Time series analysis of pulmonary tuberculosis incidence in earthquake-affected and non-affected areas of Kermanshah Province Authors: Fereshteh Davoodi, Ehsan Mardani-Talaee, Yahya Salimi, Fatemeh Heydarpour, Gholamreza Abdoli Published: 2025 ABSTRACT AbstractBackgroundEarthquakes increase disease incidence and mortality due to overcrowding, poor housing and nutrition, stress, treatment discontinuation, and poor patient management. This study examined the impact of an earthquake on pulmonary tuberculosis (TB) incidence in three earthquake-stricken areas and 11 other districts of Kermanshah Province to better inform disease control planning.MethodsMonthly pulmonary TB case counts and population data for 14 districts of Kermanshah Province were obtained from the TB Registration Office and the Kermanshah Deputy of Medical Sciences Statistics Center, covering 2009–2018. The total cohort comprised 2,622 TB cases: 207 from three earthquake-affected districts (Sar-e-Pul-e-Zahab, Qasr-e Shirin, Salas-e Babajani) and 2,415 from eleven non-affected districts (e.g., Kermanshah, Islamabad, Paveh). The study period included 8 years pre-earthquake (2009–2016) and 2 years post-earthquake (2017–2018). The earthquake’s impact on monthly TB incidence trends was assessed using segmented regression, single-group Interrupted Time Series (ITS) analysis, and multiple-group ITS models comparing affected and control regions.ResultsThe impact of the earthquake on TB incidence varied across districts. While earthquake-affected areas showed a significant increase in TB cases (ß2= 0.059, P < 0.011) following the event, the TB incidence trend showed a pronounced decline (ß3= -0.0042, P < 0.001) with 207 observations (184 before, 23 after). Similar patterns were observed in other districts, with some experiencing significant increases (e.g., Islamabad, ß2= 0.00028, P < 0.001) and others decreases (e.g., Paveh, ß2=-0.00049, P < 0.001) in TB levels following the earthquake.ConclusionThe earthquake did not result in a sustained increase in TB incidence in affected regions. The findings may reflect effective TB control efforts, post-disaster public health interventions, or underreporting. Strengthening TB surveillance and long-term monitoring in disaster zones is essential for informed response planning.Keywords: Pulmonary tuberculosis, Earthquake, Interrupted time series analysis, Disaster epidemiology, Post-disaster surveillance Introduction Tuberculosis (TB) is one of the most serious public health challenges, continuing to impose a heavy burden on healthcare systems, especially in low- and middle-income countries [1–3]. In 2023, 10.8 million people were estimated to develop TB, and approximately 1.25 million people died from it [1]. TB, which is caused by Mycobacterium tuberculosis, chiefly affects the lungs, and its transmission is closely linked to biological, socioeconomic, and behavioral factors, including diabetes, smoking, malnutrition, poverty, and overcrowding [2, 3]. Although the Directly Observed Treatment Short-course (DOTS) strategy has enhanced adherence to treatment [4], numerous regions of the world still struggle to reach TB control targets because of limited resources, inadequate political commitment, and poor coordination within the healthcare system [5]. Apart from individual and system-level factors, external determinants such as natural disasters can complicate TB control efforts. For instance, earthquakes may worsen TB transmission or reactivation due to population displacement, crowding in shelters, interruption of healthcare services, malnutrition, stress-related immunosuppression, and disruptions in medication supply [6–8]. Still, empirical results from earthquake-affected regions are mixed; whereas some studies documented short-term rises in TB after major earthquakes, others found no clear change or even declining trends, underscoring the role of healthcare system resilience and the efficacy of post-disaster responses [9–15]. Iran is located in a seismically active zone, and large earthquakes put a strain on public health infrastructure. Nevertheless, there is a dearth of evidence on the longer-term impacts of earthquakes on TB trends in Iran. The 2017 Kermanshah earthquake—one of the most devastating ones in recent decades—led to considerable human and infrastructural losses and displaced thousands of people [16, 17]. Understanding how such an event affected TB notification patterns can inform preparedness planning and boost surveillance and service continuity in future crises. Thus, the present study used interrupted time-series analysis over 12 years to assess the effect of the 2017 Kermanshah earthquake on the incidence trend of pulmonary TB in affected and non-affected districts of Kermanshah Province [6]. Method Study setting The present research was carried out in Kermanshah Province in western Iran near the Iraq border. Kermanshah comprises 14 administrative districts (counties) differing in socioeconomic, geographic, and demographic features. The intervention group included three western districts —Sar-e-Pul-e-Zahab, Qasr-e Shirin, and Salas-e Babajani—which were directly affected by the 2017 Kermanshah earthquake, a 7.3 magnitude event that struck near Ezgole on November 21, 2017. These areas experienced significant casualties, displacement, and infrastructure damage. The remaining 11 districts—Kermanshah, Islamabad, Paveh, Javanrud, Dalahu, Ravansar, Sonqor, Sahneh, Kangavar, Gilan-e