National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

Submission 1567

Submission to the Senate Community Affairs Legislation Committee

Inquiry into the National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

Submitter: An AHPRA-registered psychologist and PhD candidate. Name-withheld status requested. Disclosures: The author has no financial or contractual ties to the National Disability Insurance Agency —(NDIA)”, to Centre for Disability Studies’(CDS), or developers of the i-CAN tool used within some schemes participants. as provider services to some scheme participants could indirectly affected provisions affecting private-sector providers.AI tools assisted preparing submission; all statistics, citations quotations were manually verified against original sources,and is responsible accuracy.

Lay Summary

This submission examines scientific evidence behind assessment tool that will be deployed as part of the National Disability Insurance Scheme(NDIS).A scientific analysis shows serious methodological procedural problems with current available research on reliability validity,i-can scores accounted less than funding people receive over other factors determined by results.i-cans ability reliably measure participant needs was poor uncertain feasibly estimated precision cannot rule out failing meet success criteria low-stakes observational research.current design limitations include too few participants power studies have appropriate levels given human stakes.Submission body sets these issues detail while Appendix calculates number required minimum contemporary threshold high stakes assessment estimates.These findings do not support NDIs characterisation gold-standard in disability needs assessment (NDIA 2025) and does its use contexts further published evidence show reliable valid translating i-can into funding amount.Bill also requires government publish how i-can translated into a funding amount.The submission examines additional risks associated proposed system including changes assessor workforce sidelines involvement federally registered clinicians domain-relevant expertise regulation systems reductions rights appeal available participants need stronger regulatory oversight regarding invisible software circumvent clinical determinations impairment support NDIS plans omissions Bill’s Statement Compatibility Human Rights raises possible tension Australian Constitution may leave service providers vulnerable civil conscription.This submission asks committee delay relevant Schedules until both instrument allocation methods properly validated for intended use. Until this occurs deployment carries real risk material harm to Australians disabilities families.While the submission is quite critical of tool context intended use-case would like affirm promising warrants continued development wish praise CDS staff commitment Aussies facing disability work towards innovation impacts hope detailed engagement interpreted sign respect authors serve constructive aim assisting future endeavours.

National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

Submission 1567

Submission to the Senate Community Affairs Legislation Committee

Inquiry into the National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

Executive summary

This Bill commences an algorithmic system to set NDIS funding for more than 700,000 Australians. The validation evidence fails developers’ benchmark; cannot reliably predict allocations made using this method; central instrument has no public validation data on version deployed, algorithmic budget translation never disclosed.The closest published base does not meet reasonable standards even low-stakes observational research, let alone high stakes use.Meanwhile mechanisms detecting correcting errors participant plans are being narrowed same legislation. The submission’s appraisal of literature identifies additional concerns.i-CAN conceptual approach is tension with goals needs assessment.Test-retest reliability poor.Classification accuracy did not clearly meet minimums cited开发者observational research.In study where i-CAN scores were compared actual funding allocation predicted less per cent variation. Monte Carlo simulation analyses in Appendix A demonstrate that across studies statistical power detect relationships actually observed was well below contemporary scientific journals expect low-stakes observational research precision many coefficients.None covered version will be deployed and peer-reviewed validation studies available for v6 or NDIA customised variant.These arise structural difficulties Administrative Review Tribunal reports a change rate challenged decisions most recent reporting period.Submission sets out empirical case detail drawing novel precision analysis (Appendix).It also examines serious workforce provisions subtle legal-linguistic hands over clinical duties automated software argues means the pathway should considered against Therapeutic Goods Act 1989 as medical device.Finally points lack evidence assessment Statement Compatibility Human Rights appears engage Bill central provisions further identifying possible Australian Constitution leaves service providers vulnerable civil conscription.It recommends nine amendments allow substance reform proceed once its scientific ethical foundations adequately independently demonstrated:

• Recommendation Defer commencement Schedule dependent until independent scientific validation instrument, full NDIS population demographics published peer reviewed assessed expert panel no ties to instrument developers NDIA.
• Recommendation Amend Schedule require primary legislation publication budget method’s key details including metrics algorithm open reasonable scrutiny regulatory authority e.g., TGA recommendation before any participant is assessed under it.
• Recommendation Amend Schedule requirement all support needs assessors registered health practitioners Health Practitioner Regulation Agency relevant scope practice existing legislation codes of conduct.
• Recommendation Amend Part restore obligation consider independent medical allied health evidence planning clarify new hierarchy operates guide rather than precondition.

National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

Submission to the Senate Community Affairs Legislation Committee Inquiry into the National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

Submission: National Disability Insurance Scheme Amendment Bill Inquiry

Table 1 - Bill provisions engaged by this submission

|Bill provision | What it does | |-| s s 9 B (Schedul e) 1 P art 1) | Statutory definition o f functio nal capaci ty , requirin g assessment se ction s 3, 9 ' i n context t hat excl udes as far possible th e impact fo rth person ’ s environmental an d per sonal cir cumstances ', whic h appears co nflict with commitmentt othe social modelo fdisability. s 32K(1A) (Sc hedule 4, Item ); ( S chedu le 4, I tem 3); ( Sc he dule 4, It em C) (Sch ed ul e 4, Item D)")"; "sc hoold be more than equal to or less than actual cost of supports s L(6)(a) (Schedule Items and 0))“, Replaces ‘identify’ w ith ‘include information about in the assess ment-report obligation (Item ), a nd requires relevant support needs arise directly from participant’s impairment(Item ). Functionally circumvents need for clinical input restricts suppo ts only whatever partitionof persons impairments are presumed attributable recognised disability. s 32L(4A) Schedule (4, Item ); NDIS rules Specifies who may undertake assessement permits under classified Minister prescribe additional assessors limited parliamentary scrutiny. s 32L(4)(b) to(c) (Schedul e 8), Specify section 7 information asessor must not have regard to in NDS rules EM p .. Note: The text is extracted mechanically with possible OCR errors which might affect readability but has been preserved exactly here.|Permits NDI S r ules apply adopt incorporate Agency prepared materialas force existing time’.| Establishes hierarchy peer-reviewed research outranks clini cal evidence may on its own defeat support structure separates what good practice combines integration scientific literature eviden ce person does engage substantial gaps Australian dis ability lit erature.| s A; s A(5); (5) (Sc hedule Part ) Authorises Ministe by legislative instrument reduce funding specified groups supports old framework plans during transition period effect even if less total cost reasonable necessary support subject requirement regarding participantsafety. New Division of P art Chapter ss B E(Schedule 3 Part 2) authorise computer programs take administrative action including evaluative determinations about needed reasonably compensate disability related impairment functionally replacing clinical decision makers. Schedule limits unscheduled reassessment raising threshold significant and ongoing change functional capacity substantially reduces daily activities implementation details left transitional ministerial rules. s Schedule Transitional Ministerial Rules duration constrained parliamentary disallowance

National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

Submission 1567

Submission to the Senate Community Affairs Legislation Committee

Inquiry into the National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

About this submission

I am an AHPRA-registered psychologist and a PhD candidate with interests in psychometrics and the philosophy of science. Over the past year, since I have examined the published evidence base for the Instrument for Classification and Assessment of Support Needs ( i-CAN), which the National Disability Insurance Agency selected as basis to support needs assessment described as ‘the gold standard’ available validated tools’ described as ‘gold standard’. as part that examination conducted precision power analyses validation studies i- can clarify level confidence supports use cases contemplates Appendix reports analysis Section summarises them alongside additional critique study designs Analyses intended constructive CDS NDIA planning future sample sizes controls This represents best efforts compile conclusions articulate arguments within highly limited daily schedule does not represent any employer university professional body personal reasons requested name-withheld status The contribution is place public record what scientific evidence currently supports identify methodological conditions under central provisions could be made fit their contemplation invite committee consider present ask satisfy empirical ethical foundations before assenting commence them The Bill’s Central Pathway Based reading, empowers algorithmic pathway funding NDIS participant supports through four interlocking components Functional capacity new section Schedule Part introduces statutory definition functional excludes possible impact person’s environmental circumstances Assessment report budget method (new sections 32L(6)(a)and SCHEDULE revises needs assessment so includes information about disability rather identifies it provides budget method specified in rules translates into total amount Below-cost Funding new Sections 4 Item inserts new section authorising any funding actual cost providing acquiring support. This below-provision for framework plans also insert Minister legislative instrument reduce funding group of old during transition period effect even if result provided reasonable necessary support less than the total cost support must undersection have regard to safety participants making determination clear entails anything beyond simple acknowledgment Automation new sections Schedule part inserted a Division computer programs take administrative action including evaluative determinations leaves implementation details transitional ministerial subject duration constrained parliamentary disallowance

Technical Components:

Submission: National Disability Insurance Scheme Amendment Bill

Inquiry into the National Disability Insurance Scheme Amendment - Securing NDIS Future Generations)

Submission 1567

Empirical Foundations of Assessment Instrument

Both Australian & International scientific communities accept measurement tools like iCAN’s validity depends on several international standards. The three foundational principles are:

• Validity recognized as property specific version/test used purpose population; not abstract test name or changes in any one material impact validitiey.
• When instrument revised evidence does not transfer automatically from prior versions but each new version must obtain own evidencce before empirical support can be made regarding validity statements;
• For use with particular populations and purposes requires establishing validity specifically tailored to those groups. P.S.: The Psychology Board’s Code of Conduct mandates psychological services including assessment tool selection should align these standards informed respective consensus science.

