Susan M. Shortreed

Cognitive-behavioral treatment for comorbid insomnia and osteoarthritis pain in primary care: the lifestyles randomized controlled trial.

To assess whether older persons with osteoarthritis (OA) pain and insomnia receiving cognitive–behavioral therapy for pain and insomnia (CBT‐PI), a cognitive–behavioral pain coping skills intervention (CBT‐P), and an education‐only control (EOC) differed in sleep and pain outcomes.

Short-term improvement in insomnia symptoms predicts long-term improvements in sleep, pain, and fatigue in older adults with comorbid osteoarthritis and insomnia

Summary Sleep, pain, and fatigue were longitudinally examined in 367 osteoarthritic older adults. Two‐month sleep improvement predicted 18‐month improvements in sleep, pain, and fatigue. ABSTRACT In a primary care population of 367 older adults (aged ≥60 years) with osteoarthritis (OA) pain and insomnia, we examined the relationship between short‐term improvement in sleep and long‐term sleep, pain, and fatigue outcomes through secondary analyses of randomized controlled trial data. Study participants, regardless of experimental treatment received, were classified either as improvers (≥30% baseline to 2‐month reduction on the Insomnia Severity Index [ISI]) or as nonimprovers. After controlling for treatment arm and potential confounders, improvers showed significant, sustained improvements across 18 months compared with nonimprovers in pain severity (P < 0.001, adjusted mean difference = −0.51 [95% CI: −0.80, −0.21), arthritis symptoms (P < 0.001, 0.63 [0.26, 1.00]), and fear avoidance (P = 0.009, −2.27 [−3.95, −0.58]) but not in catastrophizing or depression. Improvers also showed significant, sustained improvements in ISI (P < 0.001, −3.03 [−3.74, −2.32]), Pittsburgh Sleep Quality Index Total (P < 0.001, −1.45 [−1.97, −0.93]) and general sleep quality (P < 0.001, −0.28 [−0.39, −0.16]) scores, Flinders Fatigue Scale (P < 0.001, −1.99 [−3.01, −0.98]), and Dysfunctional Beliefs About Sleep Scale (P = 0.037, −2.44 [−4.74, −0.15]), but no improvements on the Functional Outcomes of Sleep Questionnaire or the Epworth Sleepiness Scale. We conclude that short‐term (2‐month) improvements in sleep predicted long‐term (9‐ and 18‐month) improvements for multiple measures of sleep, chronic pain, and fatigue. These improvements were not attributable to nonspecific benefits for psychological well‐being, such as reduced depression. These findings are consistent with benefits of improved sleep for chronic pain and fatigue among older persons with osteoarthritis pain and comorbid insomnia if robust improvements in sleep are achieved and sustained. Trial Registration: ClinicalTrials.gov Identifier: NCT01142349.

Informing sequential clinical decision-making through reinforcement learning: an empirical study

This paper highlights the role that reinforcement learning can play in the optimization of treatment policies for chronic illnesses. Before applying any off-the-shelf reinforcement learning methods in this setting, we must first tackle a number of challenges. We outline some of these challenges and present methods for overcoming them. First, we describe a multiple imputation approach to overcome the problem of missing data. Second, we discuss the use of function approximation in the context of a highly variable observation set. Finally, we discuss approaches to summarizing the evidence in the data for recommending a particular action and quantifying the uncertainty around the Q-function of the recommended policy. We present the results of applying these methods to real clinical trial data of patients with schizophrenia.

The Effect of Program Design on Engagement With an Internet-Based Smoking Intervention: Randomized Factorial Trial

