Intelligent administration compliance management system for HIV patients

Abstract

The invention discloses an HIV patient intelligent administration compliance management system, which comprises an intelligent sensing layer, a data processing layer and an application service layer, and realizes data interaction through an Internet of Things (IoT) communication protocol (preferably Bluetooth 5.0, NB-IoT). The system is used for solving the problems of self-report deviation, dependence on active operation, lack of drug storage monitoring and insufficient privacy protection in HIV patient administration compliance management. Through the comprehensive action of the technical effects, the drug administration compliance rate of the HIV patient can be remarkably improved: on one hand, the conditions of missing taking, wrong taking and invalid taking of drugs are reduced through accurate monitoring and dynamic reminding; on the other hand, the problem of compliance caused by side effects, scene changes and the like can be solved in time through personalized intervention of medical staff based on real data, the risk of virus drug resistance is reduced, the long-term treatment effect of patients is finally improved, the disease deterioration and virus propagation probability is reduced, and the method has important clinical application value and public health significance.

Description

Technical Field

[0001] This invention relates to a smart medication adherence management system for HIV-infected patients who need to take medication regularly for a long period of time. Background Technology

[0002] HIV (Human Immunodeficiency Virus) infected patients need to take antiviral drugs regularly for a long period to suppress viral replication and control disease progression. Medication adherence is a core factor determining treatment effectiveness. Clinical studies have shown that when HIV patient medication adherence is below 95%, the risk of viral resistance increases significantly, leading to treatment failure, disease deterioration, and an increased probability of viral transmission. However, current HIV patient medication adherence management still faces several technical deficiencies and application challenges, as follows:

[0003] 1. Self-reporting bias leads to distorted management data. Current HIV patient medication adherence monitoring relies heavily on patient subjective reports (such as paper logs or manual recording of medication status via an app) or regular follow-up visits by healthcare staff. This model suffers from a serious "self-reporting bias"—patients may forget or conceal missed / incorrect doses (e.g., fear of reprimand from healthcare staff or avoidance of negative feedback about poor treatment outcomes), resulting in inaccurate data. Healthcare staff cannot obtain accurate medication behavior data, making it difficult to develop targeted intervention plans and ultimately affecting management efficiency.

[0004] 2. Existing monitoring technologies rely on active operation, resulting in high patient compliance barriers. Current mainstream drug administration monitoring technologies (such as button-operated smart pillboxes and medication reminder devices requiring manual QR code confirmation) all require patients to actively perform operations (such as manually pressing buttons to record medication or scanning the drug's QR code for confirmation). Some HIV patients are older, have weaker operational abilities, or experience fatigue or cognitive decline during treatment. The active operation mode can easily lead to patients abandoning use due to its perceived "cumbersome operation," and it cannot prevent inaccurate monitoring data caused by "operation by others" (such as family members pressing buttons to record medication), failing to accurately reflect the patient's actual medication behavior.

[0005] 3. Fragmented multi-dimensional data and lack of integrated analysis capabilities: Existing technologies mostly focus on "medication action recognition" (such as using a pillbox sensor to determine whether medication has been taken), which can only monitor the single behavior of "whether medication has been taken," while ignoring other key dimensions of data closely related to medication adherence:

[0006] Physiological state dimension: HIV drugs (such as nucleoside reverse transcriptase inhibitors) are prone to causing side effects such as liver damage and anemia. Current technology cannot collect patients' liver function (ALT, AST indicators), heart rate, blood oxygen and other physiological data in real time or periodically, and cannot link the causal relationship between "side effects" and "decreased drug compliance".

[0007] Behavioral habits and scenario dimensions: Scenarios such as patients seeking medical treatment in other places or traveling on business trips can disrupt the medication schedule. Current technologies lack the ability to identify patients' home / outdoor scenarios and cannot dynamically adjust reminder strategies, which can easily lead to patients missing doses.

[0008] Drug storage: Some HIV drugs (such as certain integrase inhibitors) need to be stored in strict light-proof and refrigerated (2~8℃). Improper storage conditions can lead to drug inactivation. Current technology lacks a real-time monitoring and early warning mechanism for drug storage temperature and humidity. Patients may take inactivated drugs due to improper storage, which will affect the treatment effect.

[0009] 4. Inefficient doctor-patient collaboration and lack of privacy protection mechanisms.

[0010] On the one hand, under the existing management model, patient medication data and physiological data are stored in different devices (such as smart pillboxes storing medication records locally and wearable devices storing physiological data independently). Medical staff need to obtain data through methods such as manual uploads by patients and offline submissions. Data transmission is lagging and integration is difficult, making it impossible to detect abnormal patient compliance and intervene in a timely manner. On the other hand, HIV disease involves patient privacy (such as disease information, identity information, and geographical location information). Existing technologies lack specific privacy protection designs for HIV patient data, which easily leads to the risk of data leakage, causing patients to resist using the monitoring system due to concerns about privacy exposure.

[0011] In summary, current HIV patient medication adherence management technologies cannot meet the actual needs of "passive monitoring, multi-dimensional integration, efficient collaboration, and secure privacy protection," and there is an urgent need for an intelligent management system that can solve the above-mentioned core problems. Summary of the Invention

[0012] In order to overcome the shortcomings of the prior art, one of the objectives of this invention is to provide an intelligent drug administration compliance management system for HIV patients.

[0013] One of the objectives of this invention is achieved through the following technical solution:

[0014] An intelligent drug administration adherence management system for HIV patients includes an intelligent sensing layer, a data processing layer, and an application service layer. Each layer interacts with data through Internet of Things (IoT) communication protocols (preferably Bluetooth 5.0 or NB-IoT). The system is used to address the problems of self-reporting bias, reliance on active operation, lack of drug storage monitoring, and insufficient privacy protection in HIV patient drug administration adherence management.

[0015] The intelligent sensing layer includes a multimodal data passive acquisition module, which integrates Internet of Things (IoT) devices (smart pillboxes, wearable devices, home scene sensors) and non-invasive monitoring technologies (facial recognition, weight sensing, and non-intrusive acquisition of physiological parameters) to passively collect multi-dimensional data related to HIV patient medication, without requiring patients to actively press buttons or manually input operations.

[0016] The data processing layer is deployed on a cloud server and includes a data cleaning submodule, a data integration submodule, and an analysis submodule. The data cleaning submodule is used to remove abnormal data (such as instantaneous fluctuation data from weight sensors and offline invalid data from wearable devices). The data integration submodule is used to associate multi-source data by timestamp. The analysis submodule is used to generate patient medication adherence assessment results and risk warning information.

