Spatiotemporal distribution and predictors of tuberculosis incidence in Morocco

Spatiotemporal distribution and predictors of tuberculosis incidence in Morocco Background: Tuberculosis (TB) is a major health problem in Morocco. This study aims at examining trends in TB in Morocco and identifying TB spatial clusters and TB-associated predictors. Method: Country-level surveillance data was exploited. Kendall’s correlation test was used to examine trends and an exploratory spatial data analysis was conducted to assess the global and local patterns of spatial autocorrelation in TB rates (Moran’s I and local indicator of spatial association [LISA]) at the prefecture/province level. Covariates including living in a prefecture versus living in a province, annual rainfall, annual mean temperature, population density, and AIDS incidence were controlled. An ordinary least squares regression was thus performed and both spatial dependence and heteroscedasticity were assessed. Results: A decrease in TB incidence rate was seen between 1995 and 2014 (Kendall’s tau b = − 0.72; P < 0.0001). However, while the period between 2005 and 2014 (10 last years) was considered, TB rate remained stable and as high as 84 per 100 000 population per year (95% CI: 83.7–84.3). The highest incidence rates were seen in Tanger- Assilah, Fez, Tetouen-M’diq Fnidaq, Inezgane Ait Melleoul, and Casablanca. From 2005 to 2014, while TB incidence rate was stable in Fez (P = 0.500), Tetouen-M’diq Fnidaq (P = 0.300), Casablanca (P = 0.500), Mohammadia (P = 0.146), Al Hoceima (P = 0.364), and Guelmim (P = 0.242), an increase in TB incidence rate was seen in Tanger-Assilah (Kendall’s tau = 0.49; P = 0.023) and a decrease in Salé (Kendall’s tau b = − 0,54; P = 0.014) and Inezgane-Ait Melloul (Kendall’s tau b = − 0,67; P = 0.0023). TB is strongly clustered in space (P-values of Moran’s I < 0.01). Two distinct spatial regimes that affect TB spatial clustering were identified (east and west). In the east, both annual rainfall (P = 0.003) and AIDS (P = 0.0002) exert a statistically significant effect on TB rate. In the west, only the living area (prefecture versus province) was associated with TB rate (P = 0.048). Conclusions: New information on TB incidence and TB-related predictors was provided to decision-making and to further pertinent research. Association between annual rainfall and TB may be of interest to be explored elsewhere. Keywords: TB, Meteorological data, Prefecture/province, AIDS, Population density, Morocco Multilingual abstract and middle income countries [1]. In Morocco, TB remains Please see Additional file 1 for translations of the abstract a major public health problem in spite of the efforts of the into the five official working languages of the United Ministry of health (MH) to alleviate it [2]. In 2015, 30 636 Nations. cases were reported; a total of 656 cases died from TB [3]. A national TB program was set at the end of the sev- enties to prevent, control, and eventually eliminate TB Background from Morocco. Standardized treatment regimens are Tuberculosis (TB) is one of the top 10 causes of death provided for free [4]. Two reference national laboratories worldwide [1]. In 2015, 10.4 million people around the provide testing for TB infection. In 2004, Morocco world fell ill from TB and a total of 1.8 million died from managed to reach the WHO objectives related to TB this disease. Over 95% of deaths from TB occur in low diagnosis and treatment [2]. Thus, in 2015, 83% of the * Correspondence: mina.sadeq@gmail.com cases were detected, 85% were treated for TB [2]. How- Environmental Epidemiology Unit, National Institute of Hygiene. Ministry of ever, TB incidence did not seem to decrease in Morocco. Health, Rabat, Morocco Full list of author information is available at the end of the article © The Author(s). 2018 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Sadeq and Bourkadi Infectious Diseases of Poverty (2018) 7:43 Page 2 of 13 The recent statistics showed that TB incidence in Morocco software version 1.6.7.9, March 2015, developed by Luc was as high as 89 per 100 000 population in 2015 [2]. Anselin (ASU, GeoDa Center for Geospatial Analysis and More may need to be explored about TB in Morocco. Computation, Arizona, USA), was used for these pur- Studies on spatial clusters of TB incidence that would poses. This was performed for each year under study. have given better understanding where interventions are most required are lacking in Morocco. On the other Data on potential predictors of TB hand, it is thought that TB prevails in prefectures rather Meteorological data by year than in provinces, and that the population density is a A centroid for each polygon that represents a province/ risk factor of TB in Morocco. Such claims require further prefecture was created and its GWS84 coordinates were research. Association between TB incidence and meteoro- determined. Those coordinates were used to get meteoro- logical factors has been cited elsewhere [5–7], but has not logical data (annual rainfall and annual mean temperature) been explored in Morocco yet. Cases of AIDS/ HIV are by province/prefecture. Climate monitoring data were more vulnerable to TB infection. Including AIDS/HIV obtained from the Global Climate Monitor [10]made incidence as a covariate in a regression model would best available under the Open Database License. This was predict TB incidence in Morocco. performed for each year under study. The QGIS soft- This work aimed at, first, examining trends in TB inci- ware version 2.0.1 ‘Dufour’ (FreeSoftwareFounda- dence rate in Morocco; second, examining spatial cluster- tion,Inc.,Boston,USA)was used. ing/clusters of TB incidence at the province/prefecture level; third, exploring non-spatial and spatial correlation Spatial regime between TB and some covariates in order to specify a Box maps of raw TB incidence rate were examined first model that would best predict TB in Morocco. Potential and foremost. High TB rates were seen in the west of predictors are living in a prefecture versus living in a Morocco, low TB rates in the remaining part of the province, population density, AIDS incidence, and me- country, and this may suggest the possible presence of teorological factors, i.e., annual rainfall and annual mean spatial heterogeneity in the form of spatial regimes. temperature. Diagnostics for spatial dependence and Thus, it was hypothesized that TB predictors may exert spatial heterogeneity were performed. Spatial study was a different effect across the west and east of Morocco. In limited to the last four years (i.e., 2011 to 2014) for a this study, these spatial regimes, i.e., west versus east, reason cited in “Results” section. were identified as shown in Fig. 2, and were evaluated as a dummy variable in the statistical spatial analyses. They Methods will be incorporated into the multivariate analyses that Geographical data, study area/population, and population adjust for spatial heterogeneity. density by year A polygon shapefile map of Morocco comprising 59 prov- HIV/AIDS by year inces/prefectures, developed for a previous study [8], was Data on HIV are not available and only those on AIDS used. The process of georeferencing, digitalizing, and incidence (by province/prefecture) of 2008 and 2009 are combining some provinces/prefectures is described else- [9], and this imposed a constraint as to the multivariate where [8]. Data on population size by province/prefecture regression. To deal with this, it was first opted for data were obtained from “Santé en Chiffres” files that were on AIDS of 2009 and it was checked whether the other made available by the Service of Studies in Health and potential predictors affect TB in the years between 2011 Health Information-Ministry of Health (SSHHI-MH) [9]. and 2014 and in 2009, similarly. If it is the case, AIDS The total population size was 32 187 000 inhabitants in rate can then be incorporated as an additional covariate 2011, it was 33 848 000 inhabitants in 2014. For each year in the multivariate regression to draw conclusions about under study, the population density by province/prefec- the effect of this variable on TB incidence. ture was calculated; thus, the population in a province/ prefecture was divided by the size of that province/ prefecture. Statistical analysis Trends in TB TB data by year An approximate two-sided Kendall’s rank correlation Data on both new cases and incidence of TB by province/ test was conducted to examine variation in TB incidence prefecture were obtained from “Santé en Chiffres” files from 1995 to 2014 (20 years) and from 2005 to 2014 made available by SSHHI-MH [9]. The raw incidence (10 years); the P-values and size effects of which are rates of TB by province/prefecture were calculated; out- provided. An annual Poisson incidence rate estimate of liers were looked for. Thus box map was displayed to TB and a Poisson rate confidence interval were also check for variance instability of the raw rates. GeoDa provided. The incidence rate is estimated as the number Sadeq and Bourkadi Infectious Diseases of Poverty (2018) 7:43 Page 3 of 13 of events observed divided by the time at risk of event used to perform Kendall’s rank correlation tests. Then, during the observation period. GeoDa was again used to perform bivariate Moran’s I test A Kendall’s rank test was performed to evaluate vari- to examine bivariate LISA between TB and covariates. In ation in the incidence rate from 2011 to 2014 in selected addition, an ordinary least squares (OLS) regression ana- prefectures and provinces. Statistics were calculated in lysis that took into account the previously identified spatial exact form. regimes was conducted. GeoDa was used for this purpose. These statistical methods were conducted using the Multicollinearity condition number, normality (Jarque-Bera StatsDirect statistical software version 3.0.194 (StatsDir- test), spatial dependence for weight matrix (row-standard- ect Ltd., Cheshire, UK). ized weights and Lagrange multiplier tests), and spatial heteroskedasticity (Breusch-Pagan test and Koenker-Bassett Global spatial clustering and LISA clusters of TB test) were all assessed. Then, the stability of predictors ef- The exploratory spatial data analysis approach [11–16] fect across regimes was evaluated. GeoDaSpace (ASU, was used to examine global and local patterns of spatial GeoDa Center for Geospatial Analysis and Computation, autocorrelation in TB rates and in covariates. A contigu- Arizona, USA) was used to perform a Chow test. ity raw standardized weight file was created. Queen contiguity, which defines spatial neighbours as those provinces/prefectures with shared borders and vertices, Results was chosen. Thus, the global univariate Moran’s I statis- Trend in TB incidence in Morocco from 1995 to 2014 and tic was examined. A positive and significant Moran’s I from 2005 to 2014 indicates clustering in space of similar TB rates. The A decrease in TB incidence rate was seen from 1995 local indicators of spatial association (LISA), which show to 2014 in Morocco (Fig. 1)(Kendall’stau b= − 0,72 the presence or absence of significant spatial clusters or P < 0.0001). The mean estimate of new TB cases (all outliers, was also examined. GeoDa software was used to forms confounded) was 27 642. A maximum number of perform these spatial analyses. 