By Eric Andrew Kristof, RN, CAHIMS | Healthcare IT Professional | HIMSS Arkansas Chapter Member
About This Series: The HIMSS Word of the Day is a running blog series drawing from the HIMSS Dictionary of Health Information and Technology Terms, Acronyms, and Organizations, 6th Edition (2025) — the authoritative reference for healthcare IT professionals and the primary study resource for CAHIMS and CPHIMS certification. Each post paraphrases the term in plain language, grounds it in real-world healthcare IT context, and adds a bedside nursing perspective you won’t find in a textbook.
📖 The Term: ANN (Artificial Neural Network)
Source: HIMSS Dictionary of Health Information and Technology Terms, Acronyms, and Organizations, 6th Edition (2025)
🔍 Plain-Language Definition
The HIMSS Dictionary lists ANN as Artificial Neuron Network, a computational model inspired by the structure and function of biological neural networks. Most other sources write it as artificial neural network, and this post uses that form. The Dictionary explains that information flowing through the network changes the network itself: the model adjusts, or in a sense learns, from the inputs and outputs it sees.
The Dictionary describes ANNs as nonlinear statistical modeling tools that either capture complex relationships between inputs and outputs or find patterns, and as one of the main tools in machine learning. In simple terms, an ANN is built from layers of connected nodes. Input data, such as vital signs, lab values, or image pixels, passes through the layers, and the connections are adjusted during training until the output, such as a risk score or a classification, is as accurate as possible.
ANNs differ from traditional regression in being able to capture non-linear relationships, but they are harder to explain. Many modern image tools are a type of ANN called a convolutional neural network.
| Element | Role | Example |
|---|---|---|
| Input layer | Receives the data | Age, vital signs, lab values, image pixels |
| Hidden layers | Combine inputs in non-linear ways | Learned patterns not set by hand |
| Output layer | Produces the result | Risk score or class label |
| Training | Adjusts connections to reduce error | Learning from past patient records |
| Validation | Tests on data not used in training | Held-out patients |
| Explainability | Showing why a result was produced | Heat map over an image |
⚠️ Exam Alert: An artificial neural network (ANN) is a machine learning model made of layers of connected nodes that learns relationships between inputs and outputs from data. It is a nonlinear modeling tool and a statistical one, not a set of hand-written rules. A convolutional neural network (CNN) is a type of ANN used mainly for images. An ANN’s results depend on the quality of training data and must be validated on new patients.
🌐 Why It Matters: Real-World Healthcare IT Context
ANNs appear in prediction models, diagnostic tools, and image analysis. The studies below show where they help and where care is needed.
A 2019 systematic review and meta-analysis in Injury by Hassanipour and colleagues compared ANN and logistic regression models for predicting outcomes in trauma patients. From 326 potentially relevant studies, 10 were included. The pooled area under the curve was 0.91 for ANN and 0.89 for logistic regression, and accuracy was 90.5 percent for ANN and 83.2 percent for logistic regression. The authors concluded that ANN performed better for predicting final outcomes in trauma patients. The gap in AUC is modest, and the studies varied, so the result suggests an advantage and not a guarantee.
A 2021 systematic review in Computational and Mathematical Methods in Medicine by Wei and colleagues compared a backpropagation ANN with linear regression for diagnosing digestive diseases. Nine articles published between 1991 and 2017, with sample sizes from 42 to 3,222 patients, were included. In eight, the ANN was superior based on area under the ROC curve, while one reported linear regression to be better for early diagnosis of colorectal cancer. The authors concluded that both approaches fit well, with the ANN showing higher accuracy. The mixed result is a reminder that a more complex model is not always better.
A 2023 study in Frontiers in Neurology by Yang and colleagues used ANNs to predict prolonged hospital stay after acute ischemic stroke in 2,240 patients at one stroke center, 49.2 percent of whom had a stay longer than the median of 9 days. Fourteen baseline parameters were associated with prolonged stay, and the ANN model reached a validation AUC of 0.788, with accuracy of 74.5 percent. Key factors included stroke severity score at admission, atrial fibrillation, and receipt of thrombolytic therapy. This was a retrospective, single-center analysis, so it needs testing elsewhere.
A 2026 study in Nursing Research and Practice by López-Medina and colleagues shows both promise and a pitfall. Twenty-four stoma nurses at 17 hospitals collected 1,165 images of peristomal skin, which the team used to train convolutional neural networks to tell healthy skin from lesions. With a standard data split, the best model had an accuracy of 0.889 and an AUC of 0.924. When the split was changed so that images from the same patient could not appear in both training and test sets, preventing data leakage, the best results fell to accuracy 0.778 and AUC 0.653. The authors framed the models as preliminary groundwork for automated diagnosis that could support remote care. The drop is a lesson in why test data must be truly independent of training data.
