A Bayesian framework for estimating the risk ratio of hospitalization for people with comorbidity infected by SARS-CoV-2 virus

Article activity feed

1. 

   ScreenIT

   Mar 1, 2021

   SciScore for 10.1101/2020.07.25.20162131: (What is this?)

   Please note, not all rigor criteria are appropriate for all manuscripts.

   Table 1: Rigor

   NIH rigor criteria are not applicable to paper type.

   Table 2: Resources

   No key resources detected.

   ---

   Results from OddPub: We did not detect open data. We also did not detect open code. Researchers are encouraged to share open data when possible (see Nature blog).

   ---

   Results from LimitationRecognizer: We detected the following sentences addressing limitations in the study:

   One limitation with our current analysis is that our estimated hospitalization risk for patients with a comorbidity of interest (e.g., diabetes) may be confounded with other comorbidities in the same patient (e.g., hypertension). This limitation is attributed to our lack of access to the necessary data rather than our Bayesian approach …

   SciScore for 10.1101/2020.07.25.20162131: (What is this?)

   Please note, not all rigor criteria are appropriate for all manuscripts.

   Table 1: Rigor

   NIH rigor criteria are not applicable to paper type.

   Table 2: Resources

   No key resources detected.

   ---

   Results from OddPub: We did not detect open data. We also did not detect open code. Researchers are encouraged to share open data when possible (see Nature blog).

   ---

   Results from LimitationRecognizer: We detected the following sentences addressing limitations in the study:

   One limitation with our current analysis is that our estimated hospitalization risk for patients with a comorbidity of interest (e.g., diabetes) may be confounded with other comorbidities in the same patient (e.g., hypertension). This limitation is attributed to our lack of access to the necessary data rather than our Bayesian approach per se. In this study, we relied upon two published summary statistics (i.e., COVID-NET and New York), which did not include the details of joint comorbidities in their publications3,5. For researchers who can access the complete medical records (instead of just summary statistics), they would be able to obtain the frequency of joint comorbidities (e.g., number of COVID-19 hospitalized patients with both diabetes and hypertension). Then, our Bayesian approach could be applied directly to estimate the hospitalization risk for such joint comorbidities. Specifically, instead of using the frequency of a particular comorbidity (e.g., diabetes), our model would use the frequency of the joint comorbidities (e.g., diabetes and hypertension). In addition, the joint prevalence of comorbidities (e.g., diabetes and hypertension) would be used as informative priors. The only biological assumption in our Bayesian model is that people in the general population, regardless of the status of their comorbidities, could be equally infected by the SARS-CoV-2 virus (no assumptions on the severity of the symptoms after infection were made). The assumption of equal ch...

   ---

   Results from TrialIdentifier: No clinical trial numbers were referenced.

   ---

   Results from Barzooka: We did not find any issues relating to the usage of bar graphs.

   ---

   Results from JetFighter: We did not find any issues relating to colormaps.

   ---

   Results from rtransparent:

   • Thank you for including a conflict of interest statement. Authors are encouraged to include this statement when submitting to a journal.
   • Thank you for including a funding statement. Authors are encouraged to include this statement when submitting to a journal.
   • No protocol registration statement was detected.

   ---

   Read the original source
2. Version published to 10.1093/jamia/ocaa246

   Sep 29, 2020
3. 

   ScreenIT

   Jul 30, 2020

   SciScore for 10.1101/2020.07.25.20162131: (What is this?)

   Please note, not all rigor criteria are appropriate for all manuscripts.

   Table 1: Rigor

   NIH rigor criteria are not applicable to paper type.

   Table 2: Resources

   No key resources detected.

   ---

   Results from OddPub: We did not detect open data. We also did not detect open code. Researchers are encouraged to share open data when possible (see Nature blog).

   ---

   Results from LimitationRecognizer: We detected the following sentences addressing limitations in the study:

   One limitation with our current analysis is that our estimated hospitalization risk for patients with a comorbidity of interest (e.g., diabetes) may be confounded with other comorbidities in the same patient (e.g., hypertension). This limitation is attributed to our lack of access to the necessary data rather than our Bayesian …

   SciScore for 10.1101/2020.07.25.20162131: (What is this?)