Gharb, and Harsin—were not directly impacted by the earthquake and were used as the control group (Fig. 1). These districts were selected due to their geographic proximity, demographic similarity (in terms of population size, urban–rural distribution, and healthcare access), and the availability of uninterrupted TB surveillance data, making them suitable comparators in an ITS framework. Non-affected districts were chosen as controls due to their proximity to the earthquake epicenter, similar socioeconomic and demographic features, and comparable TB incidence trends before the earthquake. All control districts were within Kermanshah Province but outside the directly impacted regions; in this way, comparable access to healthcare and TB surveillance under the same provincial health system could be ensured. Geographic distribution of study areas in Kermanshah Province, Iran Study design This was a retrospective ecological time-series (panel) analysis based on monthly district-level data on pulmonary TB notifications from 2008 to 2019. The data were retrospectively retrieved from Iran’s primary healthcare TB surveillance system. Each district represented an observational unit with repeated monthly measurements. The 2017 Kermanshah earthquake was regarded as an interruption, and its effect was evaluated via segmented regression within an Interrupted Time Series (ITS) framework. Data on confirmed pulmonary TB cases were obtained from the Tuberculosis Registration Office and the National TB Registry System at Kermanshah University of Medical Sciences. Included were newly diagnosed pulmonary TB cases with bacteriological confirmation (via smear microscopy, culture, or GeneXpert MTB/RIF) or clinical diagnosis according to WHO criteria [11]. Cases with prior TB treatment were excluded. Case validation was conducted by trained physicians through cross-referencing laboratory data, radiographic evidence, and clinical records. Population data were sourced from the Department of Health’s Statistics Center. Post-disaster population estimates were adjusted to reflect displacement dynamics using data from the Kermanshah Disaster Management Organization, supplemented by satellite imagery, shelter registration data, and census updates from 2016 to 2020. These adjusted denominators were used to calculate monthly TB incidence per 100,000 population. This study utilized routinely collected, anonymized health data and did not involve direct contact with human participants or the collection of biological samples. Consequently, the need for informed consent was waived by the Research and Technology Ethics Committee of Kermanshah University of Medical Sciences. The study protocol was reviewed and approved by the same committee (IR.KUMS.REC.1399.1157). Statistical analysis Moving average (ARIMA) models, state-space formulations, and more recent machine learning–based approaches are useful for predicting and pattern detection; nevertheless, they are not specifically designed to quantify the causal effect of a distinct intervention. ITS analysis, by contrast, provides a robust quasi-experimental framework enabling formal estimation of the immediate and sustained impacts of a well-defined event (e.g., a natural disaster) on an outcome over time. ITS offers an interpretable and transparent method to assess post-disaster epidemiological variations by explicitly modeling pre-intervention trends and controlling for temporal autocorrelation. Given these benefits, ITS was chosen as the most appropriate methodological approach for the current study [12, 13]. Two ITS models were used to investigate the impact of the earthquake on the incidence of pulmonary TB. First, the time before the earthquake was used as the baseline (control period) for each district to analyze all earthquake-affected areas and districts of Kermanshah Province separately from the Segmented Regression Model and the single-group ITSA (based on Model 1). In this method, ordinary least squares was used to fit the model, and the Newey-West method was employed to estimate the variances to handle autocorrelations among observations. To control for other changes that might manifest instead of our intended intervention (earthquakes) and affect the interpretation of causation, we used Multiple-Group Itsmatch (Model 2) to compare the intervention group (earthquake-affected districts) with the control group (11 non-affected districts in Kermanshah province) that was similar at least in the magnitude of occurrence and pre-intervention trend. Kermanshah region was selected as the control group due to its demographic similarity to the earthquake-affected areas and the absence of direct earthquake damage. The following two models were used, fitted with lag [1, 14]: In Models 1 and 2: Yt is the sum of outcomes occurring at equal intervals per time t; Tt: is from the beginning of the study; Xt: a categorical variable indicating the intervention (periods before intervention 0, other cases); XtTt: an interaction term or interaction of time and intervention. β0: the intercept, which is the average level at which the dependent variable begins. β1: the slope of the outcome variable and the slope of the control group outcome variable before the intervention in Models 1 and 2, respectively. β2: a change in the outcome level that is indeed an immediate effect in Model 1, and a change in the outcome level that occurs immediately in the period after the start of the intervention in Model 2, as