WO2021140402A1

WO2021140402A1 - Diagnostic prediction system and method

Info

Publication number: WO2021140402A1
Application number: PCT/IB2020/062431
Authority: WO
Inventors: Gerrit Johannes VAN ZYL
Original assignee: Healthbridge Usa Inc
Current assignee: Healthbridge Usa Inc
Priority date: 2020-01-10
Filing date: 2020-12-24
Publication date: 2021-07-15
Anticipated expiration: 2022-07-10

Abstract

A diagnostic prediction system which includes an interface module and a diagnostic prediction module/engine. The interface module is configured to receive/retrieve information on one or more medicines prescribed, provided and/or administered to a patient. The diagnostic prediction module/engine is configured to predict one or more possible diagnoses by using a prediction algorithm/model which utilises one or more therapeutic qualities which is/are associated with the one or more medicines. Each possible diagnosis may be associated with, or relate to, one or more specific diagnosis/diagnostic codes. The diagnostic prediction module/engine may, more specifically, be a diagnostic coding prediction module/engine which is configured to predict one or more possible diagnoses and the appropriate diagnosis/diagnostic code(s) to correctly code the diagnosis, by using the prediction algorithm/model. The system may therefore be a diagnostic coding prediction system.

Description

DIAGNOSTIC PREDICTION SYSTEM AND METHOD

BACKGROUND OF THE INVENTION

THIS invention relates to a diagnostic prediction system and method. More specifically, it relates to a diagnostic coding prediction system and method.

The cost of providing healthcare is rising much faster than the ability of the funding systems to pay for these costs. One potential solution to address this challenge is the adoption of clinical record keeping solutions that help reduce medical errors and improve treatment choices. This requires health workers to use electronic health records, electronic medical records (“EHRs” and “EMRs”) to capture a patient’s symptoms, treatments and diagnoses.

However, for the information to be useful for reimbursement and clinical analytical purposes, it is important that the diagnoses be captured accurately and with the highest degree of specificity possible.

Existing EMR / EHR solutions are typically rich in functionality, but not simple to use and take too much of the healthcare providers’ time to capture relevant information. This has a number of very negative consequences:

• Reduced time a healthcare provider has available to focus on a patient during a consult, which directly impacts on the quality of care - Doctors spend more time capturing the clinical information, instead of looking into their patients and trying to treat them. Physicians can spend up to twice as much time completing electronic health records (EHR) as they do treating patients, according to a new study by the American Medical Association.

• EHRs negatively impact doctors’ relationships with patients, according to a study published in the Journal of Innovation in Health Informatics: “It’s like texting at the dinner table”: A qualitative analysis of the impact of electronic health records on patient-physician interaction in hospitals - Kimberly D Pelland, Rosa R Baier, Rebekah L Gardner.

• Healthcare providers need to work extended hours just to keep up with admin as a direct consequence of EMR / EHR usage. Doctors can spend on average 1.4 hours after hours to work on EHRs. This leads to unprecedented high levels of stress and burnouts among healthcare providers. In recent years, “physician burnout” has skyrocketed to the top of the agenda in medicine. A 2018 Merritt Hawkins survey found that 78% of doctors in the USA suffered symptoms of burnout, with EHRs being the key reason. In Stanford Medicine’s 2018 National Physician Poll, 59% said EHRs needed a “complete overhaul”; 54% said that the systems had detracted from their professional satisfaction; and 49% said it detracted them from their clinical effectiveness.

T o deal with these consequences, a new role has been created in healthcare, commonly known as a “healthcare data scribe”. These scribes are typically nurses, or in many cases medical doctors typically in a different country with cheaper labour, trained to listen in on conversations between doctors and patients and extract and encode the relevant clinical information into the EHR systems. While this is a solution to the mandatory use of EHRs, it has a number of unintended negative consequences. The first is that this just further increases the cost of healthcare, and instead of solving the problem at the source, just shifts the workload to another party. Furthermore, it introduces an added degree of separation between the doctor and the EHR and, in doing so, prevents the doctor from encoding some of the more subtle, but nevertheless valuable data points that typically arise in a consultation.

Harvard School of Public Health calls it a public health crisis.

To alleviate these problems, it is essential to reduce the time a healthcare provider spends on capturing relevant information. The system described in US Patent Publication No. 2008/0288292 A1 has tried to address this problem, but not to an adequate extent.

The process of correctly coding the diagnosis for a complex patient encounter is a non-trivial task and one that can take many minutes of a physician’s time to do correctly. This is because accurate diagnostic coding is one of the foundations of healthcare informatics and inaccuracies lead to mistakes on the treatment of individual patients, as well as in population analyses. For this reason, the healthcare payors often link reimbursement of claims to the accuracy of the coding. This places an additional burden on the physicians.

