The problems of eligibility and endogenous confounders when assessing the mortality impact of a nationwide disease-prevention programme: The case of insecticide-treated nets in Togo

Centers for Disease Control and Prevention, Division of Reproductive Health, 4770 Buford Hwy NE, Mail Stop K-23, Atlanta, GA 30341, USA.
Population Studies (Impact Factor: 1.08). 03/2011; 65(1):57-71. DOI: 10.1080/00324728.2010.544323
Source: PubMed


Evaluation of the mortality impact of nationwide disease-prevention efforts is complicated by potential endogeneity: programme recipients may have unobserved characteristics that simultaneously make them both more likely to become recipients and more likely to survive as a result of other health practices. This population-based study assesses the mortality impact of a nationwide programme that distributed insecticide-treated nets (ITNs) to mothers of children aged 9-59 months in Togo. By comparing mortality rates before and after the programme according to households' eligibility status, we demonstrate that a one-time programme that restricts eligibility to households with a surviving child excludes some households with a high risk of child mortality. We then apply simultaneous estimation models to untangle the mortality impact of ITNs from the effects of unobserved confounders and show that among eligible households, living in a household with ITNs significantly reduces mortality for children aged 20-59 months, even after controlling for endogeneity.

11 Reads
  • [Show abstract] [Hide abstract]
    ABSTRACT: The continuously increasing array resolution of CMOS imagers poses a great challenge in combining high-frame-rate and low light detection in the same sensor. To cope with this, parallel readout architectures are needed. This paper proposes a readout architecture for 8 K stacked image sensors, which uses a novel 1D decoding readout based on block-of-pixels and incremental-sigma–delta ADCs. The proposed 1D decoding system reduces the control lines of the pixels and allows a simpler decoding, an increased parallelism, and an improved robustness over process yield. The experimental results from a test chip implemented in a standard CIS technology show that at 10 $mu{rm m}$ pixel pitch, the proposed readout architecture can achieve a high-frame-rate of 730 frames/s and a low read noise of 1.4 $e^{-}$ . In a real stacked implementation, the frame rate can further increase to about 960 frames/s at 8 K resolution, at the cost of a slight increase in thermal noise by 14 $mu{rm V}$ .
    IEEE Sensors Journal 06/2014; 14(6):1966-1973. DOI:10.1109/JSEN.2014.2307792 · 1.76 Impact Factor