Article Text
PDF
Abstract
Background Increased lactate is associated with high mortality among patients with suspected infection or trauma in the emergency department (ED), but the association with patients with other aetiologies is less well described. The aim of this study was to describe the relation between lactate, aetiology and 7-day mortality in adult ED patients.
Methods A retrospective cohort study of all adult patients who had a lactate measured within 4 h after arrival to the ED at Odense University Hospital between June 2012 and May 2013. The categorisation of suspected aetiology was based on discharge diagnoses.
Results 5360 patients were included; 51.7% were men, and the median age was 67 years (IQR 50–79). 77.2% had low lactate (0–1.9 mmol/L), 16.2% intermediate lactate (2–3.9 mmol/L), and 6.6% high lactate (≥4 mmol/L). 7-day mortality was 2.9% (95% CI 2.4% to 3.5%) for patients with low lactate, 7.8% (95% CI 6.1% to 9.8%) for patients with intermediate lactate, and 23.9% (95% CI 19.6% to 28.8%) for patients with high lactate. The association between lactate level and mortality varied across different diagnostic groups. Based on Area Under the Curve in receiver operating characteristic analysis, lactate level showed to be useful in patients with infection (0.78, 95% CI 0.73 to 0.84), trauma (0.78, 95% CI 0.65 to 0.92), cardiac diseases (0.83, 95% CI 0.75 to 0.91) and gastrointestinal diseases (0.83, 95% CI 0.68 to 0.98). Lactate level was not useful in neurological (0.58, 95% CI 0.50 to 0.67) and respiratory disease (0.64, 95% CI 0.55 to 0.74), and of uncertain value in the remaining diagnostic groups.
Conclusions Among adult ED patients, the prognostic value of lactate varies between diagnostic groups.
- acute medicine-other
- emergency department
- death/mortality
- diagnosis
Statistics from Altmetric.com
Key messages
What is already known on this subject?
Numerous observational studies have proved that increased lactate among patients diagnosed with suspected infection or trauma is associated with high mortality in the emergency department (ED). Until now, the association between other diagnosed categories in the ED with elevated lactate is sparsely described.
What might this study add?
Our study suggests that mortality risk, due to an elevated lactate, differs between different diagnostic groups in the ED, depending on the cause of hyperlactatemia.
Introduction
The lactate level is often interpreted as a marker of tissue hypoperfusion and is used in the definition of shock.1 Lactate is produced by anaerobic metabolism when cellular oxygen usage fails, but other factors, such as decreased clearance can also influence the lactate level.2 ,3
Increased lactate is associated with high mortality among patients with suspected infection or trauma in the emergency department (ED),4–11 but the risk of an adverse outcome might depend on the underlying condition of hyperlactatemia, which can differ according to aetiology.2 ,3
Elevated lactate levels have been found associated with overall mortality in a broad cohort of adult patients as well as the elderly over 65 years of age, independent of the presence of infection.12 ,13 Apart from these two studies, the association among patients with elevated lactate and aetiologies other than infection and trauma is only sparsely described. Increased time to normalise lactate have also been found to predict higher mortality rates6 and have been used in treatment guidance.14
As knowledge of the prognostic value of an elevated lactate level among undifferentiated ED patients could be clinically useful in order to identify critically ill patients for intensive resuscitation, the aim of the present study was to describe the relation between lactate, aetiology and 7-day mortality in adult ED patients.
Method
Study design and setting
We performed a hospital-based cohort-study at Odense University Hospital between 1 June 2012 and 31 May 2013. The ED at Odense University Hospital serves a mixed rural–urban population (n=288 000 adults). It provides 24-hour acute care and acts as primary hospital for the local area. The department operated as an open general ED for unselected patients as well as the primary department for virtually all patients admitted acutely to the hospital. Exceptions are patients with prehospital-identified severe heart disease, pregnant patients past 16th week of gestation, patients in haemodialysis, patients in active chemotherapy or radiation therapy, and children with medical conditions.
Participants
All patients ≥18 years of age who had an arterial blood gas (ABG) sample drawn within 4 h of arrival to the ED were included. If the same patient had more than one contact to the ED during the study period they were only included at first visit. Foreign or emigrated patients who did not have an active Personal Identification Number (PRN) that is unique to all Danish citizens,15 were excluded. So were all ABG analyses without registration of the patient’s PRN or lactate measurement.
