Machine Learning in Healthcare (no coding required!)

Name: Machine Learning in Healthcare (no coding required!)
Rating: 4.5 (873 reviews)

Learn how to apply machine learning techniques in healthcare

Created byEddie Jay

Last updated 7/2024

English

English [Auto],French [Auto],

What you'll learn

8 sections • 42 lectures • 2h 45m total length

Course Structure1:32
The slides can be downloaded at Lecture 4 Electronic Health Record section
You will find the dataset in the Assignment in Section6
Case study1:17
To make what you learn here concrete and actionable, we will use a case study - opioid dependency prediction at a clinic.
You will find the training dataset in the Assignment in Section6 of this course.

Health Insurance claims2:09
Electronic health records2:19
Research reports2:57
Public health2:17
Wearables2:40
Explore how wearable devices measure heart rate, blood pressure, pulse rate, and temperature, and how sensitivity and specificity shape false positives and false negatives in personalized medicine.
Insurance Claims
What are strengths of EHR data?
Key takeaways from sources of health data session
Structured data3:11
Explore how structured health care data use tables and relational schemas to organize diagnoses, procedures, drugs, laboratories, and vitals, with ICD-10 and ICD-9, CPT and HCPCS, and NDC codes.
Data structure
Unstructured data2:15
Features of data - Hierarchical Diagnoses4:17
Explore the hierarchical structure of health data with diagnoses, drugs, and procedure codes, and see how ICD-10 and ICD-9 roll up to chapters, enabling faster, aggregated analysis.
Features of data - Hierarchical Drugs2:36
Features of data - Hierarchical Procedures3:51
Examine the hierarchical structure of health care procedure codes, including HCPCS and CPT levels (CPT1 and CPT2), modifiers, and ICD-10-PCS/ICD-9-CM codes to link diagnoses with services.
Features of data - Hierarchical LOINCs and other2:23
Disease Etiology, Chronology, Supply vs Demand4:33
Why understanding of medicine and medical practices enhance analyses?
Errors and error handling4:22
Correcting errors in health data

Steps to Predictive Modeling1:06
7 steps to predictive modeling/Machine learning
Planning the analysis3:57
A solid analytic plan ensures you ask the important questions upfront and keeps you on the right track, improving the efficiency and effectiveness of your analysis.
Getting Buy in for the analysis7:31
Your ability to convince others to work with you, on a project, is a key skill as an analyst. Remember, without end users, what you analyze is theoretical, pointless. Market yourself and your projects like a pro!
Data use considerations1:58
Why plan?
Why do we need buy in for the project?

Initial exploration of the data2:07
as a reminder, you will find the dataset in the Assignment
Data preparation3:44
Feature engineering12:49
Learn how feature engineering turns raw health data into predictive features for modeling. Explore quantitative and domain expertise approaches that improve signal and enable actionable insights.
On features
From raw data to training dataset7:47
Going from raw data to training data set with machine learning features.

Statistical Primer intro0:58
Logistic regression - intro4:29
Logistic regression Pros and Cons5:46
Bias Variance trade off5:38
Bias Variance
Decision tree3:20
Explore decision trees, a supervised non-parametric method that splits data into homogeneous subsets using features, branches, and leaves to predict outcomes efficiently.
When to use which?
Random forest4:20
Neural net4:10
Explore neural nets, a non-parametric technique inspired by biological networks that learn patterns through layered features. Understand their strong predictive power and the healthcare explainability challenges.

Cross validation1:17
Sensitivity vs Specificity3:28
On sensitivity and specificity
Cross validation in KNIME4:38
Sensitivity/Scenario testing4:40
Analyze sensitivity and scenario testing by examining how thresholds alter predicted outcomes and false positives. Compare decision tree and logistic regression on 220 patients to show threshold and false-positive effects.
Presenting the results4:04
Present results by balancing technical and emotive content, use graphics to show impact on patient care and opioid dependency, and document data sources, assumptions, and methodology for a two-page report.
Implementation and Future improvements2:09
Apply logistic regression to new patient data to estimate opiate abuse risk, enabling real-time EHR integration or nurse outreach for high-probability cases.
Practice practice practice!1:04

5 key elements of health insurance products2:58
US Health System series
The 5 key elements of health plan design, that help you see and understand the whys behind each different design
View the complete course + others here ==> https://www.udemy.com/user/yidingjiang/
Intro to the Payors lines of business (health insurers)3:03
US Health System series
View the complete course + others here ==> https://www.udemy.com/user/yidingjiang/
Health data 101 - Health Insurance Claims2:09
Health insurance claims data are a major source of health care data. As long as the the medical services are paid by health insurers, you will see the claims. This is a relative well structured and complete source, although clinical detail can be lacking at times.
View the complete course + others here ==> https://www.udemy.com/user/yidingjiang/

This course will teach you how to work with health data, using machine learning models to find actionable insights.

Through a step-by-step guided case study, you will learn practical skills that you can apply immediately!

We will use a case study: Opioid Abuse Prediction for a clinic

Topics we will cover:

Health Data (sources, types, features, error handling)
Logistics of machine learning
What predictive model features are, and how to create them
A statistical primer, highlighting key machine learning models and concepts
Build a decision tree, logistic regression and random forest through
1. Opioid abuse prediction case study
2. KNIME (a free machine learning software, no coding required!)
Assess model performance
Output presentation and implementation