A home mental health monitoring and closed-loop care support system

Abstract

The application discloses a kind of home mental health monitoring and closed loop nursing support system, it is related to intelligent medical treatment and psychological health home nursing technical field, through intelligent mattress collection sleep interruption and off-bed data, through door magnet collection entrance opening data, through wearable equipment collection heart rate and heart rate variability data, through voice terminal collection voice emotion data;With off-bed event as time anchor point, carry out multi-source data synchronization, abnormality elimination and night state fragment chain construction, generate home psychological state feature;Again with patient historical baseline comparison, identify continuous wakefulness, heart rate variability anomaly, long time off-bed, door magnet repeatedly open and the silence psychological crisis event formed without active help-seeking, and output risk level;According to risk level, push mindful audio, CBT thinking record question or voice pacification prompt, the application realizes the monitoring, intervention, reevaluation and linkage closed loop of home psychological crisis, to improve night crisis response efficiency.

Description

Technical Field

[0001] This invention relates to the fields of smart healthcare and home-based mental health care, specifically to a home-based mental health monitoring and closed-loop care support system. Background Technology

[0002] Current home-based mental health care relies heavily on patient self-reporting, caregiver observation, or single-device data collection of sleep and heart rate. This makes it difficult to promptly detect "silent psychological crises" that are less common but have serious consequences in nighttime home settings. For example, patients may repeatedly wake up late at night and wander out of bed for extended periods, approaching entrances and exits, exhibiting persistently abnormal heart rate variability, without actively calling for help or filling out questionnaires. Meanwhile, caregivers may be asleep, making it impossible for the system to determine whether the patient is at risk of wandering off, self-harm, or experiencing an acute emotional breakdown. At the same time, existing digital treatment tools are disconnected from real-time monitoring data. CBT mindfulness recording, mindfulness training, caregiver guidance, and community / hospital assistance cannot be triggered continuously based on the same abnormal event, which can easily lead to delayed intervention, vague care recommendations, and incomplete emergency response information.

[0003] Therefore, there is a need for a home-based mental health monitoring and care support system that can form a closed loop in the home setting through multi-source non-intrusive monitoring, intelligent risk identification, proactive intervention feedback, and emergency linkage. Summary of the Invention

[0004] The purpose of this invention is to provide a home-based mental health monitoring and closed-loop care support system to address the shortcomings of the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a home-based mental health monitoring and closed-loop care support system, comprising:

[0006] The IoT monitoring module collects sleep interruption and bed exit data through a smart mattress, entrance and exit opening data through a door sensor, heart rate and heart rate variability data through wearable devices, and voice emotion data through a voice terminal.

[0007] The edge preprocessing module performs time synchronization and anomaly removal on the collected data, and generates home psychological state characteristics within the same nighttime time window.

[0008] The psychological risk level generation module compares the characteristics of the patient's psychological state at home with the patient's historical baseline. When it identifies continuous awakening, abnormal heart rate variability, getting out of bed beyond the threshold, or repeated opening of the door magnet without active calling for help, it generates silent psychological crisis events and risk levels.

[0009] The digital intervention module outputs mindfulness audio, CBT thinking record questions, or voice reassurance prompts according to the risk level, and collects patient responses, emotional changes, bed status, and heart rate variability changes to form post-intervention status characteristics.

[0010] The psychological risk assessment module reassesses the risk based on the post-intervention status characteristics. If the risk is reduced, the caregiver support module will push home observation suggestions. If the risk is not reduced, an escalation treatment instruction will be generated.

[0011] The emergency response module sends the patient's address, anomaly summary, risk level, and intervention details to the community or hospital based on the upgraded treatment instructions.

[0012] Preferably, the data collection process of the IoT monitoring module includes the following steps:

[0013] The smart mattress determines the bedtime, bedtime, number of nighttime awakenings, and duration of sleep interruption based on changes in bed pressure and body movement, and generates a bed-out event when the patient transitions from a bed-lying state to a bed-out state.

[0014] When the door sensor detects that the entrance / exit has changed from a closed state to an open state, it generates a door sensor opening event and records the opening time, opening duration, number of openings, and the corresponding entrance / exit location.

[0015] Wearable devices continuously collect heart rate and heart rate variability data before and after sleep interruption, bed exit events, or door magnetic sensor opening events, and generate physiological stress data when heart rate variability remains abnormal.

[0016] When both bed-off events and physiological stress data are present, the voice terminal outputs voice questions and generates voice emotion data based on the patient's response status, speech rate, volume, pause duration, and emotional tendency.

[0017] Preferably, the time synchronization of the edge preprocessing module includes the following steps:

[0018] The system receives monitoring data packets sent by smart mattresses, door sensors, wearable devices, and voice terminals. The monitoring data packets include device identifiers, collection time, event types, and collection values.

[0019] Using the start time of leaving the bed and the time of returning to the bed in the bed-leaving event as time anchors, the entrance and exit opening data, heart rate and heart rate variability data and voice emotion data are written into the same event cache queue;

[0020] The time between leaving the bed is extended forward by 120 seconds and the time returning to the bed is extended backward by 300 seconds, forming a preceding and following adjacent interval. Data falling within the preceding and following adjacent intervals are arranged in the order of occurrence to obtain the calibrated monitoring event sequence.

[0021] Preferably, the anomaly removal process performed by the edge preprocessing module includes the following steps:

[0022] Determine whether the entry / exit opening data falls within the adjacent interval before and after any bed leaving event. If it does not fall within the interval, mark the entry / exit opening data as a low-confidence event.

[0023] Determine whether the time interval between heart rate and heart rate variability data and sleep interruption and bed exit data exceeds 300 seconds. If it does, mark the heart rate and heart rate variability data as a low-confidence event.

[0024] If, within 600 seconds of a low-confidence access point opening data event, the voice terminal does not respond, displays negative voice emotion, or the user leaves the bed again, the access point opening data event is converted into a valid event.

[0025] If, within 300 seconds of a low-confidence heart rate and heart rate variability data point, the duration of time spent out of bed exceeds 600 seconds or the data point for opening an entrance or exit is obtained, the heart rate and heart rate variability data point will be converted into a valid event. Low-confidence events that are not verified will not participate in the formation of home-based psychological state characteristics.

[0026] Preferably, the edge preprocessing module generates home-based psychological state characteristics, specifically including the following steps:

[0027] The purified monitoring event sequence was spliced ​​together in the order of occurrence of a sleep interruption, a continuous process of getting out of bed, the corresponding opening status of the entrance and exit, the corresponding heart rate and heart rate variability, and the corresponding voice emotion response state to form a nighttime state segment.

