Journey map-based method for identifying and optimizing pain points of elderly patients in outpatient clinic

Abstract

This invention discloses a method for identifying and optimizing pain points in outpatient medical visits for elderly patients based on a "journey map," relating to the field of pain point identification and optimization for outpatient medical visits for elderly patients. The method includes: constructing a structured digital map containing all key nodes, paths, and their attributes based on the hospital's physical layout and standard medical procedures; building a voice-guided medical model based on the structured digital map, and providing precise navigation according to the elderly patient's voice needs; fabricating a wearable mobile sensing terminal, binding it to the elderly patient's identity information, and constructing a cloud-based and lightweight sensing terminal architecture; and conducting test experiments with a control group to verify and evaluate the credibility and user acceptance of the voice-guided medical model. The advantages of this invention are: it achieves precise, accompanying navigation for elderly patients' outpatient medical visits, significantly reducing the difficulty and physical exertion for elderly patients in finding their way, and improving the medical experience and efficiency.

Description

Technical Field

[0001] This invention relates to the field of pain point identification and optimization for elderly patients seeking outpatient medical care, specifically to a method for identifying and optimizing pain points for elderly patients seeking outpatient medical care based on a "journey map". Background Technology

[0002] With the accelerating aging of the global population, the high prevalence of chronic diseases and frequent medical visits among the elderly make them the primary users of healthcare services. However, the outpatient environment, as the front line of healthcare services, presents significant challenges due to its spatial complexity, professional procedures, and information overload, especially for elderly patients whose cognitive and physical abilities are declining. Traditional outpatient service models rely on static signage systems, paper guidance sheets, and manual inquiry desks. These methods present information in a fragmented manner, are outdated, and have limited capacity, making it difficult to handle the massive and dynamic medical information. In recent years, the wave of "smart hospital" construction has driven a series of digital solutions, such as hospital apps, WeChat official accounts, and in-hospital self-service terminals, aiming to improve service efficiency. These technologies are generally based on the mobile internet and smartphone ecosystem, with the core logic being patient-initiated inquiries and self-service operations. However, for many elderly people who are not adept at using smart devices and face a "digital divide," complex interfaces, cumbersome registration and login processes, and interactive designs requiring good eyesight and responsiveness may actually exacerbate their anxiety and helplessness during the medical process. Meanwhile, the mature application of indoor positioning technologies (such as Wi-Fi and Bluetooth beacons), path planning algorithms, and natural language processing technologies theoretically provides the technological possibility for indoor navigation. However, medical scenarios are highly specialized and dynamic, and general technical solutions cannot be directly adapted. The ultimate goal of medical navigation is not simply "from point A to point B," but rather to understand "what health needs (A) require which professional service node (B) to go to," and to avoid congestion and reduce waiting time during the journey. This requires the system to deeply integrate medical professional knowledge graphs, real-time hospital operational data, and high-precision spatial information to form an intelligent companion system that understands needs, perceives the environment, and makes dynamic decisions. Currently, how to organically integrate these disparate technologies and design a service loop that truly centers on elderly users and bridges the digital divide has become a key issue that urgently needs to be addressed in the field of smart healthcare.

[0003] Existing technological solutions rely on patients operating smartphone apps themselves, which poses a significant obstacle for many elderly people: their vision is declining and they cannot see small print and icons clearly; their fingers are not dexterous and they cannot accurately touch the screen; they are confused and afraid of multi-level menu structures and complex verification processes. Essentially, it requires users to actively adapt to the technology, rather than the technology actively adapting to the user. This interaction model excludes a large number of elderly patients who need the most help from the service. Summary of the Invention

[0004] To address the aforementioned technical issues, this paper presents an optimization method for identifying pain points in outpatient medical visits for elderly patients based on a "journey map." This technical solution resolves the problem mentioned in the background that requires users to actively adapt to the technology, rather than having the technology actively adapt to the user. This interaction mode excludes a large number of elderly patients who are most in need of assistance from the services.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: An optimization method for identifying pain points in outpatient medical visits for elderly patients based on a "journey map" includes: Based on the hospital's physical layout and standard medical procedures, a structured digital map containing all key nodes, paths, and their attributes is constructed. Based on structured digital maps, a voice-guided medical model is constructed, and precise navigation is provided through the voice-guided medical model according to the voice needs of elderly patients; Create wearable mobile sensing terminals, bind elderly patients' identity information, and build a cloud-based and lightweight sensing terminal architecture; By setting up a control group, a test experiment was conducted to verify and evaluate the credibility and user acceptance of the voice-guided medical model.

[0006] Preferably, the construction of a structured digital map containing all key nodes, paths, and their attributes based on the hospital's physical layout and standard medical procedures specifically includes: Key nodes include at least: entrance / exit, registration / check-in point, waiting areas for each department, consultation room, examination / laboratory, payment window, pharmacy and auxiliary services; Define the function, physical location, associated departments, and standard service duration for each key node; Define the start point, end point, physical length, path type, key signs and auxiliary facilities along the route for each path; Based on the hospital's key nodes, paths, and attributes, a structured digital map was constructed using CAD software. The structured digital map constructed using CAD software includes: a base layer, a node layer, a path layer, a marker layer, a facility layer, and data association.

[0007] Preferably, the step of constructing a voice-guided medical model based on a structured digital map, and providing precise navigation based on the elderly patient's voice needs, specifically includes: A voice-guided medical model is constructed based on a structured digital map; the voice-guided medical model takes the voice needs of elderly patients as input and navigation guidance instructions as output. The voice-guided medical model includes voice-to-text conversion, text recognition, path planning, and guidance functions. The speech-to-text function refers to converting the speech needs of elderly patients into text features. The text recognition function refers to matching the converted text features with a medical text database to identify the text features that meet the speech needs of elderly patients. The path planning function refers to obtaining a guidance path based on the text features of the elderly patient's speech needs, using location coordinates and a structured digital map. The guidance function refers to providing guidance based on the acquired guidance path, through voice broadcasting and / or the presentation of a 3D digital map.

