Forecasting in Pharmaceutical Industry – Part 2 (drugs with a limited supply of raw material)
This article was published as a part of the Data Science Blogathon
In my previous article, we learned about patient-level forecasting for new products. In this next article of the series, we will learn about the in-market forecast of those products which have a limited supply of raw material.
In-market products are those that are already present in the market. One simple method is to use time series forecasting on volume and patient number and calculate net forecasted sales. But there are certain drugs or therapies which have a limited supply of raw material and it may happen that demand exceeds the availability of the drug. In that case, we need a different method to calculate drug forecast.
We will take an example of PDT i.e plasma-derived therapy market in which the raw material is human plasma and hence it is limited in supply. In this case what should our approach while forecasting?
We will answer this in our article.
What is Plasma?
We can refer to the below figure and conclude that if we remove RBC, WBC, and platelets from the blood then the remaining portion is called plasma. It has multiple clinical uses and required in various therapies. It cannot be produced in the laboratory.
Multiple products can be derived from plasma and can be used for various therapies.
The below image is taken from Takeda. If you follow the embedded link you will get to know various PDT products across different countries.
Steps to Forecast
We will learn the forecast of limited raw material supply drugs in the below steps.
Calculation of Patient Number
We will start with the epidemiology section and calculate the number of total treated patients using prevalence rate, diagnosis rate, and treatment rate. In case you are not familiar with these terms, you can refer to my previous article to know more.
Total Treated Patients = Total Patients*Diagnosis rate*Treatment rate
In the below screenshot I have done calculations taking some dummy data for the total treated patients.
Conversion of patients into volume
Once we find out the total number of patients then we will check the annual utilization in grams per patient. Using annual utilization and total treated patients we can calculate the total required volume.
Again for simplicity, I have not included free goods or stocking here. Also, I have excluded parallel trades in this exercise and assume that we have a 100% market share.
Now total market share is obtained by calculating total patients on our drug divided by the total treated patients.
Total Volume(in grams) = Annual Utilization (gram/patient) * Total market share
Till now the calculation is the same as we have performed for any other drug. However, an important point in this forecasting is to track how much volume is allocated because here forecasting will not be driven by demand.
A simple example of this is the corona vaccines forecast.
Companies know that entire India needs to be vaccinated and they know the population demographics as well but they are not able to produce enough vaccines as the raw material is limited(though it is for a short period of time). Hence they might not be forecasting the vaccines based on demand but rather on supply of raw material.
Once we calculated allocated volumes and total volume we will convert the same in Gross Sales.
Conversion of volume into sales
We can consider two different types of sales here:
- Tender market – Drugs are procured via tendering
- Non-tender market – Drugs are not procured via tendering. It is called the prescription market as well.
However, it is only done when we are informed about the different types of sales as clients keep an eye on both the market type. In case there is no provision of tender and non-tender market i.e one of them is 0% then we can simply convert volume into sales by multiplying the list price of drug with volume.
Gross Sales = If allocated volume<total volume then allocated volume*list price else total volume*list price
Market Share Calculation
The market share calculation approach is similar to what we have discussed in the previous article. Usually, the normal user entry share method is used in this type of forecast where business users input desired market share based on business data using linear interpolation.
Refer to the below screenshot where the simple user share method is depicted. Let us assume that our product is Prod A and competitor product is Prod B and Prod C. Till FY 2021 we have market data available. Now we need to forecast market share from FY 2022 to FY 2025. Business users need to enter FY 2022 market share based on underlying calculated data and also for FY 2025 using share projection. Now for FY 2023 and FY 2024, we will use linear interpolation to calculate market share.
In a complex model, we use market share to calculate volume.
FY 2023 market share = [market share (FY 2025)- market share(FY 2022)]/3+market share (FY 2022)
FY 2024 market share = [market share (FY 2025)- market share(FY 2022)]/3+market share (FY 2023)
- Normal case and High case forecast
A good practice to follow is to create a normal case and high case forecasts. There isn’t much difference between normal case and high case except the high case is an optimistic version of forecast where the company believes that they would be able to procure more allocated volume or demand volume would be high. Clients love to compare both normal cases and high cases.
One can visualize the numbers for both normal case and high case forecast to impress the clients. 🙂
We discussed how we can forecast those drugs which have a limited supply of raw materials. The method which I have shared here is widely used across the pharmaceutical industry. In the next part of this series, we will learn about methodologies of in-market forecasts with synthesized data.
Hope you found this article informative!
Himanshu is a data science consultant helping business stakeholders across various domains with actionable data insights and enabling clients’ processes with ML models. You can connect with me on LinkedIn.
The media shown in this article are not owned by Analytics Vidhya and are used at the Author’s discretion.