Skepticism about measurement: Hospital beds edition

by Shubho Roy.

The Covid 19 pandemic has motivated many studies based on data about the disease and the response. However, measurement in India is often weak. There is a need for greater caution before using such data. In this article, we look closely at one such issue: estimating health care capacity based on hospital beds.

The health care infrastructure response to Covid 19 has been to ramp up the number of hospital beds. However, severe and critical care patients in India may often need ventilators and ICU beds. There are no measures available about the number of ventilators or ICU beds in India. Researchers have taken to guesswork in order to address this gap. How reasonable are these estimates?

Extrapolation from the count of hospital beds

It makes sense to use hospital beds as a standardised measure of hospital capacity. Infrastructure, equipment and manpower standards for hospitals have been built on a per bed basis. The number of nurses, doctors, equipment, and even floor space is a function of the number of beds in the hospital. For example, according to the Indian government, a hospital should ideally have 80 to 85 sq m of plinth area per bed; there should be a toilet for every six beds; and one operation theatre for every 50 beds in the general ward. This makes estimating the availability of health facilities easier. Count the number of hospital beds in a country, and you have a sense of the overall health care capacity. Both the World Health Organization and the World Bank track the number of hospital beds per 1000 population as a measure of health care capacity.

Estimates

If we work within such quantification of hospital capacity based on the number of beds, how many ventilators and ICU beds might be present in India? The central government formulated the IPH Standards in 2012, to improve capacity in government health services (run mostly by the state governments). The standards for district-hospitals (at pg.5) requires 300 district-hospital beds per million population. IPH standards for district-hospitals states that five to ten per cent of the total beds in a district-hospital should be ICU beds (See page 25) and each ICU bed should have a ventilator (amongst other equipment). If the entire country were to be up to IPHS standards, there should be 416,189 district hospital beds and between 41,618 to 20,809 ventilators in the country (in the government system).

In reality, the numbers will probably be lower than the standards. Experts have tried to estimate the availability of hospital beds, ICUs and ventilators for the present epidemic. Rajagopalan and Choutagunta have estimated the availability of hospital beds (both in government and private sector) in various Indian states. Singh et al. use a 2008 paper by Yeolekar and Mehta, which estimates that there are around 5-8% ICU beds in government hospitals. Singh et al. assume that around 50% of the ICU beds have a ventilator. This gives them a range of 35,699 to 57,119 ventilators for the entire country (in the government system). Similarly, Kapoor et al. estimate that there should be around 35,699 ICU beds and 17,850 ventilators for the country.

For U.P., they estimate 3,813 ICU beds and 1,907 ventilators.

The ground reality in Uttar Pradesh

How do the government aspirations and expert estimates stack up against the ground reality? In 2019, India’s supreme audit institution (the CAG), carried out a performance audit of Hospital Management in Uttar Pradesh. U.P. has 75 districts and 174 district hospitals (in 2018). The CAG covered seven districts (out of 75) for its audit of 16 government hospitals. In addition, three out of 11 district-hospitals of Lucknow were audited (See Table 45 at pg. 93). The seven districts which were fully covered by the audit (for district-hospitals) are distributed across the five administrative regions of U.P. The CAG found that the reality was far away from the aspiration or estimates.

The seven districts have a population of 25.9 million. As per IPH standards, they should have 7,700 district-hospital beds. The CAG found 2,275 beds, a shortfall of over 70%. If the districts follow IPH standards, there should be 385 to 770 ICU beds for the ideal 7,700 district-hospital beds. Even if IPH standards were maintained on a base of 2,275 beds, there should have been 113 to 228 ICU beds in the sampled district hospitals. The CAG found that ` 10 out of the 11 district hospitals had no ICU beds’. Only the Gorakhpur district hospital had 13 ICU beds (3% of its total beds). In short, of the seven districts examined by the CAG, there were 13 ICU beds, all located in one district.

Even where there were ICU beds, the CAG found shortages of equipment considered essential for an ICU bed (as per IPH standards). The CAG noted:

“audit observed that only six High-end Monitors were available against the requirement of 14, seven Infusion pumps were available against the requirement of 14, while Ventilators, Ultrasound for invasive procedures and Arterial Blood Gas (ABG) analysis machine were not available at all in D.H. Lucknow. Similarly, in D.H. Gorakhpur, there were no Ventilators, Infusion Pumps, Ultrasound for invasive procedures and ABG analysis machine.”

