Skip to main content

Predictors of latent tuberculosis infection treatment completion in the US private sector: an analysis of administrative claims data



Factors that affect latent tuberculosis infection (LTBI) treatment completion in the US have not been well studied beyond public health settings. This gap was highlighted by recent health insurance-related regulatory changes that are likely to increase LTBI treatment by private sector healthcare providers. We analyzed LTBI treatment completion in the private healthcare setting to facilitate planning around this important opportunity for tuberculosis (TB) control in the US.


We analyzed a national sample of commercial insurance medical and pharmacy claims data for people ages 0 to 64 years who initiated daily dose isoniazid treatment between July 2011 and March 2014 and who had complete data. All individuals resided in the US. Factors associated with treatment completion were examined using multivariable generalized ordered logit models and bivariate Kruskal-Wallis tests or Spearman correlations.


We identified 1072 individuals with complete data who initiated isoniazid LTBI treatment. Treatment completion was significantly associated with less restrictive health insurance, age < 15 years, patient location, use of interferon-gamma release assays, non-poverty, HIV diagnosis, immunosuppressive drug therapy, and higher cumulative counts of clinical risk factors.


Private sector healthcare claims data provide insights into LTBI treatment completion patterns and patient/provider behaviors. Such information is critical to understanding the opportunities and limitations of private healthcare in the US to support treatment completion as this sector’s role in protecting against and eliminating TB grows.

Peer Review reports


Up to 13 million people in the US have latent tuberculosis infection (LTBI) [1, 2]. These people are infected with Mycobacterium tuberculosis yet do not have active tuberculosis (TB) disease; they are asymptomatic and cannot transmit TB. Without treatment 5–10% of people with LTBI will develop TB over time, with higher progression risk among immunocompromised persons [3]. Although LTBI treatment does not eliminate the risk of progression to active TB, completion of a proven LTBI treatment regimen (e.g., 6 or 9 months of daily isoniazid, 4 months of daily rifampin, 12 doses of weekly isoniazid and rifapentine) dramatically decreases that risk [4]. The US’ strategic plan to eliminate domestic TB includes risk-targeted identification and treatment of people with LTBI [5]. This strategy is supported by the US Preventive Services Task Force’s (USPSTF) recent “Grade B” rating for LTBI testing in high-risk populations, which indicates to primary care providers that targeted LTBI testing and treatment afford moderate health benefit with little risk [6, 7].

Public health agencies have traditionally provided most TB control and prevention services in the US [8,9,10,11]. However, the USPSTF’s rating and current policy will likely drive increased involvement by private sector providers as health insurers are now required to cover TB/LTBI testing in high-risk populations with no patient cost sharing [12]. At the same time, the uninsured rate in the US is decreasing [13] and health insurance coverage is associated with increased use of primary and other private sector health care [14]. These shifts present an opportunity to coordinate public/private approaches to TB prevention. Factors associated with LTBI treatment completion are seldom studied outside of public health settings [15, 16]. Differences in patient risks, provider and patient incentives, and care processes in the private sector suggest a need for more information about the factors associated with treatment completion in this increasingly important arena.

We used a national sample of commercial claims data to examine private sector LTBI treatment across the US as a step toward filling this gap. Insurance claims offer a window into private healthcare practice patterns [17]. We aimed to use these data to identify factors associated with the completion of daily dose isoniazid LTBI treatment in the private sector setting.


Data source and analytic sample

We analyzed de-identified medical and pharmacy claims from Optum Clinformatics® Data Mart (formerly called the National Research Database) which includes claims for approximately 30.6 million commercially insured individuals – about 19% of the commercially insured US population [18]. We analyzed data for a randomly selected sample of 4 million people who were ages 0 to 64 years. Additional details about this sample are described elsewhere [19]. We used a claims-based algorithm to identify individual 6 to 9 month daily dose isoniazid courses of treatment for LTBI [19], which have been the most commonly used LTBI treatment regimens [20]. We examined treatment initiated between July 2011 and March 2014. In addition to requiring that data be available to determine if treatment was completed (as specified in the algorithm) [19] we required non-missing socio-demographic variables (i.e., the percent of foreign-born in county, patient location category, percent of households in county living under the federal poverty level (FPL), and state TB rate).


Outcome variable

The outcome of interest was completion of daily isoniazid treatment for LTBI [21]. Patients may be prescribed a 6 or 9-month isoniazid regimen [4]. While our data do not indicate whether the 6 or 9-month regimen was prescribed, we could determine how many doses of isoniazid were dispensed. Thus, we grouped isoniazid treatments into three mutually exclusive ordinal categories: 1) non-completion (< 180 doses received within 9-months), 2) completion of the 6-month regimen but not the 9-month regimen (180 to 269 doses received within 9-months), or 3) completion of the 9-month regimen (≥ 270 doses received within 12-months) [20]. These increasing levels of completion are important because, while isoniazid treatment completion at any duration does not necessarily imply LTBI cure, the risk of progression to active TB decreases as the duration of isoniazid treatment increases [22].

