Dead cause chain auxiliary inference method and device

Abstract

The invention discloses a death chain auxiliary inference method and device. The method comprises the following steps: acquiring a death investigation record text; generating an ICD code list associated with the current death event based on a timeline according to the death investigation record text; determining a cause chain compiling rule, a cause chain compiling case, each cause chain and the probability thereof related to the current death event according to the ICD coding list; and the Prompt is utilized to prompt the large language model to determine the cause chain and the root cause of the current death event in combination with the cause chain compilation rule, the cause chain compilation case, the cause chain and the probability of the cause chain. By utilizing the scheme of the invention, the extraction of the dead cause information and the auxiliary deduction of the dead cause chain can be completed intelligently and automatically, and the precision and efficiency of the dead cause monitoring work are improved.

Description

Technical Field

[0001] The present invention relates to the field of cause of death monitoring statistics, and in particular to a method and device for auxiliary inference of cause of death chains. Background Art

[0002] Death registration and reporting, as well as statistics on causes of death, are fundamental tasks that involve the continuous and systematic collection of mortality data and comprehensive analysis to study mortality levels, causes of death, and trends and patterns. Health indicators such as life expectancy, maternal mortality, and infant mortality rates generated from this analysis provide crucial information reflecting national and regional socioeconomic levels and cultural development, providing a scientific basis for national socioeconomic development goals and health policy formulation.

[0003] Whether residents' death information is complete and the cause of death is accurately determined is the foundation of cause-of-death monitoring. There are two core issues involved in determining the cause of death: 1. Appropriate and unified classification standards, that is, what is the basis for determining the cause of death; and 2. How to determine the cause of death. When collecting cause-of-death statistics, if only one disease is involved, the classification of the cause of death is relatively simple; however, in most cases, death is caused by two or more diseases, and only one cause of death can be selected for cause-of-death statistics. Therefore, the World Health Organization defines "underlying cause" as: (a) the earliest disease or injury in a series of morbid events that directly led to death, or (b) the accident or violent situation that caused the fatal injury. In determining the underlying cause of death, a concept often involved is "whether there is a reasonable sequence." This sequence is the cause-of-death chain, which refers to the series of diseases or injuries that led to death recorded on the "Medical Certificate of Death" and have a logical temporal relationship and can be reasonably explained.

[0004] In cause-of-death statistics, the death chain is undoubtedly the most core information source. This chain, meticulously organized and linked according to a specific logical sequence, clearly outlines the development and inherent laws of a disease, presenting the complex causes of death in an orderly manner.

[0005] Compiling a high-quality cause-of-death chain requires a high level of expertise. Its successful implementation requires not only a solid foundation of clinical medical knowledge but also a precise grasp of the detailed rules and professional guidance for disease and death coding established by the World Health Organization. However, in reality, most physicians engaged in clinical diagnosis and treatment have a limited grasp of the relevant rules and guidance for disease and death coding, influenced by factors such as their daily work priorities and differences in their professional training systems. This undoubtedly poses certain challenges to the accurate and high-quality compilation of cause-of-death chains. Furthermore, in actual practice, a large number of death certificate reports have indeed been found to contain quality defects in the completion of cause-of-death chains. Such defective cause-of-death chain completion can directly mislead government decision-making and lead to inappropriate allocation of health resources. Summary of the Invention

[0006] The present invention provides a method and device for auxiliary inference of cause of death chains, which intelligently and automatically extract cause of death information and complete auxiliary inference of cause of death chains, thereby improving the accuracy and efficiency of cause of death monitoring work.

[0007] To this end, the present invention provides the following technical solutions:

[0008] A method for assisting inferring a chain of death causes, the method comprising:

[0009] Obtain text of death investigation records;

[0010] Generating a timeline-based ICD code list associated with the current death event according to the death investigation record text;

[0011] Determining the cause-of-death chain compilation rules, cause-of-death chain compilation cases, and each cause-of-death chain and its probability related to the current death event based on the ICD code list;

[0012] The Prompt language model is used to combine the cause of death chain compilation rules, the cause of death chain compilation cases, the cause of death chain and its probability to determine the cause of death chain and the root cause of death of the current death event.

[0013] Optionally, generating a timeline-based ICD code list associated with the current death event according to the death investigation record text includes:

[0014] extracting medical terms from the death investigation record text;

[0015] Performing semantic retrieval on the ICD code knowledge base according to the medical terminology to obtain ICD code retrieval results;

[0016] A timeline-based ICD code list associated with the current death event is generated according to the ICD code search result.

[0017] Optionally, generating a timeline-based ICD code list associated with the current death event based on the death investigation record text further includes:

[0018] Determine the optimal ICD code by using a large model driven by a prompt word engineering and combining the death investigation record text and the ICD code search results;

[0019] The step of generating a timeline-based ICD code list associated with the current death event according to the ICD code search result includes:

[0020] A timeline-based ICD code list associated with the current death event is generated according to the optimal ICD code.

[0021] Optionally, the medical terminology includes the following information: disease diagnosis, symptoms and signs.

[0022] Optionally, determining the cause of death chain compilation rules, cause of death chain compilation cases, and each cause of death chain and its probability related to the current death event according to the ICD code list includes:

[0023] According to the category of the ICD code in the ICD code list, a cause of death chain compilation rule library is searched to determine a cause of death chain compilation rule related to the current death event;

[0024] According to the ICD code in the ICD code list, searching the death cause chain compilation case library to determine the death cause chain compilation case related to the current death event;

[0025] The ICD code list is input into the N-Gram probability model to determine the death cause chains and their probabilities related to the current death event.

