Identifying potentially undiagnosed nontuberculous mycobacterial lung disease among patients with chronic obstructive pulmonary disease: Development of a predictive algorithm using claims data.

BACKGROUND: Nontuberculous mycobacterial lung disease (NTMLD) is a debilitating disease. Chronic obstructive pulmonary disease (COPD) is the leading comorbidity associated with NTMLD in the United States. Their similarities in symptoms and overlapping radiological findings may delay NTMLD diagnosis in patients with COPD. OBJECTIVE: To develop a predictive model that identifies potentially undiagnosed NTMLD among patients with COPD. METHODS: This retrospective cohort study developed a predictive model of NTMLD using US Medicare beneficiary claims data (2006 - 2017). Patients with COPD with NTMLD were matched 1:3 to patients with COPD without NTMLD by age, sex, and year of COPD diagnosis. The predictive model was developed using logistic regression modeling risk factors such as pulmonary symptoms, comorbidities, and health care resource utilization. The final model was based on model fit statistics and clinical inputs. Model performance was evaluated for both discrimination and generalizability with c-statistics and receiver operating characteristic curves. RESULTS: There were 3,756 patients with COPD with NTMLD identified and matched to 11,268 patients with COPD without NTMLD. A higher proportion of patients with COPD with NTMLD, compared with those with COPD without NTMLD, had claims for pulmonary symptoms and conditions, including hemoptysis (12.6% vs 1.4%), cough (63.4% vs 24.7%), dyspnea (72.5% vs 38.2%), pneumonia (59.2% vs 13.4%), chronic bronchitis (40.5% vs 16.3%), emphysema, (36.7% vs 11.1%), and lung cancer (15.7% vs 3.5%). A higher proportion of patients with COPD with NTMLD had pulmonologist and infectious disease (ID) specialist visits than patients with COPD without NTMLD (≥ 1 pulmonologist visit: 81.3% vs 23.6%, respectively; ≥ 1 ID visit: 28.3% vs 4.1%, respectively, P < 0.0001). The final model consists of 10 risk factors (≥ 2 ID specialist visits; ≥ 4 pulmonologist visits; the presence of hemoptysis, cough, emphysema, pneumonia, tuberculosis, lung cancer, or idiopathic interstitial lung disease; and being underweight during a 1-year pre-NTMLD period) predicting NTMLD with high sensitivity and specificity (c-statistic, 0.9). The validation of the model on new testing data demonstrated similar discrimination and showed the model was able to predict NTMLD earlier than the receipt of the first diagnostic claim for NTMLD. CONCLUSIONS: This predictive algorithm uses a set of criteria comprising patterns of health care use, respiratory symptoms, and comorbidities to identify patients with COPD and possibly undiagnosed NTMLD with high sensitivity and specificity. It has potential application in raising timely clinical suspicion of patients with possibly undiagnosed NTMLD, thereby reducing the period of undiagnosed NTMLD. DISCLOSURES: Dr Wang and Dr Hassan are employees of Insmed, Inc. Dr Chatterjee was an employee of Insmed, Inc, at the time of this study. Dr Marras is participating in multicenter clinical trials sponsored by Insmed, Inc, has consulted for RedHill Biopharma, and has received a speaker's honorarium from AstraZeneca. Dr Allison is an employee of Statistical Horizons, LLC. This study was funded by Insmed Inc.

Required sample size to detect mediation in 3-level implementation studies.