While the doctor may have a clear idea of the diagnosis in physiological terms, the process of converting this diagnosis into diagnosis/diagnostic codes in one or more of the industry coding systems (such as the ICD-10- CM coding system) is often non-trivial. For example, the diagnosis of “influenza” maps to the following diagnosis codes:

J09 Influenza due to certain identified influenza viruses

J09.X Influenza due to identified novel influenza A virus

J09.X1 . with pneumonia

J09.X2 . with other respiratory manifestations

JG9.X3 . with gastrointestinal manifestations

J09.X9 . with other manifestations

J10 Influenza due to other identified influenza virus

J10.Q Influenza due to other identified influenza virus with pneumonia

J 10.00 Influenza due to other identified influenza virus with unspecified type of pneumonia

J 10.01 Influenza due to other identified influenza virus with the same other identified influenza virus pneumonia J 10.08 Influenza due to other identified influenza virus with other specified pneumonia J10.1 Influenza due to other Identified Influenza virus with other respiratory manifestations

J10.2 Influenza due to other identified influenza virus with gastrointestinal manifestations

J10.8 Influenza due to other identified influenza virus with other manifestations

J 1081 Influenza due to other identified influenza virus with encephalopathy

J 1082 Influenza due to other identified influenza virus with myocarditis

J 10.83 Influenza due to other identified influenza virus with otitis media

J 10.89 Influenza due to other identified influenza virus with other manifestations

J11 Influenza due to unidentified influenza virus

J11.0 Influenza due to unidentified influenza virus with pneumonia J11.00 Influenza due to unidentified influenza virus with unspecified type of pneumonia J11.Q8 Influenza due to unidentified influenza virus with specified pneumonia

J11.1 Influenza due to unidentified influenza virus with other respiratory manifestations

J11.2 Influenza due to unidentified influenza virus with gastrointestinal manifestations

J11.8 Influenza due to unidentified influenza virus with other manifestations

J11.81 Influenza due to unidentified influenza virus with encephalopathy

J11.82 Influenza due to unidentified influenza virus with myocarditis

J11.83 Influenza due to unidentified influenza virus with otitis media J11.89 Influenza due to unidentified influenza virus with other manifestations

Similarly “pneumonia” maps to the following ICD-10-CM codes:

J12 Viral pneumonia, not elsewhere classified

J12.0 Adenoviral pneumonia J12.1 Respiratory syncytial virus pneumonia J12.2 Parainfluenza virus pneumonia J12.3 Human metapneumovirus pneumonia J12.8 Other viral pneumonia

J 12.81 Pneumonia due to SARS-associated coronavirus J 12.89 Other viral pneumonia J12.9 Viral pneumonia, unspecified J14 Pneumonia due to Hemophilus influenzae

J15 Bacterial pneumonia, not elsewhere classified

J15.Q Pneumonia due to Klebsiella pneumoniae J15.1 Pneumonia due to Pseudomonas J15.2 Pneumonia due to staphylococcus J 15.20 . unspecified

J15.21 Pneumonia due to staphylococcus aureus

J 15.211 Pneumonia due to Methicillin susceptible Staphylococcus aureus

J15.212 Pneumonia due to Methicillin resistant Staphylococcus aureus

J 15.29 Pneumonia due to other staphylococcus J15.3 Pneumonia due to streptococcus, group B J15.4 Pneumonia due to other streptococci J15.5 Pneumonia due to Escherichia coli J15.6 Pneumonia due to other Gram-negative bacteria J15.7 Pneumonia due to Mycoplasma pneumoniae J15.8 Pneumonia due to other specified bacteria J15.9 Unspecified bacterial pneumonia J16 Pneumonia due to other infectious organisms, not elsewhere classified

J16.0 Chlamydial pneumonia

J16.8 Pneumonia due to other specified infectious organisms J18 Pneumonia, unspecified organism

J18.0 Bronchopneumonia, unspecified organism J18.1 Lobar pneumonia, unspecified organism Ji8.2 Hypostatic pneumonia, unspecified organism J18.8 Other pneumonia, unspecified organism J18.9 Pneumonia, unspecified organism

Where the accuracy of ICD10 codes is enforced, the reaction of the industry has been to use administrative specialists called medical coders for this task. These specialists are trained in correctly coding claims and optimising the billing for doctors. This also has unintended consequences. The medical coders are incentivised to get claims paid, not to code correctly for the patient and often these goals diverge.

The Inventor wishes to address at least some of the problems mentioned above.

SUMMARY OF THE INVENTION

In accordance with a first aspect of the invention there is provided a diagnostic prediction system wherein the system includes: an interface module which is configured to receive/retrieve information on one or more medicines prescribed, provided and/or administered to a patient; and a diagnostic prediction module/engine which is configured to predict one or more possible diagnoses by using a prediction algorithm/model which utilises one or more therapeutic qualities which is/are associated with the one or more medicines.

Each possible diagnosis may be associated with, or relate to, one or more specific diagnosis/diagnostic codes. The diagnostic prediction module/engine may, more specifically, be a diagnostic coding prediction module/engine which is configured to predict one or more possible diagnoses and the appropriate diagnosis/diagnostic code(s) to correctly code the diagnosis, by using the prediction algorithm/model.

The system may therefore be a diagnostic coding prediction system.

The diagnosis/diagnostic code may relate to, or be based on, a specific diagnostic coding system, such as the ICD-10 or ICD-10-CM coding system. The interface module may, more specifically, be configured to provide a list of predicted diagnosis/diagnostic codes which are most likely associated with the patient. The interface module may more specifically be configured to send the list to a computing terminal, such as a computer or a mobile communication device.