Explanatory variables

Explanatory variables were constructed from the medical and pharmacy claims data (see Additional file 1 for details). Socio-demographic variables included sex, age, census region, and a patient location variable based on the National Center for Health Statistics urban-rural classification [23]. The percentage of households living under the federal poverty level in a patient’s county served as a proxy for household income [24]. Additional variables included insurance type (health maintenance organization [HMO], point of service [POS], or preferred provider organization [PPO]), prescription size (the supply of isoniazid received when the first prescription was filled; < 2 months or ≥ 2 months), year, and the type of LTBI diagnostic test received in the 6 months before treatment initiation. Non-clinical variables related to risk of LTBI or progression to active TB were included, such as the state TB rate. While country of birth was unavailable, we included prevalence of foreign-born individuals in the patient’s county as a proxy [25, 26]. Clinical risk factors included diabetes, tobacco use, HIV, immunosuppressive medication use, contact with or exposure to TB, and a history of or late effects of TB [27]. A simple count of each patient’s clinical risk factors represented cumulative risk (i.e., 0, 1, or > 1 risk factor).

Statistical analyses

We calculated the proportion of individuals in each of three categories of treatment completion (i.e., < 6 months, 6 to < 9 months, ≥9 months) and examined the bivariate relationships between the explanatory variables and completion using Kruskal-Wallis tests and Spearman correlations. We explored the adjusted association between these variables and treatment completion category using multivariable generalized ordered logit models. Variables meeting the parallel-lines assumption were constrained to have equal effects; the odds ratios for non-completion versus completing ≥6 months of treatment and those for completing < 9 months of treatment versus ≥9 months of treatment were the same. Variables violating the assumption were not constrained and consequently have different odds ratios for completion category comparisons [28]. We ran two multivariable generalized ordered logit models. In Model 1 we examined the relationship between completion and cumulative risk. Model 2 explored the relationship between completion and individual clinical risk factors.

We also ran a multivariable logit model with completion of ≥6 months of treatment as the outcome measure and all predictors from the more detailed Model 2 as explanatory variables. This logit model was used to examine the reduction of variance in the treatment completion variable attributable to each predictor, which provided insight into the importance of the variables with respect to model predictions of completing ≥6 months of treatment [29, 30].

We conducted two sets of post hoc analyses. First, in order to assess the robustness of our findings we conducted sensitivity analyses using variations on our treatment completion outcomes measure. We ran four multivariable logistic regression models to explore characteristics associated with completion of ≥5 months of treatment and compare the results to the characteristics associated with ≥6 months of treatment in Models 1 and 2. Four models were used because we had two sets of explanatory variables (see descriptions of Models 1 and 2 above), and we defined completion two ways: 1) 150 doses in 9 months, and 2) 150 doses in 8 months. We explored the data using two definitions because we identified no previous studies or clinical practice guidelines defining a time period in which 150 doses (5 months) of isoniazid would be considered completed treatment.

Second, we explored our findings related to the LTBI testing variable. We ran a frequency distribution which contained additional details about the LTBI tests received. Additionally, to clarify differences between the results in our bivariate and multivariable analyses, we conducted post hoc bivariate analyses exploring the relationship between the explanatory variables and the type of LTBI diagnostic test using chi square tests for categorical variables and ANOVAs for continuous variables.

We used Stata 14.2 for most statistical testing [31] but used IBM SPSS Modeler 17 to complete the importance analysis [32]. All statistical testing was two-sided, and significance was tested at p < .05.


Two (0.2%) of 1074 individuals identified with the algorithm as having initiated isoniazid LTBI treatment were excluded due to missing geographic variables. Of the remaining 1072 almost half (46.2%) completed ≥6 months of treatment. The balance (53.8%) initiated but did not complete the minimum 6-months course. Roughly equal proportions completed ≥6 but < 9 months treatment or ≥ 9 months (23.6 and 22.6% of all patients, respectively; Table 1).

Table 1 Completion of daily-dose isoniazid treatment for latent tuberculosis infection. N = 1072

Tables 2 and 3 describe relationships between the explanatory variables and the likelihood of treatment completion from bivariate analyses and multivariable models, respectively. Significant unadjusted non-clinical factors associated with completion included younger age, PPO insurance, larger prescription size, and residing in a county with < 15% of households below FPL. Similarly, in the multivariable models younger people (ages 0 to 14 years) had higher adjusted odds of treatment completion than older people. Compared to people in large central metropolitan counties, those in large fringe metropolitan counties had lower adjusted odds of completing ≥6 months of treatment, although this association was not seen with completing ≥9 months of treatment. Residing in a county with ≥15% of households below FPL was significantly associated with lower adjusted odds of completion. Detailed adjusted odds ratios for the associations described above are found in Table 3.