[0017] The application service layer includes a patient-side APP, a medical staff-side management platform, and a smart medicine box local interaction unit, which are used to provide management services based on the analysis results of the data processing layer and meet privacy protection requirements through encryption technology.

[0018] The multimodal data passive acquisition module includes an upgraded smart pillbox, which comprises:

[0019] Face recognition module: It adopts a deep learning-based face recognition algorithm (referencing the facial feature extraction architecture of CN209884764U, adding a liveness detection function, and verifying through blinking and head shaking to avoid photo deception). It pre-stores patient facial feature templates, and collects the facial image of the person picking up the medicine in real time when picking up the medicine, and compares it with the template (the comparison threshold is ≥95% to be considered successful). If the comparison is successful, the medicine box can be opened; if the comparison fails, an identity abnormality warning is triggered.

[0020] Weight sensing module: Employs a high-precision pressure sensor (detection accuracy ±0.1mg) to monitor the total weight of the medication in the pillbox in real time. It identifies the entire "medication dispensing-administration" process through continuously collected weight change curves: When the weight reduction equals the single prescription dose (preset in the system), it is determined as "valid medication dispensing"; if the weight reduction is not equal to the single dose or there is no secondary weight change within 10 minutes after dispensing (excluding medication being returned), it is determined as "abnormal medication administration".

[0021] Temperature and humidity sensing module: adopts digital temperature and humidity sensors (temperature measurement range -10℃~60℃, accuracy ±0.5℃; humidity measurement range 0~100%RH, accuracy ±3%RH), preset suitable storage thresholds for HIV drugs (temperature 2~8℃, humidity 30%~60%RH in a light-protected environment, the specific thresholds can be adjusted according to the type of drug). When the monitored value exceeds the threshold for more than 5 minutes, a storage abnormality warning is triggered.

[0022] The triggering methods for the identity anomaly warning and storage anomaly warning include: the smart medicine box emits a buzzer (frequency 2kHz, lasting 10 seconds) and flashes an LED light (red, at 1-second intervals) locally, while simultaneously pushing text warning information to the patient's APP.

[0023] As a further improvement to the above technical solution:

[0024] The multimodal data passive acquisition module also includes a wearable device linkage unit. The wearable device linkage unit establishes a stable connection with a smartwatch / bracelet via Bluetooth 5.0 protocol to collect the patient's physiological data in real time. The collection frequency is: heart rate once per minute, blood oxygen once every 5 minutes, and sleep stages once every 30 seconds. The wearable device linkage unit uploads the collected physiological data to the data processing layer in batches of 15 minutes. Before uploading, CRC verification is used to ensure data integrity.

[0025] The wearable device linkage unit is also compatible with a portable liver function testing device, which uses a dry biochemical sensor (based on the principle of enzyme-catalyzed reaction-optical colorimetry, with ≥2 detection channels) to detect the patient's ALT (alanine aminotransferase) and AST (aspartate aminotransferase) levels. The wearable device linkage unit connects to the liver function testing device via a USB-Type-C interface and automatically synchronizes the test data according to a preset cycle (default once a week, which can be adjusted by the medical staff to every 3 days / every 2 weeks). After synchronization, the data is associated with the patient's medication record for that day. The analysis submodule of the data processing layer presets the normal reference range for ALT to 0-40 U / L and the normal reference range for AST to 0-40 U / L. When the test value exceeds the range, it is marked as "abnormal liver function index" and pushed to the medical staff management platform.

[0026] The multimodal data passive acquisition module further includes an environment and behavior perception unit, which includes:

[0027] Home scene sensors include smart door lock sensors (detecting door lock open / close status and opening / closing time, sampling once per second) and refrigerator door magnetic sensors (detecting refrigerator door open / close status and continuous opening time, sampling once per second). When the smart door lock has no opening record for 8 consecutive hours and the refrigerator door is opened ≥3 times per day, the patient is determined to be in "home status". Conversely, if the smart door lock is opened ≥2 times in 24 hours and the refrigerator door is opened ≤1 time, the patient is initially determined to be "not at home".

[0028] GPS positioning module: Integrated into the patient's app, it uses differential positioning technology (positioning accuracy of 5-10 meters). After collecting the patient's location information, it performs privacy desensitization through "coordinate offset + regional coding" (offsetting the actual latitude and longitude by 50-100 meters, or converting it to the city-level administrative region code, without retaining the precise coordinates).

[0029] The environment and behavior perception unit uploads the "home / non-home" determination result and the desensitized location information to the data processing layer every hour. If the desensitized location is detected to be ≥50 kilometers away from the patient's permanent residence (preset in the system) for more than 24 hours, it is determined to be a "business trip / medical treatment scenario".

[0030] The analysis submodule of the data processing layer includes a scene analysis subunit. After receiving the scene determination results from the environment and behavior perception unit, the scene analysis subunit generates a reminder strategy adjustment plan based on the patient's preset medication plan (including daily medication time, dosage, and drug type).

[0031] For "business trip / medical treatment in another city": advance the medication reminder time by 1 hour (to avoid travel conflicts) and automatically adjust the reminder time according to the time zone of the other city (e.g., if the patient is traveling from Beijing to New York, the reminder time will be adjusted according to the New York time zone).

[0032] If the patient is in "home status": the original medication schedule will be followed. If the smart lock detects that the patient has left the house (the door is open and not closed), a second reminder will be sent after a 10-minute delay (to avoid missing the patient).

[0033] The notification strategy adjustment plan is pushed to the patient-side APP and smart pillbox in real time at the application service layer.

[0034] The analysis submodule of the data processing layer also includes a compliance assessment subunit. This compliance assessment subunit uses a 7-day assessment cycle and constructs an assessment model using the following multi-dimensional data as input parameters:

[0035] Dosing behavior data: Dosing on time rate (the deviation between the actual dosing time and the planned time is ≤30 minutes is considered on time), dosage accuracy rate (the actual amount of medicine dispensed is consistent with the planned dosage is considered accurate).

[0036] Physiological data: Number of abnormal heart rate / blood oxygen levels collected by wearable devices, and whether liver function test indicators are normal;

[0037] Scenario matching data: the degree of matching between drug administration behavior and "home / remote" scenarios (e.g., administering drugs according to the adjusted reminder time in a remote scenario is considered a match);

[0038] The assessment model categorizes compliance levels into four levels: Excellent (overall score ≥ 90 points), Good (75–89 points), Average (60–74 points), and Poor (< 60 points). It generates a compliance assessment report containing scores for each parameter and explanations of any anomalies, which is pushed to the application service layer every 7 days.

