A method and system for generating and managing execution of an exercise prescription for the elderly based on stages and dimension coverage

Abstract

The application discloses a kind of old age movement prescription generation and execution management method and system based on stage and dimension coverage (Ruomu engine), solve the problem that existing scheme lacks stage template, dimension is not even, no pre-training hard check, no cross-discipline closed loop, evaluation lag.Method: S1 binds 6 base prescription templates according to effective stage effectiveStage;S2 generates coverage score table according to strength, endurance, flexibility, balance, coordination, six-dimensional score of heart and lung;S3 is supplemented from action library when coverage is lower than threshold δ ;S4 is linearly decomposed into week and subgoal according to 12 weeks target;S5 is forced seven measurement hard check before training;S6 triggers cross-discipline work order when physical sign is out of bounds;S7 is evaluated based on Barthel and evaluation score every month and adjusts file.Prescription is embedded in evaluation report chapter field.Four closed loops of generation, check, execution and feedback.

Description

Technical Field

[0001] This invention relates to the field of computer-aided decision-making technology for smart elderly care and rehabilitation training, specifically to a method and system for automatically generating exercise prescriptions and managing their execution based on an age-related functional stage and a dimensional coverage model. The applicant has named the specific device for implementing the method and system described in this invention the "RuoMu Engine" (hereinafter referred to as "RuoMu Engine"). "RuoMu" is merely the applicant's internal code name and commercial identifier for the engine and does not constitute a limitation on the technical solution and scope of protection of this invention. Background Technology

[0002] As my country's aging population continues to deepen, exercise prescription, as a core intervention for functional maintenance and disability rehabilitation in the elderly, faces the following technical shortcomings in its digital implementation: 1. Lack of staged templates: Current prescription recommendations are mostly categorized by a single "exercise intensity" (low / medium / high), without using differentiated templates based on the comprehensive assessment stages of the elderly. The same prescription template is given to both active A0 elderly and deeply disabled A5 elderly, which cannot guarantee safety and effectiveness.

[0003] 2. Imbalance in Dimensional Coverage: Rehabilitation practice requires prescriptions to cover six dimensions: strength, endurance, flexibility, balance, coordination, and cardiopulmonary function. Existing systems only record the total number of movements and total duration, without verifying the coverage of dimensions. This results in some prescriptions consisting entirely of "strength training" while completely lacking "balance training," rendering fall prevention ineffective.

[0004] 3. Lack of pre-training verification: The existing system only records the trainer's subjective assessment before training begins, without mandating the collection of seven vital signs: blood pressure, heart rate, blood oxygen, GCS, basal body surface area, skin condition, and body temperature. If an accident occurs during training, the organization cannot prove that the necessary safety assessment has been conducted.

[0005] 4. Lack of interdisciplinary closed-loop management for boundary violations: Abnormal signs detected before training are often recorded but ignored, and are not automatically assigned to the corresponding departments such as swallowing, wound care, psychiatry, cardiology, and urology for further treatment. There is no closed-loop audit for boundary violation incidents.

[0006] 5. Delayed evaluation of efficacy: The current system only evaluates the efficacy of prescriptions at the end of the 12-week cycle, without the ability to adjust treatment levels monthly during the intermediate period. If an elderly patient needs to be downgraded immediately due to changes in their condition, manual intervention is required.

[0007] 6. Fragmented Prescription Data: The existing system stores prescriptions in a separate business table, separate from the assessment report. After a re-evaluation, the prescriptions are often still the old version, and the prescriptions seen by family members and trainers are inconsistent with the latest assessment conclusions.

[0008] 7. Implementation disconnect: One day after payment, the trainer, family members, and elderly people do not know what to do, when to do it, or at what frequency. The implementation of the lesson plan relies entirely on the trainer's personal experience.

[0009] Existing publicly available solutions mainly focus on recommending exercise intensity or single movements, without disclosing a complete closed loop combining six elements: staged templates, dimensional coverage verification, pre-training hard verification, cross-disciplinary work orders, monthly effect evaluation, and prescription-embedded assessment reports. This invention addresses these shortcomings by providing a method and system for an exercise prescription engine based on stage and dimensional coverage. Summary of the Invention

[0010] 3.1 Purpose of the Invention This invention provides a method and system that automatically generates prescriptions based on the comprehensive assessment stage of the elderly, performs the first seven hard verification tests, automatically dispatches cross-disciplinary work orders when exceeding the scope, and evaluates the effectiveness monthly based on Barthel and assessment scores, triggering grade adjustments. This constitutes a complete closed loop of "generation-verification-execution-feedback", minimizing the reliance on trainers' experience and implementing safety responsibilities at the system level.

