The research team, led by clinicians at the Department of Child and Adolescent Psychiatry at Ege University Faculty of Medicine in İzmir, reviewed the medical records of children and adolescents aged 6 to 21 who had been diagnosed with ASD after their sixth birthday. Every diagnosis had been confirmed using the Autism Diagnostic Observation Schedule, Second Edition (ADOS-2), a standardized observational assessment conducted by a senior child and adolescent psychiatrist with more than two decades of autism expertise. The team then extracted data using structured case report forms that drew on clinical records, parental reports, and a literature-informed checklist of camouflaging-related features, scoring masking, camouflaging, and compensatory strategies on a 0-to-10 visual analog scale.

The results paint a picture of striking diagnostic complexity. Attention-Deficit/Hyperactivity Disorder (ADHD) was by far the most common co-occurring psychiatric condition, present in 89.1 percent of the sample, followed by anxiety disorders at 26.1 percent, obsessive-compulsive disorder and major depressive disorder at 17.4 percent each, and specific learning disorder at 15.2 percent. This prevalence of ADHD is consistent with the concept of diagnostic overshadowing, in which the visible symptoms of one condition, such as hyperactivity and inattention, draw clinical attention away from the subtler social communication difficulties that define autism. Previous research has shown that children first diagnosed with ADHD experience longer delays before an eventual ASD recognition, and the new findings reinforce that pattern in a pediatric clinical setting.

Sex-related differences in presentation emerged as a central theme. Masking-related social adaptation behaviors, defined in the study as a child’s attempts to imitate peers in order to communicate and play, or to participate in social settings when invited even without intrinsic inclination, were the only camouflaging-related characteristic that differed significantly between girls and boys, and they were markedly more frequent among girls (p = 0.001). Giftedness and easy temperament, categorized as compensatory factors, were also observed more frequently in girls and were identified among the factors associated with delayed recognition. In contrast, boys showed significantly greater behavioral difficulties than girls, a pattern suggesting that more overt, externalizing symptoms may lead to earlier recognition in boys, while girls, who often display more subtle and internalized patterns, may face greater risk of delayed diagnosis.

The gendered nature of camouflaging strategies has been documented in prior research. Autistic girls are more likely to engage in socially oriented camouflaging, such as imitating peers’ gestures and facial expressions, memorizing conversational scripts, and closely observing social interactions to learn norms. Autistic boys, by contrast, tend to suppress atypical behaviors like stimming and limit social engagement rather than actively imitate. The new study’s finding that social adaptation behavior was the only masking-related factor differing between sexes aligns with the hypothesis that gendered socialization, which often encourages girls to appear more social and to model socially appropriate behaviors, may increase the risk that core autistic features in girls remain hidden, thereby delaying or obscuring diagnosis.

When families were asked directly about the reasons for the delayed diagnosis, the most frequently cited factors were the failure of consulted physicians to diagnose ASD (58.7 percent), high intelligence (43.5 percent), delayed referral to a psychiatrist (39.1 percent), difficulty accepting the diagnosis by families (34.8 percent), and having at least one family member with similar traits (30.4 percent). Camouflaging itself was identified as a reason for delay in 78.3 percent of cases. Only 39.1 percent of families had suspected ASD before the child turned six, and among those who had, 64.3 percent reported receiving a non-ASD diagnosis instead. These findings echo earlier work showing that delayed diagnoses can occur even when evaluations are carried out by multidisciplinary teams, underscoring the need for post-graduate education focused on less overt autism presentations.

The clinical stakes of late diagnosis are considerable. Early identification of autism is closely linked to access to appropriate support services, educational resources, and community programs, and it reduces parental stress while enhancing adult independence. Missed diagnoses restrict access to crucial early interventions during critical developmental periods. In the new study, the initiation of appropriate treatment and interventions following the late ASD diagnosis was associated with a significant decrease in Clinical Global Impression severity scores, from 4.45 to 2.80, and a significant increase in Global Assessment Scale functioning scores, from 52.71 to 73.04 (p < 0.001 for both). These retrospective changes suggest potential clinical benefits associated with accurate diagnosis and appropriate management, even when recognition comes years after the first referral.