31 771 cases was reported in 1996, a minimum number of 25 473 cases in 2008. While the period between 2005 and Specifying a regression model of TB 2014 was considered, TB incidence rate was stable (Ken- Global clustering of potential predictors was examined first dall’stau= − 0.16; P = 0.242). During this 10-year period, and foremost. GeoDa was used to perform Moran’s I test. TB incidence rate mean was 85.3 per 100 000 population, Then, potential multicollinearity and linear correlation SD = 3.5. The Poisson annual incidence rate estimate was between TB and predictors were analysed. StastDirect was 84 per 100 000 population per year at 95% CI:83.7–84.3. Fig. 1 Incidence rate (Log10 scale) of tuberculosis, Morocco, 1995–2014. This figure depicts secular trends of tuberculosis in Morocco from 1995 to 2014. A decrease was shown during this 20-year period (Kendall’s tau = − 0.72; P = 0.0001). TB incidence rate remained relatively stable at 85.3 per 100 000 population from 2005 to 2014 Sadeq and Bourkadi Infectious Diseases of Poverty (2018) 7:43 Page 4 of 13 Descriptive analysis and trends in TB incidence in for each of the years under study (Fig. 3), we considered prefectures and provinces that rate smoothing is not necessary. Global univariate The highest incidence rate was shown in the prefecture Moran’s I statistics, univariate LISA cluster maps, and of Tanger-Assilah, identified as the unique outlier in univariate LISA significance maps of TB rates by year Figs. 2 and 3. Figure 4 shows TB incidence rates in all were all illustrated in Fig. 5. For each of the years under the prefectures of Morocco. High TB rates were seen in study, spatial clustering of high TB incidence rates was Tanger-Assilah, Fez, Tetouen-M’diq Fnidaq, Inezgane- shown in the north-west part of Morocco (Global Moran’s Ait Melloul, Casablanca, Mohammadia, and Salé. The I statistics were all statistically significant at 0.01 level or respective means (expressed per 100 000 population) less). Larache and Kenitra formed high spatial clusters and standard deviation for the period between 2005 and during the four years, Skhirate-Témara and Benslimane 2014 were 188 ± 12, 156 ± 12, 133 ± 13, 153 ± 26, 134 ± from 2011 to 2013, Salé from 2012 to 2014, Mohammadia 13, 122 ± 10, and 125 ± 8. As to provinces, high TB inci- in 2012, Tétouen- M’diq Fnidaq in 2013, Meknes and dence rates were seen in Al Hoceima (119 ± 7), Guelmim Khemissat in 2014, and Tanger-Assilah from 2013 to 2014 (104 ± 19), Khemissat, and Larache. From 2005 to 2014, (Fig. 5). Spatial clustering of low TB incidence rates was while TB incidence rate was stable in Fez (P = 0.5), located throughout the eastern and the southern part of Tetouen-M’diq Fnidaq (P = 0.300), Casablanca (P =0.5), the country as indicated in blue in Fig. 5. The significance Mohammadia (P = 0.146), Al Hoceima (P = 0.364), and level was tightened even more (to P = 0.01 instead of 0.05) Guelmim (P = 0.242), an increase in TB incidence rate to detect spatial clusters of low TB rates. For each of the was seen in Tanger-Assilah (Kendall’stau=0.49; P =0.023) years under consideration, the provinces of Errachidia and and a decrease in Salé (Kendall’stau b= − 0,54; P = 0.014) Ouarzazate were consistently significant even at that more and Inezgane-Ait Melloul (Kendall’stau b= − 0,67; demanding level. Two spatial outliers were identified, P = 0.0023). namely, Guelmim province and the prefecture of Fahs Anjra (Fig. 5). Global spatial clustering and LISA of TB incidence between 2011 and 2014 Univariate spatial autocorrelation, linear correlation, As TB incidence rate was approximatively stable from bivariate spatial autocorrelation, and regression model 2005 to 2014, spatial study was restricted to the last four There was evidence of a significant spatial pattern of years (i.e. 2011 to 2014). Data on recent TB cases reported annual mean temperature as well as of annual rainfall in in 2015, 2016 and 2017 are still not publicly available. each of the years under study (Fig. 6); spatial clustering Since box maps of TB raw rate showed only one outlier of AIDS incidence rates was also evident in 2009 (Global Fig. 2 TB raw rate box maps (using 1.5 as hinge). 2011–2014. Morocco. Examining raw TB rate box maps revealed that more than 25% of data were located in the west part of the country where high TB levels were shown (in yellow) while low TB levels were seen in the west part of the country (in green), and this identified two distinctive spatial regimes Sadeq and Bourkadi Infectious Diseases of Poverty (2018) 7:43 Page 5 of 13 Fig. 3 Descriptive analysis of TB incidence between 2011 and 2014. Morocco. The highest incidence rate, a unique outlier, was shown in the prefecture of Tanger-Assilah. High TB rates were also shown in the prefectures of Fez, Tetouen-M’diq Fnidaq, and Casablanca Moran’s I = 0.345; P = 0.001)(Table 1 and Fig. 7). The tau b tests were not statistically significant) (Table 1). prefecture of Casablanca formed the unique cluster of Annual rainfall showed spatial correlation stronger than population density in the country. linear correlation (Bivariate I >Kendall’s tau b), suggesting Annual rainfall, population density, and AIDS rates that this correlation is determined by geographic location. were strongly and positively correlated with TB, the The population density showed spatial correlation weaker mean annual temperature was not (all related Kendall’s than linear correlation, which indicates that this correlation Fig. 4 TB incidence rate by year (2011–2014), prefectures, Morocco. Legend: This figure depicts variation of TB incidence rate in the prefectures of Morocco. During the period from 2011 to 2014, the Highest TB incidence rates were seen in Tanger-Assilah, Fez, Tetouen-M’diq Fnidaq, Inezgane- Ait Melloul and Casablanca, all showed in the map. We looked more closely at TB variation in these prefectures during the period from 2005 to 2014. An increase in TB incidence rate was seen in Tanger-Assilah (Kendall’stau = 0.49; P = 0.023) while a decrease was seen in Inezgane-Ait Melloul (Kendall’stau b = − 0,67; P = 0.0023). TB incidence rate was stable in Fez (P = 0.5), Tetouen-M’diq Fnidaq (P = 0.300), and Casablanca (P = 0.5) Sadeq and Bourkadi Infectious Diseases of Poverty (2018) 7:43 Page 6 of 13 Fig. 5 LISA cluster maps and LISA significance maps of TB incidence, Morocco, 2011–2014. LISA indicates the presence or absence of significant spatial clusters or outliers for each province/prefecture. The province of Larache and the province of Kenitra formed spatial clusters of high TB incidence from 2011 to 2014, the prefecture of Skhirate-Témara and the province of Benslimane from 2011 to 2013, the prefecture of Salé from 2012 to 2014, and the prefecture of Tanger-Assilah from 2013 to 2014. Other areas formed temporary spatial clusters, including the prefectures of Mohammadia (2012), Tétouen- M’diq Fnidaq (2013), Meknes (2014), and the province of Khemissat (2014). All these cited provinces/prefectures were shown in red. Significant spatial clusters of low TB incidence were located in the east spatial regime. The provinces of Errachidia and Ouarzazatewere consistently significant even at more demanding level (i.e., P = 0.01 instead of 0.05). Two spatial outliers (in pink) were identified, namely the prefecture of Fahs Anjra (in the north) and the province of Guelmim (in the south) is partly determined by location. Table 1 also showed a sta- also seen between the population density and area (pre- tistically significant dispersion of TB rates correlated with fecture versus province) in 2009, 2011, 2012, 2013, and AIDS rates (Global Moran’s I = − 0.139; P = 0.019). 2014. Kendall’s tau b were 0.59, 0.61, 0.60, 0.61, 0.61, Multicollinearity was seen between annual rainfall and respectively; all of which were statistically significant at mean annual temperature in 2009, 2011, 2012, 2013, and less than 0.0001 level. 2014. Kendall’s tau b were − 0.41, − 0.60, − 0.59, − 0.44, Taking into account both correlation and multicolli- and − 0.28, respectively; all of which were statistically nearity, we chose a regression model that includes only significant at less than 0.001 level. Multicollinearity was the annual rainfall and the spatial regimes as predictors Sadeq and Bourkadi Infectious Diseases of Poverty (2018) 7:43 Page 7 of 13 Fig. 6 LISA cluster maps of annual rainfall and annual mean temperature, Morocco, 2009 and 2011–2014. Annual mean temperature as well as annual rainfall showed spatial clustering in all years under study. The south of Morocco has low rainfall and high temperatures. High annual rainfall clusters were shown in red (maps in left) of TB incidence rates for each of the five years under (Table 3), but annual rainfall has relatively stable effect study; thus, we performed OLS regression (Table 2). An- across the east and west (P > 0.05) (Table 3). This sug- nual rainfall was consistently related to TB only in 2013 gests the influence of underlying variables that exert a (P = 0.008), the spatial regime in all the years under different effect across the west and the east regimes. study (P ≤ 0.03 were statistically significant), indicating a Diagnostics of multicollinearity condition numbers did potential different effect of predictors across the west not suggest problems with the stability of the regression and the east regimes (Table 2). This led us to perform a results, that may be due to multicollinearity (all numbers spatial Chow test (Table 3). It was statistically significant related to the study years were less than 8) (Table 4). (P < 0.03) for each of the years under consideration The models assume normality of the errors, as indicated Sadeq and Bourkadi Infectious Diseases of Poverty (2018) 7:43 Page 8 of 13 Table 1 Non-spatial and spatial correlation between TB and its were 0.014 and 0.003, respectively) (Table 8); spatial de- potential predictors pendence was not present. Variable Univariate I (P)b(P) Bivariate I (P) Both annual rainfall and area affect TB in the years between 2011 and 2014 and in 2009, similarly (Tables 5 Outcome (year) and 6). This led us to incorporate AIDS rates as an add- TB (2009) 0.169 (0.035) itional covariate in the regression model (Table 9). In the TB (2011) 0.229 (0.009) east, annual rainfall as well as AIDS exerts a statistically TB (2012) 0.230 (0.007) significant effect on TB (respective P-values were 0.003 TB (2013) 0.389 (0.001) and 0.0002). In the west, only the living area was statisti- TB (2014) 0.387 (0.001) cally significant (P = 0.048), suggesting that individuals living in prefectures are more exposed to TB transmis- Covariates (year) sion than those living in provinces (Table 9). Rainfall (2009) 0.895 (0.001) 0.31 (< 0.001) 0.419 (0.001) Rainfall (2011) 0.894 (0.001) 0.196 (0.03) 0.350 (0.001) Rainfall (2012) 0.909 (0.001) 0.203 (0.03) 0.354 (0.001) Discussion Rainfall (2013) 0.854 (0.001) 0.380 (< 0.0001) 0.498 (0.001) To our knowledge, this is the first study to explore trends in TB incidence rate in Morocco, its spatial pat- Rainfall (2014) 0.832 (0.001) 0.373 (< 0.0001) 0.491 (0.001) terns and predictors, and the spatial patterns of its pre- Temperature (2009) 0.703 (0.001) −0.006 (0.95) −0.092 (0.09) dictors. TB incidence rate was stable at the country Temperature (2011) 0.724 (0.001) −0.097 (0.29) −0.207 (0.003) level. A close look at TB incidence in the prefectures Temperature (2012) 0.721 (0.001) −0.087 (0.33) −0.190 (0.001) and provinces that had the highest rates was informative. Temperature (2013) 0.622 (0.001) −0.027 (0.76) −0.091 (0.078) Included in such group are those that form spatial clus- Temperature (2014) 0.576 (0.001) 0.015 (0.87) −0.065 (0.190) ters, including Tanger-Assilah, Tetouen-M’diq, Salé and Guelmim, and those that do not form spatial clusters, Pop_Density (2009) 0.258 (0.007) 0.44 (< 0.0001) 0.225 (0.002) including Fez, Casablanca, Inezgane-Ait Melloul, Moha- Pop_Density (2011) 0.250 (0.006) 0,53 (< 0.0001) 0.246 (0.005) madia, and Al Hoceima. A significant increase in TB Pop_Density (2012) 0.252 (0.004) 0,49 (< 0.0001) 0.250 (0.004) incidence rate was seen in Tanger-Assilah. It is