|Year : 2021 | Volume
| Issue : 3 | Page : 234-242
Spatiotemporal analysis and seasonality of tuberculosis in Algeria
Schehrazad Selmane1, Mohamed L’hadj2
1 L’IFORCE, IFORCE, Faculty of Mathematics, University of Sciences and Technology Houari Boumediene, Algiers, Algeria
2 Beni Messous University Hospital Centre, Ministry of Health, Population and Hospital Reform, Algiers, Algeria
|Date of Submission||15-May-2021|
|Date of Acceptance||19-Jun-2021|
|Date of Web Publication||03-Sep-2021|
L'IFORCE, Faculty of Mathematics, University of Sciences and Technology Houari Boumediene, Algiers
Source of Support: None, Conflict of Interest: None
Background: This study aimed to describe the spatiotemporal distribution, to build a forecasting model, and to determine the seasonal pattern of tuberculosis (TB) in Algeria. Methods: The Box–Jenkins methodology was used to develop predictive models and GeoDa software was used to perform spatial autocorrelation. Results: Between 1982 and 2019, the notification rate per 100,000 population of smear-positive pulmonary TB (SPPTB) has dropped 62.2%, while that of extrapulmonary TB (EPTB) has risen 91.3%. For the last decade, the mean detection rate of PTB was 82.6%. At around, 2% of PTB cases were yearly reported in children under 15 years old, a peak in notification rate was observed in the elderly aged 65 and over, and the sex ratio was in favor of men. Between 52% and 59% of EPTB cases were lymphadenitis TB and between 15% and 23% were pleural TB. About two-third of EPTB cases were females and around 10% were children under the age of 15. The time series analysis showed that (1,1, 2) × (1, 1, 0)4 (respectively (0, 1, 2) × (1, 1, 0)4, (3, 1, 0) × (1, 1, 0)4) offered the best forecasting model to quarterly TB (respectively EPTB, SPPTB) surveillance data. The most hit part was the Tell followed by high plateaus which accounted for 96.6% of notifications in 2017. Significant hot spots were identified in the central part for EPTB notification rate and in the northwestern part for SPPTB. Conclusions: There is a need to reframe the set objectives in the state strategy to combat TB taking into account seasonality and spatial clustering to ensure improved TB management through targeted and effective interventions.
Keywords: Algeria, Box–Jenkins, seasonality, spatial autocorrelation analysis, spatial clustering, spatial distribution, tuberculosis
|How to cite this article:|
Selmane S, L’hadj M. Spatiotemporal analysis and seasonality of tuberculosis in Algeria. Int J Mycobacteriol 2021;10:234-42
| Introduction|| |
Tuberculosis (TB) is a communicable disease caused by bacillus Mycobacterium TB which commonly affects the lungs leading to pulmonary TB (PTB) but may also reach other bodily organs leading to extra-PTB (EPTB). The disease remains a major global health issue and ranks tenth among the top causes of death worldwide. In 2019, around 1.4 million people died and 10 million developed TB.
In former times, a country with a high TB prevalence, Algeria, has been able to join in the 1980s the group of countries with moderate prevalence by dint of a National TB Control Programme (NTBCP) setup in the 1960s. All cases of TB are treated and monitored by specialist staff in one of the 251 respiratory disease units (RDU) spread across the country, and all costs including care and TB drugs are fully borne by the Government.,
The spatiotemporal distribution and seasonality of TB have been investigated in several countries but not in Algeria; they will be addressed for the first time in the present study.,,,,,,,
| Materials|| |
The study area
Located in North Africa, Algeria is characterized by three geographic areas. The Tell, in the north, whose area represents 3.6% of the country's area is made up of 22 provinces and is home to an estimated 23,117,102 people with a population density of 265 persons/km2 as 2017. The High Plateaus, whose area represents 13.3% of the country's area, is made up of 17 provinces with an estimated population of 14,638,159 and a population density of 53 persons/km2 as of 2017. The Sahara, whose area represents 83.1% of the country's area, is made up of nine provinces with an estimated population of 3,994,737 and a population density of two persons/km2 as of 2017.
