key: cord-0760876-jdyhys5l authors: Reyes-Sánchez, Francisco; Basto-Abreu, Ana; Torres-Alvarez, Rossana; Canto-Osorio, Francisco; González-Morales, Romina; Dyer, Dwight; López-Ridaura, Ruy; Zaragoza-Jiménez, Christian A.; Rivera, Juan A.; Barrientos-Gutiérrez, Tonatiuh title: Fraction of COVID-19 hospitalizations and deaths attributable to chronic diseases date: 2021-12-16 journal: Prev Med DOI: 10.1016/j.ypmed.2021.106917 sha: f9327d293b6118c77dc4f5638339c9687d7c0b68 doc_id: 760876 cord_uid: jdyhys5l Evidence shows that chronic diseases are associated with COVID-19 severity and death. This study aims to estimate the fraction of hospitalizations and deaths from COVID-19 attributable to chronic diseases associated to poor nutrition and smoking among adults who tested positive to COVID-19 in Mexico. We analyzed 1,006,541 adults aged ≥20 who tested positive for COVID-19 from March 23 to December 5, 2020. Six chronic diseases were considered: obesity, chronic obstructive pulmonary disease (COPD), hypertension, diabetes, cardiovascular disease, and chronic kidney disease (CKD). We calibrated the database using a bias quantification method to consider undiagnosed disease cases. To estimate the total impact of multiple diseases, we defined a multimorbidity variable according to the number of diseases. Risks of hospitalization and death were estimated with Poisson regression models and used to calculate population attributable fractions (PAFs). Chronic diseases accounted for to 25.4% [95% CI: 24.8%–26.1%], 28.3% (95% CI: 27.8%–28.7%) and 15.3% (95% CI: 14.9%–15.7%) of the hospitalizations among adults below 40, 40–59, and 60 years and older, respectively. For COVID-19-related deaths, 50.1% (95% CI: 48.6%–51.5%), 40.5% (95% CI: 39.7%–41.3%), and 18.7% (95% CI, 18.0%–19.5%) were attributable to chronic diseases in adults under 40, 40–59, and 60 years and older, respectively. Chronic diseases linked to poor nutrition and smoking could have contributed to a large burden of hospitalization and deaths from COVID-19 in Mexico, particularly among younger adults. Medical and structural interventions to curb chronic disease incidence and facilitate disease control are urgently needed. Chronic diseases have been associated with COVID-19 severity and death. Meta-analytical evidence shows that people with chronic obstructive pulmonary disease, diabetes mellitus, hypertension and cardiovascular disease experience a higher risk of developing severe COVID-19. 1 Similarly, obesity, diabetes mellitus, hypertension, and cardiovascular disease have been associated with an increased risk of death. 2 Yet, the population fraction of COVID-19 cases and complications attributable to chronic diseases remains understudied. Estimating the fraction of COVID-19 deaths attributable to chronic diseases could help to inform the heterogeneity in fatality rates observed across countries and to estimate the burden of disease potentially attributable to chronic diseases. 3, 4 In Mexico, prior population attributable fraction (PAF) estimates have shown that 1.1%, 14.3% and 16.8% of deaths from COVID-19 were attributable to diabetes, hypertension, and obesity, respectively among ambulatory patients. 5 Estimating PAFs for different risk factors is informative, yet insufficient to capture the total burden of multiple chronic diseases. 6 Few studies have estimated the impact of multiple chronic diseases or multimorbidity on COVID-19. [7] [8] [9] Estimating the PAF for chronic diseases is challenging, as the PAF for each disease cannot be simply added, given their potential overlap. 6 A second challenge is misclassification bias, considering that a large proportion of the population who suffer chronic conditions may not have been diagnosed. For example, in Mexico 40% of the people with hypertension are unaware of their condition. 10 Most studies rely on self-reported disease diagnosis, misclassifying a large proportion of the population. To our knowledge, no study has attempted to estimate the impact of this bias on COVID-19 attributable fractions. We aimed to estimate the fraction of hospitalizations and deaths from COVID-19 potentially attributable to chronic diseases associated with poor nutrition and smoking among patients positive to COVID-19 in Mexico considering obesity, chronic obstructive pulmonary disease (COPD), hypertension, diabetes, cardiovascular disease, and chronic kidney disease (CKD). These diseases were selected given their prevalence and the wide range of interventions available to reduce their impact through public policy or individual-level interventions. [11] [12] [13] [14] We also implemented bias quantification methods to consider the impact of misclassification induced by self-reported data. J o u r n a l P r e -p r o o f Journal Pre-proof Data was obtained from the publicly available COVID-19 national registry provided by the Health Ministry of Mexico. 