Much of environmental epidemiology focuses on the causal pathways connecting environmental exposures and risk factors to health outcomes. For example, we may seek to understand the degree to which variability in exposure to air or water contamination, chemical agents, disease vectors, or radiation explains variability in outcomes within a population. In other cases, we may seek to understand the degree to which risk or protective factors such as physical activity mediate the strength of symptoms resulting from a given level of exposure. Unfortunately, assessing these connections can be difficult using traditional instruments, as self—reporting of average exposures over long timeframes is often inaccurate, and fails to capture the effects of non—linear dose—response relationships. At the same time, detailed self—reporting of these matters is often prohibitively burdensome, and often incomplete, and frequently otherwise inaccurate.
Ethica provides smartphone—based solutions for many challenges in traditional environmental epidemiology, using three distinct features:
Leveraging the smartphone’s strong set of sensors.
Elicit information from participants using ecological momentary assessments (EMAs) and environmentally—triggered surveys.
Ability to allow for study—specific user interfaces.
Many of such assessments rely heavily on location data cross linked with other external data sources. For example, location data collected by Ethica for a given study can be cross linked to databases or live data feeds from environmental monitoring stations or sampling results providing such information as UV, SOx, NOx and CO levels, enterococcus levels, or mosquito population counts. In other cases, the databases linked to by location data may be quasi—static, for example, locations of tobacco advertising, or soil or groundwater contamination levels. Alternatively, some investigations may be focusing on exposure to certain environments such as to a specific feed barn known to exhibit adverse air quality. In cases in which there are time—dependent exposure effects, it may be helpful to rely on the smartphone’s active sensing of the frequency with which the participant took outside breaks.
In addition to assessing such broad average levels of exposures, some studies benefit from active participant reporting of materialized risks. For example, discovery of a tick on the participant’s body, exposure to pronounced levels of diesel or other fumes or tobacco smoke, skin exposure to petrochemicals, or even having encountered certain types of messaging or promotion (e.g., for tobacco, fast—food, etc.).
In addition to the above information on exposures, explicating the causal pathways affecting outcomes often relies upon evidence concerning protective and risk behaviors. For example, when a participant travels outside during the evening in an area exhibiting a high burden of mosquito—borne illness, Ethica’s on—phone questionnaires might enquire as to whether that participant is wearing mosquito repellent and appropriate protective clothing. During high—risk tick season in which a participant’s location indicates presence in a rural or wilderness area, Ethica may enquire as to which, if any, personal protective practices a respondent has applied, such as use of long pants and sleeves, tucking pants into socks, checking themselves for ticks. When a participant is recorded as likely being outside during the mid—day, Ethica might enquire about the use of protective levels of sunscreen, hats and protective clothing. Alternatively, when considering exposure to chemicals, Ethica may enquire about the use of respirators and protective headgear.
Another key domain of evidence that addressed by the Ethica Health system concerns the health and behavioral outcomes involved. In terms of health outcomes, while clinical outcomes are often of great interest, in some cases such outcomes are captured effectively by other datasets. On the phone, questionnaires and proactive participant self—reporting can capture not just clinical but also subclinical outcomes. Examples include occurrences of wheezing and coughing, a tell—tale “bulls—eye” rash associated with Lyme Disease, sunburn, stomach ache, nausea/vomiting, and headaches that are too mild to trigger care—seeking by the respondent, but adversely affect quality of life and which may sometimes (e.g., with sunburn) pose risk for later complications. While Ethica Health studies can allow participants to report such information through the app, often considerable advantage is secured by posing questionnaires to enquire about occurrence of such health outcomes, either as EMAs triggered at random times or on certain schedules (e.g., in summer evenings for mosquitoes, or in peak tick season), or in EMAs triggered by particular factors (e.g., detection of WiFi networks associated with health providers indicating possible care—seeking).
For some investigations in the environmental health area, behavioural outcomes are also of considerable interest, for example, those involving protective and risk behaviors. For example, researchers may wish to study the impact of posted or media advisories on exposures, such as coastal bathing, presence outdoors at dawn and dusk, or personal protective behaviors against ticks.
The capacity of Ethica Health to elicit information on health exposures, risk and protective behaviours, and both subclinical and clinical outcomes can allow for very powerful insights. For example, we can study the degree to which exposure to feed barns elevates reported occurrence of coughing and wheezing, or which chemical agents are an important precursor to headaches. Alternatively, for that participant we may be able to sense the degree to which occurrence of sunburn is associated with failure to adhere to personal protective behaviours (e.g., use of a hat, sun screen and appropriate clothing).
Example Problems and Questions That Can be Addressed
To what degree does variability in exposure explain variability in outcomes within the population?
How proximate is the impact of exposure on symptomology in different subgroups?
To what degree do other stressors or risk or protective factors (e.g., physical activity) mediate the strength of symptoms seen from a given exposure?
With dynamic modeling:
How much would an intervention (removal of point source X, use of personal protective gear, enhanced ventilation, etc.) help reduce the burden of infection in the population?
Detecting exposure to environmental risks such as poor air quality [e.g., diesel fumes], ticks, mosquitoes, pro—tobacco promotion or messaging.
Auto detecting some risk status such as being in a feed barn, being outside, swimming in the ocean. Typically the severity of this risk status is detected by cross—linking with other database (e.g,. UV information, reports of pollutant levels from environmental sensors).
Allowing manual reporting of certain risk status that cannot be automatically detected such as wearing short sleeves, wearing mosquito repellent.
Experience of subclinical symptoms that would otherwise go unreported, for example, wheezing and coughing, stomach ache, sunburn, nausea/vomitting, headaches.
More reliable information on context of exposures through sensors (location, time, bluetooth): where, when, with whom.