Cost Driver Decomposition

Overview

This skill breaks down total healthcare cost changes into their component drivers — utilization rate changes, unit cost (price) changes, case-mix/severity shifts, population composition changes, and benefit design effects. It applies actuarial decomposition methods, Kitagawa-Blinder-Oaxaca analysis, and service-category drill-downs to produce transparent, auditable explanations of cost trends.

When to Use

• Explaining year-over-year or quarter-over-quarter PMPM cost changes to leadership
• Preparing cost trend analysis for actuarial reviews or rate-setting
• Identifying the root causes of cost growth for targeted intervention
• Comparing cost drivers across plans, lines of business, or provider networks
• Supporting VBC contract negotiations with transparent cost trend data

Required Inputs

| Input | Description | Format | |-------|-------------|--------| | Claims data | Allowed amounts, service dates, CPT/HCPCS, revenue codes, DRG | Claims detail | | Enrollment data | Member months, plan type, LOB, age/sex | Enrollment file | | Service categories | Inpatient, outpatient, professional, pharmacy, ancillary | Category mapping | | Network data | In-network vs. OON, contracted rates, provider type | Network file | | Benefit design | Copay/coinsurance structure, deductible levels, formulary tier | Plan design | | Prior period data | Same data elements for comparison period(s) | Historical claims |

Methodology

Step 1 — Normalize to PMPM Basis

Convert all cost data to per-member-per-month (PMPM) for trend comparability:

• Total allowed PMPM = Total allowed amount / Total member months
• Calculate PMPM for each service category (IP, OP, Professional, Rx, Other)
• Adjust for incomplete periods: apply completion factors for claims lag (IBNR)
• Standardize benefit design impacts to allowed-amount basis (remove member cost-sharing)

Step 2 — Decompose Cost into Price × Utilization × Mix

Apply the fundamental cost identity:

Cost PMPM = Utilization Rate × Unit Cost × Case Mix Index

For each service category, separate:

• Utilization rate: Admits/1000, visits/1000, scripts/1000, units/1000
• Unit cost: Cost per admit, cost per visit, cost per script, cost per unit
• Case mix: Average DRG weight (IP), average RVU (professional), average AWP (Rx)

Calculate the contribution of each component to total cost change using the decomposition:

ΔCost = ΔUtil × Price₀ × Mix₀ + Util₀ × ΔPrice × Mix₀ + Util₀ × Price₀ × ΔMix + interaction terms

Step 3 — Service Category Drill-Down

Decompose each major service category into sub-categories:

Inpatient:

• Medical vs. surgical admissions
• Top 10 DRGs by cost contribution and trend
• Length of stay trends (geometric mean LOS vs. expected)
• Readmission-driven costs

Outpatient:

• Facility outpatient vs. ambulatory surgery center
• ED costs (visit volume × cost per visit)
• Advanced imaging (MRI, CT, PET) utilization and unit cost
• Site-of-service shift analysis (hospital outpatient → freestanding)

Professional:

• E&M visit trends by complexity level (99213 vs. 99214/99215)
• Specialist referral patterns and cost per referral episode
• Telehealth substitution effects

Pharmacy:

• Generic vs. brand vs. specialty drug mix
• Top 10 drugs by cost and trend (new-to-market drugs driving growth)
• Biosimilar adoption rates
• Specialty pharmacy concentration (top 1% of claims driving what % of Rx cost)

Step 4 — Population Mix Adjustment

Isolate cost changes due to population composition shifts:

• Age/sex factor adjustment: apply prior-period age/sex cost relativities to current population
• Risk score adjustment: compare average RAF score shifts between periods
• Enrollment mix: proportion changes across LOB, plan design, geographic area
• New member vs. continuing member cost differential

Step 5 — Identify Actionable Cost Drivers

Categorize each driver by controllability:

| Category | Controllable? | Intervention Lever | |----------|---------------|-------------------| | Unit price increases | Partially | Network contracting, reference-based pricing | | Utilization growth | Partially | UM programs, care management, prior auth | | Case mix severity | Limited | Risk adjustment accuracy, documentation | | Population aging | Not controllable | Adjust expectations and pricing | | New technology/drugs | Limited | Formulary management, clinical pathways | | Site-of-service shift | Yes | Benefit design, steering programs | | Waste/low-value care | Yes | Choosing Wisely campaigns, clinical decision support |