[0026] Optionally, the using of the Prompt large language model in combination with the cause of death chain compilation rules, the cause of death chain compilation case, and the cause of death chain and its probability to determine the cause of death chain and the root cause of death of the current death event includes:

[0027] Setting the role and inference rules of the Prompt prompt large language model;

[0028] According to the set reasoning rules, the cause of death chain compilation rules, the cause of death chain compilation cases, and the cause of death chain and its probability are input into the Prompt prompt large language model, and the cause of death chain and the root cause of death of the current death event are obtained based on the reasoning of the Prompt prompt large language model.

[0029] Optionally, the inference rules include: input information and input order, processing flow, output requirements, and constraints.

[0030] Optionally, the method further includes:

[0031] Determine the accuracy of the chain of causes and underlying causes of death in the current death;

[0032] If inaccurate, the stated chain of causes of death and the underlying cause of death will be corrected;

[0033] The corrected cause of death chain and underlying cause of death are added to the cause of death chain compilation case library.

[0034] A device for assisting inferring a chain of death causes, the device comprising:

[0035] A receiving module is used to obtain the text of the death investigation record;

[0036] An ICD code list generating module is used to generate an ICD code list associated with the current death event based on the timeline according to the death investigation record text;

[0037] An information determination module is used to determine the cause of death chain compilation rules, cause of death chain compilation cases, and each cause of death chain and its probability related to the current death event based on the ICD code list;

[0038] The comprehensive inference module is used to use the Prompt language model to combine the cause of death chain compilation rules, the cause of death chain compilation cases, and the cause of death chain and its probability to determine the cause of death chain and the root cause of death of the current death event.

[0039] Optionally, the ICD code list generation module includes:

[0040] a term extraction unit, configured to extract medical terms from the death investigation record text;

[0041] A retrieval unit, configured to perform a semantic search on an ICD code knowledge base based on the medical terminology to obtain an ICD code retrieval result;

[0042] A list generating unit is used to generate an ICD code list associated with the current death event based on a timeline according to the ICD code retrieval result.

[0043] Optionally, the ICD code list generation module further includes:

[0044] A screening unit, configured to determine an optimal ICD code by utilizing a large model driven by a prompt word engineering, in combination with the death investigation record text and the ICD code search result;

[0045] The list generating unit generates a timeline-based ICD code list associated with the current death event according to the optimal ICD code.

[0046] Optionally, the information determination module includes:

[0047] a rule determination unit, configured to search a cause of death chain compilation rule base according to the category of the ICD code in the ICD code list, and determine a cause of death chain compilation rule related to the current death event;

[0048] a case determination unit, configured to search a cause of death chain compilation case database according to the ICD code in the ICD code list, and determine a cause of death chain compilation case related to the current death event;

[0049] The probability determination unit is used to input the ICD code list into the N-Gram probability model to determine each cause of death chain related to the current death event and its probability.

[0050] Optionally, the comprehensive inference module includes:

[0051] A setting unit, configured to set the role and inference rules of the Prompt large language model;

[0052] The reasoning unit is used to input the cause of death chain compilation rules, the cause of death chain compilation cases, the cause of death chain and its probability into the Prompt prompt large language model according to the set reasoning rules, and obtain the cause of death chain and root cause of death of the current death event based on the reasoning of the Prompt prompt large language model.

[0053] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which executes the steps of the death cause chain assisted inference method when the computer program is executed by a processor.

[0054] The death cause chain auxiliary inference method and device provided by the present invention are based on the diagnosis and treatment information reported in the deceased's hospitalization medical record and the key text information contained in the death investigation record. They comprehensively utilize pre-trained large language models, thought chains, knowledge base retrieval enhancement generation, N-Gram probability models and other technical means to intelligently extract the relevant disease ICD (International Classification of Diseases) codes from the descriptions of a series of diseases, pathological conditions or injuries that cause or promote death, as well as the accidents or violence that caused such injuries, from the text materials of the deceased's hospitalization medical record homepage and the death investigation record, and infer the logical relationship and causal order between the codes to provide a possible death cause chain and root cause of death.

[0055] Compared with the existing technology, the present invention not only utilizes the description of death events from a medical perspective by clinical diagnosticians or death investigators, but also combines the World Health Organization's rules and guidelines on disease and death coding, as well as the prior probabilities of the order of various diseases in the cause-of-death chain trained from historical cause-of-death chain sample data, thereby effectively improving the accuracy of the intelligently generated cause-of-death chain, improving the quality of cause-of-death monitoring data, and improving the accuracy and efficiency of cause-of-death monitoring work. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be derived from these drawings without inventive effort.

[0057] Figure 1 This is a diagram showing the basic format of the cause of death chain in the "Resident Death Medical Certificate";

[0058] Figure 2 This is a flow chart of a method for assisting inferring a cause of death chain provided by an embodiment of the present invention;

[0059] Figure 3 This is a flowchart of generating an ICD code list based on a death investigation record text in an embodiment of the present invention;

[0060] Figure 4 It is a structural diagram of a death cause chain auxiliary inference device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0061] The following describes the specific embodiments of the present invention in detail with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only used to illustrate and explain the present invention and are not intended to limit the present invention.