[0028] The start time of sleep interruption is used as the starting point of the nighttime state segment, and the latest of the following—the time of returning to bed, the time of the end of voice interaction, or the time of the last valid event—is used as the end point of the segment.

[0029] Arrange multiple nighttime state segments from morning to night according to their starting points to form a nighttime state segment chain;

[0030] Continuous arousal features, persistent bed-leaving features, entrance / exit proximity features, persistent physiological stress features, and abnormal voice response features are extracted from the nighttime state fragment chain and encapsulated into home psychological state features according to the chronological order of the nighttime state fragment chain.

[0031] Preferably, the psychological risk level generation module establishes a patient's historical baseline and compares it with similar time periods, specifically including the following steps:

[0032] Data from the patient’s most recent 14 effective nighttime windows were used as historical baseline data. An effective nighttime window is defined as a nighttime record in which there are continuous data collection records from smart mattresses, door magnets, and wearable devices, and the duration of missing data for any single type does not exceed 10% of the total duration of the nighttime window.

[0033] Divide the nighttime time window into several similar time periods;

[0034] The median values ​​of sleep interruption, heart rate variability, duration of getting out of bed, and number of times the entrance / exit was opened were taken within each similar time period to obtain the continuous wakefulness baseline, heart rate variability baseline, duration of getting out of bed baseline, and entrance / exit opening baseline.

[0035] By comparing the corresponding features in the current nighttime state segment chain with the historical baseline of the same time period, the results of continuous wakefulness deviation, physiological stress deviation, bed exit deviation, and entrance / exit deviation are obtained.

[0036] Preferably, the psychological risk level generation module generates silent psychological crisis events by including the following steps:

[0037] The nighttime state segment that first satisfies the abnormal result of continuous wakefulness deviation is identified as the initial abnormal segment;

[0038] Following the sequence of nighttime state fragments, search for fragments with abnormal heart rate variability within 30 minutes after the initial abnormal fragment, search for fragments with persistent abnormality after leaving the bed within 30 minutes after the abnormal heart rate variability fragment, and search for fragments with repeated opening and closing of entry and exit points within 10 minutes after the fragments with persistent abnormality after leaving the bed.

[0039] When continuous awakening, abnormal heart rate variability, getting out of bed beyond the threshold, and repeated opening of the door sensor occur in the above sequence, a nighttime crisis behavior chain is formed.

[0040] Verify whether there are any records of voice calls for help, button calls for help, or mobile calls for help within the same nighttime time window. If not, mark the nighttime crisis behavior chain as a silent psychological crisis event and write the abnormal segments into the event summary in the order of occurrence.

[0041] Preferably, the psychological risk level generation module generates risk levels by including the following steps:

[0042] A Level 1 risk is generated when the event summary contains only continuous arousal abnormalities and heart rate variability abnormalities, and the interval between the two is no more than 30 minutes.

[0043] Based on Level 1 risk, Level 2 risk is generated when the duration of a single bed leave is not less than 900 seconds, or when the current bed leave duration increases by not less than 600 seconds relative to the bed leave duration baseline.

[0044] Based on the level 2 risk, a level 3 risk is generated when the entrance / exit is opened no less than twice within 600 seconds after getting out of bed and there is no record of active distress calls.

[0045] The risk level and event summary are sent together to the digital intervention module, which then selects the corresponding intervention content according to the risk level.

[0046] Preferably, the digital intervention module generates post-intervention state characteristics by including the following steps:

[0047] When the risk level is Level 1, output a 5-minute mindfulness audio; when the risk level is Level 2, output 3 CBT thought recording questions related to sleep interruption, reasons for getting out of bed, and current thoughts; when the risk level is Level 3, output reassuring voice prompts including identity verification, current location verification, and safety guidance.

[0048] An intervention start marker is written when the intervention content is first output. The intervention start marker includes the intervention start time, risk level, event summary number, and intervention content type.

[0049] Starting from the intervention initiation marker, patient responses, changes in verbal emotion, return-to-bed status, and changes in heart rate variability were collected within 15 minutes. Under level 3 risk, the 15 minutes were divided into three consecutive 5-minute segments.

[0050] The patient's response, mood improvement, return-to-bed status, heart rate variability, and physiological stress mitigation results were written into the same intervention feedback segment and associated with the pre-intervention home psychological state characteristics according to the event summary number to form the post-intervention state characteristics.

[0051] Preferably, the psychological risk assessment module and the emergency response module, based on the post-intervention state characteristics, implement closed-loop management, specifically including the following steps:

[0052] The psychological risk assessment module determines whether the post-intervention status characteristics simultaneously meet the criteria of timely or delayed response, emotional relief, stable return to bed, relief of physiological stress, and no re-opening of the entrance / exit during the reassessment period. If these criteria are met, the risk is considered reduced, and the caregiver support module will push home observation recommendations.

[0053] If any two of the following two conditions are present after intervention: no response, worsening of emotions, failure to return to bed, failure to alleviate physiological stress, or re-opening of the entrance / exit during the reassessment period, the risk is deemed not to have been reduced, and an escalation treatment instruction is generated.

[0054] The emergency response module generates a tiered rescue information package based on the upgraded response instructions. The first-level information package includes the patient number, home address, current risk level, and time of abnormal occurrence. The second-level information package includes an event summary, intervention content, and post-intervention status characteristics. The third-level information package includes suggested communication scripts, a summary of the nighttime baseline for the past 14 days, and emergency contact information.

[0055] The emergency linkage module first sends a Level 1 information packet to the community or hospital, and then sends a Level 2 information packet after receiving confirmation. Before the rescuers arrive, it continuously updates information such as whether the patient has returned to bed, whether the door sensor has been triggered again, whether there is a voice response, and whether the heart rate variability has recovered.

[0056] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0057] This invention uses sleep interruption and bed-leaving events collected by a smart mattress as time anchors. Data collected from door sensors, wearable devices, and voice terminals are written into the same event cache queue. Through adjacent intervals, subsequent verification of low-confidence events, and nighttime state fragment chains, the sequence "continuous awakening - physiological stress - prolonged bed-approaching entrances / exits - abnormal voice" is constructed into a traceable continuous event. This solves the problems of fragmented multi-source data, false triggering of door sensors, and misjudgments caused by momentary drift of wristbands in existing technologies. The system can still identify silent psychological crisis precursors at night even when patients do not actively fill out questionnaires or seek help, improving the accuracy and continuity of home monitoring.

