Digital transformation is sweeping across all industries, including healthcare and insurance. One area where digital transformation can have a significant impact is in predicting member churn. Member churn, also known as member attrition, is a problem for healthcare insurance companies because it can lead to revenue loss, reduced customer satisfaction, and increased marketing expenses. In this blog post, we’ll explore how digital transformation can be used to predict member churn in healthcare insurance.
Understanding Member Churn in Healthcare Insurance
Before we dive into how digital transformation can be used to predict member churn, let’s first understand what member churn is and why it is a problem for healthcare insurance companies. Member churn is when a healthcare insurance company loses members, either because they switch to another insurer or because they decide to go without insurance. Member churn is a problem for healthcare insurance companies for several reasons.
First, member churn leads to revenue loss. When members leave, they take their premium payments with them, reducing the insurer’s revenue. Second, member churn can lead to reduced customer satisfaction. When members leave, they may do so because they are dissatisfied with the service or coverage provided by the insurer. Finally, member churn can lead to increased marketing expenses as insurers work to attract new members to replace those who have left.
Predicting Member Churn with Digital Transformation
Now that we understand why member churn is a problem for healthcare insurance companies, let’s explore how digital transformation can be used to predict member churn. Digital transformation refers to the use of digital technologies to transform business processes and operations. In the context of healthcare insurance, digital transformation can be used to analyze large amounts of data to identify patterns and trends that can help predict member churn.
The first step in using digital transformation to predict member churn is to gather data. Healthcare insurance companies have access to a wealth of data about their members, including demographic information, medical history, claims data, and more. This data can be used to create a profile of each member, which can be used to identify potential churn risks. Once the data has been collected, it can be analyzed using machine learning algorithms. Machine learning is a type of artificial intelligence that allows computers to learn from data without being explicitly programmed. By analyzing historical data, machine learning algorithms can identify patterns and trends that can help predict member churn.
There are several factors that can be used to predict member churn, including:
- Claims history – Members who have a history of making frequent claims may be more likely to churn.
- Demographic information – Members who are younger or have a lower income may be more likely to churn.
- Health status – Members with chronic conditions may be more likely to churn.
- Engagement – Members who are less engaged with their healthcare insurance company may be more likely to churn.
By analyzing these factors, healthcare insurance companies can create a churn risk score for each member. This score can be used to identify members who are at risk of churning and take proactive steps to retain them.
Taking Action to Prevent Churn
Once healthcare insurance companies have identified members who are at risk of churning, they can take proactive steps to prevent it. This can include:
- Personalized outreach – Healthcare insurance companies can reach out to at-risk members with personalized messages that address their specific concerns and needs.
- Improved communication – Healthcare insurance companies can improve communication with at-risk members by providing them with more information about their coverage and benefits.
- Incentives – Healthcare insurance companies can offer incentives, such as discounts or rewards, to at-risk members who remain with the company.
- Improved service – Healthcare insurance companies can improve their service to at-risk members by providing them with better access to healthcare providers or improving their claims processing times.
The Challenges of Customer Churn
Wovenware’s data science team recently began working with a major healthcare provider so we needed to take a closer look at machine learning in healthcare. We wanted to help our customer better predict customer churn and more proactively prevent it. Companies are committed to keeping customer churn as low as possible because the cost of acquiring new customers is actually higher than the cost to retain existing customers. They realize that any improvement in customer churn has a big impact on revenue.
Challenges to Addressing Customer Churn
Our healthcare client has a few peculiarities that make it a challenge to keep customer churn in check. Its customers can choose to change their service provider at any time, but it is notified at the end of the month when it’s too late for any remedial action. This limits its ability to identify the customer’s reason for leaving. In addition to that, the nature of the business also limits the value of the data related to customer behavior. Consider this, if a customer is using his health insurance does it means that he’s happy with the service or that he is just sick? On the other hand , what about a customer that hardly ever uses his health insurance, does it means that he is unhappy with the service or that he is just healthy?
Our strategy to leverage machine learning in healthcare to address our client’s customer churn, given the limitations mentioned above, was to build a predictive deep learning model to help it know which customers were at a higher risk of canceling their subscription in the upcoming month. Data that helped build the model included existing customer demographic data and health insurance claims, such as dollar amounts and type of claims. The resulting live predictions would give the provider enough time to contact the high-risk clients and address any need they have before they cancel their membership.
How Did We Address It?
So how did we accomplish this? First, we processed the claims data because it consisted of millions of data points with multiple entries, per day, for each customer. We also consolidated the claims data of each customer by month (since this is the timeframe the client uses to measure customer churn). Then, we analyzed the consolidated data to find patterns that could help us identify valuable features to train the deep learning model. We compared the data points of customers that stopped using the service to data points of customers that continued using the service, and found that all demographic and claims data followed the same distribution, which was a roundabout way to find that we had no meaningful features to train a deep learning model.
Given this setback, we decided to engineer new features by performing arithmetic operations on other claims features, which turned out to be valuable. We also used Pearson Correlation Coefficients to determine the strength of the relationships between features and kept the features with the strongest relationships as the indicators of customer churn.
What we found is that the occurrence of a customer leaving is actually rare, and leads to an unbalanced dataset, which is a problem when training a deep learning model. A model trained with an imbalanced dataset could learn to correctly predict the prevalent case and perform poorly when presented with a rare case, which for us is the case of interest.
The architecture we employed used three fully connected layers, a single neuron at the output, and the sigmoid activation function. We optimized a binary cross entropy loss using a sigmoid output. A portion of the dataset was used for training and an another portion was used to test the trained model. The portion of the dataset used for testing is called the holdout set. It was especially important to handle the holdout set with care because we wanted to avoid statistical bias on our results.
In the second blog post of this three-part series, read about our approach to model validation with the holdout set.