15 The surveillance system contains information on all suspected cases of COVID-19, people with at least one major symptom (cough, fever, dyspnea, or headache) and one minor symptom (myalgia, arthralgia, odynophagia, chills, chest pain, rhinorrhea, anosmia, dysgeusia, conjunctivitis). 16 The registry is based on the national hospital and sentinel surveillance systems, detailed elsewhere. 17 Briefly, all patients with a diagnosis of a severe acute respiratory infection are hospitalized and submitted to testing using reverse transcription polymerase chain reaction (PCR). For ambulatory patients, only a varying random sample of a minimum of 10% is tested by PCR analysis. PCR results and deaths are updated with varying time delays. Data available includes all clinical and epidemiological information obtained at the time of registry. For this analysis, we included individuals with a positive test for COVID-19 registered in the surveillance system during the community transmission phase of the pandemic (started March 23 rd , with last update December 7 th , 2020), with a sample size of 1,121,541subjects. We used December as the cut-off date to consider a period before vaccination in Mexico, which began on December 24th, 18 since vaccination is expected to change hospitalization and death risks. Subjects with missing values in obesity, hypertension, diabetes, COPD, cardiovascular disease, and CKD were excluded (n=4,619). We excluded subjects under 20 years (n=47,831) because multimorbidity in that age group is rare. Still, we performed a sensitivity analysis in adolescents 12 to 19 years (Table A .7 in the appendix). We excluded patients registered fifteen days before the cut-off date (n=48,852) to avoid death censoring. Finally, we excluded subjects that reported implausible delay times in seeking medical attention (n=13,698): reported 0 days from symptom onset to healthcare admission but were admitted with a serious condition (death, pneumonia, or intensive care unit). The final sample presented 1,006,541 adults who tested positive for COVID-19 from March 23rd to November 24th, 2020. Figure Two outcomes were defined: 1) hospitalization, refers to a COVID-19 case that required inpatient care; 2) death/fatality, defined as the death of a person who tested positive to COVID-19, as recorded in the database. Note that the "death" outcome is related to the case fatality rate since it is defined among diagnosed cases. Six diseases associated with poor nutrition or smoking were considered: obesity, COPD, hypertension, diabetes, cardiovascular disease, and CKD. Information was self-reported, obtained by the medical unit's epidemiologist when the person first sought medical attention. An individual could have more than one disease, which could lead to double-counting of preventable cases if fractions of each disease are simply added. 6 To estimate the overall burden of chronic diseases, we defined multimorbidity as the number of chronic diseases in four independent categories: no diseases, one disease, two diseases, or three or more diseases. Quantitative bias analysis. Since chronic disease information was self-reported, we conducted a quantitative bias analysis for diabetes and hypertension to consider undiagnosed cases. We limited the analysis to these two diseases because, except for obesity, are the most prevalent chronic diseases in Mexico, and where data was available for bias quantification. We had data to adjust obesity prevalence, but as obesity is reversible, the adjustment would not be appropriate. First, we estimated the proportion of undiagnosed cases by age group for diabetes and hypertension by comparing the 2016 National Health and Nutrition Survey (ENSANUT-2016) self-reported data against a gold standard (HbA1c and fasting blood glucose for diabetes, and systolic and diastolic blood pressure for hypertension). 19 The question used to assess self-reported diagnosis of diabetes or hypertension in ENSANUT-2016 was similar to the question used in the COVID-19 case-assessment format: "Has a doctor ever diagnosed you with [name of chronic disease]?". Sensibilities from ENSANUT-2016 are summarized in Table A .1 in the appendix. We assumed a specificity of 100% for all diseases, because the effect of treatment could affect HbA1c or BP; thus, no means to validate a positive answer exist. The adjusted prevalence of diabetes/hypertension by age groups was estimated with Rogan and Gladen's formula 20 (see the appendix for more details). Table A .2 in the appendix presents the prevalence of diabetes and hypertension before and after applying the adjustment formula. We used a sample balancing method "raking" to replicate the adjusted prevalence of diabetes and hypertension. Raking is a statistical method that adjusts a set of data so that its marginal (not adjusted) totals match control (adjusted) totals, 21 which is a simpler way to estimate risk ratios and attributable fractions compared to prior methods of misclassification analysis. 