   Please note, not all rigor criteria are appropriate for all manuscripts.

   Table 1: Rigor

   NIH rigor criteria are not applicable to paper type.

   Table 2: Resources

   No key resources detected.

   ---

   Results from OddPub: We did not detect open data. We also did not detect open code. Researchers are encouraged to share open data when possible (see Nature blog).

   ---

   Results from LimitationRecognizer: We detected the following sentences addressing limitations in the study:

   One limitation with our current analysis is that our estimated hospitalization risk for patients with a comorbidity of interest (e.g., diabetes) may be confounded with other comorbidities in the same patient (e.g., hypertension). This limitation is attributed to our lack of access to the necessary data rather than our Bayesian approach per se. In this study, we relied upon two published summary statistics (i.e., COVID-NET and New York), which did not include the detail of joint comorbidities in their publications3,5. For researchers who can access the complete medical records (instead of just summary statistics), they would be able to obtain the frequency joint comorbidities (e.g., number of COVID-19 hospitalized patients with both diabetes and hypertension). Then, our Bayesian approach could be applied directly to estimate the hospitalization risk for such joint comorbidities. Specifically, instead of using the frequency of a particular comorbidity (e.g., diabetes), our model would use the frequency of the joint comorbidities (e.g., diabetes and hypertension). In addition, the joint prevalence of comorbiditie (e.g., diabetes and hypertension) would be used as informative priors. The only biological assumption in our Bayesian model is that people in the general populatio regardless of the status of their comorbidities, could be equally infected by the SARS-CoV-2 vir (no assumptions on the severity of the symptoms after infection were made). The assumption o equal chance of infection by the virus is based on the rationale that SARS-CoV-2 is a newly emerged virus to the human population; thus, nobody is particularly immune to the virus. For example, it has been reported that viral loads were similar in asymptomatic and symptomatic patients6, and young children and adults were both similarly infected by the virus7. If future research shows that people with specified comorbidities do have a different chance of being infected, our Bayesian approach can be modified to accommodate that difference (e.g., using a scaling factor in our model to reflect the differential degree of infection probability). Materials and Methods Bayesian modeling The overview of our Bayesian model is depicted in the figure 2. We classify people in the general population into three categories: (1) h - the people without any known comorbidities, (2 - the people with a particular comorbidity of interest (e.g., diabetes), who may or may not carry other types of comorbidities, and (3) o - the people with some other types of comorbidities excluding c (e.g., any other types of comorbidities excluding hypertension). Let Nh, Nc, and No denote the number of hospitalized COVID-19 patients for these three categories, respectively. N denote the total number of COVID-19 patients in the hospital, i.e., N = Nh + Nc + No. Then the vector (Nh, Nc, No) is a multinomial random variable8: (Nh, Nc, No) ~ Multinomial (q, N) (1) where q denotes a vector of unobserved multinomial probabilities (qh, qc, qo) corresponding to N Nc, and No, respectively. Let th, tc, and to denote the unknown probabilities of hospitalization for people infected by SARS-CoV-2 virus in the category of h, c, and o, respectively, in the general population. For example, if 50,000 out of a total of 200,000 infected people with a particular comorbidity of interest (i.e., in the category of c) were eventually hospitalized, tc would be equal to 0.25 (i.e., 50,000/200,000). We define tc/th to be the risk ratio of the probability of hospitalization for the infected people with a particular comorbidity of interest c (e.g., diabetes) versus the probability hospitalization for the infected people without any comorbidities. It’s important to note that th, tc and to are different from qh, qc, and qo. However, we derived an algebraic relationship between tc/th and qc/qh, as shown below. Let kh and kc denote the proportion of people without any medical conditions and the people with a comorbidity of interest c (e.g., diabetes), respectively, in the general population of size Npop. Let rh, rc, and ro denote the probabilities of being infected by SARS-CoV-2 virus for peop in the categories of h, c, and o, respectively. Then, q can be expressed as follows: qh = thrhkhNpop/(thrhkhNpop + tcrckcNpop + toro(1-kh - kc)Npop) qc = tcrckcNpop/(thrhkhNpop + tcrckcNpop + toro(1-kh - kc)Npop) (2) (3) In Eq. (2), the numerator (thrhkhNpop) corresponds to the expected number of hospitalized COVID-19 patients without any comorbidities. Specifically, out of the general population of size Npop, khNpop is the expected number of people without any comorbidities given the definition of k rhkhNpop is the expected number of people without any comorbidities infected by the virus given the definition of rh; thrhkcNpop is the expected number of infected people without any comorbidities who were hospitalized given the definition of th. Similarly, in Eq. (3), the numerato (tcrckcNpop) corresponds to the expected number of hospitalized COVID-19 patients with the comorbidity of interest c. In