[0008] Preferably, the process of fabricating a wearable mobile sensing terminal, binding it to the elderly patient's identity information, and constructing a cloud-based and lightweight sensing terminal architecture specifically includes: The hardware integration of the wearable mobile sensing terminal includes: an identity binding module, an indoor positioning module, a lightweight interaction module, and a data communication module. Identity binding module: It has a built-in, non-rewritable, unique device identifier, which is softly bound to the patient's ID for that visit through the hospital information system interface when the patient picks up the device. Indoor positioning module: integrates UWB and / or Bluetooth AoA chips, and collects the three-dimensional coordinates of wearable mobile sensing terminals in real time under the coverage of the positioning base station in the hospital. Lightweight interactive module: The front of the device features a physical help button, a miniature speaker, and / or a 3D display; Data communication module: integrates Bluetooth Low Energy and Wi-Fi to encrypt and upload location data and interaction events between the wearable mobile sensing terminal and the cloud system; A cloud-based and lightweight sensing terminal architecture is constructed. The cloud layer integrates a structured digital map and a voice-guided medical model for reasoning and computation, while the lightweight sensing terminal is used to collect patient voice and output guidance instructions.

[0009] Preferably, the step of setting up a control group and conducting test experiments to verify and evaluate the credibility and user acceptance of the voice-guided medical model specifically includes: Two groups of elderly patients with the same number of participants were set up in the experimental group, one group did not wear the mobile sensing terminal and the other group wore the mobile sensing terminal. Set the same starting point, ending point, and requirements, conduct test experiments, and record the first-time completion rate, active help rate, movement path distance, and task completion time for each person in each group; For each tester, obtain their evaluation score for this task on a 10-point scale; Based on the results of the test experiments, the credibility and user acceptance of the voice-guided medical model were verified and evaluated.

[0010] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention proposes a method for identifying and optimizing pain points in outpatient medical visits for elderly patients based on a "journey map." Based on a voice-guided medical model, it calculates the patient's real-time location, target nodes, and real-time hospital business data to generate a dynamically optimal path that integrates physical distance and real-time waiting time at each node. By constructing a structured digital map fully aligned with hospital operations, it accurately maps the elderly patient's spoken needs to specific spatial nodes. Combined with UWB centimeter-level positioning and a lightweight wearable terminal, it achieves a closed-loop service from needs understanding and path planning to accompanying guidance. This method enables precise planning and real-time control of the individual medical journey for elderly patients, ensuring that patients dynamically avoid congested windows and busy paths while traveling to their target nodes, prioritizing the shortest overall travel time. This avoids patients wasting energy and time in ineffective queues and detours, greatly improving the efficiency of elderly patients finding their way in complex outpatient environments and increasing their first-time completion rate. It significantly reduces their anxiety and physical burden during the medical process, while also providing data-driven decision support for hospitals to optimize service processes and alleviate spatial congestion, comprehensively improving the age-friendliness and overall operational efficiency of outpatient services. Attached Figure Description

[0011] Figure 1 This is a flowchart of an optimization method for identifying pain points in outpatient medical treatment for elderly patients based on a "journey map" according to the present invention. Figure 2 This is a flowchart of the text recognition function of the present invention; Figure 3 This is a flowchart of the path planning function of the present invention; Figure 4 This is a structural diagram of the electronic device proposed in this invention; Figure 5 This is a schematic diagram of the structure of the computer-readable storage medium proposed in this invention. Detailed Implementation

[0012] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0013] Reference Figure 1 As shown, an optimization method for identifying pain points in outpatient medical visits for elderly patients based on a "journey map" includes: Based on the hospital's physical layout and standard medical procedures, a structured digital map containing all key nodes, paths, and their attributes is constructed. Based on structured digital maps, a voice-guided medical model is constructed, and precise navigation is provided through the voice-guided medical model according to the voice needs of elderly patients; Create wearable mobile sensing terminals, bind elderly patients' identity information, and build a cloud-based and lightweight sensing terminal architecture; By setting up a control group, a test experiment was conducted to verify and evaluate the credibility and user acceptance of the voice-guided medical model.

[0014] To explain, the "journey map" mentioned in this solution refers to a structured digital map built based on the hospital's physical layout and standard medical procedures, encompassing all key nodes, paths, and their attributes, for the entire outpatient process of elderly patients. This map is used to quantitatively analyze the spatiotemporal behavior and experiential pain points of elderly patients at each stage of their medical journey. By constructing this structured digital map, a "dynamic traffic network" model for outpatient services is established, comprehensively depicting the real-time status of each node and the path network topology, providing a quantitative evaluation benchmark for intelligent scheduling. Furthermore, a voice-guided medical model is developed as a personalized dynamic planning engine, generating a globally optimal medical path for each elderly patient based on the real-time situation, integrating the shortest path and minimum waiting time. Simultaneously, a lightweight wearable terminal and cloud-based collaborative architecture are deployed to achieve high-precision real-time positioning and low-latency intelligent guidance. Through control group experiments, this solution verifies the achievement of a closed-loop process from overall hospital situational awareness to individual dynamic path optimization and accompanying execution guidance, effectively addressing the core pain points of elderly patients in complex outpatient environments, such as difficulty finding their way, ineffective waiting, and physical exhaustion caused by information asymmetry and unreasonable path planning.