the control group. β3: In Models 1 and 2, it indicates the difference between the outcome slope before and after the intervention and the change in the outcome slope in the control group after the intervention, respectively. β4: The difference in the baseline level (intercept) is the outcome variable between the intervention and control groups. β5: the difference between the slope (trend) of the outcome variable between the intervention group and the control group before the intervention. β6: the difference in the level of the outcome variable between the intervention group and the control group during the intervention. β7: the difference between the slope (trend) of the outcome variable between the intervention and control groups after the intervention. For the analysis, a time variable representing the number of months since the start of the study (ranging from 1 to 131) was created. The earthquake variable was coded as the intervention indicator, with “0” denoting the pre-earthquake period and “1” the post-earthquake period. The pulmonary TB notification rate (i.e., number of reported new TB cases per 100,000 population) was calculated by dividing the number of monthly notified pulmonary TB cases by the estimated population in each district. As this study relied on routinely collected surveillance data, the term’ notification rate’ is used in place of ‘incidence’ (hereinafter) to more accurately reflect case reporting rather than true population-level disease burden. Prior to model fitting, exploratory data analysis (EDA) was conducted to assess the assumptions of linearity, stationarity, and residual autocorrelation. Time series plots of TB notification rates were visually inspected for non-linear trends or abrupt structural changes. No strong non-linear patterns were evident, supporting the use of segmented linear regression. Stationarity was assessed both visually and via the Augmented Dickey-Fuller test, which showed no significant unit roots in select districts. Since some districts (e.g., Dalahu) reported very few TB cases, especially in the post-earthquake period, monthly counts were aggregated at the district level, and log-transformed TB incidence rates were utilized as the dependent variable. This transformation helped stabilize variance and reduce the influence of sparse data. A sensitivity analysis excluding districts with very low counts provided similar estimates, thereby verifying the robustness of the results. In order to evaluate autocorrelation, we checked residual plots, examined autocorrelation function (ACF) and partial ACF (PACF) plots, and computed the Durbin–Watson statistic. The diagnostics provided evidence of first-order autocorrelation but no significant higher-order lags. Therefore, a lag(1) structure was applied in all models. Including higher-order lags did not enhance model fit (AIC/BIC) and occasionally created instability in parameter estimation. We applied Newey–West robust standard errors with lag(1) across all models to account for any residual autocorrelation and heteroskedasticity. We assessed model adequacy using several diagnostic criteria. The Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) were employed to compare alternative model specifications. Residual independence was assessed using the Durbin–Watson statistic and autocorrelation plots, whereas residual normality and homoscedasticity were inspected visually and checked with standard statistical tests. Based on these diagnostics, the final ITS models provided an adequate fit to the data. To improve the visual clarity of time series plots, simple exponential smoothing was applied to the monthly TB data. The smoothing parameter (α) was optimized to minimize mean squared error. This smoothing technique was used only for graphical representation—not for statistical inference—to reduce short-term noise and better visualize long-term trends. This approach is commonly employed in epidemiological time series analysis to assist interpretation, especially in the absence of strong seasonal effects [15]. Statistical modeling was performed using both single-group Interrupted Time Series Analysis (ITSA) and matched multiple-group ITS models [14, 16]. While our ITS models inherently account for secular trends through the time variable (Tt), we note that Kermanshah Province exhibits minimal seasonal TB variation due to its arid climate, as demonstrated in prior regional studies [17]. Furthermore, the use of monthly aggregated data inherently mitigates short-term fluctuations unrelated to the earthquake. Model diagnostics were conducted to assess the validity of ITS assumptions. All the statistical analyses, with statistical significance set at less than 0.05 and 95% confidence interval (CI), were carried out using Stata v. 14.2 (Stata Corp, College Station, TX, USA). Results During the study period (2008–2019), 2,622 pulmonary TB cases were reported in 14 districts of Kermanshah Province. The overall annual incidence rate demonstrates a gradual decline before the earthquake, followed by short-term fluctuations after the earthquake (Fig. 1). At the provincial level, the ITS analysis demonstrated a significant immediate reduction in TB incidence post-earthquake (level change = β₁, p < 0.05), followed by a non-significant variation in the post-earthquake trend (slope change = β₂, p > 0.05). Similar patterns were found in affected and non-affected districts, suggesting that the earthquake had merely a temporary