At 1 year after Hurricane Katrina, the health burden of enrollees increased significantly more versus a comparison group. Emergency department visits and hospitalizations remained elevated.
To assess the effects of Hurricane Katrina on mortality, morbidity, disease prevalence, and service utilization during 1 year in a cohort of 20,612 older adults who were living in New Orleans, Louisiana, before the disaster and who were enrolled in a managed care organization (MCO).
Hurricane Katrina has had profound and ongoing effects on the health of inhabitants of the city of New Orleans, Louisiana, and surrounding areas. Numerous reports on the immediate health effects of Hurricane Katrina have focused on surveillance of the following: mortality, injury, and illness1,2; mental health problems3; health services utilization4,5; disruption to the healthcare delivery system4,6,7; implications for patient care during disasters8; and suggestions for health policy change.9 Although the disaster literature contains many studies on the immediate health effects of natural disasters, few (including the Hurricane Katrina literature) have tracked health outcomes during the long term or have been able to compare health status before an event versus after an event.10,11 Older adults with a heavy burden of chronic conditions require effective care delivery systems12; however, few studies to date have focused on the health effects following a major disruption to care delivery such as occurred with Hurricane Katrina.
The first objective of this study was to document mortality associated with Hurricane Katrina in a defined population of older adults. The second objective was to explore the relationship between prehurricane health risk status and the effects of the hurricane on morbidity and health service use in a population of continuously enrolled older persons from a New Orleans–based Medicare Advantage plan.
The study population consisted of all noninstitutionalized PH enrollees who lived in 4 parishes in the New Orleans metropolitan area served by PH providers (Orleans, Jefferson, St Tammany, and Plaquemines parishes) before Hurricane Katrina. All PH managed care beneficiaries enrolled as of September 1, 2004 (1 year before Hurricane Katrina [n = 27,871]) comprised the cohort for the survival analysis. Enrollees who survived and remained continuously enrolled through August 31, 2006 (n = 20,612) comprised the cohort for the morbidity and utilization analyses. A subset (n = 728) was randomly selected for structured telephone interviews among subjects who had been stratified by health status using prospective adjusted clinical groups (ACGs) (described herein). Half of them were in the highest risk category, a quarter in a moderate risk category, and another quarter in the lowest risk categories. Telephone interviews occurred from December 2006 through February 2007.
We compared the following characteristics of several subgroups of the study population using χ2 test or t test as appropriate: (1) those who lived in heavily disrupted Orleans Parish (n = 7242) with those in more moderately disrupted parishes (n = 13,120) and (2) those surveyed by telephone who completed interviews (n = 303) with those who refused to participate in interviews (n = 425). To assess the generalizability of the total study population to the total population of New Orleans 65 years and older, recently published data on age and sex from US Census 200026 and from the 2005 American Community Survey, New Orleans,27 were used. To assess generalizability of the total study population to a national sample, ACG weights were benchmarked to a nationally representative random sample of Medicare beneficiaries.28
Monthly counts of mortality for 1 year before and 1 year after Hurricane Katrina were compared in survival analysis, controlling for age, sex, and race/ethnicity, and were evaluated using χ2 test. A Cox proportional hazards regression model was developed. Different relative risk parameters were allowed for death happening before versus after Hurricane Katrina, with a change point for the time of the hurricane.
Alternative explanations for the increase in ACG scores during each year were tested. “Coding creep” was ruled out by ascertaining that the direction and amount of change in the score moved in parallel with service utilization. We compared summary morbidity scores before and after Hurricane Katrina by parish and by race/ethnicity using t test to determine statistically significant differences.
To assess the effect of aging alone, we used a data set of enrollees 65 years and older (n = 77,603) from several MCOs that did not experience Hurricane Katrina.29 Using propensity scoring techniques, we created a comparison population matched to the study population by age, sex, and ACG score before Hurricane Katrina. Several weighting schemes were tested, with the goal of showing a similar percentage change from the preyear to the prediction year. At the outset, the ACG score of the unmatched comparison population was 0.80 compared with the study population score of 1.05. Propensity scoring matching for age and sex yielded a comparison group that had a 3.1% change in summary health scores from preyear to postyear compared with the study population change of 12.6% from preyear to postyear. Adding the ACG scoresefore Hurricane Katrina to the propensity scoring resulted in a smaller change in summary health scores from preyear to postyear for the comparison population.