More specifically, the diagnostic prediction module/engine may be configured to predict one or more possible diagnoses which is/are the most likely and, optionally, to propose the most appropriate codes to correctly describe the diagnosis, by using the prediction algorithm/model. Information on the therapeutic qualities of each of a plurality of medicines may be stored on a database. The diagnostic prediction module/engine may be configured to utilise the information stored in the database within the prediction algorithm/model, in order to predict one or more possible diagnoses which is/are the most likely.

The interface module may be configured to receive/retrieve information on a plurality of medicines prescribed, provided and/or administered to the patient. The interface module may be configured to receive the information on the one or more medicines prescribed, provided and/or administered to a patient via a computing terminal, such as a computer or smart device (e.g. a tablet or smart phone). A user may for example enter the information on the medicines prescribed, provided and/or administered on the computing terminal. The user may be an authorized user, such as a medical practitioner or other medical assistant/support which is authorized/qualified to enter/provide the information. The information may however also be received/retrieved from another system or a database (e.g. a shared database), through natural language processing, or from an OCR module/program (e.g. an OCR of a script a doctor wrote).

The interface module may be configured to provide/present the one or more possible diagnoses to a user through/via a visual display, such as a display screen (e.g. a computer screen). The interface module may be configured to send the information on the diagnosis and the proposed coding of the diagnosis to an external system, such as an existing billing system. The interface module may be configured to send information on the one or more possible diagnoses back to the terminal from which the information on the one or more medicines prescribed, provided and/or administered was received.

The diagnostic prediction module/engine may be configured to rank the list based on probability, by using the prediction algorithm/model. The interface module may more specifically be configured to send the ranked list back to the computing terminal. The interface module may be configured to prompt a user (e.g. a doctor) with additional questions to narrow the list of potential diagnoses.

The interface module may be configured to receive a selection of one of the possible diagnoses from the user, which should be used as the correct diagnosis of the patient. More specifically, the interface module may be configured to receive a selection of specific diagnostic codes from the user, which should be used as an input to determine the correct way to code the diagnosis of the patient. The prediction module/engine may be configured to store the selection on a database.

The interface module may be configured to receive/retrieve any one or more of the following information: information on one or more consumables used during the medical consultation; information on one or more treatment actions performed on/for the patient (e.g. by a medical practitioner); and/or information on one or more diagnostic tests performed or laboratory orders placed for the patient.

The interface module may also be configured to receive/retrieve any one or more of the following information: age and/or sex of the patient; previous diagnoses of the patient existing medical conditions of the patient (such as chronic conditions); laboratory results associated with the patient; environmental conditions relevant to the condition of the patient; and/or previous treatments of the patient.

The interface module may be configured to receive the information, listed in the previous two sentences, from the user via a computing terminal (e.g. a computer, tablet or smart phone) or through other means such as the database of an existing EMR system or a medicine prescribing system.

The prediction algorithm/model may be configured to utilise any one or more of the following information which is received/retrieved by the interface module, together with the one or more therapeutic qualities which is/are associated with the one or more medicines: information on one or more consumables used during the medical consultation; information on one or more treatment actions performed on/for the patient (e.g. by a medical practitioner); information on one or more diagnostic tests performed or laboratory orders placed for the patient; age and/or sex of the patient; previous diagnoses of the patient; laboratory results associated with the patient; and/or previous treatments of the patient, environmental conditions that may have an impact on the coding of the diagnosis.

The consumables may, for example, include injections, sutures and/or bandages.

The prediction algorithm/model may be an artificial intelligence/machine learning algorithm/model. The algorithm/model may be trained by using a dataset which includes any one or more of the following: consumables historically used during various medical consultations with patients (e.g. injections, sutures and/or bandages); prescriptions historically provided to various patients (e.g. prescribed medicines); historical claims which are associated with various patients; one or more treatment actions historically applied to/on patients; medical history data of various patients (e.g. previous medical conditions/illnesses); and historical claim acceptance/rejection information from health insurers.

In accordance with a second aspect of the invention there is provided a diagnostic prediction method wherein the method includes: a. receiving/retrieving, by using a processor, information on one or more medicines prescribed, provided and/or administered to a patient; and b. predicting, by using a processor, one or more possible diagnoses by implementing/using a prediction algorithm/model which utilises one or more therapeutic qualities which is/are associated with each of the one or more medicines.

Each possible diagnosis may be associated with, or relate to, one or more specific diagnosis/diagnostic codes. Step b may therefore more specifically include predicting, by using a processor, one or more possible diagnoses and an appropriate diagnosis/diagnostic code(s) to correctly code the diagnosis, by implementing/using the prediction algorithm/model.

The method may therefore be a diagnostic coding prediction method. The prediction algorithm/model may be a code prediction algorithm/model.

The diagnosis/diagnostic code may relate to, or be based on, a specific diagnostic coding system, such as the ICD-10 or ICD-10-CM coding system.

The method may therefore include providing a list of predicted diagnostic codes which are most likely associated with the patient to a user via a visual display. The method may include sending the list to a computing terminal (e.g. a computer or smart device), which then displays the list on a display screen.

The method may include ranking, by using a processor, the list based on probability, by using the prediction algorithm/model. The method may include sending the ranked list back to the computing. Step (b) may more specifically include predicting, by using a processor, one or more possible diagnoses which is/are the most likely, by using the prediction algorithm/model. Information on the therapeutic qualities of each of a plurality of medicines may be stored on a database. Step (b) may therefore include utilising the information stored in the database within the prediction algorithm/model, in order to predict one or more possible diagnoses which is/are the most likely.

