Nontuberculous Mycobacteria–Associated Lung Disease, United States in Hospitalized Persons, 1998–2005

Megan E. Billinger; Kenneth N. Olivier; Cecile Viboud; Ruben Montes de Oca; Claudia Steiner; Steven M. Holland; D. Rebecca Prevots


Emerging Infectious Diseases. 2009;15(10) 

In This Article


Data Source and Study Population

We used data from the Agency for Healthcare Research and Quality's Healthcare Cost and Utilization Project (HCUP), specifically the State Inpatient Databases (SID). The SIDs provide record-level data, without personal identifiers, on nearly 100% of community hospital discharges in participating states. Records were included for hospitalizations that had an International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM), code associated with pulmonary NTM (031.0) as a primary or secondary discharge diagnosis. The study population included all records for persons hospitalized with pulmonary NTM as a primary or secondary diagnosis in the 11 states participating in HCUP (Arizona, California, Colorado, Florida, Illinois, Iowa, Massachusetts, New Jersey, New York, Washington, and Wisconsin) during the years specified.[17] These states represented 42% of the US population during the study period.

Data Analysis

Data elements available in the HCUP dataset included year of hospitalization, age when hospitalized, sex, state where hospitalization occurred, type of NTM infection (pulmonary, disseminated, cutaneous, unspecified, or other) and up to 29 possible secondary diagnoses. No information on mycobacterial species is available in this dataset. Because NTM is known to be a common opportunistic infection among people with AIDS, particularly before the widespread availability of combination antiretroviral medications,[18] we limited our analysis to non-AIDS NTM using the code for HIV/AIDS (042), which indicates hospitalizations where AIDS was known to be an underlying illness. Additionally, we restricted our analysis to the 1998–2005 study period to avoid misclassification among types of NTM because the ICD-9-CM code for disseminated NTM was introduced in 1997. Before implementation, hospitalizations associated with disseminated NTM may have been included in the 4 other NTM categories (pulmonary, cutaneous, unspecified, other). We examined prevalence trends in pulmonary NTM by age and sex and described the most frequently associated underlying illnesses. To analyze the most frequent secondary underlying illnesses, we grouped the following conditions/codes as chronic obstructive pulmonary disease (COPD): obstructive chronic bronchitis with and without exacerbation (ICD-9-CM 491.21, 491.22); emphysema not elsewhere classified (492.8); chronic obstructive asthma (493.20); and chronic airway obstruction not elsewhere classified (496).

To estimate prevalence of hospitalizations, we used age- and sex-specific US census data for participating states during the study period; both individual years and midpoint population (average of 2001–2002 census population estimates) were used as appropriate. Although prevalence more often refers to the number of persons with a condition in a population at a determined time, we use it here to describe the number of hospitalizations among persons with NTM. To compare prevalence among states, we calculated age-and sex-adjusted rates using the US census 2000 reference population; χ2 tests were used to determine significance among groups at a significance level of p<0.05. Data analyses were calculated using SAS 8.0 and 9.1 (SAS, Cary, NC, USA) and EpiInfo version 3.4 (Centers for Disease Control and Prevention, Atlanta, GA, USA). The average annual percent increase in prevalence and the significance of these trends were estimated by use of Poisson regression models. Prevalence was modeled as a function of time, with prevalence as the dependent variable and time as the independent variable; Pearson's scale factor was used to account for overdispersion. Model fit was assessed by the value of the scaled Pearson χ2, which equals the value divided by the degrees of freedom (value/DF); a value of 1 indicates that the model is a good fit. Wald 95% confidence limits were estimated as well. For modeling trends by age and sex for all 11 states combined, separate models were fit for each age and sex group. Prevalence was defined as the number of observed cases in a given age and sex group for each year as the numerator and the estimated annual population for the specified age and sex group for that year as a denominator, modeled in SAS as the observed count data with a log population offset. For estimation of average annual percent change for men and women in 3 states (California, Florida, and New York), age-adjusted prevalence was the dependent variable, modeled as expected number of cases with a log population offset; time (year) was the independent variable. Models were fit separately for men and women. A constant term was included as part of these equations.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as: