Excess deaths directly and indirectly attributable to COVID-19 using routinely reported mortality data, Bishkek, Kyrgyzstan, 2020: a cross-sectional study

Abstract

Objectives

Studies on excess deaths (ED) show that reported deaths from COVID-19 underestimate death. To understand mortality for improved pandemic preparedness, we estimated ED directly and indirectly attributable to COVID-19 and ED by age groups.

Design

Cross-sectional study using routinely reported individual deaths data.

Settings

The 21 health facilities in Bishkek that register all city deaths.

Participants

Residents of Bishkek who died in the city from 2015 to 2020.

Outcome measure

We report weekly and cumulative ED by age, sex and causes of death for 2020. EDs are the difference between observed and expected deaths. Expected deaths were calculated using the historical average and the upper bound of the 95% CI from 2015 to 2019. We calculated the percentage of deaths above expected using the upper bound of the 95% CI of expected deaths. COVID-19 deaths were laboratory confirmed (U07.1) or probable (U07.2 or unspecified pneumonia).

Results

Of 4660 deaths in 2020, we estimated 840–1042 ED (79–98 ED per 100 000 people). Deaths were 22% greater than expected. EDs were greater for men (28%) than for women (20%). EDs were observed in all age groups, with the highest ED (43%) among people 65–74 years of age. Hospital deaths were 45% higher than expected. During peak mortality (1 July –21 July), weekly ED was 267% above expected, and ED by disease-specific cause of death were above expected: 193% for ischaemic heart diseases, 52% for cerebrovascular diseases and 421% for lower respiratory diseases. COVID-19 was directly attributable to 69% of ED.

Conclusion

Deaths directly and indirectly associated with the COVID-19 pandemic were markedly higher than reported, especially for older populations, in hospital settings, and during peak weeks of SARS-CoV-2 transmission. These ED estimates can support efforts to prioritise support for persons at greatest risk of dying during surges.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• This is the first study in Central Asia of excess deaths associated with COVID-19 using actual individual-level mortality data (not modelled or aggregated) abstracted directly from death registers and cross-validated across multiple sources.

• This study estimated percentage of excess death directly and indirectly related to the COVID-19 pandemic and 12 other causes of death, by age groups, sex, place of death and deaths across.

• Percentage above the expected threshold level was reported, which is a more conservative metrics, and therefore, likely underestimate the true ED percentage.

• COVID-19 deaths might have been attributed to diseases, such as cardiovascular diseases or diabetes, that are known risk factors for severe COVID-19, or such diseases might have caused death in those coincidentally infected with COVID-19.

• Changing population size in Bishkek could have impacted our findings, however, we reviewed annual population data to confirm that no large or unexpected shifts occurred during the study period.

Introduction

COVID-19 was first detected in Kyrgyzstan on 17 March 2020. Until June 2020, COVID-19 testing in the country was limited, and the national case surveillance system only captured laboratory-confirmed cases. Starting in July 2020, cases of unspecified pneumonia were added as probable cases for COVID-19 surveillance. Of the 26 877 cases of COVID-19 officially registered in Bishkek in 2020, 14 952 (56%) were laboratory confirmed, and 11 952 (44%) were probable cases. The case fatality rate (CFR) was 2.5% for all cases, but the CFR differed between laboratory confirmed and probable cases (0.9% vs 4.5%, respectively; unpublished data provided by the Bishkek City Surveillance Department). The CFR for laboratory-confirmed COVID-19 in Kyrgyzstan was 20%–50% of that reported in China, Italy, Russia, Germany and the USA.^{1 2} In addition, two external analyses using aggregated data and modelling estimated that COVID-19 deaths in Kyrgyzstan were 5.2 and 12.5 times higher than the official COVID-19 reported deaths.^{2 3} According to aggregated data estimations, all-cause deaths in Kyrgyzstan in 2020 were 25% higher than expected.⁴

Accurate mortality rates for COVID-19 require high coverage of quality laboratory tests and a robust case and mortality surveillance system. However, the COVID-19 mortality rate alone does not capture the impact of the COVID-19 pandemic. Deaths from other causes not directly attributed to COVID-19 also increased during the pandemic, but these may have been prevented if appropriate interventions had been implemented.⁵ The surges of COVID-19 stressed healthcare systems and public health infrastructure, which negatively affected the provision of medical care. Shortages in supplies and providers, delays in elective and preventive services, and psychosocial and economic stressors have been associated with significant increases in excess deaths (ED).⁶