Tuberculosis is classified as exclusively PTB, exclusively EPTB, or EPTB with concurrent PTB involvement. The PTB is further disaggregated as sputum smear-positive PTB (SPPTB) and sputum smear-negative PTB. The sites of EPTB include pleural, lymphadenitis, bone and joint, urogenital, primary, and the remaining sites are named others. An EPTB case is notified bacteriologically confirmed (case from whom the taken biological specimen is positive by smear microscopy, culture or World Health Organization (WHO)-endorsed rapid diagnostics) or bacteriologically not confirmed. Specimens taken from relatively inaccessible sites are mostly paucibacillary, decreasing thus the sensitivity of diagnostic tests. In that case, various clinical or paraclinical examinations in addition to histopathological findings of lesion sites help the doctor to set the diagnosis, and the patient is notified as bacteriologically not confirmed. The notification rate was estimated as the total number of notified cases in a year divided by the estimated mid-year population and expressed in 100,000 people.
Any diagnosed or suspected patient with TB is referred to the nearest RDU as the only empowered to treat TB patients and to provide TB drugs. Each public health department in each of the country's 48 provinces routinely collects TB data from RDU. All that stuff is sent to the Ministry of Health (MH) and the NTBCP Committee on a monthly basis.
Data sources were MH, NTBCP, and WHO for TB notification cases and National Statistics Office for population data.,, No ethical permission for TB data was required for this retrospective study; the statutorily collected data were anonymized.
Time series analysis
The quarter-wise notification cases of TB, EPTB, and SPPTB for the period 2008-2017 were used to explore the disease pattern and to develop a forecasting model using the Box–Jenkins approach.,,, A thorough description of SARIMA model and the various statistical tests used to select the best model are given in Selmane and L'hadj papers., Data analysis and modeling were performed using Eviews 9 software (IHS EViews, Irvine, CA, USA) (https://www.eviews.com/).
Spatial autocorrelation analysis
The Moran's I statistic was used to analyze the global spatial autocorrelation. It supplies a unique measure of spatial autocorrelation for an attribute in a region as a whole. This statistic is a cross-product statistic between a variable and its spatial lag.
To specify the spatial neighborhood relationship, the first order Queen's contiguity rule was used. Values of Moran's I range from −1 to 1; the value 1 indicates highly positive autocorrelation and the value −1 indicates highly negative autocorrelation, while 0 indicates spatially random distribution. The higher the value is, the stronger the autocorrelation is. Its statistical significance is determined by the z-score and its associated P value. This statistic allows checking whether objects with similar attribute values are close to each other.
The Moran Index does not supply guidance on spatial patterns. To identify local clusters and local spatial outliers, we made recourse to the local Moran's I statistic developed by Anselin in 1995. To identify the location and types of clusters, the local indicators of spatial association derived by Anselin were used.
A positive value for I indicates that adjacent locations have similar high or low values which form clusters. A negative value for I indicates that adjacent locations have dissimilar values. Outliers in which a high value is surrounded primarily by low values (high–low), and outliers in which a low value is surrounded by high values (low–high).,,
Spatial autocorrelation analysis and mapping were carried out using GeoDa software (https://geodacenter.github.io/).
| Results|| |
Trends in tuberculosis in Algeria
[Figure 1] depicts the yearly evolution of the notified cases and the notification rate per 100,000 inhabitants of TB, SPPTB, and EPTB from 1982 to 2019. Of all 654,658 reported TB cases over the 38-year period, 271,661 (41.5%) were SPPTB, 301,567 (46.1%) were EPTB cases. The mean ± standard deviation (SD) was 17,227.84 ± 4146.13 cases (95% confidence interval [CI], 15,865.04–18590.64). The coefficient of variation (CV), defined as the ratio of the SD to the mean, for TB (CV = 0.24), EPTB (CV = 0.52), and SPPTB (CV = 0.14) indicates that there is more variability in EPTB in comparison to TB and SPPTB. The case notification rate per 100,000 population of SPPTB has dropped 62.2% from 33.2 in 1982 to 12.6 in 2019, while that of EPTB has risen 91.3% from 17.3% in 1982 to 33.1% in 2019. Between 1982 and 2000, SPPTB has been in the lead. In 2001, EPTB cases took over and have stayed ahead since. A gradual downward trend in case notification rate of all TB forms was observed until 1992 to resume to a gradual increasing trend until 2007 and since then, a slight TB fluctuation prevails around 55 cases per 100,000 people.
