Blood Lead Testing in Flint Before and After Water Contamination

This study used the complete set of Medicaid claims data for children born in Michigan between 2013 and 2015, linked to their birth record data, which contained maternal census block of residence during pregnancy. These data allowed us to identify children who did not have any lead test claims. The resulting longitudinal dataset tracked 206 001 children from 2013 to 2017. Because we could not follow all children until they reached age 3 and, given Medicaid requirements that children receive a lead test by age 24 months, we followed children until 30 months.

A child was “affected” by the Flint water crisis if maternal residence at birth was within Flint. The comparison group was children born to mothers in other large cities in Michigan and, separately, children born to mothers in Genesee County, but outside Flint city limits.

Sample averages were plotted to visualize trends in lead tests in Flint and other Michigan cities from 2013 to 2017.^*

The association between the issuance of a lead advisory in Flint and the proportion of children who received a lead test by a specific age was estimated by using difference-in-differences linear regression, comparing children whose mothers resided in Flint with those whose mothers resided in other Michigan cities before and after the lead advisory (Supplemental Information, Supplemental Tables 1–3, and Supplemental Figures 6–10). This method adjusted for other contemporaneous factors that may have affected testing in Michigan, such as the implementation of the Affordable Care Act or state-wide campaigns against lead paint.

The analysis was at the individual level, controlling for the child’s sex and birth weight, maternal age, education, race, and smoking during pregnancy, and paternal age and education. These characteristics may be associated with the underlying likelihood of receiving a test. The analysis included time and city fixed effects to account for general trends in lead testing receipt over time in Michigan and time-invariant aspects of each city. It also included birth year and birth month fixed effects to control for cohort characteristics. Standard errors were clustered by city of birth.

Additionally, analyses were disaggregated into monthly differences to investigate the effect of the advisory on lead tests among children aged 10 months (before) and 13 months (after the recommended annual test) separately. Analyses were performed in Stata 16.

The intervention studied was a public lead advisory and subsequent declaration of a public health emergency from the City of Flint in September 2015.^17,23  Previous studies have revealed that the water in Flint had higher levels of lead after the initial switch from Lake Huron to the Flint River in April 2014.^{17–19,24,25}  However, city officials, including the city manager and Environmental Protection Agency officials, continually reassured residents that the water was safe to drink.^26,27  The first official confirmation of higher than usual lead levels in residential water was issued by Genesee County in a press release in September 2015, singling out young children as particularly vulnerable.^28,29  The advisory coincided with a highly publicized study finding higher rates of blood lead levels among children in Flint,^18,20,28,29  which may have induced higher rates of lead testing of children in Flint. This advisory, although specific to Flint, may have had a smaller effect in other Michigan cities, as well, especially other areas in Genesee County. Thus, estimates were a lower bound because the control group may be partially treated.

The main outcome measure was a blood lead test claim (current procedural terminology code: 83655) by 12 to 24 months of age and by 0 to 24 months of age.

The sample consisted of 62 650 children tracked for an average of 35.1 months in Flint and other large cities in Michigan. The Genesee County comparison tested whether children living closer to Flint were more affected by the lead advisory than those in other cities. The Genesee sample tracked 10 523 children who were born to mothers residing in the county, outside of Flint.

The first outcome of interest was the unadjusted monthly means of lead test receipt for Flint, other cities, and Genesee County excluding Flint. Figure 1 reveals the percentage of young children with a blood lead claim in each calendar month for all eligible Medicaid-enrolled children 0 to 3 years old. Whereas the 3 lines reveal similar levels and trajectory (∼3%–4%) until the city’s lead advisory in September 2015, the line for Flint increased to almost 5% in October whereas the rest of Michigan decreased slightly. After the Governor’s public apology for the water contamination in January 2016 and the declaration of a state of emergency,³⁰  Flint blood lead tests increased to 13% and remained high through spring 2016. Other cities’ blood lead tests remained approximately constant from 2016 to 2017. Genesee County experienced a smaller increase in lead tests to ∼7% in February 2016. The 3 lines began to follow similar levels and patterns to each other by mid-2016.

