Creating a Data Set for Elderly Diabetic Patients

Diabetes Database for Older Patients
A diabetes database must capture information that is directly related to the disease and some of the valuable elements include age, A1C results, skin fold, blood pressure, weight, date of first symptoms, and blood glucose level. Age of the patient is valuable because it assists in determining how old the patient is and it can also be used to determine the best cause of treatment to be used especially for diabetes. The age element will have a data type of number since age is recorded in figures. A1C results are obtained after the patient has undergone testing for diabetes. Theses element is valuable because the results will help establish if the patient is diabetic or prediabetic (Balas & Boren, 2000). A1C results are recorded in number format because the results are normally obtained in percentage. Skinfold element is used to determine if the patient has any skin disease that is related to diabetes. Skinfold will be recorded as a binary because it requires a yes or no. Blood pressure is closely monitored for diabetic patients because a high blood pressure can be lethal especially for diabetic patients. Blood pressure will be a text value because it is measured in terms of systolic pressure over diastolic pressure. Weight element offers information regarding the BMI of the patient, which is vital for checking if the patient is overweight or underweight. Weight will be a number value. Date of first symptoms assists in establishing the onset of the disease and this can be beneficial for treatment management. The value will be a date. Blood glucose level will establish the patient's blood glucose level at first visit and assist in determining the appropriate treatment. This will be a number value in the database.
Type of Data Integration
This patient population would require a virtual integration. This is because the information stored in the database should only be viewed by others and not modified. A virtual integration allows for the unified view of data from a single logical source, which reduces the complexity of accessing data across distributed systems or data sources that have different models (Weaver, Ball, Kim, & Kiel, 2016). A majority of the other systems would only need to have an overview of the information and since they might be using different models, the best way is to allow them to access only the information they require. This would also increase the privacy of the patient and ensure that the data is not modified by unauthorized personnel. Virtual integration has zero latency of data updates and there is no need to have separate storage for the consolidated data. This is because the data is only provided as stored in the other dataset and no modification is allowed to the data.
Since all the two systems would be using the same database, virtual integration would be highly recommended. This is because there will be no need to have a separate storage of the data before it can be accessed by either of the systems. Virtual integration has a limitation in that it works only if the separate systems accessing the information are all running the same database. The EHR database would be able to facilitate this integration in that it would allow for the administration system to access only the information relevant for the administrators and the clinical system would have access to information relevant to the clinicians. The information would be virtual and there is no risk of having duplicate records in case an update is made on one side of the system.
Potential Clinical Question
The potential question that could be answered with data mining is what is the potential for skin disease development in elderly diabetic patients? The data mining techniques that would be used for this challenge include association, clustering, and prediction. Association technique involves making a correlation between two or more items in order to identify patterns. Association is a straightforward data mining technique and it would assist us to identify if there is a correlation between the elderly diabetic patients and skin disease. Clustering technique involves the grouping of individual pieces of data together in order to form a structured opinion (Tomar & Agarwal, 2013). At the most, basic level clustering would involve identifying a cluster and correlating its results. Prediction technique mainly involves analyzing trends, pattern matching, classification, and relation. This technique is normally used together with other data mining techniques. By analyzing the past events one is able to make predictions about what might happen in the future. For our case, this technique would allow us to determine how likely an elderly patient is likely to suffer from skin disease.
Individual Components to the Question
The individual components to our question are the age of the patient, is the patient diabetic, and skin disease. The age of the patient is vital because it will help us determine if the patient is elderly, and the information would be extracted from the age element of the database. Our question is only geared towards diabetic patients. Therefore, it is vital that we determine the patient is diabetic and this will be done by extracting information from A1C results, blood pressure, and glucose level elements. This information will assist in establishing the level of diabetes for the patient. Skin disease component will be extracted from the skin fold, and skin disease element of the database. This information is beneficial because it offers information about the condition of the patient's skin on admission or visit to the hospital.


References
Balas, E. A., & Boren, S. A. (2000). Managing clinical knowledge for health care improvement. Yearbook of medical informatics 2000: Patient-centered systems.
Tomar, D., & Agarwal, S. (2013). A survey on Data Mining approaches for Healthcare. International Journal of Bio-Science and Bio-Technology, 5(5), 241-266.
Weaver, C. A., Ball, M. J., Kim, G. R., & Kiel, J. M. (2016). Healthcare information management systems. Cham: Springer International Publishing.