[0011] 3.2 Technical Solution The core solution of this invention, as described in independent claims 1 and 9, includes seven steps (S1 to S7) on the method side and eight modules on the system side. Its core innovation lies in: Innovation Point 1 — Staged Template Binding: This invention binds six stages to six sets of basic prescription templates. A0 uses P00 for general maintenance, A1 uses P05 for pre-frailty prevention, A2 uses P10 for moderate pre-frailty, A3 uses P15 for frailty recovery, A4 uses P16 for moderate to severe, and A5 uses P16A for deep disability. The binding relationship is stored in a configuration table, allowing for version control. When the stage of the same elderly person changes, the prescription template automatically follows, eliminating the need for manual reconfiguration.

[0012] Innovation Point 2 — Six-Dimensional Coverage Verification: This invention explicitly defines six dimensions—strength, endurance, flexibility, balance, coordination, and cardiopulmonary function—as the integrity dimensions of the prescription. Each movement is pre-scored by a team of rehabilitation therapists on four levels (0 / 1 / 2 / 3), forming a dimensional scoring table for the movement library. When the prescription is generated, the six-dimensional coverage vector V is obtained by summing the columns. If any dimension is lower than the stage threshold δ_j, it is automatically supplemented from the movement library to ensure the balance and completeness of the prescription.

[0013] Innovation Point 3 — Seven Pre-Training Hard Verifications: This invention uses the collection of seven vital signs—blood pressure, heart rate, blood oxygen, GCS, basal body surface area, skin condition, and body temperature—as hard preconditions for starting training. Each of the seven indicators is judged according to preset thresholds, resulting in a pass, warning, or rejection. Training can only begin when all seven indicators pass or at most one indicator issues a warning. The record of passing these hard verifications constitutes complete audit evidence of the organization's safety responsibility.

[0014] Innovation Point 4 — Closed-Loop Interdisciplinary Work Orders: This invention establishes an automatic mapping relationship between seven out-of-boundary events and five types of interdisciplinary work orders (swallowing, wound, mental health, cardiology, and urology). When any vital sign exceeds the boundary, the engine automatically generates a work order, assigns it to the corresponding department according to priority, notifies the family and responsible caregiver, and writes back the closed-loop status after the work order is processed. Out-of-boundary events are no longer overlooked.

[0015] Innovation Point 5 – 12-Week Target Linear Decomposition and Monthly Adjustment: This invention uses a 12-week prescription cycle, linearly decomposing the overall target into specific, actionable indicators on a weekly and per-prescription basis. Every 30 days, an EVAL is calculated based on the Barthel Index increment ΔBI and the assessment score increment ΔCS. If EVAL is greater than the upgrade threshold up_th, the prescription level is reduced by one level in the next cycle (capacity improvement); if it is less than the downgrade threshold down_th, the prescription level is increased by one level in the next cycle (capacity decline requires conservatism). A change notification is generated and sent to family members for confirmation before any upgrade or downgrade is implemented.

[0016] Innovation Point 6 — Prescription Embedded Evaluation Report: This invention stipulates that prescriptions do not have a separate business table, but are embedded in the chapters field of the evaluation report in JSON format. Both the reader and the displayer obtain the prescription through the evaluation report, avoiding the data synchronization lag problem caused by a separate table, and also avoiding the mismatch scenario of "updated evaluation but outdated prescription".

[0017] 3.3 Beneficial Effects (1) Stage safety: Different templates are bound to different stages. A5 deep disability and A0 full vitality do not share the same prescription, thus avoiding excessive or insufficient strength from the source.

[0018] (2) Dimensional balance: The six-dimensional coverage is explicitly verified to avoid prescription bias, and the three aspects of fall prevention, cognitive maintenance and cardiopulmonary endurance are taken into account.

[0019] (3) Pre-training evidence: The passing records of the seven hard test verifications constitute a complete chain of evidence for the organization's safety responsibility. In the event of an accident, the organization can prove that it has been fully assessed.