Step (a) may more specifically include receiving/retrieving information on a plurality of medicines prescribed, provided and/or administered to the patient. Even more specifically, step (a) may include receiving the information on the one or more medicines prescribed, provided and/or administered to a patient via a computing terminal, such as a computer or smart device (e.g. a tablet or smart phone). The method may therefore include entering the information on the medicines prescribed, provided and/or administered on the computing terminal. The information may however also be received/retrieved from another system or a database (e.g. a shared database), through natural language processing, or from an OCR module/program (e.g. an OCR of a script a doctor wrote).

The method may include providing/presenting the one or more possible diagnoses to a user through/via a visual display, such as a display screen (e.g. a computer screen). The method may include sending information on the one or more possible diagnoses back to the terminal from which the information on the one or more medicines prescribed, provided and/or administered was received.

The method may include receiving a selection, preferably via a computing terminal, of one of the possible diagnoses from the user, which should be used as the correct diagnosis of the patient. More specifically, the method may include receiving a selection of a specific diagnostic code from the user, which should be used as the correct diagnosis of the patient. The method may include storing the selection on a database. The method may including receiving/retrieving, by using a processor, any one or more of the following information: information on one or more consumables used during the medical consultation; information on one or more treatment actions performed on/for the patient (e.g. by a medical practitioner); and/or information on one or more diagnostic tests performed or laboratory orders placed for the patient.

The method may also include receiving/retrieving, by using a processor, any one or more of the following information: age and/or sex of the patient; previous diagnoses of the patient; laboratory results associated with the patient; and/or previous treatments of the patient.

The method may include receiving the information, listed in the previous two sentences, from the user via a computing terminal (e.g. a computer, tablet or smart phone).

The prediction algorithm/model may be configured to utilise any one or more of the following received/retrieved information, together with the one or more therapeutic qualities which is/are associated with the one or more medicines: information on one or more consumables used during the medical consultation; information on one or more treatment actions performed on/for the patient (e.g. by a medical practitioner); information on one or more diagnostic tests performed or laboratory orders placed for the patient; age and/or sex of the patient; previous diagnoses of the patient; laboratory results associated with the patient; and/or previous treatments of the patient. The consumables may, for example, include injections, sutures and/or bandages.

The prediction algorithm/model may be an artificial intelligence/machine learning algorithm/model. The method may include training the algorithm/model by using a dataset which includes any one or more of the following: consumables historically used during various medical consultations with patients (e.g. injections, sutures and/or bandages); prescriptions historically provided to various patients (e.g. prescribed medicines); historical claims which are associated with various patients; one or more treatment actions historically applied to/on patients; medical history data of various patients (e.g. previous medical conditions/illnesses); and historical claim acceptance/rejection information from health insurers.

BRIEF DESCRIPTION OF THE DRAWINGS

The invention will now by described, by way of example, with reference to the accompanying diagrammatic drawings. In the drawings:

Figure 1a shows a schematic illustration of how doctors typically operate; Figure 1b shows a schematic illustration of how the diagnostic prediction system, in accordance with the invention, turns the process shown in Figure 1 a around;

Figure 2 shows a schematic layout of a diagnostic prediction system in accordance with the invention;

Figure 3 shows a functional layout of the system shown in Figure 2; Figure 4 shows a schematic layout of a process flow of the system shown in Figure 2;

Figure 5 shows a schematic illustration of a modelling process of a diagnostic prediction engine of the system shown in Figure 2; and

Figure 6 shows a schematic illustration of a prediction process of the diagnostic prediction engine of the system shown in Figure 2.

DESCRIPTION OF PREFERRED EMBODIMENTS

The present invention relates to a diagnostic prediction system. More specifically, the invention relates to a diagnostic code prediction system.

Currently, a doctor would typically identify a patient’s health/medical issue based on a proper diagnosis. Based on the diagnosis, a list of actions are taken by a doctor. For example, the doctor may prescribe certain medication, perform certain medical procedures/investigations and/or require certain medical tests to be undertaken. Reference is in this regard made to Figure 1 a.

The present invention however effectively turns this process around whereby the system, based on the prescribed medication, medical procedures/investigations/tests performed/required by the doctor, is capable of predict/recommend the correct coding of a medical diagnosis to a relatively high degree of accuracy. Reference is in this regard made to Figure 1b.

In Figures 2-4, reference numeral 10 refers generally to a diagnostic coding prediction system in accordance with the invention.

The system 10 includes a server 12 which is configured (e.g. by way of a processor 13 and appropriate software) to provide an interface module 14. The interface module 14 is configured to communicate with a doctor 50 (or other user or system 81 ) via a communication network 100. More specifically, a doctor 50 would typically use a computing terminal (e.g. a laptop 102 or smart device 104) in order to send information to the server 12. The information typically includes information on: medicine(s) prescribed, provided and/or administered to a particular patient; one or more consumables used during the medical consultation; one or more treatment actions performed on/for the patient (e.g. by a medical practitioner); and/or one or more diagnostic tests performed or laboratory orders placed for the patient.