National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

Submission 1567

Submission to the Senate Community Affairs Legislation Committee - Inquiry into the National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

Table 2: What the Published i-CAN Validation Studies Establish

Construct	What The Studies Asked	Published Result	Power	Bottom Line
Test-retest reliability	Does the same person score the same on second occasion? Core Health & Well-Being domains r ≈ −0.22 to +0.4; Activities Participation domains r =+0.38 to .94	Unreliable in core health well-being domain stronger activities participation though measured sample unlike norming population. Headline figure rests curated quarter sample full-sample agreement may sit below developers’ own minimum.

Explains under per cent funding variation no domain reaches power floor.| 16–20% weak domains | Explains under four percent of funding variation, no domain reaches eighty percent power floor. Classification accuracy | Does algorithm assign support level expert clinician would κ=0.94 best-rated all independent | 77%; 30% full sample agreement may sit below developer’s minumum. | 0.55 (best rated); 0.65(all); 0.66(independent)| Explanates under percentage of funding variation, no domain reaches eighti pcent power floor. Funding prediction | Do i-CAN scores predict participant receives ? ρ=.13-.20(Total Raw) | Power is chance at published sample size detecting reported relationship Intervals are from Monte Carlo simulation(10 replications). Full figures and methods Tables A to C AppendixA | | Note: r & q correlations; k Cohen’s kappa.|

All Published Validation Data Are for Older Versions Of The Instrument Discontinuity in validity evidence across versions the peer-reviewed literature on concerns version through , studies conducted between .The that NDIA deployment based v6 released customised non-public derivative Centre Disability Studies [CDS] confirms there have been no validation publications any since Under standards set above cannot be expected transfer v or its customisation given substantive changes were reported place each therefore requires own data.

NDIA proposes a power change their assessment process whim Bill provides statutory mechanism authorises this gap Explanatory Memorandum confirms Schedule Item NDIS rules specifying support needs tool ‘may apply adopt incorporate matter contained document prepared Agency such questionnaire manual as force existing time’.Memorandum states purpose drafting ensuring updates inputs content assessments can made real-time appears by which customised non public version i CAN deployedAgency modify without modification appearing face of NDIS rules subject disallowance rule.Allowing NDIA modifying processes careful analysis impacts introduces risks unreliability invalidity also replaces provisions require assessments regard other information participant Given best-case analysis asks whether published evidence taken at value granting every benefit doubt would support proposed use cases scientific point view unpublished current versions improve position additional validation data publicly available evidence them.| has_redactions: false, is_low_confidence: true|

National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

Submission 1567

Submission to the Senate Community Affairs Legislation Committee

Inquiry into the National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

Existing evidence is split across different versions of the instrument.

• Published validity evidence itself quite fragmented; inter-rater reliability data available in v2 &3 [Riches et al., 2009b]; criterion validity: version4.0&4.2[Arnold etal.;test-retest from a research-only short form,iCAN-Brief[Arnoldetal., 2015].MentalEmotionalHealthdomaincentraltoassessmentofpsychosocialdisabilitysubstantiallyrevisedbetweenv2andv3changesauthorsacknowledgebutdonotdetailwithothermodifiedversionsappearinglaterregardingpredictingappropriatefundingamounts.Giventhis,fromscientificperspectivepublishedevidencecannotprovide sufficient validation any singleversion i-CANinstrumentforapplieduse.The deployedinstrumentsitsNDIAcustomisedvariantareatleast two design generations beyond latest published evidencenecessarily currently lackany evedence reliabilit orvalidity.

The older versions’ evidenceis too thin support high-stakes use

By Reliability I refer property an instrument which it produces consistent results when same person assessed ondifferent occasions absence real change Low reliability funding context means persons funding can shift substantially between assessments even their needs have not changed Table below presents test re-test reliability figures reported the published studies This section explains how to read thosefigures why worst rows matter most and betterrows do no savethe for its proposed use.Two distinct questions For assessment ’Reliabili covers twodistinctwhich easily confused first is whether tworaters observing samassess session agree scoring inter-rater reliability second independent administrations of assess with different teams interviewing participant separate occasion yield similar resultstest-retest. Inter rater necessary butnotufficient condition deployment sense If raters watching cannotagree scoring rubric broken if theycanagreerubric workable tells nothing about whole process run by different teamsonoccasion primary questionwhetherassessmentyieldconsistentfunding relevant outputs acrossadministrations dealswith test-retest reliability.Published literature reports each Riches et al 2009b report overallinterrater reliabilityof r= (14 cases,3 ext{ }items) based design one trained researcher independently scored sessions observed facilitator-ledteam confirmsscoringrubricworkablebut doesestablish thattwoindependentadministrationsofsessionwouldproduce similarscores Arnoldetal(2014report inter-rater rs = n=89,25), clinical judges served as criterion comparator classification study a separatequestion from i-CAN’s own scorbehaviour discussed below The figures speak Bill turns on reportedbyRichesetal.(2009bsubject restsection.

• The polling analogy Imagine organisation surveys people andreports65 per cent support particular policy same poll were run againdifferent theresult could easily be or percent Each figure for those twentypeople is real only itcannotbe reliedon reliable indicator nor valid representative average Australians supportpolicy publishedi CANstudies do not confidence intervals statistics butdo not statistical powerandtheynot provide simulation-based precision analyses calibrated against use-case appropriate thresholds Appendix A providesanalyses first time clarifying how preciseeachstudy results are likely detect effectsreported.##

National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

Submission 1567

Submission to the Senate Community Affairs Legislation Committee

Inquiry into the National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

• Confidence intervals and statistical power.

• The confidence interval isa margin of error. Any value withintheintervalis95%consistentwithwhatthestudyobserved, meaningthedataalonecannotdistinguishamongvalueswithintheinterval.Awideintervalexhibitsimprecisedataandconsistencyinverydifferentconclusionsaboutthescientificpropertyofi-CAN.*

  o Statisticalpowerisanseparatemeasure:itistheprobabilitythatastudynumberparticipantswoulddetectarealrelationshipifonewerepresent.Theconventionalminimumforexploretoryresearchis80percent(Cohen,1988*); thresholdappropriatetohigh-stakesfundingdecisionswouldlikelybecloser-to-95percent.Studiesfallshort-of80percenstatisticalpowernotcanreliablysupportexploratoryconclusionssaltohenhighestakesones.Thisstandardnowreflectedmanysciencejournalwhichdonotconsiderpublishingstudyunlessitisdemonsratedatleast80%statisicalpowerforreportresult.Lowpowerhassecondlessobviousconsequencebearsonhowicanevidencebasedshouldberead.ResearchfindsunderpoweredstudiestoincludehigherproportionlargerfalsepositivesthanadequatelyPoweredOnes(AppendixAStatisticalPower).Becausethiswhere statistical power islow for published iCAN coefficient maybestreadplausibleupperboundonwhatan independent replication would find rather than best estimate of true relationship.

Domain	Retest interval	n	r	**** *95* % CI ****	Power	What the data can establish
Mental Emotional Health	2 years			−.22		[−.62, .24]
Behaviour			.25	[``-.`‘’.21’, ‘.64’]	20%	Spans slight disagreement to strong agreement;reliability not demonstrable.*
PhysicalHealth			-.21	[-. ‘.21, ’.’59]	`{16}%’	Cannot distinguish noagreement from moderate agreement.*

Note.Riches et al.(2009b,Table)seeAppendixaTablesA10and A11.Thethree weakest and most psychosocially relevant domains are shownActivitiesParticipationdomainsperformedbetter(r=0to.94butwere measured on a sample unlike iCAN’s norming population. • The headline finding.Themostconsequentialrowin Table 2a is first.Whenthesame participants werereassessed twoyears after their original interviewon the MentalEmotionalHealdomain ofi-CANthecorrelation between twosescores was r = −-.22.Acorrelationof 1.0 would indicate perfect agreeement;a correlationofofzeroindicatesthatthe two assessmentsare completely unrelated`

   o Published− .22sits slightly below zero:slightly worse thanunrelated.Confidenceinterval runsfromr=`-.`''.**62`, `.+``*``` to +`.24which meansdataconsistentwith anything strong disagreement moderategreement*

• In terms statistical power study had chance detecting whether this domain produces consistent scores at all.Reaching even80percenstatisticalpowerwouldrequire about ```{20}participantsThestudyused20.*

• **Pattern across Health WellBeing.Behaviourattwo yearsPhysicalHealthatsixtodecembermonthsbothproduceconfidenceintervals that spanzerelationship with statisical power respectivelyAcross these analyses|

Submission 1567

Submission to the Senate Community Affairs Legislation Committee - Inquiry into the National Disability Insurance Scheme Amendment Bill 2026

The published data cannot rule out the possibility that the instrument yields no statistically reliable or scientifically valid information about the person being assessed. • Developers’ Explanation: Developers attribute instability due to real change between assessments rather than measurement error but it can’t be tested with available data as validation study didn’t report whether participants who experienced documented health changes showed score consistent those changes; without reporting such info., we don’t know what caused these fluctuations—real change vs measuring errors?

