Correction: Adult mortality trends in Qatar, 1989-2015: National population versus migrants

[This corrects the article DOI: 10.1371/journal.pone.0203996.].


Introduction
The major contributor to Qatar's total population growth has been the influx of healthy male migrant labor [1] to build Qatar's workforce [2,3]. Consequently, Qatar's total population increased rapidly reaching a population growth rate of 22% in 2005-2010 while it was <4% prior to 2000 [4]; with a resultant change in its age pyramid (S1a and S1b Fig). For 2015, 90% of Qatar's total population aged 15 years and above consisted of non-Qataris [5]. Qatar and other countries of the Gulf Cooperation Council possess unique demographics with high proportions of migrants in their populations, reaching over 80% in the United Arab Emirates [6]. These countries appear to be part of one of the leading labor-importing regions in the world [7] especially from the neighboring Arab and Asian countries [8]. In 2016, 64% of Qatar's total population was from five countries, namely India, Nepal, Bangladesh, Philippines, and Egypt [9]. Recently, worldwide attention has focused on an increased number of deaths among migrant workers within Qatar [10,11].
Parallel to the demographic challenges, Qatar is grappling with high rates of diabetes and obesity. Qatar is among the top ten countries with the highest diabetes prevalence worldwide [12], at 17% of adult Qataris in 2012 [13]. Additionally, adult obesity in Qatar was 41% in 2012 [13], while the global estimate was 13% [14]. The burden of adult obesity in Qatar is also higher than the other countries of the Gulf Cooperation Council (40% in Kuwait [15], 36% in Bahrain [16], 33% in UAE [17], 30% in Oman [18], and 28% in Saudi Arabia [19]). Qatar has overall benefited from socioeconomic and healthcare system developments [20]. The Global Burden of Disease Study 2015 (GBD 2015) showed decreased all-cause mortality (death from any cause) and cause-specific mortality (death from a specific cause) rates since 1990 in Qatar [21]. Remarkably, these GBD 2015 mortality trends refer to both the Qataris (nationals) and non-Qataris (migrants).

Materials and methods
For assessing the impact of demographic changes, socioeconomic and healthcare system development, and health-based interventions, measuring the magnitude of deaths by population is necessary. Therefore, we evaluated trends over 27 years of all-cause mortality and cause-specific mortality for the three commonest causes of death, namely cardiovascular and circulatory disease, neoplasms, and injuries, in Qataris and non-Qataris [21]. Additionally, we investigated mortality differentials between Qataris and non-Qataris.
Publicly available data on Qatar's yearly population size by sex, five-year-age group (from 15 year old to over 75 year old), and nationality were retrieved from MDPS's Census, Population, Housing, and Establishments annual reports for the years 2006-2014 [25]. Annual population growth rates in Qataris were computed by sex and age group for the period 2006-2014. Using these estimated population growth rates, Qatari population size between 1989-2005 and in 2015 was extrapolated assuming constant increase in population between 1989-2015 as observed between 2006-2014. Publicly available data on Qatar's total population size (combining Qataris and non-Qataris) was retrieved by sex and five-year-age group, in 1990,1995,2000,2005,2010, and 2015 from the World Population Prospect 2015 Revision by United Nations (UN) Population Division [4]. Annual population growth rates were computed by sex and age group for the periods 1990-1995, 1995-2000, 2000-2005, 2005-2010, 2010-2015. To estimate non-Qatari yearly population size by sex, we assumed that sex-specific non-Qatari population growth was similar to sex-specific Qatar's population growth, as the non-Qatari population constituted the vast majority of Qatar's total population-73% in 1986 to 90% in 2015 [5,26]. Thus, using these estimated sex-specific population growth rates and sex-specific population size in non-Qataris between 2006-2014, yearly sex-specific non-Qatari population size between 1989-2005 and in 2015 was extrapolated. Computed Qatar's total population size between 1989-2005 and in 2015 was the sum of Qatari and non-Qataris population sizes.
Estimated ASMRs and their standard errors were imported into the United States Surveillance, Epidemiology, and End Results Joinpoint Trend Analysis Software (version 4.4.0.0) [30]. Using Joinpoint regression, we assessed ASMR time trends by estimating annual percent change (APC) and average annual percent change (AAPC) in ASMR by period, nationality, and sex. Joinpoint software cannot process records with dependent variables (death rates) equal to zero [30]. As such, the years with mortality rates equal to zero were removed from the trends. When a trend was not significantly increasing or decreasing, it was considered constant. The significance of mortality trend changes were tested using Monte Carlo Permutation method (probability (p)-value threshold = 0.05). We provided contextual information that could explain all-cause mortality trends in Qataris and non-Qataris by reporting pertinent dates that marked Qatar's socioeconomic and healthcare system development [31,32].