(pg. 34 of the CAG Report)

In the seven districts where the CAG audited district hospitals, the CAG found no ventilators. Even in the district-hospitals in Lucknow (outside the seven districts), 2% of the beds were ICU, and there were no ventilators.

The CAG report only covers seven districts out of U.P.’s 75 districts. U.P. is one of India’s poorest states. The government of U.P. also operates some super-speciality hospitals where facilities might be better, but they will be few. These seven districts constitute 11% of U.P.’s population. While the findings may not be representative of India, they are not inconsequential. The gap between central government standards, expert estimations, and reality is vast. Table 1 shows the gap between these numbers for the 11 districts of U.P.

Target and reality gap between hospital beds, ICUs and ventilators

Measurement approach	Hospital Beds	ICU Beds	Ventilators
IPH standards	7,700	385 - 770	385 - 770
Expert estimation (from UP numbers)	8,339	419	210
If existing beds maintained ICU ratio	2,275	113 - 228	113 - 228
CAG findings (Reality)	2,275	13	0

Goodhart’s law

Why is there such a large discrepancy between the IPH standards, expert estimates and the reality observed by the CAG? We may conjecture that Goodhart’s Law is at work. Goodhart’s law states: “When a measure becomes a target, it ceases to be a good measure”.

For too long, the academic and policy literature, in India, has emphasised one metric: the number of beds available in government hospitals. Press articles regularly criticize the government for India’s low bed to population ratio (See here, here and here). An easy way out for politicians and officials is to look good in such measurement, while skimping on other elements of health care. Between 2014-2018 the Central Government spent Rupees 8.5 billion for the country out of which Rupees 1.5 billion was spent in U.P., under the National Health Mission, to upgrade facilities in state government hospitals. The result was a rapid expansion in the number of beds (the measure which health policy makers are sensitive to), and not much else.

This problem is not limited to hospital beds. It extends to other parts of the health sector. The central government operates a detailed database called the Health Management Information System. As an example, Smriti Sharma shows that there are significant discrepancies in the database. Numbers which portray the system in poor light are under-reported while the numbers which show the health system positively are inflated.

Using the hospital bed measure to estimate the availability of health care capacity is misleading. Even when the government sets up new facilities, measurement is being done on the basis of new beds. Till April 11, the government had set aside 100,000 hospital beds and 11,500 ICU beds in 586 hospitals. On May 15, the Maharashtra government planned to set aside another 100,000 beds, just in Mumbai with an additional 1,000 ICU beds. In thinking about the situation in health care, this is not enough information. We need to know the facilities and personnel that will be available for these beds.

A general philosophy in India is to be very careful about using data. Researchers need to gain confidence in the quality of the measurement process. This is particularly critical where the agency which performs a function is also the source of data about the same function. When the underlying data is weak, no amount of cleverness in statistics can rescue the distortion of our view of what is going on.

References

Assessing Healthcare Capacity in India. Shruti Rajagopalan and Abishek Choutagunta, Mercatus Working Paper, Mercatus Center at George Mason University, Arlington, VA, April 2020.

Covid 19 in India: State-wise estimates of current hospital beds, intensive care unit (ICU) beds and ventilators, Geetanjali Kapoor, Aditi Sriram, Jyoti Joshi, Arindam Nandi, and Ramanan Laxminarayan, Center for Disease Dynamics, Economics and Policy, Princeton University April 2020.

Covid 19 | Is India’s health infrastructure equipped to handle an epidemic?, Prachi Singh, Shamika Ravi and Sikim Chakraborty, Up Front, Brookings, March 24, 2020.

Covid-19 in India in the coming months: The puzzles faced by leaders of health care organisations, Ajay Shah, The Leap Blog, June 2020.

Hospital Management in Uttar Pradesh, Comptroller and Auditor General of India, 2019.

Problems of the Health Management Information System (HMIS): the experience of Haryana., Smriti Sharma, The Leap Blog, June 2016.

Prudent public health intervention strategies to control the coronavirus disease 2019 transmission in India: A mathematical model-based approach., Sandip Mandal, Tarun Bhatnagar, Nimalan Arinaminpathy, and Anup Agarwal Indian Journal of Medical Research. 2020 10.4103/ijmr.IJMR_504_20.

 

The author is a researcher at the University of Chicago and would like to thank Renuka Sane and Rajeswari Sengupta for their valuable inputs.

Covid-19 in India in the coming months: The puzzles faced by leaders of health care organisations

by Ajay Shah.