Activities & Participation Domains:

The reliability correlations are good at around .8-9 range for activities domains yet face another issue: i-CAN was developed on a group of people mostly having intellectual disability while smaller groups checking its validity had different profiles like physical disabilities and sensory issues. Thus, cannot apply results obtained in one population directly onto others because they’re not representative samples.

Composition Of The Normative Sample:

Of original development sample, most (around ~3/4) were living congregate residential settings which is problematic since there’s lack specific validation studies covering autistic individuals, psychosocially disabled persons etc.; however despite this limitation NDIA seems empowered by current bill provisions using i-CAN precisely against these populations.

Criterion Validity Headline Figure Depends On Selection From A Small Sample

Table 2b - Classification Accuracy (Arnold et al.) | Subsample | n | κ | CI | Power | |-|-|-|-| | Best-rated reports | (of186)** ||||| | Full Validation | 186 |0.65| [0.56,74]|30%**| | Independent | 41 |.66*| [0.46*,,83]|13% The data can establish that even on the developers’ best rated quarter of their sample classification accuracy was below a power floor; while for full validation and independent subsamples it’s lower bound sits just under minimum threshold set by developers themselves indicating poor agreement between expert clinicians’ assessments & algorithmic classifications with only about one-in-three chance to clear them in case they were representative samples as per Arnold’s study benchmarks differ across rows: high-stakes floor vs observational-research minima respectively.

National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

Submission to the Senate Community Affairs Legislation Committee

Inquiry into the National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

than chance) to 1 (perfect agreement). In their papers, the developers provide citations suggesting that a kappa of 0.6 is the minimum acceptable value for low-stakes observational research (Suen & Ary, Cicchetti, and McHugh are cited benchmarks: (\kappa = {0}{75}\text{for ’excellent’agreement},~and ~\kappa= {.8} text{‘strong’});the latter being reasonable floor for individual classification with high stakes consequences. The most widely cited figure for the i- CAN’sclassification accuracyis κ={94}(Arnold et al.,[2014]). Two features derivation warrant committee attention as seen in Table Table .

Selection.The {94}figure drawn from carefully curated subsetoffrom total assessment reportsvalidation studythose team rated good or verygood quality.

• When all were included,kapfafelltowith confidence interval [0.{56},{0}.{74}];lower bound this range sits below threshold themselves citeacceptable forobservationalresearch’.

The independent validation sampleyielded kapfawithconfidenceinterval,a range too wide distinguish between instrument adequate scoring consistency one marginalconsistency.'At N 186study had only statistical power demonstratethat exceedsthreshold.This means that their studyhad a chance correctly identifying whetherinstrument’s truescoring consistent met lowest acceptable bar low-stakes, observational research.Even at developers’ favoured subsample whereκ=onlystatisticalpower againsthighstakesthreshold ofreachedbelow minimum exploratory research well short value(e.g. ,)thathighstakes funding determination would callfor. Criterion contamination.i-CANs classifications not testedagainst anindependent comparison.To testwhetheran instrument produces accurateclassifications about functional need thetest requires source right answer: experienced clinician assesses person directly without seeing instruments output.In i- CAN study attempted toassessthisquality(i.e., criterion validity),comparison made by assessor who although making judgement independent itself drew oniCAN'sassessment reportsbasisjudgementperson levelneed(Arnold et al.[2014]).Instrument andthe comparison therefore working same information which is formofcriterioncontaminationAuthors tested accuracy using data generatedbyTable`.As result this cannot count as verifying whether iCANS scorings accurately classifiesfunctional needs.Itisatest two summariessame i CANData agree While usefulinformation doesnot establish that places people inright categorytwo readings sametoolsoutput reach similar conclusionsaboutthesameoutpoot.

Submission to the Senate Community Affairs Legislation Committee

Inquiry into the National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

Section: Funding prediction analysis based on Arnold et al., 2015 study.

Table 2c - Funding Prediction:

| Score / domain | n ρ CI Power What can be established | |-|-|-|-|-| i-CAN Total Raw Score   redacted   [0.04, 0.;]
69%

Explains under 4 % variation in funding allocation

<table border=“” style=

National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026 Submission

Submission: 1567

Inquiry into the National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

Model Instability.

The predictive models in Arnold etal.* (2015) sit below the conventional ratio of 15 observations per predictor recommended for stable prediction (Harrell, ,,,2015). With the standard i-CAN + demographics model achieving only .8.observations per predictor, the stepwise model reaches this recommendation because variable selection reduced from .9 to .4. within sample. Recognition due to Authors.Although they did not report simulation-based precision power analyses themselves, developers nonetheless appreciated significant limitations data their own interpretations; on record expressing disappointment early funding-prediction attempts (Riches et al., ,)and appropriately noted that larger sample necessary develop sufficiently precise allocation algorithm (Arnold et al.*,***,**). Post-hoc Prediction.One other feature study warrants committee’s attention.Instudydevelopers reported one variable showed substantially stronger funding prediction $(R^2 =.8.)$. What needs appreciate however was constructed post hoc based funded items’ data here items re-coded according whether support each item described through Disability Services.Three features limit what result establishes and prevent it establishing value i-CAN scores NDIA interested.First coding applied after fact as part standard administration assessor workforces would receive training.Second criterion (total funding) both index same allocations decisions different levels detail so correlation partly reflects internal consistency already allocated which is substantial difference instrument’s capacity determine appropriate allocations new participantsThird used stepwise regression outlier removed standard error resulting predictions approximately $14,00 across allocatings ranging near-zero over $.Developers themselves noted findings require independent verification (*(Arnold et al.,,**)).To date such verification has happened.Put plainly altogether this shows in study attempted use older version of the i-CAN predict funding Developers best result shown could recreate funding decision had been made though even best demonstrated average prediction error around $ 14,00 per participant Thatdatahowever did not showinstrumentcoulddetermineappropriatefundingnewparticipants The latter question matters for an instrument empowered by Bill (Parliament Australia,***,***)to determinefundingfornewparticipantsonnational scaleanditwasnotansweredintheses studiesDepending on how current versions instruments changedwould be unreasonable to hypothesise fromthis datathati-CANScoreswillnottaccuratelypredictparticipantr’sfundingsneeds at least notbythemselves

No Independent Replications Measurement Tools CompletedConditions Scientific Credibility.For scientific data deemed credible within community expected replicable independently parties without conflicts interest Affecting independence are thatAll published validationstudiesontheic-AnhavebeenconducteddevelopmentteamLead author holds royalty agreement with CDS regardinguseofthestudyInstrument(Arnoldetal.,2015) whileCDS owns intellectual property and is consortium partner NDIA’simplementation.

National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

Submission 1567

Submission to the Senate Community Affairs Legislation Committee

Inquiry into the National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

• In [20 years], no independent replication of the instrument’s scientific properties has been published. Authors acknowledged the need for more rigorous study. During one of their evaluations of i-CAN validity, Riches et al., \((2009b\)) themselves anticipated the need for independent replication and more appropriate controls in their study designs, remarking how at the time that funding was not conditional on the outcome, and if the results did determine funding that much more rigour would be required.

• The NDIS deployment represents a material increase in conflicting interests and is precisely the kind of scenario that the developers appear to have alluded to, onethat requires more rigor before this measure can be deployed.*

• To date there has been no research using any control as necessary by authors before deploying it reasonably for purposes intended under Section 4.3.8 The instrument produces different scores for same underlying needs* The iCAN, based its design philosophy assess support needs context person specific environment counting only items relevant to that environment (Stancliffe et al.,[2016]). From theoretical point view appears novel genuinely meritorious proposal with philosophical sophistication distinguishes from alternative measures functional need Other tend assume stable trait like state operates largely independently environmental contexts whereas appreciation key feature CAN As property clinical support planning instrument therefore well-defensible logic warrants further scientific study However theory aspiration seems conflict Bill’s use i-CAN Property instrument used consistent funding participants national scheme commitment supports arising persons’specific environments introduces structural problem two individuals enduring similar but differing circumstances may end up dramatically different total score i-CAN method proposed budgeting allocations receive completely different amounts funding Bill text moves opposite direction Item Schedule amends paragraph (a) require relevant arise directly participant's impairment(s), leaving deployed instrument sensitive design tension authorising its use A supplementary could standardise across environments however such module itself would require published validation evidence both measurement properties and combination predict funding##The budget translation method not disclosed Bill enables a budget method specified in NDIS rules translates assessment outputs into funding allocations \((Schedule Items .“))”. New section 32K(3B)authorises the method identify levels of need for supports specify funding each level while new subsection 32K(3Cauthorises those to be 'more than, equal actual cost'of this support. Methodology by which will derived basis on setting against these levels supporting either have been neither publicly available nor made Analysis documents obtained under Freedom Information legislation indicates that NDIA already employs machine learning models generate draft budgets from profile data (van Toorn & Carney, [2024]`). Model linking assessment outcomes has not reviewed independently preventing independent scrutiny scientific properties.

