Disease trajectory browser for exploring temporal, population-wide disease progression patterns in 7.2 million Danish patients.

We present the Danish Disease Trajectory Browser (DTB), a tool for exploring almost 25 years of data from the Danish National Patient Register. In the dataset comprising 7.2 million patients and 122 million admissions, users can identify diagnosis pairs with statistically significant directionality and combine them to linear disease trajectories. Users can search for one or more disease codes (ICD-10 classification) and explore disease progression patterns via an array of functionalities. For example, a set of linear trajectories can be merged into a disease trajectory network displaying the entire multimorbidity spectrum of a disease in a single connected graph. Using data from the Danish Register for Causes of Death mortality is also included. The tool is disease-agnostic across both rare and common diseases and is showcased by exploring multimorbidity in Down syndrome (ICD-10 code Q90) and hypertension (ICD-10 code I10). Finally, we show how search results can be customized and exported from the browser in a format of choice (i.e. JSON, PNG, JPEG and CSV).

Age-stratified longitudinal study of Alzheimer's and vascular dementia patients.

INTRODUCTION: Similar symptoms, comorbidities and suboptimal diagnostic tests make the distinction between different types of dementia difficult, although this is essential for improved work-up and treatment optimization. METHODS: We calculated temporal disease trajectories of earlier multi-morbidities in Alzheimer's disease (AD) dementia and vascular dementia (VaD) patients using the Danish National Patient Registry covering all hospital encounters in Denmark (1994 to 2016). Subsequently, we reduced the comorbidity space dimensionality using a non-linear technique, uniform manifold approximation and projection. RESULTS: We found 49,112 and 24,101 patients that were diagnosed with AD or VaD, respectively. Temporal disease trajectories showed very similar disease patterns before the dementia diagnosis. Stratifying patients by age and reducing the comorbidity space to two dimensions, showed better discrimination between AD and VaD patients in early-onset dementia. DISCUSSION: Similar age-associated comorbidities, the phenomenon of mixed dementia, and misdiagnosis create great challenges in discriminating between classical subtypes of dementia.

Incorporating symptom data in longitudinal disease trajectories for more detailed patient stratification.

OBJECTIVE: Use symptoms to stratify temporal disease trajectories. MATERIALS AND METHODS: We use data from the Danish National Patient Registry to stratify temporal disease pairs by the symptom distributions they associate to. The underlying data comprise of 6.6 million patients collectively assigned with 7.5 million symptoms from chapter XVIII in the WHO International Classification of Disease version 10 terminology. RESULTS: We stratify 33 disease pairs into 67 temporal disease-symptom-disease trajectories from three main diagnoses (two diabetes subtypes and COPD), where the symptom significantly changes the risk of developing the subsequent diseases. We combine these trajectories into three temporal disease networks, one for each main diagnosis. We confirm apparent relations between diseases and symptoms and discovered that multiple symptoms decrease the risk for diabetes progression. CONCLUSION: Symptoms can be used to stratify disease trajectories, and we suggest that this approach can be applied to temporal disease trajectories systematically using structured claims data. The method can be extended to also use text-mined symptoms from unstructured data in health records.

Comorbidity landscape of the Danish patient population affected by chromosome abnormalities.

PURPOSE: Most chromosome abnormality patients require long-term clinical care. Awareness of mosaicism and comorbidities can potentially guide such health care. Here we present a population-wide analysis of direct and inverse comorbidities affecting patients with chromosome abnormalities. METHODS: We extracted direct and inverse comorbidities for the 11 most prevalent chromosome abnormalities from the Danish National Patient Registry (covering 6.9 million patients hospitalized between 1994 and 2015): trisomy 13, 18, and 21, Klinefelter (47,XXY), triple X, XYY, Turner (45,X), Wolf-Hirschhorn, Cri-du-chat, Angelman, and Fragile X syndromes (FXS). We also performed four sub-analyses for male/female Down syndrome (DS) and FXS and non-mosaic/mosaic DS and Turner syndrome. RESULTS: Our data cover 9,003 patients diagnosed with at least one chromosome abnormality. Each abnormality showed a unique comorbidity signature, but clustering of their profiles underlined common risk profiles for chromosome abnormalities with similar genetic backgrounds. We found that DS had a decreased risk for three inverse cancer comorbidities (lung, breast, and skin) and that male FXS and non-mosaic patients have a much more severe phenotype than female FXS and mosaic patients, respectively. CONCLUSION: Our study underlines the importance of considering mosaicism, sex, and the associated comorbidity profiles of chromosome abnormalities to guide long-term health care of affected patients.

Survival prediction in intensive-care units based on aggregation of long-term disease history and acute physiology: a retrospective study of the Danish National Patient Registry and electronic patient records.

BACKGROUND: Intensive-care units (ICUs) treat the most critically ill patients, which is complicated by the heterogeneity of the diseases that they encounter. Severity scores based mainly on acute physiology measures collected at ICU admission are used to predict mortality, but are non-specific, and predictions for individual patients can be inaccurate. We investigated whether inclusion of long-term disease history before ICU admission improves mortality predictions. METHODS: Registry data for long-term disease histories for more than 230 000 Danish ICU patients were used in a neural network to develop an ICU mortality prediction model. Long-term disease histories and acute physiology measures were aggregated to predict mortality risk for patients for whom both registry and ICU electronic patient record data were available. We compared mortality predictions with admission scores on the Simplified Acute Physiology Score (SAPS) II, the Acute Physiologic Assessment and Chronic Health Evaluation (APACHE) II, and the best available multimorbidity score, the Multimorbidity Index. An external validation set from an additional hospital was acquired after model construction to confirm the validity of our model. During initial model development data were split into a training set (85%) and an independent test set (15%), and a five-fold cross-validation was done during training to avoid overfitting. Neural networks were trained for datasets with disease history of 1 month, 3 months, 6 months, 1 year, 2·5 years, 5 years, 7·5 years, 10 years, and 23 years before ICU admission. FINDINGS: Mortality predictions with a model based solely on disease history outperformed the Multimorbidity Index (Matthews correlation coefficient 0·265 vs 0·065), and performed similarly to SAPS II and APACHE II (Matthews correlation coefficient with disease history, age, and sex 0·326 vs 0·347 and 0·300 for SAPS II and APACHE II, respectively). Diagnoses up to 10 years before ICU admission affected current mortality prediction. Aggregation of previous disease history and acute physiology measures in a neural network yielded the most precise predictions of in-hospital mortality (Matthews correlation coefficient 0·391 for in-hospital mortality compared with 0·347 with SAPS II and 0·300 with APACHE II). These results for the aggregated model were validated in an external independent dataset of 1528 patients (Matthews correlation coefficient for prediction of in-hospital mortality 0·341). INTERPRETATION: Longitudinal disease-spectrum-wide data available before ICU admission are useful for mortality prediction. Disease history can be used to differentiate mortality risk between patients with similar vital signs with more precision than SAPS II and APACHE II scores. Machine learning models can be deconvoluted to generate novel understandings of how ICU patient features from long-term and short-term events interact with each other. Explainable machine learning models are key in clinical settings, and our results emphasise how to progress towards the transformation of advanced models into actionable, transparent, and trustworthy clinical tools. FUNDING: Novo Nordisk Foundation and Innovation Fund Denmark.