Interrupted time-series analysis of active case-finding for tuberculosis during the COVID-19 pandemic, Zambia

Abstract Objective To evaluate the impact of the coronavirus disease 2019 (COVID-19) pandemic and the subsequent implementation of tuberculosis response measures on tuberculosis notifications in Zambia. Methods We used an interrupted time-series design to compare monthly tuberculosis notifications in Zambia before the pandemic (January 2019 to February 2020), after implementation of national pandemic mitigation measures (April 2020 to June 2020) and after response measures to improve tuberculosis detection (August 2020 to September 2021). The tuberculosis response included enhanced data surveillance, facility-based active case-finding and activities to generate demand for services. We used nationally aggregated, facility-level tuberculosis notification data for the analysis. Findings Pre-pandemic tuberculosis case notifications rose steadily from 2890 in January 2019 to 3337 in February 2020. After the start of the pandemic and mitigation measures, there was a −22% (95% confidence interval, CI: −24 to −19) immediate decline in notifications in April 2020. Larger immediate declines in notifications were seen among human immunodeficiency virus (HIV)-positive compared with HIV-negative individuals (−36%; 95% CI: −38 to −35; versus −12%; 95% CI: −17 to −6). Following roll-out of tuberculosis response measures in July 2020, notifications immediately increased by 45% (95% CI: 38 to 51) nationally and across all subgroups and provinces. The trend in notifications remained stable through September 2021, with similar numbers to the predicted number had the pandemic not occurred. Conclusion Implementation of a coordinated public health response including active tuberculosis case-finding was associated with reversal of the adverse impact of the pandemic and mitigation measures. The gains were sustained throughout subsequent waves of the pandemic.


Introduction
Since the beginning of the global coronavirus disease 2019 (COVID-19) pandemic in early 2020, there were fears that it would devastate fragile health systems in resource-limited settings and erase hard-fought gains for several public health priorities, including tuberculosis. [1][2][3][4] As COVID-19 continued to spread throughout the world in 2020, tuberculosis service infrastructure and resources were diverted towards the pandemic response. Many vulnerable individuals with undiagnosed tuberculosis had difficulty using tuberculosis services due to limited access (clinic closures, health worker shortages or to avoid crowds), fear of contracting COVID-19 or stigma related to tuberculosis and COVID-19. 5 In a survey of 567 tuberculosis health professionals from 64 low-and middle-income countries, 233 (41%) said it was much harder or impossible for tuberculosis patients to seek care at facilities since the start of the pandemic. Likewise, 162 (29%) of respondents said it was difficult or nearly impossible to provide tuberculosis diagnostic services. 6 Zambia is a high tuberculosis burden country where tuberculosis is a leading cause of mortality, especially among people living with human immunodeficiency virus (HIV). 7,8 Following the identification of the first two COVID-19 cases in Zambia on 18 March 2020, the government quickly implemented several public health measures to prevent and mitigate the spread of severe acute respiratory syndrome coronavirus 2, (the virus causing . In line with early reports from other countries, preliminary data in Zambia showed large reductions in tuberculosis notifications following the pandemic and related transmission risk mitigation measures. To address the adverse effects of the pandemic on tuberculosis outcomes and services, the National Tuberculosis and Leprosy Programme worked with implementation partners to design and implement a series of measures to improve tuberculosis case detection. Strategies included an enhanced surveillance system with active case-finding and activities to generate demand for services. As of December 2021, there were 210 195 confirmed COVID-19 cases and 3667 confirmed COVID-19-related deaths among the general population in Zambia. 9 However, it remains unknown how tuberculosis notifications in Zambia were affected by the pandemic and how notifications may have changed following the roll-out of several targeted tuberculosis-related activities to mitigate its impact. Therefore, we undertook an interrupted time-series analysis of national tuberculosis notification data to evaluate the impact of the pandemic and subsequent tuberculosis response measures and Objective To evaluate the impact of the coronavirus disease 2019 (COVID- 19) pandemic and the subsequent implementation of tuberculosis response measures on tuberculosis notifications in Zambia. Methods We used an interrupted time-series design to compare monthly tuberculosis notifications in Zambia before the pandemic (January 2019 to February 2020), after implementation of national pandemic mitigation measures (April 2020 to June 2020) and after response measures to improve tuberculosis detection (August 2020 to September 2021). The tuberculosis response included enhanced data surveillance, facility-based active case-finding and activities to generate demand for services. We used nationally aggregated, facility-level tuberculosis notification data for the analysis. Findings Pre-pandemic tuberculosis case notifications rose steadily from 2890 in January 2019 to 3337 in February 2020. After the start of the pandemic and mitigation measures, there was a −22% (95% confidence interval, CI: −24 to −19) immediate decline in notifications in April 2020. Larger immediate declines in notifications were seen among human immunodeficiency virus (HIV)-positive compared with HIV-negative individuals (−36%; 95% CI: −38 to −35; versus −12%; 95% CI: −17 to −6). Following roll-out of tuberculosis response measures in July 2020, notifications immediately increased by 45% (95% CI: 38 to 51) nationally and across all subgroups and provinces. The trend in notifications remained stable through September 2021, with similar numbers to the predicted number had the pandemic not occurred. Conclusion Implementation of a coordinated public health response including active tuberculosis case-finding was associated with reversal of the adverse impact of the pandemic and mitigation measures. The gains were sustained throughout subsequent waves of the pandemic. assess whether effects differed among key subgroups and by province.

