Healthcare Inflation Attribution Engine

python
# Healthcare Inflation Attribution Engine
def decompose_trend(claims_current_year, claims_prior_year, population_current, population_prior):
    # Step 1: Calculate aggregate trend
    total_cost_current = sum(c.paid_amount for c in claims_current_year)
    total_cost_prior = sum(c.paid_amount for c in claims_prior_year)
    
    pmpm_current = total_cost_current / len(population_current) / 12
    pmpm_prior = total_cost_prior / len(population_prior) / 12
    
    total_trend = (pmpm_current - pmpm_prior) / pmpm_prior
    
    # Step 2: Standardize service mix to isolate unit cost vs utilization
    service_categories = ['inpatient', 'outpatient', 'professional', 'pharmacy', 'other']
    
    component_attribution = {}
    
    for category in service_categories:
        # Filter claims by category
        current_claims = [c for c in claims_current_year if c.category == category]
        prior_claims = [c for c in claims_prior_year if c.category == category]
        
        # Unit cost trend (price per service, holding utilization constant)
        current_unit_cost = sum(c.paid_amount for c in current_claims) / len(current_claims) if current_claims else 0
        prior_unit_cost = sum(c.paid_amount for c in prior_claims) / len(prior_claims) if prior_claims else 0
        
        unit_cost_trend = (current_unit_cost - prior_unit_cost) / prior_unit_cost if prior_unit_cost > 0 else 0
        
        # Utilization trend (services per member, holding price constant)
        current_utilization = len(current_claims) / len(population_current) / 12
        prior_utilization = len(prior_claims) / len(population_prior) / 12
        
        utilization_trend = (current_utilization - prior_utilization) / prior_utilization if prior_utilization > 0 else 0
        
        # Case mix trend (severity/complexity shift)
        # Use DRG weight, CPT RVU, or diagnosis severity score as proxy
        current_avg_severity = sum(c.case_mix_weight for c in current_claims) / len(current_claims) if current_claims else 0
        prior_avg_severity = sum(c.case_mix_weight for c in prior_claims) / len(prior_claims) if prior_claims else 0
        
        case_mix_trend = (current_avg_severity - prior_avg_severity) / prior_avg_severity if prior_avg_severity > 0 else 0
        
        # Contribution to total trend
        category_pmpm_prior = sum(c.paid_amount for c in prior_claims) / len(population_prior) / 12
        category_weight = category_pmpm_prior / pmpm_prior if pmpm_prior > 0 else 0
        
        component_attribution[category] = {
            'unit_cost_trend': unit_cost_trend,
            'utilization_trend': utilization_trend,
            'case_mix_trend': case_mix_trend,
            'total_category_trend': (1 + unit_cost_trend) * (1 + utilization_trend) * (1 + case_mix_trend) - 1,
            'contribution_to_total_trend': category_weight * ((1 + unit_cost_trend) * (1 + utilization_trend) * (1 + case_mix_trend) - 1)
        }
    
    # Aggregate components
    total_unit_cost_contribution = sum(v['contribution_to_total_trend'] * (v['unit_cost_trend'] / (v['total_category_trend'] or 1)) 
                                       for v in component_attribution.values() if v['total_category_trend'] != 0)
    total_utilization_contribution = sum(v['contribution_to_total_trend'] * (v['utilization_trend'] / (v['total_category_trend'] or 1)) 
                                        for v in component_attribution.values() if v['total_category_trend'] != 0)
    total_case_mix_contribution = sum(v['contribution_to_total_trend'] * (v['case_mix_trend'] / (v['total_category_trend'] or 1)) 
                                     for v in component_attribution.values() if v['total_category_trend'] != 0)
    
    # Provider-specific unit cost attribution (top drivers)
    provider_inflation = {}
    for claim in claims_current_year:
        if claim.provider_id not in provider_inflation:
            provider_inflation[claim.provider_id] = {'current_costs': [], 'prior_costs': []}
        provider_inflation[claim.provider_id]['current_costs'].append(claim.paid_amount)
    
    for claim in claims_prior_year:
        if claim.provider_id in provider_inflation:
            provider_inflation[claim.provider_id]['prior_costs'].append(claim.paid_amount)
    
    provider_trends = []
    for provider_id, data in provider_inflation.items():
        if data['prior_costs']:
            avg_current = sum(data['current_costs']) / len(data['current_costs'])
            avg_prior = sum(data['prior_costs']) / len(data['prior_costs'])
            trend = (avg_current - avg_prior) / avg_prior
            
            provider_trends.append({
                'provider_id': provider_id,
                'provider_name': lookup_provider_name(provider_id),
                'unit_cost_trend': trend,
                'claim_volume': len(data['current_costs'])
            })
    
    provider_trends.sort(key=lambda x: abs(x['unit_cost_trend']), reverse=True)
    
    return {
        'total_trend': total_trend,
        'unit_cost_contribution': total_unit_cost_contribution,
        'utilization_contribution': total_utilization_contribution,
        'case_mix_contribution': total_case_mix_contribution,
        'category_breakdown': component_attribution,
        'top_provider_inflators': provider_trends[:20]
    }

# Example: Decompose 2024 trend
current = load_claims('2024')
prior = load_claims('2023')
pop_current = load_population('2024')
pop_prior = load_population('2023')

result = decompose_trend(current, prior, pop_current, pop_prior)

print("Total Medical Trend: {:.1%}".format(result['total_trend']))
print("\nComponent Attribution:")
print("  Unit Cost Inflation: {:.1%}".format(result['unit_cost_contribution']))
print("  Utilization Change: {:.1%}".format(result['utilization_contribution']))
print("  Case Mix Shift: {:.1%}".format(result['case_mix_contribution']))

print("\nCategory Breakdown:")
for cat, data in result['category_breakdown'].items():
    print("  {}: {:.1%} total ({:.1%} unit, {:.1%} util, {:.1%} mix)".format(
        cat.upper(), data['total_category_trend'], 
        data['unit_cost_trend'], data['utilization_trend'], data['case_mix_trend']))

print("\nTop 5 Provider Inflators:")
for p in result['top_provider_inflators'][:5]:
    print("  {}: {:.1%} unit cost trend ({} claims)".format(
        p['provider_name'], p['unit_cost_trend'], p['claim_volume']))