Skip to main content

Stoploss Provisions

A stoploss provision is a contract clause that changes how a payer reimburses a claim once it crosses a threshold. These are not inherently reported in price transparency data, but are often described in hospital additional notes. Remits analyses can also help triangulate high dollar reimbursement trends.

Strategy​

Stoploss terms are sparsely and inconsistently documented, so the pipeline is designed to maximize coverage while preserving confidence.

Step 1 — Collect direct observations. Hospital price transparency files (MRF) are parsed for notes that describe stoploss terms and aggregated to one row per provider-payer-network. Claims remittance data (Komodo) is processed in parallel to identify providers whose large claims are consistently paid at a stoploss-level percentage, aggregated to one row per provider-payer.

Step 2 — Build the universe. Every provider-payer-network combination that could plausibly have a stoploss clause is enumerated into a single table. This becomes the target — every row in this universe will receive a stoploss rate.

Step 3 — Fill gaps through progressive inference. Not every provider-payer-network has direct evidence. The pipeline applies nine inference methods in descending confidence order: first broadcasting a provider's own observed terms across their other networks, then borrowing from health system peers, then aggregating across similar contracts in the same network and state, and finally falling back to national patterns. Each method assigns a confidence score (0–10); the highest-scoring row wins per contract group.

Step 4 — Combine. All sources are unioned and deduplicated, keeping the best row per contract group.

Classification​

Every stoploss provision is described by these fields:

FieldValuesDescription
stoploss_typefirst dollar, second dollarWhether the rate applies to the entire claim or only the excess above the threshold
threshold_typedollar, LOS, noneWhat triggers the provision — a dollar amount or a length-of-stay (days)
threshold_valueDOUBLEDollar amount (e.g., 100000) or day count (e.g., 5)
reimbursement_typepercentage, per diemRate structure once the threshold is crossed
reimbursement_valueDOUBLEPercentage (e.g., 75) or per diem amount (e.g., 1500)
nte_typedollars per day, NULLCap type, when present
nte_valueDOUBLECap amount per day, when present

Confidence Score​

Every row in the combined table has a canonical_score (0–10). When multiple sources cover the same contract group, the highest-scoring row wins.

Scorecanonical_sourceDescription
10mrf_fullDirect MRF extraction — both stoploss_type and at least one rate field successfully parsed
9mrf_note_onlyMRF extraction with a qualifying note, but stoploss_type could not be classified
8mrf_cross_networkProvider's own MRF data applied to all their networks with the same payer
7remitsProvider's dominant claims payment pattern applied to all their networks
6health_systemTerms inherited from a peer hospital in the same health system
5health_system_cross_networkHealth-system-inherited terms applied to all networks for that provider-payer
4networkMost common structure among ≥5 providers in the same payer, network, and state
3payer_network_state_aggMost common structure across all providers in the same payer, network, and state
2payer_network_national_aggSame as Score 3, applied nationally across the same payer and network
1state_aggMost common structure across all providers in the same state
0national_aggUniversal fallback — always present for every row in the universe

Architecture​

SOURCE SYSTEMS
├─ Hospital Rates (MRF)         glue.hospital_data.hospital_rates
└─ Claims Remits (Komodo)       tq_intermediate.external_komodo.remits

        ↓

① OBSERVATION
  MRF Path
  ├─ provisions_stoploss_mrf_base
  ├─ provisions_stoploss_mrf_extracted
  └─ provisions_stoploss_mrf_aggregated      scores 9–10
  Remits Path
  ├─ provisions_stoploss_remits_base
  └─ provisions_stoploss_remits_aggregated

        ↓

② PROVIDER UNIVERSE
  └─ provisions_stoploss_provider_universe   all provider × payer × network

        ↓

③ INFERENCE
  ├─ provisions_stoploss_mrf_crossnet_broadcast        · 8
  ├─ provisions_stoploss_remits_broadcast              · 7
  ├─ provisions_stoploss_healthsystem                  · 6
  ├─ provisions_stoploss_hs_crossnet_broadcast         · 5
  ├─ provisions_stoploss_network                       · 4
  ├─ provisions_stoploss_agg_payer_net_state           · 3
  ├─ provisions_stoploss_agg_payer_net_national        · 2
  ├─ provisions_stoploss_agg_state                     · 1
  └─ provisions_stoploss_agg_national                  · 0

        ↓

④ COMBINED
  └─ provisions_stoploss_combined            best score wins per group

Layer 1: Observation​

MRF Notes​

Hospital price transparency notes are processed in three steps to produce structured stoploss terms for each provider, payer, and network combination.

Step 1 — Base (provisions_stoploss_mrf_base)

Pulls all commercial hospital rate records whose notes mention stoploss-related terms and contain dollar amounts.

Filter criteria
  • At least one note field (additional_generic_notes, additional_payer_notes, negotiated_algorithm) contains any of: stop, loss, threshold, discount, charges, outlier
  • At least one note field contains a digit
  • payer_class_name = 'Commercial'
  • Network ID is not null, or plan_name is not null

Step 2 — Extracted (provisions_stoploss_mrf_extracted)

Filters out non-stoploss notes, then parses each qualifying note into structured fields: stoploss type, threshold, reimbursement rate, and cap.