Methods
We undertook a retrospective analysis of Zambia's tuberculosis case notifications recorded and reported from January 2019 through September 2021. All individuals diagnosed with drugsusceptible tuberculosis and registered for treatment initiation (notified) were included in the analysis, regardless of age or tuberculosis type (that is, new or retreatment).

Data sources
For the analysis we used nationally aggregated, routinely captured tuberculosis programme data supplied by the National Tuberculosis and Leprosy Programme. Every health facility in Zambia uses a paper-based system to document tuberculosis case notifications. Once an individual with tuberculosis starts treatment, their data are recorded in a facility-level register: demographic characteristics (including age, sex, HIV status), tuberculosis laboratory results (such as the Xpert® MTB/RIF assay or smear microscopy status), and diagnosis and treatment start dates. Every month, facilities send tuberculosis notification data to the district, where they are aggregated. Data are then sent to the province, where they are aggregated again before being sent to the National Tuberculosis and Leprosy Programme. Next, a monitoring and evaluation team reviews the data; any incomplete or irregular data are clarified and corrected before undergoing final national-level data aggregation. To illustrate the relationship with COVID-19 infections over the same period, we also used publicly available data on the daily number of COVID-19 cases reported in Zambia. 10 This study was a population-level analysis without the use of any patient identifiers. The University of Zambia biomedi-cal research ethics committee reviewed the study and classified it as exempt from human subjects' research review.

Setting
Measures to mitigate the risk of CO-VID-19 transmission were implemented nationally in late March 2020. Mitigation measures comprised travel restrictions, closures of restaurants, bars and educational institutions, and limitations of public gatherings to less than 50 people. While many of these measures had been scaled down by September 2020, most individuals in Zambia resumed their normal daily activities by August 2020, in large part because they could not afford the direct and indirect costs of continued compliance with the measures. As the pandemic progressed, mitigation measures were scaled up and eased in response to subsequent surges in COVID-19 infections (Fig. 1). 11 During the initial period in which pandemic mitigation measures were in place in Zambia, tuberculosis diagnostic and treatment services remained open throughout the country. Adaptations to existing tuberculosis services were made to protect vulnerable patients and staff and avoid overcrowding at facilities, including pausing community-based activities, providing prolonged tuberculosis medication refills, and undertaking telephone-based household contact tracing and tuberculosis treatment assessments. HIV services made similar adaptations. HIV services already provided extended multi-month dispensing of antiretroviral therapy (ART) for up to 6 months to clinically stable patients. This strategy was therefore substantially scaled up across Zambia at the onset of the pandemic. 12 In July 2020, following 3 months of declining tuberculosis case notifications (April to June), the National Tuberculosis and Leprosy Programme worked with several stakeholders to design and begin implementing a multicomponent strategy to improve tuberculosis surveillance and tuberculosis detection during the ongoing COVID-19 pandemic (Box 1).

