Measuring the NISQ Decoherence Threshold in AI-Orchestrated Quantum Optimization for Pediatric Combination Therapy: A Synthetic Clinical Case Study

This preprint describes a synthetic clinical case study exploring the feasibility of using quantum computing to select the most balanced combination therapy for a child with co-occurring developmental conditions. The work was performed on a synthetic digital twin D - a virtual patient pre-loaded with all available clinical information. No real patients were involved, and no personal data were used. Why a quantum computer was used. The starting point was a deep investigation of all known therapeutic approaches relevant to the case - not only medications, but also psychotherapy, dietary supplements, vitamins, and micronutrients. These options differ in kind: some are drugs with interaction and safety risks, some are non-pharmacological modules, some are nutritional supports. A clinician therefore faces not a single choice but a combination problem - which components, in what number, and with which trade-offs. Once the full body of information was collected and structured, it became clear that the number of plausible combinations far exceeds what can be handled by informal reasoning in a short clinical visit. The researcher then decided to use quantum computing as an experimental tool to systematize and analyze this volume of information and to search for the best possible combination under explicit constraints. The quantum computer does not replace clinical judgment and is not a "treatment optimizer"; it is an instrument for exploring the practical limits of current quantum technology in this kind of task. Why this is a quantum-type problem. The clinical decision space is combinatorial. From 1,247 candidate options, 25 binary decision variables were retained, each representing inclusion or exclusion of a therapeutic component - whether a medication, a psychotherapy module, a supplement, a vitamin, or a micronutrient. The task was to minimize a weighted sum of included components plus pairwise interaction and safety penalties, subject to a limit on the number of concurrent components. Even 25 binary variables give more than 33 million possible subsets before constraints are applied, making this a discrete constrained optimization problem. QAOA was used as an approximate sampler for this constrained problem, with special initialization and mixing rules to keep the solution within the allowed number of active components. In the broader clinical setting, each option may also have multiple dose levels; including dose levels would turn the problem into a multi-choice optimization, increasing the variable space multiplicatively. This preprint did not perform dose optimization; it addressed component selection only. Comparison with traditional recommendations. In this study, the results obtained via quantum computing were compared with traditional clinical recommendations - primarily guideline-based stepped care and monotherapy. The goal was not to replace these recommendations but to understand whether a quantum approach could complement them, help organize a large number of alternatives, and make the selection process more transparent and auditable. Traditional frameworks were used as a reference point and as safety constraints. What was shown. On small test problems, a distinguishable signal was obtained: the quantum system more often selected the same logical combinations than random noise would. On a real current quantum device with 8 parameters, the signal was weak but statistically consistent. With 25 parameters, the result essentially dissolved into noise. This localized the practical limit of current quantum computing for this class of clinical tasks. The clinically useful output is not a ready-made prescription but an auditable shortlist of hypothetical options for multidisciplinary discussion. Purpose and significance. This work was conducted exclusively for research purposes - to encourage leading specialists in the field to initiate their own studies in this direction and to move toward personalized medicine, where quantum-enhanced AI systems may in the future help clinicians tailor therapy to each patient's individual characteristics. To the best of our knowledge, this is one of the first attempts to use quantum computing to optimize a large volume of clinical data when selecting combination therapy in clinical medicine, demonstrated here on a pediatric synthetic case. Quantum computing has already been applied to specific medical tasks, such as radiotherapy optimization, so we do not claim this is the first work in the history of medicine; it is one of the first attempts in this particular clinical scenario. The current result is only a proof-of-concept on synthetic data and is not validated for clinical use. Disclaimer (EU / Legal Notice): This is a preprint and has not been peer-reviewed. It is intended for research and discussion purposes only. This work does not constitute medical advice, a medical recommendation, or clinical decision support (CDS) within the meaning of Regulation (EU) 2017/745 (MDR) or Regulation (EU) 2024/1689 (AI Act). It is not a medical device and has not been certified by any notified body. The clinical case described is a synthetic digital twin (D). No real patient data, including any personal data as defined under the General Data Protection Regulation (EU) 2016/679 (GDPR), were used. No identifiable patient information is included. The candidate combinations presented are hypothesis-generating only and must not be interpreted as dosing guidance, treatment orders, or prescribing information. Final clinical decisions remain the sole responsibility of licensed healthcare professionals. Computational platforms. The 5-variable and 25-variable probes were executed on the BlueQubit CPU simulator, and the 8-variable hardware scaling probe was executed on IBM Fez. Use of these platforms is subject to their respective terms of service. In accordance with IBM Quantum citation guidelines, the IBM Quantum service should be cited as: IBM Quantum. https://quantum.cloud.ibm.com/, 2026. The quantum processing unit used in this work is IBM Fez (ibm_fez). BlueQubit's terms of use are available at https://www.bluequbit.io/. The quantum computing results are derived from free-tier, noisy intermediate-scale quantum (NISQ) hardware and simulators. They are not validated for clinical deployment. This preprint explicitly limits its scope to computational methodology and does not claim any diagnostic or therapeutic utility.