The 'Acute Admission Database' is a clinical database comprising data from acute admissions from the ED at Hillerød University Hospital, which is one of the largest specialized regional hospitals in the Capital Region of Denmark. The hospital is a 24-hour acute care hospital offering emergency, level-2 trauma, medical, surgical, and intensive care services for 310.000 citizens in North Zealand. The ED has approximately 50.000 patient contacts annually. We retrieved 6279 unique patients from the Acute Admission Database in the period from September 22, 2009 to February 28, 2010. Inclusion criteria were all patients aged > 16 years admitted from the ED, either to the ED observationary unit or to a general ward in-hospital. Patients with minor complaints and injuries that did not result in admission were excluded. Patients admitted more than once during the study period were only represented by the latest admission. Selection of the cohort is depicted in Figure 1.
Data collection and quality
The data in the 'Acute Admission Database' derives from 3 sources (Figure 2):
1) Triage data: Data from initial assessment upon arrival, including time and date, vital signs, presenting complaint and triage category. Data was manually entered in the 'Acute Admission Database' as described below.
2) Blood sample results: Sampling performed at admission, including a venous acid-base (VAB) status retrieved from the hospital database, LABKA ll, version 1.4.2.H5 Computer Sciences Corporation (CSC).
3) Patient outcome: Data retrieved from the hospital database, OPUS Arbejdsplads, version 2.5.0.0 ©2010 Computer Sciences Corporation (CSC).
The three data sources were merged using the Central Personal Registry (CPR) number, which uniquely identifies gender and date of birth of all Danish citizens. The back-end database was designed in MSSQL and the front-end data-collecting tool was designed as a web application, enabling staff to report data directly from any workstation in the hospital using normal security validation. All data types were defined as Boolean, integer or numeric. The primary key was the unique CPR number. The secondary key consisted of the date and time of admission, making cross-reference possible with other health related databases. The final data set was of SQL type, making it suitable for export to almost any table format. We used IBM-SPSS Statistics v 18.0 (IBM Corporation) for data analysis.
Triage system and triage data
Hillerød hospital started using 'Hillerød Adaptive Process Triage' (HAPT) [10] in May 2009, having no previous experience with formalized triage. HAPT is inspired by the Swedish Adaptive Process Triage model, ADAPT [11], and has subsequently evolved into the 'Danish Emergency Process Triage' [12], which is currently under implementation at several hospitals across the country. The triage system ranks patients into five colour-coded triage categories. Each patient is assigned a triage level for each of the two main descriptors 1) vital signs and 2) presenting complaint. The variable of the two associated with the more urgent triage category determines the final colour-coded triage level, which in turn determines the level of patient monitoring, treatment and re-evaluation. The triage categories are 1) red (resuscitation, re-evaluation every 0 minutes (min)), 2) orange (emergent, re-evaluation every 10 min), 3) yellow (urgent, re-evaluation every 60 min), 4) green (non-urgent, re-evaluation every 180 min) and blue (minor injuries or complaints, re-evaluation every 240 min). Patients in blue triage category were not admitted and therefore not included in the Acute Admission Database. For further information about the details in the triage system, please consult references [10, 12]. Data collected from the admission process was manually transferred from the triage form into the database. Predefined validation rules and filters controlled the quality of data entry. There were two validation rules for every manually entered numeric value: 1) An absolute range. Data outside this range was not allowed into the database. 2) An uncertainty range, requiring an extra validation from the user. The triage category for the presenting complaint was assigned according to a presenting complaint algorithm [10, 12], defining the acuity of each complaint within 29 predefined main categories. If the presenting complaint algorithm did not adequately cover the patient's condition, the patient was scored as 'no adequate category' and the final triage category was determined solely by the patient's vital signs. If the presenting complaint was not scored by the triage nurse or physician, the medical students responsible for database entry were allowed to assign a category according to the nurse's notes in the triage form (from one of the 29 predefined categories). A three-week test period was used to familiarize the students with the database and to reinforce the nursing staff in recording triage data as completely as possible. During the project period, random samples were taken to ensure correct triage of the patients as well as correct data entry into the database. All triage documents were stored in hard copy in order to be able to retrieve original data if necessary. All data entered into the database was subsequently reviewed by one of two authors (CB and MPL) by comparing data on the triage sheet with the data entered in the database for each patient. The two reviewers did not participate in the primary data entry. Review of patient admission data was performed consecutively, and the reviewers were unaware of patient outcome at the time of registration.
Blood sample data
Blood sampling was performed by a team of medical students, who had a four-day course on procedure and technique before starting the project. They were instructed to minimize tourniquet time and to release the tourniquet as soon as the needle was within the vein. The blood sampling procedure followed a predefined sequence to minimize the risk of pre-analytical errors. The hospital standard procedures included a VAB for all patients in red, orange and yellow triage categories, and this was the final sample obtained from the cubital vein. Blood sampling was performed as soon as possible after patient admission and primary triage. All blood sample results were included in the database. An arterial blood gas was obtained and analyzed on physician request and staffs were instructed to use arterial blood gas results if in need of acute acid-base status.
Outcome measures
Predefined outcomes were retrieved from the hospitals computerized information system. LOS, admission to ICU, and mortality during admission and within 7 or 28 days after admission were registered. Furthermore the receiving ward and the discharge diagnosis were retrieved.
Data safety
The Acute Admission Database was placed on a MS SQL Server 2005, maintained by the Department of Information Technology at Hillerød Hospital. Due to the private nature of the data, user interaction with the database was logged. Transaction logging was done every half hour. All medical students authorized to enter the database were obliged by duty of confidentiality. All had an extensive course in the use of the database and a personal log in. The two reviewers had extended rights and were the only ones with access to change data after logging in order to be able to correct errors in the process of data entry.
Ethics
The study was approved by The Danish National Committee on Biomedical Research Ethics, J.nr. H-A-2009-006, and the Danish Data Protection Agency, Copenhagen, J.nr. HIH 2009-2 Akutdatabasen.