Written by | Wen Yehao Edited by | Wu Xianzhi
For many patients, leaving the hospital marks the end of in-hospital treatment, but the journey to recovery has only just begun.
Holding pages of discharge instructions filled with professional jargon, patients often have a vague understanding of medication and dietary restrictions. Once home, they face questions about wound care, metric monitoring, and warning signs for symptoms. Without real-time doctor guidance, most rely solely on subjective feelings to make judgments. Yet medical risks often lurk beyond these subjective sensations.
The service boundary of hospitals has long been confined to in-hospital settings. Continuous medical services frequently suffer from disconnection after patients are discharged.
Over the past few decades, the healthcare system has addressed shortages of doctors and medicines, improved efficiency through informatization, and connected patients with medical resources via the Internet. However, there has always been a lack of systematic, scalable solutions to meet health needs after patients leave the hospital.
Tao Xiaodong, President of iFLYTEK Healthcare, stated that the construction of Healthy China requires continuous strengthening of universal health management, disease prevention and control, and capabilities for continuous medical services.
The direction of continuous healthcare has long been clear. In general industry practice, primary healthcare institutions and hospitals have explored post-discharge management. The true bottleneck for the entire service chain lies in the rigid constraints of labor supply and service costs. As AI healthcare moves deeper into clinical applications, the service radius of medical staff is effectively extended, allowing limited manpower to handle the routine health management needs of patients after discharge.
From Shortage to Continuity: Four Evolutions of Medical Services
The service boundary of hospitals is essentially determined by supply capacity. Technology can push this boundary outward bit by bit. Looking back over the past few decades, each generation of technology solved many problems, but the segment "after patient discharge" remained relatively weak.
Specifically, Stage 1.0 was characterized by shortages of doctors and medicines, focusing primarily on treatment.
For a long period after the founding of the PRC, medical resources were relatively scarce, making it difficult to see a doctor due to shortages of personnel and supplies. The top priority of the healthcare system was building hospitals, expanding beds, and training doctors. Barefoot doctors and the three-tier medical prevention network brought basic healthcare to villages and towns.
For hospitals, the mission at this stage was simple: cure the patients who came through the door. Beds, operating tables, and emergency rooms consumed almost all of doctors' energy. The boundary of medicine was within the hospital, focused on acute diseases and life-saving; patients spent most of their time at home, relying largely on self-healing and experience for illnesses.
Stage 2.0 involved informatization, leading to efficiency improvements.
In the 1980s, large hospitals began exploring standalone information systems. In 1995, the Golden Health Project launched, and China's Hospital Information System (CHIS) went online. Subsequently, Electronic Medical Records (EMR) and PACS (Picture Archiving and Communication Systems) became widespread. In 2009, the new healthcare reform further increased investment.
Informatization served as an efficiency revolution within hospitals, moving registration, billing, pharmacies, and medical records from paper to computers. However, relevant systems were often built vertically, with inconsistent standards and incompatible interfaces. Inter-departmental communication within hospitals was difficult, let alone cross-hospital integration.
Stage 3.0 saw the birth of Internet healthcare, establishing connectivity.
Starting in 2010, online appointments and light consultations took the lead. 2014 was considered Year One of mobile healthcare, with over 80 financing deals related to the sector throughout the year. In 2015, Ping An Good Doctor launched, and Wuzhen Internet Hospital issued its first prescription. In 2018, the direction of "Internet + Healthcare" was officially endorsed.
The arrival of Internet healthcare unlocked the value of "connectivity," allowing patients to find doctors, medicines, and services online. However, "connectivity" does not equate to "continuity." The former is more about point-to-point connection—ask and leave—with costs dropping rapidly as technology advances. The latter, as a sustained behavior, sees costs accumulate over time.
Taking the hospital execution side as an example, follow-ups mainly relied on nurses making phone calls or creating WeChat groups. Doctors, busy with surgeries and ward rounds, had the will but lacked the capacity to manage out-of-hospital patients.
