Detecting Insomnia in Patients With Low Back Pain
Detecting Insomnia in Patients With Low Back Pain
Insomnia is prevalent in patients with LBP. Identifying a measure to detect insomnia that is accurate, brief and easy to administer in this population is essential so that clinicians and researchers can gain a more complete understanding of these common co-morbidities. We evaluated the discriminatory properties of four self-report sleep measures (the Pittsburgh questionnaire, Insomnia index, Epworth scale and the Roland item). These measures were selected because they are widely employed to assess sleep quality and they assess some insomnia symptoms such as sleep disturbance and day-time sleepiness. Our findings suggest that the Pittsburgh questionnaire and the Insomnia index are accurate instruments for screening insomnia in patients with LBP, while the Epworth and the Roland have unacceptably low accuracy.
The Epworth scale and Roland item showed poor accuracy to distinguish between patients with insomnia and those without. This may be because each of these questionnaires only assesses a single criterion of insomnia. The Roland item measures sleep quality (the impact of LBP on a patient's sleep), while the Epworth scale measures day-time sleepiness. Although sleep disturbance and day-time sleepiness are considered to be important traits of insomnia diagnosis, they are inadequate for an insomnia diagnosis.
Although the Pittsburgh questionnaire is designed to assess sleep quality and sleep disturbance, it additionally includes the assessment of day-time impairment related to disturbed sleep: day-time sleepiness and the influence of disturbed sleep on a patient's social activity. Likewise the Insomnia index measures the severity of insomnia and its effect on a patient's day-time functioning. The ROC analysis indicated that both the Pittsburgh questionnaire and Insomnia index were able to accurately distinguish patients with insomnia from those without. These findings suggest that when the goal is to diagnose insomnia a measure that combines several insomnia criteria is more accurate than a single criterion measure. This is supported by the work of Sanford et al. who evaluated the accuracy of the Epworth scale to diagnose insomnia in a community sample. They found that although participants with insomnia reported statistically significantly higher scores of sleepiness than those without insomnia, the ROC curve analysis demonstrated that the scale had poor accuracy to diagnose insomnia. The authors concluded that a measure that combines several insomnia criteria is likely to provide a more accurate diagnosis.
Identifying the optimal cut-off score of a questionnaire is essential to identify patients who require further sleep evaluation and those who may require intervention. Our analysis of sensitivity and specificity for the Pittsburgh questionnaire using the recommended global score of > 5 showed that although the questionnaire yielded a high sensitivity of 100%, the specificity was low at 44%. The optimal cut-off score of > 6, identified from the ROC curve, resulted in a slightly higher specificity of 49% with 100% sensitivity. This finding concurs with the finding of Backhaus et al. who examined the psychometric properties of the Pittsburgh questionnaire for patients with primary insomnia. The authors reported that the cut-off score of > 6 increased the questionnaire's capacity to rule out patients without clinical insomnia. Although both scores yielded a high sensitivity (100%), which indicated that all of the patients with insomnia have been detected, both cut-off scores had low specificity suggesting that half of the patients had been incorrectly identified as having insomnia. Higher sensitivity is an important characteristic of a screening tool in the primary care setting as it is often more important not to miss any patients with the condition than to incorrectly identify some without it. On the other hand, for research purposes high specificity may be required to rule out people without the condition and therefore provide homogenous sample of individuals with a high probably of having the condition. It has been suggested that cut-off score of > 10 for the Pittsburgh questionnaire would increase the questionnaire's accuracy to detect people who have difficulty in initiating and maintaining sleep, which is an important trait of insomnia. Our data showed the cut-off score > 10 resulted in a specificity of 85% with a sensitivity of 50%. The rate of insomnia accordingly declined to 24%, which was closer to the estimate reported by the sleep diary (25%).
