Types of Data Analysis: Real-World Applications in Tackling Drug-Resistant Tuberculosis

Why Data Analysis Matters Beyond the Buzzwords

Data analytics is crucial to the decision-making process for businesses, researchers, state-owned enterprises, and many other organisations that use data. In previous weeks, we looked at the importance of defining the problem and formulating your questions. Questions guide you on the type(s) of analysis you need to do to gain the insights

In this blog, we will look at one study’s problem definitions to show how different types of analysis might look in practice. We will focus in particular on the drug resistance problem in tuberculosis (TB) to show how various types of analyses play a role in addressing the study’s problem.

Study

Drug-resistant tuberculosis is a global challenge, especially in the public health sector. Southern Africa has some of the highest reported cases of drug resistance in the world. Rural clinics often struggle to detect and manage multidrug-resistant TB (MDR-TB). Data analysis can shed light on the scale of the problem, its underlying causes and possible solutions. (Keep in mind, yes, this is a real problem;  however, we are using this scenario to illustrate the different types of analysis.)

Illustration of a tuberculosis-affected human torso highlighting the lungs, with visible lesions typically associated with TB infection.

Defined Problem: The prevalence of multidrug-resistant TB among patients receiving first-line treatment in rural clinics in Southern Africa remains poorly quantified.

Goal/Aim: Quantify and understand the prevalence of MDR-TB in rural clinics.

Descriptive Analysis: What is Happening

Question: What proportion of patients receiving first-line TB treatment in rural clinics have MDR-TB?

Application: Descriptive analysis examines patient records, lab test results, and treatment data to determine prevalence rates.

Example:

• 15% of patients in rural clinics tested positive for MDR-TB

• Resistance rates vary by clinic: 8% in one, 18% in another.

Insight: Descriptive analysis quantifies the problem, showing how widespread MDR really is.

Diagnostic Analysis: Why is it happening?

Question: Are certain age groups, genders, or comorbidities linked to higher resistance?

Application: Diagnostic analysis digs further into patterns and associations

• Patients with HIV co-infection showed a 1.7x higher odds of MDR-TV

• Younger patients (18-25) have a lower adherence rate, correlating to resistance.

  
  

Insight: This type of analysis identifies risk factors and explains why certain groups are more vulnerable.

Predictive Analysis: What Could Happen Next?

Question: Can we predict which patients are at risk of developing resistance?

Application: Using historical clinical data, machine learning methods (logistic regression or random forest) to predict future MDR cases based on factors such as:

• Age, comorbidities, prior treatment history, and treatment adherence patterns

• The model predicts that 30% of patients with poor adherence and HIC co-infection may develop MDR-TB within the next treatment cycle.

Insight: Predictive analysis forecasts who is more likely to develop resistance so that earlier interventions are made.

Prescriptive analysis: What should we do about it?

Question: How can we improve outcomes and reduce resistance?

Application: This type of analysis uses optimisation and simulation to recommend interventions:

• Increase adherence monitoring for high-risk groups

• Allocate more diagnostic resources to clinics that have the highest predicted resistance rates

• Test different policies, such as a weekly or monthly check-up, to determine which one is the most effective approach.

Insight: Prescriptive analysis answers, what action will most effectively reduce MDR-TB prevalence

Exploratory Analysis: What we Don’t Know Yet?

Question: Are there hidden patterns in the data that we haven’t considered?

Application: Exploratory analysis uncovers  unexpected relationships:

• Patients in certain rural regions showed high levels of resistance unrelated to demographics

• Preliminary trends suggest seasonal patterns in treatment in treatment interruptions

Insight: Exploratory Analysis involves discovering new questions we didn’t know to ask.

Summary Table

Type of Analysis	Key Question	Purpose/Application	Example Insight
Descriptive Analysis	What is happening?	Quantify prevalence from data	15% of patients tested positive for MDR-TB
Diagnostic Analysis	Why is it happening?	Identify risk factors and associations	HIV co-infection raises MDR-TB risk by 1.7 times
Predictive Analysis	What could happen next?	Forecast future cases using historical data	30% with poor adherence and HIV may develop MDR-TB next cycle
Prescriptive Analysis	What should we do about it?	Recommend interventions through optimisation	Increase adherence monitoring and resource allocation
Exploratory Analysis	What don’t we know yet?	Uncover hidden patterns and new questions	Seasonal patterns and regional resistance unrelated to demographics

Why is this important to know?

Knowing the types of analysis is important because it guides the methods you use when conducting each type. Ultimately, this provides a clear outline of what you need to do, how to do it, and why.

As you may have noticed, a single analysis can involve a combination of multiple types of analysis. With exploratory analysis, new questions arise. This links us back to question formulation; it is an iterative process. As you do these analyses, you may identify new problems and questions.

Further Reading:

1. "Data Science for Business" by Foster Provost and Tom Fawcett

2. "An Introduction to Statistical Learning" by Gareth James, Daniela Witten, Trevor Hastie, and Robert Tibshirani

3. World Health Organization (WHO) Reports on Tuberculosis and Drug Resistance

4. "Practical Statistics for Data Scientists" by Peter Bruce and Andrew Bruce

5. "Applied Predictive Modeling" by Max Kuhn and Kjell Johnson