believed Pop_Density (2013) 0.252 (0.006) 0,56 (< 0.0001) 0.257 (0.004) that poverty and housing conditions in some communes Pop_Density (2014) 0.259 (0.004) 0,50 (< 0.0001) 0.182 (0.011) within this prefecture are the main risk factors of TB, AIDS (2009) 0.345 (0.001) 0.26 (0.006) −0.139 (0.019) and this may be subjective due to the scarcity of existing I = Moran’s Index statistic; P = P-Value; b = Kendall’s tau research that explores such issues in this prefecture or in others in Morocco. TB incidence rates in the prefec- ture of Fahs Anjra, a low-high outlier located between by the Jarque-Bera test (all P-values were > 0.05) (Table 4). Tanger-Assilah and Tetouen M’ diq, were influenced by Apart from that of the year 2009, all the P-values corre- TB rates in these neighbouring prefectures. TB incidence sponding to Heteroskedasticity-related tests were small, rates in Tetouen-M’diq, Mohamadia, El Hoceima, indicating heteroscedasticity was present (P < 0.05) Guelmim, and Casablanca remained stable, and this sug- (Table 4). Diagnostics for spatial dependence for weights gests that more efforts and research may be required. A matrix (row-standardized weights) did not indicated the high incidence cluster surrounded by low incidence clus- presence of spatial dependence. Lagrange Multiplier tests ters was observed in Guelmim Province; it was not seen in (lag and error tests) were not statistically significant 2009 (Fig. 7). Casablanca has a population density ≥ 150 (P > 0.05) (Table 4). inhabitants per km [17], so have other prefectures [17]; Given the strong heteroscedasticity and the results of however, only Casablanca was found to form a spatial the Chow test, we examined regression of TB incidence cluster of population density in Morocco, and this was not rates on covariates in the west and the east regimes, separ- expected. Besides, Casablanca did not form a spatial ately. Since multicollinearity was present between annual cluster of TB as it is believed. Attention has to be paid to rainfall and mean annual temperature, and between area all these previously cited areas that may require further and population density for each of the studied years in research and new efficient strategic measures. both spatial regimes, we performed OLS regression of TB rates on annual rainfall and area (Tables 5 and 6). In the In Morocco, TB spatial distribution is not random. An east, there was no evidence of spatial dependence (P-value enduring spatial clustering of high TB incidence rates related to lag tests were > 0.05) or heteroscedasticity (P- was seen in the north-western part of the country, that value related to error tests were > 0.05) (Table 7). In the of low TB incidence rates throughout the south and the west, heteroscedasticity was present only in 2012 (P-values east parts of the country. This supports our hypothesis related to Koenker-Bassett test and Breusch-Pagan test suggesting a different effect of potential predictors on Sadeq and Bourkadi Infectious Diseases of Poverty (2018) 7:43 Page 9 of 13 Fig. 7 LISA cluster maps and LISA significance maps of AIDS incidence (upper maps) and TB incidence (lower maps), 2009, Morocco. Some provinces, including Taroudant, Guemim, and Layoune and some prefectures, including Inezgane Ait Melloul and Agadir Ida Outanane form spatial clusters of high AIDS incidence rates. The prefecture of Meknes formed a high-low outlier. Other provinces, including Essaouira, Tata, and Essmara and the prefecture of Chtouka Ait Baha constitute a low-high clusters. A total of 10 spatial clusters of low AIDS incidence rates were identified. However, while examining AIDS clusters at more demanding significant level (i.e., 0.01 instead of 0.05), only low incidence clusters located in the east, namely the provinces of Khenifra, Errachidia, and Figuig were identified; no high incidence cluster was seen Table 2 Ordinary Least Squares Regresssion of TB rates in 2009 TB across the two key distinctive spatial regimes, i.e., and between 2010 and 2014. Morocco the west and the east, that were identified in this study. Covariate (year) Coefficient t-Statistic Probability Spatial patterns of further predictors were explored. Both annual rainfall and mean annual temperature Intercept (2009) 24.179 1.944 0.057 showed spatial clustering and were negatively correlated, Intercept (2011) 23.874 2.123 0.038 Intercept (2012) 22.568 1.957 0.055 Table 3 Spatial Regimes (West & East) Diagnostics. Chow test. Intercept (2013) 23.395 2.922 0.005 2009 and between 2010 and 2014, Morocco Intercept (2014) 23.863 2.186 0.033 Variable DF Value Probability Annual Rainfall (2009) 0.0587 1.710 0.093 Intercept (2009) 1 0.253 0.615 Annual Rainfall (2011) 0.037 1.465 0.149 Intercept (2011) 1 5.776 0.016 Annual Rainfall (2012) 0.051 1.691 0.096 Intercept (2012) 1 1.872 0.171 Annual Rainfall (2013) 0.058 2.743 0.008 Intercept (2013) 1 3.030 0.082 Annual Rainfall (2014) 0.060 1.808 0.076 Intercept (2014) 1 0.455 0.500 Spatial regime (2009) 29.713 2.229 0.030 Annual Rainfall (2009) 1 0.435 0.510 Spatial regime (2011) 45.149 4.240 < 0.0001 Annual Rainfall (2011) 1 0.185 0.667 Spatial regime (2012) 40.340 3.817 < 0.001 Annual Rainfall (2012) 1 0.096 0.757 Spatial regime (2013) 36.664 3.700 < 0.001 Annual Rainfall (2013) 1 0.471 0.493 Spatial regime (2014) 30.152 2.474 0.016 Annual Rainfall (2014) 1 3.613 0.057 Value F-statistic P (F-statistic) Global Chow test (2009) 2 6.767 0.034 Adjusted R Squared (2009) 21.25% 8.691 < 0.001 Global Chow test (2011) 2 22.603 < 0.0001 Adjusted R Squared (2011) 28.72% 12.686 < 0.0001 Global Chow test (2012) 2 18.348 0.0001 Adjusted R Squared (2012) 26.67% 11.548 < 0.0001 Adjusted R Squared (2013) 39.53% 19.962 < 0.000001 Global Chow test (2013) 2 18.009 0.0001 Adjusted R Squared (2014) 32.77% 15.138 < 0.00001 Global Chow test (2014) 2 14.359 < 0.001 Sadeq and Bourkadi Infectious Diseases of Poverty (2018) 7:43 Page 10 of 13 Table 4 Regression diagnosis by year Table 5 Ordinary Least Squares Regression of TB rates in 2009 and between 2010 and 2014. East regime. Morocco Multicollinearity condition number: Covariate (year) Coefficient t-Statistic Probability Year Number Intercept (2009) 30.949 2.384 0.027 2009 5.54 Intercept (2011) 21.399 2.753 0.012 2011 4.68 Intercept (2012) 23.721 2.677 0.014 2012 4.96 Intercept (2013) 26.060 3.918 0.001 2013 3.97 Intercept (2014) 36.031 3.685 0.002 2014 7.08 Annual Rainfall (2009) 0.026 0.605 0.552 Normality of errors (Jarque-Bera test) Annual Rainfall (2011) 0.041 1.990 0.060 Year DF Value P-Value Annual Rainfall (2012) 0.041 1.632 0.118 2009 2 0.522 0.770 Annual Rainfall (2013) 0.037 1.592 0.126 2011 2 0.404 0.817 Annual Rainfall (2014) 0.008 0.250 0.805 2012 2 0.471 0.790 Prefecture vs province (2009) 39.323 1.166 0.257 2013 2 2.858 0.240 Prefecture vs province (2011) 30.375 1.319 0.202 2014 2 1.289 0.525 Prefecture vs province (2012) 35.128 1.369 0.186 Heteroskedasticity Koenker-Bassett test Breusch-Pagan test coefficients: Prefecture vs province (2013) 39.445 1.703 0.103 Year DF Value P-Value) Value P-Value Prefecture vs province (2014) 30.400 1.202 0.243 2009 2 4.4506 0.108 5.338 0.069 Adjusted R Squared < 0.2 for all studied years; p (F-statistic) > 0.05 for all years except for the year 2011 for which P = 0.044) 2011 2 10.823 0.004 12.867 0.002 2012 2 9.534 0.009 11.400 0.003 2013 2 9.007 0.011 10.799 0.005 2014 2 6.340 0.042 4.311 0.116 Table 6 Ordinary Least Squares Regresssion of TB rates in 2009 and between 2010 and 2014. West regime. Morocco Spatial dependance for weight matrix (row-standardized weights) Covariate (year) Coefficient t-Statistic Probability MI/DF Value P-Value Intercept (2009) 40.299 1.703 0.099 Moran’s I (error) test Intercept (2011) 61.578 3.232 0.003 2009 −0.079 −0.408 0.683 Intercept (2012) 46.563 2.170 0.038 2011 −0.046 −0.002 0.998 Intercept (2013) 45.894 3.304 0.002 2012 −0.050 −0.049 0.961 Intercept (2014) 19.455 0.700 0.489 2013 0.015 0.702 0.483 Annual Rainfall (2009) 0.052 1.080 0.289 2014 0.174 2.497 0.013 Annual Rainfall (2011) 0.015 0.368 0.716 Lagrange Multiplier (lag) test Annual Rainfall (2012) 0.053 1.024 0.313 2009 1 0.363 0.547 Annual Rainfall (2013) 0.058 2.047 0.049 2011 1 0.062 0.804 Annual Rainfall (2014) 0.104 1.899 0.067 2012 1 0.065 0.799 Prefecture vs province (2009) 41.335 2.673 0.012 2013 1 0.442 0.506 Prefecture vs province (2011) 42.693 2.908 0.007 2014 1 3.460 0.063 Prefecture vs province (2012) 38.846 2.724 0.010 Lagrange Multiplier (error) test Prefecture vs province (2013) 34.896 2.917 0.006 2009 1 0.727 0.394 Prefecture vs province (2014) 31.151 2.630 0.013 2011 1 0.246 0.620 2012 1 0.292 0.589 Value F-statistic p (F-statistic) 2013 1 0.028 0.867 Adjusted R Squared (2009) 19.17% 4.914 0.014 2014 1 3.564 0.059 Adjusted R Squared (2011) 16.85% 4.446 0.020 Unit of analysis = 59; MI Moran’sI, DF Degree of Freedom Adjusted R Squared (2012) 16.84% 4.442 0.020 Adjusted R Squared (2013) 26.67% 7.182 0.003 Adjusted R Squared (2014) 25.12% 6.704 0.004 Sadeq and Bourkadi Infectious Diseases of Poverty (2018) 7:43 Page 11 of 13 Table 7 Regression diagnosis by year. East spatial regime Table 8 Regression diagnosis by year. West spatial regime Multicollinearity condition number: Multicollinearity condition number: Year Number Year Number 2009 3.64 2009 7.00 2011 3.31 2011 5.66 2012 3.20 2012 6.62 2013 2.56 2013 5.01 2014 3.66 2014 11.08 Normality of errors (Jarque-Bera test) Normality of errors (Jarque-Bera test) Year DF Value P-Value Year DF Value P-Value 2009 2 7.837 0.020 2009 2 1.833 0.399 2011 2 1.250 0.535 2011 2 0.721 0.697 2012 2 0.894 0.640 2012 2 2.038 0.361 2013 2 0.441 0.802 2013 2 0.276 0.802 2014 2 1.219 0.544 2014 2 0.670 0.715 Heteroskedasticity Koenker-Bassett test Breusch-Pagan test Heteroskedasticity Koenker-Bassett test Breusch-Pagan test coefficients: coefficients: Year DF Value P-Value Value P-Value Year DF Value P-Value Value P-Value 2009 2 1.952 0.377 4.112 0.128 2009 2 5.123 0.036 5.123 0.077 2011 2 1.738 0.419 1.680 0.432 2011 2 5.133 0.077 5.996 0.050 2012 2 1.260 0.533 1.207 0.547 2012 2 8.586 0.014 11.499 0.003 2013 2 0.672 0.714 0.586 0.746 2013 2 4.741 0.093 5.583 0.061 2014 2 2.591 0.274 1.389 0.499 2014 2 2.710 0.258 1.794 0.408 Spatial dependance for weight matrix (row-standardized weights) Spatial dependance for weight matrix (row-standardized weights) MI/DF Value P-Value MI/DF Value P-Value Moran’s I (error) test Moran’s I (error) test 2009 −0.042 0.320 0.749 2009 −0.188 −1.024 0.305 2011 0.06 1.020 0.308 2011 −0.121 −0.509 0.610 2012 0.145 1.414 0.157 2012 −0.143 −0.705 0.481 2013 0.033 0.742 0.458 2013 −0.073 −0.082 0.934 2014 0.358 2.942 0.003 2014 −0.0767 −0.166 0.868 Lagrange Multiplier (lag) test Lagrange Multiplier (lag) test 2009 1 0.148 0.700 2009 1 2.633 0.105 2011 1 0.274 0.600 2011 1 1.244 0.265 2012 1 0.234 0.629 2012 1 1.493 0.222 2013 1 0.093 0.760 2013 1 0.191 0.662 2014 1 0.844 0.358 2014 1 0.076 0.783 Lagrange Multiplier (error) test Lagrange Multiplier (error) test 2009 1 0.061 0.805 2009 1 0.419 0.518 2011 1 0.130 0.718 2011 1 0.912 0.340 2012 1 0.484 0.487 2012 1 1.266 0.261 2013 1 0.039 0.843 2013 1 0.333 0.564 2014 1 0.473 0.491 2014 1 0.366 0.545 Unit of analysis = 59; MI Moran’sI, DF Degree of Freedom Unit of analysis = 59; MI Moran’sI, DF Degree of Freedom Sadeq and Bourkadi Infectious Diseases of Poverty (2018) 7:43 Page 12 of 13 Table 9 Ordinary Least Squares Regresssion of TB rates in 2009. [18, 19]. In Morocco, inequalities in socio-economic Morocco conditions, including housing conditions, vary by prov- Covariate (year) Coefficient t-Statistic Probability ince/prefecture as well as inside prefectures. Increasing rainfall may affect over-crowded dwellings and cause Spatial East Regime dwellers to be at greater risk of developing