One of the more technically intriguing results concerns the relationship between camouflaging and post-diagnosis functioning. Clinician-rated camouflaging scores demonstrated a moderate positive correlation with post-diagnosis Global Assessment Scale scores (r = 0.38, p = 0.009), a finding that contrasts with adult literature linking camouflaging to increased psychological burden and autistic burnout. However, this association was no longer statistically significant after controlling for baseline GAS using partial correlation, suggesting that the observed relationship may partly reflect pre-existing differences in functioning rather than a direct effect of camouflaging itself. Camouflaging rate was also correlated with follow-up CGI score (r = -0.36), CARS score (r = 0.43), and baseline CGI score (r = 0.33). In a multiple linear regression model predicting follow-up CGI scores, camouflaging rate, CARS score, and age at ASD diagnosis were all significantly associated with the outcome, while baseline CGI was not.

The study also documented the broader mental health context of this late-diagnosed group. Nine cases (19.6 percent) had a history of non-suicidal self-injury, most commonly self-cutting, two cases had a history of suicide attempts, and 17.3 percent reported passive suicidal thoughts. Previous research has shown that autistic individuals are at markedly increased risk of self-injury and suicidality compared to the general population, and such behaviors are often under-recognized in individuals with late or overshadowed diagnoses, where emotional and behavioral difficulties may mask core autistic features. The authors suggest that ASD screening is critical for cases presenting with self-harm and suicide attempts. Sensory sensitivities were present in 78.3 percent of participants, spanning visual, auditory, olfactory, gustatory, and tactile domains, a rate consistent with prior estimates ranging from 45 to 95 percent and a potential clinical clue during the diagnostic process.

The authors are careful to frame these findings as exploratory. The retrospective design, the small sample size of 46 participants, the single cultural context, and the use of a clinician-developed case report form rather than a standardized pediatric camouflaging instrument all limit generalizability. The gender imbalance in the sample, with 23.9 percent female and 76.1 percent male, may also have constrained the ability to detect sex-related differences in age at diagnosis. Notably, no significant difference was found between the diagnosis ages of girls and boys in this sample, a result that contrasts with previous findings of longer delays in females and may reflect the younger age of the cohort or the sample’s gender imbalance. The authors call for further prospective research using developmentally appropriate pediatric measures of camouflaging to clarify how these behaviors unfold across childhood and adolescence, and whether targeted assessment tools could capture the nuances of less overt autism presentations to improve diagnostic equity for all children on the spectrum.

Subject of Research: Clinical characteristics and camouflaging behaviors in children and adolescents with late-diagnosed autism spectrum disorder: a retrospective descriptive study

Article Title: Clinical characteristics and camouflaging behaviors in children and adolescents with late-diagnosed autism spectrum disorder: a retrospective descriptive study

Article References: Özbaran, N. B., Şener, Z. İ., Sarı, S., & Korkmaz Şahin, G. (2026). Clinical characteristics and camouflaging behaviors in children and adolescents with late-diagnosed autism spectrum disorder: a retrospective descriptive study. Frontiers in Child and Adolescent Psychiatry, 5, Article 1910194. https://doi.org/10.3389/frcha.2026.1910194

Image Credits: AI Generated

DOI: 10.3389/frcha.2026.1910194

Keywords: Clinical, characteristics, camouflaging, behaviors, children, adolescents, late-diagnosed, autism, spectrum, disorder, retrospective, descriptive

The research team, led by Armand Mbanya and Felipe Lobelo of The Southeast Permanente Medical Group and Emory University, analyzed the charts of 2,071 patients with laboratory-confirmed COVID-19 diagnosed between March 3, 2020, and September 11, 2020, during the first pandemic wave in metro Atlanta. Rather than lumping patients into the familiar binary of diabetes versus no diabetes, the investigators stratified them into five groups based on hemoglobin A1c (HbA1c), the glycated hemoglobin measure that reflects average blood glucose over roughly three months. Patients were classified as normoglycemic (HbA1c at or below 5.6 percent), prediabetic (5.7 to 6.4 percent), or diabetic with good control (below 7 percent), moderate control (7 to 9 percent), or poor control (above 9 percent), the last threshold aligned with the Healthcare Effectiveness Data and Information Set definition of poor diabetes management.

