Real-Word Studies (RWS) are studies based on real situations that participants encounter in daily clinical practice. To obtain accurate research results, high-quality structured data is often required. In the face of massive text data, how do we get high-quality structured data? Today I will share with you a simple and efficient data processing method.
01
—
Data source analysis of RWS
Existing data sources mainly include:
1. Electronic medical record (EMR)
2. Electronic health record (EHR)
3. Medical insurance data (claims data)
4. Birth and death registration
5. Public health monitoring data and regionalization
6. Medical data, etc.
Researchers need to extract key information from participants' medical records in order to better understand patients' conditions and treatment outcomes.
1. Basic information: such as age, gender, medical history, family medical history, etc.
2. Diagnosis information: disease diagnosis, condition description, patient complaints, etc.
3. Treatment information: treatment plan, drug treatment, surgical treatment, etc.
4. Observation data: such as blood pressure, blood sugar, body temperature and other data.
5. Disease progress: the evolution process of the disease, the change of symptoms, etc.
6. Adverse reactions: adverse drug reactions, postoperative complications, etc.
7. Quality indicators: such as patient satisfaction, treatment effectiveness and other indicators.
The above information is only part of the extraction of case information, and the specific extraction methods and information content need to be determined according to the purpose and problems of the research.
02
—
Structured processing of data
The required information is often hidden in these texts and cannot be directly used for statistical research. How to extract the information we need from these texts? The traditional method is manual or engineering method using natural language processing, and the cost of these two methods is often relatively high.
Is there a simpler and feasible solution now?
Today we will take the above text as an example and use "Knowledge AI" for structured data processing.
The following is a piece of case information excerpted from "Lilac Garden":
Patient, female, 75 years old.
The chief complaint was "cough, expectoration, fever for 1 week, chest tightness and dyspnea for 5 days". One week ago, the patient suffered from runny nose, stuffy nose, itchy throat, cough, yellow phlegm, and fever, with a maximum body temperature of 38.0°C, accompanied by chills and shivering. Five days ago, the patient developed chest tightness, located behind the middle of the sternum, which was oppressive, persistent, and gradually developed dyspnea. He took compound acetaminophen granules and cefixime. No chest and back pain, no hemoptysis, no abdominal pain, bloating, diarrhea, vomiting, headache, limb hemiplegia, etc. The patient had a medical history of "hypertension" in the past. Physical examination on admission: body temperature 38.0°C, pulse 84 beats/min, respiration 22 beats/min, blood pressure 122/70mmHg. Fingertip Pulse Oxygen 98%. Sober, answer relevant questions. Distended jugular veins. Coarse breath sounds in both lungs, more moist rales can be heard in both lungs, and dry rales are not heard. There is no expansion of the heart circle, the rhythm is regular, the heart sound is low and blunt, and no murmur is heard. There was no edema in both lower extremities. Examination of abdomen and nervous system (-). Laboratory examination: hypersensitive troponin T: >50000 pg/ml (0-34.2). CK-MB: 184.7 U/L (0-25); CK: 2671 U/L (25-173). pro-BNP 22063.1 pg/ml. D-dimer: 4567 ug/L (0-1000). Blood gas analysis: pH: 7.3, PCO2: 34mmHg, PO2: 71mmHg, HCO3: 27.2mmol/L, lactic acid: 1.1mmol/L. Blood routine: white blood cells 6.66x10^9/L, neutrophils 82.1% (50-70), lymphocytes 9.9% (20-40), hemoglobin 111g/L, platelets 110x10^9/L. Hypersensitive C-reactive protein: 100 mg/L (0-10). Procalcitonin was normal. Biochemistry: serum potassium: 3.84mmo/L. Creatinine 119 umol/L.Liver function: AST: 326 lU/L; ALT: 117 lU/L; serum albumin: 32.1 g/L. There was no obvious abnormality in blood sugar, blood lipid, and nail function.
Today we will take the above text as an example, and use Zhiwu AI for structured data processing.
The processing idea we adopt is a processing strategy from the whole to the part, from coarse to fine.
1. Large category extraction
We input and extract age, gender, diseases that need to be identified, medical history, family medical history, disease diagnosis, condition description, patient complaints, treatment plan, drug treatment, surgical treatment, adverse reactions, postoperative complications, patient satisfaction with medical treatment, treatment Effectiveness, laboratory test
It is displayed in the form of a table, and the header is content and result.
2. Refined extraction for each category
(1), extraction laboratory test
(2), extraction of illness and chief complaint
(3), extraction of history of present illness and past history
(4), extraction of medication information
You can open "Knowledge AI" to experience it yourself, and you can enter "Please help me extract medication information and display it in a table + the text you want to extract". You can get what you want.
03
—
Prospects for real-world research
A trend in medical research today is Real-world evidence (RWE). Compared with previous clinical trials, RWE better reflects the real situation of diseases and treatment options in the real world, and has wider application value. In the future, with the continuous upgrading and improvement of RWE technology and methods, it will play a more important role in evaluating treatment effects, formulating health policies and guidelines, and supporting drug approval and supervision.
Treatment efficacy in critically ill patients remains an open question in real-world studies. In emerging segments, such as precision medicine, personalized medicine and digital medicine, RWE will also play a vital role in providing patients with more precise and personalized treatment plans to maximize their benefits.
In general, the application prospects of real-world research are very broad, and will play an increasingly important role in medical treatment, drug approval and supervision, health policy and guideline formulation, etc., to promote the continuous advancement and progress of medical research, and provide Provide more accurate and effective medical services to patients.
| Product Experience |
The above screenshots are all from our AI assistant: Do you know AI Q&A? If you are interested, you can click the link to view it. You can log in directly by scanning the QR code on WeChat. No other complicated verification is required. New users can try it for free for 3 days.
Click here to experience it directly!
| Recommended reading |
1. How to make a weight loss plan efficiently
2. How to quickly generate a copy
3. How to successfully complete the graduation thesis
4. How to become an invincible programmer in the universe
5. Quick Cheats for Essays Helped by ChatGPT
6. My mother gave birth to me, and ChatGPT eliminated my father
7. Use ChatGPT to easily get children's bedtime stories
8. How to use ChatGPT as a language translation tool
9. Teach you to use ChatGPT to assist in writing papers
10. Overview of Real World Studies (RWS)
11. How to become a GPT engineer with an annual income of 100W