Development, implementation and evaluation of an ECG algorithm for ambulance service
In order to facilitate and accelerate the interpretation of ECG images in preclinical emergency medicine, an algorithm was developed to test this. In an online test, 357 staff of emergency medical service tested the functionality of the analysis procedure. Defibrillation-worthy ECG images are not included in this work because they belong to the resuscitation, has already been extensively published. The development of the algorithm involved three levels: The first was to define the properties of preclinical ECG images based on questions that needed to be answered dichotomically, such as "Are P-waves present?" "Yes" or "No". The answers were summarized in a table, from which the graphic processing of the information in the form of a search tree took place as a second level of development. The algorithmic query always started with the question "Did the patient have a pulse?" And ended with a defined ECG target image. This created a complex primary algorithm. As a third stepp, the primary algorithm was compressed to make it practical. The 11 questions are tested in this study. The study population was guided online through a web-based test. Each study was randomly assigned to the group with or without algorithm. In the case of using the secondary algorithm, it appeared of the screen throughout the duration of the interpretation. The test persons had to interpret 15 ECG images in succession, each with 5 predefined answers. The results of the query were collected in a database and formed the basis of the statistical calculations. From 15.05.2013 to 10.07.2013, a total of 387 people took part in the online survey. n = 357 subjects (92.25%) completed the complete ECG study with input questions and the interpretation of 15 images. All of the following results refer exclusively to the participants with closed interpretation: group without algorithm (μ0) n = 175: correct answers x ̅ = 11,70; σ = 2.024 and total time in sec. x ̅ = 392.35; σ = 294,687. Group with algorithm (μ) n = 182: correct answers x ̅ = 12,17; σ = 2.877 and total time in sec. x ̅ = 316.83; σ = 149.56. The statistical mass consisted of 154 (43.1%) paramedics, 77 (21.6%) paramedics, 63 (17.6%) paramedics and 35 (9.8%) specialists. The remaining participants had a qualification as first responder or fireman. 62.7% of the study participants worked full-time in the rescue service. The average age was 31.6 years. 62.7% interpreted an ECG 1 to 10 times a month, 15.7% 11 to 20 a month. 82.3% of the subjects rated their ECG interpretability as medium good. The hypothesis that the number of correct answers in the interpretation of preclinical ECG images increases with the aid of the algorithm was again rejected in favor of the null hypothesis - for lack of significance α = 0.072. By contrast, the analysis time for ECG interpretation decreases as expected by using the developed algorithm. H1 was retained, with α = 0.003, validated by t-test in independent samples. Neither age nor gender influenced the number of correct answers or interpretation time. Rs = 0.387 with α = 0.05 between the medical qualification and the number of correct answers. Furthermore, with rs = -0.261 and α = 0.05, a correlation between the medical qualification and the analysis time was to be recorded. rp = 0.362; Significance 2-sided α = 0.05 was found in the self-assessment of the subjects. This means that they have generally assessed their ECG interpretation skills correctly. For emergency responders there was a higher number of correct responses with α = 0.05, but no faster interpretation time. The frequency of the ECG interpretation in each case exercised had an influence on the number of correct answers when using the algorithm. rs = 0.362; Significance 2-sided α = 0.05. The time span since the acquisition of the last and highest medical qualification has a highly significant influence of α = 0.005 on the number of correct answers when using the algorithm. That is, the shorter the distance to the end of the training, the more correct answers. In addition, it could be shown that subjects with a low medical qualification using the algorithm were able to increase the number of correct interpretations most strongly (α = 0.05). In summary, the interpretation of preclinically relevant ECG images using the newly developed algorithm is recommended on this data basis. All occupational groups involved in the study benefited from it.