Can the GAD-7 be used to create valid, reliable, feasible and actionable outcome-focused paediatric anxiety quality measures for use across specialty and paediatric treatment settings? Authors: Tami L Mark, Lissette M Saavedra, Sydney Kirsch, Ye Z Pogue, Frances L Lynch, Corinna Koebnick, Antonio A Morgan Lopez, Anna C Yaros, Emily K Juel Published: 2026 ABSTRACT AbstractIntroductionPaediatric anxiety disorders are among the most common psychiatric disorders of childhood and adolescence, affecting between 15% and 30% of youth. Rates rose sharply during the COVID-19 pandemic, and untreated anxiety is associated with impaired functioning and increased risk of depression, suicidality and substance use in adulthood. Evidence-based treatments such as cognitive-behavioural therapy and selective serotonin reuptake inhibitors are highly effective yet remain underused, while non-recommended treatments, including benzodiazepines, are often prescribed. Despite the public health burden, there are currently no endorsed outcome-focused quality measures for paediatric anxiety. Developing such measures would allow health systems and providers to track outcomes, reduce disparities and drive quality improvement.Methods and analysisThis study will develop and test two complementary outcome-focused paediatric anxiety quality measures based on the seven-item Generalised Anxiety Disorder scale: (1) remission (clinically significant reduction in anxiety symptoms to below threshold) and (2) response (clinically significant improvement, even if remission is not achieved). Using electronic health record and claims data from Kaiser Permanente Northwest and Kaiser Permanente Southern California from 2017 to 2024, we will establish measure inclusion and exclusion criteria, evaluate the application of symptom weighting, test measure reliability and validity and implement casemix adjustment models. Analyses will examine disparities by race, ethnicity and social risk factors through linkage indices capturing local social determinants of health. Qualitative interviews with providers and electronic medical record vendors will be conducted to assess feasibility and usability. Measures will be prepared and submitted for endorsement by a national quality measurement body.Ethics and disseminationThis study has been reviewed and approved by the institutional review boards at RTI International and Kaiser Permanente. Findings will be disseminated through peer-reviewed publications, scientific conferences, stakeholder workshops and submissions to a national quality endorsement body. STRENGTHS AND LIMITATIONS OF THIS STUDY To our knowledge, this is among the first US-based studies to develop and validate casemix-adjusted paediatric anxiety outcome quality measures at scale. To our knowledge, this is among the first US-based studies to develop and validate casemix-adjusted paediatric anxiety outcome quality measures at scale. The study leverages a large and diverse dataset (>19 000 youth aged 12–17 across more than 600 providers). The study leverages a large and diverse dataset (>19 000 youth aged 12–17 across more than 600 providers). Advanced psychometric approaches, including symptom weighting using moderated non-linear factor analysis and measurement error-corrected multilevel modelling, will be evaluated as an alternative to standard scoring approaches. Advanced psychometric approaches, including symptom weighting using moderated non-linear factor analysis and measurement error-corrected multilevel modelling, will be evaluated as an alternative to standard scoring approaches. Analyses explicitly assess disparities by socioeconomic risk factors, advancing equity in paediatric mental health. Analyses explicitly assess disparities by socioeconomic risk factors, advancing equity in paediatric mental health. The study is limited to Kaiser Permanente integrated health systems, which may affect generalisability. The study is limited to Kaiser Permanente integrated health systems, which may affect generalisability. Introduction Anxiety disorders are among the most common psychiatric disorders in children and adolescents, with prevalence estimates ranging from 15% to 30% globally.1 2 During the COVID-19 pandemic, multiple large meta-analyses and population-based studies have documented a marked increase in anxiety in youth, with prevalence estimates higher than pre-pandemic.35 Paediatric anxiety is associated with impairment in relationships, physical health, social interaction, academic performance, daily functioning and reduced quality of life.68 Left untreated, anxiety frequently follows a chronic course and is associated with an elevated risk of depression, substance use, suicidality and chronic medical conditions in adulthood.913 Cognitive-behavioural interventions, particularly exposure-based interventions that constitute a core component of evidence-based treatment, are often underutilised and non-recommended treatments such as benzodiazepines continue to be prescribed in many settings.1416 Furthermore, treatment use remains low relative to need.5 Among youth who receive care, treatment response is highly variable. Some youth improve