Evaluating Accuracy and Completeness Of Recent Homelessness In a Tuberculosis Case Management and Surveillance System, New York City, 2001-2012

BACKGROUND:   Homeless history during the year before tuberculosis (TB) diagnosis (recent homelessness) is one variable used by the Centers for Disease Control and Prevention to determine funding for TB control programs. In September 2010, the New York City (NYC) Department of Health and Mental Hygiene Bureau of Tuberculosis Control replaced its mainframe-based TB registry (registry) with an enhanced case management and surveillance system (Maven). We compared accuracy and completeness of the recent homelessness variable in the registry with Maven.

METHODS:   NYC TB cases counted between January 1, 2001 and June 30, 2012 were evaluated. Cases counted before September 1, 2010 were classified as registry cases. To assess accuracy of recent homelessness, recent homelessness responses were compared to responses to another variable: homeless at report of TB diagnosis. If a case was entered as being homeless at report of TB diagnosis, the case was considered to be recently homeless. Using this definition cases considered to be recently homeless, but not recorded as recently homeless in the registry or Maven, were considered misclassified. Completeness was evaluated by comparing the proportion of unknown responses for the recent homelessness variable between registry and Maven cases.

RESULTS:   Of 10,698 TB cases in the study period, 9,474 (89%) were registry and 1,224 (11%) were Maven cases. Although 330 (3%) cases were indicated as being recently homeless, the accuracy assessment identified an additional 198 recently homeless cases, suggesting the true prevalence of recent homelessness was 5%. Of those who were misclassified, 186 (94%) were registry cases, and 12 (6%) were Maven cases. Regarding completeness, recent homelessness was unknown for 36% of registry cases and 3% of Maven cases. 

CONCLUSIONS:   Although the findings of this evaluation cannot definitively be attributed to Maven, completeness and accuracy of the recent homelessness variable did improve after Maven implementation. While Maven has enhanced our ability to collect TB surveillance and case management data, its complexity has introduced new challenges including the continued need for internal validation and quality assurance checks in a more complex system. Ensuring accuracy and completeness of the recent homelessness variable is important as it has implications for program funding, case management activities, and our understanding of homelessness as a risk factor for TB in NYC.