Department of Emergency Medicine, The University of Texas Health Science Center, Houston, TX 77030, USAUnited States
Also available as PDF (figures and tables appear in the PDF).
Introduction: Existing literature is missing a description of a displaced population in the after math of Hurricane Katrina, who were seen and discharged from emergency departments of a Houston hospital system 10 years ago. Hypothesis/Problem: Health effects of Hurricane Katrina are an important public health topic that is not sufficiently discussed in the existing literature. Failure to provide this information is largely due to the lack of appropriate, representative data and absence of a systematic data capture process. Methods: A retrospective Electronic Health Record review of ‘Katrina evacuees’, obtained from Houston Fire Department run call data, was used to identify: visit type, top three ICD-9-coded diagnoses, medical insurance, number of visits and emergency medical service utilization. Results: The majority of patient visits were by Black, female gender and adults between 19 and 44 years. The leading diagnosis was hypertension. Circulatory system related diagnoses were nearly three times higher among Katrina evacuees than national data from 2005 and 2007. Most patients used emergency medical service services [815(60%)], had one emergency department visit [570(70%)], and reported Medicaid [577(40%)] or self-pay [425(30%)] as the insurance source. Conclusion: Disaster planning for the aftermath of natural disasters would benefit from knowledge pertaining to known chronic and non-chronic care needs of populations in pre-specified areas. Variance in primary diagnoses suggests the need for published data reporting annual primary diagnoses in local EDs by region. Access to this information via the internet contributes to estimating the likelihood of ED volume of chronic and non-chronic visit demand,1 providing foundational information for disaster preparedness plans nationwide.
Keywords: Emergency medicine; Hurricane; Disaster preparedness; Natural disasters
The cohort of patients for this retrospective chart review was obtained from the Houston Fire Department’s (HFD) KErun calls to a Memorial Hermann Hospital (MHH) System ED. HFD, the third largest fire department in the United States, is responsible for providing Emergency Medical Services (EMS) to a population of more than 2 million in an area totalling 654 square miles.29 Run calls were designated as KEs by HFD if they had: (1) HFD electronic tablet records designating subject(s) as FEMA, flood evacuee, New Orleans, refugee or Katrina, (2) run calls made to temporary housing shelters housing Hurricane Katrina evacuees (i.e. Houston Reliant Center), or (3) run calls made to a zip code where FEMA maps indicated a high concentration of Hurricane Katrina evacuees. MHH is the largest not-for-profit health system in Southeast Texas and largest health system in Houston with 12 satellite hospitals, one of the nation’s busiest Level 1 trauma centers and a total of 3781 beds. Thus, run calls to MHH following Hurricane Katrina serves as an indicator of resource utilization required post the translocation of a population following a natural disaster. The study was reviewed and approved by the Institutional Review Board of the University of Texas – Health Science Center at Houston and 18,19 MHH (HSC-MS-07-0519).
The cohort utilized in this study were designated as KE if they had a: (1) Gulf coast state address, (2) Gulf coast state phone number, (3) notation on their medical record describing them as a KE or hurricane evacuee, and (4) Out-of-State insur-
ance. Patient-visits from the cohort of HFD run calls between August 2005 and August 2006 (landfall of Hurricane Katrina to one year post -storm), were cross-referenced with the MHH Electronic Health Record (EHR). Data were collected on standardized electronic abstraction forms created in Microsoft Ex- ® cel spreadsheet Version 14.0.7116.5000 (Microsoft Corporation, 2010, Redmond, Washington, USA).
Ten data abstractors received a tutorial from the PI, which included collection a mock data abstraction session. Training was supplemented with a PowerPoint tutorial (made available for review). Routine meetings took place between abstractors and investigators to ensure productive and consistent rates of data abstraction. The PI implemented quality assurance methods and reviewed the first five patient-visit-data abstractions of each abstractor. This strategy ensured achievement of performance goals. Retraining took place when errors were identified.
Abstractors included patient visits found in the MHH EHR between August 2005 and August 2006. Additional visits absent from the HFD dataset were also included. For example, if the HFD dataset shows two visits between August 2004 and August 2006 for a patient and the MHH EHR listed even visits for that same patient during that time interval, then all seven patientvisits were recorded in the study dataset. Missing, conflicting, or ambiguous data were marked as unknown by the abstractors and later reviewed by senior members of the research team for quality control/assurance purposes.
