Healthcare Analytics Made Simple电子书

Title Page

Copyright and Credits

Healthcare Analytics Made Simple

Dedication

Packt Upsell

Why subscribe?

PacktPub.com

Foreword

Contributors

About the author

About the reviewer

Packt is searching for authors like you

Preface

Who this book is for

What this book covers

To get the most out of this book

Download the example code files

Download the color images

Conventions used

Get in touch

Reviews

Introduction to Healthcare Analytics

What is healthcare analytics?

Healthcare analytics uses advanced computing technology

Healthcare analytics acts on the healthcare industry (DUH!)

Healthcare analytics improves medical care

Better outcomes

Lower costs

Ensure quality

Foundations of healthcare analytics

Healthcare

Mathematics

Computer science

History of healthcare analytics

Examples of healthcare analytics

Using visualizations to elucidate patient care

Predicting future diagnostic and treatment events

Measuring provider quality and performance

Patient-facing treatments for disease

Exploring the software

Anaconda

Anaconda navigator

Jupyter notebook

Spyder IDE

SQLite

Command-line tools

Installing a text editor

Summary

References

Healthcare Foundations

Healthcare delivery in the US

Healthcare industry basics

Healthcare financing

Fee-for-service reimbursement

Value-based care

Healthcare policy

Protecting patient privacy and patient rights

Advancing the adoption of electronic medical records

Promoting value-based care

Advancing analytics in healthcare

Patient data – the journey from patient to computer

The history and physical (H&P)

Metadata and chief complaint

History of the present illness (HPI)

Past medical history

Medications

Family history

Social history

Allergies

Review of systems

Physical examination

Additional objective data (lab tests, imaging, and other diagnostic tests)

Assessment and plan

The progress (SOAP) clinical note

Standardized clinical codesets

International Classification of Disease (ICD)

Current Procedural Terminology (CPT)

Logical Observation Identifiers Names and Codes (LOINC)

National Drug Code (NDC)

Systematized Nomenclature of Medicine Clinical Terms (SNOMED-CT)

Breaking down healthcare analytics

Population

Medical task

Screening

Diagnosis

Outcome/Prognosis

Response to treatment

Data format

Structured

Unstructured

Imaging

Other data format

Disease

Acute versus chronic diseases

Cancer

Other diseases

Putting it all together – specifying a use case

Summary

References and further reading

Machine Learning Foundations

Model frameworks for medical decision making

Tree-like reasoning

Categorical reasoning with algorithms and trees

Corresponding machine learning algorithms – decision tree and random forest

Probabilistic reasoning and Bayes theorem

Using Bayes theorem for calculating clinical probabilities

Calculating the baseline MI probability

2 x 2 contingency table for chest pain and myocardial infarction

Interpreting the contingency table and calculating sensitivity and specificity

Calculating likelihood ratios for chest pain (+ and -)

Calculating the post-test probability of MI given the presence of chest pain

Corresponding machine learning algorithm – the Naive Bayes Classifier

Criterion tables and the weighted sum approach

Criterion tables

Corresponding machine learning algorithms – linear and logistic regression

Pattern association and neural networks

Complex clinical reasoning

Corresponding machine learning algorithm – neural networks and deep learning

Machine learning pipeline

Loading the data

Cleaning and preprocessing the data

Aggregating data

Parsing data

Converting types

Dealing with missing data

Exploring and visualizing the data

Selecting features

Training the model parameters

Evaluating model performance

Sensitivity (Sn)

Specificity (Sp)

Positive predictive value (PPV)

Negative predictive value (NPV)

False-positive rate (FPR)

Accuracy (Acc)

Receiver operating characteristic (ROC) curves

Precision-recall curves

Continuously valued target variables

Summary

References and further reading

Computing Foundations – Databases

Introduction to databases

Data engineering with SQL – an example case

Case details – predicting mortality for a cardiology practice

The clinical database

The PATIENT table

The VISIT table

The MEDICATIONS table

The LABS table

The VITALS table

The MORT table

Starting an SQLite session

Data engineering, one table at a time with SQL

Query Set #0 – creating the six tables

Query Set #0a – creating the PATIENT table

Query Set #0b – creating the VISIT table

Query Set #0c – creating the MEDICATIONS table

Query Set #0d – creating the LABS table

Query Set #0e – creating the VITALS table

Query Set #0f – creating the MORT table

Query Set #0g – displaying our tables

Query Set #1 – creating the MORT_FINAL table

Query Set #2 – adding columns to MORT_FINAL

Query Set #2a – adding columns using ALTER TABLE

Query Set #2b – adding columns using JOIN

Query Set #3 – date manipulation – calculating age

Query Set #4 – binning and aggregating diagnoses

Query Set #4a – binning diagnoses for CHF

Query Set #4b – binning diagnoses for other diseases

Query Set #4c – aggregating cardiac diagnoses using SUM

Query Set #4d – aggregating cardiac diagnoses using COUNT

Query Set #5 – counting medications

Query Set #6 – binning abnormal lab results

Query Set #7 – imputing missing variables

Query Set #7a – imputing missing temperature values using normal-range imputation

Query Set #7b – imputing missing temperature values using mean imputation

Query Set #7c – imputing missing BNP values using a uniform distribution

Query Set #8 – adding the target variable

Query Set #9 – visualizing the MORT_FINAL_2 table

Summary

References and further reading

Computing Foundations – Introduction to Python

Variables and types

Strings

Numeric types

Data structures and containers

Lists

Tuples

Dictionaries

Sets

Programming in Python – an illustrative example

Introduction to pandas

What is a pandas DataFrame?

