Tool Mechanism & Limitations 1

During a structured dynamic conversation or interview with our Health AI, approximately 25 questions are asked over 5 minutes to collect information about users’ symptoms, demographics, medical history, risk factors, and other related information.

The conditions presented on the screen at the end of the health chat are ordered by the estimated prevalence of the conditions in our database in relation to the symptoms and attributes reported during the health chat. The  list is a) not exhaustive and b) does not constitute a diagnosis:

  1. Using supervised machine-learning models, our algorithms were initially trained on an EMR dataset of two million patients of an HMO in Israel. Information from digital medical interviews and the collective expertise of medical providers on the platform was then leveraged to retrain the algorithm, which now covers 208 common adult primary care conditions, including the top reasons for primary care visits globally. The model does not present conditions deemed to be ‘non-serious’ or ones calculated by the algorithm to have a probability of ~5% or less (depending on the condition). Conditions not included in the 208 conditions covered by the algorithm are likewise excluded. (See the full list of K conditions below. For an additional explanation of the development of the algorithms, see the ‘Studies’ section below.)
  2. The potential conditions listed at the end of the health chat are for informational purposes only. They do not replace clinicians’ judgment or constitute a diagnosis. This is because, among other reasons, the data available to the diagnostic models is limited compared to what clinicians can access – prior encounters included. Moreover, the models do not incorporate external data, making it challenging to capture real-world trends. Instead, this list provides insight into how people with similar symptoms and specific shared biomarkers were diagnosed and treated, after which users can chat with a provider that practices on the K Health platform for proper diagnosis and care by clicking the relevant button.

Studies

  • Zeltzer D, Herzog L, Pickman Y, et al. Diagnostic Accuracy of Artificial Intelligence in Virtual Primary Care. Mayo Clinic Proceedings: Digital Health. 2023; 1(4):480-489. (available at  https://www.sciencedirect.com/science/article/pii/S2949761223000706)
  • Koren G, Souroujon D, Shaul R, et al. “A patient like me” – an algorithm-based program to inform patients on the likely conditions people with symptoms like theirs have. Medicine (Baltimore). 2019;98(42):e17596 (available at https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6824689

***No prospective clinical trials have been conducted to test the performance of these treatment models or the AI system functionalities described above.

Appendix 1: list of covered K conditions

Condition Name
Abdominal pain
Acne
Acute Atrial Fibrillation (PAF)
Acute diverticulitis
Acute hepatitis
Acute sinusitis
Allergic reaction
Allergic rhinitis
Amaurosis Fugax
Anal fissure
Anemia – General
Angina stable
Anxiety disorder
Appendicitis
Asthma
Athletes foot
Atrophic vaginitis
Bacterial vaginosis
Balanitis
Bell’s palsy
Benign positional vertigo
Bladder infection
Blepharitis
Bowel obstruction
BPH
Carcinoid syndrome 
cardiac arrhythmia unspecified
carpal tunnel disease
Cat-scratch disease
Cataract
Celiac disease
Cellulitis
Cervical radiculopathy
Cervical sprain
CHF
CHF decompensation
Cholecystitis
Cirrhosis
Claudication
Cluster headache
Colorectal cancer
Condyloma accuminata 
Conjunctivitis
Constipation
COPD
COPD exacerbation
Corneal abrasion
COVID-19
Crohn’s disease
Cushing’s Syndrome
Dehydration
Dental infection/abscess
Dermatitis
Diabetes Insipidus 
Diabetes type 1
Diabetes type 2
Diabetic foot
Diabetic neuropathy
Diverticulosis
Dry eye
DVT
Dysmenorrhea
Ectopic pregnancy
Endometriosis
Epididymitis
Epilepsy
Episcleritis
Erectile dysfunction
Erysipelas
Erythema nodosum
Fatigue
Fibromyalgia
Gastritis
Gastroenteritis
Gerd
GI parasite infection
Gouty arthritis/gout
Graves disease
Guillain-Barré syndrome (GBS)
Headache
Heat stroke
Hemorrhoids
Hernia – uncomplicated
Herpes genital
Herpes Simplex UNS
Hives
Hordeolum
Hyperopia
Hyperthyroidism
Hypothyroidism
IBS
Impacted cerumen
Incarcerated hernia
Influenza
Ingrown toenail
Lactose intolerance
Local benign breast condition
Low back strain
Lumbar radiculopathy
Lupus
Lyme disease (acute phase)
Lymphadenitis
Lymphedema
Macular degeneration
Major depression
Mastitis
Meniere’s disease
Meningitis
Menopause
Migraine
Mittelschmerz
Mononucleosis
Multiple myeloma
Multiple sclerosis
Musculoskeletal chest pain
Myasthenia gravis
Myocardial infarction
Narrow angle glaucoma
Nearsightedness
Nephrotic syndrome
New pregnancy
Obstructive sleep apnea
Optic neuritis
Oral thrush
Orthopedic Elbow Conditions
Orthopedic Finger Conditions
Orthopedic Shoulder Conditions
Orthopedics condition Ankle Pain
Orthopedics condition Hip Pain
Orthopedics condition Knee Pain
Orthostatic hypotension
Osteoarthritis
Otitis externa
Otitis media
Ovarian cyst
Ovarian torsion
Pancreatitis
Panic attack (nee disorder)
Paronychia
Parotitis
Pelvic inflammatory disease
Peptic ulcer disease
Perianal abscess
Pericarditis
Peripheral vascular disease
Pheochromocytoma
pilonidal abscess
Plantar fasciitis
Pneumonia
Pneumothorax
Polycystic ovarian syndrome
Polymyalgia rheumatica
Postconcussive syndrome
Premenstrual syndrome
Presbycusis
Presbyopia
Prolapse rectum
Prostatitis
Pulmonary embolism
Pyelonephritis
Raynaud’s disease
Renal colic
Restless leg syndrome 
Retinal detachment UNS 
Rheumatoid arthropathy
Sarcoidosis
Scleroderma
Sensorineural hearing loss
Septic arthritis
Shingles
Submandibular sialadenitis
Sjogren’s syndrome
Spinal stenosis
Ankylosing spondylitis
STD
Stomatitis aphthous
Strep pharyngitis
Stroke
Subarachnoid hemorrhage
Subconjunctival hemorrhage
Subdural hemorrhage
Superficial vein thrombosis 
Temporal arteritis
Tension headache
Testicular cancer
Testicular torsion
TIA
Tinnitus
TMJ
Torticollis
Tremor benign essential
Trigeminal neuralgia
Ulcerative colitis
Ulnar neuropathy
Upper respiratory infection
Urethritis UNS
Uterine leiomyoma
Uterovaginal prolapse
Uveitis
Vaginal yeast infection
Varicocele
Varicose veins
Vasovagal syncope
Vestibular neuritis
Viral pharyngitis
Vitreous floater
Vocal cord polyp/nodule
Vulvovaginitis

 1Model development supported by the following medical literature:

Harrison’s principles of internal medicine. (1998). New York: McGraw-Hill, Health Professions Division.

Hopcroft, K. and Forte, V. (2020) Symptom sorter. 6th Edition, CRC Press, Boca Raton.