The cut-off score of > 14 for the Insomnia index produced optimal discrimination between patients with insomnia and those without (sensitivity 60% and specificity 86%). This result is similar to that of Smith and Trinder who found that a cut-off score of > 14 provided optimal discrimination between people with primary insomnia and the control group. Although this score maximized both sensitivity and specificity to 94%, with an AUC value of 0.97, the study was a case–control design which might have led to overestimation of test accuracy. Savard et al. who examined the discriminatory properties of the Insomnia index in a sample of 1670 patients with cancer, similarly reported that the cut-off score of >14 provided optimal discrimination between patients with insomnia and those without insomnia (sensitivity 51% and specificity 91%).
In clinical practice, likelihood ratios are more useful than sensitivity and specificity for characterising test accuracy. Likelihood ratios indicate likely the test result is in people with the disease compared to those without the disease. The likelihood ratio analysis (Table 4) suggests that the Insomnia index was marginally more accurate than the Pittsburgh questionnaire for detecting patients with insomnia. The analysis demonstrated that a score of 15 points or more on the Insomnia index is 4.43 times more likely in someone with insomnia than someone without insomnia. In addition to its accuracy, the Insomnia index is briefer, easier to administer and to score than the Pittsburgh questionnaire. The Insomnia index therefore appears to be most useful for insomnia assessment of patients with LBP.
This study had limitations that should be addressed. First, although the sleep diary is considered to be a useful tool to diagnose insomnia, its subjective nature is a disadvantage. It has been suggested that patients with sleep problems have the tendency to misperceive their sleep. These patients tend to underestimate their sleep duration and overestimate sleep latency and duration of waking after sleep onset. This limitation potentially influenced sleep variables derived from the sleep diary and therefore the study findings. Second, the study did not investigate sleep-related disorders that may occur along with insomnia, for example sleep apnea and periodic limb movement, which might confound the insomnia assessment. A strength of this study is however that we followed the guidelines for assessing insomnia and the recommendations for designing studies for a diagnostic testing. Additionally, it is the first study to provide guidelines for the optimal assessment of insomnia in patients with LBP using self-reported questionnaires. Finally, the inclusion of patients with LBP who were seeking health care for their LBP, as well as those who were not, increases the representativeness of the sample.
Discussion
Insomnia is prevalent in patients with LBP. Identifying a measure to detect insomnia that is accurate, brief and easy to administer in this population is essential so that clinicians and researchers can gain a more complete understanding of these common co-morbidities. We evaluated the discriminatory properties of four self-report sleep measures (the Pittsburgh questionnaire, Insomnia index, Epworth scale and the Roland item). These measures were selected because they are widely employed to assess sleep quality and they assess some insomnia symptoms such as sleep disturbance and day-time sleepiness. Our findings suggest that the Pittsburgh questionnaire and the Insomnia index are accurate instruments for screening insomnia in patients with LBP, while the Epworth and the Roland have unacceptably low accuracy.
The Epworth scale and Roland item showed poor accuracy to distinguish between patients with insomnia and those without. This may be because each of these questionnaires only assesses a single criterion of insomnia. The Roland item measures sleep quality (the impact of LBP on a patient's sleep), while the Epworth scale measures day-time sleepiness. Although sleep disturbance and day-time sleepiness are considered to be important traits of insomnia diagnosis, they are inadequate for an insomnia diagnosis.
Although the Pittsburgh questionnaire is designed to assess sleep quality and sleep disturbance, it additionally includes the assessment of day-time impairment related to disturbed sleep: day-time sleepiness and the influence of disturbed sleep on a patient's social activity. Likewise the Insomnia index measures the severity of insomnia and its effect on a patient's day-time functioning. The ROC analysis indicated that both the Pittsburgh questionnaire and Insomnia index were able to accurately distinguish patients with insomnia from those without. These findings suggest that when the goal is to diagnose insomnia a measure that combines several insomnia criteria is more accurate than a single criterion measure. This is supported by the work of Sanford et al. who evaluated the accuracy of the Epworth scale to diagnose insomnia in a community sample. They found that although participants with insomnia reported statistically significantly higher scores of sleepiness than those without insomnia, the ROC curve analysis demonstrated that the scale had poor accuracy to diagnose insomnia. The authors concluded that a measure that combines several insomnia criteria is likely to provide a more accurate diagnosis.