[0039] The patient-side APP of the application service layer includes the following functional modules:

[0040] The reminder module receives and displays medication reminders (text + ringtone reminders) and abnormal warnings (storage abnormalities, identity abnormalities, and physiological indicator abnormalities), and supports patients in viewing historical reminder records.

[0041] Report module: Displays individual adherence assessment reports, marks abnormal items (such as "Insufficient medication dosage at 18:00 on 2024-XX-XX, dosage accuracy 80%), and provides improvement suggestions (such as "It is recommended to check the single dose shown on the medicine box when picking up the medication").

[0042] Privacy settings module: Supports access permissions for patient management data (e.g., allowing only the attending physician to view liver function data and denying access to other personnel);

[0043] The application service layer's healthcare management platform includes a patient data dashboard (displaying multi-dimensional data categorized by patient ID), an assessment report viewing module, and an intervention plan generation module. Healthcare staff can generate personalized intervention plans based on assessment reports (such as increasing weekly follow-ups for patients with poor compliance and adjusting medication dosages for patients with abnormal liver function) and push the plans to the patient's app.

[0044] The application service layer also includes a privacy protection module, which employs a full-link encryption mechanism for transmission, storage, and access.

[0045] Data transmission encryption: All uploaded / downloaded data is encrypted using the AES-256 symmetric encryption algorithm, and the key is automatically updated every 24 hours;

[0046] Data storage encryption: Patient personal information (name, ID number, HIV condition information) is stored on a cloud server using the RSA asymmetric encryption algorithm, and only the encrypted ciphertext is retained;

[0047] Access control: Patients access personal data through dual authentication of "fingerprint recognition + APP password", and medical staff verify access rights through "employee ID + dynamic password (updated every 60 seconds)". The system records all access logs (including access personnel, time and data type), and the logs are retained for ≥1 year.

[0048] The facial recognition module of the smart pillbox also includes an anomaly handling unit: when the comparison fails three times in a row, the pillbox is automatically locked (locked for 30 minutes), and a "multiple identity verification failures" warning message is pushed to the medical staff management platform; if the patient's facial condition changes (such as wearing a mask or slight facial swelling) and the comparison fails, the patient can send an "identity reset application" through the patient's APP, and the facial feature template will be updated after the medical staff approves it.

[0049] The weight sensing module of the smart pillbox also includes an anomaly analysis unit: when the following anomalies are detected, a corresponding anomaly warning is triggered and uploaded to the data processing layer:

[0050] Abnormal medication dispensing: Medication dispensing amount < 80% of the single prescription dose (judged as "insufficient dose") or > 120% (judged as "excessive dose");

[0051] Delayed medication: No change in weight within 30 minutes after medication is obtained (determined as "medication not taken after obtaining medication");

[0052] Medication Remaining Abnormality: The amount of medication remaining in the medicine box is less than the total prescription dose for 3 days (judged as "insufficient medication"), triggering a "Medication Replenishment Required" reminder;

[0053] The abnormal warning information is simultaneously pushed to the patient's mobile app and the medical staff's management platform. The medical staff can proactively contact the patient to confirm the situation based on the warning information.

[0054] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0055] (i) Eliminate reliance on active operation to improve the authenticity of monitoring data and the ease of system use.

[0056] The multimodal data passive acquisition module of the intelligent sensing layer of this invention upgrades the smart pillbox through "face recognition + weight sensing"—the face recognition module automatically verifies the identity of the person picking up the medicine (to prevent children from accidentally ingesting it or others from taking it on their behalf), and the weight sensing module automatically identifies the entire process of "picking up medicine and taking it" through the drug weight change curve. There is no need for patients to manually press buttons, scan codes or perform other active operations, which completely solves the problem of "high threshold for active operation". It is especially suitable for HIV patients with weak operation ability or with treatment side effects, and significantly improves the ease of use of the system.

[0057] The passive data collection mode avoids data distortion caused by "operation by others" and "subjective concealment". Medical staff can obtain the patient's real medication behavior data (such as medication time, dosage, and whether the medication was taken), providing accurate data support for compliance assessment and intervention, and effectively reducing management errors caused by self-reporting bias.

[0058] (ii) Multi-dimensional data integration and analysis to achieve full-link management of "behavior-physiology-scenario-storage".

[0059] Physiological state correlation management: The system collects heart rate, blood oxygen, and sleep stage data in real time through wearable device linkage units, and is compatible with portable liver function testing devices to synchronize ALT and AST indicators. It can directly link "abnormal physiological indicators" with "drug administration behavior" - for example, when the ALT index is detected to be outside the normal range, medical staff can quickly determine whether the patient's compliance has decreased due to drug side effects, and then adjust the medication plan in a timely manner to avoid the deterioration of compliance caused by the failure to intervene in side effects in a timely manner;

[0060] Contextualized dynamic reminders: The environment and behavior perception unit determines the patient's home status through home scene sensors (smart door locks, refrigerator door magnets), and combines privacy-desensitized GPS positioning to identify business trips / medical treatment scenarios in other places. The data processing layer automatically adjusts the medication reminder strategy based on the scenario results (such as reminding patients 1 hour in advance and calibrating the time zone in other places scenarios, and delaying the second reminder in home and outing scenarios), effectively reducing the risk of missed doses caused by scenario changes.

[0061] Medication storage safety assurance: The smart pillbox has a built-in temperature and humidity sensor module that monitors the storage conditions of the medication in real time. When the temperature and humidity exceed the suitable range for HIV medication (such as temperature > 8℃, humidity > 60%RH), an alarm will be triggered through local beeping, LED flashing, and APP push notifications to prevent patients from taking ineffective medication due to improper storage and to ensure the treatment effect.

[0062] (III) Optimize the efficiency of doctor-patient collaboration and achieve a closed loop of "real-time monitoring - timely intervention - personalized management".

[0063] The system's data processing layer is deployed on a cloud server. Through data cleaning, integration, and analysis sub-modules, it associates multi-source data (drug administration behavior, physiological indicators, and scenario information) by timestamp, automatically generates compliance assessment reports (divided into four levels: excellent, good, average, and poor, with abnormal items marked), and pushes them to the medical staff management platform in real time. Medical staff do not need to manually collect and integrate data, and can quickly grasp the patient's compliance status, significantly improving collaboration efficiency.