Data analysis
We undertook an interrupted timeseries analysis using an ordinary leastsquares segmented linear regression model. We used the Cumby-Huizinga general test and autocorrelation plots (up to lag order = 12) to assess for autocorrelation and possible seasonality. We used Newey-West standard errors to account for autocorrelation. The user-written ITSA command within Stata version 17.0 (Stata Corp., College Station, United States of America) was used to conduct interrupted time-series analyses. 13 We defined three time periods for all interrupted time-series analyses: before the pandemic (Period 1, January 2019 to February 2020); during the pandemic, at the start of pandemic mitigation measures (Period 2, April 2020 to June 2020); during the pandemic, after implementation of tuberculosis casefinding measures (Period 3, August 2020 to September 2021). March 2020 and July 2020 represented phase-in periods and were not included in the data analysis. For March, the first COVID-19 cases were announced on 18 March 2020 and transmission mitigation measures were put into place later that month. For July, the scale-up of active tuberculosis case-finding activities across provinces did not begin until later in the month. Therefore, the effects of such measures on tuberculosis notifications in March and July, respectively, may not have been apparent until the following complete calendar month. Thus, two interruptions occurred between February and April 2020 (interruption 1) and between June and August 2020 (interruption 2).
We estimated several pre-and post-interruption measures of monthly tuberculosis case notifications and corresponding 95% confidence intervals (CI), including: (i) baseline trends in tuberculosis notifications before the pandemic (Period 1 trend); (ii) the absolute number of tuberculosis notifications and the relative per cent difference (compared with the counterfactual) in notifications immediately following interruptions 1 and 2 (change in level); (iii) the absolute trend in tuberculosis notifications during Periods 2 and 3 and the trend in these periods relative to the previous period (change in trend for Periods 2 and 3).
We also estimated the predicted number of tuberculosis case notifica-

Box 1. Activities implemented to mitigate the impact of COVID-19 on tuberculosis diagnosis and treatment services in Zambia Tuberculosis situation room
Each week, staff at the National Tuberculosis and Leprosy Programme held an online meeting to review national tuberculosis notification data. Participants were staff from implementing partner organizations and those involved in tuberculosis programming from the facility level through to the national level. Centralized teams and higher-performing facilities and districts (selected based on comparisons to set tuberculosis notification targets) were able to share best practices and provide support to lower-performing facilities and districts. To support this initiative, the National Tuberculosis and Leprosy Programme developed a weekly data reporting tool; built local capacity among facility-based health-care workers to improve record-keeping at all patient entry points; and began reporting data weekly rather than monthly. The permanent secretary of the health ministry officially launched this initiative to enhance support by leadership at each level. Since the initiative's launch in July 2020, weekly meetings have been sustained throughout the pandemic.

Facility-based active case-finding
Beginning in July 2020, we designed a strategy to motivate public health facilities across all provinces to begin systematic symptoms-based screening of all individuals attending the facility for active tuberculosis disease, regardless of the reason for presentation. We initially prioritized facilities contributing 80% of notifications at the district level. Activities included (i) training and reorienting health-care staff at each facility using standard operating procedures and standardized training slides; (ii) increasing the availability of digital chest X-rays for use as a triage test; and (iii) improving access to urine tuberculosis tests (lateral flow urine lipoarabinomannan assay). Over the subsequent months, we scaled up the number of tuberculosis diagnostic facilities providing facility-based active case-finding. Once saturation was achieved at diagnostic facilities, we introduced systematic, symptoms-based screening at tuberculosis treatment-only facilities (those with no tuberculosis diagnostic capacity). Sputum samples were collected from patients who screened positive for tuberculosis and sent to local diagnostic facilities for microbiological testing. We gave technical supervision, support and mentorship around this initiative to individual health facilities through weekly provincial-level tuberculosis situation room meetings. Since initial scale-up, all facilities that we have trained have sustained systematic symptoms-based screening throughout the pandemic.