Table 2 Frequency distribution of patient characteristic variables for people initiating daily-dose isoniazid treatment and the proportion of people completing treatment by each characteristic. Treatment completion was categorized as 1) less than 6 months completed, 2) at least 6 months but less than 9 months completed, and 3) 9 or more months completed
Table 3 Results of two multivariable generalized ordered logit modelsa with partial proportional odds which examine associations between patient characteristics and the completionb of daily-dose isoniazid treatment for latent tuberculosis infection (N = 1072)

Insurance type and prescription size were also significantly associated with completion. The adjusted odds of a PPO-insured patient completing ≥6 months of treatment were 1.8 to 1.9 times that of an HMO-insured patient, and the odds of a PPO-insured patient completing ≥9 months were 2.8 to 2.9 times that of an HMO-insured patient. Larger prescription size was associated with higher adjusted odds of completing ≥9 months of treatment, although this association was not seen for completing ≥6 months of treatment.

IGRA testing, HIV, and immunosuppressive medication use each had statistically significant bivariate associations with treatment completion. In the multivariable model, people with HIV had an adjusted 2.5 times greater odds of an increased level of completion relative to those without. Additionally, both unadjusted and adjusted likelihood of completion was significantly associated with cumulative clinical risk. Compared to people with no clinical risk factors, those with one risk factor had 1.5 times greater adjusted odds and those with more than one risk factor had 1.8 times greater adjusted odds of an increased level of treatment completion. The importance analysis indicated that the most important variable in predicting treatment of ≥6 months of treatment was patient location, followed closely by immunosuppressive medication use (Fig. 1; see Additional file 2 for logistic regression model results).

The results of the sensitivity models examining ≥5 months of treatment were quite similar to the primary analyses wherein completion was defined as ≥6 months of treatment (see Additional file 3 for detailed sensitivity model results). All findings were directionally identical and odds ratios were of similar magnitude. While most variables were consistent in terms of statistical significance, there were two exceptions. Some age group and insurance type categories that were significant in the primary analyses were not significant in the sensitivity analyses. However, the p-values for these categories approached significance, ranging from p = 0.052 to p = 0.072. Based on these results we concluded that the results of our primary analyses were robust to variations in the definition of treatment completion.

Additional post hoc analyses indicated that 34.9% of the individuals initiating LTBI treatment had no procedure or diagnostic code in the medical claims data specifically indicating that an LTBI test occurred, although the majority of these individuals had a diagnosis of LTBI (Table 4). We also identified significant associations between LTBI diagnostic test type and our model’s explanatory variables (Table 5). Diagnostic test type was significantly associated with age, region, patient location, insurance plan type, year, clinical risk factor count, history of or late effects of TB, HIV, diabetes, tobacco use, and immunosuppressive medication use.

Table 4 Frequency distribution of evidence of latent tuberculosis infection (LTBI) testing occurring in the 6 months prior to LTBI treatment initiation with isoniazid (n = 1072)
Table 5 Bivariate associations between Mycobacterium tuberculosis test type and other patient characteristics. Includes people initiating daily-dose isoniazid treatment (N = 1072)


We used commercial insurance claims data to identify important individual, clinical, and system factors associated with the completion of LTBI treatment with isoniazid. Most striking were significant associations between a patient’s insurance plan type and treatment completion, suggesting that benefit design is a potential means to modify patient behaviors and ultimately TB risk. HMO plans, the most tightly managed insurance design, were associated with the lowest likelihood of completion; PPO plans, the least restrictive plans, were associated with the highest. Completion differences may be due to differences in access or cost sharing, as such health plan characteristics are associated with continued adherence to other types of medications [32].

The lower completion rates for HMO-insured individuals suggest a need for HMOs to monitor and conduct quality improvement initiatives that improve enrollees’ LTBI treatment completion rates. Such activities would not be unusual – HMOs in most states are required to operate quality assurance programs that involve monitoring and conducting activities to improve care processes and clinical outcomes, such as improving medication adherence rates [33]. As private sector LTBI treatment becomes more common, the National Committee for Quality Assurance (NCQA) should consider incorporating an LTBI treatment completion measure into its standard set of quality performance measures (Healthcare Effectiveness Data and Information Set [HEDIS]) [34]. Health plans’ quality improvement activities often focus on improving HEDIS rates, as many states consider quality assurance requirements met if plans maintain NCQA accreditation [33] and plans are required to calculate HEDIS measures to attain and maintain accreditation [35].

Pharmacy benefit design and prescribing offer similar opportunities to decrease TB risk through improved treatment completion. Individuals filling larger prescriptions (≥ 2 months supply) had greater odds of completing a 9-month regimen. Although we cannot be certain given data limitations, completion of the longer regimen may be due to the use of mail order pharmacies with automatic refill programs. Many insurers disallow community pharmacies from providing a > 1-month supply of a medication. However, enrollees may be able to use mail order pharmacies to receive up to a 90-day supply [36], and mail order pharmacies are more likely to have automatic refill programs [37]. These programs address patient passivity and transportation barriers by mailing prescription refills at regular intervals. Thus, encouraging patients to fill larger prescriptions and use automated mail order programs may increase 9-month isoniazid completion rates so long as appropriate clinical monitoring to avoid hepatotoxicity and other complications is ensured [21].