[0062] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0063] The cause of death chain is the core information basis for cause of death statistics. The basic format of the cause of death chain is as follows: Figure 1 As shown in the following figure, it can be divided into three parts:

[0064] ① Part I of the cause of death: This is the main content of the basic format. It is necessary to fill in the disease that caused the death and the earlier causes. It is a required part and requires the doctor to describe the sequential relationship of a series of diseases that caused someone's death.

[0065] ② Part II of the cause of death: This is a supplement to the content of Part I, and it is filled in with other meaningful circumstances that contributed to death but are not related to the disease or condition that caused death.

[0066] ③ The approximate time interval from the onset of each reported symptom or condition to death (time units: minutes, hours, days, weeks, months or years) can help determine the relationship between various diseases.

[0067] When determining whether the contents of Part I of the death certificate are in a reasonable order, the clinician's input should be fully considered as much as possible. Physicians need to fully understand and learn the requirements for completing a death certificate so that their input can express a reasonable order, that is, a reasonable chain of causes of death.

[0068] However, in actual work, we often encounter situations where "a reasonable order exists", "more than one order exists", and "no order exists", which are explained below with examples.

[0069] Example 1: There is a reasonable order, as shown in Table 1 below.

[0070] Table 1

[0071]

[0072] Example 2: There is more than one sequence, as shown in Table 2 below.

[0073] Table 2

[0074]

[0075]

[0076] Example 3: There is no order, as shown in Table 3 below.

[0077] Table 3

[0078]

[0079] Due to the existence of the above-mentioned situations, it also brings certain difficulties to the accurate determination of the chain of causes of death. To this end, the embodiment of the present invention provides a method and device for auxiliary inference of the chain of causes of death, which is based on the diagnosis and treatment information reported in the deceased's hospitalization medical record and the key text information contained in the death investigation record. It comprehensively utilizes pre-trained large language models, thinking chains, knowledge base retrieval enhancement generation, N-Gram probability models and other technical means to intelligently extract the relevant disease ICD codes from the first page of the deceased's hospitalization medical record and the text materials of the death investigation record for a series of diseases, pathological conditions or losses that cause or promote death, as well as accidents or violence that cause such injuries, and infer the logical relationship and causal order between the codes to provide possible chains of causes of death and root causes of death. It can be used to assist professional and technical personnel of medical and health institutions and disease prevention and control departments in reporting and inferring chains of causes of death, and improve the quality of cause of death monitoring data.

[0080] like Figure 2 FIG. 1 is a flow chart of a method for assisting inferring a cause of death chain according to an embodiment of the present invention, comprising the following steps:

[0081] In step 201, the death investigation record text is obtained.

[0082] In step 202, a timeline-based ICD code list associated with the current death event is generated according to the death investigation record text.

[0083] The process of generating the ICD code list is as follows Figure 3As shown, the following steps are included:

[0084] Step 301: extract medical terms from the death investigation record text.

[0085] Compared with traditional NLP (Natural Language Processing) technology, the large language model relies on the advanced Transformer architecture (an artificial intelligence model mainly used to understand and generate language) and has excellent capabilities for unsupervised learning of massive amounts of text. It can learn the subtle and critical long-distance dependencies within the language and achieve in-depth interpretation of the text.

[0086] To this end, in embodiments of the present invention, a pre-trained large language model can be used to extract medical terminology from death investigation records. For example, an industry-leading large language model pre-trained on large-scale, multi-domain text can be selected to ensure a broad and solid foundation of linguistic knowledge. For example, the Tongyi Qianwen large language model has been pre-trained on billions of texts covering fields such as medicine and technology, accumulating extensive experience in semantic understanding. Text to be interpreted, such as diagnosis and treatment information reported in hospital medical records and death investigation records, is fed into the large language model. Leveraging its deep understanding of language, the large language model analyzes the text structure sentence by sentence, identifying grammatical components such as subject, predicate, object, attributive, adverbial, and complement, and restoring the core context of the text. For example, for the medical text "The patient recently developed symptoms of heart failure due to long-term, uncontrolled hypertension, initially experiencing dyspnea and subsequently worsening lower limb edema," the large language model can accurately parse the patient's underlying disease, the secondary symptoms it causes, and the order in which these symptoms appear, providing a clear framework for subsequent analysis.

[0087] While pre-trained large language models possess powerful text parsing capabilities, enabling in-depth analysis of medical text, accurately extracting a patient's underlying illness and the resulting secondary symptoms from conventional language descriptions, and even clarifying the order in which symptoms presented based on textual clues, they also have certain limitations. When faced with highly specialized disease and death coding rules, rigorous inference rules for cause-of-death chains, and the precise probability calculation required to determine the order in which diseases or conditions led to death, large language models alone still cannot achieve satisfactory results.

[0088] Therefore, in an embodiment of the present invention, a pre-trained large language model can be used to extract medical terms from the death investigation record text to obtain all disease-related medical terms recorded in the death investigation record text, and these medical terms are arranged based on a time chain and have a chronological relationship.

[0089] Step 302: Perform semantic search on the ICD code knowledge base according to the medical terminology to obtain ICD code search results.

[0090] The ICD, developed by the World Health Organization (WHO), is an internationally recognized standard for the classification of diseases and health issues. It categorizes diseases into organized groups based on their etiology, pathology, clinical manifestations, and anatomical location, and uses coding to represent these groups.