|Figure 1: Yearly evolution of overall tuberculosis, smear-positive pulmonary tuberculosis, and extrapulmonary tuberculosis notification cases and notification rates in Algeria from 1982 to 2019|
Click here to view
The detection rate of PTB, defined as the percentage of SPPTB in diagnosed PTB cases and which measures the effectiveness of NTBCP, varied from province to province from 53% to 100% and the national target set at 85% was reached only in 2008 [Figure 2]. Between 2008 and 2017, around 2% of PTB cases were yearly reported in children under 15 years old and a peak in notification rate was observed in the elderly aged 65 and over. The PTB sex ratio was in favor of men (around 1.4). More than half of the EPTB cases (varying between 52% and 59%) were lymphadenitis. The pleural TB comes in second place with between 15% and 23% of EPTB cases. About two-third of EPTB cases were females and around 10% were children under the age of 15.
|Figure 2: Pulmonary tuberculosis detection rate and bacteriological status of extrapulmonary tuberculosis from 2008 to 2017|
Click here to view
Tuberculosis in Algeria during COVID-19 pandemic
A total of 17,736 TB cases were notified in 2020, of which 12,405 were EPTB and 5331 were PTB, thereby recording a fall of 15%, 13%, and 19.5% in TB, EPTB, and PTB notification cases, respectively, compared to the previous year. The notification rate was estimated at 40.8 for TB, at 28.6 for EPTB, and at 12.3 for PTB. The detection rate of PTB was 77%, the cure rate of new SPPTB cases was 78.2%, the treatment success rate was 88.8%, and 76.8% of EPTB cases were confirmed bacteriologically.
Times series analysis
Given the incubation period of active TB, the data were aggregated into quarterly data and are listed in [Table 1]. The mean ± SD was 5389.53 ± 683.18 cases (95% CI, 5180–5600) for TB, 3092.85 ± 635.04 cases (95% CI, 2900–3290) for EPTB, and 1878.85 ± 232.86 cases (95% CI, 1810–1950) for SPPTB. The peak of notification cases was observed in the second-quarter and the seasonality was manifest [Figure 3].
|Figure 3: Hodrick–Prescott Filter with smoothing parameter λ =1600 for overall tuberculosis, extrapulmonary tuberculosis, and smear-positive pulmonary tuberculosis quarterly cases from 2008 to 2017|
Click here to view
|Table 1: Frequency of tuberculosis, extrapulmonary tuberculosis, and smear-positive pulmonary tuberculosis quarterly reported cases from 2008 to 2017|
Click here to view
The Box–Jenkins methodology was applied to find out the best fitting model using quarterly data from 2008 to 2017. One differencing step was required to achieve a stationary condition of the original time series TB, EPTB, and SPPTB. The model (1, 1,2) × (1, 1, 0)4 was selected as the best model with dependent variable DTB = D (TB). The model (0, 1, 2) × (1, 1, 0)4 was selected as the best model with dependent variable DEPTB = D (EPTB). The model (3, 1, 0) × (1, 1, 0)4 was selected as the best model with dependent variable DSPPTB = D (SPPTB).
The model outputs are displayed in [Table 2]. The three models are invertible, no root lies outside the unit circle. The residuals are normally distributed and white noise. These findings attest to the adequacy of the three models. The estimated quarterly DTB, DEPTB, and DSPPTB cases for the period 2008–2017 computed using estimation equation were closely approximated to the quarterly recorded cases [Figure 4] with an estimated R2 of 88.9% for DTB, 90.6% for DEPTB, and 81.2% for SPPTB, which suggests the adequacy of the selected models for TB forecasting.
