One More Time About a Synergy of AI and OR/MS
I have recently attended a webinar organized by the NY Metro INFORMS Chapter on "Synergizing Artificial Intelligence and Operations Research: Perspectives from INFORMS Fellows on the Next Frontier." https://us06web.zoom.us/webinar/register/WN_HI7YOf3YSaGPYyuC5w6TzQ#/registration
The panelists aggressively promoted findings from their recent INFORMS Journal on Data Science paper
Synergizing Artificial Intelligence and Operations Research: Perspectives from INFORMS Fellows on the Next Frontier
This paper revisits Herbert Simon's 1987 vision of a partnership between artificial intelligence (AI) and operations research/management science (OR/MS) to enhance decision-making.
Two Heads Are Better than One: The Collaboration between AI and OR
I have mixed feelings about this webinar and its conclusions.
Therefore, I decided to test myself what value could be added by AI to OR/MS.
I used, as a test-bed, my consulting work that I had performed some time ago. This example is instructive because it illustrates a typical request for analysis written in plain language with some unstructured numbers by a healthcare manager who is an analytic layman. This is a rather typical example of a problem in a healthcare setting that requires, first, some interpretation and making some reasonable assumptions to perform the analysis using OR/MS.
I used three popular AI platforms (OpenAI GPT-4o Mini by OpenAI; Grok 3 Mini by xAI; and Free ChatGPT by OpenAI). I used, as input, the same consulting request text provided in the attached document. The AI outcomes and recommendations are also presented in the attached document.
The bottom line:
· In my opinion, the presenters of the webinar failed to present a business case or a specific example that would clearly demonstrate the AI value added to OR/MS approach.
· Simon's vision, " we should aspire to increase the impact of MS/OR by incorporating the AI kit of tools that can be applied to ill-structured non-quantitative decision domains, " was and still remains a big beautiful dream.
· There were no convincing demonstrations of the value added or additional insights to OR/MS methodology by any of the three tested AI platforms. In fact, the AI recommendations were to use OR/MS methods, such as Excel spreadsheets or simulation modeling. Such recommendations were trivial and generic.
This is the original consulting request:
"Center Manager-Wockhardt Eye Hospital. Location: Aurangabad, Maharashtra, India.
Dear Mr.Kolker, let me discuss my problem in detail.
I work in an ophthalmology hospital. The ideal situation I wish is: we treat 120 patients a day with 5 doctors from 9 am to 1 pm, 2 doctors would see a preliminary examination of the eye and dilate the patient for 30 minutes, whereas 3 doctors would start seeing the final examination after the dilation and will dispose the patient. We promise a delivery of 2 hours to 80% of patients who come for watering & irritation, normal checkups, headaches, glasses, and redness. The rest 20% would have cataract, glaucoma, and other complex disease which is bound to take more than 2 hours of investigation. We can monitor the waiting time of more than 2 hours in each clinic. But this 'ideal' situation hardly happens because one doctor would be in operation theatre, one doctor would be on leave, or in ward rounds or in camp, so I am left with only 3 doctors in the unit from 9 am to 1 pm.
With 3 doctors, the bulk of patients arrive at 10 am (40 patients/hour) all these patients have to go for vision examination by nurses (5 min each patient) and 3 nurses would do it.
The challenges I have observed are:
* arrival of complicated cases, where a thorough examination of the posterior segment of the eye is needed (could take up to 30 min/ patient)
* patients who find it hard to place their chin in front of the slit lamp or language barrier, more inquisitive patients who keep loading the doctors with questions
* patients who would go for nature's call or coffee or to the lab for blood tests and will arrive to see their successors in the doctor's consultation room before them.
* trauma case that needs complete attention before shifting the patient to OT
* doctors who are purposefully slow to make sure that they see fewer patients than their colleagues
* missing case sheets by the nurse who has to spend all her time searching the case sheets.
Now, how would I monitor the patient waiting time in a clinic of 120 patients with all these parameters, to make sure patients are seen as early as possible rather than shouting, and arguing, thus disturbing the entire atmosphere of a peaceful work environment?
