About

Andrew Ho is the Charles William Eliot Professor of Education at the Harvard Graduate School of Education. He is a psychometrician whose research aims to improve the design, use, and interpretation of test scores in educational policy and practice. Ho is known for his development of methods for measuring educational progress and educational inequality. He is a developer of a national archive of student achievement data (SEDA) and advocates for using educational tests for low-stakes monitoring in multiple-measures systems. Ho has served as the Immediate Past President of the National Council on Measurement in Education, a trustee of the Carnegie Foundation for the Advancement of Teaching, and a member of the Governing Board for the National Assessment of Educational Progress. He also serves on Technical Advisory Committees advising state testing programs for seven states. At HGSE, he teaches courses in statistics and psychometrics and advises students in various education policy and evaluation programs. Ho holds a Ph.D. in Educational Psychology and an M.S. in Statistics from Stanford University. Before his academic career, he taught middle school creative writing in Honolulu, Hawaii, and high school Physics and AP Physics in Ojai, California.

Research topics

• Medicine
• Computer science
• Statistics
• Psychology
• Mathematics education

Selected publications

• Olfactory decline in aging: longitudinal trajectories and associations with cognitive decline and postmortem neuropathology

  medRxiv · 2026-04-28

  articleOpen access

  Importance: Decline in olfactory function may be used as a predictor of cognitive decline, to enhance early detection models, improve risk stratification, and enable early intervention. Objective: To assess the longitudinal association between olfactory decline, cognitive decline, and postmortem neuropathology. Design setting and participants: Retrospective longitudinal analysis with clinicopathological correlations of a prospective population-based cohort study using data from the Arizona Study of Aging and Neurodegenerative Disorders (AZSAND) and its Brain and Body Donation Program. Participants included cognitively unimpaired individuals without parkinsonism that converted to mild cognitive impairment (MCI) and/or dementia or remained cognitively stable. Main Outcomes and Measures: longitudinal change in olfaction, neuropsychiatric symptoms, motor function and memory, conversion to MCI/dementia, postmortem neuropathology. Results: Over a mean follow-up period of 7.7 ± 5.4 years, out of 922 participants who were cognitively unimpaired at the first cognitive conference, 643 remained cognitively unimpaired, 279 converted to MCI, and 82 developed dementia. Of these, 633 individuals had at least 2 olfactory tests.Converters showed reduced olfactory function (t=-12.6, p <0.0001), faster progression in neuropsychiatric symptom burden (t=3.42, p < 0.001), and faster decline in memory (t= -7.33, p <0.0001) prior to conversion while no significant differences were observed in motor scores between converters and non-converters. Using ROC analysis, olfactory decline, increased neuropsychiatric symptom burden, as well as motor and memory decline predicted conversion to MCI with a consistent accuracy of ~ 70% up to 5 years before conversion, while UPSIT alone had an accuracy of ~ 60%. Longitudinal decline in olfaction was associated with a higher burden of a-synuclein (t= -8.21, p <0.0005), tau tangle (t= -2.66, p < 0.01) and amyloid plaque burden (t= -2.85, p < 0.005) and a faster decline over time was associated with a higher burden of tau (t=5.66, p<0.0001). Conclusions and Relevance: A reduction in olfactory identification ability is observed up to a decade prior to conversion to MCI and is associated with underlying burden of neuropathology markers, underscoring the value of incorporating olfactory testing in cognitively unimpaired individuals to identify those at-risk of future cognitive decline.

  PublisherOA PDFDOI

• Corrigendum to “Accelerating high-order continuum kinetic plasma simulations using multiple GPUs” [J. Comput. Phys. 540 (2025) 114271]

  Journal of Computational Physics · 2026-04-16

  article1st authorCorresponding
  PublisherDOI

• P17.12.B INVESTIGATING SHORT TERM (&amp;lt;6 MONTHS) SURVIVORS OF GLIOBLASTOMA - AN ANALYSIS OF THE HISTO-MOL GBM COLLABORATIVE DATABASE.

