RESPONSE LETTER TO DRS. MARTIN AND ALESSI

To the Editor: Drs. Martin and Alessi raise valid concerns about three Minimum Data Set (MDS) measures. Each concern was already discussed, at some length, in the original report. Strengths and limitations of MDS-based research derive in part from the simplicity of its measures. Debate about the role of the MDS in research is almost as old as the instrument itself but has not prevented many worthwhile studies from being performed. We welcome Martin and Alessi’s new data but cannot agree with the conclusions they make, which seem to reflect some misconceptions about insomnia, its measurement, and consequences of imprecise measurement. Insomnia is a patient complaint not generally defined by any criterion-standard objective measure. Insomnia is ‘‘a complaint of difficulty initiating sleep, difficulty maintaining sleep, or waking up too early or sleep that is chronically nonrestorative or poor in quality.’’ The complaint occurs despite adequate opportunity for sleep and leads to daytime impairment, as defined by any of nine symptoms. In their own study, Martin and Alessi instead defined insomnia according to observed daytime sleepiness, actigraph-estimated sleep time, and an arbitrarily assumed sleep opportunity of 10 hours (letter above) or 8 hours. This is particularly problematic, because many individuals who obtain little nocturnal sleep have no complaints about it, experience no daytime symptoms, or know that to alleviate such symptoms they only need to get more sleep. The MDS may well identify insomnia that is really a complaint and a problem more effectively than actigraphy can. Actigraphy does not record any physiological parameter that defines sleep. Although agreement with laboratory-based polysomnography is excellent in normal subjects, performance is considerably worse in insomniacs, and its use in insomnia has not been fully validated. Substantial differences in numbers of insomniacs identified using the MDS, actigraphy, and the Pittsburgh Sleep Quality Index (PSQI) are not surprising. The PSQI was designed to assess sleep quality and was validated in good and poor sleepers, not strictly insomniacs. The instrument includes items about daytime sleepiness, enthusiasm, bad dreams, and use of hypnotics. It generates a continuous score likely to show a unimodal distribution that would make any dichotomized classification highly sensitive to the chosen cutoff. Martin and Alessi confirmed that each of their own MDS-identified hypnotic users used hypnotics but found that the MDS missed several actual hypnotic users. However, their subjects were selected for hypersomnolence and may have used hypnotics with high frequency, because sleepiness is one of the main side effects. Five of the 6 months for which our study enumerated falls were assessed using the MDS item that showed moderate rather than fair agreement with chart abstraction. Even if all the new data cited by Martin and Alessi are interpreted as they do, the risk raised by some missed cases of insomnia, hypnotic use, and falls is most likely a conservative bias against detection of the reported associations. Had it been possible to identify missed cases of insomnia, hypnotic use, or falls, our conclusions would have been substantially different only if missed cases showed systematically different relationships between the variables under study. Martin and Alessi offer no specific reason to suspect that this occurred. A widely generalizable, longitudinal investigation of 34,163 subjects (more than 94% of those targeted for inclusion) and many potential confounds could not be performed with actigraphy, thorough chart review, or polysomnography. As noted in the report, the conclusions should not be overinterpreted to justify hypnotic use in nursing homes unless the findings are confirmed in future clinical trials. In addition, investigators whose own research efforts are predicated on the ineffectiveness and dangers of hypnotics in older people should not exaggerate the findings. For example, we did not ‘‘conclude that . . . reduction of insomnia using . . . hypnotics may actually reduce the risk for subsequent falls,’’ although we didFand still doFnote that the data ‘‘could support (this) speculation.’’