|Figure 4: Quarterly actual and fitted cases of the first difference of tuberculosis, extrapulmonary tuberculosis, and smear-positive pulmonary tuberculosis and the corresponding correlogram of residuals: 2008–2017|
Click here to view
Spatial distribution of tuberculosis
The distribution of TB notification rate in Algeria is geographically imbalanced [Figure 5]. Descriptive statistics of TB, EPTB, and SPPTB notification rates by province are displayed in [Table 3]. During the decade 2008–2017, Médéa province recorded in 2017 the highest TB and EPTB notification rates (135.4 and 114.3 cases/100,000 inhabitants respectively), and Oran province recorded in 2009 the highest SPPTB notification rate (73 cases/100,000 inhabitants). In 2017, out of 48 provinces, 18 (respectively 18, 20) had a TB (respectively EPTB, SPPTB) notification rate higher than the national notification rate estimated at 53.4 (respectively 35.4, 14.8) per 100000 inhabitants. At provincial levels, TB notification cases were highly correlated to the population size and the population density; the Spearman correlation coefficients are r = 0.83 and 0.71, respectively.
|Figure 5: Geographical distribution of the overall tuberculosis notification rate per 100, 000 people, smear-positive pulmonary tuberculosis, and extrapulmonary tuberculosis notification cases in Algeria at the provincial level for the year 2017|
Click here to view
|Table 3: Descriptive statistics of the annual notification rates of tuberculosis forms|
Click here to view
Spatial autocorrelation analysis
Based on the results displayed in [Table 4], all Moran Index values were positive and ranged from 0.29 to 0.58 (0.28–0.61, 0.30–0.61) for TB (EPTB, SPPTB) notification rates. Moreover, all z-scores were greater than the critical value 1.64 at a level of significance of 0.05. Consequently, the notification rates exhibited spatial autocorrelation globally.
|Table 4: Spatial autocorrelation analysis of tuberculosis notification rate results (2008-2017)|
Click here to view
On the basis of local spatial autocorrelation analysis outcomes for SPPTB notification rate, the most significant hot spots were located in the northwest of the country, numbering no more than 8 and no <4. Seven out of the ten northwestern provinces of the country were permanently in the SPPTB hot spots and this since 2011. In contrast, the provinces numbering no more than 11 and no <6, exhibiting significant cold spots (low–low), were located in the east part of the country. Regarding the EPTB notification rate, the pattern of clustering was not stable; there was a fluctuation in the number and location of significant hot and cold spots. The hot spots varied from 1 to 7 and were located in the western part of the country then shifted to the central part of the country. Some provinces changed from nonsignificant to significant hot spots and vice versa. For instance, Ain Defla, Blida, and Msila provinces changed from nonsignificant in 2008 and 2009 to significant hot spots from 2010 to 2017. On the other hand, some provinces changed permanently of spots as was the case for Bouira province. The number of provinces in nonsignificant spots decreased from 38 in 2008 to 31 in 2017 [Figure 6].
|Figure 6: Spatial clusters and spatial outliers map for local spatial autocorrelation analysis of overall tuberculosis, extrapulmonary tuberculosis, and smear-positive pulmonary tuberculosis notification rate from 2008 to 2017. Blue (respectively red, pink, light blue, gray) corresponds to low–low (respectively high–high, high–low, low–high, not significant) clusters|
Click here to view
| Discussion|| |
Since 1962, TB has become a health priority in Algeria. It was one of the first diseases to benefit from an NTBCP. Over six decades, considerable efforts were made to improve the care, the detection rate of PTB, and treatment of TB. Nonetheless, EPTB is emerging as a major contributor to TB burden and the rise in EPTB cases is hampering the decrease in the overall TB notification cases. In 1982, the percentage of EPTB and SPPTB cases were 25.6% and 49.1%, respectively, while in 2019, these figures were changed to 68.3% and 26%, respectively. The change in SPPTB to EPTB ratio shows a downward trend over 1982–2019; a similar pattern was observed in Iraq and Oman.,,, Globally in 2019, 16% of the notified TB cases were EPTB, the lowest notification was estimated in the WHO Western Pacific Region (8%), and the highest was estimated in the WHO Eastern Mediterranean Region (24%) and varied from country to country within regions.
According to the WHO, the impact of the COVID-19 pandemic on TB services has been severe and a sharp drop in TB notification cases in 2020 was reported in some countries. The TB control in Algeria has been impacted by the spread of COVID-19 but not in a harsh way as in other countries. A delay in the assigned objectives was reported. The estimated TB notification rate in 2020 was at its lowest since 1982. The detection rate of PTB was 77%, while the average rate during the last decade was 82%. Urgent actions are needed to get the program on track, especially in the context of this pandemic; otherwise, we risk jeopardizing valuable progress made in the fight against TB over four decades.
