Evaluating Difficult Decisions in Public Health Surveillance: Striking the Right Balance between Timeliness and Completeness.

Introduction: State-based surveillance programs play a key role in birth defects planning, prevention, education, support, and research activities. High-quality data are essential to all of these functions, and a key indicator of quality is timeliness. The Florida Birth Defects Registry (FBDR)-one of the largest population-based state registries in the United States-faces challenges with timeliness, as evidenced by its 18-month lag time. The goal of this study was to determine if the timeliness of the FBDR could be improved without significantly reducing the completeness of birth defect ascertainment.

Methods: Using 2006-2011 data from the FBDR, we first investigated the timing of diagnosis of birth defects by estimating the effect of different periods of follow-up on prevalence rates reported by the FBDR. We achieved this through retrospective reconstructions of the FBDR under 5 different scenarios with progressively narrower follow-up windows for each infant, and by comparing recalculated rates to the rate of the current FBDR with 1 year of follow-up. We then considered scenarios in which the time lag used to construct the FBDR was reduced (15, 12, 9, and 6 months) by using less data (from 7 to 4 quarters). Recalculated rates were again compared to the current FBDR constructed with 2 years of data and an 18-month lag. Analyses were performed overall and for 44 specific defects.

Results: During the 6-year study period, the FBDR identified more than 27,000 infants with a defect detected during the first year of life. Restricting follow-up from 1 year to 9 months would only result in a loss of 1.4% of cases. Cutting follow-up in half to 6 months would miss 3.2% of cases, although there was significant variation across defects. Improving timeliness had a small impact on completeness of ascertainment. Overall, compiling the FBDR with only 6 quarters of Florida Agency for Health Care Administration data (as opposed to 8 quarters) would improve timeliness by approximately 6 months, resulting in a registry that is 99.4% complete.

Discussion: Six-to-nine month improvements in timeliness were achievable with a minimal sacrifice in completeness (0.6%-1.7%). Efforts to enhance data quality through the assessment of timeliness and completeness indicators are not unique to birth defects surveillance programs. Other programs, particularly those with similar passive case ascertainment protocols, can use our findings to consider a more timely release of registry data, or to design similar investigations of their own.