[0064] The healthcare management platform at the application service layer supports the generation of personalized intervention plans—for example, increasing the frequency of follow-up visits for patients with poor compliance, and adjusting the medication dosage for patients with abnormal liver function. The plans can be directly pushed to the patient's app, realizing a closed-loop management system of "monitoring-assessment-intervention-feedback" and avoiding untimely interventions due to data lag.

[0065] (iv) End-to-end privacy protection to eliminate patient concerns.

[0066] In response to the significant privacy protection needs of HIV patients, this invention employs a full-link encryption mechanism across the entire transmission-storage-access process in its application service layer privacy protection module.

[0067] During the data transmission phase, AES-256 symmetric encryption and automatic key updates every 24 hours ensure transmission security.

[0068] During the data storage phase, sensitive information such as patient names, ID numbers, and HIV status are stored using RSA asymmetric encryption, retaining only the encrypted data.

[0069] Access permissions are verified through dual authentication of "patient fingerprint + APP password" and "medical staff ID + dynamic password", combined with access log retention (≥1 year) to achieve traceability, completely eliminating the risk of data leakage, eliminating patients' concerns about privacy and resistance to using the system, and improving system acceptance and actual application rate.

[0070] (v) Reduce treatment risks and improve long-term treatment outcomes for HIV patients

[0071] Through the combined effects of the above technologies, this invention can significantly improve HIV patient medication adherence: on the one hand, accurate monitoring and dynamic reminders reduce missed doses, incorrect doses, and the use of ineffective drugs; on the other hand, personalized intervention by medical staff based on real data can promptly address adherence issues caused by side effects and changes in circumstances, reduce the risk of viral drug resistance, and ultimately improve long-term treatment outcomes for patients, reduce disease deterioration and the probability of viral transmission, thus having significant clinical application value and public health significance.

[0072] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described in detail below with reference to the accompanying drawings. Attached Figure Description

[0073] Figure 1 This is a diagram illustrating the overall system architecture and data flow of this embodiment;

[0074] Figure 2 This is a flowchart illustrating the workflow of the smart pillbox in this embodiment.

[0075] Figure 3 This is a flowchart of the compliance assessment and intervention process in this embodiment;

[0076] Figure 4 This is a flowchart of the scene-adaptive reminder process in this embodiment;

[0077] Figure 5 This is a flowchart of the entire privacy protection process in this embodiment. Detailed Implementation

[0078] Next, in combination with the accompanying drawings and specific embodiments, the present invention will be further described. It should be noted that, on the premise of no conflict, any combination of the following-described embodiments or technical features can form a new embodiment.

[0079] It should be noted that when a component is referred to as "fixed to" another component, it can be directly on the other component or there may also be an intermediate component. When a component is considered to be "connected to" another component, it can be directly connected to the other component or there may be an intermediate component at the same time. When a component is considered to be "disposed on" another component, it can be directly disposed on the other component or there may be an intermediate component at the same time. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are only for the purpose of illustration.

[0080] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which this invention belongs. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more of the related listed items. Specific Embodiment

[0081] In this embodiment, taking a tertiary hospital's HIV specialized treatment center as an example, 60 HIV patients (meeting the inclusion criteria: aged 22 - 65 years, diagnosed with HIV for ≥1 year, taking fixed-dose antiviral drugs for a long time, and having basic smartphone operation ability) were selected as experimental subjects. The system deployment period was 3 months to comprehensively verify the function implementation and effectiveness of the "intelligent perception layer - data processing layer - application service layer".

[0082] 1.1 Specific Implementation of Each Layer of the System

[0083] 1.1.1 Intelligent Perception Layer (Multimodal Data Passive Acquisition Module)

[0084] (1) The upgraded intelligent medicine box

[0085] Hardware Configuration:

[0086] The main body of the medicine box is made of ABS anti-drop material (size 15 cm × 10 cm × 5 cm) and is内置以下模块:

[0087] Face recognition module: It adopts the HiSilicon Hi3519V100 image processing chip, integrates a 2-megapixel camera (supports infrared liveness detection to avoid photo / video spoofing), pre-stores patient facial feature templates (collects 3 facial images from different angles to build a template library), and sets the comparison threshold to 95% (comparison time ≤ 0.5 seconds); if the comparison fails 3 times in a row, the medicine box is automatically locked for 30 minutes and a "verification abnormality" warning is pushed to the medical staff.

[0088] Weight sensing module: Employs an HX711 high-precision pressure sensor (detection accuracy ±0.1mg), installed at the bottom of the medicine box, to collect the total weight of the medicine in real time (sampling frequency 1 time / second); identifies the process through weight change curves.

[0089] If the weight reduction is equal to the preset single dose (e.g., if a patient's prescription is "Raltelapir Potassium Tablets 400mg / time", then the reduction corresponds to the weight of 400mg of medication), and there is no secondary weight change within 10 minutes after the reduction (excluding "returning the medication after taking it"), it is determined to be "effective medication administration".

[0090] If the weight reduction is less than 80% of a single dose (e.g., only 200mg is taken), it is considered "underdose"; if the weight reduction is greater than 120% of a single dose (e.g., 500mg is taken), it is considered "overdose".

[0091] If the weight returns to its original weight within 30 minutes after the medication is taken (if the medication is returned without being taken), it is considered "medication taken but not taken".

[0092] Temperature and humidity sensing module: Employs SHT30 digital sensor (temperature measurement range -10℃~60℃, accuracy ±0.5℃; humidity range 0~100%RH, accuracy ±3%RH), preset HIV drug storage threshold (based on patient medication type, such as "emtricitabine tenofovir tablets" set to "temperature 2~8℃, humidity 30%~60%RH"); when the monitored value exceeds the threshold for 5 minutes, a local alarm is triggered (2kHz buzzer for 10 seconds + red LED flashing at 1-second intervals), and a "Drug storage abnormality" text reminder is pushed to the patient's APP.

[0093] (2) Wearable device linkage unit

[0094] Equipment selection and data acquisition:

[0095] It connects to the Huawei Watch GT3 smartwatch (supports Bluetooth 5.0 protocol) to collect physiological data in real time: heart rate (1 beat / minute), blood oxygen saturation (1 time / 5 minutes), and sleep stages (1 time / 30 seconds, distinguishing between light sleep / deep sleep / REM sleep); the data is uploaded to the patient's APP via Bluetooth every 15 minutes, and the integrity is ensured by CRC16 verification before uploading (if the verification fails, the data is collected again).