addition, the term toro(1-kh - kc)Npop in the denominator of both Eq. and (3) corresponds to the expected number of hospitalized COVID-19 patients with some othe types of comorbidities excluding c. Therefore, the denominator in both Eq. (2) and (3) corresponds to the expected total number of hospitalized COVID-19 patients, regardless of the status of their comorbidities. If we assume that people in the general population, regardless of the status of their comorbidities, have an equal chance of being infected by SARS-CoV-2 virus (see Discussion), then rh = rc = ro and the Eq. (2) and (3) can be simplified by canceling out rh, rc, ro, and Npop from both the numerator and denominator. The simplified Eq. (2) and (3) are shown as follows qh = thkh/(thkh + tckc + to(1-kh - kc)) qc = tckc/(thkh + tckc + to (1-kh - kc)) (4) (5) Dividing Eq. (5) by Eq. (4), we obtain the expression of tc/th as follows: tc /th = khqc /kcqh (6) To estimate the posterior probability of tc/th, we need to sample from the following posterior distribution: Prob(kh, kc, q | N, Nh, Nc, No) ∝ Multinomial(Nh, Nc, No | q, N)Prior (q)Prior(kh)Prior(kc) (7) To specify the prior distribution for q, we chose a Dirichlet distribution8 as it is commonly use as the conjugate prior of the multinomial likelihood described in Eq. (1). (qh, qc, qo) ~ Dirichlet (a1, a2, a3) (8) where a1, a2, a3 correspond to the shape parameters of Dirichlet distribution. For this study, we set a1 = a2 = a3 = 1 to set a uniform prior, although those values can be adjusted to more accurately reflect the probabilities of hospitalization for each category of patients if more data becomes available in the future. To specify the prior distributions for kh and kc, we chose beta distributions as they are commonly used to model proportions8. kh ~ Beta (ah, bh) kc ~ Beta (ac, bc) (9) (10) where ah, bh, ac, and bc denote shape parameters of the corresponding beta distributions. Usin the method of moments8, these parameters can be expressed as follows: ah = bh = ac = bc = µh(µh(1-µh)/sh2 -1) (1-µh)(µh(1-µh)/sh2 -1) µc(µc(1-µc)/sc2 -1) (1-µc)(µc(1-µc)/sc2 -1) (11) (12) (13) (14) where µh and sh2, and µc and sc2 represent the mean and variance of the proportions of the healthy people and people with the comorbidity of interest c, respectively, in the general population. Then, by plugging in the priors in Eq. (7), the posterior distribution becomes the following: Prob(kh, kc, q | N, Nh, Nc, No) ∝ Multinomial(Nh, Nc, No | q ) ×Dirichlet(a1, a2, a3) ×Beta(ah, bh) ×Beta(ac, bc) (15) In summary, the foundation of our approach is based on our derived algebraic relationship (Eq. 6) between the quantity of tc/th (the risk ratio) and the quantities of kh, kc, and q. Using a uniform Dirichlet distribution, q is modeled by a noninformative prior; q is related to the observe data (N, Nh, Nc, and No) in hospitalized COVID-19 patients through the multinomial likelihood as described in Eq. (1). kh and kc are modeled by informative priors using beta distributions whose shape parameters were expressed using the published prevalence rates for comorbidities in th general population. Through sampling from the posterior distribution of kh, kc, and q, we were ab to estimate the posterior distribution of tc/th as a derived quantity of khqc /kcqh. We used WinBUGS9 (version 1.4.3) to implement the above models. The posterior distributions of risk ratios for different comorbidities were estimated with the Markov Chain Mon Carlo (MCMC) sampling strategy implemented in WinBUGS10 using the following parameters: t number of chains of four, the number of total iterations of 100,000, burn-in of 10,000, and thinn of 4. Convergence and autocorrelations were evaluated with trace/history and autocorrelation plots. Multiple initial values were applied for MCMC sampling. Comorbidity data for hospitalized COVID-19 patients For the above Bayesian approach, the following two types of data are required: (1) the frequency of the comorbidity of interest (e.g., diabetes) in COVID-19 patients in hospitals, and ( the prevalence of the comorbidity in the general population. For the comorbidity frequency of hospitalized COVID-19 patients, we used a large-scale dataset, available at COVID-NET5, collected from 154 acute care hospitals in 74 counties in 13 states in U.S. from March 1 to May this COVID-NET dataset, 314 had asthma, 266 had COPD, 859 had cardiovascular diseases, 819 had diabetes, 1154 were obese, 1428 had hypertension, and 336 had no known medical conditions. Besides the COVID-NET dataset, we also used a published dataset from the state o New York3 collected from 12 hospitals in New York City, Long Island, and Westchester County from March 1 to April 4, 2020. Among a total of 5700 hospitalized COVID-19 patents in this Ne York dataset, 479 had asthma, 287 had COPD, 966 had cardiovascular diseases, 1808 had diabetes, 1737 were obese, 3026 had hypertension, and 350 had no known medical conditions In both the COVID-NET and New York datasets, cardiovascular disease referred to coronary artery disease and congestive heart failure. For the prevalence of comorbidities in general U.S. adult population, the following estimates (mean ± standard error) by the U.S. public health government agencies were used: asthma (7.7%±0.22%)11, cardiovascular disease (5.6%±0.14%)12, COPD (5.9%±0.051%)13, diabetes (13%±5.6%)14, obesity (42.4%±1.8%)15, and hypertension (49.1%±1.5%)16. The proportion of healthy adults in the U.S. who have no medical conditions was estimated to be 12.2% (95% CI: 10.9–13.6)17.