Figure 2 reveals the percentage of children who had ever received a lead test in Flint (Fig 2A), other cities (Fig 2B), and Genesee County (Fig 2C), by the specific age in months denoted on the horizontal axis, before and after September 2015. Flint revealed a leftward shift in lead test timing. Lead testing increased by 35 percentage points (PPT) from 10 to 12 months after September 2015, compared with a 45-PPT increase before the advisory. In other cities, no shift occurred after September 2015. The rest of Genesee County experienced a smaller shift in lead test receipt before 12 months. Importantly, this increase in Flint appeared to be only a shift in the timing of the first lead test receipt; no overall increase in the proportion of children receiving a test occurred.

Figure 3 provides difference-in-differences estimates for subsamples by age. The dependent variable was a binary variable indicating if the person had ever received a lead test. Fig 3A contains results for Flint compared with other Michigan cities; Fig 3B contains results for Flint versus the rest of Genesee County. Fig 3A coefficients were positive and statistically significant for ages <17 months, with the largest magnitudes for 6 to 11 months, showing lead testing started at an earlier age after the lead advisory. The differences were slightly larger in magnitude than those from Fig 2 after adjusting for covariates. They suggest that lead test receipt among children in Flint from age 6 to 11 months increased by ∼15 to 23 PPT. From 12 to 30 months, lead testing continued to be higher among children in Flint compared with other cities, although the difference was attenuated, and dropped off sharply after 30 months. Panel B provides similar but slightly smaller effects, reflecting the increase in children ever receiving lead testing in other parts of Genesee County.^†

Figure 4 provides unadjusted rates of ever receiving a lead test among children for a monthly cross-section of 10-month-old (Fig 4A) and 13-month-old children (Fig 4B). There was a large increase in the proportion of individuals who received a lead test in Flint beginning in late 2015. The proportion of 10-month-old children ever receiving a lead test increased from <10% to >40% by the end of the first quarter of 2016. Among 10-month-olds in other Michigan cities, the proportion remained constant between 15% to 20% across the sample period. Genesee County results were slightly noisier but revealed evidence of a smaller increase in early 2016 compared with Flint. Changes among 13-month-old children were less clear in Flint. Although the average was between 45% to 60% in the period before the lead advisory, it dipped to <40% in the summer of 2015. After the advisory, the rate of any lead test receipt for 13-month-old children ranged mostly between 60% and 80%, with large fluctuations. For other cities, these rates ranged between 40% and 50% throughout the sample.

Figure 5 presents these results in a regression framework in which each point estimated the monthly difference between Flint and the control group by age cohort. Comparing Flint to other cities, the point estimates suggested an increase in lead testing for 10-month-olds of between 30 and 50 PPT in 2016. This was consistently 5 PPT higher than the comparison between Flint and the rest of Genesee County in Fig 5C. Evidence of an increase in lead testing for 13-month-olds was more mixed. The persistence of the increase in testing in Fig 5A and 5C was surprising given the half-life of lead is only 2 to 8 weeks, and the surveillance system was not set up for children <12 months of age.³¹ 

The main findings of this study were that blood lead testing occurred earlier after the lead advisory for children in Flint, Michigan, but the cumulative rate of testing by 24 months of age was unchanged. These results suggest that earlier testing occurred among children whose parents would have initiated testing by age 2 years irrespective of the Flint water crisis.

Medicaid mandates lead testing for children aged 12 to 24 months;^15,32  however, testing rates are typically low.^16,33,34  Children in Michigan have long had among the highest rates of elevated blood lead levels in the country and low rates of follow-up testing for those with previously elevated blood lead levels.^15,35,36  The individual-level Medicaid data used in this article revealed only approximately one-half of the children who resided in Flint were tested by 13 months of age and 28% were not tested by 24 months.