[0091] The ICD coding knowledge base can be a corresponding database provided by the WHO query platform or the national medical insurance information business coding standard data, and can be queried through the query interface of the corresponding platform or through the ICD disease coding query tool.

[0092] Step 303: Generate a timeline-based ICD code list associated with the current death event according to the ICD code search result.

[0093] In order to better ensure the accuracy of the generated ICD code list, in some embodiments, a prompt word engineering-driven large model can be used to combine the death investigation record text and the ICD code retrieval results to determine the optimal ICD code; then, a timeline-based ICD code list associated with the current death event can be generated based on the optimal ICD code.

[0094] Prompt word engineering is a key technology in natural language processing. It involves designing and constructing input prompts to effectively guide a large language model to produce the desired output. The large model driven by prompt word engineering can be a Reasoning Large Language Model (LLM). It should be noted that the solution of the present invention does not rely on a specific large language model; any large language model with reasoning capabilities will be acceptable.

[0095] For example, in a non-limiting embodiment, the prompt words are set as follows:

[0096] #Character Setting

[0097] You are a professional medical coding reviewer, tasked with accurately coding medical death certificates based on the WHO ICD-10 coding rules. You possess knowledge of forensic pathology and clinical diagnostic reasoning skills.

[0098] #Processing Flow

[0099] Please strictly follow the following order:

[0100] 1. Timeline reconstruction

[0101] When parsing death investigation records, a structured timeline must be established:

[0102] - Chronological arrangement of all key medical events

[0103] -Mark the time of each event (accurate to day / hour)

[0104] - Identify the causal logic chain between events

[0105] -Mark abnormal points in time intervals (such as sudden deterioration)

[0106] 2. Dynamic code matching

[0107] Based on the candidate encoding list, perform:

[0108] a) Time series mapping: associate the event at each time node with the corresponding ICD code

[0109] b) Causal verification: Confirm that the coding follows the ICD causal chain rules (e.g., infection → sepsis → MODS)

[0110] c) Stage markers: distinguish the code sets corresponding to the initiating event, progress event, and terminal event

[0111] d) Pathway integrity: retaining the necessary intermediate codes in the course of disease

[0112] 3. Decision output

[0113] Output line by line according to the timeline and causal chain relationship, with each line outputting an ICD code and the corresponding time.

[0114] #Input data

[0115] Investigation record text:

[0116] Candidate ICD code list:

[0117] #Special Case Handling

[0118] Special instructions are required in the following situations:

[0119] - If poisoning / injury occurs, the external cause must also be recorded (add XX code)

[0120] - Tumors should be distinguished between primary and secondary sites

[0121] - Perinatal deaths require the use of special codes P95-P96

[0122] - When there are multiple codes of equal priority, "Code Ambiguity" should be marked.

[0123] It should be noted that, in specific implementation, other processing flows may also be adopted, which is not limited in the embodiment of the present invention.

[0124] Continue to refer to Figure 2 In step 203, the cause of death chain compilation rules, cause of death chain compilation cases, and each cause of death chain and its probability related to the current death event are determined according to the ICD code list.

[0125] Specifically, a cause-of-death chain compilation rule base can be retrieved based on the category of the ICD code in the ICD code list to determine the cause-of-death chain compilation rule relevant to the current death event. This cause-of-death chain compilation rule base can be developed by the cause-of-death coding authority based on international ICD coding rules and domestic practices. For example, it may include, but is not limited to, a list of ICD codes that cannot be used as the underlying cause of death.

[0126] Then, according to the ICD code in the ICD code list, the cause of death chain compilation case library is searched to determine the cause of death chain compilation case related to the current death event. The cause of death chain compilation case library can be established by the death cause coding business department and organized by experts for review.

[0127] Then, the ICD code list is input into the N-Gram probability model to determine the death cause chains and their probabilities related to the current death event.

[0128] The N-Gram probability model is a widely used statistical language model in natural language processing. Its core principle is to segment text into N consecutive characters or words, forming N-Gram units. The N-Gram probability model uses these N-Gram units to predict the next word or character in the text. Based on the statistical information of large amounts of text data, it calculates the frequency of occurrence of different N-Gram units to infer the probability of the next element appearing after a given preceding text sequence. For example, in a large news corpus, the term "Beijing Tiananmen" appears frequently. When the model encounters "Beijing," based on previously calculated N-Gram frequency information, it can predict that the next word with a high probability of appearing is "Tiananmen."

[0129] The cause of death chain is composed of a disease diagnosis sequence. For example, an example of a cause of death chain is as follows:

[0130] (a) I46.1 Sudden cardiac death

[0131] (b) I50.9 Heart failure

[0132] (c) I24.9 Acute coronary syndrome

[0133] (d) I25.1 Coronary atherosclerotic heart disease

[0134] The probability of a chain of causes of death can be calculated using the chain rule as follows:

[0135] A death chain sample C consisting of at most 4 diseases is recorded as:

[0136] C=c1,c2,c3,c4;

[0137] Then its probability P(C) can be calculated using the following formula:

[0138] P(C)=P(c1)P(c2|c1)P(c3|c1c2)P(c4|c1c2c3);

[0139] The meaning of each parameter is as follows:

[0140] P(c1) is the probability that the first disease diagnosis or injury or poisoning event in the chain of death is C1;

[0141] P(c2|c1) is the conditional probability that the second disease diagnosis or injury or poisoning event in the death chain is C2 when the first disease diagnosis or injury or poisoning event is C1;