TB. This might Intercept −13.562 −1.009 0.326 explain annual rainfall association with TB in the east re- Annual Rainfall 0.130 3.377 0.003 gime where the provinces of Morocco prevail, and may also Prefecture versus province 34.599 1.444 0.165 explain the connection between high spatial clusters of TB AIDS 20.538 4.549 0.0002 and rainfall in some prefectures in the west regime. Further Spatial West Regime research that uses new scientific approaches such as map- ping diseases and risk analysis at a small-area level are likely Intercept 20.643 0.591 0.559 to be required in the suspected areas. Annual Rainfall 0.0843 1.314 0.199 More focus on AIDS/HIV related to TB may be needed Prefecture versus province 35.565 2.059 0.048 in Morocco. In a previous study, AIDS was reported to be AIDS 4.890 0.770 0.447 prevalent in Agadir, Marrakesh and Casablanca [20]. A particular stakeholders’ focus may have been made on Value F-statistic P (F-statistic) these prefectures. Our study revealed the presence of five Adjusted R Squared (East) 0.495 8.19 3 0.001 high spatial clusters of AIDS incidence rates, all located in Adjusted R Squared (West) 0.181 3.431 0.029 the west regime. Examining spatial patterns of AIDS inci- East: Koenker-Bassett Statistic: 7.63 (P = 0.054); Jarque-Bera statistics: 0.08 dence at a more demanding significant level (i.e., 0.01 in- (P = 0.96); Multicollinearity condition number: 5.5; Spatial autocorrelation of stead of 0.05) yielded only low incidence clusters, formed residuals: Moran’s I: 0.24 (P = 0.81); Langrange multiplier (lag): 0.09 (P = 0.77); by Khenifra, Errachidia, and Figuig, all located in the east Langrange multiplier (error): 0.11 (P = 0.74) West: Koenker-Bassett Statistic: 7.80 (P = 0.19); Jarque-Bera statistics: 1.99 regime. This supports our previous finding with regard to (P = 0.37); Multicollinearity condition number: 11.0; Spatial autocorrelation of the association between TB and AIDS seen only in the residuals: Moran’s I: -0.96 (P = 0.34); Langrange multiplier (lag): 2.52 (P = 0.11); Langrange multiplier (error): 0.84 (P = 0.18) east spatial regime, and may require particular decision making’s attention and further small-area studies that con- and this was expected. The south of Morocco generally sider not only AIDS/HIV incidence but also socio- has low rainfall and high temperatures, the north of the economic and meteorological factors as well. country high rainfall and low temperatures. Previous This study has several strengths. It is based on reliable studies carried out elsewhere [5–7] pointed to a connec- data, obtained from “Santé en Chiffres” files, that are tion between temperature and TB and to a negative representative of the population of Morocco; the Climate correlation between rainfall and TB, and this is not con- monitoring data, obtained from the Global Climate sistent with the findings of this study that suggested that Monitor, was also used. Both data sources have been annual mean temperature was not correlated with TB, proven in this study to be reliable for epidemiological re- and that annual rainfall was positively correlated with search. Exploring spatial clustering of TB, including TB. Differences in meteorological factors between coun- spatial effect, and identifying two distinctive spatial re- tries might be one potential explanation. In our study, gimes at the country level are other important strengths annual mean temperatures varied approximatively be- of this study. This study has a few limitations. Like any tween 10 °C and 25.6 °C [data not shown], annual mean country-level surveillance data, some TB cases may not rainfall between 10.1 and 723 mm [data not shown]. In a be reported, but they may be minor as all the physicians study undertaken in China, the monthly average in the country are conscious of the importance of temperature varied between − 13.4 °C and 20.4 °C, the reporting TB cases, and TB diagnosis and treatment are monthly rainfall between 0 and 195.1 mm [5]. This free and available at the primary healthcare centres, raises the question of whether the bacterium Mycobac- which are the first level health facility. On the other terium tuberculosis can grow and spread especially well hand, aggregate rather than individual level data were in particular conditions of temperature, humidity and considered, which may be another common limitation of probably other meteorological factors. Another question this ecological study. In spite of this, this study did yield of whether there are underlying factors that may serve new informative TB-related findings that may contribute as an intermediary factors between meteorological con- to guide decision-making in Morocco and to urge to fur- ditions and TB trends is also raised. Previous ecological ther pertinent research studies on TB. studies have suggested that “broad socio-economic de- velopment, rather than the success of TB control pro- Conclusions grammes, is the main determinant behind the declining In Morocco, TB is not randomly distributed in space. trends of TB observed in many regions of the world” Two distinctive spatial regimes that affect TB spatial Sadeq and Bourkadi Infectious Diseases of Poverty (2018) 7:43 Page 13 of 13 clustering were identified. Planning of TB control and %20progr%C3%A8s%20d%C3%A9fis%20et%20perspectives.pdf. 2017. Accessed 11 Jul 2017. prevention measures/strategies that focuses on both sus- 3. Maroc. Ministère de la Santé . Direction de l'Epidémiologie et de Lutte pected prefectures and provinces may be required and Contre les Maladies. Situation Epidémiologique de la Tuberculose au Maroc would best address this health problem. A critical need – Année 2015. http://www.sante.gov.ma/Documents/2016/03/Situation_ %C3%A9pidimio_de_la_TB_au_Maroc__ is to conduct research that considers all the herein stud- 2015%20Fr%20V%2020%20%20mars.pdf (2016). Accessed 11 Apr 2017. (in ied risk factors and uses new scientific approaches such French). as geographic information systems (GIS) and risk ana- 4. Maroc. Ministère de la santé. Plan National d'accélération de la réduction de l'incidence de la tuberculose 2013–2016. (in French). http://www.sante.gov. lysis at a small-area level. More publicly available aggre- ma/Documents/Actualites/FICHE%20ACC%C3%89L%C3%89RANT%20VF.pdf. gated data on HIV or other health outcomes will Accessed 11 Apr 2017. definitely be a useful tool with regard to epidemiological 5. Rao HX, Zhang X, Zhao L, Yu J, Ren W, Zhang XL, et al. Spatial transmission and meteorological determinants of tuberculosis incidence in Qinghai research in Morocco. Questions are arisen about under- Province, China: a spatial clustering panel analysis. Infect Dis Poverty. lying risk factors linked to rainfall that may influence TB 2016;5:45. incidence and about association between annual rainfall 6. Khaliq A, Batool SA, Chaudhry MN. Seasonality and trend analysis of tuberculosis in Lahore, Pakistan from 2006 to 2013. J Epidemiol Glob Health. and TB, and this may be of interest to be explored 2015;5:397–403. elsewhere. 7. Reza Beiranvand R, Karimi A, Delpisheh A, Sayehmiri K, Soleimani S, Ghalavandi S. Correlation assessment of climate and geographic distribution of tuberculosis using geographical information system (GIS). Iran J Public Additional file Health. 2016;45(1):86–93. 8. Sadeq M. Spatial patterns and secular trends in human leishmaniasis Additional file 1: Multilingual abstract in the five official working incidence in Morocco between 2003 and 2013. Infect Dis Poverty. 2016;5:48. languages of the United Nations. (PDF 480 kb) 9. Maroc. Ministère de la Santé/Direction de la Planification et des Ressources Financières/Direction de la Planification et des Etudes/Service des Etudes et de la Planification Sanitaire. (Maroc.MS/DPRF/DPE/SEIS). Santé en chiffres. Abbreviations From 2004 ed to 2014 ed. http://www.sante.gov.ma/publications/Etudes_ AIDS: Acquired Immune Deficiency Syndrome; LISA: Local Indicators of enquete/Pages/default.aspx. Accessed 07 Jan 2017. (in French). Spatial Association; MH: Ministry of Health; SD: Standard Deviation; 10. Climatic Research Unit, University of East Anglia. Global Climate Monitor SSHHI: Service of Studies in Health and Health Information-Ministry of Health; http://www.globalclimatemonitor.org. Accessed 07 Jan 2017. TB: Tuberculosis; WHO: World Health Organization 11. Anselin L. Exploratory spatial data analysis and geographic information systems. In: Painho M, editor. New tools for spatial analysis. Luxembourg: Acknowledgements Eurostat; 1994. p. 45–54. We would like to thank the Service of Studies in Health and Health 12. Anselin L. Local indicators of spatial association-LISA. Geograph. Anal. 1995; Information-Ministry of Health. Rabat. Morocco. 27:93–115. 13. Anselin L. Interactive techniques and exploratory spatial data analysis. In: Availability of data and materials Longley PA, Goodchild MF, Maguire DJ, Rhind DW, editors. Geographical Reported cases of TB for all years under study and those of AIDS in 2009 information systems: principles, techniques, management and applications. were made available by the Service of Studies in Health and Health New York: Wiley; 1999. p. 251–64. Information-Ministry of Health (SSHHI-MH) [9]. 14. Anselin L. Exploratory spatial data analysis in a geocomputational environment. In: Longley PA, Brooks SM, McDonnell R, Macmillan B, editors. Geocomputation, a primer. New York: Wiley; 1998. p. 77–94. Authors’ contributions 15. Haining R. Spatial data analysis in the social and environmental sciences. MS conceived and designed the study, performed the spatial and statistical Cambridge: University Press; 1990. data analysis and data interpretation and led the writing of this manuscript. 16. Bailey TC, Gatrell AC. Interactive spatial data analysis. Harlow: Longman; JEB contributed to data interpretation, and critically revised the draft version of the paper. Both authors read and approved the final version of the 17. Maroc. HCP. RGPH2014. http://www.hcp.ma/Demographie-population_r142. manuscript. html. Accessed 11 Apr 2017. (in French). 18. Dye C, Lonnroth K, Jaramillo E, Williams BG, Raviglione M. Trends in Ethics approval and consent to participate tuberculosis incidence and their determinants in 134 countries. Bull World Not applicable. Health Organ. 2009;87:683–91. 19. Obermeyer Z, Abbott-Klafter J, Murray CJ. Has the DOTS strategy improved Competing interests case finding or treatment success? An empirical assessment. PLoS One. The authors declare that they have no competing interests. 2008;3:e1721. 20. Kouyoumjian SP, Mumtaz GR, Hilmi N, Zidouh A, El Rhilani H, Alami K, et al. Author details The epidemiology of HIV infection in Morocco: systematic review and data Environmental Epidemiology Unit, National Institute of Hygiene. Ministry of synthesis. Int J STD AIDS. 2013;24:507–16. Health, Rabat, Morocco. University Hospital Center. Moulay Youssef Hospital, Rabat, Morocco. Faculty of Medicine and Pharmacy, University Mohammed V, Rabat, Morocco. Received: 27 October 2017 Accepted: 19 April 2018 References 1. WHO. Tuberculosis. Fact sheet N° 104. http://www.who.int/tb/publications/ factsheets/en/. Accessed 11 Apr 2017. 2. Maroc. Ministère de la santé. Lutte Antituberculeuse au Maroc : progrès, défis et perspectives. (in French). http://www.sante.gov.ma/Documents/ 2017/03/Lutte%20Antituberculeuse%20au%20Maroc_ http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Infectious Diseases of Poverty Springer Journals

Spatiotemporal distribution and predictors of tuberculosis incidence in Morocco

Free
13 pages
Loading next page...
 
/lp/springer_journal/spatiotemporal-distribution-and-predictors-of-tuberculosis-incidence-hNC4V0C9se
Publisher
BioMed Central
Copyright
Copyright © 2018 by The Author(s).