Peering into the next six months

How might the pandemic play out in India in coming months? There are newspaper reports about some important statistical evidence from ICMR about the spread of Covid-19 in 70 districts of India (caveat). Based on antibody testing, it appears that about a third of the people in containment zones in some large cities had antibodies in late April. We can cautiously expect significant progress towards herd immunity, in containment zones, by today, i.e. mid-June. Recent stories from Dharavi in Bombay are consistent with such an argument (while also being a testimony to the public health capability of the municipal authorities).

In most of India, however, the picture is quite different. E.g. while about a third of the people in the containment zones in Bombay had antibodies in end-April, the fraction of persons in Bombay as a whole who had antibodies is small. In most of India, the bulk of the epidemic lies in the future.

There is a public health problem (how to slow down the spread of the disease) and there is a health care problem (how to care for the people who get sick). In this article, we focus on the health care problem. For the leadership of health care organisations in most of India, this is an extremely important moment, when they need to plan for this coming surge. In this article, we think about the pandemic from their point of view. A given facility might appear to be relatively unruffled today, but it is important to envision the coming surge of demand for health care, and to lay the groundwork for faring better at the peak of the pandemic in the catchment of the facility. What are the issues, and what are the potential actions that can be taken?

About 70 per cent of health care in India is in the private sector. In this article, we place ourselves in the shoes of the leaders of health care organisations of all kinds, but we have an accent on private organisations as this is where the bulk of the action will lie.

Equipment

Oxygen therapy is a key element. There has been a lot of talk, internationally, about ventilators. However, from the viewpoint of both efficacy and cost, ventilators are a poor solution. The skill required of health care workers, to use a ventilator, is substantial and this will limit scale up. It is more useful to develop a strategy that involves oxygen cylinders and oxygen concentrators. The former is associated with the problem of managing the supply chain for oxygen cylinders. All hospital beds should be equipped with oxygen ports.

In many hospitals, there is a need to introduce physical isolation and establish a dedicated wing in which Covid-19 patients will be treated.

Beds in the ICU are a scarce resource. It is useful to establish `step down beds', where patients exiting the ICU can be safely placed, when they require a high standard of care but no longer require to be in the ICU. This will improve the extent to which the ICU is available to the patients who need it most.

Internal management

The medical community in each city needs to debate and agree on the clinical protocols that will be put into play, that are feasible and cost-effective under their local conditions. This will reduce fumbling and recrimination in the surge. Conversations and documents around rules of triaging will help.

Economies of scale and cost reductions can be obtained by establishing `eICUs', where a central command centre has skilled staff which monitors the data coming in from a remote ICU. This is a more feasible path to scaling up ICU capacity, particularly in places when the skilled staff in ICUs is hard to find.

When the surge comes, the management processes of the hospital will be tested. Every element of the process requires analysis from the viewpoint of coping with a surge environment. Enhancing non-medical staff and processes, ahead of time, will help cope with the surge.

Health care workers

A key problem concerns health care workers (HCWs), who face the risk of high dose exposure to the virus. While some HCW are driven to serve the community, many may retreat from work when the surge gathers momentum. At precisely the time when the most capacity is required, the capacity could degrade, thus increasing the chances of an organisational rout.

The leadership needs to undertake many measures which will be fair to HCW and reinforce their commitment to hold the ranks:

1. It is penny wise, pound foolish, to skimp on the quality and quantity of PPE. If a few ward boys get sick, word of this will leak to other ward boys. For ward boys to feel safe, their training and consumables have to be of high quality.
2. HCW and their families need to be reassured that there will be ample effort on giving them treatment if required.
3. PPE and training is required not just in the ICU but also for the primary care providers, who are the first point of contact for patients when they reach the facility.
4. Periodic antibody testing for all HCW will be particularly useful: (a) In assessing the extent to which infection and immunity has come about, (b) Generate metrics of the class of situations where new infections are coming about and feed back to process improvements, and (c) Increase the confidence of HCW as a stream of process improvements are visible, and when it is seen that the infection rate and severity of the disease is low.

Community initiatives

These elements (equipment, management, HCW) constitute a reasonable work plan to gear up for the surge. But many or most hospitals today are beset with difficulties. In the best of time, their management bandwidth was limited. Covid-19 has induced a financial crisis with a decline in non-Covid revenues, and the Indian financial system is not able to engage effectively with most hospitals. The thin capabilities have been adversely affected by the retreat of HCW. The puzzle lies in finding the energy and resources to actually pull off a significant amount of preparatory work.

