About

Noémi Kreif is an Assistant Professor at the University of Washington School of Pharmacy, affiliated with the CHOICE Faculty in the Department of Pharmacy. Her expertise encompasses big data, causal inference, health economics, health outcomes, health policy, and machine learning. She holds a PhD in Health Economics from the London School of Hygiene and Tropical Medicine (LSHTM), University of London, along with an MA in Economics from the Central European University in Hungary and an MSc in Economics from Corvinus University of Budapest, Hungary. Her research focuses on causal inference methods for cost-effectiveness analysis and health policy evaluation, as well as the quantitative evaluation of national health policy reforms. She is involved in incorporating machine learning techniques into health policy evaluation, health economics, and outcomes research, with an emphasis on personalized and stratified decision making. Kreif's work aims to advance methodologies for evaluating health interventions and policies, contributing to evidence-based decision making in healthcare.

Research topics

• Economics
• Political Science
• Sociology
• Medicine
• Computer Science
• Economic growth
• Mathematics
• Psychology
• Management science
• Finance
• Algorithm
• Applied psychology
• Management
• Programming language
• Business
• Data science
• Public administration
• Cognitive psychology
• Medical education

Selected publications

• Depression Symptoms, Employment, and Household Economic Outcomes: Evidence from the MIND-ECON Trial in South Africa

  SSRN Electronic Journal · 2026-01-01

  preprintOpen access
  PublisherDOI

• Transferability of Real World Evidence to Support HTA Recommendations in Lower Income European Countries

  Health Science Reports · 2026-01-27 · 1 citations

  articleOpen access

  Background and Aims: Lower income European countries (LIECs) have more limited financial resources to cover high-cost technologies in rare diseases than higher income European countries (HIECs). Our study explores how treatment recommendations in myelodysplastic syndrome (MDS) can be supported in LIECS by transferring real-world evidence (RWE) generated by target trial emulation (TTE) method in HIECs. Method: In the HTx project transferability aspects of the MDS case study were considered upfront. HTA agency consortium partners set expectations for the MDS case study team on how to integrate the new TTE methodology into the routine work of HTA bodies. In consecutive workshops consortium members and external HTA experts identified the main challenges of transferring evidence generated by TTE method to LIECs and made conclusions on how to overcome these challenges. Results: The lack of local real-world data before making reimbursement decisions is an important challenge to apply the TTE method to LIECs. Differences in patient pathways and comparator technologies, limited expertise and resources for adapting international HTA methods are significant barriers of transferring RWE from other countries.Still, transferring RWE to LIECs from other countries based on the TTE methodology represents an improvement to the current standard HTA methods, especially if joint clinical assessment provides the unbiased judgement on the relative effectiveness of orphan medicines. The TTE approach also provides an opportunity to LIECs to judge the value of high-cost technologies for different patient subgroups. However, HTA professionals in LIECs need training about advanced methodologies. Conclusion: This is the first study to explore how RWE generated by the TTE method can be transferred to optimize treatment decisions of patients with a rare disease in countries with limited HTA capacities. Five general concluding statements were made on the novelty of the TTE method and on how to overcome main challenges of transferring TTE results to HTA systems in LIECs.

  PublisherOA PDFDOI

• Real-world comparative effectiveness of novel doublet therapies for metastatic hormone-sensitive prostate cancer.

  Journal of Clinical Oncology · 2026-03-01

  article

  70 Background: Metastatic hormone-sensitive prostate cancer (mHSPC) is an advanced form of prostate cancer with a 5-year survival below 40%. Treatment guidelines recommend androgen deprivation therapy (ADT) combined with androgen receptor pathway inhibitors (ARPIs), docetaxel, or both. However, no head-to-head randomized trials have compared available doublet therapies, and real-world evidence remains limited. This study evaluated the real-world comparative effectiveness of ADT-based doublet regimens. Methods: Using the Komodo Health Claims Database, we identified US men aged ≥18 years with mHSPC who initiated ADT within 4 months of diagnosis. Eligible patients received index treatment with abiraterone, apalutamide, enzalutamide, or docetaxel within 4 months of ADT initiation between January 2019 and December 2023, with follow-up through June 6, 2025. Patients receiving triplet therapy, multiple index treatments, or with other cancer types were excluded. Primary outcomes were progression-free survival (PFS) and overall survival (OS). Progression was defined as time from index treatment date to metastatic castration-resistant prostate cancer (mCRPC), based on treatment initiation and hormone-resistance diagnosis code. Cox proportional hazards models were used to estimate hazard ratios and 95% confidence intervals using overlapping propensity score weights generated from machine learning-based generalized boosted models for confounding adjustment. Results: This study included 9,141 US men with mHSPC (abiraterone + ADT = 3,968; apalutamide + ADT = 1,096; enzalutamide + ADT = 1,902; docetaxel + ADT = 2,175). Each ADT-based ARPI doublet (abiraterone, apalutamide, and enzalutamide) demonstrated superior PFS compared to docetaxel + ADT, primarily due to prolonged time to mCRPC progression. However, no significant differences in PFS were detected among the three ARPI regimens. For OS, no significant differences were observed between any ARPI doublet and docetaxel + ADT, nor within the ARPI regimens. Full survival results are shown in Table 1. Conclusions: Real-world evidence shows superior PFS with each ARPI + ADT regime versus docetaxel + ADT for mHSPC, with comparable OS. Given similar survival outcomes among ARPI doublets, treatment selection may be guided by cost, toxicity profiles, and patient-specific factors. PFS and OS adjusted hazard ratios (aHR). Treatment Comparator PFS: aHR (95% CI) OS: aHR (95% CI) ABI + ADT DOC + ADT 0.55 (0.49 – 0.62) * 0.99 (0.86 – 1.14) APA + ADT DOC + ADT 0.59 (0.54 – 0.65) * 0.91 (0.75 – 1.11) ENZ + ADT DOC + ADT 0.59 (0.51 – 0.67) * 0.88 (0.75 – 1.04) APA + ADT ABI + ADT 0.99 (0.88 – 1.12) 0.92 (0.77 – 1.10) ENZ + ADT ABI + ADT 0.93 (0.84 – 1.03) 0.89 (0.78 – 1.03) APA + ADT ENZ + ADT 1.07 (0.93 – 1.22) 1.03 (0.85 – 1.25) ABI: abiraterone, ADT: androgen deprivation therapy, APA: apalutamide, DOC: docetaxel, ENZ: enzalutamide. *Significant findings.