substantially, others show minimal change, and some worsen despite receiving nominally appropriate interventions.17 This treatment gap highlights the need for quality measures in paediatric anxiety care.18 The National Academies of Science and the Centers for Medicare and Medicaid Services (CMS) have called for outcome-based quality measures in mental health.19 Outcome measures provide actionable data for quality improvement and accountability, particularly when they are casemix-adjusted (eg, adjusting for differences in child demographic and clinical characteristics) to allow fair comparisons across providers and systems. Process measures indicate whether services were delivered but not whether care led to symptom or functional gains or declines. Many process quality measures are based on procedure billing codes captured in insurance claims, which reduces the reporting burden. However, in mental health, procedure codes are broad, for example, indicating that psychotherapy was delivered but not the type of psychotherapy (eg, cognitive-behavioural exposure therapy). This limits the usefulness of procedure codes as quality measures. Moreover, the manner in which the cognitive therapy was delivered, such as the level of therapeutic alliance, cannot be captured with administrative data such as insurance claims but may greatly influence effectiveness and outcomes.9 In contrast to process measures, outcome-focused quality measures assess whether care leads to symptom improvement. When patient-reported outcome-focused quality measures are aggregated to the provider or system level, they can support performance feedback, benchmarking and targeted quality improvement efforts.20 Outcome-focused measures are also essential for identifying inequities in care, as disparities often emerge more clearly in treatment response and recovery than when studying access alone. Without validated outcome-focused measures, health systems lack scalable tools to evaluate effectiveness and ensure that improvements in access translate into improvements in patient well-being. Despite this need, outcome-focused quality measures in the US are limited. For youth mental health, the primary example is the nationally endorsed outcome measure of depression treatment remission and response in youth, which is based on the Patient Health Questionnaire-9 (PHQ-9).21 Currently, there are no outcome-based quality measures for paediatric anxiety, one of the most prevalent psychiatric disorders in youth.22 The present study addresses this critical gap by developing and testing two outcome-focused paediatric anxiety quality measures based on the widely used Generalised Anxiety Disorder–7 (GAD-7).23 The GAD-7 was selected as the basis for an anxiety outcome measure because (1) it has strong psychometric properties in both adult and adolescent populations23 24; (2) it is responsive to change and thus can be used to measure progress in reducing anxiety symptoms,25 and (3) it is feasible to implement at scale because it is brief and as evidenced by the fact that it is already used widely in US health systems, including Kaiser Permanente, the Veterans Health Administration and the Department of Defense,2629 in the UK and Australia’s mental health systems.30 31 By leveraging routinely collected electronic health record (EHR) data from two large integrated health systems, applying advanced psychometric methods and following national standards for quality measure development and endorsement, this project aims to create practical, scalable tools for monitoring outcomes, supporting quality improvement and advancing equity in paediatric mental healthcare. The specific study aims are as follows. To develop and test two complementary casemix-adjusted paediatric anxiety quality measures. To develop and test two complementary casemix-adjusted paediatric anxiety quality measures. To assess whether paediatric anxiety outcomes vary by individual demographic and socioeconomic characteristics. To assess whether paediatric anxiety outcomes vary by individual demographic and socioeconomic characteristics. To determine how the measures can be implemented and used for quality improvement through provider and EHR organisation feedback. To determine how the measures can be implemented and used for quality improvement through provider and EHR organisation feedback. To submit the measures for endorsement to a consensus-based entity quality measurement review organisation. To submit the measures for endorsement to a consensus-based entity quality measurement review organisation. Methods This is a protocol for a mixed-methods study that combines retrospective analysis of EHRs and claims data with qualitative interviews. The quantitative component will focus on measuring development, psychometric testing and descriptive analyses of measure distribution. The qualitative component will assess feasibility and usability through interviews with providers and EHR vendors. The study was awarded in October 2024, and the anticipated project duration is 4 years. Outcome measurement instrument The percentage of a clinician’s patients who achieved recovery or progress towards recovery