How to Improve the Patient Experience with AI?

Patient satisfaction surveys give you valuable insights into service and safety-related issues within your department. While some of these insights can be observed from the structured ratings in the surveys (see example in Figure 1), you get the full picture only by analyzing the free-text comments where patients provide a lot more detail on what they actually experienced.

Figure 1: Structured, overall physician ratings

Why did this physician receive only 4.8 stars?

If you were wondering why the physician from Figure 1 only received a 4.8/5 rating on several of the aspects (e.g. confidence, likelihood to recommend, concern), the best place to look for more context is within the free-text comments.

Here’s an actual snippet from the free-text comments for this physician that can give you clues as to what is happening:

…I think this provider is competent, but my confidence in the provider is not high because of the way he interacts with me. He doesn’t waste any energy on the common social graces or polite interactions, nor did he spend any time showing me he cared about my problem. If my provider doesn’t even try to pretend he cares about me or my issues, then I don’t care to go to her with my issues in the future….

From this one comment you can see that the patient is not completely satisfied because of the “interaction issues” which affects their confidence in the provider. Now, imagine if we performed a Text Analysis of ALL the survey comments, spanning several years. You will start seeing really interesting themes without having to read each and every comment.

What AI can reveal about the patient experience?

Below, we describe the types of insights you can get from a detailed analysis of patient comments leveraging machine learning and natural language processing technologies. These insights can provide context around the structured ratings from patient satisfaction surveys and help you improve your overall patient experience.

#1 Safety issues in your department

Comments from patients can often open your eyes to potential safety issues happening in your department. For example, if Emergency Room patients were generally left to wait over 2 hours while they were complaining of pain, this is a potential red flag which has to be addressed.

While the structured ratings can inform you about long wait times at your department, the free-text comments give you the context (e.g. pain while waiting, wait-time of 2 hours, not attended to while in pain) which can help you take the appropriate corrective action.

In isolation, each of the safety related complaints may not seem like an issue at all. However, when the safety issue appears as a recurring pattern where the data is able to show it, this becomes an obvious RED FLAG.


#2 Perception of support staff

Patient feedback would sometimes also include comments about support staff if something really stood out in a positive or negative way. Analysis of free-text patient comments on a departmental level can reveal how patients perceive your support staff. Was scheduling difficult? Were patients adequately greeted? Were the medical assistants rude? Were patients being informed about delays in a timely fashion and so on. All of this can help you understand areas that your department needs to improve on to make the whole patient experience a positive one.



#3 Physician weaknesses

While physicians in your department may be fairly competent, there will always be areas that they could work on. For example, if over 200 patient comments, a trend appears that at least 15 patients spoke about the physician making a paperwork error, this could be an area for the physician to work on. Perhaps more “technical” skills training is required or training on how to properly use the EMR software at your hospital. A detailed text analysis can capture a variety of issues including problems with bedside manners, timeliness, competency, diagnosis errors and more.



#4 Remarkable physician traits

Every physician has a unique style. Why do patients favor one physician over another? Text Analysis of volumes of patient comments can reveal interesting physician traits. For example, one physician may be loved for their friendliness while another could be loved for intelligence and competence. This information can be used in creative ways.

For instance, you can use known physician traits to help patients find a provider based on their personal preferences. You can also reward physicians based on certain traits that they demonstrate towards patients. Another use of these interesting traits could be for marketing the quality of your physicians and your department.



#5 Issues with hospital facility and location

When you aggregate all comments from patients within one hospital facility, Text Analysis can reveal important issues at the hospital level. For example, if parking was really limited, patient complaints relating to parking would become a recurring theme and the analysis would capture that. Other issues you may find could relate to the location of the hospital, cleanliness issues, food availability and others that would be very specific to your hospital.

As you make changes to address issues discovered, repeating the analysis in subsequent years will allow you to monitor if the issue has been adequately addressed or if there are still lingering complaints.




We described 5 important insights you can get from Text Analysis of volumes of free-text patient comments coming from surveys. These insights can help you improve the overall patient experience by allowing you to focus on the real issues at different levels of your organization.

When you pair these patient comments with patient reviews on the Web (e.g. RateMDs), you can get an even better view into what is going on in your department and hospital.

By analyzing all these patient comments on a regular cadence through a service like ours or with your own software tools, you are opening up new opportunities for addressing issues you never knew existed in your department or with a particular physician.

Such analyses also gives you a measurable way to assess if the problems at your department are slowly being resolved (e.g. wait time, parking issues, unpleasant staff, dirty facility, uncomfortable lobby and etc.).

Keep Reading: A case study of how we helped a 300-bed hospital gain insights from their patient comments


Scroll to Top