[0058] This invention further directly links risk levels with event summaries and incorporates intervention initiation markers when outputting mindfulness audio, CBT thought recording questions, or voice reassurance prompts. Subsequently, it collects patient responses, changes in voice and emotion, bed-return status, and heart rate variability changes to form post-intervention status characteristics, which are then used to determine whether home observation or escalation of treatment is necessary. This closed-loop approach solves the problems in existing technologies where intervention content is disconnected from real-time risk, caregivers lack clear treatment guidelines, and information received by communities or hospitals is incomplete. It enables psychological intervention results to participate in risk reassessment and sends the patient's address, abnormal summary, risk level, and intervened content in a structured manner to external assistance, thereby improving the efficiency of nighttime crisis response. Attached Figure Description

[0059] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0060] Figure 1 This is a flowchart of a home-based mental health monitoring and closed-loop care support system according to the present invention. Detailed Implementation

[0061] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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.

[0062] Example 1, please refer to Figure 1 As shown in this embodiment, a home-based mental health monitoring and closed-loop care support system includes:

[0063] The IoT monitoring module collects sleep interruption and bed exit data through a smart mattress, entrance and exit opening data through a door sensor, heart rate and heart rate variability data through wearable devices, and voice emotion data through a voice terminal.

[0064] In one specific embodiment, the IoT monitoring module is installed in the patient's home environment to continuously collect data on the patient's nighttime sleep status, out-of-bed behavior, entrance and exit activities, physiological stress state, and vocal emotional state without relying on the patient actively filling out questionnaires or reporting. The IoT monitoring module includes a smart mattress, door magnetic sensors, wearable devices, and a voice terminal. Each device collects different types of home status data and transmits the results wirelessly to an edge preprocessing module for subsequent formation of unified home psychological state characteristics. This setup corresponds to the technical approach described in the document of "collecting data using various wearable devices such as smart mattresses, door magnetic sensors, and wristbands, as well as voice interaction terminals."

[0065] Specifically, the smart mattress is installed on the patient's usual sleeping bed and contains a pressure sensing unit, a body movement detection unit, and a sleep state recognition unit. The pressure sensing unit detects whether the patient is in bed, the body movement detection unit detects the patient's turning over, limb movements, or abnormally frequent body movements during sleep, and the sleep state recognition unit determines the patient's bed-in time, bed-out time, continuous bed-lying duration, number of nighttime awakenings, and duration of sleep interruptions based on pressure and body movement changes. When the smart mattress detects that the patient has changed from a bed-lying state to a bed-out state within a preset nighttime period, it generates a bed-out event. When a bed-out event occurs multiple times in a short period of time, or when the patient does not return to a bed-lying state within a preset time after getting out of bed, the smart mattress marks the bed-out event as abnormal bed-out data and sends it along with the occurrence time to the edge preprocessing module.

[0066] The door magnetic sensor is installed at the patient's bedroom door, front door, balcony door, or other entrances / exits that may form a path out of the home, to collect the door's open and closed status. When the door magnetic sensor detects a change from a closed to an open state, it generates a door magnetic opening event and records the opening time, duration, number of openings, and corresponding door location. If multiple door magnetic openings are detected within the same nighttime time window, or if the front door or balcony door is detected to be open after the smart mattress has generated abnormal bed-leaving data, the door magnetic sensor marks the door magnetic opening event as abnormal entrance / exit activity data. Therefore, the data collected by the door magnetic sensor can be temporally correlated with the bed-leaving data collected by the smart mattress to determine whether the patient is engaging in risky behavior such as approaching or attempting to pass through entrances / exits late at night after leaving bed.

[0067] The wearable device, worn on the patient's wrist or other suitable location for collecting physiological signals, includes a heart rate acquisition unit and a heart rate variability analysis unit. The heart rate acquisition unit collects the patient's heart rate data at a preset sampling frequency, while the heart rate variability analysis unit obtains heart rate variability data based on changes in adjacent heartbeat intervals and generates physiological indicators reflecting the patient's physiological stress state. Specifically, the wearable device continuously collects heart rate and heart rate variability data before and after events such as sleep interruption, getting out of bed, or door magnetic sensor opening. When the patient's heart rate is higher than their historical nighttime baseline, or their heart rate variability is lower than their historical nighttime baseline and persists for more than a preset time, the wearable device generates abnormal physiological stress data. This abnormal physiological stress data, along with the bed-getting data and door magnetic sensor opening data, is used to determine whether the patient is in a state of tension, fear, anxiety, or other abnormal psychological state.

[0068] The voice terminal is positioned in the patient's bedroom, living room, or near the bedside and can collect the patient's voice through wake words, proactive inquiries, or triggering by abnormal events. The voice terminal includes a microphone array, a voice recognition unit, and a voice emotion analysis unit. The microphone array collects the patient's voice signal, the voice recognition unit identifies whether the patient responds and the content of the response, and the voice emotion analysis unit generates voice emotion data based on speech rate, volume, pause duration, tone changes, and keyword information. For example, when the smart mattress detects that the patient has been out of bed for an extended period at night and the wearable device detects abnormal heart rate variability, the voice terminal can automatically output a voice inquiry such as "Do you need help?" or "Are you feeling unwell?" If the patient does not respond, the response time is too long, the voice volume is significantly reduced, the speech rate is abnormally fast, or negative emotion keywords appear, the voice terminal generates abnormal voice emotion data and sends it to the edge preprocessing module.

[0069] Furthermore, the smart mattress, door magnetic sensor, wearable device, and voice terminal all possess unique device identifiers and timestamp recording capabilities. When collecting data, each device encapsulates the device identifier, collection time, event type, and collected value into a monitoring data packet. For example, the monitoring data packet generated by the smart mattress includes the patient identifier, mattress device identifier, bed exit start time, bed exit duration, and body movement intensity; the monitoring data packet generated by the door magnetic sensor includes the door magnetic device identifier, door position, opening time, and number of openings; the monitoring data packet generated by the wearable device includes heart rate value, heart rate variability value, and duration of continuous abnormality; and the monitoring data packet generated by the voice terminal includes voice response status, voice emotion tendency, and voice interaction result. These monitoring data packets are transmitted to the edge preprocessing module via Bluetooth, Wi-Fi, ZigBee, cellular networks, or other low-power wireless communication methods.

[0070] The edge preprocessing module performs time synchronization and anomaly removal on the collected data, generating home psychological state characteristics within the same nighttime time window.

[0071] The edge preprocessing unit receives monitoring data packets from the smart mattress, door magnetic sensor, wearable device, and voice terminal. Each monitoring data packet includes the patient identifier, device identifier, acquisition time, event type, and acquisition value. The preferred nighttime time window is from 22:00 to 06:00 the following day; if the median bedtime of a patient over 14 consecutive days is later than 22:00, the start point of the nighttime time window is set to 30 minutes before this median, and the end point is set to 8 hours after the start point. This allows the nighttime monitoring range for different patients to closely match their actual sleep schedules.

