Home / Tutorials

Hospital Accessibility Analysis in ArcGIS Online

Timely access to hospital services can be critical for mitigating morbidity and mortality for many acute conditions, such as heart attack and stroke. Convenient access can also improve quality of life for people who require hospital admission for elective medical procedures.

Hospital closures in the USA have made timely access to hospital services a challenge in many rural areas, especially for elderly and low-income individuals (Wishner and Solleveld 2016).

This tutorial will focus on visualization and analysis of rural hospital closure data in ArcGIS Online.

Literature Review

The Hill-Burton program was created by the US Congress in 1946 to provide federal funding for construction of rural hospitals, most notably in the South. However, limits on Medicare spending imposed by Congress in the early 1980s led to the closure of many rural hospitals. This trend was slowed by the Medicare Rural Hospital Flexibility Program of 1997, which authorized higher "reasonable cost basis" reimbursement at rural hospitals designated as "Critical Access Hospitals." (Wishner and Solleveld 2016).

However, closures began to increase again during the Great Recession of 2008-2009.

• Between 2010 and 2021, around 136 rural hospitals closed in the US, with at least one closure in every state. The pressures are manifest in the economics, with low patient volume and high reliance on government reimbursement for services making rural hospitals especially vulnerable. Financial pressures associated with COVID-19 became especially acute beginning in 2020 (Ellison 2021).
• The closures during the 2010s resulted in as much as one percent of the US population being unable to access a hospital with a 15 minute drive time, although that effect diminishes with 30- or 60-minute drive times. The Deep South was the area most deeply affected (McCarthy et al. 2021).
• Not all closures are the same, with some nearby facilities capable of absorbing demand with little increase in access time. Closures generally did not affect EMS response time, but did affect transport times (Miller et al. 2020).

A number of researchers have studied the public health effects of increased travel time to emergency rooms and found that the implications are complex and dynamic:

• Jang et al. (2021) assert that different acute conditions have "optimal" travel times, under which decreased travel times do not substantially improve outcomes:
  • Intracranial injury (TBI and stroke) 71–80 min
  • Acute myocardial infarction (heart attack) 31–40 min
  • Other acute ischemic heart disease 70–80 min
  • Fracture of the femur, 41–50 min
  • Sepsis. 61–70 min
• Ripley et al. (2015) found that after adjusting for the patient, treatment, and facility characteristics, travel times of more than 90 minutes significantly increased odds of in-hospital mortality over travel times of less than 30 minutes.
• Shen and Hsia (2011) found that lost access to hospital services within less than 10 minutes resulted is small increases in heart attack mortality, but were considerably more substantial when access time increased by over 30 minutes. These effects diminish over time, indicating that communities eventually adapt to the changed conditions.
• Yamashita and Kunkel (2010) used a multiple regression model to observe that travel distance to the hospital becomes a less significant determinant of heart disease mortality when social factors like poverty rate, employment rate and rural population are considered. This leads to the possibility that the issue is less about travel time and more about the health characteristics of people who live in rural areas.
• Beyond transport times, the timeliness of response upon arrival at the ER is also a major factor, indicating that quality of service may be as important as access to that service. Almekhlafi (2021) estimated that for every 10-minute delay between arrival at the emergency room (ER) and starting stroke treatment, patients with severe strokes may lose eight weeks of healthy life.

Acqire the Data

Hospital List

The definitive list of hospitals in the US is the list of Hospital General Information available from the Centers for Medicare and Medicaid Services, which is part of the US Department of Health and Human Services. This data contains street and city information, but no lat/long, thus necessitating geocoding if you want to use it for mapping.

When working in ArcGIS Online, the easier source for that data is the Living Atlas layer Hospitals Registered with Medicare from ESRI's Federal User Community organization.

The default symbology for the hospital layer includes multiple categories, two of which represent inpatient facilities that are most relevant for the analysis in this example.

Acute Care Hospitals provide "inpatient medical care and other related services for surgery, acute medical conditions or injuries (usually for a short term illness or condition)" (Centers For Medicare and Medicaid Services 2013).

Critical Access Hospitals have no more than 25 inpatient beds and are at least 15 miles by secondary road or 35 miles by primary road from the hearest hospital. This category was created by the aforementioned Medicare Rural Hospital Flexibility Program of 1997 (Wishner and Solleveld 2016).