[0020] (4) All violations will be investigated: Violations will be automatically dispatched to the corresponding department for handling to avoid omissions. The closed-loop status of work orders can be audited afterward.

[0021] (5) Dynamic adjustment: Monthly effect evaluation allows prescriptions to be adjusted in a timely manner according to changes in the elderly’s condition, without having to wait for the 12-week cycle to end.

[0022] (6) Data consistency: The embedded evaluation report in the prescription ensures that the values ​​read by multiple terminals are consistent, avoiding the mismatch between the evaluation update and the prescription.

[0023] (7) Implementation: The process from payment to trainer execution is fully defined. Trainers follow the system prompts in sequence, significantly reducing reliance on personal experience. Attached Figure Description

[0024] To facilitate understanding of the present invention by those skilled in the art, the accompanying drawings are briefly described below. All drawings are black and white line drawings, using only boxes, connecting lines, arrows, and Arabic numerals; the Chinese meanings of each node are given in the following text descriptions, and the drawings themselves do not contain textual annotations.

[0025] Figure 1 shows the overall architecture and four-stage closed-loop flowchart of the prescription engine of this invention. The figure shows from top to bottom: [1] effectiveStage input; [2] basic prescription template P_base; [3] six-dimensional coverage verification; [4] 12-week target linear decomposition; [5] pre-training seven-piece hard verification; [6] cross-boundary cross-science work order dispatch (side branch); [7] training execution and Barthel recording; [8] monthly effect evaluation and file adjustment. The output of node [5] is divided into two paths: entering [7] through the branch and entering [6] through the cross-boundary branch. Node [8] is connected to [1] in a loop to form a complete closed loop.

[0026] Figure 2 illustrates the staged template binding and six-dimensional coverage. The left side vertically lists six stages, A0 to A5, while the right side lists the corresponding six basic prescription templates: P00, P05, P10, P15, P16, and P16A. The templates are bound together horizontally on both sides. A six-dimensional radar chart is also shown in the lower right corner, with the six axes labeled as strength, endurance, flexibility, balance, coordination, and cardiopulmonary function. Shaded areas in the chart indicate the actual coverage of a prescription; a depression in a certain axis indicates that that dimension needs to be supplemented.

[0027] Figure 3 is a time series diagram of the linear decomposition of the 12-week target. The horizontal axis represents the week numbers W1 to W12, and the vertical axis is divided into three rows: the top row represents the number of training sessions this week, the middle row represents the cumulative duration this week, and the bottom row represents the dimensionality coverage increment this week. The values ​​in the three rows are arranged in a linearly increasing or uniformly distributed manner. The right side of the figure is labeled "At the end of 12 weeks, the difference between the actual cumulative completion rate and the target is used to trigger the calibration of the next cycle".

[0028] Figure 4 shows the flowchart of the seven-piece hardware verification process before training. A rectangle in the center is labeled "Seven-piece Hardware Verification," from which seven branches extend outwards. Each branch ends at the corresponding feature's bounding threshold. The seven branches converge back to the center and output a diamond-shaped decision node asking "All seven items passed or at most one warning?" The yes branch outputs "Training Allowed," and the no branch outputs "Training Rejected," connecting to the work order process in Figure 5.

[0029] Figure 5 shows the structure of cross-departmental work order dispatch. The central node, "Cross-Departmental Event," radiates outwards to five branches: Swallowing Department (triggered by Kubota), Wound Care Department (triggered by Skin), Psychiatry Department (triggered by GCS), Cardiology Department (triggered by BP / HR), and Urology Department (triggered by Urinary Incontinence). Each branch ends with a work order box, and below each work order box is a vertical line connecting to the "Notify Family + Responsible Manager + On-Duty Personnel" convergence node. Below the convergence node is a write-back node, "Closed-Loop Status Write-back to Training Record." Detailed Implementation

[0030] Example 1: Complete prescription generation for elderly people in stage A2 Subject: Ms. Li, 72 years old, most recent assessment report effectiveStage = A2.

[0031] Step S1: The stage reading module reads effectiveStage = A2 from the assessment report and binds it to the P10 moderate pre-fainting intervention template according to the configuration table. The initial action card set of the P10 template is C = {c_101 elastic band upper limb, c_102 sitting training, c_103 walking training}, 3 times a week, 35 minutes each time, 3 sets per session, with 60 seconds of rest between sets.