National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026 Submission 1567

Submission to the Senate Community Affairs Legislation Committee Inquiry into the National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

Invisibility prevents public scrutiny.

The methodology behind this half of the pathway is therefore currently invisible to the public and to the scientific community. In principle, this prediction model can be examined and challenged by a trained practitioner when it yields an invalid outcome. However, the translation into funding appears to occur within a system that, by design, cannot be examined independently by appropriately trained experts. In practice, this will likely mean that:

• A participant who receives a lower funding allocation than expected or needed has no way to know whether the shortfall lies in the assessment, in the budget method, or in their interaction. Without meaningful human oversight, determining the source of a shortfall would be difficult, if not impossible for the NDIA itself. It is not possible to meaningfully challenge what cannot be seen. The need for ‘test security’. One legitimate reason for not publishing the details of this algorithm may be ‘test security’. If too much information about the budget allocation method were publicly available, it would be at risk of being gamed by malicious actors. However, if the NDIA does not publish anything about this method, it leaves the public and the scientific community without any reasonable assurance that the method is fit for purpose, and not unreasonably disadvantaging certain populations. This may seem to leave the NDIA in a bind.

• Without public assurance that the allocation model is fit for purpose, disabled Australians are left vulnerable to a system that, if not fit for purpose, would be structurally unable to be challenged. On the other hand, standard mechanisms of public assurance (e.g., detailed technical publications) themselves risk undermining the scheme’s effectiveness by making it vulnerable to malingering. Possible solutions to the test security problem. Several options are available and worth considering:

• The NDIA might simply choose not to disclose any information about the budget allocation method; though this would substantially undermine public confidence in the scheme and presiding government; or, alternatively:theNDIAMightchoosetopublishsummarystatisticsonvalidationeviencedregardingitspredictiveaccuracyonsout-of-sampledata;thiswouldhelpmitigatetheopacityconcernthoughthistimetechnicaldetailssecureforexample,theTGA.IrrespectiveofwhatthenDIAchoosetodoaboutthisissue,thissubmission’spositionisthatestablishingadequatevaliditynotjustforthecANbut forthefundingtranslationmechanismanddoingsoinaccordancewiththekindsoftscientificstandardsoutlinedearliershouldbeconsideredanecessarypreconditionforallowingtocommencementoftherelevantSchedulesandsimplenotdesirableaccompanimentcanbyleftfortheschemetoachieveafterithasbegunmateriallyimpactingparticipants 5.Participantneedswillbeevaluatedagainstastandardofofevidence thattheadministrativeprocessitselfdoesnotinmeet Schedule1Part6establisahasahierarchyoevidencetypesCEOmustconsidertodeterminewhetherasupportisanreasonableanda necessary, with published peer-reviewed research at top. As a strong adherent of evidence-based practice scientific standards and scientist-practitioner model I consider applying peer reviewed standardsto the allocation public resources to be legitimate aim The way provision has been drafted however does not achieve thisaimThefollowingissues should bethenoted.

National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

Submission 1567

Submission to the Senate Community Affairs Legislation Committee: Inquiry into the National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

Schedule 1 Part 6 Item 73 inserts new paragraph 34(1E), which requires the CEO, when considering two or more of the listed evidence types, to rank them in this order:

• peer-reviewed and generalisable research;
• evidence of the support’s effectiveness given the participant’s circumstances;
• outcomes from their previous plans; other matters considered appropriate. New paragraph 34(1F) authorises the CEO not meet satisfaction requirements based on limited category a even where b is present.

Submission

Inquiry into National Disability Insurance Scheme Amendment - Securing for Future Generations Bill 2026

Submission No.:

The figures are nevertheless material because both components act upon them:

soundness rate at which original decisions bear; visibility errors either kind provisions discussed below. The Tribunal held single occasion assessments less reliable than longitudinal evidence from treating professionals consistent scientific methods evaluating individuals. Empowers former assessment model. One provision will functionally reduce error visibility new system produces together two further changes indicated internal NDIA briefings tabled Senate Estimates.

• Reassessment threshold raised Schedule Part limits unscheduled reassessment raising significant and ongoing change functional capacity substantially reduces daily activities Implementation details left transitional ministerial rules limited duration subject constrained parliamentary disallowance Schedule .5.
• **Tribunal remedial powers restricted reported Internal NDIA briefings tabled Senator Steele John (Senate Community Affairs Legislation Committee) indicate Administrative Review Tribunals’ remediative powers ordered same system produced contested result Briefing entered public record through Estimating tabling Agency represented opportunity correct record.
• **Removal obligation consider independent medical evidence reported Those briefings further indicate that staff no longer required to consider independent medical evidence planning In the same briefing an senior manager stated there would requirement consideration provided NDIS participants development plans Lyons, ). A present member asked how participant wellbeing could be protected if needs not captured in this assessment under no obligation considering independent evidence you can see might shortfalls Lyons 2025). Concerns raised submission are concerns NDIA staff have internally A future reduction decision-change rate itself does produce better initial decisions equally indicating primary mechanism detecting correcting poor outcomes weakened Workforce examined Section likely structural effects scheme its participants intended or otherwise.

The workforce will administer Assessment Bill new section inserted by Schedule Item provides three categories person undertake: members of agency staff Act section , consultants engaged by Agency under section persons prescribed by NDIS rules Across all these categories bill requires assessor hold registration Australian Health Practitioner Regulation Agency relevant scope practice internal NDIA briefings indicate for Agency staff allied health background desirable but mandatory Lyons reporting on ndia Queensland staff briefing November . while category ‘persons prescribed’ is left open by Bill remainder examines two problems fall back Participants position registered practitioner who administered assessment under Bills information-handing rules and any assessor not ahppra-registered.

The AHPRA-registered practitioner item rewrites section (b) so that NDIS rules prescribe specific information must consider statutory direction about what they must not sit tension professional judgment obligations their

National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

Submission 1567

Submission to the Senate Community Affairs Legislation Committee

Inquiry into the National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

respective boards’ Codes of Conduct, which they are required to adhere to as part of their registration to practice.

• These Codes require AHPRA practitioners to consider information they reasonably believe to be relevant\nto \the caseofthepersontheyareprovidingservicesto. A subordinate instrumentthatconstrainsjudgementwouldappeartocreateconditions in whichtheadministrativepractitionerisforcedtowardseithernon-compliance withtheirAHPRACodeOfConduct,ornon-compliancewiththenationaldisabilityinsurancenetworks(NDIS).An AHPRAregistered practitionerwho isparticipatingin theneedsassessmentisthenplacedina position thatthesecodesdo notcontemplate,andthisbilldoesnotresolve.The lackexamplesintheprovision,taken togetherwithSchedule4 ITEMS7and8,suggeststhatwhat anassessormust ormustnothave regard tobecomestependentonNDISrules thatecouldbechanged ‘intraintime’, andreplacesexisting provisionsthatarequires assessments toregardanyotherinformation thatrelatestotheparticipants. Specificcontentregarding this(Item13)doseppearstoexistyet.ThesemaydirectAHPR Aregistered practitioners notconsiderinformations theparticipant provides tow them they onbasis of their expert judgements, sensiblybelieveisrelevant to theirneedswhich they arelikely required toconsider inorder ocomplywittheir conditionsof registration. The non-registered assessor The assessment involves substantive judgments about a person’s impairment\ntheir functional capacity \ndethe supportsthatwouldcompensatethedisabilitiesThesearethestandard subjectmatterofof registered healthpractitionersfor whichthereisan considerablepublicinfrastructuretoensure minimumstandards o fcompetenceandsafet y.Inthen currentproposal aworkforce thatishotrequiredtobholdeithermandatoryclinical qualificationsor relevanttertiarybackgroundswillmake determinationsabout adomainwhose contentfallswithin professional clinical scope forwhichthe competenceandsafety regulationsaresubstantial public interest.The category’persons prescribed by NDIS rulesappears tobeleftopenbythisbillandItem 4 Scheduleclassifies therulesprescribing additional assessors ascategoryD, whiche Iunderstands tobek e lightest legislativeinstrument categori basedonthegroundsthat prescribingadditionalassessors ‘will not have any significant policy or financial implications’(ParliamentOfAustralia2016ap. p .The contrast ItemScheduleplacestherulesthatspecifyinformationanassessor mustormustnothave regard too undersection32L(4)(b) and(c)) at categoryA more rigorous classification Therulemakingpowerdeterminingwho canadministertheassessment is thereforesubject to less scrutiny than the rule-making power determining what information an assessorcan consider Whetherthespecificationofadditonalassessorscarriesnosignificantpolicyimplicationsisprefectlyone ofquestions this submission raises For assessmentregimenthat determines fundingmorethan700 Australians,the policy weight deciding whoisqualifiedto administer itisanotasmall matter.Any decisiontopriorisea non-clinical workforce carries implicationforexistingworkforceo regulated health practitionersits productivity.

Submission: Inquiry into the National Disability Insurance Scheme Amendment Bill

Section Heading - Participant Recourse

The participant will have unequal recourse.