from anxiety will be measured using the GAD-7, a brief, seven-item self-report questionnaire that captures the core symptoms of worry along with related symptoms such as restlessness and irritability. Data sources The study will be conducted using EHR and claims from Kaiser Permanente Northwest (KPNW) and Kaiser Permanente Southern California (KPSC), accessed through the Mental Health Research Network’s Virtual Data Warehouse, a governed research infrastructure that standardises and de-identifies data across participating health systems. No individually identifiable patient information is accessed or retained by the research team. KPNW serves approximately 650 000 individuals in Oregon and Southwest Washington. KPSC serves approximately 4.6 million individuals in Southern California. In KPNW and KPSC, the GAD-7 is routinely administered in both specialty and primary care settings. The study will use data from 2017 through 2024. Sample size and power Between 2017 and 2024, approximately 42 000 youth completed at least one GAD-7 assessment across KPNW and KPSC. Of these, 19 919 had ≥2 assessments, and 12 659 had baseline scores ≥10, indicating moderate or severe anxiety. Power simulations indicate that with >600 providers, the study will have adequate power (≥0.80) to detect clinically meaningful variation in remission and response rates between providers, as well as disparities across racial/ethnic and socioeconomic subgroups of youth. Quality measure specification development The quality measures will be developed and tested following the Centres for Medicare and Medicaid Services (CMS) guidance for quality measure development. These steps involve conceptualising the measures, specifying the measures (eg, determining denominator and numerator criteria), testing the measure to ensure validity and assessing the measures for feasibility and usability. Case identification (measure denominator specification) The study will use feedback from advisory panel members, review of the literature on the psychometric properties of the GAD-7, review of how the GAD-7 is used in other systems as a quality measure and data analyses to determine the most appropriate patient inclusion criteria. We will determine the age cut points (eg,10 vs 13 years old), the cut point for having anxiety as indicated by an elevated GAD-7 score, the types of anxiety disorders that should be included and whether any other disorders should be excluded. Defining anxiety, remission and response (measure numerator criteria) We will determine a threshold for defining remission and response. Using an adult sample, the developers of the GAD-723 determined that cut points of 5, 10 and 15 could be interpreted as representing mild, moderate and severe levels. A study in adolescents32 determined that a cut point of 11 had the best sensitivity and specificity for identifying cases of moderate anxiety. We will determine the population sizes and outcomes that result from using different cut points. We will explore whether remission and response should be calculated using differential symptom weighting.3335 Using item response theory and moderated nonlinear factor analysis (MNLFA), we will examine whether all GAD-7 symptoms contribute equally to total anxiety scores or if certain symptoms carry greater clinical weight. Prior item response theory research shows that equal weighting does not fit the data well and that items like ‘not able to stop worrying’ and ‘worrying too much’ have a greater influence on overall severity.36 Using equal weights can misclassify patients, treating those with the same total score as having equivalent anxiety even when their symptom patterns differ. To address this, we will estimate parallel GAD-7 scores using this MNLFA framework. This approach accounts for differential symptom weighting and measurement variation across populations and clinical presentations. MNLFA also provides person-specific reliability estimates, allowing more precise evaluation of individual change and recovery. Defining a period for measuring remission and response We will determine at what period remission or response should be measured, such as 6 months±60 days and 12 months±60 days, as is used for the depression remission measure. Handling lost-to-follow-up cases We will determine how to classify patients who do not complete follow-up GAD-7 assessments. For example, should these patients be considered as non-responders (as in Minnesota’s PHQ-9 measures)37 or excluded from both numerator and denominator?30 38 These decisions have implications for fairness and interpretability. Clinician attribution criteria Anxiety outcome quality measures will be created for all clinicians who treated at least 10 youth who meet study eligibility criteria. This criterion ensures sufficient sample size at the provider level to allow meaningful comparisons across clinicians and clinics and prevents re-identification. We will determine how ‘treatment’ should be defined. For example, should the quality measures be assigned to any clinician who delivered a GAD-7, any clinician who diagnosed anxiety in the clinical record or any clinician who treated the patient for