COVID-19 — Impact on Healthcare Accessibility for Vulnerable Populations

Why Do We Care About Access to Health Care

COVID 19 is the biggest challenge of our times. It has become the catalyst that intensified the existing fragile relationship between healthcare providers and the at-risk population. These needs include treatment for underlying medical conditions such as asthma, diabetes, heart disease, high blood pressure, and obesity. Those with chronic diseases in the underserved population must continue medical care as they have a higher chance of contracting COVID-19.

A range of hundreds of indicators shows a worrying disruption in the World’s basic health services as governments focused on containing the spread of COVID-19. These include curtailed immunization schedules, restricted inpatient, outpatient and emergency treatment for infectious and non-communicable diseases, reduced laboratory investigations, and lowered access to mental health treatment

Despite additional funding, the continued lack of medical investment and healthcare infrastructure will present challenges to mounting an effective response against the COVID-19 pandemic. Moreover, the significant inefficiency, dysfunction, and acute shortage of healthcare delivery systems in the public sector do not match up with the growing needs of the population.

The workforce in the long-term skill facilities is faced with challenges that include staff shortages, frequent turnovers, significant resident to staff ratio, not enough personal protective equipment (PPE), and, most importantly, lack of training and education for a large portion of the staff.

How do you define a Vulnerable Group

In this project by Omdena, we used publicly available data to look at countries around the world and analyze the healthcare accessibility of vulnerable populations impacted by the COVID-19 pandemic.

One of the important tasks was to identify the vulnerable population who would be impact by COVID and also population subset who would not be able to get required medical attention, due to resources getting diverted to COVID. The interesting aspect was that the criterion was not just economic but consisted of indicators listed below :

• Health Access and Quality Index
• Physicians per 10000 Population
• Critical Beds per 10000 Population
• Basic Sanitization Service % of People
• Life Expectancy
• Probability of Dying from Chronic Diseases
• Adult Mortality Rate per 1000 Population
• Population Over Age 65 Percentage
• Nurses per 10000 Population
• Population Density
• Health Expenditure as % of GDP

Data was collected from the University of Oxford for the majority of countries. This is where the first challenge in terms of getting the most recent and accurate data emerged. There were countries with the most recent socio-economic, COVID test, Infection, Recovery, and death rates reporting their data and there were countries that had a paucity of data either for technical or political perspective. At this stage, we have to take a decision to select countries that had reliable data and some of the countries with dated/unreliable data were taken out of scope with regards to the project.

Getting Started on Identifying Patterns and Magic Equation

Figure 1 Healthcare Indicators Cluster Plot

A sample of 34 countries was chosen based on geographic location, economy level, COVID-19 numbers, population, policy measurement as well as data availability. COVID-19 death and recovery rate variables are also calculated by dividing both deaths and recovered numbers with total confirmed cases number respectively. Thus, a dataset with healthcare-related features and 34 countries as the entries were completed. Finally, a K-means clustering model Figure 1, was built using the seven healthcare variables. The cluster plot shows seven groups of countries. One of the interesting insights that came out was that although most countries formed logical clusters there were few interesting exceptions as well. The United States for example came in another cluster of countries that had good infrastructure but not the best in class. This was a validation of the analysis because the US was overwhelmed with its medical infrastructure especially in the less developed parts of the country.

Figure 2 Healthcare spend and Critical Beds Plot

We also started analyzing deeply the healthcare spend and the medical infrastructure. Figure 2 demonstrates an interesting perspective where some of the advanced economies like the US, Italy, and Sweden are spending a higher % of GDP on their healthcare but have not invested enough on Critical beds. On the other hand countries like Japan, Germany, Hungary, Argentina, etc have a higher Critical bed ratio for their population. This has strategic implication when the caseloads increased and hospital infrastructure was overwhelmed,

This is further validated with Death and Recovery cases for selected countries with regards to health infrastructure consisting of Critical Bed. Countries with lower Hospital beds etc and higher infection rates had higher death rates and lower recovery rates.

Figure 3 Critical Beds and Recovery / Death rate Plot

The key outcome from Figure 2 and Figure 3 is that it is not just investment in Healthcare that matter but also these investments have to be targeted in critical areas such as Critical beds, ventilators, etc. There are major differences in how different countries with different economic and social parameters spend their resources on health and these have had a major impact during the COVID19 pandemic.