Fig. 1

Bar chart depicting the importance of variables in predicting completion of ≥6 months of isoniazid treatment for latent tuberculosis infection (LTBI). Longer bars represent greater importance

Our analysis suggests that private sector providers are likely sensitive to and communicating the importance of treatment completion for LTBI patients at high risk of active TB. Patients with serious known risk factors such as HIV and immunosuppressive medication use [27] are more likely to complete treatment than others, and immunosuppressive medication use is of particular importance in predicting adherence. Correspondingly, completion was increasingly likely as the total number of clinical risk factors increased. Nevertheless, there are opportunities to improve completion in high-risk private sector patients, as nearly half of those with one clinical risk factor and 45.3% of those with > 1 risk factor did not complete at least 6 months of LTBI treatment. As shorter-course regimens (e.g., 3 months of weekly isoniazid and rifapentine; 4 months of daily rifampin) typically have higher completion rates [38, 39], the use of these regimens would likely increase treatment completion rates. We also found that TST is much more likely to be used among young children than IGRA. This is consistent with the CDC guidelines [40] and suggests that private providers are receiving CDC messaging related to best practices [21] and are following these practices.

We found that likelihood of completing ≥6 months of treatment varied by patient location, with individuals in large fringe metro counties (i.e., suburban counties [23]) having a lower likelihood of completion than those in large central metro counties (i.e., counties containing an inner-city [23]). These findings are in contrast to recent research examining chronic condition medication adherence for rural, suburban, and urban populations in which no significant differences were found [41]. The differing LTBI treatment completion rates that we identified may be due to differences in provider familiarity with LTBI treatment best practices. Increased provider awareness of best practices and more years of experience are associated with increasing provider adherence to best practices [42, 43]. As TB incidence is much higher in urban areas than other areas [44], providers in urban areas have likely had more exposure to patients in need of LTBI treatment, more exposure to LTBI treatment guidelines, and a greater awareness of the benefits of LTBI treatment completion. Claims data do not allow us to investigate providers’ knowledge of LTBI treatment best practices, so additional research is warranted to confirm the cause of the location-related differences. Even so, given the suburbanization of the US population [45] and the importance of this variable in identifying patients likely to complete < 6 months of treatment (see Fig. 1), our findings identify an important opportunity to improve LTBI treatment completion rates in patients treated by private sector providers in suburban areas.

Our finding that IGRA is associated with greater likelihood of treatment completion aligns with anecdotal reports that IGRA testing may yield greater diagnostic confidence for patients and providers relative to TST. However, the association is only significant in our unadjusted analysis. LTBI test type is also associated with many other variables, including clinical risk factors, census region, insurance plan type, and year. After adjusting for these other variables, there is no significant association between the receipt of an IGRA and treatment completion. It is unclear if the use of IGRA facilitates completion or if IGRA testing is more common in patients with other characteristics associated with completion.

Claims are a rich source of information about commercial insurance-covered LTBI treatment occurring across the US, but they have limitations. These data generally accurately reflect diagnoses and treatment [17], but accuracy varies with the clarity of coding instructions and guidelines [46]. There is ambiguity in the diagnostic and procedure coding for LTBI. For example, providers may be using the “contact with or exposure to tuberculosis” diagnosis code to represent LTBI status rather than known recent contacts. This might explain inconsistencies between our findings and prior reports of better completion rates among TB contacts [47,48,49,50]. Conversely, many of our findings regarding LTBI treatment completion are consistent with past research, including associations with younger age and higher income [15, 16]. Additionally, claims data only reflect information submitted to a third party payer for the purposes of reimbursement [17]. Our finding that LTBI testing procedure codes were not present in the claims for over a third of the individuals initiating isoniazid treatment suggests that some providers are either not billing for LTBI testing or some patients are receiving LTBI testing and treatment in different settings. For example, a patient might be diagnosed for LTBI in a workplace, school, or public health department that does not bill third party payers but subsequently seek treatment or fill prescriptions in the private sector using insurance benefits.

Due to limitations of claims data we cannot precisely determine treatment intent or adherence, and conclusions about provider and patient behavior are based on inference, not direct report. For instance, it is unclear whether a 6 or 9-month treatment regimen was prescribed for a given patient. Further, we cannot know if a filled prescription is actually consumed, and it is possible that those enrolled in automatic refill programs may receive refills even if they have discontinued their treatment. Of course, the uncertainty related to medication consumption applies to all medication adherence research not involving direct observation [51]. Fortunately, numerous studies have illustrated that medication adherence as measured by filled prescriptions is significantly correlated with both medication consumption and drug serum levels [52]. Consequently, claims-based methods of evaluating medication adherence are widely used in health services research and quality assurance monitoring [53,54,55,56,57,58,59,60,61,62].