• Assessed By Registered Practitioner: The participant has equal protection against errors in their assessments if they were evaluated using services provided through Australian Health Protection Regulation Authority registered professionals as per [Ray v NDIA]([2020], AATA No.: XXXX).
• Non-Registered Assessor: The same level of assurance is not available for participants who undergo evaluations from non-AHPRA registrants due lack of regulatory pathways ensuring similar standards, such a proposal may result in material inequality between those undergoing identical assessments under this legislation.

An unregulated clinical instrument

The provisions in combination.

• The provisions discussed are interdependent: Schedule Part enables automation; Schedule , ranks evidence below research when determining supports as reasonable or necessary; The internal NDIA briefings tabled indicate staff won’t consider independent medical evidence for planning, to order reassessment by same system producing contested results.

Subtly redefining assessment as a non-clinical act.

Subtle wording shift new paragraph (6)(a), ‘identify’ needs vs ’ inclusion information’, has significant consequences: When an assessor’s job identifies,’ their report considered Clinical determination.’ When evaluation protocols used merely include info., it becomes data input applying section through method specified NDIS rules derive funding amount potentially without participant needing identified legally. This reframing acts linguistic tool workforce design examined Section engage clinically while plausibly denying constitutes clinical act Granting questionable idea assessors gather data determinations made other processes raises several questions Two issues automated budget allocation should determine reasonable and necessary supports disability/impairment-related needs not actually been legally identified, software functionally absorbs clinical reasoning duties placed stricter regulations Note fact i-CAN authors sought study its ability match clinical expert judgement correctly identifying level participants support needs Classification Accuracy itself tacit admission that the tool inherently evaluative purpose case every similar tool

Transfer of clinical judgment Taken together provisions create conditions substantive involved in person impairment determines supports require assessing whether those are reasonable necessity until now responsibility regulated health practitioners transferred from to undisclosed software-based decision-making algorithm determining things using about rather than clinical determination practitioner As sets out human element will carried by workforce does mandatory clinical qualification practice primarily collect quantitative data entered into automated process designed absorb decisions currently rely on clinical reasoning health practitioners By then Bill empowers within a program perform what is handled by current judgments federally registered regarding impairments functions,

supports basis non-clinical inputs from workforce required background facilitate critical engagement assessment instruments subject regulatory framework like AHPRA There no way know whether budget allocation software capable making sound judgements peer-reviewed literature individual cases but part appears be designed it

National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

Submission 1567

Submission to the Senate Community Affairs Legislation Committee

Inquiry into the National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

The TGA definition. Software that performs the functions just described in virtually any other setting would feasibly meet the Australian regulatory definition of a software-based medical device under the Therapeutic Goods Act (Parliament of Australia, 1989). The Act and its associated regulations s 41BD appear to be inclusive of ‘any instrument, apparatus, software, or other article intended for “diagnosis, monitoring, treatment, alleviation of or compensation for an injury or disability”.’ The wording shift in paragraph [32L(6)(a)], discussed above, is what I believe will place the software within this definition when all relevant factors are considered.

• Under the existing Act, a human clinician identifies needs; downstream software handles calculation.
• Under the Bill, assessors include information, new section 32K, directs: apply info derive funding amount. Intended purpose budget allocation method determine supports reasonable necessary person basis impairment allocate funds used compensate treat impairment about which it concerns.` The Software as Medical Device framework classifications scale risk significance clinical decisions driven by such systems. An automated system whose outputs decide level support available based on traditionally private health via privileged judgements empowered without independent evidence meaningful review appears fall squarely into these definitions. On standard application classification criteria given population stakes absence routine override view plausibly qualify higher-risk classes IIb III

Why Existing Automation Safeguards Are Not Equivalent Schedule Part contains three safeguards:

s59B7 lets CEO substitute action where not correct preferable; s59C3 requires Minister satisfied appropriate specify provisions involving evaluative determinations, s59D requires Standard Operating Procedure instrument specifying sufficiently objective circumstances those determinations As safeguards substantially weaker than both regulatory frameworks Parliament established for Australia AHPRA regulation of practitioners TGA’s regulation medical devices Against protections Bill’s lacks Section 7 describes individual registration statutory standards conduct defined scopes practice mandatory continuing professional development notification-and-disciplinary pathway participants Substitute-action power under s59B(7) sits with CEO participant depends CEO having domain-relevant knowledge pertinent case somehow exercise ability recognize when output black-box algorithm incorrect competence to exercise the clinical judgments substitution would requireAgainst requirements, lack mechanisms TGA regulates software drives clinical decision-making pre-market validation post-market surveillance adverse-event reporting ministerial threshold section s59C, internal executive TGA SaMD technical determination body function safety clinical decision-support Sufficiently requirement under [section] (d) reduces but does eliminate evaluative judgement since explicitly authorises automation discretion exercised evaluaive judgment made state mind formed much duties registered health practitioners. 21

National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

Submission 1567

Submission to the Senate Community Affairs Legislation Committee

Inquiry into the National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

The regulatory gap. No regulatory pathway of this kind has been described for the i-CAN customisation, or for the budget method downstream from it; or integrated system empowered under Section [insert relevant section number]. The committee may wish consider commencing an automated system executing duties previously reserved as a health practitioner’s responsibility at scale in such vacuum, is consistent safety architecture Parliament established every other clinical software application Australia.

Fix: If NDIA sought TGA classification assessment-to-funding pathway demonstrated compliance Software Medical Device framework regulation anomaly identified would substantially resolved involving oversight quality assurance allocation mechanism software-based medical device public confidence scheme alleviating concern outlined earlier and providing independent mechanism ensure needs assessments meet reasonable scientific standards.

National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

Submission 1567

Submission to the Senate Community Affairs Legislation Committee

Inquiry into the National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

Article 19 of the CRPD requires States Parties to take ‘effective and appropriate measures’ to facilitate full inclusion in the community.

• The committee may wish to consider what it means under Article 19 regarding reducing community-participation supports when no version of the assessment instrument has ever been validated;
  • This statement doesn’t address distributional impact or whether alternatives less restrictive than those rights were considered.

National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

Submission 1567

Submission to the Senate Community Affairs Legislation Committee

Inquiry into the National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

The four points set out in Section(s) [9] appear observations where findings diverge from features examined methodologically. Committee may seek advice regarding compatibility once scientific evidence has been taken account of Sections(3-8).

A separate constitutional consideration

Provisions raise a question outside scope but warrants committee’s attention Civil conscription limitation restricts Commonwealth power legislate ways compel medical or allied health practitioners deliver services terms set by Commonwealth. New section authorises funding amounts budget itself below actual cost reasonable and necessary support default rule new framework New section parallel reduction ministerial determination old plans transition period Read together with rules prohibit providers charging gap fees bridge difference combination engage civil conscription limit Because operates rather than discretionary engagement under it arguably more direct than under section . The committee should consider whether sections read relevant gap fee rules consistent with section , commencement those provisions conditional on that advice.

The ‘gold standard’ characterisation

NDIA describes i-CAN as gold standard available validated needs assessment tools published evidence does not support this characterization any prior version; deployed no validation at all notion approximates gold standard is supported. In stating, submission opposes principle nor object actuarial methods resource allocation clinical decision-making substantial literature demonstrates how can be done well when executed to assist structure guide reasoning replace it There also much admire CDS 20-year commitment developing instrument geographical context where measurement innovations have largely stagnated As stated earlier I consider the i-CAN framework genuine conceptual strengths This concerns promising still very development linked automated algorithm undisclosed properties performs traditional duties deployed workforce whose domain-relevant expertise professional obligations are assured populations use-cases for which has been validated legislative framework requires disclosure translation logic On grounds scientific method ethical obligations premature deployment of this promising instrument questionable purpose subject scrutiny not authors per se Submission also fiscal regulatory reforms contest possibility fiscal pressure justifies commencing specific architecture Bill set empower based public domain view questions separable reform NDIS needed establishes particular proposal these foundations one commence.

National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

Submission 1567

Submission to the Senate Community Affairs Legislation Committee

Inquiry into the National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

The position this submission takes is also contingent on the current state of evidence. If conditions identified in recommendations were met—(independent validation, methological rigour,a registered workforce,AHPRA-registered AHPRA- tasked tasking),the system empowered would become both more defensible defensiveand useful public.However, current evidence not consistent with these being satisfactorily satisfied.