EDs are a key measure of the impact of the COVID-19 pandemic at the population level. To more accurately understand the burden of COVID-19, countries need to not only estimate overall ED from all causes but also identify those, directly and indirectly, attributable to COVID-19.⁷ Standardised approaches to measuring ED that account for differences in age, sex, and other factors affecting mortality make it possible to compare COVID-19-related mortality across regions.^7–9 ED estimates contribute to an improved understanding of the public health and demographic consequences of the pandemic. Stratifying ED by specific causes of death also helps identify people most at risk of dying during the pandemic. This knowledge can help guide efforts to protect high-risk populations during future surges. Although ED in Kyrgyzstan was not measured prior to this study, the ED per cent in the USA was estimated as 23% (522 368 ED) in 2020,⁹ and the range of the ED per cent by countries in 2020 after the pandemic began varied from the lowest in Finland and Costa Rica (2%) and Estonia (4%) to highest in Ecuador (60%) and Peru (79%).⁴

Understanding the actual burden of death associated with the COVID-19 pandemic in Kyrgyzstan is essential for strengthening the public health system, making timely adjustments to mitigation measures, and improving medical care during future outbreaks. For this purpose, we aimed to determine the number and proportion of ED in Bishkek, the capital city of Kyrgyzstan, from 1 January 2020 to 31 December 2020, by comparing weekly observed deaths with historical totals. We also sought to identify gaps in mortality reporting to help strengthen the national mortality surveillance system.

Methods

Data sources

Kyrgyzstan’s national mortality surveillance uses a paper-based system to capture data. Data are reported aggregately per region by month and analysed at a national level annually. We abstracted data from paper death registers from 2015 to 2020 from all reporting healthcare facilities in Bishkek. These included 21 outpatient healthcare facilities (19 family health centres (FHCs) and 2 clinics that provide care to specialty populations outside of the general healthcare system—The Polyclinic of the Presidential Administration and the Railroad Transport Polyclinic). The mortality data collected from the 21 healthcare facilities included deceased who had a residential address in Bishkek and were assigned to territorial FHCs or to polyclinics for the specialty population (online supplemental figure 1). Statisticians from FHCs checked the list of people who died in territorial offices of civil statistics monthly. Deaths among non-Bishkek residents (people who were treated in Bishkek-area hospitals and died there but had residence outside of Bishkek or were only temporarily living in Bishkek) were not included in FHC records and our database. Residents of Bishkek who died in other regions of Kyrgyzstan or outside the country or who died before 2015 or after 2020, were excluded from the database.

We cross-compared these data with mortality data reported by several sources (online supplemental figure 1), including the following:

• Maternal mortality data from the National Centre for Maternity and Child Welfare.

• Aggregated annual death statistics for the FHC catchment area reported on Form No 12—‘Healthcare Facility Activities Report’ (F.12) as compiled by the Ministry of Health.

• Form No 14 ‘Hospital Activity Report’ (F.14), which includes statistics for deaths that occur in hospitals as compiled by the Ministry of Health.

• Deaths reported by the National Statistics Committee of Kyrgyzstan.

• Deaths reported by the Republican Pathoanatomical Bureau and the Republican Forensic Medical Examination Centre, including deaths of homeless people.

• Deaths reported by the Bishkek territory city offices of vital statistics of the National Statistics Committee.

We collected data on the date of death, location of death, residential address, date of birth, sex, diagnosis, underlying death cause code, place of death, whether the case was accounted for in the official statistics, year of death registration, and FHC at the place of residence and compiled these data in Microsoft Excel. Population demographic data used as the denominator were obtained from the Bishkek City Statistics Department of the National Statistics Committee of Kyrgyzstan.

