Table of Contents
The Integral Route for School Inclusion (RIIE): study protocol for a stepped-wedge cluster randomised evaluation of a seven-phase institutional procedure in Colombian schools
Andrés Alexander Paternostro Correa Paternostro Academy, Barranquilla, Colombia · Barranquilla, Colombia · Paternostro Academy, Barranquilla, Colombia
Document type: study protocol, submitted for peer review prior to data collection. Status:
Abstract
widely described as ceremonial: individualised plans are produced and filed, external clinical reports are archived unread, and support, once assigned, is rarely withdrawn. Existing instruments address either whole-school self-evaluation (the Index for Inclusion ) or tiered academic support (multi-tiered systems of support), but no widely used framework in the region specifies the decision sequence for an individual student from detection to the dignified closure of support, and none regulates withdrawal at all.
Inclusion (RIIE) — increases students’ sense of school belonging and improves the quality of individualised planning, relative to comparable institutions not applying it.
cross over to the intervention at randomised time points, with five measurement waves and separate measurement of implementation outcomes.
sequences of six, spanning public and private sectors and contrasting socioeconomic contexts. The analysis cohort is defined at baseline by a universal screen applied identically in all institutions, independently of the framework, and then frozen.
measured with a validated instrument.
rather than filing; referrals to special education and their socioeconomic distribution; school persistence; teacher burnout; and proportion of supports formally closed rather than indefinitely prolonged.
following Proctor’s taxonomy, so that a null result can distinguish an ineffective route from a route never actually used.
implementation fidelity, the upper bound of the 95% confidence interval for the between-condition difference in belonging falls below the pre-specified minimum difference of interest of 0.20 SD, the framework’s central hypothesis will be considered false as formulated and reported as such. Failure to detect an effect is not equivalent to excluding one; only the latter refutes the framework. An inconclusive interval will not be reported or cited by the author as supportive evidence.
Keywords: inclusive education; sense of belonging; implementation science; institutional decoupling; reasonable accommodations; Latin America; de-implementation.
Background
1.1 The implementation gap Article 24 of the Convention on the Rights of Persons with Disabilities (2006) and General Comment No. 4 (2016) establish inclusive education as a right requiring transformation of culture, policy and practice. Latin American states have legislated accordingly: Colombia’s Decree 1421 of 2017, Argentina’s CFE Resolution 311/16, Chile’s Decree 83/2015, Mexico’s 2019 General Education Law and Peru’s regulatory reform of 2021 all mandate barrier removal and reasonable accommodation.
The gap between mandate and practice is well documented but under-theorised in this field. This protocol adopts the neo-institutional account of Meyer and Rowan: organisations adopt formal structures ceremonially to secure legitimacy and then decouple them from actual work, sustaining the gap through a logic of confidence that discourages close inspection. On this reading, the filed individualised plan is not a failure of compliance but a predictable organisational product.
1.2 What existing instruments do and do not address
field, adapted in Colombia by the Ministry of Education in 2008 — evaluates the institution across cultures, policies and practices. It does not specify what a school should do on Monday with a particular student. Multi-tiered systems of support specify escalation but are largely oriented to academic response and have limited penetration in Spanish-speaking school systems. A scoping review of six regional frameworks (Colombia, Argentina, Chile, Mexico, Peru, Spain), conducted under PRISMA-ScR with a protocol deposited in an open repository and cited here by that deposit, found that none regulates the withdrawal of support . Until the review itself is published, this claim is to be read as a documentary review by the author with a registered protocol, not as a peer- reviewed finding. Entry into support is procedurally detailed; exit is unaddressed in all six. The most recent national instrument available at the time of writing reproduces the same asymmetry: Colombia’s Ministry of Education Directive 005 of 24 July 2026, on curricular flexibilisation, specifies preventive work, pedagogical assessment, characterisation, the mandatory content of the individualised plan, implementation and monitoring, and contains no provision on the withdrawal of an individual support. It also extends flexibilisation beyond disability to any student whose circumstances place their right to education at risk, which is the same activation logic the framework proposes and for which no procedure is specified.
1.3 The framework under evaluation The RIIE comprises a permanent Phase 0 (preventive culture) and six active phases: early activation from documented classroom friction rather than clinical diagnosis; integral characterisation; a tiered technical decision; an individualised plan; implementation with short-cycle monitoring; and dignified closure , in which every support is assigned a withdrawal horizon at the moment it is granted.
Its four distinctive elements, relative to the instruments above, are: (i) activation without a diagnostic gatekeeper; (ii) formalised de-implementation; (iii) explicit ownership of the interface between school, family and external clinicians — in ecological terms, the mesosystem; and (iv) a support role defined by its own planned obsolescence.
