Baseline

Learning personal emotional patterns over time

Overview

Unlike wearables that measure objective physiology (heart rate, HRV), therapy data systems track subjective emotional experience.

Baseline in therapy data means learning what "normal" emotional patterns look like for this individual, so deviations can be noticed.

The Challenge of Emotional Baselines

Wearables baseline: straightforward

// Objective measurements
const restingHeartRate = 62 bpm;
const typicalHRV = 55 ms;
 
// Deviations are measurable
if (currentHR > baseline + 20) {
  // Something changed
}

Easy because:

Objective numbers
Individual physiology is relatively stable
Clear metrics

Therapy baseline: complex

// Subjective self-reports
const typicalMood = ??? // What's "normal" for this person?
const usualMoodRange = ??? // How much fluctuation is typical?
 
// Deviations are subjective
if (currentMood < typicalMood - ???) {
  // Is this concerning, or just a bad day?
}

Hard because:

Subjective scaling (your "3" ≠ my "3")
Emotional fluctuation is normal
Context-dependent (bad week ≠ disorder)
Individual differences enormous

How to Learn Emotional Baseline

1. Initial Baseline Period

Collect data before making conclusions.

❌ Don't:

// First day of mood tracking
if (mood === 2) {
  alert("Persistent low mood detected!"); // Can't know baseline yet
}

✅ Do:

// Wait for baseline period
const BASELINE_PERIOD_DAYS = 14;
 
if (totalMoodEntries < BASELINE_PERIOD_DAYS) {
  return "Still learning your patterns. Keep tracking!";
}
 
const baseline = calculateBaseline(moodEntries);

Recommended baseline period: 14-30 days

Why:

Captures weekly cycles (weekday vs. weekend)
Smooths out daily noise
Establishes reasonable "normal range"

2. Calculate Personal "Normal Range"

Find the typical mood range for this person.

function calculateBaseline(entries: MoodEntry[]): Baseline {
  const moods = entries.map(e => e.mood);
  
  return {
    average: mean(moods),           // e.g., 3.2
    median: median(moods),           // e.g., 3.0
    stdDev: standardDeviation(moods), // e.g., 0.8
    range: {
      low: percentile(moods, 25),   // e.g., 2.5
      high: percentile(moods, 75)   // e.g., 4.0
    }
  };
}

Insights from baseline:

Average mood: What's typical?
Standard deviation: How much does mood normally fluctuate?
Range (25th-75th percentile): "Normal" variation zone

Use case:

const baseline = calculateBaseline(last30Days);
 
if (currentMood < baseline.range.low) {
  // Below typical range — might be worth noting
} else if (currentMood > baseline.range.high) {
  // Above typical range — positive deviation
}

3. Account for Weekly/Monthly Cycles

Some patterns are predictable.

// Weekly patterns
const mondayAvg = 2.5;
const saturdayAvg = 4.0;
 
// Don't flag Saturday's 4.0 as unusual — it's typical for weekends

Common cycles:

Weekday vs. weekend differences
Monthly hormonal cycles (menstruation affects mood)
Seasonal patterns (some people have lower mood in winter)

Approach:

function calculateWeekdayBaselines(entries: MoodEntry[]): WeeklyBaseline {
  const byDay = groupBy(entries, e => e.date.getDay());
  
  return {
    monday: mean(byDay[1]),
    tuesday: mean(byDay[2]),
    wednesday: mean(byDay[3]),
    // ...etc
  };
}
 
// Compare today to typical [day of week]
const todayBaseline = weeklyBaseline[currentDayOfWeek];
if (currentMood < todayBaseline - 1) {
  // Unusual for a Monday
}

4. Update Baseline Over Time

People change. Baseline should adapt.

❌ Don't:

// Lock baseline to first 30 days forever
const baseline = calculateBaseline(entries.slice(0, 30));

✅ Do:

// Rolling baseline using recent data
const baseline = calculateBaseline(last60Days);
 
// OR weighted: recent data counts more
const baseline = calculateWeightedBaseline(allEntries, recentWeight=0.7);

Why:

Life circumstances change
Successful treatment shifts baseline upward
Chronic conditions may shift baseline downward
Seasonal changes

Update frequency: monthly or quarterly

What Baseline DOES

✅ Notices Personal Deviations

Detect when someone is outside their typical range.

const baseline = getUserBaseline(userId);
 
if (currentMoodAvg < baseline.average - 2 * baseline.stdDev) {
  return `
    You've been feeling lower than usual lately.
    Consider talking to someone about how you're feeling.
  `;
}

Use for:

Personal insights ("You've been lower than usual")
Encouraging help-seeking
Identifying possible need for professional support

✅ Recognizes Positive Changes

Notice when things are improving.

const recentAvg = mean(last7Days);
const baselineAvg = baseline.average;
 
if (recentAvg > baselineAvg + 0.5) {
  return "Your mood has been better than usual this week. What's been helping?";
}

Use for:

Celebrating progress
Identifying what's working
Positive reinforcement

✅ Informs Professional Conversations

Give therapists useful data.

// In therapy session
Therapist: "How have you been?"
User: "I don't know, about the same?"
 
// App data shows:
User's baseline: 3.2/5
Last 2 weeks: 2.1/5
 
// More informed conversation
User: "Actually, looking at my app, I've been about 1 point lower than usual lately."
Therapist: "That's helpful to know. What's been different?"

Use for:

Concrete data vs. vague recollection
Treatment progress tracking
Noticing patterns to discuss

What Baseline DOES NOT Do

❌ Diagnose Disorders

Deviation from baseline ≠ mental illness.