Data limitations left us unable to identify important TB risk factors. Patient-level income and country of birth were unavailable. While 59% of foreign-born people in the US have private health insurance [13], claims data do not identify nativity. However, county-level nativity and FPL rates were included as proxies. Our data also did not detail treatment-related out-of-pocket costs for isoniazid or office visits, nor did it provide insight into insurance benefit plan design or network adequacy. Our analysis examining the importance of the variables in the model should be interpreted with these limitations in mind, as the results only assess the relative importance of variables available within the administrative claims data. Other, unavailable variables may be of great importance in predicting treatment completion. Nevertheless, claims data provide unique opportunities to better understand LTBI treatment occurring in a setting of increasing importance for TB prevention in the US.


In the US, patient risks, provider and patient incentives or barriers, benefits design, and care processes in private healthcare differ substantially from that of public health programs. Our findings illustrate that many of these factors have an impact on LTBI treatment completion. This new information enables the development of evidence-based LTBI private sector treatment strategies. Such work is critical as more private healthcare providers provide LTBI treatment and as public health authorities consider the opportunities and limitations of private healthcare as a partner to US TB elimination efforts.



Centers for Disease Control and Prevention


Federal poverty level


Healthcare Effectiveness Data and Information Set


Human immunodeficiency virus


Health maintenance organization


Interferon-gamma release assays




Latent tuberculosis infection


National Committee for Quality Assurance


Point of service


Preferred provider organization




Tuberculin skin test


United States


United States Preventive Services Task Force


  1. 1.

    Mancuso JD, Diffenderfer JM, Ghassemieh BJ, Horne DJ, Kao TC. The prevalence of latent tuberculosis infection in the United States. Am J Respir Crit Care Med. 2016;194(4):501–9.

    Article  PubMed  CAS  Google Scholar 

  2. 2.

    Miramontes R, Hill AN, Yelk Woodruff RS, Lambert LA, Navin TR, Castro KG, LoBue PA. Tuberculosis infection in the United States: prevalence estimates from the National Health and nutrition examination survey, 2011-2012. PLoS One. 2015;10(11):e0140881.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  3. 3.

    Kahwati LC, Feltner C, Halpern M, Woodell CL, Boland E, Amick HR, Palmieri Weber R, Jonas DE: Screening for Latent Tuberculosis Infection in Adults: An Evidence Review for the U.S. Preventive Services Task Force. Evidence Synthesis No. 142. AHRQ Publication No. 14–05212-EF-1. Rockville, MD: Agency for Healthcare Research and Quality. In., vol. 14–05212-EF-1; 2016.

  4. 4.

    Centers for Disease Control and Prevention: Treatment regimens for latent TB infection (LTBI). 2017.

  5. 5.

    Centers for Disease Control and Prevention: Division of Tuberculosis Elimination Strategic Plan 2016–2020. 2015.

  6. 6.

    US Preventive Services Task Force: Update on Methods: Estimating Certainty and Magnitude of Net Benefit. 2017.

  7. 7.

    US Preventive Services Task Force, Bibbins-Domingo K, Grossman DC, Curry SJ, Bauman L, Davidson KW, Epling JWJ, Garcia FA, Herzstein J, Kemper AR, et al. Screening for latent tuberculosis infection in adults: US preventive services task force recommendation statement. JAMA. 2016;316(9):962–9.

    Article  Google Scholar 

  8. 8.

    Sterling TR, Bethel J, Goldberg S, Weinfurter P, Yun L, Horsburgh CR, Tuberculosis Epidemiologic Studies C. The scope and impact of treatment of latent tuberculosis infection in the United States and Canada. Am J Respir Crit Care Med. 2006;173(8):927–31.

    Article  PubMed  Google Scholar 

  9. 9.

    Ehman M, Flood J, Barry PM. Tuberculosis treatment managed by providers outside the public health department: lessons for the affordable care act. PLoS One. 2014;9(10):e110645.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  10. 10.

    Institute of Medicine Committee on the Elimination of Tuberculosis in the United States. Ending neglect: the elimination of tuberculosis in the U.S. Washington (DC): national academy of sciences; 2000.

    Google Scholar 

  11. 11.

    Balaban V, Marks SM, Etkind SC, Katz DJ, Higashi J, Flood J, Cronin A, Ho CS, Khan A, Chorba T. Tuberculosis elimination efforts in the United States in the era of insurance expansion and the affordable care act. Public Health Rep. 2015;130(4):349–54.

    Article  PubMed  PubMed Central  Google Scholar 

  12. 12.

    The Patient Protection and Affordable Care Act, Sec. 2713, Coverage of Preventive Services. 2010.

  13. 13.