We evaluated differences in prevalences of diseases using McNemar test. The mean preyear and postyear utilizations were compared using paired t test. For the telephone survey population, we defined disruption as positive responses to questions related to events that occurred early or in the long term at the thresholds already defined. Logistic regression analysis was used to determine the odds of self-reported physical health decline given short-term, long-term, and both shortterm and long-term disruption, controlling for age, sex, race/ethnicity, Medicaid eligibility, and health status before Hurricane Katrina health status as measured by the ACG concurrent risk score.
gives characteristics of the study population who were continuously enrolled in PH and of the telephone survey subset. Comparing characteristics of enrollees who lived in the heavily disrupted Orleans Parish with those in the moderately disrupted 3 other parishes revealed several statistically significant differences. Orleans Parish residents were more likely to be older, female, Medicaid eligible, and nonwhite. Nonwhite subjects were overwhelmingly black (94.2%) and did not differ from other nonwhite subjects by Medicaid eligibility or by health risk before Hurricane Katrina.
For the telephone survey, 728 persons were selected in the stratified sample: 303 (41.6%) completed interviews. The 425 persons (58.4%) who did not complete the telephone survey included 23.4% refusals, 13.3% who could not be located, 10.3% ineligible or physically or mentally unable to respond, and 11.4% who were locatable but could not be reached.
Health status of the telephone survey population sampled differed from that of the total study population, reflecting the intentional oversampling of the sickest population. Table 1 gives the concurrent health risk scores before Hurricane Katrina standardized to persons 65 years and older from a national fee-for-service Medicare sample.28 The score of 1.08 indicates that the total study population had an 8% greater health burden during that year than the average Medicare beneficiary. Orleans Parish enrollees showed less disease burden before Hurricane Katrina than enrollees in the other parishes. For the telephone survey population, disease burden was considerably higher, reflecting the sampling strategy.
However, there was little difference between telephone survey respondents and nonrespondents in the health risk scores before Hurricane Katrina (1.80 and 1.75, respectively). There were sex and age differences. Respondents were more likely to be female (59.4% vs 48.2%) (P = .003) and were younger: 51.8% versus 46.4% were aged 65 to 74 years, 42.6% versus 40.9% were aged 75 to 84 years, and 5.6% versus 12.7% were 85 years and older (P = .006). There were no differences in the proportions of respondents versus nonrespondents who lived in Orleans Parish before Hurricane Katrina.
Disenrollment for the year before Hurricane Katrina was 4.2% and rose to 11.3% (n = 3072) in the year after Hurricane Katrina. Those who disenrolled were older (mean age, 75.6 vs 74.8 years) and had higher summary health risk scores in the years before Hurricane Katrina (7.1 vs 6.5), and greater proportions were female (64.6% vs 58.4%), nonwhite (50.6% vs 33.8%), Medicaid eligible (15.0% vs 8.9%), and residents of Orleans Parish (57.6% vs 35.6%); all differences were significant at P <.001.
with men was 0.63 (95% CI, 0.57-0.70). The relative risk of mortality was not significantly different by race/ethnicity or by parish. Cause of death was not available.
gives morbidity scores by race/ethnicity and by parish standardized to the total study population before Hurricane Katrina (adjusted for age and sex) and by whether individuals were displaced out of state at any time in the year following Hurricane Katrina. The pattern of the scores indicates that morbidity increased after Hurricane Katrina for all race/ethnicity and parish groups, with a 12.6% increase overall but with a 15.9% increase for nonwhite subjects in Orleans Parish. This change contrasts with a 3.4% increase in morbidity in 1 year among the matched sample of Medicare managed care enrollees. Nonwhite subjects in Orleans Parish had scores indicating less morbidity before Hurricane Katrina than white subjects in the same parish and maintained lower scores, despite a greater morbidity increase from before the hurricane to after the hurricane. The smallest increase was for nonwhite subjects in the other 3 parishes (4.8%). For all subgroups, those displaced experienced greater increases in morbidity compared with those not displaced.
Among the continuously enrolled population, gives the prevalences of treated morbidities per 1000 members and the percentage changes from before the hurricane to after the hurricane. Although the absolute numbers are small, the change in prevalence of cardiac arrest was highest (135.3%) among the conditions hypothesized to increase. Other conditions such as acute myocardial infarction and sleep problems also demonstrated significant increases in prevalence. Several conditions hypothesized a priori not to increase in prevalence (the control conditions) experienced nonsignificant change.