Can I work this situation with all the challenges in an Excel sheet or simulation software? How should I go about it? Kindly assist me."
OR/MS perspective
(see section 3.6 in the book "Healthcare Management Engineering in Action", Springer, 2024.
https://link.springer.com/chapter/10.1007/978-3-031-53663-2_3 )
This problem description should first be converted into a more specific and formal description that would make it amenable to quantitative OR/MS analysis.
Below is my own problem summary that I made without any AI input:
- clinic is open 4 hours per day, from 9 am to 1 pm. The daily patient volume is 120.
- 80% of patients are regular ones (normal checkups, minor irritation, etc.). They are all first preliminary examined, then dilated, then finally examined and discharged.
- 20% of patients are complicated (cataract, glaucoma, etc.). They are first preliminary examined, and then proceed for final examination by a doctor skipping dilation.
- Preliminary examination takes 5 min, with the range 4-6 min.
- Dilation takes to work about 30 min with a range of 20 - 35 min.
- Dilated patients wait in a separate area until they are seen by the next available doctor for final examination.
-Final examination by the doctor for regular patients takes about 20 min with a range from 15 min to 30 min.
- Final examination by the doctor for complicated patients is in the range from 30 min (posterior segment of eye) to 120 min (2 hrs) (cataract, glaucoma) with the most likely being 1 hour (65 min).
- Staffing for base (ideal case) model: 2 doctors for preliminary examination and dilation, and 3 doctors for final examination. Total doctors' staffing: 5. No nurses.
- Staffing for the actual case is 3 nurses for preliminary examination and dilation and 3 doctors for final examination.
Operations performance metrics target: 80% or more of regular patients wait less than 2 hours. Because of a lack of more specific information "the bulk of patients arriving at 10 am…" was interpreted as some patients (20) arriving from 9 to 10 am, while the remaining 100 (bulk) arrive at 10 am (actually spread from 10 am to 12:30 pm).
The question in the problem description document was "How would I monitor the patient waiting time in a clinic of 120 patients with all these parameters to make sure patients are seen as early as possible?"
From the OR/MS standpoint, the question should have been interpreted as 'Given patient arrival pattern uncertainty, what staffing would be needed to meet the target operational performance?'
OR/MS Solution
Obviously, an Excel spreadsheet would not be sufficient for analyzing such a dynamic problem. Discrete event simulation modeling is the methodology that should be used to determine the optimal staffing of doctors that meets the operational performance criteria. The first model would be a baseline one (an ideal case) that includes two preliminary exam doctors and three final exam doctors. Again, I suggested this approach without any AI help. As an OR/MS expert, I did not need any help. Details of a discrete event simulation model and simulation results were presented in section 3.6 in the book "Healthcare Management Engineering in Action", Springer, 2024.
https://link.springer.com/chapter/10.1007/978-3-031-53663-2_3
The main take-away was that the number of doctors (not nurses) was the bottleneck. Adding full and part-time doctors will elevate this bottleneck. The model suggested the required number of full- and part-time doctors.
It was possible to simulate many other scenarios using this type of simulation modeling to find out the most satisfactory staffing level and shifts for meeting the operational performance targets. The following simulation scenarios were suggested: (i) the effect of additional part-time exam doctors scheduled for different shifts with various availability; (ii) restriction of the total daily patient volume; (iii) changing the percentage of regular and complicated patients; (iv) various patient arrival patterns; (v) extension of clinic hours or allowing overtime; (vi) or combinations of the above.
AI perspective.
I used the above plain language description as an input into several popular open AI platforms to test what value AI can add to OR/MS approach or improve or enhance it in any way.
The following AI platforms have been used:
OpenAI GPT-4o Mini By OpenAI
GPT-4o Mini is OpenAI's most advanced model in the small models category. It is multimodal (accepting text or image inputs and outputting text), and has higher intelligence than gpt-3.5-turbo but is just as fast. It is meant to be used for smaller tasks, including vision tasks.
