Longitudinal Data Sources Research Articles

Estimating the Comparative Effectiveness of Feeding Interventions in the Pediatric Intensive Care Unit: A Demonstration of Longitudinal Targeted Maximum Likelihood Estimation.

Longitudinal data sources offer new opportunities for the evaluation of sequential interventions. To adjust for time-dependent confounding in these settings, longitudinal targeted maximum likelihood based estimation (TMLE), a doubly robust method that can be coupled with machine learning, has been proposed. This paper provides a tutorial in applying longitudinal TMLE, in contrast to inverse probability of treatment weighting and g-computation based on iterative conditional expectations. We apply these methods to estimate the causal effect of nutritional interventions on clinical outcomes among critically ill children in a United Kingdom study (Control of Hyperglycemia in Paediatric Intensive Care, 2008–2011). We estimate the probability of a child’s being discharged alive from the pediatric intensive care unit by a given day, under a range of static and dynamic feeding regimes. We find that before adjustment, patients who follow the static regime “never feed” are discharged by the end of the fifth day with a probability of 0.88 (95% confidence interval: 0.87, 0.90), while for the patients who follow the regime “feed from day 3,” the probability of discharge is 0.64 (95% confidence interval: 0.62, 0.66). After adjustment for time-dependent confounding, most of this difference disappears, and the statistical methods produce similar results. TMLE offers a flexible estimation approach; hence, we provide practical guidance on implementation to encourage its wider use.

Constrained Agency in Later Working Lives: Introduction to the Special Issue

In the social sciences there is an ongoing debate about the primacy of structure or agency in shaping individual behaviors. This Special Issue presents a comprehensive set of papers on structure and agency in the area of later working lives and retirement, specifically focusing on situations where agency is constrained. The papers (a) study a diverse set of labor market situations in which individual agency is constrained, (b) focus on individual- and country-level predictors of constrained agency, and (c) provide insights regarding consequences of constrained agency in different areas of life. A broad range of country contexts are covered in the articles and rich—often longitudinal—data sources are used to advance our understanding of agency in later working lives. The current Introduction to the Special Issue elaborates on the theoretical background linking the different papers, briefly describes the different papers, and discusses directions for future research to follow from the full set of papers.

New Zealand Longitudinal Census development

ABSTRACT
 ObjectivesThe aim for the New Zealand Longitudinal Census (NZLC) is to link data from seven recent censuses (1981-2013) and develop weight methodology to reduce the potential for link bias in the NZLC.
 ApproachThe most workable way to create a longitudinal census dataset was to link adjacent pairs of censuses and then to link these census pairs together creating six linked pairs covering the seven censuses from 1981 to 2013. We refined the deterministic linking in the six pairs and then added probabilistic linking, where possible, for the remaining records in each of the theoretically linkable populations. The linking of the 2006 and 2013 Censuses was methodological challenge as the interval between censuses was 7 years instead of the normal 5 years between censuses. The challenge was heightened by having a question that asked for respondent address five years ago (ie in 2008).
 In the linking of six censuses from 1981 to 2006, we had included indicators that identified completely linked and partially linked families. To link 2013-2006 census pairs, more variables were added in the probabilistic linking stage including geographic variables and new derived variables which included the combination of the role types within family group, family number, and sex variables. In developing weights for the linked pairs, all variables were changed into numeric values and then categorised. The data was then divided into male and female and logistic regression was used to predict probabilities of being liked.
 ResultsThe size of linked populations is relatively stable across similar time spans. We linked the majority of linked records in the deterministic stage which contributed links of for two-thirds of the theoretical populations. The subsequent stages included deterministic linking and probabilistic linking which added approximately 5 percentage points to the link.
 ConclusionThe linking of census datasets from 1981 to 2013 to form a longitudinal data source allowed us to understand important aspects of population change across time which cannot be derived from simple cross. It provides a major new source of the analysis for insights into New Zealand demographic and social outcomes. Currently we are creating a set of generic weights for each individual record that may be applied in various research scenarios. Weighting of longitudinal data, as a part of NZLC, requires substantial in-depth statistical analysis of bias and coverage.

Psychosocial screening and management of young people aged 18-25 years with diabetes.