The server 12 is also configured (e.g. by way of software) to provide a diagnostic prediction module/engine, more specifically a diagnostic coding prediction module/engine (hereinafter referred to as “diagnostic coding prediction engine 16”) which is configured to predict one or more possible diagnoses and the appropriate diagnostic codes for the diagnoses by using a prediction algorithm/model (hereinafter referred to as “prediction algorithm”) which utilises, amongst others, one or more therapeutic qualities which is/are associated with each of the medicines prescribed, provided and/or administered to a particular patient. More specifically, the prediction algorithm/model implemented by the diagnostic coding prediction engine 16 is an Al (artificial intelligence) based/machine-learning based model.

Modellinq of the diaqnostic codinq prediction enqine

Reference is in this regard made to Figure 5.

The diagnostic coding prediction engine 16 is modelled (see block 17) using a learning dataset 18 derived from a community of doctors/physicians. The historical dataset includes:

• prescriptions historically provided to various patients (e.g. prescribed medicines);

• consumables historically used during various medical consultations with patients (e.g. injections, sutures and/or bandages);

• one or more treatment actions historically applied to/on patients; and

• historical claims which are associated with various patients (e.g. claim acceptances/rejections (optional). In addition to the information above, a patient medical history dataset 20 could be used. Information on whether a patient has or does not have medical insurance information may also be included.

By using these datasets 18, 20, the prediction algorithm is trained. Optionally, a cost optimization function 22 can also be used during the training. This cost optimization function 22 may include metrics that measure the accuracy of predictions and/or the level of payment received from a medical insurer.

Additionally, a domain ontology 24 (such as the medical SNOMED ontology) can be added to the training of the prediction algorithm.

The training set out above produces a diagnostic coding prediction engine 16 which can be used to make diagnostic predictions.

As mentioned, in practice, doctors are required to utilise a diagnostic coding system, such as the ICD-10-CM coding system, to enter the correct diagnostic code which is associated with a particular diagnosis. The diagnostic coding prediction engine is therefore also trained using a particular coding system (e.g. the ICD-10-CM coding system) in order to allow it to not only make diagnostic predictions, but provide a doctor with a list of predicted, diagnostic codes from which a doctor can make a selection. The prediction algorithm can therefore be referred to as a diagnostic code prediction algorithm.

Use of the diaqnostic codinq prediction enqine durinq prediction

The process to predict a list of diagnoses/diagnostic codes is illustrated in Figure 6.

Information from the patient encounter (see block 30) is entered on a computing terminal of a doctor (e.g. a laptop 102 or smart device 104) and sent to the server 12, which stores it on a database 21. The information is then used by the prediction algorithm (see block 32) in order to produce a ranked list of the most probable diagnoses described by one or more diagnostic codes (see block 34). This information could include:

• A chief complaint articulated at the start of the visit;

• T reatment actions performed by the doctor;

• Any medicines dispensed, prescribed and/or administered; and/or

• Any diagnostic tests performed or laboratory orders placed.

This information could also be supplemented by additional information relating to the patient’s insurance status and medical history (see block 36), such as:

• Age and sex;

• Previous diagnoses;

• Laboratory results;

• Previous treatments;

• Environmental conditions;

• Patient genetic markers;

• Etc.

It should be appreciated that the information may however also be received/retrieved from another system or a database (e.g. a shared database), through natural language processing, or from an OCR module/program (e.g. an OCR of a script a doctor wrote).

With regards to the medicines dispensed, prescribed and/or administered, the therapeutic qualities of each medicine is determined by using a medical repository/database 23. The medical repository typically includes details of the therapeutic qualities of each of a large collection of medicines. These therapeutic qualities are then used, together with the other information, by the prediction algorithm in order to produce a ranked list of the most probable diagnoses and associated diagnostic codes.

The ranked list is sent back to the laptop 102 or smart device 104 (e.g. through the interface module 14), which is displayed on a display screen of the laptop 102 or smart device 104. The doctor (or other user) then selects the correct diagnostic codes, which is communicated back to the prediction algorithm (see block 38). The algorithm then makes an additional prediction. The process continues until the user is satisfied that the diagnosis has been fully coded.

The prediction process will now be described in more detail, with reference to a few specific examples. In the examples below, “ICD” refers to the specific ICD code and “Prob” refers to the probability of the associated code.

Example 1 : Patient with respiratory issues:

A decongestant is a very common medicine to give to a patient that has a respiratory issue:

Input treatment tokens:

D-0197-RO 1A (Medicine - DECONGESTANTS AND OTHER NASAL PREPARATIONS FOR TOPICAL USE)

ICD Prob Description

J06 - 0.289 - Acute upper respiratory infections of multiple and unspecified sites

J30 - 0.263 - Vasomotor and allergic rhinitis

J01 - 0.157 - Acute sinusitis

J11 - 0.035 - Influenza, virus not identified

R04 - 0.016 - Haemorrhage from respiratory passages

J00 - 0.015 - Acute nasopharyngitis [common cold]

J45 - 0.011 - Asthma

B34 - 0.010 - Viral infection of unspecified site

The output from the diagnostic coding prediction engine 16 provides a list of the likely diagnoses. One possibility is that the origin of the respiratory issue is an allergic reaction (like rhinitis). In these cases, the doctor adds an antihistamine to the list:

