EPI UPDATE

The Department of Health has a home on the World Wide Web at http://www.doh.state.fl.us

For information on diseases and conditions of public health importance go to MyFlorida.com, click on Health and Human Services, then Consumers--Diseases and Conditions.

In this issue:

1. Racial Differences in Mortality among Patients Hospitalized for Meningococcal Disease
2. Influenza Virus Surveillance Summary Update
3. Schools of Public Health Web link
4. Weekly Disease Table

Zuber Mulla, Ph.D., Bureau of Epidemiology

2/26/2002

Recent studies have reported that Black patients hospitalized for various diseases have a lower risk of hospital mortality than white patients even after adjusting for various confounders [1, 2]. Black patients hospitalized throughout Florida for invasive group A streptococcal disease were much less likely than whites to die in the hospital (Odds ratio [OR] = 0.31, p=0.07) [1]. A recent study of racial differences in hospital mortality among men hospitalized in the United States Veterans Affairs health care system found that Black patients admitted for common medical diagnoses such as pneumonia and diabetes had lower mortality rates than White patients [2]. The protective effect of Black race persisted even after adjusting for several factors including diagnosis, age, financial assets, length of stay, presence of comorbid disease (Adjusted relative risk=0.74, 95% CI: 0.64-0.86). The authors considered several explanations for this result but could not identify a definite cause or causes.

These aforementioned studies prompted the current analysis of race and mortality in patients hospitalized throughout Florida for certain invasive meningococcal infections.

A retrospective cohort study was conducted in which race (Black and white) was the exposure and hospital mortality was the outcome.

Public use data on hospital discharges occurring in calendar years 1992 through 1998 were obtained from the Florida Agency for Health Care Administration (AHCA). Principal and secondary diagnoses in the AHCA database are coded using the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM). A case of meningococcal disease was defined as patient whose principal diagnosis code was 036.0 (meningococcal meningitis), or 036.1 (meningococcal encephalitis), or 036.2 (meningococcemia).

Cases that were discharged to another hospital were excluded from the analysis so as to minimize the inclusion of duplicate patient records. Asians, Black Hispanics, white Hispanics, and those of other races were excluded from the analysis.

The final sample size was 839 patient records (150 Black, 689 white).

Statistical analysis

The SAS System (Release 8.02 for Windows) was used to perform simple and multiple logistic regression [3]. Two models were developed, a reduced model and a full model. The reduced model only contained the Race term. The full model contained Race and the following potential confounders: Age (a continuous variable in years), Sex, Diabetic status (a secondary discharge diagnosis code – a comorbid condition), Health insurance (Yes/No), and Length of hospital stay (a binary variable: ³ 5 days compared to <5 days). The length of stay was used as a proxy measure of disease severity.

ORs from the reduced and full models were calculated along with 95% confidence intervals (CI) in the traditional manner.

A total of 511 records had "meningococcal meningitis" recorded as the principal discharge diagnosis. Twenty-two patients had meningococcal encephalitis and 306 patients meningococcemia. The overall mortality was 7.4% (62/839).

Table 1 shows the crude and adjusted ORs (from the reduced and full models, respectively) for mortality among patients hospitalized for meningococcal disease. The crude OR indicates that Blacks might be less likely to die in the hospital; however, the result was not statistically significant. After adjusting for five factors, the OR was closer to the null value of no association. There is a ³ 10% difference between the crude and adjusted ORs, which indicates confounding by the five covariates was present [4].

Table 1. Odds Ratios for Mortality (Blacks compared to whites) in Patients Hospitalized for Meningococcal Disease
(N=839 records)

* Adjusted for age, sex, diabetic status, length of stay, and insurance status.

Discussion

This simple analysis showed no association between race and the risk of mortality in patients hospitalized for meningococcal disease. This AHCA database did not contain information on the medications that were administered to the patients during their hospital stay, e.g., antibiotics and vasopressors. These are important data. If there is a racial difference in the administration of these drugs, then the adjusted OR reported above would be confounded by the treatment regimen.

The adjusted OR reported above may also be confounded by variables such as the patient’s marital status/social support network, income, and educational level. All of these factors affect hospital mortality. If there are racial differences in these variables, then these factors satisfy the definition of a confounder. The AHCA database unfortunately does not contain these data.

A limitation of using hospital discharge data to examine short-term survival of any condition is the possibility of excluding ill individuals who died outside of the hospital, that is, those who died before being admitted to the hospital and those who died soon after discharge. In the current study, the median length of stay of Black patients was 8 days. This figure for whites was 7 days. Had this study found a protective effect of Black race and if Blacks had a shorter length of stay, then one could hypothesize that the reason for the protective effect of Black race could simply be racial differences in discharge practices. Blacks may have been "pushed" out of the hospital prematurely and died elsewhere, for example, at home or in a skilled nursing facility. These deaths would not be captured by a study of hospital mortality.

Administrative databases, such as AHCA and Medicare inpatient datasets, have been used extensively in the past by outcomes researchers; however, they were not created for this use and therefore have several limitations such as miscoded diagnoses. The reader is referred elsewhere for a discussion of this topic [5, 6]. Future researchers interested in the clinical epidemiology of meningococcal disease should collect detailed data from several hospitals.

