Cryptosporidium spp. were detected in the specimens and differentiated by PCR and restriction fragment length polymorphism analysis of the small subunit rRNA gene (11). C. hominis was subtyped by sequence analysis of the 60-kDa glycoprotein gene (12). Each specimen was analyzed at least 2× by PCR, with positive and negative controls in each run. Prevalence rates and 95% CIs were computed; the χ² test was used to test differences. Odds ratios (ORs) and 95% CIs were calculated.

Among the 6,284 patients, 102 were positive for Cryptosporidium spp.: 90 from hospital I (2.8%, 95% CI 2.2–3.3), 3 from hospital II (0.6%, 95% CI 0–1.3), and 9 from hospital III (0.4%, 95% CI 0.1–0.6) (p<0.01). Ward assignment was available for 1,592 of 3,245 patients in hospital I. In most of the 12 wards, the infection rate was 0%–2.3%; in ward A, it was 51.4% (p<0.01) (Table).

In hospital I, children <6 months old had a significantly higher positive rate (8.4%, 95% CI 5.6–11.2) than older children (1.9%, 95% CI 1.4–2.4) (p<0.01; data not shown). This was mainly because of a high infection rate among the age group in ward A (61.5%, 95% CI 36.9–86.2) versus those in other wards (40.0%, 95% CI 19.0–61.0). No age-associated difference in infection rates was found in other wards (p = 0.80; data not shown).

Cryptosporidiosis was more prevalent during February–July 2008 (p<0.01). Prevalence rates remained at ≈6% in the monthly distribution of the 2 main C. hominis subtypes in hospital I; however, when adequate numbers of patients were sampled, rates of Cryptosporidium infection in ward A remained >28% in most study months.

C. hominis was identified in 90.2% (92/102) of Cryptosporidium-positive patients in the 3 hospitals, of whom 86 were patients in hospital I. In hospital I, C. hominis was detected only in ward A; C. meleagridis was isolated from 4 patients in other wards (Table). In contrast, C. canis (1 case) and C. hominis (2 cases) were identified in hospital II patients, and C. canis (1 case), C. hominis (4 cases), C. felis (2 cases), and C. meleagridis (2 cases) were identified in hospital III patients (Table).

Six C. hominis subtypes were found at the 3 hospitals; 4 were in 73 specimens from hospital I (Table). Of those 73 specimens, 71 (97.3%) were subtype IaA14R4 or IdA19, and they were mostly found in ward A and unknown wards (Table). Other subtypes (IbA19G2 and IdA14) were not found in ward A (Table). With 1 exception, subtypes in hospital I were not found in other hospitals; subtype IaA14R4 was found in 2 patients in hospital III. Likewise, subtypes IaA18R4 (in 1 patient in hospital II) and IgA14 (in 1 patient in hospital III) were not found in hospital I.

The source of the cryptosporidiosis outbreak is unknown. Most of the 12 wards in hospital I were located in the main building; ward A, the smallest ward, was in an adjacent building and was for children from a welfare institute. Hired caregivers cared for children in ward A; family members were the primary caregivers for patients in other wards. Thus, poor diaper-changing and hand-washing practices by caregivers could be responsible for the persistence of C. hominis infections in ward A. However, the facts that most of the patients were examined for Cryptosporidium infection only once and that many of the specimens were not submitted immediately after patients were hospitalized prevented us from concluding with certainty whether the infections were acquired in the hospital or in the welfare institute. The likelihood for widespread foodborne and waterborne transmission of cryptosporidiosis in hospital I was small because children in ward A and other wards shared the same source for food and drinking water. The likelihood of direct transmission of cryptosporidiosis among ward A patients was also small because 80% of patients were <1 year old and mostly stayed in cribs and beds.

This cryptosporidiosis outbreak has several key features. First, it was lengthy, lasting >14 months (November 2007–December 2008); only limited sampling was done before November 2007; and Cryptosporidium spp. were still present in December 2008. The longest previous outbreak was 4 months (14). Second, the number (>38) of involved patients was high. Judged by the low occurrence of the 2 subtypes in other wards, most of the 32 IaA14R4- and IdA19-positive patients with missing ward information were probably also from ward A. Thus, >60 children might have been part of the outbreak. Third, this outbreak was caused concurrently by 2 C. hominis subtypes, of which IaA14R4, but not IdA19, was significantly associated with diarrhea. The observed difference in virulence is consistent with data from a community study in Peru (15), in which subtype family Ia, of which IaA14R4 is a member, was more virulent than Id, of which IdA19 is a member.