BACKGROUND: Statistical tests of mediation are important for advancing implementation science; however, little research has examined the sample sizes needed to detect mediation in 3-level designs (e.g., organization, provider, patient) that are common in implementation research. Using a generalizable Monte Carlo simulation method, this paper examines the sample sizes required to detect mediation in 3-level designs under a range of conditions plausible for implementation studies. METHOD: Statistical power was estimated for 17,496 3-level mediation designs in which the independent variable (X) resided at the highest cluster level (e.g., organization), the mediator (M) resided at the intermediate nested level (e.g., provider), and the outcome (Y) resided at the lowest nested level (e.g., patient). Designs varied by sample size per level, intraclass correlation coefficients of M and Y, effect sizes of the two paths constituting the indirect (mediation) effect (i.e., X→M and M→Y), and size of the direct effect. Power estimates were generated for all designs using two statistical models-conventional linear multilevel modeling of manifest variables (MVM) and multilevel structural equation modeling (MSEM)-for both 1- and 2-sided hypothesis tests. RESULTS: For 2-sided tests, statistical power to detect mediation was sufficient (≥0.8) in only 463 designs (2.6%) estimated using MVM and 228 designs (1.3%) estimated using MSEM; the minimum number of highest-level units needed to achieve adequate power was 40; the minimum total sample size was 900 observations. For 1-sided tests, 808 designs (4.6%) estimated using MVM and 369 designs (2.1%) estimated using MSEM had adequate power; the minimum number of highest-level units was 20; the minimum total sample was 600. At least one large effect size for either the X→M or M→Y path was necessary to achieve adequate power across all conditions. CONCLUSIONS: While our analysis has important limitations, results suggest many of the 3-level mediation designs that can realistically be conducted in implementation research lack statistical power to detect mediation of highest-level independent variables unless effect sizes are large and 40 or more highest-level units are enrolled. We suggest strategies to increase statistical power for multilevel mediation designs and innovations to improve the feasibility of mediation tests in implementation research.

From Data to Causes III: Bayesian Priors for General Cross-Lagged Panel Models (GCLM).

This article describes some potential uses of Bayesian estimation for time-series and panel data models by incorporating information from prior probabilities (i.e., priors) in addition to observed data. Drawing on econometrics and other literatures we illustrate the use of informative "shrinkage" or "small variance" priors (including so-called "Minnesota priors") while extending prior work on the general cross-lagged panel model (GCLM). Using a panel dataset of national income and subjective well-being (SWB) we describe three key benefits of these priors. First, they shrink parameter estimates toward zero or toward each other for time-varying parameters, which lends additional support for an income → SWB effect that is not supported with maximum likelihood (ML). This is useful because, second, these priors increase model parsimony and the stability of estimates (keeping them within more reasonable bounds) and thus improve out-of-sample predictions and interpretability, which means estimated effect should also be more trustworthy than under ML. Third, these priors allow estimating otherwise under-identified models under ML, allowing higher-order lagged effects and time-varying parameters that are otherwise impossible to estimate using observed data alone. In conclusion we note some of the responsibilities that come with the use of priors which, departing from typical commentaries on their scientific applications, we describe as involving reflection on how best to apply modeling tools to address matters of worldly concern.

Mapping Activity Patterns to Quantify Risk of Violent Assault in Urban Environments.

BACKGROUND: We collected detailed activity paths of urban youth to investigate the dynamic interplay between their lived experiences, time spent in different environments, and risk of violent assault. METHODS: We mapped activity paths of 10- to 24-year-olds, including 143 assault patients shot with a firearm, 206 assault patients injured with other types of weapons, and 283 community controls, creating a step-by-step mapped record of how, when, where, and with whom they spent time over a full day from waking up until going to bed or being assaulted. Case-control analyses compared cases with time-matched controls to identify risk factors for assault. Case-crossover analyses compared cases at the time of assault with themselves earlier in the day to investigate whether exposure increases acted to the trigger assault. RESULTS: Gunshot assault risks included being alone (odds ratio [OR] = 1.6, 95% confidence interval [CI] = 1.3, 1.9) and were lower in areas with high neighbor connectedness (OR = 0.7, 95% CI = 0.6, 0.8). Acquiring a gun (OR = 1.4, 95% CI = 1.1, 1.6) and entering areas with more vacancy, violence, and vandalism (OR = 1.7, 95% CI = 1.1, 2.7) appeared to trigger the risk of getting shot shortly thereafter. Nongunshot assault risks included being in areas with recreation centers (OR = 1.2, 95% CI = 1.1, 1.4). Entering an area with higher truancy (OR = 1.6, 95% CI = 1.1, 2.5) and more vacancy, violence, and vandalism appeared to trigger the risk of nongunshot assault. Risks varied by age group. CONCLUSIONS: We achieved a large-scale study of the activities of many boys, adolescents, and young men that systematically documented their experiences and empirically quantified risks for violence. Working at a temporal and spatial scale that is relevant to the dynamics of this phenomenon gave novel insights into triggers for violent assault.