[0072] The edge preprocessing unit first performs time synchronization. Using sleep interruption and bed-leaving data collected by the smart mattress as time anchors, entrance / exit opening data, heart rate and heart rate variability data, and voice emotion data within the same nighttime time window are written into the same event cache queue. When the smart mattress generates a bed-leaving event, it records the bed-leaving start time and bed-return time. The bed-leaving duration is obtained by subtracting the bed-leaving start time from the bed-return time. If no bed-return time has been detected, the real-time bed-leaving duration is obtained by subtracting the bed-leaving start time from the current collection time. A contiguous interval is defined by 120 seconds prior to the aforementioned bed-leaving start time and 300 seconds prior to the bed-return time. When no bed-return time is detected, a contiguous interval is defined by 120 seconds prior to the bed-leaving start time and 300 seconds prior to the current collection time. The collection times of the door magnetic sensor, wearable device, and voice terminal are all mapped to this contiguous interval. Data whose collection times fall within this interval are arranged chronologically to form a calibrated monitoring event sequence.

[0073] Anomaly removal is then performed. If entry / exit opening data does not fall within the immediate preceding or following interval of any bed-leaving event, it is initially marked as a low-confidence event and not immediately discarded. Similarly, if the time interval between heart rate and heart rate variability data and sleep interruption and bed-leaving data exceeds 300 seconds, it is also initially marked as a low-confidence event. Heart rate variability is preferably represented by the root mean square of the difference between adjacent normal heartbeat intervals. The calculation first obtains the difference between adjacent normal heartbeat intervals within a consecutive 5-minute period, then squares each difference and calculates the average. Finally, the square root of the average is taken to obtain the heart rate variability value. If the current heart rate variability value is lower than 80% of the median heart rate variability for the same time period over 14 consecutive nights, and this remains true for 180 seconds, then this segment of physiological data is considered to have a continuous relationship.

[0074] Low-confidence events require subsequent data verification before being classified as valid events. The verification rules are as follows: if, within 600 seconds of a low-confidence entrance / exit opening data point, there is no response from the voice terminal, negative emotional tone in the voice, or the individual leaves the bed again, the entrance / exit opening data point is classified as a valid event. Similarly, if, within 300 seconds of a low-confidence heart rate or heart rate variability data point, the individual remains out of bed for more than 600 seconds or an entrance / exit opening data point is recorded, the physiological data is classified as a valid event. Unverified low-confidence events are only recorded and do not participate in the formation of home-based psychological state characteristics. After the above processing, a purified monitoring event sequence is obtained, which reduces interference from daily accidental door triggering, momentary physiological drift caused by loose door fittings, and accidental environmental sound recordings on subsequent judgments.

[0075] Based on the purified monitoring event sequence, the edge preprocessing end concatenates a sleep interruption, a duration of getting out of bed, the corresponding entrance / exit opening status, the corresponding heart rate and heart rate variability changes, and the corresponding voice emotion response status into nighttime state segments according to their occurrence sequence. Each nighttime state segment includes a segment start point, segment end point, number of sleep interruptions, duration of getting out of bed, number of entrance / exit openings, duration of sustained abnormal heart rate variability, voice response status, and voice emotion tendency. The segment start point is the start time of the sleep interruption, and the segment end point is the latest of the following: the time of returning to bed, the time of the end of the voice interaction, or the time of the last valid event. Multiple nighttime state segments are arranged from earliest to latest according to their segment start points, forming a nighttime state segment chain.

[0076] The edge preprocessing unit then extracts home psychological state features from the nighttime state segment chain. Continuous arousal features are obtained from the number of sleep interruptions within the same nighttime time window; persistent bed-out features are obtained by summing the duration of bed-out in each nighttime state segment; entrance / exit proximity features are obtained from the number of entrance / exit openings within 600 seconds of bed-out; persistent physiological stress features are obtained by summing the duration of persistent abnormal heart rate variability; and abnormal voice response features are obtained by summing the number of no-response instances, the number of response delays exceeding 30 seconds, and the number of negative voice emotions. These five types of features are encapsulated into home psychological state features according to the chronological order of the nighttime state segment chain, with corresponding segment start and end points. This allows subsequent psychological risk assessments to directly identify the continuous process of "sleep interruption - bed-entrance / exit proximity - physiological stress - abnormal voice," thus providing a time-consistent, noise-controlled, and traceable data foundation for the identification of silent psychological crisis events.

[0077] The psychological risk level generation module compares the characteristics of the patient's psychological state at home with the patient's historical baseline. When it identifies continuous awakening, abnormal heart rate variability, getting out of bed beyond the threshold, or repeated opening of the door magnet without active calling for help, it generates silent psychological crisis events and risk levels.

[0078] After receiving the home-based psychological state characteristics established in the previous processing phase, the patient's historical baseline is retrieved. The historical baseline is formed from the patient's most recent 14 effective nighttime time windows. An effective nighttime time window is defined as a nighttime record in which continuous data collection occurs from the smart mattress, door magnetic sensor, and wearable device, and the duration of missing data for any single type does not exceed 10% of the total duration of the nighttime time window. If there are fewer than 14 effective nighttime time windows, the most recent 7 effective nighttime time windows are used; if there are fewer than 7, all effective nighttime time windows after enrollment are used, and only observation prompts are output before reaching 7, without outputting level 3 risk.

[0079] Continuous wakefulness baselines were established based on similar time periods. The nighttime time window was divided into four periods: 22:00-00:00, 00:00-02:00, 02:00-04:00, and 04:00-06:00. The number of sleep interruptions recorded in the historical data was counted for each period, and the median was used as the continuous wakefulness baseline. The heart rate variability baseline was represented by the median heart rate variability value within the same time period; the duration of being out of bed baseline was represented by the median duration of being out of bed within the same time period; and the entrance / exit opening baseline was represented by the median number of times entrances / exits were opened within the same time period. Using the median instead of the mean reduces the impact of occasional insomnia, temporary water intake, or family members opening the door on the historical baseline.

[0080] The continuous arousal characteristics, physiological stress persistence characteristics, bed-out persistence characteristics, and entrance / exit proximity characteristics in the current nighttime state segment chain are compared with the historical baselines for the corresponding time periods. The continuous arousal deviation is calculated as follows: the number of sleep interruptions in the current time period minus the continuous arousal baseline, yielding the number of arousal increases; when the number of arousal increases is not less than 2 and the number of sleep interruptions in the current time period is not less than 3, the continuous arousal deviation is considered abnormal. The physiological stress deviation is calculated as follows: the current heart rate variability value divided by the heart rate variability baseline, yielding the heart rate variability maintenance ratio; when this ratio is not higher than 0.8 and persists for more than 180 seconds, the physiological stress deviation is considered abnormal. The bed-out deviation is calculated as follows: the current bed-out duration minus the bed-out persistence baseline, yielding the bed-out increase duration; when the bed-out increase duration is not less than 600 seconds and the current single bed-out duration is not less than 900 seconds, the bed-out deviation is considered abnormal. The method for calculating the deviation of the entrance / exit is as follows: the number of times the entrance / exit is opened within 600 seconds after the current bed exit is subtracted from the baseline number of times the entrance / exit is opened to obtain the number of times it is opened; when the number of times it is opened is not less than 2, and the number of times the entrance / exit is opened within 600 seconds after the current bed exit is not less than 2, the deviation of the entrance / exit is recorded as abnormal.