Closed Rural Hospitals

The Cecil G. Sheps Center for Health Services Research at The University of North Carolina at Chapel Hill has been tracking closure of rural hospitals since 2005.

The Sheps Center makes their closure list available as an Excel file which has street addresses but no lat / long coordinates. A geocoded snapshot of that list as of 11 January 2022 is available as the Minn 2022 Closed Rural Hospitals layer from the University of Illinois ArcGIS Online organization.

Spatial Analysis

Filter Open Hospitals

The CMS hospital list will likely contain facilities that the Sheps Center has found to be closed completely or closed for inpatient services. You will need to filter only the open facilities so you can have a clear view of the areas around closed hospitals that do or do not have reduced hospital accessibility.

1. Click the Perform Analysis icon on the hospitals layer, Then Find Locations and Find Existing Locations.
2. Add Expression and choose not within a distance of 0.5 miles from the closed hospitals layer features.
3. Provide a meaningful Result layer name.
4. Zoom your map to cover the area you are analyzing and make sure Use current map extent is checked.
5. Click Show credits to make sure your analysis will consume a reasonable number of credits. Unless you are covering a multi-state region, your count of credits should be under one.
6. Run Analysis. This may take a few minutes, depending on the number of features being analyzed and the amount of current activity on the cloud server.

Circular Buffers

Since 30 minutes of travel time to the hospital seems to be an approximate threshold for improved health outcomes (see the literature review above), we use circular buffers of 15 miles geodesic distance to get a rough estimate of 30-minute service areas to hospitals.

1. On the open hospital layer, click the Perform Analysis icon and select Use Proximity and Create Buffers.
2. Enter buffer size (15 miles).
3. Give a meaningful Result layer name.
4. Show credits to make sure you have a number in single digits or less.

Drive-Time Buffers

You can now identify potential areas of reduced access by looking for closed hospitals points that are not covered by the 15-mile buffers.

In this example, St. Mary's Hospital in Streator, IL, which closed in 2016, is outside the buffers for facilities to the north and south.

To get a more realistic assessment of timely emergency healthcare access around the closed facility, we put 15-minute drive-time buffers on the adjacent facilities to see what areas are not covered.

We choose to do drive-time analysis on a more limited collection of facilities because drive-time analysis consumes a considerably higher number of credits per feature than circular buffers, and this reduces the cost of analysis and the possibility of account credit exhaustion.

1. Zoom in to the area around the closed facility, but stay zoomed out so that the central points of adjacent facility are within the current extent (area of view).
2. On the open hospital point (not buffer) layer, click the Perform Analysis icon and select Use Proximity and Create Drive-Time Areas.
3. For the Measure, enter 15 minutes of Driving Time.
4. For Travel direction select Towards Facility.
5. Enter a meaningful Result layer name.
6. Make sure Use current map extent is checked and Show credits to make sure the maximum credits is under 20 and, preferably, less than 10.
7. Hide the circular buffers and verify that the area around the closed facility is still outside the accessibility areas for the open facilities.

In the case of the area around the closed St. Mary's Hospital in Streator, IL, we find that the adjacent facilities fail to offer 15-minute access to the area around the closed facility.

Changing the distance to 30 minutes increases the coverage to include the area around the closed facility, indicating that while the closure was not desirable, the 30-minute access to emergency services noted as important in the literature is still available in the area around the closed hospital.

Overlaying the two or more different buffers on top of each other forms an approximation of an isarithmic map, where the buffer colors blend together visually to connect areas of similar value. The effect can be improved by using the dissolve option to get rid of the boundary lines when areas overlap.

Community Analysis

Demographics

We can use the US Census Bureau's data.census.gov data portal to find comparative demographic information on the community.

For Streator, IL in the 2019 American Community Survey, we see characteristics typical of rural communities that lose local hospital access:

1. The total town population of 12,500 is clearly rural in character.
2. The median age of 42.6 years is older than 38.6 years for Illinois as a whole.
3. The median household income of $39,736 is considerably lower than $69,187 in the state as a whole.

Google Street View

Google Street View can be used to "ground truth" your data, although the date that the Street View camera visited a location can vary widely and may not reflect current conditions.