Input treatment tokens:

D-0197-R06A (Medicine - ANTIHISTAMINES FOR SYSTEMIC USE)

ICD Prob Description

J30 - 0.335 - Vasomotor and allergic rhinitis

J06 - 0.290 - Acute upper respiratory infections of multiple and unspecified sites

J01 - 0.262 - Acute sinusitis

J11 - 0.019 - Influenza, virus not identified

J00 - 0.016 - Acute nasopharyngitis [common cold]

B34 - 0.014 - Viral infection of unspecified site

J10 - 0.012 - Influenza due to other identified influenza virus

R04 - 0.010 - Haemorrhage from respiratory passages

The diagnostic coding prediction engine 16 then processes this and the probability for diagnosis J3 (Vasomotor and allergic rhinitis) jumps to the top of the list. Also high on the list are the viral diagnoses, like “Influenza” and “Acute nasopharyngitis [common cold]”. These viruses create a similar allergic reaction in the nasal passages.

If the doctor adds an antibiotic to the list it is usually an indication that the patient has a bacterial infection, or the patient is at a high risk of developing a bacterial infection. The bacterial infection could be present in one or more of:

• Pharynx

• Sinuses

• Tonsils

• Chest (bronchioles) • Middle ear (Otitis media)

Input treatment tokens:

D-0197-J01C (Medicine - BETA-LACTAM ANTIBACTERIALS, PENICILLINS)

D-0197-R01A (Medicine - DECONGESTANTS AND OTHER NASAL PREPARATIONS FOR TOPICAL USE)

D-0197-R06A (Medicine - ANTIHISTAMINES FOR SYSTEMIC USE)

ICD Prob Description

J06 - 0.335 - Acute upper respiratory infections of multiple and unspecified sites

J01 - 0.274 - Acute sinusitis

J20 - 0.032 - Acute bronchitis

J03 - 0.030 - Acute tonsillitis

J02 - 0.014 - Acute pharyngitis

H66 - 0.013 - Suppurative and unspecified otitis media

J22 - 0.012 - Unspecified acute lower respiratory infection

L01 - 0.010 - Impetigo

The diagnostic coding prediction engine 16 then correctly predicts the list of likely diagnoses.

In some cases the patient may have a history of asthma. In these cases, a lower respiratory infection could trigger a potentially dangerous flare up. In these cases, the doctor may give the patient a drug to reduce the likelihood of trigger such a flare up.

Input treatment tokens:

D-0197-R03C (Medicine - ADRENERGICS FOR SYSTEMIC

USE)

D-0197-J01C (Medicine - BETA-LACTAM

ANTIBACTERIALS, PENICILLINS) D-0197-RO 1A (Medicine - DECONGESTANTS AND OTHER NASAL PREPARATIONS FOR TOPICAL USE)

D-0197-R06A (Medicine - ANTIHISTAMINES FOR SYSTEMIC USE)

ICD Prob Description

J06 - 0.304 - Acute upper respiratory infections of multiple and unspecified sites J45 - 0.088 - Asthma

J22 - 0.055 - Unspecified acute lower respiratory infection

J20 - 0.051 - Acute bronchitis

J21 - 0.025 - Acute bronchiolitis

J41 - 0.019 - Simple and mucopurulent chronic bronchitis

J40 - 0.015 - Bronchitis, not specified as acute or chronic

In this case the diagnosis “Asthma” rises to second place in the list.

Example 2: Patient with severe back pain (dorsalgia):

Patients with back pain are usually treated with a combination of analgesics and anti-inflammatories.

The doctor starts with a standard analgesic (like paracetamol/acetaminophen) :

Input treatment tokens:

D-0197-N02B (Medicine - OTHER ANALGESICS AND ANTIPYRETICS)

ICD Prob Description

G44 - 0.150 - Other headache syndromes

R51 - 0.094 - Headache

M79 - 0.093 - Other soft tissue disorders, not elsewhere classified M54 - 0.067 - Dorsalgia

J06 - 0.061 - Acute upper respiratory infections of multiple and unspecified sites

R10 - 0.031 - Abdominal and pelvic pain

M25 - 0.019 - Other joint disorders, not elsewhere classified

G43 - 0.019 - Migraine

The diagnostic suggestions are appropriately broad.

As soon as the doctor adds an anti-inflammatory the diagnostic coding prediction engine 16 becomes more specific:

Input treatment tokens:

D-0197-N02B (Medicine - OTHER ANALGESICS AND ANTIPYRETICS)

D-0197-M01A (Medicine - ANTIINFLAMMATORY AND ANTIRHEUMATIC PRODUCTS, NON-STEROIDS)

ICD Prob Description

M54 - 0.220 - Dorsalgia

M79 - 0.160 - Other soft tissue disorders, not elsewhere classified

G44 - 0.053 - Other headache syndromes

M25 - 0.051 - Other joint disorders, not elsewhere classified

M13 - 0.042 - Other arthritis

R51 - 0.039 - Headache

R52 - 0.024 - Pain, not elsewhere classified

When the doctor also gives an anti-inflammatory injection, the probability of dorsalgia increases and other diagnoses like gout appropriately are included in the list:

Input treatment tokens:

