Uncovering Sepsis Through Data: A Tableau Project in Healthcare Analytics

Sepsis is a life-threatening condition caused by the body's extreme response to infection. It can lead to tissue damage, organ failure, and death if not recognized early. It kills more people annually than many other well-known diseases. Early detection is critical - every hour of delay in treatment increases mortality risk by 7-10%.

As part of an 8-team collaborative analytics project, we analyzed a comprehensive dataset of over 1 million records from 40,336 patients (2,932 with sepsis and 37,404 without) to uncover patterns that could improve sepsis detection and treatment. Our goal was to build a consolidated Tableau dashboard revealing critical insights into sepsis onset, progression, and patient outcomes.

Project Structure: A Sprint-Based Approach

The entire project was broken into 6 sprints, each focusing on a critical aspect of sepsis:

• Biomarker Analysis

• Demographics , Sepsis Onset, ICU Type and Patient Volume

• Correlation of Key Biomarkers

• Organ Function Analysis ( Heart, Kidney, Lungs, Liver)

• Systemic Inflammatory Response Syndrome (SIRS) Criteria

• APACHE, MOD, SOFA, ABG , Septic Shock - Analysis

My Contributions: Core Analytics Work

As part of a 4-member team, I led analysis in several critical areas throughout the sepsis analytics project. I focused on biomarker pattern analysis, patient demographic studies, and clinical scoring system validation across multiple sprints. My work involved analyzing large healthcare datasets, identifying key correlations between vital signs, and developing insights into sepsis progression patterns that contributed to the team's overall findings.

Sprint 1: Biomarker Foundation Analysis

Overview

This sprint established foundational understanding of critical biomarkers in sepsis progression. The team analyzed vital signs patterns, normal physiological ranges, and sepsis-induced deviations across multiple organ systems, focusing on EtCO2, O2Sat, HR, MAP, temperature, and blood pressure measurements.

Analysis

• Investigated vital signs including End-Tidal Carbon Dioxide (EtCO2), Oxygen Saturation (O2Sat), Heart Rate (HR), Mean Arterial Pressure (MAP), Temperature (Temp), Diastolic Blood Pressure (DBP), and Systolic Blood Pressure (SBP).

• Examined normal ranges and deviations caused by sepsis.

• Assessed how changes in these biomarkers affect organs such as the brain, heart, lungs, kidneys, liver, and gastrointestinal tract.

  
  

Key Insights

• EtCO2 falls below 25 mmHg in sepsis due to poor tissue perfusion and metabolic acidosis, impacting overall metabolism.

• O2Sat decreases with sepsis-induced hypoxia, leading to tissue damage; excess oxygen can worsen injury.

• Elevated HR and low MAP increase the risk of organ failure.

• Fever (>38°C) or hypothermia (<36°C) are strong sepsis markers; low DBP and SBP reduce blood flow, risking organ damage.

Sprint 2: Patient Volume & Demographics Analysis

Overview

This sprint provided comprehensive analysis of patient demographics, admission timing patterns, ICU type distributions, and healthcare resource utilization. The focus included hourly admission tracking across MICU and SICU units, sepsis incidence mapping, and healthcare load assessment.

Analysis: 

• Tracked patient admissions by hour in Medical ICU (MICU) and Surgical ICU (SICU).

• Compared sepsis vs. non-sepsis patient volumes over time.

• Evaluated admission sources including direct ICU admission, hospital-to-ICU transfers, and ICU-to-hospital discharges.

  
  

Key Insights

• Peak patient admissions were identified by hour.

• Sepsis patients were predominantly admitted during specific high-volume periods.

• The majority of sepsis cases originated from ICU - hospital transfers during hospitalization.

Sprint 3: Biomarker Correlation Analysis

Overview

This sprint explored complex interrelationships between vital biomarkers to understand sepsis severity progression. Advanced statistical correlation analysis was conducted using sophisticated visualization methods including scatterplots, bubble charts, and Sankey diagrams to reveal predictive patterns.

Analysis

• Correlated Temperature, HR, Respiratory Rate, EtCO2, MAP, O2Sat, SBP, and DBP.

• Assessed variations by age and gender.

• Used scatterplots, bar graphs, and Sankey diagrams for visualization.

  
  

Key Insights

• Respiratory rate increases with severity, negatively correlates with EtCO2, and positively correlates with HR and temperature.

• Low EtCO2 signals worsening sepsis and correlates positively with O2Sat.

• HR shows moderate correlation with other vitals and varies by age and gender.

• MAP strongly correlates with SBP and DBP; lower MAP indicates increased sepsis severity.

• Age and gender influence biomarker patterns, with elderly patients and males exhibiting more severe deviations.

