We retrospectively compared patients with identified MDR-TB to those with non MDR-TB to identify differences in health system cost stratified by patient outcomes. Latvia’s national tuberculosis control program, the Nacionala Tuberkulozes Apkarošanas Programma (NTAP) treats tuberculosis patients and provides other public health protections as well; the NTAP and its activities are well described elsewhere [18, 19]. We obtained study data from NTAP’s national tuberculosis registry and patient medical records.
Study subjects received all tuberculosis treatment through Latvia’s TB control program. Records for all non-imprisoned, tuberculosis patients aged >=18 with treatment start dates during calendar year 2002 were eligible for inclusion. From these, we randomly selected every second MDR-TB patient and every 13th non MDR-TB patient from the TB registry, starting from a random set point. These were included where basic and comparable data points were available; where such information was insufficient, we selected the next eligible subject from the registry.
We trained public health clinic staff within the NTAP to abstract clinical and outcome data from patient records using a standard methodology and data collection instrument. Demographic, clinical, and health system utilization data were abstracted. These included gender, age, geographic location, treatment delivery type, duration, and current treatment outcomes. Outcome measures were identified from the clinical impressions reflected in the patient records and were aggregated into three categories--cured, death, or not cured (included treatment failure and default). Patients with a history of prior treatment were also identified. Health system utilization data abstracted included duration of anti-tuberculosis treatment in days; the type and number of outpatient clinical encounters; duration of inpatient treatment; and the other medical supplies, services, or procedures consumed during anti-tuberculosis therapy, such as anti-tuberculosis drugs, laboratory tests, and imaging services. Susceptibility testing in Latvia is standard during diagnosis, and we identified resistance on that basis . Data on extensively drug resistant (XDR) TB and quinolone resistance was not captured.
Cost data were obtained by study staff from clinic and other administrative records. Treatment cost was measured as a proportional share of the total annualized cost for all infrastructure, wages, utilities, supplies, and other goods and services consumed by inpatient and outpatient services. Outpatient service costs were denominated by encounter, e.g. the cost of one directly observed therapy visit for one patient. Inpatient care cost was denominated by patient day, and was estimated in a similar fashion to costs for outpatient care. Where staff, facilities, or other resources were shared between tuberculosis and other care, the cost of tuberculosis care was estimated based on the proportion of the resource consumed by tuberculosis care. For instance, the cost of facility infrastructure assigned to tuberculosis care was estimated as a function of the physical area used in that care. Costs were converted from the Lat to 2002 U.S. dollars (USD) for analysis and report.
Estimates of cost and outcomes are limited to those incurred by the public health system during the course of tuberculosis treatment. We did not attempt to estimate the economic costs or value associated with personal health losses or gains, such as acute illness, death, or cure. We did not estimate societal or other nonhealth system costs such as wage or other personal losses to a patient during treatment. Cost estimations that occur during the course of treatment are near term, representing periods of generally two years or less, and we did not discount or adjust these for time.
We used multivariate regression analysis to identify healthcare cost and utilization patterns associated with a diagnosis of MDR-TB. Negative binomial regression was used to model utilization outcomes (number of hospital days, clinic visits, specialist visits, x-rays and c computed tomography (CT)) adjusting for vital status, sex, age, and cured status. Linear regression analysis was used to examine inpatient, outpatient, prescription medication and overall costs in USD adjusting for type of TB infection, sex, age, and outcome. Robust standard errors are reported and all p-values were two-tailed. A p-value < 0.05 was considered statistically significant. All data, cost, and outcome measures are reported in the aggregate, and no personal or patient identifiers were collected, retained or reported. This program evaluation was determined to be nonhuman subject’s research by both the Office of the Associate Director for Science, U.S. Centers for Disease Control and Prevention, and by the University of North Texas Health Science Center’s Office for the Protection of Human Subjects; these determinations were provided, reviewed, and approved by program personnel within the NTAP.