Healthcare industry worldwide have been wrestling with shrinking margins, rising cost, declining market share, huge lawsuits due to medical errors, fraud and abuse as well as dramatic legislative changes especially in US. And there is huge pressures to improve on financial performance, deliver better quality service at the lower cost, better and quicker respond to customer's demand and simply keep in pace with the changing environment – which is often driven by customers as well as by competitors.
And then there are the abundance and richness of data. There are patient's demographic data, hospital data, physician notes, lifestyle data, billing data and medical and medication history. One of the biggest problems in data mining in medicine is that the raw medical data is big. And often in free-text format which means that text mining algorithms need to be used to parse and transform such data before using conventional data mining techniques. All these data components can have a major impact on diagnosis, prognosis and treatment of the patient, and should not be ignored. However, center of analysis doesn't always have to involve patient, but can be hospitals, doctors or even specific healthcare providers and schemes. Below is an overview of common applications of data mining and healthcare:
Disease management concerns with predictive as well as descriptive aspects of specific disease. What is likely probability of specific disease outcome, and what are the factors associated with these outcomes with the focus on actionable factors. One must separate the effects from causes for specific disease, and that can be done by separating the event period from the period of input data collection. Disease management can involve explicit aspect of the disease whose resolution can be beneficial to not only health-providers but more importantly to the patient. Descriptive component of disease management involves desirable as well as undesirable patterns – and acting on these patterns involves either supporting them or breaking them and then measuring effects of these actions for the purpose of achieving specific disease management goals.
Some of the examples of disease management questions:
- If surgical procedure "X" is done, then 45% of the time infection "Y" occurs within two weeks- Why, reasons, contributing factors?
- What, if any nosocomial infections exist in emergency room and contributing factors?
- Why do some congestive heart failure (CHF) patients return to the heart clinic after bypass surgery for care within 3 months, while others don't?
- Compare and contrast high length of stay patient groups based upon bed location, nursing teams, and treatment modalities.
- Compare and contrast treatment results or glucose levels for type II diabetic patients for a given time period, by physician, gender, age group, etc.
- What practice patterns for managing primary mammogram candidates will yield the best outcomes in terms of survival rates or complication rates at the least cost?
- What percentage of women in membership between the ages 40 - 60 have had a mammogram in the last 12 months?
- What is the comparative mean value of hypertension levels within a certain group or population of patients and does it fall within acceptable statistical levels? Do variations in clinical practice patterns have a cause and effect relationship?
Outcomes Analysis: Clinical and Financial
A Clinical Outcome is the result of medical or surgical intervention or nonintervention. It can refer to, but is not limited to the following:
- Re-admittance rates
- Changes in birth and death rates for a global population, for example, residents of a state
- The outcome of a given diagnostic procedure, lab result or medical test
- The results for a patient after care, for example, how long it took to restore the patient's ability to walk, or to work, or how long and to what degree did the patient have pain
- Did the patient recover, how long did it take
- The patient's own perception of their care and progress. It is thought that through a historical record of outcome experiences, caregivers will know better which treatment modalities result in consistently better outcomes for patients. Effective outcomes Management often relies on a successful data warehousing strategy designed to track historical outcome experiences in many areas such as epidemiological studies, lab results, responses to treatments, mortality and morbidity rates, length of patient stay and clinical effectiveness measures and then using these patterns to improve processes, treatments, diagnostics procedures, etc.
- The definition of a financial outcome varies depending upon an organization's goals and overall strategy. As an example, financial outcomes might cover measures such as hospital length of stay, net margins, cost breakouts, number of doctor or hospital visits, office visits - just to name a few. Knowing patterns of financial outcomes are key in improving not just key financial indicators in terms of the revenue but to – more importantly – improve patients experience in dealing with specific healthcare institution or entity.
The definition of a financial outcome varies depending upon an organization's goals and overall strategy. As an example, financial outcomes might cover measures such as hospital length of stay, net margins, cost breakouts, number of doctor or hospital visits, office visits - just to name a few. Knowing patterns of financial outcomes are key in improving not just key financial indicators in terms of the revenue but to – more importantly – improve patients experience in dealing with specific healthcare institution or entity.