Subject
Medicine & Public Health; Infectious Diseases; Tropical Medicine; Public Health
eISSN
2049-9957
D.O.I.
10.1186/s40249-018-0429-0
Publisher site
See Article on Publisher Site

Abstract

Background: Tuberculosis (TB) is a major health problem in Morocco. This study aims at examining trends in TB in Morocco and identifying TB spatial clusters and TB-associated predictors. Method: Country-level surveillance data was exploited. Kendall’s correlation test was used to examine trends and an exploratory spatial data analysis was conducted to assess the global and local patterns of spatial autocorrelation in TB rates (Moran’s I and local indicator of spatial association [LISA]) at the prefecture/province level. Covariates including living in a prefecture versus living in a province, annual rainfall, annual mean temperature, population density, and AIDS incidence were controlled. An ordinary least squares regression was thus performed and both spatial dependence and heteroscedasticity were assessed. Results: A decrease in TB incidence rate was seen between 1995 and 2014 (Kendall’s tau b = − 0.72; P < 0.0001). However, while the period between 2005 and 2014 (10 last years) was considered, TB rate remained stable and as high as 84 per 100 000 population per year (95% CI: 83.7–84.3). The highest incidence rates were seen in Tanger- Assilah, Fez, Tetouen-M’diq Fnidaq, Inezgane Ait Melleoul, and Casablanca. From 2005 to 2014, while TB incidence rate was stable in Fez (P = 0.500), Tetouen-M’diq Fnidaq (P = 0.300), Casablanca (P = 0.500), Mohammadia (P = 0.146), Al Hoceima (P = 0.364), and Guelmim (P = 0.242), an increase in TB incidence rate was seen in Tanger-Assilah (Kendall’s tau = 0.49; P = 0.023) and a decrease in Salé (Kendall’s tau b = − 0,54; P = 0.014) and Inezgane-Ait Melloul (Kendall’s tau b = − 0,67; P = 0.0023). TB is strongly clustered in space (P-values of Moran’s I < 0.01). Two distinct spatial regimes that affect TB spatial clustering were identified (east and west). In the east, both annual rainfall (P = 0.003) and AIDS (P = 0.0002) exert a statistically significant effect on TB rate. In the west, only the living area (prefecture versus province) was associated with TB rate (P = 0.048). Conclusions: New information on TB incidence and TB-related predictors was provided to decision-making and to further pertinent research. Association between annual rainfall and TB may be of interest to be explored elsewhere. Keywords: TB, Meteorological data, Prefecture/province, AIDS, Population density, Morocco Multilingual abstract and middle income countries [1]. In Morocco, TB remains Please see Additional file 1 for translations of the abstract a major public health problem in spite of the efforts of the into the five official working languages of the United Ministry of health (MH) to alleviate it [2]. In 2015, 30 636 Nations. cases were reported; a total of 656 cases died from TB [3]. A national TB program was set at the end of the sev- enties to prevent, control, and eventually eliminate TB Background from Morocco. Standardized treatment regimens are Tuberculosis (TB) is one of the top 10 causes of death provided for free [4]. Two reference national laboratories worldwide [1]. In 2015, 10.4 million people around the provide testing for TB infection. In 2004, Morocco world fell ill from TB and a total of 1.8 million died from managed to reach the WHO objectives related to TB this disease. Over 95% of deaths from TB occur in low diagnosis and treatment [2]. Thus, in 2015, 83% of the * Correspondence: mina.sadeq@gmail.com cases were detected, 85% were treated for TB [2]. How- Environmental Epidemiology Unit, National Institute of Hygiene. Ministry of ever, TB incidence did not seem to decrease in Morocco. Health, Rabat, Morocco Full list of author information is available at the end of the article © The Author(s). 2018 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Sadeq and Bourkadi Infectious Diseases of Poverty (2018) 7:43 Page 2 of 13 The recent statistics showed that TB incidence in Morocco software version 1.6.7.9, March 2015, developed by Luc was as high as 89 per 100 000 population in 2015 [2]. Anselin (ASU, GeoDa Center for Geospatial Analysis and More may need to be explored about TB in Morocco. Computation, Arizona, USA), was used for these pur- Studies on spatial clusters of TB incidence that would poses. This was performed for each year under study. have given better understanding where interventions are most required are lacking in Morocco. On the other Data on potential predictors of TB hand, it is thought that TB prevails in prefectures rather Meteorological data by year than in provinces, and that the population density is a A centroid for each polygon that represents a province/ risk factor of TB in Morocco. Such claims require further prefecture was created and its GWS84 coordinates were research. Association between TB incidence and meteoro- determined. Those coordinates were used to get meteoro- logical factors has been cited elsewhere [5–7], but has not logical data (annual rainfall and annual mean temperature) been explored in Morocco yet. Cases of AIDS/ HIV are by province/prefecture. Climate monitoring data were more vulnerable to TB infection. Including AIDS/HIV obtained from the Global Climate Monitor [10]made incidence as a covariate in a regression model would best available under the Open Database License. This was predict TB incidence in Morocco. performed for each year under study. The QGIS soft- This work aimed at, first, examining trends in TB inci- ware version 2.0.1 ‘Dufour’ (FreeSoftwareFounda- dence rate in Morocco; second, examining spatial cluster- tion,Inc.,Boston,USA)was used. ing/clusters of TB incidence at the province/prefecture level; third, exploring non-spatial and spatial correlation Spatial regime between TB and some covariates in order to specify a Box maps of raw TB incidence rate were examined first model that would best predict TB in Morocco. Potential and foremost. High TB rates were seen in the west of predictors are living in a prefecture versus living in a Morocco, low TB rates in the remaining part of the province, population density, AIDS incidence, and me- country, and this may suggest the possible presence of teorological factors, i.e., annual rainfall and annual mean spatial heterogeneity in the form of spatial regimes. temperature. Diagnostics for spatial dependence and Thus, it was hypothesized that TB predictors may exert spatial heterogeneity were performed. Spatial study was a different effect across the west and east of Morocco. In limited to the last four years (i.e., 2011 to 2014) for a this study, these spatial regimes, i.e., west versus east, reason cited in “Results” section. were identified as shown in Fig. 2, and were evaluated as a dummy variable in the statistical spatial analyses. They Methods will be incorporated into the multivariate analyses that Geographical data, study area/population, and population adjust for spatial heterogeneity. density by year A polygon shapefile map of Morocco comprising 59 prov- HIV/AIDS by year inces/prefectures, developed for a previous study [8], was Data on HIV are not available and only those on AIDS used. The process of georeferencing, digitalizing, and incidence (by province/prefecture) of 2008 and 2009 are combining some provinces/prefectures is described else- [9], and this imposed a constraint as to the multivariate where [8]. Data on population size by province/prefecture regression. To deal with this, it was first opted for data were obtained from “Santé en Chiffres” files that were on AIDS of 2009 and it was checked whether the other made available by the Service of Studies in Health and potential predictors affect TB in the years between 2011 Health Information-Ministry of Health (SSHHI-MH) [9]. and 2014 and in 2009, similarly. If it is the case, AIDS The total population size was 32 187 000 inhabitants in rate can then be incorporated as an additional covariate 2011, it was 33 848 000 inhabitants in 2014. For each year in the multivariate regression to draw conclusions about under study, the population density by province/prefec- the effect of this variable on TB incidence. ture was calculated; thus, the population in a province/ prefecture was divided by the size of that province/ prefecture. Statistical analysis Trends in TB TB data by year An approximate two-sided Kendall’s rank correlation Data on both new cases and incidence of TB by province/ test was conducted to examine variation in TB incidence prefecture were obtained from “Santé en Chiffres” files from 1995 to 2014 (20 years) and from 2005 to 2014 made available by SSHHI-MH [9]. The raw incidence (10 years); the P-values and size effects of which are rates of TB by province/prefecture were calculated; out- provided. An annual Poisson incidence rate estimate of liers were looked for. Thus box map was displayed to TB and a Poisson rate confidence interval were also check for variance instability of the raw rates. GeoDa provided. The incidence rate is estimated as the number Sadeq and Bourkadi Infectious Diseases of Poverty (2018) 7:43 Page 3 of 13 of events observed divided by the time at risk of event used to perform Kendall’s rank correlation tests. Then, during the observation period. GeoDa was again used to perform bivariate Moran’s I test A Kendall’s rank test was performed to evaluate vari- to examine bivariate LISA between TB and covariates. In ation in the incidence rate from 2011 to 2014 in selected addition, an ordinary least squares (OLS) regression ana- prefectures and provinces. Statistics were calculated in lysis that took into account the previously identified spatial exact form. regimes was conducted. GeoDa was used for this purpose. These statistical methods were conducted using the Multicollinearity condition number, normality (Jarque-Bera StatsDirect statistical software version 3.0.194 (StatsDir- test), spatial dependence for weight matrix (row-standard- ect Ltd., Cheshire, UK). ized weights and Lagrange multiplier tests), and spatial heteroskedasticity (Breusch-Pagan test and Koenker-Bassett Global spatial clustering and LISA clusters of TB test) were all assessed. Then, the stability of predictors ef- The exploratory spatial data analysis approach [11–16] fect across regimes was evaluated. GeoDaSpace (ASU, was used to examine global and local patterns of spatial GeoDa Center for Geospatial Analysis and Computation, autocorrelation in TB rates and in covariates. A contigu- Arizona, USA) was used to perform a Chow test. ity raw standardized weight file was created. Queen contiguity, which defines spatial neighbours as those provinces/prefectures with shared borders and vertices, Results was chosen. Thus, the global univariate Moran’s I statis- Trend in TB incidence in Morocco from 1995 to 2014 and tic was examined. A positive and significant Moran’s I from 2005 to 2014 indicates clustering in space of similar TB rates. The A decrease in TB incidence rate was seen from 1995 local indicators of spatial association (LISA), which show to 2014 in Morocco (Fig. 1)(Kendall’stau b= − 0,72 the presence or absence of significant spatial clusters or P < 0.0001). The mean estimate of new TB cases (all outliers, was also examined. GeoDa software was used to forms confounded) was 27 642. A maximum number of perform these spatial analyses. 