[0081] After obtaining the four types of deviation results, the initial abnormal segment is searched from the nighttime state segment chain. The initial abnormal segment is the nighttime state segment that first satisfies the abnormal result of continuous arousal deviation. After identifying the initial abnormal segment, the search continues along the chronological order of the nighttime state segment chain for abnormal segments of physiological stress, persistent abnormality of getting out of bed, and repeated opening of entrances and exits. The abnormal segment of physiological stress must appear within 30 minutes after the initial abnormal segment; the persistent abnormality of getting out of bed must appear within 30 minutes after the abnormal segment of physiological stress; and the repeated opening of entrances and exits must appear within 10 minutes after the persistent abnormality of getting out of bed. When the four types of abnormal segments meet the above chronological order, a nighttime crisis behavior chain is formed. This chain is used to express the continuous process of "physiological stress after continuous arousal, followed by prolonged getting out of bed and approaching entrances and exits," avoiding the identification of a single nighttime awakening or a single door opening as a psychological crisis.

[0082] After a nighttime crisis behavior chain is established, it is verified whether there are any records of proactive calls for help within the same nighttime time window. Proactive calls for help records include any of the following semantic expressions recognized by the voice terminal: "Help me," "I feel unwell," or "I need to go to the hospital," a pressed emergency call button, or a request for help issued by the patient's mobile device. The confirmation method for voice calls for help is: the voice text contains any of the above-mentioned semantic expressions for help, and the voice acquisition time is within 300 seconds after the start of the nighttime crisis behavior chain and before the end of the chain. If no such proactive calls for help are found, the nighttime crisis behavior chain is marked as a silent psychological crisis event and written into an event summary in chronological order. The event summary includes the start time of the abnormal segment, the duration of abnormal heart rate variability, the longest single time spent out of bed, the number of times entrances / exits were opened within 600 seconds after getting out of bed, the voice response status, and the proactive call verification result.

[0083] Risk levels are determined progressively based on the completeness of the abnormalities in the event summary. If only continuous arousal abnormalities and physiological stress abnormalities exist, and the interval between the two does not exceed 30 minutes, the risk level is recorded as Level 1. Building upon Level 1 risk, if a single period of being out of bed lasts at least 900 seconds, or if the extended period of being out of bed exceeds 600 seconds, the risk level is recorded as Level 2. Building upon Level 2 risk, if the entrance / exit is opened at least twice within 600 seconds of being out of bed, and there is no record of active distress calls, the risk level is recorded as Level 3. If the patient has active distress call records at any stage, it is not marked as a silent psychological crisis event; instead, the event summary is directly transmitted to the emergency response center or caregiver.

[0084] Through the above processing, the characteristics of home-based psychological states are transformed into silent psychological crisis events and risk levels with chronological order, historical baseline reference, and verification of distress calls. This processing method can distinguish between ordinary nighttime awakenings, brief nighttime wakings, accidental door magnet activation, and silent crisis precursors, enabling subsequent digital interventions to trigger mindfulness audio, CBT thought recording questions, voice reassurance prompts, or escalation measures based on level one, level two, or level three risk, respectively.

[0085] The digital intervention module outputs mindfulness audio, CBT thinking record questions, or voice reassurance prompts based on the risk level, and collects patient responses, emotional changes, bed status, and heart rate variability changes to form post-intervention status characteristics.

[0086] After receiving the risk level and event summary, the digital intervention module first reads the start time of the abnormal segment, the duration of abnormal heart rate variability, the duration of bed alighting, the status of entrance / exit opening, and the voice response status from the event summary, and then selects intervention content based on the risk level. Level 1 risk corresponds to mindfulness audio, with an optimal audio duration of 5 minutes and an initial volume of 70% of the patient's average nighttime voice terminal playback volume over the past 7 days; Level 2 risk corresponds to CBT thought recording questions, prioritizing the output of 3 questions related to sleep interruption, reasons for bed alighting, and current thoughts; Level 3 risk corresponds to voice reassurance prompts, with prompts including 3 categories of content: identity verification, current location verification, and safety guidance, with a 10-second interval between each category to avoid continuous repetition causing stimulation.

[0087] The digital intervention module writes an intervention start marker when sending intervention content. The intervention start marker includes the intervention start time, risk level, event summary number, and intervention content type. The intervention start time is based on the time when the first segment of content is output by the voice terminal or the patient's mobile device; if mindfulness audio is selected, the first second of the audio playback is used as the intervention start time; if CBT thought recording questions are selected, the end time of the first question display or broadcast is used as the intervention start time; if voice reassurance prompts are selected, the end time of the first reassurance prompt broadcast is used as the intervention start time. The edge preprocessing module starts the feedback collection period from the intervention start marker, preferably 15 minutes; under level 3 risk, the feedback collection period is divided into three consecutive 5-minute segments to more quickly obtain changes in the patient's condition.

[0088] During the feedback collection period, the voice terminal records the patient's response. A response within 60 seconds of the intervention start mark is recorded as an timely response; a response after 60 seconds but not exceeding 180 seconds is recorded as a delayed response; and no response after 180 seconds is recorded as no response. Changes in voice emotion are determined by the difference in voice emotion tendency before and after the intervention. Voice emotion tendency is assigned a 5-level scale: stable (0), mild tension (1), anxiety (2), fear (3), and crying or semantic confusion (4). Before the intervention, the most recent voice emotion tendency assignment from the event summary relative to the intervention start mark is used; after the intervention, the last valid voice emotion tendency assignment within the feedback collection period is used. The difference between the former and the latter yields the emotion improvement value. An emotion improvement value of 1 or more is recorded as a alleviated emotion; an emotion improvement value of 0 is recorded as no change in emotion; and an emotion improvement value less than 0 is recorded as an aggravated emotion.