  Neuro-Oncology · 2025-10-01

  articleOpen access

  Abstract BACKGROUND Population level survival for glioblastoma (GBM) remains limited at only 9-12 months. However, whilst a minority of patients have extended survival &amp;gt;24 months (long term survivors, LTS), a significant proportion survive &amp;lt;6 months (short term survivors, STS) and the characteristics of STS are poorly defined. Further, the survival benefit with concurrent temozolomide seen in clinical trials only emerges &amp;gt;6 months, with no separation in the survival curves before this. MATERIAL AND METHODS Using a large multicentre retrospective database of pathologically confirmed IDH wildtype (IDHwt) GBMs diagnosed in 2021, we investigated differences in the clinical and treatment characteristics of STS compared to LTS. Differences were assessed by independent samples T-test and survival was assessed using Kaplan-Meier methodology. RESULTS Of 1612 GBM patients, there were 520 STS (32%) and 243 LTS (15%). Age at diagnosis was increased in STS compared to LTS (median 65 [interquartile range: 59-72] vs 58 years [interquartile range: 50-67], p&amp;lt;0.001), whilst gender balance was not changed. STS presented with more motor (47% vs 30%, p&amp;lt;0.001), cognitive (30% vs 21%, p=0.007) and behavioural (10% vs 4%, p=0.002) symptoms, but fewer seizures (20% vs 30%, p=0.003) and headaches (31% vs 39%, p=0.027). Performance status (PS) significantly differed (p&amp;lt;0.001) with a greater proportion of PS&amp;gt;2 (36% vs 9%) and a lower proportion of PS0 (18% vs 47%) in STS. STS tumours were more commonly midline (10% vs 3%, p&amp;lt;0.001), multifocal (30% vs 14%, p&amp;lt;0.001) and contrast enhancing (97% vs 92%, p=0.002), but less frequently MGMT promoter methylated (40% vs 69%, p&amp;lt;0.001). Surgical strategy differed (p&amp;lt;0.001), with more biopsies (50% vs 13%) and fewer gross total resections (20% vs 56%) in STS. Oncological treatment also differed (p&amp;lt;0.001), with less STS receiving oncological treatment (37% vs 98%) and fewer patients received chemoradiotherapy (hypofractionated 10% vs 14%, or conventional 12% vs 80%). In STS who underwent adjuvant radiotherapy (n=206), there was an increase in the size of the planning target volume (median 386cc vs 268cc, p&amp;lt;0.001) and an increased proportion of patients had evidence of progression between surgery and radiotherapy (20% vs 11%, p&amp;lt;0.001). In STS patients who received radical radiotherapy (60Gy/30# or 40Gy/15#, n=156), there was improved survival with the addition of concurrent temozolomide (median survival 4.2 months vs 4.9 months, p=0.002). CONCLUSION We present a well characterised clinical cohort of pathologically confirmed IDHwt GBM STS, demonstrating clear differences from LTS. STS represent a third of pathologically confirmed GBM patients, plus additional patients who were not felt suitable for pathological confirmation. Further research to reliably predict STS at diagnosis would help patients make more informed decisions about their treatment.

  PublisherOA PDFDOI

• AN UNCOMMON CAUSE OF DIZZINESS IN A YOUNG PATIENT

  CHEST Journal · 2025-10-01

  articleOpen accessSenior author
  PublisherOA PDFDOI

• Identification of novel small molecule chaperone activators for neurodegenerative disease treatment

  Biomedicine & Pharmacotherapy · 2025-04-14 · 2 citations

  articleOpen access1st authorCorresponding

  A pathological hallmark of neurodegenerative disease is the accumulation of aberrant protein aggregates which contribute to the cytotoxicity and are therefore a target for therapy development. One key mechanism to manage cellular protein homeostasis is heat shock proteins (HSPs), protein chaperones which are known to target aberrant protein accumulation. Activation of HSPs target aberrant TDP-43, tau and amyloid to rescue neurodegenerative disease. As an attempt to target HSP activation for neurodegeneration therapy, we here develop a drug screening assay to identify compounds that will activate the master regulator of HSPs, the transcription factor heat shock factor 1 (HSF1). As HSF1 is bound by HSP90 which prevents its activation, we developed a NanoBRET assay, which allows us to monitor and quantify the HSF1-HSP90 interaction in living cells to screen for compounds disrupting this interaction and thereby releasing HSF1 for activation. After the optimisation and validation of the assay, a two thousand compound library was screened which produced 10 hits including two known HSP90 inhibitors. Follow-up functional study showed that one of the hits oxyphenbutazone (OPB) significantly reduces the accumulation of insoluble TDP-43 in a cell model, eliciting no signs of stress or toxicity. Overall, this study demonstrates a viable strategy for new drug discovery in targeting aberrant proteins and identifies potential candidates for translation into neurodegenerative disease treatment.

  PublisherDOI

• Efficacy and Safety of Ustekinumab and Vedolizumab for Crohn’s Disease of the Pouch