The AI output was:
"Optimizing patient flow and monitoring waiting times in a busy ophthalmology clinic, particularly with the challenges you've described, requires a structured approach. Utilizing tools like Excel for tracking and managing data or simulation software for modeling can be effective strategies. Here's a comprehensive outline to help you navigate this issue:
1. Data Collection and Monitoring Setup:
Using Excel:
- Patient Tracking Sheet:
Create a log where you input patient arrival times, initial examination start, dilation times, and final exam times.
- Use columns for patient ID, arrival time, start time for prelim exams, dilation duration, start time for final exams, and total waiting time.
- Categorization:
- Separate patients based on their complexity (routine vs. complicated cases) in different sheets/tabs.
- Assign a complexity code (e.g., 1 for simple and 2 for complex) to each patient for sorting and analysis.
- Nurse and Doctor Schedules:
create a schedule that indicates the availability of each doctor and nurse. Make a note of any planned leaves or availability that may affect patient processing.
- Incorporating Break Times/Disruptions:
- Maintain a log for each doctor to note any disruptions (emergency cases, breaks, etc.) to identify patterns affecting wait times.
2. Data Analysis:
- Calculate Waiting Times:
- Analyze the sheet daily or weekly to calculate the average waiting time, broken down by patient complexity.
- Identify instances where patients exceed the 2-hour waiting window.
- Bottleneck Analysis:
- Identify common bottlenecks (e.g., wait time at vision examination, delays due to complications) and their causes.
- Performance Tracking:
- Track the performance of each doctor and each shift. Analyze if some doctors are consistently slower than their colleagues and identify reasons.
3. Optimizing Patient Flow:
Suggestion Batching:
- Establish a triage system upon patient arrival:
- *Fast-Track Routine Patients:* Identify lower complexity patients who can quickly proceed with exams.
- *Complex Cases Protocol:* Set aside specific days or sessions for complex cases to reduce their impact during peak hours.
Improvement Measures:
- Patient Education:
- Have educational materials ready for patients (pamphlets, videos) to reduce questions and prepare them for examinations.
- Checking Mechanism:
- Ensure nurses check all case sheets and necessary forms are ready before patients are called to see doctors.
- Queue Management:
- Use a digital queue management system (or even a physical digital board) to manage patient queues.
4. Simulation Software:
Using simulation software can help model various scenarios, predict patient flow, and identify optimal staffing for different times.
1. Creating a Simulation Model:
- Input your clinic's workflow, including patient arrival rates, doctor availability, exam times, etc.
- Simulate different scenarios (varying the number of doctors, adding a patient education session, etc.) to see impacts on waiting times and patient satisfaction.
Grok 3 Mini
The latest fast model from xAI. A lightweight model that thinks before responding. Fast, smart, and great for logic-based tasks that do not require deep domain knowledge.
The output was:
"I understand your situation in the ophthalmology hospital-it sounds frustrating to manage patient flow with so many variables and disruptions, especially when you're short-staffed and dealing with unpredictable challenges like complicated cases, patient behaviors, and administrative issues. The good news is that you can absolutely monitor patient waiting times effectively, even with all these parameters, by using tools like Microsoft Excel for basic tracking and analysis, or simulation software for more advanced modeling. This will help you identify bottlenecks, optimize workflows, and maintain a peaceful environment without resorting to shouting or arguments.
I'll break this down step by step: first, I'll explain how to approach the monitoring process, then discuss using Excel as a starting point, and finally, suggest when and how to use simulation software. My goal is to help you create a system that's practical, data-driven, and scalable.
Step 1: Define Your Monitoring Goals and Process
Before diving into tools, let's outline a clear plan. You want to track waiting times for 120 patients per day (from 9 am to 1 pm) while accounting for the challenges you mentioned. Here's how to structure this:
- Key Metrics to Track:
- Arrival Time: When the patient arrives (e.g., many at 10 am in bulk).
- Preliminary Exam Time: Time for vision exam by nurses (5 minutes each) and dilation (30 minutes total per patient).
- Final Exam Time: Time with doctors for simple cases (aim for under 2 hours total) vs. complex cases (over 2 hours).
- Waiting Time: Total time from arrival to start of preliminary exam, plus time from preliminary to final exam.
- Total Turnaround Time: From arrival to departure, including any delays.