Step 6 — Benchmark Cost Drivers

Compare observed drivers against external references:

• Medical CPI components for price trend expectations
• Milliman HCG or Optum normative databases for utilization trends
• PwC/Deloitte annual medical cost trend surveys (typically 6-8% total trend)
• Plan-specific historical trend corridors

Step 7 — Construct Cost Driver Narrative

Synthesize findings into a structured explanation:

Example narrative structure:
"Total medical cost PMPM increased 7.2% ($485 → $520). This was driven by:
 • Unit price increases: +3.8% (hospital rate escalators +4.2%, offset by Rx generic conversions)
 • Utilization changes: +2.1% (ED visits +8%, offset by IP admits −3%)
 • Case mix: +0.9% (higher acuity surgical cases, new specialty drug starts)
 • Population mix: +0.4% (aging, risk score increase)
 Of the 7.2% trend, approximately 3.5 points are addressable through site-of-service
 steering, ED diversion programs, and specialty pharmacy management."

Output Specification

Cost Driver Report:
├── Executive Summary (total PMPM trend, top 3 drivers, actionable share)
├── PMPM Trend Summary (by service category, current vs. prior)
├── Decomposition Waterfall (price, utilization, mix contributions)
├── Service Category Deep Dives (IP, OP, Professional, Rx)
├── Population Mix Analysis (age/sex, risk score, enrollment changes)
├── Benchmark Comparison (observed vs. expected by driver)
├── Controllability Assessment (actionable vs. non-actionable drivers)
├── Cost Reduction Opportunity Inventory (estimated savings by lever)
└── Methodology and Data Notes (IBNR factors, normalization methods)

Analysis Framework

Cost Trend Decomposition Waterfall

Present as a waterfall chart data structure:

| Component | Contribution | Running Total | |-----------|-------------|---------------| | Starting PMPM | $485.00 | $485.00 | | + Price changes | +$18.43 | $503.43 | | + Utilization changes | +$10.19 | $513.62 | | + Case mix shift | +$4.37 | $517.99 | | + Population mix | +$1.94 | $519.93 | | + Interaction | +$0.07 | $520.00 | | = Ending PMPM | | $520.00 |

Examples

Example 1 — Annual Cost Trend for Commercial Plan Decompose a 9.1% PMPM increase for a 120,000-member commercial plan. Price contributed +4.2% (hospital contract escalators), utilization +2.8% (driven by 14% ED growth and specialty referral increases), pharmacy mix +1.6% (two new specialty drugs), population +0.5%. Identify $14M addressable through ED diversion ($4M), site-of-service steering ($6M), and specialty pharmacy step therapy ($4M).

Example 2 — MA Plan Cost Driver Analysis Explain why a 40,000-member MA plan's medical loss ratio increased from 84% to 87%. Decompose: inpatient cost +11% (3 high-cost transplant cases contributed 40% of IP increase), outpatient +6% (advanced imaging), pharmacy +9% (Part B drugs). After removing outlier cases, underlying trend is 5.8%, within actuarial expectations.

Guidelines

• Always apply IBNR completion factors before trend comparison; incomplete data understates recent costs
• Separate one-time events (catastrophic claims, COVID-19 surges) from underlying trend
• Use allowed amounts (not paid amounts) for trend analysis to neutralize benefit design changes
• Apply large-claim pooling (truncate at $250K or plan-specific threshold) for stable trend estimates
• Report both per-member and total-dollar perspectives — they tell different stories

Validation Checklist

• [ ] PMPM calculations use correct member-month denominators
• [ ] IBNR/completion factors are applied and documented
• [ ] Decomposition components sum to total observed change (reconciliation check)
• [ ] Large claims are identified and impact is quantified separately
• [ ] Service category mappings are consistent across periods
• [ ] Benchmark sources are current and relevant to population type
• [ ] Controllable vs. uncontrollable classification is clinically sound

HIPAA Compliance

This skill processes aggregate financial and utilization data. When patient-level claims are used in decomposition, all processing must comply with HIPAA Privacy and Security Rules. Apply minimum necessary data access. Large-claim analyses that reference specific cases must be de-identified in reports per 45 CFR §164.514. Cost reports shared with brokers, consultants, or external actuaries require Business Associate Agreements. Never include member identifiers in cost trend outputs.