Filtering and extraction detail

Multi-field exclusions

Notes matching any of these patterns — based on combinations of note fields — are dropped before extraction:

PatternCondition
Hospital System Supply IdentifierGeneric note contains "Hospital System Supply Identifier" ending with "Lawson ID", no payer notes
Term LineGeneric note ends with "Term Line" + number, no payer notes
Outpatient billingGeneric note is exactly the standard outpatient billing setting note, no payer notes
Re-evaluatedPayer or generic note starts with "Re-evaluated:", no "threshold" keyword
CPT + contract paymentGeneric note is a CPT code; payer note starts with "contract indicates payment"
Bare number + % of chargesGeneric note is a bare number; payer note is exactly "N% OF TOTAL BILLED CHARGES"
Rev/Proc codeGeneric note matches "Rev NNN Proc NNNNN" format, no payer notes
Reimb exceeds chargesGeneric note is exactly "Reimb exceeds charges. Trim to Lessor Of Claim"
OPPS APCAny note starts with "OPPS" + whitespace/; + "APC"

Per-note exclusions — 10 families of note text that indicate non-stoploss content

Notes containing any of the following patterns are excluded on a per-note basis:

Carveouts

  • Cost Plus Markup
  • Starts with Implants and Devices
  • Contains implant threshold
  • implants on a single charge line
  • Starts with If > $2500
  • Starts with Eligible charges greater than $1,090
  • Contains at cost +
  • Contains documented acquisition cost
  • Contains invoice cost

Revenue-code-specific thresholds

  • Contains qualifying revenue code
  • Contains applicable revenue code charges exceed
  • Contains revenue code charges exceed
  • Contains threshold for each revenue code
  • Contains charges for specified revenue codes exceed
  • Matches (rev code NNN) charges in excess of
  • Matches Rev 0NNN ... in addition to IP rate
  • Contains revenue code charges included in per diem

Transplant / global case rates

  • Starts with Global case rate listed refers to the fixed fee payment made to provider...
  • Starts with Patients who expire during treatment or who do not complete the transplant process
  • Contains transplant case rate includes/refers to/is inclusive
  • Contains transplant global case rate
  • Contains transplant inlier case rate/period
  • Contains transplant inpatient stays exceeding
  • Contains case rate applies to the mother's stay
  • Contains highest case rate is paid
  • Contains packaged/bundled into a separate rate

Fee schedules / OPPS / Medicare

  • Starts with APC grouping requires price history
  • Starts with Paid under OPPS;
  • Contains OPPS CCR
  • Contains MEDICARE OPPS OUTLIER
  • Starts with This is being calculated using the IPPS logic from CMS.
  • Matches Medicare NNN/NNN Reimbursement:
  • Starts with Medicare OP Reimbursement:
  • Starts with TRICARE OP Reimbursement:
  • Contains contractual percentage of Medicare
  • Starts with MEDICAL/SURGICAL Procedures: Fee Schedules
  • Contains to determine the federal payment amount for each IRF patient

Code-level / lookup rates

  • Matches DRG NNN DRG Case Weight
  • Contains of billable gross charges for specified code
  • Contains aggregated billable gross charges for specific code
  • Starts with Procedure Lookup Tables:
  • Starts with (Sum: ) Procedure Lookup:
  • Starts with FormulaCPE:
  • Starts with Group (Sum Of)
  • Starts with Group (Lesser Of)
  • Contains primary procedure pays at

Long-stay / outlier structures

  • Contains Long Stay Outlier
  • Starts with High Cost Outlier:
  • Contains Capital Addon
  • Starts with Admin Days:
  • Contains outlier payments are provided on a service basis
  • Contains outlier terms do not apply
  • Contains adjusted base rate indicated...further adjusted for transfers and outliers
  • Contains additional reimbursement may apply if total charges exceed threshold
  • Matches lesser of: [(if transferred

Non-stoploss rate structures

  • Matches N% of cost
  • Contains reimbursed at ... of invoice cost when billed charges exceed
  • Contains reimbursement rate exceeds billed charges
  • Contains uncontracted outpatient services will be paid
  • Contains services not specifically listed in contract

Specialty / clinical billing

  • Contains anesthesia is calculated by adding together base units
  • Contains anesthesia procedures will divide billed minutes
  • Contains midlevel providers and midlevel modifiers are paid
  • Contains skilled nursing provision
  • Contains use this provider number in place of provider number from claim

Algorithm notes / scripting artifacts

  • Starts with Interim Value:
  • Starts with Can be subject to stoploss provisions
  • Starts with Re-evaluated: HLB. (without threshold keyword)
  • Contains End If
  • Contains Processing Action: Note Only
  • Matches {Charges In Service Type} or {Units In Service Type} (unfilled template variable)

Carveout exclusion format

  • Contains \nExclusion - without Rate Threshold

Extraction patterns — regex and keyword rules for parsing thresholds, rates, and caps from raw note text

Text is normalized before extraction: lowercased, hyphens removed, % → "percent".