[0089] The smart mattress continuously records the patient's return-to-bed status during the feedback collection period. If the patient transitions from an out-of-bed state to a bed-lying state and remains in bed for at least 300 seconds, the return-to-bed status is recorded as "returned to bed." If the patient only briefly touches the bed surface and remains in bed for less than 300 seconds, the return-to-bed status is recorded as "not stably returned to bed." If no restoration of bed surface pressure is detected, the return-to-bed status is recorded as "not returned to bed." The wearable device synchronously records changes in heart rate variability (HRV). The calculation method is as follows: Let T0 be the time corresponding to the intervention start marker. The heart rate variability data collected in the period from 5 minutes before T0 to T0 are HRV1, HRV2, ..., HRVn, respectively. Then, the pre-intervention value HRV is: HRV = (HRV1 + HRV2 + ... + HRVn) / n; where n is the number of effective heart rate variability data in the 5 minutes before the intervention.

[0090] Heart rate variability (HRV) data collected between 10 and 15 minutes after intervention T0 are HRV'1, HRV'2, ..., HRV'm, respectively. The post-intervention value of HRV is then calculated as: HRV_after = (HRV'1 + HRV'2 + ... + HRV'm) / m; where m is the number of valid HRV data points collected between 10 and 15 minutes after intervention. The change in heart rate variability ΔHRV is calculated using the following expression: ΔHRV = HRV_after - HRV_before; where a positive ΔHRV indicates that the patient's heart rate variability increased after intervention compared to before intervention, and the patient's physiological stress tended to ease; a ΔHRV of 0 indicates that there was no change in heart rate variability before and after intervention; a negative ΔHRV indicates that the patient's heart rate variability decreased after intervention compared to before intervention, and the patient's physiological stress did not ease or even worsened.

[0091] In a preferred embodiment, when ΔHRV > 0 and the HRV reaches more than 80% of the patient's baseline heart rate variability for the same nocturnal period, it is recorded as heart rate variability recovery; otherwise, it is recorded as heart rate variability not recovered.

[0092] When the change in heart rate variability is positive and the value after intervention reaches more than 80% of the baseline heart rate variability for the same time period, it is recorded as physiological stress mitigation.

[0093] The edge preprocessing module writes patient response results, changes in voice and emotion, return-to-bed status, and changes in heart rate variability into the same intervention feedback segment in the order of occurrence after the intervention initiation marker. The intervention feedback segment includes at least the intervention start time, intervention content type, response result, emotion improvement value, return-to-bed status, heart rate variability, and physiological stress mitigation result. Subsequently, the intervention feedback segment is correlated with pre-intervention home psychological state characteristics according to event summary numbers to form post-intervention state characteristics. Post-intervention state characteristics retain pre-intervention characteristics such as continuous arousal, persistent bed-leaving, entrance / exit proximity, persistent physiological stress, and abnormal voice response, and add response results, emotion improvement value, return-to-bed status, heart rate variability, and physiological stress mitigation result, enabling the psychological risk assessment unit to re-determine the risk level based on continuous data before and after the same event. This processing method ensures that mindfulness audio, CBT thought recording questions, and voice reassurance prompts are no longer limited to one-way output, but rather transform the patient's actual response, emotion changes, whether they returned to bed, and physiological stress changes into data that can be reassessed.

[0094] The psychological risk assessment module reassesses the risk based on the post-intervention status characteristics. If the risk is reduced, the caregiver support module will push home observation suggestions; if the risk is not reduced, an escalation instruction will be generated.

[0095] After receiving the post-intervention status characteristics, the psychological risk assessment module matches them with the event summary generated before the intervention and reassesses the risk based on the status changes of the same silent psychological crisis event before and after the intervention. Post-intervention status characteristics include patient response outcomes, changes in voice and emotion, bed rest status, changes in heart rate variability, changes in entrance / exit status, and changes in bed leave persistence. The psychological risk assessment module uses the intervention start marker as the time starting point and selects data within 15 minutes after the intervention as reassessment data. When the risk level is level three, the 15-minute reassessment data is split into three 5-minute segments: the first, second, and third 5-minute segments. The most recent 5-minute segment is used first for status confirmation to avoid delayed handling of nighttime crisis situations.

[0096] During reassessment, the psychological risk assessment module first judges the patient's response. A patient responding verbally within 60 seconds of the intervention initiation marker is recorded as an immediate response; a response between 60 and 180 seconds is recorded as a delayed response; and no response after 180 seconds is recorded as no response. Next, changes in verbal emotion are assessed. For several prognoses, a decrease of one or more levels in verbal emotion compared to pre-intervention levels is recorded as emotional mitigation; no change in level is recorded as emotional instability; and an increase in level is recorded as emotional aggravation. Then, the patient's return to bed status is assessed. If the smart mattress detects the patient's transition from an out-of-bed state to a bedridden state and the patient remains in bed for more than 300 seconds, it is recorded as stable return to bed; less than 300 seconds of continuous bed rest is recorded as brief return to bed; and no restoration of bed pressure is detected, resulting in no return to bed. Finally, changes in heart rate variability are assessed. For several prognoses, an average heart rate variability of more than 80% of the baseline heart rate variability for the same time period between the 10th and 15th minutes, and higher than the average for the 5 minutes before intervention, is recorded as physiological stress mitigation.

[0097] The psychological risk assessment module generates a reassessment result based on the above judgments. When the reassessment result simultaneously meets the criteria of timely or delayed response, emotional easing, stable return to bed, and reduced physiological stress, and the entrance / exit is not reopened during the reassessment period, the risk is considered reduced. If several pre-existing conditions were level three risk, the risk is reduced to level two or level one; if several pre-existing conditions were level two risk, the risk is reduced to level one. After the risk is reduced, the caregiver support module pushes home observation suggestions to the caregiver's mobile device. These suggestions include whether the patient has responded, returned to bed, changes in mood, changes in heart rate variability, suggested observation duration, and the time to review again. The preferred observation duration is 30 minutes, and the preferred time to review again is 15 minutes after the push notification.

[0098] If the reassessment results show any two of the following: no response, worsening emotional state, failure to return to bed, unresolved physiological stress, or re-opening of the exit / exit during the reassessment period, the risk is considered not reduced. Similarly, if the original risk level was Level 3 and either no response or re-opening of the exit / exit occurred, the risk is also considered not reduced. When the risk has not decreased, the psychological risk assessment module generates an escalation instruction. This escalation instruction includes the patient's identity information, home address, event summary, intervention content, post-intervention status characteristics, current risk level, and recommended treatment method, and is transmitted to the caregiver support module and the emergency response module.

[0099] As a result, the actual feedback from patients after the intervention is transformed into actionable risk reassessment criteria, enabling home observation and escalation of treatment to be determined based on the continuous changes before and after the same abnormal event.

[0100] The emergency response module sends the patient's address, anomaly summary, risk level, and intervention details to the community or hospital based on the upgraded treatment instructions.