[0096] It is compatible with Shenzhen Aikon Biotechnology iCARE-100 portable liver function tester (based on enzyme-catalyzed reaction-optical colorimetry principle, with 2 detection channels) and is linked to a watch via USB-Type-C interface; the preset testing cycle is once a week (the medical staff can adjust it to every 3 days / 2 weeks according to the patient's historical liver function data). During the test, the patient only needs to add 20μL of fingertip blood, and after 5 minutes, the ALT (normal range 0~40U / L) and AST (normal range 0~40U / L) data are automatically synchronized to the system.

[0097] (3) Environmental and behavioral perception unit

[0098] Home scene sensors:

[0099] Deploy Xiaomi Smart Door Lock E10 (detects door lock open / close status, sampling frequency 1 time / second) and Xiaomi Refrigerator Door Magnetic Sensor (detects refrigerator door open / close status and continuous open time, sampling frequency 1 time / second); Judgment Logic:

[0100] If the smart door lock has no opening record for 8 consecutive hours and the refrigerator door is opened ≥3 times per day, it is determined to be in "home status".

[0101] If the smart door lock is opened ≥2 times within 24 hours and the refrigerator door is opened ≤1 time, it is preliminarily determined that the user is not at home.

[0102] GPS location and privacy anonymization:

[0103] Integrated into the patient's APP (developed based on Android 12 / iOS 16 system), it uses differential GPS positioning (accuracy of 5~10 meters); the desensitization method is "coordinate offset + regional coding": the actual latitude and longitude are offset 80 meters to the east, and at the same time converted to the city-level administrative region code (such as Beijing code 110000, without retaining the precise coordinates); if the straight-line distance between the desensitized location and the patient's permanent residence (preset as "Haidian District, Beijing") is ≥50 kilometers and lasts for 24 hours, it is judged as "business trip / medical treatment scenario".

[0104] 1.1.2 Data Processing Layer (Cloud Server)

[0105] Hardware configuration: Alibaba Cloud ECS server (4 cores, 8GB memory, 500GB SSD storage, supports elastic expansion), operating system is CentOS 8.0.

[0106] Core functionality implementation:

[0107] Data cleaning submodule:

[0108] Abnormal data is removed using the "3σ criterion": For example, instantaneous fluctuations in weight sensing (single fluctuation > 5mg and duration < 1 second), and invalid off-line data of wearable devices (blank data segments with Bluetooth disconnected > 30 minutes); Missing data is completed using the "linear interpolation method" (for example, if the missing blood oxygen data ≤ 2 times, it is completed with the average value of adjacent data before and after).

[0109] Data integration sub-module:

[0110] Multi-source data is associated according to the principle of "timestamp alignment": For example, the medication-taking record at "2024-06-15 08:00", the synchronous heart rate data, and the door lock status data are marked as the same time node, generating a "patient's single-dose administration event data packet".

[0111] Analysis sub-module:

[0112] Scene analysis sub-unit: After receiving the determination result of "at home / away", the reminder strategy is adjusted in combination with the preset medication plan (for example, a certain patient plans to take medicine at 08:00 and 20:00 every day):

[0113] Away scene (such as the patient traveling from Beijing to Shanghai): The reminder time is advanced by 1 hour (07:00, 19:00), and it is automatically calibrated according to the Shanghai time zone (to avoid missing doses due to time difference);

[0114] At-home and out scene (door lock opened > 30 minutes without closing): The first reminder is sent at the original reminder time (08:00), and the second ringtone reminder is sent 10 minutes later (the ringtone type is "crescendo" to avoid missing in a noisy environment).

[0115] Compliance assessment sub-unit: Taking 7 days as a cycle, a weighted scoring model (total score 100 points) is constructed, and the input parameters and weights are as follows: <�

[0116]

[0117] Assessment level classification: Excellent (≥90 points), Good (75 - 89 points), Medium (60 - 74 points), Poor (<60 points), generating an assessment report including "scores of each index, description of abnormal items (such as "insufficient dose on June 10th, accuracy rate 85.7%")".

[0118] 1.1.3 Application service layer (patient-side APP + medical staff-side management platform)

[0119] (1) Patient-side APP (name: "HIV Smart Medicine Manager")

[0120] Core function module:<�

[0121] The reminder module displays medication reminders (text "400mg of raltegravir potassium tablets need to be taken at 08:00 today" + "ding-dong" sound), abnormal warnings (red icons indicate "storage abnormality / identity abnormality"), and supports viewing historical reminder records within 3 months.

[0122] Reporting module: A compliance assessment report (including score, grade, abnormal items and improvement suggestions, such as "dosage accuracy is low, it is recommended to check the single dose on the medicine box display screen when picking up the medicine) will be pushed out every Monday).

[0123] Privacy settings module: Patients can select "Allow healthcare personnel to access data" (only the attending physician can view liver function data, while other healthcare personnel can only view compliance reports), and can revoke permissions at any time.

[0124] (2) Medical staff management platform (Web version, access address: https: / / hiv-mgmt.hospital.com)

[0125] Core functional modules:

[0126] Patient data dashboard: Displays "real-time medication status (taking medication / not taking medication), physiological indicator trend chart (ALT change curve in the past month), and scene distribution (home / remote location percentage)" categorized by patient ID.

[0127] Intervention plan generation module: For patients with poor compliance (e.g., score <60 points), it automatically recommends "add one telephone follow-up per week"; for patients with abnormal liver function (ALT>40U / L), it recommends "adjust the medication dosage (e.g., reduce from 400mg to 300mg)", which can be directly pushed to the patient's APP after confirmation by medical staff.

[0128] (3) Privacy Protection Module

[0129] End-to-end encryption implementation:

[0130] Transmission encryption: All data (such as patient identity information and physiological data) is encrypted using the AES-256 symmetric encryption algorithm. The key is automatically updated every 24 hours after two-way authentication between the cloud and the APP.

[0131] Encrypted storage: Patient's name, ID number, and HIV condition information are stored using RSA-2048 asymmetric encryption (public key encryption, private key held only by the patient and authorized medical staff), and only the encrypted text is stored in the cloud.

[0132] Access control: Patients log in using a dual authentication method of "fingerprint recognition + 6-digit APP password"; medical staff are verified using "employee ID + dynamic password (updated every 60 seconds and sent by the hospital's OA system)"; the system records all access logs (including access personnel, time, and data type), and the logs are retained for 2 years for traceability.

