Life Unleaded: measuring the impact of lead reduction in water supplies

Lead is a very toxic element that can have adverse consequences on babies and children's health, even at low concentrations. In the 1970s, the majority of people - including pregnant mothers - living in Glasgow would drink tap water that contained 5 times the level of lead than is currently accepted. When the toxicity of lead started to be widely recognised, sometimes after campaigns by civic groups, the UK regulators set up programmes and policies to reduce and eventually phase out lead from water pipes and petrol. This project will examine the influence of specific interventions to remove lead from the water supply on pregnancy outcomes (e.g., live births, birth weight, stillbirths, miscarriage) and infant mortality by combining historical and administrative health data. The analysis will study two water treatment programmes that successfully reduced lead content in tap water in Glasgow in 1978 and 1989. The project has been approved by the Public Benefits and Privacy Panel and a data sharing agreement between Scottish Government and the University of Glasgow has been signed. The processing of these data meet the Data Protection Act Schedule 2, section 6 and Schedule 3, section 8. Glasgow receives its water from Loch Katrine reservoir in the Trossachs. Lead piping combined with the soft water of the loch meant that lead levels in Glasgow's water were considered unsafe. Two lead treatments were carried out to raise the pH of the water supply, so that lead levels were reduced. The first, in 1978, was the addition of lime to the water supply. The second, in 1989, was lime combined with orthophosphate. These successfully reduced lead levels in Glasgow's water. However, no research was carried out on what the effects of this lead treatment were for the population of Glasgow. Using a unique dataset, I will exploit the variation in individual's exposure to lead contaminated water to examine the impact of this on a range of outcomes.
Information on pregnancy outcomes such as live births, stillbirths and miscarriages and will be gathered from different records: SMR02, SMR01 (ISD), and Death, Birth and Stillbirths Registrations (NRS). The data will cover the period from 1975 to 2000. Live births will be linked to death registration to identify if the child died before age 5. Every outcome of interest, such as birth weight, gestational age, miscarriages, death before the age 5 will be analysed separately. Variables that will be useful as controls are gender and mother's characteristics such as age, height, smoking history and previous obstetric history (if available). These records will be linked to birth records (NRS) when possible to provide information on parental occupation and backgrounds that could be used as additional controls. With an effort to improve the set of confounders, additional information related to Carstairs scores from 1981 and 1991 will be linked to postcode sector. Further data will be sought on education and crime outcomes that can be linked to birth records. This will be used to understand the wider effects of the lead treatment beyond health outcomes. Lead has been shown to affect cognitive development in children which may affect educational outcomes. There is also some evidence that lead exposure reduces inhibitions against violence and leads to loss of impulse control. Reduced exposure to lead may therefore cause a reduction in individual's propensity to commit a crime, especially violent crime.
A secondary aim of the research will be to use the findings of the effects of lead to predict where high lead levels may still be present in the water supply. Lead contamination in water has been found in new houses built in Edinburgh in the early 2000's as well as in the highly publicised case of Flint, Michigan. Using modern machine learning techniques, I will build a forecasting model that aims to find where there is likely to be high levels of lead in the water supply.