“Complex pharmacy transactions, especially for patients who have to fill multiple medications periodically, is a challenge.”- Reni Thomas
Reni Thomas, Vice President- Analytics Partner, Strategic Healthcare and Lifesciences Clients at Genpact spoke at the third edition of the Machine Learning Developers Summit (MLDS 2021).
In her talk titled, “Transforming Patient Pharmacy Experience”, Thomas covered topics, such as how healthcare business problem can be turned into a statistical problem, the challenges patients who take multiple medications face, how to improve patient pharmacy experience and roles data and analytics play.
Thomas explained consequences of a bad pharmacy experience:
- Patients don’t take their medication (Non-adherence to medication): Complex pharmacy transactions and bad pharmacy experience, especially for patients who have to fill multiple medications periodically, could result in non-adherence to medication. This typically happens to those patients who take more than three medications in a day.
- Increase in adverse events/ complications and hence the overall cost of care: There is a huge fallout in prescriptions taken versus prescriptions written. Hence, there is a strong correlation between positive patient experience and better health outcomes.
While discussing the patient pharmacy experience, the speaker mentioned some of the points which can help in improving the patient pharmacy experience;
- Home delivery of medication
- Coordinating prescriptions
- Automatic refill of the prescription
- 90-day refills
According to Thomas, behavioral economics plays an important role in healthcare analytics. The analytical process includes three major parts: understanding the drivers of home delivery adoption; defining the target members; and measuring the uptake in home-delivery and thereby experience.
Thomas gave an example of an AI model that can help in improving the patient pharmacy experience. For training and testing the models, various data sources are being considered in healthcare, including medical claims data, provider data, past interventions and contact data member plan and enrollment data, among others.
Next, the speaker explained the approach of the AI model in healthcare analytics::
- Data Preparation: Defining population, data wrangling using internal and external datasets.
- Exploratory Analysis: This helps in understanding the patient profiles with respect to age, demographics, socio-economic condition, mail vs retail, out of pocket with regards to pharmacy utilisation.
- Predictive Model: This serves to identify active patients likely to convert to online pharmacy. Classification models like XGBoost, Random Forest etc works well in healthcare analytics.
- Opportunity vs Engagement: Incorporate patients with highest opportunity with the results from those with the highest likelihood to engage with the online pharmacy.
- Targeting: Defining patient profiles for targeting.
Thomas concluded the session by discussing key KPIs and hypotheses in healthcare analytics along with the impacts and steps to increase medication adherence.
The key KPIs and hypotheses include patient demographic factors, patient’s medical condition, financial factors and patient engagement metrics. The key metrics for medication adherence are-
- Improve overall patient experience
- Overall MPR improvement
- Reduction in hospital admissions and ER visits
- The decrease in the cost of care.