Skip to main content

Interpreting Results

This guide will help you understand and effectively utilize the differential diagnosis results returned by the Knidian AI API.

Understanding the Response Structure​

The API response contains two main sections:

  1. Diagnosis: Contains the differential diagnosis with multiple potential diseases
  2. Triage: Provides urgency assessment and recommendations

Let's explore each section in detail.

Diagnosis Section​

The diagnosis section contains an array of diseases, each representing a potential diagnosis based on the provided clinical history.

Disease Categories​

Diseases are categorized into three groups:

  1. Most-Likely ("category": "most-likely"): Diagnoses with the highest probability based on the clinical presentation. These should be considered primary diagnostic considerations.

  2. Expanded ("category": "expanded"): Less likely but still possible diagnoses that should be considered if the most likely diagnoses are ruled out.

  3. Can't-Miss ("category": "cant-miss"): Critical conditions that require immediate consideration due to their potentially serious or life-threatening nature, even if their probability is lower.

Probability Scores​

Each disease includes a probability score (probability) ranging from 0.00 to 1.00:

  • High probability (0.70-1.00): Strong evidence supporting this diagnosis
  • Moderate probability (0.40-0.69): Reasonable evidence supporting this diagnosis
  • Low probability (0.00-0.39): Limited evidence supporting this diagnosis

Note: Probability scores are relative within the context of the provided clinical history and should not be interpreted as absolute diagnostic certainty.

Diagnostic Reasoning​

The why field for each disease provides the reasoning behind its inclusion in the differential diagnosis. This explanation includes:

  • Which symptoms and findings support the diagnosis
  • Relevant risk factors
  • Why this diagnosis fits the clinical picture

This information is valuable for understanding the AI's reasoning process and can help clinicians evaluate the validity of each suggested diagnosis.

Diagnostic Tests​

The diagnostic_tests array lists recommended tests to confirm or rule out each diagnosis. Tests are generally listed in order of priority, with more critical or cost-effective tests appearing first.

Consider these factors when evaluating recommended tests:

  • Clinical urgency
  • Test availability
  • Patient-specific factors
  • Cost-effectiveness
  • Invasiveness

Standard of Care​

The standard_care array provides evidence-based interventions for each diagnosis, often including:

  • Clinical practice guidelines
  • Treatment protocols
  • Management recommendations

Each standard of care item may include a URL linking to the source of the recommendation, allowing for further exploration of the evidence base.

Medications​

When applicable, the drugs array lists recommended medications for each diagnosis, including:

  • Medication name
  • Drug class
  • General dosage information
  • Important considerations
  • Whether it's considered a first-line treatment

Important: Medication recommendations are for informational purposes only and should not replace clinical judgment. Always consider patient-specific factors, contraindications, and current prescribing information.

Triage Section​

The triage section provides an assessment of the urgency of the case and recommendations for the appropriate level of care.

Triage Levels​

There are two possible triage levels:

  1. RED: Urgent cases requiring immediate medical attention
  2. WHITE: Non-urgent cases that can be managed in a routine clinical setting

Triage Justification​

The triage_justification field explains why the particular triage level was assigned, referencing specific symptoms, vital signs, or risk factors that influenced the urgency assessment.

Triage Recommendation​

The triage_recommendation field provides specific guidance on the next steps, which may include:

  • Recommended clinical setting (emergency department, urgent care, primary care)
  • Immediate interventions to consider
  • Monitoring recommendations
  • Timeframe for follow-up

Practical Application​

Clinical Decision Support​

The differential diagnosis should be used as a clinical decision support tool, not as a replacement for clinical judgment. Consider:

  1. Evaluating each suggested diagnosis in the context of the specific patient
  2. Prioritizing diagnostic workup based on clinical urgency and pre-test probability
  3. Using the API's reasoning to complement your own diagnostic process

Documentation Support​

The API response can be valuable for clinical documentation:

  1. Use the structured differential diagnosis to enhance your clinical notes
  2. Reference the diagnostic reasoning to support your assessment
  3. Cite the recommended tests and interventions in your plan

Educational Use​

The differential diagnosis can be a powerful educational tool:

  1. Compare the AI-generated differential with your own
  2. Study the reasoning behind each diagnosis
  3. Review the standard of care recommendations to stay updated on evidence-based practices

Example Interpretation​

Consider this simplified response excerpt:

{
  "diagnosis": {
    "diseases": [
      {
        "name": "Acute Myocardial Infarction",
        "probability": 0.85,
        "category": "most-likely"
      },
      {
        "name": "Aortic Dissection",
        "probability": 0.25,
        "category": "cant-miss"
      }
    ]
  },
  "triage": {
    "triage_level": "RED"
  }
}

Interpretation:

  • The most likely diagnosis is Acute Myocardial Infarction (high probability)
  • Aortic Dissection is a critical "can't-miss" diagnosis that must be considered despite its lower probability
  • The RED triage level indicates this is an emergency requiring immediate attention

Limitations​

When interpreting results, keep these limitations in mind:

  1. The API generates differential diagnoses based solely on the provided clinical history
  2. Physical examination findings, laboratory results, and imaging studies not included in the clinical history cannot influence the results
  3. The API does not have access to the patient's complete medical record
  4. Regional variations in disease prevalence may not be fully accounted for
  5. The differential diagnosis is not a definitive diagnosis

Best Practices​

  1. Provide comprehensive clinical histories for more accurate results
  2. Review all categories of diagnoses, not just the most likely
  3. Pay special attention to "can't-miss" diagnoses, even if their probability is lower
  4. Use the API's reasoning to enhance your own diagnostic process
  5. Consider the triage recommendation in the context of your clinical setting and resources