Measures: Data was collected on the following variables: name, Medical Record Number (MRN), Date of Birth (DOB), race, arrival date, discharge date, visit type, chief complaint, and top three diagnoses, along with ICD 9 codes, medical insurance, and listed address were recorded for each patient visit. Patient visit data was collected from each patient’s face sheet in the EHR. Highest hospital location and EMS usage data were obtained by a review of discharge summaries or ED physician and nurse documentation.
An inter-rater reliability assessment was done on 20 patient visits from each of the 74 datasets. Patient visits were randomly selected and abstracted again by a single abstractor. Data from the abstractors were compared and the inter-rater reliability was assessed using Cohen’s Kappa co-efficient with Mi- ® crosoft Excel® spreadsheet Version 14.0.7116.5000 (Microsoft Corporation, 2010, Redmond, Washington, USA) (See Table 1). Abstractor agreement for each variable (number of agreements divided by the agreements plus disagreements) was illustrated. Only patient visits that concluded in the Emergency Department were utilized for data analysis, all inpatient, observation and outpatient visits to MHH were excluded.
Gender 100 1.00(1.00-1.00) Race 99.8 0.99(0.99-1.00) Medical Insurance 99.1 0.99(0.98-0.99) Visit Type 91.9 0.87(0.85-0.89) EMS Used 88.8 0.81(0.79-0.84) Highest Hospital Acuity 85.4 0.78(0.75-0.81) ‘CI’ refers to Confidence Interval
Table 1. Summary of study statistics and inter-rater reliability agreement.
The abstraction process flow is mapped in Figure 1.
HFD cohort data contained 11,305 entries and dates of transport ranged from August 2005 to January 2007. ‘Entries’ are line items by HFD personnel and confirm an ambulatory phone call was made. Missing patient identification data, such as date of birth or social security number [4(.04%)] and duplicate entries [1,023(9%)] were removed. ‘Patient visits’ reflect patient-encounters to the ED. In some cases, a single patient had multiple patient visits. ‘Patients’ are the individual person present at a patient visit. ‘MHH destination’ refers to MHH. Frequency data reported in Figure 1 reflect the variable described as the numerator and the preceding cell as the denominator.
Abstractor agreements in Table 1 were highest in categories with well-defined, objective criteria such as gender and race (99% to 100% agreement; 0.99-1.00). However, fair abstractor agreements were noted in subjective categories such as visit type, highest hospital acuity (85% to 91% agreement; 0.78- 0.91).
The study population, 815 KE, accounted for 1354 patient visits. Most patient visits were by those of Black race, female gender, and adults 19-44 years (Table 2). Approximately, one-third of ED visits did not utilize EMS services. Fifty-one percent (N=735) of the KE had Medicare/Medicaid. Eighteen percent (N=146) had private insurance. Seventy percent (N=570) had a single ED visit. Thirteen percent (N=106) had more than two visits. Only ~7 % (N=57) had greater than three visits. Merely 0.7% (N=6) utilized the ED more than ten times.
Chronic conditions, including hypertension (ranked st rd 1st) and diabetes mellitus (ranked 3rd), were leading diagnoses among KE visits (Table 3). Non-chronic conditions, including headaches, back pain, chest pain, and abdominal pain were in the top six individual diagnoses. The percent distribution of ICD-9-CM diagnostic code categories when compared to those found in the published National Hospital Ambulatory Medical 30,31 Care Survies (Table 4) showed an increase in 3 categories: Musculoskeletal system & connective tissue (709.3-739.9),
0 to 15 99 7% 15 to 24 389 29% 25 to 44 495 37% 45 to 64 268 20% 65 to 74 51 4% 75 and over 52 4%
Male 470 35% Female 884 65%
Black 1147 85% White 147 11% Asian/South Pacific 19 1% Latino 3 0% Other 38 3%
Yes 815 60% No 519 38% Unknown 20 1%
Medicare Only 69 5% Medicaid Only 577 40% Medicare & Medicaid 89 6% Self-Pay 425 30% Private 268 18% Worker’s Compensation 6 <1%
Table 2. Descriptive analysis of Katrina evacuees.
Endocrine, nutritional & metabolic, immunity (204-279.9) and Circulatory system (388.7-459.19) with a decrease in the Injury & Poisoning category (800-999.9). Compared to the leading primary diagnosis in US Emergency department in 2005 and 2007 (Table 5), spinal disorders (720-724), arthropathies and related disorders (710-719), contusions with intact skin surface (920- 924) and Headache (784) were the leading ICD-9 diagnoses.
