An international prospective cohort study of mobile phone users and health (COSMOS): Factors affecting validity of self-reported mobile phone use

2017 · Unclear · Moderate evidence

← Research stream

UnclearModerate evidencePaper 2311

Study snapshot

Publicationmanual · 2017

Study typecohort

Reported directionUnclear

Evidence ratingModerate

Abstract

An international prospective cohort study of mobile phone users and health (COSMOS): Factors affecting validity of self-reported mobile phone use Toledano MB, Auvinen A, Tettamanti G, Cao Y, Feychting M, Ahlbom A, Fremling K, Heinävaara S, Kojo K, Knowles G, Smith RB, Schüz J, Johansen C, Poulsen AH, Deltour I, Vermeulen R, Kromhout H, Elliott P, Hillert L. An international prospective cohort study of mobile phone users and health (COSMOS): Factors affecting validity of self-reported mobile phone use. Int J Hyg Environ Health. 2017 Sep 20. pii: S1438-4639(17)30321-8. doi: 10.1016/j.ijheh.2017.09.008. Highlights • Agreement for self-reported phone use was higher for call duration than frequency. • Subjects tended to underestimate rather than overestimate mobile phone use. • Agreement for self-reported call frequency was higher in men and older subjects. • Subjects who reported symptoms were more likely to overestimate low call duration. Abstract This study investigates validity of self-reported mobile phone use in a subset of 75 993 adults from the COSMOS cohort study. Agreement between self-reported and operator-derived mobile call frequency and duration for a 3-month period was assessed using Cohen's weighted Kappa (κ). Sensitivity and specificity of both self-reported high (≥10 calls/day or ≥4h/week) and low (≤6 calls/week or <30min/week) mobile phone use were calculated, as compared to operator data. For users of one mobile phone, agreement was fair for call frequency (κ=0.35, 95% CI: 0.35, 0.36) and moderate for call duration (κ=0.50, 95% CI: 0.49, 0.50). Self-reported low call frequency and duration demonstrated high sensitivity (87% and 76% respectively), but for high call frequency and duration sensitivity was lower (38% and 56% respectively), reflecting a tendency for greater underestimation than overestimation. Validity of self-reported mobile phone use was lower in women, younger age groups and those reporting symptoms during/shortly after using a mobile phone. This study highlights the ongoing value of using self-report data to measure mobile phone use. Furthermore, compared to continuous scale estimates used by previous studies, categorical response options used in COSMOS appear to improve validity considerably, most likely by preventing unrealistically high estimates from being reported. ncbi.nlm.nih.gov Excerpts The target population for COSMOS was adult mobile phone users, aged 18-69 years, in 5 European countries: Denmark, Finland, the Netherlands, Sweden and the UK, and recently a 6th cohort has been initiated in France. This analysis focuses on participants recruited into the study in Finland, Sweden and the UK between 2007 and 2010. Participants were asked to report frequency and duration of mobile phone voice calls for the preceding three months via the following two questions: "Over the last three months, how often did you talk on a mobile phone?" with the response options: <1 call per week (Finland and Sweden only; the UK web-based questionnaire filtered out these respondents in a previous question), 1-6 calls per week, 1-9 calls per day, ≥10 calls per day. "Over the last three months, on average, how much time per week did you spend talking on a mobile phone?" with the response options: <5 min, 5-29 min, 30-59 min, 1-3 h, 4-6 h, >6 h. Participants were asked if they experienced symptoms ("no symptoms, headache, dizziness, numbness in hands, nausea, hearing loss, tinnitus/ringing sound in ear, warming sensation on face and/or ear") whilst using, or shortly after using, a mobile phone. Approximately 20% of participants spent at least 4 h per week on calls and/or made at least 10 calls per day, and were thus defined as having high mobile phone use (Table 1). We found that a considerable proportion of respondents misclassified their mobile phone use (approximately 60% and 40% for call duration and frequency, respectively) (Table 2, Supplementary Tables 1 & 2). Approximately a third of the participants underestimated their mobile phone call duration and frequency. The proportion of participants overestimating mobile phone use was much lower (23% for duration and 5% for call frequency among one- phone users) (Table 2). Agreement between self-reported and operator call duration was significantly lower among those who reported experiencing symptoms whilst (or shortly after) using a mobile phone (κ = 0.44 (95% CI: 0.43, 0.46)) compared with those without symptoms (κ = 0.50 (95% CI: 0.49, 0.50)), primarily because those with symptoms were more likely to overestimate low call duration (sensitivity = 65% (95% CI: 62%, 67%) vs. 78% (95% CI: 77%, 79%) for those with and without symptoms respectively) (Table 4). A similar pattern was observed for call frequency, but the differences were smaller. In this largest validation study to date, we found fair to moderate agreement between self-reported and operator-derived data on mobile phone use. The sensitivity of self-report was generally high for correctly identifying those with the smallest amount of mobile phone use, but lower for identifying heavy mobile phone use, in line with our observation that respondents in this study were more likely to underestimate than overestimate their mobile phone use. Our findings demonstrate that those who experience symptoms when using a mobile phone are more likely to overestimate light mobile phone use, particularly call duration, compared to those without symptoms. This suggests that an individual's experience and/or perception of their health may influence the self-reporting of mobile phone use, likely affecting the validity of such exposure assessments. More specifically, it is possible that rumination bias (a form of information bias), whereby those with symptoms overestimate (consciously or subconsciously) their phone use in an effort to explain their symptoms, could be occurring in this subset of individuals. This finding has potential implications for the interpretation of previous cross-sectional studies investigating associations between mobile phone use and the symptoms reported here (Mortazavi et al., 2007; Soderqvist et al., 2008; Sandstrom et al., 2001). Overestimation of mobile phone use among those who report such symptoms would likely bias cross-sectional risk estimates away from the null, even if a true association does not exist (Armstrong, 1998), thus potentially distorting any observed associations.