[0101] After receiving an escalation command, the emergency response module first structures and organizes the information in the escalation command to form an assistance request for the community or hospital. The escalation command includes at least the patient's identity information, home address, event summary, current risk level, intervention content, and post-intervention status characteristics. The patient's identity information includes the patient's name or patient number, age, previous mental health management category, and emergency contact person; the home address includes the community name, building number, unit number, house number, and pre-registered home visit assistance information; the event summary includes the initial abnormal segment time, continuous arousal status, duration of abnormal heart rate variability, duration of being out of bed, access point status, voice response status, and active call for help verification results; the intervention content includes the output mindfulness audio, CBT thought record questions or voice reassurance prompts, and their corresponding output times.

[0102] Before generating a rescue request, the emergency response module performs a necessity check on the escalation command. If the current risk level is Level 3, and any one of the following post-intervention characteristics is unresponsive, not returned to bed, heart rate variability not recovered, or entrance / exit reopened, the rescue request is sent directly to both the community and hospital. If the current risk level is Level 2, and both unresponsiveness and not returned to bed are present, the rescue request is sent to the community and simultaneously copied to the hospital. If the current risk level is Level 1, no rescue request is automatically sent; only a continued observation suggestion is sent to the caregiver. This triage method avoids indiscriminately pushing alleviated nocturnal abnormalities to medical resources while ensuring that unresolved silent psychological crises can enter the external rescue process.

[0103] The emergency request is sent in a tiered information packet format. The first-level packet is used for rapid location tracking, including the patient's ID, home address, current risk level, and time of the incident. The second-level packet assists in judgment, including an event summary, intervention details, and post-intervention status characteristics. The third-level packet is used for on-site handling reference, including suggested communication scripts, a summary of the patient's baseline data for the past 14 nights, and emergency contact information. The emergency response module first sends the first-level packet, and after receiving confirmation from the community or hospital, it sends the second-level packet. If confirmation is not received within 30 seconds, the first-level packet is sent again, repeated twice, and then a backup communication channel is used. Backup communication channels include SMS gateways, cellular data links, or pre-bound duty terminal channels.

[0104] After receiving a distress request, the emergency response module continuously updates the patient's status at the community or hospital level. Updates include whether the patient has returned to bed, whether the door sensor has been triggered again, whether a voice response has been received, whether heart rate variability has recovered to more than 80% of the historical baseline for the same period, and whether the voice terminal remains interactive. The preferred update frequency is once every 60 seconds; however, if the entrance / exit is detected to have reopened, the patient remains unresponsive for 300 seconds, or abnormal heart rate variability persists for more than 600 seconds, the update frequency is adjusted to once every 30 seconds. Each update includes the data collection time, allowing the community or hospital to assess whether the on-site risk continues to escalate.

[0105] Before rescuers arrive, the emergency response module will keep the voice terminal or the patient's mobile device interactive. If the patient responds, a voice summary will be transmitted to the community or hospital, and the voice terminal will play low-stimulation reassuring prompts, such as confirming the current location, guiding the patient away from the entrance / exit, prompting them to return to the bedside and sit down, or waiting for family members to check on them. If the patient does not respond, the voice terminal will play a short confirmation prompt every 60 seconds, with each prompt not exceeding 10 seconds, to avoid causing further anxiety to the patient due to continuous high-intensity broadcasts at night.

[0106] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A home-based mental health monitoring and closed-loop care support system, characterized in that, include: The IoT monitoring module collects sleep interruption and bed exit data through a smart mattress, entrance and exit opening data through a door sensor, heart rate and heart rate variability data through wearable devices, and voice emotion data through a voice terminal. The edge preprocessing module performs time synchronization and anomaly removal on the collected data, and generates home psychological state characteristics within the same nighttime time window; The psychological risk level generation module compares the characteristics of the patient's psychological state at home with the patient's historical baseline. When it identifies continuous awakening, abnormal heart rate variability, getting out of bed beyond the threshold, or repeated opening of the door magnet without active calling for help, it generates silent psychological crisis events and risk levels. The digital intervention module outputs mindfulness audio, CBT thinking record questions, or voice reassurance prompts according to the risk level, and collects patient responses, emotional changes, bed status, and heart rate variability changes to form post-intervention status characteristics. The psychological risk assessment module reassesses the risk based on the post-intervention status characteristics. If the risk is reduced, the caregiver support module will push home observation suggestions. If the risk is not reduced, an escalation treatment instruction will be generated. The emergency response module sends the patient's address, anomaly summary, risk level, and intervention details to the community or hospital based on the upgraded treatment instructions.

2. The home-based mental health monitoring and closed-loop care support system according to claim 1, characterized in that, The IoT monitoring module collects data through the following steps: The smart mattress determines the bedtime, bedtime, number of nighttime awakenings, and duration of sleep interruption based on changes in bed pressure and body movement, and generates a bed-out event when the patient transitions from a bed-lying state to a bed-out state. When the door sensor detects that the entrance / exit has changed from a closed state to an open state, it generates a door sensor opening event and records the opening time, opening duration, number of openings, and the corresponding entrance / exit location. Wearable devices continuously collect heart rate and heart rate variability data before and after sleep interruption, bed exit events, or door magnetic sensor opening events, and generate physiological stress data when heart rate variability remains abnormal. When both bed-off events and physiological stress data are present, the voice terminal outputs voice questions and generates voice emotion data based on the patient's response status, speech rate, volume, pause duration, and emotional tendency.

3. The home-based mental health monitoring and closed-loop care support system according to claim 2, characterized in that, The edge preprocessing module performs time synchronization in the following steps: The system receives monitoring data packets sent by smart mattresses, door sensors, wearable devices, and voice terminals. The monitoring data packets include device identifiers, collection times, event types, and collection values. Using the start time of leaving the bed and the time of returning to the bed in the bed-leaving event as time anchors, the entrance and exit opening data, heart rate and heart rate variability data and voice emotion data are written into the same event cache queue; The time between leaving the bed is extended forward by 120 seconds and the time returning to the bed is extended backward by 300 seconds, forming a preceding and following adjacent interval. Data falling within the preceding and following adjacent intervals are arranged in the order of occurrence to obtain the calibrated monitoring event sequence.

