RESUMEN
OBJECTIVE: It is challenging to identify sepsis in the emergency department, in part due to the non-specific presentation of septic patients. Current clinical sepsis screening tools rely on vital signs but many patients present with near normal vital signs and are therefore not identified as septic. This suggests that variables, e.g. signs and symptoms, need to be included to improve sepsis detection in the emergency department. Our hypothesis was that the presentation of sepsis differs based age and sex. The potential differences in presentation could be used to apply to future sepsis screening tools. The aim was to analyze the prevalence of keywords reflecting the presentation of septic patients in the emergency department in relation to age and sex. METHOD: Retrospective cross-sectional study. Keywords reflecting sepsis presentation to the emergency department were quantified and compared between age categories and the sex. 479 patients admitted to the emergency department of Södersjukhuset, Stockholm during 2013 and discharged with an ICD-10 code consistent with sepsis were included. We adjusted for multiple comparisons by applying Bonferroni-adjusted significance levels for all comparisons. RESULT: "Pain" and "risk factors for sepsis" were significantly more common among patients younger than 65 years as compared with those 75 years and older: (n = 87/137; 63.5% vs n = 99/240; 41.3%, P-value < 0.000) and (n = 74/137; 54.0% vs 55/240; 22.9%, P-value < 0.000) respectively. "Risk factors for sepsis" was also significantly more common among patients between 65 and 74 years as compared with those 75 years and older: (n = 43/102; 42.2% vs 55/240; 22.9%, P-value < 0.000). "Pain" and "gastrointestinal symptoms" were significantly more common among women as compared with men: (n = 128/224; 57.1% vs n = 102/255; 40.0%, P-value < 0.000) and (n = 82/244; 36.6% vs n = 55/255; 21.6%, P-value < 0.000) respectively. CONCLUSION: The keywords "pain" and "risk factors for sepsis" were more common among younger patients and "pain" and "gastrointestinal symptoms" were more common among women. However, most keywords had a similar prevalence irrespective of age and sex. The results could potentially be used to augment sepsis screening tools or clinical decision tools.
Asunto(s)
Sepsis , Choque Séptico , Masculino , Humanos , Femenino , Estudios Retrospectivos , Choque Séptico/diagnóstico , Estudios Transversales , Servicio de Urgencia en Hospital , Sepsis/diagnóstico , Sepsis/epidemiologíaRESUMEN
BACKGROUND: Sepsis is a life-threatening condition, causing almost one fifth of all deaths worldwide. The aim of the current study was to identify variables predictive of 7- and 30-day mortality among variables reflective of the presentation of septic patients arriving to the emergency department (ED) using machine learning. METHODS: Retrospective cross-sectional design, including all patients arriving to the ED at Södersjukhuset in Sweden during 2013 and discharged with an International Classification of Diseases (ICD)-10 code corresponding to sepsis. All predictions were made using a Balanced Random Forest Classifier and 91 variables reflecting ED presentation. An exhaustive search was used to remove unnecessary variables in the final model. A 10-fold cross validation was performed and the accuracy was described using the mean value of the following: AUC, sensitivity, specificity, PPV, NPV, positive LR and negative LR. RESULTS: The study population included 445 septic patients, randomised to a training (n = 356, 80%) and a validation set (n = 89, 20%). The six most important variables for predicting 7-day mortality were: "fever", "abnormal verbal response", "low saturation", "arrival by emergency medical services (EMS)", "abnormal behaviour or level of consciousness" and "chills". The model including these variables had an AUC of 0.83 (95% CI: 0.80-0.86). The final model predicting 30-day mortality used similar six variables, however, including "breathing difficulties" instead of "abnormal behaviour or level of consciousness". This model achieved an AUC = 0.80 (CI 95%, 0.78-0.82). CONCLUSIONS: The results suggest that six specific variables were predictive of 7- and 30-day mortality with good accuracy which suggests that these symptoms, observations and mode of arrival may be important components to include along with vital signs in a future prediction tool of mortality among septic patients presenting to the ED. In addition, the Random Forests appears to be a suitable machine learning method on which to build future studies.