anxiety using mental health interventions? Casemix adjustment Following CMS endorsement body guidelines, we will develop a casemix-adjustment approach. Casemix adjustment will adjust for factors such as patient severity, which influence the outcome but are independent of the quality of care. To develop a casemix-adjustment approach, we will first draw on the research and clinical literature and our team’s expertise in anxiety disorders. We will also review casemix-adjustment models currently used by organisations for the depression response and remission measure. The measure steward of those measures, the Minnesota Community Measurement, uses an ‘actual to expected’ methodology for casemix adjustment.37 In this methodology, each provider’s depression remission and recovery rates are compared with an ‘expected rate’ based on the specific characteristics of patients seen by the provider compared with the total patient population. Expected values are calculated using multinomial logistic regression, which controls for patient characteristics (eg, comorbid psychiatric and medical diagnoses, past hospitalisation or emergency department visit for a mental health condition, age and gender). A population proportions test is used to determine whether there is a statistically significant difference between the expected and actual rates. Benchmarking Benchmarking refers to the identification of a ‘performance leader’ whose outcomes and practices can be understood and emulated. Benchmarks should encourage improvement but also be attainable and perceived as such by providers. In this study, we will recommend an aggregate level of remission and response that providers should aim to achieve, such as 50%. Measure statistical testing Once we decide on the specification of the quality measures, we will conduct descriptive analyses, reliability testing and validity testing. Descriptive analyses will include calculation of the distribution of remission and response rates across patients, clinicians and (if feasible) clinics. We will also evaluate the effect of casemix adjustment. Reliability testing will consist of signal-to-noise analysis, intraunit reliability and Adams’s rho. Predictive validity will be evaluated by examining whether higher provider remission and response rates are associated with lower rates of subsequent mental health-related emergency department visits and hospitalisations. Face validity will be assessed through stakeholder feedback. Stakeholders will evaluate whether the measures address meaningful gaps in paediatric mental health quality measurement. Their input will help determine whether observed differences in remission and response rates are likely to reflect true quality variation rather than measurement error or random fluctuation. Usability and feasibility testing will be conducted through semistructured interviews with EHR vendors and provider organisations. Interviews will assess the feasibility of data collection, workflow integration and data standards, as well as usability for quality improvement initiatives. Template or matrix analysis will be used to code transcripts and identify themes related to feasibility and usability. Interviewees will be asked to consider how these measures might be adopted in clinical practice and quality improvement initiatives and what unintended consequences should be anticipated. For example, could providers become reluctant to treat youth with more severe anxiety if outcomes are publicly reported without adequate risk adjustment? Stakeholders will also identify strategies to mitigate such risks. Patient and public involvement A central feature of this project is structured, recurring engagement with a Technical Expert Panel (TEP) and a Youth patient Advisory Board (YAB). Together, these groups represent the perspectives of researchers, clinicians, quality measurement experts and families with lived experience of paediatric anxiety. Their contributions ensure that the measures we develop are not only psychometrically rigorous but also clinically meaningful, feasible to implement and acceptable to those most affected. The TEP and YAB will convene at least once annually throughout the 4-year study. Each meeting will be structured around the current stage of measure development, allowing the research team to present interim findings and draft specifications and to solicit targeted feedback. Meetings will be complemented by smaller ad hoc consultations with subgroups of the panel as specific technical or clinical issues arise. The TEP will include national experts in behavioural health quality measurement, with experience serving on measure endorsement committees and developing endorsed measures. These individuals bring technical expertise in psychometrics, risk adjustment and measure specification. The panel will also include researchers who specialise in paediatric anxiety, ensuring that developmental, diagnostic and treatment-specific issues are adequately considered. The YAB will include parents of children with anxiety disorders and adolescents with lived experience of anxiety treatment. Their contributions will ensure that the measures reflect