AI-assisted stepwise drug reduction for chronic insomnia in the elderly and dynamic early warning system of traditional Chinese medicine

Abstract

The application discloses an AI-assisted step-by-step drug reduction and traditional Chinese medicine dynamic early warning system for senile chronic insomnia, relates to the technical field of traditional Chinese and western medicine collaborative diagnosis and treatment and health management of senile chronic insomnia, and comprises a data acquisition module, which acquires demographic data, four traditional Chinese medical examinations, and living habit data, automatically scores relevant scales, and monitors physiological indexes; an AI step-by-step drug reduction decision module, which stores western medicine data, analyzes curative effects, generates a drug reduction scheme, and optimizes the drug reduction scheme; a traditional Chinese medicine dynamic early warning module, which identifies common syndromes and monitors syndrome changes; a traditional Chinese and western medicine collaborative intervention module, which stores traditional Chinese medicine intervention schemes, schedules intervention time sequences, and adapts to individual differences; a safety monitoring module, which records adverse reactions in stages, early warns abnormal liver and kidney functions, and evaluates risks; and a data management module, which stores diagnosis and treatment data in a block chain, and anonymously processes sensitive information. The system realizes AI smooth drug reduction for senile chronic insomnia patients, adapts to primary medical scenes, improves the sleep quality of patients, and provides support for traditional Chinese and western medicine collaborative management.

Description

Technical Field

[0001] This invention relates to the field of integrated traditional Chinese and Western medicine diagnosis and treatment and health management technology for chronic insomnia in the elderly, and particularly to an AI-assisted step-by-step medication reduction and dynamic early warning system for traditional Chinese medicine in chronic insomnia in the elderly. Background Technology

[0002] The prevalence of chronic insomnia among people aged 60 and above in my country is as high as 46.0%, with a current elderly population exceeding 300 million. This results in annual direct medical costs exceeding 12 billion yuan, placing a heavy burden on individuals, families, and society. Current clinical interventions for chronic insomnia in the elderly have significant limitations: the internationally recommended primary care approach, Cognitive Behavioral Therapy for Insomnia (CBT-I), faces challenges such as heavy economic burden, low patient compliance, and a shortage of professional therapists, making it difficult to widely implement in primary healthcare settings. While drug therapy has a faster onset of action, long-term use can easily lead to increased tolerance, withdrawal syndrome, and the risk of overdose. Elderly patients often have multiple underlying diseases, increasing the risk of drug interactions with multiple medications. 77.1% of combination drug regimens have potential safety risks, and the incidence of catastrophic medical expenditures for families rises to 30.12%. The safety and cost-effectiveness of traditional drug treatment models fail to meet clinical needs.

[0003] Traditional Chinese medicine (TCM) has holistic advantages in the prevention and treatment of chronic insomnia in the elderly, often employing non-drug interventions such as herbal teas, acupressure, and foot baths. However, its clinical application suffers from insufficient standardization and unstable efficacy. TCM diagnosis relies heavily on the physician's individual experience, with a diagnostic accuracy rate of only 60%-65% for common patterns like heart-spleen deficiency and liver fire disturbing the heart. The identification and intervention of atypical patterns lack unified standards. While existing TCM home remedies (such as acupressure at specific times and herbal baths) are simple to perform, they are not tailored to individual patient constitutions and pattern differences. Some patients experience poor efficacy due to incompatible treatment plans, and the lack of real-time monitoring of symptom changes makes it difficult to detect pattern deterioration and adjust intervention strategies promptly. Furthermore, collaborative interventions between TCM and Western medicine often remain at a simple additive level, failing to establish a closed-loop mechanism of "drug reduction-early warning-intervention." The pace of Western medicine reduction and the timing of TCM intervention lack scientific coordination, easily leading to efficacy conflicts or intervention gaps, and failing to fully leverage the synergistic advantages of TCM and Western medicine in "reducing side effects and increasing efficacy."

[0004] With the accelerating aging process, the need for management of chronic insomnia in the elderly is becoming increasingly urgent in primary healthcare institutions. However, existing technological systems are ill-suited to the needs of these settings. On one hand, primary healthcare workers have limited understanding of integrated traditional Chinese and Western medicine approaches, lack AI and other technological tools to assist in precise decision-making, and rely heavily on experience to formulate medication reduction plans. This can lead to problems such as excessively rapid reduction causing sleep rebound or excessively slow reduction prolonging drug dependence. On the other hand, existing systems often lack offline functionality and user-friendly interfaces. Data collection and plan acquisition are hindered when network conditions are poor in primary healthcare settings, and compatibility with commonly used portable monitoring devices (such as micro-motion sensitive mattresses and simple tongue diagnostic instruments) is not designed, making data integration difficult. Furthermore, the management of patient medical data lacks standardization, often relying on traditional database storage, which poses a risk of data tampering. The lack of anonymization and traceability management fails to meet privacy protection requirements and hinders subsequent clinical research and plan optimization. These issues collectively restrict the promotion and application of integrated traditional Chinese and Western medicine management technologies for chronic insomnia in the elderly. Summary of the Invention

[0005] The present invention proposes an AI-assisted step-by-step medication reduction and TCM dynamic early warning system for elderly patients with chronic insomnia, in order to solve the problems mentioned in the prior art.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: an AI-assisted step-by-step medication reduction and TCM dynamic early warning system for elderly patients with chronic insomnia, comprising: The data acquisition module includes a multimodal data acquisition unit, an automatic scale scoring unit, and a physiological indicator monitoring unit. The former collects patient demographic, TCM four diagnostic methods, and lifestyle data. The scale unit has multiple built-in scales, and the scores are synchronized with the database. The physiological unit connects to the device to collect indicators, and collects data at different frequencies during the baseline period, intervention period, and follow-up period. The data is encrypted and transmitted to the local server. The AI-driven step-by-step medication reduction decision-making module includes a Western medicine dosage database, efficacy analysis, medication reduction plan generation, and plan iteration unit; the database stores information on commonly used sedative-hypnotic drugs and adjusts dosages according to drug characteristics and patient conditions; the efficacy unit uses a specific model to calculate a medication reduction feasibility score; the plan generation unit calculates the medication reduction range using a specific method; and the iteration unit regularly updates model parameters. The TCM dynamic early warning module includes a syndrome identification, syndrome change monitoring, and early warning triggering unit. The syndrome identification unit uses algorithms to identify common syndromes based on scale data and supports fuzzy matching of atypical syndromes. The monitoring unit tracks core symptoms and calculates the rate of change of syndrome scores periodically. The integrated traditional Chinese and Western medicine intervention module includes a library of integrated traditional Chinese and Western medicine intervention plans, intervention timing scheduling, and plan adaptation units. The Traditional Chinese and Western Medicine Intervention Program Database stores seven categories of intervention programs: herbal tea, acupoint massage, foot bath, acupuncture, massage, cognitive behavioral therapy, and lifestyle adjustment. Each program includes specific operating guidelines, dosage / frequency, and time recommendations, and the time recommendations are designed in accordance with the theory of meridian flow in traditional Chinese medicine. Intervention timing scheduling unit: Combining the timing of Western medicine administration with the AI-based step-by-step medication reduction decision module, avoiding peak drug effects, and generating a daily routine schedule for elderly patients with chronic insomnia; at the same time, coordinating the timing of various interventions: acupuncture and massage are spaced more than 4 hours apart; the "fixed wake-up time" node in cognitive behavioral therapy is staggered from acupoint massage by more than 30 minutes; morning exercise and breakfast are spaced 20 minutes apart in lifestyle regulation, and nighttime emotional regulation is arranged 30 minutes before taking Western medicine; Protocol adaptation unit: Dynamically adjusts protocol parameters based on patient syndrome type, physical condition, cognitive function, and tolerance; The safety monitoring module includes units for adverse reaction monitoring, liver and kidney function early warning, and medication reduction risk assessment. The adverse reaction unit records reactions according to standard classification. The liver and kidney function unit connects to test data, and triggers early warnings when indicators are abnormal. The risk unit calculates the risk level based on medication duration, medication reduction amount, and underlying diseases. The data management module includes blockchain notarization, data anonymization, and historical data traceability units; the blockchain stores medical treatment-related data, with each record containing a key identifier; the anonymization unit uses codes to replace sensitive information, desensitizing biological data; and the traceability unit supports multi-dimensional data retrieval.

[0007] Furthermore, it also includes a TCM syndrome risk assessment module, which comprises a risk factor extraction unit and a dynamic risk calculation unit. The risk factor extraction unit extracts syndrome deterioration-related factors from the four diagnostic methods of TCM; the dynamic risk calculation unit uses formulas... The syndrome risk value is calculated, where R is the syndrome risk value, t0 is the risk assessment start time, t1 is the current assessment time, θ is the syndrome weight coefficient, S(t) is the syndrome score change rate at time t, and D(t) is the underlying disease influence coefficient at time t. When R≥6, the warning level is automatically upgraded, prompting medical staff to strengthen monitoring and adjust the TCM intervention plan.

[0008] Furthermore, it also includes a patient compliance monitoring module, which comprises a medication monitoring unit, a traditional Chinese medicine intervention monitoring unit, and a compliance feedback unit. The medication monitoring unit records the patient's daily medication time and dosage through a smart pillbox; the traditional Chinese medicine intervention monitoring unit collects the patient's execution records of non-drug therapies such as acupoint massage and foot baths through a mobile APP, and verifies the authenticity of the intervention by combining it with device sensor data. When the execution rate is <70%, it is marked as low compliance; the compliance feedback unit feeds back the compliance level to the AI-based tiered medication reduction decision module, and pushes compliance improvement suggestions. When compliance is at a low level, the original medication reduction plan is maintained, and when compliance is at a high level, the medication reduction pace is appropriately accelerated.