- Delays and Challenges: Note reasons for delays (e.g., complicated cases, patient interruptions, missing case sheets, doctor absences).
- Bottlenecks: Track how factors like doctor availability (e.g., only 3 doctors), nurse workload, or interruptions affect flow.
- Target Outcomes:
- Ensure 80% of simple cases (watering & irritation, checkups, etc.) are seen within 2 hours.
- Monitor the 20% complex cases (e.g., cataract, glaucoma) separately, as they exceed 2 hours.
- Use data to forecast staffing needs, reduce disruptions, and improve patient satisfaction.
- Data Collection Process:
- Who Collects Data? Assign a dedicated staff member (e.g., a nurse or admin) to log data in real-time using a simple form or app. This could be done on paper initially and transferred to Excel, or directly into a digital tool.
- How Often? Track data for every patient daily, then analyze it weekly or monthly to spot patterns.
- Handle Challenges in Tracking:
- For complicated cases (e.g., 30+ min exams), add a field to note the extra time.
- For disruptions (e.g., patient delays for coffee or trauma cases), log them as "interruption events" with timestamps.
- To address doctor variability (e.g., slow doctors), track individual doctor consultation times anonymously to identify trends without blame.
- For missing case sheets, integrate a quick check-in system to ensure files are ready.
By focusing on these, you'll create a "feedback loop": collect data, analyze it, and use insights to adjust schedules (e.g., stagger patient arrivals or allocate more time for complex cases).
Step 2: Using Excel for Monitoring
Excel is an excellent, accessible tool to start with-it's free (if you have Microsoft Office), easy to learn, and powerful enough to handle your scenario without needing advanced software. You can use it to log data, calculate wait times, and visualize trends. Here's how to set it up:
A. Set Up Your Excel Spreadsheet
Create a template like this (I'll describe it; you can replicate it in Excel):
1. Columns for Patient Data:
- Patient ID: A unique number for each patient (e.g., auto-generated).
- Arrival Time: Timestamp (e.g., 10:00 AM).
- Patient Type: Categorize as "Simple (80%)" or "Complex (20%)" based on initial assessment.
- Preliminary Exam Start Time: When the nurse starts (e.g., after vision check and dilation).
- Preliminary Exam End Time: Approximately 30 minutes later.
- Final Exam Start Time: When the doctor begins.
- Final Exam End Time: When the doctor finishes.
- Departure Time: When the patient leaves.
- Delay Reasons: A dropdown or text field for notes (e.g., "Complicated case - posterior exam," "Patient interruption," "Doctor in OT," "Missing case sheet").
- Total Wait Time: Calculated field (e.g., =Final Exam Start Time - Arrival Time).
- Total Turnaround Time: Calculated field (e.g., =Departure Time - Arrival Time).
Example of a basic row:
| Patient ID | Arrival Time | Patient Type | Prelim Start | Prelim End | Final Start | Final End | Departure Time | Delay Reasons | Total Wait Time | Total Turnaround |
|------------|--------------|--------------|--------------|------------|-------------|-----------|----------------|---------------|-----------------|------------------|
| 001 | 10:00 AM | Simple | 10:05 AM | 10:35 AM | 11:00 AM | 11:30 AM | 11:45 AM | None | 1 hour | 1 hour 45 min |
2. Formulas for Calculations:
- Use Excel's time functions to automate calculations:
- Total Wait Time: =TEXT(Final Start - Arrival Time, "h:mm") (subtracts times and formats as hours:minutes).
- Total Turnaround Time: =TEXT(Departure Time - Arrival Time, "h:mm").
- Conditional Formatting: Highlight cells where Total Turnaround Time > 2 hours in red, so you can quickly spot issues.
- Averages: Add a summary section at the bottom with formulas like =AVERAGE(Total Wait Time column) to calculate daily averages.
3. Incorporate Challenges into the Sheet:
- Variable Patient Arrival: Use a separate tab to log arrival patterns (e.g., 40 patients at 10 am). Then, use graphs to show peaks.
- Complicated Cases: Add a column for "Exam Duration Override" where you manually enter extra time (e.g., +30 min for posterior exams).