Asunto(s)
Servicio de Urgencia en Hospital , Mortalidad Hospitalaria , Sepsis , Anciano , Anciano de 80 o más Años , Estudios Transversales , Femenino , Humanos , Aprendizaje Automático , Masculino , Persona de Mediana Edad , Estudios Retrospectivos , Sepsis/mortalidad , SueciaRESUMEN
BACKGROUND: Current sepsis screening tools are predominantly based on vital signs. However, patients with serious infections frequently present with normal vital signs and there has been an increased interest to include other variables such as symptoms in screening tools to detect sepsis. The majority of patients with sepsis arrive to the emergency department by emergency medical services. Our hypothesis was that the presentation of sepsis, including symptoms, may differ between patients arriving to the emergency department by emergency medical services and patients arriving by other means. This information is of interest to adapt future sepsis screening tools to the population in which they will be implemented. The aim of the current study was to compare the prevalence of keywords reflecting the clinical presentation of sepsis based on mode of arrival among septic patients presenting to the emergency department. METHODS: Retrospective cross-sectional study of 479 adult septic patients. Keywords reflecting sepsis presentation upon emergency department arrival were quantified and analyzed based on mode of arrival, i.e., by emergency medical services or by other means. We adjusted for multiple comparisons by applying Bonferroni-adjusted significance levels for all comparisons. Adjustments for age, gender, and sepsis severity were performed by stratification. All patients were admitted to the emergency department of Södersjukhuset, Stockholm, and discharged with an ICD-10 code compatible with sepsis between January 1, and December 31, 2013. RESULTS: "Abnormal breathing" (51.8% vs 20.5%, p value < 0.001), "abnormal circulation" (38.4% vs 21.3%, p value < 0.001), "acute altered mental status" (31.1% vs 13.1%, p value < 0.001), and "decreased mobility" (26.1% vs 10.7%, p value < 0.001) were more common among patients arriving by emergency medical services, while "pain" (71.3% vs 40.1%, p value < 0.001) and "risk factors for sepsis" (50.8% vs 30.8%, p value < 0.001) were more common among patients arriving by other means. CONCLUSIONS: The distribution of most keywords related to sepsis presentation was similar irrespective of mode of arrival; however, some differences were present. This information may be useful in clinical decision tools or sepsis screening tools.
RESUMEN
INTRODUCTION: Sepsis is an acute, life-threatening condition caused by a dysregulated systemic response to infection. Early medical intervention such as antibiotics and fluid resuscitation can be life-saving. Diagnosis or suspicion of sepsis by an emergency call-taker could potentially improve patient outcome. Therefore, the aim was to determine the keywords used by callers to describe septic patients in South Africa when calling a national private emergency dispatch centre. METHODS: A retrospective review of prehospital patient records was completed to identify patients with sepsis in the prehospital environment. A mixed-methods design was employed in two-sequential phases. The first phase was qualitative. Thirty cases of sepsis were randomly selected, and the original call recording was extracted. These recordings were transcribed verbatim and subjected to content analysis to determine keywords of signs and symptoms telephonically. Once keywords were identified, an additional sample of sepsis cases that met inclusion and exclusion criteria were extracted and listened to. The frequency of each of the keywords was quantified. RESULTS: Eleven distinct categories were identified. The most prevalent categories that were used to describe sepsis telephonically were: gastrointestinal symptoms (40%), acute altered mental status (35%), weakness of the legs (33%) and malaise (31%). At least one of these four categories of keywords appeared in 86% of all call recordings. CONCLUSION: It was found that certain categories appeared in higher frequencies than others so that a pattern could be recognised. Utilising these categories, telephonic recognition algorithms for sepsis could be developed to aid in predicting sepsis over the phone. This would allow for dispatching of the correct level of care immediately and could subsequently have positive effects on patient outcome.