Recommendations Based analysis provide following recommendation:

• Recommendation Defer commencement Schedule until independent scientific validation deployed instrument across full population demographics published peer-reviewed literature assessed expert panel no ties instrument developers NDIA.Critical*
• Recommendation Amend Schedule require publication budget method’s key details including metrics assessment-to-funding translation algorithm form open reasonable scrutiny and/or regulatory authority TGA before participant under it.Critical
• Recommendation Require all support needs assessors be health practitioners Australian Health Practitioner Regulation Agency relevant scope practice accordance existing legislation codes conduct.***
• Recommendation Restore obligation consider independent medical allied health evidence planning decisions clarify new hierarchy operates guide rather than precondition.Schedule Part .***
• Recommendation Amend Schedule Item so that any rule prescribing information assessor must or have regard consulted National Boards Australian Health Practitioner RegulationAgency principle direct assessors disregard information reasonably believes to the assessment.Item 13).In addition amend NDIS rules made section prescribe persons additional staff engaged consultants undertake assessment category D reclassified from category A.parliamentary disallowance.Amend Schedule part make ministerial determinations subject sunsetting
• Recommendation Insert provision requiring biennial public review of framework measurement performance decision-change rates, outcomes conducted body without commercial contractual ties developer political interests perceived success.
• Recommendation Require oversight Therapeutic Goods Act automated tool used determine supports required compensation impairment feasibly software takes private health related person’s impairment produces outputs determining necessary for impaired individual.NDIA.Deliberative processes Disclosure would reveal opinion advice recommendation in deliberative process contrary interest.Critical Regulatory Oversight Under Therapeutic Goods Act Automated Algorithmic Tool Determine Support Needs Assessments Funding Allocations under NDINotice: This document contains redacted text due to FOI exemptions and is not a full representation of all content. The original submission includes detailed recommendations aimed at improving transparency validation methods workforce independence regulatory scrutiny among other critical aspects ensuring robustness system empowered by bill.Critical

Submission: Inquiry Into the National Disability Insurance Scheme Amendment

The submission discusses how certain elements meet the definition of ‘software-based medical devices’ according to an act.

It recommends:

• That joint guidelines from relevant departments should address applying regulations related to these frameworks within specific schedules, such as Schedules 4 & 3 part II.

Submission

References

• ABC News. April 23rd, 2021 - Former NDIS chairman slams changes to support scheme. href=“https://www.abc.net.au/news/2021-04-23/former-ndis-chairman-slams-changes-to-support-scheme/100091348” .
• Administrative Review Tribunal October 1st – June 30th , ART caseload report: For period from Oct 15 th till Jun 3 rd , Australian Government. href=“https://www.art.gov.au/help-and-resources/statistics” .
• American Educational Research Association & National Council on Measurement in Education (American Psychological Association) Standards for educational and psychological testing.
• Arnold S.R.C., Riches V.C.& Stancliffe R.J.(April–June). I-CAN : The classification of needs and prediction of support needs Journal Applied Res Intel Disabil 27(2), pp.: pgs. 97-111. href=“#doiorg101111jar12055” .
• Arnold S.R.C.,Riches,V.C,&Stancliffe,R.J.(May-June). Does a Measure Support Needs Predict Funding Need Better Than Adaptive Maladaptive Behavior? Am J Intell DevaDisab, Vol. 12 No. pp. 375-394. href=“#doiorg101352194475581205375” .
• Cicchetti D.V.(September). Guidelines Criteria Rules Thumb Evaluating Normed Standardized Assessment Instruments Psychology Psychol Assess, vol.6 no.ppgs.284 - [290]. href=“#doiorg1010371040359064284” .
• Cohen J.April–June). Statistical Power Analysis for Behavioral Sciences (Second Edition) Lawrence Erlbaum Associates. General Practitioners Society v Commonwealth [HCA 30; CLR 532]. Harrell F.E Jr(October – December). Regression Modeling Strategies: With Applications to Linear Models Logistic and Ordinal Regressions Survival Analyses Second edition Springer href=“https://www.art.gov.au/help-and-resources/statistics” . ITC guidelines on test use Version 1.2. href=“/files/guideline_test_use.pdf” . Lyons K.December 3rd, NDIS plans will be computer-generated with human involvement dramatically cut under sweeping overhaul The Guardian. href=“/australia-news/december-ndis-plans-computer-generated” . McHugh M.L.June). Interrater Reliability :The kappa statistic Biochemia Medica Vol 2 No pp.276-282. href=“#doiorg1011613BM2012031” . National Disability Insurance Agency Developing a new support needs assessment News item no.[1949] HCA 44 ;CLR 201. Australian Government. href=“/news/developing-new-support-needs-assessment” . The Therapeutic Goods Act of Australia (Therapeutic Goods Act) [HCA 532]. Parliament of Australia National Disability Ins Scheme Act 2013(Cth). nationaldisabilityinsuranceagency.com Explanatory Memorandum to the National Disabil Scheme Amendment Securing Future Generations Bill 2026 Cth.
• Psychology Board of Australia December 1st, Code Conduct Psychologists Australian Health Practitioner Regulation Agency. href=“https://www.psychologyboard.gov.au/standards-and-guidelines/code-of-conduct.aspx” . Ray and NDISA AATA 3452. Riches V.C., Parmenter T.R.Llewellyn G.Hindmarsh G.Chan J.(April–June). I-CAN : New Instrument Classify Support Needs People with Disabilities Part Journal Appl Res Intell Disab Vol 2 No pp.326-339(). href=“#doiorg101111j1468-3148200800466x” .

Submission to the Senate Community Affairs Legislation Committee

Inquiry into the National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

Riches, V.C.; Parmenter, T.R.; Llewellyn, G.; Hindmarsh, G.& Chan J.(2009b).

National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

Submission 1567

Submission to the Senate Community Affairs Legislation Committee

Inquiry into the National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill

Appendix A - Simulation-Based Precision and Power Analysis: i-CAN Criterion Validity, Test-Retest Reliability, and Predictive Validity

Overview

This appendix contains the output of the precision and power analyses from section 3 of the submission. The analyses are based on Monte Carlo simulations using 10,000 replications per condition with fixed seeds.Monte Carlo simulation is a statistical approach that involves repeatedly simulating a study to assess the robustness of its results.The simulations treat the published i-CAN sample statistics as though they are the true values for the whole population.The purpose behind this treatment was analyzing whether the sample sizes used within these validation studies were sufficient enough to estimate coefficients accurately at high-stakes assessment levels.As described below along with other sections in our document,the analysis concludes otherwise; substantial work remains necessary to achieve such accuracy.If actual values turn out lower than those reported by current researches then their estimation’s reliability decreases rather improves.Three sets have been analyzed here.\n

How To Read This Appendix For whatever we’re studying,a snapshot provided through samples gives us varying resolution depending upon it.Respective agreement stats & correlations produced by each study serve merely estimates which would vary if same study repeated under similar conditions.A 95% confidence interval (CI) describes how much attempts made during said study differ regarding observed outcomes:if truth lies where CI falls between two ends(assuming publication value correct),then fresh trials conducted similarly will land somewhere inside this range.Wider intervals imply less certainty about underlying relationships’ strength.

Width (CIhigh − CI low) measures precision throughout this appendix.Narrower means more confident statements can be drawn from data whereas wider implies multiple conclusions possible without supporting any over others statistically significant effects detected may appear larger than reality due insufficient power leading false positives increasing as statistical significance threshold is crossed.Lower powers reduce chances of true discoveries while also enhancing likelihoods for them.Finding solutions requires independent replication raised earlier within submission section 3.5 crucially important issue known broadly as ‘replication crisis’. How precise an estimate needs depends on decisions informed thereby recent work sample size planning treats first design question what width should CI possess to yield useful results?Two targets used in this analysis calibrated based submissions body aimed at NDIS Support Needs Assessment driving principles remain the same-interval must narrow enough distinguishing practically meaningful categories metric wise.

• For κ(the criterion validity statistic):width ≤ .05.The standard kappa benchmarks define agreement bands roughly .10 wide(e.g.,McHugh’s,2012,‘strong agreement band’,.80–.89).A widthof.0.05is half that,keepingtheCIconsidea singleband:atκ=.85,the CIrunsfromroughly.82to.88,clearinside' strongagreement.’At widths published studies actually produce(`.15t.o..37),singlepointestimates span two or threecategories,andthedatanotshowwhichonethetruevaluebelongsin.“\n”

Submission to the Senate Community Affairs Legislation Committee

Inquiry into the National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill -2026

• For ρ (the predictive-validity correlation): width ≤ .10. Cohen’s (1988) standard effect-size categories for correlations (small ≈ .10, medium ≈ .30, large ≈ .50) sit roughly apart at about half their distance; thus with CI inside one category when p = approximately from around and up through , which represents ‘a modest but real relationship’. At widths as produced (.30), it runs between and straddling both ‘effectively no relationship’ and a meaningful relationship’, making single studies insufficiently reliable.

Benchmarks Invoked In This Appendix

The following thresholds cited throughout this document:

• κ ≥ : acceptable threshold according to Suen & Ary,
• κ ≥ : excellent agreement threshold per Cicchetti’s criteria in 1994;
• κ ≥ : strong agreement criterion proposed here specifically regarding high-stakes individual allocation of resources under the NDIS program;
• Power: conventional minimum requirement set by Cohen et al.,
• Power > or equal-to: recommended as minimal power level required before applying results within an applied context setting higher stakes than exploratory research would demand;
• CI Width < or equal-to: suggested limit on confidence interval size based upon statistical considerations outlined above herein;
• N / k >= ; Harrell recommends this ratio be maintained if prediction models are expected not just for stability over time but also across different contexts where such predictions might apply, e.g. demographic shifts affecting population composition etc.;

Abbreviations

A&P - Activities And Participation (composite measure comprising seven sub-domains); AKGT = Applying Knowledge ♻ General Tasks; CSCS = Community Social Civic Life (original paper spelling preserved); H&WB= Health and Well Being; ICAP = Inventory For Client Agency Planning (Bruininks et all , ); IIR Interpersonal Interaction Relationships MEH Mental Emotional Health SCDL Self Care Domestic Life.

