This presentation reports on how one organization improved its patients' experiences through sentiment analysis. For follow up see http://tellmymd.com.
8. Example of a Complaint
“The Drs seem to
feel they know more about my body than
I do. I feel put off when they give me a
strange look and tell me , "No, I doubt
that. I would like to be taken seriously
and not given weird looks when
describing my situation.”
9. Example of a Positive Comment
“I love this clinic. The staff is amazing,
always beyond helpful and friendly. I
transferred from my family Dr. because I
had such a wonderful experience here.
Don't change anything. Dr. X is
awesome! I am one happy patient.”
This is Farrokh Alemi. I would like to tell you a brief story about how on critical access organization has improved patient experiences
This presentation describes how Southern Humboldt Community Healthcare District, which includes both the clinic and the Jerold Phelps Community Hospital, improved patient satisfaction. The Jerold Phelps Community Hospital is one of thirty‐two Critical Access Hospitals in the California. It is located in Garberville, a very charming small town with a main street with restored shop fronts and restaurants. A quiet place that is not always on forefront of change and innovation. Extraordinary things are happening in this rural community. There is a lot going here, from new technology, to better finances, to culture of dedication to patients, to quality awards from California Hospital Association. It is a long list of recent achievements. I am here to tell you a small part of the transformation that has taken place. I am here to tell you how they improved patient satisfaction.
It is not easy to improve patient satisfaction. CMS has asked acute care hospitals to conduct CAHPS surveys. This is required by urban hospitals and optional to rural hospitals. CAHPS is a 27 item survey of a randomized sample of patients that takes a long time to collect and has irritated both hospital administrators and patients. No one wants to fill them, forcing hospitals to hire firms to call patients and remind them to complete the survey. Sometimes patients have to be called up to 5 times to remind them to complete the survey. The data are available after long delays. Lots of data are collected, reams of reports are prepared which sit on a manager’s desk and do not lead to real and lasting improvements. A lot of effort goes into measurement and very little into actually improving the patient’s experience.
We set out to reinvent satisfaction surveys. To make it more efficient, more flexible. We did not want to lead the client to specific answers. We decided to use the shortest possible survey. We asked just one question.
This is an open ended question. It allows us to capture the clients’ views in their own voice. We do not need to reformulate it or force the responses to a preset menu of options. The patient says what he wants and in words he wants.
We can collect the information through personal interviews, from web comments, from complaints to risk management department, through emails, kiosks, IVR phone calls, postcards and comment card. Early on, Southern Humboldt Community Healthcare District decided to focus on comment cards and kiosks at strategic locations. Here you see the back of the comment card that patients use.
This is the front. Patients can leave the cards in comment boxes or stamp and mail them.
Here is an example complaint that was posted. Here the patient is talking about how the doctor make him or her feel put off. The patient wants to be taken seriously. This complaint points to a problem in the relationship or empathy shown to the patient.
Here is an example of a positive comment. The patient’s words such as amazing, beyond helpful, friendly and happy communicate that this is a positive comment.
We use sentiment analysis to classify comments into positive or negative comments. This is a patent pending artificial intelligence process that classifies sentences based on the keywords and their context. We classify the topic into complaint or praise about getting in to be seen, wait time, physician issues, nursing issues, staff issues, cost of care, facility issues, privacy and other topics. This allows further analysis of the data.
Here are some example of words used in complaints received early on about a year ago.
Here are examples of words used more recently, reflecting a shift in the comments.
To provide managers with quantitative benchmarks we examine time to next complaint. One would expect that over time fewer and fewer complaints are seen. Here we see consecutive complaints increased around the 25th day. When consecutive complaints go beyond a control limit then an email is sent to the manager alerting him or her to potential adverse events occurring.
Managers are also provided with benchmark comparisons showing how they stack up against other units within the organization. Here the ER is doing worst than the clinic.
The data is analyzed by the computer. Reports and graphs are written by the computer and as a consequence the analysis is available in real time, when the manager wants to examine the data. As new comments come in the analysis and the reports change. This enables improvement teams to make changes and track their impact. It allows employees to make changes and see how it is affecting their performance. The receptionist can see if a few changes in her or his interactions with the patient leads to more praise and fewer complaints.
This graph shows complaints and praises over the last 500 comments received within the Southern Humboldt’s Clinic. On the X-axis is time. On the Y-axis we see number of comments. Green shows positive and red shows negative comments. So at start there were 77 comments of which 13 were complaints. That is a rate of 16%. In the next time period, the number of comments declined to 38 of which 9 were complaints. This is a rate of 23%.
In the next set of time periods we see the number of comments climbing to 103 of which 15 were complaints, a rate of 15%.
The we see a period in which the clinic ran out of comment cards and the number of comments declined to low of 17.
Finally more recently we see the number of positive comments increasing and the number of negative comments staying low. Look at all the red on the left and the lack of red on the right side of the graph.
The rate of complaints declined from starting values of 16% of comments to recent values of 3%. We have seen a measurable and statistically significant improvement in patient experiences at Southern Humboldt Health District over time.
This has been made possible by the great people who work at Southern Humboldt. What we have contributed are analytics and business intelligence that engages every employee in making change.
We have designed a process that costs a fraction of standard CAHPS surveys.
It reports in real time, when data and information matter.
It is less biased than asking patients to answer preset leading questions. It is something else when you get feedback in the customer’s words than some standardized questionnaire menu.
We have been delighted to help Southern Humboldt Health District improve over time. Urban healthcare organizations have a lot to learn from Southern Humboldt. It is not just measurement that made the difference. They have been able to engage their staff in improvement efforts. They have a culture of quality that focuses on patient experiences. I have been in many organizations who want to improve patient experiences. It is rare to see real lasting improvements. Most try but few succeed. Southern Humboldt is succeeding. Its patients are having more positive experiences and at the end of the day that is what really matters.