic imaging have been de- veloped such as the pulmonary embolism rule-out criteria [PERC], Wells criteria and Revised Geneva Score (RGS) [2]. To decrease diagnos- tic uncertainty and reduce unnecessary imaging D-dimer has been used

* Corresponding author.

E-mail address: [email protected] (M. Wilson).

to help rule-out pulmonary emboli. D-dimer has shown very high sensitivity for the presence of PE (sensitivity 80-100%) but a low spec- ificity (specificity 23-63%). This makes D-dimer relatively useful in ruling out PE in intermediate and low risk groups, but the test retains a high false positive rate. The specificity of D-dimer decreases with age, thus age-adjusted cutoffs may increase the specificity of D-dimer in this population and reduce unnecessary diagnostic testing [4].

Several previous studies on age adjusted D-dimer levels have been performed which have generally found an increased specificity with varying degrees of sensitivity when applied retrospectively [5,6]. Prospectively, large multi-center studies have found increased specific- ity without a reduction in sensitivity. The multi-center European ADJUST-PE study was a prospective study that utilized age adjusted D-dimer to rule out PE and followed patients for 3 months after the study period. They found no increase in false-negatives and were able to exclude PE on the basis of D-dimer alone in 23% more patients

https://doi.org/10.1016/j.ajem.2020.10.067 0735-6757/(C) 2020 Published by Elsevier Inc.

when utilizing age adjusted cut-offs among patients 75 or older [7]. A large systematic review of 12,497 patients with non-high probability of PE were examined with age adjusted cutoff and found to have in- creased specificities over all age categories (above 50 in 10 year inter- vals) with sensitivity above 97% in all categories [8].

• 1. Goals of this investigation

The objective of this study was to see if implementation of auto- matic electronic health record (EHR) age-adjustment of D-dimer cut- offs led to a change in the proportion of imaging studies ordered at a regional medical system in the mid-Atlantic comprising six hospitals and 400,000 annual ED visits. Additionally, the study aimed to evalu- ate the safety of real-world implementation of age-adjusted D-dimer levels by screening for PEs identified in the ED that were missed using the new cutoffs, as well as PE related 30 day mortality for the index hospitalization.

1. Methods
   1. Study design

We performed a retrospective cohort study utilizing data from six hospital EDs in a mid-Atlantic health system encompassing both large urban academic centers and small community hospitals. The course of the study period was from September 28, 2015 to September 11, 2018. The positive reference value D-dimer cutoff of 0.5 ug/mL was adjusted in the electronic health record (EHR) (Cerner First Net; Kan- sas City, MO) to an automated age adjusted cutoff using the formula: Age (years) x 0.01 ug/mL as of March 21, 2017. Following implemen- tation, D-dimer results flagged as abnormal based only on age- adjusted cutoffs, not traditional D-dimer cutoffs (Fig. 1). Data from

1.5 years prior to the implementation was matched to data for the

1.5 years following.

Clinicians ordered D-dimers based on their own practice to screen for either PE or DVT. Of note, the EHR contains decision support tools prompting physicians and APPs to first risk stratify for pulmonary em- bolism using the Modified Geneva Score and Pulmonary Embolism Rule-out Criteria or use Well’s Score for suspected DVT with rec- ommendations for D-dimer testing prior to imaging as indicated by risk level. These were in place one year before the start of this study to en- courage a standardized evidence based approach to risk assessment. Serum D-dimer testing was performed using ELISA D-dimer (STA Liatest D-di) assay with a manufacturer reported detection limit of 0.27 ug/mL. The primary outcome was the proportion of patients who underwent imaging (CTPA or V/Q scanning) as a follow up to D-dimer testing. Sec- ondary outcome measures include diagnostic accuracy, missed PE dur- ing the index ED-visit as identified on chart review of radiology records, and PE related 30 day mortality for the index hospitalization. Relevant data were extracted from the systemwide EHR into Microsoft Excel (Redmond, WA) for analysis. A diagnosis of “acute pulmonary em- bolism” in the absence of a “history of pulmonary embolism” in dis- charge Diagnostic coding was identified via language processing to determine Diagnostic test accuracy. Data analysis was performed in Stata 15 (College Station, TX) with chi-square testing for significance.

Fig. 1. Adjusted EHR display so that 73 year old patient with D-Dimer of 0.72 micrograms/ mL does not flag as abnormal.

1. Results

There were a total of 22,302 D-dimer tests performed in the pre- implementation group and 25,082 D-dimer tests post-implementation as presented in Table 1. Of the D-dimer tests performed pre implemen- tation, 10,837 (48.6%) were positive using the standard cutoff which re- sulted in 7218 imaging studies (7218/22302 = 32.4%).