Data analysis

EDs were calculated as the difference between the observed number of deaths in 2020 and the expected numbers of deaths in 2020, based on an analysis of deaths from 2015 to 2019. For increased robustness of our estimates, we used two approaches to estimate expected deaths: (1) the baseline method, which used the average number of registered deaths in the previous years, and (2) the threshold method, which used the upper bound of the 95% CI of registered deaths in the preceding years.^8–10

We calculate ED in weekly intervals, and the total ED was calculated by adding ED for each week of 2020. The population SD formula from the WHO recommended statistical calculator was used to estimate the total ED and ED by age, sex, place of death and cause of death coding for pulmonary, cardiovascular, and cerebrovascular diseases, malignancies, and deaths from intended and unintended events according to the International Classification of Diseases, Tenth Revision (ICD-10) (online supplemental table 1).^{10 11}

Since the laboratory diagnosis of COVID-19 was not widely available in the early stage of the pandemic in Kyrgyzstan, the COVID-19 cases without laboratory confirmation could be registered as COVID-19-like diseases (unspecified pneumonia or influenza). Therefore, we classified deaths as being directly attributable to COVID-19 if the cause of death was listed as: (1) a COVID-19 diagnosis from a SARS-CoV-2 positive PCR (U07.1), (2) a probable COVID-19 diagnosis (U07.2, B 34.2) or (3) ED due to unspecified pneumonia or influenza (J09–J18). People with missing causes of death were excluded from cause specific subanalysis. We estimated the number of ED indirectly associated with COVID-19 by subtracting deaths directly attributed to COVID-19 from the total number of ED. We report 2020 ED rates due to all causes and by age and sex per 100 000 population using the 2020 midpoint group-specific population. ED per cent is expressed as ED count divided by expected death count using the threshold method.⁹

Patient and public involvement

There were no patients or public involvement in this study including design development, study conduct, reporting or dissemination plans.

Results

There were 23 995 deaths in Bishkek from 2015 to 2020 based on records that included missing deaths of people experiencing homelessness (46), maternal deaths (75) and deaths data from the Polyclinic of the Presidential Administration (455) and the Railroad Transport Polyclinic (527). From this, we excluded 479 records, resulting in the inclusion of 23 516 records in our analysis (online supplemental table 2).

ED estimates

Of 4660 deaths registered in 2020 in Bishkek, 840–1042 deaths were ED (table 1). Deaths were 22% above expected, with an ED rate of 79–98 ED per 100 000 people.

Table 1.

Excess death (ED) by age, sex and place of death, Bishkek, Kyrgyzstan, 2020

Characteristics	Observed deaths (count)	Expected deaths (count)*	ED (count)*	ED rate (per 100 000)*	ED per cent above expected (%)*
Overall	4660	3618–3820	840–1042	79–98	22%–29%
Sex
Male	2223	1621–1742	481–602	96–121	28%–37%
Female	2437	1896–2039	398–541	70–96	20%–29%
Place of death†
Home	2880	2406–2557	323–474	30–45	13%–20%
Hospital	1571	963–1081	490–608	46–57	45%–63%
Age
Age<65	1790	1373–1467	323–417	32–42	22%–30%
Age≥65	2870	2186–2345	525–684	814–1061	22%–31%
0–14	279	239–264	15–40	5–13	6%–17%
15-34‡	99	71–88	13–28	4–9	15%–39%
35–44	169	119–149	20–50	13–32	13%–32%
45–54	387	271–309	78–116	67–99	25%–43%
55–64	856	526–604	252–330	299–391	42%–63%
65–74	1056	637–736	320–419	751–984	43%–66%
75–84	1069	823–886	183–246	1093–1469	21%–30%
85+	745	599–677	68–146	1326–2847	10%–24%

Disaggregated results do not sum to overall numbers because of underlying imputations. EDs are calculated as the difference between observed and ED numbers for the threshold and average (baseline).

*The range of values is based on the average (baseline) and the upper bound of the 95% CI (threshold) of historical deaths (2015–2019). For simplicity, we included only ED per cent above the threshold in text and graphics.

†One hundred and forty-two people who died in another place and 67 who died in a day hospital were not included.

‡Due to the small number of deaths (25) in the age group (15–24), we combined this age group with the age group (25–34) to calculate the ED%.

EDs were 28% higher than expected for men and 20% higher than expected for women. ED per cent rose steadily by age, from 6% in those under 15 to 43% in those 65 years and older (table 1). ED peaks were observed from 8 July to 14 July, when it reached 267% above expected (figure 1). This peak was also observed for both men and women and people who died at home and in a hospital. Notably, hospital deaths were 562% above expected at the peak. By age group, ED was highest among people 55–64 years, old, where it peaked at 550% above expected (online supplemental figure 2).