[0015] The construction of a structured digital map based on the hospital's physical layout and standard medical procedures, including all key nodes, paths, and their attributes, specifically includes: Key nodes include at least: entrance / exit, registration / check-in point, waiting areas for each department, consultation room, examination / laboratory, payment window, pharmacy and auxiliary services; Define the function, physical location, associated departments, and standard service duration for each key node; Define the start point, end point, physical length, path type, key signs and auxiliary facilities along the route for each path; Based on the hospital's key nodes, paths, and attributes, a structured digital map was constructed using CAD software. The structured digital map constructed using CAD software includes: a base layer, a node layer, a path layer, a marker layer, a facility layer, and data association.

[0016] To explain this, accurately guiding elderly patients to their desired location or specific destination requires first constructing an accurate, structured digital map containing all key nodes, paths, and their attributes. In a specific implementation case, the data collection scope is defined, clearly defining the building area of ​​the hospital's outpatient building (including multi-building interconnected scenarios), including all floors (several floors above ground and several floors underground), public passageways (main corridors, branch corridors, elevator lobbies, stairwells), and functional areas (treatment areas, examination areas, payment areas, medication dispensing areas, rest areas, and restrooms). The data is then structured around the sequence: "Patient admission → registration / check-in → waiting → consultation → examination / testing → payment → medication dispensing → discharge." The standard process is at its core, and it simultaneously covers derivative process nodes such as "midway consultation (service desk), temporary rest, and restroom use" to ensure that the map matches the actual medical trajectory. The hospital's coordinate system is constructed with the bottom of the central pillar in the outpatient hall as the global coordinate origin, the X-axis pointing along the main corridor into the depths of the outpatient department, the Y-axis perpendicular to the main corridor, and the Z-axis representing the floor height. All length data is in meters, retained to two decimal places, and time data is in minutes, retained to one decimal place. Location coordinates are represented by a three-dimensional coordinate system of "(X, Y, Z)," retained to two decimal places. Standardized terminology and unified attribute naming are used for node functions, path types, and identifier types. The key nodes and their classification methods are shown in Example Table 1:

[0017] Table 1. Examples of Key Nodes and Their Classification Methods To achieve accurate quantitative analysis of the outpatient journey of elderly patients, the continuous medical space and service process must first be deconstructed into discrete, functionally defined key nodes. Based on the core functions and spatial attributes of the nodes, this scheme establishes a multi-level key node classification system as shown in Table 1. This system follows the following construction logic: Process integrity principle: The node sequence must cover the complete medical treatment flow from "entry" to "exit" to ensure no breakpoints. All standard outpatient procedures (registration, waiting, consultation, examination, payment, and medication) are abstracted into corresponding core process nodes. Function-space correspondence principle: Each node is both the execution point of a service function (e.g., "registration") and a clearly defined physical spatial area (e.g., "3 meters in front of manual windows 1-3 in the outpatient hall"). This provides a precise coordinate mapping basis for subsequent location-based spatiotemporal data binding; The principle of age-appropriateness: In addition to the core process nodes, a special category of "auxiliary service nodes" has been added, which covers supportive spaces that are crucial to the physical and physiological needs of elderly patients, such as restrooms (especially accessible restrooms), drinking water areas, and rest areas, to ensure that the journey map can comprehensively reflect various environmental factors that affect the medical experience of the elderly. The principle of mutually exclusive categories and clear hierarchy: Node categories (such as "registration / check-in nodes" and "payment / settlement nodes") are functionally mutually exclusive to avoid confusion in data statistics. At the same time, the classification system presents a clear hierarchical structure, which facilitates data analysis at different granularities (such as analyzing the overall efficiency of "waiting nodes" or conducting in-depth analysis of the congestion of "internal medicine waiting areas"). As shown in Example Table 1, this classification system constitutes a unified data model and reference framework for all subsequent quantitative analyses (such as node dwell time, transfer paths between nodes, and voice guidance). The function, physical location, associated department, and standard service duration of each key node are defined. Taking the No. 1 manual registration window node in the outpatient hall as an example, its structured attributes are defined as follows: Node ID: NODE-101-Registration Window-1; Core functions: outpatient registration, medical insurance card linking; Additional functions: medical card application, registration information inquiry; Physical coordinates: (X: 10.50, Y: 5.20, Z: 3.00) (a local coordinate system established based on the origin of the outpatient hall); Location description: Located on the east side of the outpatient hall, about 15 meters from the main entrance, next to the first window on the north side of the self-service registration machine area; Related departments: Outpatient registration and payment office, information technology department (system support); Standard service duration: Single person service duration: 2.5 minutes, peak waiting time: 15.0 minutes (occurring time: 8:30-9:30 am on weekdays); Based on all attribute information, generate a unique identifier ID for each node and associate it with the corresponding location on the floor plan in the digital map; The path categories and lists are shown in Example Table 2:

[0018] Table 2. Path Classification and List Examples The specific implementation steps are as follows: Step 1: Based on the node distribution, identify all valid paths between nodes (excluding closed channels and employee-only channels) to ensure that there is at least one main path and one alternative path between any two adjacent functional nodes; Step 2: Use a laser rangefinder to measure the length segment by segment along the center line of the path. For paths that cross floors, it is necessary to measure the actual walking distance (not vertical height) of the elevator / stairs. Step 3: Record the signs along the route one by one, take photos of the signs with a panoramic camera, and simultaneously label the signs with their coordinates, color, shape and other attributes; Step 4: Record the location coordinates, facility type, quantity, and functional attributes of auxiliary facilities along the route, and associate them with unique facility identifiers (IDs); Step 5: Generate a unique path identifier ID based on the path attributes, draw the path centerline on the digital map, and associate the starting point, ending point nodes, and signs and facilities along the way; The construction of structured digital maps in CAD software specifically includes: Basic layer: Outline of the outpatient building, walls, doors, windows, columns and other physical structures; Node layer: Location, boundary, and attribute annotations for all key nodes; Path layer: Centerline, length, and attribute labels for all paths; Signage layer: Location, style, and text information of key signs along the route; Facility layer: Location, type, and attribute labels for auxiliary facilities; Data association: Establish cross-layer associations through node ID, path ID, and facility ID to realize interactive functions such as "get associated paths through nodes" and "get signs / facilities along the way through paths"; When a hospital undergoes physical renovations (such as department relocation or the establishment of new examination rooms), authorized administrators can update the corresponding layers (nodes, paths, facilities) using CAD software editing tools. The system automatically assigns a unique ID to the new element and records the version and time of the change.