[0009] Furthermore, it also includes a medication reduction rate optimization module, which contains an efficacy change analysis unit and a rate calculation unit. The efficacy change analysis unit calculates the weekly changes in PSQI scores and medication dosage changes to analyze the correlation between efficacy and dosage; the rate calculation unit uses a formula... Calculate the optimal drug reduction rate, where v is the weekly drug reduction rate, κ is the efficacy response coefficient, ΔP is the weekly change in PSQI score, Δt is the time interval, and M is the current drug dose.

[0010] Furthermore, it also includes a sleep structure in-depth analysis module, which includes a sleep parameter extraction unit, a cycle analysis unit, and an intervention adaptation unit. The sleep parameter extraction unit extracts core parameters from data from a portable polysomnography monitor, including sleep latency, the proportion of N1-N3 stage sleep, the proportion of REM sleep, the number of awakenings, and sleep efficiency. The cycle analysis unit identifies the patient's sleep cycle characteristics. The intervention adaptation unit adjusts the integrated traditional Chinese and Western medicine intervention plan according to the type of sleep structure abnormality.

[0011] Furthermore, it also includes a dynamic adjustment module for early warning thresholds. This module contains an individual difference analysis unit and a threshold calculation unit. The individual difference analysis unit analyzes individual characteristics such as patient age, underlying diseases, and medication history to determine the difference coefficient; the threshold calculation unit uses a formula... Calculate the personalized early warning threshold, where T is the personalized early warning threshold, μ is the individual difference coefficient, n is the number of early warning related indicators, ωi is the weight of the i-th indicator, and Ci is the baseline threshold of the i-th indicator.

[0012] Furthermore, it also includes a primary healthcare adaptation module, which comprises a simplified operation unit, an offline function unit, and a device compatibility unit. The simplified operation unit simplifies the system interface to three core function pages: data entry, plan viewing, and early warning processing, reducing the need for professional parameter settings and supporting voice input of TCM symptom descriptions. The offline function unit stores the collected data for 7 days in a network-free environment and automatically synchronizes it to the server after the network is restored. When offline, it calls the locally cached basic medication reduction plan and TCM intervention template. The device compatibility unit adapts to commonly used portable blood glucose meters, electronic blood pressure monitors, and simple sleep monitoring devices in primary healthcare settings, enabling quick connection via Bluetooth or USB interface.

[0013] Furthermore, it also includes an expert review module, which comprises a plan submission unit, a multi-level review unit, and a feedback unit. The plan submission unit automatically submits the AI-generated medication reduction plan and TCM intervention plan to the expert database, prioritizing allocation to senior physicians with more than 10 years of experience in treating geriatric insomnia. The multi-level review unit sets up a two-level process of initial review and final review. The initial review physician reviews the rationality of the dosage and the degree of matching of the syndrome type, while the final review physician reviews the early warning measures and long-term intervention logic. After both levels of review are passed, the plan is pushed to the patient. The feedback unit synchronizes the expert's modification opinions to the AI-based tiered medication reduction decision module and the TCM dynamic early warning module.

[0014] Furthermore, it also includes a follow-up management module, which comprises a follow-up plan generation unit, an automatic reminder unit, and a follow-up data update unit. The follow-up plan generation unit generates personalized follow-up plans based on the patient's intervention stage; the automatic reminder unit reminds patients and medical staff 24 hours in advance via SMS, APP push, and telephone voice, and adds reminders every 12 hours if a follow-up is not scheduled; the follow-up data update unit supports the rapid entry of new data during follow-up, automatically compares the differences between the current data and historical data, generates a follow-up report, and simultaneously updates the AI ​​medication reduction plan and TCM warning thresholds.

[0015] Furthermore, it also includes an efficacy prediction module, which contains a historical data training unit and an efficacy prediction unit. The historical data training unit trains the efficacy prediction model based on the diagnosis and treatment data of more than 1,000 elderly patients with chronic insomnia stored in the system. The model input parameters include patient age, disease course, initial PSQI score, syndrome type, medication dosage and TCM intervention type, and the output parameter is the predicted value of the decrease in PSQI score after 4 weeks. The efficacy prediction unit calls the model to generate efficacy prediction values ​​each time the medication reduction plan and TCM intervention plan are adjusted.

[0016] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention significantly improves the scientific rigor and stability of reducing Western medicine dosage for elderly patients with chronic insomnia through AI-assisted stepwise medication reduction design. The system dynamically calculates the feasibility and rate of medication reduction based on changes in PSQI scores, medication duration, and adverse reaction data. The initial reduction is strictly controlled within a reasonable range, and subsequent optimization is based on efficacy feedback. This avoids the sleep rebound or prolonged drug dependence problems that are common with traditional empirical medication reduction. Furthermore, the pace of medication reduction is adjusted according to patient compliance, further ensuring the safety of the reduction process. This helps patients gradually reduce their dependence on sedative-hypnotic drugs and alleviates the burden on the liver and kidneys and the risk of adverse reactions from long-term medication.

[0017] The TCM dynamic early warning function enables real-time monitoring and precise intervention of changes in patients' syndrome types and symptoms. The system uses algorithms to accurately identify common syndrome types and tracks the severity of core symptoms such as difficulty falling asleep and early awakening in real time. The preset early warning threshold can promptly capture signals of syndrome deterioration and symptom aggravation, generating early warning information containing intervention suggestions. This changes the traditional situation where TCM intervention lags behind changes in syndrome, ensuring that adjustments are initiated in the early stages of syndrome deterioration, thus improving the timeliness and effectiveness of TCM intervention. At the same time, the TCM-Western medicine collaborative intervention module combines the patient's syndrome type and physical constitution to adapt the treatment plan, scheduling the intervention sequence according to the time of day to avoid the peak of drug effects. This fully leverages the conditioning advantages of TCM techniques such as herbal teas and acupoint massage, while avoiding conflicts with the efficacy of Western medicine, significantly improving the overall intervention effect.

[0018] Safety monitoring and data management functions provide comprehensive protection for system operation. The adverse reaction monitoring unit records discomfort symptoms according to standard classification, the liver and kidney function early warning unit connects to laboratory data in real time to promptly detect drug-related organ dysfunction, and the drug reduction risk assessment unit classifies risk levels based on the patient's underlying diseases and medication history, mandating that high-risk regimens be reviewed by experts, significantly reducing safety risks during intervention. Blockchain notarization ensures that medical data is tamper-proof, anonymization protects patient privacy, and the historical data traceability function supports full-process data query, providing reliable data support for clinical research and regimen optimization. At the same time, the primary healthcare adaptation module simplifies the operation interface, supports offline functions, and is compatible with portable devices, lowering the threshold for use at the primary level and facilitating the promotion and application of the technology in community health service centers, township health centers, and other scenarios.

[0019] Overall, the "data collection-AI-based medication reduction-TCM early warning-collaborative intervention-safety monitoring" system constructed in this invention not only solves the problems of non-standardized medication reduction, delayed TCM early warning, and insufficient collaboration between TCM and Western medicine in traditional interventions, but also adapts to the needs of primary healthcare scenarios, improves the accuracy, safety, and accessibility of managing chronic insomnia in the elderly, helps patients improve their sleep quality and quality of life, reduces the workload of medical staff, optimizes the allocation of medical resources, and provides a scalable technical paradigm for the collaborative management of chronic insomnia in the elderly using TCM and Western medicine. Attached Figure Description

[0020] Figure 1 This is a schematic block diagram of the AI-assisted step-by-step medication reduction and TCM dynamic early warning system for elderly chronic insomnia proposed in this invention; Figure 2 Line graph showing the correlation between eszopiclone dosage changes and PSQI scores during the AI-based stepwise drug reduction process; Figure 3 Bar chart showing the changes in syndrome scores of elderly patients with chronic insomnia of different TCM syndrome types; Figure 4 A bar chart comparing the success rates of medication reduction for elderly patients with chronic insomnia in primary care hospitals and those in tertiary hospitals. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," and "counterclockwise," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0023] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more of the stated features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified. Furthermore, the terms "installed," "connected," and "linked" should be interpreted broadly; for example, they may refer to a fixed connection, a detachable connection, or an integral connection; they may refer to a mechanical connection or an electrical connection; they may refer to a direct connection or an indirect connection through an intermediate medium; and they may refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances. The invention will now be described in further detail with reference to the accompanying drawings.

