Numerical Parameters and Quality Indicators in a Medical Emergency Department

Increasingly, patients are admitted to hospital as emergency cases beyond the remit of outpatient emergency care. The university medical center in Erlangen has registered annual increases of up to 10%. Thanks to the continuous availability of medical services, a large proportion of patients even present to hospitals independently (1, 2).

The organizational structures of emergency departments vary widely, partly—but not entirely—owing to the links with rescue services and the general environment of the hospital. Different organizational approaches range from outpatient admission units with triage systems to special emergency outpatient clinics (“chest pain unit”) or admission units for emergencies where possible referrals/transfers to specialist wards or units are decided only after a time delay—for example, the next morning (1, 3–6).

Although triage instruments have become the quality standard in emergency departments, and time intervals—for example, from the patient’s arrival to initial contact with a doctor/nursing staff—can be measured in a replicable fashion, this does not allow any conclusions about the quality—in the sense of correctness/appropriateness of measures—or about the time it takes to reach a diagnosis.

Ultimately, quality indicators are lacking in the heterogeneous area between outpatient care delivered by general practitioners and inpatient care (4, 7).

The main requirements for an emergency department are, however, homogenous. They consist of recognizing the acutely life threatening condition of the patient, providing appropriate treatment, or referring the patient for such treatment. This process comprises primary diagnostics and risk stratification on the basis of lead symptoms, initial care, and organization of further specialist diagnostic tests and treatment (1, 2, 6, 8).

The present study aims to investigate whether on the basis of data from the hospital information system—data whose documentation is mandatory and that are available in an electronic format—objectifiable numerical parameters of an emergency department can be calculated and used as potential quality indicators.

Method
In a retrospective design, we included in the study all patient contacts at the medical emergency admissions department at the university medical center in Erlangen over a period of 12 months. Patients’ movements—in a daily or weekly rhythm pattern—discharge diagnoses, and the respective treatment times in the emergency department were analyzed.

The medical emergency department at the university medical center is responsible for surgical and neurological admissions of patients from the Erlangen area, with a population of more than 100 000, and from the surrounding areas. Further, the university medical center is of supraregional importance; patients from other regions are admitted to hospital through this emergency department. The university medical center comprises 23 hospitals with about 1400 beds. As a rule, patients are delivered by emergency services or emergency physicians to one of the emergency wards.

Primary examination of the patient then takes place in the medical emergency department, followed by diagnostic tests and treatment on the basis of the main symptoms, and in adherence to internal standards.

The emergency department is staffed by an at least third year resident with a qualification in emergency medicine or “rescue services,” and a minimum of 1 year’s experience in intensive care medicine; the staff work in shift patterns. Continuous backup is provided by on-call medical specialists. Altogether, 8 monitoring stations are available as acute treatment units. Short term inpatient care is not available in the emergency department. Further interventions—for example, emergency endoscopies—are conducted in the endoscopy ward or the intensive care unit.

Data management
We obtained our data from the university medical center’s centralized data warehouse, where data from central information systems as well as from different departmental systems are collected and made available for integrated evaluations. The evaluations took place by using a direct SQL database query as well as an integrated user interface with graphical tools for analyzing the data (9).

The entire cohort of medical emergency admissions was identified on the basis of organizational units, changes in types of cases, and main diagnoses according to the ICD-10, and the resultant dataset was analyzed under several clinical aspects.

Types of diagnoses
Admission and discharge diagnoses and main hospital diagnoses correspond to the definition of the German coding guidelines and were taken from the hospital’s clinical information system. The diagnoses were analyzed according to diagnostic groups (one-digit ICD code) or by individual diagnoses (three-digit ICD code).

Diagnosis related length of hospital stay
The length of hospital stay describes a time interval counted in minutes and is calculated from the interval between the time a patient is registered by the administrators to the time that patient is transferred into another organizational unit from the emergency ward, as documented in the clinical information system. For the purposes of this study, the length of a patient’s hospital stay—related to their diagnosis at admission—will be referred to as the diagnosis related length of stay.