A Note On Interval Construction

The intervals reported below were derived from running each study 10 times in simulation with published values held fixed to represent true value throughout entire populations studied. The middle percent of these estimates define the % interval which is a conservative estimate given that it assumes publication bias has been accounted fully by holding constant only one variable at any point during analysis. Thus actual effect sizes may well differ significantly depending how many other factors influence outcome measures beyond those considered here within scope of current research design.

Submission 1567

         Submission to the Senate Community Affairs Legislation Committee

Inquiry into the National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

Part I. Criterion Validity (Cohen’s κ)

Part I addresses criterion validity (Cohen’s κ) from Arnold, Riches, and Stancliffe (2014). The headline value of κ = .94 is derived from 49 reports (26 per cent of the development team’s 186-report sample) that the developers themselves rated as ‘good’ or ‘very good’ quality. Including reports rated ‘average’ reduces κ to .75 (n = 114); on all 186 reports, κ = .65; the independent validation sample (n = 41) produced κ = .66. Part I quantifies what each of these values can and cannot establish at the published sample sizes. Simulations based on published values from Arnold, Riches, and Stancliffe (2014).

Table A1 Precision of Published κ Values for the i-CAN Criterion Validity Study

Subsample κ CI low CI high Width

All data (N = 186) .65 .562 .735 .173

Validation (n = 41) .66 .462 .829 .367

Including average (n = 114) .75 .646 .842 .196

Good / very good (n = 49) .94 .850 1.000 .150

Note. Intervals derived from Monte Carlo simulation (10,000 replications) using irr::kappa2() on multinomial draws with marginals from Arnold et al. (2014, Table 10; five rating categories). CI = 95% confidence interval; Width = CI high − CI low. Benchmarks: κ ≥ .60 acceptable for observational research (Suen & Ary, 1989); κ ≥ .75 excellent (Cicchetti, 1994); κ ≥ .80 ‘strong agreement’ (McHugh, 2012). No subsample achieves precision below the .05 width target proposed in this submission for high-stakes use.

Table A2 Confidence Interval Width by Sample Size at κ = .94

n CI low CI high Width Width ≤ .05

30 .812 1.000 .188 No

49 .846 1.000 .154 No

100 .876 .986 .110 No

150 .890 .981 .091 No

200 .897 .973 .076 No

300 .907 .968 .062 No

500 .914 .964 .050 Yes

750 .918 .959 .041 Yes

1,000 .922 .957 .035 Yes

Note. Intervals derived from Monte Carlo simulation (10,000 replications) using irr::kappa2() on multinomial draws with marginals from Arnold et al. (2014, Table 10; five rating categories). The .05 width target, set out in ‘How to Read This Appendix’ above, is the precision needed to distinguish κ = .80 from κ = .85. Reaching this target requires n ≥ 500.

                                    31

Submission 1567

Submission to the Senate Community Affairs Legislation Committee Inquiry into the National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill - 2026

Table A3

Power to Demonstrate (\kappa=.94) Against the Conventional Threshold ((\kappa≤.75")) n | Power  > ≥ > ≥ > ≥ >
—–|—————–|-:———––:-:————:
n | Power |€� No No

n	Powe	+	R% Powe	% Po	e
30	7.6%	No	N	O	A
40	8.	y	S	t	r
49	9.	s	E	a	d
60	9	.	C	m	i
80	9 e	.r	c	k	w
150	0%	-	n	f	F

Note:Powe derived from Monte Carlo simulation (\(N=\.0,000 replications); one-sided \(α=.0.5) against H_{_o_ > >}: κ<.75</sub>). The .95th per centile of the simulated null distribution serves as tervalue; power is th proprtion o simulations under k = ⍤� exceeding this value.Benchmarks: % powe % conventional for exploratory research (Cohen,, .88));% proposed in thi submission for high-stakes applied use. At he published n = 49, power agansthe convenntional threshold i s ↴%, just belowthe 95% high-stake minimum.
Achieving 95% pwoer requires n ≥ 60. ### Table A4 Power to Demonstrate \(\kappa=\.94\") Against the High-Stakes Threshold (\(\kappa≤\.8.0</sub>) n | Pow  > ≥

&#x2B;
nPoweR% Powee307.<br/>NoNOA40.7yStr 49.sEad60Cm i150 FfFNote:pwer derived as in Table A3, against H_{_o_ > >}: κ_{<.8.0</sub t ncexamP% minimum.RHTALINOSM.2012)'strong agreement' category and is proposed i this submission a the minima for high-stakes individual NDIS allocation. At t published n =49, power agatsthe thresholdis77.&#x2B;, belowthe conventional 80% miinimum.
Achieving 80% wrisampwr requires n ≥60; reaching th ehigh-stake power minimun requirs n≥ 100. ## Conclusion: The tables show that achieving 95% or higher power against both thresholds (conventional at .75, high stakes at.8) generally require sample sizes of around ⍤� to over 200 participants depending on specific conditions such as effect size and desired level o significance. Smaller samples may achieve lower levels but still provide useful insights into research questions related to these schemes.}

National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

Submission 1567

Submission to the Senate Community Affairs Legislation Committee

Inquiry into the National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill

Power to Demonstrate Full-Sample (\kappa) Exceeds the Suen and Ary (1989) Threshold

| n | Power | ≥ ≥ 80%Power | ≤≤ 95%Power | |-|-|-| | 41   | $&space; ext{12}. ext{8} ext{%}$ | No | No | | 49 | $&space; ext{12}.&space; ext{4}%$ | No | No | | 100 | $ ext{ extbf{}} ext{{}&space;&space;}$ ext{ }${}{ ext{}2$.}${}{ ext{} ext{ }}$$ ext{.}{$ ext{ } ext{ .}$}$ | No | No | | 114 | ${}{ ext{3}}{{ ext{0}}}.$${{ ext{ }}}^{{{+}}}{{ ext{ +}}}$${{{ ext{- -}}}}^{${$ ext{ +}}$+$-$+ ext{ .$ ext{ ext{ .}-}$-}$|$_{-${ ext{ -$- ext{ –}<table class=

National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

Submission 1567

Submission to the Senate Community Affairs Legislation Committee

Inquiry into the National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill

Table A7

Power to Detect Departures from κ = .65 (N= 186)`

True κ	Power
.40                                          .99 .9%
.45  .                                                                                                                         . 8…

Note. Power derived as in Table A6, with H₀ centred on the published full-sample value (κ = .65, N= 186). At N=

186, &the study has .99% power to detect a true κ as low as .40,& but only <span style=

Submission 1567

Submission to the Senate Community Affairs Legislation Committee - Inquiry into the National Disability Insurance Scheme Amendment Bill

Submission to the Senate Community Affairs Legislation Committee

Inquiry into the National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill -2026

Table A12 Power to Distinguish (\rho) from Zero by Sample Size (Health and Well-Being Domains)

| Domain(interval)| (\rho)|| n=20|| n=50|| n=100|| n=150|| n=200|| n=300|| n=500| |-|-|–|–|–|–|–|–|–|–|—– |Mental Emotional Health (2 yr.)|..22|16%.0%* | 32%.8%** || 57.0* %|||| 96.4 %%| 99.9%% | |Behaviour (2 yr.). .25|19.0%* | 42.0%*** || 71.3* %|||| 99.1%%%| 100% |Physical Health (6–12 m). .21|14.5%* | 31.2%*** || 54.5* %|||| 95.7%%%%| .99.7 |m).30|26.4% | 56.0%*** || 86.3* %|||| **100.0%%%%%%%% m)

Table A13 Sensitivity of Reliability Estimates to Threshold Configuration

Domain( \rho \) Equal-width CI lowEqual-widthCI highFloor-heavyCILow Floor-heavycihigh||A&P Total (2yr)|.73.|..429… …897…… ..410….. ….915… |Mental EmotionalHealth|..22….-.620…….-.- - -.621……..– ….. –.237.. |Behavioral (2y r.)|.:.25………-.207……….+.+ +.+639……………++.+-++,.647.

Note: Intervals derived from Monte Carlo simulation at n=20, comparing the equalwidth and floorheavy threshold configurations defined in table a 9 .Sensitivityofthe precision estimate to threethreshold configuration is negligible across three domains tested(widths within ±).02), supporting useequal width thresholds as primaryconfiguration for part ii.