Procedure for lactate measurement
Lactate measurements were performed using ABG samples according to a clinical decision based on the patient's condition. All ABG samples were performed bedside, analysed and registered in an ABL machine (ABL 800 Flex, Radiometer medical ApS, Åkandevej 21, DK-2700 Brønshøj, Denmark).
Quantitative variables
On the basis of previous studies and definitions of shock, initial arterial lactate level was a priori defined as low (<2 mmol/L), intermediate (2–3.9 mmol/L) or high (≥4 mmol/L).5 ,7 ,8 ,10 In order to assess potential confounders, we analysed age, gender and comorbidities as risk factors for mortality. Charlson Comorbidity Index was used to generate the comorbidity score (0, 1–2, >2).16 Age was grouped into three categories in a natural order (18–49, 50–79 and ≥80 years).
Data sources
Data was extracted from the ABL machines located at the ED as well as from the electronic patient records. Using the Danish unique PRN, information on included patients was also retrieved from the Danish National Patient Register and the Civil Registration System.15 ,17 From the National Patient Register, we identified all discharge diagnoses (action-diagnoses) based on the International Classification of Diseases 10th revision (ICD-10),18 for the last 10 years, which we used for the Charlson Comorbidity Score. From the Civil Registration System, we gathered information regarding gender and age as well as 7-day mortality.
Definition of disease categories
The presumed aetiological categories were based on the patient discharge diagnosis for the entire admission period. A protocol of criteria for categorisation was made a priori (see online supplementary appendix 1). Two of the researchers, a priori, performed categorisation of all 982 different discharge diagnoses independently and blinded to each other, as well as the individual patient records (see online supplementary appendix 2). The κ-score for interobserver agreement, in classification of the ICD-10 codes, was 0.95 (95% CI 0.94 to 0.97).
The categorisation was defined according to a clinical approach, physiological features and theoretical pathogenesis of lactate production,3 ,19 including either a disease in an organ system or a systemic condition. Diseases in organ systems were defined as neurological, intestinal, cardiac, respiratory, endocrine and haematological diseases. Nephrological and hepatological diseases were considered as one category due to the function of lactate clearance.19 Systemic conditions were defined and prioritised as infection, trauma, hypovolemic condition, malignant diseases, intoxication and allergic conditions. Systemic conditions were prioritised over diseases related to organ systems. For example, a patient with pneumonia was allocated to the category of infection instead of respiratory diseases, and a patient with gastrointestinal bleedings was allocated to the category of hypovolemic conditions that covers all acute bleedings, dehydration, burns and tissue corrosions regardless of which organ is affected. Discharge diagnoses that did not match the criteria for the above-mentioned groups, were categorised into ‘other diseases’. The categories haematology and allergy did not meet an a priori-defined criterion of a minimum number of 50 patients in a category and were transposed to the category ‘other diseases’.
Statistical methods
Baseline characteristics are presented for patients with low, intermediate and high lactate levels. Categorical variables are expressed as percentages with 95% CIs based on the binominal distribution. All continuous variables were non-normally distributed, and are described as medians with IQR. Pearson's χ2 test was used to compare categorical variables.
A logistic regression model was used to assess the relationship (OR) between 7-day mortality and lactate level. For all calculations of OR, the low lactate level was used as reference. A multivariate logistic regression model was performed including age, sex, comorbidity and discharge categories as predefined potential confounding variables. An interaction analysis was performed to test the interaction between lactate and discharge categories in the multivariate regression model. In the different discharge categories, Cuzick's test for trend was used to test the trend in 7-day mortality for increasing lactate values. p Values <0.05 were considered statistically significant.
Area Under the Curve (AUC) regarding discharge categories and lactate levels was used to evaluate lactate as a prognostic marker. AUC >0.75 and a lower CI 95% >0.5 was considered useful. AUC <0.65 and a lower CI 95% <0.50 was considered unlikely to be useful.
All statistical analyses were performed with STATA software (V.13,0 Stata Corporation LP, Texas, USA).