[0142] P(c3|c1c2) is the conditional probability that the third disease diagnosis or injury or poisoning event in the death chain is C3 when the first disease diagnosis or injury or poisoning event in the death chain is C1 and the second disease diagnosis or injury or poisoning event is C2;

[0143] P(c4|c1c2c3) is the conditional probability that the fourth disease diagnosis or injury or poisoning event in the death chain is C4 when the first disease diagnosis or injury or poisoning event is C1, the second disease diagnosis or injury or poisoning event is C2, and the third disease diagnosis or injury or poisoning event is C3;

[0144] These probabilities can be estimated using the frequencies in the death chain training sample library as follows:

[0145] 1. Count N-Gram frequencies: Count the number of occurrences of the 2-4 cause-of-death codes (i.e., N-Grams) that make up the cause-of-death chain in the training cause-of-death chain sample library, and generate a lookup table of all possible N-Grams and their corresponding frequencies.

[0146] 2. Based on the maximum likelihood estimation (MLE), the N-Gram probability is expressed as conditional probability. For example, P(c4|c1c2c3) represents the conditional probability as follows: the number of samples with the cause of death chain (c1, c2, c3, c4) divided by the number of samples with the first three digits of the cause of death being (c1, c2, c3).

[0147] In step 204, the Prompt language model is used to combine the cause of death chain compilation rules, the cause of death chain compilation cases, the cause of death chain and its probability to determine the cause of death chain and the root cause of death of the current death event.

[0148] Prompt is a technology based on artificial intelligence (AI) instructions. It explicitly and specifically directs the output of a language model to determine the task or type of text the user wants the model to complete. Prompt encompasses three main elements: task, instruction, and role to ensure the generated text meets the user's needs.

[0149] In some embodiments, the accuracy of the cause-of-death chain and underlying cause of death for the current death event can be manually determined (e.g., by a team of experts). If inaccurate, the cause-of-death chain and underlying cause of death are corrected and added to the cause-of-death chain compilation case library. This provides more cases for the cause-of-death chain compilation case library, providing more samples for the Prompt large language model's reasoning and decision-making.

[0150] The death chain auxiliary inference method provided in the embodiment of the present invention integrates information from different sources and uses a combination of large models and probability models to infer the death chain and root cause of death of a death event, thereby improving the accuracy and efficiency of death cause inference.

[0151] The following further illustrates the process of automatically generating a cause of death chain based on a certain death investigation record text using the cause of death chain auxiliary inference method of the present invention.

[0152] The text of the hypothetical death investigation record is as follows:

[0153] The patient was previously diagnosed with hypertension (the specific diagnosis time and diagnosis unit are unknown), and was diagnosed with coronary heart disease more than 20 years ago (the specific diagnosis time and diagnosis unit are unknown); on January 1, 2025, he was sent to the emergency department of *** Hospital by 120 due to "impaired consciousness and speech difficulty for 8 hours". After clinical + physical and chemical examinations, he was diagnosed with: lung infection, large-area cerebral infarction, and heart failure. After active treatment, his condition did not improve. On January 8, 2025, his condition further worsened, and he developed ventricular fibrillation and sudden cardiac death. His family refused all rescue measures, and he was declared clinically dead at 17:00 on January 8, 2025.

[0154] Step 1: Extract medical terms from the death investigation record text and perform semantic retrieval based on the ICD code knowledge base to obtain ICD code retrieval results.

[0155] First, medical terms were extracted from the death investigation record text. The extracted medical terms are as follows:

[0156] Hypertension, coronary heart disease, impaired consciousness with speech impairment, lung infection, large-area cerebral infarction, heart failure, ventricular fibrillation, and sudden cardiac death.

[0157] Secondly, perform semantic search. The main search results are as follows:

[0158] I10 Hypertension

[0159] I11 Hypertensive heart disease

[0160] I12 Hypertensive kidney disease

[0161] I13 Hypertensive heart and kidney disease

[0162] I15 Secondary hypertension

[0163] I25.1 Coronary atherosclerotic heart disease

[0164] R40.2 Unspecified coma

[0165] J98.4 Pulmonary infection

[0166] I63.9 Large-area cerebral infarction

[0167] I50.9 Heart failure

[0168] I49.0 Ventricular fibrillation

[0169] I46.1 Sudden cardiac death

[0170] Step 2: Generate a timeline-based ICD code list associated with the current death event based on the search results, that is, output the ICD code list associated with the current death event according to the timeline.

[0171] For example, using the prompt word engineering-driven large model combined with the death investigation record text and ICD code search results, the ICD code list output by timeline is as follows:

[0172] I10 Hypertension (Timeline starting point: history of more than 20 years ago)

[0173] I25.1 Coronary heart disease (Timeline starting point: medical history more than 20 years ago)

[0174] J98.4 Pulmonary infection (2025-01-01 emergency department visit date)

[0175] I63.9 Large-area cerebral infarction (acute onset time: 2025-01-01)

[0176] I50.9 Heart failure (acute exacerbation time on January 1, 2025)

[0177] I49.0 Ventricular fibrillation (terminal event on January 8, 2025)

[0178] I46.1 Sudden cardiac death (clinical death date: 2025-01-08)

[0179] Step 3: Based on the generated ICD code list, the death chain compilation rule knowledge base is searched to extract the death chain compilation rules related to the current death event.