  PublisherDOI

• Public Policies and Femicides during the COVID-19 Pandemic: Evidence from São Paulo, Brazil

  Economics & Human Biology · 2025-07-29 · 2 citations

  articleOpen access
  PublisherOA PDFDOI

• EE84 Cost-Effectiveness Analysis of Xanomeline and Trospium Chloride for Schizophrenia in the United States

  Value in Health · 2025-07-01

  articleSenior author
  PublisherDOI

• Psychological distress in adolescence and later economic and health outcomes in the United States population: A retrospective and modeling study

  PLoS Medicine · 2025-01-16 · 8 citations

  articleOpen accessCorresponding

  BACKGROUND: Federal policy impact analyses in the United States do not incorporate the potential economic benefits of adolescent mental health policies. Understanding the extent to which economic benefits may offset policy costs would support more effective policymaking. This study estimates the relationship between adolescent psychological distress and later health and economic outcomes and uses these estimates to determine the potential economic effects of a hypothetical policy. METHODS AND FINDINGS: This analysis estimated the relationship between psychological distress in those aged 15 to 17 years in 2000 and economic and health outcomes approximately 10 years later, accounting for an array of explanatory variables using machine learning-enabled methods. The cohort was from the National Longitudinal Study of Youth 1997 and nationally representative of those aged 12 to 18 years in 1997. The cohort included 3,343 individuals under age 18 years in round 4 who completed the Mental Health Inventory-5 (MHI-5). Round 1 captured 50 explanatory variables that covered domains of potential confounders, including basic demographics, neighborhood environment, family resources, family processes, physical health, school quality, and academic skills. The exposure included a binary variable of clinically significant psychological distress (MHI-5 score of less than or equal to 3) and a categorical variable of symptom severity on the MHI-5. Outcomes covered domains of employment, income, total assets at age 30 years, education, and health approximately 10 years later. Forty-seven percent of the cohort were black and Hispanic, and 4.4% had past-month clinically significant psychological distress. Past-month clinically significant psychological distress in adolescence led to a 6-percentage-point (95% confidence interval [CI] [-0.08, -0.03]) reduction in past-year labor force participation 10 years later and $5,658 (95% CI [-6,772, -4,545]) USD fewer past-year wages earned. We used these results to model the labor market impacts of a hypothetical policy that expanded access to mental health preventive care and reached 10% of youth who would have otherwise developed clinically significant psychological distress. We found that the hypothetical policy could lead to $52 (95% credible interval [51,54]) billion USD in federal budget benefits over 10 years from labor supply impacts alone. This study faced limitations, including potential unmeasured confounding, missing data, and challenges to generalizability. CONCLUSIONS: Our findings showed the impacts of adolescent mental health policies on the federal budget and found potentially large effects on the economy if policies achieve population-level change.

  PublisherOA PDFDOI

• Estimating the Causal Effect of Realistic Treatment Strategies Using Longitudinal Observational Data