Studies conducted in Algeria were TB units level epidemiological studies, laboratory studies, and very few used mathematical approaches.,,,, Our literature review found no studies conducted on spatial clustering of TB in Algeria and no studies conducted at a national level to determine the possible seasonal pattern of TB, except one study carried out in 2019 on TB data from Medea province. The purpose of this retrospective spatiotemporal study was to bring word on TB seasonality, spatial distribution, and spatial clustering in the country.
The seasonality was shown with a peak in spring. Similar results have been reported in Lahore, Pakistan. However, the peak period differs from one region to another. Seasonality of TB has been reported in Kuwait, India, and China where summer was shown as the peak season, while in Iran, a high number of notifications were reported in spring and in summer.,,,
Analysis and interpretation of collected data are important components in monitoring and surveillance of a disease., Additional tool used for monitoring the course of a disease is time series analysis.,,,, The outcomes of the time series analysis were upbeat for forecasting TB, EPTB, and SPPTB by means of the TB surveillance system data in Algeria. However, the fact remains that the built models were based only on notified cases, without taking into account other variables that may have an impact on TB incidence. The impact of factors influencing the TB incidence needs to be identified and selected factors should be integrated into the modeling process to improve the forecasting.
TB notification cases were highly correlated to population density; the population density cannot be considered as a factor affecting the spreading of TB, but it is rather overcrowded cities and densely populated areas that contribute to its spread.
The notification of TB cases was not evenly distributed across the country; the most hit part, in 2017, was the Tell (64.7% of cases) followed by the high plateaus (31.9% of cases). The global and local Moran's indexes were used to investigate the geographical clustering patterns of TB. So far, no study has analyzed the spatiotemporal dynamics of clusters of TB in Algeria. Significant hot spots were identified for EPTB notification rate in the central part and for SPPTB in the northwestern part of the country. These findings could contribute to targeted and effective management TB control. Studies on TB clustering have been undertaken in Morocco to identify TB spatial clusters and associated predictors, in Guangxi Zhuang autonomous region in China to determine spatial and temporal pattern of TB transmission, in Uganda to identify areas where both TB and HIV disease co-cluster.,,
Although this study met its stated goals, there are some limitations. First, TB data are stored at the province level which represents a broad aggregate level. The findings can be more relevant and informative if the spatial analysis is performed at the municipality level, a finer geographical level. Second, some potential risk factors data, such socioeconomic, environmental factors, and climate factors which may be related to TB clustering, were not included due to nonavailability of such data.
| Conclusions|| |
We are seeing a rising trend in EPTB forms and a downward trend in PTB. Add to this, the highest EPTB clusters were located in the north central part of the country and vary year by year, while the highest SPPTB clusters were located in northwestern part of the country. Plus, the seasonality of the disease has been demonstrated with a peak in the spring. Consequently, there is a need to reframe the set objectives and the actions to be carried out taking into account seasonality and spatial clustering of TB to make the current system more effective and to ensure, in real time, a better management of the disease by targeted and effective interventions.
It is mentioned in Tuberculosis data subsection: No ethical permission for TB data was required for this retrospective study; the statutorily collected data were anonymized.
Financial support and sponsorship
Conflicts of interest
There are no conflicts of interest.
| References|| |
Tuberculosis Control Handbook for Medical Staff. National Tuberculosis Control Program, Prevention Department, INSP. Edition 2011, Ministry of Health, Population and Hospital Reform, Algeria.
L'Hadj M, Amrane R, Fissah O. 139-PA10 Tuberculosis morbidity from 1982 to 1994 in Algeria. Tuber Lung Dis 1995;76:4.
Sadeq M, Bourkadi JE. Spatiotemporal distribution and predictors of tuberculosis incidence in Morocco. Infect Dis Poverty 2018;7:43.
Cui Z, Lin D, Chongsuvivatwong V, Zhao J, Lin M, Ou J, et al.
Spatiotemporal patterns and ecological factors of tuberculosis notification: A spatial panel data analysis in Guangxi, China. PLoS One 2019;14:e0212051.