Diagnostic agreement and diagnostic efficiency
Diagnosis at admission is based on a patient’s medical history, clinical examination by the admitting doctor in the medical emergency department, emergency investigations (for example, laboratory testing, radiology, and electrocardiography), and further diagnostic measures—recorded in the clinical information management system as a diagnostic code. The discharge diagnosis—also coded in the information system—is based on the total information and data gathered during the patient’s inpatient stay, and on the diagnosis at -admission. For the purposes of this article, this serves as the reference diagnosis to calculating the diagnostic efficiency (DE). Assuming that an admission diagnosis (ICD) made in the emergency department is confirmed as the discharge diagnosis, the diagnostic agreement (DA) can be calculated within a group as the relative frequency of agreement between the diagnosis made in the emergency department and the main diagnosis made in the hospital.

On the basis of the DA and the respective ICD based length of stay in the emergency ward (shortest, longest, and mean length of stay in minutes), then the ICD based DE can be calculated for the shortest stay (DE-short), for the longest stay (DE-long), and for the mean stay (DE-mean) according to the formula shown in Figure 1 (gif ppt).

Statistics
The collected data were tabled and evaluated by using the statistical package SPSS version 13. The descriptive results of continuous variables were reported as

• Minimum
• Maximum
• Mean
• Median
• Standard deviation (SD)
• Skewness
• Standard error of skewness (SE).

Results
During the study period, 6683 patients were treated in the emergency department. 4321 patients (64.6%) received further care as hospital inpatients. The mean age of the patients was 67.4 years (minimum 18, maximum 104); 52.4% of patients were female. Of those admitted as inpatients, 270 (6.2%) died during further treatment. 2362 patients (mean age 49.2 years; minimum 18, maximum 96; 45.7% female) were discharged directly by the emergency department after diagnosis and treatment.

Patient contact in the daily and weekly profile
The daily profile showed the largest increase in patients—independently of what day of the week it was— between 8 am and 11 am, with peaks occurring at 11 am, 2 pm, and 7 pm (Figure 2 gif ppt).

The number of inpatient admissions developed in parallel to the number of patient contacts and independently of the time of day. Of the patients admitted to hospital, 5% were admitted to wards not specializing in internal medicine. 14% of patients were admitted to intensive care units.

Range of outpatient and inpatient diagnoses
The treating physicians documented 8626 admission diagnoses (ICD one-digit to three-digit codes). At the time of admission, 4973 of these diagnoses were in 4321 inpatients and 3653 were in 2362 outpatients. Table 1 (gif ppt) shows the three most common admission -diagnoses, subcategorized by specialty.

Quality indicators
We present data on admission and discharge diagnoses from the clinical information system, as well as the diagnosis related length of stay in the emergency department—some of them subcategorized by specialty—as numerical parameters and the resulting agreement and efficiency as potential quality indicators.

• Diagnoses at admission and discharge
Of a total of 4973 admission diagnoses, 71% were confirmed as discharge diagnoses. In 29% of patients, an additional 2160 treatment diagnoses were also coded. The frequency spread of these diagnoses was wide. The most common were -gastrointestinal hemorrhage (K92.2), dehydration, and syncope and collapse (R55), dyspnea (R06.0), unstable angina (I20.0), left ventricular failure (I50.1), urinary tract infection (N39.0), acute posthemorrhagic anemia (D62), and acute subendocardial myocardial infarction (I21.4).

• Diagnosis related length of stay
The mean length of stay in the emergency department was 116 minutes (minimum 1 minute, maximum 952 minutes, SD 77 minutes, skewness 2, SE 0.038). For the specialty specific length of stay, the following treatment times applied:

Gastroenterology/hepatology—In this group, patients with gastrointestinal hemorrhage (K92) were transferred after receiving treatment for 80 minutes, those with acute pancreatitis (K85) after 88 minutes, other diseases of biliary tract (K83) after 98 minutes, alcoholic liver disease (K70) after 90 minutes, non-infectious gastroenteritis or colitis (K52) after 114 minutes, and calculus of bile duct with cholangitis/ cholelithiasis (K80) after 130 minutes.

