System for evaluating the reliability of information generated based on hearing

Abstract

Evaluating the reliability of the content heard from a doctor regarding the prescribing situation of pharmaceuticals. 【Solution means】The method according to the present invention includes a first estimation step of estimating the number of patients under treatment, which is the number of patients for whom the doctor in charge has prescribed a certain pharmaceutical, based on the result of an interview regarding the number of patients for whom a certain pharmaceutical has been prescribed among the patients in charge of a single doctor; a second estimation step of estimating the number of patients under treatment based on information other than the interview result; a similarity calculation step of calculating the similarity of the results of the first estimation step and the second estimation step; and an output step of outputting information based on the calculated similarity.

Description

Technical Field

[0001] The present invention relates to a technique for evaluating the reliability of information generated based on hearings (especially hearings conducted for doctors and the like regarding the prescription status of pharmaceuticals).

Background Art

[0002] For pharmaceutical companies, it is very important to know which doctors are prescribing their pharmaceuticals and to what extent. However, generally, pharmaceutical companies can only grasp the sales (number of sales) of their pharmaceutical products on a facility (such as a hospital) basis. Therefore, in order to find out which doctors among the multiple doctors belonging to that facility have prescribed that pharmaceutical product to what extent (for how many patients), pharmaceutical information officers (also called MRs, etc.) visit doctors regularly to exchange information and conduct interviews (hearings), and in the course of these exchanges, obtain information about patients who have newly been prescribed that pharmaceutical product (also called "new patient numbers", etc.). And based on the information obtained, the sales volume for each doctor is estimated. Also, it is conceivable to perform this estimation based on receipt data (Patent Document 1), but in the first place, pharmaceutical companies do not always have access to accurate receipt data.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, doctors are very busy and it is difficult to conduct hearings frequently. Therefore, hearings are conducted once every few months, and reports may be made for several months at a time. Also, it is not uncommon for a doctor to be in charge of a large number of patients and handle a large number of pharmaceuticals, so it is rare for them to accurately remember the number of new patients for individual pharmaceuticals. In addition, there are facilities where patients are examined under a team system in a department, and in some cases, the number of new patients including the same patient may be obtained from multiple doctors. Also, depending on the relationship with other pharmaceutical companies or the trust relationship with pharmaceutical information staff, there are doctors who do not convey accurate new patient numbers. As a result, there are always doubts, to some extent, about the reliability (accuracy) of the content of information (hearing data) generated based on the results of hearings from doctors.

[0005] Furthermore, apart from the reliability of the content of the hearing itself, since pharmaceutical information staff manually input the content of the hearing from doctors into the system, there may be cases where "intentionally" incorrect information is input (intentional mistakes), or "carelessly" (simple input mistakes) occur. That is, intentional misinput is the case where pharmaceutical information staff input a larger or smaller number for reasons such as wanting to present their activities in a better or worse light. The case of "carelessly" is when the definition to be input is different, such as when, instead of inputting the number of new patients, all the patients for whom the doctor is currently prescribing the product are input. Or, there may be cases where an incorrect value is entered during manual input, such as one more "0".

[0006] However, there was no method to objectively evaluate the reliability of hearing data. When the reliability of hearing data is suspected, it is conceivable for pharmaceutical information staff to manually correct the information based on other information, experience, or intuition, but such corrections are time-consuming and it is difficult to determine objectively whether the corrected content is appropriate.

[0007] An object of the present invention is to evaluate the reliability of information obtained from hearings with relevant parties regarding the prescription of pharmaceuticals.

Means for Solving the Problem

[0008] In one aspect, the present invention provides a program for causing a computer to execute a first estimation step of estimating the number of patients currently prescribed with a drug based on hearing information generated based on the content of an interview about the number of patients prescribed with the drug, a second estimation step of estimating the number of patients currently prescribed with the drug based on non-hearing information other than the hearing information, a similarity calculation step of calculating the similarity between the results of the first estimation step and the second estimation step, and an output step of outputting information based on the calculated similarity.

Advantages of the Invention

[0009] According to the present invention, the reliability of information obtained by interviewing persons related to the prescription of drugs can be evaluated.