Ethics committee approval
The study was approved by the Danish Data Protection Agency (J No 2013-41-2036) and the Danish National Board of Health (J No 3-2013-355). No further ethical approval is needed for register-based research in Denmark. The reporting of this study conforms to the STROBE (STrengthening the Reporting of OBservational studies in Epidemiology) statement.20
Results
Baseline characteristics
A total of 12 142 ABG tests were performed in the ED during the observation period, and 5360 different patients were included in the study (figure 1).
Flow chart illustrating the inclusion process. N, number; ED, emergency department; PRN, Personal Identification Number.
Median age of the included patients was 67 years (IQR 50–79); 51.7% were men, and the overall mortality in the cohort was 5.1% (95% CI 4.5% to 5.7%). Median time for ABG measurement was 1 h after arrival. Of the 5360 patients, 4139 (77.2%) had a low lactate level (median 1.0 mmol/L; IQR 0.7–1.3), 870 patients (16.2%) had an intermediate lactate level (median 2.5 mmol/L; IQR 2.2–3.0), and 351 patients (6.6%) had a high lactate level (median 5.9 mmol/L; IQR 4.7–9.0) (table 1). Patients with intermediate and high lactate levels were more frequently men (p=0.0001). A higher proportion of multiple comorbidities was found in patients with a high lactate level compared to a low lactate level (p=0.002).
Baseline characteristics
Seven-day mortality and lactate levels
For low, intermediate and high lactate levels, 7-day mortality was 2.9% (95% CI 2.4% to 3.5%), 7.8% (95% CI 6.1% to 9.8%), and 23.9% (95% CI 19.6% to 28.7%), respectively. Univariate comparisons between survivors and non-survivors showed that intermediate lactate (OR 2.8; 95% CI 2.1 to 3.9) and high lactate (OR 10.5; 95% CI 7.8 to 14.3), high age, and a Charlson Comorbidity Score >0 were significantly associated with increasing 7-day mortality (table 2). Comparing the discharge categories univariate, using the category infection as reference, cardiac diseases and malignant diseases were significantly associated with higher risk of mortality. Intoxication, endocrine diseases and other diseases were significantly associated with a lower risk of 7-day mortality.
Univariate and multivariate analyses of the variables lactate, gender, age, comorbidities and discharge category as a function of 7-day mortality
In the multivariate logistic regression model, the adjusted OR for intermediate lactate was 3.0 (95% CI 2.2 to 4.1), and for high lactate 11.5 (95% CI 8.1 to 16.2). A test for interaction between lactate and discharge category in the multivariate analysis, showed that the OR for mortality, according to low lactate level, varied between 0.22 (95% CI 0.10 to 0.51) and 4.10 (95% CI 1.77 to 9.53), intermediate lactate 0.68 (95% CI 0.24 to 1.97) and 6.12 (95% CI 2.73 to 13.72), and high lactate 4.75 (95% CI 1.72 to 13.08) and 70.17 (95% CI 12.01 to 409.88) in relation to discharge categories.
Aetiology, lactate level and 7-day mortality
The most frequent categories of discharge diagnoses were infection (21.1%) followed by respiratory diseases (11.8%), neurological diseases (7.3%) and cardiac diseases (6.7%) (table 2). Within the categories, the overall absolute risk of 7-day mortality differed between 17.1% (95% CI 9.7% to 27.0%) among malignant diseases, and 0.4% (95% CI 0.0% to 2.3%) among endocrine diseases (table 2).
All discharge categories, except neurology and hepatology/nephrology, showed a significant trend towards increasing mortality with increasing lactate level (figure 2). The highest absolute risk of mortality among patients with high lactate was found in the category of malignant diseases at 66.7% (figure 2). Patients with high lactate levels and infection had a 26.4% 7-day mortality, and patients with trauma with a high lactate level had a 29.4% 7-day mortality (figure 2).
Absolute risk of 7-day mortality as a function of the 12 different disease categories and lactate levels in percentage (%) with 95% CIs. *Indicates a significant trend of higher mortality with increasing lactate levels, within a disease category (p<0.05).
High lactate levels were most frequently registered among patients with neurological discharge diagnoses at 16.5% (table 1), but 7-day mortality among neurological patients with high lactate level was low, only 8.2% (figure 2).