Aturinde A, Farnaghi M, Pilesjö P, Mansourian A. Spatial analysis of HIV-TB co-clustering in Uganda. BMC Infect Dis 2019;19:612.
Khaliq A, Batool SA, Chaudhry MN. Seasonality and trend analysis of tuberculosis in Lahore, Pakistan from 2006 to 2013. J Epidemiol Glob Health 2015;5:397-403.
Akhtar S, Mohammad HG. Seasonality in pulmonary tuberculosis among migrant workers entering Kuwait. BMC Infect Dis 2008;8:3.
Kumar V, Singh A, Adhikary M, Daral S, Khokhar A, Singh S. Seasonality of tuberculosis in Delhi, India: A time series analysis. Tuberc Res Treat 2014;2014:514093.
Luo T, Sumi A, Zhou D, Kobayashi N, Mise K, Yu B, et al.
Seasonality of reported tuberculosis cases from 2006 to 2010 in Wuhan, China. Epidemiol Infect 2014;142:2036-48.
Moosazadeh M, Khanjani N, Bahrampour A. Seasonality and temporal variations of tuberculosis in the north of Iran. Tanaffos 2013;12:35-41.
National Statistics Office. Available from: https://www.ons.dz/
. [Last accessedon 2020 Apr 24].
Allard R. Use of time-series analysis in infectious disease surveillance. Bull World Health Organ 1998;76:327-33.
Box GE, Jenkins GM. Time Series Analysis, Forecasting and Control. San Francisco: Holden-Day; 1976.
Selmane S, L'Hadj M. Forecasting and prediction of scorpion sting cases in Biskra province, Algeria, using a seasonal autoregressive integrated moving average model. Epidemiol Health 2016;38:e2016044.
Selmane S. Dynamic relationship between climate factors and the incidence of cutaneous leishmaniasis in Biskra Province in Algeria. Ann Saudi Med 2015;35:445-9.
Anselin L. Local indicators of spatial association-LISA. Geograph Anal 1995;27:93-115.
Ali Z, Al-Obaidi M, Fadhaa S, Mankhi A, Misha'al K, Jassim I, et al. Epidemiological profile of tuberculosis in Iraq during 2011-2018. Indian Journal of Tuberculosis. Available online 19 January 2021. 10.1016/j.ijtb.2021.01.003.
Gaifer Z. Epidemiology of extrapulmonary and disseminated tuberculosis in a tertiary care center in Oman. Int J Mycobacteriol 2017;6:162-6.
] [Full text]
Joob B, Wiwanitkit V. Tuberculosis case detection rate: An observation on the change in 12-year period (2005–2016). Biomed Biotechnol Res J 2019;3:192-5. [Full text]
Al Mayahi ZK, AlAufi I, Al Ghufaili B, Al Balushi Z, Al Mughazwi Z, Mohammed E, et al.
Epidemiological profile and surveillance activity of tuberculosis in South Batinah, Oman, 2017 and 2018. Int J Mycobacteriol 2020;9:39-47.
Selmane S. The impact of treatment of latent tuberculosis on the incidence: The case of Algeria. World Academy of Science, Engineering and Technology. Int J Math Comput Sci 2014;8:961-6
Selmane S, L'Hadj M. Epidemiology and clinical characteristics of tuberculosis in Leon Bernard tuberculosis unit in Algeria. Int J Mycobacteriol 2020;9:254-60.
] [Full text]
L'Hadj M, Selmane S. Time series analysis of tuberculosis in Medea Province in Algeria. J Eng Technol Appl Sci 2019;4:85-94.
Ifticene M, Kaïdi S, Khechiba MM, Yala D, Boulahbal F. Genetic diversity of Mycobacterium tuberculosis
strains isolated in Algeria: Results of spoligotyping. Int J Mycobacteriol 2015;4:290-5. [Full text]
Ifticene M, Gacem FZ, Yala D, Boulahbal F. Mycobacterium tuberculosis
genotype Beijing: About 15 strains and their part in MDR-TB outbreaks in Algeria. Int J Mycobacteriol 2012;1:196-200. [Full text]
[Figure 1], [Figure 2], [Figure 3], [Figure 4], [Figure 5], [Figure 6]
[Table 1], [Table 2], [Table 3], [Table 4]