Quality You Can Measure

With autism now identified in approximately 1 in 31 children (Shaw et al., 2025), the demand for applied behavior analysis (ABA) services continues to grow faster than the workforce available to provide it (Behavior Analyst Certification Board, 2026). ABA is supported by decades of evidence documenting gains in communication, adaptive, and social outcomes for autistic children (Smith & Iadarola, 2015), and it is now a covered standard of care across many public and private payers. Naturalistic Developmental Behavioral Intervention (NDBI) is a form of ABA that integrates behavior-analytic procedures with developmental science, teaching skills within play and everyday routines rather than apart from them. As access expands, a central question becomes more important: how does an organization know, and demonstrate, that the care children receive is delivered with quality and produces the outcomes it promises?

Implementation quality is itself a clinical variable. What a child receives in each session depends on how consistently treatment procedures are implemented (DiGennaro Reed et al., 2011; Fiske, 2008), and an organization’s measurement practices determine whether it can evaluate its own effectiveness (Cooper et al., 2020). Outcome reporting requires the same discipline. Satisfaction measures are valuable, but they answer a different question than clinical progress. The percentage of clients who improved shows direction but says nothing about magnitude. Percentage change can also overstate progress on criterion-referenced instruments when baseline scores are low. Stronger reporting pairs the magnitude of change with the number of measurements behind each result and presents outcomes alongside evidence that treatment was delivered as designed. The same logic extends to cost. Programs that produce faster skill acquisition and durable reductions in interfering behavior position children for earlier titration and discharge, and the projected lifetime savings of effective early intervention are substantial (Jacobson et al., 1998; Chasson et al., 2007).

This paper has two aims: 1) to describe how ANNA structures its clinical model and assures the quality of its delivery, and 2) to report client outcomes across the organization’s complete assessment record, with every analytic convention stated in full. ANNA provides center-based services to autistic children ages 1 through 6 across six Massachusetts locations. Families who entrust us with their children’s early intervention deserve to see both the outcomes of care and the clinical system designed to produce them. The same transparency matters to the clinicians and funders who make that care possible.

ANNA’s clinical model is documented in the ANNA Clinical Practice Guidelines, which translate professional standards into everyday clinical decisions and align with the BACB Ethics Code for Behavior Analysts (Behavior Analyst Certification Board, 2020), the CASP ABA Practice Guidelines (Council of Autism Service Providers, 2024), and the standards of the Autism Commission on Quality (ACQ). The model is rooted in Naturalistic Developmental Behavioral Interventions (Schreibman et al., 2015) and blends the Early Start Denver Model (Rogers & Dawson, 2010), Project ImPACT (Ingersoll & Dvortcsak, 2019), and Reciprocal Imitation Training (Ingersoll & Schreibman, 2006) around each child’s individual assessment picture.

The guidelines span the full arc of a child’s care and include how medical necessity, service model, and treatment intensity are determined; how each child’s presentation is assessed and conceptualized; and how treatment plans, goals, and protocols are developed from that picture. They set expectations for assent and neurodiversity-affirming practice, active treatment procedures, high-risk procedures, treatment settings, and caregiver coaching, and they define how care is sustained and concluded responsibly through supervision and fidelity, caseload management, cultural and language responsiveness, documentation standards, an appeals process, coordination of care through transition and discharge, and quality assurance and compliance monitoring. The sections are written so that a clinician facing a decision, at any point from intake to discharge, finds written guidance rather than improvising alone.

Assent deserves particular mention, because assent-based practice is central to ANNA’s model. Caregiver consent authorizes treatment; client assent governs whether a specific activity proceeds in the moment, and both are required. The guidelines define assent and its withdrawal in observable terms, specify how staff respond, and require that every child’s program include a goal supporting advocacy, protest, or refusal. Staff are trained to treat the absence of affirmative engagement as a signal to adjust, so that every child has an effective and respected way to communicate no.

A written model matters only if the people delivering it are trained to the model and then verified against it. Every clinician, Behavior Technician (BT) and Board Certified Behavior Analyst (BCBA) alike, completes comprehensive onboarding training in the ANNA model before working independently with children. All training is competency-based and covers all aspects of the ANNA Clinical Practice Guidelines, including NDBI practices, assent standards, safety procedures, and documentation expectations. Each competency must be demonstrated to criterion, with instruction and practice continuing until it is. Verification then carries the same expectations into everyday operations. Behavior Technicians are regularly observed delivering the NDBI practices the model calls for. BCBAs are observed across the full span of their clinical work through a defined set of fidelity checks including a dedicated check on assessment implementation, in which the BCBA is observed conducting the assessment process itself, along with checks covering program oversight, BT supervision, caregiver coaching, and fidelity to the specific NDBI protocols ANNA has adopted. Every treatment plan is reviewed and approved by a Clinical Director before it is submitted. Every clinical session note is audited daily by an automated system against documentation and medical-necessity expectations. Clinical Directors carry a defined set of standing oversight activities, verified monthly with executive attestation.