[0180] For example, the search results are as follows:

[0181] 1) Sudden cardiac death (I46.1) is usually the direct cause of death, and the primary cause of sudden cardiac death needs to be found.

[0182] 2) Coronary atherosclerotic heart disease (I25.1) and hypertension (I10) may both be primary causes. The disease with a greater impact on death should be selected as the main cause of death.

[0183] Step 4: Search the death chain compilation case knowledge base based on the generated ICD code list to extract the death chain compilation cases related to the current death event.

[0184] For example, the search results are as follows:

[0185] Case 1: The patient was previously diagnosed with hypertension, diabetes, and Alzheimer's disease (the specific diagnosis time and diagnosis unit are unknown); more than 10 years ago, he was diagnosed with coronary heart disease (the specific diagnosis time and diagnosis unit are unknown); on December 31, 2024, he was sent to the emergency department of XXX Hospital by 120 due to "fever and shortness of breath for 3 days". Clinical + physical and chemical examinations diagnosed him with: heart failure, acute coronary syndrome, lung infection, and respiratory failure. After active treatment, his condition did not improve. On January 1, 2025, his condition further worsened. After active rescue treatment, he was declared clinically dead at 14:13 on January 1, 2025.

[0186] The chain of death in this case is compiled as follows:

[0187] I: (a) I46.1 Sudden cardiac death

[0188] (b) I50.9 Heart failure

[0189] (c) I24.9 Acute coronary syndrome

[0190] (d) I25.1 Coronary atherosclerotic heart disease

[0191] II: (1) J98.4 Pulmonary infection

[0192] (2) J96.9 Respiratory failure

[0193] (3) I10 Hypertension

[0194] Underlying cause of death: I25.1 Coronary atherosclerotic heart disease.

[0195] Step 5: Use the N-Gram probability model to calculate the probability of the order of the diseases involved in the current death event in the cause of death chain, that is, calculate the several cause of death chains with the highest probability under each code combination in the ICD code list obtained in step 2.

[0196] For example, the two death chains with the highest calculated probability are:

[0197] Cause of death chain 1:

[0198] (a) I46.1 Sudden cardiac death

[0199] (b) I49.0 Ventricular fibrillation

[0200] (c) I50.9 Heart failure

[0201] (d) I25.1 Coronary atherosclerotic heart disease

[0202] Death Chain 2:

[0203] (a) I46.1 Sudden cardiac death

[0204] (b) I50.9 Heart failure

[0205] (c) I63.9 Large-area cerebral infarction

[0206] (d) I10 Hypertension

[0207] Step 6: The Prompt language model combines the cause-of-death chain compilation rules and compilation cases related to the current death event with the cause-of-death chain and its probability determined based on the N-Gram model to determine the cause-of-death chain and root cause of death of the current death event.

[0208] Specifically, the role and reasoning rules of the Prompt prompt large language model can be set; according to the set reasoning rules, the cause of death chain compilation rules, cause of death chain compilation cases, and the cause of death chain and its probability are input into the Prompt prompt large language model, and the cause of death chain and the root cause of death of the current death event are obtained based on the reasoning of the Prompt prompt large language model.

[0209] For example, the role settings are as follows:

[0210] As a World Health Organization-certified ICD-10 coding expert, you are responsible for constructing a logical cause-of-death chain based on clinical data that complies with international statistical standards for causes of death. You must strictly adhere to the WHO "Rules for the Medical Certification and Coding of Causes of Death" while also taking into account local statistical requirements.

[0211] The inference rules may include, but are not limited to: input information and input order, processing flow, output requirements, and constraints, etc. For example, the following inference rules are set:

[0212] #Input data

[0213] Initial coding sequence: <Please insert temporal / causal ordering of ICD-10 preliminary coding list here>

[0214] Rule Base:

[0215] (1) Core rules:

[0216] The selection of the root cause of death should satisfy the causal chain of "start → promotion → end"

[0217] Treatment complications (Y40-Y84) need to be excluded as the underlying cause

[0218] Trauma should follow the order of "external cause → injury → complication"

[0219] (2) Local Supplementary Rules: <Please insert specific supplementary terms here>

[0220] A chain of more likely causes of death for reference: <insert a chain of more likely causes of death here>

[0221] Reference Cases:

[0222] <Please insert typical success case description and code chain here>

[0223] #Processing Flow

[0224] 1. Logic verification stage:

[0225] Check whether the initial sequence complies with the principle of "causal inversion does not exceed 1 order"

[0226] Verify whether there are any intermediate links that should be excluded (such as treatment complications)

[0227] 2. Case reference stage:

[0228] Refer to previous typical cases as supplementary knowledge.

[0229] 3. Rule-based decision-making:

[0230] When statistical recommendations conflict with coding rules, the rules take precedence.

[0231] When there are multiple legal options, choose past cases and conditional probabilities.

[0232] #Output requirements

[0233] Final chain of cause of death:

[0234] I:(a) Output line a here

[0235] (b) Output line b here

[0236] (c) Output line c here

[0237] (d) Output line d here

[0238] II:(1) (2) (3)

[0241] Root cause of death: Output the root cause of death here

[0242] Decision Log:

[0243] #Constraints

[0244] The clinical entity diagnosis cannot be modified, only the coding order can be adjusted

[0245] According to the above-mentioned reasoning rules, the final chain of causes of death and the root cause of death are as follows:

[0246] I: (a) I46.1 Sudden cardiac death

[0247] (b) I49.0 Ventricular fibrillation

[0248] (c) I50.9 Heart failure

[0249] (d) I25.1 Coronary atherosclerotic heart disease

[0250] II: (1) I10 Hypertension

[0251] (2) J98.4 Pulmonary infection

[0252] (3) I63.9 Large-area cerebral infarction

[0253] Underlying cause of death: I25.1 Coronary atherosclerotic heart disease.

