Validation And Minimum Important Difference Of The Pittsburgh Sleep Quality Index

BackgroundThe changes of subjective indicators to determine treatment effects have been one of the important methods of the evaluation of curative effect of traditional Chinese medicine(TCM). A grown number of diseases, especially chronic diseases of subjective experience, apply subjective indicators for the clinical evaluation. The reasonable application of these indicators in order to reliabilitily, validitily and accurately improve the evaluation of clinical outcomes has accumulated professinals’attention in the area of medical care during recent years. Take insomnia as an example, a number of scales have been developed to assess insomnia from different aspects. However, when researchers are selecting primary or secondary effect indicators for their studies, it is the commonly used measurement easier to be selected than the proper ones. Therefore, further studies on the properties of instruments are desired to provide enough information in helping researchers selecting suitable assessment tools for their studies. In addition, TCM for insomnia are known to have quite a long history and as we can find corresponding descriptions in the "huangdi neijing"(an acient book of TCM). TCM has been playing an importment role in the treatment of insomnia and has brought significant attentions. Judging from the results of recently completed clinical trials and systematic reviews, TCM has shown encouraging effect in the treatment of insomnia. However, from which aspect can the effectiveness of TCM be sensitively or specificly reflected, or among the existed instruments, which can better reflect the effects of TCM in regards of the properties of reliability, validity and accuracy, are still unknown. Further more, most of the scales applied in assessments of insomnia are subjective measurements, and belong to subjective evaluation. How to regulate evaluation process and reasonably interprctate evaluation results, especially for the intervention of TCM, still needs further research. Object iveTo assess the reliability, validity, responsiveness and minimum important difference(MID) of the Pittsburgh Sleep Quality Index(PSQI).DesignThere are two parts of our study, one is data analysis of existed clinical trial and another is a cross-sectional survey.SettingThree out-patient departments of the second affiliated hospital of Guangzhou University of Chinese Medicine located in different areas of Guangzhou.ParticipantsFrom July2011to October2011,544convenience samples were recruited from three different settings. The selected cases were those meeting the inclusive criteria as follows:patients with insomnia:(1)18to65years old;(2) education level is above primary;(3) have difficulties in falling asleep and in maintaining sleep; those who has the above symptoms more than3nights a week, for at least1month; those who has insomnia caused distress or impairment in social contacts or work. Non-insomnia participants:(1) not meet the symptom criteria of insomnia (see above);(2) be18to65years old;(3) education level is above primary. Exclusion criteria:those who unwilling or unable to cooperate the survey.Methods(1) All participants were assessed using PSQI. General socioeconomic characteristics were recorded.(2) Split-half reliability was evaluated by Spearman--Brown formula. Cronbach’s alpha coefficient and item-total correlation were applied to evaluate internal consistency. We used confirmatory factor analysis (CFA) to assess the construct validity. In order to exhibit the convergent validity, Spearman correlation coefficient between PSQI score and1SI score were calculated. By analyzing the correlations between the PSQI and questionnaire type, interview date and interview department, the discriminant validity were evaluated. In addition, For individuals who with or without insomnia, we assessed the discriminant validity by ROC curve.(3) Using anchor-based methods and distribution-based methods to evaluate the MID of the PSQI.(4) Data obtained from a clinical trial (ID:200HBAI13B062) were included to assess the responsiveness of the PSQI. Patients were randomly allocated to treatment or control group. Patients in the treatment group were treated by standardized herbal decoction and acupuncture, those who were in the control group were treated with decoction placebo and sham acupuncture. All Participants were educated with sleep hygiene instruction, PSQI were completed at baseline, day29th in the treatment period and day28th at follow-up Statistical analysis was conducted with paired t-test, effect sizes and standardized response mean to assess the responsiveness of the PSQI.ResultsThe PSQI had good recovery and response rate. A test-retest reliability (24hours interval) was done among29participants. Intra-class correlation coefficient of the PSQI was0.994. Split-half reliability of the PSQI was0.824. Overall the Cronbach’s alpha coefficient was0.845. Confirmatory factor analysis showed that the fitting indicators for the (?)2/df, RMSEA, NNFI, CFI and GFI were4.83,0.09,0.96,0.98and0.97. The correlation between the PSQI and the Insomnia Severity Index (ISI) was0.842. The correlations between the PSQI and questionnaire type, interview date and interview department were-0.046,0.276and-0.331. Grouping by individuals with or without insomnia, the area under the ROC(receiver operator characteristic curve) curve for the PSQI was0.944(95%CI0.925-0.963).Based on the anchor-based methods suggested that the MID of the PSQI was estimated to be1.63and3.98, respectively. That is, before and after treatment, the patients perceived a PSQI improvement of21%as beneficial or deterioration of50%as harmful. Using the effect size and the standard error of measurement approach, the MID of PSQI was estimated as0.76and1.80, respectively.The differences of the PSQI domains and total score were significantly different by comparing the scores before with those after the treatment (29days)(P<0.05). The fffect size of PSQI total score was0.99and all the seven domains ranged from0.29to1.32. The standardized response mean of the PSQI total score was0.67and those for the seven domains were ranging from0.20to1.01. Di fferences for the PSQI domains and total score were significant at day29th and day56th (P<0.05) by comparing the scores before with those after the treatment. Effect size of the PSQI total score was0.34and for the seven domains were lower than0.5. The standardized response mean of the PSQI total score was0.51and the domains’standardized reponse moan were lower than0.5.Conelusions(1) The PSQI had good reliability and validity, and could be used for sleep quality evaluation for patients with insomnia.(2)The PSQI and its domains had sufficient responsiveness especially at the29th day.(3) Anchor-based and distribution-based methods can app] ied for MID determi nation, and these approaches may be used in other areas regarding clinical practice and scale research.