A case was defined as illness in any guest or employee who stayed at the tourist resort during the period July 1–31 and who had diarrhea (three or more loose stools in any 24-hour period) or vomiting (at least one episode) or both, in the same period. Case finding was done by checking records of the resort medical center; after July 20, a door-to-door search was initiated. Demographic data and information on symptoms were collected in face-to-face interviews by the medical staff of the local health unit and the University of Bari.

Because of the high number of cases in resort staff members, a retrospective cohort study was performed to assess risk factors associated with illness in this group. Persons eligible for the study were staff members employed at the resort from July 1 to 31. Standard questionnaires were sent to all 224 staff members in the first week of August. Information requested included name, date of birth, sex, room number, job type, date of onset and type of symptoms, and water and food preferences. A month had elapsed between onset of symptoms and distribution of the questionnaires. We did not inquire about actual food history and activities of staff members during the outbreak but rather about their food preferences and usual activities.

The questionnaires from guests and staff members were returned to the Istituto Superiore di Sanità, where the data were analyzed by using SPSS Base 10.0 (SPSS Inc., Chicago, IL) and Epi-Info 6.04 (Centers for Disease Control and Prevention, Atlanta, GA). Information collected on cases was used to construct the epidemic curve and describe the clinical presentation of the disease. Attack rates, denominator data, personal characteristics, and clinical symptoms of cases were compared between guests and staff members by chi square or Fisher exact test when appropriate; the Mann-Whitney U-test was used for comparisons of age. The room location of ill persons was plotted on a map of the resort that included water pipelines in an attempt to identify any clustering of cases along the pipeline. Statistical test for clustering was performed by the cluster k-means method with SPSS Base 10.0 (SPSS Inc.).

In the cohort study, the attack rate was calculated for the total staff and also by specific job type. Relative risks and 95% confidence intervals were also calculated for job type, behaviors and activities, and food preferences.

From July 18 to 28, samples (28 fecal and 2 vomit specimens) were collected from 30 participants whose illness met the case definition. Part of each specimen was stored at -20°C until examination for viral particles and free fecal cytotoxins, and the rest was refrigerated and processed within 12 hours of collection.

Ova and parasites were detected by direct microscopy, and Salmonella, Shigella, Campylobacter, Yersinia enterocolitica, Staphylococcus aureus, and enteropathogenic E. coli were sought by standard methods (14). The presence of Clostridium perfringens enterotoxin (CPE) was determined either by assaying the cytopathic effect on Vero cells or by reverse passive latex agglutination (RPLA) test (Oxoid Italia Spa, Garbagnate Milanese, Milan) according to the manufacturer’s instructions (14).

Stool and vomit suspensions were examined by NLV-specific reverse transcription/polymerase chain reaction (RT/PCR) with generic primers JV12–JV13 to a consensus sequence on the RNA polymerase segment of the genome shared by most NLV strains (15). For confirmation of the diagnosis, gels were further analyzed by Southern blot with a mixture of NLV-specific probes (15). The 327-bp amplification product was subjected to sequence analysis with PCR primers, and the sequences obtained were aligned with those in the European Molecular Biology Laboratory Nucleotide Data Bank.

After July 13, water samples were repeatedly collected from the main public water supply and from various points outside and inside the resort. Samples from food in the kitchen and the refrigerators were collected and sent to the University of Bari on July 18. Water and food samples were subjected to culture tests for enteric bacterial pathogens, according to standard methods.

Of 344 cases identified from July 1 to 31, 69 (20%) were in staff members. Information on personal characteristics and clinical presentation was available for 248 ill persons (Table 1). Diarrhea, vomiting, and abdominal pain were observed in >70% of all cases. Five patients were hospitalized; all recovered rapidly and were discharged within a few hours. None of the patients had any further sequelae. Attack rates did not differ by age, sex, or symptoms for cases in guests or staff members. For cases in guests, the median interval from the arrival date at the resort and onset of symptoms was 4 days, and symptoms developed in 77% within 5 days.

The epidemic curve shows three distinct peaks in each of the 3 weeks, beginning on July 12 (70 cases), July 18 (26 cases), and July 27 (55 cases). Over the total outbreak period, 275 cases occurred in guests and 58 in staff (Figure 3). Fifty-seven percent of cases in staff members occurred before July 15. The outbreak lasted 24 days, and no cases were observed after July 31.

Because of the rapid turnover at the resort, attack rates for guests were calculated separately for each week: an attack rate of 102 (10.5%) of 970 was observed in week 1; 66 (8.7%) of 760 in week 2; and 105 (10.1%) of 1,034 in week 3. Ill guests occupied 157 of the resort’s 456 rooms. No significant evidence of either clustering by the cluster k-means methods (p=0.392) or increased frequency of cases in rooms near the water pipeline was observed. Attack rates by sex, age group, and week of stay were similar.

For the analysis of risk factors in the cohort study, 181 questionnaires from 224 staff members were completed and analyzed. The attack rate in this group was 69 (38.1%) of 181. The lowest attack rates were observed in staff members who worked in the kitchen or the office, and the highest were in waiters, sports trainers, entertainers, and cleaning staff (i.e., staff members who are have close contact with guests) (Table 2). Staff members who took showers on the beach or consumed drinks with ice were more likely to become ill than those who did not. No association was found between disease and eating any particular type of food or with being at work on July 8–11 (the first days of the outbreak) (Table 3).

Stool samples from 28 patients were negative for ova and parasites and bacterial enteropathogens. Of the 28 stool samples examined by NLV-specific RT-PCR, 22 had an amplified DNA of the size expected for NLV. The 327-bp amplification product was also confirmed for all samples by Southern blot hybridization with NLV-specific probes. Vomit specimens from two other subjects were negative.

A readable common sequence of 290 bp was obtained with sequence analysis and found to be the same for eight samples, indicating a single outbreak virus strain. The sequence was analyzed against the European Molecular Biology Laboratory Nucleotide Data Bank, yielding a best fit with the RNA polymerase sequence of the Lordsdale strain of NLV (16). Nucleotide identity between the two strains was 93.1% (270/290 residues), indicating that the outbreak NLV strain belongs to GGII.

When the stool supernatants stored at -20°C were examined by the Vero cell assay for free bacterial toxins, a CPE consistent with that of C. perfringens enterotoxin was induced by seven samples. The RPLA test confirmed the presence of C. perfringens enterotoxin in all seven samples. The positive specimens had been collected July 18–21, and all were also positive for NLV (Table 4).

All food samples tested were negative for enteropathogenic bacteria. Water samples collected on July 13 from faucets in the bar, the kitchen, and a guest room (Figure 2, point 6) had high levels of coliforms (up to 130 CFU/mL) and fecal streptococci (up to 22 CFU/mL). The same level of contamination was observed in water samples from the pipe connecting the resort to the public water supply (Figure 2, point 2); samples collected from the public water supply outside the resort (Figure 2, point 1) were always negative. After July 15, when chlorine was added to the tank, the level of contamination of tap water inside the resort steadily decreased; no contamination was detectable after superchlorination on July 22.