Verification counts for little unless its findings lead to action. ANNA’s Quality Assurance and Improvement program reviews performance quarterly across five domains: treatment fidelity, case management, clinical documentation, client outcomes, and the experience and satisfaction of families and staff. Findings that fall short of expectations become named improvement initiatives with owners, defined actions, and re-measurement dates. Safety incidents are trended within the same review so that clinical and operational signals are read together, and survey results carry defined escalation pathways when they decline. When fidelity observations identify an area of practice that needs strengthening, the response is training and curriculum improvement rather than individual blame. The whole system exists so that the quality of a child’s experience at ANNA never depends on chance. Guiding that work from outside the organization’s own chain of command, ANNA’s Community Advisory Board of autistic self-advocates, family members, and allied professionals advises on how quality is defined and measured, keeping autonomy, communication, felt safety, belonging, and family partnership alongside skill acquisition in the definition of a good outcome.

The record includes every current or discharged ABA client with at least one verified standardized assessment between January 2024 and September 2026: 95 children who received center-based services across ANNA’s six Massachusetts clinic sites, all between the ages of 1 and 6 at enrollment. Comprehensive and focused treatment models are both represented, with an average of 21.2 delivered direct ABA therapy hours per week (median 20.3, range 0.7 to 37.1). Intensity is determined child by child through medical necessity rather than a standard package, spanning the comprehensive and focused ranges described in the CASP guidelines (Council of Autism Service Providers, 2024). The 48 clients contributing comparative measurements had a median observation span of 11.6 months between first and latest assessments.

The dataset comprises 799 verified assessment administrations across 95 clients and seven instrument domains, compiled from the source documents in each client’s clinical file. Of 1,212 candidate score entries, 385 were duplicates of the same administration and 28 could not be verified, and the unverifiable entries were excluded and logged, leaving the 799 analyzed here. A comparative measurement is a client’s reassessment in a domain against that client’s immediately preceding administration in the same domain. The record contains 269 comparative measurements from 48 clients. Assessments follow a target 6-month reassessment cadence. Observed intervals tracked close to that target for criterion-referenced instruments (median 5.6 months pooled across the criterion-referenced instruments) and ran near twelve months for the annually normed Vineland-3 (median 11.9 months), and rates are therefore normalized per month between administrations. Four conventions govern all reporting: raw-score change in the instrument’s native units is the primary metric, with percentage change secondary because low starting scores inflate it; medians accompany every mean; movement counts as improved only when it strictly moved in the clinically desired direction, with unchanged scores reported as their own category; and record counts appear with every figure. No client, cohort, or period was excluded; the small number of individual administrations that could not be verified against source records is described in Section 6.

Three limitations shape how the results in this report should be read. First, the design is a pre-post single-organization service record with no comparison group. Maturation therefore contributes to criterion-referenced gains in young children, and no causal attribution is claimed. The norm-referenced Vineland-3 partially addresses maturation by construction. Second, the record reflects the documents that could be verified. A small number of score totals and dates could not be confirmed against source records, and those items were excluded and logged for review rather than estimated. The Vineland-3 Maladaptive scales are not yet represented for the same reason. Third, some domains remain thin, with ESDM at n = 8 and Barriers at n = 30, and reassessment intervals differ by instrument (near six months for criterion-referenced instruments and near twelve months for the Vineland-3), so per-interval gains are not directly comparable across instruments; the per-month rates in Table 3 are provided for that reason. These constraints are labeled and addressed rather than hidden.

The next iteration of this report is already taking shape, and four commitments will strengthen it. First, successive periods will be pooled into larger cohorts reported with standardized effect sizes. Second, outcomes will be disaggregated by treatment tenure and by measures of intensity and fidelity exposure. This will test whether sustained high fidelity compounds gains. Third, lived-experience outcome measures will be co-developed with the Community Advisory Board. These measures will then be added to the standing assessment battery. Fourth, maintenance of gains will be followed through titration of service hours and discharge.