Effects of Achieving Target Measures in Rheumatoid Arthritis on Functional Status, Quality of Life, and Resource Utilization: Analysis of Clinical Practice Data.

OBJECTIVE: To evaluate associations between achieving guideline-recommended targets of disease activity, defined by the Disease Activity Score in 28 joints using C-reactive protein level (DAS28-CRP) <2.6, the Simplified Disease Activity Index (SDAI) ≤3.3, or the Clinical Disease Activity Index (CDAI) ≤2.8, and other health outcomes in a longitudinal observational study. METHODS: Other defined thresholds included low disease activity (LDA), moderate (MDA), or severe disease activity (SDA). To control for intraclass correlation and estimate effects of independent variables on outcomes of the modified Health Assessment Questionnaire (M-HAQ), the EuroQol 5-domain (EQ-5D; a quality-of-life measure), hospitalization, and durable medical equipment (DME) use, we employed mixed models for continuous outcomes and generalized estimating equations for binary outcomes. RESULTS: Among 1,297 subjects, achievement (versus nonachievement) of recommended disease targets was associated with enhanced physical functioning and lower health resource utilization. After controlling for baseline covariates, achievement of disease targets (versus LDA) was associated with significantly enhanced physical functioning based on SDAI ≤3.3 (ΔM-HAQ -0.047; P = 0.0100) and CDAI ≤2.8 (-0.073; P = 0.0003) but not DAS28-CRP <2.6 (-0.022; P = 0.1735). Target attainment was associated with significantly improved EQ-5D (0.022-0.096; P < 0.0030 versus LDA, MDA, or SDA). Patients achieving guideline-recommended disease targets were 36-45% less likely to be hospitalized (P < 0.0500) and 23-45% less likely to utilize DME (P < 0.0100). CONCLUSION: Attaining recommended target disease-activity measures was associated with enhanced physical functioning and health-related quality of life. Some health outcomes were similar in subjects attaining guideline targets versus LDA. Achieving LDA is a worthy clinical objective in some patients.

When one spouse has an affair, who is more likely to leave?

OBJECTIVE: We examine whether having an affair around the time a marriage broke up is associated with being the person who wanted the divorce more or the person who was left. We also examine predictors of having an affair around the end of the marriage. METHODS: We use the National Survey of Families and Households, using each ex-spouse's reports of which spouse wanted the divorce more and whether either was having an affair around the end of the marriage. We combine latent class models with logistic regression, treating either spouse's report as a fallible indicator of the reality of whether each had an affair and who wanted the divorce more. RESULTS: We find that a spouse having an affair is more likely to be the one who wanted the divorce more. We find little gender difference in who has affairs preceding divorce. CONCLUSIONS: Results suggest that it is more common to leave because one is having an affair, or to have an affair because one has decided to leave, than it is to discover one's spouse having an affair and initiate a divorce.

Fears of violence during morning travel to school.

PURPOSE: Children's safety as they travel to school is a concern nationwide. We investigated how safe children felt from the risk of being assaulted during morning travel to school. METHODS: Children between 10 and 18 years old were recruited in Philadelphia and interviewed with the aid of geographic information system (GIS) mapping software about a recent trip to school, situational characteristics, and how safe they felt as they travelled based on a 10-point item (1 = very unsafe, 10 = very safe). Ordinal regression was used to estimate the probability of perceiving different levels of safety based on transportation mode, companion type, and neighborhood characteristics. RESULTS: Among 65 randomly selected subjects, routes to school ranged from 7 to 177 minutes (median = 36) and .1-15.1 street miles (median = 1.9), and included between 1-5 transportation modes (median = 2). Among students interviewed, 58.5% felt less than very safe (i.e., <10) at some point while traveling to school and one-third (32.5%) of the total person time was spent feeling less than very safe. Nearly a quarter of students, or 24.6%, felt a reduction in safety immediately upon exiting their home. The probability of reporting a safety of >8, for example, was .99 while in a car and .94 while on foot but was .86 and .87 when on a public bus or trolley. Probability was .98 while with an adult but was .72 while with another child and .71 when alone. Also, perceived safety was lower in areas of high crime and high density of off-premise alcohol outlets. CONCLUSIONS: Efforts that target situational risk factors are warranted to help children feel safe over their entire travel routes to school.