22 Raking was performed using the "survey" package on R Statistical Software [23] [24] [25] after eliminating missing data for diabetes, obesity and hypertension, and subjects under 20 years but before any other exclusions. Covariates included age (years), sex (male/female), state of residence (Mexico City as reference), cases recorded by type of epidemiological surveillance system (sentinel/no sentinel, given the differences in COVID-19 testing procedures), health system and delay in seeking medical attention. where and denote the sampling weight and the relative risk of death/hospitalization of the -th individual in the sample. PAFs compared the observed risks scenario to a counterfactual scenario where all individuals in the sample had zero diseases (relative risk = 1.0). Four sensitivity analyses were performed. The first one aimed to assess the impact of Demographic and health-related characteristics of the population are presented in Table A The proportion of severe COVID-19 attributable to multimorbidity was studied in the UK, where 51% of severe COVID-19 cases were attributed to unhealthy behaviors (13% to smoking, 9% to physical inactivity, and 30% to overweight and obesity) 8 . In the US, this fraction was estimated in 36% (14% to smoking, 12% to physical activity, and 16% to diet). 9 In another US study, 63% of hospitalizations were attributable to the joint effects of diabetes, J o u r n a l P r e -p r o o f obesity, hypertension and heart failure. In our study, 27.5% of COVID-19 hospitalizations were attributable to six chronic diseases associated with smoking and poor nutrition. We considered different diseases and employed a multimorbidity approach to combine diseases and avoid double counting, which is likely the reason why our estimate is smaller. Latin America has been heavily affected by COVID-19 fatality in young adults. Developing countries show a different pattern of fatality rate in comparison with developed countries, where COVID-19 deaths among young adults are rare. 34 We found that up to 50% of COVID-19 fatalities were attributable to chronic diseases in younger adults, compared to 18.7% in adults 60 years of age and older. A recent study showed that the differential burden of deaths in young adults between developed and developing countries could be attributable to a lower recovery rate once infected, driven by chronic diseases and poorer healthcare access. 34 We did not explore healthcare access in our analysis, and that could certainly contribute to a higher fatality rate. Future studies should explore other causes linked to increased fatality rates, such as quality of care. The data used in our analysis was produced in the context of the epidemiological surveillance system of respiratory diseases in Mexico and, as we have shown, it is subject to several sources of bias that could affect the estimation of risk ratios and, consequently, the PAF. Data on chronic diseases is self-reported and a large proportion of the population in Mexico with chronic diseases has not yet been diagnosed. We used a bias quantification approach to adjust our estimates, producing an increase in the PAF from 18.5% to 27.5% for hospitalization and from 23.3% to 35.7% in deaths (Table 2) . Unfortunately, we did not have information to adjust the misclassification for all diseases considered in our analysis. Thus, we are still likely underestimating the PAF, 35 particularly given the large impact of obesity on COVID-19 that has been reported in the literature. We did not have information about chronic disease control, which could further increase fatality. 36, 37 Our study could also be subject to selection bias. Our sample is composed mainly by subjects with severe COVID-19, since all severe cases of respiratory disease are subject to the COVID-19 test, while only a subsample of mild cases is tested. Then, the global association captures both the causal factors of COVID-19 and the imbalance in access to testing, among other biases; unfortunately, we did not have access to an estimate of the total number of mild respiratory disease, which could have allowed us to produce a better estimate. Also, we do not know the mortality coverage of the surveillance system. The death registry in Mexico has important delays, and a significant J o u r n a l P r e -p r o o f proportion of deaths have not been reported in the official COVID-19 registry, as evidenced by excess mortality estimates. 