Brief Description of the Drawings

[0010]

Figure 1

Figure 2

Figure 3

Embodiments for Carrying Out the Invention

[0011] In the following example, for a specific indication (a disease or symptom for which a drug is recognized to be effective and its use is approved), when estimating how many patients are currently being prescribed (administered) by a certain doctor (the number of patients currently being prescribed), the reliability of the results obtained from a hearing survey (interview) conducted with a person in charge of medical information, etc., about the prescription status of the drug for that indication (including the number of patients being prescribed (prescribed in the past, to be prescribed, etc.)) for that one doctor is verified.

[0012] In addition, when there are multiple indications for the pharmaceutical product, the purpose of the above estimation may be the number of patients currently prescribed the pharmaceutical product for multiple indication groups rather than the number of patients currently prescribed for each indication, or it may be to estimate the number of patients currently prescribed the pharmaceutical product by the doctor regardless of the indication and other prescription situations. Also, when the doctor who conducts the interview is in a position to know the prescription situations by other multiple doctors, information on the total number of patients prescribed the pharmaceutical product by multiple doctors may also be the subject of the interview. In other words, the doctor or other related person who conducts the interview and the doctor or other related person who prescribed the pharmaceutical product subject to the interview do not have to be the same. In short, the information to be collected only needs to be related to the number of patients for whom the pharmaceutical product is the subject of prescription. Hereinafter, the case of conducting an interview with and obtaining information from a single doctor regarding the number of patients prescribed a single pharmaceutical product for a single indication among the patients under the doctor's care will be described. Note that the content of the pharmaceutical product, symptoms, purpose of prescription (clinical trial, treatment, etc.) and form are not limited.

[0013] <Configuration example> The information processing apparatus 10 is implemented as a general-purpose or dedicated personal computer, smartphone, other mobile terminal, or server, which includes a processor (such as a CPU or GPU), a storage unit (memory), and an input / output device (such as a keyboard, mouse, display, and communication interface for exchanging information with an external device).

[0014] The information processing apparatus 10 includes a control unit 200, a storage unit 100, an input unit 300, an output unit 400, and a communication unit 500. The control unit 200 is realized by a processor 20. The storage unit 100 is a memory for storing various types of information read by the control unit 200. The storage unit 100 stores a program for realizing the functions of the information processing apparatus 10 described later. In addition, it stores information (hearing information) generated by a person in charge of pharmaceutical information or the like based on an interview with a doctor or the like. "Generated based on an interview" means any information that is generated upon the occurrence of an interview and can be associated with the content of the interview, regardless of whether the content of the interview is reflected in the information. Also, the storage unit 100 stores various types of data and algorithms used in the control unit 200. Specifically, it stores hearing information and information regarding the time-series change of the continuous prescription rate (such as a survival curve) used in the estimation process in the first estimation unit 201, information other than the hearing result and a learned model used in the estimation process in the second estimation unit 202, and a similarity determination algorithm used in the determination process in the comparison unit 203.

[0015] The input unit 300 is an input device such as a keyboard or a mouse, and is used to input hearing information. The output unit 400 is an information output device having, for example, a display or an image processing processor, and generates information based on the processing result of the comparison unit 203 and outputs it in the form of characters, images, sounds, or other formats. The communication unit 500 is an interface for exchanging information with an external device by wire or wirelessly, and acquires information used in the calculation performed by the control unit 200 from other servers or databases via the Internet or the like as necessary.

[0016] The control unit 200 includes a first estimation unit 201, a second estimation unit 202, and a comparison unit 203.

[0017] The first estimation unit 201 estimates the currently prescribed patients based on the hearing information. This will be specifically described with reference to FIG. 2.

[0018] In this example, a pharmaceutical information staff member conducts an interview with the attending physician regularly (in this example, once a month), and inquires about the number of patients (new patients) for whom the target drug has been newly prescribed for a specific indication in the past month (from the previous interview to the current interview). Here, the interview results for the most recent N months are 5 patients, 1 patient in the previous month, and 4 patients in the month before that... Such results have been obtained. Due to the nature of the interview results, they are discrete values.

[0019] In addition, the first estimation unit 201 estimates the number of patients currently (more precisely, at the time of N months in the current month) under prescription based on the continuous prescription rate of the drug for that indication. Generally, during survival, the prescription is continued until the disease condition is completely cured, so the continuous prescription rate often essentially coincides with what is known as the survival curve. The survival curve is a graph showing the survival rate over time in a certain population, with the vertical axis representing the proportion of survivors (or the number of survivors) and the horizontal axis representing time (days, months, years, etc.), indicating how the population decreases over time. Note that even during survival, there may be cases where the prescription is stopped due to side effects of the drug or where the prescription is stopped because the upper limit value of the prescribed dosage, etc., has been reached, so the survival curve and the continuous prescription rate do not always refer to the same thing.