- Interruptions: Create a dropdown list in the Delay Reasons column (e.g., options: "Patient Delay," "Trauma Case," "Doctor Absence"). This helps quantify how often these occur.
- Doctor and Nurse Workload: Track which doctor/nurse handled each patient and how long it took. Use pivot tables to analyze if one doctor is consistently slower.
- Bulk Data Entry: If you have many patients, use Excel's data validation for quick entry (e.g., dropdowns for Patient Type).
B. Daily Workflow with Excel
- Morning Prep: Open the Excel file at 9 am. Have staff enter data as patients are processed.
- Real-Time Monitoring: Use a shared Excel file (via OneDrive or Google Sheets for collaboration) so you can check wait times on the fly. If a patient is approaching 2 hours, alert the team.
- End-of-Day Analysis: After 1 pm, review the data. For example:
- Pie chart for patient types (80% simple vs. 20% complex).
- Line graph for wait times over the day to see spikes at 10 am.
- This will help you answer questions like: "How many patients waited over 2 hours today, and why?"
C. Limitations and Tips for Excel
- Pros: It's simple, cost-effective, and you can start immediately.
- Cons: It doesn't handle randomness well (e.g., simulating "what if a doctor is absent?"), so it's better for tracking real data than predicting scenarios.
- Tips:
- Train your team on basic Excel use (e.g., 1-hour session).
- Back up data daily to avoid loss.
- If you have 120 patients, use Excel's filters and sorting to manage the volume.
With Excel, you can monitor your current situation effectively and start making improvements, like adjusting nurse assignments based on arrival peaks.
Step 3: When and How to Use Simulation Software
If Excel feels insufficient (e.g., you want to simulate "what if" scenarios with your challenges), move to simulation software. This is ideal for modeling complex systems like yours, where variables (e.g., doctor absences, patient interruptions) create uncertainty.
Why Use Simulation Software?
- It allows you to create a virtual model of your clinic, incorporating all your challenges to predict wait times and test solutions.
- For example, you could simulate: "If we have only 3 doctors and 40 patients arrive at 10 am, how does that affect wait times?" Or, "What if we add a buffer for complicated cases?"
Recommended Software:
- Free/Entry-Level Options:
- Excel's own simulation tools (e.g., using Monte Carlo add-ins) for simple scenarios.
- AnyLogic Personal Learning Edition (free for basic use) or Simul8 Community Edition-these are user-friendly for healthcare simulations.
- Paid Options: Arena or ExtendSim if your hospital can invest (they're more powerful but might require training).
How to Set It Up:
1. Choose Software: Start with AnyLogic-it's intuitive for beginners.
2. Model Your Process:
- Input Parameters: Define variables based on your description:
- Patient arrival: Bulk at 10 am (40/hour), with 80% simple and 20% complex.
- Resources: 3 doctors, 3 nurses; each preliminary exam takes 5-30 minutes; final exam varies (e.g., 30 minutes for simple, up to 60+ for complex).
- Challenges: Add random events, like a 10% chance of a trauma case interrupting, or a 20% chance of patient delays.
- Simulation Logic: Create a flowchart:
- Step 1: Patient arrives → Goes to nurse for preliminary exam.
- Step 2: If dilated, queues for doctor.
- Step 3: Doctor examines; add delays for interruptions.
- Output: Simulated wait times for 120 patients.
3. Run Scenarios:
- Baseline: Simulate your current setup.
- "What If?": Test changes, like hiring a temp doctor or staggering arrivals.
4. Analyze Results: The software will output reports, e.g., average wait times, probability of meeting the 2-hour promise.
5. Integration with Excel: Use Excel to import simulation results for further analysis.
Getting Started with Simulation:
- Learning Curve: If you're new, start with online tutorials (e.g., YouTube for AnyLogic). It might take a few days to build a basic model.
- Cost and Time: If your hospital can't afford this, stick with Excel for now and scale up later.
- When to Switch: If you notice patterns in Excel data that need deeper analysis (e.g., frequent delays from complicated cases), simulation will help.
