Non-cardiac chest pain patients in the emergency department: Do physicians have a plan how to diagnose and treat them? A retrospective study

Background Non-cardiac chest pain is common and there is no formal recommendation on what diagnostic tests to use to identify underlying diseases after an acute coronary syndrome has been ruled out. Objective To evaluate the diagnostic tests, treatment recommendations and initiated treatments in patients presenting with non-cardiac chest pain to the emergency department (ED). Methods Single-center, retrospective medical chart review of patients presenting to the ED. Included were all medical records of patients aged 18 years and older presenting to the ED with chest pain and a non-cardiac discharge diagnosis between January 1, 2009 and December 31, 2011. Information on the diagnosis, diagnostic tests performed, treatment initiated and recommendation for further diagnostic testing or treatment were extracted. The primary outcomes of interest were the final diagnosis, diagnostic tests, and treatment recommendations. A formal ACS rule out testing was defined as serial three troponin testing. Results In total, 1341 ED admissions for non-cardiac chest pain (4.2% of all ED admissions) were analyzed. Non-specific chest pain remained the discharge diagnosis in 44.7% (n = 599). Identified underlying diseases included musculoskeletal chest pain (n = 602, 44.9%), pulmonary (n = 30, 2.2%), GI-tract (n = 35, 2.6%), or psychiatric diseases (n = 75, 5.6%). In 81.4% at least one troponin test and in 89% one ECG were performed. A formal ACS rule out troponin testing was performed in 9.2% (GI-tract disease 14.3%, non-specific chest pain 14.0%, pulmonary disease 10.0%, musculoskeletal chest pain 4.7%, and psychiatric disease 4.0%). Most frequently analgesics were prescribed (51%). A diagnostic test with proton pump inhibitor (PPI) was prescribed in 20% (mainly in gastrointestinal diseases). At discharge, over 72 different recommendations were given, ranging from no further measures to extensive cardiac evaluation. Conclusion In this retrospective study, a formal work-up to rule out ACS was found in a minority of patients presenting to the ED with chest pain of non-cardiac origin. A wide variation in diagnostic processes and treatment recommendations reflect the uncertainty of clinicians on how to approach patients after a cardiac cause was considered unlikely. Panic and anxiety disorders were rarely considered and a useful PPI treatment trial to diagnose gastroesophageal reflux disease was infrequently recommended.


Introduction
The top priority in patients presenting with chest pain to the emergency department (ED) is to rule out a potentially life-threatening disease such as an acute coronary syndrome (ACS), pulmonary embolism, aortic dissection, or pneumonia. After a thorough diagnostic work-up, an acute myocardial ischemia can be ruled out for 60% to 90% of patients presenting with chest pain [1][2][3][4]. While in specialized units, including cardiac care units and intensive care units, the proportion of patients with ACS may be higher [5], the percentage of patients in the ED with ACS decreased in the US from 23.6% in 1999-2000 to 13.0% in 2007-2008 [6]. When no specific disease causing the chest pain can be identified, patients are usually discharged with the diagnosis of non-cardiac chest pain (NCCP).
Patients with NCCP can be categorized in patients with and without an identifiable underlying disease (i.e. non-specific chest pain). It has been suggested that up to 50% of the patients discharged with NCCP have an underlying gastrointestinal reflux [7] or a psychiatric disease [8,9]. Further, chest pain is frequently the result of musculoskeletal diseases [10]. Whereas the mortality rates among patients discharged with NCCP from the ED is low [11], 90% complained of persisting symptoms and impaired quality of life at a 4-year follow-up [12]. Despite normal coronary angiograms, 44% of patients with NCCP still believed they suffer from an underlying cardiac disease and 50% reported limitations in performing their daily activities [13]. Therefore, a primary focus on ruling out cardiovascular disease in patients with NCCP may result in overtesting without improving the patients' confidence. Further, elevated troponin test results can be found in patients without chest pain or ischemic electrocardiographic changes and, in a retrospective study, elevated troponin test results had no clinical utility but resulted in downstream testing [14].
Therefore, the clinical challenge is to determine which diagnostic tests to apply in patients with chest pain after a cardiac disease has been ruled out in order to discriminate between patients with non-specific chest pain and other underlying diseases presenting with NCCP. For example, a high dose proton pump inhibitor (PPI) treatment trial may be useful to identify patients with underlying gastroesophageal reflux disease (GERD) and screening tools may identify patients with an underlying panic or anxiety disorder [15]. To date, the diagnostic processes and the treatment recommendations in patients discharged from the ED with a diagnosis of NCCP are poorly investigated and mainly based on the physicians' personal beliefs and experiences.
The objective of this retrospective study was to gather knowledge about the diagnostic steps in the ED and the treatment recommendations for patients discharged from the ED with a diagnosis of NCCP. We analyzed the frequency of discharge diagnoses, the performed diagnostic tests, the initiated treatments, and the treatment recommendations. We hypothesized that the majority of patients were discharged with a musculoskeletal or non-specific disease and the diagnostic assessment focused mainly on ruling out of an ACS. Further, we hypothesized that a high-dose PPI treatment trial to identify patients with GERD related chest pain was infrequently used and psychiatric diseases rarely considered.