[0032] Step S2: The dimension coverage verification module checks the action library dimension scoring table. | Movement | Strength | Endurance | Flexibility | Balance | Coordination | Cardiorespiratory | | c_101 Elastic Band Upper Limb | 3 | 1 | 1 | 0 | 1 | 0 | | c_102 Sitting Training | 2 | 2 | 0 | 1 | 1 | 1 | | c_103 Walking Training | 1 | 3 | 0 | 2 | 1 | 3 | The six-dimensional coverage vector V = (6, 6, 1, 3, 3, 4).

[0033] Step S3: A2 stage threshold δ_j = 5. Flexibility dimension v_3 = 1 < 5, needs to be supplemented; balance dimension v_4 = 3 < 5, needs to be supplemented; coordination dimension v_5 = 3 < 5, needs to be supplemented. Retrieve movements with high flexibility scores and a safety rating of A from the movement library, and add c_201 stretching combination (flexibility 3, balance 2, coordination 2); add c_202 series standing exercises (balance 3, coordination 2, strength 1). After supplementation, V = (7, 6, 4, 8, 7, 4), flexibility is still 1 point short, cardiopulmonary function is still 1 point short, so add c_203 breathing exercises (flexibility 2, cardiopulmonary function 2). Finally, V = (7, 6, 6, 8, 7, 6), all targets are met, forming a complete stage prescription P_full.

[0034] Step S4: The target calibration module reads the evaluation cycle start t_0 = 2026-06-01, T_end = 2026-08-24 (12 weeks later). Weekly target g_w: number of training sessions this week = 3, cumulative duration this week = 105 minutes, dimensionality coverage increment this week = ΔV_target / 12. Heart rate zone for each target g_s: 72 years old, [ (220 − 72 ) ×0.55, ( 220 − 72 ) × 0.70 ] = [ 81 , 104 ] bpm.

[0035] Step S5: June 1, 2026 Before the first training session, trainer Zhang led Li through seven tests: BP 138 / 85 → Normal; HR 82 → Normal; SpO2 97% → Normal; GCS 15 → Normal; Kubota Grade 1 → Normal; No skin lesions → Normal; Body temperature 36.6℃ → Normal. All seven items passed, training permitted.

[0036] Step S7: Training is complete. The trainer records the Barthel Index as 85 points, which falls into the BTL timeline.

[0037] 30 days later, on July 1, 2026, the performance evaluation was as follows: BTL starting point 85, ending point 90, ΔBI = 5; evaluation scores CS_new = 63, CS_prev = 60, ΔCS = 3. α = 0.6, β = 0.4, EVAL = 0.6 × 5 + 0.4 × 3 = 4.2. The up-th threshold is up_th = 5, the down-th threshold is down_th = -3, EVAL falls within [-3, 5], maintaining the current stage A2.

[0038] The complete prescription for the stage, P_full, is embedded in the chapters field of Li's vitality function assessment report in JSON format: chapters.prescription = { "template": "P10", "stage": "A2", "duration": 35, "freq": 3, "sets": 3, "rest": 60, "actions": [ "c_101", "c_102", "c_103", "c_201", "c_202", "c_203" ], "dimension_coverage": { "strength": 7, "endurance": 6, "flexibility": 6, "balance": 8, "coordination": 7, "cardio": 6}, "weekly_goal": { "sessions": 3, "duration_min": 105, "coverage_delta": "ΔV / 12"}, "session_goal": { "duration_min": 35, "hr_range": [ 81, 104 ]}, "start_at": "2026-06-01", "end_at": "2026-08-24" } Example 2: Rehabilitation prescription for elderly people with severe disability in stage A5 Subject: Mr. Wang, male, 81 years old, with a 5-year history of Alzheimer's disease, effective Stage = A5.

[0039] Step S1: Bind the P16A deep disability rehabilitation template, 3 times a week (hard limit), 20 minutes each time, with 12 movement cards in the action card set C = {c_A5_01 passive flexion and extension of the hand, c_A5_02 sitting balance in bed, c_A5_03 chewing tongue exercise}.