impact on TB notifications as opposed to a sustained disruption. Most areas indicated consistent pre-earthquake declining trends in TB incidence when stratified by district. Some smaller districts (e.g., Dalahu and Eslamabad-e-Gharb) exhibited short-term irregularities, mainly ascribed to small case numbers. Due to the small number of TB cases in these districts, caution should be exercised in interpreting the corresponding estimates, since they may reflect random rather than meaningful variations. Still, all coefficients are presented for transparency and completeness. There was no evidence of a province-wide long-term rise in TB incidence. Model diagnostics showed that the final ITS models yielded an adequate fit to the data. The AIC and BIC values were lowest for the chosen models, and residual analyses indicated no evidence of significant heteroscedasticity or autocorrelation. These results confirm the appropriateness and robustness of the models utilized for assessing earthquake-related changes in TB incidence. Tables 1 and 2 list all detailed regression coefficients and confidence intervals for district-level and comparative models, guaranteeing full transparency of the statistical results while keeping the text concise. Table 1Estimation of the segmented regression model coefficients for the notification rate of tuberculosis (100,000 people) in earthquake-stricken areas and other districts of Kermanshah Province from March 2009 to February 2020districtTotal number of tuberculosis casesModel coefficientStandard deviationtFp>|t|Confidence interval 95%MinimumMaximumEarthquake affectedImpact of rapid change in levelβ20.0590.0232.57389.730.0110.0130.105Interplay between time and interventionβ3-0.00420.00084-5.080.001-0.0059-0.0026IslamabadImpact of rapid change in levelβ20.000280.0000793.60102.570.0010.000120.00044Interplay between time and interventionβ3-0.000183.16e− 6-5.760.001-0.000024-0.000012PavehImpact of rapid change in levelβ2-0.000490.00012-3.9825.420.001-0.00074-0.00024Interplay between time and interventionβ3-0.0000346.28e− 6-5.480.001-0.000046-0.000021JavanrudImpact of rapid change in levelβ2-0.000330.00010-3.2453.200.002-0.00054-0.00013Interplay between time and interventionβ3-0.0000363.26e− 6-11.190.001-0.000042-0.00003DalahuImpact of rapid change in levelβ2-0.0170.0036-4.96167.530.001-0.025-0.010Interplay between time and interventionβ3-0.000410.00023-1.710.089-0.000880.000064RavansarImpact of rapid change in levelβ2-0.000180.00006-2.73274.320.007-0.00032-0.000051Interplay between time and interventionβ3-0.0000373.54e− 6-10.700.001-0.000044-0.000030SonqorImpact of rapid change in levelβ2-0.0000660.000052-1.2788.260.208-0.000170.0000370Interplay between time and interventionβ3-0.0000182.44e− 6-7.630.001-0.000023-0.000013SahnehImpact of rapid change in levelβ2-0.000300.00013-2.27111.720.025-0.00056-0.000038Interplay between time and interventionβ30.0000355.88e− 65.990.0010.0000230.000046KermanshahImpact of rapid change in levelβ20.0270.0251.09190.180.277-0.0220.078Interplay between time and interventionβ3-0.00650.0011-5.500.001-0.0089-0.0042KangavarImpact of rapid change in levelβ2-0.0110.0020-5.84246.420.001-0.015-0.0078Interplay between time and interventionβ3-0.00010-0.00063-1.580.115-0.00022-0.00025Gilan-e GharbImpact of rapid change in levelβ21.60e− 60.000110.0148.240.98-0.000210.00022Interplay between time and interventionβ3-0.0000545.02e− 610.860.001-0.000064-0.000044HarsinImpact of rapid change in levelβ2-0.000520.00010-5.28132.470.001-0.00072-0.00033Interplay between time and interventionβ30.0000283.78e− 6-7.650.001-0.000036-0.000021Regression using standard deviation Newey-West; Maximum lag:1; Number of observations = 131; df (3 and 127) Prob > F = 0.001 Estimation of the segmented regression model coefficients for the notification rate of tuberculosis (100,000 people) in earthquake-stricken areas and other districts of Kermanshah Province from March 2009 to February 2020 Regression using standard deviation Newey-West; Maximum lag:1; Number of observations = 131; df (3 and 127) Prob > F = 0.001 Table 2Estimation of the segmented regression model coefficients for the f of tuberculosis (100,000 people) in earthquake-stricken areas and other districts of Kermanshah Province from March 2009 to February 2020Total number of tuberculosis casesModel coefficientStandard deviationtp > tConfidence interval 95%MinimumMaximumInterceptβ01.060.02936.120.0011.011.12Pre-intervention consequence slopeβ1-0.00390.00041-9.420.001-0.004-0.003Impact of rapid change in levelβ20.0270.0251.090.27-0.0220.078Interplay between time and interventionβ3-0.0060.0011-5.500.001-0.0089-0.004Difference in the level between treatment and control prior to interventionβ4-0.0430.034-1.240.21-0.1110.025Difference in the slope between treatment and control prior to interventionβ5-0.000090.0005-0.190.84-0.00110.000912Difference in the level between treatment and control in the period immediately following intervention initiationβ60.0320.0340.930.35-0.0350.099Difference between treatment and control in the slope after initiation of the intervention compared with pre-interventionβ70.00220.00141.560.119-0.00050.0051Post-intervention linear trend (Treated unit)β1+ β3+ β5+ β7-0.00830.0008-10.510.001-0.0099-0.0068Post-intervention linear trend (Controls)β1+ β3-0.01050.0011-9.400.001-0.0127-0.0083Difference of linear trend after intervention between control and treatmentβ5+ β70.00220.00141.590.111-0.00050.0049Regression