1. Mulla ZD et al. Invasive group A streptococcal infections in Florida [abstract 61]. 39^th Annual Meeting of the Infectious Diseases Society of America, October 2001, San Francisco, CA.
2. Jha AK et al. Racial differences in mortality among men hospitalized in the Veterans Affairs health care system. Journal of the American Medical Association, 2001; 285:297-303.
3. SAS Institute Inc. The LOGISTIC Procedure. In: SAS/STAT User’s Guide, version 6. Fourth edition. V. 2. Cary, NC: SAS Institute Inc., 1989; 1071-1126.
4. Greenland S. Modeling and variable selection in epidemiologic analysis. American Journal of Public Health, 1989; 79:340-349.
5. Jencks SF et al. Assessing hospital-associated deaths from diagnosis data. Journal of the American Medical Association, 1992; 260:2240-2246.
6. Iezzoni LI et al. Comorbidities, complications, and coding bias. Journal of the American Medical Association, 1992; 267:2197-2203.

For information on the profession of public health and a current list of accredited schools of public health in the United States and Puerto Rico visit the website of the Association of Schools of Public Health: http://www.asph.org/

Carina Blackmore, M.S. Vet. Med., Ph.D.

National report: During week 9 (February 24-March 2, 2002), 463 (24.5%) of 1,886 specimens tested by the World Health Organization (WHO) and National Respiratory and Enteric Virus Surveillance System (NREVSS) collaborating laboratories across the United States were positive for influenza. During the past three weeks (weeks 7-9) the highest proportion of positive influenza cultures (34-36%) were reported from the East North Central, South Atlantic and West South Central region of the United States. Since September 30, a total of 59,035 specimens for influenza viruses have been tested and 8,385 (14.2%) specimens from 50 states were positive. Of the 8,385 isolates identified, 8,231 (98%) were influenza A viruses and 154 (2%) were influenza B viruses. Two thousand four hundred and thirty-eight (30%) of the influenza A viruses were subtyped, 2,412 (99%) were H3 viruses and 26 were H1 viruses. So far this season, CDC has characterized 305 influenza A viruses antigenically. All viruses were similar to the flu A strains in the 2001-2002 vaccine. Influenza B viruses can currently be divided into 2 antigenically distinct lineages, B/Yamagata/16/88 and B/Victoria/2/87. The B component of the current influenza vaccine belongs to the B/Yamagata lineage and is expected to provide lower levels of protection against viruses of the B/Victoria lineage. At this time, 7 of the 36 influenza B viruses CDC has characterized were the B/Victoria strain. In the United States B/ Victoria has been identified in Hawaii, Maryland, New Jersey and New York, but the strain has also been detected in Asia, Canada and Europe. Influenza B/Victoria will replace the B/Yamagata strain in the 2002-03 vaccine. The proportion of patient visits to sentinel physicians for influenza-like illness (ILI) overall was 2.9%, which is above the national baseline of 1.9%. The proportion of deaths attributed to pneumonia and influenza as reported by the vital statistics offices of 122 U.S. cities was 8.9% during week 9. This percentage is above the epidemic threshold of 8.3% for this time. Influenza activity was reported as widespread in 12 states (Arizona, Colorado, Minnesota, Nebraska, New York, South Dakota, Tennessee, Texas, Utah, Vermont, Virginia and Wisconsin), regional in 25 states (Connecticut, Delaware, Georgia, Idaho, Indiana, Iowa, Kentucky, Louisiana, Maine, Maryland, Massachusetts, Michigan, Mississippi, Montana, Nevada, New Hampshire, New Jersey, New Mexico, North Carolina, North Dakota, Ohio, Oklahoma, South Carolina, Washington and Wyoming) this week. Sporadic activity was reported from 13 states.

Florida: Influenza activity, calculated based on the proportion of patients with influenza-like illness (ILI) seeking care by physicians participating in the Florida Sentinel Physicians Surveillance Network was 1.0% this week, the lowest reported activity since Mid-December (week 51). The activity reached a peak (2.7%) in mid-January (week 4). Influenza-like illness activity was detected in 16 of 26 participating counties from Walton to Monroe. Higher flu activity than expected for this time of year (>2%) was reported by physicians in Duval, Monroe, Palm Beach, Polk, and Walton Counties. Three cases of influenza were laboratory confirmed this week. Influenza A (H1N1) was confirmed from Duval (1) County. Influenza A of unknown subtype was detected from patients in Miami-Dade County. Between September 4 and Mar 8, influenza A (H3N2) was isolated from 125 patients residing in Broward, Collier, Duval, Escambia, Hillsborough, Indian River, Lake, Leon, Levy, Marion, Monroe, Osceola, Palm Beach, Pinellas, Polk, Santa Rosa, Sarasota and St. John’s Counties. Influenza A (H1N1) from 3 patients in Duval and Palm Beach Counties and influenza A of unknown subtype was diagnosed in patients in Broward, Gadsden, Lee, Martin, Orange, Pinellas, Palm Beach and Hillsborough Counties. Influenza B has been recovered from patients in Broward (1), Hillsborough (2) and Palm Beach (2) Counties. In addition, positive rapid antigen tests were reported from Duval County, Escambia, Hillsborough, Palm Beach, Lee, Marion, Miami-Dade, Okaloosa, Pinellas and Volusia Counties.

* The column of data representing the "3-year average to week ##" is the average of years 1999, 2000 and 2001 cases to the current listed week (##).