Overall perceived health predicts risk of hospitalizations and death in adults with heart failure: a prospective longitudinal study.

BACKGROUND: Patient overall perception of health may provide an effective early warning for risk of hospitalization and death among heart failure patients. OBJECTIVE: Determine whether overall perceived health predicts all-cause hospitalization or death in heart failure patients after adjusting for confounding factors in a sample of adults with heart failure. DESIGN: Prospective, longitudinal, observational study. SETTINGS: Three outpatient urban settings in the northeast United States between 2007 and 2010. PARTICIPANTS: Adults with chronic Stage C heart failure confirmed by echocardiographic and clinical evidence. METHODS: A secondary analysis was conducted using data collected on 273 Stage C patients with heart failure. Participants in the parent study were followed for 6 months. Overall perceived health was measured by self-report. Hospitalization and death were assessed from electronic hospital records and confirmed with county death records as needed. Cox proportional hazards models were used to examine the association between perceptions of health and rates of hospitalization and death. RESULTS: Patients with poor or fair perceived health had over 5.5 times the rate of death or hospitalization over the 6-month period (hazard ratio; 95% confidence interval: 2.0-15.6; p=0.001) after controlling for model covariates. The predictive ability of perceived health attenuated over time such that at 30-days patients who reported poor or fair perceived health had only 1.2 times the rate of an event and virtually no difference in event rate by 60-days. CONCLUSIONS: Overall perceived health is a powerful indicator of impending events and can be a quick tool for prioritizing heart failure patients who are at highest risk of imminent death and hospitalization. Questions about perceived health need to be asked of patients regularly in order to have clinical utility.

Stimulants and cardiovascular events in youth with attention-deficit/hyperactivity disorder.

OBJECTIVE: This study examined associations between stimulant use and risk of cardiovascular events and symptoms in youth with attention-deficit/hyperactivity disorder and compared the risks associated with methylphenidate and amphetamines. METHOD: Claims were reviewed of privately insured young people 6 to 21 years old without known cardiovascular risk factors (n = 171,126). A day-level cohort analysis evaluated the risk of cardiovascular events after a diagnosis of attention-deficit/hyperactivity disorder in relation to stimulant exposures. Based on filled stimulant prescriptions, follow-up days were classified as current, past, and no stimulant use. Endpoints included an emergency department or inpatient diagnosis of angina pectoris, cardiac dysrhythmia, or transient cerebral ischemia (cardiac events) or tachycardia, palpitations, or syncope (cardiac symptoms). RESULTS: There were 0.92 new cardiac events and 3.08 new cardiac symptoms per 1,000,000 days of current stimulant use. Compared with no stimulant use (reference group), the adjusted odds ratios of cardiac events were 0.69 (95% confidence interval 0.42-1.12) during current stimulant use and 1.18 (95% CI 0.83-1.66) during past stimulant use. The corresponding adjusted odds ratios for cardiac symptoms were 1.18 (95% CI 0.89-1.59) for current and 0.93 (95% CI 0.71-1.21) for past stimulant use. No significant differences were observed in risks of cardiovascular events (2.14, 95% CI 0.82-5.63) or symptoms (1.08, 95% CI 0.66-1.79) for current methylphenidate use compared with amphetamine use (reference group). CONCLUSIONS: Clinical diagnoses of cardiovascular events and symptoms were rare and not associated with stimulant use. The results help to allay concerns over the cardiovascular safety of stimulant treatment for attention-deficit/hyperactivity disorder in young people without known pre-existing risk factors.