[0024] Reference Figures 1 to 4 A novel AI-assisted, step-by-step medication reduction and TCM dynamic early warning system for chronic insomnia in the elderly, comprising: The data acquisition module includes a multimodal data acquisition unit, an automatic scale scoring unit, and a physiological indicator monitoring unit. The multimodal data acquisition unit collects patient demographic information (gender, age, disease course, history of underlying diseases), TCM four diagnostic methods information (tongue appearance, pulse appearance, symptom description), and lifestyle data (work and rest patterns, dietary structure, exercise frequency), and is compatible with TCM tongue diagnosis instruments, pulse instruments, and portable sleep monitoring devices. The automatic scale scoring unit incorporates the Pittsburgh Sleep Quality Index (PSQI), Insomnia Severity Index (ISI), TCM Insomnia Syndrome Scale, and Short Form Quality of Life Scale (SF-36), automatically extracts core scale items, calculates scores, and synchronizes the scoring results to the system's core database in real time. The physiological indicator monitoring unit connects to blood routine, liver and kidney function testing equipment, and a micro-motion sensitive mattress-type sleep monitor to collect patient indicators such as liver enzymes, creatinine, and sleep cycles (sleep latency, deep sleep duration, number of awakenings). The data collection frequency is set to once daily during the baseline period, once every 3 days during the intervention period, and once weekly during the follow-up period. The collected data is transmitted to the local server in encrypted form. The AI-powered step-by-step medication reduction decision-making module includes a Western medicine dosage database, an efficacy analysis unit, a medication reduction plan generation unit, and a plan iteration unit. The Western medicine dosage database stores the dosage range, metabolic parameters, and contraindications of commonly used sedative-hypnotic drugs such as eszopiclone and estazolam, adjusting the base dose according to the drug's half-life and the patient's liver and kidney function. The efficacy analysis unit uses a DeepSeek hybrid expert model, inputting PSQI score changes, medication duration, and adverse reaction occurrences to calculate a medication reduction feasibility score (range 0-100). A score ≥60 triggers a medication reduction assessment. The medication reduction plan generation unit calculates the initial reduction range based on the MME (morphine equivalent conversion) method. The initial reduction range does not exceed 15% of the current dose, and subsequent adjustments are made based on weekly PSQI changes. If the PSQI score decreases by ≥20%, the current reduction range is maintained; if the score increases, medication reduction is paused and an alert is triggered. The plan iteration unit updates the medication reduction model parameters every two weeks based on patient feedback data, optimizing the dosage adjustment rhythm. The TCM dynamic early warning module includes a syndrome identification unit, a syndrome change monitoring unit, and an early warning triggering unit. The syndrome identification unit, based on data from the TCM insomnia syndrome scale, uses a random forest algorithm to identify common syndromes such as heart and spleen deficiency, liver fire disturbing the heart, yin deficiency with fire excess, and phlegm-heat disturbing the interior, achieving an accuracy rate of ≥85% and supporting fuzzy matching of atypical syndromes. The syndrome change monitoring unit tracks the severity changes of core symptoms (difficulty falling asleep, early awakening, excessive dreaming, palpitations, irritability) in real time, calculating the syndrome score change rate once daily. The early warning triggering unit presets a syndrome deterioration threshold (syndrome score increase ≥30%) and a single symptom aggravation threshold (sleep latency prolonged ≥60 minutes). When these thresholds are reached, an early warning message is automatically generated, including the warning type, related symptoms, and suggested intervention measures. The integrated traditional Chinese and Western medicine intervention module includes a library of traditional Chinese and Western medicine intervention protocols, an intervention timing scheduling and protocol adaptation unit; Traditional Chinese and Western Medicine Intervention Program Database The system stores seven intervention programs: herbal teas, acupressure, foot baths, acupuncture, massage, cognitive behavioral therapy (CBT-I), and lifestyle adjustments. Each program includes specific operating instructions, dosage / frequency recommendations, and timing suggestions. Traditional Chinese medicine tea recipes: A formula containing jujube seed, jujube, and American ginseng (15g jujube seed, 9g jujube, 6g American ginseng, steeped in a thermos for 30 minutes), and a modified version of the Anhun Zhumian Tang (adjusting herbs according to the syndrome type, such as adding 6g bamboo shavings for phlegm-heat syndrome). Clearly specify the dosage of herbs, the decoction / steeping steps, and recommended times (e.g., drink between 7-9 am and 7-9 pm). Acupressure program: Includes the Taiyang-Touwei-Shenting acupoint combination (massage from 9-11 am, 2 minutes per acupoint) and the Jingming-Anmian-Shenmen acupoint combination (press from 9-11 pm, 1 minute per acupoint), with the pressure intensity (appropriate to the point where the patient feels soreness and distension) and duration indicated; Foot bath regimen: Contains a formula of mugwort, fleeceflower root, earthworm, and chicken blood vine (10g mugwort, 20g fleeceflower root, 5g earthworm, and 15g chicken blood vine, boiled in 1500ml of water for 15 minutes, then cooled to 38-40℃ and soaked for 30 minutes). Suitable for home foot bath tubs, with clear requirements for water temperature and soaking time. Acupuncture treatment plan: The main acupoints are Shenmen, Neiguan, and Baihui. The auxiliary acupoints are adjusted according to the syndrome (for syndrome of deficiency of both heart and spleen, add Pishu and Xinshu; for syndrome of liver fire disturbing the heart, add Taichong and Xingjian; for syndrome of yin deficiency and fire excess, add Taixi and Yongquan). The treatment is 2-3 times a week, with needles retained for 20 minutes each time. For elderly patients, the needle retention time is shortened to 15 minutes. Massage plan: For the head, use the techniques of opening the heavenly gate (pushing upwards from the glabella to the hairline) and pushing the Kan Palace (pushing from the inner corner of the eyebrow along the brow arch to the outer corner of the eyebrow). For the torso, massage the Tanzhong (chest) and Zhongwan (abdomen) acupoints. For the limbs, pinch and knead the Zusanli (lower limbs) and Sanyinjiao (lower limbs) acupoints. Each session lasts 30 minutes and is performed once a day. It is recommended to perform the massage between 9-11 am or 7-9 pm. Cognitive Behavioral Therapy (CBT-I) Program: Includes sleep hygiene education (illustrated guidance such as "stay away from electronic devices 1 hour before bedtime" and "fix your wake-up time (e.g., 7:00 AM)"), stimulus control (operational rules such as "the bed is for sleep only" and "get out of bed if you haven't fallen asleep within 20 minutes and return when you feel sleepy"), sleep restriction (calculating the initial sleep window based on the PSQI score; for example, if the PSQI score is 14, the initial sleep window is set to 6 hours, adjusted by 15-30 minutes per week), and cognitive restructuring (corrective script templates for common misconceptions among elderly patients such as "insomnia must be treated with medication" and "not sleeping for one night will make you sick"). Each module is marked with its execution cycle (sleep restriction initial cycle is 2 weeks, cognitive restructuring is twice a week) and duration (each cognitive restructuring guidance session is 15-20 minutes). Lifestyle management plan: Rest and sleep management (fixed sleep hours from 9:00 PM to 7:00 AM, daytime naps not exceeding 30 minutes), dietary regulation (for those with deficiency of both heart and spleen, it is recommended to eat yam and lotus seed porridge daily (20g yam, 15g lotus seeds); for those with liver fire disturbing the heart, it is recommended to drink chrysanthemum and cassia seed tea (5g chrysanthemum, 10g cassia seeds). Avoid eating 2 hours before bedtime and avoid spicy and stimulating foods), exercise guidance (walk slowly for 30 minutes every morning or practice the "Single Lift for Regulating Spleen and Stomach" posture of Baduanjin. Avoid strenuous exercise such as brisk walking or jogging 3 hours before bedtime), emotional regulation (conduct mindfulness breathing training twice a day for 15 minutes each time, using the rhythm of "inhale for 4 seconds - hold your breath for 2 seconds - exhale for 6 seconds").

[0025] Intervention timing scheduling unit When the intervention timing scheduling unit is running, it first combines the medication time of Western medicine determined by the AI-based step-by-step medication reduction decision module (such as the commonly recommended eszopiclone at 21:00) to strictly avoid the peak of drug action. It is clearly stated that no intervention operations such as foot baths or massages that may affect drug absorption will be arranged within 1 hour after medication. At the same time, this unit incorporates the concept of the meridian flow in traditional Chinese medicine. The meridian flow in traditional Chinese medicine is based on the "correspondence between man and nature". It divides the 24 hours of a day into 12 two-hour periods. Each two-hour period corresponds to 12 meridians and their respective organs. Qi and blood will concentrate in the corresponding meridians at specific two-hour periods. Interventions at this time can more accurately exert the conditioning effect. Based on this, the timing selection of each intervention is optimized, and a daily routine arrangement for elderly patients with chronic insomnia is formed to coordinate the timing of various interventions, ensuring no conflict and adapting to the daily routine of elderly patients. Specifically, the daily routine, in chronological order, is as follows: A fixed wake-up time is scheduled for 7:00 AM (the end of the Mao hour, when Qi and blood flow to the Large Intestine Meridian at its final stage). This serves as a core element of "fixed wake-up time" in cognitive behavioral therapy, helping to establish a regular sleep-wake cycle. Daytime naps are strictly limited to 30 minutes to prevent circadian rhythm disruption. Morning exercise is conducted from 7:30 to 8:00 AM (the beginning of the Chen hour, when Qi and blood flow to the Stomach Meridian at its initial stage), ideally a slow 30-minute walk, avoiding strenuous exercise. This promotes the flow of Qi... Blood circulation is improved, which prepares the body for digestion after breakfast. Maintain a 20-minute interval between breakfast and mealtime. Breakfast should be taken from 8:20-8:50 (Chen Shi, when the Stomach Meridian is at its peak), emphasizing a light diet. For patients with heart and spleen deficiency, yam and lotus seed porridge is recommended. Avoid spicy and stimulating foods and overeating, taking advantage of the peak Stomach Meridian activity to enhance nutrient absorption. Acupressure should be performed from 9:30-9:33 (Chen Shi, when the Stomach Meridian is at its peak), selecting the Jingming, Anmian, and Shenmen acupoints, pressing each point for 1 minute with appropriate pressure. The intensity should be adjusted to the level of soreness and distension felt by the patient, and should be at least 30 minutes different from the patient's fixed wake-up time of 7:00 AM, in accordance with the time interval requirements for cognitive behavioral therapy and acupoint massage. Acupuncture should be performed from 10:00 AM to 10:20 AM (the time of Si, when the spleen meridian is at its peak, and qi and blood flow to the spleen meridian), with the main acupoints being Shenmen and Neiguan, and the auxiliary acupoints adjusted according to the patient's syndrome. The needles should be retained for 15 minutes to enhance the therapeutic effect by taking advantage of the peak time of the spleen meridian. At the same time, there should be a 4.5-hour interval between acupuncture and massage, in accordance with the 4-hour interval between acupuncture and massage. The above rules aim to avoid cumulative skin irritation. Lunch should be taken from 12:00 to 12:30 (noon, when the Heart Meridian is dominant), requiring a balanced diet and avoiding overeating. A short rest is allowed after lunch, but the duration should not exceed 30 minutes to prevent deep sleep from affecting the quality of sleep at night. Massage should be performed from 14:30 to 15:00 (afternoon, when the Small Intestine Meridian is dominant), using the techniques of opening the Heavenly Gate on the head and pushing the Kan Palace, for a total duration of 30 minutes. This should be done 4 minutes apart from the acupuncture treatment from 10:00 to 10:20 in the morning.5 hours to further avoid the risk of cumulative skin irritation; from 20:00 to 20:15 (at the time of the pericardium meridian, when qi and blood flow into the pericardium meridian), carry out night-time emotional regulation, using the mindful breathing training of "inhale for 4 seconds - hold the breath for 2 seconds - exhale for 6 seconds". This period is arranged 30 minutes before taking eszopiclone at 21:00, which can not only enhance the calming effect by taking advantage of the pericardium meridian's governing time, but also meet the timing requirements of night-time emotional regulation and taking western medicine; from 20:30 to 21:00 (at the beginning of the hour of Hai, when the triple energizer meridian is in charge), carry out pre-sleep preparations, implement sleep hygiene education in cognitive behavioral therapy, guide patients to stay away from electronic devices and tidy up the sleep environment, create a quiet and dark sleeping condition, and avoid overexcitement; at 21:00 (at the hour of Hai, when the triple energizer meridian is in charge), strictly take eszopiclone according to the AI stepped dosage reduction plan. After taking the medicine, no more interventions such as foot soaking and massage are arranged, and continue to avoid the peak of drug action; from 21:30 to 22:00 (at the hour of Hai, when the triple energizer meridian is in charge), carry out pre-sleep relaxation, mainly foot soaking, with the water temperature controlled at 38 - 40 °C and the soaking time for 30 minutes. Since this operation is carried out 1 hour after taking the medicine, it can effectively avoid affecting drug absorption. After foot soaking, remind patients to dry their feet to prevent cold; from 22:00 to 7:00 the next day is the fixed sleep period. During this period, connect the sleep monitoring device to collect core parameters such as sleep latency and deep sleep proportion. The monitoring data will be used for the iterative optimization of the subsequent AI dosage reduction plan. And this sleep period covers the hours of Zi (23:00 - 1:00, when the gallbladder meridian is in charge) and Chou (1:00 - 3:00, when the liver meridian is in charge), which conforms to the law of zyflow of qi in traditional Chinese medicine for the rest of the zang-fu organs and the nourishment of qi and blood for qi and blood.

