Revealed versus potential spatial accessibility of healthcare and changing patterns during the COVID-19 pandemic

Abstract

Background: Timely access to healthcare is essential but measuring access is challenging. Prior research focused on analyzing potential travel times to healthcare under optimal mobility scenarios that do not incorporate direct observations of human mobility, potentially underestimating the barriers to receiving care for many populations.

Methods: We introduce an approach for measuring accessibility by utilizing travel times to healthcare facilities from aggregated and anonymized smartphone Location History data. We measure these revealed travel times to healthcare facilities in over 100 countries and juxtapose our findings with potential (optimal) travel times estimated using Google Maps directions. We then quantify changes in revealed accessibility associated with the COVID-19 pandemic.

Results: We find that revealed travel time differs substantially from potential travel time; in all but 4 countries this difference exceeds 30 minutes, and in 49 countries it exceeds 60 minutes. Substantial variation in revealed healthcare accessibility is observed and correlates with life expectancy (⍴=-0.70) and infant mortality (⍴=0.59), with this association remaining significant after adjusting for potential accessibility and wealth. The COVID-19 pandemic altered the patterns of healthcare access, especially for populations dependent on public transportation.

Conclusions: Our metrics based on empirical data indicate that revealed travel times exceed potential travel times in many regions. During COVID-19, inequitable accessibility was exacerbated. In conjunction with other relevant data, these findings provide a resource to help public health policymakers identify underserved populations and promote health equity by formulating policies and directing resources towards areas and populations most in need.

Fig. 1. Revealed travel time to healthcare facilities using a passenger vehicle.

The inequality ratio is shown in brackets after the country name and illustrated in colored bars with colors representing income information based on World Bank classifications (see legend). Blue dots represent the median travel time in a country and vertical bars represent percentiles: 10th, 25th, 75th, and 90th. Countries are sorted by the 90th percentile. Data for other modes of transportation are presented in Supplementary Fig. 1.

Fig. 2. Relative change in median travel time and median distance traveled compared to the pre-pandemic baseline (corresponding quarter in 2019).

The three rows represent data for passenger vehicles (a, b), public transportation (c, d), and walking (e, f), respectively. The panels on the left (a, c, e) depict change in travel time, and those on the right (b, d, f) depict change in distance traveled. Box-plots summarize the revealed time statistics across countries. Boxplots edges represent the 50th (center line), 25th, and 75th percentile (box limits), whereas the whiskers extend to the minimum and maximum values but no further than 1.5 times the interquartile range.

Fig. 3. Pre-pandemic median revealed travel times to healthcare and absolute changes therein during the COVID-19 pandemic.

Panels a, c and e, depict median revealed travel times pre-pandemic (2019), whereas panels b, d, and f depict absolute changes during the COVID-19 pandemic (2020–2021, see Methods). Gray color indicates countries or regions not included in the study; see Fig. 1 and Supplementary Table 3 for the list of studied countries.

Fig. 4. Spearman’s rank correlation of revealed accessibility metrics with health outcomes.

a The correlation between the 90th percentile of the revealed travel time in a passenger vehicle, with life expectancy in 2017. b The correlation between the median revealed travel time and infant mortality rate in 2018. c The correlation between the inequality ratio and life expectancy in 2017. d The correlation between average increase in public transport travel time during the COVID-19 pandemic, and life expectancy in 2017. Note that the red line is the linear line of best fit shown for illustrative purposes; the Spearman correlation is based on the rank and is not linear with respect to these axes. Quantitative analyses of health outcomes and revealed accessibility while controlling for potential accessibility and wealth is presented in Supplementary Table 5.

Fig. 5. Estimated potential travel time for different modes of transportation.

Gray indicates countries for which data were not available due to privacy and data availability/quality (Methods). Comparisons between potential and revealed travel time are presented in Supplementary Fig. 2a.

Fig. 6. Population-weighted difference between 90th percentile of the revealed travel time and potential travel times (in minutes) for the time period of Jan 2019–Dec 2019.

The factors by which revealed travel times differ from potential are shown in parentheses after the country name. Corresponding analysis for revealed and potential travel distances is presented in Supplementary Fig. 2b.