Tian Yanghua, President of the First Affiliated Hospital of University of Science and Technology of China (Anhui Provincial Hospital), did the math: patients often traveled specifically to register and visit the hospital just to view a single lab result. This hospital reaches a daily outpatient peak of 30,000 visits. If one-third of these needs could be diverted online, approximately 10,000 in-person visits could be reduced. Including accompanying family members, parking, and transportation, significant social costs could be saved.
The problem is that the supply of medical personnel is limited. Take a leading Grade-A tertiary hospital with millions of annual outpatient visits: despite having over 8,000 employees, there are only 2,400 clinical doctors. Supporting such a massive outpatient volume while providing refined services creates immense pressure. Scaling continuous services remains difficult. This long-standing contradiction, which hindered the industry, gradually showed signs of resolution as large models moved into clinical use and met the conditions for scaled application.
In Tao Xiaodong's view, this step has industrial inevitability. After ten to twenty years of informatization, internal data interoperability and operational efficiency have matured. Future investments naturally shift toward AI and intelligent technologies.
Facts support this. The arrival of large models fundamentally changed the cost structure of continuous services: repetitive labor is handled by AI, massive amounts of information are processed by AI, and service scope is expanded by AI, allowing medical staff to focus on professional judgment.
Han Hui, Vice President of Qilu Hospital of Shandong University, views this as a lever to implement the full chain of "prevention, diagnosis, treatment, and rehabilitation," pulling hospitals that previously focused on treatment back to the complete chain of prevention, diagnosis/treatment, and post-discharge follow-up.
However, possessing technical conditions does not equal service implementation. Continuous services require a complete solution spanning diagnosis, data, and services. This solution cannot be simply assembled from single-point capabilities. This is precisely the challenge that iFLYTEK Healthcare's accumulated expertise over the past years addresses.
AI Healthcare: Different Paths, Same Goal
As "continuity" becomes a common goal for various hospitals, vendors with different strengths offer their own answers based on their respective advantages.
From the hospital perspective, AI healthcare exploration generally follows several paths:
Some enter through mass health service gateways, leveraging traffic and ecosystem capabilities to connect users and cover broad health needs like consultations and chronic disease management. Others extend from in-hospital digital infrastructure, using hospital information systems and data accumulation to expand service boundaries outward. Still others start with breakthroughs in AI technology for specific diseases or single steps, focusing on improving efficiency and precision in specific diagnostic scenarios.
The common trend in the industry is stringing together scattered capabilities into a complete service chain.
Internet giants, relying on hundreds of millions of users, payment, and e-commerce ecosystems, have packaged health consultations, report interpretations, and chronic disease management into super apps. These players primarily address how services reach patients and how patients are organized.
Vendors that have long served hospital information systems and data middle platforms, acting as builders of hospital digitalization, are now extending services outside the hospital along the flow of data.
A batch of vendors started with imaging—the earliest standardized single-point algorithm—delivering department-level tools and expanding services from image diagnosis to more clinical scenarios, addressing efficiency and precision in specific diagnostic links.
While paths differ, goals align: making medical services truly continuous and complete. However, looking at the industry as a whole, after various attempts, a gap still exists between professional diagnosis and sustained patient service.
After all, real-world healthcare is inherently continuous. When patients are in the hospital, doctors can see the complete condition. But once discharged, subsequent treatment and rehabilitation progress are hard to capture. When patients return for visits, lacking context such as daily medication, metrics, and symptoms, judgments rely only on current complaints and examinations.
Therefore, what hospitals need is a product capable of continuously collecting and accumulating patients' daily out-of-hospital information—allowing hospitals to see recovery status after discharge and handing doctors the complete patient context during visits.
iFLYTEK Healthcare fills exactly this gap.
According to Tao Xiaodong, when iFLYTEK Healthcare was established in 2016, industry attention focused on imaging-assisted diagnosis and Internet healthcare. Imaging data was voluminous and highly standardized, while the Internet had already proven its efficacy in other industries.
He did not shy away from discussing iFLYTEK's path difference: "We do not have particularly strong Internet genes." In Tao Xiaodong's view, rather than capturing traffic first and then screening for commercial opportunities among users, iFLYTEK prefers to start from the essence of healthcare, standing from a clinical perspective to identify true patients and true users, achieving commercial success in the process.