38 It is difficult to predict the direction of this bias, as it will depend on whether the differential registry is informed by chronic diseases. While we tried to quantify the most salient biases in our analyses, future studies with better designs will be needed to avoid the pitfalls of registry data and improve our understanding of the link between chronic diseases and COVID-19 severity and mortality. A sizable proportion of the hospitalizations and deaths from COVID-19 were associated with diseases caused by poor nutrition and tobacco use in Mexico, particularly among young adults. Mexico has been dealing with the double challenge of high demand for hospital services due to COVID-19 and treating chronic diseases. Individual efforts to control and reduce chronic diseases are direly needed in the short term. In the longer term, implementing structural interventions such as taxes on tobacco, sugary beverages, and high-energy foods of low nutritional value, warning labels, advertisement bans to non-nutritional food and beverages and smoke-free spaces could be critical to reduce the burden of chronic diseases and COVID-19. Figure 1 Writing -review & editing. Francisco Canto-Osorio: Investigation, Methodology, Visualization, Writing -original draft, Writing -review & editing. Romina González-Morales: Investigation, Methodology, Visualization, Writing -original draft, Writing -review & editing. Dwight Dyer: Data Curation, Conceptualization, Investigation, Methodology, Supervision, Writing -review & editing. Ruy López Ridaura: Data Curation, Conceptualization, Investigation Project administration, Resources, Supervision, Writingreview & editing. Tonatiuh Barrientos-Gutiérrez: Conceptualization, Funding acquisition, Investigation, Methodology, Project administration, Supervision, Validation, Visualization, Writing -original draft, Writing -review & editing Declaration of interests ☒ The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work Predictors of COVID-19 severity: a systematic review and meta-analysis Prevalence and Associated Risk Factors of Mortality Among COVID-19 Patients: A Meta-Analysis Infection fatality rate of COVID-19 inferred from seroprevalence data. Bulletin of the World Health Organization Population attributable fraction: names, types and issues with incorrect interpretation of relative risks Diabetes and Obesity, Major Risk Factors for Death in Patients With COVID-19 in Mexico. Archives of Medical Research Why population attributable fractions can sum to more than one Estimating excess mortality in people with cancer and multimorbidity in the COVID-19 emergency Lifestyle Risk Factors for Cardiovascular Disease in Relation to COVID-19 Hospitalization: A Community-Based Cohort Study of 387,109 Adults in UK. medRxiv Reducing COVID-19 hospitalization risk through behavior change Hipertensión arterial en adultos mexicanos: prevalencia, diagnóstico y tipo de tratamiento. Ensanut MC 2016. Salud Pública de México World Heart Federation. Enfermedades Cardiovasculares en México La enfermedad renal crónica en México. Hacia una política nacional para enfrentarla Datos Abiertos Dirección General de Epidemiología Actualización de la Definición Operacional de Caso Sospechoso de Enfermedad Respiratoria Viral Dirección General de Epidemiología: Lineamiento estandarizado para la vigilancia epidemiológica y por laboratorio de la enfermedad respiratoria viral Calendario de vacunación Diseño metodológico de la Encuesta Nacional de Salud y Nutrición de Medio Camino 2016. Salud Pública de México ESTIMATING PREVALENCE FROM THE RESULTS OF A SCREENING TEST Practical Considerations in Raking Survey Data. Survey Practice Applying Quantitative Bias Analysis to Epidemiologic Data Analysis of complex survey samples survey: analysis of complex survey samples. R package version 3 R: A Language and Environment for Statistical Computing [Internet]. R Foundation for Statistical Computing Doubly robust estimation of the generalized impact fraction Potential Impact Fraction and Population Attributable Fraction for Cross-Sectional Data The Association of Obesity, Type 2 Diabetes, and Hypertension with Severe Coronavirus Disease 2019 on Admission Among Mexican Patients Increased Risk of Hospitalization and Death in Patients with COVID-19 and Pre-existing Noncommunicable Diseases and Modifiable Risk Factors in Mexico. Archives of Medical Research Characterizing a two-pronged epidemic in Mexico of non-communicable diseases and SARS-Cov-2: factors associated with increased case-fatality rates Predicting Mortality Due to SARS-CoV-2: A Mechanistic Score Relating Obesity and Diabetes to COVID-19 Outcomes in Mexico Collider bias undermines our understanding of COVID-19 disease risk and severity. medRxiv The Younger Age Profile of COVID-19 Deaths in Developing Countries Obesity is the comorbidity more strongly associated for Covid-19 in Mexico. A case-control study Outcomes in Patients With Hyperglycemia Affected by COVID-19: Can We Do More on Glycemic Control? Diabetes Care Risk factors for COVID-19-related mortality in people with type 1 and type 2 diabetes in England: a population-based cohort study Grupo interinstitucional para la estimación del exceso de mortalidad por todas las causas This work was supported by Bloomberg Philanthropies (https://www.bloomberg.org/; JRD received the grant). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Ethics approval was not required since the work used only publicly available data.Ethical compliance: Approval was not required since the work used only publicly available data.