We also developed a heatmap between different economic and health parameters each column on the dataset where connections between each variable can be seen more clearly

Figure 4 Correlation Heat-map of Healthcare Indicators

Some of the insights:

• Health access quality index, a number of critical beds, and life expectancy are positively correlated with recovery rate. This means higher values of these indicators would likely result in a high recovery rate.
• The probability of dying from chronic diseases is negatively correlated with the recovery rate. Thus, countries with a lower probability of death might have a higher recovery rate.
• The percentage of the population aged over 65 is the only indicator which has a logically acceptable correlation with the death rate.

There is no single variable that has dominant effects on both COVID-19 recovery and death rate. In other words, each country might have different effects regarding the number of these variables to their COVID-19 recovery and death rate. Hence, some combinations of these variables are needed for deeper analysis. Finally, considering the clusters, policy intake, and measurement, as well as healthcare infrastructure varieties, eleven countries were chosen to represent the analysis.

Healthcare Indicators

Figure 5 Healthcare Indicators vs Recovery and Death Rate

The above visualization reflects the Death and recovery rates for various countries against parameters such as the Beds / 10000 population, Doctors per 10000 population, Probability of Dying, and HAQ Index. Some of the interesting insights from the visualizations are :

• Italy has a high ratio for Doctors per 10000 population, High HAQ index yet it is suffering from a high death rate that is because it has a higher ratio of vulnerable population over 65 and a bottleneck in terms of the number of beds once the infections increased.
• A similar scenario is playing out for the United States and the UK as well where in spite of having better Healthcare infrastructure in terms of doctors and HAQ the relatively high population over 65 and lower number of beds seem to have spiked up the death rate. Italy has 22.7% of its population over the age of 65 and in the United Kingdom, it is 18%.
• South Korea was able to lower the number of new infections with a high number of availability of critical beds and doctors per ten thousand population, a high health access quality (HAQ) index

From the success of South Korea, Japan, and Australia, we can see that a strong healthcare infrastructure plays a key role in combating COVID-19 from the country. Also on the negative side, a higher % of the population over 65 can result in lower recovery and higher death rates in spite of having high medical infrastructure. This also borne by the fact that developing countries with lesser healthcare infrastructure were able to resist the COVID onslaught due to lower under 65 population

Effect of mobility restrictions on Infection rates

Figure 6 Mobility and Lockdown Policy Case Scenarios

We also look at the effect of Mobility on daily cases. Italy’s stringent lockdown policy resulted in a fall in mobility by 52.8% and in the United Kingdom by 40.43%. A higher fall in mobility is a good indicator of the success of lockdown policies.

Some of the interesting observations are :

1. There is a lagging impact of mobility and new infection cases
2. The stringency with which mobility is implemented matters. In Italy, the lockdown was quite stringent in the second phase resulting in the decline of new cases. In the USA the lockdown implementation varied across states had a consistent increase of new cases
3. Australia which has the lowest population over 65 benefited the most from the lockdown.
4. The curtailment in mobility indicates lower access to healthcare institutions which directly affects the recovery and death rates.

Let's Look Real countries with Real Impact

1) India Showcase — Lockdown Policies Impact on Access to Healthcare

As the COVID-19 cases embarked in India, the government imposed stringent lockdown policies, which resulted in a steep decline in the mobility of the population by 33.9%, Figure 7. The data released by National Health Mission shows that with the stringent lockdown in place, a vast majority of the population might have missed potential life-saving medical treatment (like diabetes, hypertension, malaria, children immunization, maternal healthcare services, and emergencies) as the novel coronavirus pandemic spread and a lockdown came into force, thus causing a hindrance for the population to easily access healthcare facilities, Figure 7.

A strong positive correlation can be observed between the mobility changes and medical treatments and a negative correlation between the lockdown stringency index with the medical treatments. These numbers clearly show that people have simply not been able to easily access healthcare since the lockdown.