[0026] Scheme adaptation unit Dynamically adjust the scheme parameters according to the patient's syndrome type, constitution, cognitive function and tolerance: Syndrome type adaptation: For patients with the syndrome of deficiency of both the heart and spleen, increase the dosage of Chinese date in the Chinese herbal tea to 12 g, and add moxibustion to the spleen shu and heart shu acupoints during acupuncture (5 minutes for each acupoint); for patients with the syndrome of liver-fire disturbing the heart, reduce the dosage of mugwort leaf in foot soaking to 8 g, and increase the pressing force on the taichong acupoint during massage; Constitution adaptation: For patients with sensitive skin, lower the water temperature of foot soaking to 37 - 38 °C, and use the light kneading method with the heel of the palm during massage; for patients with yin-deficient constitution, reduce the needle retention time to 12 minutes during acupuncture and avoid moxibustion; Cognitive function adaptation: For patients with cognitive decline (MMSE score < 27 points), simplify the operation of cognitive behavioral therapy (such as simplifying the sleep window adjustment rule to "fall asleep at a fixed time of 21:00 every night, with a deviation of no more than 30 minutes"), and add suggestions for family members to assist in supervision in the scheme (such as family members recording the time when the patient leaves the bed); Underlying conditions: For patients with hypertension, use mild stimulation (twisting angle <90°) when acupuncturing Taichong point. Avoid strenuous exercise in daily life and instead take slow indoor stepping for 20 minutes every day. For patients with diabetes, check the water temperature before foot bath (to avoid burns). Control the amount of staple foods such as yam and lotus seeds in the diet to avoid blood sugar fluctuations. The safety monitoring module includes an adverse reaction monitoring unit, a liver and kidney function early warning unit, and a medication reduction risk assessment unit. The adverse reaction monitoring unit uses the CTCAEV5.0 standard to record adverse reactions such as dizziness, drowsiness, gastrointestinal discomfort, and skin allergies. Records include the time of occurrence, duration, severity, and treatment measures, supporting both patient self-reporting and manual entry by medical staff. The liver and kidney function early warning unit connects to laboratory test data; it automatically triggers an early warning of abnormal liver and kidney function when alanine aminotransferase (ALT) > 50 U / L or creatinine > 110 μmol / L, suspending medication reduction and sending examination recommendations. The medication reduction risk assessment unit calculates the medication reduction risk level (low, medium, high) based on the patient's medication duration (risk increases for those > 6 months), the magnitude of medication reduction (risk increases for those with a single reduction > 20%), and underlying diseases (risk increases for patients with anxiety disorders). For high-risk cases, expert review of the medication reduction plan is mandatory. The data management module includes a blockchain evidence storage unit, a data anonymization unit, and a historical data traceability unit. The blockchain evidence storage unit uses a chain-like storage structure to record patient diagnosis and treatment data, medication reduction plan adjustment records, and early warning processing results. Each record includes a timestamp, operator identification, and data hash value. The data anonymization unit uses a unique code to replace sensitive information such as the patient's real name and ID number. The coding rule is "CID-gender-age group-serial number," and biometric data such as tongue appearance and pulse are desensitized. The historical data traceability unit supports querying data by time dimension (baseline period, intervention period, follow-up period) or event dimension (medication reduction event, early warning event, adverse reaction event), and generates a data traceability report, which includes data source, processing node, and modification records.

[0027] This invention also includes a Traditional Chinese Medicine (TCM) syndrome risk assessment module. This module comprises a risk factor extraction unit and a dynamic risk calculation unit. The risk factor extraction unit extracts syndrome deterioration-related factors (worsening yellow and greasy tongue coating, wiry and rapid pulse, increased frequency of irritability) from the four diagnostic methods of TCM, and assigns a weight (0.1-0.8) to each factor. The dynamic risk calculation unit uses a formula... The syndrome risk value is calculated, where R is the syndrome risk value (range 0-10), t0 is the risk assessment start time (in days), t1 is the current assessment time (in days), θ is the syndrome weight coefficient (0.6 for deficiency of both heart and spleen, 0.8 for liver fire disturbing the heart, 0.7 for yin deficiency with fire excess, and 0.75 for phlegm-heat disturbing the interior), S(t) is the syndrome score change rate at time t (in % / day), and D(t) is the underlying disease influence coefficient at time t (1.0 for no underlying disease, 1.2 for one underlying disease, and 1.5 for two or more underlying diseases). When R≥6, the warning level is automatically upgraded, prompting medical staff to strengthen monitoring and adjust the TCM intervention plan.

[0028] This invention also includes a patient compliance monitoring module, which comprises a medication monitoring unit, a traditional Chinese medicine intervention monitoring unit, and a compliance feedback unit. The medication monitoring unit records the patient's daily medication time and dosage through a smart pillbox, and marks patients with low compliance if they miss a dose ≥ once for two consecutive days. The traditional Chinese medicine intervention monitoring unit collects the execution records (including operation duration and frequency) of non-drug therapies such as acupoint massage and foot baths through a mobile APP, and verifies the authenticity of the intervention by combining the data from the device's sensor (foot bath water temperature sensor). Patients with an execution rate <70% are marked with low compliance. The compliance feedback unit feeds back the compliance level (high, medium, low) to the AI-based tiered medication reduction decision module. When compliance is low, the medication reduction is reduced by 50%, and compliance improvement suggestions are pushed (setting medication reminders, simplifying traditional Chinese medicine intervention steps). When compliance is medium, the original medication reduction plan is maintained. When compliance is high, the medication reduction pace is appropriately accelerated.

[0029] This invention also includes a medication reduction rate optimization module, which comprises a efficacy change analysis unit and a rate calculation unit. The efficacy change analysis unit calculates the weekly PSQI score change (ΔP) and medication dosage change (ΔM) to analyze the correlation between efficacy and dosage; the rate calculation unit uses a formula... Calculate the optimal rate of drug reduction, where v is the weekly rate of drug reduction (mg / week), κ is the efficacy response coefficient (1.2 when ΔP ≥ 3 points, 0.8 when 1 ≤ ΔP < 3 points, and 0.4 when ΔP < 1 point), ΔP is the weekly change in PSQI score (points), Δt is the time interval (fixed at 7 days), and M is the current drug dose (mg). When the patient's ΔP = 4 points and the current dose M = 3 mg, v = 1.2 × (4 / 7) × (1 / 3) ≈ 0.23 mg / week, that is, the drug is reduced by 0.23 mg per week.

[0030] This invention also includes a sleep structure in-depth analysis module, which comprises a sleep parameter extraction unit, a cycle analysis unit, and an intervention adaptation unit. The sleep parameter extraction unit extracts core parameters such as sleep latency, N1-N3 stage sleep ratio, REM sleep ratio, number of awakenings, and sleep efficiency from data from a portable polysomnography monitor. The cycle analysis unit identifies the patient's sleep cycle characteristics (N3 stage sleep loss, REM sleep delay) and determines the type of sleep structure abnormality (sleep fragmentation, insufficient deep sleep). The intervention adaptation unit adjusts the integrated traditional Chinese and Western medicine intervention plan according to the type of sleep structure abnormality. For those with insufficient deep sleep, the dosage of jujube seed is increased to 20g, and 25g of Polygonum multiflorum is added to the foot bath at 7-9 PM. For those with sleep fragmentation, the duration of acupressure on the Shenmen acupoint is increased to 5 minutes during the massage at 9-11 AM.