    Barnett JCV, Marina S: Health insurance coverage in the United States. 2016

  14. 14.

    Foutz J, Damico A, Squires E, Garfield R. The Uninsured: A Primer - Key Facts about Health Insurance and the Uninsured Under the Affordable Care Act. The Henry J. Kaiser Family Foundation. 2017.

  15. 15.

    Hirsch-Moverman Y, Daftary A, Franks J, Colson PW. Adherence to treatment for latent tuberculosis infection: systematic review of studies in the US and Canada. The international journal of tuberculosis and lung disease. 12(11):1235–54.

  16. 16.

    Stuurman AL, Noordegraaf-Schouten MV, Van Kessel F, Oordt-Speets A, Sandgren AA, Van der Werf MJ. Interventions for improving adherence to treatment for latent tuberculosis infection: a systematic review. BMC Infect Dis. 2016;16:257.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  17. 17.

    Virnig BA, McBean M. Administrative data for public health surveillance and planning. Annu Rev Public Health. 2001;22(1):213–30.

    Article  PubMed  CAS  Google Scholar 

  18. 18.

    Optum: Clinformatics Data Mart Database. 2014.

  19. 19.

    Stockbridge EL, Miller TL, Carlson EK, Ho CS. Tuberculosis prevention in the private sector: using claims-based methods to identify and evaluate latent tuberculosis infection treatment with isoniazid among the commercially insured. J Public Health Manag Pract. 2017. Epub Ahead of Print

  20. 20.

    Horsburgh CR Jr, Goldberg S, Bethel J, Chen S, Colson PW, Hirsch-Moverman Y, Hughes S, Shrestha-Kuwahara R, Sterling TR, Wall K, et al. Latent TB infection treatment acceptance and completion in the United States and Canada. Chest. 2010;137(2):401–9.

    Article  PubMed  Google Scholar 

  21. 21.

    Centers for Disease Control and Prevention: Latent Tuberculosis Infection: A Guide For Primary Health Care Providers. 2013.

  22. 22.

    Comstock GW. How much isoniazid is needed for prevention of tuberculosis among immunocompetent adults? Int J Tuberc Lung Dis. 1999;3(10):847–50.

    PubMed  CAS  Google Scholar 

  23. 23.

    Centers for Disease Control and Prevention: NCHS Urban-Rural Classification Scheme for Counties. 2017.

  24. 24.

    United States Census Bureau: Small Area Income and Poverty Estimates. 2015.

  25. 25.

    United States Census Bureau: American Community Survey (ACS). 2018.

  26. 26.

    United States Department of Agriculture: Atlas of Rural and Small-Town America. 2018

  27. 27.

    Centers for Disease Control and Prevention. Targeted tuberculin testing and treatment of latent tuberculosis infection. American Thoracic Society. MMWR. 2000;49(RR-6):1–51.

    Google Scholar 

  28. 28.

    Williams R. Generalized ordered logit/partial proportional odds models for ordinal dependent variables. Stata J. 2006;6(1):58–82.

    Google Scholar 

  29. 29.

    Saltelli A, Tarantola S, Campolongo F, Ratto M. Sensitivity analysis in practice: A Guide to Assessing Scientific Methods. West Sussex, England: John Wiley & Sons Ltd; 2004.

    Google Scholar 

  30. 30.

    IBM Corporation: IBM SPSS modeler 17 algorithms guide. Chicago, IL: IBM software Group; 2015.

    Google Scholar 

  31. 31.

    StataCorp: Stata Statistical Software: Release 14. College Station, TX: StataCorp LP; 2016.

    Google Scholar 

  32. 32.

    Horgan CM, Merrick EL, Stewart MT, Scholle SH, Shih S. Improving medication management of depression in health plans. Psychiatr Serv. 2008;59(1):72–7.

    Article  PubMed  Google Scholar 

  33. 33.

    Alexander BD, Broccolo BM, Daniels AR, SM DV, Hutto CB, Colborn-Loepere C, Matyas DE, Mayo TW, Miles JJ, Polikoff EA, et al. Fundamentals of Health Law, vol. 5th. Washington, DC: American Health Laywers Association; 2011.

    Google Scholar 

  34. 34.

    National Committee for Quality Assurance: HEDIS & Performance Measures. 2018.

  35. 35.

    National Committee for Quality Assurance: Health Plan Accreditation. 2017.

  36. 36.

    Pharmacy Benefits Management Institute: 2015–2016 Prescription Drug Benefit Cost and Plan Design Report. 2017.

  37. 37.

    Brown TA, Rickles NM. Mail-order versus local pharmacies on adherence: study methods make for unfair comparison. Am J Manag Care. 2010;16(3):226–7. author reply 227-228

    PubMed  Google Scholar 

  38. 38.