[0133] 1.2 Example of System Workflow (Taking a Patient Traveling to Another Location as an Example)

[0134] Scenario Trigger: Patient Li (permanent residence: Beijing) traveled to Shanghai for business. After the GPS module of the APP collected the location information, it was desensitized to "Huangpu District, Shanghai" after being offset by 80 meters. The straight-line distance from the permanent residence (Haidian District, Beijing) was greater than 50 kilometers and lasted for 24 hours. The environmental and behavioral perception unit determined it to be a "business trip scenario" and uploaded it to the cloud.

[0135] Strategy Adjustment: After receiving the results, the scenario analysis subunit of the data processing layer generates a reminder strategy based on Li's plan to take medication at 08:00 and 20:00 every day: the reminder time is advanced by 1 hour (07:00 and 19:00) and calibrated according to the Shanghai time zone (UTC+8).

[0136] Reminder execution: The application service layer pushes the adjusted reminder to Li's APP. At 07:00, the APP emits a "ding-dong" ringtone and text reminder, while the smart medicine box LED light flashes green.

[0137] Medication monitoring: When Li took the medicine, the facial recognition module successfully matched the face (takes 0.3 seconds) and the medicine box was opened; the weight sensor module detected that the medicine had decreased by 400mg (equal to a single dose), and there was no weight change within 10 minutes, which was determined to be "effective medication", and the data was synchronized to the cloud.

[0138] Synchronized with medical staff: The medical staff management platform updates Li's "medication taken" status in real time. If Li does not take his medication between 07:00 and 07:30, the system will push a "missed dose warning" to the medical staff at 07:30, and the medical staff can proactively contact Li to confirm.

[0139] II. Experimental Procedure

[0140] 2.1 Experimental Design

[0141] Experimental subjects: 60 patients from a certain HIV treatment center were selected. Inclusion criteria:

[0142] Age 22-65 years, diagnosed with HIV ≥1 year ago;

[0143] Take a fixed regimen of antiviral drugs (such as "rectilavir potassium tablets + emtricitabine tenofovir tablets") for ≥3 months;

[0144] No severe cognitive impairment (MMSE score ≥24 points), and possesses basic smartphone operation skills;

[0145] Sign an informed consent form, agreeing to participate in a 3-month experiment.

[0146] Grouping method: The "self-controlled before-and-after" design was adopted (to avoid the influence of individual differences). The first month was the "baseline period" (using traditional management methods: paper medication log + monthly follow-up), and the following two months were the "intervention period" (using this system).

[0147] Experiment period: March 1, 2024 to May 31, 2024 (3 months in total).

[0148] 2.2 Experimental Procedure

[0149] Step 1: Baseline period data collection (March 1st to March 31st)

[0150] Patients are provided with paper medication logs and are required to record the time and dosage of their medications daily.

[0151] At the end of each month, an offline follow-up visit will be conducted to collect logs and calculate the adherence rate (number of times medication was taken on time / total number of times medication was taken).

[0152] Liver function indicators (ALT, AST) are collected once every 2 weeks, and the time when abnormal indicators are discovered is recorded (the time from the abnormality of the indicators to medical intervention).

[0153] The baseline missed dose rate (number of missed doses / total number of doses) and the self-reported deviation rate (number of deviations between log records and actual doses / total number of doses, verified with the assistance of a home camera) were calculated.

[0154] Step 2: System Deployment and Intervention Data Collection (April 1st to May 31st)

[0155] Equipment deployment: Smart pillboxes (with pre-stored patient facial templates) and Huawei Watch GT3 watches were distributed to 60 patients, and they were instructed to install the "HIV Smart Medication Manager" APP and complete the privacy permission settings; management platform accounts were configured for medical staff (permissions were assigned according to professional titles: attending physicians could view all data, while nurses could only view compliance reports).

[0156] Data Acquisition: The system automatically collects drug administration behavior data (timeliness rate, dosage accuracy rate), physiological data (heart rate, blood oxygen, liver function), and scenario data (home / remote), and uploads them to the cloud in real time;

[0157] Intervention implementation: For patients with "moderate / poor" compliance (score <75 points), medical staff generate personalized intervention plans through the management platform (e.g., "poor" patients receive an additional weekly telephone follow-up, and "moderate" patients receive improvement suggestions).

[0158] Security monitoring: Daily checks of system privacy protection logs to record any data leaks, unauthorized access, or other incidents.

[0159] Step 3: Indicator Statistics and Analysis

[0160] The core indicators for the baseline period and the intervention period were statistically analyzed separately: compliance rate, missed dose rate, self-reported bias rate, intervention time for abnormal liver function, and number of privacy and security incidents.

[0161] Statistical analysis was performed using SPSS 26.0 software. Quantitative data were expressed as mean ± standard deviation (x ± s). Paired t-tests were used for before-and-after comparisons, and P < 0.05 was considered statistically significant.

[0162] 2.3 Monitoring Indicators

[0163] Key efficacy endpoints: medication adherence (primary endpoint), missed dose rate;

[0164] Secondary efficacy endpoints: self-reported bias rate, time to intervention for abnormal liver function;

[0165] Security metrics: number of privacy and security incidents (such as data breaches and unauthorized access), and device malfunctions (such as the number of times the smart pillbox fails).

[0166] III. Experimental Data and Results Analysis

[0167] 3.1 Comparison of core indicators (n=60)

[0168]

[0169] Results Explanation:

[0170] The intervention significantly improved medication adherence (from 72.3% to 96.8%) and reduced the missed dose rate to 2.1%, with statistically significant differences (P < 0.001), indicating that the system can effectively improve patient adherence.

[0171] The self-reported bias rate dropped from 25.7% to 1.2% because the system uses passive data collection (no manual recording is required), completely eliminating bias caused by subjective concealment or forgetting.

[0172] The intervention time for abnormal liver function has been shortened from 48.2 hours to 6.5 hours. Because the system synchronizes physiological data in real time, medical staff can quickly detect abnormalities and intervene.

[0173] No privacy or security incidents occurred during the experiment. Only one instance of data upload delay occurred due to the smart pillbox running out of power (it returned to normal after the battery was replaced). The device security was good.

[0174] 3.2 Typical Case Analysis

[0175] Case subject: Patient Wang (male, 45 years old, diagnosed with HIV 3 years ago, taking "Ralteiravir potassium tablets 400mg bid", baseline adherence rate 68.5%, missed dose rate 22.3%).