[0019] The process of constructing a voice-guided medical model based on a structured digital map, and providing precise navigation based on the voice needs of elderly patients, specifically includes: A voice-guided medical model is constructed based on a structured digital map; the voice-guided medical model takes the voice needs of elderly patients as input and navigation guidance instructions as output. The voice-guided medical model includes voice-to-text conversion, text recognition, path planning, and guidance functions. The speech-to-text function refers to converting the speech needs of elderly patients into text features. The text recognition function refers to matching the converted text features with a medical text database to identify the text features that meet the speech needs of elderly patients. The path planning function refers to obtaining a guidance path based on the text features of the elderly patient's speech needs, using location coordinates and a structured digital map. The guidance function refers to providing guidance based on the acquired guidance path, through voice broadcasting and / or the presentation of a 3D digital map.

[0020] The purpose of building the voice-guided medical model is to create a closed-loop intelligent service system that can understand, make decisions, and execute, in order to replace traditional static signage or manual inquiry, and provide end-to-end accompanying navigation for elderly patients. This model achieves this goal through the collaborative work of four functional modules. The speech-to-text function serves as a perception entry point, aiming to transform the most natural interaction method (speech) of elderly patients into information that can be processed by computation. This function is achieved by integrating existing mature adaptive speech recognition services in the medical field (such as cloud-based ASR services based on deep learning architecture) or lightweight speech recognition engines that can be deployed offline. For example, it can be integrated and converted online by calling standard API interfaces such as Alibaba Cloud Intelligent Voice Interaction, Microsoft Azure Speech Services, or iFlytek Open Platform. As the core of cognition, the text recognition function aims to bridge the "semantic gap" between the colloquial expressions of elderly patients and the terminology in the hospital's standardized database. Through intent recognition and semantic vector matching, it accurately maps vague user needs to specific nodes in a structured digital map, transforming unstructured language into structured spatial targets. As a decision engine, the route planning function aims to formulate a safe, efficient, and physiologically appropriate route for the elderly after obtaining a clear spatial goal. This function not only considers the shortest distance but also incorporates the constraint of minimizing the total waiting time of the route, aiming to minimize the overall time and physical exertion of elderly patients and achieve optimal route planning. The guidance and instruction function serves as the execution and interaction terminal. Its purpose is to transform the abstract path planning results into multimodal guidance that is easy for the elderly to perceive and follow. Its function is to call the navigation instruction sequence and path session ID generated by the path planning function, convert the instruction sequence into voice broadcast through speech synthesis technology, and / or highlight the corresponding path in the 3D digital map interface, and update the patient's location in real time. Based on the matching of the patient's real-time coordinates with the coordinates of key points in the instruction sequence, the next voice instruction is dynamically triggered to achieve 'wherever you go, I will say it' companion guidance.

[0021] Reference Figure 2 As shown, the text recognition function specifically includes: Define and include standard terms for key nodes, paths and their attributes in structured digital maps, as well as commonly used expressions for elderly patients in historical data, and build a structured terminology library; Based on publicly available and compliant medical text data, combined with transcripts of real conversations between elderly patients and hospital staff and nurses, a basic training sample set was constructed. The BERT pre-trained model based on the Transformer architecture automatically mines the semantic associations between intent types and key nodes from the training base sample set; Construct a mapping subset table between intent types and key nodes, and automatically match the corresponding node subset based on the intent type labels output by the BERT pre-trained model; Based on the matched node subset, and using the nearest neighbor search method based on the structured terminology database, the cosine similarity between the query semantic vector and all term vectors belonging to the node subset in the terminology database is calculated. Set a similarity threshold and determine whether the obtained cosine similarity is greater than the threshold: if yes, bind the node to the terms in the library in a consistent manner and record and store them in descending order; if no, it means that the node lacks a corresponding term in the library, and the top N nodes in descending order of similarity will be used as a candidate node list and an abnormal warning will be issued. An expert review process is set up, which uses expert experience to review the patient's verbal expression, the core entities extracted, the intent type, the candidate node list, and the location of the corresponding node on the digital map. The data reviewed by experts is re-incorporated into the training base sample set and structured terminology library, and feedback labels are set to optimize and update the BERT pre-trained model; the feedback labels include: pass, correct, and reject.

