Notes: Counties are categorized based on proximity to the hurricane: direct path, near path, remote identified as a disaster area by FEMA, and remote not identified as a disaster area by FEMA. Data may be suppressed due to confidentiality or not applicable if there were no counties classified in a proximity category. See Data Notes & Methods.
Note: Counties are categorized based on proximity to the hurricane: direct path, near path, remote identified as a disaster area by FEMA, and remote not identified as a disaster area by FEMA. Data may be suppressed (§) for confidentiality or marked as not applicable (NA) if there were no counties classified in a proximity category. See Data Notes & Methods.
This Fast Stats topic provides general descriptive statistics on changes in rates of hospital utilization following historical U.S. hurricanes. Information on hospital utilization is based on data from the HCUP State Inpatient Databases (SID) and State Emergency Department Databases (SEDD). Information about the proximity of counties to hurricane paths was derived from the National Oceanic and Atmospheric Association (NOAA) Best Track datasets and the Federal Emergency Management Agency (FEMA) Disaster Declaration Summary database. In order to calculate hospital utilization rates, estimates of the resident population were taken from the U.S. Census Bureau's American Community Survey (ACS). The hospital utilization statistics reported here may be influenced by a number of factors such as hurricane-related evacuations and hospital closures for which the source data have not been adjusted. These limitations may cause imprecision in the estimates. Please refer to the "Caveats on Data Analysis" section below for more information. HurricanesEleven hurricanes that impacted the mainland of the United States between 2005 and 2017 are included. States that were affected by these hurricanes were identified based on information on hurricane activity from the National Oceanic and Atmospheric Association (NOAA) and information on counties identified by the Federal Emergency Management Agency (FEMA) as major disaster areas caused by the hurricane. The start date of hurricane activity was identified by the date of the first record that indicated hurricane activity for a State in the NOAA Storm Best Track dataset. Hurricane activity was defined as any one of the following: a tropical cyclone of hurricane intensity with winds greater than 64 knots, a tropical cyclone of tropical storm intensity with winds of 34-64 knots, or an extratropical cyclone with winds of at least 34 knots.
Hospital utilization data from neighboring States not impacted by a hurricane may be included with utilization data from the States impacted by the hurricane. If a patient who resided in a hurricane-impacted State was treated at a hospital in another State, the hospital encounter is included in the utilization count for the patient's county of residence if the hospital was located within 250 miles of the patient's residence. Distance was determined based on the centroids of the ZIP Codes of the hospital and patient's residence. Hurricane ProximityCounties in States impacted by the hurricane were classified into one of four proximity categories: direct path, near path, remote/FEMA disaster, and remote/not disaster. These proximity designations were derived from two data sources:
Counties were classified into one of the four hurricane proximity categories as follows:
The "direct path" designation was based on the hurricane path only, whether or not the county was declared a major disaster area by FEMA. It is possible, but rare, for counties in the direct path of the hurricane to not be designated as major disaster areas by FEMA. Note that one or more proximity categories may not be applicable for some hurricanes if there were no counties classified in a proximity category. For example, all the counties in the affected States could be classified into the direct path, near path, or remote/FEMA disaster proximity categories, and none into the remote/not disaster proximity category. Hurricane-specific maps displaying the areas designated as direct, near, remote/FEMA disaster, and remote/not disaster are available in an exported data file, which can be downloaded by expanding "Show Data Export Options". The maps document the population at risk for the hurricane and what hospital utilization data were available from the HCUP State databases.
3 https://www.fema.gov/disasters Patient County AssignmentInpatient stays and emergency department (ED) visits were classified into hurricane proximity categories using the patient's county of residence. Patient county was assigned based on the ZIP Code of the patient's residence using the SAS function for ZIP Code to county assignment. For ZIP Codes that cross county boundaries, the SAS function used the geographic centroid of the ZIP Code to assign the county.6 A sensitivity test using the 35.4 million records in the 2016 SID demonstrated that the SAS function assigned a county different from the county with the population centroid of the ZIP code in 0.6 percent of SID records. If the patient ZIP Code indicated the patient was homeless (the HCUP data element ZIP = "H"), then the patient county was assigned to be the same as the hospital county. Records for patients with a ZIP Code that was missing, invalid, or indicated the person was from a foreign country were excluded from the tabulated counts. This exclusion dropped less than 1 percent of records in any year. Additional information on using HCUP data for county-level analyses is available in Method Series Report #2019-04: Conducting County-Level Analyses With HCUP Data: Approaches and Methodological Considerations.
