Claims data also known as administrative data. They're the electronic transaction data between patient and healthcare providers. They're required by insurance company because for doctors or any healthcare provider to get paid by an insurance company, they have to provide documentation and proof. Claims are the most important. The proof the doctor can provide to get paid by the insurance company. A different type of claims data. At the top level depending on the insurance types, we have Medicare claims data, those are national insurance program in the US, and we also have a commercial or private insurance data. Those are, for example, your employers provide healthcare to you and those claims are considered commercial or private insurance. Let's look at what information are in those claims. For the Medicare claims, there are four different types or different parts, Part A, Part B, Part C, and Part D. They capture different type of clinical visits. Part A is about inpatient visit, there's a hospital insurance. Part B is about medical insurance that covered outpatient visit, physician services, ambulance visit. Part C is very specialized Medicare advantage plan, so that's a capitation plan that individual can buy to augment their Medicare insurance. Part D is about prescription drugs. Commercial or private claims data have similar information as Medicare claims data, and it has this medical claims and pharmacy claims. For medical claims, they have inpatient claims that captures hospital visit, nursing facility visit, nursing home and hospice. Outpatient visit is another type of claims, it consists of ambulatory visit, outpatient center and facility fee. Then we have a professional claims, those are associated with doctors fee. Pharmacy claim is another type of claims and it merely captures the prescription drugs. What are the information in the claims data? What are the data elements? Obviously it has a membrane information, it has a date when this claims happen and the places, which hospital they visit. It has a lot of structure information such as diagnosis code, ICD decode, procedure information, what they call the HCPCS. That's really a procedure code like CPT and other type of information as well. Then you have a drug related information that document with NDC code. The financial information like how much money was charged or paid, or what's the total expenditure for this? What are the properties of claims? On the positive side, it's a very large volume, plenty of claims data are available. It provide a holistic view of patients, at least for that time period. All patient interactions with providers are captured in the claims. It can be used to somehow understand the medication compliance, because you can see all the filled prescriptions. The disadvantage, there could be coding errors. The claims data, not a clinical documentation. The main purpose is for billing purpose, so the quality can vary. It can have inaccurate data, and so it's an imperfect reflection of the actual patient status. It can have a time lags because claims filing may be delayed, sometimes over a few weeks or months, so it's not a real-time information. They're temporarily limited. What I mean by that is, depending on how long the patient have the same insurance, it may be difficult to track individual over long period of time, especially if the individual change insurance plan. If over a year they change the insurance plan, then the information will be difficult to capture in claims data from one insurance company. But of course, if you aggregate everything, some company does that, then that claims can be a more complete data about individuals. Claims versus EHR, they contain a lot of similar informations. Let's compare the differences between these two important patient information. In terms of scope of data, claims contains all transaction of the patient and provider interaction, while the EHR capture only the portion of care provided by a specific provider, because keep in mind EHR is hospital specific. If patient visit multiple hospital, then each hospital will have their own partial view of this patient. Medication data, for claims data you know the medication at our field, you don't know whether they take the medication but you know that at least they paid that medication copay and they got the medication. Well, the medication prescription, that's where the EHR data contains. That information only covers the prescriptions and you don't know whether they actually filled that prescription. The internal data richness and the claims data will be limited. It only have structured codes like diagnosis code, CPT code, and NDC code, while the EHR data are a lot richer. It has clinical notes, lab results, vital sign, problem lists, social history, imaging data, and so on. It's a lot richer than the claims data. Continuous signals is another important data that commonly collected in the hospital, especially in the intensive care unit. It monitors patients information with various sensors. For example, here we're showing the ECG monitoring screens and then the oxygen saturation data, heart rate monitoring, blood pressure, EEG, so those are all commonly monitored information that generate these continuous signals. The data are coming in real time, so if you want to analyze such data, you really need to provide real-time clinical decision support. What are the pros and cons of continuous signals? The pros are, there are very detailed data. Continuous monitoring data, they sample at very high frequency, for example, 200 hertz, so you will see a very detailed data from that sensor. It's also objective measures. There's no human opinion imorphed, it's not like if you looking at diagnosis, that's the opinion of a doctors after looking at that patient. Here the data output from the sensors, so they're more objective. What are the cons? It can be very noisy, depending on the sensor placements, or interference, or other patient movement, the data may be noisy. It's a very large volume, as I said, they can be sampled at very high frequency, so if you accumulate such data over a long period of time, over many patients, it could be a vast volume of data that need to be processed in real time.