There are many problems which are hard to address at the level of one hospital. Consider a city like Nagpur. There is significant value in constructing a Coalition of hospitals and of the local business community, which can work towards many initiatives -- without any government involvement -- which will reduce the damage caused to the city from the epidemic. Examples of such collaborative initiatives are :

1. Nagpur requires facts, through random sampling, about the state of infection and antibodies in Nagpur. The weekly or monthly construction of these facts is vital for health care organisations to know the planning horizon that they face, before the surge. The citizenry requires these facts to make decisions about the economic and social activities that are safe. The Indian state does not produce this information. Better planning by health care organisations is good for them and for the citizenry and economy of Nagpur. It would be valuable if such a work program can be put together by the Coalition.
2. The Coalition can collaborate with the medical testing industry to establish capacity, and negotiate bulk rates.
3. The Coalition can establish a process of discussion and drafting of appropriate clinical protocols which can then by used by all HCW in the city.
4. When an individual requires health care or a bed, there is chaos during the surge, with patients running around across multiple facilities looking for spare capacity. The Coalition should establish a shared information system and call centre for patients to use. This will reduce the operational overheads and queues outside facilities. This will increase bed utilisation and improve the allocation of facilities based on the condition of the patient.
5. The Coalition can pool resources to do bulk buying and inventory management on medical supplies such as PPE or oxygen cylinders, and dynamically respond to the shortages of consumables that are discovered at future dates.
6. There has been significant friction between health care organisations and the government. The Coalition could be more effective in addressing inappropriate behaviour of various arms of the Indian state. The Coalition is a natural locus for addressing bad behaviour by some health care actors, and can head off such problems so as reduce the probability of the state getting involved.
7. The Coalition can be more effective in overcoming the frictions faced by hospitals in empanelment with the various government sponsored health insurance schemes (GSHIS) and address frictions associated with dealing with health insurance companies.
8. Some health care organisations may falter in their commitment to stay in this fight. The Coalition will be valuable in exerting peer pressure, and in helping transmit management knowledge to some organisations who are sitting on the fence.

To the extent that the health care problem in a city like Nagpur is worked out well, individuals will feel more safe, and will get back to working and consuming, thus bolstering the economy. If the health care system gets crushed, there will be greater reticence on the part of the citizenry to spend or work, and the economy will be more adversely affected. Supply chains will get disrupted if there is a lockdown in the future in Nagpur. There is thus ample self interest which should drive the business community, and the health care community, to come together, and expend financial and management resources on building such a Coalition.

There are severe financial problems in many health care organisations today, as the traditional revenue stream has dried up as a consequence of fearful households. The leadership of many health care organisations is firefighting a financial crisis, which is exerting a tax upon their management bandwidth, at a time when they should primarily be working on laying the groundwork for the surge. The Indian financial system works poorly and is not able to perform its role, of efficiently supplying capital. There is a need for owners to bring in equity capital to alleviate this problem. It is also in the self interest of business interests in a given city, to offer loans to health care organisations, so as to diminish the organisational rout of health care organisations in the surge, which would adversely impact upon the economy of the city.

Conclusion

The bulk of the discussions surrounding Covid-19 in India are focused on public policy. But state capacity in India is low, and we should have low expectations for what the state can do. As the de-lockdown progresses, the pandemic will accelerate. We are now at the last barricade: Health care.

How things work out in 2020 will now be shaped by the sagacity and leadership qualities of the senior managers of health care organisations across the country. There are about 10,000 important hospitals in India, and about 50,000 key persons who make up the leadership of these hospitals. All eyes are on the actions of these 50,000 people, which will have an impact upon millions of lives. Seldom has so much depended on so few.

In this article, we have shown some areas of planning and preparedness that are required in health care organisations. All large hospitals in the country, private or public, need to plan for the surge. A key theme we have emphasised is negotiation and collaboration between private persons. The Indian state is generally not able to usefully intermediate in the interactions between private persons. A key feature of the way forward lies in organising communities, in privately negotiated local solutions. It is in the best interests of the citizenry and the health care community of (say) Nagpur to take their future in their own hands, to plan their best way forward.

Information about COVID-19 in India

By Natasha Agarwal and Harleen Kaur.

The presence of timely and reliable data enables informed decision-making by government organisations and individuals. When a machine-readable dataset is released on a website, it is non-rival, and thus has characteristics of a public good. There is a case for state financing or production of information. As Carl Malamund says, "Government information is a form of infrastructure, no less important to our modern life than our roads, electrical grid or water systems". Open Data Governance (ODG) are structured datasets produced by government institutions that are released in a machine-readable format. These datasets contain information such as statistics, plans, maps, environmental data, spatial data, materials of agencies, ministries, parliamentary data, budgetary data, and laws.