In response to questions regarding disruption in the first 4 months following Hurricane Katrina, the greatest concern reported was damage to the home, with 68.6% reporting a house moderately or severely damaged or destroyed. Just over half (52.1%) reported that their house was unlivable for more than 3 weeks. Only 5.9% reported having trouble getting healthcare in the first 4 months after Hurricane Katrina. Fewer respondents reported that disruption remained at 1 year. Approximately 28% said that their house remained unlivable or destroyed or that their financial situation was worse. Few respondents (5.6%) reported early or long-term problems getting healthcare.
gives telephone survey responses to questions regarding early and long-term personal disruption. The most common early effect was damage to respondents’ homes. Markers of long-term disruption were less frequently reported. Survey respondents who experienced short-term and longterm disruption had more than 3 times greater odds of reporting worse health (odds ratio [OR], 3.6; 95% CI, 1.4-9.5). Higher K6 scores, which are indicative of distress, were also significant (OR, 1.1; 95% CI, 1.1-1.2). Other control variables (age, sex, race/ethnicity, Medicaid eligibility, and health risk before Hurricane Katrina) were not significant.
particularly those from Orleans Parish, had dramatic increases in morbidity scores compared with those who were not nondisplaced. This may be an artifact of the morbidity measure, which depended in part on utilization, and may indicate greater access to healthcare or more thorough diagnostic workups after the disaster. It may also reflect differences in practice patterns between physicians seeing patients known to them in the setting of a Medicare managed care plan versus physicians seeing newly displaced older patients with multiple chronic conditions previously unknown to them in a fee-forservice encounter.
Although there was extensive hurricane damage to 2 key network hospitals, PH had signed 3 new local hospital agreements, which brought the number of contracted facilities up to 6 in the original service area within 1 year. For those displaced outside of the New Orleans area, the efforts made by PH to contact enrollees after Hurricane Katrina and to assure out-of-network providers of payment may have attenuated worse health consequences. The increased utilization suggests that the decision by the Centers for Medicare & Medicaid Services to require Medicare Advantage plans to continue payments without out-of-service area penalties was an important factor in providing ongoing access to care in the face of high and unexpected relocation following the disaster.
Our findings about the effect of Hurricane Katrina on chronically ill patients add a valuable perspective to that reported earlier by Kessler and colleagues,5 whose focus was on a population of persons who had contacted the American Red Cross or were defined as eligible for assistance from the Federal Emergency Management Agency. Many (30%-40%) reported that they had no health insurance, and there were few older persons (245 responses from those χ60 years) in that population. Diseases and service utilization were self-reported. Our study population originally lived in several parishes, one third in the hardest hit parish of Orleans, affording a comparison of the health effect between those most heavily affected by the disaster and those less affected. All had health insurance through the MCO. Documented service utilization remained high after the first month.
Results from the small sample of telephone-interviewed PH enrollees mirrored the morbidity finding in the larger study population and provided a link between self-reports of heavy disruption and health decline. However, the low response rate to the telephone survey introduced the possibility of selection bias yielding reports of health outcomes and disruption effects that did not represent those of the larger population. This possibility is somewhat attenuated by the fact that respondents and nonrespondents had similar health status before Hurricane Katrina, as well as that similar proportions came from Orleans Parish, presumably experiencing similar disruption. Locating persons after a disaster is a well-known problem. That only 13.3% could not be located is positive and testifies to the close contact maintained between enrollees and the health plan. The low proportion that reported having trouble getting healthcare in the first 4 months following Hurricane Katrina or in the longer term may be unique to the telephone survey respondents. Alternatively, it may relate to the fact that housing issues were so overwhelming or may be a tribute to the efforts made by the MCO to keep their medical system intact and to notify patients displaced from the New Orleans area that the costs of care would be covered by PH.
The study has several major strengths. First, the study focused on a continuously enrolled population with health data available on all subjects before and after Hurricane Katrina. Second, we utilized ACG methods, which provided a rigorous and well-established summary measure of health status, enabling comparisons before and after the event. The MCO continued operating with only minor interruption throughout the period following Hurricane Katrina, lending credibility to the complete capture of claims data. Third, unlike most studies in the disaster literature, this study focused on older adults and had l year of follow-up. Fourth, the mortality measure was reliable and accurate.