This study reveals the difficulty of inducing changes in early-childhood preventive health care.^‡ The fact that lead testing rates were unchanged despite the emergency created by the water contamination suggests that primary prevention should remain the focus of policy; it also points to the need for a more proactive approach to lead testing, including reducing initial exposure through home visit and testing programs,^9,37  school-based testing, and other lead screening policies,^30,36,37,38  federal interventions,^39–41  and community outreach programs using mobile health platforms.⁴²  These policies need to be balanced with prevention guidelines and should be based on evidence of the costs and benefits of these efforts.^7,15,43,44 

Recent interventions in Flint and nationally have focused on reducing early childhood lead exposure through primary prevention, such as targeted home remediation programs and reduced exposure to lead pipes, as documented by the Flint Lead Free workgroup reports, recent legislative sessions, and legal settlements.^{37–41,45–50}  After the Flint water crisis, reports revealed that, in 2016, >82% of all homes undergoing environmental inspection were those of children with blood lead levels of >5mg/dL, highlighting the importance of lead testing and the interdependence of primary and secondary prevention.^12,50 

Beyond testing programs, Medicaid has urged states to do better in lead screening and reporting for children in a recent informational bulletin.¹²  The recommendations included a 2-pronged approach of reinforcing the lead testing mandate and more outreach to increase awareness of the latest mandates and recommendations.¹²  Despite these recommendations,^32,34  increasing lead screening among Medicaid-enrolled children remains a challenge, as our results reveal. Yet, lead screenings remain an important tool for reducing racial and economic disparities in exposure to environmental hazards,^16,34  as well as serving to connect children and their families to a medical home for ongoing care.²¹  Lead screenings during the coronavirus disease 2019 pandemic decreased by nearly 40% in Flint and across Michigan,^51,52  highlighting the relationship between interrupted care and lead testing. Although some states, including Michigan,⁵³  have considered enacting universal lead testing for all infants,⁵⁴  concerns about false positives, lack of expert consensus, and lack of positive benefit–cost calculations in areas with low rates of lead exposure have compelled states to focus only on testing for infants at a high risk of exposure to lead.^{15,35,54–56} 

This study has several limitations. It focused on children with Medicaid claims data through age 3. This omitted children with private insurance and those who were uninsured; Medicaid was the primary payer for 43.3% of all births in Michigan, and 51% of births in the cities used in this study, but 87% of all births in Flint. Thus, this study focused on a vulnerable population, one more likely to live in areas with older housing and a higher likelihood of lead exposure.^{16,34,35,57,58} 

We assigned children to the Flint group on the basis of the maternal address at the time of birth. To the extent that individuals were more likely to move post lead advisory from Flint and, thus, less likely to fully internalize the lead advisory than other cities, our results would be a lower bound, although previous literature revealed, if anything, a decrease in the rate of moving from Flint compared with other cities.²¹ 

Using claims data allowed the investigation of the receipt of blood lead tests, an important aspect of secondary prevention, but provided limited information on whether parents or clinicians were driving the receipt of these tests,³⁸  or on primary prevention methods.¹⁰  A recognized limitation of Medicaid claims data are that they do not contain follow-up information on children who disenrolled from Medicaid or who moved to a different state. Previous literature revealed no significant change in mobility out of state and county⁵⁹  and decreased movement out of Flint.²¹ 

The contamination in Flint is one of several recent high-profile water lead contaminations (Washington, District of Columbia, Newark, New Jersey, Benton Harbor, Michigan),^60,61  and elevated water lead levels are more common across municipalities in the United States than these high-profile cases suggest.⁶²  Policies focusing on secondary prevention must wrestle with the challenge of providing incentives for following expert recommendations among the harder-to-reach set of children, clinicians, and parents.

The events studied in this analysis illuminate that concern over lead exposure, media coverage, and active public health campaigns do little to affect long-term lead testing rates. Therefore, increased primary prevention and designing proactive public health outreach methods that increase lead testing^39–41  and increase adherence to screening standards through Medicaid-related incentives should be a focus of future public health policy.

We thank Glenn Copeland and Tamara Weaver of the Vital Records and Health Statistics Division and Matthew Schneider of the Medicaid Division of the Michigan Department of Health and Human Services for providing vital statistics data and facilitating receipt of Medicaid claims data linked to vital records. The authors gratefully acknowledge manuscript editing and comments from Antony Hsu, MD (St. Joseph Medical Center, Ypsilanti, MI), Steven Korzeniewski (Wayne State University, Detroit, MI), and Anne Byrne, PhD (US Department of Agriculture).

PPT

percentage points