She left, he left: how employment and satisfaction affect women's and men's decisions to leave marriages.

Studies examining determinants of divorce have largely ignored differences between factors that elevate wives' and husbands' initiation of divorce. The authors use longitudinal data and a latent class model embedded in a competing-risks event history model to assess distinct predictors of wives and husbands leaving marriages. They find that when men are not employed, either spouse is more likely to leave. When wives report better-than-average marital satisfaction, their employment affects neither spouse's exit. However, when wives report below-average marital satisfaction, their employment makes it more likely they will leave. The authors' findings suggest that theories of divorce require "gendering" to reflect asymmetric gender change.

Physical and mental health, cognitive development, and health care use by housing status of low-income young children in 20 American cities: a prospective cohort study.

OBJECTIVES: We assessed the independent effect of homeless and doubled-up episodes on physical and mental health, cognitive development, and health care use among children. METHODS: We used data from 4 waves of the Fragile Families and Child Wellbeing Study, involving a sample of 2631 low-income children in 20 large US cities who have been followed since birth. Multivariate analyses involved logistic regression using the hybrid method to include both fixed and random effects. RESULTS: Of the sample, 9.8% experienced homelessness and an additional 23.6% had a doubled-up episode. Housing status had little significant adverse effect on child physical or mental health, cognitive development, or health care use. CONCLUSIONS: Family and environmental stressors common to many children in poverty, rather than just homeless and doubled-up episodes, were associated with young children's poor health and cognitive development and high health care use. Practitioners need to identify and respond to parental and family needs for support services in addition to housing assistance to effectively improve the health and development of young children who experience residential instability, particularly those in homeless families.

Combined analysis of retrospective and prospective occurrences in cohort studies: HIV-1 serostatus and incident pneumonia.

BACKGROUND: The authors show how information collected on retrospective occurrence times may be combined with prospective occurrence times in the analysis of recurrent events from cohort studies. METHODS: We demonstrate how the observed data can be expanded from one to two records per participant and account for the within-individual dependence when estimating variances. We illustrate our methods using data from the Women's Interagency HIV Study, which recorded 384 retrospective and 352 prospective occurrences of pneumonia in 9478 retrospective and 7857 prospective person-years among 2610 adult women. RESULTS: The hazard of non-Pneumocystis carinii pneumonia among the 2056 HIV-1 infected women was 2.24 times (95% confidence limits: 1.74, 2.89) that of the 554 uninfected women, independent of age. This hazard ratio was homogeneous across retrospective and prospective occurrences (P for interaction = 0.96) and combining occurrence types increased the precision by reducing the standard error by about a fourth. CONCLUSIONS: As expected, HIV-1 infection increases the hazard of pneumonia, with more precise inference obtained by combining information available on bidirectional occurrences. The proposed method for the analysis of bidirectional occurrence times will improve precision when the estimated associations are homogeneous across occurrence types, or may provide added insight into either the data collection or disease process when the estimated associations are heterogeneous.

Mortality after the hospitalization of a spouse.