Governments across the globe have been actively opening their data through national and regional data transparency portals recognising the need for making data available to the public. The process is informed by ODG principles. There are three main reasons for opening government data; increasing transparency, releasing the social and commercial value of the data, and to encourage participatory governance (Attard et al. (2015)). As an example, the COVID-19 pandemic is best controlled through behavioral changes by each individual. To support such changes, the governments need to open their data about the pandemic at an individual and community level.

The ODG principles defining best practices of data sharing include; i) identifying and publishing high-value datasets in a standardised format (such as a directory of medical professionals, tests conducted and results and information about surveillance), ii) adopting open data scheme protocol to share human and machine-readable, non-proprietary format and include universal resource identifier and linked data to provide access, iii) removing barriers to data access such as requirements of establishing an account, of proving identity, or payments for data access, and iv) making information available in perpetuity by not deleting/changing data permanently.

In this article, we examine the information systems on COVID-19 in India from the viewpoint of these issues in the design of a high performance statistical system.

Data.gov.in and its limitations

In India, an open data policy the National Data Sharing and Accessibility Policy (NDSAP) was announced in 2012 to open government data to the public by following ODG principles.

The policy requires all ministries, departments, subordinate bodies, organisations, and autonomous bodies of the Indian Government to share all publicly generated non-sensitive data in both human-readable and machine-readable formats. The data is disseminated through a common government data platform deployed and managed by the National Informatics Centre (NIC), Ministry of Communications and Information Technology. It mandated that datasets be periodically updated by government agencies along with comprehensive meta-data which enables data discovery and access through departmental portals.

Furthermore, NDSAP requires the Department of Information Technology (DIT) to publish guidelines to implement NDSAP. The implementation guidelines provide details of the data contribution process including; the role and responsibilities of the data controller, approval, publishing process for catalogs and resources, and management of published datasets.

In compliance with NDSAP, India's national data transparency website, data.gov.in was launched in 2012. Accordingly, data.gov.in provides a unified catalog of datasets allowing users to browse the dataset catalog, view the meta-data associated with each dataset, comment on and rank various datasets, download available datasets, submit suggestions and queries on the published dataset, and submit a request for those that are not available yet (Chattapadhyay (2013)).

Despite the comprehensiveness of the policy and the accompanying guidelines, agencies have responded predictably, i.e. they neither comply with NDSAP nor with the implementation guidelines. As a result, data.gov.in contains issues such as the absence of databases, duplicate datasets, lack of follow-up, or meta-data (Agarwal (2016) and Buteau et al. (2015)). The terms 'policy document' and 'guidelines' which are often used in India are ineffective in that they do not constrain the executive. Hence, these documents amount to exhortations that have little impact on the incentives of officials in favour of greater opacity, reduced work, or gaining power through the control of data.

Ministry of Health and Family Welfare (MoHFW) and COVID-19 data

We examine the data in the public domain emanating from MoHFW during the ongoing COVID-19 pandemic. To understand the availability of resources for healthcare, we searched for a directory of healthcare providers (both institutions and individuals). The latest hospital directory available on data.gov.in was for 2016 and the latest data for the number of registered allopathic doctors and dental surgeons was available for the year 2013.

The MoHFW is disseminating limited data on the spread of COVID-19 through the data.gov.in portal. For example, as of 1st June 2020, the data reported under mygov.in (not in data.gov.in) contains information on three variables namely (i) total number of persons infected with COVID-19; (ii) COVID-19 infected persons who have been cured/discharged/migrated; and (iii) COVID-19 infected persons who have died. The state-wise distribution of these three variables is available for a given date "T = Today". This data cannot be downloaded. The meta-data for this information is also not available. On the other hand, the data.gov.in only releases daily factsheets in a pdf format summarising this data.

The dissemination of COVID-19-related data by the MoHFW has problems. It gathers detailed COVID-19-related data from the National Centre for Disease Control (NCDC) (surveillance data from the field) and Indian Council of Medical Research (ICMR) (data through the testing laboratory network), which is not reflected in data.gov.in.

The NCDC, under the Integrated Disease Surveillance Project (IDSP), consists of union, state, and district-level units responsible for the surveillance of infectious diseases in India. Although it releases weekly outbreak reports notifying the status of infectious diseases in India, the reports are available only on its website and not integrated on data.gov.in. On the COVID-19 pandemic, the weekly outbreak report dated 10th-16 February, 2020 was the latest available report under IDSP as of 8 June, 2020.