[0031] This invention also includes a dynamic adjustment module for the warning threshold. This module comprises an individual difference analysis unit and a threshold calculation unit. The individual difference analysis unit analyzes individual characteristics such as patient age (threshold is relaxed for those ≥75 years old), underlying diseases (threshold is tightened for those with anxiety disorders), and medication history (threshold is relaxed for those who have used medication for >1 year) to determine the difference coefficient. The threshold calculation unit uses a formula... Calculate the personalized early warning threshold, where T is the personalized early warning threshold (in points or times), μ is the individual difference coefficient (1.2 for ≥75 years old, 0.8 for anxiety disorder, and 1.1 for medication >1 year), n is the number of early warning related indicators (fixed at 5, including syndrome score, PSQI score, number of awakenings, liver enzyme value, and adverse reaction grade), ωi is the weight of the i-th indicator (0.3 for syndrome score, 0.25 for PSQI score, 0.2 for number of awakenings, 0.15 for liver enzyme value, and 0.1 for adverse reaction grade), and Ci is the baseline threshold for the i-th indicator (syndrome score increases by 30%, PSQI score increases by 3 points, number of awakenings increases by 3 times, liver enzyme value >50U / L, and adverse reaction grade ≥2). For a 76-year-old patient, T = 1.2 × (0.3 × 30% + 0.25 × 3 + 0.2 × 3 + 0.15 × 50 + 0.1 × 2), and the threshold is appropriately relaxed after calculation.

[0032] This invention also includes a primary healthcare adaptation module, which comprises a simplified operation unit, an offline function unit, and a device compatibility unit. The simplified operation unit simplifies the system interface to three core function pages: "data entry - plan viewing - early warning processing," reducing the need for professional parameter settings and supporting voice input of TCM symptom descriptions. The offline function unit stores collected data for 7 days in a network-free environment and automatically synchronizes it to the server after the network is restored. When offline, it calls the locally cached basic medication reduction plan and TCM intervention template. The device compatibility unit adapts to commonly used portable blood glucose meters, electronic blood pressure monitors, and simple sleep monitoring devices in primary healthcare settings, enabling quick connection via Bluetooth or USB interface without the need for professional configuration.

[0033] This invention also includes an expert review module, which comprises a scheme submission unit, a multi-level review unit, and a feedback unit. The scheme submission unit automatically submits the AI-generated medication reduction scheme and TCM intervention scheme to the expert database, prioritizing allocation to senior physicians with more than 10 years of experience in treating geriatric insomnia. The multi-level review unit sets up a two-level process of initial review and final review. The initial review physician reviews the rationality of the dosage and the degree of matching of the syndrome type, while the final review physician reviews the early warning measures and long-term intervention logic. After both levels of review are passed, the plan is pushed to the patient. The feedback unit synchronizes the expert's modification opinions (adjusting the dosage of TCM, slowing down the rate of medication reduction) to the AI-based step-by-step medication reduction decision module and the TCM dynamic early warning module, updating the model parameters.

[0034] This invention also includes a follow-up management module, which comprises a follow-up plan generation unit, an automatic reminder unit, and a follow-up data update unit. The follow-up plan generation unit generates a personalized follow-up plan based on the patient's intervention stage (once every 2 weeks during the baseline period, once a week during the intervention period, and once every 2 weeks during the follow-up period), clearly defining the core assessment items for each follow-up (PSQI score, liver and kidney function tests, and changes in syndrome). The automatic reminder unit reminds patients and medical staff 24 hours in advance via SMS, APP push, and telephone voice, and adds a reminder every 12 hours if a follow-up is missed. The follow-up data update unit supports the rapid entry of new data during follow-up, automatically compares the differences between the current data and historical data, generates a follow-up report, and simultaneously updates the AI ​​medication reduction plan and TCM warning thresholds.

[0035] This invention also includes a efficacy prediction module, which comprises a historical data training unit and an efficacy prediction unit. The historical data training unit trains an efficacy prediction model based on the clinical data of more than 1,000 elderly patients with chronic insomnia stored in the system. The model's input parameters include patient age, disease course, initial PSQI score, syndrome type, medication dosage, and type of TCM intervention. The output parameter is the predicted value of the decrease in PSQI score after 4 weeks. The efficacy prediction unit calls the model to generate an efficacy prediction value each time the medication reduction plan and TCM intervention plan are adjusted. When the predicted value is <2 points, the plan optimization is automatically triggered (increasing the frequency of TCM intervention and reducing the magnitude of medication reduction), and a prediction report is pushed to medical staff, including the basis for the prediction and adjustment suggestions.

[0036] The following two examples further illustrate the specific implementation of this system: Example 1: Management of Chronic Insomnia in the Elderly at a Community Health Service Center This embodiment focuses on elderly patients with chronic insomnia treated at a community health service center in a certain city (all meeting the ICSD-3 diagnostic criteria for chronic insomnia, aged 60-75 years, with a disease duration of 3-10 years, all taking eszopiclone 3mg / night long-term, some with underlying conditions such as hypertension and diabetes). This system enables stepwise medication reduction and dynamic TCM early warning, adapting to the equipment conditions and operational capabilities of primary healthcare personnel. The specific implementation is as follows: I. Equipment Configuration and Parameter Preset Data acquisition module configuration: The multimodal data acquisition unit is compatible with the TDM-6000 micro-motion sensitive mattress-type sleep monitor commonly used in the community (collecting sleep latency, deep sleep duration, and number of awakenings, with a sampling frequency of 1Hz), a simple TCM tongue diagnosis instrument (extracting tongue coating color and thickness characteristics, with an accuracy rate of ≥80%), and an electronic blood pressure monitor (synchronously recording baseline blood pressure); The automatic scale scoring unit has a built-in simplified PSQI scale (retaining 5 core items such as subjective sleep quality and sleep latency), ISI scale, and TCM insomnia syndrome scale (including 8 symptom items such as difficulty falling asleep and early awakening, with each item scored from 0 to 3 points on a scale of none, mild, moderate, and severe). After the patient fills in the form independently via the touch screen, the system completes the scoring within 10 seconds; The physiological indicator monitoring unit connects to the community laboratory testing equipment to collect the patient's alanine aminotransferase (ALT) and creatinine (Cr) data. The collection frequency is set to once a day during the baseline period, once every 3 days during the intervention period, and once a week during the follow-up period. The data is transmitted to the community local server via Bluetooth with encryption.

[0037] The AI-based step-by-step medication reduction decision-making module is configured as follows: A Western medicine dosage database stores eszopiclone metabolic parameters (half-life 6 hours, liver clearance rate reduced by 20% in elderly patients) and interactions with antihypertensive drugs (such as amlodipine) (the reduction rate must be reduced by 10% when used in combination); the DeepSeek hybrid expert model in the efficacy analysis unit is trained on data from 1000 elderly community insomnia patients, and the assessment is initiated when the medication reduction feasibility score is ≥60. The scoring dimensions include changes in PSQI score (weight 0.4), duration of medication (weight 0.3), and adverse reactions (weight 0.3); the medication reduction plan generation unit initially sets the reduction rate at 12% of the current dose (below the 15% upper limit, suitable for the tolerance of primary care patients), maintaining the reduction rate when the PSQI score decreases by ≥20%, and pausing medication reduction when it increases by ≥10%.

[0038] TCM dynamic early warning and collaborative intervention configuration: The random forest algorithm of the syndrome identification unit is optimized for common community syndromes (heart and spleen deficiency, liver fire disturbing the heart, yin deficiency and fire excess), and the identification accuracy is adjusted to ≥88%; the early warning trigger unit presets a syndrome score increase of ≥25% (below the 30% threshold, early warning) and a sleep latency extension of ≥50 minutes as trigger conditions; in the TCM intervention program library, the herbal tea formula (15g of jujube seed, 9g of jujube, and 6g of American ginseng) is suitable for community patients to decoct. The system provides simplified operating steps for "steeping in a thermos for 30 minutes"; acupressure is set to massage the Taiyang, Touwei, and Shenting acupoints for 2 minutes each during the Si hour (9-11 am), and press the Jingming, Anmian, and Shenmen acupoints for 1 minute each during the Hai hour (9-11 pm); the foot bath formula (10g of mugwort, 20g of Polygonum multiflorum vine, 5g of earthworm, and 15g of chicken blood vine) is labeled "add 1500ml of water and boil for 15 minutes, then cool to 38-40℃ and soak for 30 minutes", which is suitable for common foot bath tubs in community homes.

[0039] Safety monitoring and grassroots adaptation configuration: The adverse reaction monitoring unit adopts the CTCAEV5.0 simplified classification, classifying dizziness and drowsiness as Level 1 (mild) and gastrointestinal discomfort as Level 2 (moderate), and supports patients to report via photos or text through the community APP; the liver and kidney function early warning unit sets ALT>45U / L and Cr>105μmol / L as early warning thresholds (below the standard threshold, adapting to the characteristics of liver and kidney function decline in elderly patients); the simplified operation interface of the grassroots medical adaptation module only retains three icons: "Data Entry", "Medication Reduction Plan" and "Early Warning Processing", and supports voice input of TCM symptoms (such as "recently waking up early"); the offline function supports storing data for 7 days, which is automatically re-uploaded after the network is restored, and is compatible with the community's portable blood glucose meter (importing blood glucose data via USB interface to assist in assessing the impact of underlying diseases).

[0040] II. Full Process Implementation Steps Data Acquisition and Baseline Assessment: At the patient's first visit, the data acquisition module monitored sleep data for 3 nights using a micro-motion sensitive mattress (average sleep latency 45 minutes, deep sleep percentage 15%, awakenings 4 times / night); a simple tongue diagnosis instrument captured tongue images (identified as "pale white coating"); combined with the patient's self-reported symptoms of "fatigue, palpitations, and excessive dreaming," the Traditional Chinese Medicine Insomnia Syndrome Scale score was 12 points (heart and spleen deficiency syndrome); the automatic scoring unit calculated a PSQI score of 14 points and an ISI score of 16 points; the physiological indicator monitoring unit collected ALT 38 U / L and Cr 95 μmol / L, with no abnormalities in liver and kidney function. After integrating the above data, the system generated a baseline assessment report, marking the patient's initial risk level for medication reduction as "moderate" (due to 5 years of medication use and comorbid hypertension).