    Sandgren AA, Noordegraaf-Schouten MV, van Kessel F, Stuurman A, Oordt-Speets A, van der Werf MJ. Initiation and completion rates for latent tuberculosis infection treatment: a systematic review. BMC Infect Dis. 2016;16(1):204.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  39. 39.

    Sharma SK, Sharma A, Kadhiravan T, Tharyan P. Rifamycins (rifampicin, rifabutin and rifapentine) compared to isoniazid for preventing tuberculosis in HIV-negative people at risk of active TB. Cochrane Database Syst Rev. 2013;7:Cd007545.

    Google Scholar 

  40. 40.

    Lewinsohn DM, Leonard MK, LoBue PA, et al. Official American Thoracic Society/Infectious Diseases Society of America/Centers for Disease Control and Prevention clinical practice guidelines: diagnosis of tuberculosis in adults and children. Clin Infect Dis. 2017;64(2):e1-e33.

    Article  PubMed  Google Scholar 

  41. 41.

    Arbuckle C, Tomaszewski D, Aronson BD, Brown L, Schommer J, Morisky D, Linstead E. Evaluating factors impacting medication adherence among rural, urban, and suburban populations. J Rural Health. 2018;

  42. 42.

    Francke AL, Smit MC, de Veer AJE, Mistiaen P. Factors influencing the implementation of clinical guidelines for health care professionals: a systematic meta-review. BMC Medical Informatics and Decision Making. 2008;8(38)

  43. 43.

    McKinlay JB, Link CL, Freund KM, Marceau LD, O’Donnell AB, Lutfey KL. Sources of variation in physician adherence with clinical guidelines: results from a factorial experiment. J Gen Intern Med. 2007;22(3):289–96.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  44. 44.

    Oren E, Winston CA, Pratt R, Robison VA, Narita M. Epidemiology of urban tuberculosis in the United States, 2000-2007. Am J Public Health. 2011;101(7):1256–63.

    Article  PubMed  PubMed Central  Google Scholar 

  45. 45.

    Gardner T: Urban-suburban migration in the United States, 1955–2000. US Census Bureau Center for Economic Studies Paper No. CES-WP-16-08: 2016.

  46. 46.

    Centers for Disease Control and Prevention: Official Authorized Addenda: Human Immunodeficiency Virus Infection Codes and Official Guidelines for Coding and Reporting ICD-9-CM. 1994.

  47. 47.

    Li JJ. Adherence to treatment of latent tuberculosis infection in a clinical population in new York City. Int J Infect Dis. 2010;14(4):e292–7.

    Article  PubMed  Google Scholar 

  48. 48.

    Nyamathi AM, Christiani A, Nahid P, Gregerson P, Leake B. A randomized controlled trial of two treatment programs for homeless adults with latent tuberculosis infection. Int J Tuberc Lung Dis. 2006;10(7):775–82.

    PubMed  CAS  Google Scholar 

  49. 49.

    Fiske CT, Yan FX, Hirsch-Moverman Y, Sterling TR, Reichler MR. Risk factors for treatment default in close contacts with latent tuberculous infection. Int J Tuberc Lung Dis. 2014;18(4):421–7.

    Article  PubMed  PubMed Central  Google Scholar 

  50. 50.

    Young KH, Ehman M, Reves R, Peterson BL, Khan A, Chorba TL, Jereb J. Tuberculosis contact investigations--United States, 2003-2012. MMWR Morb Mortal Wkly Rep. 2016;64(50–51):1369–74.

    Article  PubMed  Google Scholar 

  51. 51.

    McGinnis B, Kauffman Y, Olson KL, Witt DM, Raebel MA. Interventions aimed at improving performance on medication adherence metrics. Int J Clin Pharm. 2014;36(1):20–5.

    Article  PubMed  Google Scholar 

  52. 52.

    Steiner JF, Prochazka AV. The assessment of refill compliance using pharmacy records: methods, validity, and applications. J Clin Epidemiol. 1997;50(1):105–16.

    Article  PubMed  CAS  Google Scholar 

  53. 53.

    Agency for Healthcare Research and Quality: Measure Summary: Medication management for people with asthma (NQMC: 010514). 2015.

  54. 54.

    Agency for Healthcare Research and Quality: Measure Summary: Antidepressant medication management (effective acute phase treatment) (NQMC: 010531). 2015.

  55. 55.

    Agency for Healthcare Research and Quality: Measure Summary: Antidepressant medication management (effective continuation phase treatment) (NQMC: 010532). 2015.

  56. 56.

    Agency for Healthcare Research and Quality: Measure Summary: Adherence to antipsychotic medications for individuals with schizophrenia (NQMC: 010540). 2015.

  57. 57.

    Sueta CA, Rodgers JE, Chang PP, Zhou L, Thudium EM, Kucharska-Newton AM, Stearns SC. Medication adherence based on part D claims for patients with heart failure after hospitalization (from the atherosclerosis risk in communities study). Am J Cardiol. 2015;116(3):413–9.

    Article  PubMed  PubMed Central  Google Scholar 

  58. 58.