[0022] This can be explained by the fact that, in order to enable elderly patients to conveniently obtain accurate navigation through natural language, it is necessary to bridge the "semantic gap" between their colloquial expressions and the standardized location information of the hospital. To this end, this solution lays the foundation for semantic alignment by constructing a standardized terminology library that is fully aligned with the hospital's structured digital map and incorporating the high-frequency colloquial expressions of elderly patients. Based on the BERT pre-trained model with the Transformer architecture, the patient's speech is converted into structured intent labels and semantic vectors. Then, the search range is quickly locked by mapping subsets of intent types and key nodes, and the semantic vector similarity calculation is used to achieve intelligent matching from vague colloquialisms to precise locations. This automatically transforms unstructured user needs into navigation instructions that can be recognized by the structured digital map, forming the core decision engine of the intelligent triage service. It should be explained that the training base sample set specifically includes: The intent types include: department search, doctor search, examination item search, facility search, process inquiry, location navigation, report collection inquiry, and payment consultation. Several labeled sample libraries were constructed, with samples specific to elderly patients accounting for ≥60%, covering all intent types throughout the process. A cross-validation mechanism of dual annotation and third-party arbitration was adopted to include data with annotation consistency of over 90% into the sample set. The data was then divided into training set, validation set, and test set in a ratio of 7:1:2. The BERT pre-trained model is a single-task model. Its core task is to classify the intent of the input sentence. The model takes the sentence as input, and after encoding, the final hidden state of the [CLS] flag is used as input for two parts: one is the intent classification head, which maps the [CLS] vector to the probability distribution of intent type through a fully connected layer; the other is to directly extract the [CLS] vector as the query semantic vector for subsequent vector similarity matching. Using intent type as the classification label, output the probability distribution of the core intent type of the statement; With intent classification as the objective, a loss function is constructed using cross-entropy loss. Performance metrics: Intent classification accuracy ≥ 95%; if not, supplement labeled samples and retrain. The mapping subset table between intent types and key nodes is shown in Example Table 3:

[0023] Table 3. Example of a subset of intent types mapped to key nodes. As shown in Table 3, the intent type labels output by the BERT pre-trained model are used to automatically match the corresponding node subsets, and subsequent matching is only performed within the subsets. This limits the number of cosine similarity calculations for the subsets, reduces the amount of computation, and improves computational efficiency. The implementation process of the text recognition function is as follows: The text features of the speech input by elderly patients are obtained through speech-to-text conversion and then preprocessed. The preprocessed text features are input into the trained BERT pre-trained model, and the model outputs the intent type label and the query semantic vector ([CLS] vector). Based on the intent type label with the highest probability, the corresponding node subset is obtained through the mapping subset table between intent type and key node; Within the scope of this node subset, calculate the cosine similarity between the semantic vector of the query statement and the vectors of all corresponding terms in the terminology database, and match the corresponding terms in the database according to the similarity threshold. The matching results are verified and feedback is provided through expert verification. In actual implementation, the expert verification process needs to be translated into user feedback, and corrections and optimizations are made based on actual user feedback. It should be noted that expert verification prioritizes high-frequency inquiry content (such as medical conversations of elderly patients and high-frequency questions in service desk consultation records), while other scenarios are continuously adjusted based on actual user feedback. The initial baseline value for the similarity threshold is determined as follows: Based on a validation set with real labels, for each query, the cosine similarity between its query semantic vector and the uniquely correct standard term vector is calculated to obtain a set of positive sample similarity scores. The distribution of these scores is analyzed, and the 95th percentile is taken as the initial baseline threshold to ensure that the system can cover the vast majority (95%) of known correct matches. After system deployment, the similarity scores of valid user queries are continuously collected for statistical analysis.

[0024] Reference Figure 3 As shown, the path planning function specifically includes: The real-time coordinates of the elderly patient are obtained as the starting point of the path, and the coordinates and unique ID of the target node are obtained from the text recognition results. Based on the coordinates of the starting point and the ending point, relevant path network topology, length, type and facility information are extracted from the path layer and node layer of the structured digital map; The optimal path is determined by minimizing both distance and total waiting time. Based on the A* algorithm, the path node with the smallest A* algorithm result is selected to determine the route. The continuous path is decomposed into several straight navigation segments according to key nodes, and the guidance instructions of all navigation segments are combined in sequence to form a complete navigation instruction sequence. The basic layer, node layer, path layer, sign layer, facility layer, and their associated attribute identifiers in the structured digital map are simultaneously bound to the navigation instruction sequence. Generate a globally unique path session ID and store the relevant data together for real-time access by the guidance instruction function. The relevant data includes: path geometric coordinate sequence, structured guidance instruction sequence, estimated total time and total distance, starting point, ending point coordinates and node ID.