Demand generation
Initially, community-based volunteers were working within health facilities to generate demand for tuberculosis services among community members already attending health facilities for any reason. After initial COVID-19 mitigation measures were scaled down in September 2020, we extended these activities to communities to raise further awareness about tuberculosis and COVID-19 and to encourage people to seek care when symptomatic. We developed a standardized guide to support these sensitization activities, focused on (i) raising awareness about tuberculosis symptoms and risks; (ii) educating individuals to request tuberculosis screening; and (iii) providing information about where tuberculosis screening and testing was offered.

Household contact tracing
To safely resume home visits, we provided community-based health-care workers with refresher training during lunchtime meetings, using standardized training guides. Trainings were led by staff of the health ministry and implementing partner organizations. Community-based healthcare workers were provided with specific education on COVID-19 infection and prevention control measures and were also supplied with washable masks and hand sanitizer.
Research COVID-19 and tuberculosis, Zambia Patrick S Lungu et al.
tions in September 2021 had the pandemic not occurred and had the subsequent tuberculosis response measures not been implemented (the counterfactual) and compared this with the number of tuberculosis case notifications predicted for September 2021 after accounting for both interruptions (that is, observed notifications). We conducted all analyses overall (at the national level) and according to age (≥15 years, < 15 years), HIV status (positive, negative), tuberculosis classification type (microbiologically confirmed pulmonary tuberculosis, clinically diagnosed pulmonary tuberculosis or extrapulmonary tuberculosis) and by province. We wanted to evaluate whether possible impacts on tuberculosis notifications differed by sex, but due to the way routine data are aggregated and reported in Zambia, monthly sex-disaggregated data were not available for interrupted time-series analyses. We therefore descriptively assessed quarterly sex-disaggregated notification data to evaluate whether differential impacts by sex may be present.

Before pandemic
Before the pandemic started in Zambia, tuberculosis notifications were slowly increasing by about 32 cases (95% CI: 22 to 42) notified per month, from a baseline of approximately 2890 cases (95% CI: 2804 to 2975) in January 2019 to 3337 cases (95% CI: 3252 to 3421) in February 2020 (Table 1; Fig. 1). Although trends in monthly tuberculosis notifications varied across subgroups and provinces, notifications were not declining in any subgroup or province before COVID-19 cases were found in Zambia (Table 1; Fig. 2; Fig. 3; Fig. 4).

After implementation of measures
The relationship between daily CO-VID-19 cases and monthly tuberculosis notifications is shown in Fig. 1 Fig. 4). Tuberculosis notifications among children were less immediately impacted than among adults ( Table 1; Fig. 2; Fig. 3). The immediate impact of pandemic measures differed substantially at the provincial level, with significant negative impacts observed in eight of 10 provinces. Lusaka (−32%; 95% CI: −37 to −27) and Southern provinces (−30%; 95% CI: −34 to −26) reported the largest immediate declines in tuberculosis notifications.
In the three months following the start of the pandemic and after a large initial decline in national tuberculosis notifications, notifications remained steady on a month-to-month basis and did not substantially differ from pre-pandemic monthly trends (Table 1; Fig. 1). Following large immediate declines in tuberculosis notifications among people living with HIV, the monthly trend in notifications increased compared with pre-pandemic trends ( Table 1; Fig. 4). In contrast, after a smaller immediate decline among HIV-negative individuals, monthly tuberculosis notifications decreased substantially compared with pre-pandemic trends. Significant declines in month-tomonth tuberculosis notification trends compared with pre-pandemic levels were observed among children, for microbiologically confirmed tuberculosis cases and across several provinces.
The trend in monthly national tuberculosis notifications between August 2020 and September 2021 remained steady (20; 95% CI: −3 to 44; Table 1). Stable or small increases in monthly tuberculosis notification trends during this period was observed across all subgroups and provinces. The proportional distribution of tuberculosis notifications according to sex did not significantly differ across each of the three periods between January 2019 and September 2021 (range of total proportion of notifications among males: 64% to 68%; Table 2).
In September 2021, the overall number of tuberculosis notifications was 4107 cases (95% CI: 3923 to 4292). This figure was not significantly different to the estimated number of notifications assuming a continuation of prepandemic trends (difference: 229; 95% CI: −100 to 558; Table 3; Fig. 1). The number of tuberculosis notifications in September 2021 in each subgroup and in nine of 10 provinces were similar to or exceeded the number of tuberculosis notifications predicted for September 2021, assuming a continuation of prepandemic trends.