Over the past decade, iFLYTEK has gradually built an integrated layout covering Government (G), Business (B), and Consumer (C), forming a complete link where "diagnosis generates data, foundations accumulate archives, and services catch patients."
For patients, what iFLYTEK catches during their post-discharge time is a series of interconnected services.
Regarding patient management, following patients through the entire journey from hospitalization to recovery, the AI Patient Management Platform has iterated to version 3.0, expanding from single-specialty diseases to hospital-wide stratification.
Before discharge, large models read through pre-surgical assessments, admission summaries, ward round records, and tests/checks. Combined with clinical guidelines, they profile the patient, turning "incomprehensible medical records" into "readable recovery plans."
After returning home, the platform stratifies by need: low-intervention requirements are automatically handled by AI for phone follow-ups, appointment reminders, and health education, free and available for consultation anytime. For key diseases during the golden window of rehabilitation, health managers collaborate with attending doctors, offering text/image consultations, video visits, and regular assessments.
Currently, the platform has accumulated over 160 standardized specialized disease pathways, covering more than 30 clinical departments, managed jointly by doctors, nurses, nutritionists, and rehabilitation therapists. Relying on intelligent follow-up Agents and dynamic evidence-based reasoning, plans automatically adjust with recovery progress. When risks emerge, graded interventions are triggered.
Take a real story as an example: A 69-year-old patient with a history of cerebral infarction and pulmonary embolism underwent his first coronary intervention in May 2026. In June, family members reported chest tightness on the platform. By September, symptoms occurred frequently accompanied by fatigue. AI identified risk clues, verified by health managers. Doctors from the Department of Cardiovascular Medicine at Sichuan Provincial People's Hospital judged that readmission was necessary, completing subsequent interventions according to a phased plan.
Such individual cases are being replicated in more hospitals.
It is reported that the AI Patient Management Platform has currently landed in over 300 hospitals nationwide, cumulatively serving 500,000 discharged patients.
iFLYTEK Xiao Yi, as an AI Health Butler, targets higher-frequency and more dispersed daily health needs.
In fact, the vast majority of moments related to health in a person's life occur outside the hospital, such as unexplained abdominal pain, arrows on physical examination reports, uncertainty about which department to visit, or wanting to confirm drug side effects. These fragmented scenarios rarely warrant an outpatient visit but are ever-present.
When users feel unwell, iFLYTEK Xiao Yi can conduct multiple rounds of active inquiry, probing continuously like a doctor to understand symptoms, duration, triggers, and associated manifestations layer by layer, then judging health directions and recommending departments.
Tian Yanghua cited examples: Many hospitals have an "Allergy Department," but ordinary people don't know what diseases it treats, often confusing "neurosis" with "psychosis." The same disease may belong to different departments in different hospitals. In the AI era, this complex medical knowledge is being conveyed to the vast majority of ordinary users through dialogue.
Furthermore, after uploading test reports and physical examination results, AI can interpret abnormal indicators item by item and prompt re-examination. iFLYTEK Xiao Yi has further upgraded comprehensive interpretation of multiple medical reports, linking reports from different times, connecting indicators, and comparing trends, rather than just reading a few arrows on a single report.
Building on this, iFLYTEK Xiao Yi added health risk insights and comprehensive health management plans. Based on dynamic health archives, it identifies risks and generates complete plans covering medication, re-examination, diet, and exercise, dynamically adjusting with indicator changes. For users preferring Traditional Chinese Medicine (TCM), it launched tongue/face constitution identification, recommending wellness plans such as diets, exercises, and tea beverages accordingly.
Following these capabilities downward, the form of health services is changing—extending from a single consultation to a lifetime.
Decade of B/G Accumulation: Realization and Revaluation of C-End Value
Whether continuous services can be solidified depends on the depth of co-construction with hospitals and industrial partners.
On the surface, competition occurs at the level of C-end products and traffic. At the bottom, however, it tests understanding of hospital scenarios, data accumulation, and thickness of trust. To provide good continuous services, hospitals need an industrial partner that can walk alongside them for the long term.
In this regard, iFLYTEK Healthcare's past years of accumulation on the B-end and G-end constitute exactly the foundation most needed for continuous services.
First is data. To serve a patient well, one must truly understand them. Previously, complete in-hospital diagnostic data for patients was scattered across different systems, requiring B-end integration to connect them.
Take imaging as an example. When a patient travels between several hospitals, they often rely on a bag of portable imaging films and paper reports. Whether films are complete and reports are lost directly determines what doctors at the next hospital can see. If materials are incomplete, repeated checks and spending are inevitable.
Carrying and safeguarding these materials fell entirely on patients. The original images doctors truly needed were locked within the systems of various hospitals.
Guangxi Medical Insurance Imaging Cloud aims to untie this knot. Co-built by iFLYTEK Healthcare, it aggregates images scattered across various medical institutions onto a single platform under patient authorization, becoming a continuous record following the patient.
This means that CT scans and MRIs taken at one hospital can be directly accessed by receiving doctors at another hospital for review or referral, allowing them to view layers and analyze comparisons.
As of September 2026, Guangxi Medical Insurance Imaging Cloud has connected 469 medical institutions, aggregated over 12 million imaging cases, and facilitated over 2.2 million cross-hospital accesses. Broadening the view, iFLYTEK Imaging Cloud overall serves over 260 million person-times, achieving full coverage in provinces like Anhui and Guangxi.
Model professionalism is equally important as data continuity.
In 2023, the iFLYTEK Spark Medical Large Model was released. At that time, it was the only medical large model in the industry trained on fully domestic computing power. In March 2025, the industry's first medical deep-reasoning large model, Spark Medical Large Model X1, landed, introducing evidence-based reasoning and knowledge reflection, significantly reducing medical hallucinations.
Subsequently, based on a medical knowledge base of nearly 1 billion entries and real service data averaging over 2.2 million interactions per day from both doctors and patients, the Spark Medical Large Model continued to iterate.
According to Tao Xiaodong, in actual applications, cases where AI recommendations differed from doctor judgments and were not modified by doctors are selected for expert team review. If AI is correct, this judgment is reinforced. If AI is wrong, it becomes a new data point feeding back into AI.
This cycle of "Use—Feedback—Adjudication—Retraining" drives the model forward. At the 2026 iFLYTEK Global 1024 Developer Conference, the Spark Medical Large Model iterated to version V3.7, achieving multimodal fusion of semantics, imaging, and voice. It ranked first in the industry with a score of 99.6 in the latest MedBench 5.0 Agent Evaluation released by Shanghai AI Laboratory.
Additionally, for the heavily regulated healthcare industry, trust and compliance cannot be ignored. In this aspect, iFLYTEK's long-standing B-end and G-end background serves as a major endorsement.
Ultimately, the fusion of data, models, and trust constitutes the long-term partnership between iFLYTEK and hospitals.
The most direct evidence of whether hospital-side demand is genuine is whether patients and hospitals are willing to pay continuously for it. iFLYTEK Healthcare's financial reports provide proof of this.
In 2025, iFLYTEK Healthcare's revenue from patient management services reached CNY 273 million, growing 29.3%. In the first half of 2026, iFLYTEK Healthcare's revenue was CNY 446 million, growing 49.4%, with AI health assistant revenue reaching CNY 182 million, growing 74.9%, becoming one of the fastest-growing business segments.
For instance, the "Qilu · Spark" Full-Course Management Large Model co-built by iFLYTEK Healthcare and Qilu Hospital of Shandong University has cumulatively managed over 270,000 patient-visits, reducing follow-up workload for medical staff by 90% and achieving patient satisfaction over 98%. The "Zhi Yi Sui Xing" Large Model co-built with the First Affiliated Hospital of USTC covers over 100 disease types, with about half of discharged patients included in management.
Through these numbers, iFLYTEK Healthcare's role becomes increasingly clear. It stands at the forefront of technology while walking alongside doctors, embedding AI into existing chains of diagnosis and rehabilitation, extending limited professional capabilities to the post-discharge time previously unserved—ensuring that everyone, whether inside or outside the hospital, receives continuous care and concern.
At its core, the story of continuous healthcare is just beginning. And the AI answer sheet submitted by iFLYTEK Healthcare is making quality healthcare truly accessible, sustainable, and trustworthy.