20–21 Nov 2026
Online Conference
Asia/Karachi timezone

A Co-occurrence Analysis of Comorbidities Using Cosine Similarity on MIMIC-IV Diagnosis Data

20 Nov 2026, 21:10
5m
Online Conference

Online Conference

Online Conference

Speaker

Safiya Arif

Result

Three groups of conditions that go together:
Cardiometabolic: Hypertension, High cholesterol, Diabetes, Tobacco use and Anemia. Hypertension with High cholesterol is the strongest pair (about 0.7).
Kidney and heart: Acute kidney failure, CKD, Heart failure and Anemia. Acute kidney failure with CKD and with Anemia are both around 0.65.
Acute or critical illness: Sepsis, Acidosis, Thrombocytopenia, Pneumonia and Acute kidney failure. Sepsis with Acidosis is about 0.6.

Aim & Purpose

To identify what diseases occur together so that we can potentially prevent or discourage the other diseases affiliated with it,

Discussion (methods + interpretation)

Data source: Johnson, A., Bulgarelli, L., Pollard, T., Horng, S., Celi, L. A., & Mark, R. (2023). MIMIC-IV Clinical Database Demo (version 2.2). PhysioNet. RRID:SCR_007345. https://doi.org/10.13026/dp1f-ex47 . Method: Codes were joined
to their long titles and consolidated into short condition names, yielding 1,310 distinct
conditions. Each patient was encoded as a binary vector, and conditions present in fewer than 5% of patients were excluded. Pairwise co-occurrence counts were computed from the patient-by-condition
matrix and normalized with cosine similarity to reduce the influence of prevalence.

Background

Application of linear algebra to patient records of a hospital to find the co-occurrences of diseases and identifying the pattern of certain diseases co-occurring together in populations.

Email address 24G-BCS026@agu.edu.pk

Author

Co-author

Presentation materials