6Additional information from SAS on the geocode procedure: Unit of AnalysisThe unit of analysis is the hospital discharge (i.e., the hospital inpatient stay) or an emergency department (ED) visit, not a person or patient. This means that a person who is admitted to the hospital or visits the ED multiple times in one year is counted each time as a separate discharge from the hospital or a separate visit in the ED. For Fast Stats, all stays and visits are counted one time only, regardless of the number of relevant diagnosis or procedure codes that appear on the record. For instance, when identifying injury-related inpatient stays and ED visits, a record may include more than one of the injury-specific codes; in such a case, the record is only included once in the injury counts. Percent Change from Pre-Hurricane Average for Inpatient Stays or ED VisitsThe percent change in the rate of inpatient stays or ED visits compares hospital utilization during and post-hurricane to the pre-hurricane average utilization rates. Time periods are calculated based on the State-specific start date of the hurricane activity (documented under the section on Hurricanes).
The rates by hurricane proximity were population-weighted to account for the different sizes of counties. Population data were obtained from the U.S. Census Bureau, American Community Survey (ACS) overall and by specific age groups. The rate of inpatient stays or rate of ED visits includes the HCUP number of stays or ED visits in the numerator and the U.S. resident population in the denominator (with a multiplier of 10,000). For age, population rates were based on the population for that age group. Population-based rates by hurricane proximity always include data from two or more counties and two or more hospitals. Counties and hospitals may be within the same State or from different States. The percent change from the pre-hurricane weekly average is presented as the baseline value 0 for the pre-hurricane period. The percent change for the hurricane and each post-hurricane week is calculated from the pre-hurricane weekly average and demonstrates how utilization varied each week from the pre-hurricane period. Detailed information on the population-based rates and percent change are available in the exported data file (under "Show Data Export Options"). For the pre-hurricane period, the export file includes the 4-week average rate, the average number of weekly encounters (i.e., the average numerator count of HCUP inpatient stays or ED visits), and the Census population count (the population denominator for the rate). For the hurricane week, the export file includes the number of encounters, the population-based rate, and the percent change from the pre-hurricane average rate to the rate for the hurricane week. For each post-hurricane week, the export file includes the number of encounters, the population-based rate, and the percent change from the pre-hurricane average rate to the rate for the post-hurricane week. Counts are rounded to the nearest 10 discharges or ED visits, with any counts less than or equal to 10 or representing fewer than two hospitals suppressed for confidentiality. The exception is raw counts of 11-14, which are rounded to 11. Suppression Rules for ConfidentialityIf the average number of encounters in the pre-hurricane period is less than or equal to 10, or represents fewer than two hospitals, the percent changes for the hurricane week and each post-hurricane week are suppressed. This will result in one or more missing trend lines in a graph. If the number of encounters for the hurricane week or any post-hurricane week is less than or equal to 10, or represents fewer than two hospitals, the percent change is suppressed. This will result in one or more missing data points in a graph and will cause a discontinuity in the trend lines. When three or more data points are suppressed, the trend line is omitted from the graph and none of the data values are provided in the underlying data tables and exported data file. Suppression of multiple trend lines is particularly common in the inpatient setting for age 0-17 years. For some hurricanes, all trend lines may be omitted from certain graphs due to data suppression; in these instances, the graph includes a note indicating that "Data are insufficient for presentation." Inpatient StaysStatistics on inpatient stays for each hurricane are from the HCUP State Inpatient Databases (SID) and quarterly data if available. Information based on quarterly data should be considered preliminary. Quarterly data will be replaced by the State's complete annual SID for the year when it is available. As a result, previously reported statistics for a given hurricane may change. For this analysis, the SID are limited to patients treated in community hospitals in the State. Community hospitals are defined as short-term, non-Federal, general, and other hospitals, excluding hospital units of other institutions (e.g., prisons). Included among community hospitals are obstetrics and