Sprint 4: Kidney Analysis

Overview

This sprint investigated organ dysfunction patterns across all major systems affected by sepsis progression. The analysis covered heart, kidney, lung, and liver function markers with detailed examination of organ-specific biomarkers. Our team specifically focused on comprehensive kidney function analysis, tracking creatinine, bicarbonate, and pH markers over time to understand acute kidney injury progression.

Analysis

• Explored the trends of key kidney function markers including Creatinine, Bicarbonate (HCO3), and pH over time.

• Examined the distribution and severity of kidney dysfunction (Moderate and Severe) across various age ranges.

• Visualized the data using line charts to track trends and tree maps to show the distribution of dysfunction severity by age group.

  
  

Key Insights

• Creatinine, Bicarbonate, and pH fluctuate widely, indicating unstable kidney function.

• Moderate kidney dysfunction is most prevalent in patients aged 71-80 and 61-70.

• Severe kidney dysfunction also peaks in older age groups, confirming that advanced age is a major risk factor.

• Younger patients show fewer cases of kidney dysfunction and less severe outcomes.

Sprint 5: SIRS Analysis

Overview

This sprint validated the effectiveness of SIRS identification criteria across diverse patient populations. Comprehensive validation was conducted across 40,336 patient records, analyzing biomarker trend patterns and onset timing to assess SIRS criteria reliability in sepsis detection.

Analysis

• Reviewed 40,336 patient records to classify SIRS-positive (≥2 criteria) and non-SIRS patients. • Analyzed distribution by age groups and ICU type (MICU vs SICU).

• Monitored biomarker trends and time to SIRS onset.

• Examined transition hours of biomarker abnormalities across age groups.

  
  

Key Insights

• 24,321 patients were classified as SIRS-positive; the highest incidence was in the 61-70 age group.

• Earlier SIRS onset was observed in SICU patients.

• Biomarker trends indicate distinct abnormality patterns, with the highest transition activity in the 61-80 age group.

• Visualization tools demonstrated patient flow through SIRS stages and age-related differences in biomarker activity.

Sprint 6: Clinical Scoring System Analysis

Overview

This sprint integrated multiple clinical scoring systems to create comprehensive predictive models for sepsis outcomes. The analysis incorporated APACHE, MOD, and SOFA scoring systems alongside ABG analysis and septic shock indicators to enhance mortality risk prediction and ICU length of stay forecasting. Our team specifically concentrated on APACHE and MOD scoring system integration to develop enhanced predictive analytics for patient outcomes.

Analysis

• Assessed mortality risk and ICU length of stay (LOS) relative to APACHE and MOD scores.

• Analyzed organ dysfunction prevalence across systems (heart, kidney, liver, lung).

• Studied demographic distribution and biomarker patterns by scoring categories.

  
  

Key Insights

• Mortality rates increase with higher APACHE scores; ICU LOS positively correlates with score severity.

• Key high-risk biomarkers identified include Creatinine, Heart Rate, MAP, Oxygenation, pH, Potassium, Respiratory Rate, Temperature, and WBC.

• The majority of sepsis patients exhibit organ dysfunction, predominantly affecting the heart and kidneys.

• MOD scores peak in the first ICU week, with mortality risk declining over time.

• The age group 41-70 is most affected; gender distribution is relatively balanced across APACHE categories.

Final Presentation

• Team Consolidation: All eight teams consolidated their specialized sprint analyses into a comprehensive, unified Tableau workbook with multiple storyboards, representing months of intensive healthcare analytics work with cross-team validation

• 12-Dashboard Analytics System: Created an integrated sepsis system featuring Demographics and Sepsis Overview dashboards, four Organ Function dashboards (Lungs, Kidney, Liver, Heart), and five Medical Scoring Systems (SIRS, Septic Shock, MOD, SOFA, ABG)

• Comprehensive Data Integration: Successfully integrated patient records from 8 research teams, ensuring consistency in findings and eliminating analytical contradictions across all team contributions

• Interactive Framework: Each dashboard provided interactive controls and real-world medical insights, offering view of sepsis from biomarkers to outcomes with standardized visualization templates.

Key Learning Outcomes

Domain Expertise Gained

• Deep understanding of sepsis pathophysiology and progression

• Experience with clinical scoring systems (APACHE, MOD, SOFA, SIRS)

• Healthcare data quality challenges and solutions

• Translation of analytics into clinically actionable insights

Technical Skills Development

• Advanced Tableau visualization techniques for healthcare data

• Multi-team project coordination and data integration

• Statistical analysis of complex datasets

• Clinical data interpretation and validation methods

Project Management Experience

• Sprint-based healthcare analytics methodology

• Cross-functional team leadership and coordination

Final Thoughts

This project was a true demonstration of how data analytics can save lives. With a collaborative approach, clinical understanding, and visualization tools, we transformed a complex dataset into actionable sepsis insights. I'm proud of what my team contributed and the role I played in consolidating the final dashboard.