As COVID-19 spreads rapidly, identifying most at-risk communities and providing targeted social assistance and prevention measures can be challenging.
The social and economic footprint of the COVID-19 pandemic has put enormous pressure on decision-makers to respond quickly and efficiently. As the virus spreads rapidly, identifying most at-risk communities and providing targeted social assistance and prevention measures can be challenging.
Combined with statistical modeling and machine learning tools, satellite imagery and survey data can help identify potential hot spots and facilitate local responses. We recently collaborated with Fraym, a geospatial analysis firm, on a mapping exercise for Pakistan. The analysis combined geotagged household survey data (Demographic and Health Survey 2018) with satellite imagery to create localized population information at the level of one-square-kilometer.
Satellite imagery and survey data can help identify potential hot spots in Pakistan and facilitate local responses to the COVID-19 pandemic
We highlight three specific examples that can help decision-makers in Pakistan, as well as other data analysts working in similar contexts:
Risk of exposure and disease transmission
Evidence shows that living arrangements, exposure to large groups of people, and access to basic sanitation and hygiene contribute to the transmission of COVID-19. For example, populations living in overcrowded dwellings such as intergenerational households, in high-density areas such as city slums, and areas with poor access to personal hygiene, may be at greater risk.
High-resolution estimates of exposure risk can help identify potential hotspots and develop targeted prevention measures at the community level within provinces
The COVID-19 exposure risk distribution shows that high population density and limited access to basic services in Pakistan translate into high exposure risk for a large share of the population. Zooming in, one notices this is particularly true for communities in Khyber Pakhtunkhwa and Sindh provinces, and for those living in highly populated districts and neighborhoods within these provinces.
Community behavior and access to information
Timely information about recommended protocols and behaviors for containing the disease is crucial. The access to information indicator shows the percentage of the population without a radio, television, or internet or who does not regularly consume these information sources.
Khyber Pakhtunkhwa and Sindh provinces, where the COVID-19 exposure risk is relatively high, seem to have the lowest access to communication channels. Within these two provinces, the southern districts with the highest exposure also have the lowest access to information sources, making them particularly susceptible to disease spread.
Within Khyber Pakhtunkhwa and Sindh provinces, the southern districts with the highest exposure also have the lowest access to information sources
Many countries, including Pakistan, have used various dissemination channels to promote mitigating behaviors among their citizens (e.g., television, radio, and digital media). Using high-resolution maps to understand how at-risk populations consume information can inform which communication channels are most appropriate for targeted campaigns.
Vulnerability to socioeconomic impacts
COVID-19 impacts go beyond immediate public health concerns and can trigger prolonged economic hardship. It is important to identify the populations living in areas hit hard by the pandemic to provide targeted support. The socioeconomic impact index shows that the most socioeconomically vulnerable communities in Pakistan are concentrated in southern Punjab, interior parts of Sindh, and most of Khyber Pakhtunkhwa. In these areas, lockdowns have drastically reduced economic activity and employment. As disease transmission slows down and policy focus shifts to economic recovery, populations living in these areas may be among those who need targeted support.
These high-resolution maps can inform targeted responses to COVID-19. This can be done not only at the provincial or district level, but at the neighborhood level as well. For example, by layering different risk and vulnerability maps, decision-makers can identify nuanced responses and channel their resources to specific areas with the most dire needs. For questions related to the underlying data, machine learning tools or the maps, Fraym’s data scientists can be reached at global.dev@fraym.io.
Join the Conversation