Methods
Single-center, retrospective medical chart review of patients presenting to one of the ten largest hospitals in Switzerland, the Cantonal Hospital Winterthur, between January 1, 2009 and December 31, 2011. The study period was chosen because an outpatient clinic opened in 2012 and therefore, many patients eligible for this study were potentially treated elsewhere. The hospital is affiliated to the University of Zurich and covers the medical services for approximately 200'000 persons (15 percent of the inhabitants of the canton Zürich).

Eligibility criteria
Included were all medical records of patients age 18 years and older presenting to the ED with chest pain and a non-cardiac discharge diagnosis between January 1, 2009 and December 31, 2011. Excluded were patients with chest pain of cardiovascular origin, pregnant women, trauma patients or life-threatening conditions, malignant disease, current fracture, renal replacement therapy or severe kidney failure (creatinine clearance of less than 30ml/min/ 1.73m 2 ) as well as patients with incapacitation or records of patients which opted out of releasing their records for scientific purposes.

Data extraction procedure
All records identified by this search were screened by two researchers (TD, SM) for inclusion or exclusion. In case of uncertainty, the records were discussed with the principal investigator (MW) and disagreement was resolved within the research group. Each patient included in the study was assigned a unique de-identified number. We defined the first presentation for chest pain to the ED as the index consultation for the first episode. During the following three months each presentation (to the ED, outpatient consultation, hospitalization) was considered potentially related to the index consultation and was defined as a follow-up consultation. Presentations to the ED or hospitalizations after more than three months due to chest pain were defined as a new index visit of a second episode.
Variables of interest were predefined and the extraction form was pilot-tested in 20 records.
To ensure high quality in the data extraction, TD and SM were trained and monitored by MW and an extraction manual was used. We extracted information on general characteristics (age, gender), cardiovascular risk factors, signs and symptoms at presentation, preexisting comorbidities, medications, clinical findings at presentation, blood analyses, ECG, imaging studies, coronary angiography, non-invasive testing (e.g. treadmill testing, cardiac scintigraphy, echocardiography) and other tests/investigations. Further information on discharge medications, discharge diagnosis, recommended procedures / investigations after discharge were extracted.

Study endpoint
The main study endpoints of interests were the final diagnosis, performed diagnostic tests, and treatment recommendations. The final diagnosis was based on the discharge diagnosis extracted from the discharge letters. In patients with re-visits to an outpatient clinic or the ED any additional follow-up assessment and reports were reviewed and screened for changes in the discharge diagnosis. In patients with differences between the discharge diagnosis and the diagnosis on follow-up visits, the final diagnosis was adjudicated by a research committee (JS, UH, JB, MW)-blinded to the details of index visit-based on the results of the follow-up evaluation and records of re-hospitalizations. Each final diagnosis was assigned to one of the five categories: musculoskeletal chest pain, gastrointestinal chest pain, pulmonary chest pain, chest pain in psychiatric diseases, and non-specific chest pain.
We defined a formal ACS rule out testing as serial troponin tests performed at presentation, a second (after 3 to 6 hours), and a third (beyond 6 hours) [16]. Additional endpoints were: recommendations on further evaluation after discharge, and re-visits to the ED.

Data quality and statistical analysis
The quality of the data extraction was assessed by a researcher not involved in the extraction process (BK). In total, six predefined parameters (troponin test result, pain reproducible by movement, coronary angiography, recommendation for further diagnostic evaluation, recommendation for further treatment, and the discharge diagnosis) in 379 ED visits were reviewed. The quality of data extraction was high with an error rate of 5.4% (95% CI 4.5-6.4).
We calculated median and interquartile ranges for continuous variables, numbers and percentages of total for binary or categorical variables. A chi-squared test was used for group comparisons for categorical variables and Kruskal-Wallis test was used for continuous variables between groups. Differences between the diagnostic categories were visualized using bar plots. All analyses were performed with the statistical software R [17].

Ethical review board approval
Due to the retrospective nature of the study, data extraction did not interfere or influence the treatment of patients. The study was approved by the independent Ethics Committee of the Canton Zurich, Switzerland (KEK-ZH number 2014-0506, approved in December 2014) and complied with international standards including the declaration of Helsinki, good clinical practice, and the Swiss law for research in human subjects.

