Stream-Level Cost Benchmarking

This content is for 2025. Switch to the latest version for up-to-date documentation.

This tutorial builds on the cost-vs-price analysis by breaking results down by care stream (acute admitted, ED, mental health, subacute, outpatient) and comparing each stream against NHCDC national reference costs. This mirrors the methodology used in the Australian Hospital Patient Costing System (AHPCS) standards.

Multi-stream analysis workflow

1. Load packages and prepare data

import pandas as pd
import numpy as np
from nwau_py.calculators import (
    calculate_acute, AcuteParams,
    calculate_ed,     EDParams,
    calculate_mh,     MHParams,
    calculate_subacute, SubacuteParams,
    calculate_outpatients, OutpatientParams,
)
from nwau_py.pricing_constants import get_nep

np.random.seed(2025)

# Acute
acute_df = pd.DataFrame({
    "DRG": np.random.choice(["E62A", "E62B", "F62A", "F62B", "G02A"], size=50),
    "LOS": np.random.randint(1, 10, size=50),
    "ICU_HOURS": 0, "ICU_OTHER": 0,
    "PAT_SAMEDAY_FLAG": 0, "PAT_PRIVATE_FLAG": 0,
})

# ED
ed_df = pd.DataFrame({
    "AECC": np.random.choice(["AECC01", "AECC02", "AECC03", "AECC04"], size=80),
    "COMPENSABLE_STATUS": 2, "DVA_STATUS": 2,
})

# Mental health
mh_df = pd.DataFrame({
    "AMHCC": np.random.choice(["1A01", "1A02", "1B01", "2A01"], size=30),
    "LOS": np.random.randint(1, 20, size=30),
    "STATE": 1, "PAT_PRIVATE_FLAG": 0, "PAT_SAMEDAY_FLAG": 0,
    "PAT_PUBLIC_FLAG": 1,
    "priceCat": 1,
})

# Subacute
sa_df = pd.DataFrame({
    "ANSNAP": np.random.choice(["4A01", "4A02", "4B01", "4B02"], size=25),
    "ADM_DATE": pd.Timestamp("2025-03-01"),
    "SEP_DATE": pd.Timestamp("2025-03-10"),
    "LEAVE_DAYS": 0, "BIRTH_DATE": pd.Timestamp("1970-06-15"),
    "STATE": 1, "PAT_PRIVATE_FLAG": 0, "PAT_PUBLIC_FLAG": 1,
    "CARE_TYPE": 1,
})

# Outpatients
op_df = pd.DataFrame({
    "TIER2_CLINIC": np.random.choice([20.01, 20.02, 30.01, 40.01], size=60),
    "SERVICE_DATE": pd.Timestamp("2025-06-01"),
    "BIRTH_DATE": pd.Timestamp("1980-01-01"),
})

2. Calculate NWAU per stream

result_acute  = calculate_acute(acute_df, AcuteParams())
result_ed     = calculate_ed(ed_df, EDParams())
result_mh     = calculate_mh(mh_df, MHParams())
result_sa     = calculate_subacute(sa_df, SubacuteParams())
result_op     = calculate_outpatients(op_df, OutpatientParams())

nwau_col = "NWAU25"
streams = {
    "Acute":       result_acute,
    "ED":          result_ed,
    "MentalHealth": result_mh,
    "Subacute":    result_sa,
    "Outpatient":  result_op,
}

3. Calculate notional funding and build comparison table

nep = get_nep("2025")

cost_benchmarks = {
    "Acute":       5_800,
    "ED":            620,
    "MentalHealth": 9_200,
    "Subacute":     7_400,
    "Outpatient":     520,
}

rows = []
for stream_name, df_result in streams.items():
    total_nwau = df_result[nwau_col].sum()
    notional_funding = total_nwau * nep
    n_episodes = len(df_result)
    synthetic_obs_cost = n_episodes * cost_benchmarks[stream_name]
    cost_ratio = (synthetic_obs_cost / notional_funding) if notional_funding > 0 else float("inf")
    rows.append({
        "Stream":    stream_name,
        "Episodes":  n_episodes,
        "Total NWAU":  round(total_nwau, 4),
        "Notional Funding": round(notional_funding, 2),
        "Observed Cost":    round(synthetic_obs_cost, 2),
        "Cost Ratio":       round(cost_ratio, 4),
    })

benchmark_table = pd.DataFrame(rows)
benchmark_table

The table shows each stream’s aggregate NWAU, notional funding, synthetic observed cost, and cost ratio. Some streams may show NWAU25 = 0 if the synthetic classification codes do not match the published price-weight tables — this realistically reflects what happens with incomplete reference data.

4. Export the comparison

benchmark_table.to_csv("stream_benchmarking_summary.csv")

NHCDC and AHPCS context

NHCDC provides the national cost benchmarks that inform NWAU price weights and the NEP.
AHPCS ensures methodological consistency in cost allocation across hospitals and jurisdictions.
Without AHPCS compliance, inter-hospital cost comparisons may be misleading.
In practice, non-admitted and subacute streams use NEC (not NEP) pricing, which has a fixed-plus-variable cost structure.

AHPCS methodology notes

Scope alignment: Only include episodes that fall within the NHCDC scope. Exclude non-admitted ED, S100 pharmaceuticals, and other out-of-scope activity.
Cost allocation: Use the AHPCS standard cost buckets (direct, indirect, on-costs) and apply the same overhead allocation ratios used in the NHCDC.
Grouper version: Ensure the DRG grouper version used for pricing matches the version used by NHCDC for the same reference year. For AR-DRG, that means the same ICD-10-AM, ACHI, and ACS versions plus the same licensed grouping tables or software release.
Outlier policy: Apply the NHCDC trim-point / outlier policy before computing per-episode averages. Unadjusted means are not comparable.

Caveats for real-world use

Not all streams use NEP — subacute and non-admitted activity are more commonly priced via NEC (National Efficient Cost), which has a fixed-plus-variable cost structure.
NHCDC cost data is lagged — NHCDC data typically reflects costs from 1–2 years prior.
Risk adjustment — real benchmarking should adjust for case mix, remoteness, and indigenous status.
Jurisdiction variation — states and territories may apply different price multipliers, loadings, and activity funding policies.
Synthetic data only — all values shown are illustrative and do not reflect real hospital performance.

What to try next

Return to Calculating Notional Funding from Activity Data to review the pricing step, or Comparing Observed Costs to Efficient Prices for the baseline cost-vs-price analysis.