Routine psychosocial screening and management of people with diabetes is recommended. To profile demographic, medical and psychosocial characteristics of young people with diabetes, and to develop a screening tool and care pathway for routine use. Indices of diabetes control and recorded diabetes complications were complimented by psychosocial screening tools assessing psychological, diabetes specific and perceived stress (Kessler 10, Problem Area in Diabetes, Perceived Stress Scale), well-being (World Health Organization Well Being Index-5), disordered eating (Eating Disorder Risk Inventory-3 Risk Composite), compensatory behaviour questionnaire, social support (Multidimensional Scale of Perceived Social Support), resilience (Connor Davidson Resilience Scale - 2 item) and financial concerns. Service provision and demographic data were also collected. Diabetes and mental health clinicians then identified a subset of measures to use for routine screening along with care pathways. Psychosocial screening was well accepted. Participants (151) had suboptimal glycaemic control (glycated haemoglobin 8.0 interquartile range 1.8%/64 interquartile range 22 mmol/mol). Severe diabetes-related distress (Problem Area in Diabetes ≥40) was found in 19.4% and 26.0% reported difficulties managing healthcare costs. A mental health disorder was likely in 9.7%, whilst 23.4% had high Kessler 10 scores. Low World Health Organization Well Being Index-5 scores (≤13) were seen in 29.0%. Risk for an eating disorder (Eating Disorder Risk Inventory-3 Risk Composite) was 12.7%, whereas approximately 36.0% had disturbed eating behaviours. Psychosocial screening of young adults with diabetes identified complex needs. A brief psychosocial screening tool and associated care pathways were developed for routine use in a young adult tertiary referral diabetes clinic. The tool assesses constructs, such as diabetes distress, depression, anxiety, well-being, hypoglycaemia-unawareness, fear of hypoglycaemia, social support, weight, shape and eating concerns and financial concerns. This will provide a longitudinal data source for further research to inform clinical practice.

Changes in Veteran Tobacco Use Identified in Electronic Medical Records

Electronic medical records represent a new source of longitudinal data on tobacco use. Electronic medical records of the U.S. Department of Veterans Affairs were extracted to find patients' tobacco use status in 2009 and at another assessment 12-24 months later. Records from the year prior to the first assessment were used to determine patient demographics and comorbidities. These data were analyzed in 2015. An annual quit rate of 12.0% was observed in 754,504 current tobacco users. Adjusted tobacco use prevalence at follow-up was 3.2% greater with alcohol use disorders at baseline, 1.9% greater with drug use disorders, 3.3% greater with schizophrenia, and lower in patients with cancer, heart disease, and other medical conditions (all differences statistically significant with p<0.05). Annual relapse rates in 412,979 former tobacco users were 29.6% in those who had quit for <1 year, 9.7% in those who had quit for 1-7 years, and 1.9% of those who had quit for >7 years. Among those who had quit for <1 year, adjusted relapse rates were 4.3% greater with alcohol use disorders and 7.2% greater with drug use disorders (statistically significant with p<0.05). High annual cessation rates may reflect the older age and greater comorbidities of the cohort or the intensive cessation efforts of the U.S. Department of Veterans Affairs. The lower cessation and higher relapse rates in psychiatric and substance use disorders suggest that these groups will need intensive and sustained cessation efforts.

Health Information Exchange Organizations and Their Support for Research: Current State and Future Outlook

Federal investment spurred health information exchange organization (HIO) development and maturation to provide third-party approaches to electronic health information exchange across disparate electronic health record (EHR) systems. By creating opportunities for data aggregation across multiple medical institutions, HIOs also spur research. Using data from a 2015 national web-based survey of HIOs (N = 64), we identified HIOs supporting or not supporting research, and compared characteristics of the 2 groups. We found that 15 (23%) of the 64 HIOs reported supporting research, 30 (47%) reported planning to support research, and 19 (30%) did not support research. Research-supporting HIOs were more likely than nonresearch supporting HIOs to offer advanced functionality, such as allowing users to query and retrieve data from multiple sources. Our study offers encouraging preliminary evidence that HIOs are supporting research, which could offer a solution to current challenges in creating comprehensive longitudinal clinical data sources for research.

Longitudinal Emergency Medical Technician Attributes and Demographic Study (LEADS) Design and Methodology.