Together, these studies suggest that ANNs can match or beat simpler models in some tasks, but the size of the gain depends on the task, the data, and how carefully the model is tested.
Common ANN problems:
- Training data that is small, biased, or not representative
- Test data that overlaps with training data, which inflates results
- Results that are hard to explain to clinicians and patients
- Models that perform worse at another site or on another population
- Complex models used when a simpler one would do as well
- No plan to monitor performance after launch
What supports responsible use of ANNs:
- Independent test data, including splits by patient
- Comparison with simpler baseline models
- External validation at other sites
- Explanation methods that clinicians can review
- Clinician involvement in choosing outputs and thresholds
- Monitoring of accuracy and fairness over time
🏥 The Nurse’s Perspective: Clinical, Bedside, & Workflow
Nurses contribute data to ANNs, receive their outputs, and in some projects collect the images and measurements used to train them.
At the Bedside
Accurate, timely documentation of vital signs, assessments, and photographs feeds the models that estimate risk. A nurse who understands that a score reflects the quality of the data entered can see why complete and consistent charting matters.
In Clinical Workflow
Nurse informaticists help decide how model outputs should appear, who responds to them, and what the response should be. They also question models that have not been tested on the unit’s own patient population.
In EHR Implementation (Go-Live Perspective)
Prediction scores and alerts are tested with nursing staff before an Epic Go-Live. At-the-elbow support analysts note when a score is ignored, misunderstood, or triggers too often, and those observations help leaders adjust thresholds and training.
🎓 CAHIMS / CPHIMS & HIMSS Perspective
CAHIMS Exam Domain Mapping
ANNs appear in analytics, decision support, and AI governance questions:
| CAHIMS Domain | How ANNs Appear |
|---|---|
| Healthcare & Technology Environments | AI in healthcare, regulation, ethics |
| Clinical Informatics | Prediction models, decision support, image analysis |
| Healthcare Information & Systems Management | Model deployment, data pipelines, monitoring |
| Management & Leadership | Governance of AI tools, evaluation, change management |
Key Terms to Know Alongside ANN
- Machine Learning — methods that let computers learn patterns from data
- Deep Learning — machine learning using neural networks with many layers
- Convolutional Neural Network (CNN) — neural network type used mainly for images
- Logistic Regression — statistical model for predicting a yes or no outcome
- Data Leakage — overlap between training and test data that inflates performance
- AUC — measure of how well a model separates cases from non-cases
HIMSS Organizational Position
HIMSS offers education and research on health information and technology topics through its Resource Center. Members can find related information on the HIMSS Resource Center page.
🔗 Explore Further: External References
- Hassanipour et al. (2019), Injury — Comparison of Artificial Neural Network and Logistic Regression Models for Prediction of Outcomes in Trauma Patients: A Systematic Review and Meta-Analysis https://pubmed.ncbi.nlm.nih.gov/30660332/
- Wei et al. (2021), Computational and Mathematical Methods in Medicine — Comparison of Diagnosis Accuracy between a Backpropagation Artificial Neural Network Model and Linear Regression in Digestive Disease Patients: An Empirical Research https://pubmed.ncbi.nlm.nih.gov/33727951/
- Yang et al. (2023), Frontiers in Neurology — Risk Factor Identification and Prediction Models for Prolonged Length of Stay in Hospital after Acute Ischemic Stroke Using Artificial Neural Networks https://pubmed.ncbi.nlm.nih.gov/36846116/
- López-Medina et al. (2026), Nursing Research and Practice — Identification of Peristomal Skin Alterations Using Convolutional Artificial Neural Networks https://pubmed.ncbi.nlm.nih.gov/42305646/
- HIMSS — Resource Center https://www.himss.org/resource-center
Term source: HIMSS Dictionary of Health Information and Technology Terms, Acronyms, and Organizations, 6th Edition (2025). All definitions in this series are paraphrased for editorial purposes. Readers are encouraged to consult the primary source for exact language.
Eric Andrew Kristof, RN, CAHIMS is a Healthcare IT professional and HIMSS Arkansas Chapter member based in Hot Springs Village, AR, with hands-on Epic Go-Live experience and a diverse healthcare and IT background. He writes at kristof.org at the intersection of clinical care and healthcare technology.
Leave a Reply