Notable public health and medical challenges resulted from Hurricane Katrina, rendering several large hospitals immobilized and KEs with chronic and non-chronic health conditions without access to usual sources of health care. This is one of few studies to 1) analyze summary data captured by community-
1 Hypertension 2 Headache 3 Diabetes Mellitus 4 Back pain unspecified 5 Chest pain 6 Abdominal pain 7 Convulsions 8 Fever 9 Injuring passenger in MVC other than motorcycle E812.1 52 1.6 10 Shortness of breath
Table 3. Top 10 diagnoses among Katrina evacuees.
Infectious & Parasitic Diseases 1 Neoplasms 137 Endocrine, Nutritional & Metabolic, Immunity 240 Blood & Blood-Forming disorders 280 Mental Disorders 290 Nervous System and Sense Organs 316 Circulatory System 388.7 459.19 296 9.2 3.5 3.6 Respiratory System 459.2 519.9 178 5.5 11.0 9.8 Digestive System 520 Genitourinary System 579 Complications of Pregnancy, Childbirth, & the Puerperium 628.2 674.8 108 3.3 -- -- Skin & Subcutaneous Tissue 674.9 709.02 52 1.6 4.0 4.2 Musculoskeletal System & Connective Tissue 709.3 739.9 346 10.7 5.5 6.3 Congenital Anomalies 740 Certain Conditions Originating in the Perinatal Period 760 Symptoms, Signs & Ill-Defined Conditions 779.4 799.9 835 25.9 19.3 21.6 Injury & Poisoning 800 V Codes V01 E Codes E800 E999.1 284 8.8 -- -- -- notes that the value was not listed.
Table 4. Katrina evacuee diagnoses by system and ICD-9 code compared to national data (reference).
based, emergency medical services (EMS) healthcare providers and 2) assess disaster survivors’ healthcare needs in comparison to a national population. Study findings provide, for the first time, the prevalence of EMS and ED utilization of a displaced population as well as the common discharge diagnoses among a displaced population seeking emergent care in a urban large health system post landfall of one of the worst natural disasters in US history, Hurricane Katrina.32 Though different, our findings add to the body of work noted from those that sought treatment in New Orleans.19 Our outcomes support findings from Chicago15 indicating health care providers need to prepare to manage chronic conditions. These conditions may worsen in the absence of healthcare access, full insurance coverage, and suf- 19 33-35 ficient income. Attention to spinal conditions/injuries is also important, as presentation would be different from the typical
401.9 207 6.4 784 88 2.7 250 73 2.3 724.5 72 2.2 786.5 66 2.0 789 62 1.9 780.39 59 1.8 780.6 53 1.6
786.05 51 1.6
136.9 45 1.4 3.0 2.8 239.9 3 0.1 0.2 0.2 279.9 148 4.6 1.5 1.8 289.9 50 1.5 -- -- 315.9 150 4.6 3.5 3.5 388.69 51 1.6 5.1 5.2
578.9 75 2.3 6.0 6.1 628.1 139 4.3 4.6 5.4
759.9 1 0.0 -- -- 779.3 0 0.0 -- --
999.9 356 11.0 24.9 22.5 V86.1 109 3.4 2.6 2.7
1 Abdominal pain 2 Chest pain 3 Contusion with intact skin surface 4 Acute upper respiratory infection, excluding pharyngitis 460–461,463–466 45 1.4 3.7 3.2 5 Spinal disorders 6 Open wound, excluding head 7 Cellulitis and abscess 8 Fractures, excluding lower limb 9 Urinary tract infection, site not specified 10 Sprains and strains, excluding ankle and back 11 Sprains and strains of neck and back 12 Otitis media and Eustachian tube disorders 13 Open wound of head 14 Rheumatism, excluding back 15 Diseases of the teeth and supporting structures 16 Headache 17 Arthopathies and related disorders 18 Pyrexia of unknown origin 19 Acute pharyngitis 20 Asthma
-- notes that the value was not listed
Table 5. The 20 leading primary diagnosis in the ED: United States, 2007.
reasons patients visit the ED.
Findings suggest the majority of the displaced population had a single ED visit, half had government insurance, and two of the top three diagnoses, hypertension and diabetes, were chronic conditions left unmanaged in the wake of the storm. Comparatively, non-chronic issues, body aches/pains and headaches, were likely related to environmental and/or social stressors post disaster.
Study findings contribute to the existing body of literature by providing additional information, aiding in overall understanding of disaster-displaced populations, which is described as critical for a society to prepare and respond effectively.8 The data is representative of a larger displaced population that utilized EMS services prior to presentation to a health system. Primary diagnoses of KE in this study (Table 3) are vastly different from the primary diagnoses reported on a national scale (Table 5). Variance in primary diagnoses suggests the need for published data reporting annual primary diagnoses in local EDs by region. Access to this information via the internet contributes to estimating the likelihood of ED volume of chronic and non-chronic visit demand,1 providing foundational information for disaster preparedness plans nationwide.