Figure 1.

Deaths per week by sex, age and place of death, Bishkek, 2020 (n=4660). Figure 1 plots on the left show registered deaths overall and by sex, place of death, age group (dark solid line), the upper and lower limits (95% CI) of historical average (grey area) and per cent of excess death (shown on the right) compared with the threshold calculated based on 2015–2019 historical data (dark dashed line). *value extends above axis. ED, excess death.

ED by cause of death

ED was 291% above expected for COVID-19-like diseases (influenza and unspecified pneumonia) (table 2). COVID-19 deaths totalled 697 in 2020, including 222 laboratory confirmed COVID-19 deaths, 335 probable COVID-19 deaths, and 140 ED from COVID-19-like diseases. Of 2020 ED, 31% of deaths were not directly caused by COVID-19 but were attributable to the COVID-19 pandemic.

Table 2.

Excess death (ED) by causes of death, Bishkek, Kyrgyzstan, 2020

Causes of death*	Observed deaths (count)	Expected deaths (count)†‡	ED (count)‡	ED per cent above expected (%)
Indirectly associated with COVID-19§	3963	3534–3704	259–429	7%–12%
Influenza and pneumonia (COVID-19 like diseases)	168	28–43	125–140	291%–500%
Chronic lower respiratory diseases	75	30–44	31–45	70%–150%
Diabetes	33	15–21	12–18	57%–120%
Other diseases of the circulatory system	115	77–96	19–38	20%–49%
Chronic IHD	1725	1299–1440	285–426	20%–33%
IHD	1889	1445–1585	304–444	19%–31%
Myocardial infarctions	132	93–115	17–39	15%–42%
Perinatal death (during 0–6 days of life)	166	132–151	15–34	10%–26%
Cerebrovascular diseases (ischaemic, haemorrhagic, lacunary and other brain circulatory disorders)	361	311–344	17–50	5%–16%
Malignancies	700	615–675	25–85	4%–14%
Stroke	247	210–239	8–37	3%–18%
Intended and unintended traumas, poisoning, suicide and other adverse events	142	130–141	1–12	1%–9%

*Other diseases of the respiratory system, hypertension and heart failure were not included due to the small number of deaths (up to 10).

†Expected deaths were calculated as the difference between observed and ED numbers for threshold and average (baseline).

‡The range of values is based on the average (baseline) and the 95% upper CI (threshold) of historical deaths (2015–2019). For simplicity, we only report ED per cent above the threshold in the narrative and in graphics.

§Total observed deaths not directly attributable to COVID-19 (3963) were calculated as: total deaths (4660)−COVID-19 deaths (697). COVID-19 deaths included laboratory confirmed (222)+probable (335)+deaths due to unspecified pneumonia (140).

IHD, ischaemic heart disease.

Weekly analysis shows that ED from COVID-19-like diseases peaked 8 July –14 July at 2900% above expected. Deaths from chronic lower respiratory tract diseases were 70% above expected for the year and peaked at 445% above expected in the first week of September (figure 2).

Figure 2.

Deaths per week by cause of death groups, Bishkek, 2020 (n=4660). Figure 2 registered deaths by causes of deaths (dark solid line), average level based on 2015–2019 historical data (grey dash line), the upper and lower limits (95% CI) of historical average (grey area). The maximum weekly excess death number and per cent above the threshold is pointed in each figure. *excess death (ED) count and per cent above expected based on upper 95%CI of historical average. ED, excess death.

For causes of death other than respiratory and COVID-19-like diseases, an increase in ED percentages were observed for diabetes (57%), acute myocardial infarction (15%), chronic ischaemic heart disease (20%), other diseases of the circulatory system (20%) and perinatal death (10%) (table 2). The peak period of ED for these groups was in the middle of July and in October–December for diabetes, other diseases of the circulatory system, and perinatal death (figure 2). ED was also observed to a lesser extent for other causes of death, including unintended traumas, poisoning and other adverse events, and malignancies. There were no peaks in July for these causes of death as observed with COVID-19-like diseases and circulatory and cerebrovascular disease. Unnatural deaths (traumas, poisoning and adverse events) were less than expected during most weeks.

Characteristics of COVID-19 deaths

Of 725 COVID-19 (laboratory confirmed and probable) and unspecified pneumonia deaths, 73% occurred in hospitals, 56% were men and 49% were 65 years old and above (table 3).

