As the director of a survey research consultancy based in Cape Town, who puts a lot of faith in the authority and persuasiveness of accurate and well-sourced data, I was relieved and excited when it became clear that the government is taking a science- and data-driven approach to tackling the coronavirus crisis. It made the strict lockdown regulations look purposeful and the lesser evil compared to the alternatives.
Unfortunately, over the past few days I have grown worried that the government’s positive intentions and initiatives in this regard are about to hit a reality wall, costing us precious time that was bought with the lockdown while effective and affordable options are falling by the wayside.
The problem with current screenings
On Easter Monday Professor Salim Abdool Karim, the chairperson of Health Minister Zweli Mkhize’s Covid-19 advisory group, gave a detailed presentation about the scientific rationale guiding the government during this crisis. The projections he made are sound and reasonable, but the truth is that we still know too little about the virus and how it spreads to make any firm predictions.
I do not worry about the quality of the work done by the scientists but I am concerned about the reliance on modelling itself that seems to inform much decision-making these days as we simply do not know enough about this virus and local contexts vary too much to make precise predictions at this point. There is no doubt that scientific modelling plays a key role in understanding, predicting and controlling the epidemic, but until we have enough data to feed the models, and they are starting to converge, putting our fate into the hands of scientific models alone is a gamble we cannot afford to take.
There is an alternative: the collection of data to measure the infection rate. While the latest number of infections is publicly updated on a daily basis, we do not know the real figure because of the way we are testing. With the start of the lockdown there have been repeated announcements by the government that it will seriously ramp up screening and testing. One of the most tangible signs of these efforts are the community screening and testing campaigns carried out in selected neighbourhoods.
My neighbourhood, Bo-Kaap in Cape Town, happens to have been one of the first places selected for community screening, and it is my participation in this exercise that sparked my curiosity and subsequent nervousness about the government’s data strategy. In my neighbourhood, screening meant that groups of fieldworkers were going from house to house and asking exactly six questions (name, age, four symptoms) for everyone who lived there. Those referred for testing were given a slip with a time slot and an instruction to make their way to the local testing site.
There are many problems with this approach, first of which is that it will take too long – it took a week to complete screening in the Bo-Kaap, including a four-day break over Easter.
But what really filled me with trepidation, as I dutifully spelt my strange-sounding name to the friendly and professional fieldworker, was the sight of the paper sheets she was using for recording our responses. I might be a bit of a snob when it comes to face-to-face data collection methods, but there really is no reason why in this day and age we are still seeing fieldworkers equipped with clipboards and piles of papers instead of tablets or smartphones.
Apparently, the CSIR is working on an app for data capture which will be rolled out to fieldworkers in the coming weeks, which is mind-boggling considering that every day in lockdown costs us a sizable chunk of our economy. Existing solutions for this sort of data capture and processing already exist and are deployed on a daily basis. Instead, we try to build our own app and, in the meantime, send fieldworkers out with piles of paper.
In fact, I argue that we should not do any in-field screening at all. Instead of sending fieldworkers from door to door, the bulk of this work could and should be done via WhatsApp chatbots, SMS and USSD surveys and robocalls. Working with the mobile network providers, there are many ways to ensure quality control, participation and feasibility (e.g. comparing location data and personal details with RICA, making participation data-free, incentivising participation with free airtime etc).
The entire screening exercise as it is currently done is only good for referring symptomatic cases for testing (versus asymptomatic individuals who can still transmit the virus). Arguably, symptomatic individuals are likely to show up at a local testing site or clinic on their own account if their symptoms worsen.
This approach will not help us get an accurate picture of the infection rate because we will never cover the entire country – testing only the obvious cases while a very large chunk of infections go undetected. Nor will we learn anything new about the virus.
The case for a household survey
So what do we do? Screen and test as many individuals as possible, right? Yes and no. Ideally, we would want to test as many people as possible, so that positive cases are separated from the rest of the population and given the care they need. But we simply won’t be able to test a large proportion of the population anytime soon, even if all the stars align. In the meantime, it would still make a huge difference to know whether our currently low infection rate reflects reality or whether the virus has managed to silently spread to large parts of the population beneath the radar of public testing and community screening.
To achieve this, I am arguing that we must take the field resources freed up by shifting face-to-face screenings to digital, and invest them into a representative, nationwide testing survey. The idea here is that if we apply the same sampling procedures used for representative household surveys to test for Covid-19, we will be able to get a precise estimate of the real infection rate fairly quickly. What this means is that we go to the homes of a statistically random selection of South Africans and test them, regardless of whether they show any symptoms or not.
Ideally, we would aim to conduct the survey once lockdown has ended and confirmed cases in clinics and hospitals start to rise exponentially once again. Not only will this enable us to know with a greater level of certainty the percentage of people infected in our country, but if we ran it as an ongoing survey over the course of the epidemic, adding variables of interest as we go along, we could learn so much more about the virus.
This way we would not only have a reliable measure of the spread of the virus at any given time, but we could also start making inferences about its mode of transmission and keep a finger on the pulse of a nation in distress.
This is not a maverick idea: Germany, USA and Austria at least are planning to roll out a representative testing survey in the near future (Austria already conducted a relatively small one to be repeated on a larger scale) for the same reasons provided above. The government’s own plan to introduce national surveillance days where 5% of selected schools, mines, big companies and prisons will be tested (or screened) on a monthly basis seems to follow a similar logic, but it will not replace a representative survey as it still leaves out a large proportion of the population and is not very efficient.
There are statistical and practical hurdles we would need to overcome, but we do have a real advantage over many other places in rolling out such a survey.
While face-to-face household surveys have been largely replaced by online and telephonic surveys in developed countries, we do have a highly capable, experienced and innovative face-to-face survey industry prepared to get the job done quickly and efficiently – if given the chance. DM