Baseline presentation
The majority of patients were female (55%), married (58%), and the median age was 46 years (IQR 33-60, Table 1). Cardiovascular disease was known in 34.2% (previous acute myocardial infarction in 8.7%), a history of peripheral arterial disease in 0.8%, stroke in 1.9%, and diabetes mellitus in 5.1%. Overall, 32% of the patients met the definition for multimorbidity (� two known diseases). Further details are summarized in Table 1 and S1 Table. ED evaluation Diagnostic evaluation was mainly performed on an outpatient basis (90.2%, Table 2). The diagnostic work-up with regards to vital signs was comparable in all categories: blood pressure

Medication use
Diabetes mellitus therapy 0.37  Table). The formal ACS rule out testing was performed in patients with GI-tract diseases (14.3%), non-specific chest pain (14.0%), pulmonary diseases (10.0%), musculoskeletal chest pain (4.7%), and psychiatric diseases (4.0%).   The most frequently performed additional diagnostic test was a chest x-ray (58.9%). Around 10% of all patients were hospitalized (0.4% in an intensive care unit). Additional evaluation using a tread mill test was performed in 2.6%, echocardiography in 4.8%, MIBI scintigraphy in 0.7%, and coronary angiography in 3.0%. Fig 2 shows the differences in the use of these test for each diagnostic groups. In non-specific chest pain, tread mill tests, MIBI scintigraphy, and coronary angiography were most often performed. Echocardiography was most often performed in patients with a psychiatric discharge diagnosis.

Recommendations at discharge and initiated treatments
Overall, 76 different recommendations at discharge were identified. In Table 3 the most frequent recommendations and initiated treatments are summarized. In 17% of the patients further cardiac evaluation (treadmill testing 9.1%, least often coronary angiography 0.2%) was recommended followed by a GP follow-up or a GP initiated assessment/action (16.6%). In patients with non-specific chest pain cardiac evaluation and GP follow-up was most often recommended (26.5% and 20.8%, respectively). Whereas a psychiatric evaluation was recommended in very few patients (<1%), anxiolytics or psychological treatment was most often initiated in patients with psychiatric discharge diagnosis (18.7%).
Newly initiated acetylsalicylic acid (ASS) treatment was found in 3.5% and for statins in 2.3% of patients. Analgesics were used by 51% of all patients (musculoskeletal chest pain category 66%). The most frequently newly prescribed analgesic was paracetamol (30.4%), followed by NSAIDs (23.2%), and metamizole (17.2%). PPI use was found in 20.5% of patients. Recurrent visits were recorded in 13.1% mainly with emergency readmissions (8.1%) and outpatient visits (3.3%). Whereas 50% were related to the first emergency visit, 50% were due to other reasons. The highest proportion of related visits was found in patients with pulmonary chest pain (23.3%) and a GI-tract disease (20%). Technical evaluations in the outpatient setting at the same hospital were found in less than 5% of the patients.

Discussion
The major finding of this study is that chest pain of non-cardiac origin accounted for 4.2% of all ED visits and the diagnostic evaluation included in a minority of patients a formal cardiologic work-up with sequential cardiac troponin testing. In the majority of patients musculoskeletal chest pain or a non-specific chest pain was the discharge diagnosis and psychiatric diseases were rarely considered. Over 72 different recommendations at discharge were given, ranging from no further measures to extensive cardiac evaluation. Despite the recommendation for cardiologic follow-up evaluation in one fifth of the patients, ASS treatment was initiated in only a small proportion of those patients. The most frequently initiated treatment was analgesics where mainly paracetamol was prescribed. A diagnostic test with proton pump inhibitor was prescribed in 20% of patients without specific recommendations about the follow-up assessment.