Importing data

Importing data into pandas from Python data structures

Importing data into pandas from a flat file

Importing data into pandas from a database

Common operations on DataFrames

Adding columns

Adding blank or user-initialized columns

Adding new columns by transforming existing columns

Dropping columns

Applying functions to multiple columns

Combining DataFrames

Converting DataFrame columns to lists

Getting and setting DataFrame values

Getting/setting values using label-based indexing with loc

Getting/setting values using integer-based labeling with iloc

Getting/setting multiple contiguous values using slicing

Fast getting/setting of scalar values using at and iat

Other operations

Filtering rows using Boolean indexing

Sorting rows

SQL-like operations

Getting aggregate row COUNTs

Joining DataFrames

Introduction to scikit-learn

Sample data

Data preprocessing

One-hot encoding of categorical variables

Scaling and centering

Binarization

Imputation

Feature-selection

Machine learning algorithms

Generalized linear models

Ensemble methods

Additional machine learning algorithms

Performance assessment

Additional analytics libraries

NumPy and SciPy

matplotlib

Summary

Measuring Healthcare Quality

Introduction to healthcare measures

US Medicare value-based programs

The Hospital Value-Based Purchasing (HVBP) program

Domains and measures

The clinical care domain

The patient- and caregiver-centered experience of care domain

Safety domain

Efficiency and cost reduction domain

The Hospital Readmission Reduction (HRR) program

The Hospital-Acquired Conditions (HAC) program

The healthcare-acquired infections domain

The patient safety domain

The End-Stage Renal Disease (ESRD) quality incentive program

The Skilled Nursing Facility Value-Based Program (SNFVBP)

The Home Health Value-Based Program (HHVBP)

The Merit-Based Incentive Payment System (MIPS)

Quality

Advancing care information

Improvement activities

Cost

Other value-based programs

The Healthcare Effectiveness Data and Information Set (HEDIS)

State measures

Comparing dialysis facilities using Python

Downloading the data

Importing the data into your Jupyter Notebook session

Exploring the data rows and columns

Exploring the data geographically

Displaying dialysis centers based on total performance

Alternative analyses of dialysis centers

Comparing hospitals

Downloading the data

Importing the data into your Jupyter Notebook session

Exploring the tables

Merging the HVBP tables

Summary

References

Making Predictive Models in Healthcare

Introduction to predictive analytics in healthcare

Our modeling task – predicting discharge statuses for ED patients

Obtaining the dataset

The NHAMCS dataset at a glance

Downloading the NHAMCS data

Downloading the ED2013 file

Downloading the list of survey items – body_namcsopd.pdf

Downloading the documentation file – doc13_ed.pdf

Starting a Jupyter session

Importing the dataset

Loading the metadata

Loading the ED dataset

Making the response variable

Splitting the data into train and test sets

Preprocessing the predictor variables

Visit information

Month

Day of the week

Arrival time

Wait time

Other visit information

Demographic variables

Age

Sex

Ethnicity and race

Other demographic information

Triage variables

Financial variables

Vital signs

Temperature

Pulse

Respiratory rate

Blood pressure

Oxygen saturation

Pain level

Reason-for-visit codes

Injury codes

Diagnostic codes

Medical history

Tests

Procedures

Medication codes

Provider information

Disposition information

Imputed columns

Identifying variables

Electronic medical record status columns

Detailed medication information

Miscellaneous information

Final preprocessing steps

One-hot encoding

Numeric conversion

NumPy array conversion

Building the models

Logistic regression

Random forest

Neural network

Using the models to make predictions

Improving our models

Summary

References and further reading

Healthcare Predictive Models – A Review

Predictive healthcare analytics – state of the art

Overall cardiovascular risk

The Framingham Risk Score

Cardiovascular risk and machine learning

Congestive heart failure

Diagnosing CHF

CHF detection with machine learning

Other applications of machine learning in CHF

Cancer

What is cancer?

ML applications for cancer

Important features of cancer

Routine clinical data

Cancer-specific clinical data

Imaging data

Genomic data

Proteomic data

An example – breast cancer prediction

Traditional screening of breast cancer

Breast cancer screening and machine learning

Readmission prediction

LACE and HOSPITAL scores

Readmission modeling

Other conditions and events

Summary

References and further reading

The Future – Healthcare and Emerging Technologies

Healthcare analytics and the internet

Healthcare and the Internet of Things

Healthcare analytics and social media

Influenza surveillance and forecasting

Predicting suicidality with machine learning

Healthcare and deep learning

What is deep learning, briefly?

Deep learning in healthcare

Deep feed-forward networks

Convolutional neural networks for images

Recurrent neural networks for sequences

Obstacles, ethical issues, and limitations

Obstacles

Ethical issues

Limitations

Conclusion of this book

References and further reading

Other Books You May Enjoy

Leave a review - let other readers know what you think