   ---

   Results from Barzooka: We did not find any issues relating to the usage of bar graphs.

   ---

   Results from JetFighter: We did not find any issues relating to colormaps.

   ---

   About SciScore

   SciScore is an automated tool that is designed to assist expert reviewers by finding and presenting formulaic information scattered throughout a paper in a standard, easy to digest format. SciScore is not a substitute for expert review. SciScore checks for the presence and correctness of RRIDs (research resource identifiers) in the manuscript, and detects sentences that appear to be missing RRIDs. SciScore also checks to make sure that rigor criteria are addressed by authors. It does this by detecting sentences that discuss criteria such as blinding or power analysis. SciScore does not guarantee that the rigor criteria that it detects are appropriate for the particular study. Instead it assists authors, editors, and reviewers by drawing attention to sections of the manuscript that contain or should contain various rigor criteria and key resources. For details on the results shown here, including references cited, please follow this link.

   Read the original source
4. 

   ScreenIT

   Jul 30, 2020

   SciScore for 10.1101/2020.07.25.20162131: (What is this?)

   Please note, not all rigor criteria are appropriate for all manuscripts.

   Table 1: Rigor

   NIH rigor criteria are not applicable to paper type.

   Table 2: Resources

   No key resources detected.

   ---

   Results from OddPub: We did not detect open data. We also did not detect open code. Researchers are encouraged to share open data when possible (see Nature blog).

