Assessment of Disease Codes Reliability in Emergency Department Patients

For more than a century, the international classification of diseases (ICD) has had a wide range of global uses. Through data reported and coded with ICD, critical information on the scope, causes, and consequences of diseases and deaths is provided globally (1, 2). The statistics obtained from these data support service planning, quality and safety management, healthcare research, and payment systems (2). Therefore, the reliability of the data is critically important. Problems in data entry due to excessive workload, insufficient staff, or lack of experience threaten data reliability. There is no study in our country testing the reliability of these data that will provide a source for the planned national database application. In this study, we aimed to assess the reliability of emergency department diagnoses by investigating the accuracy of diagnoses of presenting patients where high-intensity service delivery occurs.

After obtaining data usage permission from the Local Ethics Committee (University of Health Sciences Turkey, Gaziosmanpaşa Training and Research Hospital Clinical Research Ethics Committee, date: 16.08.2023, no: 118), records of patients who requested urology consultations after presenting to the emergency department within the past year prior to the approval date were retrospectively screened through the hospital automation system (Sarus^® Hospital Information System, Technoritma Software Services Inc., Ankara, Turkey). The records were classified separately as emergency area presentations in the categories of green, yellow, red, and other (internal medicine, surgery, pediatrics), taking into account the time of day. Data such as age, gender, and the ICD codes entered in the consultation request and response were recorded separately. The compatibility of the patients’ emergency department diagnoses by the triage physician with the diagnoses made after further examination and evaluation in the urology clinic was evaluated.

Statistical Analysis

In this study, statistical analyses were performed using the NCSS (Number Cruncher Statistical System) 2007 Statistical Software (Utah, USA) package. In addition to descriptive statistical methods (mean, standard deviation), the compatibility between the groups was evaluated using Cohen’s kappa analysis. The results were evaluated at a significance level of p<0.05.

In the past year, urology consultation requests were made for 2.218 patients from emergency areas, but urology consultations were not conducted for 21 patients. A total of 2.197 patients [1.660 (75.56%) men and 537 (24.44%) women] with an average age of 50.59±23.10 (range: 1-98) years were included in the study. Of the patients seen in the urology clinic, 637 (28.99%) were referred from the green zone, 703 (32.04%) from the yellow zone, 221 (10.10%) from the red zone, and 636 (28.86%) from other branches. Of the patients, 1.623 (73.87%) presented between 08:00-16:00, 406 (18.48%) between 16:00-00:00, and 168 (7.64%) between 00:00-08:00, (Table 1). Patients were most frequently consulted with 13 different ICD codes, with “Hematuria; R31” being the most common, and after urological evaluation, the most common diagnosis was “Renal colic; N23” (Table 2). The ICD codes of patients who requested urology consultations from emergency areas were found to have a statistically significant compatibility with the ICD codes of patients after urological evaluation (kappa: 0.863, p<0.05). This significant compatibility was also observed in the classified evaluations in terms of the unit and hours of consultation requests (Table 1). Although the decrease was statistically significant, it was found that the compatibility rate decreased during peak hours of patient volume (08:00-16:00 kappa: 0.856, 16:00-00:00 kappa: 0.878, 00:00-08:00 kappa: 0.887).

Data coded and reported using the ICD offer standard terminology and classification, in a conceptual framework independent of language and culture (1). Through this, ICD-based national data systems have been developed to determine priorities by analyzing data on incidence and mortality rate, especially in situations threatening public health like cancer. After the National Cancer Institute of America launched the SEER (Surveillance, Epidemiology, and End Results) program in 1973, the World Health Organization published the International Classification of Diseases-Oncology (ICD-O) in 1976 (3, 4). ICD-O was approved in our country in 2013 and came into force in 2016 (5). However, the recent pandemic has revealed that systems collecting only cancer cases are insufficient. Comprehensive databases providing data on acute and chronic diseases affecting public health are still lacking. In our study, the reliability of ICD codes in automation systems that will also provide sources for non-oncology databases was evaluated through a pilot study. Even in departments with relatively higher patient volume, such as the emergency department, diagnostic reliability was found to be high.

In programs with comprehensive and continuous data entry, like national databases, ensuring data reliability at the beginning is important, but it is also crucial to design the process in a way that allows for continuous testing of data reliability (6). Our study found that, although statistically significant, data reliability slightly decreased during peak hours of patient volume. The high accuracy of urological diagnoses was thought to be related to the lower patient volume in the urology consultation area compared to clinics. In addition to facilitating data entry into automation systems, it was concluded that employing a sufficient and competent workforce might be effective in ensuring data security. Providing optimal workforces and physical conditions was thought to facilitate the monitoring of dynamic data flow.

Study Limitations

The data from our study are too limited to support the claim that sufficient reliability has been achieved for the creation of national database programs. This necessitates multicenter, high-volume studies. Our study conducted in the busy emergency department serves as a pilot study that will pave the way for similar studies in different clinics.

The reliability of diagnosis codes is high even in high-volume areas like emergency departments. These results have inspired the creation of secure databases that will provide a source for diagnosis code-based studies for non-cancer entities affecting public health in our country.