using standard deviation Newey-West; Maximum lag:1; Number of observations = 129; F(3 and125) = 205.89; Prob > F = 0.001 Estimation of the segmented regression model coefficients for the f of tuberculosis (100,000 people) in earthquake-stricken areas and other districts of Kermanshah Province from March 2009 to February 2020 β1 + β3 + β5 + β7 β1 + β3 β5 + β7 Regression using standard deviation Newey-West; Maximum lag:1; Number of observations = 129; F(3 and125) = 205.89; Prob > F = 0.001 Overall, these results suggest that although the 2017 Kermanshah earthquake induced a temporary fluctuation in TB notification rates, the long-term trend of TB incidence in the province remained stable under the existing TB control system. Earthquake-affected areas In earthquake-affected districts, TB incidence showed an immediate rise post-earthquake (level change, β₂ = 0.059, P = 0.011), followed by a significant declining trend during the post-disaster period (trend change, β₃ = − 0.0042, P < 0.001). In total, 207 monthly observations were analyzed (184 before and 23 after the earthquake). Islamabad district In investigating the monthly pulmonary TB notification rate in Islamabad following the earthquake, a noteworthy rise in the levels was found (ß2= 0.00028; P < 0.001), with 163 patients (138 before and 25 after the earthquake). Nonetheless, the trend in pulmonary TB notification rate exhibited a more pronounced downward trajectory after the earthquake than before (ß3= -0.00018; P < 0.001) (Table 1; Fig. 2.B). Monthly pulmonary tuberculosis notification rate (per 100,000 population) before and after the 2017 earthquake in Kermanshah Province, Iran (2009–2020). Each subpanel (A–L) displays Interrupted Time Series (ITS) results for a specific district or region in Kermanshah Province. The vertical dashed line marks the timing of the earthquake (November 2017). Solid lines represent fitted regression slopes before and after the earthquake. Dots show observed monthly TB notification rate values. Panel A: Aggregated data from the three earthquake-affected areas (Sar-e-Pul-e-Zahab, Qasr-e Shirin, Salas-e Babajani). Panels B–L: Data from individual districts in the province Paveh district In examining the monthly pulmonary TB notification rate, with 122 patients observed (105 before and 17 after the earthquake) in Paveh, a significant decrease in the level was observed after the earthquake (ß2=-0.00049; P < 0.001). However, the trend in pulmonary TB notification rate exhibited a further decline after the earthquake compared to before (ß3= -0.000034, P < 0.001) (Table 1; Fig. 2.C). Javanrud district In investigating the monthly pulmonary TB notification rate in Javanrud, a significant decline in the level was found after the earthquake, with 117 patients (96 before and 21 after the earthquake) (ß2=-0. 00033; P < 0.002). Furthermore, the trends in pulmonary TB notification rate exhibited a more pronounced downward trajectory after the earthquake than before (ß3=-0.000036; P < 0.001) (Table 1; Fig. 2.D). Dalahu district In examining the monthly pulmonary TB notification rate, with 122 patients observed (105 before and 17 after the earthquake) in Paveh, a significant decrease in the level was observed after the earthquake (ß2=-0.00049; P < 0.001). However, the trend in pulmonary TB notification rate exhibited a further decline after the earthquake compared to before (ß3= -0.000034, P < 0.001) (Table 1; Fig. 2.C). Javanrud district In investigating the monthly pulmonary TB notification rate in Javanrud, a significant decline in the level was found after the earthquake, with 117 patients (96 before and 21 after the earthquake) (ß2=-0. 00033; P < 0.002). Furthermore, the trends in pulmonary TB notification rate exhibited a more pronounced downward trajectory after the earthquake than before (ß3=-0.000036; P < 0.001) (Table 1; Fig. 2.D). Dalahu district Analyzing the monthly notification rate of pulmonary TB in Dalahu revealed a substantial reduction in the level after the earthquake, supported by 32 total observations (29 before and 3 after the earthquake), with a significant decrease (ß2 = -0.017; P < 0.001). However, the post-earthquake trend in pulmonary TB notification rate showed a further decline compared to the pre-earthquake period, although this difference was not statistically significant (ß3 = -0.00041; P < 0.089) (Table 1; Fig. 2.E). Ravansar district A noticeable decrease in the monthly notification rate of pulmonary TB was observed after the earthquake in Ravansar. There were 60 observations (54 before and 6 after the earthquake), showing a significant decline (ß2= -0.00018; P < 0.007). Additionally, the trend in pulmonary TB notification rate displayed a more pronounced downward trajectory after the earthquake compared to before (ß3= -0.000037; P < 0.001) (Table 1; Fig. 2.F). Sonqor district In studying the monthly notification rate of pulmonary TB, 70 observations were made (60 before and 10 after the earthquake) in Sonqor. There was a decline in the level after the earthquake; however, it was not statistically significant (ß2=- 0.000066; P < 0.208). Additionally, the changes in the trend of pulmonary TB notification rate showed a more pronounced downward trend after the earthquake compared to before (ß3= -0.000018; P < 0.001) (Table 1; Fig. 2.G). Sahneh district In studying the monthly notification rate of pulmonary TB in Sahneh, a significant decrease in the level was observed after the earthquake, as