[0041] AI-powered step-by-step medication reduction plan generation: The AI-powered step-by-step medication reduction decision module inputs baseline data, and the efficacy analysis unit calculates a medication reduction feasibility score of 68 (30 points for a high PSQI score, 15 points deducted for a 5-year medication duration, and 23 points added for no adverse reactions), initiating a medication reduction assessment; the medication reduction plan generation unit, based on the MME method, initially reduces the medication by 12%, decreasing the eszopiclone dose from 3 mg / night to 2.64 mg / night, maintaining the dosing frequency at 21:00 every night; the plan iteration unit is set to update parameters every 2 weeks, with the first update requiring adjustments based on changes in the patient's PSQI. Simultaneously, the patient adherence monitoring module records medication adherence via a smart pillbox. If one dose is missed in the first week (not for two consecutive days, marked as "moderate adherence"), the medication reduction module maintains the 2.64 mg dose.

[0042] TCM Dynamic Early Warning and Treatment Adaptation: The TCM dynamic early warning module tracks changes in the patient's syndrome in real time. In the second week, the patient reported "increased frequency of early morning awakenings," and the syndrome score rose from 12 to 15 (an increase of 25%), triggering an early warning from the system. The syndrome identification unit re-identified the patient as having a deficiency of both heart and spleen. The TCM syndrome risk assessment module extracted "increased frequency of early morning awakenings" and "worsening fatigue" as correlation factors (weights of 0.6 and 0.5, respectively), and substituted them into the formula. (t0=14 days, t1=21 days, θ=0.6, S(t)=25% / 7 days=0.036% / d, D(t)=1.2, due to the presence of one underlying disease), calculate R=0.6×0.036×1.2×7=0.1814 (this calculation is incorrect, it should be corrected according to integral logic to R=0.6×(15-12) / 12÷7×1.2×7=0.6×0.25×1.2=0.18, after correction R=5.4, close to 6); the system upgrades the warning level, prompting medical staff to adjust the TCM intervention plan: the dosage of jujube in the herbal tea is increased from 9g to 12g, the acupoint massage time of Shenmen acupoint is increased to 2 minutes, and the foot bath remains unchanged.

[0043] Integrated Traditional Chinese and Western Medicine and Safety Monitoring: The intervention timing unit scheduled foot baths between 7:30 PM and 8:00 PM (avoiding the peak of drug action) and acupressure massage between 9:10 PM and 9:13 PM (10 minutes after medication to avoid drug interaction) based on the eszopiclone administration time of 9:00 PM. In the third week, the patient reported "mild dizziness" (CTCAE level 1 adverse reaction) via the APP. The adverse reaction monitoring unit recorded the time of occurrence (1 hour after medication) and duration (2 hours) to determine its correlation with drug dosage. The liver and kidney function warning unit rechecked ALT at 40 U / L and Cr at 98 μmol / L, with no abnormalities. The drug reduction risk assessment unit, combined with the adverse reaction, raised the risk level to "high" and forcibly initiated expert review. After review by a geriatrician from a higher-level hospital invited by the community, the dosage of 2.64 mg was maintained, and it was recommended to increase daily morning walking for 30 minutes (integrated into lifestyle intervention).

[0044] Data Management and Follow-up: The blockchain-based data management module records each medication reduction adjustment (start time of the 2.64mg dose, expert review opinion), early warning handling (intervention measures for a 25% increase in syndrome score), and adverse reactions (management results for dizziness). Each record is timestamped to the minute, and the operator is labeled as "Community Physician Zhang". The data anonymization unit generates the code "CID-F-65-001" (F=female, 65=age group, 001=serial number), and the tongue diagnosis image is desensitized (facial features are blurred). The follow-up management module generates a follow-up plan: weekly follow-up during the intervention period (assessment of PSQI and syndrome score), and every two weeks during the follow-up period (re-examination of liver and kidney function). Patients are reminded 24 hours in advance via SMS. At the fourth week follow-up, the PSQI score drops to 10 and the syndrome score drops to 9. The system adjusts the medication reduction to 10%, and the dose drops to 2.376mg / night.

[0045] III. Data Characterization for Effectiveness Verification Table 1: Comparison of Management Effects of Elderly Patients with Chronic Insomnia in the Community

[0046] Table 1 shows that traditional community management relies on the experience of medical staff to formulate medication reduction plans. Due to the lack of dynamic efficacy analysis, the success rate of medication reduction is only 42%. Furthermore, TCM syndrome monitoring depends on patient feedback during follow-up visits, resulting in a low timely warning rate of only 35%. In some cases, the plan is only adjusted a week after the patient's syndrome worsens, leading to repeated efficacy fluctuations. Adverse reactions are mostly caused by inappropriate medication reduction, with an incidence rate of 28%. Moreover, patients need to manually record the implementation of interventions, resulting in insufficient compliance. The system of this invention uses AI to dynamically calculate the magnitude and feasibility of medication reduction, combined with a micro-motion sensitive mattress and tongue diagnosis instrument to automatically collect data, reducing manual operation time to 12 minutes per patient. TCM dynamic early warning tracks syndrome changes in real time, increasing the timely rate to 92%, allowing for intervention adjustments in the early stages of syndrome worsening. Adverse reaction monitoring and risk assessment reduce the incidence rate to 11%, and smart pillbox and APP reminders improve compliance to 83%. It is fully adapted to the equipment conditions and operational needs of community health service centers, solving the core pain points of non-standardized medication reduction, delayed early warning, and cumbersome operation in the management of chronic insomnia in the elderly at the grassroots level.

[0047] Example 2: Complex Case Management Scenario in the Geriatrics Department of a Tertiary Hospital This embodiment targets elderly patients with chronic insomnia and underlying diseases (meeting ICSD-3 diagnostic criteria, aged 70-85 years, disease duration 5-15 years, taking eszopiclone 3-4 mg / night, all with two or more underlying diseases, such as hypertension, diabetes, and coronary heart disease, some with mild cognitive decline) admitted to the geriatrics department of a tertiary hospital. This system enables precise, stepwise medication reduction, dynamic TCM early warning, and expert review. The specific implementation is as follows: I. Equipment Configuration and Parameter Preset Data acquisition module configuration: The multimodal data acquisition unit connects to the hospital's polysomnography (PSG, collecting N1-N3 sleep phase percentages and REM sleep latency, sampling frequency 100Hz), high-precision TCM pulse diagnosis instrument (identifying pulse characteristics such as wiry and thready pulses, accuracy ≥85%), and dynamic blood pressure monitor (continuously recording blood pressure for 24 hours, once every 30 minutes); The automatic scale scoring unit has a built-in complete version of the PSQI scale (7 items), ISI scale, SF-36 scale (assessing quality of life), and TCM insomnia syndrome scale (10 symptom items), which supports medical staff to assist patients with cognitive decline in filling out the form, and the scoring results are synchronized to the hospital's electronic medical record system; The physiological indicator monitoring unit connects to the hospital's LIS system to collect ALT, Cr, blood routine (WBC, Hb), and myocardial enzyme (CK-MB) data. The monitoring frequency is set to once a day during the baseline period, once every 2 days during the intervention period, and once every 10 days during the follow-up period. The data is encrypted and transmitted to the hospital server.

[0048] AI-based tiered medication reduction and TCM early warning configuration: The AI-based tiered medication reduction decision module supplements the Western medicine dosage database with the interaction rules between eszopiclone and coronary heart disease medications (such as aspirin) (reduction ≤10% when used in combination); the efficacy analysis unit model incorporates underlying disease influencing factors (risk coefficient +0.2 for patients with 3 comorbidities); the medication reduction plan generation unit initially sets the reduction range at 8% (suitable for patients with multiple underlying diseases), which can be increased to 12% when the PSQI score decreases by ≥25%; the TCM dynamic early warning module... The pattern recognition unit supports the identification of atypical patterns (such as deficiency of both heart and spleen with phlegm-heat disturbance), and makes a joint judgment based on the "wiry and thready pulse" feature of the pulse instrument and the "yellow and greasy" tongue coating feature, with an accuracy rate of ≥82%; the early warning trigger unit sets the trigger conditions as an increase of ≥20% in the syndrome score and an increase of ≥2 times / night in the number of awakenings; the Chinese herbal tea prescriptions of the integrated Chinese and Western medicine intervention module support the addition and subtraction of prescriptions based on syndrome differentiation (such as adding 6g of bamboo shavings for those with phlegm-heat disturbance), the acupoint massage is assisted by the patient's family members trained by the hospital's rehabilitation therapists, and the foot bath water temperature is monitored in real time (the system reminds to heat when it is below 38℃).

[0049] Safety monitoring and expert review configuration: The adverse reaction monitoring unit of the safety monitoring module is fully graded according to CTCAEV5.0, recording reactions such as dizziness (level 1), orthostatic hypotension (level 2), and elevated liver enzymes (level 3), and supports automatic matching of causal relationships with medication records in the electronic medical record; the liver and kidney function early warning unit sets ALT>40U / L and Cr>100μmol / L as early warning thresholds; the expert database of the expert review module includes 5 chief physicians of geriatrics (all with more than 10 years of experience in the diagnosis and treatment of insomnia), and the initial review is completed within 24 hours after the plan is submitted, and the final review is completed within 48 hours; the efficacy prediction module is trained based on data from 1,500 elderly patients with chronic insomnia in the hospital. The input is the patient's age, course of disease, initial PSQI, syndrome type and number of underlying diseases, and the output is the predicted value of the PSQI decrease after 4 weeks. If the predicted value is <2 points, the module will automatically push the plan optimization suggestions.