Stoploss Type (first match wins):

PriorityValueSignal
1first dollarfirst dollar, 1st dollar, entire claim, entire admission, entire stay
2second dollarsecond dollar, 2nd dollar, in addition to, over the threshold, charges over, charges in excess, after the threshold, above the threshold, for each day past, days after, excess of, over and above (with carveout-language guards)
3first dollarcharges exceeding followed by threshold or outlier
4first dollarNote contains revert (with guards)
5first dollarNote starts with charges greater than

Dollar Threshold:

PatternExample
k-notation$50k, 1.5k
Keyword-anchored (exceeds, threshold of, greater than + ≥5-digit number)exceeds $100,000
Structured fixed amount (+ Fixed Amount N)+ Fixed Amount 56687.00
Numeric fallback (any ≥5-digit number)$25000

LOS Threshold: trigger word (exceeds, >, past, after, etc.) + number + day(s) — e.g., exceeds 5 days

Reimbursement — Percentage:

PatternExample
Structured reimbursement rate (Stoploss First/Second Dollar + Reimbursement Rate N)Stoploss First Dollar + Reimbursement Rate 60.00
Structured POC rate (initial rate is zero, followed by + Percent Of Charge + Reimbursement Rate N)+ Percent Of Charge + Reimbursement Rate 54.25
Structured rate threshold (percentage on same line as Rate Threshold in auto-generated note)Rate Threshold $125,000.00...reimbursed at 18.6%
Decimal (N.N percent)60.5%
Integer (N percent)75%

Reimbursement — Per Diem:

PatternExample
Amount-before (N per diem)1500 per diem
Rate-of (per diem rate of N)per diem rate of $1500
Bare (per diem N)per diem $1500

Cap: not to exceed / NTE / up to a maximum of + number + (optional k) + per day

Quality gate and confidence scoring

A note is only kept if:

  1. Contains stoploss, threshold, excess, outlier, or exceed — or starts with charges greater than
  2. At least one field extracted: percentage, per diem reimbursement, dollar threshold ≥ $50K, or per diem threshold

Confidence score (sum of all fired pattern weights, max 15):

CategoryPatternPoints
Stoploss Typestoploss_type is not null1
Dollar Thresholdk-notation, keyword-anchored, or structured2
Dollar Thresholdnumeric-fallback1
Dollar Thresholdvalue between 100K–100K–300K (bonus)1
LOS Thresholddays-keyword1
LOS Thresholdvalue between 3–30 days (bonus)1
Percentagestructured auto-generated format3
Percentagedecimal or integer2
Percentagevalue between 20%–70% (bonus)1
Per Diemamount-before or rate-of2
Per Diembare1
Per Diemvalue between 500–500–5,000 (bonus)1
Capeither pattern2

Step 3 — Aggregated (provisions_stoploss_mrf_aggregated)

Produces one row per provider-payer-network. When multiple qualifying notes exist for the same group, the best is selected — preferring notes shared by more rate IDs and notes with more extracted fields.

Rows where both the threshold and reimbursement type are fully classified get source_label = 'mrf_full' (score 10). Rows where the note qualifies but the stoploss type could not be determined get source_label = 'mrf_note_only' (score 9).

Output schema
ColumnTypeDescription
mrf_idVARCHARprovider_id-payer_id-network_id-setting
provider_idVARCHAR
payer_idVARCHAR
network_idBIGINT
settingVARCHAR'Inpatient' or 'Outpatient'
source_labelVARCHAR'mrf_full' or 'mrf_note_only'
stoploss_typeVARCHAR'first dollar', 'second dollar', or NULL
threshold_typeVARCHAR'dollar', 'LOS', or NULL
threshold_valueDOUBLEDollar amount or day count
reimbursement_typeVARCHAR'percentage', 'per diem', or NULL
reimbursement_valueDOUBLEPercentage or per diem dollar amount
nte_typeVARCHAR'dollars per day' or NULL
nte_valueDOUBLECap amount per day
dollar_thresholdDOUBLERaw extracted dollar threshold
per_diem_thresholdDOUBLERaw extracted day count threshold
percentage_reimbursementDOUBLERaw extracted percentage
per_diem_reimbursementDOUBLERaw extracted per diem amount
ref_matched_textVARCHARPipe-delimited substrings that matched
ref_regex_patternVARCHARPipe-delimited pattern labels
ref_rate_countBIGINTDistinct rate IDs sharing the canonical note
source_idBIGINTMinimum rate ID from the contributing group
source_noteVARCHARNote text used for extraction
traceabilityVARCHARJSON with source record ID, note type, rate count, note text

Remits​

Claims remittance data from Komodo is used to infer stoploss reimbursement rates where MRF notes are unavailable.

Step 1 — Pre-Processing (remits_lines_by_year)

Raw claims are filtered, linked to hospital and payer reference data, and grouped by year to produce the input for the clustering step.

Pipeline steps
StepTableDescription
1remits_spinesJoin remits to hospital and payer spines in a single pass
2remits_stoploss_rollupRoll up stoploss provisions to (provider, payer) level
3remits_headerAggregate to remit_id grain; threshold driven by MRF with $150K fallback
4remits_linesLine-level table scoped to remit_ids in header, with applied_threshold per line
5remits_lines_filteredFilter to quality gates and 10–90% payment ratio for stoploss signal
6remits_lines_by_yearGroup filtered lines by year — final output consumed by remits_base

Input tables: tq_intermediate.external_komodo.remits, tq_intermediate.external_komodo.medical_headers, tq_production.spines.spines_provider_hospitals, and payer mapping tables.

