Diagnosis trajectories of prior multi-morbidity predict sepsis mortality

Sepsis affects millions of people every year, many of whom will die. In contrast to current survival prediction models for sepsis patients that primarily are based on data from within-admission clinical measurements (e.g. vital parameters and blood values), we aim for using the full disease history to predict sepsis mortality. We benefit from data in electronic medical records covering all hospital encounters in Denmark from 1996 to 2014. This data set included 6.6 million patients of whom almost 120,000 were diagnosed with the ICD-10 code: A41 ‘Other sepsis’. Interestingly, patients following recurrent trajectories of time-ordered co-morbidities had significantly increased sepsis mortality compared to those who did not follow a trajectory. We identified trajectories which significantly altered sepsis mortality, and found three major starting points in a combined temporal sepsis network: Alcohol abuse, Diabetes and Cardio-vascular diagnoses. Many cancers also increased sepsis mortality. Using the trajectory based stratification model we explain contradictory reports in relation to diabetes that recently have appeared in the literature. Finally, we compared the predictive power using 18.5 years of disease history to scoring based on within-admission clinical measurements emphasizing the value of long term data in novel patient scores that combine the two types of data.


Sepsis patients stratified into more than 2,200 different trajectories
We observed a long tailed distribution, where a few trajectories were followed by many patients, while most trajectories were followed by fewer patients. We found that 21 of the top 45 trajectories belonged mainly to 'Diseases of the circulatory system', confirming that many sepsis patients had chronic cardio-vascular problems (eTable 2). 3,366 patients followed the most common trajectory, where patients had 'Angina Pectoris' before 'Chronic ischemic heart disease' with 'Pneumonia' third. Another 15 of the top 45 trajectories belonged to the 'Endocrine, nutritional and metabolic diseases' chapter. More specifically, these 15 trajectories all contained two or three of the diabetes ICD-10 codes ('IDDM', 'NIDDM', and 'unspecific diabetes mellitus'), confirming, that many patients with sepsis had diabetes mellitus prior to their sepsis diagnosis.
Many patients followed several pathways, especially trajectories containing similar diagnoses within the same chapter. We therefore decided to examine the number of patients that followed each trajectory in a manner where each patient only was counted in the most populated trajectory he/she followed thereby visualizing better the diversity across the sepsis population.
We then observed several new chapters occurring, including 'Diseases of the genitourinary system', 'Diseases of the respiratory system', 'Neoplasms', 'Diseases of the blood and blood-forming organs and certain disorders involving the immune mechanism' and 'Diseases of the musculoskeletal system and connective tissue'. This again indicated that patients getting sepsis were relatively diverse segregating into several major, but largely independent patient sub-groups.

Mortality measure
30-day mortality is the most widely used measure, although some studies use inhospital or 90-day mortality. The in-hospital and 30-day mortality for sepsis patients are similar (24), whereas the 90-day mortality may include late adverse effects of sepsis and related therapies (25). For simplicity, we chose 30-day mortality for this analysis.

Unspecific diagnosis
In our data, unspecific diagnoses like "Other anemias", "Other septicaemia" and "Bacterial pneumonia, not elsewhere classified" were much more frequently used than specific diagnoses like "Anemia in chronic diseases classified elsewhere", "Streptococcal sepsis" or "Pneumonia due to Streptococcus pneumonia". This might be explained by the fact that specific diagnoses often require more and longer patient examination, but it may also be that doctors have to assign diagnoses outside of their main area of expertise.
Analyzing registry data containing 120,000 sepsis patients can lead to extremely low p-values, which proves a significant difference in the mean of the groups. However, it is important to emphasize the importance of the change in sepsis mortality rather than only focusing on the significance level. Sepsis patients with 'anemias in diseases classified elsewhere' or 'other Anemias' had an RR sepsis dead of 1.6 and 1.5 (p-values = 2.96⋅10 -35 and 2.51⋅10 -110 ), respectively. The difference in p-value is uninteresting, whereas the similar change in mortality indicates an equal medical relevance.
We confirmed the pattern of comorbidities found in the Danish data by comparing it to a Swedish comorbidity study, using similar Swedish electronic patient records 8 .