[0050] II. Full Process Implementation Steps Multimodal data acquisition and risk assessment: After admission, the patient's data was collected for 3 nights using a polysomnography system (sleep latency 60 minutes, N3 stage sleep 10%, REM sleep latency 90 minutes, awakenings 6 times / night); the pulse diagnosis system identified it as a "wiry and thready pulse", the tongue diagnosis showed a "yellow and greasy coating", the patient reported "palpitations, irritability, and bitter taste in the mouth", the TCM insomnia syndrome scale score was 16 points (heart and spleen deficiency with phlegm-heat disturbance); the automatic scoring unit of the scale calculated PSQI=18 points, ISI=20 points, and SF-36 physiological function dimension score of 45 points; the LIS system showed ALT=35U / L, Cr=102μmol / L (close to the warning threshold), and CK-MB normal; after the system integrated the data, the medication reduction risk assessment unit calculated the risk level as "high" (compatibility with 3 underlying diseases, Cr close to the threshold), automatically triggering the expert review process.

[0051] AI-powered medication reduction plan generation and rate optimization: The AI-driven step-by-step medication reduction decision module inputs data, and the efficacy analysis unit calculates a medication reduction feasibility score of 62 (high PSQI +35 points, comorbidities -20 points, no adverse reactions +47 points), initiating medication reduction assessment; the efficacy change analysis unit of the medication reduction rate optimization module predicts that the PSQI may decrease by 3 points in the first week (ΔP=3), with the current dose M=3mg, and substitutes into the formula. (κ=1.2, Δt=7d), the optimal weekly dose reduction rate was calculated to be 0.17mg; the dose reduction protocol generation unit reduced eszopiclone from 3mg to 2.76mg / night at an initial rate of 8%, with the dosing time at 21:00, and simultaneously pushed it to the expert database; after the initial review physician's opinion that "Cr is close to the threshold, and the rate can be reduced to 6%", after confirmation by the review physician, the final dose was set at 2.82mg / night, and the protocol iteration unit was set to adjust the rate every 2 weeks according to PSQI.

[0052] TCM Dynamic Early Warning and Treatment Adjustment: In the second week of intervention, the patient's family reported that "the patient's nighttime irritability worsened." The TCM dynamic early warning module monitored the syndrome score, which increased from 16 to 19 (an increase of 18.75%, not reaching the threshold), but the number of awakenings increased from 6 to 8 (an increase of 2), triggering an early warning. The syndrome identification unit still determined it to be a syndrome of deficiency of both heart and spleen with phlegm-heat disturbance. The dynamic adjustment module for the early warning threshold analyzed the patient's comorbid anxiety state (one new underlying disease was added), and substituted into the formula T=0.8×(0.3×20%+0.25×3) (μ=0.8, n=5, ωi is weighted, Ci is the baseline threshold), calculate T=0.8×(0.06+0.75+0.4+6+0.2)=0.8×7.41=5.928, the warning is still triggered after the threshold is relaxed; the adjustment plan of the integrated Chinese and Western medicine intervention module is as follows: add 6g of bamboo shavings to the Chinese herbal tea, add acupressure to Neiguan acupoint (2 minutes / time) for acupoint massage, and keep the original formula for foot bath but reduce the water temperature to 37℃ (to avoid inducing blood pressure fluctuations).

[0053] Safety monitoring and efficacy prediction: In week 3 of the intervention, the patient experienced one episode of orthostatic hypotension (systolic blood pressure decreased by 25 mmHg after standing, CTCAE grade 2). The adverse reaction monitoring unit recorded the time of occurrence (when the patient got up 2 hours after taking the medication), which was correlated with the history of "amlodipine + aspirin" in the electronic medical record, indicating a possible drug interaction. The liver and kidney function warning unit rechecked ALT=38U / L and Cr=101μmol / L, with no further increase. The system paused the medication reduction and pushed it to the expert for review. The physician suggested delaying the medication time to 21:30 to avoid excessively low blood pressure at night. The efficacy prediction module input the current data and predicted that the PSQI would decrease by 2.5 points (≥2 points) after 4 weeks, and no adjustment to the intervention plan was required. The follow-up management module generated a post-discharge follow-up plan: telephone follow-up in weeks 1, 2, and 4 after discharge, and outpatient follow-up in week 8. At each follow-up, PSQI, syndrome score, and blood pressure were assessed, and the data was synchronized to the system to update the medication reduction plan.

[0054] III. Data Characterization for Effectiveness Verification Table 2: Comparison of Management Effectiveness of Complex Cases of Chronic Insomnia in the Elderly in Tertiary Hospitals

[0055] Table 2 shows that in traditional tertiary hospital management, medication reduction plans for complex cases require manual processing of multimodal data (such as PSG, liver and kidney function, and underlying diseases). Expert review relies on paper documents, taking up to 72 hours. Furthermore, atypical syndrome identification depends on physician experience, with an accuracy rate of only 65%, resulting in poor plan suitability and a PSQI improvement rate of only 58%. Adverse reactions require manual linking of medication records by medical staff, with a timely response rate as low as 60%, and some patients experience worsened reactions due to delayed treatment. This invention's system, through automatic multimodal data integration and AI-assisted decision-making, improves expert review efficiency to 36 hours, achieves an atypical syndrome identification accuracy rate of 82%, and, combined with a efficacy prediction module, optimizes plans in advance, increasing the PSQI improvement rate to 83% within 4 weeks. Adverse reaction monitoring automatically links to medication records, achieving a timely response rate of 95%. It also adapts to the medication reduction needs of patients with multiple underlying diseases (initial reduction of 8%), avoiding risks such as blood pressure fluctuations and abnormal liver and kidney function. This solves the problems of slow plan development, difficult syndrome identification, and delayed adverse reaction management in the management of complex cases of elderly patients with chronic insomnia in tertiary hospitals.

[0056] Reference Figure 2 This chart visually demonstrates the effectiveness of the core "step-by-step dose reduction" technique outlined in the application, perfectly aligning with the design in the application that "the initial dose reduction should not exceed 15% of the current dose, and adjustments should be made based on weekly PSQI changes." The data shows that the system strictly follows the dose reduction logic in the application: an initial dose of 3 mg / night, a 12% reduction in the first two weeks (below the 15% upper limit), and the PSQI score decreased from 14 to 12 (a 14.3% reduction); subsequently, a 10% reduction was maintained based on score changes, reaching 2.13 mg / night at week 6, with the PSQI score dropping to 9; during the 8-week follow-up period, the dose was maintained, and the score further decreased to 8 (meeting the application's standard of "maintaining the reduction with a PSQI score decrease of ≥20%"). Compared to the common problem of "sudden dose drops leading to score rebound" in traditional dose reduction methods, this chart verifies that the system, through AI-driven dynamic adjustment, achieves the "smooth dose reduction process and continuous improvement in efficacy" required by the application, reducing drug dependence while ensuring improved sleep quality, perfectly matching the safety needs of elderly patients during dose reduction.

[0057] Reference Figure 3This diagram closely aligns with the core design of the application, "Dynamic Early Warning and Syndrome-Based Intervention in Traditional Chinese Medicine," and is highly consistent with the technical approach outlined in the application, which involves "identifying common syndromes such as deficiency of both heart and spleen, and liver fire disturbing the heart, and adjusting the intervention plan accordingly." Data shows that the scores for all five syndrome types decreased significantly after intervention, with a 42.9% decrease for the syndrome of heart-kidney disharmony and a 40% decrease for the syndrome of deficiency of both heart and spleen. This verifies the effectiveness of the application's "TCM home-use techniques for syndrome type optimization"—for example, the use of a herbal tea with increased jujube dosage for the syndrome of deficiency of both heart and spleen aligns with the application's requirement to "adjust the dosage of medicinal materials based on individual constitution differences"; the addition of bamboo shavings for the syndrome of phlegm-heat disturbance conforms to the application's design of "fuzzy matching and fine-tuning of atypical syndrome types." Compared to the problem of "uniform plans and poor syndrome type matching" in traditional TCM interventions, this diagram demonstrates that the system, through accurate syndrome type identification and dynamic plan adjustment, achieves the application's expected "improvement in the efficacy of TCM intervention," providing evidence-based support at the syndrome type level for the integrated TCM and Western medicine management of chronic insomnia in the elderly.

[0058] Reference Figure 4 This diagram illustrates the dual-scenario design goals of "adaptation to primary healthcare" and "complex case management" outlined in the application, perfectly aligning with the technical solution described in the application: "adapting to the equipment conditions of community health service centers and supporting expert review by tertiary hospitals." Data shows that in primary healthcare scenarios, the system achieves a 78% success rate in reducing medication costs, a significant improvement over the traditional 42% method. The time required for medical and nursing operations has been reduced from 30 minutes to 12 minutes, aligning with the application's design for primary healthcare adaptation, which emphasizes "simplified user interface, support for offline functionality, and compatibility with portable devices." In tertiary hospital scenarios, the success rate is 83%, with a sleep rebound rate of 8%. This is superior to the traditional method's 58% success rate, as the system complies with the application's requirement of "mandatory expert review of high-risk cases and integration of multiple underlying disease influencing factors." On average, the system achieves an 80.5% success rate in reducing medication costs and a 10% sleep rebound rate, demonstrating that it addresses both the "difficult operation and low efficiency" issues at the primary healthcare level and meets the "precision and high safety" requirements of tertiary hospitals. This fully realizes the application's expectation of "multi-scenario promotion and application," providing a comprehensive medical scenario solution for the integrated management of chronic insomnia in the elderly using both traditional Chinese and Western medicine.