D-0197-N02B (Medicine - OTHER ANALGESICS AND ANTIPYRETICS) D-0197-M01A (Medicine - ANTIINFLAMMATORY AND ANTIRHEUMATIC PRODUCTS, NON-STEROIDS)

C-0201 -M01A (Injection - ANTIINFLAMMATORY AND ANTIRHEUMATIC PRODUCTS, NON-STEROIDS)

ICD Prob Description

M54 - 0.344 - Dorsalgia

M79 - 0.143 - Other soft tissue disorders, not elsewhere classified

M25 - 0.061 - Other joint disorders, not elsewhere classified

M13 - 0.057 - Other arthritis

M10 - 0.036 - Gout

M15 - 0.026 - Polyarthrosis

Example 3: Diabetes and hyperlipidemia

Diabetes and hyperlipidemia are very often correlated. This example illustrates how the diagnostic coding prediction engine 16 is able to predict the correct diagnosis based on the treatments.

Input treatment tokens:

P-4050 (Procedure - Blood glucose test with a urine dipstick)

ICD Prob Description

110 - 0.163 - Essential (primary) hypertension E11 - 0.141 - Type 2 diabetes mellitus

ZOO - 0.085 - General examination and investigation of persons without complaint and reported diagnosis

E78 - 0.031 - Disorders of lipoprotein metabolism and other lipidaemias

R53 - 0.029 - Malaise and fatigue

Z01 - 0.015 - Other special examinations and investigations of persons without complaint and reported diagnosis The diagnosis coding prediction engine correctly predicts a number of diagnoses that are associated with blood glucose tests. The system correctly includes the code ZOO (General examination) in the list because doctors routinely perform glucose tests during routine physical examinations.

If the doctor adds a statin medicine to treat hyperlipidemia, then the predictions change accordingly:

Input treatment tokens:

P-4050 (Procedure - Blood glucose test with a urine dipstick)

D-0197-C10B (Medicine - LIPID MODIFYING AGENTS, COMBINATIONS)

ICD Prob Description

E78 - 0.663 - Disorders of lipoprotein metabolism and other lipidaemias

110 - 0.105 - Essential (primary) hypertension E11 - 0.084 - Type 2 diabetes mellitus 125 - 0.031 - Chronic ischaemic heart disease E10 - 0.006 - Type 1 diabetes mellitus E14 - 0.004 - Unspecified diabetes mellitus

Finally, if the doctor adds a blood glucose lowering medicine the diagnosis of “Type 2 diabetes mellitus” rises to the top:

Input treatment tokens:

P-4050 (Procedure - Blood glucose test with a urine dipstick)

D-0197-C10B (Medicine - LIPID MODIFYING AGENTS, COMBINATIONS)

D-0197-A 10B (Medicine - BLOOD GLUCOSE LOWERING DRUGS, EXCL. INSULINS) ICD Prob Description

E11 - 0.736 - Type 2 diabetes mellitus

E78 - 0.137 - Disorders of lipoprotein metabolism and other lipidaemias

110 - 0.084 - Essential (primary) hypertension E14 - 0.014 - Unspecified diabetes mellitus E10 - 0.010 - Type 1 diabetes mellitus R73 - 0.007 - Elevated blood glucose level

The correct coding for this patient’s condition would likely be a combination of E11 and E78.

Reference is also now made to Figure 4 which provides an overview of the system 10. As illustrated, a doctor or another user 50 typically enters the following information on a computing terminal, such as a laptop 102:

• A chief complaint articulated by the patient at the start of the visit;

• T reatment actions performed by the doctor on the patient;

• Any medicines dispensed, prescribed and/or administered to the patient; and/or

• Any diagnostic tests performed on, or laboratory orders placed for, the patient.

This information is then sent to the server 12 via the communication network 100. The server 12 then utilises a medical repository 23 (as described above) in order to determine the therapeutic qualities of the medicines dispensed, prescribed and/or administered (see block 42). These therapeutic qualities are then used by the prediction algorithm, together with the other information received, in order to generate a ranked list (based on probability) of diagnoses/diagnostic codes (see block 43). In order to do so, the prediction algorithm typically utilises a diagnosis repository 44 and/or a diagnosis coding repository 46. The ranked list is then displayed to the doctor or another user 50 which then either confirms the diagnosis/diagnostic code from the list or adjusts the diagnosis/ diagnostic code.

The diagnostic coding prediction engine 16 can typically be configured to use any adjustment made by a doctor 50 to train the prediction algorithm further.

It should be noted that the system 10 could be adapted/developed further, based on a particular prescribing behavior of the specific healthcare provider (e.g. doctor) to whom the most likely diagnosis are presented, and/or a community of healthcare providers (e.g. a practice, a hospital, providers of the same or similar specialty, segments by age, geography, etc.) for a particular medical diagnosis, to automatically and continuously improve the diagnosis prediction, and automatically learn the evolving prescribing behaviour(s).

The system 10 could also be developed further to suggest diagnostic codes from a new and potentially unknown code set to a doctor, based on the old code and the therapeutic actions of the medicines prescribed. This could apply in situations where there is a transition from an old code set to a new code set (e.g. from ICD10 to ICD11 ).