[0040] Step S3: A5 stage threshold δ_j = 3, a lenient threshold to avoid introducing high-risk actions to fill in the dimensions.

[0041] Step S5: Seven pre-training tests: BP 158 / 95 → Warning; HR 88 → Normal; SpO2 96% → Normal; GCS 14 → Warning; Kubota Grade 2 → Normal; Grade I pressure sore on right heel → Warning; Body temperature 36.9 → Normal. If three warnings exceed the "maximum of one warning" limit, training is refused.

[0042] Step S6: The cross-departmental work order dispatch module generates three work orders: ① Cardiovascular department work order (BP out of bounds), high priority; ② Psychiatric work order (GCS boundary overstepping), high priority; ③ Wound Care Work Order (Skin Beyond Boundaries), Priority: mid.

[0043] The three work orders are combined into a single comprehensive work order R = {workorder_2026070101} and dispatched to the on-duty doctor's workstation. At the same time, Ms. Wang, the family member, and Mr. Chen, the responsible housekeeper, are notified via SMS.

[0044] After the treatment was completed, the doctor updated the work order status to "BP blood pressure medication adjusted, GCS acute event ruled out, wound treated" and linked it to the original training record t_2026-07-01. The next day, before attempting training again, the seven tests were re-executed. If all tests were passed, training was allowed.

[0045] Example 3: Monthly performance evaluation triggers upgrade Test subject: Zhao, male, 68 years old, starting from A3 stage, Barthel Index 65.

[0046] 30 days later, the two most recent records on the BTL table changed from 65 to 78, with a ΔBI of 13; the assessment score changed from 45 to 52, with a ΔCS of 7. EVAL = 0.6 × 13 + 0.4 × 7 = 10.6, which is greater than up_th = 5, triggering an upgrade.

[0047] Announcement of Changes in the Document Generation Phase of the Document Transfer Module: { "elder": "Zhao Mou", "from_stage": "A3", "to_stage": "A2", "eval_score": 10.6, "delta_bi": 13, "delta_cs": 7, "next_template": "P10", "effective_at": "2026-08-01" } A notification will be sent to the family member for confirmation. After the family member confirms, the prescription template for the next 12 weeks will switch from page 15 to page 10. If the family member does not confirm within 7 days, it will be considered as acceptance.

[0048] Example 4: Cross-Science Work Order Aggregation Subject: Ms. Sun, 78 years old, tested seven times on the same day for BP exceeding the limit and skin exceeding the limit.

[0049] The cross-departmental work order dispatch module detected two out-of-bounds entries with the same elder_id and the same date, and merged them into a single comprehensive work order according to the work order aggregation rules. { "workorder_id": "wo_2026080315", "elder_id": "sun_xxx", "issues": [ { "type": "cardiovascular", "source": "BP", "value": "165 / 95", "prio": "high"}, { "type": "wound", "source": "skin", "value": "sacral I degree", "prio": "mid"} ], "target_depts": [ "cardio", "wound_care" ], "assigned_to": [ "doctor_202", "nurse_305" ], "status": "open" } The integrated work order reduces the workload of on-call doctors, allowing two issues to be addressed together during the same ward round.

[0050] Example 5: Consistency Assurance of Prescription Embedded Assessment Reports Scenario: A trainer adjusts a senior citizen's prescription on the management platform (adding an action card). 30 minutes later, the headquarters auditor checks the senior citizen's assessment report in the management backend and needs to see the latest version of the prescription.

[0051] If the prescription is stored in a separate business table: after the trainer makes a modification, the prescription field of the assessment report needs to be refreshed synchronously. If the synchronization fails, the prescription seen by the headquarters will still be the old version.

[0052] The prescription embedding module of this invention allows trainers to modify only the `chapters.prescription` field of the assessment report; the modification is completed immediately upon being saved to the database, without any synchronization link. Headquarters auditors read the assessment report and immediately access the latest prescription, ensuring strong consistency between the two ends.