Step 2 — Base (provisions_stoploss_remits_base)

Groups payment rates from claims history into clusters to identify the dominant rate for each provider-payer pair, giving extra weight to surgical claims.

Clustering algorithm
  • Input range: Lines with line_perc_allowed between 0.25 and 0.95 (25%–95%)
  • Revenue code weighting: Surgical codes (revenue_code not in 025x, 026x, 027x, 063x, 064x) are weighted 3x
  • Clustering: Rates sorted ascending; a new cluster starts when gap > 0.05 (5 percentage points). Each cluster is labeled by its minimum rate value
  • Cross-year aggregation: Clusters scored by n_years, n_rev_codes, n_surgical_codes, lines_at_percentage, and lines_at_percentage_weighted

Step 3 — Aggregated (provisions_stoploss_remits_aggregated)

Selects the dominant payment rate cluster per provider-payer. The winning cluster must have at least 25 claims and represent more than 25% of total lines. If no MRF threshold is available, $150,000 is used as the fallback.

Traceability fields
FieldDescription
ref_lines_at_pctNumber of lines at the dominant rate
ref_pct_of_totalDominant rate's share of all lines for this provider-payer
ref_total_linesTotal lines across all clusters
ref_stoploss_thresholdApplied threshold (from MRF or $150K default)
ref_threshold_source'mrf' or 'default'
ref_rate_distributionJSON map of all rate clusters and their line counts

Layer 2: Provider Universe​

Defines every provider-payer-network combination that will receive a stoploss row. All inference methods in Layer 3 fill values into this universe.

Built from three sources — commercial rate records, MRF-observed stoploss rows, and remits-observed rows — then enriched with provider and payer attributes.

Source tables and enrichment columns

Sources (deduplicated, clear_rates preferred over mrf when both present):

SourceDescription
clear_ratesAll hospital commercial rate records with canonical_rate_score > 1 and a non-null network_id
mrfAll rows from provisions_stoploss_mrf_aggregated
remitsAll rows from provisions_stoploss_remits_aggregated (null network_id; pass through as-is)

Enrichment joins:

  • Provider attributes (health system, location, bed count, revenue) from spines_provider_hospitals + spines_provider_healthsystems
  • Payer commercial market share by state from policy_reporter_state
Hospital attributes
ColumnDescription
healthsystem_id, healthsystem_nameHealth system affiliation
provider_state, provider_city, provider_countyLocation
provider_latitude, provider_longitudeCoordinates
provider_type, provider_subtypeProvider classification
total_licensed_bedsHospital size
net_patient_revenueHospital revenue
national_net_patient_revenue_rankNPR rank nationally (RANK() OVER ORDER BY net_patient_revenue DESC)
state_net_patient_revenue_rankNPR rank within state
Payer coverage data
ColumnDescription
payer_state_covered_livesPayer's covered lives in the provider's state
payer_state_covered_lives_pct_of_totalPayer's share of total commercial covered lives in that state

Layer 3: Inference​

Methods are listed in confidence order (highest score first). The highest-scoring row per contract group wins.

Some methods adjust dollar thresholds using each hospital's cost-to-charge ratio (CCR), so that thresholds are comparable across hospitals with different charge structures.

CCR adjustment formula

adjusted_threshold = ROUND(source_median_threshold × recipient_ip_ccr_relativity, 0)

The source threshold is first normalized by the source hospital's CCR relativity, then scaled by the recipient hospital's CCR relativity. If a relativity is outside [0.1, 10.0] or unavailable, the raw median is used instead.

Broadcasts​

Take a single observed rate per provider-payer and apply it to all networks for that pair — no aggregation or inference, just expansion of what's already known.

Score 8 — MRF Crossnet Broadcast (provisions_stoploss_mrf_crossnet_broadcast)​

Extends a provider's own MRF stoploss terms to all their networks with the same payer. For example, if BCBS Preferred has data, those terms are applied to BCBS Select and BCBS HMO where no direct data exists.

Source: MRF — provider's own stoploss data, best row per provider-payer
Recipients: All networks for that provider-payer pair
Expanded to: (provider, payer) → (provider, payer, network)

Selection logic
  1. From provisions_stoploss_mrf_aggregated, select the best row per (provider, payer): prefer source_label = 'mrf_full' (score 10) over 'mrf_note_only' (score 9); tiebreak by dollar_threshold DESC NULLS LAST
  2. Join to the provider universe on (provider, payer) — one source row fans out to all networks for that provider-payer
  3. Provider's own observed network is included (no self-exclusion)
Traceability fields
FieldDescription
source_network_id, source_network_nameThe specific network the source MRF row came from
source_label'mrf_full' or 'mrf_note_only'
source_noteOriginal MRF note text used for extraction

Score 7 — Remits Broadcast (provisions_stoploss_remits_broadcast)​

Extends the payment rate inferred from claims history to all networks for that provider-payer pair.

