It has been observed repeatedly in the United States that during influenza epidemics the number of deaths from pneumonia and influenza exceeds expected values for several weeks (1-3). Therefore, as part of the regular influenza surveillance system, CDC obtains from 121 U.S. cities reports of all deaths due to pneumonia and influenza in order to assess the extent and impact of influenza activity. The following is a description of CDC's system of collecting, analyzing, and interpreting data on deaths due to pneumonia and influenza.

Each week, 121 cities throughout the United States relay mortality data by postcard to the National Morbidity and Mortality Statistical Activity of the CDC. The numbers of deaths occurring in these cities are reported separately for all causes, influenza, and pneumonia. A death is attributed to pneumonia if it appears on Part 1 of the death certificate as an immediate or underlying cause of death. An influenza death is reported if influenza appears anywhere in Part 1 or 2 of the certificate, taking precedence over other causes of death. The data are reported for the following age groups: under 28 days, 28 days to 1 year, 1-14 years, 15-24 years, 25-44 years, 45-64 years, and greater than 64 years. Deaths are reported to CDC by the number reported to the health departments the previous week rather than by date of occurrence.

The reported numbers of deaths are shown as dots connected by line segments (Fig. 1). The solid line for each mortality category is the expected number of deaths. The broken line is the epidemic threshold, the derivation of which is described below, which is set at a point where significant deviations from the expected number can be seen. The charts are drawn to a scale that allows the distance between the expected and threshold levels to be con

stant for every curve. This device allows one to compare the influenza activity between regions by glancing at the regional chart. Although the vertical labels are different, comparison of the absolute distance on the chart between observed and threshold levels between regions shows whether the mortality is significantly higher in 1 region than another. This is accomplished by allowing 0.3 inches on the original full-size chart to represent 1.65 standard errors of measurement for each graph that is drawn.

Each year before the respiratory disease season begins, equations to describe the expected number of deaths are generated for deaths due to all causes by age group and deaths due to either pneumonia or influenza by region. The expected mortality level is calculated by using weekly data for the previous 5-year period, omitting epidemic weeks, and fitting a regression model which includes terms to compensate for a general upward or downward linear trend, a semiannual cycle (26 weeks), and an annual cycle (52 weeks) about a general mean value. The equation used (shown below) derives the expected mortality level as a function of week number (t). All other values in the equation are constants solved for by the regression equation:

=u+r++A]Coo룸플++ Boln 룸플+ + A2000 + B2017 꿀.

These equations are then used to evaluate current mortality statistics reported by the 121 cities in terms of previous reporting trends or expected number of deaths. Except for resulting in a slightly smoother curve and yielding a standard error, which forms the basis of the epidemic threshold, the regression procedure may be thought of as a means of averaging the deaths for corresponding weeks over the 5-year period and using the average as the expected number for the next year.

Influenza - Continued

The epidemic threshold is used to identify weeks in which the difference between the observed number of deaths and expected number of deaths cannot be explained by chance (4). Ninety-five percent of all the weekly reports should, according to the model, fall beneath the threshold line, which is 1.65 standard errors (SE) above the expected values (2-tailed test). Since 1 in every 20 weeks by chance alone is expected to pass through the threshold, the definition of an epidemic requires 2 or more successive weeks greater than 1.65 SE away from the mean. The probability that this would occur by chance alone is approximately 1 in 400 (0.052).

The statistical model used in the analysis of mortality data generates expected values so that the sum of the differences between the expected and observed values will equal zero after epidemic weeks are eliminated. Two or more successive weeks falling above the threshold is interpreted as indicating a mechanism other than chance alone. Experience has shown that when the observed numbers exceed the epidemic threshold for 2 or more weeks, this reflects substantial influenza activity occurring in the country or region.

Excess deaths caused by an influenza epidemic are calculated from this equation as the difference between the ob

Epidemiologic Notes and Reports

Influenza - Wisconsin

Three isolates made at the University of Wisconsin from a 22-year-old man have been confirmed by CDC as influenza A/New Jersey/76. The isolates came from throat washings from the patient, who had a mild influenza-like illness beginning on November 24. An investigation is in progress to determine if there has been secondary spread of the virus to any of his contacts. Specimens were obtained from swine on this man's farm, and A/New Jersey/76 influenza virus was isolated from 6 of the 8 clinically ill swine. There has been slight seasonal increase in upper respiratory illness in the community but only sporadic cases of influenzalike illness. Extensive ongoing laboratory virus surveillance systems in Wisconsin that screen more than 60 specimens

served and expected values for those weeks falling above the threshold line when 2 or more consecutive weeks are above the threshold line.