[0053] Example 6: Hardware and Software Implementation Environment The applicant has named the specific implementation device of this invention the "Ruomu" engine. The Ruomu engine's server-side implementation uses a NestJS 10, Prisma 5, and PostgreSQL 16 architecture, deployed on a dual-instance cluster. The front-end family client and trainer client use the native WeChat Mini Program framework. The action library is stored in the `exercise_bank` table, the action dimension score is stored in the `exercise_dimension_score` table, the boundary violation rules are stored in the `pretrain_check_rule` table, the Barthel timeline is stored in the `barthel_timeline` table, and the work order is stored in the `referral_ticket` table. The `engine_version` field of the prescription JSON is labeled with a string like "ruomu-2.3" to indicate the Ruomu engine version number, facilitating cross-version auditing and regression. A scheduled task triggers the monthly performance evaluation every day at midnight via a cron expression. The above hardware and software stack, as well as the name "Ruomu," are merely illustrative and commercial identifiers; any equivalent technical solution should fall within the protection scope of this invention.

[0054] Example 7: Batch prescription repair Scenario: Early versions of the engine did not enforce the supplementation of dimension thresholds during the dimension coverage verification process, resulting in some elderly people's prescriptions having missing dimensions.

[0055] To fix the batch prescription replenishment script, follow these steps: 1. Query the set R of report records in the evaluation report table where the chapters.prescription field exists and any dimension of the dimension_coverage vector is below the stage threshold; 2. For each record, rerun the dimension coverage check using the current engine version and supplement the actions; 3. Write back the supplemented prescription chapters.prescription, and record the "supplement prescription" event in the audit_log table, marking the old action list and the new action list; 4. If more than 2 new movements are added, a trainer notification will be triggered. The trainer should communicate the reason for the addition of movements to the family before the next training session.

[0056] The prescription refill process reuses the engine's main workflow, eliminating the need to write separate prescription refill logic. All prescription refill events are logged and can be audited later.

[0057] VI. Conclusion of the Instruction Manual The method and system described in this invention, using the "Ruomu" engine as the specific implementation device, have been deployed in a senior care service institution, covering 480 elderly people, generating over 4,200 prescriptions, executing over 12,000 seven-item tests, issuing 320 cross-disciplinary work orders, and triggering 68 monthly performance evaluations for grade adjustments. No training-related accidents occurred during the six months of operation. Compared to traditional exercise prescription programs, the safety responsibility of trainers is implemented at the system level, and the visibility of the prescription process to family members is significantly improved. "Ruomu" is merely the applicant's commercial code and internal identifier and does not constitute a limitation of the claims; the scope of protection of this invention is determined by the claims, and all equivalent modifications fall within the scope of protection of this invention.

Claims

1. A method for generating and managing exercise prescriptions for the elderly based on stage and dimension coverage, characterized in that, Includes the following steps: S1: Read the effectiveStage of the test subject U, where the effectiveStage is one of the six-stage set {A0, A1, A2, A3, A4, A5}; bind the basic prescription template P_base according to the stage, where the basic prescription template P_base includes at least the action card set C = {c_1, c_2, ..., c_k}, the total duration of each training session, the weekly training frequency freq, the number of sets per session, and the rest between sets; S2: For each action c_i in the action card set C of the basic prescription template P_base, score s_ij according to the six dimensions of strength, endurance, flexibility, balance, coordination, and cardiopulmonary function, where j belongs to {1, 2, 3, 4, 5, 6}, forming a dimension coverage score matrix M, and summing the columns to obtain a six-dimensional coverage vector V = (v_1, v_2, v_3, v_4, v_5, v_6), where v_j = Σ_i s_ij; S3: If there is a dimension j in the six-dimensional coverage vector V such that v_j is less than the stage-related threshold δ_j, then retrieve the candidate action with a high score in that dimension from the stage-related action library L and add it to the basic prescription template P_base until the coverage of all dimensions reaches the threshold, forming a complete stage prescription P_full. S4: Read the subject's current assessment cycle start point t_0 and target cycle T_end (default T_end = t_0 + 12 weeks), and linearly decompose the 12-week target into weekly target g_w and individual target g_s. The weekly target includes three items: number of training sessions this week, cumulative duration this week, and dimensional coverage increment this week. The individual target includes three items: training duration this week, number of sets this week, and heart rate zone this week. S5: Before each training session, the subject undergoes a mandatory hard check (CHK) of seven tests, which include at least seven items: systolic blood pressure (BP), resting heart rate (HR), blood oxygen saturation (SpO2), Glasgow Coma Scale (GCS), Kubota water swallowing test, skin integrity examination, and body temperature (T). Each item is judged according to preset boundary rules as pass, warning, or rejection. Training can only begin when all seven items are passed or at most one item is warned. S6: If any vital sign is determined to be out of bounds in step S5, a cross-departmental work order R is automatically generated according to the out-of-bounds type. The work order R includes at least five fields: work order type, target department, priority, source vital sign, and source subject U. The work order type supports at least five categories: swallowing department, wound care department, psychiatry department, cardiology department, and urology department. S7: After training is completed, record the Barthel Index and volume score according to the preset scale and write them into the Barthel Timeline (BTL) table; every 30 days, perform the prescription effect evaluation (EVAL) based on the BTL table and the most recent evaluation report. If the score improvement exceeds the threshold up_th, a grade increase is triggered; if the score decrease exceeds the threshold down_th, a grade decrease is triggered. The grade increase / decrease results are used as the basis for selecting prescription templates for the next 12 weeks. The complete prescription P_full for that stage is embedded in the chapters field of the vitality function assessment report and stored in JSON format. The prescription does not have a separate business table, and reading and display are both completed through the chapters structure.

