Matlab, a subdistrict 57 km from Dhaka in Bangladesh, is the site of the oldest demographic surveillance site in a low and middle income country. It’s famous for the world-changing trials conducted there of using oral rehydration solution to treat childhood diarrhoea; vaccines for cholera, rotavirus infection, and other infections; and family planning. The use of tetanus toxoid in the control group of a cholera vaccine trial showed a dramatic drop in neonatal mortality among the offspring of pregnant women given the toxoid and led to its global adoption. And the hospital attached to the research centre provides such high quality and free treatment for diarrhoea that people travel as far as 100 km to the hospital, passing many government and private hospitals on the way (and the government hospitals are free). When I visited a couple of months ago, we discussed with Dr Md Al Fazal Khan, the head of Matlab, both the glorious past and the hope for an equally glorious future. (I should declare that I was until recently the unpaid chair of the board of icddr,b [formerly International Centre for Diarrhoeal Disease, Bangladesh], the parent body of Matlab, for five years).

We travelled to Matlab from Dhaka by car and speedboat. The speedboat trip is both beautiful and exciting. The Gumti River is much wider than the Thames but a minor river by Bangladesh standards, a country that drains both the Brahmaputra and the Ganges rivers into the Bay of Bengal. The river is filled with water hyacinths, and at times it looked as if there was no way through them; but our driver sped over them, reversing when we were through to unwrap the hyacinths from our propeller. We passed lone fishermen in small wooden boats, huge, brightly coloured barges and ferries, and boats filled with waving children on their way to school. We also saw boats that are the homes of a nomadic ethnic group Bede who make their living by snake charming and dancing at weddings.

When we arrive in Matlab a large group was assembled to meet us. They are used to visitors: people come from all over the world to visit Matlab. We were presented with flowers and taken to the guest house. Over a late breakfast we discussed the extraordinary dataset that researchers at Matlab have created over more than 50 years. There are high quality data on every birth, death, marriage, divorce and much more, including verbal autopsies on the cause of death, on 238 000 people. Data were collected from every household monthly, but might soon be collected quarterly. Families are linked together, and we later saw a picture of a family of four generations included in the dataset. During the 50 years there have been a famine and a dramatic demographic and epidemiological transition with family size dropping precipitously and deaths from non-communicable increasing from 10% in 1986 to 80% in 2006. The data are all digitised. There is also another dataset on maternal and child death. In addition, there have been many studies asking other health and demography questions that have contributed data.

The data have been used in thousands of studies, but we all agreed that they had much more to offer, not least in social sciences. The challenge to the management is to know what is required to get the dataset in an optimal form for sharing and to understand the market for these data, “the new oil,” as they are often described. The hope is that the data might be used to produce many more insights and generate income to help pay the costs of the hospital.

Later we visited a centre where one of the community health research workers was collecting information from a household. Each worker covers about 50 households. Data were collected on paper from 1966 to 2011, but are now collected electronically. The original paper records are stored in the hospital.

It has become increasingly difficult to fund the data collection. It has been funded mainly by icddr,b core money with partial support from the large trials that were conducted there. There continue to be large trials—for example of Hepatitis E vaccine—but it’s a challenge to find funding. The big killer now is not infection, but non-communicable disease (NCD), and a large cohort is being studied with the University of Cambridge of the burden of NCD and associated risk factors. The hope is to conduct more trials, probably of treating NCD, including mental health, as well as vaccine trials, and Matlab has an unequalled record of conducting high quality trials.

As part of its contract with the local people Matlab Hospital provides treatment for diarrhoea and maternity care. We visited the hospital, which despite being filled with patients with diarrhoea has no smell. The hospital is seeing an 8% annual growth in patients with diarrhoea, most of whom now come from outside Matlab. This is a tribute to the quality of care offered to the patients, but presents a problem. The capacity of the hospital is being reached, and there are usually patients in the corridors overnight. The care is provided free, but it may be necessary to introduce a charge for patients from outside Matlab—not so much to make income, but to deter patients. But there is something unsatisfactory about having to divert patients to poorer quality care.

We saw as well the maternity care, including a delivery room. Services are provided not only in the main hospital, but also in some 35 fixed site clinics and three substations. We visited both a clinic and a subcentre: the clinics are staffed by community health workers and provide vaccination, family planning, and treatment of minor illnesses; the subcentres are staffed by nurses and paramedics and offer outpatient services, antenatal care, and normal vaginal deliveries. We learnt that the Caesarean section rate is rising rapidly and is now about 40%. One reason for this is that agents of private clinics hover outside the hospital and subcentres and encourage women to have Caesarean sections. Matlab Hospital itself does not offer Caesarean sections. We agreed that once the rate has risen it’s hard to reduce it.

The work at Matlab began in the 60s, and we visited the barge that was the original centre. Scientists, mainly Americans, slept and ate on the barge. In those days there was no phone and no air conditioning. Now Matlab has Wifi and a comfortable guest house where we stayed. The challenge is to attract new work and funding and not only maintain but enhance a unique dataset. icddr,b is currently trying to recruit a chief scientist to develop the work. The scientist will inherit an extremely well run machine and be presented with the chance to do world-class research and improve the health of people in Matlab, Bangladesh, and other low and middle income countries.

Richard Smith was the editor of The BMJ until 2004.

Competing interest: RS has been the unpaid chair of the board of trustees of icddr,b for the past five years but has now stepped down. icddr,b paid his expenses to travel to Dhaka for the board meeting and on to Matlab.