It should be emphasized that this method of influenza surveillance is based on data from 121 urban centers in the United States, most of whose populations exceed 100,000 and whose total population constitutes approximately onethird of the U.S. population. Thus, these numbers represent only an index of mortality due to pneumonia and influenza, but they nevertheless serve as the most readily available and sensitive indicator of influenza activity for the country. As can be seen on Figure 1, through week 47 there is no evidence of excess deaths due to pneumonia and influenza for the country.

Reported by National Morbidity and Mortality Statistical Activity, Bureau of Epidemiology, CDC.

References

1. Collins SD, Lehmann J: Excess deaths from influenza and pneumonia and from important chronic diseases during epidemic periods, 1918-51, Public Health Monogr 10 (PHS Publication 213). Washington, Government Printing Office, .1953

2. MMWR 14:8-11, 1965

3. Center for Disease Control: Influenza Surveillance Report No. 90, 1973-1974 and 1974-1975. Issued February 1976

4. Serfling RE: Methods for current statistical analysis of excess pneumonia-influenza deaths. Public Health Rep 78:494-506,1963

and Washington, D.C.

per week from persons with upper respiratory illness have not recovered other influenza viruses.

An influenza B/Hong Kong/72 virus isolated from a nasopharyngeal specimen from a 34-month-old child in the Washington, D.C., area is the only other current evidence of influenza activity in the United States.

Reported by R Pawlisch, DVM, Brodhead, Wis; B Easterday DVM, PhD, Dept of Veterinary Medicine, DB Nelson, MS, State Laboratory of Hygiene, University of Wisconsin; HG Skinner, MD, State Epidemiologist, Wisconsin Dept of Health and Social Services; Martin E Levy, MD, State Epidemiologist, District of Columbia Community Health and Hospital Administration; and the National Influenza Immunization Program, CDC.

Two hundred eighty-four cases of trichinosis, including 1 fatality, were reported to CDC in 1975. This annual total is nearly 2.5 times higher than the mean number of cases reported during the previous 5 years. It also represents the highest annual incidence since 1961. The increase is associated with an unusually large number of common-source outbreaks.

While 24 states, the District of Columbia, and New York City reported at least 1 case, 85.6% (243) of the cases orignated from only 7 states lowa, New Jersey, Illinois, Alaska, Connecticut, Massachusetts, and New York. The incidence of reported cases was 1.3 per million population for the entire United States.

Pork, mainly prepared as sausage, continued to be the major source of trichinosis infections in the United States; it accounted for 73.2% of the 254 cases for which a source was identified. Non-pork products were identified as the probable source of infection for the remaining 68 cases. Meat from wild animals (walrus and black bear) was responsible for 33 cases, and ground beef was identified as the probable source of infection for 34 cases. Although cattle are not considered a natural reservoir of Trichinella spiralis because they are herbivorous, beef products may be adulterated with pork through intentional or inadvertent mixing of beef and pork. Twenty-eight of the cases attributed to ground beef consumption were reported from New Jersey - a marked increase over previous years. To determine the frequency of pork adulteration of ground beef, the New Jersey State Health Department conducted a survey of retail meat markets in the state. It found evidence of pork in ground beef samples purchased at 3 of 4 stores which had been the suspected sources of products associated with cases and at 5 of 10 stores with no such association.

For 241 cases, the place where the suspect meat was obtained was reported. Of these, the source of meat for 181 cases (75.1%) was a supermarket, butcher shop, or other commercial outlet. Twenty-six (10.8%) patients had consumed the suspect meat items at a restaurant or other public eating place; illness in 18 of these cases was attributed to ground beef. The walrus and the black bear which accounted for 33 cases were obtained by hunting or trapping. No cases were attributed to meat obtained directly from a family farm.

Twenty separate common-source outbreaks, each in

volving 2 or more cases, accounted for 69.4% of all cases reported in 1975 (Table 1). Ingestion of trichinae-infected walrus meat caused 28 cases in Alaskan Eskimos in what appears to have been the first reported outbreak caused by this host in North America. The meat items implicated as the causes of 10 of the outbreaks involving 132 cases were pork products, usually prepared as sausage. Ground beef was implicated as the cause of 5 outbreaks, including 1 involving 15 persons who became ill after eating rare hamburgers at a country club restaurant in New Jersey.

Common-source outbreaks due to pork products revealed that persons of German, Italian, Polish, or Portuguese ethnic ancestry were represented in disproportion to their numbers in the general United States' population. That certain national ancestry, particularly German or Italian, increases the risk of exposure to the parasite has been previously observed in the United States (1) and has been explained as a result of these group's preferences for pork products, expecially sausage, prepared in a traditional manner which often does not result in the destruction of T. spiralis.

Pork Spare Ribs

Ground Beef

Unknown

Pork Sausage

Unknown

Ground Beef Pork Sausage