outcomes that matter to patients and families and that implementation strategies are patient-centred.39 Together, TEP and YAB provide complementary expertise. The TEP emphasises technical, clinical and methodological rigour, while the YAB emphasises patient priorities, acceptability and equity. This combined structure ensures that the measures are both scientifically robust and meaningful in practice. Ethics and dissemination This study relies on secondary analysis of existing EHR and administrative data collected as part of routine clinical care and does not involve experimental manipulation of treatment or patient assignment. This study was reviewed by RTI International’s Institutional Review Board and determined to be exempt under Category 4 (secondary research on data or specimens; STUDY00022880). All data use will comply with the Health Insurance Portability and Accountability Act and applicable data governance requirements of participating health systems. The data used in this study are de-identified, aggregated EHR and administrative claims data derived from routine clinical care. No individually identifiable patient information is accessed, stored or analysed. Because the proposed quality measures are intended to support quality improvement rather than inform individual patient care, results will be reported only in aggregate at the clinician, clinic or system level with minimum denominator thresholds applied to prevent indirect re-identification. Dissemination and pathway to adoption Dissemination of the paediatric anxiety remission and response measures will occur through both the formal quality measurement infrastructure and the scientific literature. Following development and testing, the measures will be submitted for endorsement through the CMS consensus-based entity, which will make the specifications publicly available through national measure repositories and facilitate uptake by health systems, payers and EHR vendors. In parallel, study findings will be disseminated through peer-reviewed publications and presentations focused on (1) national benchmarks for paediatric anxiety outcomes, (2) psychometric and methodological foundations of the measures, (3) disparities in outcomes by race, ethnicity and social risk factors and (4) methodological advances in symptom weighting and reliable change estimation. Because the measures will be specified using Fast Healthcare Interoperability Resources standards, they will be designed for direct implementation within existing EHR systems, reducing barriers to adoption.40 Beyond accountability applications, the measures are intended to support audit-and-feedback and continuous quality improvement initiatives, enabling health systems to identify unwarranted variation in outcomes, target improvement strategies and track progress over time. Strengths and limitations The study’s strengths include access to a large, diverse paediatric population across two major integrated health systems, rigorous methodological approaches and alignment with the CMS endorsement framework. Another strength is our exploration of differential symptom weighting, which addresses limitations of using total GAD-7 scores that assume all symptoms contribute equally to anxiety severity. Prior research and our planned MNLFA analyses suggest that symptoms such as ‘not able to stop worrying’ and ‘worrying too much’ carry greater diagnostic weight than others.41 Ignoring these differences can distort estimates of improvement and recovery, particularly when comparing individuals or groups with distinct symptom patterns. By applying MNLFA-based scoring, we can produce more accurate estimates of underlying anxiety severity and account for variation in measurement across populations and clinical contexts.41 This approach enhances interpretability and supports more precise tracking of change, offering a stronger foundation for evaluating intervention effects. Although GAD-7 is widely used, reliance on a single instrument may not capture the full heterogeneity of paediatric anxiety presentations. Also, missing follow-up data is an inherent challenge in outcome measurement, and decisions about how to classify patients without repeat GAD-7s may affect results. Finally, while casemix adjustment improves fairness, it may also limit our understanding of disparities if not carefully implemented. By including place-based variables via the Child Opportunity Index, we will attend to this by measuring contextual factors, including children’s access to neighbourhood opportunity and other structural disadvantages that can impact treatment success.42 If successful, the paediatric anxiety remission and response measures developed here could serve as national models for wide adoption, similar to PHQ-9–based depression outcome measures. Health systems could use them to monitor outcomes, identify high‐performing providers, target areas for quality improvement and reduce inequities in paediatric mental healthcare. The project also lays the groundwork for expanding outcome measurement to other paediatric conditions, advancing the broader goal of embedding patient‐reported outcomes into quality frameworks.