[0025] One explanation is that simply generating navigation instructions based on the shortest physical path on a map may not provide the optimal medical experience in complex and dynamic environments like hospitals. For elderly patients with limited mobility and physical strength, a route that is "shortest" but requires a long wait in crowded windows may actually consume far more physical and time resources than a route that is "slightly longer" but allows for "instant service." Therefore, the core contradiction that needs to be addressed is: how to find the best balance between the two mutually constraining goals of "walking distance" and "waiting time" to meet the actual needs of elderly patients. Based on the A* algorithm, with the dual optimization objectives of "optimal distance" and "shortest total path waiting time", the final route is determined by selecting the path node combination with the minimum algorithm calculation result (total cost), specifically including the following: Define the core optimization objective and cost function: The overall cost function of the A* algorithm is constructed by assigning the physical length cost of the path to "optimal distance" and the real-time waiting time cost of the target node (e.g., registration window, payment window, medication dispensing window) to "shortest total path waiting time". ,in, The cumulative actual cost from the starting point to the current node n is obtained by weighted summation of two parts (which need to be normalized to eliminate the influence of dimensions): one is the physical walking distance from the starting point to the current node, and the other is the actual waiting time of all business windows passed before the current node (obtained and accumulated in real time through the hospital information system interface). The weights are set according to the needs of elderly patients (e.g., distance weight 0.4, waiting time weight 0.6, prioritizing reducing waiting time). To estimate the cost from the current node n to the destination, a design based on static spatial information is used, which only reflects an optimistic estimate of the physical distance of the path. Manhattan distance correction is adopted (adapted to the grid layout of the hospital), i.e., h(n) = Manhattan distance(n, destination), to ensure that h(n) never overestimates the actual walking distance from the current node to the destination, satisfying the acceptability requirement of the heuristic function of the A* algorithm and ensuring the search capability for the optimal solution; Initialize path search parameters: Using the starting point of the elderly patient's visit (such as the main entrance of the outpatient department) as the initial node and the target node (such as a department clinic or pharmacy window) as the endpoint, load the node and path network in the hospital's structured digital map. At the same time, obtain the current waiting time data of each window in real time through the hospital information system interface. Initialize an open list to store nodes to be evaluated and a closed list to store nodes that have been evaluated. Add the starting point to the open list and set its cumulative actual cost to zero. Its total cost is the value of the above-mentioned estimated cost part. Iterative search for the optimal path node: Select the node with the smallest total cost f(n) from the open list as the current node for expansion. If the current node is the destination, the search is successful and terminates. Otherwise, traverse all adjacent reachable nodes of the current node (excluding obstacle nodes such as walls and closed areas). For each adjacent node: calculate its new cumulative actual cost g(n), which is the cumulative actual cost of the current node plus the incremental distance cost generated by moving from the current node to the adjacent node. If the adjacent node is a service window, the real-time waiting time cost of the window needs to be added in addition (the waiting time at the current moment is obtained through the hospital information system interface and used as a temporary cost item for the node). The incremental part is calculated according to the weight and normalization rules; its estimated cost h(n) is calculated: determined based on the Manhattan distance between the position of the adjacent node and the destination; the total cost of the adjacent node is obtained; if the node is not in the open list, it is added to the open list, and its predecessor node is recorded as the current node and the updated g(n), h(n), f(n); if the node is already in the open list, the newly calculated f(n) is compared with the original recorded total cost. If the new value is smaller, its cost data and predecessor node are updated. After processing all adjacent nodes, the current node is moved into the closed list. This selection, expansion, and update process is repeated until the destination node is found or the open list is empty. Determining the final route: After the search terminates, starting from the endpoint node, backtracking its predecessor nodes in sequence until returning to the starting point, thus forming a complete and ordered sequence of path nodes, such as "Outpatient Main Entrance → Self-service Registration Machine → Elevator → 3rd Floor Internal Medicine Waiting Area → Internal Medicine Clinic → Elevator → 1st Floor Payment Window → Outpatient Pharmacy". The total cost of the A* algorithm corresponding to this path node sequence is the minimum among all possible paths, where g(n) accumulates the total actual walking distance and the waiting time cost of the windows already passed, and h(n) ensures the optimality of the search process. Under the set weight system, the best balance between walking distance and total waiting time is achieved, providing elderly patients with a navigation route with the best overall cost.

[0026] The process of creating a wearable mobile sensing terminal, binding it to the elderly patient's identity information, and constructing a cloud-based and lightweight sensing terminal architecture specifically includes: The hardware integration of the wearable mobile sensing terminal includes: an identity binding module, an indoor positioning module, a lightweight interaction module, and a data communication module. Identity binding module: It has a built-in, non-rewritable, unique device identifier, which is softly bound to the patient's ID for that visit through the hospital information system interface when the patient picks up the device. Indoor positioning module: integrates UWB and / or Bluetooth AoA chips, and collects the three-dimensional coordinates of wearable mobile sensing terminals in real time under the coverage of the positioning base station in the hospital. Lightweight interactive module: The front of the device features a physical help button, a miniature speaker, and / or a 3D display; Data communication module: integrates Bluetooth Low Energy and Wi-Fi to encrypt and upload location data and interaction events between the wearable mobile sensing terminal and the cloud system; A cloud-based and lightweight sensing terminal architecture is constructed. The cloud layer integrates a structured digital map and a voice-guided medical model for reasoning and computation, while the lightweight sensing terminal is used to collect patient voice and output guidance instructions.

[0027] The fundamental purpose of taking the steps of "creating wearable mobile sensing terminals" and building a "cloud and lightweight terminal architecture" is to establish a reliable physical carrier and data foundation for the intelligent triage system that can bridge the digital divide and ensure the universality of services. Through this dedicated terminal, three core objectives are aimed at achieving: First, to provide objective and continuous spatiotemporal and behavioral data sources for quantitative analysis, transforming "pain point identification" from subjective research to data-driven approaches; second, to provide elderly patients who are not proficient in operating smart devices with a zero-learning-cost service entry point of "one-click help and voice interaction," ensuring the accessibility of technology; and third, through the collaborative architecture of "cloud intelligence and lightweight terminal," to control terminal costs and power consumption while ensuring the powerful functions and sustainable iteration of the system, ensuring the engineering feasibility of large-scale deployment. In the specific implementation case, a complete interaction will be used as an example for illustration: Patients can pick up a terminal at the entrance or scan the terminal's QR code to bind and activate it with their current visit ID in the hospital information system; After the terminal is activated, the UWB positioning module is started, and the three-dimensional coordinates are collected in real time under the coverage of the positioning base station in the hospital and the positioning is completed online. The initial position is reported to the cloud simultaneously. The patient presses the help button and asks, "How do I get to the ECG room?" The terminal receives the recorded speech and uploads it to the cloud-based voice guidance model via Wi-Fi for processing. The cloud generates a sequence of navigation instructions based on the voice recording, and sends the instructions to the terminal, which then broadcasts or displays them through its miniature speaker and / or 3D display. The cloud continuously verifies in the background whether the terminal is following the path based on the terminal's real-time location data, and triggers replanning when deviation occurs; The system automatically determines whether the patient has successfully reached the target node based on the terminal's positioning data. The success of the navigation and whether the patient needs to seek help again midway are all recorded and stored in the cloud. The system obtains feedback tags from elderly patients regarding their chosen route guidance through voice input, which are then received and stored in the cloud.