In order to identify significant differences in all-cause mortality between Qataris and non-Qataris, we estimated Comparative Mortality Figures (CMFs), which correspond to the ratio of ASMR in Qataris divided by the ASMR in non-Qataris [27]. We also computed 95% confidence intervals (CI) of CMFs and considered the ratio between ASMRs in Qataris and non-Qataris to be significant if the unity was not included within the 95% CI limits. Additionally, years of life lost (YLL), which is a summary measure of premature death were estimated by sex, age, and nationality due to all causes of death, neoplasms, cardiovascular and circulatory disease, and injuries in 2014 and 2015. YLL was estimated using WHO standard life expectancy at age of death, which provides life expectancy in years (L) in each age group of a standard population [33]. YLL (YLL = Number of deaths x L), which gives a higher weight to death at younger ages, brings the attention to those causes of death that are more common in younger age groups.

Results
During the study period, 26,673 deaths in over-20-year-old adults were recorded (15244 in non-Qataris and 11,429 in Qataris). Deaths in males were 2.7 times more frequent than in females (19,367 and 7,306 deaths). The annual average number of all-cause deaths was 1,159 deaths ranging from 632 (1989) to -2,070 (2014). In non-Qataris, annual total number of deaths increased about four-times, from 296 deaths (1989) to 1,438 deaths (2015). In Qataris, annual total number of deaths also increased but at a lower magnitude as compared to non-Qataris (from 336 deaths in1989 to 560 in 2015). In 2014, in Qatar's total population, injuries were responsible for 19.7% of deaths (92.6% in males), followed by cardiovascular and circulatory disease 17.8% (69.9% in males), and neoplasms 15.8% (46% in females).
In 2015, we estimated 60,716 YLL (82% in males) in Qatar's total population (S1 Table). In 2014, we estimated a similar total number and sex-distribution of YLL (S2 Table). During 2014, 29.2% of YLL were attributed to injuries, followed by cardiovascular and circulatory disease (14.4%), and neoplasms (11.8%). The first cause of premature mortality in Qatari and non-Qatari males was injury counting for 25.1% (2,043 YLL) and 34.9% (14,686 YLL) of the YLL due to all causes (8,132 YLL and 42,144 YLL, respectively) in 2014. The same year, the first cause of premature mortality in Qatari and non-Qatari females was neoplasms counting for 33.5% (1,724 YLL) and 25.1% (1,636 YLL) of the YLL due to all causes (5,140 YLL and 6,520 YLL, respectively).
Between 1989-2015, all-cause ASMR (per 100,000) significantly decreased in non-Qataris (from 346 to 229 in females and 402 to 133 in males); and in Qataris (from 2,754 to 382 in females and 4,086 to 506 in males; Fig 1a and 1b, Table 1). For 1989-2015, AAPC was -3.3% in non-Qataris and -7.7% in Qataris (both sexes), which led to an AAPC at -5.5% in Qatar's total population (p-values<0.05). Even if ASMR decreased two-times faster in Qataris than in non-Qataris, all-cause ASMRs in Qataris were always significantly higher than in non-Qataris (both sexes, CIF Q/NQ ranging from 7.7 in 1989 to 3.1 in 2015). However, the magnitude of the ratio decreased over time (both sexes). Thus, all-cause ASMR in Qatar's total population was always lower than in Qataris because of the mortality rates in non-Qataris, which includes mainly migrant workers. Overall, healthy worker effect seems to influence ASMR trends in Qatar's total population.
