I completed this project for the Maven Healthcare Challenge. I analyzed nine years worth of survey results from the Hospital Consumer Assessment of Healthcare Providers and Systems (HCAHPS). The HCAHPS survey is a standardized instrument for measuring patients' perspectives on hospital care, and one of its main goals is to create incentives for hospitals to improve their quality of care.

Analysis Questions

• Have hospitals' HCAHPS scores improved over the past 9 years?
• Are there any specific areas where hospitals have made more progress than others?
• Are there any major areas of opportunity remaining?
• What recommendations can you make to hospitals to help them further improve the patient experience?

Analysis Process

The Data

The data includes results by release period and measure at both a national and state level. The results are broken down by what percentage patients chose the most positive ("top-box response"), intermediate ("middle-box response"), and least positive ("bottom-box response") answer choice for each measure (some measures are comprised of multiple questions). In addition, the data contains the questions that feed each measure, the survey period that corresponds to each release period, and the response rate for each hospital.

Exploring & Analyzing the Data

I started by exploring and cleaning the data in Excel. One of the first things I noticed is that for the composite (multiple questions) measures 1, 2, 3, and 5, as well as for the measures that assess the cleanliness and quietness of the hospital environment, the bottom-box answer is "usually" and the middle-box answer is "sometimes or never." "Usually" is more positive than "sometimes or never", and so it appeared that middle-box answer is actually the least positive for these measures. I created a new column to identify the true "bottom-box" response for each measure for accurate analysis. In Tableau, I created a calculated field to pull the middle-box response rate for the measures identified above when I analyzed bottom-box response trends.

There are a total of ten measures, and I wanted to create larger buckets that I could work with in my dashboard for easier viewing of the measures. I created a new column in the source data in the Measures table to place each measure in one of the following categories: Communication, Hospital Environment, Discharge & Care Transition, or Overall. In my Tableau dashboard, I color-coded these categories to help organize the measures.

Finally, I added a release period year column as a date field in the Reports table in the source data. The release period column is a text field and is a primary key in the Reports table and a foreign key in several other tables. I wanted to preserve the original field in order to relate the tables, so I created the new date field to pull into the visual when I wanted to display the data as trends over time.

Tableau Dashboard

View my interactive dashboard on my Tableau Public Profile.

The first image below is the top portion of the Tableau Dashboard. The map chart can be used as a filter for the Measure Scores and Response Rate charts to the right of it. The titles (gray boxes) for the Measure Scores and Response Rate charts are sheet selectors--click on the title to switch between the charts. On the map, hover over a state to view additional details in the tool tips. To view measure scores and response rates for a single state, click on the state on the map. To zoom in on a region, use the Region filter. Finally, to view one or a few measures at a time on the Measure Scores line chart, use the Measure filter above the map.

I decided to show the year-over-year trend for the measures individually instead of an overall average top-box score of all measures for a couple of reasons. First, I thought it would be interesting and useful to be able to compare measure scores to one another. If you click on the first (top) "i" icon above the Measure Scores line chart, it will show the legend for the measure color categories I created. This way, you can compare, for instance, how the measures related to communication (blue lines) compare to the measures related to hospital environment (gray lines). You can also use the Measure drop-down filter above the map chart to show fewer measures at a time. I realize having so many lines is initially a lot to take in, so I tried to make it easier to understand with the color categories, color legend menu that can be shown and hidden with the "i" icon, additional details in tool tips, and filter menu options.

The two info ("i") icons on the top right of the dashboard next to the title will show additional details about the measure categories and the response categories when selected. This information can be hidden again by clicking the "X" that will replace the "i" when the information is showing.

The next image below includes the charts on the bottom portion of the Tableau dashboard. The Survey Measures & Questions chart shows the measure and corresponding questions that feed that measure, along with the measure category color indicator (dot on the right). This chart is filtered by the National Results line charts on the bottom left. Select a measure from the filter, and the year-over-year scores will show on the line chart, and the questions will show on the Survey Measures & Questions chart. The title for the National Results charts are also sheet selectors--click on the title to switch from top-box percentage trends to bottom-box percentage trends.

The Top-box Response Change bar charts (bottom right, also sheet selectors) show state-level increases and decreases in top-box percentage scores for all measures and for specific measures (by using the Measures filter). There is one chart to show the percent change from the first release year (2015) to the latest release year (2023) and one chart to show the percent change from the 2022 to 2023 release years. I chose to include 2022 and 2023 because there is an average overall decrease in all top-box measure scores nationally, but, there are some states that showed improvement during this time. I wanted to highlight that information, plus have an easy way for the user to view general trends in changes, from most to least positive, on a state level.

