Our healthcare system speaks about the society we live in. It is well narrated by Jared Diamond in his book Guns, Germs and Steel that in the course of human history, the world has always been divided into haves and have-nots where socio-economic inequality is linked to geographical regions and populations. As socioeconomic status is an important determinant of health, the factors that influence socioeconomic status inadvertently influence health and contribute to healthcare inequities.
The high cost of providing healthcare is one of the biggest challenges facing the industry. Hospitals are focusing on providing high-quality care without increasing costs. Some innovative models have been brought into healthcare delivery from other areas to address the issues of cost and quality. One such model is Risk Adjustment, which helps us understand medical spending and healthcare utilization while accounting for population risk. Risk is calculated from the diagnosis and conditions already existing in the population or an individual. Risk adjustment is well used in programs such as Medicare, Medicaid, and various other business models.
Risk adjustment was initially used to access the risk score for the enrollees to set health insurance premiums. Currently, risk adjustment is used to move dollars from low-risk payers to relatively high-risk payers based on the health condition/diagnosis of the patient. The Affordable Care Act (ACA) diversified risk adjustment to the individual and small group health insurance markets starting in 2014.
The risk adjustment models can be prospective and concurrent/retrospective based on the date periods and weights. Concurrent is calculated based on the already rendered service data whereas the prospective model predicts the future looking at the same data. Some of the risk adjustment models are influenced by the financial movement between insurance plans rather than the actual care delivered or the quality of care. Healthcare reforms brought many changes to risk adjustment models with regards to data collection and analysis, but we still have improvements to make to the models by taking into account health determinants that drive up the cost such as childhood obesity, maternal health, and infectious and chronic diseases. We have analyzed some topics during our research with the risk models such as:
Assessing high-risk clients: Risk assessment is key for risk adjustment and it depends mostly on the traditional parameters such as demographic data (age, sex, etc.) and insurance claims data (ICD codes, treatment codes, etc.). This data is analyzed and the population is segmented into a cost group. Prospective risk predicts the financial benefits from high to low-cost plans through these groups. However, a major challenge to this prediction lies in collecting the correct data. Data analysis of ICD codes explains only the cost rather than the intensity of the disease. The purpose of the assessment is mostly restricted to the number of visits rather than the actual health condition and prognosis of the patient.
Risk Adjustment Models: Many models are in place, but CMS-Hierarchical Conditional Categories (HCC) is the most sophisticated diagnosis-based tool used for calculating the risk for Medicare plans, both current and prospective. HCC currently has around 80 aggregated Conditional Categories (CC) along with ICD codes. With the change in ICD 10, there are around 70000 diagnosis code sets established for quality in data for better measurement of healthcare utilization. Meanwhile, the HCC risk model does not increase CC for analyzing the risk. With such a gap between the ICD data and HCC, it is difficult to have a better prediction for healthcare risk.
Data Availability: Getting the most out of demographics and claims data is the source for risk adjustment. The challenge with the prospective risk is it predicts based on the current year data and does not take the developing health conditions of the client into consideration. Data used for the models are missing critical diagnostic information due to claim processing time and gaps in Electronic Health Records. Current healthcare data is incomplete as the patient data is scattered among different providers and specialties services. As prospective risk adjustment is mostly for the benefit of plans to accommodate high-risk members, predictions using the incomplete data will create gaps in the exchange of financials between insurance plans.
Interoperability: One goal of healthcare informatics is to have the right data available to the right people at the right time. The need for interoperable systems should grow to enable administrative, clinical and diagnostic data integrate together as this will give a better predictive risk score for the patients and insurance plans alike. Risk score assessed on the basis of a complete picture of the care of a member will lead to better quality and incentives towards plan and providers.
Considering Co-morbidities Through Data Mining: Health risks happen throughout a client’s life cycle. Data on comorbidities is a key element to predict hospital stays, costs, morbidity, and mortality rate. In order to gather this data for the prospective risk adjustment, one has to mine historical claims data along with considering the current year data. An example of mining the data is looking for factors that cause obesity, which increases the frequency of cardiovascular disease and diabetes in the client.
Considering SES and Other Non-traditional variables: Socio-economic status is becoming a vital source in healthcare as it describes some of the diseases brought on by factors related to income, lifestyle, behavior, food habits, geography, and educational status. As many socio-economic factors are not directly associated with a disease they do not exist in the insurance claims data, but these factors play a major role in health risk assessments.
Rise of Mobile Health Data: With the growing technological advancements, the demand for wearable fitness devices is growing as well. Collecting this data from different apps and social media sites can help us fill in the gaps in clinical data and electronic health records which can help assess risk with chronic diseases.
The growing demand for data may be overwhelming when it comes to data collection, but analyzing data gaps can potentially improve the quality of care while keeping the costs low. Apart from structured data, unstructured data such as clinical condition notes is significant in healthcare. I hope that the healthcare industry will integrate the current fragmented process of data collection and analysis ensuring a thorough understanding of risk causes, thereby leading to improved healthcare quality without raising out of pocket expenditures for patients.