Figure 7 India Lockdown Policies and Access to Healthcare[37][38]

2) England and Wales Showcase

England and Wales Non-COVID19 Deaths and The Vulnerable Population

ONS (Office for National Statistics -UK) is one of the few organizations which has real-time data regarding healthcare for both COVID and Non-Covid. We leveraged the data from UK and Wales to get insights into COVID on both COVID as well as Non-Covid population who were vulnerable.

Figure 8 Analyzing Non-COVID19 Total Deaths in England and Wales

Significant fluctuations of England-Wales’s chronic diseases patients’(Non-COVID19) mortality records have been observed for over the past 5 months of 2020 and compared with an average mortality of the last 5 years as shown in Figure 8.

In fact, total Non-COVID deaths from Mar 20th to May 1st were 85,630, an increase of 18.7% on the same time interval of the average past 5 years.

Figure 9 Non-COVID19 Deaths by Age in England and Wales

The Office for National Statistics (ONS) data showed the majority of Non-COVID deaths victims were people aged over 65 years old as illustrated in Figure 9. This age class accounted for about 85.76% of total Non-COVID deaths occurred in England and Wales amid the COVID19 crisis, indicating that they may be a huge biased approach in handling patients and treatment amid this crisis.

Homes and Care Homes are No Longer Safe

Figure 10 Non-COVID19 Deaths by Place in England and Wales

The majority of Non-COVID patient deaths took place in hospitals (33.7%) followed by care homes(30.3), private houses(29.3%), and communal places(6.7%) respectively during the crisis period.

The analysis also shows an absolute upward trend in Non-COVID deaths at care homes and homes over the period. This indicates our hypothesis that deaths on vulnerable people treatments or lack of transportation availability due to the Government’s internal movement restrictions policy or medical triage.

Correlation of Non-COVID19 Mortality, Government Policies, and Public Mobility Change

In the amid of COVID19, Government policies such as Mobility restriction and healthcare infrastructure are effectively established to curb the spread of the diseases and helps severely affected COVID19 patients. But for the Non-COVID19 population, there is a decrease in appointment by -42.8% due to Lockdown policies.

Figure 12 Government Policies Impact on Hospitals’ Appointments in England and Wales[33][37][38][39]

Thus, in Figures 12 and 13, there is a linear associative correlation between Non-COVID19 patient’s deaths and Government stringency index & mobility index. The Non-COVID19 patient’s mortality rate spiked from Mar 27 as government policies and mobility restriction came into force. So this analysis showcases, there is an indirect impact on Non-COVID19 patients due to Government policies and mobility restrictions imposed for the COVID19 crisis.

Figure 13 Government Policies Impact on Non-COVID19 Total Deaths in England and Wales [37][38][39]

Key Findings of the analysis

• There are no particular variables that have unique dominant effects on COVID-19, as healthcare-related variables widely vary for each country. Age over 65 and hospital infrastructure have a stronger correlation on death and recovery rates. For example, Italy has been overwhelmed due to high infection rates among over 65 population and the hospital beds beyond a certain point could not take the patient load, Australia on the other hand had a lower population over 65 and it’s hospital infrastructure is was able to handle the load.
• Socio-Economic and Health parameters can’t be clustered to predict outcomes for a set of countries. One of the major findings is that each country has a different culture of enforcement, work practices, etc which would result in different outcomes although the input parameters might be the same. One of the prime examples is similar countries in the EU having different death/recovery rates although they can be governed by similar socio-economic policies.
• The number of Chronic diseases/Non-COVID19 patients’ mortality in England-Wales is high amid the COVID19 crisis as compared to average deaths in previous years. There is a direct correlation between canceled/missed appointments and higher death rates along with mobility restriction
• Population over 65 is suffering higher death rates in places other than hospitals due to mobility restrictions. This has implications for the future model of healthcare

What made our life difficult during the 8-week Challenge

• Data availability: Lack of recent health indicators amid COVID19.
• Data sparsity: Not recording data at regular time intervals.
• Bias: Lack of targeted data for vulnerable groups.

Appreciation where it’s due

This work would not have been possible without the help of all team members. Special thanks to the domain experts who provided the insights and challenged us to operate out of conventional thinking.

Rohet Sareen SRIDATT MORE Nikolaus Siauw Hunar Batra Mohammed Ba Salem Kushal Vala AjaykumarPalaniswamy