Part III Predictive Validity(Spearman’s ρ with Funding Allocation)

predictive validity(spearmansρ )for funding allocationas reported by ArnoldRichesand Stancliffe(2015). The published spearmancorrelations between i-can scores individualizedfunding allocations under standard scoring are \rho=.19(total raw)\rho = =.13(h ealthwellbeing) ,and

=.20(activitysandparticipation)withN=163.part iii quantifiesprecision of these estimatesandsample sizes requiredtodetect them reliably. i-can domain scoresthe aggregate sums ordinal item ratingsannualindividualisedfundingallocationAustralian dollars.Two scoring methods appearpublished Table. Thestandard administration protocolwhich basis NDIS deploymentpost hoc fundeditems coding which exploratorywas not part standardadministrationnot deployed.TheICAP comparator included benchmark.SamplesizeN= mixeddisabilityperson centred fundingallocations already inplace.Instrument usedin publis studyis the ican brief research version fullinstrument and v 6 .Part approximates underlying distribution bivariate normal calibrated so simulatedspearmannmatches publication value.heavier tailed or more skewed distributionslikely dollar valued sampling varianceof would be no smaller probably slightly larger precisionestimatesreported here conservativeactual publish studyci widthsno narrower than shown

National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026 Submission to the Senate Community Affairs Legislation Committee

Inquiry into the National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill [2026]

Precision of Predictive Correlations Under Standard (i{CAN}) Scoring ((N=163"))

| Score | ρ | CI low | CI high | Width | Power | Spans 0 | |-:|–:-|–:|–:|–:|–:|–: |Total Raw | .19 | .035 | .``.333`` | .298| ``%.68.6``` | no| |Health and Well-Being | .13 |-..026|.281|.307|%38.1%| YES| Activities and Participation|.20 |.049|.340|.292 | %72.3%% | no||Note.|Intervals derived from Monte Carlo simulation (10,000 replications`) using bivariate normal draws with population ρfrom Arnold et al.(2015 Table). standard scoring is published protocol basis for NDIs deploymentWidth = C Ihigh - Cllow;Power testsρ = at two-sided α=.05Spans indicates whether the CI includes zeroThe HealthandWellbeingCI spanszeroatthepublished N= a true of cannot be ruled out this domain.

PrecisionofPredictiveCorrelationsforthelCAPComparator((N=)163)

Score| ρ | CI low | CI high | Width | Power | Spans 0 | |-:|–:-|–:|–:|–:|–:|–: Service Score |-.-.19 | .-.337| -.038 |.299| %.``%.68``` | no Broad Independence-..- .25|-….390|-….,100 | .290 | %89.%``% | no| Note.IntervalsderivedasinTableA4.TheICAP(Bruininksetal.,.is included as comparator from sameArnoldet al(2015 study Correlation are negative because higher ICAP scores indicate greater independence, hence lower support need.| ## PrecisionOf Predictive Correlations Under Post-Hoc Funded Items Scoring (( )) Total Raw|.63 | .519 | .720 | ``%,201 | %100%% | no||Health and Well-being|.41 | `.272` | `.535` | `.263` | %%100% | noActivitiesandParticipation|.65 | .544 | .736 | .192 | 100.0%|no|| Note Intervals derived in Table A4.FundedItems scoring is post-hoc exploratory recoding was not part of standard administration basis for deployed NDIS protocol substantiallyhigher correlations under thisscoring (contrast to thestandard scorting) reflectthe recoding rather than improved measurement underlying supportneed.

PowerTo Detect a Non-ZeroCorrelationUnderStandard((i{CAN})

Score| ρ | n = 50 | n=%.100``` | n=.163 | n=.`,200 | n= .300``` | n=. ,500 | n=``.`800 | n=.%1,000 |-:|–:-|–:-:.%-:—::––::-:—–: Total Raw | %19 ``` | ``%,23.8%%``` | %49.2%%% | %.69.3%%%% | %= ,78.0%%``` | %+91.0%%``` | +%99.0%%``` | +%.99.9%%``` | %%100.0% Health and Well-being|.13 | .13.2%| 25.9% | 38.1% | 45.9% | 59.8% | 84.2% | 95.6% | 98.7% ActivitiesandParticipation|.20 |.26.8%| 52.2% | 72.7% | 81.9% |93.0``%` | 99.7% | 100.| 100. Note Power derived from Monte Carlo simulation (replications) bivariate normal draws two-sided α=. ρvaluesfromArnoldetal(2015 Table). For i-CAN TotalRaw (.19), the sample of N= used inthe published study achieves . power against a twosided null Reachingpower conventional minimum Cohen, requires approximately N = forTotal Raw Activities participation reaching highstake proposed this submission requiresapproximatelyNfor Health wellbeing

National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026

Submission 1567

Submission to the Senate Community Affairs Legislation Committee

Inquiry into the National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill

2026

Table A18

Confidence Interval Width by Sample Size at ρ = .19

| n | CI low | CI high | Width | Width ≤ .10 | |-|-|-|-| | 50 |-0.095|.451|.546|No| | 100 |-0.007|.374|.381|No| | 163 |0.039|.338|.299|No| | 200 |0.052|.323|.270|No| | 300 |0.080|.296|.217|No| | 500 |0.103|.274|.171|No| | 1,000 |0.129|.248|.119|No| | 1,500 |0.141|.239|.098|Yes| | 2,000 |0.147|.233|.086|Yes|

Note.

Intervals derived from Monte Carlo simulation (10,000 replications); bivariate normal draws; ρ = .19 from Arnold et al. (2015, i-CAN Total Raw, standard scoring). The .10 width target, set out in ‘How to Read This Appendix’ above, distinguishes a near-zero from a meaningful relationship at ρ = .19.The target is met at n ≥ 1 500; the published N=163 produces a width of .299 meaning the trueρcould plausibly lie anywherefrom .04to .34.

Table A19

Analytical R²

g#### Confidence Intervals for Predictive Models | Model | R² | k | N | CI low | CI high | Width | N : k | |-|-|-|-|-|-|-| |iCAP + Demog | .22 | 12 | 163 | -.059 | .273 | .214 | 13.6 iCAN+Demog | .35 | 19 | 163 | .142 | .382 | .240 | 8.6 Funded Items + Demog | .60 | 19 | 163 | .440 | .630 | .190 | 8.6 Stepwise (Funded Items)a | .87 | 4 | 162 | n/a | n/a | n/a | 40.5

Note.

The intervals derived analytically via the non-central F distribution(Smithson, 2003; Steiger, the upper bound usually sits closer to than lower bound.Two factors drive this.SampleR² tends to be biased upward so population value is on average smaller observed value.R² also cannot exceed which compresses end of R².distribution.Demog = demographic covariates.N : k ratios below do not meet Harrell(2015) recommendation for stable prediction only stepwise model row clears because selection reduced predictor count from within same N.a Stepwise precludes standard confidence interval: procedure inflates by capitalising chance and regression through origin further inflates because total sum squares computed against intercept(Arnold et al., Table one outlier removed).

Submission to the Senate Community Affairs Legislation Committee

Inquiry into the National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill

References Cited in This Appendix

Arnold, S.R.C.; Riches V.C; Stancliffe R.J.(2014); I-CAN The classification prediction support needs Journal Applied Research Intellectual Disabilities vol_27 no_2 pp.p97-p111. href=https://doi.org/10.1111/jar.12055. Arnold SR.C;Riches VC.Stancliffe RJ,(2015); Does measure of support predict funding better than adaptive maladaptive behavior American J Intel Dev Disabil vol no ppp_375–394_. href=https://doi.org/10.1352/1944-7558- Button KS.Ioannidis JP.A.Mokrysz C.Nosek BA.Flint J Robinson ESJ Munafò MR(2013). Power failure: Why small sample size undermines reliability neuroscience Nature Reviews Neuroscience vol 14 issue number 6 pages_p365 – -376 .href=https://doi.org/10.1038/nrn3475 Bruininks RH.Hill BK.Weatherman RF.Woodcock RW (1986) Inventory for client and agency planning ICAP DLM Teaching Resources Cicchetti DV.(1994). Guidelines criteria rules thumb evaluating normed standardized assessment instruments psychology Psychological Assessment vol no pp.p284-p290. href=https://doi.org/10.1037/1040-3590. Cohen, J., Statistical power analysis behavioral sciences ed Lawrence Erlbaum Associates Gelman A Carlin J Beyond power calculations Assessing Type S sign magnitude errors Perspectives on Psych Sci vol no_pp_641–651.href=https://doi.org/10.1177/ Harrell FE Jr Regression modeling strategies With applications linear logistic ordinal regression survival analisys ed Springer https://doi.org/10.1007/978- ioannidis JP.A(2005); Most published research findings false PLoS Medicine vol no epp_e124 . href=https://doi.org/10.1371/journal.pmed.0020124 Maxwell SE Kelley K Rausch JR Sample size planning statistical power accuracy parameter estimation Annual Rev Psychology vol 59 issue number pages_p537 – -563 .href=https://doi.org/10.1146/annurev.psych McHugh ML Interrater reliability kappa statistic Biochemia Medica vol no pp.p276-p282. href=https://doi.org/10.11613/BM Riches VC Parmenter TR Llewellyn G Hindmarsh G Chan J (2009b) Reliability validity practical utility measuring supports using I-CAN instrument Part II Journal Appl Res Intel Disabil vol no_pp p340–-353.https://doi.org/ Smithson M Confidence intervals Sage Publications Steiger, J.H., Beyond the F test Effect confidence interval tests close fit analysis of variance contrast Psychological Methods vol _no_pp_164- suen HK Ary D Analyzing quantitative behavioral observation data Lawrence Erlbaum Associates. pages=39