Healthcare Data Security for Beginners

Explore healthcare data security and privacy through key standards and regulations, and learn data analysis basics—distributions, missing values, categorical features, and demographic data set analysis for dataset representativeness.

Explore data security and privacy from patient, provider, and data user perspectives, highlighting electronic health records, breaches, privacy regulations, and the roles of payors and government payers.

Explore how data privacy and security affect patients, illustrated by real breaches, including a payer disclosing an HIV case, and the risks of identity theft and impersonation.

Medical records offer detailed identity data and slower breach notifications, enabling hackers to commit identity theft and file false medical claims or large sums of money.

Medical record data holds higher value for hackers than credit cards or Social Security numbers, making it easier to commit identity theft.

Stay compliant with HIPAA by keeping PHI secure and up to date; avoid common violations like emailing PHI to external accounts, leaving devices unattended, and failing to encrypt data.

Protect data with encryption to keep it from being used by others if a device is lost or stolen.

Explore key regulations and standards governing health care data privacy and security, focusing on U.S. health care regulations for electronic health records.

Compare HIPAA, HITECH, GDPR, and the UK DPA, and learn the core tenets that shape privacy and security of patient data across regions.

Explore how the 1996 legislation and the 2009 high tech act laid the foundation for electronic health records in the US, driving digitization of records to electronic format.

Define PHI, or protected health information, including names, addresses, and electronic medical records. Ensure PHI complies with data security and privacy standards, guiding transmission, storage, and model-building use.

Identify covered entities as health insurance plans, providers, or clearinghouses that transmit protected health information and comply with hip regulations on those transmissions.

Identify business associates as entities that perform functions involving the use or disclosure of protected health information on behalf of or providing services to covered entities.

Understand how business associates disclose data under privacy law for permitted purposes and apply the business associate agreement addendum (BAA) between a government entity and a business associate.

Encrypt PHI data at rest, use a VPN, and secure data transmission to follow best practices in healthcare data security.

Choose a cloud service provider that is HIPAA compliant and approved, and regularly verify their compliance to protect healthcare data.

Covered entities limit access to the minimum necessary level to accomplish the objective, with greater oversight on use and limited latitude on how to use data.

Explore HIPAA de-identification methods for datasets by applying expert determination or Safe Harbor, removing identifiers such as names and addresses to reduce privacy risks while enabling research.

Explore the importance of exploratory data analysis, or EDA, in the data science process, and its role within the cross-industry standard process for data mining.

Explore why EDA is important for uncovering data leakage in features and transformations, saving time, and guiding modeling objectives and evaluation metrics.

Explore value distributions, including the normal bell-curve distribution, labeled as the gasoline distribution, and the uniform distribution. Also examine skewed, unbalanced, and balanced distributions where a subset dominates.

Analyze demographic data to represent the general population in healthcare research, ensuring age, gender, ethnicity, employment, education, and marital status are included for meaningful trials and fair machine learning outcomes.

Explore how data analytics transforms raw data into meaningful insights, using qualitative/quantitative and structured/unstructured data to answer problems with statistical models.

Identify data types by distinguishing quantitative numbers from qualitative attributes, and classify data as structured, semi-structured, or unstructured. Use relational databases and JSON to show fixed versus flexible models.

Identify the three v's of big data—volume, velocity, and variety—and decide when to apply big data tools versus traditional data processing to handle diverse, real-time or historical data.

Explore the evolution of analytics from batch and historical BI to real-time and predictive analytics, using data warehouses and machine learning to forecast behavior and detect threats.

Select the right source data—from data warehouses, relational tables, NoSQL, CSV, or text files—and tailor processing to the analytics problem, then collect, store, analyze, and visualize with cleaning and preprocessing.

Ingest data from diverse sources using AWS tools like Kinesis and Firehose, and securely connect on-premises to the cloud; use SQS to query data for processing and analytics.

Explore AWS storage options for analytics, including S3 for durable data, Redshift for data warehousing, DynamoDB for NoSQL, and Glacier for infrequent access, integrated with analytic services.

Explore AWS big data and analytics to process batch data with EMR and Hadoop, enable real-time insights with Kinesis and Lambda, and visualize with ElasticSearch, Kibana, and data pipelines.

Learn how machine learning, a subfield of artificial intelligence, enables computers to learn automatically from data, improve over time, and use patterns to make informed predictions without ongoing human intervention.

Explore Arthur Samuel's 1959 definition of machine learning as a field of artificial intelligence that enables computers to learn without explicit programming.

Present Tom Mitchell's formal definition of machine learning: a computer program learns from experience with respect to a task and a performance measure, improving its performance with more experience.

Discover how machine learning drives business value through iterative models that continually optimize predictions, with applications in recommendation engines, demand forecasting, computer vision, fraud detection, and medical diagnosis.

Explore foundational machine learning terminology used in healthcare data security, including features, labels, dataset, instance, and sample, and how data items are tagged with classifications.

Explore terminology in the dataset by examining three houses with features such as number of rooms, size, levels, and color, where the label represents the purchase price.

Explore the machine learning lifecycle from dataset collection and feature availability to model training, selecting an algorithm, and testing for accuracy to ensure useful predictions.

Explore the differences between supervised and unsupervised learning, including classification, regression, clustering, and anomaly detection, within healthcare data security training.

Explore supervised learning by training models on a dataset with the correct answers to predict the answers for new data points.

Learn how supervised learning uses an algorithm to analyze past observations, train on labeled data, and map inputs to a predicted output.

Learn how supervised learning uses training data to create mapping functions, distinguishing classifiers for discrete outputs from regression models for continuous outputs, with fraud detection and temperature forecasting as examples.

Explore supervised learning algorithms for classification problems and regression problems, including linear regression, decision trees, support vector machines, and naive Bayes.

Learn to perform supervised training by splitting data into training and testing sets, extracting feature vectors, building a predictive model, evaluating accuracy against labels, and deploying it for new data.

Discover how supervised machine learning answers business problems by selecting training data sources, including home sales, viewing statistics, malignant cancers, and loan default data.

Master supervised training by splitting data into training and test sets (80/20 or 75/25), building models, and evaluating accuracy with linear regression, logistic regression, and support vector machines.

Explore unsupervised learning to train models that uncover patterns in a dataset with no labeled answers. Discover how to use these patterns to help find answers from the data.

Explore unsupervised learning with unlabeled data in healthcare data security contexts, discover patterns through clustering, anomaly detection, and association discovery, and learn how model restructuring supports predictions with new data.

Explore unsupervised training to identify patterns, relationships, and correlations in data, including clustering, anomaly detection, and association discovery across instances and features.

Identify a clustering pattern in travel data, where Mary consistently travels to or from New York on the first of each month at the cost of thirty dollars.

Identify anomalies in healthcare data by spotting unusual instances, such as Mary spending two thousand on entertainment in New York on a day with no travel costs.

Explore association discovery by identifying feature rules that span across instances in a dataset, revealing potential features and associations.

Explore unsupervised learning and clustering concepts with examples like cane means hierarchal and gasoline mixture used for clustering analysis.

Explore unsupervised training in healthcare data security, where models learn from feature vectors without labels, identifying cluster patterns, anomalies, and associations to classify data.

Explore unsupervised learning to solve business problems by leveraging diverse data sources, such as customer profiles, transactions, and purchase history, for fraud detection and product association analysis.

Explore unsupervised training, its key characteristics, how it is used, and example algorithms. The training dataset contains only examples and no specific label or outcome.

Explore unsupervised learning by training on feature vectors without labels, identifying cluster patterns, anomalies, and association words, and using the trained model to classify data insights.