Source: Remits — provider's dominant payment rate cluster, best per provider-payer
Recipients: All networks for that provider-payer pair
Expanded to: (provider, payer) → (provider, payer, network)

Selection logic
  1. From provisions_stoploss_remits_aggregated, select one row per (provider, payer) — the dominant cluster already chosen in Layer 1
  2. Join to the provider universe on (provider, payer) — fans out to all networks
  3. No stoploss type or threshold type is assigned; only reimbursement_type = 'percentage' and reimbursement_value are set from the cluster rate
Traceability fields
FieldDescription
ref_lines_at_pctNumber of claims lines at the dominant rate cluster
ref_pct_of_totalDominant cluster's share of all lines for this provider-payer
ref_total_linesTotal lines across all clusters
ref_stoploss_thresholdDollar threshold used to identify stoploss claims (from MRF or $150K default)
ref_threshold_source'mrf' or 'default'
ref_rate_distributionJSON map of all rate clusters and their line counts

Health System​

Borrow stoploss terms from a peer hospital in the same health system that has direct MRF data.

Score 6 — Health System (provisions_stoploss_healthsystem)​

Copies stoploss terms from a peer hospital in the same health system that has price transparency data. The peer with the highest dollar threshold is used as the representative. Terms are copied directly with no adjustment.

Source: MRF (Health System) — stoploss data from a peer hospital in the same health system
Recipients: All hospitals in that health system without their own MRF data for that payer and network
Expanded to: (health system, payer, network) → (provider, payer, network)

Selection logic
  1. For each (healthsystem, payer, network), find all MRF rows where source_note IS NOT NULL and the provider has a healthsystem_id
  2. Rank by dollar_threshold DESC NULLS LAST, percentage_reimbursement DESC NULLS LAST — select the highest-threshold row as the representative
  3. All other providers in the same healthsystem receive that representative's stoploss terms for the same (payer, network)
  4. Source provider is excluded from receiving its own terms (provider_id <> source_provider_id)

All stoploss fields are copied directly from the representative MRF row.

Traceability fields
FieldDescription
source_hs_id, source_hs_nameHealth system the source provider belongs to
source_provider_id, source_provider_nameThe specific provider whose MRF data was inherited
source_network_id, source_network_nameNetwork of the source row
source_payer_id, source_payer_namePayer of the source row
source_noteOriginal MRF note text

Score 5 — Healthsystem Crossnet Broadcast (provisions_stoploss_hs_crossnet_broadcast)​

Takes the stoploss terms inherited from a health system peer (Score 6) and applies them across all networks for that provider-payer pair, using the same logic as Score 8.

Source: MRF (Health System) — health-system-inherited stoploss data, best row per provider-payer
Recipients: All networks for that provider-payer pair
Expanded to: (provider, payer) → (provider, payer, network)

Selection logic
  1. From provisions_stoploss_healthsystem, select the best row per (provider, payer): tiebreak by dollar_threshold DESC NULLS LAST
  2. Join to the provider universe on (provider, payer) — fans out to all networks
  3. Mirror of Score 8 logic, but sourced from health-system-inherited data rather than direct MRF data
Traceability fields
FieldDescription
source_hs_id, source_hs_nameHealth system the original source provider belongs to
source_provider_id, source_provider_nameThe peer hospital whose MRF data was inherited in Score 6
source_network_id, source_network_nameNetwork of the Score 6 source row
source_payer_id, source_payer_namePayer of the Score 6 source row
source_noteOriginal MRF note text

Network​

Infer stoploss terms from the most common structure seen across providers in the same payer, network, and state — with a high-confidence bar.

Score 4 — Network Inference (provisions_stoploss_network)​

Infers stoploss terms from the most common structure seen among providers in the same network and state. Requires at least 5 providers with 80% agreement on the dominant structure. Dollar thresholds are adjusted for each hospital's charge structure.

Source: MRF — ≥5 providers in the same payer, network, and state with ≥80% agreement on structure
Recipients: All providers in that payer, network, and state
Expanded to: (payer, network, state) → (provider, payer, network)

Eligibility and calculation

Eligibility requirements:

  • dollar_threshold IS NOT NULL AND > 0
  • percentage_reimbursement IS NOT NULL
  • network_id IS NOT NULL
  • setting = 'Inpatient'
  • The (payer, network, state) cluster has ≥ 5 contributing providers
  • The modal structure (type combination + banded values) has ≥ 80% agreement across contributors

Clustering:

  • Contributors are bucketed into type-bands: FLOOR(dollar_threshold / 25000) (25K bands) and FLOOR(percentage_reimbursement / 5) (5-point bands)
  • The dominant cluster (largest count, tiebroken alphabetically) is selected
  • type_agreement_rate = modal_cluster_count / total_contributor_count

Threshold calculation:

  1. Find the median contributor within the modal cluster (by raw dollar_threshold)
  2. Apply CCR adjustment: COALESCE(median_norm_threshold × recipient_ip_ccr_relativity, median_raw_threshold)
    • median_norm_threshold = median_raw_threshold / source_ip_ccr_relativity
    • Final value = normalized threshold × recipient's CCR relativity
  3. Falls back to raw median if CCR relativity is unavailable or out of range
Traceability fields
FieldDescription
ref_n_contributorsNumber of providers in the modal cluster
ref_n_hs_contributorsNumber of distinct health systems in the modal cluster
ref_agreement_rateFraction of total (payer, network, state) contributors that matched the modal cluster
source_provider_idProvider ID of the median contributor
source_threshold_valueRaw median threshold before CCR adjustment
calc_threshold_valueFinal CCR-adjusted threshold
ref_ccr_relativityRecipient provider's CCR relativity used in adjustment

Aggregated​

Broad fallbacks using the most common stoploss structure across progressively wider populations. Applied when no provider- or network-specific data is available.