Discussion
We found that national tuberculosis notifications in Zambia immediately declined by 22% following the confirmation of COVID-19 cases and implementation of mitigation measures designed to stem further COVID-19 transmission. There were substantial differences in the immediate and subsequent impact of the  Notes: We analysed data from nationally aggregated, facility-level routine tuberculosis notification data. Period 1 was before any COVID-19 cases were reported. Period 2 was immediately following confirmation of COVID-19 cases and implementation of national pandemic mitigation measures. Period 3 was after roll-out of tuberculosis response measures across provinces, including tuberculosis active case-finding activities. March 2020 and July 2020 represented phase-in periods and were not included in the data analysis. No. of cases is the number of tuberculosis case notifications. Percentage difference is the relative difference in tuberculosis notifications immediately following a time interruption compared with the counterfactual. The data underpinning the calculations are visually represented in Fig. 1, Fig. 2, Fig. 3 and Fig. 4, where the observed number of tuberculosis notifications in the first month immediately following a time interruption are compared against the predicted counterfactual number of tuberculosis notifications in that same month assuming that there was no change in the trend of monthly tuberculosis notifications from the previous period.
Research  and tuberculosis, Zambia Patrick S Lungu et al.
pandemic on tuberculosis notifications among subgroups and across provinces in Zambia. The immediate effect was most pronounced among people living with HIV, who had threefold higher declines than HIV-negative individuals, and in Lusaka province, which accounts for more than 40% of tuberculosis notifications in Zambia. 7 Following the initial scale-up of several tuberculosis response measures in July 2020 tuberculosis notifications immediately increased by 45%. Notifications then remained stable over the subsequent months, despite two additional, larger surges of COVID-19, and were similar to pre-pandemic levels in September 2021. These data highlight the importance of careful and continued surveillance of important public health problems during the COVID-19 and future pandemics. The results point to the feasibility and positive impact associated with implementing coordinated public health responses to alleviate the detrimental effect of the pandemic. Given the success associated with the activities implemented, such initiatives will be continued as part of Zambia's national tuberculosis programmatic strategy. Studies in several high-tuberculosis burden countries in sub-Saharan Africa [14][15][16] and other parts of the world 17,18 also found an immediate detrimental impact of the pandemic and associated mitigation measures on tuberculosis diagnostic and treatment outcomes. The impact on national-level tuberculosis notifications in Zambia (22%) is slightly lower than what was reported from facilities in Nigeria (34%) 15 and Uganda (43%). 14 However, we found substantial differences in the effects of the pandemic on tuberculosis notifications in Zambia that ranged from no immediate impact in some provinces to immediate adverse impacts exceeding 30% in others. The reasons underpinning such heterogeneity may in part reflect urban versus rural differences. People in more rural settings may have perceived themselves at lower risk for COVID-19 and thus their health-seeking behaviours were less initially impacted. There may also be a differential impact of temporarily suspending community-based tuberculosis activities, including sensitization activities and household case-finding.
We also found that immediate declines in tuberculosis notifications were more pronounced among people living with HIV compared with HIV-negative patients. However, this difference likely reflects far fewer people living with HIV attending facilities during the early stages of the COVID-19 pandemic. A national campaign had been initiated to make early contact with all people living with HIV and to provide extended ART refills (up to 6 months) in March and early April 2020. 