The distribution of the cohort itself tells a story. Nearly half of the infected patients, 986 individuals or 47.6 percent, had no diabetes. But 658 patients, 31.8 percent, had prediabetes, a condition affecting roughly one in three American adults that rarely triggers the clinical vigilance reserved for diagnosed diabetes. Another 183 patients (8.8 percent) had well-controlled diabetes, 162 (7.8 percent) had moderately controlled disease, and 82 (4 percent) had HbA1c above 9 percent. The prediabetes figure is striking on its own: nearly a third of all confirmed COVID-19 cases in this integrated health system occurred in patients whose glucose metabolism was already dysregulated, even if no physician had yet sounded the alarm.

The gradient in outcomes was unmistakable. Hospitalization rates rose in lockstep with worsening glycemic control, a pattern that held with high statistical significance (p < 0.0001). Patients with HbA1c above 9 percent fared worst: 38.3 percent required hospitalization and 14.8 percent were admitted to the intensive care unit, the highest proportions in the study. Normoglycemic patients without diabetes were hospitalized significantly less often than every other group. Prediabetic patients occupied an uncomfortable middle ground, hospitalized less often than those with moderately or poorly controlled diabetes, but their raw mortality proportion was the highest of any group in the sample, a finding the authors attribute in part to the structure of the health system itself. Diagnosed diabetic patients at Kaiser Permanente are followed closely, while many prediabetic patients may be undiagnosed or recently identified, leaving them with less proactive metabolic care heading into infection.

To move beyond raw associations, the team built multivariable logistic regression models adjusting for a remarkable breadth of confounders: age, sex, race and ethnicity, insurance type, household income, the neighborhood deprivation index, comorbidity burden measured by the Charlson Comorbidity Index, medications, quality-of-care metrics, and lifestyle behaviors including self-reported weekly exercise and smoking. In the fully adjusted model, prediabetes was independently associated with 1.43 times the odds of hospitalization (95 percent confidence interval 1.09 to 1.87), while insulin use carried 3.02 times the odds (95 percent confidence interval 1.95 to 4.68). Age, smoking (odds ratio 1.72), obesity (1.38), and physical inactivity (1.32) each independently raised hospitalization risk, while living in northeast Atlanta counties with higher median incomes and better housing quality was associated with lower risk (odds ratio 0.66).

When the investigators compared glycemic subgroups directly against one another, the dose-response relationship sharpened further. Relative to normoglycemic patients, those with prediabetes had 1.64 times the adjusted odds of hospitalization, patients with moderately controlled diabetes had 2.24 times the odds, and those with HbA1c above 9 percent had 3.3 times the odds. Most tellingly, prediabetic patients required hospitalization at rates not statistically distinguishable from patients with well-controlled diabetes, whose adjusted odds ratio of 1.04 relative to normoglycemia was essentially neutral. In other words, on the hospitalization endpoint, untreated glucose dysregulation in the prediabetic range behaved like established diabetes that happens to be well managed, a comparison the authors describe as prediabetes being a significant risk factor for severe COVID-19, comparable to poorly controlled diabetes in its clinical weight.

The biology underlying these statistics is increasingly well mapped. SARS-CoV-2 enters human cells through the angiotensin-converting enzyme 2 (ACE2) receptor, with the transmembrane serine protease TMPRSS2 priming the viral spike protein for entry and replication. Both pathways are dysregulated in cardiometabolic disease, and glycated ACE2 in diabetic tissue may act as an open door for viral entry into cardiomyocytes. High glucose levels directly increase viral replication in human monocytes, and glycolysis sustains that replication once it begins. Hyperglycemia also blunts the effectiveness of COVID-19 immunotherapy such as tocilizumab, drives endothelial dysfunction and hypercoagulability, and amplifies the cytokine storm that defines severe disease. Insulin itself may contribute to this inflammatory cascade by stimulating cytokine release, one proposed mechanism behind the threefold hospitalization odds observed among insulin users, who by definition carry the most advanced metabolic disease.