Table 3.

COVID-19-associated deaths including unspecified pneumonia by place of death, age and sex, Bishkek, 2020

Characteristics	Total		Laboratory confirmed COVID-19		Probable COVID-19		Unspecified pneumonia
	N	%	N	%	N	%	N	%
Total	725		222		335		168
Place of death
Home	122	16.8	36	16.2	42	12.5	44	26.2
Hospital	530	73.1	167	75.2	243	72.5	120	71.4
Day-patient facility	64	8.8	19	8.6	45	13.4	0
Other	9	1.2			5	1.5	4	2.4
Sex
Male	403	55.6	132	59.5	180	53.7	91	54.2
Female	322	44.4	90	40.5	155	46.3	77	45.8
Age
0–17	8	1.1	0		1	0.3	7	4.2
18–64	359	49.5	105	47.3	162	48.4	92	54.8
65+	358	49.4	117	52.7	172	51.3	69	41.1

Deaths from all causes and from COVID-19 were elevated during July and coincided with the first wave of COVID-19 (figure 3). During the second COVID-19 wave from 7 October to 23 December, which lasted almost 3 months, deaths did not increase as precipitously (figure 3).

Figure 3.

Weekly registered COVID-19 cases, deaths and excess deaths (A) and death due to COVID-19 (B) per week, Bishkek, 2020. Until July 15, clinical cases of COVID-19 were not included in the official statistics, and on the graph—6293 cases of unspecified pneumonia are not shown in (A). Deaths from COVID-19 included laboratory confirmed (n=222) and probable (n=335) COVID-19 and excess death due to unspecified pneumonia using baseline level (n=140) (B). A dramatic increase in the number of COVID-19 cases, as well as the total number of registered deaths, excess mortality (A), mortality from COVID-19 (B), occurred in the same period, in the summer of 2020 (the first wave). The second wave of COVID-19 cases in October–December was twice lower than the first wave and was not accompanied by a large increase in mortality.

Discussion

Our analysis using individual-level mortality data found that mortality during the pandemic was 22% higher than expected. We estimate that 69% of ED were directly due to COVID-19 and 31% were due to other causes of death that are likely indirectly attributable to the pandemic. These results support those from other mortality studies in Kyrgyzstan that relied on aggregated data.^{2 3} Our results fall within the ranges of those reported by other countries. They are similar to estimates from Kazakhstan (27%) and Russia (24%), but higher than estimates from Uzbekistan (13%).⁴ Our results are also consistent with estimated ED per cent in European countries (15%–20%) and the USA (23% in 2020), the country with the highest case burden globally.^{12 13} Our estimated ED rates for Bishkek (79–98 per 100 000) were lower than rates in Kazakhstan (160 per 100 000) and Russia (245 per 100 000), but higher than rates in Uzbekistan (52 per 100 000).⁴

We estimate that most ED (69%) were deaths directly attributable to COVID-19. This estimate is within the range (66%¹⁴ −72.4%¹⁰) of that reported in other studies. About 1 in 10 ED were from COVID-19-like diseases that were not laboratory confirmed COVID-19 (125–140 ED). These deaths were likely COVID-19 deaths lacking laboratory confirmation due to poor laboratory capacity.

Like in some countries, ED also differed by sex and age group in our study. Specifically, we found that men had a 1.4 times higher ED rate than women. This difference was observed in Lithuania (1.4),¹⁵ Poland (1.6),¹⁵ the Czech Republic (2.0),¹⁵ in New York City, USA (2.0).¹⁶ Sex differences are not unique to COVID-19, and increased ED has been observed among men for several other viruses, including influenza.¹⁷ Age differentials were also expected, and we found substantially higher ED among people 55–64 and 65–74 years old compared with younger groups. In the European region, there was a substantial increase of ED observed for people 45 years and older.¹⁸ Older adults (>65 years old) are known to have the greatest risk of severe disease, and the majority of COVID-19 deaths have occurred in this population globally.¹⁵

More than three-quarters of ED occurred over a short period from 24 June to 28 July. This peak in ED coincided with the dates of the largest number of cases of COVID-19 registered in the city. Overload of the healthcare system occurred during this period. Hospitals had greater number of patients than usual, including patients from outside the city. Additionally, resources for routine non-COVID-19 care were diverted to care for COVID-19 patients. This may have contributed to increased non-COVID-19 ED for Bishkek residents. During the second wave of COVID-19 which occurred from October to December 2020, morbidity was half what it was in July and less acute (cases were more spread out over time). This is possibly the result of better system preparedness during this time, and evidence-based treatment recommendations being available.

