SMVaddadi/propensity-data-apra-phi-gap

apra_phi_gap — mirror

A convenience mirror of a public dataset, pushed by propensity.data.hf_mirror so that benchmark runs do not depend on a flaky upstream origin.

This is a mirror, not a grant of rights

Copyright and licence remain with the original source named below. Mirroring moves bytes; it does not and cannot change the terms they arrived under. If you intend to redistribute this further, or to use it commercially, read the original terms — not this card, and not the license: field above, which is our best reading and may be wrong.

Provenance

• Source: (c) Australian Prudential Regulation Authority (APRA). Quarterly Private Health Insurance Medical Gap and Medical Services statistics, June 2026. Licensed CC BY 3.0 Australia per the notice inside each workbook's own Notes sheet (APRA's website states CC BY 4.0 International and its NCPD publication page states 4.0; the in-file notice governs the file it ships inside). APRA does not endorse this repository or any use of this data -- non-endorsement is a condition of the grant -- and accepts no responsibility for accuracy, completeness or currency. WHAT THIS IS, AND WHAT IT IS NOT: an EXPENDITURE AGGREGATE with NO PAYER ENTITY. No insurer is named anywhere in either Data sheet; the entity axis is state x gap type x MBS fee band only. It is therefore a FEATURE/covariate source and NEVER an insurance outcome source -- the cells are pre-tabulated dollar and service-count totals with no claim line, no dispute and no adjudication decision. An earlier triage in this programme called it payer-claims-adjudication and an outcome source; the audit overturned both and this card does not re-assert them. MEASURED: medical_gap arm 13,748 rows x 13 cols over 124 quarters, 1995-09-30 to 2026-06-30 -- the longest time axis in the health collection. Four co-observed outcomes in one row (amount charged, Medicare benefit, fund benefit, number of services) plus patient_out_of_pocket. medical_services arm 163,620 x 9 over 77 quarters, 8 states x 18 specialties x 3 gap types x 5 line items. THE OUT-OF-POCKET RESIDUAL IS APRA'S OWN PUBLISHED FIGURE, verified rather than assumed: it reproduces APRA's Gap column to 0.000000 across 45 published figures in sheets T1-T9, and the derived no-gap proportions reproduce APRA's own TA table to 0.0e+00 across 117 figures. But it is NOT clean at row level -- 696 of 13,748 rows (5.06%) have a NEGATIVE residual from APRA's own rounding (its Notes sheet says details may not add to totals), and of the 3,612 rows that are zero to within half a cent only 3,275 are exactly 0.0, so == 0 is the wrong test by 337 rows. FOUR MORE MEASURED CAVEATS. (1) The gap-type x fee-band cross is 14 pairs, NOT 15: Known gap agreement never co-occurs with Up to MBS Fee, structurally, so a reader expecting 120 rows per quarter reads 112 as missing data. (2) An INTERIOR 10-quarter hole: ACT is absent entirely from 2007-Jun through 2009-Sep in both arms (114 quarters against 124), so a pivot silently NaNs and a lag straddles 2.5 years. (3) THE TWO ARMS DO NOT RECONCILE -- they publish the same four quantities by quarter x state x gap type, and on the 1,818 overlapping cells 8.69% differ by more than 1%, the worst by 75.2%, with only 134 agreeing to the cent. Do not stack or join them. (4) Count columns are not integral (7 fractional values in number_of_services), and in the services arm No of gap services exceeds Total number of services on 1,291 of 32,724 cells -- APRA's file violating its own containment -- so feeding both wide puts a gap-rate model's numerator in its features. Checked and ABSENT: no suppression markers at all (0 of 13,748 and 0 of 163,620 value cells fail numeric coercion, and there is no Notes column), no total/all-states aggregate row, no constant columns, no label re-casing, and one single schema across all 124 quarters. WHAT THIS IS NOT: not claim-grain, not an insurer panel, not an outcome source, and not a grant of rights by this repository.
• Origin URL: https://www.apra.gov.au/news-and-publications/quarterly-private-health-insurance-statistics
• Loader: propensity.data.public_loaders.load_apra_phi_gap
• Extract: full pull from origin
• Reported source: apra_phi_gap
• Rows pushed: 13,748
• Columns: 13

Licence and redistribution

• Licence as found: other (cc-by-3.0-au)
• Redistribution verdict: open
• Basis: APRA, CC BY 3.0 AU. Read 2026-11-10 VERBATIM from EACH workbook's own licence sheet: "(c) Australian Prudential Regulation Authority (APRA) 2024 / This work is licensed under the Creative Commons Attribution 3.0 Australia Licence (CCBY 3.0). / This licence allows you to copy, distribute and adapt this work, provided you attribute the work and do not suggest that APRA endorses you or your work." Found in the Notes sheet of Quarterly Private Health Insurance Medical Gap June 2026.xlsx (4,224,785 bytes, SHA-256 7dd93337...) and in the Notes sheet -- WITH A TRAILING SPACE, a real difference between the two files -- of Quarterly Private Health Insurance Medical Services June 2026_0.xlsx (10,557,105 bytes, SHA-256 fc89a536...), and independently out of each xl/sharedStrings.xml. Attribution AND non-endorsement both travel; both are conditions of the grant. VERIFIED PER FILE, not inherited: APRA's website says CC BY 4.0 International, its workbooks say 3.0 Australia, and its NCPD publication page says 4.0 -- so a publisher-level verdict would record the wrong VERSION for some file whichever it chose. The in-file notice governs the file it ships inside, so 3.0 AU is recorded here: the LESS permissive reading. TWO HONESTY BOUNDS. The (c) year is 2024 on a June-2026 publication, so the notice is stale boilerplate rather than a per-edition re-assertion. And the two files type the licence URL differently (http://creativecommons.org/... versus www.creativecommons.org/...), which shows they are separately maintained notices rather than one template -- one more reason not to inherit. No personal data: pre-tabulated dollar and service-count totals, no payer entity, geography is state/territory.

Citation

Australian Prudential Regulation Authority (2026). Quarterly Private Health Insurance Medical Gap statistics / Medical Services statistics, June 2026. Licensed CC BY 3.0 AU; APRA does not endorse this use.

Loading

from propensity.data.hf_mirror import load_mirrored

result = load_mirrored("apra_phi_gap", org="SMVaddadi")
result.frame.head()

load_mirrored falls back to the original loader (origin, then its own offline replica) if this mirror is unreachable, so it is safe to depend on.