Similarly, ICMR, the designated body under the National Disaster Management Act to coordinate the testing strategy for COVID-19 has been releasing its data through its website and not through data.gov.in. Through its website, ICMR releases information on two parameters, the total number of samples tested for COVID-19 over time, and in the last 24 hours.

Therefore, data.gov.in is not being utilised by the union government agencies for releasing information. Individuals and researchers interested in the government data on the pandemic have to access information available in different silos according to their skills and knowledge. Moreover, none of the information shared is available in a machine-readable or standardised format. This leads to a weak information base on Covid-19 available to the public and to researchers, which hampers the decision making of individuals on the appropriate care that they should take, and hampers policymaking by government organisations for want of data and research.

Data disseminated by state governments

The union agencies are not the only government source on COVID-19 information. We now study the data dissemination protocols for COVID-19 as followed by the states.

We could not find state data on COVID-19 on the data.gov.in website. As a result, the following information was collected through individual COVID-19 portals set up by the states. Table 1 shows that there is heterogeneity in reporting across states. The information shared by the states is classified into three categories; "state-level", "district-level" and "individual-level".

Table 1: State-level reporting parameters for COVID-19 (As of 9 June, 2020)

	Parameters	Delhi	Kerala	Maharashtra	Gujarat	Karnataka	Madhya Pradesh
State-level data	Total COVID-19 confirmed cases	Y	Y	Y	Y	Y	Y
	Active cases	Y	Y	Y	Y	Y	Y
	Total COVID-19 tests conducted	N	Y	N	Y	Y	N
	Hospitalisation status of positive cases	Y	Y	N	N	Only ICU patients	N
	Isolated/ quarantined patients	Y	Y	N	Y	Y	N
	Total recovered patients	Y	Y	Y	Y	Y	Y
	Total deaths	Y	Y	Y	Y	Y	Y
District-level data	Number of people under observation	N	Y	N	N	Y	N
	Number of quarantined/ isolated people	N	Y	N	N	Y	N
Individual-level data	Age	N	Y	N	N	N	N
	Gender	N	N	N	N	Y	N
	Comorbidity	N	Y	N	N	N	N

Table 1, placed above, shows the data sharing protocol for COVID-19 in selected states. We may point out a few facts that influence the interpretation of this table:

1. Data as of 10th June, 2020. Sources: Delhi, Kerala, Maharashtra, Gujarat, Karnataka and Madhya Pradesh.

2. Maharashtra, Gujarat, and Karnataka share information about the same parameters at the State and District level. The information depicted here is about parameters in addition to the duplicate information.

3. In the studied states, Gujarat and Delhi inform about the number of patients on ventilators at the state level. However, the information on available hospital beds and ventilators in Delhi is shared under a separate website, https://coronabeds.jantasamvad.org/.

4. District-level information in Kerala is available for patients hospitalised, symptomatic patients hospitalised, the chronology of positive cases, and hotspots. No other states releases data on these parameters.

5. Karnataka is the only state which shared anonymised patient data related to their travel history, district, and location of isolation. It also has a dedicated patient case number for individual patients for whom information is shared.

6. Madhya Pradesh had a dedicated website for individual-level data which was discontinued from 11th May 2020 onwards following the raising of privacy concerns over social media.

We find that in most states, the baseline data includes overall state data about testing rates, persons infected, deaths, and recovery data. However, some states provide additional information such as the number of COVID-19 tests conducted, the number of isolated/quarantined persons, the counts of patients on ventilators, and stable patients. While some states like Maharashtra report data at the district level along with the overall state data, others like Karnataka share information at the individual level. There is a high variation in the type of data shared by the states. For instance, at an individual level, Karnataka reports anonymised information about the demographic details in addition to the baseline data. On the other hand, Madhya Pradesh used to share the name and addresses of the suspected COVID-19 patients to the public while reporting individual-level data. Similarly, Kerala, Maharashtra, and Gujarat report their data at the district level. Kerala reports its surveillance data which is not reported by Maharashtra, and Gujarat. Some states provide daily reports in English, while others do not. For example, Gujarat provides daily reports only in Gujarati.