In Qatari males and females, all-cause ASMR trends have been steadily decreasing since 1989 (APC = -7.9% and -7.6%, p-values<0.05). In non-Qatari females, the decreasing trend was steady since 1989 but at a slower pace (APC = -2.2%, p-values<0.05). This decrease in mortality appears to follow Qatar's gradual healthcare system and socioeconomic development. Remarkably, this steady decrease in mortality in non-Qatari females occurred while  [22], MDPS's Census, Population, Housing, and Establishments annual reports (http://www.qix.gov.qa/portal/page/portal/QIXPOC/ Documents/QIX%20Knowledge%20Base/Publication/Labor%20Force%20Researches/labor%20force%20sample%20survey) [25], population growth in this population did not fluctuate in the same magnitude as in non-Qatari males. In non-Qatari males, all-cause ASMR trend decreased significantly between 1998-2010 (APC = -4.6, p-values<0.05) and reached a plateau thereafter. This significant decrease in mortality appears to have occurred after massive labor migration started. Between 2005-2010, population growth reached a peak at 28% in males. Interestingly, within the same period the decrease in mortality in non-Qatari males was the highest (AAPC = -13.7%, p-value<0.05). When population growth in males was below 5% in [1990][1991][1992][1993][1994][1995][1995][1996][1997][1998][1999][2000], and 2010-2015, AAPCs for the same periods were not significant. The large increase in non-Qatari population within a short span of time could partially explain decreased ASMRs. Of note, ASMR is a weighted average of age-specific rates [27]. So, if the denominators (population size) within age groups were increasing dramatically during a short period of time due to the influx of young and/or healthy migrants while the numerators remained minimally affected, age-specific rates would also decrease. Healthy worker effect seems to influence ASMR trends in non-Qataris.
Between 1991-2014, cause-specific ASMR for cardiovascular and circulatory disease was higher in Qataris than in non-Qataris (both sexes), which resulted in higher ASMRs in Qataris compared to ASMRs in Qatar's total population (Tables 2 and 3). During this period, mortality trend declined significantly in non-Qatari females and in Qataris females : AAPC were statistically significant (p-value<0.05). 1 : We assumed that non-Qatari population growth was similar to Qatar's population (total) growth, [4] as this population constituted the vast majority of Qatar's total population (73% in 1986 [26]  For cerebrovascular disease, cause-specific ASMR was lower in non-Qataris than in Qataris until 2011 in females and males, which resulted in lower ASMRs in Qatar's total population compared to ASMRs in Qataris. Mortality in non-Qatari males decreased significantly between 1998-2014 (APC = -12.7%), while in Qataris it decreased significantly between 1991-2001 (APC = -10.8%). In females, mortality was higher in Qataris than in non-Qataris (11.4 and 8.4/ 100,000, respectively) in 2014. Mortality rates in females were higher than in males (3.8 and 2.7/100,000, respectively).
For ischemic heart disease, cause-specific ASMR was lower in non-Qataris than in Qataris (both sexes), which resulted in lower ASMRs in Qatar's total population compared to ASMRs in Qataris. ASMRs decreased significantly in non-Qatari males since 2001 (p-value<0.05), while in Qataris (both sexes) and in non-Qatari females, decreased ASMRs were observed a decade earlier (p-value<0.05).
For hypertensive disease, cause-specific ASMR was lower in non-Qataris than in Qataris (both sexes) between 1990-2014, which resulted in lower ASMRs for Qatar's total population compared to ASMRs in Qataris. Similar decreasing and significant trends were observed in Qatari females and males and in non-Qatari females (APC = -12.3%; -11.5%; and -7.3%, respectively; p-value<0.05). In non-Qatari males the trend was constant until 2001 and a significant decrease in mortality was observed thereafter (APC = -8.6%; p-value<0.05).
For neoplasms, cause-specific ASMR was lower in non-Qataris than in Qataris (both sexes), which resulted in lower ASMRs in Qatar's total population compared to ASMRs in Qataris. A significant decrease was observed in non-Qatari males and in Qataris females and males since 1991 (APC = -5.4%; -8.6%; and -6.7%. respectively, p-value<0.05). In non-Qataris females, ASMRs were constant.
For female breast cancer, cause-specific ASMR was lower in non-Qataris than in Qataris, which resulted in lower ASMRs in Qatar's total population compared to ASMRs in Qataris. As mortality trends were constant in non-Qataris and significantly declining in Qataris (APC = -5.6%; p-value<0.05), in 2014, ASMRs in non-Qataris and Qataris were similar at 22.5 and 23.7/100,000, respectively. These trends resulted in a constant mortality rate trend for Qatar's female population. For prostate cancer, cause-specific ASMR was lower in non-Qataris than in Qataris, which resulted in lower ASMRs in Qatar's total population compared to ASMRs in Qataris. Mortality in non-Qataris, of which the vast majority is constituted of young people, remained stable. However, in Qataris, mortality trend declined significantly since 2000 (APC = -8.3%; p-value<0.05).