gynecology, otolaryngology, orthopedic, cancer, pediatric, public, and academic medical hospitals. Excluded are community hospitals that are also rehabilitation and long-term acute care facilities. If a patient was transferred from one community hospital to another, then the SID records for both the transferring and receiving hospitals were included in the analysis. In any data year for the hurricane-impacted States, an average of 0.6 percent of inpatient stays from community hospitals7 are missing from the SID (range of 0.0 to 4.7 percent). These missing discharges represent an average of 3.0 percent of community hospitals7 (range of 0.0 to 31.4 percent). One State is missing a total of 4.7 percent of discharges, due to missing 31.4 percent of the community hospitals in the state7 (predominately small hospitals with fewer than 50 beds). 7Excluded are community hospitals that are also rehabilitation and long-term acute care facilities. Emergency Department VisitsEmergency department (ED) visits are defined as ED encounters that do not result in a hospital admission to the same hospital (i.e., treat-and-release ED visits). Statistics on treat-and-release ED visits for each hurricane are from the HCUP State Emergency Department Databases (SEDD) and quarterly data if available. Information based on quarterly data should be considered preliminary. Quarterly data will be replaced by the State's complete annual SEDD for the year when it is available. As a result, previously reported statistics for a given hurricane may change. The SEDD are limited to patients treated in community hospital-owned EDs in the State. Excluded are community hospitals that are also rehabilitation and long-term acute care facilities. If a patient was transferred from the ED, then records for both the transferring and receiving facilities were included in the analysis. There would be a SEDD record from the transferring ED. Most of the time (91 percent), ED transfers result in an inpatient stay. In these cases, the record for the receiving hospital would be included in the SID; otherwise, there would be a second SEDD record. In any data year for the hurricane-impacted States, an average of 0.4 percent of the ED visits from community hospital-owned EDs8 are missing from the SEDD (range of 0.0 to 1.7 percent). These missing ED visits represent an average of 1.0 percent of community hospital-owned EDs8 (range of 0.0 to 3.1 percent). Information on ED utilization is not presented for all hurricanes. Gustav, Ike, Isaac, and Rita lack ED utilization information because none of the impacted states provided ED data corresponding to the hurricane time period. Additionally, some states impacted by hurricanes Harvey, Irene, Matthew, and Sandy provided inpatient data, but no ED data. In this situation, the ED information incorporates data from a smaller set of states than the inpatient information; accordingly the population count is smaller for the ED setting than the inpatient setting. 8Excluded are community hospitals that are also rehabilitation and long-term acute care facilities. Transition in Clinical Coding Used to Define ConditionsIn October 2015, the United States transitioned coding systems for reporting diagnoses and inpatient procedures from the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) to the International Classification of Diseases, Tenth Revision, Clinical Modification/Procedure Coding System (ICD-10-CM/PCS). The following hurricanes would have had diagnoses and procedures reported using ICD-9-CM: Dennis, Rita, Wilma, Gustav, Ike, Irene, Isaac, and Sandy. The following hurricanes would have had diagnoses and procedures reported using ICD-10-CM/PCS: Matthew, Harvey, and Irma. No hurricane has clinical data that crossed coding systems. More information about the use of data across the two coding system may be found on the HCUP User Support (HCUP-US) web page for ICD-10-CM/PCS Resources. CirculatoryCirculatory is defined using the following Clinical Classifications Software (CCS) for ICD-9-CM categories or Clinical Classifications Software Refined (CCSR) for ICD-10-CM default categorization scheme for the principal (or first-listed) diagnosis as appropriate for the time period of the hurricane. The circulatory condition must be reported as the principal diagnosis on an inpatient stay or the first-listed diagnosis on an emergency department visit. The principal or first-listed diagnosis is used so that a record is only assigned to one specific clinical condition. Percent change in population rates for circulatory conditions is not reported for age 0-17 years because these conditions are relatively uncommon for this age group. CCSR for ICD-10-CM Diagnoses Starting October 1, 2015
InfectionsInfection is defined using the following Clinical Classifications Software (CCS) for ICD-9-CM categories or Clinical Classifications Software Refined (CCSR) for ICD-10-CM default categorization scheme for the principal (or first-listed) diagnosis as appropriate for the time period of the hurricane. An infection must be reported as the principal diagnosis on an inpatient stay or the first-listed diagnosis on an emergency department visit. The principal or first-listed diagnosis is used so that a record is only assigned to one specific clinical condition. CCSR for ICD-10-CM Diagnoses Starting October 1, 2015