2. The method as described in claim 1, characterized in that, The binding relationship between the six stages mentioned in step S1 and the basic prescription template P_base is as follows: ① Bind the P00 general maintenance template to the A0 phase, twice a week, 45 minutes each time; ② In the A1 phase, bind the P05 pre-frailty prevention template 2 to 3 times a week, 40 minutes each time; ③ The A2 phase is bound to the P10 moderate pre-frailty intervention template, 3 times a week, 35 minutes each time; ④ In the A3 phase, bind the P15 frailty recovery template, 3 times a week, 30 minutes each time; ⑤ A4 stage is bound to the P16 moderate to severe weakness recovery template, 3 times a week, 25 minutes each time; ⑥ The A5 stage is bound to the P16A deep disability rehabilitation template, with a strict limit of 3 times per week (not to be exceeded), each session lasting 20 minutes; The binding relationship is stored in the configuration table stage_prescription_binding, which allows for version management.

3. The method as described in claim 1, characterized in that, The scoring rules for the six dimensions mentioned in step S2 are as follows: For each action c_i, the rehabilitation therapist expert group pre-scored it on four levels: 0, 1, 2, and 3: 0 represents that the action does not contribute to the dimension; 1 represents low contribution.

2. Representatives' contributions; 3 represents the main contribution; the sum of the scores for each action across the six dimensions is no less than 3; the scoring results are stored in the action library table exercise_dimension_score and can be visually edited by operations personnel in the management backend.

4. The method as described in claim 1, characterized in that, The dimension coverage threshold δ_j mentioned in step S3 is configured differently according to the stage: δ_j is 6 for stages A0 and A1; δ_j is 5 for stages A2 and A3; δ_j is 4 for stage A4; and δ_j is 3 for stage A5. When supplementing actions, the candidate actions are sorted by a combination of the target dimension score and the safety rating to avoid introducing high-risk actions to supplement a single dimension.

5. The method as described in claim 1, characterized in that, The 12-week target linear decomposition described in step S4 further includes: ① The number of training sessions this week for the weekly target g_w is directly taken from the freq field of the basic prescription; ② This week's cumulative duration g_w.duration = freq × duration; ③ This week's dimensional coverage increment is 1 / 12 of the total target increment ΔV_target over 12 weeks; ④ The heart rate zone for each target g_s is dynamically calculated based on resting heart rate and age using a preset formula, which is [(220 − age) × 0.55, (220 − age) × 0.70]; ⑤ At the end of 12 weeks, the target calibration for the next cycle is triggered based on the difference between the actual cumulative completion rate and g_w.

6. The method as described in claim 1, characterized in that, The specific rules for the seven tests mentioned in step S5 regarding exceeding the limits are as follows: ① Systolic blood pressure (BP): Normal [90, 140], Warning [140, 160] or [80, 90], Reject > 160 or < 80; ② Resting heart rate (HR): Normal [55, 100], Warning [100, 120] or [45, 55], Reject >120 or <45; ③ SpO2: Normal ≥ 95%, Warning [92%, 95%), Reject < 92%; ④ Glasgow Coma Scale (GCS): Normal 15 points, Warning 13 or 14 points, Rejection < 13 points; ⑤ Kubota drinking water test: Normal level 1 or 2, warning level 3, rejection level 4 or 5; ⑥ Skin integrity: Normal without damage, warning of minor damage, rejecting second-degree or higher pressure sores or wounds with active exudation; ⑦ Body temperature T: Normal [36.0, 37.3] ℃, warning [37.3, 38.0] or [35.5, 36.0] ℃, refuse > 38.0 or < 35.5 ℃; The seven out-of-bounds rules are stored in a configuration table, allowing for versioning adjustments.

