Modern BI Data Analytics for the Healthcare Industry

"The efficiency of commercial medicine is achieved through analytics—comprehensive, timely, and integrated analytics. While not a sufficient condition, it is an essential one. Without other factors such as experienced medical staff, modern treatment tools and methods, a healthcare institution loses its essence. However, it is analytics and the metrics obtained that allow balancing between the conflicting goals of commerce and medicine.

What is modern data analytics?

From our experience, we see at least 3 generations in the development of analytics within an organization.

The first generation involves extracting data from databases and generating reports in tabular form (Excel or Google Sheets). Subsequently, an analysis of the obtained data takes place in a semi-manual mode. Often, each department creates its own summary reports. For example:

Marketing focuses on the number of leads and their cost.
Sales or call center departments focus on the number of incoming calls and their effectiveness.
The finance department collects VAT and management reporting.
The medical department looks at basic metrics such as the number of patients, treatment duration, and medical staff availability.

Next is a regular review, including a breakdown of the metrics received by top management with each department. However, information from other departments is usually not fully communicated.

So, what are the downsides?

Decision-making speed is at the level of several weeks.
Data is not integrated with other departments (data silos).
Costs for data collection and analysis for each department manager responsible for analysis (on average, one day a week per manager).
Comprehensive (cross-functional) metrics using multiple data sources are rarely applied."

The second and third generations are based on the creation of a unified data repository. This can be referred to as a data lake for large organizations where it is impractical to link all data into a single data model (hence the need for one virtual repository for databases, collected models, and tables). If there is one data model stored in a single, preferably columnar, database, it is more appropriate to speak of a data warehouse.

But this is all semantics. The essence is that it is possible to create one place for collecting data needed for analytics."

How to Create a Data Warehouse?

It is necessary to set up regular extraction, transformation, and data merging in a separate database. After that, analysis is conducted using a BI tool for visualization and calculations. The analyst can interact with the system and create analyses on demand, but the data model is already provided. It is also possible to create a data warehouse and separate data marts for each function. The main data model is assembled, but new tables can be added for even deeper analysis, testing not only KPIs and cross-functional analysis but also Data Science methods.

Having a unified storage location where data is interconnected, you can begin analyzing and obtaining metrics.

For instance, suppose the directory of all patients is linked to the actual table from the Medical Information System (MIS) for services provided and the actual payment table from 1C. Now you can calculate accounts receivable at the patient level and its terms.

What metrics become possible?

Looking at it from the perspective of data sources, we can identify three types of metrics:

Metrics entirely based on one data source. For example, the sum of services provided, which can be taken from the Medical Information System (MIS).
Metrics that require data from multiple sources. These can be called cross-functional analytics metrics.
Metrics based on creating new information within the dataset.

Suppose we found clusters using some static method and then began tracking their proportions—this is already the third type of metrics. It can also be expressed in indexes or probabilities. For example, the customer churn risk index or the probability of revenue decline.

If the third type of metrics is related to Data Science and the first is more basic analytics, then the second type is something in between Data Science and vanilla analytics, but with a greater emphasis on data engineering. After all, you cannot find metrics from different sources without connecting them to the data warehouse. Also, it is with the third type of metrics that the third generation of data analysis development in the company begins.

Cross-functional analytics (from the perspective of a data analyst) involves combining data from two different sources to uncover new information and insights.

Let's explore the second type of analytics (cross-functional) based on our reports, as it is precisely this type that can answer various business questions in the field of medicine that were previously inaccessible.

Starting with more pragmatic questions:

How does the cost of customer acquisition through channels correlate with their Lifetime Value?
How does current revenue relate to departmental and daily plans?
Which keywords and advertising campaigns generate the highest percentage of target calls?
What is the current accounts receivable from customers?

Moving on to more strategic inquiries:

Which demographic groups are most valuable for the business?
In which medical specialties and diagnoses do customers trust us the most?
Which department and which doctor provide the highest amount of medical care?
Which services are most in demand based on types of medical history?
How does the staffing of medical personnel correlate with the dynamics of provided services per patient?
What average feedback do patients leave after treatment from different types of medical histories, demographic groups, or after being observed by different doctors?
Or how does their pain rating on a 10-point scale change before and after treatment?

At the bottom are examples of possible outcomes. Here, you find metrics of both the first and second types

‍

The 'Annual Report' dashboard from QompaX displays total and service-specific revenue, the number of calls, hospitalizations, their average cost and duration, the number of services per hospitalization, the number of patients, and services per patient, as well as bed occupancy compared to the previous year

‍

The QompaX report provides a gender-age analysis based on revenue and the number of hospitalizations, as well as differences (absolute values): hospitalizations and the number of services per patient, revenue per service, bed occupancy, services per hospitalization, and its duration.

‍

The 'Average Patient Portrait' dashboard from QompaX provides information on the average bill, the average cost of hospitalization, the average number of visits, and the average duration of the patient lifecycle for top specialties and age groups among women and men.

‍

This report from QompaX provides an overview of monthly/yearly revenue by specialties and service groups, the number of hospitalizations per month/year, revenue from hospitalizations per month/year, broken down by types of hospitalizations (initial/repeat), and the duration of completed hospitalizations per month/year.

‍

Understanding that there is quite a lot of information for analysis within data sources, including the Medical Information System (MIS), is crucial.

The reports presented above are based on randomized data. However, the implementation of such analytics cases has been successfully conducted by us in both commercial networks of medical clinics and commercial inpatient facilities. Based on these cases, we optimized our methods for creating data warehouses, data models, and subsequent calculations and visualizations specifically for managing a medical holding. The key point is that now the speed is measured not in dozens of months with significant accompanying budgets but in a couple of months.

Speed and experience are particularly important in uncertain times when protracted projects with long payback periods become less attractive for using financial resources. Among all possible investment expenditures, projects for several months with a clear and visible return on investment, significantly exceeding the investments, should be the priority. You can read more about why a data lake is like an asset with growing value on our website.

We have compiled several pages with metrics on our website that, in our view, can be analyzed in any efficient medium to large-sized medical institution – Metric Playbook.

An accompanying positive side effect is the gradual improvement in data quality due to their constant replenishment and verification.

It is important to remember that analytics is not the only necessary condition, and it is precisely with its advent that processes and methods of work in the company should simultaneously change...