1.4 Theoretical position, stated explicitly
The framework is stratified rather than unified. It draws on general systems theory for its unit of analysis (the interface), on cybernetics for its control logic (requisite variety, negative feedback, a stopping criterion), on neo-institutional theory for its account of failure (decoupling and isomorphism), on Bronfenbrenner’s bioecological model for the mesosystem and proximal processes, and on the capability approach for its normative grounding (conversion factors). It does not claim that any of these levels reduces to another.
Objectives and hypotheses
reported sense of school belonging among students facing barriers than the same institutions during their control periods, with a pre-specified minimum difference of interest of 0.20 SD at 24 months, computed on the pooled baseline standard deviation of the full student sample. Three outcomes are possible and only three. Confirmation: the lower bound of the 95% confidence interval exceeds zero and the point estimate reaches 0.20 SD. Refutation: the upper bound falls below 0.20 SD. This is an equivalence-style criterion and therefore demands more power than a superiority test, a consequence stated here rather than discovered later: with twenty-four clusters, a truly null effect on the institution- level outcome would most plausibly yield an interval too wide to exclude 0.20 SD, and would be classified inconclusive rather than refuting. The falsification test is consequently pre-specified on the co-primary plan-quality outcome, whose unit of analysis is the individualised plan rather than the institution and which therefore has several hundred units. A design analysis reporting the probability of reaching each of the three verdicts under true effects of 0, 0.10, 0.20 and 0.30 SD will be computed and published with this protocol: a study that promises refutation without reporting its capacity to refute is promising something it has not verified it can deliver. Inconclusive: the interval spans 0.20 SD
and it is pre-registered here that an inconclusive result will not be presented, cited or summarised by
the author as supportive evidence in any venue.
a higher proportion of individualised plans associated with observable classroom change; (b) no increase in special-education referrals, and reduced socioeconomic disproportionality among them; (c) higher school persistence among students facing barriers; (d) no increase in teacher burnout; and (e) a measurable proportion of supports formally closed.
the target construct; achievement gains, if any, would be secondary and are not the basis on which the framework should be judged.
Methods
3.1 Design
Two-year stepped-wedge cluster randomised design: twenty-four institutions in four sequences of six, five measurement waves, with the timing of crossover randomised. Every institution contributes both control and intervention person-periods, which increases efficiency at a fixed number of clusters; randomising when rather than whether an institution adopts removes the self-selection that was the principal limitation of the earlier design; and no institution is asked to forgo the intervention. Calendar time is modelled as a fixed effect, since secular trends are confounded with time in this design. Reporting follows the CONSORT extension for stepped-wedge trials. Simple parallel randomisation of adoption is not feasible; institutions self-select into implementation, and comparison institutions are matched post hoc on size, sector, socioeconomic index and baseline outcome values.
This is the principal design limitation and is acknowledged rather than mitigated away: self- selection means implementing institutions differ systematically from comparison institutions in ways matching cannot fully capture, most obviously in leadership commitment. 3.2 Participants
Twenty-four institutions in four sequences of six. The analysis cohort is defined at baseline by a universal screen applied identically in all institutions and independently of the framework, and is then frozen. This is a substantive safeguard rather than a technicality: Phase 1 activates support from documented classroom friction rather than diagnosis, and therefore changes who is identified as facing barriers. Without a frozen cohort, institutions under the framework would identify more students with milder presentations while control periods identified fewer and more severe, and the resulting composition difference would favour the framework. Students activated by the route after baseline are analysed separately as an implementation outcome, never within the primary effect estimate. A random sample of the general student body is measured for comparison of belonging scores.
Composition requirements: at least half of implementing institutions must be public; at least a third must serve populations in the lowest two socioeconomic quintiles. A study conducted only in well- resourced private institutions would test the framework under conditions in which almost anything works. Stopping rule: if fewer than twenty institutions are recruited, the study is reclassified as a feasibility and implementation study, the falsification test is not conducted, and this is reported as the study’s status. An underpowered trial does not produce weak evidence; it produces uninterpretable evidence that will be used in whichever direction the reader prefers.
3.3 Outcomes and instruments
(the Psychological Sense of School Membership scale, Goodenow 1993, or the OECD PISA belonging index, allowing national and international comparison).
scored against a rubric published before the first wave by assessors blind to institution, condition and wave, with inter-rater reliability reported on a minimum 20% double-scored subsample against a pre- specified acceptability floor. This outcome is co-primary because its unit of analysis is the plan rather than the institution, yielding hundreds of units rather than twenty-four, and it will therefore be adequately powered irrespective of institutional recruitment. Secondary: administrative referral data with socioeconomic disaggregation; enrolment persistence; a validated teacher burnout measure; and support-closure records.
