A detailed description of study participants, population, and methodology is reported elsewhere (20). In brief, health department records were reviewed for all culture-confirmed patients who met the surveillance case definition (21) and were reported from the seven sites (Arkansas, California [six counties], Maryland, Massachusetts, Michigan, New Jersey, and Texas [four counties]) from January 1996 through December 2000. Contacts (of culture-confirmed patients in the sentinel areas) with active TB were identified through routine public health investigations, as defined by local contact and source-case investigation policies and practices at each study site. Source case investigations were undertaken for all patients <5 years of age. Two sites also routinely performed source case investigations on children <5 years of age. Information about epidemiologically related patients identified from public health investigations was gathered with a standardized data collection form that included the direction of transmission (i.e., whether the contact was a source patient or secondary patient in relation to the index patient or whether the direction of transmission was unknown), the relationship between patients (shared a household, nonhousehold friends or relatives, co-worker, or common source), and the exposure setting (correctional, school or day-care center, workplace, emergency shelter, group quarters, hospital, nursing home, other long-term care facility, or other setting). Data were entered into Epi Info version 6d (22) databases and routinely sent to CDC. State TB registry numbers for patients in the multisite genotyping network database were matched against the CDC’s national TB surveillance registry to obtain sociodemographic, behavioral, clinical, treatment, and drug-susceptibility information, which is routinely reported for all TB patients on the Report of Verified Patients of Tuberculosis (23). Project activities described in this paper were determined by CDC’s institutional review board to be exempt from full committee review since genotyping of isolates was considered a public health surveillance activity and all other data used in the analysis of this paper were previously collected.

Genotyping of M. tuberculosis isolates was conducted in accordance with standardized study protocols (20). IS6110-based restriction fragment length polymorphism (RFLP) analysis was performed on all available isolates. Because low-copy numbers of IS6110 reduce test specificity, isolates containing six or fewer IS6110 copies were further analyzed by spacer oligonucleotide typing (spoligotyping) (24). Patients were determined to have concordant genotypes if their isolates contained seven or more IS6110 bands with identical patterns or six or fewer IS6110 bands with identical patterns and matching spoligotypes.

Our investigation focused on culture-confirmed patients <5 years of age; TB in young children represents recent transmission, and source patient investigations are routinely conducted for this group. A source patient was defined as a confirmed TB patient who was identified by chart abstraction as the likely source of infection for another reported TB patient. A secondary patient was defined as a confirmed TB patient who was infected by an identifiable source. Epidemiologically related source patients and secondary patients identified through routine public health investigations were considered suspected patient pairs. Because some source patients transmitted TB to more than one secondary patient, the number of suspected patient pairs does not equal the number of source patients. A secondary patient, however, could have only one designated source patient.

Genotypes for isolates from suspected patient pairs were compared, and patient pairs were categorized as 1) confirmed patient pairs, if isolates had concordant genotypes 2) refuted patient pairs, if isolates had discordant genotypes, and 3) undetermined patient pairs, if genotypes were unavailable for the patient pair.

Data in the multisite genotyping network database were analyzed with SAS version 8.0 (25) and Epi Info version 6d (22) software packages. Patients were excluded from analysis when records were not available for review or lacked complete information from public health investigations, including three patients <5 years of age from one site, for whom source patients were not identified but who were entered into the database as the source for an adult case. Because young children are not typically considered to be infectious (26) and records for these patients were not available for further examination at the time of this analysis, information was determined to be incomplete for these patients.

Univariate analysis was conducted to examine factors associated with the identification of source patients for young children and to investigate associations between key variables and the identification of refuted patient pairs. Differences in proportions were assessed with the chi-square statistic or 2-tailed Fisher exact test. Relative risks (RR) and 95% confidence intervals of point estimates were generated where appropriate. Differences in the means of continuous data were tested with the Wilcoxon rank-sum test when sample sizes were small. Unless otherwise noted, p values <0.05 were interpreted as statistically significant differences for all statistical tests.

Genotypes of isolates from young children without a known source patient were matched against the genotyping network project database to find previously unidentified adult TB patient(s) whose genotype matched the child’s. Since the sentinel study sites represented geographically dispersed states that did not necessarily share a common border, genotype matches were limited to patients from the same site.

To better describe the unique characteristics of patients who transmit TB to young children, source patients (in confirmed patient pairs) who transmitted TB to young children were compared with those who transmitted to adolescents or adult patients. Since source patients who infected children 5 years of age or older may be very similar to source patients who infected children newly born to 4 years of age, two different comparison groups were identified 1) source patients for all secondary patients ≥5 years of age and 2) source patients for secondary patients ≥15 years of age (excluding source patients that transmitted to children 5–14 years of age).

From 1996 to 2000, a total of 15,035 TB patients were reported from the seven sentinel surveillance sites; 11,923 (79%) were culture confirmed, and isolates from 10,752 (90%) culture-confirmed patients were genotyped. Of all patients in the study, 518 (3%) patients were <5 years of age. Culture was attempted in 270 (52%) patients <5 years of age, and 122 (45%) of these patients were culture confirmed. Isolates from 114 (93%) culture-confirmed children <5 years of age were genotyped.