[0020] This information on the continuous prescription rate may be derived by known statistical processing such as the Kaplan - Meier method, or may be calculated by a predetermined algorithm based on data from clinical trials. Whether it is published or not (whether it is only held by the pharmaceutical company) does not matter. Also, it may be calculated based on the number of new patients obtained from other physicians for the same drug. Further, when there is no data for the same indication and drug as the target, data for similar indications or drugs may be adopted. The continuous prescription rate, by its nature, is a continuous value (probability distribution) as shown in the figure.

[0021] The first estimation unit 201 calculates, based on the number of continuous prescription patients, the probability distribution of the number of patients under prescription, which is a continuous value, from the number of new patients, which is a discrete value, obtained during a predetermined period (for example, the most recent one year). In this example, it is calculated as a normal distribution with an expected value of 9.5 people and a standard deviation of 2.3. That is, the number of patients under prescription is calculated to be approximately 7 to 12 people.

[0022] The second estimation unit 202 estimates the current number of patients under prescription based on information other than hearings (non-hearing information) collected in advance.

[0023] The non-hearing information includes at least one of the following information, for example. · Information regarding the amount or content of hearing activities (the number of visits to the doctor, the time spent on each visit, the amount of information obtained in the hearing other than the number of new patients (information regarding the content and amount of information obtained other than the number of new patients), the status of the activities of the pharmaceutical information staff who conducted the hearing (the amount and content of hearings to other doctors regarding the pharmaceutical product and / or hearings conducted regarding other pharmaceutical products)) · Attributes of the doctor The attributes include, for example, the department to which the doctor belongs, specialty (whether a specialist or not), position, rank, age, etc. This is because the prescription policy may vary depending on the attributes of the doctor. · Information regarding the medical institution to which the doctor belongs For example, it includes information such as the size of the medical institution (number of beds, number of employees), the number of patients the doctor is in charge of, and location information. This is because there may be differences in the tendency for the pharmaceutical product to be prescribed depending on the scale and region of the medical institution. · Others For example, information obtained from receipt (medical fee statement) information may include information that may be related to the prescription of pharmaceutical products.

[0024] Note that the above information may be information regarding the prescription of the target drug itself, or may be information regarding the prescriptions of drugs other than the target drug. In short, any information that is considered to be related to the prescription of the target drug can be utilized.

[0025] Based on the non-hearing information, the second estimation unit 202 estimates the probability distribution of the number of patients under prescription. Specifically, after excluding outliers from the data set obtained as the non-hearing information, grouping is performed according to a predetermined criterion. Then, for each group, a learning model that outputs the number of patients under prescription as a probability distribution from the input information is constructed in advance. As the prediction model, a probability distribution fitting model, a Gaussian process using the Cauchy distribution, or the like can be adopted. By inputting the non-hearing information into the learning model thus constructed, the probability distribution of the number of patients under prescription is obtained.

[0026] The comparison unit 203 compares the first estimation result calculated by the first estimation unit 201 with the second estimation result calculated by the second estimation unit 202, and calculates the similarity (distance between the two probability distributions). When at least one of the first estimation result and the second estimation result is a Gaussian distribution, the distance between the distributions can be directly calculated using a known method such as KL divergence or JS divergence. Note that when the distance between the distributions cannot be calculated analytically or is practically too difficult, sampling is performed from each probability distribution to construct a point cloud, and the point clouds are compared. The greater the difference between the two distributions (that is, the lower the similarity between the distributions, the greater the distance between the distributions), the more the two estimation results can be considered to be inconsistent.

[0027] The comparison unit 203 calculates and outputs information based on the similarity. The information based on the similarity (distance between two probability distributions) can be expressed, for example, as one index (distance; score). Alternatively, it may be expressed in a category to which the similarity or consistency or the reliability of the hearing result belongs (in the form of "high similarity / no problem with consistency / no problem with the reliability of the hearing result", "low similarity / slightly doubtful about consistency / slightly doubtful about the reliability of the hearing result", "low similarity / no consistency / considerable doubt about the reliability of the hearing result"). Alternatively, when the similarity is below a predetermined value, a message such as "There is doubt about the reliability of at least one of the data." may be generated and notified to the user. That is, the method of secondarily processing the result of comparing one estimation result is arbitrary.