Final Thoughts
You can definitely manage this with Excel as your primary tool-it's straightforward and will give you immediate insights into waiting times while accounting for your challenges. Start by creating the spreadsheet I outlined, collect data for a week, and review it to see improvements. If things get more complex, transition to simulation software for predictive modeling.
This approach will help you foster a more efficient, less chaotic environment by focusing on data rather than reactions. If you share more details (e.g., your current workflow or specific Excel questions), I can refine this further. Good luck-I'm here to assist!"
Free ChatGPT by OpenAI
Access Free ChatGPT instantly through our OpenAI API-powered interface. We've made ChatGPT's powerful AI capabilities freely available to everyone, with no registration required and no usage limits.
The output was :
"To effectively monitor patient waiting times in your ophthalmology clinic, especially given the complexities and challenges you've outlined, a systematic approach is important. While Excel can provide a relatively simple way to track and analyze data, a simulation software tool might be better suited for modeling the dynamic and complex nature of your clinic. Here's a step-by-step guide for both approaches:
Option 1: Using Excel
1. Data Collection:
- Gather data on patient arrivals, the type of cases, time spent on examinations, and outcomes. You will need to track:
- Arrival times
- Time taken by nurses for preliminary examination
- Time taken by doctors for final examination
- Types of cases (routine or complex)
- Any delays or disruptions (e.g., nature's call, language barriers)
2. Create a Tracking Template:
- Design an Excel sheet with columns such as:
- Patient ID
- Arrival Time
- Preliminary Examination Start Time
- Preliminary Examination End Time
- Doctor Consultation Start Time
- Doctor Consultation End Time
- Type of Case
- Total Waiting Time
- Notes (for delays or other disruptions)
3. Formulas:
- Use formulas to calculate waiting times. For instance:
- Total Waiting Time = (Doctor Consultation End Time - Arrival Time)
4. Data Analysis:
- Use pivot tables or graphs to analyze waiting times based on different parameters (e.g., types of cases, doctor efficiency).
- Highlight any trends, such as peak times leading to delays or specific issues causing longer waiting times.
5. Regular Reviews:
- Set a schedule to review the data regularly, and adjust workflows or resource allocation based on findings.
Option 2: Using Simulation Software
1. Choose a Simulation Tool:
- Select software that specializes in healthcare operations, such as AnyLogic, Simul8, or Arena, which can model patient flow.
2. Model Your Clinic:
- Create a model that includes:
- The flow of patients from arrival to consultation.
- Time distributions for examinations (e.g., 5 minutes for preliminary, 15-30 minutes for final depending on complexity).
- Random arrivals based on historical data.
- Include factors for delays, such as patients taking breaks, language barriers, and others.
3. Input Data:
- Input your collected data (arrival rates, processing times, etc.) into the simulation model.
4. Run Simulations:
- Simulate various scenarios to observe how changes in staffing levels, patient arrival patterns, and case complexities impact waiting times.
- Explore "what-if" scenarios, such as what happens when one doctor is unavailable or if more nurses are added for preliminary exams.
5. Output Analysis:
- Analyze the results from the simulation to identify bottlenecks and areas for improvement.
- Use the insights gained to inform changes in scheduling, staffing, and processes.
General Recommendations
1. Standard Operating Procedures (SOPs):
- Standardize processes for routine examinations and consultations to minimize variations caused by different doctors.
2. Patient Education:
- Provide clear instructions to patients before their appointment regarding what to expect, helping them to be prepared (e.g., time for blood tests or what to do with delays).
3. Communication:
- Implement a system for keeping patients informed about expected waiting times and managing their expectations.
4. Resource Allocation:
- Consider hiring additional staff during peak hours or cross-training existing staff to fill in as needed.
5. Feedback Mechanism:
- Create a system for collecting patient feedback on waiting times and experiences to continuously improve the process.
By utilizing either Excel for basic tracking or a more advanced simulation for in-depth analysis, you can gain valuable insights into your clinic's operations, ultimately leading to improved patient satisfaction and efficient use of resources."
------------------------------
Alexander Kolker
ge healthcare
MILWAUKEE WI
------------------------------