Testing data: A powerful ally in the fight against Covid-19

Covid-19 cases are on the rise in Bangladesh again. The average number of new cases over the last seven days was 1,854 per day, compared to an average of 1,444 a week earlier—that's a 28 percent increase. Some startling statistics recently emerged showing huge spikes in cases in border districts, raising the alarm that we may be seeing a spread of the Delta variant (popularly known as "Indian variant") spilling over the Indian border. However, to understand the true picture of Covid-19 in these districts, there is an urgent need for more numbers. It's been over a year since the head of the World Health Organization stressed the need to "test, test, test" and make widespread testing, isolation and contact tracing the "backbone" of the global response. But are enough tests being conducted in these districts to be able to rely on the statistics and form an effective response?

Since the outbreak of the pandemic, Bangladesh has conducted just over six million Covid tests in total, which sounds like a lot until you realise that in a country with an estimated population of 166 million people, that represents a testing rate of 36 per thousand (or one test for every 27 people). Now, compare this to other countries in the region: Pakistan has a testing rate of one test for every 16 people. For Nepal and Sri Lanka, it is one test for every nine and six people respectively, and in India, one test for every four people. Indeed, with the exception of Afghanistan, Bangladesh has the lowest test rates in South Asia. Yet, the latest GDP per capita figures show the country has the highest growth in the region. Given the remarkable economic success of the country, it is disheartening that widespread testing has still not been adopted.

A more direct comparison can help give a better perspective of these testing numbers. West Bengal has an estimated population of 100 million (10 crores), equivalent to approximately 60 percent of the population of 166 million (16.6 crores) of Bangladesh. And yet, the number of tests being conducted in the state is approximately 70,000 per day, according to Indian media, compared to an average of around 16,500 per day in Bangladesh over the last two weeks. In short, West Bengal is conducting seven times as many tests per person per day, despite West Bengal itself having been criticised for one of the lowest state-wise testing rates across India.

Bangladesh has a proven capacity to conduct more testing though. Around the peak of the second wave, for several days between April 4 and 13, the number of tests exceeded 30,000 per day, with the highest number just short of 35,000 on April 12. Indeed, for the two months from March 12 through to May 12, the average (median) number of tests per day was 24,000 compared to the median of 15,500 daily tests over the last six months (and 16,500 over the last two weeks).

In fairness, the successes must be highlighted, namely that Bangladesh does have the lowest death rate per capita in the region (although each of the nearly 13,000 deaths is a tragedy)—so why is more testing even important? Because, knowledge is power and effective decision-making relies on being informed. If we are to avoid the tragedy happening in India, the relevant authorities need to understand and be able to act on the situation as it unfolds, not two weeks later.

Insufficient testing has limited the usefulness of the reported numbers of cases throughout the pandemic in Bangladesh. It has never been known at any one time how many people are infected, how widespread infections are, and importantly, where they are. More cautious commentators have relied on the less direct measure of positivity rate (the proportion of tests which return a positive diagnosis) to vaguely track the situation and general direction of change. Day-on-day increases in positivity suggest an increased rate of infection, such as is happening now—the average national positivity rate over the last week now being 10.8 percent, whereas it was 8.4 percent two weeks ago. Which leads us to the border districts.

This week, analysis by the Directorate General of Health Services (DGHS) reported that bordering districts such as Rajshahi, Chapainawabganj and Satkhira (amongst others) have positivity rates in excess of 30 percent. These numbers are very alarming, but how reliable are they?

In statistics, there is a concept called sampling error, the important part here being that working with small numbers can lead to extreme statistical values, and trying to draw too much insight from a small sample is not wise. Take, for example, the case of Rajshahi. The DGHS's Covid-19 dashboard shows that on June 7, the positivity rate in Rajshahi hit a staggering 48 percent. However, in absolute numbers this was 108 cases, which means there were 225 tests, in a district with a population of over two million (20 lakh). From such small numbers of tests, it cannot be known whether this represents a substantial level of transmission across broad areas of the district, or just a targeted testing of a single infected area. In other districts, the testing numbers are even lower, with an average of 72 and 52 tests per day in Satkhira and Naogaon respectively over the last month. At these numbers, a swing of 10 percent in positivity can be achieved by testing of a single infected family of six. Only through routine widespread testing can we be confident that these are not isolated biases in the sampling.

It should be stressed here that an increase in the rise of cases is not in dispute—these figures clearly show a concerning rise in the transmission rate. The problem is that the true extent of this rise is not knowable from the small numbers of tests being conducted.

With the outbreak in India clearly becoming uncontrollable by mid-April and with the knowledge that completely closing the porous borders is more or less impossible, stepping up testing in border districts would have been a sensible approach. We know that given the typical delay of a couple of weeks between infection and hospitalisation, by the time hospitals are under pressure, as they are now, the transmission is already well-established. But routine testing in these districts would be a significant weapon in the authorities' arsenal in catching and isolating transmission of cases at the early stages, which is crucial in preventing the spread of new variants. However, so far, that has not been the case. If we take the example of Rajshahi again, we find that the average daily tests in April did increase to 266 during the second wave, up from 186 in March, but then decreased back down to a daily average of 183 in May. Other border districts tell a similar tale.

Data is an extremely powerful tool and has transformed the world in recent decades. It is encouraging to see the authorities acknowledge this by investing in the publication of high-quality Covid-19 data dashboards for public consumption, and the recent announcement of waiving the test fees in specific districts, to encourage more tests, is a step in the right direction. Without a doubt, the best and most effective decisions are informed decisions, and data is the key to that. However, the usefulness of data is only as good as its quality. We must invest in more testing to truly understand and reduce the transmission of Covid-19.

Dr Timothy S. Green is a data scientist with a PhD in astrophysics. He lives in Dhaka.