The study’s socioeconomic findings add a viral-news dimension that extends beyond the laboratory. Patients with the worst glycemic control lived in neighborhoods with the most unfavorable deprivation index (0.42 versus 0.04 for normoglycemic patients), had the lowest household incomes, and reported the least weekly exercise, averaging 52.5 minutes compared with 98.2 minutes for normoglycemic patients. Among prediabetic patients, higher neighborhood deprivation independently increased hospitalization odds (1.35), while commercial insurance, a proxy for healthcare access, halved them (0.49). Among patients with well-controlled diabetes, Black race was associated with 3.08 times the hospitalization odds, echoing umbrella reviews showing that Black and Hispanic individuals faced higher COVID-19 infection and hospitalization risks in the pre-vaccination era, at least partly attributable to higher comorbidity rates. The message is that glucose dysregulation and social disadvantage compound each other, concentrating severe COVID-19 risk in communities already burdened by obesity, inactivity, and limited access to care.

The authors are candid about limitations. The retrospective design, drawn from charts within a single integrated health system during the first pandemic wave before vaccines and antivirals were available, limits causal inference and generalizability. HbA1c values were the most recent measurements within 12 months of diagnosis, so glycemic control may have shifted between testing and infection, risking misclassification. Some social determinants data were measured at the area rather than individual level, and the small number of deaths (27 total, 1.3 percent) constrained statistical power for mortality endpoints. Notably, the prediabetes mortality finding, while statistically significant in unadjusted comparisons, did not survive as an independent predictor in the regression models, and the authors themselves flag the limited power around deaths as a key gap.

Still, the implications are hard to ignore. Previous landmark studies, including the CORONADO cohort and a single-center US analysis of hospitalized diabetic patients, found no association between HbA1c measured at admission and COVID-19 outcomes, leading some investigators to recommend against using HbA1c for in-hospital risk stratification. A meta-analysis by Prattichizzo and colleagues resolved part of the contradiction: pre-infection HbA1c was associated with COVID-19 mortality, but the association vanished when analysis was restricted to studies measuring HbA1c at admission. The Kaiser Permanente study, by anchoring glycemic status before infection, lands squarely on the side of the evidence suggesting that chronic glucose regulation, not acute admission hyperglycemia alone, determines who progresses to severe disease. With one in three American adults prediabetic and COVID-19 still circulating, the authors argue for proactive management, early detection of glucose dysfunction, vaccination outreach prioritized for metabolically at-risk populations, and prospective studies with more frequent glycemic monitoring. The quiet warning of this study is that the pandemic’s most vulnerable patients were never only those with diabetes. They were the millions walking around with glucose levels already drifting toward danger, unrecognized and untreated, long before the virus arrived.

Subject of Research: Glycemic control and COVID-19 outcomes across the continuum of glucose regulation: a retrospective cohort study

Article Title: Glycemic control and COVID-19 outcomes across the continuum of glucose regulation: a retrospective cohort study

Article References: Mbanya, A., Leung, S., Bienvenida, A., Leslie, E., Shin, S. R., Koplan, K. E., & Lobelo, F. (2026). Glycemic control and COVID-19 outcomes across the continuum of glucose regulation: a retrospective cohort study. Frontiers in Clinical Diabetes and Healthcare, 7, Article 1862850. https://doi.org/10.3389/fcdhc.2026.1862850

Image Credits: AI Generated

DOI: 10.3389/fcdhc.2026.1862850

Keywords: Glycemic, control, COVID-19, outcomes, across, continuum, glucose, regulation, retrospective, cohort, scientific research