evidenced by the 68 observations (53 before and 15 after the earthquake) (ß2= -0.00030, P < 0.025). However, the trend in pulmonary TB notification rate exhibited a more pronounced increase after the earthquake compared to before (ß3= 0.000035, P < 0.001) (Table 1; Fig. 2.H). Kermanshah district There was no significant change in the monthly notification rate of pulmonary TB in Kermanshah after the earthquake [1532 observations (1337 before and 195 after the earthquake)]. Moreover, there was no change in the level (ß2= 0.027; 95% CI= [-0.022 and 0.078]; P < 0.277). Nevertheless, the trend in pulmonary TB notification rate demonstrated a more pronounced downward trajectory after the earthquake than before (ß3= -0.0065) (P < 0.001), as illustrated in Table 1; Fig. 2.I. Kangavar district In analyzing the monthly notification rate of pulmonary TB in Kangavar after the earthquake, a notable reduction in the level was witnessed (ß2 = -0.011; P < 0.001) based on 81 observations (81 observations; 67 before and 14 after the earthquake). Furthermore, the changes in the trend of pulmonary TB notification rate did not show a significant decline after the earthquake compared to before (ß3= -0.00010; P < 0.115) (Table 1; Fig. 2.J). Gilan-e Gharb district In Gilan-e Gharb, an analysis of the monthly notification rate of pulmonary TB, consisting of 57 observations (51 before and 6 after the earthquake), revealed no significant change in the level (ß2= 1.60e (−6); 95% CI= [-0.000021 to 0.00022]; P < 0.98). Nevertheless, there was a more pronounced downward trend in pulmonary TB notification rate following the earthquake compared to before (ß3 = -0.000054; P < 0.0001) (Table 1; Fig. 2.K). Harsin district In Harsin, examining the monthly notification rate of pulmonary TB, involving 113 observations (98 before and 15 after the earthquake), demonstrated a notable decrease in the level after the earthquake (ß2 = -0.00052), which was statistically significant (P < 0.001). Additionally, there was a more distinct downward trend in pulmonary TB notification rate after the earthquake than before (ß3=-0.000028; P < 0.001) (Table 1; Fig. 2.H). Comparison of earthquake-affected and control areas The notification rate of pulmonary TB was compared between earthquake-affected areas (Sar-e-Pul-e-Zahab, Qasr-e Shirin, and Salas-e Babajani) and an appropriate control group (the 11 control cities in Kermanshah province). The starting point of the cases was estimated to be at 1.06. The pulmonary TB trend in the control group (Kermanshah) declined significantly before the earthquake (ß1 = 0.0039, 95% CI= -0.004 and − 0.003, P < 0.001). There was no significant increase in the level of pulmonary TB in the control group after the earthquake ((ß2= 0.027, 95% CI= -0.022 and 0.078, P < 0.011). However, the changes in pulmonary TB notification rate trend in the control group were more downward after the earthquake than before (ß3= 0.006, 95% CI= -0.008 and − 0.004, P < 0.001). There was no significant difference in TB onset between the control and intervention groups (ß4= -0.043, 95% CI= -0.011 and 0.025, P < 0.011). Besides, there was no significant difference in the disease trend between the control and intervention groups before the earthquake (ß5 = 0.00009, 95% CI = 0.0011 and 0.0009, P < 0.84). No significant difference in the disease level was observed between the earthquake-affected areas and non-earthquake-affected districts while the earthquake occurred (ß6= 0.032, 95% CI= -0.035 and − 0.099, P < 0.35). There was no significant difference between the disease rate in the earthquake-affected areas and the control group after the earthquake (ß7= 0.0022, 95% CI=- 0.0005 and 0.005, P < 0.011); furthermore, the linear trend difference of pulmonary TB after the earthquake between the control and earthquake-affected groups was not significant (Table 2; Fig. 3). Comparative time trends in pulmonary TB notification rate (per 100,000 population) before and after the earthquake in intervention (earthquake-affected areas) and control group (Kermanshah district). Regression lines indicate monthly trend estimates derived from ITS analysis, with pre- and post-earthquake slopes shown separately for each group Discussion This study examined the impact of the 2017 earthquake on the notification rate of pulmonary tuberculosis (TB) in districts in Kermanshah province, Iran. Our findings indicate that the earthquake’s effects on the TB notification rate varied across districts. While some districts showed an immediate increase in TB cases, others experienced a significant decrease in notification rate and/or trend after the earthquake. These results suggest possible regional differences in the earthquake’s short-term impact on TB notifications; nevertheless, these patterns should be interpreted with caution due to the limited number of cases in some districts and the exploratory nature of district-level analyses. In Kermanshah district, TB notifications gradually decreased post-earthquake, which may demonstrate temporary disruptions in diagnostic and reporting systems instead of a true decline in transmission. The earthquake caused severe infrastructure damage and led to the temporary closure of health facilities, which probably delayed case detection and follow-up. Similar patterns have been witnessed in post-disaster settings in Nepal and Japan [7, 18], where health service interruption caused short-term declines in TB notifications, followed by recovery