   ---

   Results from LimitationRecognizer: We detected the following sentences addressing limitations in the study:

   One limitation with our current analysis is that our estimated hospitalization risk for patients with a comorbidity of interest (e.g., diabetes) may be confounded with other comorbidities in the same patient (e.g., hypertension). This limitation is attributed to our lack of access to the necessary data rather than our Bayesian …

   SciScore for 10.1101/2020.07.25.20162131: (What is this?)

   Please note, not all rigor criteria are appropriate for all manuscripts.

   Table 1: Rigor

   NIH rigor criteria are not applicable to paper type.

   Table 2: Resources

   No key resources detected.

   ---

   Results from OddPub: We did not detect open data. We also did not detect open code. Researchers are encouraged to share open data when possible (see Nature blog).

   ---

   Results from LimitationRecognizer: We detected the following sentences addressing limitations in the study:

   One limitation with our current analysis is that our estimated hospitalization risk for patients with a comorbidity of interest (e.g., diabetes) may be confounded with other comorbidities in the same patient (e.g., hypertension). This limitation is attributed to our lack of access to the necessary data rather than our Bayesian approach per se. In this study, we relied upon two published summary statistics (i.e., COVID-NET and New York), which did not include the detail of joint comorbidities in their publications3,5. For researchers who can access the complete medical records (instead of just summary statistics), they would be able to obtain the frequency joint comorbidities (e.g., number of COVID-19 hospitalized patients with both diabetes and hypertension). Then, our Bayesian approach could be applied directly to estimate the hospitalization risk for such joint comorbidities. Specifically, instead of using the frequency of a particular comorbidity (e.g., diabetes), our model would use the frequency of the joint comorbidities (e.g., diabetes and hypertension). In addition, the joint prevalence of comorbiditie (e.g., diabetes and hypertension) would be used as informative priors. The only biological assumption in our Bayesian model is that people in the general populatio regardless of the status of their comorbidities, could be equally infected by the SARS-CoV-2 vir (no assumptions on the severity of the symptoms after infection were made). The assumption o equal chance of infection by the virus is based on the rationale that SARS-CoV-2 is a newly emerged virus to the human population; thus, nobody is particularly immune to the virus. For example, it has been reported that viral loads were similar in asymptomatic and symptomatic patients6, and young children and adults were both similarly infected by the virus7. If future research shows that people with specified comorbidities do have a different chance of being infected, our Bayesian approach can be modified to accommodate that difference (e.g., using a scaling factor in our model to reflect the differential degree of infection probability). Materials and Methods Bayesian modeling The overview of our Bayesian model is depicted in the figure 2. We classify people in the general population into three categories: (1) h - the people without any known comorbidities, (2 - the people with a particular comorbidity of interest (e.g., diabetes), who may or may not carry other types of comorbidities, and (3) o - the people with some other types of comorbidities excluding c (e.g., any other types of comorbidities excluding hypertension). Let Nh, Nc, and No denote the number of hospitalized COVID-19 patients for these three categories, respectively. N denote the total number of COVID-19 patients in the hospital, i.e., N = Nh + Nc + No. Then the vector (Nh, Nc, No) is a multinomial random variable8: (Nh, Nc, No) ~ Multinomial (q, N) (1) where q denotes a vector of unobserved multinomial probabilities (qh, qc, qo) corresponding to N Nc, and No, respectively. Let th, tc, and to denote the unknown probabilities of hospitalization for people infected by SARS-CoV-2 virus in the category of h, c, and o, respectively, in the general population. For example, if 50,000 out of a total of 200,000 infected