For the most common individual diagnoses, it was striking that, for example, patients with acute bleeding gastric ulcers or duodenal ulcers (K25.0, K26.0) were transferred after 55 minutes and 47 minutes, but patients with chronic duodenal ulcers with hemorrhage (K26.4) were transferred only after 138 minutes.

Pneumology—In this group, patients with pneumonitis due to solids and liquids (J69) were transferred after 88 minutes’ treatment, those with acute respiratory failure (J96) after 102 minutes, with chronic obstructive pulmonary disease (J44) after 108 minutes, and with pneumonias (J18, J15) after 112 minutes. Patients with pulmonary embolism and signs of right ventricular failure (I26.0) were treated in the emergency department for 98 minutes, and those without signs of right ventricular failure (I26.9) for 156 minutes.

Cardiology—Examination of the most common diagnostic codes led to the conclusion that for cardiovascular disorders (I) with atrial fibrillation and flutter (I48), the mean length of stay was 109 minutes, and for congestive heart failure (I50) it was 123 minutes.

Among the individual diagnostic codes, the picture is more differentiated. For acute transmural myocardial infarction of the anterior wall (I21.0) and acute transmural myocardial infarction of the inferior wall (I21.1), the length of stay was 26 and 20 minutes.

Diagnostic agreement and diagnostic efficiency
Among the 4321 patients receiving inpatient treatment, the DA between admission and discharge diagnosis was 0.71. Table 2 (gif ppt) provides an overview of the DA of the three most common admission diagnoses. The DA was highest for patients with atrial fibrillation and flutter (I48), at 0.92, followed by a DA of 0.89 for patients with acute coronary syndrome (I20/21), and a DA of 0.89 for patients with chronic obstructive pulmonary disease with acute lower respiratory infection (J44).

The mean diagnostic efficiency (DE) was highest for patients with acute coronary syndrome, at 0.85/min; and similarly high for patients with acute pancreatitis, at 0.84/min; followed by 0.81/min for patients with atrial fibrillation or flutter.

The lowest values of 0.16/min, 0.19/min, and 0.37/min applied to patients with an admission diagnosis of acute respiratory failure (J96), volume depletion (E86), or other disorders of fluid, electrolyte and acid–base balance (E87).

Relative to the total departmental cohort, the DE was 0.61%/min (DE-short 71.00/min, DE-long 0.07/min).

Discussion
One of the challenges that an emergency department is faced with is to decide whether inpatient admission is indicated. Findings or symptoms are assigned to an admission diagnosis by the admitting emergency physician; this diagnosis justifies inpatient admission and has to stand up to inspection by, for example, the Medical Services of Statutory Health Insurance Bodies in Germany (Medizinischer Dienst der Krankenversicherung, MDK).

Primary misadmissions have to be recorded, as do discharges for financial reasons (1, 2, 3, 6, 10, 11). The indication for admission in this study was independent of the time of day and developed in parallel with the total number of patient contacts per time unit.

The number of patients (Figure 2) follows a circadian rhythm that is typical for emergency departments and is not influenced by the severity of the illness (1, 4, 10, 12, 13). Patients receiving outpatient treatment have a wide range of diagnoses, and symptoms including a raised temperature, nausea/vomiting, cough, or unspecific pain in the upper abdomen—these patients account for 44%. Patients receiving inpatient care have clearly defined specialist diagnoses in 59% of cases. Symptoms and abnormal laboratory findings are in fourth position among inpatient diagnoses, whereby syncope and collapse take pole position. A lead symptom oriented approach in primary diagnostics therefore seems a useful standard in the emergency setting (1, 6, 8, 13).

A correct diagnosis is a quality indicator of an emergency department (14). In the present study, this was estimated retrospectively by means of the DA. Our study found that specific diagnoses—such as cardiac arrhythmia, acute coronary syndrome, chronic -obstructive pulmonary disease, or acute pancreatitis—were correctly diagnosed in the emergency department to a higher than average extent. By contrast, unspecific transfer diagnoses—such as disturbances to the electrolyte balance or respiratory failure—were mostly not confirmed.