    Seabury SA, Lakdawalla DN, Dougherty JS, Sullivan J, Goldman DP. Medication adherence and measures of health plan quality. Am J Manag Care. 2015;21(6):e379–89.

    PubMed  Google Scholar 

  59. 59.

    Kirkman MS, Rowan-Martin MT, Levin R, Fonseca VA, Schmittdiel JA, Herman WH, Aubert RE. Determinants of adherence to diabetes medications: findings from a large pharmacy claims database. Diabetes Care. 2015;38(4):604–9.

    PubMed  PubMed Central  Google Scholar 

  60. 60.

    Doshi JA, Lim R, Li P, Young PP, Lawnicki VF, Troxel AB, Volpp KG. Synchronized prescription refills and medication adherence: a retrospective claims analysis. Am J Manag Care. 2017;23(2):98–104.

    PubMed  Google Scholar 

  61. 61.

    Medicare Payment Advisory Commission: Chapter 7: Measuring the effects of medication adherence for the Medicare population. Report to the Congress: Medicare and the Health Care Delivery System. 2014.

  62. 62.

    Andrade SE, Kahler KH, Frech F, Chan KA. Methods for evaluation of medication adherence and persistence using automated databases. Pharmacoepidemiol Drug Saf. 2006;15(8):565–74. discussion 575-567

    Article  PubMed  Google Scholar 

Download references


The authors gratefully acknowledge the support of the US Centers for Disease Control and Prevention’s Division of Tuberculosis Elimination and its Tuberculosis Epidemiologic Studies Consortium (Atlanta, GA, USA) which provided valuable intellectual and other contributions. Additionally, the research reported in this publication was developed in collaboration with Magellan Health, Inc. (Scottsdale, AZ, USA). We thank Magellan for their invaluable contributions to this work. The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the United States Centers for Disease Control and Prevention (CDC) or Magellan Health, Inc. Mention of company names or products does not imply endorsement by the CDC or Magellan.


No funding was received for this study. Dr. Stockbridge is a contractor for a commercial company: Magellan Health, Inc. Magellan Health provided support in the form of salaries for Dr. Stockbridge and access to the data, but did not have any additional role in the study design, analysis, decision to publish, or preparation of the manuscript. No other authors have financial disclosures to report.

Availability of data and materials

The data used in this study were licensed from Optum by Magellan Health, Inc. These data cannot be made freely available due to the nature of the data (specifically, it contains dates related to individuals and their healthcare utilization) and due to the licensing agreement between Optum and Magellan. Researchers interested in obtaining these data may contact Mike Crowley at Optum ( in order to request clearance to use the data and to obtain a license for use of the data.

Author information




ELS conceptualized the project, designed the methods, conducted data transformations and analyses, interpreted the results, drafted the manuscript, and approved the final version of the manuscript. TLM conceptualized the project, designed the methods, interpreted the results, drafted the article, and approved the final version. EKC contributed to the methodology design, interpreted results, revised the article, and approved the final version. CH designed the methods, reviewed and approved the billing code lists, interpreted results, revised the article, and approved the final version.

Corresponding author

Correspondence to Erica L. Stockbridge.

Ethics declarations

Ethics approval and consent to participate

The institutional review board of the University of North Texas Health Science Center approved this project as exempt category research. The data analyzed in the study consisted of medical and pharmacy claims data collected for non-research purposes. The data were de-identified and fully compliant with the US Health Insurance Portability and Accountability Act of 1996. This research did not involve the collection, use, or transmittal of individually identifiable data.

Competing interests

The authors have no competing interests to declare. Dr. Stockbridge is a contractor for a commercial company: Magellan Health, Inc. This affiliation does not represent a competing interest and does not alter the authors’ adherence to BMC Public Health publication policies.

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Additional files

Additional file 1:

Excel file detailing the billing codes used in the analyses. Each tab provides information about a different variable. (XLS 385 kb)

Additional file 2:

Results of a logistic regression model which examines associations between patient characteristics and the completion of ≥6 months of daily-dose isoniazid treatment for latent tuberculosis infection (N = 1072). (XLSX 11 kb)

Additional file 3:

Results of logistic regression models which examine associations between patient characteristics and the completion of at least 5 months of daily-dose isoniazid treatment for latent tuberculosis infection (N = 1072). (XLS 31 kb)

Rights and permissions

Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (, which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver ( applies to the data made available in this article, unless otherwise stated.

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Stockbridge, E.L., Miller, T.L., Carlson, E.K. et al. Predictors of latent tuberculosis infection treatment completion in the US private sector: an analysis of administrative claims data. BMC Public Health 18, 662 (2018).

Download citation


  • Latent tuberculosis infection
  • LTBI
  • Treatment completion
  • Claims data
  • Administrative data
  • Isoniazid
  • Epidemiology
  • Health service delivery
  • Public health practice
  • Medication adherence