people with a particular comorbidity of interest (i.e., in the category of c) were eventually hospitalized, tc would be equal to 0.25 (i.e., 50,000/200,000). We define tc/th to be the risk ratio of the probability of hospitalization for the infected people with a particular comorbidity of interest c (e.g., diabetes) versus the probability hospitalization for the infected people without any comorbidities. It’s important to note that th, tc and to are different from qh, qc, and qo. However, we derived an algebraic relationship between tc/th and qc/qh, as shown below. Let kh and kc denote the proportion of people without any medical conditions and the people with a comorbidity of interest c (e.g., diabetes), respectively, in the general population of size Npop. Let rh, rc, and ro denote the probabilities of being infected by SARS-CoV-2 virus for peop in the categories of h, c, and o, respectively. Then, q can be expressed as follows: qh = thrhkhNpop/(thrhkhNpop + tcrckcNpop + toro(1-kh - kc)Npop) qc = tcrckcNpop/(thrhkhNpop + tcrckcNpop + toro(1-kh - kc)Npop) (2) (3) In Eq. (2), the numerator (thrhkhNpop) corresponds to the expected number of hospitalized COVID-19 patients without any comorbidities. Specifically, out of the general population of size Npop, khNpop is the expected number of people without any comorbidities given the definition of k rhkhNpop is the expected number of people without any comorbidities infected by the virus given the definition of rh; thrhkcNpop is the expected number of infected people without any comorbidities who were hospitalized given the definition of th. Similarly, in Eq. (3), the numerato (tcrckcNpop) corresponds to the expected number of hospitalized COVID-19 patients with the comorbidity of interest c. In addition, the term toro(1-kh - kc)Npop in the denominator of both Eq. and (3) corresponds to the expected number of hospitalized COVID-19 patients with some othe types of comorbidities excluding c. Therefore, the denominator in both Eq. (2) and (3) corresponds to the expected total number of hospitalized COVID-19 patients, regardless of the status of their comorbidities. If we assume that people in the general population, regardless of the status of their comorbidities, have an equal chance of being infected by SARS-CoV-2 virus (see Discussion), then rh = rc = ro and the Eq. (2) and (3) can be simplified by canceling out rh, rc, ro, and Npop from both the numerator and denominator. The simplified Eq. (2) and (3) are shown as follows qh = thkh/(thkh + tckc + to(1-kh - kc)) qc = tckc/(thkh + tckc + to (1-kh - kc)) (4) (5) Dividing Eq. (5) by Eq. (4), we obtain the expression of tc/th as follows: tc /th = khqc /kcqh (6) To estimate the posterior probability of tc/th, we need to sample from the following posterior distribution: Prob(kh, kc, q | N, Nh, Nc, No) ∝ Multinomial(Nh, Nc, No | q, N)Prior (q)Prior(kh)Prior(kc) (7) To specify the prior distribution for q, we chose a Dirichlet distribution8 as it is commonly use as the conjugate prior of the multinomial likelihood described in Eq. (1). (qh, qc, qo) ~ Dirichlet (a1, a2, a3) (8) where a1, a2, a3 correspond to the shape parameters of Dirichlet distribution. For this study, we set a1 = a2 = a3 = 1 to set a uniform prior, although those values can be adjusted to more accurately reflect the probabilities of hospitalization for each category of patients if more data becomes available in the future. To specify the prior distributions for kh and kc, we chose beta distributions as they are commonly used to model proportions8. kh ~ Beta (ah, bh) kc ~ Beta (ac, bc) (9) (10) where ah, bh, ac, and bc denote shape parameters of the corresponding beta distributions. Usin the method of moments8, these parameters can be expressed as follows: ah = bh = ac = bc = µh(µh(1-µh)/sh2 -1) (1-µh)(µh(1-µh)/sh2 -1) µc(µc(1-µc)/sc2 -1) (1-µc)(µc(1-µc)/sc2 -1) (11) (12) (13) (14) where µh and sh2, and µc and sc2 represent the mean and variance of the proportions of the healthy people and people with the comorbidity of interest c, respectively, in the general population. Then, by plugging in the priors in Eq. (7), the posterior distribution becomes the following: Prob(kh, kc, q | N, Nh, Nc, No) ∝ Multinomial(Nh, Nc, No | q ) ×Dirichlet(a1, a2, a3) ×Beta(ah, bh) ×Beta(ac, bc) (15) In