Since in clinical emergency medicine, time is an essential factor for the prognosis, we used the diagnosis related length of stay in addition to the DA as a further index (15–18).

Diagnosis related length of stay as the only quality indicator, however, entails the risk of misjudgement when assessing the efficiency of an emergency department (19, 20). For example, the much cited “golden hour” or “door to balloon time” prioritize the time aspect only for patients with a confirmed diagnosis (17, 18). From the perspective of a departmental cohort, however, falsely negatively diagnosed cases are not captured, and a mean length of stay can therefore not be determined (19, 20). By contrast, it has to be borne in mind that this does not do any justice to the acrobatics performed by emergency departments—namely, the correct diagnosis per unit of time. By using the DE as a new, combined criterion, both indices are included in the quality assessment. Theoretically, values between 0 and 100 for DE can be measured (if the length of treatment >1 min). Since the diagnostic process requires steps that take a minimum amount of time even in the best-case scenario (for example, deep vein thrombosis in the leg: physical examination, D-dimer tests, ultrasonography), the maximum value of 100 cannot be achieved in practice. The optimum treatment case in each diagnostic group is calculated by means of the DE-short and the case with the longest stay by means of the DE-long.

Optimization of the diagnostic process becomes possible only by means of systematically identifying such special cases.

In the present study, the DA was 0.71 and the DE 0.61/min for the entire cohort.

The highest diagnosis related DE—0.85/min (DE-short=88.97, DE-long=0.11) was calculated for patients with acute coronary syndrome. The lowest DE—0.16/min (DE-short=18.92, DE-long=0.06)—applied to patients with unspecific diagnoses, such as respi-ratory failure, which are often indicative of an unclear diagnosis at the time the patient is transferred. A bed ward assigned to the emergency department would probably contribute to increasing the DA, and possibly the DE, for this treatment area (21, 22).

However, dependence from local conditions should have a subordinate role because, as mentioned earlier, for the treatment of certain diagnoses, time allowances (“door to needle time”) exist that have to be supported in different hospitals by organizational and technical measures (for example, standby service for heart catheterization) (15, 23, 24).

A limitation of our study is the fact that the data were taken retrospectively from the documentation of the clinical information system. Prospective studies are needed to find out whether prospective data capture for the purpose of quality documentation would result in significant deviations from the patterns we identified. Further, the correctness (reliability and correct allocation/assignation) of the diagnostic codes is subject to the coding quality. As is known from postmortem studies, the clinical and diagnostic correctness in making diagnoses is subject to a certain error ratio, which in this study—as in most other clinical studies (19, 20)—was not taken into consideration. A further limitation of the study lies in the fact that we did not measure the length of time from the patient’s admission to their transfer but the length of time from administrative instruction to transfer. We did not include in the analysis the type of admission (for example, admission by emergency physician), the time of admission (day, night), or the doctors’ level of medical qualification and experience, and how all these affect the DE. For the less than 5% of pathologies that were not related to internal medicine, no DE was calculated. How quality indicators affect important primary outcome measures—for example, the length of time until the patient responds to treatment, the duration of the hospital stay, the recovery success, or mortality—was not investigated. In order to establish the newly developed quality indicators that were presented in this study, validation is therefore required. However, the numerical parameters came from the clinical information system and can be electronically generated; they characterize how emergency admissions work and are available for all emergency departments. We intentionally chose clearly defined parameters—such as time intervals and coded diagnoses—which would stand up to external assessment.

Conflict of interest statement
The authors declare that no conflict of interest exists according to the guidelines of the International Committee of Medical Journal Editors.

Manuscript received on 6 November 2008, revised version accepted on
14 September 2009.

Translated from the original German by Dr Birte Twisselmann.

Corresponding author
Priv.-Doz. Dr. med. Harald Dormann
Klinikum Fürth
Jakob Henle Str. 1
90766 Fürth, Germany
harald.dormann@klinikum-fuerth.de