Use of Mobile Device Data To Better Estimate Dynamic Population

Sep 20, 2017 - *E-mail: [email protected]; phone: 0061 417287582. ... data is made available by participants in Crossref's Cited-by Linking servi...
0 downloads 0 Views 4MB Size
Article pubs.acs.org/est

Use of Mobile Device Data To Better Estimate Dynamic Population Size for Wastewater-Based Epidemiology Kevin V. Thomas,*,†,‡ Arturo Amador,§ Jose Antonio Baz-Lomba,† and Malcolm Reid† †

Norwegian Institute for Water Research (NIVA), Gaustadalléen 21, NO-0349 Oslo, Norway Queensland Alliance for Environmental Health Sciences (QAEHS), The University of Queensland, 39 Kessels Road, Coopers Plains, Queensland 4108, Australia § Telenor ASA, Snarøyveien 30, NO-1360 Fornebu, Norway ‡

S Supporting Information *

ABSTRACT: Wastewater-based epidemiology is an established approach for quantifying community drug use and has recently been applied to estimate population exposure to contaminants such as pesticides and phthalate plasticizers. A major source of uncertainty in the population weighted biomarker loads generated is related to estimating the number of people present in a sewer catchment at the time of sample collection. Here, the population quantified from mobile device-based population activity patterns was used to provide dynamic population normalized loads of illicit drugs and pharmaceuticals during a known period of high net fluctuation in the catchment population. Mobile device-based population activity patterns have for the first time quantified the high degree of intraday, week, and month variability within a specific sewer catchment. Dynamic population normalization showed that per capita pharmaceutical use remained unchanged during the period when static normalization would have indicated an average reduction of up to 31%. Per capita illicit drug use increased significantly during the monitoring period, an observation that was only possible to measure using dynamic population normalization. The study quantitatively confirms previous assessments that population estimates can account for uncertainties of up to 55% in static normalized data. Mobile device-based population activity patterns allow for dynamic normalization that yields much improved temporal and spatial trend analysis.



catchment.9 A static population, estimated by a variety of methods, has been used to normalize the biomarker loads in the vast majority of the WBE studies performed to date. At certain WWTPs, municipalities annually provide the number of individuals present in a catchment when everybody is at home (de jure) or used the most recent census data. Another approach has been to use the level of one or more selected hydrochemical parameters (i.e., biological oxygen demand (BOD), chemical oxygen demand (COD), nitrogen (N), phosphorus (P), and ammonium (NH4+)) to estimate the contributing population based from accepted per-capita loads of each parameter.11−16 A number of studies seeking to improve the accuracy of such measurements have been carried out. Been and colleagues16 recently normalized their drug loads to NH4+ which showed a distinctive pattern associated with short- and long-term population fluctuations and provided the first dynamic measurements to be used in WBE. Another approach for determining the de facto population has been to

INTRODUCTION The quantitative analysis of the endogenous and exogenous biomarkers excreted by humans in wastewater has been shown to be an effective approach for assessing a range of different (exposure related) factors in defined communities.1−4 The approach, referred to as wastewater-based epidemiology (WBE), has predominantly been used to quantify and compare the use of illicit drugs in populations defined by their connectivity to a wastewater treatment plant (WWTP).5,6 The quantitative data (population normalized loads) generated by the approach are subject to a number of sources of uncertainty;7−10 an important factor when comparing data spatially and temporally. Considerable efforts have been made to increase the comparability of data through understanding and reducing the different sources of uncertainty through the use of a harmonized approach. By following a best-practice protocol, the sources of uncertainty in generating population weighted biomarker loads are typically between 5% and 10% from sample collection (related to sampling mode, frequency, and characteristics of the catchment), up to 26% from analytical variability, preferably