summary, the foundation of our approach is based on our derived algebraic relationship (Eq. 6) between the quantity of tc/th (the risk ratio) and the quantities of kh, kc, and q. Using a uniform Dirichlet distribution, q is modeled by a noninformative prior; q is related to the observe data (N, Nh, Nc, and No) in hospitalized COVID-19 patients through the multinomial likelihood as described in Eq. (1). kh and kc are modeled by informative priors using beta distributions whose shape parameters were expressed using the published prevalence rates for comorbidities in th general population. Through sampling from the posterior distribution of kh, kc, and q, we were ab to estimate the posterior distribution of tc/th as a derived quantity of khqc /kcqh. We used WinBUGS9 (version 1.4.3) to implement the above models. The posterior distributions of risk ratios for different comorbidities were estimated with the Markov Chain Mon Carlo (MCMC) sampling strategy implemented in WinBUGS10 using the following parameters: t number of chains of four, the number of total iterations of 100,000, burn-in of 10,000, and thinn of 4. Convergence and autocorrelations were evaluated with trace/history and autocorrelation plots. Multiple initial values were applied for MCMC sampling. Comorbidity data for hospitalized COVID-19 patients For the above Bayesian approach, the following two types of data are required: (1) the frequency of the comorbidity of interest (e.g., diabetes) in COVID-19 patients in hospitals, and ( the prevalence of the comorbidity in the general population. For the comorbidity frequency of hospitalized COVID-19 patients, we used a large-scale dataset, available at COVID-NET5, collected from 154 acute care hospitals in 74 counties in 13 states in U.S. from March 1 to May this COVID-NET dataset, 314 had asthma, 266 had COPD, 859 had cardiovascular diseases, 819 had diabetes, 1154 were obese, 1428 had hypertension, and 336 had no known medical conditions. Besides the COVID-NET dataset, we also used a published dataset from the state o New York3 collected from 12 hospitals in New York City, Long Island, and Westchester County from March 1 to April 4, 2020. Among a total of 5700 hospitalized COVID-19 patents in this Ne York dataset, 479 had asthma, 287 had COPD, 966 had cardiovascular diseases, 1808 had diabetes, 1737 were obese, 3026 had hypertension, and 350 had no known medical conditions In both the COVID-NET and New York datasets, cardiovascular disease referred to coronary artery disease and congestive heart failure. For the prevalence of comorbidities in general U.S. adult population, the following estimates (mean ± standard error) by the U.S. public health government agencies were used: asthma (7.7%±0.22%)11, cardiovascular disease (5.6%±0.14%)12, COPD (5.9%±0.051%)13, diabetes (13%±5.6%)14, obesity (42.4%±1.8%)15, and hypertension (49.1%±1.5%)16. The proportion of healthy adults in the U.S. who have no medical conditions was estimated to be 12.2% (95% CI: 10.9–13.6)17.

   ---

   Results from Barzooka: We did not find any issues relating to the usage of bar graphs.

   ---

   Results from JetFighter: We did not find any issues relating to colormaps.

   ---

   About SciScore

   SciScore is an automated tool that is designed to assist expert reviewers by finding and presenting formulaic information scattered throughout a paper in a standard, easy to digest format. SciScore is not a substitute for expert review. SciScore checks for the presence and correctness of RRIDs (research resource identifiers) in the manuscript, and detects sentences that appear to be missing RRIDs. SciScore also checks to make sure that rigor criteria are addressed by authors. It does this by detecting sentences that discuss criteria such as blinding or power analysis. SciScore does not guarantee that the rigor criteria that it detects are appropriate for the particular study. Instead it assists authors, editors, and reviewers by drawing attention to sections of the manuscript that contain or should contain various rigor criteria and key resources. For details on the results shown here, including references cited, please follow this link.

   Read the original source
5. 

   Version published to 10.1101/2020.07.25.20162131v1 on medRxiv

   Jul 28, 2020