when the services resumed. Thus, the observed variations in our study may indicate decreased case detection capacity rather than epidemiological enhancement. Notably, the earthquake-affected districts of Islamabad showed a short-term increase in TB notification rate after the disaster, which subsequently declined. This pattern is consistent with previous studies suggesting that increases in TB cases after disasters may be due to increased detection and intensified case-finding efforts during emergency relief phases [19–21]. Consistent with our results, previous research has shown that earthquakes do not uniformly lead to TB outbreaks. For example, studies in southern Italy, Haiti, and Wenchuan (China) have reported no significant long-term increases in TB notification rate after major earthquakes and have indicated stable or even declining trends over time [7, 19–23]. These studies suggest that well-established public health systems and targeted disaster responses can reduce the burden of infectious diseases even in crisis settings. This similarity may demonstrate the relative resilience of Iran’s healthcare system and the quick reactivation of TB diagnostic and treatment services post-earthquake. The national TB control program continued routine case finding and treatment supervision even in the affected districts, with the support of provincial health authorities. On the other hand, countries such as Haiti and China reported temporary rises in TB cases after major earthquakes, probably due to prolonged population displacement, destruction of laboratories, and disruption of medication supply chains [19–21]. Conversely, Italy and Japan witnessed minimal changes in TB trends, consistent with stronger health infrastructure and coordinated disaster response systems. These cross-country differences underscore the significance of health system preparedness and continuity of care in alleviating the post-disaster burden of TB [7, 10, 22, 24]. Our finding of no sustained increase in TB notification rate post-earthquake aligns with the chronic nature of TB, which typically has a long incubation period. Consequently, short-term disruptions may not immediately translate into higher notification rates [20, 22]. Moreover, post-disaster humanitarian efforts often involve enhanced case-finding, strengthened DOTS programs, and improved service delivery, which can counteract expected TB surges [25, 26]. Biologically, TB reactivation risk may increase due to immunosuppression caused by stress and malnutrition. However, this risk can be offset by post-disaster support such as improved nutrition, shelter, and medical care [27]. The continued downward trend in TB notification rate may also reflect the sustained impact of TB control strategies and socio-economic improvements after the earthquake [20, 22]. From the potential surveillance bias aspect, infrastructure damage (e.g., destruction of clinics, disruption of laboratory services) in earthquake-affected areas may have temporarily reduced TB case detection, especially in the immediate post-disaster period. While we accounted for population displacement and used data from the National Tuberculosis Registration System (which collects reports from mobile clinics and temporary shelters), lower-than-expected detection due to strained surveillance systems cannot be ruled out. For example, reduced diagnostic capacity at the main Sar-e-Pul-e-Zahab hospital—which served as a major TB referral center—may have delayed case confirmation. However, the sustained downward trend in TB notification rate in most districts over 2 years suggests that improved health services have reduced this bias in later stages. The immediate increase in TB cases observed in some areas could also reflect improved detection due to increased medical staff and mobile health units deployed during the relief phase. Given the incubation period of TB, it is unlikely that the earthquake itself was the direct cause of the short-term increase. Similar patterns were observed in Japan, where non-coastal areas showed a decrease in TB notification rate after the disaster. Disruptions in transportation and the dispersion of health care, particularly in rural and mountainous areas, may have hampered the recording and follow-up of TB cases [10]. Migration from earthquake-stricken areas to non-affected districts is another possible contributor to post-disaster TB patterns. Migrant populations often face poor housing, low-quality healthcare, and higher TB risk. Health system overload in host districts may disrupt TB services due to resource reallocation and increased burden on health workers [28]. Interestingly, in some districts, post-earthquake TB trends slowed or declined more sharply than before. This raises two critical questions for future investigation: (1) Did weakened disease management lead to underreporting or underdiagnosis? (2) Alternatively, did post-disaster rebuilding efforts and increased public health investment contribute to better TB control? Improvements in infrastructure, nutrition, and health access may act as protective factors in the longer term. Since TB is closely linked to socioeconomic status, economic revitalization after the earthquake may also have contributed to declining notification rate trends. Our comparative analysis found no statistically significant differences in TB levels or trends between earthquake-affected and control areas post-disaster. These results are consistent