Insights

• On a national level, the average top-box (most positive) response rate has either decreased somewhat or remained the same for all measures as of the 2023 release year (compared to the 2015 release year, which is the first release year in the data).
• All measures experienced an overall average decrease in top-box response rates between the 2021 and 2023 release periods. These release periods included survey data from October 2019 through September 2022. It is possible that the strain placed on the healthcare system by the Covid-19 pandemic negatively impacted hospitals' quality of care and patient experience during this time. (Note: Explore bottom-box response rates on the dashboard. These show a similar trend--an increase in more negative responses on all measures between 2021-2023 release periods).
• There were six states that were able to increase their average top-box response rate between the 2022 and 2023 release years: Alaska, Arkansas, Montana, Nevada, Montana, and Texas. Other states showed increased top-box response averages on specific measures. For example, Arkansas had a 2% increase in top-box response rates for the Care Transition measure between the 2022 and 2023 release periods. To explore more state increases (and decreases) between the last two release period years, overall or by measure, see the Top-box Response Change (2022-2023) chart in the interactive dashboard.
• Although the national trend showed either stagnant or slightly decreased top-box response percentage scores in the 2023 release period compared to the 2015 release period, there are some states that increased their overall average scores and/or specific measure scores. For example, North Dakota had a 3.3% increase in average top-box response rate in 2023 compared to 2015. In addition, North Dakota had a 4% increase in top-box response rates for the care transition measure specifically (and is above the national average as of 2023 release year). To see more states that had increased (and decreased) scores overall and for specific measures between 2015 and 2023, see Top-box Response Change (2015-2023) chart in the interactive dashboard.
• Note that the states displayed as having the largest positive increases in top-box response rates (overall or by measure) on the dashboard are not necessarily the states with the highest average top-box response score. Since The HCAHPS survey aims to improve healthcare quality, I wanted to highlight improvements (and also decreases, to identify areas in need of improvement) relative to each state instead of the overall top scores. The states with the highest average top-box scores overall can be identified on the map (darker grey indicates a higher percentage).
• The average response rate has been low across all nine years and has steadily decreased. It was the highest in the 2015 release period at 27.9% and lowest in the 2023 release period at 19.7%. The response rate can be explored in further detail based on specific states and regions in the dashboard.
• The Care Transition measure consistently had the lowest positive scores across all nine years of data. In contrast, the Discharge Information measure had much higher positive scores. I found this interesting, since these two measures have some overlapping aspects, and I even grouped them together when creating my measure categories based on the survey questions.
• The Discharge Information measure consistently had the highest rate of most positive responses, followed by the Communication with Doctors and Communication with Nurses measures.

Recommendations

• Due to consistently low rates of positive responses for the Care Transition measure, hospitals should consider prioritizing quality of care improvement efforts in this area. Since the the Discharge Information measure has some similar elements (e.g., both measures are about what happens when the patient leaves the hospital), it may be useful to further investigate what is contributing to higher patient satisfaction for the Discharge Information measure to inform efforts to improve patient experience with Care Transition.
• Facilitate a process for hospitals to share successes with quality improvement efforts with one another and also to ask for support in areas where they are experiencing challenges. People in the healthcare industry value helping others, and so this strategy will likely appeal to them. In addition, a forum such as this can serve as an opportunity to network and receive peer recognition for successes and innovative solutions to quality improvement.
• Response rates to the surveys have always been low and have decreased from 2015 to 2023. For various reasons, people have different preferences for how they provide feedback. Some patients may be more likely to fill out a survey on their mobile phone or computer, some may prefer to fill out a paper survey, and some may be more comfortable answering survey questions over the phone or being interviewed by someone before they leave the hospital. Consider offering multiple options for survey completion and assess if this increases response rate.
• In addition to different modes of survey completion, hospitals may consider offering incentives to patients for survey completion. It takes time and thoughtfulness to provide feedback, and getting something in return for that effort may increase interest and motivation in filling out the survey. For example, hospitals can enter the patient's name into a drawing for a gift certificate or gift basket in exchange for a survey response.
• Survey responses should be anonymous to encourage participation and honest feedback.