[0028] The process of setting up a control group and conducting test experiments to verify and evaluate the credibility and user acceptance of the voice-guided medical model specifically includes: Two groups of elderly patients with the same number of participants were set up in the experimental group, one group did not wear the mobile sensing terminal and the other group wore the mobile sensing terminal. Set the same starting point, ending point, and requirements, conduct test experiments, and record the first-time completion rate, active help rate, movement path distance, and task completion time for each person in each group; For each tester, obtain their evaluation score for this task on a 10-point scale; Based on the results of the test experiments, the credibility and user acceptance of the voice-guided medical model were verified and evaluated.

[0029] The explanation is that setting up a control group to conduct test experiments provides objective and quantifiable empirical evidence for the effectiveness of the voice-guided medical model through rigorous scientific comparison. Specifically, firstly, it assesses absolute effectiveness by comparing the differences in task completion rate, time spent, and walking distance between the wearing group and the non-wearing group, quantifying the actual effectiveness in solving core pain points such as "difficulty in finding a way" and "ineffective walking." Secondly, it assesses user experience and acceptance by combining the rate of users actively seeking help and subjective ratings to measure the clarity, ease of use, and overall satisfaction of the elderly population, ensuring the "age-friendliness" of the technology. This combines objective data with subjective feedback, providing a solid scientific basis for the practical application value of this solution. It should be noted that the elderly patients in the two experimental groups, with the same number of participants, were all elderly patients without prior medical experience at this hospital (or with no recent medical records) to eliminate interference from prior environmental cognition.

[0030] Furthermore, the method according to the embodiments of this application can also be achieved by means of... Figure 4 The architecture of the electronic device shown is used to implement this. For example... Figure 4 As shown, the electronic device 500 may include a bus 501, one or more CPUs 502, a read-only memory (ROM) 503, a random access memory (RAM) 504, a communication port 505 connected to a network, an input / output component 506, a hard disk 507, etc. The storage device in the electronic device 500, such as the ROM 503 or the hard disk 507, may store the "journey map"-based pain point identification and optimization method for elderly patients' outpatient visits provided in this application. The electronic device 500 may also include a user interface 508. Of course, Figure 4 The architecture shown is merely exemplary and can be omitted as needed when implementing different devices. Figure 4 One or more components in the illustrated electronic device.

[0031] Figure 5This is a schematic diagram of a computer-readable storage medium structure provided in one embodiment of this application. Figure 5 The diagram illustrates a computer-readable storage medium 600 according to one embodiment of this application. The computer-readable storage medium 600 stores computer-readable instructions. When executed by a processor, the computer-readable instructions can perform an optimization method for identifying pain points in outpatient medical visits for elderly patients based on a "journey map," as described in the above figures, according to an embodiment of this application. The storage medium 600 includes, but is not limited to, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0032] In summary, the advantages of this invention are as follows: by constructing a structured digital map that integrates the physical space of the hospital with the medical process, and combining a voice-guided medical model adapted to the speech habits of the elderly with a lightweight wearable sensing terminal, accurate and accompanying navigation for elderly patients seeking outpatient medical treatment is achieved. The effectiveness of the solution has been verified through controlled experiments, which significantly reduces the difficulty of finding the way and the physical exertion for elderly patients seeking medical treatment, and improves the medical experience and efficiency.

[0033] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A "journey map" based method for identifying and optimizing pain points of geriatric patients in outpatient clinics, characterized in that, include: Based on the hospital's physical layout and standard medical procedures, a structured digital map containing all key nodes, paths, and their attributes is constructed. Based on structured digital maps, a voice-guided medical model is constructed, and precise navigation is provided through the voice-guided medical model according to the voice needs of elderly patients; Create wearable mobile sensing terminals, bind elderly patients' identity information, and build a cloud-based and lightweight sensing terminal architecture; By setting up a control group, a test experiment was conducted to verify and evaluate the credibility and user acceptance of the voice-guided medical model.

2. The method for identifying and optimizing pain points in outpatient medical visits for elderly patients based on a "journey map" as described in claim 1, characterized in that, The construction of a structured digital map based on the hospital's physical layout and standard medical procedures, including all key nodes, paths, and their attributes, specifically includes: Key nodes include at least: entrance / exit, registration / check-in point, waiting areas for each department, consultation room, examination / laboratory, payment window, pharmacy and auxiliary services; Define the function, physical location, associated departments, and standard service duration for each key node; Define the start point, end point, physical length, path type, key signs and auxiliary facilities along the route for each path; A structured digital map was constructed using CAD software based on the hospital's key nodes, paths, and attributes. The structured digital map constructed using CAD software includes: a base layer, a node layer, a path layer, a marker layer, a facility layer, and data association.

3. The method for identifying and optimizing pain points in outpatient medical visits for elderly patients based on a "journey map" as described in claim 2, characterized in that, The process of constructing a voice-guided medical model based on a structured digital map, and providing precise navigation based on the voice needs of elderly patients, specifically includes: A voice-guided medical model is constructed based on a structured digital map; the voice-guided medical model takes the voice needs of elderly patients as input and navigation guidance instructions as output. The voice-guided medical model includes voice-to-text conversion, text recognition, path planning, and guidance functions. The speech-to-text function refers to converting the speech needs of elderly patients into text features. The text recognition function refers to matching the converted text features with a medical text database to identify the text features that meet the speech needs of elderly patients. The path planning function refers to obtaining a guidance path based on the text features of the elderly patient's speech needs, using location coordinates and a structured digital map. The guidance function refers to providing guidance based on the acquired guidance path, through voice broadcasting and / or the presentation of a 3D digital map.