Most states disseminate data through their COVID-19 websites. However, some resort to reporting through social media. For example, the Maharashtra government website on COVID-19 does not provide information other than that reported in table 1. However, the Maharashtra government has been releasing daily reports providing COVID-19-related information across age, gender, comorbidities amongst other variables through Twitter. While twitter can amplify the transmission of information in a public statistical system, it should not supplant the foundational systems. Data disseminated through a tweet cannot be traced to any government website. Besides, there is inconsistency in the reports shared by the Maharashtra government through twitter. For example, the report dated 22nd April 2020 provides for district-wise distribution of COVID-19 cases in Maharashtra which is not available in the report dated 1st April 2020. The data is a "delete-tweet" away from not being available.

There is also variation in the data sharing format. Most state governments provide data in human-readable formats like pdf. However, some state governments provide some data in machine-readable formats. For example, district-wise data on variables available on dashboard for Gujarat which contains the total number of cases tested for COVID-19, positive cases, patients recovered, people under quarantine, and total deaths can be exported to a csv document. Nevertheless, demographic details of COVID-19 patients or data patients on ventilator/stable, are only available in daily reports in pdf format.

We find that the states do not share their COVID-19 data through the data.gov.in framework. Users have to look for multiple information sources about COVID-19 data to access this data. Within the framework of stand-alone websites providing information, there are two concerns. The first concern is the lack of standardised parameters for information releasing. For instance, few states share the hospitalisation status and the availability of beds which would be useful for the general public in case of emergency. The second concern is the quality of data shared by the states. As discussed, most states share human-readable data and not machine-readable, downloadable data. Meta-data is not available for any state studied making it difficult to interpret. Moreover, the lack of data standardisation makes data non-interoperable. The state-level historical information is unavailable for most states. Therefore, not all data shared by the states is permanent.

Difficulties of CoVID-19 data release seen elsewhere in the world

So far, we have documented variation in what data is being released, and how the same is disseminated, in India. This is a global concern for COVID-19. We map the data reported by selected countries in table 2 below. We find that countries are using two forms of data distribution methods. These are daily updates and dashboards. While daily updates are usually pdf documents, dashboards provide progress of COVID-19 over time. The type of information shared by countries can broadly be classified according to the level of data as "country-level" and "individual-level". Country-level data consists of aggregate information such as the total number of tests conducted, the total number of COVID-19 positive patients, the number of patient hospitalised and deaths, etc. Some countries also share aggregate surveillance data which consists of information about individuals isolated, quarantined, and contact traced. At an individual level, we see a wide variation of data shared by the countries. While India does not provide individual-level data through its Ministry of Health, other countries share demographic information such as age, gender, race/ethnicity, and occupation. A comparison of data disclosed by selected countries is shared in table 2.

Table 2: Country-level data parameters for COVID-19 (As of 10 May, 2020)

Country	Daily updates (DU) or Dashboard (DB)	Total Number of tests conducted	Total Number of COVID-19 +ve patients	Total Number of patients hospitalised	Total Number of deaths	Surveillance data	Individual level data
							Age	Gender	Race/ Ethnicity	Occupa-tion
India	DU and DB	Y	Y	N	Y	N	N	N	N	N
USA	DU and DB	Y	Y	N	Y	Y	Y	N	Y	N
UK	DU and DB	Y	Y	N	Y	N	Y	Y	Y	Y
South Korea	DU and DB	Y	Y	N	Y	Y	Y	N	N	N
Singapore	DU and DB	Y	Y	Y	Y	Y	N	N	N	N
Canada	DB	Y	Y	Y	Y	Y	Y	Y	N	N
Australia	DU and DB	Y	Y	Y	Y	Y	Y	Y	N	N

It can be seen from the above table that most countries report testing data (information about the number of tests conducted), and the number of positive cases and deaths. At the national level, India only reports these minimum consistent variables. Some countries report more variables to the public. For instance, the US, South Korea, Singapore, Canada, and Australia report surveillance data in varying details. A few countries like Canada share their database in a downloadable format. This includes information about quarantined and isolated individuals and details about contact tracing and source of infection. Singapore, Canada, and Australia also report data on the number of cases hospitalised. The UK has recently started reporting information about COVID-19 deaths, disaggregated into deaths inside and outside hospitals. Individual-level data such as age, gender, race/ethnicity, and occupation, is visible in some countries, as is the case in some states (though not the union government) in India. The US releases data about age and race, while the UK releases information about age, gender, race, and occupation. South Korea releases age details for only severe cases and Singapore releases individual-level data only in the event of the death of the individual. Canada releases data about age and pre-existing conditions of the individuals and Australia releases information about age and gender.

Therefore, we find that data release for COVID-19 has issues of lack of standardisation and inter-operability globally. In India, the union and state governments have important deficiencies.

