ABPM Clinical Informatics cheat sheet
ABPM Clinical Informatics · Keep the key ideas within reach.
Clinical Decision Support and Evidence-Based Care
Five Rights of CDS: right information, right person, right format, right channel, right time in workflow.
CDS intervention hierarchy (least to most disruptive): passive display, suggestion, on-demand alert, interruptive alert, hard stop/order block.
Alert fatigue metrics: override rate, alert burden per clinician-day, and proportion of alerts acted upon; override rates above ~90% suggest poor specificity.
Grade strength and evidence quality separately (e.g., strong recommendation with moderate-quality evidence); strength reflects net benefit and values, quality reflects study design and bias risk.
Number needed to treat = 1 / absolute risk reduction; absolute risk reduction = control event rate minus intervention event rate.
Sensitivity = TP/(TP+FN); specificity = TN/(TN+FP); PPV depends on prevalence and equals TP/(TP+FP).
Likelihood ratio positive = sensitivity/(1-specificity); LR negative = (1-sensitivity)/specificity.
Rule-based CDS encodes explicit logic; machine-learning CDS requires validation, drift monitoring, and explanation for clinicians.
Health Information Systems and Electronic Health Records
EHR core functions: results management, order entry, clinical documentation, medication management, and clinical decision support.
CPOE reduces medication errors primarily at prescribing; barcode medication administration intervenes at administration.
Closed-loop medication management links ordering, pharmacy verification, dispensing, administration, and monitoring.
Master patient index (MPI) links records across systems; enterprise MPI spans the organization, regional MPI spans organizations.
Data warehouse supports retrospective analytics; operational data store supports near-real-time reporting; clinical data repository supports care delivery.
CPOE with basic CDS shows larger error reduction than CPOE alone; adding advanced CDS increases benefit further.
Downtime procedures require read-only access, paper order forms, and reconciliation steps on system restoration.
HL7 v2 messaging remains the dominant transport for ADT, orders, and results in legacy interfaces.
Data Analytics, Quality Measurement, and Population Health
Quality measure numerator = patients meeting the criterion; denominator = eligible population after exclusions.
HEDIS-style measures are reported as rates; eCQM specifications use value sets (e.g., VSAC) to define populations.
Risk adjustment compares observed to expected outcomes; standardized ratios compare actual to predicted events.
Population health stratification uses risk scores (e.g., HCC, LACE) to target care management resources.
Attribution assigns patients to providers or panels; assignment method affects performance measurement.
Data quality dimensions: completeness, accuracy, timeliness, consistency, and validity.
Sensitivity analysis tests how results change under different assumptions; confidence intervals quantify uncertainty.
Small denominators produce unstable rates; minimum denominator thresholds and confidence intervals should be applied.
Interoperability and Health Information Exchange Standards
HL7 v2 uses pipe-delimited segments; ADT messages track admissions, discharges, and transfers (A01 admit, A03 discharge).
FHIR resources (Patient, Observation, MedicationRequest) are RESTful; profiles constrain base resources for use cases.
C-CDA documents (CCD, discharge summary) are XML-based and use templates and coded value sets.
Terminology: SNOMED CT for clinical findings; LOINC for observations/labs; RxNorm for medications; ICD-10-CM for billing diagnoses.
HIE exchange patterns: directed push, query-based pull (XDS), and event notification (ADT alerts).
US Core Data for Interoperability (USCDI) defines required data classes for certified exchange.
IHE XDS uses document registries and repositories; the registry holds metadata, the repository holds documents.
Consent models for HIE: opt-in, opt-out, and no-consent; each changes data availability and privacy posture.
Privacy, Security, and Regulatory Compliance
HIPAA Privacy Rule governs use/disclosure of PHI; Security Rule governs administrative, physical, and technical safeguards for ePHI.
HIPAA identifiers include 18 categories (e.g., names, dates except year, geographic subdivisions smaller than state, MRNs).
Minimum necessary standard limits PHI use to what is needed for the purpose; treatment disclosures are exempt.
Breach notification: individual notice without unreasonable delay and no later than 60 days after discovery; HHS notified per size thresholds.
Security triad: confidentiality, integrity, availability; controls are administrative, physical, and technical.