Several limitations prevent us from generalizing our findings to the entire area affected. We did not have a comparison Medicare population from New Orleans. In lieu of health data on the general population of New Orleans, we used benchmarking to a national older population and a propensity scoring technique to estimate year-to-year health decline in an unaffected population. Differences in outcomes may have been related to the uniqueness of this managed care population. There is some evidence in the literature that Medicare Advantage plan enrollees are healthier.32 In addition, voluntary disenrollment of patients from the MCO may have skewed the findings because more of those who disenrolled were from Orleans Parish and more likely to be nonwhite, female, and enrolled in Medicaid. Both of these phenomena would have led to the study population’s appearing healthier, suggesting that the total population may have had even greater decline in health.
Diagnostic codes, on which the ACG method is based, depend in part on utilization and access to care and on diagnostic coding habits of health service providers. However, several methods were used to assess alternative explanations for increases in morbidity and the effects of aging in this continuously observed cohort.
Other problems experienced by the study population such as mental health issues documented elsewhere3 and disruption to social networks and living arrangements—all of which may have an effect on the health of older individuals who rely on others for assistance—were beyond the scope of this analysis. However, these issues need to be considered in understanding the long-term effects of a disaster on the health of older persons.Acknowledgments
firstname.lastname@example.org. Louisiana Department of Health and Hospitals Web site. Death rate in New Orleans area drops slightly following Katrina. 2007. http://www.dhh.state.la.us/news.asp?Detail=1112. Accessed July 1, 2007.
3. Kessler RC, Galea S, Jones RT, Parker HA; Hurricane Katrina Community Advisory Group 2006. Mental illness and suicidality after Hurricane Katrina. Bull World Health Organ. 2006;84(12):930-939.
5. Kessler RC; Writing Committee for the Hurricane Katrina Community Advisory Group, 2007. Hurricane Katrinaâ€™s impact on the care of survivors with chronic medical conditions. J Gen Intern Med. 2007;22(9):1225-1230.
7. Rudowitz R, Rowland D, Shartzer A. Health care in New Orleans before and after Hurricane Katrina. Health Aff (Millwood). 2006;25(5):w393-w406.
9. Lambrew JM, Shalala DE. Federal health policy response to Hurricane Katrina: what it was and what it could have been. JAMA. 2006;296(11):1394-1397.
11. Yzermans CJ, Donker GA, Kerssens JJ, Dirkzwager AJ, Soeteman RJ, ten Veen PM. Health problems of victims before and after disaster: a longitudinal study in general practice. Int J Epidemiol. 2005;34(4):820-826.
13. Super N, Biles B. Displaced by Hurricane Katrina: issues and options for Medicare beneficiaries. Henry J. Kaiser Family Foundation Medicare Policy Brief. 2005. http://www.kff.org/medicare/7437.cfm. Accessed April 2007.
15. Starfield B, Weiner J, Mumford L, Steinwachs D. Ambulatory care groups: a categorization of diagnoses for research and management. Health Serv Res. 1991;26(1):53-74.
17. Fowles J, Weiner J, Knutson D, Fowler E, Tucker A, Ireland M. Taking health status into account when setting capitation rates: a comparison of risk adjustment methods. JAMA. 1996;276(16):1316-1321.
19. Kario K, McEwen BS, Pickering TG. Disasters and the heart: a review of the effects of earthquake-induced stress on cardiovascular disease. Hypertens Res. 2003;26(5):355-367.
21. Leor J, Poole WK, Kloner RA. Sudden cardiac death triggered by an earthquake. N Engl J Med. 1996;334(7):413-419.
23. Logue JN, Hansen H. A case-control study of hypertensive women in a post-disaster community: Wyoming Valley, Pennsylvania. J Hum Stress. 1990;6(2):28-34.
25. Kessler RC, Andrews G, Colpe LJ, et al. Short screening scales to monitor population prevalences and trends in non-specific psychological distress. Psychol Med. 2002;32(6):959-976.
27. US Census Bureau Web site. American Community Survey (ACS). Data tables: 2005 Gulf Coast area data profiles: Louisiana, inside FEMA designated IPA area data profiles.
28. Centers for Medicare & Medicaid Services, 1998. Standard Analytic File, 5% sample. CMS Data User Reference Guide, ResDAC. http://www.resdac.umn.edu. Accessed January 20, 2006.
30. Holohan J, Ghosh A. Dual Eligibles: Medicaid Enrollment and Spending for Medicare Beneficiaries in 2003, Kaiser Commission on Medicaid and the Uninsured. Washington, DC: Kaiser Commission on Medicaid and the Uninsured; July 2005.
32. Hellinger FJ, Wong HS. Selection bias in HMOs: a review of the evidence. Med Care Res Rev. 2000;57(4):405-439.