Source-linked AI summary
Correlation between COVID-19 morbidity and mortality rates in Japan and local population density, temperature and absolute humidity
Sachiko Kodera, Essam A. Rashed, Akimasa Hirata
TL;DR
Prior research had not jointly assessed population density and ambient conditions in relation to COVID-19 morbidity and mortality. The authors analyzed prefectural data from Japan using correlation and multivariate regression, finding mild associations and adjusted model fits that varied by outcome. They note that cross-correlated ambient parameters and weak mortality fit when nosocomial infections were excluded constrain interpretation.
Problem
Prior studies had not simultaneously considered population density and ambient conditions in relation to morbidity and mortality across cities.
Source: §1 · p. 2Method
The study related prefectural COVID-19 morbidity and mortality to population and weather data using correlation analysis and multivariate linear regression.
Source: §1 · p. 3, §2.1 · pp. 3–4Results
Adjusted R2 was 0.69 (p < 0.01) for positive cases and 0.53 (p < 0.05) for confirmed deaths including nosocomial infection.
Source: §4 · p. 12Takeaways & Limitations
The findings may be useful when considering measures for potential future pandemics.
Source: §4 · p. 12Ambient parameters were cross-correlated, and the adjusted R2 for confirmed deaths excluding nosocomial infection was 0.15 (p = 0.25).
Source: §4 · p. 12
Abstract
from arXiv · showhide
This study analyzed the morbidity and mortality rates of the COVID-19 pandemic in different prefectures of Japan. Under the constraint that daily maximum confirmed deaths and daily maximum cases should exceed 4 and 10, respectively, 14 prefectures were included, and cofactors affecting the morbidity and mortality rates were evaluated. In particular, the number of confirmed deaths was assessed excluding the cases of nosocomial infections and nursing home patients. A mild correlation was observed between morbidity rate and population density (R2=0.394). In addition, the percentage of the elderly per population was also found to be non-negligible. Among weather parameters, the maximum temperature and absolute humidity averaged over the duration were found to be in modest correlation with the morbidity and mortality rates, excluding the cases of nosocomial infections. The lower morbidity and mortality are observed for higher temperature and absolute humidity. Multivariate analysis considering these factors showed that determination coefficients for the spread, decay, and combined stages were 0.708, 0.785, and 0.615, respectively. These findings could be useful for intervention planning during future pandemics, including a potential second COVID-19 outbreak.
1. Introduction
Prior studies gave inconsistent findings on weather and COVID-19 spread, while broader country-level analyses were not directly applicable to local regulation. This study examines how ambient conditions and population factors relate to morbidity and mortality across Japanese prefectures.
1. Introduction
- Case counts may be less reliable than mortality rates because mild or weakly symptomatic patients may not be tested, and testing criteria could delay reported cases.Japan relied partly on chest CT because PCR testing was limited; some infected people also lacked common symptoms.Source: §1 · p. 2
- Weather findings were inconsistent across studies, and country-level analyses were not directly applicable to local regulation.Prior work reported differing temperature and humidity associations, while some studies found no evidence that cases decline as weather warms.Source: §1 · p. 2
- Earlier studies did not simultaneously consider population density and ambient conditions, motivating the question of how both relate to morbidity and mortality across cities.The authors frame Japan as a relatively homogeneous case study in health care and household wealth.Source: §1 · pp. 2–3
- The study evaluates ambient temperature and humidity alongside population density and composition as correlates of prefectural COVID-19 morbidity and mortality.The authors describe this as the first study, to their knowledge, to highlight environmental factors during COVID-19 in Japan.Source: §1 · p. 3
- The authors suggest the findings may inform city-specific lockdown planning and future pandemic measures if local population and ambient conditions are significantly associated with spread.They note that prior modeling studies faced difficulty specifying weather- and population-related parameters for a novel virus.Source: §1 · pp. 2–3
2. Material and Methods
The study combines prefectural COVID-19, population, and weather data, then uses correlation and multivariate regression to assess associations across pandemic stages.
2.1. Data Source
- COVID-19 case and death data were counted through 25 May 2020, and prefectures with more than 4 deaths and 10 daily positive cases were classified as infected.Confirmed deaths were also prepared both including and excluding nosocomial or nursing-facility infections.Source: §2.1 · p. 3
- Some prefectures were excluded from comparisons because nosocomial or nursing-facility infection data were unclear, humidity data were unavailable, or geography complicated comparison.Gunma data excluded Isezaki City, where substantial nosocomial infections were reported.Source: §2.1 · pp. 3–4
- The study assembled population and area data for 14 prefectures and one city; it set an age threshold of 65, citing that more than 90% of victims were older than 60 and Japan’s retirement age is 65.Source: §2.1 · p. 4
- Weather data were averaged over spread, decay, and combined pandemic periods defined by prefecture-specific stage dates.The averaged measures included temperature, absolute humidity, wind velocity, and daylight hours.Source: §2.1 · p. 4, Table 2 · p. 5, Table 3 · p. 5
2.2. Statistical Analysis
- Spearman rank correlations and multivariate linear regression after considering multicollinearity were used to analyze demographic and environmental factors, with significance set at p < 0.05.The analysis also calculated partial-correlation probabilities and confidence intervals.Source: §2.2 · p. 6
3. Results
Across Japanese prefectures, population density and elderly share were associated with morbidity and mortality, while ambient conditions showed modest correlations. A multivariate model combined population density, elderly share, and absolute humidity to estimate cases and deaths.