[0254] Accordingly, the embodiment of the present invention also provides a death cause chain auxiliary inference device, such as Figure 4 Shown is a structural schematic diagram of the device.

[0255] The death cause chain auxiliary inference device 400 includes the following modules:

[0256] Receiving module 401, used to obtain the death investigation record text;

[0257] An ICD code list generating module 402 is configured to generate a timeline-based ICD code list associated with the current death event based on the death investigation record text;

[0258] An information determination module 403 is configured to determine, based on the ICD code list, a cause-of-death chain compilation rule, a cause-of-death chain compilation case, and each cause-of-death chain and its probability related to the current death event;

[0259] The comprehensive inference module 404 is used to use the Prompt language model 50 to combine the cause of death chain compilation rules, the cause of death chain compilation cases, and the cause of death chain and its probability to determine the cause of death chain and the root cause of death of the current death event.

[0260] It should be noted that, in specific implementation, the Prompt large language model 50 can be preset to the death cause chain auxiliary inference device 400, or a corresponding interface can be set in the death cause chain auxiliary inference device 400, and the comprehensive inference module 404 uses the interface to call the external Prompt large language model 50 to realize the inference of the death cause chain and the root cause of death of the current death event.

[0261] The specific structure of an embodiment of the ICD code list generation module 402 may include the following units:

[0262] a term extraction unit, configured to extract medical terms from the death investigation record text;

[0263] A retrieval unit, configured to perform a semantic search on an ICD code knowledge base based on the medical terminology to obtain an ICD code retrieval result;

[0264] A list generating unit is used to generate an ICD code list associated with the current death event based on a timeline according to the ICD code retrieval result.

[0265] In addition to the aforementioned units, another embodiment of the ICD code list generation module 402 may further include a screening unit configured to utilize a prompt word engineering-driven macromodel, combined with the death investigation record text and the ICD code search results, to determine the optimal ICD code. Accordingly, in this embodiment, the list generation unit may generate a timeline-based list of ICD codes associated with the current death event based on the optimal ICD code.

[0266] By determining the optimal ICD code through the screening unit, the final ICD code list associated with the current death event can be made more accurate.

[0267] A specific structure of the information determination module 403 may include the following units:

[0268] a rule determination unit, configured to search a cause of death chain compilation rule base according to the category of the ICD code in the ICD code list, and determine a cause of death chain compilation rule related to the current death event;

[0269] a case determination unit, configured to search a cause of death chain compilation case database according to the ICD code in the ICD code list, and determine a cause of death chain compilation case related to the current death event;

[0270] The probability determination unit is used to input the ICD code list into the N-Gram probability model to determine each cause of death chain related to the current death event and its probability.

[0271] It should be noted that, in specific implementation, the N-Gram probability model can be pre-set in the cause of death chain auxiliary inference device 400, or the probability determination unit can call an external N-Gram probability model through a corresponding interface to determine the cause of death chains and their probabilities related to the current death event. This embodiment of the present invention does not limit this.

[0272] A specific structure of the comprehensive inference module 404 may include: a setting unit and an inference unit.

[0273] The setting unit is used to set the role and inference rules of the Prompt large language model; the inference rules may include: input information and input order, processing flow, output requirements, and constraints, etc.

[0274] The reasoning unit is used to input the cause of death chain compilation rules, the cause of death chain compilation cases, and the cause of death chain and its probability into the Prompt prompt large language model according to the set reasoning rules, and obtain the cause of death chain and root cause of death of the current death event based on the reasoning of the Prompt prompt large language model.

[0275] For more explanations on the modules and units in the death cause chain auxiliary inference device 400, please refer to the description in the previous embodiment of the method of the present invention, which will not be repeated here.

[0276] The death chain assisted inference method and device provided by the embodiment of the present invention fully utilizes the deep understanding of language by the pre-trained large language model. Based on the information recorded in the medical record or the death investigation record, it parses the patient's underlying disease, the secondary symptoms caused by it, and the order in which the symptoms appear, providing a clear framework for subsequent analysis. By performing enhanced retrieval and generation on the World Health Organization's professional knowledge base on the rules and guidelines for disease and death coding, the death chain compilation rules related to the current death event are extracted, and the N-Gram probability model is called to calculate the probability of the order of the diseases involved in the current death event in the death chain, making full use of the prior probabilities that can be provided by previous data. By comprehensively utilizing technologies such as large language models, thought chains, and retrieval enhancement generation, the present invention intelligently extracts relevant disease ICD codes from a series of descriptions of diseases, pathological conditions or injuries that cause or promote death, as well as the accidents or violence that cause such injuries, and infers the logical relationship and causal order between the codes to provide possible death chains and root causes of death.

[0277] The solution of the present invention can be used to assist professional and technical personnel of medical and health institutions and disease prevention and control departments in filling in and inferring the cause of death chain, thereby improving the quality of cause of death monitoring data.