12 Notably, in the first 3 months of the pandemic in Zambia, and following a large immediate decline in tuberculosis notifications, we found that case numbers either stayed flat or continued to decline among most subgroups and in most provinces. These  data have important implications, as even short disruptions in tuberculosis services and transient declines in tuberculosis notifications may result in thousands of additional tuberculosisrelated deaths and many new incident tuberculosis cases due to prolonged periods of infectiousness. Such a result could reverse hard-won progress in tuberculosis care by several years. 1,2,4,14 We implemented enhanced tuberculosis response activities, scaled up facility-based active case-finding measures, and increased access to improved tuberculosis diagnostic tools. These activities appeared to be associated with a marked improvement in tuberculosis notifications during the pandemic, and as the activities were sustained, so too was their positive impact on notifications as the pandemic continued. Due to our quasi-experimental study design, we cannot discern whether improvements in tuberculosis notifications are the sole result of these activities or may be due to other secular trends. 19 For example, around the same time tuberculosis response measures were initially being scaled up, there may have been more individuals presenting to health facilities in the context of either greater population mobility due to defiance of COVID-19 transmission control measures or less fear of contracting COVID-19. However, this explanation seems unlikely given that COVID-19 cases were increasing in Zambia while active tuberculosis case-finding strate-gies were being rolled out. Notably, following the implementation of carefully coordinated steps to bolster tuberculosis diagnoses, there was a large, immediate increase in tuberculosis notifications. This positive effect was seen across nearly all subgroups and provinces and was directly preceded by relatively stable tuberculosis notification trends. Collectively, this provides compelling evidence that the implementation of tuberculosis response measures was responsible for increasing tuberculosis notification rates to pre-pandemic levels.
The strengths of this analysis include the use of routine, national-level programmatic data and of disaggregated analyses among several key subgroups and in all Zambian provinces. This method allowed us to assess for possible differential trends in tuberculosis notifications before and during the COVID-19 pandemic. Furthermore, we analysed notification trends across three periods covering 33 months, including multiple waves of the pandemic, which allowed us to determine that gains in notifications following implementation of tuberculosis response activities were sustained. There were some limitations, however. We were unable to determine what factors underpinned the immediate decrease in tuberculosis notifications in Zambia following the emergence of COVID-19 cases. However, the change likely reflects both individual-and health-system-related factors. For example, individuals may have been reluctant to seek care for their symptoms due to fear or stigma. 20 Additionally, during the early pandemic, some laboratory technicians refused to process specimens over safety concerns. This issue was resolved Table 3.

Comparison of predicted and actual number of monthly tuberculosis case notifications in Zambia by key subgroups and provinces, September 2021
after providing education on COVID-19 infection control measures and additional personal protective equipment. Under-notification and underreporting may have also contributed to declines in tuberculosis notifications during this period. Finally, due to the prolonged nature of tuberculosis treatment, coupled with the time required for reporting and aggregation of treatment data, we were unable to assess the impact of the pandemic on tuberculosis treatment completion rates. It will be important to monitor and formally evaluate for any detrimental effects of the pandemic on tuberculosis treatment completion rates.
In conclusion, the COVID-19 pandemic and the associated mitigation measures had a substantial impact on tuberculosis case notifications in Zambia. A carefully coordinated public health response, including active tuberculosis case-finding strategies, was feasible to implement and was associated with a return of tuberculosis case notifications to pre-pandemic levels. The gains were sustained throughout subsequent waves of the pandemic. Continued vigilance will be required during the ongoing pandemic to ensure high tuberculosis diagnosis and treatment coverage levels. ■