  Inflammatory Bowel Diseases · 2025-06-24 · 1 citations

  articleOpen access

  BACKGROUND AND AIMS: Medically refractory ulcerative colitis may require colectomy with ileal pouch-anal anastomosis. Complications of the J-pouch include pouchitis, occurring in 50%-80% of patients, and Crohn's disease (CD) of the pouch, occurring in 3%-17%. Our aim was to evaluate the efficacy and safety of ustekinumab (UST) and vedolizumab (VDZ) in patients with CD of the pouch. METHODS: This was a retrospective, multicenter cohort study of adults with CD of the pouch treated with UST or VDZ. The primary outcome was clinical response at 3 or 6 months. Secondary outcomes included clinical remission, endoscopic response, histologic response, pouch failure or surgery, and adverse effects of therapy. Multivariable logistic regression evaluated the efficacy and safety of UST versus VDZ, adjusted for age, smoking status, disease duration, corticosteroid use, and antibiotic use. Kaplan-Meier survival analysis evaluated the durability of UST versus VDZ for CD of the pouch. RESULTS: One hundred and four patients were included in this analysis. Seventy-seven patients were treated with UST and 57 patients were treated with VDZ between 2011 and 2021. A total of 64/77 (83%) UST-treated patients and 45/57 (79%) VDZ-treated patients had prior biologic exposure. Clinical response occurred in 62% UST-treated patients and 53% VDZ-treated patients at 3 months, and in 56% and 46% at 6 months, respectively. Clinical remission occurred in 32% UST-treated patients and 18% VDZ-treated patients at 3 months and 29% and 21% at 6 months, respectively. Among those treated with UST, 41% achieved endoscopic response, 10% achieved endoscopic remission, 46% achieved histologic response, and 7% achieved histologic remission. Among those treated with VDZ, 27% achieved endoscopic response, 16% achieved endoscopic remission, 26% achieved histologic response, and 8% achieved histologic remission. Over a follow-up period of 3 years, 5% UST-treated patients had inflammatory bowel disease (IBD)-related hospitalization, and 9% required pouch-failure surgery. In total, 3% VDZ-treated patients had IBD-related hospitalization and 5% required pouch-failure surgery. Reported adverse effects were uncommon, including arthralgias (1), hair loss (1), syncope (1), and upper respiratory infection (1) for UST and wrist edema (1) and elevated transaminases (1) for VDZ. In multivariable analyses, patients on UST were more likely to have a clinical response compared to VDZ at 3 months (OR 2.73, 95% [CI] 1.13-6.56, P = .025) and 6 months (OR 2.53, 95% CI: 1.01-6.29, P = .046). UST had significantly longer durability of treatment than VDZ (log-rank P < .005). CONCLUSIONS: In one of the largest cohorts evaluating UST and VDZ for CD of the pouch thus far, these biologics were found to be safe and effective treatments for CD of the pouch.

  PublisherDOI

• THE LANDSCAPE OF SECOND LINE (AND BEYOND) TREATMENTS FOR GLIOBLASTOMA (IDH WILDTYPE) IN THE UK IN 2021: AN UPDATE FROM THE HISTO-MOL GBM COLLABORATIVE

  Neuro-Oncology · 2025-09-01

  articleOpen access

  Abstract AIMS First line treatment for glioblastoma has been standardised for the last 20 years. However, there is signifi- cant variation in choice of second (and subsequent) line treatments. In particular the role of re-resection and re-irradiation are unclear. Using a large individual patient database, we sought to describe the landscape of subsequent lines of treatment. METHODS Using the Histo-Mol GBM Collaborative database, a large individual patient database of patients diagnosed with pathologically confirmed glioblastoma in the UK in 2021, we investigated choice and outcomes of subsequent lines of treatment. RESULTS To date, 1481 patients from 43 oncology centres are included in the Histo-Mol GBM Collaborative database. From this cohort, 1078 (73%) demonstrated confirmed progressive disease (PD), of whom 476 (44%) received second line treatment. Systemic therapy (SACT) was the main second line treatment, provided in 336 patients (71%), followed by surgery (84 patients, 18%) and radiotherapy (30 patients, 6%). Only 13 patients (3%) were documented as being enrolled in a clinical trial. SACT was lomustine in 180 patients (54%), PC/PCV in 72 patients (21%), and Temozolomide in 68 (20%). Confirmed second progression occurred in 320 (67%), of whom only 123 (38%) went on to have 3rd line treat- ment. This was primarily systemic therapy (92 patients, 75%), with a minority receiving radiotherapy (15 pa- tients, 12%) or surgery (12 patients, 10%). 4th (28/71 patients, 39%) and 5th (4/16 patients, 25%) line treatments were rarely offered. In total, only 100 patients (9% of all patients with PD) received re-resection, whilst 48 patients received re- irradiation (4% of all patients with PD). Re-irradiation fully overlapped the initial PTV in 25 patients (52%), had marginal overlap in 9 patients (19%), and no overlap in 14 patients (29%). CONCLUSION Less than half of all patients receive second line treatment at progression. Re-resection and re-irradiation are infrequently used across all subsequent treatment lines.