Score 3 — Payer + Network + State (provisions_stoploss_agg_payer_net_state)​

Uses the most common stoploss structure across all observed providers in the same payer, network, and state. No minimum contributor count required. Dollar thresholds are adjusted for each hospital's charge structure.

Source: MRF — all inpatient rows with a classified stoploss type for this payer, network, and state
Recipients: All providers in that payer, network, and state
Expanded to: (payer, network, state) → (provider, payer, network)

How this differs from Score 4

How it differs from Network inference (Score 4):

  • No minimum contributor count or agreement rate threshold required
  • Uses all inpatient MRF rows with stoploss_type IS NOT NULL
  • Groups by (payer_id, network_id, provider_state) — payer-specific

Structure selection: Most common (stoploss_type, threshold_type, reimbursement_type) combination per group, tiebroken alphabetically.

Threshold calculation: Median threshold_value across contributors in the dominant structure, CCR-adjusted using the same formula as network inference.

Traceability fields
FieldDescription
ref_n_contributorsNumber of providers contributing to the dominant structure
ref_agreement_rateDominant structure's share of total providers in the group
source_idRate ID of the median contributor
source_threshold_valueRaw median threshold
calc_threshold_valueCCR-adjusted threshold
ref_ccr_relativityRecipient's CCR relativity

Score 2 — Payer + Network National (provisions_stoploss_agg_payer_net_national)​

Same as Score 3 but applied nationally rather than by state. The threshold is a raw national median with no adjustment for individual hospital charge structures.

Source: MRF — all inpatient rows for this payer and network, nationally
Recipients: All providers in that payer and network
Expanded to: (payer, network) → (provider, payer, network)

How this differs from Score 3
  • Groups by (payer_id, network_id) — no state filter
  • No minimum contributor count or agreement rate threshold required
  • Uses all inpatient MRF rows with a classified stoploss type for that payer and network
  • No CCR adjustment — raw national median threshold is used as-is

Structure selection: Most common (stoploss_type, threshold_type, reimbursement_type) combination per group, tiebroken alphabetically.

Traceability fields
FieldDescription
ref_n_contributorsNumber of providers contributing to the dominant structure
ref_agreement_rateDominant structure's share of total providers in the group
source_idRate ID of the median contributor
source_threshold_valueRaw national median threshold (no CCR adjustment applied)

Score 1 — State (provisions_stoploss_agg_state)​

Uses the most common stoploss structure across all providers in the same state, regardless of payer or network. Applied when no payer- or network-specific data is available. Dollar thresholds are adjusted for each hospital's charge structure.

Source: MRF — all inpatient rows in this state, across all payers and networks
Recipients: All providers in that state
Expanded to: state → (provider, payer, network)

How this differs from Score 2
  • Groups by provider_state only — crosses all payers and networks
  • No minimum contributor count or agreement rate threshold required
  • Uses all inpatient MRF rows with a classified stoploss type in that state
  • CCR adjustment applied — thresholds normalized for differences in hospital charge structures

Structure selection: Most common (stoploss_type, threshold_type, reimbursement_type) combination per state, tiebroken alphabetically.

Traceability fields
FieldDescription
ref_n_contributorsNumber of providers contributing to the dominant structure
ref_agreement_rateDominant structure's share of total providers in the state
source_idRate ID of the median contributor
source_threshold_valueRaw median threshold before CCR adjustment
calc_threshold_valueFinal CCR-adjusted threshold
ref_ccr_relativityRecipient provider's CCR relativity used in adjustment

Score 0 — National Fallback (provisions_stoploss_agg_national)​

Universal fallback — guarantees every provider-payer-network has at least one row.

Source: MRF — all inpatient rows nationally
Recipients: Every provider-payer-network in the universe
Expanded to: national → all rows

How this differs from Score 1
  • No geographic grouping — computed once across all inpatient MRF data
  • No minimum contributor count or agreement rate threshold required
  • No CCR adjustment — raw national median threshold is used as-is

Structure selection: Single most common (stoploss_type, threshold_type, reimbursement_type) combination across all observed inpatient MRF data, tiebroken alphabetically.

Traceability fields
FieldDescription
ref_n_contributorsTotal number of providers contributing to the national structure
ref_agreement_rateDominant structure's share of all national contributors
source_idRate ID of the median contributor
source_threshold_valueRaw national median threshold (no CCR adjustment applied)

Layer 4: Combine​

Combined Table: provisions_stoploss_combined​

All 11 sources (scores 0–10) are combined. For each provider-payer-network, the row with the highest confidence score is kept, with full provider and payer context joined back.