System level (i.e. region, county, or parish) diagnoses information should be made publicly available to facilitate prep-
789.0 62 1.9 4.0 4.9 786.5 66 2.0 3.8 4.1 920–924 93 2.9 4.2 3.7
720–724 174 5.4 2.5 3.0 874–897 42 1.3 3.5 2.7 681–682 38 1.2 2.3 2.6 800–819 22 0.7 2.0 2.1 599.0 43 1.3 1.6 1.9 840–844, 10 0.3 2.2 1.8 845.1,848 846,847 50 1.5 2.2 1.8 381–382 13 0.4 1.9 1.8 870–873 28 0.9 1.9 1.7 725–729 71 2.2 1.6 1.7 520-525 10 0.3 -- 1.5 784.0 88 2.7 -- 1.5 710-719 100 3.1 -- 1.5 780.6 53 1.6 -- 1.4 462 16 0.5 1.4 1.4 493 0 0.0 1.5 1.4 All Other reasons 2202 68.3 -- 54.2
aration in disaster recovery efforts. The MedCon:PreEvent tool was developed and published to assist disaster response planning and estimate displaced persons requiring special medical care during a disaster.25 This tool can assist disaster planners with preparing medical needs of disaster evacuees. It provides data needed to re-evaluate the medical approach to use and enhance healthcare servicesprovided.25 Importance of a clear, comprehensive, and well communicated emergency response plan, with capabilities to be tailored to varied types, sizes, and proximities of disaster is supported.15 By having access to local primary ED diagnoses, disaster preparedness planners can prepare to provide survivors with access to preventive care needs for chronic conditions such as hypertension and diabetes. This information is pertinent to cost effective strategies because a population with preventable illness largely addressed by preventive care measures will likely avert use of the ED for non-emergent conditions. It is imperative that provisions for creation and dissemination of 30,31 services and resources for survivors be prioritized by disaster preparedness groups.
Although, this study offers insight on healthcare utilization of KE, the authors hesitate to make generalizations to healthcare use of KE overall. Study findings were subject to selection bias due to the use of HFD run call data. Study team members hypothesize that HFD run data contained some non-
translocated patients; thus, more stringent guidelines were implemented when determining study inclusion criteria. It is possible that the data may include non-translocated patients, as patients may have had out-of-state insurance or gulf coast telephone numbers and present to a MHH only after Hurricane Katrina. Data on what states the participants were from was not available. Sensitivity analysis on the inclusion criteria was not possible as the necessary criteria were not collected during the data abstraction. Study data was obtained by retrospective chart review; thus, the possibility of selection and/or transcription bias is inherent. MHH provided inpatient level care to critically ill and other vulnerable patients who were evacuated directly from hospitals in Louisiana and were not included in this study; thus, the study sample is not truly representative of KE who received care from this hospital. Information about KE that did not utilize HFD services or present to the MHH system are not included in this study. Study data was limited to patients discharged from the ED. Thus, patients that were hospitalized, likely to have higher illness burden and resource utilization were excluded from this analysis. Publicly available, comparative information is limited. Access to nationally available data precluded the option of individual level data; thus, only a frequency analysis was feasible for comparative purpose.
Events associated with Hurricane Katrina revealed major gaps in available disaster preparedness for at-risk medical institutions, especially tertiary and/or quaternary care academic 2,15,36-41 centers. This study’s findings contribute to future preparedness plans by defining pertinent characteristics of a translocated population post landfall of Hurricane Katrina. By eluding to the generalizability of these characteristics, at-risk metropolitan areas at large in this vulnerable region should include this information in their disaster preparedness plans, as lessons learned and suggestions for improvement are currently reported in the 2,11,15,36,42 existing body of literature. Published reports indicate lessons have not been learned, as many recommendations have been repeated or modified;15 thus, issues thwarting recovery efforts must be confronted before the next disaster arises.43 Lessons learned from Hurricane Katrina for medical providers challenge hospitals and health systems to develop a disaster based contingency plan and to become more involved in a state and local response plan.11
The authors declare that they have no conflicts of interest.
The ‘Developing Effective Methods for Addressing Needs of Disaster Survivors (DEMANDS)’ study, HSC-MS-07-0519, was approved by the UT Health IRB as a retrospective chart review; thus, participant consent was not required.