In a recent study published in the journal Emerging Infectious Diseases, researchers analyzed a pharmacy dataset to evaluate the 20% decrease in tuberculosis (TB) cases reported in 2020 by the US National Tuberculosis Surveillance System (NTSS).
About the study
In this study, the team compared the decline in TB cases during the coronavirus disease 2019 (COVID-19) pandemic to that reported during 2016–2019.
The correlation between TB medication data and TB case counts in the NTSS was examined and the expected TB counts for 2020 were predicted using a seasonal autoregressive integrated moving average (SARIMA) model. The team also matched the trends in the TB medication data with that in NTSS data between 2006 and 2019.
The researchers used five different IQVIA databases: National Prescription Audit (NPA), NPA New to Brand (NTB), NPA Extended Insights, Total Patient Tracker (TPT), and NPA Regional. Most recent data from all pyrazinamide and isoniazid analyses were used, and azithromycin was chosen as a control antibiotic to examine the specificity of the study to TB data.
The provisional 2020 NTSS data reported in March 2021 was used. However, only cases reported in the 50 US states and the District of Columbia were used leaving out those reported by US territories and freely associated states. Cases were grouped by case date or treatment start date, with case date defined as the earliest treatment start date, drug susceptibility test date, or report date. For matching with the IQVIA dataset, cases with reports of resistance to the drugs of interest and cases where the drugs of interest were not used in the initial treatment regimen were not included in the analysis.
The team conducted correlation analyses where they made national-level pairwise comparisons for IQVIA databases and metrics for both pyrazinamide and isoniazid against NTSS TB case counts. They also generated a linear model using IQVIA data for visualization and plotted 95% prediction intervals and model estimates with the data. They performed a similar state-level analysis using NPA Regional data to identify the correlations at the regional levels.
SARIMA models were fitted with the seasonal models, and a linear model was generated to predict IQVIA 2020 case counts from NTSS 2020 case counts and compare these with actual 2020 IQVIA case counts.
The results showed that the IQVIA database trends in TB case counts correlated with NTSS trends. For both the drugs of interest, the team found the strongest correlation when IQVIA data was aggregated by the TPT projected total patient counts, NTSS data was aggregated by the start date of treatment after removal of patients with drug resistance, and the two databases were aggregated by month.
On comparing patient data in both the databases, the team found that in 2019 and 2020, more than half of the patients in both IQVIA and NTSS were over 45 years of age. NTSS and IQVIA isoniazid data showed a higher proportion of male patients compared to females in 2019 and 2020, while IQVIA pyrazinamide data had a bigger percentage of female patients than males. A very high correlation was found between NTSS case counts and IQVIA’s NPA Regional NRx for pyrazinamide (r = 0.92) and isoniazid (r = 0.91) in 2019 and isoniazid (r = 0.92) and pyrazinamide (r = 0.94) in 2020.
The results further showed a large decline in 2020 in both databases. There were fewer TB cases and TB medication prescriptions in 2020 than what would be expected based on previous trends. This decline in cases and prescriptions was particularly high during the period from April–to May 2020. These TB-related data are in line with NTSS data. While this suggested the absence of underreporting, it did not rule out underdiagnosis or actual decline.
On fitting a SARIMA model to every metric, most of the fitted data was within the 95% prediction interval. On assessing the percentage differences by month, the IQVIA isoniazid projected case counts and the NTSS monthly patient counts during April–December 2020 were below the 95% prediction interval.
Overall, the study demonstrates that pyrazinamide and isoniazid projected patient counts from IQVIA strongly correlated with NTSS TB case counts. The results show that there were actual large declines in 2020, thus ruling out the underreporting of TB cases as the reason behind the reported decline in the US TB cases in 2020.
Any evidence of underreporting would mean that public health officials were unaware of these TB cases but the patients were still receiving treatment on time, leading to decreased illness, mortality, and infectiousness. Although there is evidence against underreporting of TB cases, the strong probability of underdiagnosis suggests that public health programs should be better prepared for a potential rebound in TB case counts in the post-pandemic period. This is important because misdiagnosis and delayed diagnosis of TB could lead to increased transmission as TB patients remain infectious for long time periods.
A better understanding of the mechanisms behind the decline in reported TB cases during the pandemic will help develop better TB programs for post-pandemic cases.