For colorectal cancer, cause-specific ASMR was lower in non-Qataris than in Qataris, which resulted in lower ASMRs in Qatar's total population compared to ASMRs in Qataris. The exception to this trend was in the mid-2000s and in 2014, when ASMRs were lower in Qatari females than in non-Qatari females. Modeled trends were constant in Qatari and non-Qatari males and females.
For injuries, cause-specific ASMR was lower in non-Qataris than in Qataris (both sexes), which resulted in lower ASMRs in Qatar's total population compared to ASMRs in Qataris. Qatari males had the highest ASMRs during the study period comparing to Qatari-females and non-Qatari males and females. Mortality decreased significantly in Qatari females and males since 2000 (APC = -8.0% and -5.2%, respectively; p-value<0.05). Between 2000-2014, in non-Qatari females and males, mortality trends were constant; except for males between 2005-2010 (APC = -20.5%, p-value<0.05).
For all falls, cause-specific ASMR in males was higher in non-Qataris than in Qataris, which resulted in higher ASMRs in Qatar's total population compared to ASMRs in Qataris. Deaths from falls were recorded for Qatari females only in 2005 (ASMR = 3.9/100,000). In non-Qatari females, deaths were recorded between 2007 and 2014; and ASMRs (per 100,000) ranged from 0.4 in 2014 and 2.5 in 2010 (modeled trend was constant). In Qatari males, deaths from falls have not been recorded since 2009 (ASMR = 1.2/100,000). In non-Qataris, mortality decreased significantly in males between 2001-2014 from 5.1 to 2.6/100,000 (APC = -11.3%, p-value<0.05).
For transport injuries, cause-specific ASMR was lower in non-Qataris than in Qataris (both sexes), which resulted in lower ASMRs in Qatar's total population compared to ASMRs in Qataris. Mortality trends decreased significantly in non-Qatari females and in Qatari females and males (APC = -11.6%; -6.8%; and -5.4%, respectively; p-value<0.05), while it was constant in non-Qatari males. In 2014, in males, ASMR was 13.4 and 53.8/100,000 in non-Qataris and Qataris, respectively. In females, it was 3.7 and 11.3/100,000 respectively.

Discussion
Our analysis of Qatar's vital registration data provides valuable insight on mortality trends in Qataris and non-Qataris (migrants). As reported elsewhere [10,11], number of deaths did increase in non-Qataris but this is true also for Qataris. These increased numbers were higher in magnitude for non-Qataris. However, all-cause ASMRs in non-Qatari females like in Qataris (both sexes) has steadily decreased over the last 27 years, probably owing to socioeconomic healthcare system development. In non-Qatari males, rapid demographic changes probably influenced the reduction in ASMRs because of the healthy migrant effect especially during the migration peak in 2005-2010 [34]. From 2010, cause-specific ASMRs significantly decreased for neoplasms.
The healthy migrant effect has been described in other countries such as Australia [35], Finland [36], and Belgium [37] as a lower mortality risk in migrants comparing with people who were born in the study country. In the countries of the Gulf Cooperation Council, we demonstrated previously that country-level ASMR trends were associated with population size trends likely because of the healthy migrant effect [34]. Another study emphasized that long-term residents (more than ten years of residence) in the United Arab Emirates acculturate their lifestyle with the one in the host country, which lead to a decrease of the healthy migrant effect [38]. Remarkably, in Qatar, only 14% of the non-Qatari population was long-term resident (more than ten years of residence) [39]. The percentage of short-term resident migrants (four years or less of residence) in Qatar was about 70% in 2010-20% higher than in the United Arab Emirates [39]. Here, we demonstrate that the reduction in Qatar's total population mortality (all causes, cardiovascular and circulatory disease, and neoplasms) could be partially attributed to the healthy migrant effect because the non-Qatari population demonstrates lower mortality rates for all causes, cardiovascular and circulatory disease, and neoplasms comparing to the Qatari population. The lower ASMRs in non-Qataris resulted in an even greater decrease in Qatar's total population ASMRs that could have been at higher levels otherwise. Several factors can explain the lower mortality observed in non-Qataris: migrant health screening by Qatar (selected newcomers to stay and work are negative for HIV, hepatitis B and C, syphilis, and tuberculosis); economic migrant self-selection (physically and mentally healthy individuals willing to migrate for work); and age profile (88% of the non-Qataris were aged 20-49 years in 2015) [25,35,40,41].