Encryption at rest and in transit is an addressable HIPAA safeguard; if not implemented, document rationale.
Role-based access control limits permissions by job function; audit logs support accountability and breach investigation.
42 CFR Part 2 provides stricter protections for substance use disorder records than HIPAA.
Systems Implementation, Usability, and Human Factors
SDLC phases: requirements, design, build, test, deploy, maintain; iterative and agile methods repeat cycles.
Testing types: unit, integration, system, user acceptance, and regression; regression testing verifies unchanged functions after changes.
Usability heuristics include visibility of system status, match to real-world language, user control, consistency, and error prevention.
Nielsen's usability testing: 5 users typically uncover ~85% of usability problems.
Cognitive load theory: working memory is limited; reduce extraneous load in interfaces.
Implementation go-live strategies: big bang versus phased versus pilot; each trades risk against speed.
Usability metrics: task success rate, time on task, error rate, and satisfaction (e.g., SUS score).
Human factors models (e.g., SEIPS) address work system, processes, and outcomes including patient safety.
Leadership, Governance, and Change Management
Governance structures define decision rights, prioritization, and accountability for informatics investments.
Kotter's change model: urgency, coalition, vision, communication, empowerment, short-term wins, consolidation, anchoring.
ADKAR: Awareness, Desire, Knowledge, Ability, Reinforcement; used to diagnose individual change readiness.
Stakeholder analysis maps influence and interest to tailor engagement strategy.
Project management triple constraint: scope, time, cost; quality is affected by trade-offs among them.
ROI and business case include direct costs, indirect costs, and benefits such as safety and efficiency.
Training strategy: super-users, role-based training, and just-in-time support reduce post-go-live disruption.
Risk register documents likelihood, impact, owner, and mitigation for project risks.
Clinical Informatics Fundamentals and Core Concepts
Informatics model: data -> information -> knowledge -> wisdom (DIKW).
Core informatics competencies: clinical, technical, and organizational/leadership.
Data, information, and knowledge differ in structure and context: data are raw, information is processed, knowledge is actionable.
Clinical informatics board certification is jointly sponsored by ABPM and the American Board of Pathology.
Bioinformatics focuses on molecular/genomic data; clinical informatics focuses on care delivery.
Workflow analysis maps tasks, actors, and information flow to identify redesign opportunities.
Interoperability levels: foundational (transport), structural (format), semantic (meaning), organizational (governance).
Cognitive informatics studies how clinicians think and how systems support or hinder decision-making.
Patient Safety, Error Reduction, and Risk Management
Error types: slips (attention failure), lapses (memory failure), mistakes (rule-based or knowledge-based).
Swiss cheese model: failures align across layers of defense to produce harm; barriers reduce alignment.
Just culture distinguishes human error (console), at-risk behavior (coach), and reckless behavior (sanction).
FMEA identifies failure modes, assigns severity, occurrence, and detection ratings, and computes risk priority number (RPN = S x O x D).
RCA is retrospective and examines systems; FMEA is prospective and examines processes.
Adverse event reporting systems are voluntary and underreport events; they complement, not replace, surveillance.
High-alert medications (e.g., anticoagulants, insulin, opioids) warrant independent double checks and smart-pump libraries.
Medication reconciliation at transitions reduces errors; discrepancies include omissions, duplications, and dose changes.
Research Methods, Evaluation, and Informatics Scholarship
Study designs: RCTs test causality; cohort studies estimate incidence; case-control studies estimate association; cross-sectional studies measure prevalence.
Bias types: selection, information, recall, and confounding; randomization reduces confounding.
Qualitative methods (interviews, focus groups, observation, think-aloud) explain why and how; quantitative methods measure what and how much.
Evaluation frameworks: RE-AIM (reach, effectiveness, adoption, implementation, maintenance) and the Quadruple Aim.
Internal validity concerns causal inference; external validity concerns generalizability.
Sample size depends on effect size, alpha, power, and variance; underpowered studies risk type II error.
Publication and reporting standards (CONSORT, STROBE, SQUIRE) improve transparency and reproducibility.
Mixed methods combine quantitative and qualitative data to triangulate findings and strengthen conclusions.