[0176] Baseline issues: Due to frequent business trips (2-3 times per month), Mr. Wang often missed taking his medication due to scheduling conflicts; and because he was worried about liver damage side effects, he occasionally reduced the dosage on his own (dosage accuracy rate 75.0%), but did not inform the medical staff, with a self-reported deviation rate of 32.1%.

[0177] Improved intervention:

[0178] On April 15, Mr. Wang went on a business trip to Guangzhou. The system used GPS to identify the location and moved the reminders from 08:00 and 20:00 to 07:00 and 19:00, respectively, and pushed them to the APP and smart pillbox. Mr. Wang took his medicine on time and did not miss a dose.

[0179] On April 20, when the system synchronized liver function data, it found that ALT=52U / L (exceeding the normal range). An alert was immediately pushed to the medical staff. The attending physician contacted Wang within 2 hours and adjusted the medication dosage to 300mg bid. One week later, ALT dropped to 38U / L.

[0180] Within two months of intervention, Wang's compliance rate rose to 98.3%, the missed dose rate was 0%, the dosage accuracy rate was 100%, the self-reported deviation rate was 0%, and the compliance level remained stable at "excellent".

[0181] 3.3 Statistical Validation

[0182] A paired t-test was performed on the compliance rate of 60 patients. The mean at baseline was 72.3%, the mean at intervention was 96.8%, the mean difference was 24.5%, the standard deviation was 6.3%, t=29.87, P<0.001, indicating that the compliance rate at intervention was significantly higher than that at baseline, and the systematic effect was statistically significant.

[0183] The above embodiments are merely preferred embodiments of the present invention and should not be construed as limiting the scope of protection of the present invention. Any non-substantial changes and substitutions made by those skilled in the art based on the present invention shall fall within the scope of protection claimed by the present invention.

Claims

1. A smart medication adherence management system for HIV patients, characterized in that, The system includes an intelligent sensing layer, a data processing layer, and an application service layer. Each layer interacts with the other through Internet of Things (IoT) communication protocols (preferably Bluetooth 5.0 or NB-IoT). The system is designed to address issues such as self-reporting bias, reliance on active operation, lack of drug storage monitoring, and insufficient privacy protection in HIV patient medication adherence management. The intelligent sensing layer includes a multimodal data passive acquisition module, which integrates Internet of Things (IoT) devices (smart pillboxes, wearable devices, home scene sensors) and non-invasive monitoring technologies (facial recognition, weight sensing, and non-intrusive acquisition of physiological parameters) to passively collect multi-dimensional data related to HIV patient medication, without requiring patients to actively press buttons or manually input operations. The data processing layer is deployed on a cloud server and includes a data cleaning submodule, a data integration submodule, and an analysis submodule. The data cleaning submodule is used to remove abnormal data (such as instantaneous fluctuation data from weight sensors and offline invalid data from wearable devices). The data integration submodule is used to associate multi-source data by timestamp. The analysis submodule is used to generate patient medication adherence assessment results and risk warning information. The application service layer includes a patient-side APP, a medical staff-side management platform, and a smart medicine box local interaction unit, which are used to provide management services based on the analysis results of the data processing layer and meet privacy protection requirements through encryption technology. The multimodal data passive acquisition module includes an upgraded smart pillbox, which comprises: Face recognition module: It adopts a deep learning-based face recognition algorithm (referencing the facial feature extraction architecture of CN209884764U, adding a liveness detection function, and verifying through blinking and head shaking to avoid photo deception). It pre-stores patient facial feature templates, and collects the facial image of the person picking up the medicine in real time when picking up the medicine, and compares it with the template (the comparison threshold is ≥95% to be considered successful). If the comparison is successful, the medicine box can be opened; if the comparison fails, an identity abnormality warning is triggered. Weight sensing module: Employs a high-precision pressure sensor (detection accuracy ±0.1mg) to monitor the total weight of the medication in the pillbox in real time. It identifies the entire "medication dispensing-administration" process through continuously collected weight change curves: When the weight reduction equals the single prescription dose (preset in the system), it is determined as "valid medication dispensing"; if the weight reduction is not equal to the single dose or there is no secondary weight change within 10 minutes after dispensing (excluding medication being returned), it is determined as "abnormal medication administration". Temperature and humidity sensing module: It adopts a digital temperature and humidity sensor (temperature measurement range -10℃~60℃, accuracy ±0.5℃; humidity measurement range 0~100%RH, accuracy ±3%RH), and presets suitable storage thresholds for HIV drugs (temperature 2~8℃ and humidity 30%~60%RH in a light-protected environment; the specific threshold can be adjusted according to the type of drug). When the monitored value exceeds the threshold for more than 5 minutes, an abnormal storage warning is triggered. The triggering methods for the identity anomaly warning and storage anomaly warning include: the smart medicine box emits a buzzer (frequency 2kHz, lasting 10 seconds) and flashes an LED light (red, at 1-second intervals) locally, while simultaneously pushing text warning information to the patient's APP.

2. The intelligent drug administration compliance management system for HIV patients according to claim 1, characterized in that, The multimodal data passive acquisition module also includes a wearable device linkage unit. The wearable device linkage unit establishes a stable connection with a smartwatch / bracelet via Bluetooth 5.0 protocol to collect the patient's physiological data in real time. The collection frequency is: heart rate once per minute, blood oxygen once every 5 minutes, and sleep stages once every 30 seconds. The wearable device linkage unit uploads the collected physiological data to the data processing layer in batches of 15 minutes. Before uploading, CRC verification is used to ensure data integrity.

3. The intelligent drug administration compliance management system for HIV patients according to claim 2, characterized in that, The wearable device linkage unit is also compatible with a portable liver function testing device, which uses a dry biochemical sensor (based on the principle of enzyme-catalyzed reaction-optical colorimetry, with ≥2 detection channels) to detect the patient's ALT (alanine aminotransferase) and AST (aspartate aminotransferase) levels. The wearable device linkage unit connects to the liver function testing device via a USB-Type-C interface and automatically synchronizes the test data according to a preset cycle (default once a week, which can be adjusted by the medical staff to every 3 days / every 2 weeks). After synchronization, the data is associated with the patient's medication record for that day. The analysis submodule of the data processing layer presets the normal reference range for ALT to 0~40U / L and the normal reference range for AST to 0~40U / L. When the test value exceeds the range, it is marked as "abnormal liver function index" and pushed to the medical staff management platform.