2011), assessed through structured institutional records and periodic audit. Fidelity is defined numerically and in advance: an institution is classified as having implemented the route if it documents at least 80% of the specified milestones for each phase in at least two of three audit points, verified by external audit. This threshold is fixed before baseline. A fidelity threshold set after outcomes are known is not a criterion but an alibi with the appearance of one. 3.4 Sample size
sequences derives from a stepped-wedge power calculation whose declared inputs must appear in the submitted protocol: the minimum difference of interest of 0.20 SD; the assumed intra-cluster correlation for belonging, justified from published school-level estimates with sensitivity analysis across a plausible range; the cluster autocorrelation over time; anticipated cluster-level and within-cluster attrition; and the number of students measured per cluster per wave. The binding constraint in a cluster-assigned intervention is the number of clusters, not the number of students. The calculation, its assumptions and its sensitivity analysis will be published with the protocol. A protocol published with an unjustified sample size invites the first and most damaging reviewer objection.
3.5 Analysis
Primary analysis is by intention to treat: all institutions are analysed in the condition their randomised sequence assigned them at each period, regardless of implementation fidelity, using mixed-effects models with students nested in institutions, random effects for cluster and cluster-by-period, and a fixed effect for calendar period, adjusting for baseline values and pre-specified covariates. Multiplicity across secondary outcomes is controlled by the Benjamini–Hochberg procedure at a false discovery rate of 0.05. Missing data are handled by multiple imputation under missing-at-random with a pre- specified sensitivity analysis under a plausible departure from that assumption. The analysis plan is locked and deposited in the public registry before the first wave.
Pre-specified: the falsification condition is assessed in the intention-to-treat analysis. This is a deliberate change from the earlier draft, in which falsification was assessed in a per-protocol subgroup defined by documented fidelity. Where fidelity is a post-baseline outcome correlated with leadership commitment, that specification makes the falsification condition structurally difficult to trigger, because institutions in which the route failed are disproportionately those that will document low fidelity and would therefore be excluded from the very analysis capable of refuting the framework. The legitimate question of what happened where the route was actually applied is answered instead as a secondary analysis, by complier-average causal-effect estimation using randomised sequence assignment as the instrument.
3.6 Student participation
design of measurement instruments and in the interpretation of findings, following a published co- authorship protocol. This is an obligation of the framework being tested and therefore a condition of the study, not an optional enhancement. Because it is a condition, it constrains the timetable: baseline data collection does not begin until the student co-authorship group, convened under the published protocol, has reviewed the measurement instruments. Participation in the interpretation of findings occurs later, at analysis. This sequencing is stated because a participation requirement that the calendar renders impossible is a decorative one, and the co-authorship protocol exists precisely to prevent that. 3.7 Harms, monitoring and stopping
A protocol that imagines no harm has not examined its own intervention. This one has plausible and nameable harms, and they are measured rather than assumed away.
Anticipated harms. Premature withdrawal of support, since a framework that insists on closure will produce some closures that are too early. Labelling without diagnosis, since activation from classroom observation still names a student. Teacher overload, since Phase 0 requires collective hours that in an unslack system are taken from somewhere. Resource displacement toward milder cases at the expense of students who required the individualised level. And escalation of family conflict, since minutes that record disagreement make visible what previously stayed diffuse.
Safety outcome, measured actively. The proportion of withdrawn supports that are reopened within twelve months. This is the direct indicator of the most probable harm of the framework, and it is reported irrespective of the primary result.
Monitoring. Harms are reviewed at each of the five measurement waves by the independent evaluator, not by the author. Institutions report unintended consequences through a standing open field with no character limit, and every report is coded and published in aggregate.
Stopping criteria, pre-specified. The study is suspended, and the framework revised before continuing, if the reopening rate of withdrawn supports exceeds 25 per cent at any wave, if the referral rate to special-education services rises in intervention periods relative to control periods, or if any harm is reported that the protocol did not anticipate and that the evaluator judges attributable to the framework.
Ethics and transparency
Approval will be sought from an accredited ethics committee prior to any data collection. Because the intervention is allocated at institutional level, consent operates on two distinct objects and this protocol treats them separately, following the Ottawa Statement on the ethical design and conduct of cluster randomised trials. Consent to the intervention is given by the institutional gatekeepers — the rector and the governing council — who are named in the registration and who may commit the institution because the intervention modifies institutional practice rather than individual treatment. Consent to data collection is individual: informed consent from guardians and separate assent from students; student dissent overrides guardian consent. A student who declines is withdrawn from measurement only: no student loses any support, is moved from any classroom, or is treated differently in any respect as a result of declining, and families are told so in those words.
the registered protocol will be reported as deviations, not presented as design.
