Inuence of comorbidity on hospital outcomes among patients with COVID-19 using electronic records from a nationwide Healthcare System in Peru

Large cases reported that older age and comorbidity are predictors for poor prognosis in COVID-19 patients. Nevertheless, context-specic evidence relevant in low-and middle-income countries is still pending. Retrospective cohort study using electronic health records of conrmed cases admitted in hospitalization areas from the Peruvian Social Health Insurance. The main variable was the presence of comorbidities and the outcomes were in-hospital mortality or intensive care unit admission, and in/out hospital mortality. We used Kaplan-Meier survival curves with the Log-Rank test to compare time-to-event outcomes between comorbidities groups. Crude and adjusted Cox regression models were used to estimate hazard ratios (HR). Statistical analyses were conducted with a signicance level of 5%. of than one and patient without (Log-rank test p After adjusting for sex, age, and care network patients with one comorbidity (HR: 1.16; IC %: comorbidities (HR: IC study aims to assess the risk of ICU admission and death in hospitalized patients according to the presence and types of comorbidities in hospitals from the Peruvian Social Health Insurance.


Background
More than eight months that the World Health Organization (WHO) has declared COVID-19 as a pandemic, millions of people have died because of COVID-19 worldwide with the highest fatality rates reported in males, older adults, and people with comorbidities (1,2) . Comorbidities, including hypertension, diabetes, cardiovascular diseases, cerebrovascular disease, chronic obstructive pulmonary disease (COPD), chronic kidney disease, and malignancy, contributed signi cantly to disease severity, admission to Intensive Care Unit (ICU), acute respiratory distress syndrome (ARDS), invasive ventilation, and death (3) .
The disease has spread to almost 200 countries all around the world, but its impact varied among countries, sub-populations, and ethnicities due to biological, cultural, political, and socio-behavioral factors (4) . In the early months of the pandemic, most research about cases and deaths were reported in China, Italy, France, and the United States leaving uncertainty about the impact of the disease in low-and middle-income countries (5) . A special uncertainty arises in Latin American countries because of the lack of preparedness for the pandemic, the lack of capacity for testing, the insu cient amount of ventilator, high rates of non-communicable diseases, the building capacity of the health systems, the coexistence of other outbreaks as dengue, but also lack of leadership of some of their presidents (6) .
Peru is a Latin American country, which reported its rst case of COVID-19 on March 6, 2020; since then Peru was one of the countries with the strictest measures in our region including school and borders closures, the declaration of a national emergency with social distancing, and even the implementation of a curfew. Regarding health issues, Peru dictated measures to ensure the supply of essential health goods, and establish all the public, private, or mixed sanitary entities were under the direction of the Ministry of Health to guarantee health access (7) . Although these measures, for the second week of October, Peru ranks 3th worldwide in the con rmed COVID-19 deaths per million people (8) .
Large cases reported that older age and comorbidity are predictors for poor prognosis in COVID-19 patients (3,9,10) . Nevertheless, context-speci c evidence relevant to low-and middle-income countries is still pending. Understanding the impact and consequences of current COVID-19 may help decision-makers and governments to improve and adjust efforts in real-time. Particularly, Latin American countries need to use results from their rst months of the pandemic to settle their strategies to be prepared for a possible second wave of cases. Thus, this study aims to assess the risk of ICU admission and death in hospitalized patients according to the presence and types of comorbidities in hospitals from the Peruvian Social Health Insurance.

Study design and population
Retrospective cohort study using electronic health records from mid-and high-level care hospitals that belong to the Peruvian Social Health Insurance. Peru has a fragmented health system including public and private health institutions, the Peruvian Social Health Insurance is the second largest public health provider, which insured all the formal employed population and their families (11) . We include all patients who met the following criteria: 18 years or above subjects; con rmed SARS-CoV-2 infection; admitted in hospitalization areas between April 1 and July 19, 2020. We excluded patients hospitalized in COVID-19 Villages, a national strategy that hospitalized asymptomatic SARS-CoV-2 patients only for isolation and control of the spread of the epidemic.