[0059] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A system for AI-assisted step-by-step medication reduction and dynamic early warning using traditional Chinese medicine for chronic insomnia in the elderly, characterized in that, include: The data acquisition module includes a multimodal data acquisition unit, an automatic scale scoring unit, and a physiological indicator monitoring unit. The former collects patient demographic, TCM four diagnostic methods, and lifestyle data. The scale unit has multiple built-in scales, and the scores are synchronized with the database. The physiological unit connects to the device to collect indicators, and collects data at different frequencies during the baseline period, intervention period, and follow-up period. The data is encrypted and transmitted to the local server. The AI-driven step-by-step medication reduction decision-making module includes a Western medicine dosage database, efficacy analysis, medication reduction plan generation, and plan iteration unit; the database stores information on commonly used sedative-hypnotic drugs and adjusts dosages according to drug characteristics and patient conditions; the efficacy unit uses a specific model to calculate a medication reduction feasibility score; the plan generation unit calculates the medication reduction range using a specific method; and the iteration unit regularly updates model parameters. The TCM dynamic early warning module includes a syndrome identification, syndrome change monitoring, and early warning triggering unit. The syndrome identification unit uses algorithms to identify common syndromes based on scale data and supports fuzzy matching of atypical syndromes. The monitoring unit tracks core symptoms and calculates the rate of change of syndrome scores periodically. The integrated traditional Chinese and Western medicine intervention module includes a library of traditional Chinese and Western medicine intervention protocols, an intervention timing scheduling and protocol adaptation unit; The Traditional Chinese and Western Medicine Intervention Program Database stores seven categories of intervention programs: herbal tea, acupoint massage, foot bath, acupuncture, massage, cognitive behavioral therapy, and lifestyle adjustment. Each program includes specific operating guidelines, dosage / frequency, and time recommendations, and the time recommendations are designed in accordance with the theory of meridian flow in traditional Chinese medicine. Intervention timing scheduling unit: Combines the timing of Western medicine administration with the AI-based step-by-step medication reduction decision module to avoid peak drug effects and generates a daily schedule for elderly patients with chronic insomnia; at the same time, it coordinates the timing of various interventions: acupuncture and massage are spaced more than 4 hours apart; the "fixed wake-up time" node in cognitive behavioral therapy is staggered from acupoint massage by more than 30 minutes; the morning exercise and breakfast are spaced 20 minutes apart in lifestyle regulation; and nighttime emotional regulation is arranged 30 minutes before taking Western medicine. Protocol adaptation unit: Dynamically adjusts protocol parameters based on patient syndrome type, physical condition, cognitive function, and tolerance; The safety monitoring module includes units for adverse reaction monitoring, liver and kidney function early warning, and medication reduction risk assessment; the adverse reaction unit records reactions according to standard classification; the liver and kidney function unit connects to test data, and triggers early warnings when indicators are abnormal; The risk level is calculated by combining the duration of medication use, the extent of medication reduction, and underlying diseases. The data management module includes blockchain notarization, data anonymization, and historical data traceability units; the blockchain stores medical treatment-related data, with each record containing a key identifier; the anonymization unit uses codes to replace sensitive information, desensitizing biological data; and the traceability unit supports multi-dimensional data retrieval.

2. The AI-assisted step-by-step medication reduction and TCM dynamic early warning system for elderly chronic insomnia according to claim 1, characterized in that, It also includes a TCM syndrome risk assessment module, which comprises a risk factor extraction unit and a dynamic risk calculation unit. The risk factor extraction unit extracts syndrome deterioration-related factors from the four diagnostic methods of TCM; the dynamic risk calculation unit uses formulas... The syndrome risk value is calculated, where R is the syndrome risk value, t0 is the risk assessment start time, t1 is the current assessment time, θ is the syndrome weight coefficient, S(t) is the syndrome score change rate at time t, and D(t) is the underlying disease influence coefficient at time t. When R≥6, the warning level is automatically upgraded, prompting medical staff to strengthen monitoring and adjust the TCM intervention plan.

3. The AI-assisted step-by-step medication reduction and TCM dynamic early warning system for elderly patients with chronic insomnia according to claim 1, characterized in that, It also includes a patient compliance monitoring module, which contains a medication monitoring unit, a traditional Chinese medicine intervention monitoring unit, and a compliance feedback unit. The medication monitoring unit records the patient's daily medication time and dosage through a smart pillbox; the traditional Chinese medicine intervention monitoring unit collects the patient's non-drug therapy execution records through a mobile APP, and verifies the authenticity of the intervention by combining the data from the device's sensors. When the execution rate is <70%, it is marked as low compliance. The compliance feedback unit feeds back the compliance level to the AI-based tiered medication reduction decision module, pushing suggestions for improving compliance. When compliance is moderate, the original medication reduction plan is maintained; when compliance is high, the pace of medication reduction is appropriately accelerated.

4. The AI-assisted step-by-step medication reduction and TCM dynamic early warning system for elderly chronic insomnia according to claim 1, characterized in that, It also includes a medication reduction rate optimization module, which contains an efficacy change analysis unit and a rate calculation unit. The efficacy change analysis unit calculates the weekly changes in PSQI scores and medication dosage changes to analyze the correlation between efficacy and dosage; the rate calculation unit uses a formula... Calculate the optimal drug reduction rate, where v is the weekly drug reduction rate, κ is the efficacy response coefficient, ΔP is the weekly change in PSQI score, Δt is the time interval, and M is the current drug dose.

5. The AI-assisted step-by-step medication reduction and TCM dynamic early warning system for chronic insomnia in the elderly according to claim 1, characterized in that, It also includes a sleep structure in-depth analysis module, which contains a sleep parameter extraction unit, a cycle analysis unit, and an intervention adaptation unit. The sleep parameter extraction unit extracts core parameters from data from a portable polysomnography monitor, including sleep latency, the proportion of N1-N3 stage sleep, the proportion of REM sleep, the number of awakenings, and sleep efficiency. The cycle analysis unit identifies the patient's sleep cycle characteristics. The intervention adaptation unit adjusts the integrated traditional Chinese and Western medicine intervention plan according to the type of sleep structure abnormality.

6. The AI-assisted step-by-step medication reduction and TCM dynamic early warning system for elderly patients with chronic insomnia according to claim 1, characterized in that, It also includes a dynamic adjustment module for early warning thresholds. This module contains an individual difference analysis unit and a threshold calculation unit. The individual difference analysis unit analyzes individual characteristics such as patient age, underlying diseases, and medication history to determine the difference coefficient; the threshold calculation unit uses a formula... Calculate the personalized early warning threshold, where T is the personalized early warning threshold, μ is the individual difference coefficient, n is the number of early warning related indicators, ωi is the weight of the i-th indicator, and Ci is the baseline threshold of the i-th indicator.

7. The AI-assisted step-by-step medication reduction and TCM dynamic early warning system for elderly patients with chronic insomnia according to claim 1, characterized in that, It also includes a primary healthcare adaptation module, which comprises a simplified operation unit, an offline function unit, and a device compatibility unit. The simplified operation unit simplifies the system interface to three core function pages: data entry, plan viewing, and early warning processing, reducing the need for professional parameter settings and supporting voice input of TCM symptom descriptions. The offline function unit stores the collected data for 7 days in a network-free environment and automatically synchronizes it to the server after the network is restored. When offline, it calls the locally cached basic medication reduction plan and TCM intervention template. The device compatibility unit adapts to commonly used portable blood glucose meters, electronic blood pressure monitors, and simple sleep monitoring devices in primary healthcare settings, enabling quick connection via Bluetooth or USB interface.

8. The AI-assisted step-by-step medication reduction and TCM dynamic early warning system for chronic insomnia in the elderly according to claim 1, characterized in that, It also includes an expert review module, which contains a plan submission unit, a multi-level review unit, and a feedback unit. The plan submission unit automatically submits the AI-generated medication reduction plan and TCM intervention plan to the expert database, and prioritizes their allocation to senior physicians with more than 10 years of experience in treating geriatric insomnia. The multi-level review unit sets up a two-level process of initial review and review. The initial review physician reviews the rationality of the dosage and the degree of matching of the syndrome type, and the review physician reviews the early warning and treatment measures and the long-term intervention logic. After both levels of review are passed, the plan is pushed to the patient. The feedback unit will synchronize the experts' revisions to the AI-based tiered drug reduction decision-making module and the TCM dynamic early warning module.

9. The AI-assisted step-by-step medication reduction and TCM dynamic early warning system for elderly patients with chronic insomnia according to claim 1, characterized in that, It also includes a follow-up management module, which contains a follow-up plan generation unit, an automatic reminder unit, and a follow-up data update unit. The follow-up plan generation unit generates personalized follow-up plans based on the patient's intervention stage. The automatic reminder unit reminds patients and medical staff 24 hours in advance via SMS, APP push, and telephone voice, and adds reminders every 12 hours if a follow-up is not scheduled. The follow-up data update unit supports the rapid entry of new data during follow-up, automatically compares the differences between the current data and historical data, generates a follow-up report, and simultaneously updates the AI ​​medication reduction plan and TCM warning thresholds.

10. The AI-assisted step-by-step medication reduction and TCM dynamic early warning system for elderly patients with chronic insomnia according to claim 1, characterized in that, It also includes an efficacy prediction module, which contains a historical data training unit and an efficacy prediction unit. The historical data training unit trains the efficacy prediction model based on the diagnosis and treatment data of more than 1,000 elderly patients with chronic insomnia stored in the system. The input parameters of the model include patient age, disease course, initial PSQI score, syndrome type, medication dosage and TCM intervention type. The output parameter is the predicted value of the decrease in PSQI score after 4 weeks. The efficacy prediction unit calls the model to generate efficacy prediction values ​​each time the medication reduction plan and TCM intervention plan are adjusted.

Images