Identifying the optimal cut-off score of a questionnaire is essential to identify patients who require further sleep evaluation and those who may require intervention. Our analysis of sensitivity and specificity for the Pittsburgh questionnaire using the recommended global score of > 5 showed that although the questionnaire yielded a high sensitivity of 100%, the specificity was low at 44%. The optimal cut-off score of > 6, identified from the ROC curve, resulted in a slightly higher specificity of 49% with 100% sensitivity. This finding concurs with the finding of Backhaus et al. who examined the psychometric properties of the Pittsburgh questionnaire for patients with primary insomnia. The authors reported that the cut-off score of > 6 increased the questionnaire's capacity to rule out patients without clinical insomnia. Although both scores yielded a high sensitivity (100%), which indicated that all of the patients with insomnia have been detected, both cut-off scores had low specificity suggesting that half of the patients had been incorrectly identified as having insomnia. Higher sensitivity is an important characteristic of a screening tool in the primary care setting as it is often more important not to miss any patients with the condition than to incorrectly identify some without it. On the other hand, for research purposes high specificity may be required to rule out people without the condition and therefore provide homogenous sample of individuals with a high probably of having the condition. It has been suggested that cut-off score of > 10 for the Pittsburgh questionnaire would increase the questionnaire's accuracy to detect people who have difficulty in initiating and maintaining sleep, which is an important trait of insomnia. Our data showed the cut-off score > 10 resulted in a specificity of 85% with a sensitivity of 50%. The rate of insomnia accordingly declined to 24%, which was closer to the estimate reported by the sleep diary (25%).
The cut-off score of > 14 for the Insomnia index produced optimal discrimination between patients with insomnia and those without (sensitivity 60% and specificity 86%). This result is similar to that of Smith and Trinder who found that a cut-off score of > 14 provided optimal discrimination between people with primary insomnia and the control group. Although this score maximized both sensitivity and specificity to 94%, with an AUC value of 0.97, the study was a case–control design which might have led to overestimation of test accuracy. Savard et al. who examined the discriminatory properties of the Insomnia index in a sample of 1670 patients with cancer, similarly reported that the cut-off score of >14 provided optimal discrimination between patients with insomnia and those without insomnia (sensitivity 51% and specificity 91%).
In clinical practice, likelihood ratios are more useful than sensitivity and specificity for characterising test accuracy. Likelihood ratios indicate likely the test result is in people with the disease compared to those without the disease. The likelihood ratio analysis (Table 4) suggests that the Insomnia index was marginally more accurate than the Pittsburgh questionnaire for detecting patients with insomnia. The analysis demonstrated that a score of 15 points or more on the Insomnia index is 4.43 times more likely in someone with insomnia than someone without insomnia. In addition to its accuracy, the Insomnia index is briefer, easier to administer and to score than the Pittsburgh questionnaire. The Insomnia index therefore appears to be most useful for insomnia assessment of patients with LBP.
This study had limitations that should be addressed. First, although the sleep diary is considered to be a useful tool to diagnose insomnia, its subjective nature is a disadvantage. It has been suggested that patients with sleep problems have the tendency to misperceive their sleep. These patients tend to underestimate their sleep duration and overestimate sleep latency and duration of waking after sleep onset. This limitation potentially influenced sleep variables derived from the sleep diary and therefore the study findings. Second, the study did not investigate sleep-related disorders that may occur along with insomnia, for example sleep apnea and periodic limb movement, which might confound the insomnia assessment. A strength of this study is however that we followed the guidelines for assessing insomnia and the recommendations for designing studies for a diagnostic testing. Additionally, it is the first study to provide guidelines for the optimal assessment of insomnia in patients with LBP using self-reported questionnaires. Finally, the inclusion of patients with LBP who were seeking health care for their LBP, as well as those who were not, increases the representativeness of the sample.
Source...