Post-implementation of the age-adjusted positive reference value, 25,082 D-dimer tests were performed, of which, 10,851 (43.2%) were positive, which resulted in 7017 scans (7017/25082 = 28.0%). There was a significant reduction in the proportion of D-dimer assays reported as positive (5.4%; 48.6% vs 43.2%, P < 0.05) and a significant reduction in the proportion of patients receiving imaging post- implementation (4.4%%; 32.3% vs 28.0%, P < 0.05). This absolute risk reduction of 4.4% translates into 1104 fewer imaging studies per- formed post-implementation.

Among the 11,465 patients with a negative D-dimer pre- implementation, 600 (5.2%) imaging studies were performed. Of the 14,231 encounters with a negative d-dimer post-implementation, 602 (4.2%) imaging studies were performed. Among the entire 1202 en- counters with negative d-dimers pre- and post-implementation who had imaging performed, four were positive: three of the patients pre- implementation (one intermediate probability V/Q scan, one subsegmental PE on and one lobar PE on CTPA studies) and one post- implementation (subsegmental PE on CTPA).

All in all 1354 D-dimers were “newly negative” under the age- adjusted cutoff that would have been positive prior to age-adjustment. Of these, 97 (7.2%) underwent further imaging to rule out PE and only the one patient previously mentioned as a false negative (age of 65 with a d-dimer result of 0.55ug/mL) was diagnosed with PE (subsegmental PE on CTPA). Of the 1354 individuals with newly nega- tive d-dimers, 14 died, none with suspected pulmonary embolism. Di- agnostic test characteristics were similar for D-dimer testing pre/post intervention with a notable increase in diagnostic accuracy. Pre- intervention D-dimer testing sensitivity was 96.7% with a specificity of 52.6% and diagnostic accuracy of 53.7%. Post-intervention the test’s sen- sitivity remained at 96.1% with increased specificity of 58.2% (p < 0.05) and a significantly increased accuracy of 59.2% (p < 0.05). The number of patients diagnosed with pulmonary embolism after a false negative D-dimer test was similar in both cohorts (18 pre-intervention and 26 post-intervention).

1. Discussion

We found that EHR based automated age-adjustment of D-dimer reference values reduced CTPA imaging system wide while still main- taining acceptable safety standards with no lobar or larger PEs missed, as well as no deaths within 30 days attributed to PE in the newly- negative cohort. The absolute risk reduction in CT imaging was

Table 1

Comparison of pre and post intervention patient populations.

	Pre-intervention	Post-Intervention	p-value (as appropriate)
d-Dimer resulted	22,302	25,082
Positive d-dimer	10,837	10,851
Percentage of positive d-dimer tests Imaging performed	48.6% 7218	43.2% 7017	<0.001
Percentage of total patients with d-dimer testing imaged Median age(years)	32.4% 55	28.0% 54	<0.001 <0.001
Percentage discharged	46.5%	47.0%	0.070
Mean d-dimer level (mcg/mL)	1.3	1.3	0.49

significant but on the lower range of previous studies (range 5-18%) which we suspect may be related to increased d-dimer ordering, partic- ularly in the younger cohort, relative to European studies [9,10]. While our percentage of CT scans reduced before and after implementation may be lower than other studies, it still has a large impact as our large sample size is from a high volume healthcare system with 400,000 annual visits reflecting a great reduction in radiation for patients.

Our study did not specifically address the impact of age-adjusted

D-dimers on ED throughput or cost reduction. However, it can be reasonably extrapolated that reducing the need for imaging studies helps to decrease length of stay for many patients. This could in turn help to improve overall ED throughput, which is very relevant to the high volume centers where the study was performed. Additionally, it can be inferred that fewer imaging tests would help to reduce costs for patients and for hospitals as well.

Age-adjusting the D-dimer results did not adversely affect our diag- nostic test characteristics with improved specificity and accuracy at no significant cost of sensitivity. Our results are consistent with previous studies reported sensitivities of 80-100% [2,4,7,8,10,11], and previously reported specificities of 23-63% [4]. Also, Implementing age-adjusted reference values was well received and didn’t result in a behavior change towards over-ordering as 5.2% pre and 4.2% post received imaging despite negative d-dimers.

We had four patients who had negative age-adjusted D-dimer who received CT angiography despite their negative D-dimer who ended up having pulmonary emboli noted. These four patients had unremark- able Hospital courses and the PEs found were submassive. No significant PE-related mortality occurred in the newly negative age-adjusted cohort.