Variables
From the electronic health record, we collected data about sex, age, weight, body mass index (BMI). Then we categorized age between 18-39 years, 40-59 years, and 60 years or more, and BMI was divided under 25 kg/m2, 25 kg/m2-29.9 kg/m2, and 30 kg/m2 or more. The administrative division of the Peruvian Social Health Insurance includes 34 care networks all around the country; for this study, we grouped them into four categories. We consider the three main care networks Almenara, Rebagliati, and Sabogal located in Lima (capital of Peru), Piura as the city with the most reported cases outside the capital during the study period, and the other 30 care networks grouped all together. The severity variable was built if severe acute respiratory syndrome, respiratory insu ciency, sepsis, or pulmonary edema were present in the rst 24 hours of hospitalization; the presence of at least one of these conditions was considered as a lifethreatening illness.
To generate comorbidity variables, we relied on previously grouped lists of diagnosis codes (based on International Classi cation of Diseases codes, 10th revision) named comorbidity maps (12) . We used the comorbidity map proposed by Elixhauser, which coded 31 different comorbidities using diagnosis wrote by a physician in the electronic health records (13,14) . Then we group congestive heart failure, cardiac arrhythmias, valvular disease, peripheral vascular disorders, and coagulopathy as cardiovascular diseases; hypertension uncomplicated and complicated, were considered together just like hypertension, and the same for diabetes uncomplicated and complicated, considered just as diabetes; in neoplasia category, we considered neoplasia itself, lymphoma, metastatic cancer, and solid tumor without metastasis. We keep the following comorbidities for this study: obesity, cardiovascular diseases, hypertension, diabetes, chronic kidney disease, chronic pulmonary disease, neoplasia, liver disease, and rheumatic disease. For further analysis, we consider a variable for the number of comorbidities categorized as without comorbidities, one comorbidity, and two or more comorbidities. The primary endpoint was all caused mortality occurred in or out hospitalization, the secondary endpoint was ICU admission or in-hospital mortality, whichever occurred rst. Each individual has 60 days of follow-up.

Data management
We retrieved anonymized data obtained from patients hospitalized with COVID-19 which were routinely registered in electronic health records in mid-and highlevel complexity hospitals from the Peruvian Social Security Health System (15) . Data were processed using the package data.

Statistical methods
We described sociodemographic and clinical characteristics of all patients (discharged, with ICU admission, or death) as medians and interquartile range [IQR] for numeric variables and as proportions for categorical variables. Differences between patients by the number of comorbidities (without comorbidities, one comorbidity, and two or more comorbidities) were assessed using the chi2 test for categorical variables and the non-parametric Kruskal-Wallis test for medians. We performed a time to event analysis for mortality at any point of the follow-up (in/out hospital) and in-hospital mortality or ICU-admission. The onset of follow-up for each patient was the date of hospitalization and the end of the follow-up was the date of occurrence of death, the elapsed time of 60 days, or the end of follow-up by July 19, 2020, whichever happened rst. We used Kaplan-Meier survival curves with the Log-Rank test to compare time-toevent outcomes between groups. Crude and adjusted Cox regression models were used to estimate hazard ratios (HR) for time-to-event, using cluster variance estimation, each health center de ning a cluster. The multivariate Cox regressions evaluated the number of comorbidities adjusted by sex, age, severity, and hospital care network. Statistical analyses were conducted using Stata software (version 16.0, StataCorp LLC, College Station, TX, USA) with a signi cance level of 5%.

Ethics
This study was classi ed as a minimal risk for participants. To maintain the privacy of the patients, the informatics o ce anonymized all datasets before transfer to researchers. The study protocol was approved by the Institutional Review Board of COVID studies (91-SGRyGISDIS-IETSI-ESSALUD-2020) and was also registered in the Peruvian Health Research Projects repository (PRISA, by its acronym in Spanish) ID number 1353 (18) .