[0028] In addition, after calculating the comparison result, the comparison unit 203 may further execute a process of verifying whether either (or both) of the information based on the number of new patients and the prediction model is of low reliability. At least, it is sufficient if it is information that can assist the user's judgment that the hearing information may be of low reliability and thus cannot be utilized, or that caution is required when utilizing it.

[0029] <Operation example> FIG. 3 shows an operation example of the information processing apparatus 10. In S101, the first estimation unit 201 estimates the current number of patients under treatment based on the hearing result (the first estimation result). In S102, the second estimation unit 202 estimates the current number of patients under treatment based on the non-hearing information (the second estimation result). In S103, the comparison unit 203 determines the similarity (degree of deviation) between the first estimation result and the second estimation result, and generates and outputs information based on the similarity.

[0030] According to the above embodiment, regarding the prescription of pharmaceuticals, by comparing the content of the hearings conducted by pharmaceutical information staff and others with doctors, medical staff, and other relevant parties, with various other information outside of these hearings, the reliability of the hearing content can be objectively evaluated. By its nature, the hearing content becomes discrete information (for example, the number of patients per month), while information outside of the hearings (for example, survival curves) usually often becomes continuous information. Thus, according to the above embodiment, even for information with different natures or data structures, both can be compared. Furthermore, by collectively handling various information outside of the hearings, the efficiency of reliability determination and the improvement of determination accuracy are realized.

[0031] The functions of the information processing apparatus 10 may be distributed and executed by a plurality of information processing apparatuses. For example, a server implementing the functions of the first estimation unit 201 and a server implementing the functions of the second estimation unit 202 may be provided respectively, and the information processing apparatus 10 may substantially only have the functions of the comparison unit 203, and may request the above-described first estimation and second estimation processes from these servers, and acquire the estimation results from these servers. That is, in an information processing system composed of one or more information processing apparatuses, based on the result of a hearing regarding the number of patients who have been prescribed a pharmaceutical, a first estimation step of estimating the number of patients currently being prescribed the pharmaceutical, which is the number of patients being prescribed, a second estimation step of estimating the number of patients being prescribed based on information other than the hearing result, a similarity calculation step of calculating the similarity between the results of the first estimation step and the second estimation step, and an output step of outputting information based on the calculated similarity may be executed.

Explanation of Reference Numerals

[0032] 10: Information processing apparatus 20: Processor 100: Storage unit 200: Control unit 201: First estimation unit 202: Second estimation unit 203: Comparison unit 300: Input unit 400: Output unit 500: Communication Department

Claims

1. A first estimating step of causing a computer to estimate the number of patients currently prescribed with a drug, based on hearing information generated based on the content obtained by asking about the number of patients prescribed with the drug; A second estimating step of estimating the number of patients currently prescribed with the drug based on non-hearing information other than the hearing information; A similarity calculating step of calculating the similarity of the results of the first estimating step and the second estimating step; An output step of outputting information based on the calculated similarity A program for causing the above to be executed.

2. The results of the first estimating step and the second estimating step are expressed as probability distributions The program according to claim 1.

3. In the first estimating step, The hearing information is generated based on the results of the hearing performed multiple times, A curve showing the time-series change of the continuous prescription rate for the drug is used The program according to claim 2.

4. In the second estimating step, The non-hearing information includes information on one or more selected from information on the activity status of the person who performed the hearing, the attributes of the doctor who prescribed the drug, information on the medical institution to which the doctor belongs, and the region where the doctor is active, The number of patients currently prescribed with the drug is calculated using a learning model that outputs the number of patients currently prescribed with the drug using the non-hearing information as input information The program according to claim 1.

5. A first estimating means for estimating the number of patients currently prescribed with a drug, based on hearing information generated based on the content obtained by asking about the number of patients prescribed with the drug; A second estimating means for estimating the number of patients currently prescribed with the drug based on non-hearing information other than the results of the hearing; A similarity calculating means for calculating the similarity between the estimation result of the first estimating means and the estimation result of the second estimating means An output means for outputting information based on the calculated similarity An information processing system having the above. ​

Images