Patients with heart disease have been at increased risk of severe disease and death since the beginning of the COVID-19 pandemic.¹⁹ Ischaemic heart disease deaths (not attributed to COVID-19 based on certification) accounted for most ED. Some of these ED are likely misclassification of undiagnosed COVID-19 cases because ED from ischaemic heart disease peaked at the same time as COVID-19 peaked, and coronary heart disease is associated with nearly three times the odds of poor prognosis of COVID-19.²⁰ These EDs could also have resulted from avoidance of urgent care among people with cardiovascular conditions during surges of COVID-19, as supported by the increase of ED occurring at home.²¹ ED from unnatural causes was low and can be explained by a decrease in mobility and associated accidents because of shutdowns that were put in place in the country. Also, our study did not find an increase in deaths by suicide.²² Lastly, studies have found that the COVID-19 pandemic has exacerbated adverse maternal, and perinatal outcomes.^{23 24} Excess perinatal deaths observed in our study indicate that this is also likely the case in Bishkek.

This study is subject to at least three limitations. First, the estimated ED percentage above the expected threshold level, which we used as the primary estimate to report in the article, might underestimate the true ED percentage. Second, the use of the historical average number of deaths could inaccurately estimate ED for 2020 if there was a change in Bishkek’s population or an increase of deaths due to a certain cause, other than COVID-19. However, we reviewed annual population data, including births and migration, to confirm that no large or unexpected shifts occurred during the study period that would explain the mortality trends observed.²⁵ There were also no other major infectious disease outbreaks during this time that would impact our estimates. Third, misclassification of cause of death may have occurred. Because laboratory testing for COVID-19 was not widely available, especially in the beginning of the pandemic, deaths directly attributable to COVID-19 would have been underestimated. Also, deaths from other causes, such as cardiovascular diseases,^{26 27} hypertension²⁷ and diabetes,²⁷ which are comorbidities for COVID-19, would have been overestimated.

Our results can support efforts to improve healthcare for persons at greatest risk for COVID-19 and non-COVID-19 mortality during outbreaks. Efforts that reduce excess mortality include timely adjustments of mitigation measures, and increased testing, contact tracing, isolation and quarantine.²⁸ Mortality can also be reduced by maintaining continuity of care, including preservation of essential care and mitigation for persons with chronic conditions.²⁹ Reduced access to care in emergency and non-urgent care settings contributed to gaps in critical prevention and treatment services during the COVID-19 pandemic.^{5 30} Ensuring regular antenatal care for pregnant women and care during and after delivery for newborns, as well as vaccination during pregnancy is important to reduce ED in the perinatal and early neonatal periods.^{23 24 31} Lastly, one of the biggest weaknesses in the mortality surveillance system in Kyrgyzstan is that it does not allow for timely epidemiological analysis of deaths to make real-time evidence-based decisions during a pandemic. Our results have spurred the development of an electronic registry with individual mortality data in Kyrgyzstan.³² This registry will allow for a more timely, in-depth analysis by age, sex, place of death, and causes of death both during the pandemic and in the postpandemic period, as well as facilitate the use the methodology of EDs measurement to identify unexpected mortality due to other causes. These data can also be used to help monitor the continued efficacy of COVID-19 vaccines against deaths.³³

Data availability statement

Data may be obtained from a third party and are not publicly available. Deidentified mortality data can be made available from the State Sanitary Surveillance and Disease Prevention Department of the Ministry of Health of Kyrgyzstan. Data are available on reasonable request and after getting permission from the Director of the Department (dgsenkg@mail.ru).

Ethics statements

Patient consent for publication

Not applicable.

Ethics approval

This study was reviewed and approved as a study involving non-human subjects by the local Kyrgyz Ethical Committee of the Ministry of Health of Kyrgyz Republic (Protocol No 8, December 11, 2020). This study received a non-human subject research determination by the U.S. Centers for Disease Control and Prevention.