Output schema

Provider metadata:

ColumnTypeDescription
provider_idVARCHAR
provider_nameVARCHAR
payer_idVARCHAR
payer_nameVARCHAR
network_idBIGINT
network_nameVARCHAR
settingVARCHARAlways 'Inpatient'
healthsystem_idVARCHAR
healthsystem_nameVARCHAR
provider_stateVARCHAR
provider_cityVARCHAR
provider_countyVARCHAR
provider_latitudeDOUBLE
provider_longitudeDOUBLE
provider_typeVARCHAR
provider_subtypeVARCHAR
total_licensed_bedsINTEGER
net_patient_revenueDOUBLE
national_net_patient_revenue_rankINTEGER
state_net_patient_revenue_rankINTEGER
payer_state_covered_livesDOUBLE
payer_state_covered_lives_pct_of_totalDOUBLE

Stoploss provision fields:

ColumnTypeDescription
canonical_sourceVARCHARSource label (see score table above)
canonical_scoreINTEGER0–10
canonical_stoploss_typeVARCHAR'first dollar', 'second dollar', or NULL
canonical_threshold_typeVARCHAR'dollar', 'LOS', 'none', or NULL
canonical_threshold_valueDOUBLEDollar amount or day count
canonical_reimbursement_typeVARCHAR'percentage', 'per diem', or NULL
canonical_reimbursement_valueDOUBLEPercentage or per diem dollar amount
canonical_nte_valueDOUBLECap per day, when present
canonical_nte_typeVARCHAR'dollars per day' or NULL
traceabilityVARCHARJSON documenting source IDs, contributor counts, CCR adjustments, and original note text

Abridged Table: provisions_stoploss_abridged​

Same as combined but without traceability metadata.

Traceability Examples​

Score 10 — mrf_full

{stoploss_type} stoploss once the {threshold_type} threshold of {threshold_value} is met, with a {reimbursement_type} reimbursement of {reimbursement_value}, not to exceed {nte_value} {nte_type}. Sourced directly from {provider_name}'s MRF record (id: {source_record_id}, {source_note_type}) and shared amongst {ref_rate_count} rate(s).

Source Note: "{source_note}"

Example (percentage): First dollar stoploss once the dollar threshold of $150,000 is met, with a percentage reimbursement of 85%. Sourced directly from Mercy General's MRF record (id: 4821903, additional_payer_notes) and shared amongst 3 rate(s).

Source Note: "Stoploss: first dollar at $150,000, 85% reimbursement thereafter."

Example (per diem): First dollar stoploss once the LOS threshold of 10 days is met, with a per diem reimbursement of $1,800, not to exceed $2,000 dollars per day. Sourced directly from Mercy General's MRF record (id: 4821903, additional_payer_notes) and shared amongst 3 rate(s).

Source Note: "Stoploss: first dollar beyond 10 days, $1,800 per diem thereafter, NTE $2,000/day."

Score 9 — mrf_note_only

Stoploss once the {threshold_type} threshold of {threshold_value} is met, with a {reimbursement_type} reimbursement of {reimbursement_value}, not to exceed {nte_value} {nte_type}. Sourced directly from {provider_name}'s MRF record (id: {source_record_id}, {source_note_type}) shared amongst {ref_rate_count} rate(s).

Source Note: "{source_note}"

Example: Stoploss once the dollar threshold of $100,000 is met, with a percentage reimbursement of 80%. Sourced directly from St. Luke's Medical Center's MRF record (id: 3019284, additional_payer_notes) shared amongst 1 rate(s).

Source Note: "SL provision applies at $100,000; 80% of covered charges thereafter."

Score 8 — mrf_cross_network

{stoploss_type} stoploss once the {threshold_type} threshold of {threshold_value} is met, with a {reimbursement_type} reimbursement of {reimbursement_value}, not to exceed {nte_value} {nte_type}. Derived from {source_network_name}'s network rates in {provider_name}'s MRF (id: {source_id}) shared amongst {ref_rate_count} rate(s).

Source Note: "{source_note}"

Example: Second dollar stoploss once the dollar threshold of $200,000 is met, with a percentage reimbursement of 90%. Derived from BluePPO-National's network rates in Valley View Hospital's MRF (id: 5503812) shared amongst 2 rate(s).

Source Note: "Inpatient stoploss: second dollar, $200,000 threshold, 90% reimbursement."

Score 7 — remits

{reimbursement_type} reimbursement of {reimbursement_value} ({ref_threshold_source} threshold of {ref_stoploss_threshold} used), not to exceed {nte_value} {nte_type}. Derived from remits data for {provider_name} where {ref_pct_of_total}% ({ref_lines_at_pct} of {ref_total_lines}) of high dollar remits for this payer/provider were paid at this percentage.

Rate Distribution: {ref_rate_distribution}

Example: Percentage reimbursement of 85% (mrf threshold of $125,000 used). Derived from remits data where 15.16% (47 of 310) of high dollar remits for this payer/provider were paid at this percentage.

Rate Distribution:

RateShare
85%62%
80%38%
Score 6 — health_system

{stoploss_type} stoploss once the {threshold_type} threshold of {threshold_value} is met, with a {reimbursement_type} reimbursement of {reimbursement_value}, not to exceed {nte_value} {nte_type}. Inherited from {source_hs_name} (HS id: {source_hs_id}) peer {source_provider_name} (provider_id: {source_provider_id}) for {source_payer_name} (payer id: {source_payer_id}) / {source_network_name} (network id: {source_network_id}).

Source Note: "{source_note}"

Example: First dollar stoploss once the dollar threshold of $175,000 is met, with a percentage reimbursement of 85%. Inherited from Northside Health System (HS id: 9901) peer Northside Medical Center (provider_id: 1245678) for United Healthcare (payer id: 87023) / UHC-Choice-Plus (network id: 4412).

Source Note: "Inpatient stoploss at $175,000, 85% reimbursement, first dollar."