3.1. Effect of Population Density and Elderly Population
- R2 was 0.394 for positive cases per million versus 0.097 for confirmed deaths and 0.259 for deaths excluding nosocomial infection in correlations with population density.The authors describe the case correlation as modest and the death correlations as slight or mild.Source: §3.1 · p. 7, Figure 2 · p. 7
- After normalizing rates by population density, elderly share correlated most with confirmed deaths excluding nosocomial infections (R2 = 0.482).The authors included elderly share as a factor in the later multivariate analysis.Source: §3.1 · p. 7, Figure 3 · p. 7
3.2. Effect of Ambient Conditions
- Moderate correlations appeared between cases per population density and daily maximum temperature and diurnal absolute humidity; wind velocity and daylight hours had weak correlations.Across most parameters, averages over the total stage correlated more strongly than averages over the spread or decay stages.Source: §3.2 · p. 8, Figure 4 · p. 8, Table 4 · p. 8
3.3. Multivariate Linear Regression
- The multivariate analysis selected population density, elderly percentage, and absolute humidity with VIFs below 3.78, 3.23, and 3.68 for spread, decay, and total durations, respectively.Maximum temperature was excluded because it was strongly correlated with absolute humidity.Source: §3.3 · p. 10
- Population density contributed 74.4% for cases per million, 80.0% for deaths per million including nosocomial infection, and 84.5% for deaths per million excluding it.Predicted and actual values were reported to correlate well using averages over the three stages.Source: §3.3 · p. 10, Figure 5 · p. 10, Table 6 · p. 11
4. Discussion
In 14 Japanese prefectures, the study found associations between morbidity and mortality rates and population density, elderly share, and ambient conditions. The authors discuss limits to these estimates and to comparisons with other epidemics.
4. Discussion
- Higher population density was associated with higher morbidity and mortality, while higher temperature and absolute humidity were generally associated with lower rates.The authors note that these weather associations were mild and not consistent across all prefectures; Ishikawa and Toyama differed from the others.Source: §4 · p. 11
- Ambient-condition effects were modest, and the authors hypothesized that population density and Japan’s relatively uniform policies, health insurance, and household wealth may help explain this.This is presented as a hypothesis rather than an established cause.Source: §4 · p. 11
- The multivariate estimates were rough, with adjusted R2 of 0.69 for positive cases, 0.53 for deaths including nosocomial infection, and 0.15 for deaths excluding it.Ambient parameters were cross-correlated; the adjusted R2 for deaths excluding nosocomial infection had p = 0.25.Source: §4 · p. 12
- The analysis examined common factors across 14 prefectures, but uncertainty in measured ambient conditions could also affect the correlations.The authors contrast this approach with city-level studies, including a 122-city China analysis that found no ambient-condition correlation.Source: §4 · pp. 11–12
- Because Japan had no reported pandemic in the preceding 50 years, the authors considered comparison with other epidemics infeasible.They note that influenza records came only from fixed hospital points, making proper comparison difficult.Source: §4 · p. 12
5. Conclusions
Multivariate regression showed mild associations between COVID-19 rates and population and weather factors, with potential use in estimating cases and deaths in Japanese cities. The study cautions that the short observation period limits general conclusions.
5. Conclusions
- Adjusted coefficients of determination were 0.69 for positive cases and 0.53 for confirmed deaths in multivariate linear regression.Both were reported at p < 0.05.Source: §5 · p. 12
- Mortality and morbidity rates showed mild correlations with population density, elderly population percentage, and maximum absolute humidity averaged over the spread stage under Japanese policy.Source: §5 · p. 12
- Population and weather data may help estimate cases and deaths in Japanese cities and inform policy and medical-resource planning.The authors suggest customized guidelines for where and when measures could be applied during a future pandemic, including a potential second wave.Source: §5 · p. 12
- Because prefectural weather data were similar during the limited March–May 2020 period, further data are needed before drawing a general conclusion.Source: §5 · p. 12
- The authors suggest that population density and elderly population percentage should be included in comparisons with international cities, as their omission may contribute to controversy among previous studies.Source: §5 · p. 12