Figure 3 presents an overview of the prognostic value of lactate levels in different diagnostic groups. AUC for infection was 0.78 (95% CI 0.73 to 0.84), trauma 0.78 (95% CI 0.65 to 0.92), cardiac diseases 0.83 (95% CI 0.75 to 0.91), gastrointestinal diseases 0.83 (95% CI 0.68 to 0.98), and other diseases 0.83 (95% CI 0.74 to 0.92). This indicates that lactate can have a useful prognostic value in these diagnostic categories as they all have an AUC >0.75 and a lower 95% CI limit >0.5.
Receiver Operating Characteristics Curve for lactate according to discharge categories. AUC, Area Under the Curve.
AUC for neurological diseases was 0.58 (95% CI 0.50 to 0.67), and AUC for respiratory diseases was 0.64 (95% CI 0.55 to 0.74). This indicates that lactate has no prognostic value in these diagnostic groups, as AUC was <0.65 and the upper 95% CI was <0.75. The remaining groups had uncertain prognostic values with AUC in hypovolaemia 0.68 (95% CI 0.50 to 0.85), intoxication 0.71 (95% CI 0.27 to 1.00), malignant 0.66 (95% CI 0.47 to 0.85), nefro-hepatology 0.70 (95% CI 0.51 to 0.89), and endocrine 1.00 (95% CI 0.00 to 1.00).
Discussion
Our study shows a general significant trend towards higher mortality with increasing lactate levels in a broad population of ED patients, but the association varied across different diagnostic groups. Based on AUC in receiver operating characteristic analysis, lactate level showed to be a useful prognostic marker in patients with infection, trauma, cardiac and gastrointestinal diseases, whereas it was not useful in neurological and respiratory disease, and of uncertain value in the remaining diagnostic groups. A test for interaction between lactate and discharge category supported our findings that mortality, according to lactate level, varied between diagnostic groups.
The association between lactate level and mortality in the ED is well described among patients with suspected infection or trauma.6–8 ,10 ,11 The association between increasing lactate and increasing mortality is also well described in other less selected patient populations, mainly in the intensive care unit (ICU) settings.21–23 Two studies have described the association between lactate and mortality in a broad population of ED patients, one only including the elderly.12 ,13 In parallel with the present study and the studies from ICUs, they found increasing mortality with increasing lactate levels among undifferentiated patients with or without infection.
In the analysis of mortality among patients with low, intermediate and high lactate levels, we found a trend of increasing 7-day mortality with increasing lactate among patients with infection, trauma, hypovolemic, respiratory, cardiac, intoxication, endocrine and malignant discharge diagnoses. That AUC differs from Cuzicks test for trend regarding respiratory diseases is likely explained by the variation of the statistical method, suggesting that even though we find significant differences in absolute mortalities according to lactate levels, it does not necessarily reflect the usefulness of lactate as a prognostic marker.
In parallel with other studies, we found that mortality in the intermediate lactate level from 2 to 3.9 mmol/L was higher than in the low lactate level, and that the increased mortality was not limited to patients with a lactate of 4 mmol/L or higher.5–8 ,13 ,21 ,22 We found no association between mortality and lactate level among patients discharged with neurological diagnoses. A high proportion of these patients might have arrived after general seizures, which could explain our observation that prognosis is unrelated to the lactate level.24 Unfortunately, we have no systematic access to data regarding the presence of seizures. Patients discharged with hepatology/nephrology diagnoses showed no trend towards a higher mortality with increasing levels of lactate. The elevated lactate levels for these categories cannot be explained exclusively by circulatory failure, but suggests other mechanisms, such as compromised clearance or decreased metabolism.2 ,3 ,19
Among patients with high lactate and infection, we found a 7-day mortality of 26.4%, which is in accordance with other prospective studies of patients with infection in the ED that report 3-day and 28-day mortality between 28.4% and 26.5%.5 ,9 Among patients with trauma, we found a significant trend of increased mortality according to increased lactate levels. Other studies investigating elevated lactate in selected cohorts of patients with trauma are inconsistent, probably reflecting different aspects of selection.11 ,25
As the aetiology of the patient's condition is not always evident at arrival to the ED, and the prognosis in many diseases is related to early onset of treatment, our finding of a poorer prognosis related to increasing lactate levels in a diverse ED population is potentially important to the clinician. This information could increase awareness of the risk of deterioration among patients with suspected infection or patients with trauma. However, the present study is limited in its nature of a retrospective cohort design, and the usefulness of lactate as a prognostic tool used for treatment guidance in a broad ED population has to be evaluated in a prospective study.