[0278] It should be noted that for the aforementioned method embodiments, for simplicity of description, they are all expressed as a series of action combinations. However, those skilled in the art should be aware that the present invention is not limited by the order of the actions described, because according to the present invention, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in this specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

[0279] The present invention also provides a storage medium, which is a computer-readable storage medium having a computer program stored thereon, and the computer program can be executed when it is run. Figure 1 The storage medium may include a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc. The storage medium may also include a non-volatile memory or a non-transitory memory, etc.

[0280] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer program are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data provider to another website, computer, server or data provider via wired or wireless means.

[0281] The embodiments of the present invention are described in detail above. Specific implementation methods are used herein to illustrate the present invention. The description of the above embodiments is only used to help understand the method and apparatus of the present invention. They are only embodiments of a part of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work should fall within the scope of protection of the present invention, and the content of this specification should not be understood as limiting the present invention. Therefore, any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for assisting inferring a chain of death causes, characterized in that: The method comprises: Obtain text of death investigation records; Generating a timeline-based ICD code list associated with the current death event according to the death investigation record text; Determining the cause-of-death chain compilation rules, cause-of-death chain compilation cases, and each cause-of-death chain and its probability related to the current death event based on the ICD code list; The Prompt language model is used to combine the cause of death chain compilation rules, the cause of death chain compilation cases, the cause of death chain and its probability to determine the cause of death chain and the root cause of death of the current death event.

2. The method for assisting inferring the cause of death chain according to claim 1, characterized in that: The step of generating a timeline-based ICD code list associated with the current death event according to the death investigation record text includes: extracting medical terms from the death investigation record text; Performing semantic retrieval on the ICD code knowledge base according to the medical terminology to obtain ICD code retrieval results; A timeline-based ICD code list associated with the current death event is generated according to the ICD code search result.

3. The method for assisting inferring the cause of death chain according to claim 2, characterized in that: The step of generating a timeline-based ICD code list associated with the current death event according to the death investigation record text further includes: Determine the optimal ICD code by using a large model driven by a prompt word engineering and combining the death investigation record text and the ICD code search results; The step of generating a timeline-based ICD code list associated with the current death event according to the ICD code search result includes: A timeline-based ICD code list associated with the current death event is generated according to the optimal ICD code.

4. The method for assisting inferring the cause of death chain according to claim 1, characterized in that: The method of determining the cause of death chain compilation rules, cause of death chain compilation cases, and each cause of death chain and its probability related to the current death event according to the ICD code list includes: According to the category of the ICD code in the ICD code list, a cause of death chain compilation rule library is searched to determine a cause of death chain compilation rule related to the current death event; According to the ICD code in the ICD code list, searching the death chain compilation case library to determine the death chain compilation case related to the current death event; The ICD code list is input into the N-Gram probability model to determine the death cause chains and their probabilities related to the current death event.

5. The method for assisting inferring the cause of death chain according to any one of claims 1 to 4, characterized in that: The use of the Prompt large language model in combination with the cause of death chain compilation rules, the cause of death chain compilation case, the cause of death chain and its probability to determine the cause of death chain and the root cause of death of the current death event includes: Setting the role and inference rules of the Prompt prompt large language model; According to the set reasoning rules, the cause of death chain compilation rules, the cause of death chain compilation cases, and the cause of death chain and its probability are input into the Prompt prompt large language model, and the cause of death chain and the root cause of death of the current death event are obtained based on the reasoning of the Prompt prompt large language model.

6. The method for assisting inferring the cause of death chain according to claim 5, characterized in that: The inference rules include: input information and input sequence, processing flow, output requirements, and constraints.

7. The method for assisting inferring the cause of death chain according to claim 5, characterized in that: The method further comprises: Determine the accuracy of the chain of causes and underlying causes of death in the current death; If inaccurate, the stated chain of causes of death and the underlying cause of death will be corrected; The corrected cause of death chain and underlying cause of death are added to the cause of death chain compilation case library.

8. A device for assisting inferring the cause of death chain, characterized in that: The device comprises: A receiving module is used to obtain the text of the death investigation record; An ICD code list generating module is used to generate an ICD code list associated with the current death event based on the timeline according to the death investigation record text; An information determination module is used to determine the cause of death chain compilation rules, cause of death chain compilation cases, and each cause of death chain and its probability related to the current death event based on the ICD code list; The comprehensive inference module is used to use the Prompt language model to combine the cause of death chain compilation rules, the cause of death chain compilation cases, and the cause of death chain and its probability to determine the cause of death chain and the root cause of death of the current death event.

9. The device for assisting inferring the cause of death chain according to claim 8, characterized in that: The ICD code list generation module includes: a term extraction unit, configured to extract medical terms from the death investigation record text; A retrieval unit, configured to perform a semantic search on an ICD code knowledge base based on the medical terminology to obtain an ICD code retrieval result; A list generating unit is used to generate an ICD code list associated with the current death event based on a timeline according to the ICD code retrieval result.

10. The death cause chain auxiliary inference device according to claim 8 or 9, characterized in that: The comprehensive inference module includes: A setting unit, configured to set the role and inference rules of the Prompt large language model; The reasoning unit is used to input the cause of death chain compilation rules, the cause of death chain compilation cases, the cause of death chain and its probability into the Prompt prompt large language model according to the set reasoning rules, and obtain the cause of death chain and root cause of death of the current death event based on the reasoning of the Prompt prompt large language model.

Images