4. The home-based mental health monitoring and closed-loop care support system according to claim 3, characterized in that, The edge preprocessing module performs anomaly removal by including the following steps: Determine whether the entry / exit opening data falls within the adjacent interval before and after any bed leaving event. If it does not fall within the interval, mark the entry / exit opening data as a low-confidence event. Determine whether the time interval between heart rate and heart rate variability data and sleep interruption and bed exit data exceeds 300 seconds. If it does, mark the heart rate and heart rate variability data as a low-confidence event. If, within 600 seconds of a low-confidence access point opening data event, the voice terminal does not respond, displays negative voice emotion, or the user leaves the bed again, the access point opening data event is converted into a valid event. If, within 300 seconds of a low-confidence heart rate and heart rate variability data point, the duration of time spent out of bed exceeds 600 seconds or the data point for opening an entrance or exit is obtained, the heart rate and heart rate variability data point will be converted into a valid event. Low-confidence events that are not verified will not participate in the formation of home-based psychological state characteristics.

5. The home-based mental health monitoring and closed-loop care support system according to claim 1, characterized in that, The edge preprocessing module generates home-based psychological state characteristics through the following steps: The purified monitoring event sequence was spliced ​​together in the order of occurrence of a sleep interruption, a continuous process of getting out of bed, the corresponding opening status of the entrance and exit, the corresponding heart rate and heart rate variability, and the corresponding voice emotion response state to form a nighttime state segment. The start time of sleep interruption is used as the starting point of the nighttime state segment, and the latest of the following—the time of returning to bed, the time of the end of voice interaction, or the time of the last valid event—is used as the end point of the segment. Arrange multiple nighttime state segments from morning to night according to their starting points to form a nighttime state segment chain; Continuous arousal features, persistent bed-leaving features, entrance / exit proximity features, persistent physiological stress features, and abnormal voice response features are extracted from the nighttime state fragment chain and encapsulated into home psychological state features according to the chronological order of the nighttime state fragment chain.

6. The home-based mental health monitoring and closed-loop care support system according to claim 5, characterized in that, The psychological risk level generation module establishes a patient's historical baseline and compares it with similar time periods, specifically including the following steps: Data from the patient’s most recent 14 effective nighttime windows were used as historical baseline data. An effective nighttime window is defined as a nighttime record in which there are continuous data collection records from smart mattresses, door magnets, and wearable devices, and the duration of missing data for any single type does not exceed 10% of the total duration of the nighttime window. Divide the nighttime time window into several similar time periods; The median values ​​of sleep interruption, heart rate variability, duration of getting out of bed, and number of times the entrance / exit was opened were taken within each similar time period to obtain the continuous wakefulness baseline, heart rate variability baseline, duration of getting out of bed baseline, and entrance / exit opening baseline. By comparing the corresponding features in the current nighttime state segment chain with the historical baseline of the same time period, we can obtain the results of continuous wakefulness deviation, physiological stress deviation, bed exit deviation, and entrance / exit deviation.

7. A home-based mental health monitoring and closed-loop care support system according to claim 6, characterized in that, The psychological risk level generation module generates silent psychological crisis events through the following steps: The nighttime state segment that first satisfies the abnormal result of continuous wakefulness deviation is identified as the initial abnormal segment; Following the sequence of nighttime state fragments, search for fragments with abnormal heart rate variability within 30 minutes after the initial abnormal fragment, search for fragments with persistent abnormality after leaving the bed within 30 minutes after the abnormal heart rate variability fragment, and search for fragments with repeated opening and closing of entry and exit points within 10 minutes after the fragments with persistent abnormality after leaving the bed. When continuous awakening, abnormal heart rate variability, getting out of bed beyond the threshold, and repeated opening of the door sensor occur in the above sequence, a nighttime crisis behavior chain is formed. Verify whether there are any records of voice calls for help, button calls for help, or mobile calls for help within the same nighttime time window. If not, mark the nighttime crisis behavior chain as a silent psychological crisis event and write the abnormal segments into the event summary in the order of occurrence.

8. A home-based mental health monitoring and closed-loop care support system according to claim 7, characterized in that, The psychological risk level generation module generates risk levels through the following steps: A Level 1 risk is generated when the event summary contains only continuous arousal abnormalities and heart rate variability abnormalities, and the interval between the two is no more than 30 minutes. Based on Level 1 risk, Level 2 risk is generated when the duration of a single bed leave is not less than 900 seconds, or when the current bed leave duration increases by not less than 600 seconds relative to the bed leave duration baseline. Based on the level 2 risk, a level 3 risk is generated when the entrance / exit is opened no less than twice within 600 seconds after getting out of bed and there is no record of active distress calls. The risk level and event summary are sent together to the digital intervention module, which then selects the corresponding intervention content according to the risk level.

9. A home-based mental health monitoring and closed-loop care support system according to claim 8, characterized in that, The digital intervention module generates post-intervention status characteristics through the following steps: When the risk level is Level 1, output a 5-minute mindfulness audio; when the risk level is Level 2, output 3 CBT thought recording questions related to sleep interruption, reasons for getting out of bed, and current thoughts; when the risk level is Level 3, output reassuring voice prompts including identity verification, current location verification, and safety guidance. An intervention start marker is written when the intervention content is first output. The intervention start marker includes the intervention start time, risk level, event summary number, and intervention content type. Starting from the intervention initiation marker, patient responses, changes in verbal and emotional states, bed status, and heart rate variability were collected within 15 minutes. Under level 3 risk, the 15 minutes were divided into three consecutive 5-minute segments. The patient's response, mood improvement, return-to-bed status, heart rate variability, and physiological stress mitigation results were written into the same intervention feedback segment and associated with the pre-intervention home psychological state characteristics according to the event summary number to form the post-intervention state characteristics.

10. A home-based mental health monitoring and closed-loop care support system according to claim 9, characterized in that, The psychological risk assessment module and the emergency response module implement closed-loop management based on post-intervention status characteristics, specifically including the following steps: The psychological risk assessment module determines whether the post-intervention status characteristics simultaneously meet the criteria of timely or delayed response, emotional relief, stable return to bed, relief of physiological stress, and no re-opening of the entrance / exit during the reassessment period. If these criteria are met, the risk is considered reduced, and the caregiver support module will push home observation recommendations. If any two of the following two conditions are present after intervention: no response, worsening of emotions, failure to return to bed, failure to alleviate physiological stress, or re-opening of the entrance / exit during the reassessment period, the risk is deemed not to have been reduced, and an escalation treatment instruction is generated. The emergency response module generates a tiered rescue information package based on the upgraded response instructions. The first-level information package includes the patient number, home address, current risk level, and time of abnormal occurrence. The second-level information package includes an event summary, intervention content, and post-intervention status characteristics. The third-level information package includes suggested communication scripts, a summary of the nighttime baseline for the past 14 days, and emergency contact information. The emergency linkage module first sends a Level 1 information packet to the community or hospital, and then sends a Level 2 information packet after receiving confirmation. Before the rescuers arrive, it continuously updates information such as whether the patient has returned to bed, whether the door sensor has been triggered again, whether there is a voice response, and whether the heart rate variability has recovered.

Images