  PublisherOA PDFDOI

• Impact of model uncertainty on SPARC operating scenario predictions with empirical modeling

  ArXiv.org · 2025-06-11

  preprintOpen access

  Understanding and accounting for uncertainty helps to ensure next-step tokamaks such as SPARC will robustly achieve their goals. While traditional Plasma OPerating CONtour (POPCON) analyses guide design, they often overlook the significant impact of uncertainties in scaling laws, plasma profiles, and impurity concentrations on performance predictions. This work confronts these challenges by introducing statistical POPCONs, which leverage Monte Carlo analysis to quantify the sensitivity of SPARC's operating points [1] to these crucial variables. For profiles, a physically motivated gradient-based functional form is introduced. We further develop a multi-fidelity Bayesian optimization workflow that effectively identifies operating points maximizing the probability of meeting performance goals, which gives a significant speed-up over brute force methods. Our findings reveal that accounting for these uncertainties leads to an optimal operating point different from deterministic predictions, which balances H-mode access, confinement, impurity dilution, and auxiliary power.

  PublisherOA PDFDOI

• Item-Level Heterogeneity in Value Added Models: Implications for Reliability, Cross-Study Comparability, and Effect Sizes

  Journal of Educational and Behavioral Statistics · 2025-12-28 · 1 citations

  article

  Value added models (VAMs) attempt to estimate the causal effects of teachers and schools on student test scores. We apply Generalizability Theory to show how estimated VA effects depend upon the selection of test items. Standard VAMs estimate causal effects on the items that are included on the test. Generalizability demands consideration of how estimates would differ had the test included alternative items. We introduce a model that estimates the magnitude of item-by-teacher/school variance accurately, revealing that standard VAMs can overstate reliability and overestimate differences between units. Using 16 academic outcomes from 8 studies with item-level data, we show how standard VAMs overstate reliability by a median of 0.04 on the 0 to 1 reliability scale (mean = 0.09, SD = 0.10) and provide standard deviations of teacher/school effects that are a median of 3% too large (mean = 12%, SD = 23% points). We discuss how imprecision due to heterogeneous VA effects across items attenuates effect sizes, complicates comparisons across studies, and contributes to temporal instability, though these effects are reduced when the number of items is high. Our results suggest that accurate estimation and interpretation of VAMs may be improved using item-level data, including qualitative data about how items represent the content domain.

  PublisherDOI

• Three Metrics for Monitoring Educational Progress when Tested Populations Change

  Teachers College Record The Voice of Scholarship in Education · 2025-07-01

  articleOpen access1st authorCorresponding

  Background/Context: Public monitoring of educational progress and inequality often involves tracking changes in the percentage of “proficient” students across groups and over time. These trends are important signals of state and district provision of educational opportunity. I show how known flaws of this percentage metric, sometimes assumed to be negligible, interacted with COVID-19 pandemic conditions to create a “perfect storm,” masking real declines and growing inequality. I use this to motivate three metrics necessary for fuller documentation of educational progress and inequality when tested populations change. Purpose/Objective/Research Question/Focus of Study: I present three metrics for measuring and contextualizing changes in educational achievement over time: the “match rate,” the “fair trend,” and the “equity check.” Like doctors or pilots using multiple instruments to diagnose and navigate, I argue that these three metrics are necessary for holistic understanding of educational progress when tested populations change. I show how neglecting these metrics leads to misclassification of schools that do and do not need support. These metrics have their foundations in the statistical literature for missing data and causal inference. I adapt them to the context of public reporting of educational test scores for monitoring educational equity. Research Design: I use publicly available data from the California Department of Education from 2019 through 2022 to show how poor reporting metrics led state officials to conclude that test score gaps were closing when they were in fact widening. Drawing from statistical theory, I show how these issues generalize to other contexts. I use statistical models to define three metrics that avoid biases and provide necessary context when tested populations change. The first metric is a percentage I call the “match rate.” The second and third metrics, the “fair trend” and the “equity check,” are regression-adjusted trends for changing populations. I explain how education officials can use these metrics to improve diagnosis, like doctors supplementing a patient’s pulse with their temperature, blood pressure, and oxygen saturation. Conclusions/Recommendations: Public reporting using simple metrics like “percent proficient” only yields defensible trend interpretations under conditions that are increasingly narrow and rare: when leaders care only about progress for a single stable population of students. The predictable biases and distortions of percent-proficient metrics necessitate more sophisticated metrics, simply explained, as complements. States and testing agencies should report metrics like the three I propose for transparency in technical documentation and wield them for decision-making, particularly when monitoring equity for changing populations.

  PublisherOA PDFDOI

Frequent coauthors

• Justin Reich

  Massachusetts Institute of Technology

  66 shared

• Isaac L. Chuang

  60 shared

• Jim Waldo

  58 shared

• Daniel Seaton

  34 shared

• Sergiy Nesterko

  Harvard University Press

  30 shared

• Tommy Mullaney

  Sema4 (United States)

  29 shared

• Sean F. Reardon

  Stanford University

  10 shared

• Christina Ha

  10 shared

Labs

• Harvard Graduate School of EducationPI

Awards & honors

• Elected to the American Academy of Arts and Sciences (2026)