Machine Enhanced Logic

AI evidence extraction

Structured fields below were generated from the paper’s abstract and metadata to make the evidence easier to inspect. Automated extraction may be incomplete or incorrect; verify important details against the original paper.

Study typecohort

Effect directionunclear

PopulationAdult mobile phone users (subset of COSMOS cohort) recruited in Finland, Sweden and the UK

Sample size75993

ExposureRF · Self-reported vs operator-derived mobile call frequency and call duration over the preceding 3-month period

Evidence strengthModerateConfidence: 78% · Peer reviewed: yes

Main findings

Among one-phone users, agreement between self-reported and operator-derived use was fair for call frequency (κ=0.35, 95% CI 0.35–0.36) and moderate for call duration (κ=0.50, 95% CI 0.49–0.50) over a 3-month period. Self-report showed high sensitivity for identifying low use (87% for low call frequency; 76% for low call duration) but lower sensitivity for high use (38% for high call frequency; 56% for high call duration), consistent with underestimation being more common than overestimation. Validity was lower in women, younger age groups, and participants reporting symptoms during/shortly after mobile phone use; those reporting symptoms were more likely to overestimate low call duration.

Outcomes measured

Validity/agreement of self-reported mobile phone call frequency vs operator data (Cohen's weighted kappa)
Validity/agreement of self-reported mobile phone call duration vs operator data (Cohen's weighted kappa)
Sensitivity and specificity of self-reported high vs low mobile phone use categories compared with operator data
Differences in validity by sex, age group, and symptom reporting during/shortly after phone use

Limitations

Frequency (MHz) and SAR were not reported in the abstract/excerpts
Validity assessment focused on voice calls and a 3-month window; other phone uses not described
Agreement metrics reported primarily for users of one mobile phone; results for multiple-phone users not detailed here

View raw extracted JSON

{
  "study_type": "cohort",
  "exposure": {
    "band": "RF",
    "source": "mobile phone",
    "frequency_mhz": null,
    "sar_wkg": null,
    "duration": "Self-reported vs operator-derived mobile call frequency and call duration over the preceding 3-month period"
  },
  "population": "Adult mobile phone users (subset of COSMOS cohort) recruited in Finland, Sweden and the UK",
  "sample_size": 75993,
  "outcomes": [
    "Validity/agreement of self-reported mobile phone call frequency vs operator data (Cohen's weighted kappa)",
    "Validity/agreement of self-reported mobile phone call duration vs operator data (Cohen's weighted kappa)",
    "Sensitivity and specificity of self-reported high vs low mobile phone use categories compared with operator data",
    "Differences in validity by sex, age group, and symptom reporting during/shortly after phone use"
  ],
  "main_findings": "Among one-phone users, agreement between self-reported and operator-derived use was fair for call frequency (κ=0.35, 95% CI 0.35–0.36) and moderate for call duration (κ=0.50, 95% CI 0.49–0.50) over a 3-month period. Self-report showed high sensitivity for identifying low use (87% for low call frequency; 76% for low call duration) but lower sensitivity for high use (38% for high call frequency; 56% for high call duration), consistent with underestimation being more common than overestimation. Validity was lower in women, younger age groups, and participants reporting symptoms during/shortly after mobile phone use; those reporting symptoms were more likely to overestimate low call duration.",
  "effect_direction": "unclear",
  "limitations": [
    "Frequency (MHz) and SAR were not reported in the abstract/excerpts",
    "Validity assessment focused on voice calls and a 3-month window; other phone uses not described",
    "Agreement metrics reported primarily for users of one mobile phone; results for multiple-phone users not detailed here"
  ],
  "evidence_strength": "moderate",
  "confidence": 0.78,
  "peer_reviewed_likely": "yes",
  "keywords": [
    "COSMOS",
    "mobile phone use",
    "self-report",
    "operator data",
    "validation study",
    "Cohen's kappa",
    "sensitivity",
    "specificity",
    "exposure misclassification",
    "symptoms"
  ],
  "suggested_hubs": [
    {
      "slug": "exposure-assessment",
      "weight": 0.95,
      "reason": "Primary aim is validation of self-reported mobile phone use against operator-derived records in a large cohort subset."
    }
  ]
}

AI can be wrong. Always verify against the paper.

AI-extracted fields are generated from the abstract and metadata and may be incomplete or incorrect. This material is for informational purposes and is not medical advice.

Identifiers and source

DOI: 10.1016/j.ijheh.2017.09.008
PubMed: Not supplied

Read the original paper →

Related papers

← Back to the Research Library