Aap blood test karwate hain. Doctor kehte hain sab normal hai. Aap ghar chale jaate hain.
Jo kisi ne aapko nahi bataya, wo ye ki un numbers mein se ek pichhle saal se 15% zyada tha. Aur do saal pehle se 30% zyada. Wo abhi bhi "normal" range mein hai, isliye na koi flag, na call, na follow-up. Lekin wo hil raha hai. Aur wo kis taraf ja raha hai, ye is baat se zyada matter karta hai ki wo abhi kahan hai.
Aaj use hone wala har clinical system ek hi tarah kaam karta hai. Kya ye number ek line ke upar hai ya neeche? Haan ya na. Modern healthcare ka poora infrastructure isi sawaal ka jawaab dene ke liye bana hai. Aur ye galat sawaal hai.
Wo sawaal jo matter karta hai
Ek hi din, do charts mein 1.4 ka creatinine aata hai.
Pehle patient ka creatinine athaarah mahine pehle 0.9 tha. Phir 1.0. Phir 1.2. Ab 1.4. Ise aage project karo toh ye patient ek saal ke andar kidney disease ki taraf ja raha hai. Lekin 1.4 alert threshold se neeche hai. Koi alarm nahi bajta. Visit "teen mahine mein repeat" pe khatam hoti hai.
Doosra patient teen saal se 1.4 pe hai. Stable. Managed. Routine.
Number ek hi. Ek patient mushkil mein hai. Doosra bilkul theek hai. Aaj clinical use mein koi bhi system ye fark nahi bata sakta. Na Epic. Na Cerner. Na unke upar bane AI tools mein se koi. Sab 1.4 dekhte hain aur ek hi sawaal poochte hain: kya ye line ke upar hai?
Sahi sawaal ye hai: ye kahan tha, kahan ja raha hai, aur kitni tezi se?
Har physician ye jaanta hai. Har system ise ignore karta hai.
Jab main patients dekhta tha, toh jin best clinicians ke saath maine kaam kiya, wo kuch aisa karte the jo EHR unke liye nahi kar sakta tha. Wo purane labs nikaalte the aur kisi kaagaz ke tukde pe trend lines kheenchte the. Teen saal ki values dimaag mein rakhte the aur direction ke baare mein sochte the. Wo ye nahi poochte the ki "kya ye number normal hai?" Wo poochte the "kya ye patient badal raha hai?"
Yahi reasoning achhi clinical medicine ki buniyaad hai. Aur yahi wo cheez hai jo koi bhi bade paimaane pe deployed clinical system nahi karta. EHR billing ke liye encounters document karne ke liye bana tha. Alerts thresholds check karne ke liye bane the. AI tools unhi threshold-based systems ke upar bane. Technology ki parat pe parat, aur sabse basic clinical sawaal ka jawaab phir bhi nahi milta: kya ye patient time ke saath behtar ho raha hai ya bigad raha hai?
Trajectories na padhne ki keemat
United States mein 37 million adults ko chronic kidney disease hai. Unmein se 90% ko iska pata nahi.
Aisa isliye nahi ki data maujood nahi hai. Unka creatinine saalon se dheere dheere khisak raha hai, chart mein dikh raha hai, har visit pe dekha gaya, file hua aur bhula diya gaya. CKD jaldi pakdi jaaye toh $2,000 se $5,000 saal mein manage ho sakti hai. Der se pakdi jaaye toh dialysis ka kharcha $90,000 saal ka hai. In dono outcomes ke beech ka fark bas itna hai ki koi trajectory padh raha tha ya nahi.
Yahi pattern diabetes, heart failure, COPD, hypertension mein bhi dikhta hai. Saalon ka aisa drift jo pakda ja sakta tha, lekin un systems ko nahi dikhta jo sirf ye poochte hain ki aaj ka number line ke upar hai ya nahi.
Aur koi ye kyun nahi karta
Zyadatar clinical AI snapshots pe train hota hai. Ek visit. Ek lab draw. Time ka ek pal. Model us pal ko high-risk ya low-risk classify karna seekhta hai. Wo aaj ka creatinine 1.4 dekhta hai. Use koi andaaza nahi ki athaarah mahine pehle creatinine kya tha.
Snapshot se trajectory nahi padhi ja sakti. Ek point se dheema drift nahi pakda ja sakta. Architecture hi ise rokta hai. Snapshot model mein aur data daalne se ye theek nahi hota. Model bada karne se bhi nahi. Patient ko represent karne ka tareeka badalna hoga. Time ko main dimension banna hoga, baad ki soch nahi.
Yahi Chroneme ko market ki baaki har cheez se alag banata hai. Same data pe ek behtar algorithm nahi. Health ko represent karne ka ek buniyaadi taur pe alag tareeka. Patient saalon aur providers ke paar ek continuous signal ki tarah maujood hai, alag-alag, bikhri hui visits ki series ki tarah nahi. System padhta hai ki time ke saath cheezein saath mein kaise badalti hain, aur jaanta hai ki kab koi trajectory aisi jagah ja rahi hai jahan dhyan dena zaroori hai.
Practice mein iska matlab
Pehla patient flag hota hai. Doosra nahi. Creatinine ek hi. Trajectories alag. Recommendations alag.
Chroneme ko hum isi tarah ka badlaav pakadne ke liye bana rahe hain. Hum ise ek pre-registered protocol ke tahat validate kar rahe hain, aur jo bhi milega wo publish karenge, ye bhi ki kahan ye kam padta hai.
Jo badlaav sabse pehle matter karte hain, wo shayad hi kabhi dramatic hote hain. Wo dheeme drifts hote hain. Har quarter 0.1 chadhta creatinine. Har mahine kuch beats upar khisakti heart rate baseline. Hafton mein dheere dheere girta oxygen level. Aise patterns jinhe koi akela data point kabhi flag nahi karega. Aise patterns jinke liye threshold-based systems apne architecture se hi andhe hain.
Clinical intelligence ki agli generation yahin hai. Emergencies pakadne mein nahi. Dheemi emergencies ko tez banne se pehle pakadne mein.
Sources: CKD prevalence ke figures National Kidney Foundation se.
Naye essays aur updates, email pe.