Background Participant engagement influences treatment effectiveness, but it is unknown which intervention design features increase treatment engagement for online smoking cessation programs. Objective We explored the effects of 4 design features (ie, factors) on early engagement with an Internet-based, motivational smoking cessation program. Methods Smokers (N=1865) were recruited from a large health care organization to participate in an online intervention study, regardless of their interest in quitting smoking. The program was intended to answer smokers’ questions about quitting in an effort to motivate and support cessation. Consistent with the screening phase in the multiphase optimization strategy (MOST), we used a 2-level, full-factorial design. Each person was randomized to 1 of 2 levels of each factor, including message tone (prescriptive vs motivational), navigation autonomy (dictated vs not), proactive email reminders (yes vs no), and inclusion of personally tailored testimonials (yes vs no). The effects of each factor level on program engagement during the first 2 months of enrollment were compared, including number of visits to the website resulting in intervention content views (as opposed to supplemental content views), number of intervention content areas viewed, number of intervention content pages viewed, and duration of time spent viewing this content, as applicable to each factor. Results Adjusting for baseline readiness to quit, persons who received content written in a prescriptive tone made the same number of visits to the website as persons receiving content in a motivational tone, but viewed 1.17 times as many content areas (95% CI 1.08-1.28; P<.001) and 1.15 times as many pages (95% CI 1.04-1.28; P=.009). Time spent viewing materials did not differ among groups (P=.06). Persons required to view content in a dictated order based on their initial readiness to quit made the same number of visits as people able to freely navigate the site, but viewed fewer content areas (ratio of means 0.80, 95% CI 0.74-0.87; P<.001), 1.17 times as many pages (95% CI 1.06-1.31; P=.003), and spent 1.37 times more minutes online (95% CI 1.17-1.59; P<.001). Persons receiving proactive email reminders made 1.20 times as many visits (95% CI 1.09-1.33; P<.001), viewed a similar number of content areas as persons receiving no reminders, viewed 1.58 times as many pages (95% CI 1.48-1.68; P<.001), and spent 1.51 times as many minutes online (95% CI 1.29-1.77; P<.001) as those who did not receive proactive emails. Tailored testimonials did not significantly affect engagement. Conclusions Using a prescriptive message tone, dictating content viewing order, and sending reminder emails each resulted in greater program engagement relative to the contrasting level of each experimental factor. The results require replication, but suggest that a more directive interaction style may be preferable for online cessation programs. Trial Registration clinicaltrials.gov NCT00992264; http://clinicaltrials.gov/ct2/show/NCT00992264 (Archived by WebCite at http://www.webcitation.org/6F7H7lr3P)

Optimizing prediction of back pain outcomes.

&NA; A limited set of measures obtained from patients initiating primary care for back pain demonstrated excellent ability to identify those who had continued pain associated with moderate to severe activity interference 4 months later. &NA; An accurate means of identifying patients at high risk for chronic disabling pain could lead to more cost‐effective care, with more intensive interventions targeted to those likely to benefit most. The Chronic Pain Risk Score is a tool developed to predict risk for chronic pain. The aim of this study was to examine whether its predictive ability could be enhanced by: (1) improved measures of the constructs it assesses (Improved Chronic Pain Risk Model); and (2) adding other predictors (Expanded Chronic Pain Risk Model). Patients initiating primary care for back pain (N = 571) completed measures used in the Chronic Pain Risk Score, Improved Model, and Expanded Model, then completed the Graded Chronic Pain Scale (GCPS) 4 months later (n = 521; 91% response rate). In predicting 4‐month GCPS grade III or IV (moderate or severe pain‐related activity interference), the Improved Model performed better than did the Chronic Pain Risk Score (Net Reclassification Index [NRI] = 0.32, P = 0.003). The Expanded Model improved significantly on the prediction of the Improved Model (NRI = 0.56, P < 0.001) and demonstrated excellent discriminative ability (AUC = 0.84, 95% CI = 0.79‐0.88). The Improved Model (AUC = 0.79, 95% CI = 0.75‐0.84) and the Chronic Pain Risk Score (AUC = 0.76, 95% CI = 0.71‐0.81) showed acceptable discriminative ability. A limited set of measures may be used to predict risk for future clinically significant pain in patients initiating primary care for back pain, but further evaluation of prognostic models is needed.

Positional Estimation Within a Latent Space Model for Networks

Recent advances in latent space and related random effects models hold much promise for representing network data. The inherent dependency between ties in a network makes modeling data of this type difficult. In this article we consider a recently developed latent space model that is particularly appropriate for the visualization of networks. We suggest a new estimator of the latent positions and perform two network analyses, comparing four alternative estimators. We demonstrate a method of checking the validity of the positional estimates. These estimators are implemented via a package in the freeware statistical language R. The package allows researchers to efficiently fit the latent space model to data and to visualize the results.

Inside the black box: starting to uncover the underlying decision rules used in a one-by-one expert assessment of occupational exposure in case-control studies.

