Chapter 6 Conclusion
There are significant patterns in transactions between the pharmaceutical industry and primary care physicians in the United States.
The pharmaceutical industry spends the most money each year on payments for consulting fees, food and beverage, and compensation for services other than consulting. While consulting fees present as medically-related, the definition is extremely vague. Physicians can be invited to consult for a variety of reasons, and the payment category name is applied by the pharmaceutical industry. In conjunction with the other two dominant categories, this reflects a trend in payments for non-medical reasons.
Reviewing population density categories from extremely sparse to extremely dense, it follows that most payments and the highest average payments are going to physicians in the more dense ZIP codes. This is feasible, as a payment to a high-density physician allows for impact on more patients than a payment to a low-density physician.
The ZIP code-level analysis revealed large variations in payments over each year. The plots of log payments to all ZIP codes in the US, likewise, provided little insight into these granular trends. On the state-level, there was more continuity. Here, one could identify trends in payments to physicians in a given state over time. The state-level analysis revealed there may be more regional trends in payments, as payments to states in the Northeast seemed to decrease while payments to states in the West seemed to increase.
The regional analysis confirmed these findings. Likewise, the analysis revealed physicians in the South received the highest overall payments, which wasn’t identified at the state-level. Interestingly, there was a large jump in the average payment to physicians in the West in 2020.
Analyzing payments by quarterly aggregation revealed no high-level trends, as total payments were relatively consistent across quarters for each year. For average payments, some years yielded significant results, while others did not. As above, there was a large jump in the average payment to physicians during the third quarter of 2020.
Linear models explored how these patterns converged to predict log total payments for each study year. The models confirmed many of the findings above, with significant predictive power against the baseline for regions, density categories, and some quarters. Against the baseline of a physician who specializes in Family Medicine, all other specialties were significantly predictive of increases in log total payments, except for physicians specializing in Internal Medicine: Adolescent Medicine.
Time-series modeling of average payments per quarter over each year of the study yielded a mean absolute percent error (MAPE) of 13.43. However, this MAPE value decreased to 11.96 when 2020 was dropped from the model. This indicates that payments to primary care physicians in 2020 may deviate from the trend established from 2015-2019.
6.1 Future Work
As highlighted by the time-series model, the 2020 data presents unique trends that often diverge from established norms. One potential reason for this could be the onset of the COVID-19 pandemic. Future analysis could explore specifically how the relationship between the pharmaceutical industry and primary care physicians changed in response to and during the COVID-19 pandemic.
Likewise, this analysis examined spatial relationships and temporal relationships relatively independently. In the future, it may be interesting to examine the ways these relationships influence each other, whether through spactial-temporal modeling or other techniques.