Score 5 — health_system_cross_network

{stoploss_type} stoploss once the {threshold_type} threshold of {threshold_value} is met, with a {reimbursement_type} reimbursement of {reimbursement_value}, not to exceed {nte_value} {nte_type}. Inherited from a {source_hs_name} (HS id: {source_hs_id}) peer under their {source_network_name} network.

Source Note: "{source_note}"

Example: First dollar stoploss once the dollar threshold of $100,000 is met, with a percentage reimbursement of 80%. Inherited from an Eastside Health Partners (HS id: 7741) peer under their CignaPPO-SE network.

Source Note: "SL: $100,000 aggregate, 80% reimbursement above threshold."

Score 4 — network

{stoploss_type} stoploss once the {threshold_type} threshold of {calc_threshold_value} is met, with a {reimbursement_type} reimbursement of {reimbursement_value}, not to exceed {nte_value} {nte_type}. Inferred from {ref_n_contributors} providers across {ref_n_hs_contributors} health systems sharing this network, {ref_agreement_rate} of whom share this stoploss structure. Threshold based on the median threshold of {source_threshold_value} (from provider id {source_provider_id}), adjusted to {calc_threshold_value} using this provider's cost-to-charge relativity of {ref_ccr_relativity}.

Example: First dollar stoploss once the dollar threshold of $112,500 is met, with a percentage reimbursement of 85%. Inferred from 12 providers across 5 health systems sharing this network, 91.67% of whom share this stoploss structure. Threshold based on the median threshold of $150,000 (from provider id 1093847), adjusted to $112,500 using this provider's cost-to-charge relativity of 0.7500.

Score 3 — payer_network_state_agg

{stoploss_type} stoploss once the {threshold_type} threshold of {calc_threshold_value} is met, with a {reimbursement_type} reimbursement of {reimbursement_value}, not to exceed {nte_value} {nte_type}. Aggregated from {ref_n_contributors} providers in the same payer/network/state group, {ref_agreement_rate} of whom share this stoploss structure. Threshold based on the median threshold of {source_threshold_value} (from record id {source_id}), adjusted to {calc_threshold_value} using this provider's cost-to-charge relativity of {ref_ccr_relativity}.

Source Note: "{source_note}"

Example: First dollar stoploss once the dollar threshold of $160,000 is met, with a percentage reimbursement of 85%. Aggregated from 8 providers in the same payer/network/state group, 87.50% of whom share this stoploss structure. Threshold based on the median threshold of $200,000 (from record id 2204719), adjusted to $160,000 using this provider's cost-to-charge relativity of 0.8000.

Source Note: "Inpatient: stoploss provision at $200,000; 85% of billed charges."

Score 2 — payer_network_national_agg

{stoploss_type} stoploss once the {threshold_type} threshold of {threshold_value} is met, with a {reimbursement_type} reimbursement of {reimbursement_value}, not to exceed {nte_value} {nte_type}. Aggregated from {ref_n_contributors} providers nationally under the same payer/network, {ref_agreement_rate} of whom share this stoploss structure; median threshold of {threshold_value} sourced from record id {source_id}.

Source Note: "{source_note}"

Example: First dollar stoploss once the dollar threshold of $175,000 is met, with a percentage reimbursement of 80%. Aggregated from 21 providers nationally under the same payer/network, 80.95% of whom share this stoploss structure; median threshold of $175,000 sourced from record id 3310482.

Source Note: "First dollar stoploss at $175,000, reimbursement at 80%."

Score 1 — state_agg

{stoploss_type} stoploss once the {threshold_type} threshold of {calc_threshold_value} is met, with a {reimbursement_type} reimbursement of {reimbursement_value}, not to exceed {nte_value} {nte_type}. Aggregated from {ref_n_contributors} providers across the state, {ref_agreement_rate} of whom share this stoploss structure. Threshold based on the median threshold of {source_threshold_value} (from record id {source_id}), adjusted to {calc_threshold_value} using this provider's cost-to-charge relativity of {ref_ccr_relativity}.

Source Note: "{source_note}"

Example: First dollar stoploss once the dollar threshold of $93,750 is met, with a percentage reimbursement of 85%. Aggregated from 44 providers across the state, 79.55% of whom share this stoploss structure. Threshold based on the median threshold of $125,000 (from record id 1198023), adjusted to $93,750 using this provider's cost-to-charge relativity of 0.7500.

Source Note: "Stoploss applies at $125,000; 85% reimbursement applies thereafter."

Score 0 — national_agg

{stoploss_type} stoploss once the {threshold_type} threshold of {threshold_value} is met, with a {reimbursement_type} reimbursement of {reimbursement_value}, not to exceed {nte_value} {nte_type}. Universal national fallback aggregated from {ref_n_contributors} providers, {ref_agreement_rate} of whom share this stoploss structure; median threshold of {threshold_value} sourced from record id {source_id}.

Source Note: "{source_note}"

Example: First dollar stoploss once the dollar threshold of $150,000 is met, with a percentage reimbursement of 85%. Universal national fallback aggregated from 312 providers, 76.28% of whom share this stoploss structure; median threshold of $150,000 sourced from record id 9900001.

Source Note: "Inpatient stoploss: $150,000 threshold, 85% reimbursement, first dollar basis."

On this page: