As a Healthcare IT provider company having very large amount of patients data from Healthcare EHR and EMR systems. We will do requirement de-identification and masking for PHI data before any usage in experiments. Want to understand possible Healthcare AI use cases that we can build. Data includes patients appintments, history, diagnosis, Lab results, medications, treatments, etc.
are you asking hoow. to mask patient personal healthcare data in healthcare system? or you are wanting to present how itâs done?
I am sorry your query is bit confusing? as long as I know patient id, social security number, location(address), or any related data to patientâs personal information is encrypted when third party channels are involved. This is surely not yet practice here in India everywhere but USA and UK have strict HIPA rules to be followed by any individual handling healthcare data.
No, I am not asking about masking patient data. I want to how we can leverage power of AI when we have patient data to build useful solutions for patients and healthcare providers based on the patient data available from many EMR systems.
For example, we can suggest specific medication based on the diagnosis, lab tests data Or predict patientâs health risk.
Recent advancements in Healthcare by Generative AI has made to consumes a whole lot of healthcare data and then has been used to transcribe just by listening to patient-doctor conversation, thereby handling administrative work.
A recent study also conducted by AI model was able to provide a provisional diagnosis based on patientâs clinical history, diagnostic findings which sometimes can be a lot of help in remote areas where doctors are not available.
From IBM watson for oncology to Insilico Medicine which help to find potential drugs for idiopathic pulmonary fibrosis.
In case of drug testing, earlier patient testing of newer drugs required more documentation and supervision from medical background healthcare professionals, where as recently study AI access ready data which reviewed all the patientâs data and made faster selection of clinical trials in drug screening protocols, something Insilico Medicine is currently doing.
AI powered Tool like AIDOC detects brain bleeds, pulmonary embolism, analysis real-time medical images, flags urgent needs, reduces time to diagnosis leading to better patient outcome.
Babylon Health AI symptoms check, flags life threatening condition and advised if one require immediate medical attention leading to early diagnosis and treatment which actually recently helped a patient in rural area where there was no healthcare provider and babylon health alerted the patientâs symptom to be life threatening causing immediate attention and alerting ambulance and saving the childâs health.
As far as prescribing medication to patients, I am against this as I know in India surely medication can be blessings as well as a curse. I have seen my patients being careless when they are medicated with OTC medicine and only present to clinic on emergency purpose, so for this I donât think AI should be used.
People working in AI healthcare also highly agree doctors are irreplaceable when it comes to healthcare and AI should be only used as a collaborative tool in providing better healthcare and attention in patient care.
Regards
DP
1 Like
Notice that this gives you what medication was prescribed but not what medication could or should have been prescribed. There is a knowledge gap between that and we can suggest specific medication that you provide as a functional objective. For that you also need a pharmacy/formulary and model of disease that is quite independent of any EMR or historical patient data.
One area that can be explored from the patient record side is patient similarity / cohort extraction, which is important for personalized/ precision medicine. It was an area of active research when I worked at IBM 15 years ago, and to the best of my knowledge has not been completely solved. Do be aware, though, that companies like IBM often research ideas like this way before they are commonly practiced in the field and have patents in place on system and algorithm design, even if they were never deployed. Other than the Z mainframe, patents are IBMâs most important âproductsâ from a revenue perspective. Due diligence highly recommend.
2 Likes
I agree Deepti. Healthcare is very sensitive and responsible. Just FYI, I am working with a company located in the USA and building solutions around EHR data migration, Archiving, Payer, Patient Networks, etc. mainy dealing with interoperability. I am also considering use cases where it helps in this projects with regards to operational efficiency too. For example, in solution like Interface Engine where you need to map Orders and Results Interface types data from one source EMR system to another EMR system. Here we can use AI to map HL7 fields, generate codebase and incorporate data rules.
Regards,
Dipak
ok so your creator topic was more of information than query??
I donât know how ai can be integrated in Indian healthcare system, my focus is my home first 
, as it is challenging task but not impossible. I wish government of India was more involved in making healthcare data digital which I do see recently every state is having a separate digital health initiative department.
You probably can share how you do the data integration between multiple channels of you are sharing information here.
Hi Deepti,
Let me provide some high level quick info. In USA and in general the Interoperability (accessibility and availabilities of patient data with required privacy and compliance requirement) is the key where patient data are available to healthcare systems like Hospitals, Clinicians, Labs, Insurance companies, etc. so that they can make right decisions with regards to patientsâ historical data and the latest data.
There is a platform like CommonWell you want to explore which is basically a network and above mentioned entities become members of the platform to share data. They are exposing many relevant APIs to share patientsâ data and other details based on the access permissions. So, API is one of the options to share data between different systems. The other systems can integrate the API response onto their website or solutions and also send to data to the external systems.
Hope this helps.
Regards,
Dipak
Thanks ai_curious. Your inputs are very much valuable for me. I am deep diving certain pointer you provided to come up with possible use cases along with how much they are âREALLYâ add values for the better patient outcome.
For example, trying get more details from Perplexity about cohort extraction:
*âIn this context, âcohort extractionâ means automatically identifying and pulling out a set of patients from health data (often EHRs) who share predefined clinical characteristics, events, or phenotypes.â
Those characteristics can include diagnoses, labs, medications, demographics, genomic markers, or even freeâtext note concepts, and algorithms often use similarity metrics or clustering over these features to build âcohorts of similar patients.âââ
*
I will update more and if any questions where your guidance would be helpful in order to plan efforts to build a really good use case.
Thanks again and appreciate your time.
Regards,
Dipak
This information I already knew because I worked for USA and UK health care medical transcribing industry during my early days for almost 5 years especially regarding HIPAA and PHI.
About API, I will surely explore the platform you mentioned but do you think security, privacy related to patientâs data is just limited to API especially knowing that many API are widely available on open web.
I also came across a news that some workers who were working for these transcribing industry were caught sharing patientâs data to third parties alaraming concern with people involved in the healthcare documentation, now that there are many voice over tools to record patient-doctor conversation, i really wonder how they address or would address in India as language and dialect versatility is vast, as a doctor i cannot expect a patient from South Indian village to be good in speaking English or hindi, infact even the kannada dialect changes from north Karnataka to south Karnataka.
Where as there are so many translation tool now available and I donât know how much robust they are to generalization, noise and new real-time data.
Regards
DP
This is how cohorts were identified in the past. Big data and machine learning enable discovery of the characteristics, both clinical and social, that make a set of patients similar. This might be particularly interesting if you are indeed capable of aggregating, even virtually, across EMR instances and healthcare delivery organizations.
Wonderful Sir! I will continue to coordinate and collaborate with you about the possible use cases I will come across, what values it would add, and the AI is being used to solve the ârealâ problem!
Thanks for your time and quick directions and inputs.
Regards,
Dipak