For injuries, mortality in non-Qatari males and females was lower than in Qatari males during the study period. In 2014, mortality due to falls was 0.4/100,000 in females and 2.6/100,000 in males. Remarkably, mortality due to falls in non-Qatari males decreased significantly suggesting safety improvement in the work environment. Non-Qatari males' work environment seems to have improved over time. In 2010-2013, yearly fatal occupational rate was 1.6/ 100,000 in the only designated hospital for major injury treatment in Qatar [42,43]. This is half what is reported in the United States (3.4/100,000 in 2012) [44]. For transport accidents, mortality was lower in non-Qataris most probably reflecting the differential in the mode of transportation (collective transportation in non-Qatari workers versus individual vehicles in Qataris).
The influx of migrants has enabled Qatar to address the labor shortfall in the country. This migration emphasizes the substantial levels of economic growth observed in the country in the last decades [26]. Mortality among non-Qataris decreased, likely due to the healthy worker effect and the socioeconomic development of the country but also due to the implementation of laws to protect non-Qataris. Several Ministerial decisions were launched in the last decade to enhance the quality of life thus leading to improved health and safety of the workers in Qatar. For instance, since 2014, employers are required to provide adequate worker accommodations equipped with fire extinguishers, smoke detectors and fire alarm systems. Additionally, a health and safety officer must be appointed for each accommodation site which has 1200 workers residing on site [45]. These efforts from the government should continue along with regulatory checks in order to sustain the mortality decline in non-Qataris.
Our study demonstrated that while young adult males in Qatar die predominantly from injuries, young adult females die from neoplasms. These identified causes of death are for a large part preventable and should be addressed appropriately to lower premature mortality among young adults in Qatar. In 2013, a law was enacted to ensure the implementation of a mandatory health insurance system of basic health services to all residents of Qatar [46]. As treatment for chronic diseases is free for non-Qataris, preventing premature mortality among non-Qatari females should focus on screening for neoplasms. Scaling-up of screening is likely to increase the number of diagnosed cancer recognized at an earlier stage that can potentially be cured.
Regarding Qatari citizens, the access to free healthcare, housing grants, and subsidized education that accompanied the socioeconomic development of the country [26] has likely contributed to mortality decline. Concerning Qatari females' premature mortality, scaling up screening may be a relevant strategy to identify cancer cases at an earlier stage, as such treatment can be initiated early and the cancer cured potentially. In Qataris males, transport accidents appear to be the major killer. Lack of compliance pertaining to seat belt use and the use of mobile phones amongst Qatar's drivers was reported [47]. Hence, we recommend the implementation of more aggressive and consistent national awareness campaigns pertaining to road safety and appropriate enforcement of laws related to road safety.
The strength of our study is that we emphasize demographic specificities in Qatar that should be taken into consideration when developing, implementing, and monitoring public health programs. Policy makers may use the estimated mortality rates and be able to distinguish between nationals (Qataris) and migrants (non-Qataris) when developing strategies to address premature mortality causes. Our analysis is relatively exhaustive unlike previous reports like GBD 2015 [48], which does not provide mortality estimates by sub-populations (Qataris versus non-Qataris). As such, we were able to compare mortality in the total population with mortality among Qatari and non-Qatari populations to further demonstrate our hypothesis of the healthy migrant effect [34].
The use of these secondary mortality data has several limitations such as missing data for several years and causes of death. Data validity is also difficult to assess because all data are retrieved from the same source (MDPS) [26]. Nevertheless, according to the UN Statistics Division, the coverage rate of the civil registration of births and deaths had reached 97% since the 1990s [26,49]. Additionally, ASMRs in non-Qataris might be underestimated: non-Qataris come to Qatar for work; so, unhealthy individuals return to their country of origin when they are not able to work productively. Furthermore, individuals reaching the retirement age return to their country of origin. Misclassified cause of death was probably also reported in death certificates [50] introducing bias in vital registration data. However, these data limitations would affect both Qataris and non-Qataris. Nevertheless, the advantage of using such unique data in our analysis is to assess mortality trends at population-level. Hence, the strength of our work is the comparison of mortality trends among Qataris and non-Qataris over an extensive period.
Qatar's overall mortality trends appear to be partially attributed to the healthy migrant effect, as migrants have lower mortality rates. Nevertheless, Qatar's socioeconomic development does also appear to drive mortality decline in both Qataris (nationals) and non-Qataris (migrants). Yet, premature mortality remains and might be addressed by scaling-up screening programs and encouraging healthy lifestyle and road safety.