31 771 cases was reported in 1996, a minimum number of 25 473 cases in 2008. While the period between 2005 and Specifying a regression model of TB 2014 was considered, TB incidence rate was stable (Ken- Global clustering of potential predictors was examined first dall’stau= − 0.16; P = 0.242). During this 10-year period, and foremost. GeoDa was used to perform Moran’s I test. TB incidence rate mean was 85.3 per 100 000 population, Then, potential multicollinearity and linear correlation SD = 3.5. The Poisson annual incidence rate estimate was between TB and predictors were analysed. StastDirect was 84 per 100 000 population per year at 95% CI:83.7–84.3. Fig. 1 Incidence rate (Log10 scale) of tuberculosis, Morocco, 1995–2014. This figure depicts secular trends of tuberculosis in Morocco from 1995 to 2014. A decrease was shown during this 20-year period (Kendall’s tau = − 0.72; P = 0.0001). TB incidence rate remained relatively stable at 85.3 per 100 000 population from 2005 to 2014 Sadeq and Bourkadi Infectious Diseases of Poverty (2018) 7:43 Page 4 of 13 Descriptive analysis and trends in TB incidence in for each of the years under study (Fig. 3), we considered prefectures and provinces that rate smoothing is not necessary. Global univariate The highest incidence rate was shown in the prefecture Moran’s I statistics, univariate LISA cluster maps, and of Tanger-Assilah, identified as the unique outlier in univariate LISA significance maps of TB rates by year Figs. 2 and 3. Figure 4 shows TB incidence rates in all were all illustrated in Fig. 5. For each of the years under the prefectures of Morocco. High TB rates were seen in study, spatial clustering of high TB incidence rates was Tanger-Assilah, Fez, Tetouen-M’diq Fnidaq, Inezgane- shown in the north-west part of Morocco (Global Moran’s Ait Melloul, Casablanca, Mohammadia, and Salé. The I statistics were all statistically significant at 0.01 level or respective means (expressed per 100 000 population) less). Larache and Kenitra formed high spatial clusters and standard deviation for the period between 2005 and during the four years, Skhirate-Témara and Benslimane 2014 were 188 ± 12, 156 ± 12, 133 ± 13, 153 ± 26, 134 ± from 2011 to 2013, Salé from 2012 to 2014, Mohammadia 13, 122 ± 10, and 125 ± 8. As to provinces, high TB inci- in 2012, Tétouen- M’diq Fnidaq in 2013, Meknes and dence rates were seen in Al Hoceima (119 ± 7), Guelmim Khemissat in 2014, and Tanger-Assilah from 2013 to 2014 (104 ± 19), Khemissat, and Larache. From 2005 to 2014, (Fig. 5). Spatial clustering of low TB incidence rates was while TB incidence rate was stable in Fez (P = 0.5), located throughout the eastern and the southern part of Tetouen-M’diq Fnidaq (P = 0.300), Casablanca (P =0.5), the country as indicated in blue in Fig. 5. The significance Mohammadia (P = 0.146), Al Hoceima (P = 0.364), and level was tightened even more (to P = 0.01 instead of 0.05) Guelmim (P = 0.242), an increase in TB incidence rate to detect spatial clusters of low TB rates. For each of the was seen in Tanger-Assilah (Kendall’stau=0.49; P =0.023) years under consideration, the provinces of Errachidia and and a decrease in Salé (Kendall’stau b= − 0,54; P = 0.014) Ouarzazate were consistently significant even at that more and Inezgane-Ait Melloul (Kendall’stau b= − 0,67; demanding level. Two spatial outliers were identified, P = 0.0023). namely, Guelmim province and the prefecture of Fahs Anjra (Fig. 5). Global spatial clustering and LISA of TB incidence between 2011 and 2014 Univariate spatial autocorrelation, linear correlation, As TB incidence rate was approximatively stable from bivariate spatial autocorrelation, and regression model 2005 to 2014, spatial study was restricted to the last four There was evidence of a significant spatial pattern of years (i.e. 2011 to 2014). Data on recent TB cases reported annual mean temperature as well as of annual rainfall in in 2015, 2016 and 2017 are still not publicly available. each of the years under study (Fig. 6); spatial clustering Since box maps of TB raw rate showed only one outlier of AIDS incidence rates was also evident in 2009 (Global Fig. 2 TB raw rate box maps (using 1.5 as hinge). 2011–2014. Morocco. Examining raw TB rate box maps revealed that more than 25% of data were located in the west part of the country where high TB levels were shown (in yellow) while low TB levels were seen in the west part of the country (in green), and this identified two distinctive spatial regimes Sadeq and Bourkadi Infectious Diseases of Poverty (2018) 7:43 Page 5 of 13 Fig. 3 Descriptive analysis of TB incidence between 2011 and 2014. Morocco. The highest incidence rate, a unique outlier, was shown in the prefecture of Tanger-Assilah. High TB rates were also shown in the prefectures of Fez, Tetouen-M’diq Fnidaq, and Casablanca Moran’s I = 0.345; P = 0.001)(Table 1 and Fig. 7). The tau b tests were not statistically significant) (Table 1). prefecture of Casablanca formed the unique cluster of Annual rainfall showed spatial correlation stronger than population density in the country. linear correlation (Bivariate I >Kendall’s tau b), suggesting Annual rainfall, population density, and AIDS rates that this correlation is determined by geographic location. were strongly and positively correlated with TB, the The population density showed spatial correlation weaker mean annual temperature was not (all related Kendall’s than linear correlation, which indicates that this correlation Fig. 4 TB incidence rate by year (2011–2014), prefectures, Morocco. Legend: This figure depicts variation of TB incidence rate in the prefectures of Morocco. During the period from 2011 to 2014, the Highest TB incidence rates were seen in Tanger-Assilah, Fez, Tetouen-M’diq Fnidaq, Inezgane- Ait Melloul and Casablanca, all showed in the map. We looked more closely at TB variation in these prefectures during the period from 2005 to 2014. An increase in TB incidence rate was seen in Tanger-Assilah (Kendall’stau = 0.49; P = 0.023) while a decrease was seen in Inezgane-Ait Melloul (Kendall’stau b = − 0,67; P = 0.0023). TB incidence rate was stable in Fez (P = 0.5), Tetouen-M’diq Fnidaq (P = 0.300), and Casablanca (P = 0.5) Sadeq and Bourkadi Infectious Diseases of Poverty (2018) 7:43 Page 6 of 13 Fig. 5 LISA cluster maps and LISA significance maps of TB incidence, Morocco, 2011–2014. LISA indicates the presence or absence of significant spatial clusters or outliers for each province/prefecture. The province of Larache and the province of Kenitra formed spatial clusters of high TB incidence from 2011 to 2014, the prefecture of Skhirate-Témara and the province of Benslimane from 2011 to 2013, the prefecture of Salé from 2012 to 2014, and the prefecture of Tanger-Assilah from 2013 to 2014. Other areas formed temporary spatial clusters, including the prefectures of Mohammadia (2012), Tétouen- M’diq Fnidaq (2013), Meknes (2014), and the province of Khemissat (2014). All these cited provinces/prefectures were shown in red. Significant spatial clusters of low TB incidence were located in the east spatial regime. The provinces of Errachidia and Ouarzazatewere consistently significant even at more demanding level (i.e., P = 0.01 instead of 0.05). Two spatial outliers (in pink) were identified, namely the prefecture of Fahs Anjra (in the north) and the province of Guelmim (in the south) is partly determined by location. Table 1 also showed a sta- also seen between the population density and area (pre- tistically significant dispersion of TB rates correlated with fecture versus province) in 2009, 2011, 2012, 2013, and AIDS rates (Global Moran’s I = − 0.139; P = 0.019). 2014. Kendall’s tau b were 0.59, 0.61, 0.60, 0.61, 0.61, Multicollinearity was seen between annual rainfall and respectively; all of which were statistically significant at mean annual temperature in 2009, 2011, 2012, 2013, and less than 0.0001 level. 2014. Kendall’s tau b were − 0.41, − 0.60, − 0.59, − 0.44, Taking into account both correlation and multicolli- and − 0.28, respectively; all of which were statistically nearity, we chose a regression model that includes only significant at less than 0.001 level. Multicollinearity was the annual rainfall and the spatial regimes as predictors Sadeq and Bourkadi Infectious Diseases of Poverty (2018) 7:43 Page 7 of 13 Fig. 6 LISA cluster maps of annual rainfall and annual mean temperature, Morocco, 2009 and 2011–2014. Annual mean temperature as well as annual rainfall showed spatial clustering in all years under study. The south of Morocco has low rainfall and high temperatures. High annual rainfall clusters were shown in red (maps in left) of TB incidence rates for each of the five years under (Table 3), but annual rainfall has relatively stable effect study; thus, we performed OLS regression (Table 2). An- across the east and west (P > 0.05) (Table 3). This sug- nual rainfall was consistently related to TB only in 2013 gests the influence of underlying variables that exert a (P = 0.008), the spatial regime in all the years under different effect across the west and the east regimes. study (P ≤ 0.03 were statistically significant), indicating a Diagnostics of multicollinearity condition numbers did potential different effect of predictors across the west not suggest problems with the stability of the regression and the east regimes (Table 2). This led us to perform a results, that may be due to multicollinearity (all numbers spatial Chow test (Table 3). It was statistically significant related to the study years were less than 8) (Table 4). (P < 0.03) for each of the years under consideration The models assume normality of the errors, as indicated Sadeq and Bourkadi Infectious Diseases of Poverty (2018) 7:43 Page 8 of 13 Table 1 Non-spatial and spatial correlation between TB and its were 0.014 and 0.003, respectively) (Table 8); spatial de- potential predictors pendence was not present. Variable Univariate I (P)b(P) Bivariate I (P) Both annual rainfall and area affect TB in the years between 2011 and 2014 and in 2009, similarly (Tables 5 Outcome (year) and 6). This led us to incorporate AIDS rates as an add- TB (2009) 0.169 (0.035) itional covariate in the regression model (Table 9). In the TB (2011) 0.229 (0.009) east, annual rainfall as well as AIDS exerts a statistically TB (2012) 0.230 (0.007) significant effect on TB (respective P-values were 0.003 TB (2013) 0.389 (0.001) and 0.0002). In the west, only the living area was statisti- TB (2014) 0.387 (0.001) cally significant (P = 0.048), suggesting that individuals living in prefectures are more exposed to TB transmis- Covariates (year) sion than those living in provinces (Table 9). Rainfall (2009) 0.895 (0.001) 0.31 (< 0.001) 0.419 (0.001) Rainfall (2011) 0.894 (0.001) 0.196 (0.03) 0.350 (0.001) Rainfall (2012) 0.909 (0.001) 0.203 (0.03) 0.354 (0.001) Discussion Rainfall (2013) 0.854 (0.001) 0.380 (< 0.0001) 0.498 (0.001) To our knowledge, this is the first study to explore trends in TB incidence rate in Morocco, its spatial pat- Rainfall (2014) 0.832 (0.001) 0.373 (< 0.0001) 0.491 (0.001) terns and predictors, and the spatial patterns of its pre- Temperature (2009) 0.703 (0.001) −0.006 (0.95) −0.092 (0.09) dictors. TB incidence rate was stable at the country Temperature (2011) 0.724 (0.001) −0.097 (0.29) −0.207 (0.003) level. A close look at TB incidence in the prefectures Temperature (2012) 0.721 (0.001) −0.087 (0.33) −0.190 (0.001) and provinces that had the highest rates was informative. Temperature (2013) 0.622 (0.001) −0.027 (0.76) −0.091 (0.078) Included in such group are those that form spatial clus- Temperature (2014) 0.576 (0.001) 0.015 (0.87) −0.065 (0.190) ters, including Tanger-Assilah, Tetouen-M’diq, Salé and Guelmim, and those that do not form spatial clusters, Pop_Density (2009) 0.258 (0.007) 0.44 (< 0.0001) 0.225 (0.002) including Fez, Casablanca, Inezgane-Ait Melloul, Moha- Pop_Density (2011) 0.250 (0.006) 0,53 (< 0.0001) 0.246 (0.005) madia, and Al Hoceima. A