BACKGROUND: The illness of a spouse can affect the health of a caregiving partner. We examined the association between the hospitalization of a spouse and a partner's risk of death among elderly people. METHODS: We studied 518,240 couples who were enrolled in Medicare in 1993. We used Cox regression analysis and fixed-effects (case-time-control) methods to assess hospitalizations and deaths during nine years of follow-up. RESULTS: Overall, 383,480 husbands (74 percent) and 347,269 wives (67 percent) were hospitalized at least once, and 252,557 husbands (49 percent) and 156,004 wives (30 percent) died. Mortality after the hospitalization of a spouse varied according to the spouse's diagnosis. Among men, 6.4 percent died within a year after a spouse's hospitalization for colon cancer, 6.9 percent after a spouse's hospitalization for stroke, 7.5 percent after a spouse's hospitalization for psychiatric disease, and 8.6 percent after a spouse's hospitalization for dementia. Among women, 3.0 percent died within a year after a spouse's hospitalization for colon cancer, 3.7 percent after a spouse's hospitalization for stroke, 5.7 percent after a spouse's hospitalization for psychiatric disease, and 5.0 percent after a spouse's hospitalization for dementia. After adjustment for measured covariates, the risk of death for men was not significantly higher after a spouse's hospitalization for colon cancer (hazard ratio, 1.02; 95 percent confidence interval, 0.95 to 1.09) but was higher after hospitalization for stroke (hazard ratio, 1.06; 95 percent confidence interval, 1.03 to 1.09), congestive heart failure (hazard ratio, 1.12; 95 percent confidence interval, 1.07 to 1.16), hip fracture (hazard ratio, 1.15; 95 percent confidence interval, 1.11 to 1.18), psychiatric disease (hazard ratio, 1.19; 95 percent confidence interval, 1.12 to 1.26), or dementia (hazard ratio, 1.22; 95 percent confidence interval, 1.12 to 1.32). For women, the various risks of death after a spouse's hospitalization were similar. Overall, for men, the risk of death associated with a spouse's hospitalization was 22 percent of that associated with a spouse's death (95 percent confidence interval, 17 to 27 percent); for women, the risk was 16 percent of that associated with death (95 percent confidence interval, 8 to 24 percent). CONCLUSIONS: Among elderly people hospitalization of a spouse is associated with an increased risk of death, and the effect of the illness of a spouse varies among diagnoses. Such interpersonal health effects have clinical and policy implications for the care of patients and their families.

Estimation of cumulative odds ratios.

PURPOSE: Standard estimation of ordered odds ratios requires the constraint that the etiologic effects of exposure are homogenous across thresholds of the ordered response. We present a method to relax this often-unrealistic constraint. METHODS: The kernel of the proposed method is the expansion of observed data into "person-thresholds." Using standard statistical software, for each subject we create a separate record for each response threshold and then apply binary logistic regression to estimate generalized cumulative odds ratios for one or more exposures. RESULTS: Two examples demonstrate that the proposed method provides increased flexibility in assessing the etiologic effects of exposures. A Monte Carlo simulation study supports the proposed approach by suggesting the estimated cumulative odds ratios are unbiased with proper confidence interval coverage attained by use of generalized estimating equations. CONCLUSION: The proposed method provides simple estimates of ordered odds ratios that allow the etiologic effects of exposure to vary across levels of the ordered response.

Missing data techniques for structural equation modeling.

As with other statistical methods, missing data often create major problems for the estimation of structural equation models (SEMs). Conventional methods such as listwise or pairwise deletion generally do a poor job of using all the available information. However, structural equation modelers are fortunate that many programs for estimating SEMs now have maximum likelihood methods for handling missing data in an optimal fashion. In addition to maximum likelihood, this article also discusses multiple imputation. This method has statistical properties that are almost as good as those for maximum likelihood and can be applied to a much wider array of models and estimation methods.

Factors associated with distress in urban residential fire survivors.

PURPOSE: To identify factors associated with recovery in a sample of urban residential fire survivors. DESIGN AND METHODS: 440 survivors, of residential fires were interviewed at approximately 3, 6, and 13 months after the fire to measure psychological distress. A set of factors was identified that correlated with survivors' ability to recover from the fire event. Potential predictors of increased distress were identified. Hypotheses were that participants who were lower in socioeconomic status, who were minority members, who had less social support, who engaged in attributional thinking, and had greater concurrent life stresses would have greater psychological distress in response to a residential fire and would be less able to recover from the fire event. FINDINGS: Distress after fire was high at 3 months and decreased for the majority of participants, although one-third of survivors had higher distress at 13 months than at 3 months. Loss of control and attributional variables had the strongest influence on psychological distress over time. CONCLUSIONS: The findings are consistent with stress-response tendencies expected after a stressful event. A set of predictor variables was identified to help clinicians target survivors at high risk for psychological distress after a residential fire.