InjuriesInjury is defined using the following ICD-9-CM or ICD-10-CM diagnosis codes as appropriate for the time period of the hurricane. The injury must be reported as the principal diagnosis on an inpatient stay or the first-listed diagnosis on an emergency department visit. The principal or first-listed diagnosis is used so that a record is only assigned to one specific clinical condition. ICD-10-CM Codes Starting October 1, 2015
RespiratoryRespiratory is defined using the following Clinical Classifications Software (CCS) for ICD-9-CM categories or Clinical Classifications Software Refined (CCSR) for ICD-10-CM default categorization scheme for the principal (or first-listed) diagnosis as appropriate for the time period of the hurricane. The respiratory condition must be reported as the principal diagnosis on an inpatient stay or the first-listed diagnosis on an emergency department visit. The principal or first-listed diagnosis is used so that a record is only assigned to one specific clinical condition. CCSR for ICD-10-CM Diagnoses Starting October 1, 2015
AgeAge refers to the age (in years) of the patient at admission. Discharges or visits missing age are excluded from results reported by age. Age is grouped into three categories: 0-17 years, 18-64 years, and 65+ years. Less than 0.05 percent of records are missing information on age. Caveats on Data AnalysisIt is important to note that there were certain limitations to the data used for this analysis. Identifying the Proximity of Counties to the Hurricane's PathThe six-hour intervals between measurement points in the NOAA Best Track datasets could result in gaps in information if the storm was fast moving. The use of geographic circles defined by wind radii to classify counties as impacted by the hurricane may be overgenerous in identification (e.g., when only the outside edge of the wind radii circle touches the border of the county or one small coastal island) and may not identify counties that would have been detected if a shape other than a circle was used. For some hurricanes, there is incomplete wind radii information as a storm nears dissipation, making the end of the hurricane difficult to determine. For this analysis, NOAA Best Track data records with incomplete wind radii information are not included. Population at Risk (Denominator for the Population-Based Rates)The county-specific population data were annual counts from the U.S. Census Bureau, American Community Survey (ACS). The information would not have taken into account evacuations prior to the hurricane making landfall, seasonal migration patterns (e.g., elderly living in Florida during the winter), people who resided in a county impacted by the hurricane but who were not in the area at the time of the hurricane, or people from counties not impacted by the hurricane visiting the area at the time of the hurricane. Utilization Counts (Numerator for the Population-Based Rates)Information on hospital closures around the time of the hurricane was unavailable. In addition, hospitals may have had difficulty reporting utilization to the HCUP Partner organization (or were temporarily considered exempt from reporting) resulting in an underestimate of utilization using the SID and SEDD. Hospital care just prior to the hurricane may have been the result of preparing for the hurricane. In contrast, hospital care after the hurricane may not be hurricane related. Suppression for ConfidentialitySuppression of data points, trend lines, or entire graphs was frequently applied to condition-specific selections for age 0 to 17 years, especially in the inpatient setting. For a detailed description of the suppression rules used in this topic, refer to the section "Percent Change from Pre-Hurricane Average for Inpatient Stays or ED Visits" above. Use this export feature to download the underlying data and hurricane population risk maps.
This HCUP Fast Stats topic was developed as part of the partnership project titled "Assessing and Predicting Medical Needs in a Disaster" among the Office of the Assistant Secretary for Planning and Evaluation (ASPE), the Office of the Assistant Secretary for Preparedness and Response (ASPR), and the Agency for Healthcare Research and Quality (AHRQ). For more information, see: https://www.aspe.hhs.gov/assessing-and-predicting-medical-needs-disaster.
|
Internet Citation: HCUP Fast Stats. Healthcare Cost and Utilization Project (HCUP). May 2020. Agency for Healthcare Research and Quality, Rockville, MD. www.hcup-us.ahrq.gov/faststats/hurricane/hurricaneimpact.jsp. |
Are you having problems viewing or printing pages on this Website? |
If you have comments, suggestions, and/or questions, please contact hcup@ahrq.gov. |
Privacy Notice, Viewers & Players |
Last modified 5/26/2020 |