significant increase in TB Pop_Density (2012) 0.252 (0.004) 0,49 (< 0.0001) 0.250 (0.004) incidence rate was seen in Tanger-Assilah. It is believed Pop_Density (2013) 0.252 (0.006) 0,56 (< 0.0001) 0.257 (0.004) that poverty and housing conditions in some communes Pop_Density (2014) 0.259 (0.004) 0,50 (< 0.0001) 0.182 (0.011) within this prefecture are the main risk factors of TB, AIDS (2009) 0.345 (0.001) 0.26 (0.006) −0.139 (0.019) and this may be subjective due to the scarcity of existing I = Moran’s Index statistic; P = P-Value; b = Kendall’s tau research that explores such issues in this prefecture or in others in Morocco. TB incidence rates in the prefec- ture of Fahs Anjra, a low-high outlier located between by the Jarque-Bera test (all P-values were > 0.05) (Table 4). Tanger-Assilah and Tetouen M’ diq, were influenced by Apart from that of the year 2009, all the P-values corre- TB rates in these neighbouring prefectures. TB incidence sponding to Heteroskedasticity-related tests were small, rates in Tetouen-M’diq, Mohamadia, El Hoceima, indicating heteroscedasticity was present (P < 0.05) Guelmim, and Casablanca remained stable, and this sug- (Table 4). Diagnostics for spatial dependence for weights gests that more efforts and research may be required. A matrix (row-standardized weights) did not indicated the high incidence cluster surrounded by low incidence clus- presence of spatial dependence. Lagrange Multiplier tests ters was observed in Guelmim Province; it was not seen in (lag and error tests) were not statistically significant 2009 (Fig. 7). Casablanca has a population density ≥ 150 (P > 0.05) (Table 4). inhabitants per km [17], so have other prefectures [17]; Given the strong heteroscedasticity and the results of however, only Casablanca was found to form a spatial the Chow test, we examined regression of TB incidence cluster of population density in Morocco, and this was not rates on covariates in the west and the east regimes, separ- expected. Besides, Casablanca did not form a spatial ately. Since multicollinearity was present between annual cluster of TB as it is believed. Attention has to be paid to rainfall and mean annual temperature, and between area all these previously cited areas that may require further and population density for each of the studied years in research and new efficient strategic measures. both spatial regimes, we performed OLS regression of TB rates on annual rainfall and area (Tables 5 and 6). In the In Morocco, TB spatial distribution is not random. An east, there was no evidence of spatial dependence (P-value enduring spatial clustering of high TB incidence rates related to lag tests were > 0.05) or heteroscedasticity (P- was seen in the north-western part of the country, that value related to error tests were > 0.05) (Table 7). In the of low TB incidence rates throughout the south and the west, heteroscedasticity was present only in 2012 (P-values east parts of the country. This supports our hypothesis related to Koenker-Bassett test and Breusch-Pagan test suggesting a different effect of potential predictors on Sadeq and Bourkadi Infectious Diseases of Poverty (2018) 7:43 Page 9 of 13 Fig. 7 LISA cluster maps and LISA significance maps of AIDS incidence (upper maps) and TB incidence (lower maps), 2009, Morocco. Some provinces, including Taroudant, Guemim, and Layoune and some prefectures, including Inezgane Ait Melloul and Agadir Ida Outanane form spatial clusters of high AIDS incidence rates. The prefecture of Meknes formed a high-low outlier. Other provinces, including Essaouira, Tata, and Essmara and the prefecture of Chtouka Ait Baha constitute a low-high clusters. A total of 10 spatial clusters of low AIDS incidence rates were identified. However, while examining AIDS clusters at more demanding significant level (i.e., 0.01 instead of 0.05), only low incidence clusters located in the east, namely the provinces of Khenifra, Errachidia, and Figuig were identified; no high incidence cluster was seen Table 2 Ordinary Least Squares Regresssion of TB rates in 2009 TB across the two key distinctive spatial regimes, i.e., and between 2010 and 2014. Morocco the west and the east, that were identified in this study. Covariate (year) Coefficient t-Statistic Probability Spatial patterns of further predictors were explored. Both annual rainfall and mean annual temperature Intercept (2009) 24.179 1.944 0.057 showed spatial clustering and were negatively correlated, Intercept (2011) 23.874 2.123 0.038 Intercept (2012) 22.568 1.957 0.055 Table 3 Spatial Regimes (West & East) Diagnostics. Chow test. Intercept (2013) 23.395 2.922 0.005 2009 and between 2010 and 2014, Morocco Intercept (2014) 23.863 2.186 0.033 Variable DF Value Probability Annual Rainfall (2009) 0.0587 1.710 0.093 Intercept (2009) 1 0.253 0.615 Annual Rainfall (2011) 0.037 1.465 0.149 Intercept (2011) 1 5.776 0.016 Annual Rainfall (2012) 0.051 1.691 0.096 Intercept (2012) 1 1.872 0.171 Annual Rainfall (2013) 0.058 2.743 0.008 Intercept (2013) 1 3.030 0.082 Annual Rainfall (2014) 0.060 1.808 0.076 Intercept (2014) 1 0.455 0.500 Spatial regime (2009) 29.713 2.229 0.030 Annual Rainfall (2009) 1 0.435 0.510 Spatial regime (2011) 45.149 4.240 < 0.0001 Annual Rainfall (2011) 1 0.185 0.667 Spatial regime (2012) 40.340 3.817 < 0.001 Annual Rainfall (2012) 1 0.096 0.757 Spatial regime (2013) 36.664 3.700 < 0.001 Annual Rainfall (2013) 1 0.471 0.493 Spatial regime (2014) 30.152 2.474 0.016 Annual Rainfall (2014) 1 3.613 0.057 Value F-statistic P (F-statistic) Global Chow test (2009) 2 6.767 0.034 Adjusted R Squared (2009) 21.25% 8.691 < 0.001 Global Chow test (2011) 2 22.603 < 0.0001 Adjusted R Squared (2011) 28.72% 12.686 < 0.0001 Global Chow test (2012) 2 18.348 0.0001 Adjusted R Squared (2012) 26.67% 11.548 < 0.0001 Adjusted R Squared (2013) 39.53% 19.962 < 0.000001 Global Chow test (2013) 2 18.009 0.0001 Adjusted R Squared (2014) 32.77% 15.138 < 0.00001 Global Chow test (2014) 2 14.359 < 0.001 Sadeq and Bourkadi Infectious Diseases of Poverty (2018) 7:43 Page 10 of 13 Table 4 Regression diagnosis by year Table 5 Ordinary Least Squares Regression of TB rates in 2009 and between 2010 and 2014. East regime. Morocco Multicollinearity condition number: Covariate (year) Coefficient t-Statistic Probability Year Number Intercept (2009) 30.949 2.384 0.027 2009 5.54 Intercept (2011) 21.399 2.753 0.012 2011 4.68 Intercept (2012) 23.721 2.677 0.014 2012 4.96 Intercept (2013) 26.060 3.918 0.001 2013 3.97 Intercept (2014) 36.031 3.685 0.002 2014 7.08 Annual Rainfall (2009) 0.026 0.605 0.552 Normality of errors (Jarque-Bera test) Annual Rainfall (2011) 0.041 1.990 0.060 Year DF Value P-Value Annual Rainfall (2012) 0.041 1.632 0.118 2009 2 0.522 0.770 Annual Rainfall (2013) 0.037 1.592 0.126 2011 2 0.404 0.817 Annual Rainfall (2014) 0.008 0.250 0.805 2012 2 0.471 0.790 Prefecture vs province (2009) 39.323 1.166 0.257 2013 2 2.858 0.240 Prefecture vs province (2011) 30.375 1.319 0.202 2014 2 1.289 0.525 Prefecture vs province (2012) 35.128 1.369 0.186 Heteroskedasticity Koenker-Bassett test Breusch-Pagan test coefficients: Prefecture vs province (2013) 39.445 1.703 0.103 Year DF Value P-Value) Value P-Value Prefecture vs province (2014) 30.400 1.202 0.243 2009 2 4.4506 0.108 5.338 0.069 Adjusted R Squared < 0.2 for all studied years; p (F-statistic) > 0.05 for all years except for the year 2011 for which P = 0.044) 2011 2 10.823 0.004 12.867 0.002 2012 2 9.534 0.009 11.400 0.003 2013 2 9.007 0.011 10.799 0.005 2014 2 6.340 0.042 4.311 0.116 Table 6 Ordinary Least Squares Regresssion of TB rates in 2009 and between 2010 and 2014. West regime. Morocco Spatial dependance for weight matrix (row-standardized weights) Covariate (year) Coefficient t-Statistic Probability MI/DF Value P-Value Intercept (2009) 40.299 1.703 0.099 Moran’s I (error) test Intercept (2011) 61.578 3.232 0.003 2009 −0.079 −0.408 0.683 Intercept (2012) 46.563 2.170 0.038 2011 −0.046 −0.002 0.998 Intercept (2013) 45.894 3.304 0.002 2012 −0.050 −0.049 0.961 Intercept (2014) 19.455 0.700 0.489 2013 0.015 0.702 0.483 Annual Rainfall (2009) 0.052 1.080 0.289 2014 0.174 2.497 0.013 Annual Rainfall (2011) 0.015 0.368 0.716 Lagrange Multiplier (lag) test Annual Rainfall (2012) 0.053 1.024 0.313 2009 1 0.363 0.547 Annual Rainfall (2013) 0.058 2.047 0.049 2011 1 0.062 0.804 Annual Rainfall (2014) 0.104 1.899 0.067 2012 1 0.065 0.799 Prefecture vs province (2009) 41.335 2.673 0.012 2013 1 0.442 0.506 Prefecture vs province (2011) 42.693 2.908 0.007 2014 1 3.460 0.063 Prefecture vs province (2012) 38.846 2.724 0.010 Lagrange Multiplier (error) test Prefecture vs province (2013) 34.896 2.917 0.006 2009 1 0.727 0.394 Prefecture vs province (2014) 31.151 2.630 0.013 2011 1 0.246 0.620 2012 1 0.292 0.589 Value F-statistic p (F-statistic) 2013 1 0.028 0.867 Adjusted R Squared (2009) 19.17% 4.914 0.014 2014 1 3.564 0.059 Adjusted R Squared (2011) 16.85% 4.446 0.020 Unit of analysis = 59; MI Moran’sI, DF Degree of Freedom Adjusted R Squared (2012) 16.84% 4.442 0.020 Adjusted R Squared (2013) 26.67% 7.182 0.003 Adjusted R Squared (2014) 25.12% 6.704 0.004 Sadeq and Bourkadi Infectious Diseases of Poverty (2018) 7:43 Page 11 of 13 Table 7 Regression diagnosis by year. East spatial regime Table 8 Regression diagnosis by year. West spatial regime Multicollinearity condition number: Multicollinearity condition number: Year Number Year Number 2009 3.64 2009 7.00 2011 3.31 2011 5.66 2012 3.20 2012 6.62 2013 2.56 2013 5.01 2014 3.66 2014 11.08 Normality of errors (Jarque-Bera test) Normality of errors (Jarque-Bera test) Year DF Value P-Value Year DF Value P-Value 2009 2 7.837 0.020 2009 2 1.833 0.399 2011 2 1.250 0.535 2011 2 0.721 0.697 2012 2 0.894 0.640 2012 2 2.038 0.361 2013 2 0.441 0.802 2013 2 0.276 0.802 2014 2 1.219 0.544 2014 2 0.670 0.715 Heteroskedasticity Koenker-Bassett test Breusch-Pagan test Heteroskedasticity Koenker-Bassett test Breusch-Pagan test coefficients: coefficients: Year DF Value P-Value Value P-Value Year DF Value P-Value Value P-Value 2009 2 1.952 0.377 4.112 0.128 2009 2 5.123 0.036 5.123 0.077 2011 2 1.738 0.419 1.680 0.432 2011 2 5.133 0.077 5.996 0.050 2012 2 1.260 0.533 1.207 0.547 2012 2 8.586 0.014 11.499 0.003 2013 2 0.672 0.714 0.586 0.746 2013 2 4.741 0.093 5.583 0.061 2014 2 2.591 0.274 1.389 0.499 2014 2 2.710 0.258 1.794 0.408 Spatial dependance for weight matrix (row-standardized weights) Spatial dependance for weight matrix (row-standardized weights) MI/DF Value P-Value MI/DF Value P-Value Moran’s I (error) test Moran’s I (error) test 2009 −0.042 0.320 0.749 2009 −0.188 −1.024 0.305 2011 0.06 1.020 0.308 2011 −0.121 −0.509 0.610 2012 0.145 1.414 0.157 2012 −0.143 −0.705 0.481 2013 0.033 0.742 0.458 2013 −0.073 −0.082 0.934 2014 0.358 2.942 0.003 2014 −0.0767 −0.166 0.868 Lagrange Multiplier (lag) test Lagrange Multiplier (lag) test 2009 1 0.148 0.700 2009 1 2.633 0.105 2011 1 0.274 0.600 2011 1 1.244 0.265 2012 1 0.234 0.629 2012 1 1.493 0.222 2013 1 0.093 0.760 2013 1 0.191 0.662 2014 1 0.844 0.358 2014 1 0.076 0.783 Lagrange Multiplier (error) test Lagrange Multiplier (error) test 2009 1 0.061 0.805 2009 1 0.419 0.518 2011 1 0.130 0.718 2011 1 0.912 0.340 2012 1 0.484 0.487 2012 1 1.266 0.261 2013 1 0.039 0.843 2013 1 0.333 0.564 2014 1 0.473 0.491 2014 1 0.366 0.545 Unit of analysis = 59; MI Moran’sI, DF Degree of Freedom Unit of analysis = 59; MI Moran’sI, DF Degree of Freedom Sadeq and Bourkadi Infectious Diseases of Poverty (2018) 7:43 Page 12 of 13 Table 9 Ordinary Least Squares Regresssion of TB rates in 2009. [18, 19]. In Morocco, inequalities in socio-economic Morocco conditions, including housing conditions, vary by prov- Covariate (year) Coefficient t-Statistic