Based on the description set out above, it should be clear that the present invention helps a doctor in simplifying the process to select highly precise diagnoses/diagnosis codes very quickly and accurately in a way that both saves time and increases the quality of the codes. The system 10 allows a doctor (or other user) to simply select a diagnosis code instead of typing and searching for it in the traditional way and be able to code the correct diagnosis more accurately.

Claims

CLAIMS:

1 . A diagnostic prediction system wherein the system includes: an interface module which is configured to receive/retrieve information on one or more medicines prescribed, provided and/or administered to a patient; and a diagnostic prediction module/engine which is configured to predict one or more possible diagnoses by using a prediction algorithm/model which utilises one or more therapeutic qualities which is/are associated with the one or more medicines.

2. The system according to claim 1 wherein each possible diagnosis is associated with, or related to, one or more specific diagnosis/diagnostic codes.

3. The system according to claim 2 wherein the diagnostic prediction module/engine is a diagnostic coding prediction module/engine which is configured to predict one or more possible diagnoses and the appropriate diagnosis/diagnostic code(s) to correctly code the diagnosis, by using the prediction algorithm/model.

4. The system according to claim 3 wherein the diagnosis/diagnostic code is related to, or based on, a specific diagnostic coding system.

5. The system according to claim 4 wherein the interface module is configured to provide a list of predicted diagnosis/diagnostic codes which are most likely associated with the patient.

6. The system according to claim 5 wherein the interface module is configured to send the list to a computing terminal.

7. The system according to claim 4 wherein the diagnostic prediction module/engine is configured to predict one or more possible diagnoses which is/are the most likely and to propose the most appropriate codes to correctly describe the diagnosis, by using the prediction algorithm/model.

8. The system according to claim 7 wherein information on the therapeutic qualities of each of a plurality of medicines is stored on a database and the diagnostic prediction module/engine is configured to utilise the information stored in the database within the prediction algorithm/model, in order to predict one or more possible diagnoses which is/are the most likely.

9. The system according to claim 8 wherein the interface module is configured to receive/retrieve information on a plurality of medicines prescribed, provided and/or administered to the patient, and the interface module is further configured to receive the information on the one or more medicines prescribed, provided and/or administered to a patient via a computing terminal.

10. The system according to claim 9, wherein the prediction algorithm/model is an artificial intelligence/machine learning algorithm/model.

11 . The system according to claim 10, wherein the algorithm/model is trained by using a dataset which includes any one or more of the following: consumables historically used during various medical consultations with patients; prescriptions historically provided to various patients; historical claims which are associated with various patients; one or more treatment actions historically applied to/on patients; medical history data of various patients; and historical claim acceptance/rejection information from health insurers.

12. A diagnostic prediction method wherein the method includes: a. receiving/retrieving, by using a processor, information on one or more medicines prescribed, provided and/or administered to a patient; and b. predicting, by using a processor, one or more possible diagnoses by implementing/using a prediction algorithm/model which utilises one or more therapeutic qualities which is/are associated with each of the one or more medicines.

13. A method according to claim 12 wherein each possible diagnosis is associated with, or related to, one or more specific diagnosis/diagnostic codes and step (b) more specifically includes predicting, by using a processor, one or more possible diagnoses and an appropriate diagnosis/diagnostic code(s) to correctly code the diagnosis, by implementing/using the prediction algorithm/model.

14. A method according to claim 13 wherein the diagnosis/diagnostic code is related to, or based on, a specific diagnostic coding system.

15. A method according to claim 14 wherein the method includes providing a list of predicted diagnostic codes which are most likely associated with the patient to a user via a visual display, and sending the list to a computing terminal, which then displays the list on a display screen.

16. A method according to claim 15 wherein the method includes ranking, by using a processor, the list based on probability, by using the prediction algorithm/model and sending the ranked list to a computing terminal.

17. A method according to claim 14 wherein step (b) includes predicting, by using a processor, one or more possible diagnoses which is/are the most likely, by using the prediction algorithm/model.

18. A method according to claim 17 wherein the information on the therapeutic qualities of each of a plurality of medicines are stored on a database and step (b) includes utilising the information stored in the database within the prediction algorithm/model, in order to predict one or more possible diagnoses which is/are the most likely.

19. A method according to claim 12 wherein step (a) includes receiving/retrieving information on a plurality of medicines prescribed, provided and/or administered to the patient.

20. A method according to claim 12 wherein step (a) includes receiving the information on the one or more medicines prescribed, provided and/or administered to a patient via a computing terminal.

21. A method according to claim 18 wherein the method includes receiving a selection, via a computing terminal, of one of the possible diagnoses from the user, which should be used as the correct diagnosis of the patient.

22. A method according to claim 18 wherein the method includes receiving/retrieving, by using a processor, any one or more of the following information: information on one or more consumables used during the medical consultation; information on one or more treatment actions performed on/for the patient; and/or information on one or more diagnostic tests performed or laboratory orders placed for the patient.

23. The method according to claim 12, wherein the prediction algorithm/model is an artificial intelligence/machine learning algorithm/model.

24. The method according to claim 23, which includes training the algorithm/model by using a dataset which includes any one or more of the following: consumables historically used during various medical consultations with patients; prescriptions historically provided to various patients; historical claims which are associated with various patients; one or more treatment actions historically applied to/on patients; medical history data of various patients; and historical claim acceptance/rejection information from health insurers.