Sociodemographic and clinical characteristics
Hospitals from the Peruvian Social Health Insurance attend 15,557 hospitalized patients with con rmed COVID-19 between April 1 and July 19, 2020, of whom 4,444 (28.6 %) have one comorbidity, and 3,088 (19.8 %) two or more comorbidities. Overall, 66.75% of the con rmed cases were male and the median age was 60.3 (IQR: 22.7) with 50.65% equal or above 60 years old. The median BMI was 28.7 (IQR: 6.3) and 81.03% of the patients were overweight or obese. Table 1 shows sociodemographic and clinical characteristics by the number of comorbidities. The proportion of males are higher in all the groups by the number of comorbidities. The median age for patients with two or more comorbidities is 65.7 (IQR: 17.5), with one comorbidity is 62.6 (IQR: 21.1), and without comorbidities is 56.1 (IQR: 24.9).
Overall, the proportion of comorbidities were hypertension (24.80 %), diabetes (19.26 %), obesity (12.48 %), chronic kidney disease (6.95%), cardiovascular diseases (5.16%), pulmonary disease (3.41%), neoplasia (1.92%), liver disease (1.05%), and rheumatic disease (0.86%). In the age group of 60 years or more, 56.17% had one comorbidity and 67.03% had two or more comorbidities. Patients with comorbidities are more likely to have a worst progression of the disease, in patients with ICU admission or in-hospital death 45.99% had one comorbidity and 50.26% had two or more comorbidities. Table 2 shows characteristics between discharge, ICU admission, and death groups.

Risk to ICU admission or in-hospital mortality
The Kaplan-Meier survival curves for ICU admission or in-hospital mortality shows that having comorbidities (Fig. 1). Table 3 shows Cox regression for ICU admission or in-hospital mortality. Patients with two or more comorbidities are more likely to progress to ICU admission or death compared with those without comorbidities (p < 0.05). The presence of obesity, hypertension, diabetes, and chronic kidney disease was signi cantly associated with an increased likelihood of either, ICU admission or death. After adjusting for sex, age, severity, and hospital care network patients with one comorbidity (HR: 1.09; IC 95 %: 1.01-1.18) were at higher risk to ICU admission compared with those without comorbidities. Other variables that showed to increase the risk of ICU admission or death are male sex, age, and hospital care network.

Risk to ICU in/out hospital mortality
Using in/out hospital deaths up to 60 days as the outcome, the Kaplan-Meier survival curves (Fig. 2) showed that the overall survival of patients with two comorbidities is lower than patients with one comorbidity, and both are lowest than a patient without comorbidities (Log-rank test p = 0.001). In a Cox regression, the presence of cardiovascular diseases, hypertension, diabetes, chronic kidney disease, neoplasia, and liver disease is associated with a higher risk of mortality. After adjusting for sex, age, severity, and hospital care network patients with one comorbidity (HR: 1.16; IC 95 %: 1.04-1.31) and with two or more comorbidities (HR: 1.13; IC 95%: 1.01-1.26) are at higher risk to die compared with those without comorbidities.

Discussion
Comorbidities, including obesity, cardiovascular diseases, hypertension, diabetes, chronic kidney disease, and neoplasia contributed signi cantly to ICU admission or death in hospitalized patients with con rmed COVID-19. Having two or more comorbidities concomitantly increased the risk of ICU admission by 57% and to death by 31% compared with those without comorbidities. Male sex, older age, and comorbidities have consistently shown a higher risk of mortality, but we also found that the hospital care network contributed to adverse outcomes.
The relationship between underlying diseases and the severity of COVID-19 is explained by different pathophysiology mechanisms. It has been stated the in uence of in ammatory processes with the accumulation of immune response cells in metabolic tissues (19,20) , the deterioration of lymphocyte and macrophage function (21) , the renin-angiotensin-aldosterone system imbalance (22) , and more recently, processes of hyper-in ammation and detrimental immunothrombosis (23) . Therefore, people with chronic conditions must take extra precautions to avoid COVID-19 infection.
The results of this study are similar to those published in China (24) , the United States (25,26) , Spain (15) , and Mexico (27) . In China (24) the most prevalent underlying diseases in hospitalized patients were hypertension (16.4%), cardiovascular diseases (12.1%), and diabetes (7.9%). In the United States (25) , instead, the proportion was higher, for hypertension (63.7%) and diabetes (38.4%). In comparison, our results, the most prevalent comorbidities were hypertension 24.8%, diabetes 19.3%, and obesity 12.5%. By the number of comorbidities, our proportions of the groups with one comorbidity and with two or more comorbidities are similar to those reported in Mexico, 26.0% and 19.3%, respectively.
In Peru, non-communicable diseases such as neoplasm, cardiovascular diseases, diabetes, and chronic kidney disease rank at the top causes of death and contributed signi cantly to the burden of disease (31) . Even though the different efforts implementing policies for the management of chronic diseases such as the "Plan Esperanza" for cancer (32) or the HEARTS initiative for cardiovascular diseases (33) , Peru has challenges to accomplish health international goals (34) .
The individual itself and the health system share responsibilities for the poor control of chronic diseases. On the individual's side, there is low awareness, adherence to treatment, and control (35) . On the health system's side, we found low political commitment, insu cient nancial and human resources, as well as, barriers to access to medicine and regular appointments (36) . Although in our study we did not have information about the status of treatment or control of chronic diseases to associate them with worse prognosis, it is known that patients with uncontrolled hypertension or cardiovascular diseases may induce more complications in patients with COVID-19 (37) .