Probability ince/prefecture as well as inside prefectures. Increasing rainfall may affect over-crowded dwellings and cause Spatial East Regime dwellers to be at greater risk of developing TB. This might Intercept −13.562 −1.009 0.326 explain annual rainfall association with TB in the east re- Annual Rainfall 0.130 3.377 0.003 gime where the provinces of Morocco prevail, and may also Prefecture versus province 34.599 1.444 0.165 explain the connection between high spatial clusters of TB AIDS 20.538 4.549 0.0002 and rainfall in some prefectures in the west regime. Further Spatial West Regime research that uses new scientific approaches such as map- ping diseases and risk analysis at a small-area level are likely Intercept 20.643 0.591 0.559 to be required in the suspected areas. Annual Rainfall 0.0843 1.314 0.199 More focus on AIDS/HIV related to TB may be needed Prefecture versus province 35.565 2.059 0.048 in Morocco. In a previous study, AIDS was reported to be AIDS 4.890 0.770 0.447 prevalent in Agadir, Marrakesh and Casablanca [20]. A particular stakeholders’ focus may have been made on Value F-statistic P (F-statistic) these prefectures. Our study revealed the presence of five Adjusted R Squared (East) 0.495 8.19 3 0.001 high spatial clusters of AIDS incidence rates, all located in Adjusted R Squared (West) 0.181 3.431 0.029 the west regime. Examining spatial patterns of AIDS inci- East: Koenker-Bassett Statistic: 7.63 (P = 0.054); Jarque-Bera statistics: 0.08 dence at a more demanding significant level (i.e., 0.01 in- (P = 0.96); Multicollinearity condition number: 5.5; Spatial autocorrelation of stead of 0.05) yielded only low incidence clusters, formed residuals: Moran’s I: 0.24 (P = 0.81); Langrange multiplier (lag): 0.09 (P = 0.77); by Khenifra, Errachidia, and Figuig, all located in the east Langrange multiplier (error): 0.11 (P = 0.74) West: Koenker-Bassett Statistic: 7.80 (P = 0.19); Jarque-Bera statistics: 1.99 regime. This supports our previous finding with regard to (P = 0.37); Multicollinearity condition number: 11.0; Spatial autocorrelation of the association between TB and AIDS seen only in the residuals: Moran’s I: -0.96 (P = 0.34); Langrange multiplier (lag): 2.52 (P = 0.11); Langrange multiplier (error): 0.84 (P = 0.18) east spatial regime, and may require particular decision making’s attention and further small-area studies that con- and this was expected. The south of Morocco generally sider not only AIDS/HIV incidence but also socio- has low rainfall and high temperatures, the north of the economic and meteorological factors as well. country high rainfall and low temperatures. Previous This study has several strengths. It is based on reliable studies carried out elsewhere [5–7] pointed to a connec- data, obtained from “Santé en Chiffres” files, that are tion between temperature and TB and to a negative representative of the population of Morocco; the Climate correlation between rainfall and TB, and this is not con- monitoring data, obtained from the Global Climate sistent with the findings of this study that suggested that Monitor, was also used. Both data sources have been annual mean temperature was not correlated with TB, proven in this study to be reliable for epidemiological re- and that annual rainfall was positively correlated with search. Exploring spatial clustering of TB, including TB. Differences in meteorological factors between coun- spatial effect, and identifying two distinctive spatial re- tries might be one potential explanation. In our study, gimes at the country level are other important strengths annual mean temperatures varied approximatively be- of this study. This study has a few limitations. Like any tween 10 °C and 25.6 °C [data not shown], annual mean country-level surveillance data, some TB cases may not rainfall between 10.1 and 723 mm [data not shown]. In a be reported, but they may be minor as all the physicians study undertaken in China, the monthly average in the country are conscious of the importance of temperature varied between − 13.4 °C and 20.4 °C, the reporting TB cases, and TB diagnosis and treatment are monthly rainfall between 0 and 195.1 mm [5]. This free and available at the primary healthcare centres, raises the question of whether the bacterium Mycobac- which are the first level health facility. On the other terium tuberculosis can grow and spread especially well hand, aggregate rather than individual level data were in particular conditions of temperature, humidity and considered, which may be another common limitation of probably other meteorological factors. Another question this ecological study. In spite of this, this study did yield of whether there are underlying factors that may serve new informative TB-related findings that may contribute as an intermediary factors between meteorological con- to guide decision-making in Morocco and to urge to fur- ditions and TB trends is also raised. Previous ecological ther pertinent research studies on TB. studies have suggested that “broad socio-economic de- velopment, rather than the success of TB control pro- Conclusions grammes, is the main determinant behind the declining In Morocco, TB is not randomly distributed in space. trends of TB observed in many regions of the world” Two distinctive spatial regimes that affect TB spatial Sadeq and Bourkadi Infectious Diseases of Poverty (2018) 7:43 Page 13 of 13 clustering were identified. Planning of TB control and %20progr%C3%A8s%20d%C3%A9fis%20et%20perspectives.pdf. 2017. Accessed 11 Jul 2017. prevention measures/strategies that focuses on both sus- 3. Maroc. Ministère de la Santé . Direction de l'Epidémiologie et de Lutte pected prefectures and provinces may be required and Contre les Maladies. Situation Epidémiologique de la Tuberculose au Maroc would best address this health problem. A critical need – Année 2015. http://www.sante.gov.ma/Documents/2016/03/Situation_ %C3%A9pidimio_de_la_TB_au_Maroc__ is to conduct research that considers all the herein stud- 2015%20Fr%20V%2020%20%20mars.pdf (2016). Accessed 11 Apr 2017. (in ied risk factors and uses new scientific approaches such French). as geographic information systems (GIS) and risk ana- 4. Maroc. Ministère de la santé. Plan National d'accélération de la réduction de l'incidence de la tuberculose 2013–2016. (in French). http://www.sante.gov. lysis at a small-area level. More publicly available aggre- ma/Documents/Actualites/FICHE%20ACC%C3%89L%C3%89RANT%20VF.pdf. gated data on HIV or other health outcomes will Accessed 11 Apr 2017. definitely be a useful tool with regard to epidemiological 5. Rao HX, Zhang X, Zhao L, Yu J, Ren W, Zhang XL, et al. Spatial transmission and meteorological determinants of tuberculosis incidence in Qinghai research in Morocco. Questions are arisen about under- Province, China: a spatial clustering panel analysis. Infect Dis Poverty. lying risk factors linked to rainfall that may influence TB 2016;5:45. incidence and about association between annual rainfall 6. Khaliq A, Batool SA, Chaudhry MN. Seasonality and trend analysis of tuberculosis in Lahore, Pakistan from 2006 to 2013. J Epidemiol Glob Health. and TB, and this may be of interest to be explored 2015;5:397–403. elsewhere. 7. Reza Beiranvand R, Karimi A, Delpisheh A, Sayehmiri K, Soleimani S, Ghalavandi S. Correlation assessment of climate and geographic distribution of tuberculosis using geographical information system (GIS). Iran J Public Additional file Health. 2016;45(1):86–93. 8. Sadeq M. Spatial patterns and secular trends in human leishmaniasis Additional file 1: Multilingual abstract in the five official working incidence in Morocco between 2003 and 2013. Infect Dis Poverty. 2016;5:48. languages of the United Nations. (PDF 480 kb) 9. Maroc. Ministère de la Santé/Direction de la Planification et des Ressources Financières/Direction de la Planification et des Etudes/Service des Etudes et de la Planification Sanitaire. (Maroc.MS/DPRF/DPE/SEIS). Santé en chiffres. Abbreviations From 2004 ed to 2014 ed. http://www.sante.gov.ma/publications/Etudes_ AIDS: Acquired Immune Deficiency Syndrome; LISA: Local Indicators of enquete/Pages/default.aspx. Accessed 07 Jan 2017. (in French). Spatial Association; MH: Ministry of Health; SD: Standard Deviation; 10. Climatic Research Unit, University of East Anglia. Global Climate Monitor SSHHI: Service of Studies in Health and Health Information-Ministry of Health; http://www.globalclimatemonitor.org. Accessed 07 Jan 2017. TB: Tuberculosis; WHO: World Health Organization 11. Anselin L. Exploratory spatial data analysis and geographic information systems. In: Painho M, editor. New tools for spatial analysis. Luxembourg: Acknowledgements Eurostat; 1994. p. 45–54. We would like to thank the Service of Studies in Health and Health 12. Anselin L. Local indicators of spatial association-LISA. Geograph. Anal. 1995; Information-Ministry of Health. Rabat. Morocco. 27:93–115. 13. Anselin L. Interactive techniques and exploratory spatial data analysis. In: Availability of data and materials Longley PA, Goodchild MF, Maguire DJ, Rhind DW, editors. Geographical Reported cases of TB for all years under study and those of AIDS in 2009 information systems: principles, techniques, management and applications. were made available by the Service of Studies in Health and Health New York: Wiley; 1999. p. 251–64. Information-Ministry of Health (SSHHI-MH) [9]. 14. Anselin L. Exploratory spatial data analysis in a geocomputational environment. In: Longley PA, Brooks SM, McDonnell R, Macmillan B, editors. Geocomputation, a primer. New York: Wiley; 1998. p. 77–94. Authors’ contributions 15. Haining R. Spatial data analysis in the social and environmental sciences. MS conceived and designed the study, performed the spatial and statistical Cambridge: University Press; 1990. data analysis and data interpretation and led the writing of this manuscript. 16. Bailey TC, Gatrell AC. Interactive spatial data analysis. Harlow: Longman; JEB contributed to data interpretation, and critically revised the draft version of the paper. Both authors read and approved the final version of the 17. Maroc. HCP. RGPH2014. http://www.hcp.ma/Demographie-population_r142. manuscript. html. Accessed 11 Apr 2017. (in French). 18. Dye C, Lonnroth K, Jaramillo E, Williams BG, Raviglione M. Trends in Ethics approval and consent to participate tuberculosis incidence and their determinants in 134 countries. Bull World Not applicable. Health Organ. 2009;87:683–91. 19. Obermeyer Z, Abbott-Klafter J, Murray CJ. Has the DOTS strategy improved Competing interests case finding or treatment success? An empirical assessment. PLoS One. The authors declare that they have no competing interests. 2008;3:e1721. 20. Kouyoumjian SP, Mumtaz GR, Hilmi N, Zidouh A, El Rhilani H, Alami K, et al. Author details The epidemiology of HIV infection in Morocco: systematic review and data Environmental Epidemiology Unit, National Institute of Hygiene. Ministry of synthesis. Int J STD AIDS. 2013;24:507–16. Health, Rabat, Morocco. University Hospital Center. Moulay Youssef Hospital, Rabat, Morocco. Faculty of Medicine and Pharmacy, University Mohammed V, Rabat, Morocco. Received: 27 October 2017 Accepted: 19 April 2018 References 1. WHO. Tuberculosis. Fact sheet N° 104. http://www.who.int/tb/publications/ factsheets/en/. Accessed 11 Apr 2017. 2. Maroc. Ministère de la santé. Lutte Antituberculeuse au Maroc : progrès, défis et perspectives. (in French). http://www.sante.gov.ma/Documents/ 2017/03/Lutte%20Antituberculeuse%20au%20Maroc_

Journal

Infectious Diseases of PovertySpringer Journals

Published: Jun 7, 2018

References

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off