BACKGROUND: Although multiple instruments for measuring sleep problems are available, data to guide their interpretation in patient care settings are limited. The aims of this study were to assess known-groups validity and to determine a clinically important difference (CID) for the Medical Outcomes Study (MOS) Sleep Problems Index (SPI6), a self-administered six-item instrument that has undergone extensive psychometric evaluation (Hays, 2005). METHODS: We analyzed responses to sleep questions and the SF-36 Health Survey for 3445 pts who participated in the MOS at baseline (T0) and 1814 pts who completed follow-up at 24 mos (T24). The SPI6 uses a six-point Likert scale to measure sleep problems over the prior four weeks, and is scored on a 0-100 range (higher scores indicate greater dysfunction). To assess known-groups validity, we used multiple linear regression to compare SPI6 scores in 1) patients with and without sleep-related comorbidities, and 2) patients with and without insomnia, based on DSM-IV criteria. In addition, we benchmarked changes in the SPI6 to a concurrent five-point decrement in SF-36 scores between T0 and T24, which corresponds to a small effect size (ES~0.20) in the MOS sample (Samsa, 1999). We adjusted for demographics, body mass index, health habits, physician-identified chronic conditions, and study location. RESULTS: Adjusted SPI6 scores were significantly higher in conditions associated with poor sleep: back pain (30.8 vs 26.5, ES=.24), CHF (33.2 vs 27.9, ES=.29), and depression (40.8 vs 26.1, ES=.81). Patients with mild and severe insomnia had higher adjusted SPI6 scores than those without insomnia (33.8 and 50.6 vs 16.9, respectively). A 15-point increase in SPI6 score was associated with an adjusted decrement of at least five points in all SF-36 domains (see Table). CONCLUSIONS: The SPI6 scale discriminates effectively between patients with and without chronic conditions associated with poor sleep. A 15-point change in the SPI6 represents a clinically important difference that is comparable in magnitude to the sleep dysfunction associated with mild insomnia. These findings should be confirmed by evaluating the CID relative to patient-perceived changes in sleep quality.

ASSOCIATION BETWEEN CID FOR SPI6 AND SF-36 CHANGE SCORES

SF-36 domain		2-yr Δ score (95% CI)
PHYSICAL		-8.8 (-10.2, -7.4)
PAIN		-7.2 (-8.7, -5.6)
GEN HEALTH		-7.1 (-8.5, -5.8)
VITALITY		-9.0 (-10.2, -7.8)
SOCIAL		-9.9 (-11.5, -8.4)
MENTAL		-5.8 (-6.8, -4.8)