4. The method for identifying and optimizing pain points in outpatient medical visits for elderly patients based on a "journey map" as described in claim 3, characterized in that, The text recognition function includes: Define and include standard terms for key nodes, paths and their attributes in structured digital maps, as well as commonly used expressions for elderly patients in historical data, and build a structured terminology library; Based on publicly available and compliant medical text data, combined with transcripts of real conversations between elderly patients and hospital staff and nurses, a basic training sample set was constructed. The BERT pre-trained model based on the Transformer architecture automatically mines the semantic associations between intent types and key nodes from the training base sample set; Construct a mapping subset table between intent types and key nodes, and automatically match the corresponding node subset based on the intent type labels output by the BERT pre-trained model; Based on the matched node subset, and using the nearest neighbor search method based on the structured terminology database, the cosine similarity between the query semantic vector and all term vectors belonging to the node subset in the terminology database is calculated. Set a similarity threshold and determine whether the obtained cosine similarity is greater than the threshold: if yes, bind the node to the terms in the library in a consistent manner and record and store them in descending order; if no, it means that the node lacks a corresponding term in the library, and the top N nodes in descending order of similarity will be used as a candidate node list and an abnormal warning will be issued. An expert review process is set up, which uses expert experience to review the patient's verbal expression, the core entities extracted, the intent type, the candidate node list, and the location of the corresponding node on the digital map. The data reviewed by experts is re-incorporated into the training base sample set and structured terminology library, and feedback labels are set to optimize and update the BERT pre-trained model; the feedback labels include: pass, correct, and reject.

5. The method for identifying and optimizing pain points in outpatient medical visits for elderly patients based on a "journey map" as described in claim 3, characterized in that, The path planning function includes: The real-time coordinates of the elderly patient are obtained as the starting point of the path, and the coordinates and unique ID of the target node are obtained from the text recognition results. Based on the coordinates of the starting point and the ending point, relevant path network topology, length, type and facility information are extracted from the path layer and node layer of the structured digital map; The optimal path is determined by minimizing both distance and total waiting time. Based on the A* algorithm, the path node with the smallest A* algorithm result is selected to determine the route. The continuous path is decomposed into several straight navigation segments according to key nodes, and the guidance instructions of all navigation segments are combined in sequence to form a complete navigation instruction sequence. The basic layer, node layer, path layer, sign layer, facility layer, and their associated attribute identifiers in the structured digital map are simultaneously bound to the navigation instruction sequence. Generate a globally unique path session ID and store the relevant data together for real-time access by the guidance instruction function. The relevant data includes: path geometric coordinate sequence, structured guidance instruction sequence, estimated total time and total distance, starting point, ending point coordinates and node ID.

6. The method for identifying and optimizing pain points in outpatient medical visits for elderly patients based on a "journey map" as described in claim 5, characterized in that, The process of creating a wearable mobile sensing terminal, binding it to the elderly patient's identity information, and constructing a cloud-based and lightweight sensing terminal architecture includes: The hardware integration of the wearable mobile sensing terminal includes: an identity binding module, an indoor positioning module, a lightweight interaction module, and a data communication module. The identity binding module includes: a built-in, non-rewritable, unique device identifier, which is softly bound to the patient's current visit ID through the hospital information system interface when the patient receives the device; The indoor positioning module includes: an integrated UWB and / or Bluetooth AoA chip, which collects the three-dimensional coordinates of wearable mobile sensing terminals in real time under the coverage of the positioning base station in the hospital. The lightweight interactive module includes: a physical help button and a miniature speaker and / or a 3D display on the front of the device; The data communication module includes: integrated low-power Bluetooth and Wi-Fi, which encrypt and upload location data and interaction events between the wearable mobile sensing terminal and the cloud system; A cloud-based and lightweight sensing terminal architecture is constructed. The cloud layer integrates a structured digital map and a voice-guided medical model for reasoning and computation, while the lightweight sensing terminal is used to collect patient voice and output guidance instructions.

7. The method for identifying and optimizing pain points in outpatient medical visits for elderly patients based on a "journey map" as described in claim 6, characterized in that, The process of setting up a control group and conducting test experiments to verify and evaluate the credibility and user acceptance of the voice-guided medical model includes: Two groups of elderly patients with the same number of participants were set up in the experimental group, one group did not wear the mobile sensing terminal and the other group wore the mobile sensing terminal. Set the same starting point, ending point, and requirements, conduct test experiments, and record the first-time completion rate, active help rate, movement path distance, and task completion time for each person in each group; For each tester, obtain their evaluation score for this task on a 10-point scale; Based on the results of the test experiments, the credibility and user acceptance of the voice-guided medical model were verified and evaluated.

8. A device for identifying and optimizing pain points in outpatient medical visits for elderly patients based on a "journey map," characterized in that... The apparatus is adapted to any one of the methods described in claims 1-7, the apparatus comprising: The map modeling module is used to construct a structured digital map containing all key nodes, paths and their attributes based on the hospital's physical layout and standard medical procedures. The voice interaction module is connected to the map building module. Based on the structured digital map, a voice-guided medical model is built. Based on the voice needs of elderly patients, the voice-guided medical model is used for precise navigation. The terminal module includes a wearable mobile sensing terminal, which is bound to the elderly patient's identity information and communicates with the voice interaction module to provide real-time navigation guidance during the elderly patient's medical treatment. The verification and evaluation module, connected to the terminal module and the voice interaction module, is used to conduct test experiments by setting up a control group, collect usage feedback data from elderly patients, and verify and evaluate the credibility and user acceptance of the voice-guided medical model.

9. An electronic device, characterized in that, include: At least one processor; And a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform an optimization method for identifying pain points in outpatient medical treatment for elderly patients based on a "journey map" as described in any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements any one of the methods for identifying and optimizing pain points in outpatient medical treatment for elderly patients based on a "journey map" as described in any one of claims 1-7.

Images