4. The intelligent drug administration compliance management system for HIV patients according to claim 1, characterized in that, The multimodal data passive acquisition module further includes an environment and behavior perception unit, which includes: Home scene sensors include smart door lock sensors (detecting door lock open / close status and opening / closing time, sampling once per second) and refrigerator door magnetic sensors (detecting refrigerator door open / close status and continuous opening time, sampling once per second). When the smart door lock has no opening record for 8 consecutive hours and the refrigerator door is opened ≥3 times per day, the patient is determined to be in "home status". Conversely, if the smart door lock is opened ≥2 times in 24 hours and the refrigerator door is opened ≤1 time, the patient is initially determined to be "not at home". GPS positioning module: Integrated into the patient's app, it uses differential positioning technology (positioning accuracy of 5-10 meters). After collecting the patient's location information, it performs privacy desensitization through "coordinate offset + regional coding" (offsetting the actual latitude and longitude by 50-100 meters, or converting it to the city-level administrative region code, without retaining the precise coordinates). The environment and behavior perception unit uploads the "home / non-home" determination result and the desensitized location information to the data processing layer every hour. If the desensitized location is detected to be ≥50 kilometers away from the patient's permanent residence (preset in the system) for more than 24 hours, it is determined to be a "business trip / medical treatment scenario".

5. The intelligent drug administration compliance management system for HIV patients according to claim 4, characterized in that, The analysis submodule of the data processing layer includes a scene analysis subunit. After receiving the scene determination results from the environment and behavior perception unit, the scene analysis subunit generates a reminder strategy adjustment plan based on the patient's preset medication plan (including daily medication time, dosage, and drug type). For "business trip / medical treatment in another city": advance the medication reminder time by 1 hour (to avoid travel conflicts) and automatically adjust the reminder time according to the time zone of the other city (e.g., if the patient is traveling from Beijing to New York, the reminder time will be adjusted according to the New York time zone). If the patient is in "home status": the original medication schedule will be followed. If the smart lock detects that the patient has left the house (the door is open and not closed), a second reminder will be sent after a 10-minute delay (to avoid missing the patient). The notification strategy adjustment plan is pushed to the patient-side APP and smart pillbox in real time at the application service layer.

6. The intelligent drug administration compliance management system for HIV patients according to claim 1, characterized in that, The analysis submodule of the data processing layer also includes a compliance assessment subunit. This compliance assessment subunit uses a 7-day assessment cycle and constructs an assessment model using the following multi-dimensional data as input parameters: Dosing behavior data: Dosing on time rate (the deviation between the actual dosing time and the planned time is ≤30 minutes is considered on time), Dosing accuracy rate (the actual amount of medicine dispensed is consistent with the planned dose is considered accurate). Physiological data: Number of abnormal heart rate / blood oxygen levels collected by wearable devices, and whether liver function test indicators are normal; Scenario matching data: the degree of matching between drug administration behavior and "home / remote" scenarios (e.g., administering drugs according to the adjusted reminder time in a remote scenario is considered a match); The assessment model divides compliance levels into four levels: Excellent (overall score ≥ 90 points), Good (75~89 points), Average (60~74 points), and Poor (< 60 points). It generates a compliance assessment report containing scores for each parameter and explanations of anomalies, which is pushed to the application service layer every 7 days.

7. The intelligent drug administration compliance management system for HIV patients according to claim 6, characterized in that, The patient-side APP of the application service layer includes the following functional modules: The reminder module receives and displays medication reminders (text + ringtone reminders) and abnormal warnings (storage abnormalities, identity abnormalities, and physiological indicator abnormalities), and supports patients in viewing historical reminder records. Report module: Displays individual adherence assessment reports, marks abnormal items (such as "Insufficient medication dosage at 18:00 on 2024-XX-XX, dosage accuracy 80%), and provides improvement suggestions (such as "It is recommended to check the single dose shown on the medicine box when picking up the medication"). Privacy settings module: Supports access permissions for patient management data (e.g., allowing only the attending physician to view liver function data and denying access to other personnel); The application service layer's healthcare management platform includes a patient data dashboard (displaying multi-dimensional data categorized by patient ID), an assessment report viewing module, and an intervention plan generation module. Healthcare staff can generate personalized intervention plans based on assessment reports (such as increasing weekly follow-ups for patients with poor compliance and adjusting medication dosages for patients with abnormal liver function) and push the plans to the patient's app.

8. The intelligent drug administration compliance management system for HIV patients according to claim 7, characterized in that, The application service layer also includes a privacy protection module, which employs a full-link encryption mechanism for transmission, storage, and access. Data transmission encryption: All uploaded / downloaded data is encrypted using the AES-256 symmetric encryption algorithm, and the key is automatically updated every 24 hours; Data storage encryption: Patient personal information (name, ID number, HIV condition information) is stored on a cloud server using the RSA asymmetric encryption algorithm, and only the encrypted ciphertext is retained; Access control: Patients access personal data through dual authentication of "fingerprint recognition + APP password", and medical staff verify access rights through "employee ID + dynamic password (updated every 60 seconds)". The system records all access logs (including access personnel, time and data type), and the logs are retained for ≥1 year.

9. The intelligent drug administration compliance management system for HIV patients according to claim 1, characterized in that, The facial recognition module of the smart pillbox also includes an anomaly handling unit: when the comparison fails three times in a row, the pillbox is automatically locked (locked for 30 minutes), and a "multiple identity verification failures" warning message is pushed to the medical staff management platform; if the patient's facial condition changes (such as wearing a mask or slight facial swelling) and the comparison fails, the patient can send an "identity reset application" through the patient's APP, and the facial feature template will be updated after the medical staff approves it.

10. The intelligent drug administration compliance management system for HIV patients according to claim 1, characterized in that, The weight sensing module of the smart pillbox also includes an anomaly analysis unit: when the following anomalies are detected, a corresponding anomaly warning is triggered and uploaded to the data processing layer: Abnormal medication dispensing: Medication dispensing amount < 80% of the single prescription dose (judged as "insufficient dose") or > 120% (judged as "excessive dose"); Delayed medication: No change in weight within 30 minutes after medication is obtained (judged as "medication not taken after obtaining medication"); Medication Remaining Abnormality: The amount of medication remaining in the pillbox is less than the total prescription dose for 3 days (judged as "insufficient medication"), triggering a "replenish medication" reminder; The abnormal warning information is simultaneously pushed to the patient's mobile app and the medical staff's management platform. The medical staff can proactively contact the patient to confirm the situation based on the warning information.

Images