Abstract
We investigate patterns of COVID-19 mortality across 20 Italian regions and their association with mobility, positivity, and socio-demographic, infrastructural and environmental covariates. Notwithstanding limitations in accuracy and resolution of the data available from public sources, we pinpoint significant trends exploiting information in curves and shapes with Functional Data Analysis techniques. These depict two starkly different epidemics; an "exponential" one unfolding in Lombardia and the worst hit areas of the north, and a milder, "flat(tened)" one in the rest of the country -- including Veneto, where cases appeared concurrently with Lombardia but aggressive testing was implemented early on. We find that mobility and positivity can predict COVID-19 mortality, also when controlling for relevant covariates. Among the latter, primary care appears to mitigate mortality, and contacts in hospitals, schools and work places to aggravate it. The techniques we describe could capture additional and potentially sharper signals if applied to richer data.
Introduction
The early stages of the COVID-19 pandemic in Italy were defined by extreme heterogeneity. While regions like Lombardia faced a catastrophic "exponential" outbreak, neighboring regions like Veneto—and much of central and southern Italy—experienced a significantly milder "flattened" epidemic. Traditional epidemiological models often focus on point-in-time estimates or specific parameters (like $R_0$), which can miss the rich information contained in the shape and timing of the outbreak. This study employs Functional Data Analysis (FDA) to characterize these different "shapes" and identify the socio-demographic, infrastructural, and environmental factors that explain regional variations.
Methodology
The authors treat entire mortality curves as the basic unit of observation rather than individual data points. Key techniques include:
- probKMA (Probabilistic K-mean with local alignment): A novel algorithm used to identify recurrent "motifs" or temporal patterns, allowing for grouping regions based on outbreak shape regardless of start time.
- Functional Regression: Both function-on-function (relating time-varying predictors like mobility to mortality) and function-on-scalar (relating static regional characteristics to mortality) models were utilized.
- SsNAL-EN: A functional generalization of the Elastic Net for high-dimensional feature selection to pinpoint the most relevant socio-demographic predictors.
Data and Regional Heterogeneity
The study integrates data from the Protezione Civile (DPC) and the National Statistical Institute (ISTAT). A critical finding was the discrepancy between official COVID-19 death counts and "differential mortality" (excess deaths). In hard-hit northern areas, ISTAT records were often twice as high as official DPC records during the peak, highlighting significant undercounting.
The "Two Italies": Clustering the Outbreak
The analysis confirms two starkly different epidemic motifs. Group 1, including Lombardia and Piemonte, followed an exponential path with a steep ascent. Group 2, comprising the rest of Italy and notably Veneto, followed a "flat(tened)" path. Veneto stands out as a success story; despite cases appearing concurrently with Lombardia, its curve was successfully curbed, likely due to early aggressive testing and a robust primary care response.
Predictors of Mortality
Local mobility levels and test positivity were strong predictors of mortality peaks. Beyond these, the strength of local primary care systems (measured by the number of family doctors per capita) served as a critical mitigator. Conversely, "contagion hubs" such as large hospitals, schools, and crowded workplaces were associated with higher mortality rates. The joint functional regression model captured a significant share of mortality variance, achieving an in-sample $R^2$ of 0.90.
Conclusion
While limited by the spatial resolution of regional data and a small sample size of 20 units, this functional approach provides a more nuanced understanding of the pandemic's dynamics than point-wise estimates. Future applications of these FDA techniques to finer-grained municipal or provincial data could yield even sharper insights into the drivers of infectious disease outbreaks.
Summary updated