Kurisutina
Library/Carrying on

Sliwinski et al. · 2009

Intraindividual change and variability in daily stress processes: findings from two measurement-burst diary studies

Psychology and Aging 24(4): 828–840 (December 2009), DOI 10.1037/a0017925. Read as the NIH author manuscript (PMC2857711; accepted, not copyedited; PMC text mining permitted, not open access). Penn State and Syracuse. Provenance: papers/carry_on/sliwinski2009_stress_bursts.provenance.json.

What was read

All 622 lines of the text derived from the PMC XML: abstract, introduction, method, results, discussion, limitations, funding, all references, three figure captions and every cell of Tables 1–4. The converter checks that the text keeps every non-whitespace character of the XML. Figures 1–3 are images and were not inspected; Figure 3 carries the age-stratum stability coefficients, of which the text gives four.

Question

Is a person's emotional reactivity to daily stress a stable trait, does it change with age, and does it vary within the person from one period to the next?

  • Reactivity = the within-person slope of end-of-day negative affect (NA) on that day's summed stressor severity. The authors note that daily data cannot show that stress precedes mood; "reactivity" assumes it.

Method

  • Measurement-burst design: short runs of daily measurements, repeated months or years apart. Three levels: days within bursts within persons.
  • NSDE (National Study of Daily Experiences, a MIDUS subsample):
    • 671 adults, age 24–75 at baseline (mean 47); 58% women; 94% white.
    • 8 evening phone interviews per burst; two bursts about 10 years apart (1996–97 and about 2006). 10,273 of 10,736 possible days.
    • NA: 4 K6 items, α .81. Stressors: 5 DISE stems with severity 0–3, summed per day.
    • Neuroticism: 4 MIDUS adjectives, α .74.
  • CHAP (Cognition, Health and Aging Project):
    • 116 older adults, age 66–95 (mean 80); 72% women.
    • 6 days within 12 per burst; bursts every 6 months for two years (up to 5 bursts, 3,010 days); 78% completed all 5.
    • NA: 5-adjective checklist, α .85. Stressors: a DISE variant.
    • Perceived Stress Scale (PSS, 14 items, past month), once per burst.
  • Model: a three-level linear model with day-, burst- and person-level stress (grand-mean centred), burst, interactions, and random slopes. Stability in NSDE: separate models per burst, person-specific random reactivity effects correlated across bursts.

Results

  • Mean reactivity rises with age, within persons.
    • NSDE: burst × day stress +0.04 (p < .01) over 10 years, on a day-stress slope of 0.14.
    • CHAP: +0.06 per 6-month burst (p < .01), on 0.13.
    • Baseline age did not moderate the rate, and cross-sectional age differences in reactivity were absent in both.
    • Mean NA also rose within persons (NSDE +0.09, CHAP +0.24 per burst).
  • Reactivity is partly a stable trait.
    • NSDE 10-year retest correlation of person-specific reactivity: .37 (p < .01).
    • By age stratum: 30s .62, 40s .49, 60s .21, 70s .18. The discussion rounds these as .49–.62 for the 30s and 40s, the .30s for 50–65, and about .20 for 60–75.
    • Neuroticism × day stress +0.12 (p < .01): higher neuroticism, higher average reactivity. Neuroticism did not predict change in reactivity or in NA.
  • In old age, most variance in reactivity is within the person.
    • CHAP random-slope variances: between persons .03, between bursts within persons .08. So 27% of the variance in reactivity is stable across two years and 73% varies from one 6-month burst to the next.
    • The authors: a person's reactivity "during a given week reflects to greater degree transient characteristics specific to that particular week than enduring characteristics of the person".
  • The within-person variation is systematic.
    • Within-person PSS × day stress: +0.02 (p < .01). A person reacts more in bursts when they rate their life as more overwhelming than usual.
    • A person's average PSS did not moderate reactivity (p = .80).
    • Within-person PSS also raised the odds of reporting any stressor (OR 1.10), not the severity ratings.
  • Kindling (sensitisation from cumulative stress) is not supported: no acceleration with age, no neuroticism × burst × stress effect, no between-person PSS effect.

Limits

  • Two studies with different samples, instruments and burst spacing. NSDE has only two bursts, so its within-person variability in reactivity was fixed to zero. CHAP is small (116), old and has no personality measure.
  • The stability coefficient correlates model-based person estimates from separate fits. It is not corrected for shrinkage or estimation error (my inference; the paper does not discuss this).
  • Period effects (e.g. after 9/11) could affect the 10-year change. Daily data cannot establish that stress precedes NA.
  • Inconsistencies found:
    • The text cites "Table 1 (Model B)" and "Table 1 (Model C)" for NSDE models that are in Table 2. It cites "Table 2 (Model C)" for the CHAP PSS model, which is in Table 4.
    • Table 4 labels the between-person PSS term "PSS_BP × Burst × Day Stress". The text describes a PSS_BP × day-stress interaction (p = .80).
    • Some Table 4 random-effect SEs exceed their estimates yet carry "**" (Model B level-3 burst 0.11, SE .150; Model C level-3 day stress 0.03, SE .051). The text gives p = .01 for the person-level reactivity variance, and the table marks it "*".
    • For CHAP Model B the text says baseline age "did not significantly interact with any other variable". Table 4 has burst × age +0.03 (SE .007, **). The text probably means the stress terms.

What it means for Kurisutina

  • Question 1: a person's way of reacting exists, but it is only partly a trait.
    • This is direct evidence, for daily stress and affect, that reactivity has a stable personal component: r = .37 over 10 years, .5–.6 in the 30s and 40s.
    • In old age, three quarters of the variance in reactivity comes and goes over months with the person's current load.
    • So the slot needs two things: the person's usual reactivity, and a slow-changing state (how overwhelmed they are now) that scales it. Contents plus a fixed "way of changing" parameter would miss most of the within-person part.
  • Traits carry some information here, unlike for major events.
    • Neuroticism predicted average daily reactivity. Yap and Anusic found trait moderation of reactions to major events scattered and mostly null.
    • A plausible difference: many stressor-days per person give a reliable slope; one major event gives a single noisy reaction.
    • For LISS T1, where a reaction is one change score, the noise floor will be high. The comparison with the reliability-only correlation and with non-event intervals is essential, and a null may be a power failure.
  • What to collect from a real person. A measurement-burst design (short daily diaries, repeated months apart) is what separates stable reactivity from current state. One block of interviews cannot: a single burst's reactivity would be mostly the state of that period. This is a concrete proposal for the elicitation protocol and for the Q1/Q3 tests with the user's own data. It is not started.
  • A replica that carries on for years should change its mean reactivity too. Reactivity and NA rise with age within persons. Over long horizons, "stays the same" is not the right target for affect.

Cross-references

  • summaries/carry_on/yap2012_personality_moderation.md, summaries/carry_on/anusic2014_hilda_personality.md: trait moderation of reactions to major events (mostly null).
  • summaries/carry_on/eckstein2022_context.md: individual parameters are context-bound.
  • summaries/carry_on/nichols2019_false_memory_tasks.md: stable within a kind, weak across kinds (same pattern for memory distortion).
  • summaries/carry_on/luhmann2014_its_about_time.md: minor events; combine short and long longitudinal designs.
  • docs/research/liss_q1_design.md: T1 noise floor.

This summary is our record of the paper, written after reading the full text and published as written; links into our own repository have been removed.