❌ Don't:

if (currentMood < baseline - 2) {
  return "You have depression.";
}

✅ Do:

if (currentMood < baseline - 2) {
  return "You've been feeling lower than usual. Consider talking to a professional.";
}

Why:

Many reasons for mood shifts (stress, sleep, physical health, etc.)
Clinical diagnosis requires comprehensive assessment
Baseline deviations are observations, not diagnoses

❌ Predict Crisis

Low mood ≠ suicide risk.

❌ Don't:

if (currentMood < baseline - 3) {
  alert("CRISIS: User at high suicide risk");
}

✅ Do:

// Crisis detection based on explicit language only
if (journalContains("suicide", "kill myself", "end my life")) {
  displayCrisisResources(); // 988, Crisis Text Line, 911
}

Why:

Suicide is not predictable from mood data
Low mood is common and usually not crisis
False alarms create panic; false negatives create false security

Only trigger crisis response: Explicit crisis language from user.

❌ Replace Clinical Judgment

App baseline ≠ professional assessment.

❌ Don't:

return "Your baseline indicates you need medication.";

✅ Do:

return "Your baseline shows persistent lower mood. This could be worth discussing with a doctor or therapist.";

Why:

Treatment decisions require professional expertise
Baseline is one data point among many
Context matters (major life stressors, physical health, etc.)

Baseline Variations

High Baseline, Low Fluctuation

Example:

Average mood: 4.2/5
Standard deviation: 0.3

What it suggests:

Generally positive emotional state
Relatively stable mood

Caution: Even high-baseline people can have crises. Don't ignore concerning language just because baseline is high.

Low Baseline, Low Fluctuation

Example:

Average mood: 2.1/5
Standard deviation: 0.4

What it suggests:

Persistent lower mood
Relatively stable (but low)

Implication: Consider suggesting professional evaluation for persistent low mood.

if (baseline.average < 2.5 && baseline.stdDev < 0.5) {
  return `
    You've consistently reported lower mood over time.
    Persistent low mood can benefit from professional support.
    Consider talking to a mental health professional.
  `;
}

Moderate Baseline, High Fluctuation

Example:

Average mood: 3.0/5
Standard deviation: 1.2

What it suggests:

Wide mood swings
Emotional volatility

Implication: High fluctuation alone doesn't mean disorder, but very extreme swings might be worth discussing with a professional.

if (baseline.stdDev > 1.5) {
  return `
    Your mood varies quite a bit day to day.
    If these swings feel difficult to manage, talking to a therapist might help.
  `;
}

⚠️ Warning: Don't diagnose bipolar disorder or other conditions based on mood variability.

Implementation Example

interface Baseline {
  userId: string;
  periodStart: Date;
  periodEnd: Date;
  
  average: number;
  median: number;
  stdDev: number;
  
  range: {
    low: number;   // 25th percentile
    high: number;  // 75th percentile
  };
  
  weeklyPattern?: {
    monday: number;
    tuesday: number;
    // ...etc
  };
}
 
class BaselineManager {
  async calculateBaseline(userId: string): Promise<Baseline> {
    // Get last 60 days of mood entries
    const entries = await this.getMoodEntries(userId, 60);
    
    if (entries.length < 14) {
      throw new Error("Insufficient data for baseline (need at least 14 days)");
    }
    
    const moods = entries.map(e => e.mood);
    
    return {
      userId,
      periodStart: entries[0].date,
      periodEnd: entries[entries.length - 1].date,
      
      average: mean(moods),
      median: median(moods),
      stdDev: standardDeviation(moods),
      
      range: {
        low: percentile(moods, 25),
        high: percentile(moods, 75)
      },
      
      weeklyPattern: this.calculateWeeklyPattern(entries)
    };
  }
  
  async detectDeviation(userId: string): Promise<DeviationInsight | null> {
    const baseline = await this.getBaseline(userId);
    const recentEntries = await this.getMoodEntries(userId, 7);
    
    if (recentEntries.length < 5) return null; // Need enough recent data
    
    const recentAvg = mean(recentEntries.map(e => e.mood));
    const deviation = recentAvg - baseline.average;
    
    // Significant negative deviation
    if (deviation < -1.0) {
      return {
        type: "negative_deviation",
        magnitude: Math.abs(deviation),
        message: `
          You've been feeling lower than usual (${recentAvg.toFixed(1)} vs. typical ${baseline.average.toFixed(1)}).
          Consider reaching out to someone or practicing extra self-care.
        `
      };
    }
    
    // Significant positive deviation
    if (deviation > 1.0) {
      return {
        type: "positive_deviation",
        magnitude: deviation,
        message: `
          Your mood has been higher than usual lately! 
          What's been going well?
        `
      };
    }
    
    return null; // Within normal range
  }
}

Summary

Emotional baseline:

Individual "normal range" for mood and emotional patterns
Learned over time (14-30 days minimum)
Accounts for cycles (weekly, monthly, seasonal)
Updated over time as circumstances change

What it's for:

Detecting personal deviations
Recognizing positive changes
Informing professional conversations
Self-awareness

What it's NOT for:

Diagnosing disorders
Predicting crisis
Replacing clinical judgment

Key principles:

Baseline is personal (your normal ≠ my normal)
Deviations are observations, not diagnoses
Context always matters
Baseline supports self-awareness and professional care

Build baselines that help users understand themselves better—not pseudo-diagnostic tools masquerading as clinical assessment.

Signals Deviation Engine