Relevance on public health
The rst cases of COVID-19 in Peru were reported in March after four months of the rst cases in China. During January and February, the WHO warned countries must be prepared for a possible pandemic (38) . Peru could not be prepared timely because of past and recent health decisions of government and stakeholders, and the high mortality rate had a signi cant component of disobedience of restrictions by the population (39) . As describe above different measures are taken to face COVID-19 but more efforts are needed for vulnerable people as the elderly or with comorbidities. Many health resources including personnel, infrastructure, medicines, and budgets have been centralized on COVID-19 neglecting healthcare for people with underlying diseases. People with chronic diseases need access to treatments for proper control of their condition, the lack of attention and provision of medicines generates complications that can predispose to worse endpoints if they also acquire COVID-19. The COVID-19 pandemic will continue to be a problem in countries and until an effective cure or safe vaccine is found, it is a function of the health system to take decisions under their control and drive preventive actions tackling underlying disease instead of just COVID-19 management.

Limitations
We recognized some limitations of this study; our study design only allows describing risk factors related to comorbidities and ICU admission or mortality and serves to identify populations and characteristics that predispose to negative prognosis, we cannot establish causality. The results of this study cannot be generalized to the entire population with con rmed COVID-19 since the results only correspond to hospitalized patients. Even though electronic health records have information about clinical variables at admission, we did not include them in our result because of important proportions of missing; weight (66.9%), blood pressure (60%), body temperature (65%), oxygen saturation (96%), heart rate (51%), and breath rate (61.4%). Those missing may indicate lack of measurement or lack of report due to scarcity of human resources or high demand of patients, either way, it is important not to neglect these measurements because they may help to proper diagnosis and treatment. In some cases, the errors in patients' evaluation at admission may be differential by severity, where patients with obvious warning symptoms and signs were considered with a life-threatening illness, avoiding clinical measures, but providing prompt healthcare. We strati ed patient's characteristics by severity and adjusted Cox regression by this variable.
On the other hand, the main strength of this study is the inclusion of nationwide data that convey a wider picture concerning patients a liated with Peruvian Social Health Insurance. Another important strength is we de ned comorbidities based on the International Classi cation of Diseases at any point of the hospitalization, which provides a more reliable identi cation of the disease and reduces the risk of underestimation caused by self-report. In addition to that, we have a follow-up period that best describes the impact of COVID-19 at the beginning of the pandemic in Peru.

Conclusion
The presence of comorbidities in hospitalized patients with COVID-19 is a risk factor for ICU admission and mortality, particularly obesity, cardiovascular diseases, hypertension, diabetes, and chronic kidney disease. Also, those negative outcomes are in uenced by male sex and older age. Proper identi cation of these factors can help to identify patients at higher risk in hospital admission and provide specialized care to prevent deaths.

Declarations
Funding: This study was funded by the Instituto de Evaluación de Tecnologías en Salud e Investigación -IETSI, EsSalud -Peru.

Con icts of interest:
The authors declare no con icts of interest regarding the subject of this scienti c article.
Ethical approval: The study protocol was approved by the Institutional Review Board of COVID studies (91-SGRyGISDIS-IETSI-ESSALUD-2020) Author's contribution: JTM was responsible of data analysis and their interpretation, and writing of the rst and subsequent drafts of the paper. RAC and PSB checked data analysis and provide feedback in the writing. YHR provide methodological support and participate in the writing of the nal version of the paper. All the authors are responsible for the content of this manuscript.