• 1. Limitations

As a retrospective observational study our results are subject to lim- itations in our ability to match and compare these cohorts over time. This is somewhat compensated for by the large size of our multicenter evaluation. There was an observed difference between the number of D-dimer tests performed pre and post-implementation, which may be a result of evolving practice improvement and ordering practices. There was also a large number of positive D-dimer tests pre and post implementation that did not have follow up imaging occur within the index ED visit which we attribute to triage clinician ordering of D-dimers, D-dimer/ultrasound testing for DVT, and imaging/D-dimer testing deferred to or ordered for the inpatient setting.

EHR data extraction was challenged by site-specific terminology for CT order language. Although the natural language processor searches for CT orders with variations from standard CTPA orders, additional cases were found and added to the studied cohorts over the course of the investigation to be comprehensive. This also limits the calculations of diagnostic accuracy as chart review of the entire cohort was not pos- sible and we depended on discharge diagnosis terminology equally for final diagnostics in both pre and post intervention cohorts.

Safety endpoints are limited by assumptions required for targeted

chart review of the EHR. The imaging reviewed by the chart search algo- rithm was limited to scans that occurred during the initial ED visit. 30 day mortality follow up was limited to the index visit at our

institution and thus limited/confined to presentations within our large regional health system.

1. Conclusion

Overall, we found that age adjusted D-dimers resulted in a signifi- cant systemwide decrease of PE imaging of 4.4% without adversely af- fecting our safety endpoints. We believe that automated age-adjusted D-dimer can be implemented safely in other systems to reduce imaging usage.

CRediT authorship contribution statement

Jeffrey Dubin: Conceptualization, Methodology, Validation, Re- sources. Mary Kathleen Ratay: Validation, Investigation. Matt Wilson: Validation, Investigation. Peter Davis-Allen: Validation, Investigation. Michael Gillam: Conceptualization, Methodology, Software. Joseph Izzo: Conceptualization, Methodology, Software. Kevin Maloy: Concep- tualization, Methodology, Software. Jonathan Davis: Conceptualization, Methodology, Validation, Resources. Munish Goyal: Conceptualization, Investigation, Methodology, Validation, Resources, Supervision.

Declaration of Competing Interest

None.

References

1. Silverstein MD, Heit JA, Mohr DN, Petterson TM, O’Fallon WM, Melton LJ. Trends in the incidence of deep vein thrombosis and pulmonary embolism: a 25-year population-based study. Arch Intern Med. 1998 Mar 23;158(6):585-93.
2. Wolf SJ, Hahn SA, Nentwich LM, Raja AS, Silvers SM, Brown MD. Clinical policy: crit- ical issues in the evaluation and management of adult patients presenting to the emergency department with suspected acute Venous thromboembolic disease. Ann Emerg Med. 2018 May 1;71(5):e59-109.
3. Crawford F, Andras A, Welch K, Sheares K, Keeling D, Chappell FM. D-dimer test for excluding the diagnosis of pulmonary embolism. Cochrane Database Syst Rev. 2016;8.
4. Sharp Adam L, et al. An age-adjusted d-dimer threshold for emergency department patients with suspected pulmonary embolus: accuracy and clinical implications. Ann Emerg Med. 2016;67(2):249-57. https://doi.org/10.1016/j.annemergmed. 2015.07.026.
5. Sheele Johnathan M, et al. A retrospective evaluation of the age-adjusted d-dimer versus the conventional d-dimer for pulmonary embolism. Blood Coagul Fibrinoly- sis. 2018:1. https://doi.org/10.1097/mbc.0000000000000707.
6. Righini M, Van Es J, Den Exter PL, Roy PM, Verschuren F, Ghuysen A, et al. Age- adjusted D-dimer cutoff levels to rule out pulmonary embolism: the ADJUST-PE study. Jama. 2014 Mar 19;311(11):1117-24.
7. Schouten HJ, Geersing GJ, Koek HL, Zuithoff NP, Janssen KJ, Douma RA, et al. Diagnos- tic accuracy of conventional or age adjusted D-dimer cut-off values in older patients with suspected venous thromboembolism: systematic review and meta-analysis. Bmj. 2013 May 3;346:f2492.
8. van Es N, van der Hulle T, van Es J, den Exter PL, Douma RA, Goekoop RJ, et al. Wells rule and D-dimer testing to rule out pulmonary embolism: a systematic review and individual-patient data meta-analysis. Ann Intern Med. 2016 Aug 16;165(4): 253-61.
9. Gupta A, Raja AS, Ip IK, Khorasani R. Assessing 2 D-dimer age-adjustment strategies to optimize computed tomographic use in ED evaluation of pulmonary embolism. Am J Emerg Med. 2014 Dec 1;32(12):1499-502.
10. Crichlow A, Cuker A, Mills AM. Overuse of computed tomography pulmonary angi- ography in the evaluation of patients with suspected pulmonary embolism in the emergency department. Acad Emerg Med. 2012 Nov;19(11):1219-26.