with findings from Nepal and Japan, where no significant differences were observed between disaster and non-disaster areas in TB notification rate or treatment outcomes [7, 18, 22]. Limitations and future directions Several limitations must be acknowledged to contextualize the findings and inform future research. First, future analyses should consider stratifying TB cases into newly diagnosed and recurrent infections, as these subtypes may have differing vulnerability to post-disaster disruptions. Recurrent TB often reflects prior treatment interruption, which may be exacerbated by earthquakes. Stratified modeling could help clarify whether short-term increases in notification rate are driven by reactivation or new transmission. Second, geographic Information System (GIS) mapping could be employed in future work to visualize spatial heterogeneity in TB notification rate changes across the province. Spatial overlays of earthquake intensity, population displacement, and access to care may yield insights into geographically clustered vulnerabilities. Third, while the 2-year post-earthquake follow-up period was constrained by data availability, it is consistent with studies showing that intermediate-term disaster effects on TB typically manifest within 1–3 years [24]. However, longer-term surveillance is essential to fully capture delayed effects such as latent TB reactivation or chronic health system strain. Fourth, our primary inference was derived from the multiple-group ITS model comparing affected versus control districts. District-level analyses were exploratory in nature, intended to illustrate regional heterogeneity. We did not adjust for multiple comparisons, which may increase the risk of false positives—particularly in districts with small post-earthquake sample sizes. Another limitation of our analysis is the use of a fixed lag(1) autocorrelation structure in the segmented regression models. While this approach is commonly recommended for short to medium-length monthly health surveillance data [29]and was informed by preliminary correlogram inspection, we did not conduct a full sensitivity analysis across multiple lag structures due to constraints in statistical resources and model convergence. A more comprehensive assessment of autocorrelation specifications would enhance confidence in the robustness of time series estimates and should be explored in future work. Moreover, though seasonality is unlikely to confound results in this setting, future studies in climatically variable districts should incorporate harmonic terms into ITS models. Finally, this study measured TB trends using notified case counts from the national surveillance system. As in many low- and middle-income settings, underreporting and underdiagnoses are likely, particularly in the immediate post-disaster period due to disruptions in health service delivery and diagnostic capacity. Therefore, our findings reflect trends in TB notifications, not the full notification rate of the disease in the population. Conclusion This descriptive time-series research investigated temporal changes in pulmonary TB notifications before and after the 2017 Kermanshah earthquake. The short-term variations observed probably indicate changes in case detection, reporting capacity, and service accessibility within the primary health-care system instead of true shifts in disease transmission. After the disaster, some health interventions and outreach activities may have temporarily increased TB detection, while diagnostic and reporting activities could have been disrupted or deprioritized in certain districts. Besides, a proportion of TB cases may have been diagnosed in the private sector and not reported to the public surveillance system. Given these factors and the relatively short post-earthquake observation period (2 years), the findings should be interpreted with caution. Future studies should examine the long-term impact of disasters on TB control, evaluate the performance and resilience of primary healthcare services, and investigate the role of reporting sources to better understand disaster-related impacts on TB surveillance and management. Acknowledgements The authors thank the Kermanshah University of Medical Sciences. Author contributions Fereshteh Davoodi, Ehsan Mardani-Talaee and Gholamreza Abdoli conceived the idea of the study. Yahya Salimi, Fereshteh Davoodi and Ehsan Mardani-Talaee developed the statistical analysis plan and conducted the statistical analyses. Yahya Salimi, Fereshteh Davoodi, Fatemeh Heydarpour and Gholamreza Abdoli contributed to the interpretation of the results. Gholamreza Abdoli, Fereshteh Davoodi, and Ehsan Mardani-Talaee drafted the original manuscript. All the authors reviewed the manuscript draft and revised it critically for intellectual content. All the authors approved the final version of the manuscript for publication. Funding This research was supported by Kermanshah University of Medical Sciences (grant number: 4000090). Data availability The data analyzed in the study are available from the corresponding author upon reasonable request. Declarations Ethical approval This study was approved by the Ethics Committee of Kermanshah University of Medical Sciences) IR.KUMS.REC.1399.1157). This study was conducted based on the Declaration of Helsinki. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement The data analyzed in the study are available from the corresponding author upon reasonable request.