Texas and California sites reported 73 (60%) of the 122 culture-confirmed patients <5 years of age; the Michigan and New Jersey sites reported 18 patients each, and the remaining three sites reported ≤6 patients each. Most (65%) of the study patients were <2 years of age, and 49% were girls. Forty-three percent were black, non-Hispanic; 37% Hispanic; 15% Asian; 4% white, non-Hispanic, and 2% Native-American or Alaskan Native. Of the 11 foreign-born patients <5 years of age, 4 were from Mexico, 2 were from Kenya, and 5 were from other countries. Two thirds of the young children had pulmonary TB disease, 15% had extrapulmonary disease, and 20% had both pulmonary and extrapulmonary TB.

With some notable exceptions, culture-confirmed patients <5 years of age had demographic or clinical characteristics similar to those of the 396 young children from the surveillance area who were either culture-negative or did not have a specimen collected for culture. The culture-confirmed group was more likely to be ≤1 year old (RR=1.94, p=<0.001); whereas white, non-Hispanic children (RR=0.41, p=0.02) and those treated only by private providers (RR=0.6, p=0.002) were underrepresented in the sample of culture-confirmed patients.

Results of routine investigations used in identifying source patients for culture-confirmed children <5 years of age are presented in Figure 1. Health department records were unavailable or lacked sufficient information about investigations of 11 patients; these records were excluded. At least one epidemiologically related case was identified for 66 (59%) young children with culture-confirmed TB; 57 (86%) patients had a source patient designated, but a source could not be determined for the remaining 9 patients, although an epidemiologically related case was identified. For five of the nine patients, multiple epidemiologically related patients (ranging from 2–11 related patients) were identified.

To examine factors associated with the identification of source patients for culture-confirmed children <5 years of age, we compared young children with a suspected source patient to patients with an unknown source of infection (Table 1). Children <2 years of age were more likely to have a source patient identified from routine public health investigations; however, source patients were less frequently found for foreign-born children. No other statistically significant differences were found. Drug-susceptibility patterns for isolates from young children with any drug resistance are detailed in Table 2.

Table 3 lists characteristics of the 53 source patients identified from public health investigations. In 41 (72%) of 57 suspected patient pairs involving young children, the source patient lived in the child’s household. Of the 16 nonhousehold sources, 3 were babysitters, 4 were neighbors or visitors, 2 were relatives, and 1 attended the same church as the child’s family; the specific relationship was unknown for 6 patient pairs. Eight (15%) of the source patients resulted in disease in more than one young child (including culture-negative children and patients outside of the study population).

Of the 57 culture-confirmed patients <5 years of age for whom a source patient was identified, 91% (52) had genotyping results for both the young child and the suspected source patient (Figure 2). Forty-four (85%) of 52 suspected patient pairs had concordant genotypes, and 8 (15%) of 52 had discordant genotypes. Young children in refuted patient pairs were more likely to be older than those in confirmed patient pairs (Table 4). No association between gender, ethnicity, or foreign-born status of patients and the identification of refuted patient pairs was found. Nearly three quarters (37 of 52) of suspected patient pairs lived in the same household; however, 5 (14%) of these patient pairs had discordant genotypes. Suspected patient pairs with differing drug-susceptibilities were not associated with discordant genotypes; all three patient pairs with differing drug resistance patterns had concordant genotypes (Table 5).

For the nine young patients who had at least one epidemiologically related patient identified by public health investigations but for whom the source patient could not be determined, genotyping patterns from the isolates of the epidemiologically related cases and the young child were identical, almost without exception. The only two discordant genotypes were in young children with a single related case, not among the five young children with multiple related patients.

Genotyping also identified patients in the local surveillance site who had the same genotype as young children without an identified source patient. Isolates were genotyped from 40 of 45 patients <5 years of age without a known source patient. Of these genotyped isolates, 23 (58%) matched the strain from at least l adult pulmonary TB case in the local surveillance site. For most young children (13 [57%] of 23) without an identified source patient, at least 5 adult pulmonary TB patients with genotypes matching the child’s were identified. We found a wide range in the number of adult patients (2–128) with genotypes matching the genotypes of these young children.

To better characterize the unique attributes of patients who transmit TB to young children, characteristics of their source patients (in confirmed patient pairs) were compared with those for adults and adolescents. No significant differences were found when the comparison group for this analysis consisted of all sources to secondary patients ≥5 years of age or when the comparison group was limited to sources to secondary patients ≥15 years of age. The results of the latter comparison are presented. More than 60% (354 of 584) of the suspected patient pairs in which the secondary patient was not a child were genotyped, and 240 (68%) of these patient pairs had concordant genotypes (Figure 2). The likelihood of identifying patient pairs with discordant genotypes was more than two times higher among suspected patient pairs involving secondary patients ≥15 years of age than for those involving young children (32% vs. 15% discordant genotypes) (RR=2.09, p=0.01).

Univariate associations between source patient characteristics and transmission to young children were assessed (Table 6). Although the mean age for sources to secondary patients <5 years of age was slightly lower than the mean age of sources to the comparison group, these differences were not significant (p=0.06; Wilcoxon test). For this population, confirmed source patients to young children were more likely to be foreign-born (p=0.02), Hispanic (p<0.001), a household member (p<0.001), and not receiving directly observed therapy (p<0.01) as compared with sources for adolescents and adults.