Areas of Improvement
The model does not have a high tolerance for sudden changes to the existing system. If there were to be a sudden increase in the stress levels of students or if there were a crisis that occurred that severely affected the mental health of students, the system does not have the capability to handle that. The trade-off here is that the model has high fidelity for the given considerations and predicted barriers. If there are changes in those barriers, the model can easily accommodate the new conditions and predict new attendance numbers.
In the future, analysis by time rather than just location would be useful in scheduling Let’s Talk sessions (i.e. what attendance would look like at 9am, at 10am, at 11am, etc). Further consideration of the workload of students and what their exam schedule is would allow users to create a week by week estimation of utilization. This could account for variations in weekly attendance at a location and would be a better predictor of utilization than an average. It might also be useful to perform a cost-benefit analysis with Mental Health Services. We were most focused on improving attendance to Let’s Talk and understanding how different, controllable factors would either increase or decrease attendance to Let’s Talk. However, in our analysis we did not consider the cost of adding another counselor or another location to Let’s Talk. The allocation of an additional counselor to a Let’s Talk location could be potentially harmful and unideal for MHS because of the loss of appointment times for that additional counselor.
It would also be useful if we had added a section on our survey that discussed race. We believe that this is a metric that is of paramount importance when discussing mental health because we believe that the allocation of resources should not cut out any specific population. For example, in our analysis the Center for Diversity and Inclusion had one of the lowest attendance rates. While this could be due to a number of factors, we are loath to recommend that this location be nixed simply because almost all of the minorities who attended Let’s Talk attended this specific location. However, we are lucky in that the CDI is moving locations to the DUC which, by our predictions, would be the most attended location on campus. This means that we would be able to keep a location that serves racial minorities while also increasing its usage by the greater WUSTL population.
Another interesting angle to consider would be using Matlab as a tool to code this simulation rather than Simscript. Although Simscript worked well for our purposes, it is a program that is quite outdated and has a lot of bugs. Matlab has the capacity to encode this program and store the results in matrices much more effectively than Simscript does. Matlab’s random number generator also works much better for the purposes of our project than Simscript does. If we would have used Matlab for this project, we might have been able to do a Monte Carlo simulation by using multiple different iterations of random numbers. This would have given us more confidence in the fidelity of our model and the results that it was outputting, if they were consistent. With Simscript, since the random number generator creates the same “random” numbers each time we are not able to run a Monte Carlo simulation.