Implications for India

India's existing data infrastructure does not meet the demands of a public health emergency. The implications of this are multifaceted. For example, amid the COVID-19 pandemic, the government had to create a Covid19-warriors dashboard that provides information on doctors, nurses, ASHA workers, and others who could be deployed for immediate response. If data.gov.in had worked well, then the government would have had this information already.

Likewise, the problem of inaccurate databases highlighting data discrepancies in reporting COVID-19 infected persons could have been avoided. An available database infrastructure in data.gov.in would have avoided the need for ICMR to evolve its own data-dissemination method in the middle of the COVID19 pandemic. Besides, the problem of collecting, processing, and releasing COVID-19 data with other databases would have been eased. For example, if the existing data infrastructure had data collection and reporting standards across space like district names with their respective codes, then it would not only be easy to collect the data but also facilitate easier collation with other datasets for enabling interoperability.

Conclusion

In the present article, we highlighted one element of the public health response, the issue of data release by the Indian government authorities for COVID-19. We show that the statistical system for disease surveillance dissemination in India is in a need of reform.

The ODG platform in India, data.gov.in, can play an important role in strengthening India's public health data infrastructure. To realise the utility of public data, a data protocol framework with a legally enforceable mandate on the government is required, as is seen in countries like the US. The principles of standardising, anonymising, interoperability, meta-data release, and grievance redressal in the event of non-release should be in this legal framework.

For the union government, a data.gov.in which utilises the sound principles of OGD release could become a better foundation for data release, and thus improve India's response to an epidemic. State and city governments could choose to use the services of data.gov.in or build their own systems. An indicative list of the essential components of such a portal (as seen in NDSAP and ODG principles) are provided below:

1. Standardising data release: Standardisation of reported variables such as reporting unit, disease data, language, individual, and community-level data is required. Elements that go towards this include geotagging and coding of hospitals/labs and the adoption of International Classification of Diseases (ICD) for diagnosis and treatment of diseases.

2. Ensuring privacy: Privacy is a fundamental right in India (Supreme Court of India (2017)). Despite this, states like Madhya Pradesh and Karnataka were seen to be disseminating personally identifiable information of suspected COVID-19 patients. The government would need to adopt various tools at its disposal to protect these rights at an individual and community level. These tools include tagging appropriate data, incorporating principles of Privacy by design (PBD), anonymising and utilising appropriate fiduciary principles (Cavoukian (2011) and Bailey and Goyal (2019)).

3. Interoperability: Facilitating systems interoperability by incorporating common formats, software standards, and semantic interoperability by incorporating e-governance standards so that the meaning of data is not lost across data silos is required (Wright et al. (2010)).

4. Adopting an open data scheme: Legislators need to create the frameworks through which the executive is required to release meta-data, and release data in a machine-readable format.

5. Setting up governance framework: Union, state, and city governments have legitimate authority on how they organise their work, but greater consistency and predictability for API-based access is desirable.

References

Attard et al. (2015): Judie Attard, Fabrizio Orlandi, Simon Scerri, and Sören Auer, A systematic review of open government data initiatives, Government Information Quarterly, 2015.

Chattapadhyay (2013): Sumandro Chattapadhyay, Towards an Expanded and Integrated Open Government Data Agenda for India, IDRC Digital Library.

Agarwal (2016): Natasha Agarwal, Open Government Data: An Answer to India's Growth Logjam, SSRN, 16 August, 2016.

Buteau et al. (2015): Sharon Buteau, Aurelie Larquemin and Jyoti Prasad Mukhopadhyay, Open data and applied socio-economic research in india: An overview, IFMR Working Paper, 27 May, 2015.

Supreme Court of India (2017): Justice K.S. Puttaswamy v. Union of India, 2017 (10) SCC 1.

Cavoukian (2011): Ann Cavoukian, Privacy by design: The seven foundational principles, Information and Privacy Commissioner of Ontario, 2011.

Wright et al. (2010): Glover Wright, Pranesh Prakash Sunil Abraham, Nishant Shah, Open government data study: India, The Centre for Internet and Society, 2010.

Bailey and Goyal (2019): Rishab Bailey and Trishee Goyal, Fiduciary relationships as a means to protect privacy: Examining the use of the fiduciary concept in the Draft Personal Data Protection Bill, 2018, Data Governance Network, 2019.

 

Natasha Agarwal is an independent research economist. Harleen Kaur is a researcher at NIPFP. The authors are thankful to Ajay Shah and two anonymous referees for their valuable comments and inputs on the article.