associated materials without charge, and has a professional and reputational interest in a favourable result. This is the single strongest objection an informed reader will raise, and a declaration does not answer it. The following are therefore conditions of the study and not recommendations: data are held in custody by a third party; the primary analysis is executed by an independent evaluator named in the public registration before the first wave; the analysis plan is locked and deposited in the public registry before baseline; the author has no access to unblinded outcome data before the primary analysis is executed; analysis code and de-identified data are published alongside results; and the author does not participate in outcome adjudication. One safeguard is deliberately not promised: author blinding to sequence allocation is not feasible in this design, because the author supports implementation and will necessarily know which institution crosses over in which wave. Promising an unachievable safeguard would be worse than promising none; the six conditions listed above are the ones that actually hold, and none of them depends on a blinding this design cannot deliver. Where any of these proves unachievable with available resources, the deviation will be reported explicitly so that readers can discount accordingly.
Discussion
5.1 What a positive result would and would not establish A positive result would establish that a defined institutional procedure is associated with improved belonging in Colombian schools under conditions of documented implementation. It would not establish that the framework is superior to alternatives, that it transfers to other systems, or that its specific seven-phase structure is necessary rather than incidental.
5.2 What a null result would establish
If the intention-to-treat analysis yields an upper confidence bound below 0.20 SD, the framework’s central claim fails as formulated, and the author has committed publicly, in the book that proposes the framework, to reporting such a result in the same terms in which the claim was made. A second and distinct route to failure is stated here in advance: if outcomes are null and fidelity is low across institutions, the route has not been tested, but something else has been established and will be reported as follows — a route that real schools cannot implement is a failed route, however sound its internal logic. An intervention is not judged by what it would achieve if applied, but by what happens when it is handed to a teacher with twenty-nine other students. Both routes to failure are named here so that neither can be selected after the fact.
5.3 Contribution independent of results
Irrespective of outcome, the study would produce the first systematic data on formalised support withdrawal in a Latin American school system — a practice that, as documented above, no regional framework currently regulates.
References
To be completed in the target journal’s style. Core references: Booth & Ainscow (2000); Meyer & Rowan (1977); DiMaggio & Powell (1983); Bronfenbrenner & Morris (1998); Goodenow (1993); Proctor et al. (2011); Ashby (1956); Argyris & Schön (1978); Sen (1999); UN CRPD General Comment No.
Suggested submission targets
decoupling framing fits its core interests
Inclusive Education valued
Trials / BMC family Protocol-friendly, fast DOI Requires clear trial-like structure
Educación Inclusiva
Recommended sequence: submit the protocol in English first, and publish the Spanish version afterwards citing the English DOI. The reverse order costs you the international record. 6 Administrative information
The following items are reported to complete the SPIRIT 2013 checklist for trial protocols, which governs protocols as CONSORT governs reports.
Registration. The trial will be registered in a public registry before the first measurement wave; the registry name and identifier are inserted here upon registration and appear on the title page thereafter.
Sponsor and funding. Paternostro Academy sponsors the study. Funding sources are declared in full at registration and in every publication. No funder has, or may acquire, any influence over design, analysis, interpretation, or the decision to publish, and this condition is a term of acceptance of any funding.
Roles and responsibilities. The author designed the framework, supports implementation and drafts publications; he does not hold the data, does not execute the primary analysis and does not adjudicate outcomes. The named independent evaluator holds analytic responsibility. A third-party custodian holds the data. Data monitoring. Given the nature of the intervention, a full data monitoring committee is not proportionate; monitoring is performed by the independent evaluator against the pre-specified harm criteria of §3.7, with authority to recommend suspension.
Protocol amendments. Any amendment is versioned, dated, deposited in the registry with its rationale, and reported as an amendment in every subsequent publication. Amendments after the analysis plan is locked are permitted only for safety.
Data management and confidentiality. No student-level identifiers leave the institution. Institutional data are pseudonymised at source; the key is held by the custodian and not by the author. Handling complies with Colombian Law 1581 of 2012.
Access to data and dissemination. De-identified data and analysis code are published with the results, so that any third party may reanalyse and contradict the findings. Results are published whether favourable, adverse or inconclusive; failure to publish within twelve months of database lock is itself reported as a deviation.
Ancillary and post-trial provisions. All institutions retain the materials and the installed route at the end of the study. No institution loses support, access or standing as a result of participating, of reporting adverse findings, or of withdrawing.
Paternostro Academy · Universo Pertenecer · paternostroacademy.com · Version 1.0 · August 2026
Free and open use. The only condition is to cite the source: Paternostro Correa, A. A. (2026), Universo Pertenecer, Paternostro Academy, paternostroacademy.com. If you adapt it, state what you changed.