Results compared to the literature
NCCP account for a relevant number of emergency department visits. The prevalence of NCCP reported in this study was comparable to a previous study where patients with NCCP accounted for 5% of all ED visits [18]. In the general population, the point prevalence of NCCP may be up to 25% [19]. For example, in a population-based survey in the Olmested County, Minnesota (United States), NCCP was reported by 23% of participants [20]. In patients presenting to the ED with chest pain, 78% consulted a healthcare provider-most commonly general practitioners and cardiologists-in the 12 months previous to the ED presentation [21]. It has been suggested that approximately in two thirds an underlying disease can be identified [22][23][24]. Proton pump inhibitor treatment trials are highly effective to identify patients with underlying gastroesophageal reflux diseases [15]. Further, panic disorders were common in patients presenting with chest pain to the ED and were rarely recognized by physicians but resulted in more testing and referrals [25]. Therefore, the clinical challenge is to determine which  Evaluation of non-cardiac chest pain patients in the emergency department diagnostic tests to apply in patients with chest pain after a cardiac disease has been ruled out to discriminate between patients with non-specific chest pain and other underlying diseases presenting with NCCP. In particular because patients with NCCP experience recurrent pain and a decreased quality of life [26]. This study showed that a primary focus to rule out an acute coronary syndrome (ACS) by extending ECG and cardiac troponin testing to all patients with chest pain in an ED may result in more diagnostic tests without improving the diagnostic and treatment algorithm in the majority of patients. It is important to consider that patients can have elevated troponin levels without cardiac diseases and elevated troponin test may result in more downstream testing without clinical utility [14]. In our study population, no baseline ECG and troponin test was performed in up to 30% and formal ACS rule out testing for fewer than 20%. The diagnoses in those cases was based on the physicians' clinical assessment of the patients' history, clinical findings and risk profile. Comparable to our study, in a prospective study of 108 patients with atypical chest pain presenting to an ED in England, treadmill tests (in 9.3% vs. 2.6% in our study), echocardiography (6.5% vs. 4.8%), coronary angiography (4.6% vs. 3%), and Evaluation of non-cardiac chest pain patients in the emergency department gastroscopy (5.6% vs. 1%) were performed in a minority of patients (despite an older mean age of 60 vs. 46 years) [27].
To the best of our knowledge, this was the first study that assessed diagnostic processes and treatment recommendations in patients with NCCP. We found a lack of clinical concepts to assess and treat patients with chest pain after an ACS has been ruled out. Other strategies shown to be effective or useful were not recommended by physicians. For example, a positive response to a high dose PPI treatment trial for one or two weeks indicates an underlying GERD whereas a negative response rules out GERD and can help primary care physicians to further evaluate their patients [15,28]. According to a Cochrane review cognitive-behavioral therapy may have a short-term effect in patients with chest pain and normal angiogram [29]. However, psychological assessment was recommended in only 3.4% of all patients in this study. Despite the frequency of musculoskeletal chest pain, there is only limited evidence on how to diagnose and treat these patients [30][31][32][33][34]. Physicians use a combination of indicators including the patients' history and systematic palpation of the spine and chest wall [31]. The treatment strategy found in this study included the prescription of analgesics. Musculoskeletal evaluation and treatment by physical therapists was recommended in 1.5% of the patients. It is unclear what the natural course of patients with musculoskeletal chest pain is and whether more intensive management in some patients may be necessary. In a randomized clinical study that compared manual therapy to self-management, more than one third of the patients complained about chest pain in both groups at the one year follow-up [35].

Strengths and limitations
While this study was conducted using rigorous predefined protocols and the data extraction quality was high, there are limitations to consider.
The main limitation is the retrospective nature of this study. Patients were identified by ICD-10 codes for non-specific chest pain and therefore, patients with other diseases presenting with chest pain may have been missed. However, the prevalence of patients with NCCP reported in this study was comparable to a previous study [18]. While great care was used when extracting information from the medical charts, we cannot exclude that information was missed during the process despite a data extraction quality (average error rate of below 6%). Our findings have limited generalizability because it is based on data of one teaching hospital and may not apply to other clinical settings. Further, the quality of the discharge diagnosis depends on the clinical experience of a physician and may be revised later on. More experienced physicians may need less diagnostic tests to define a working diagnosis and initiate a treatment. However, residents and an attending physician saw most patients in the hospitals. Therefore, we believe that the study provides an accurate picture of a condition, which in ED department is managed differently compared to primary care practices. The clinical experience of physicians and the skills to communicate with patients may be particularly relevant in patients with non-specific diagnoses including non-cardiac chest pain [13].

Implications for research
This study has several implications for future research. Future studies should assess the impact of a structured evaluation and treatment recommendation in patients with chest pain after an ACS has been ruled out. To prevent overdiagnosis and overtreatment, studies that assess the efficacy of clinical prediction rules to rule out ACS should be compared to clinical judgment by emergency department physicians. It has been shown that the clinical judgment by ED physicians was at least equally accurate to rule in or out an ACS compared to the HEART score, a prediction rule developed in ED patients with chest pain [36].

Implications for clinical practice
In patients with non-cardiac chest pain, panic disorders and GERD are rarely considered in the treatment recommendations. This study underscores the need for guidance in patients with non-specific chest pain. Patients with chest pain of unknown or unspecific origin may express avoidance and anxiety symptoms [37]. A structured approach with a defined communication strategy may result in assurance and reduce stress.

Conclusion
In this retrospective study a formal work-up to rule out ACS was found in a minority of patients presenting to the ED with chest pain of non-cardiac origin. A wide variation in diagnostic processes and treatment recommendations reflect the uncertainty of clinicians on how to approach patients after a cardiac cause is considered unlikely. Panic and anxiety disorders were rarely considered and a useful PPI treatment trial to diagnose gastroesophageal reflux disease was infrequently recommended.
Supporting information S1