7. The method as described in claim 1, characterized in that, The automatic dispatch rules for cross-department work orders R mentioned in step S6 are as follows: ① Kubota's drinking water test exceeded the designated area → assigned to the swallowing department, high priority; ② Skin integrity exceeds the limit → Assigned to the wound care department, priority mid; ③ Glasgow Coma Scale score exceeding the limit or abnormal emotional assessment → assigned to psychiatry department, high priority; ④ Systolic blood pressure or resting heart rate exceeding the limits → Assigned to the cardiology department, high priority; ⑤ Records of urinary incontinence or abnormal urination → Assigned to the urology department, priority mid; Once the work order is logged, the family member, the responsible manager, and the on-duty personnel of the target department are notified simultaneously; after the work order is processed, the "closed loop" status is written back and linked to the original training record.

8. The method as described in claim 1, characterized in that, The specific rules for the monthly performance evaluation (EVAL) mentioned in step S7 are as follows: ① Take the first and last two records of the most recent 30 days from the Barthel timeline BTL table and calculate the increment ΔBI; ② Take the most recent score CS_new and the previous score CS_prev from the vitality function assessment report, and calculate the increment ΔCS; ③ The overall score EVAL = α × ΔBI + β × ΔCS, where α and β are preset weights and α + β = 1; ④ If EVAL is greater than the upgrade threshold up_th, then the prescription for the next 12 weeks will be selected by reducing the current stage by 1 level; ⑤ If EVAL is less than the downgrade threshold down_th, then the prescription for the next 12 weeks will be selected by increasing the grade by 1 from the current stage; ⑥ If EVAL is between [down_th, up_th], then maintain the current stage; ⑦ Before shifting gears, a phase change notice must be generated and sent to family members for confirmation.

9. An elderly exercise prescription engine system implementing the method of any one of claims 1 to 8, characterized in that, Includes the following interconnected modules: ① Stage reading module, used to read effectiveStage from the output of the Vitality Function Engine and bind the basic prescription template by stage; ② Dimensional Coverage Validation Module, used to score the prescription action card set according to six dimensions and calculate the coverage vector, and supplement actions when necessary; ③ Target calibration module, used to linearly decompose weekly and individual targets in 12-week cycles; ④ Pre-training hard verification module, used to perform seven tests and judge according to the boundary rules; ⑤ Cross-department work order dispatch module, used to automatically generate and dispatch work orders to the corresponding departments according to the type of cross-boundary physical signs; ⑥ Execution record and Barthel timeline module, used to record the training completion status and the evolution of the Barthel exponent; ⑦ Performance evaluation and adjustment module, used to evaluate performance every 30 days based on BTL and assessment scores and trigger upgrades or downgrades; ⑧ Prescription Embedding Module: This module embeds the complete prescription P_full for a given stage into the chapters field of the evaluation report in JSON format. Prescriptions do not have a separate business table.

10. The system as claimed in claim 9, characterized in that: ① The input to the stage reading module is the effective stage, effectiveStage. After reading, it is not necessary to distinguish whether the stage comes from an algorithm value or a manually overwritten value. ② The action library of the dimensional coverage verification module supports operators to edit three views in the management backend: by topic, by dimension, and by security rating. ③ The out-of-bounds rules of the pre-training hard verification module can be configured differently according to the organization and managed in a versioned manner. Historical rules are retained for auditing when switching. ④ The cross-disciplinary work order dispatch module supports work order aggregation, that is, multiple out-of-bounds vital signs of the same subject on the same day are merged into one work order for dispatch. ⑤ The prescription embedding module avoids the data synchronization problem caused by the independent prescription business table. Both the reader and the displayer obtain the data through the chapters field of the evaluation report, thus avoiding the mismatch scenario of "updated evaluation but outdated prescription".

Images