We focus on arterial lactate, and not venous measurements, since the ABG sample is performed upon patient arrival to the ED and contains a lactate measurement. Previous studies have shown that peripheral venous lactate demonstrates a good correlation with arterial lactate,6 and we have not distinguished whether the samples from the ABL were arterial or venous.
The suspected aetiological categories are based on the patient discharge diagnoses. We cannot be sure that the ICD-10 discharge diagnoses reflect all relevant details present at the patient arrival to the ED. However, as ICD-10 discharge diagnoses are found to have high positive predictive values for diagnostics in Scandinavian countries,18 we expect that the main discharge diagnoses reflect the main reason for the patient admittance. The κ-statistic for interobserver agreement of 0.95 represents high agreement and reliability in categorisation of the discharge diagnoses. Blinding, regarding all parameters, mortality and lactate levels included, was used in order to minimise systematic bias. We defined disease categories a priori in order to minimise selection bias. This resulted in a relatively large ‘other diseases’ category, which consists of a priori-defined groups, which were not large enough for statistical analysis as individual groups. It also contains psychiatric diagnoses and all tentative diagnoses from the ED, which were not corrected later, or specified, making it impossible to determine a specific category.
Another limitation is the lack of information regarding hypotension which, unfortunately, we have no access to presently. Our multivariate analysis lacks variables for physiological parameters and blood results prognostic for illness severity, therefore, we cannot conclude lactate as an independent predictor of mortality. Other studies have found lactate as an independent predictor of mortality regardless of BP and other variables for organ dysfunction regarding patients with infection,5 ,7 but it has not yet been investigated for other disease categories. A prospective study with well-defined criteria for different symptom categories in combination with registered vital values including blood results at arrival to the ED would be clinically meaningful, along with a relevant follow-up, based on the results from the present study.
The relatively low number of patients and outcomes in some of the disease categories (particularly for high lactate levels) make interpretation difficult and must be taken into consideration, especially relevant in the categories of trauma, endocrine diseases, malignant diseases, intoxication and hepatology/nephrology. In order to conclude further on these discharge categories, larger studies are warranted. It is reasonable to assume that patients with elevated lactate levels have received more aggressive treatment than patients with normal lactate levels, and therefore, the mortality of an elevated lactate level could be underestimated in this study.
The selection of our cohort is based on the clinical decision to perform an ABG. In our ED, there are no specific guidelines for ABG samples, as they are performed on wide indication when regarded useful in the clinical context. On average, 33 ABGs were performed per day, and we included a mean of 15 patients per day. If the clinical tradition of measuring ABGs differs in other hospitals, the selection of patients will be different. An implementation of lactate measurements as a standard procedure to all patients in the ED is probably not beneficial, as this would lower the positive predictive value of elevated lactate levels.25
In conclusion, the present study shows a general significant trend towards higher mortality, with increasing lactate levels in a broad population of ED patients, but the association varied across different diagnostic groups. Therefore, the prognostic value of lactate varies between diagnostic groups.
Acknowledgments
We would like to thank Niels Ibsgaard Agerbek for professional help with all data extraction.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online appendix 1
- Data supplement 2 - Online appendix 2
Abstract in Danish
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Abstract in Danish - Online abstract
Footnotes
MP and VSB are shared first-authorship.
Correction notice This article has been corrected since it was published Online First. The provenance and peer review statement has been corrected.
Contributors MP and VSB analysed and interpreted data, and drafted the manuscript. ATL and JGH conceived and coordinated the study, as well as performed critical appraisal of the manuscript.
Competing interests VSB and MP were financially supported by a state educational grant. Professor in Acute Medicine ATL is financially supported by an unrestricted grant given from the philanthropic TRYG foundation to University of Southern Denmark. ATL, as well as JGH, are employed by the Faculty of Health and Medical sciences, University of Southern Denmark. None of the above mentioned authors, private, or public companies have financial interests in the project.
Ethics approval The Danish Data Protection Agency (J No 2013-41-2036) and the Danish National Board of Health (J No 3-2013-355).
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement We are happy to supply supplemental information and full dataset if requested.