Objectives Evaluating occupational exposures in population-based case-control studies often requires exposure assessors to review each study participants reported occupational information job-by-job to derive exposure estimates. Although such assessments likely have underlying decision rules, they usually lack transparency, are time consuming and have uncertain reliability and validity. We aimed to identify the underlying rules to enable documentation, review and future use of these expert-based exposure decisions. Methods Classification and regression trees (CART, predictions from a single tree) and random forests (predictions from many trees) were used to identify the underlying rules from the questionnaire responses, and an experts exposure assignments for occupational diesel exhaust exposure for several metrics: binary exposure probability and ordinal exposure probability, intensity and frequency. Data were split into training (n=10 488 jobs), testing (n=2247) and validation (n=2248) datasets. Results The CART and random forest models’ predictions agreed with 92–94% of the experts binary probability assignments. For ordinal probability, intensity and frequency metrics, the two models extracted decision rules more successfully for unexposed and highly exposed jobs (86–90% and 57–85%, respectively) than for low or medium exposed jobs (7–71%). Conclusions CART and random forest models extracted decision rules and accurately predicted an experts exposure decisions for the majority of jobs, and identified questionnaire response patterns that would require further expert review if the rules were applied to other jobs in the same or different study. This approach makes the exposure assessment process in case-control studies more transparent, and creates a mechanism to efficiently replicate exposure decisions in future studies.

A multiple imputation strategy for sequential multiple assignment randomized trials

Sequential multiple assignment randomized trials (SMARTs) are increasingly being used to inform clinical and intervention science. In a SMART, each patient is repeatedly randomized over time. Each randomization occurs at a critical decision point in the treatment course. These critical decision points often correspond to milestones in the disease process or other changes in a patients health status. Thus, the timing and number of randomizations may vary across patients and depend on evolving patient-specific information. This presents unique challenges when analyzing data from a SMART in the presence of missing data. This paper presents the first comprehensive discussion of missing data issues typical of SMART studies: we describe five specific challenges and propose a flexible imputation strategy to facilitate valid statistical estimation and inference using incomplete data from a SMART. To illustrate these contributions, we consider data from the Clinical Antipsychotic Trial of Intervention and Effectiveness, one of the most well-known SMARTs to date.

Association of levels of opioid use with pain and activity interference among patients initiating chronic opioid therapy: a longitudinal study.

Abstract Little is known about long-term pain and function outcomes among patients with chronic noncancer pain initiating chronic opioid therapy (COT). In the Middle-Aged/Seniors Chronic Opioid Therapy study of patients identified through electronic pharmacy records as initiating COT for chronic noncancer pain, we examined the relationships between level of opioid use (over the 120 days before outcome assessment) and pain and activity interference outcomes at 4- and 12-month follow-ups. Patients aged 45+ years (N = 1477) completed a baseline interview; 1311 and 1157 of these comprised the 4- and 12-month analysis samples, respectively. Opioid use was classified based on self-report and electronic pharmacy records for the 120 days before the 4- and 12-month outcome assessments. Controlling for patient characteristics that predict sustained COT and pain outcomes, patients who had used opioids minimally or not at all, compared with those with intermittent/lower-dose and regular/higher-dose opioid use, had better pain intensity and activity interference outcomes. Adjusted mean (95% confidence interval) pain intensity (0-10 scale) at 12 months was 4.91 (4.68-5.13) for the minimal/no use group and 5.71 (5.50-5.92) and 5.72 (5.51-5.93) for the intermittent/lower-dose and regular/higher-dose groups, respectively. A similar pattern was observed for pain intensity at 4 months and for activity interference at both time points. Better outcomes in the minimal/no use group could reflect pain improvement leading to opioid discontinuation. The similarity in outcomes of regular/higher-dose and intermittent/lower-dose opioid users suggests that intermittent and/or lower-dose use vs higher-dose use may confer risk reduction without reducing benefits.

Sex and age differences in global pain status among patients using opioids long term for chronic noncancer pain

BACKGROUND The use of chronic opioid therapy (COT) has risen dramatically in recent years, especially among women. However, little is known about factors influencing overall pain and function (global pain status) among COT users. Characterizing the typical experiences of COT patients by age-sex group could help clinicians and patients better weigh the risks and benefits of COT. Thus, we sought to characterize global pain status among COT users in community practice by age and sex. METHODS Telephone survey of 2,163 health plan members aged 21-80 years using COT. We assessed average/usual pain (0-10 scale); pain-related interference (0-10); activity limitation days, last 3 months; and pain impact, last 2 weeks (0-11). Status on each indicator was classified as low (better pain/function), moderate, or high (worse pain/function). Global pain status was categorized as favorable if 2-4 indicators were low and 0-1 was high and unfavorable if 2-4 indicators were high and 0-1 was low. RESULTS Among female COT patients, 15% (vs. 26% of males) had favorable global pain status and 59% (vs. 42% of males) had unfavorable status. Under age 65 years, women fared more poorly than men on every indicator. Among 65- to 80-year-olds, women and men had similar global pain status. CONCLUSIONS Although pain and function among COT users vary considerably, only one in five reported low pain levels and high levels of function. Young and middle-aged women seem to be at particularly high risk for unfavorable global pain status. More research is needed about how to best manage pain in this group.