Study population
The study enrolled 15 adolescent women from a clinic-based setting with the following inclusion criteria: (1) female, (2) age 14 to 17 years, (3) willing and able to travel at least 3 blocks away from home at least once each day, (4) capable of carrying a phone with her at all times for a period of 7 days, and (5) understands and speaks English. We did not collect sociodemographic data from the young women that we recruited. According to the electronic medical records for the clinic from which we recruited, there were 1004 outpatient visits in which patients were young women 14 to 17 years old. The patients' race/ethnicity reported in these records was 32% white, 27% black, and 40% Latina. The majority (88%) were on public insurance. We obtained assent from each participant and consent from her parent/guardian. We staggered enrollment such that one to four women had phones in the field at a time so that we could modify the study protocol and cell phone programs in response to problems that arose. The research received approval from the Institutional Review Board of Indiana University – Purdue University at Indianapolis.
Procedures and measures
Orientation meeting procedure
A research assistant (RA) met the adolescent at a location of her choosing. The RA oriented the adolescent to the phone – how to turn it on, charge it, and make calls. The RA answered any questions that the adolescent had and reviewed the study requests:
1. Carry the phone with you at all times while awake
2. Always keep the phone 'on'
3. Charge the phone overnight each night
4. Travel at least 3 blocks away from home at least once each day
5. Contact the RA in the event of malfunctioning or lost study equipment (the phone service agreement includes an insurance policy providing for free replacement in the event of loss or damage)
The adolescent could make unlimited calls in the evenings and weekends. She could also use up to 150 minutes of calls between 7 am and 7 pm during the weekdays of the study period.
Exit interview/process measures
At the conclusion of the 7-day participation period, the RA collected the phone and charging devices and conducted a semi-structured interview. Next, the RA reviewed the maps created from the GPS data and asked the adolescent to comment on whether she thought the map accurately represented her locations.
Process measures included:
1. Self-reported compliance in carrying the phone
2. Self-reported compliance in charging the phone
3. Self-reported accuracy of the maps in reflecting her 7 day location data
4. Periods in which the phone had no movement for over 24 hours
5. Number of phone malfunctions
6. Whether phone was used
7. Whether the phone was returned in working order
GPS-enabled cell phone
Participants carried Blackberry 7520 GPS-enabled cell phones (cost $150/phone). The cell phone measures 11.4 (L) × 7.4 (W) × 2.8 (D) centimeters, weighs 174 grams. The phones use an assisted-GPS system in which an approximate location is identified using less accurate cell phone towers and does not need a clear view of the sky to determine location. These cell phone tower data are subsequently processed to identify a more accurate point using satellite communication. This method is desirable because it uses less battery power (allowing a phone to remain charged for at least a 16 hour period of heavy use). Accuracy of positional determination is variable depending on city and location within a city but is reported to be approximately 6 meters horizontally and 10 meters vertically.
GPS data collection
Using the BlackBerry Java Development Environment, a background module was created that ran on startup and in the background. It was therefore invisible to the phone user and could not easily be turned off. At 5 minute intervals, the program obtained a "fix" on the participant's location and, through an http call, transmitted the location coordinates, battery level, device ID, and timestamp to a java servlet running on a server dedicated to the study. All data were stored in a MySQL database. Another application monitored the database and sent an SMS message to the phone if no diary data had been sent in the last day. A similar message was sent to the researchers if no diary data had been sent in two days. The program also provided the research assistant with a summary of GPS data for each phone so that she could contact adolescents from whom data was not being regularly sent.
Diary data collection
We instructed the participant to answer several questions each evening before going to bed using a program developed for the Blackberry platform. We were not interested in the answers to the questions but rather whether participants could answer various question types easily using a Blackberry and would reliably answer them on a daily basis. As such, we have devised questions which would elicit different modes of answering (yes/no, multiple choice pull-down menu, free text response). We also asked for comments regarding the phone or the study as an additional process measure. We asked the following questions:
What time did you wake up this morning? (ENTER TIME USING CLOCK OR CALENDAR)
Did you eat lunch today? (YES/NO)
YES → Approximately what time did you eat?
Did you charge the phone last night? (YES/NO)
NO → Did the phone's battery die today?
Did you need to charge the phone during the day today?
Where were you at 4 pm today? (HOME/SCHOOL/FRIEND'S HOUSE/SHOPPING/OUTSIDE IN THE NEIGHBORHOOD/OUTSIDE BUT NOT IN THE NEIGHBORHOOD/OTHER – PLEASE SPECIFY)
What color shirt are you wearing? (KEY ENTER)
Comments or notes on the study or phone: (KEY ENTER)
[The indented questions were skip pattern questions.] These data were transmitted to a secure server in a similar manner as the GPS data when the participant pressed 'send' at the end of the diary entry.
Reimbursement for participation
Adolescents were reimbursed $20 cash following agreeing to participate and $100 cash after completion of the exit interview.
Data interpolation
We anticipated times when cell phones might encounter GPS signal reception interference (e.g. near power substation transformers) or obstruction (being enclosed in thick-walled building). When possible, we interpolated position in the event of lost GPS reception. Interpolation is a method in which missing spatial data are imputed using adjacent valid values. If there were significant temporal discrepancies between serial "fixes" on location, we recorded a "no data" observation.
When more than 5 minutes but less than 1 hour elapsed between measured data points, we imputed interim 5-minute time points. If 2 temporally adjacent points bounding a period of missing data were within 30-meters (the distance used to determine the 2 points were in the same or similar location), we assigned the missing GPS data point to the earlier point. For data points more than 30 meters apart, we imputed in 2 ways. In one case, we used the point which was closer to home and, in another, used the point which was further from home. Since results were very similar for these 2 approaches, we present the data using the closer imputed point. We performed all analyses by varying the time lapse cut-off for imputation (1 hour to 24 hours) and found similar results.