  Medical Decision Making · 2025-10-27

  articleSenior author

  BackgroundReal-world data can inform health care decisions by allowing the evaluation of nuanced treatment strategies. Longitudinal observational data enable the assessment of dynamic treatment regimes (DTRs), strategies that adapt treatment over time based on patient history, but require causal inference methods to address time-varying confounding. Longitudinal targeted minimum loss-based estimation (LTMLE) is a machine learning-based double-robust approach for improved causal effect estimation.MethodsWe applied LTMLE to longitudinal registry data to evaluate the impact of erythropoiesis-stimulating agents (ESAs) in the clinical management of low to intermediate-1 risk myelodysplastic syndrome (MDS). We defined DTRs based on clinically relevant decision rules (e.g., commencing treatment when the hemoglobin level falls below a threshold) and compared them to static treatment regimes (always or never giving ESAs). Outcomes include mortality and health-related quality of life measured by EQ-5D scores.ResultsThe static regime of never administering ESAs resulted in declining counterfactual EQ-5D scores and increasing mortality risk over time. In contrast, both the static regime of continuous administration of ESAs and the use of dynamic regimes improved the EQ-5D scores and tended to reduce mortality, although the mortality differences were not statistically significant.ConclusionsThe article provides a case study application of the LTMLE method to evaluate realistic treatment policies under time-varying confounding. The findings support the potential benefits of dynamic treatment strategies for the management of MDS, highlighting the importance of personalized treatment adaptation. The study contributes methodological insights into the applications of LTMLE in small-sample, long-follow-up settings relevant to health technology assessment and policy making.HighlightsThis study applies the longitudinal targeted minimum loss estimation (LTMLE) method to evaluate the causal effect of static and dynamic treatment strategies using longitudinal observational data.We demonstrate the use of the LTMLE method to assess the impact of erythropoiesis stimulating agents (ESAs) on quality of life and mortality in patients with low to intermediate-1 risk myelodysplastic syndromes.The findings suggest that patients treated under dynamic ESA treatment regimes show an improved quality of life measured by EQ-5D scores and survival compared with those treated under the static treatment regime of never administering ESAs.This study contributes to the methodological literature by showcasing the application of the LTMLE method in a small-sample, long-follow-up setting with time-varying confounding, informing health technology assessment and policy decisions.

  PublisherDOI

• Estimating heterogeneous impacts Of subsidised health insurance: A causal machine learning approach

  PLoS ONE · 2025-09-29

  articleOpen accessSenior author

  The evaluation of social and health policies often necessitates understanding the variations in impacts based on recipients' observed characteristics, underscoring the value of estimating treatment effect heterogeneity. In this study, we leverage predictive and causal machine learning to assess the impact of the subsidised component of Indonesia's National Health Insurance Programme ("JKN") on healthcare utilisation in 2017. We employ causal forests for estimating heterogeneous treatment effects and the super learner algorithm for prediction tasks. Our approach addresses the prevalence of zeros in the utilisation outcomes through a two-part model, which separates the outcome model into zero and non-zero counts. This allows for distinct investigation of policy impacts on the decision to seek care and the quantity of care consumed. We interpret and summarise treatment effect heterogeneity using various approaches, including data-driven subgroup analyses and linear projections, which are grounded in theory. Our results demonstrate a positive average impact on healthcare demand with evident heterogeneity; for instance, the increase in demand varies among recipients. We also find that the effect is modified by a set of theoretically motivated covariates and those identified through our data-driven approach.

  PublisherDOI

• Fast Learning of Optimal Policy Trees

  ArXiv.org · 2025-06-18

  preprintOpen accessSenior author

  We develop and implement a version of the popular "policytree" method (Athey and Wager, 2021) using discrete optimisation techniques. We test the performance of our algorithm in finite samples and find an improvement in the runtime of optimal policy tree learning by a factor of nearly 50 compared to the original version. We provide an R package, "fastpolicytree", for public use.

  PublisherOA PDFDOI

• Using Policy Learning to Inform Health Insurance Targeting: A Case Study of Indonesia

  Health Economics · 2025-09-08

  articleSenior author

  This paper demonstrates how optimal policy learning can inform the targeted allocation of Indonesia's two subsidized health insurance programmes. Using national survey data, we develop policy rules aimed at minimizing "catastrophic health expenditure" among enrollees of APBD or APBN, the two government-funded schemes. Employing a super learner ensemble approach, we use regression and machine learning methods of varying complexity to estimate conditional average treatment effects and construct policy rules to optimize program benefits, both with and without budget constraints. We find that the financial impact of APBD enrollment over APBN differs with household characteristics, particularly demographic composition, socioeconomic status, and geography. Households assigned to APBD under the policy rule are typically urban-based with better facilities, whereas rural households with less accessible healthcare are assigned to APBN-a pattern intensified under budget constraints. Both constrained and unconstrained optimal policy assignments show lower expected catastrophic expenditure risk than the current assignment strategy. This study contributes to the literature on heterogeneous treatment effects, optimal policy leaning, and health financing in developing countries, showcasing data-driven solutions for more equitable resource allocation in public health insurance contexts.

  PublisherDOI

Frequent coauthors

• Ágnes Benedict

  Ashford and St Peter's Hospitals NHS Foundation Trust

  21 shared

• Richard Grieve

  London School of Hygiene & Tropical Medicine

  18 shared

• Claudie Charbonneau

  Pfizer (France)

  16 shared

• Sreedharan Hariharan

  SRM Institute of Science and Technology

  16 shared

• Sylvie Négrier

  Université Claude Bernard Lyon 1

  16 shared

• Robert A. Figlin

  16 shared

• Marc Suhrcke

  Luxembourg Institute of Socio-Economic Research

  11 shared

• Rodrigo Moreno‐Serra

  University of York

  10 shared

Education

• Ph.D., Health Economics

  London School of Hygiene and Tropical Medicine

• M.A., Economics

  Central European University

• M.S., Economics

  Corvinus University of Budapest