The objective of this study is to describe the Longitudinal Emergency Medical Technician (EMT) Attributes and Demographic Study (LEADS) design, instrument development, pilot testing, sampling procedures, and data collection methodology. Response rates are provided, along with results of follow-up surveys of non-responders (NRs) and a special survey of Emergency Medical Services (EMS) professionals who were not nationally certified. Annual surveys from 1999 to 2008 were mailed out to a random, stratified sample of nationally registered EMT-Basics and Paramedics. Survey weights were developed to reflect each respondent's probability of selection. A special survey of NRs was mailed out to individuals who did not respond to the annual survey to estimate the probable extent and direction of response bias. Individuals who indicated they were no longer in the profession were mailed a special exit survey to determine their reasons for leaving EMS. Given the large number of comparisons between NR and regular (annual) survey respondents, it is not surprising that some statistically significant differences were found. In general, there were few differences. However, NRs tended to report higher annual EMS incomes, were younger, healthier, more physically fit, and were more likely to report that they were not practicing EMS. Comparisons of the nationally certified EMS professionals with EMS professionals who were not nationally certified indicated that nationally certified EMS providers were younger, had less EMS experiences, earned less, were more likely to be female and work for private EMS services, and less likely to work for fire-based services. These differences may reflect state and local policy and practice, since many states and local agencies do not require maintenance of national certification as a requirement to practice. When these differences were controlled for statistically, there were few systematic differences between non-nationally certified and nationally certified EMS professionals. The LEADS study is the only national, randomized, and longitudinal data source for studying EMS professionals in the United States. Although not without flaws, this study remains an excellent source of information about EMS provider demographics, attributes, attitudes, workplace issues and concerns, and how the profession has changed from 1999 to 2008. Levine R . Longitudinal Emergency Medical Technician Attributes and Demographic Study (LEADS) design and methodology. Prehosp Disaster Med. 2016;31(Suppl. 1):s7-s17.

Societal pessimism in Japan, the United States, and The Netherlands

AbstractThis paper starts out with a theoretical argument, based on panel data, that public mood in general and societal pessimism in particular should be measured from an explicitly temporal perspective. Next, based on a survey among more than 200 Japanese students and a wide array of existing (longitudinal) data sources in three different languages and covering several decades it is shown that public mood in three quite different countries – first and foremost Japan, but also the United States and The Netherlands – is quite apprehensive. In all these three countries societal pessimism can be observed during the past quarter century. Finally, utilizing a MDSD approach a few possible tentative explanations for this observed pessimism are sought.

Learning temporal weights of clinical events using variable importance.

BackgroundLongitudinal data sources, such as electronic health records (EHRs), are very valuable for monitoring adverse drug events (ADEs). However, ADEs are heavily under-reported in EHRs. Using machine learning algorithms to automatically detect patients that should have had ADEs reported in their health records is an efficient and effective solution. One of the challenges to that end is how to take into account the temporality of clinical events, which are time stamped in EHRs, and providing these as features for machine learning algorithms to exploit. Previous research on this topic suggests that representing EHR data as a bag of temporally weighted clinical events is promising; however, the weights were in that case pre-assigned according to their time stamps, which is limited and potentially less accurate. This study therefore focuses on how to learn weights that effectively take into account the temporality and importance of clinical events for ADE detection.MethodsVariable importance obtained from the random forest learning algorithm is used for extracting temporal weights. Two strategies are proposed for applying the learned weights: weighted aggregation and weighted sampling. The first strategy aggregates the weighted clinical events from different time windows to form new features; the second strategy retains the original features but samples them by using their weights as probabilities when building each tree in the forest. The predictive performance of random forest models using the learned weights with the two strategies is compared to using pre-assigned weights. In addition, to assess the sensitivity of the weight-learning procedure, weights from different granularity levels are evaluated and compared.ResultsIn the weighted sampling strategy, using learned weights significantly improves the predictive performance, in comparison to using pre-assigned weights; however, there is no significant difference between them in the weighted aggregation strategy. Moreover, the granularity of the weight learning procedure has a significant impact on the former, but not on the latter.ConclusionsLearning temporal weights is significantly beneficial in terms of predictive performance with the weighted sampling strategy. Moreover, weighted aggregation generally diminishes the impact of temporal weighting of the clinical events, irrespective of whether the weights are pre-assigned or learned.

The Vanderbilt Memory & Aging Project: Study Design and Baseline Cohort Overview.

Vascular health factors frequently co-occur with Alzheimer's disease (AD). A better understanding of how systemic vascular and cerebrovascular health intersects with clinical and pathological AD may inform prevention and treatment opportunities. To establish the Vanderbilt Memory & Aging Project, a case-control longitudinal study investigating vascular health and brain aging, and describe baseline methodology and participant characteristics. From September 2012 to November 2014, 335 participants age 60- 92 were enrolled, including 168 individuals with mild cognitive impairment (MCI, 73±8 years, 41% female) and 167 age-, sex-, and race-matched cognitively normal controls (NC, 72±7 years, 41% female). At baseline, participants completed a physical and frailty examination, fasting blood draw, neuropsychological assessment, echocardiogram, cardiac MRI, and brain MRI. A subset underwent 24-hour ambulatory blood pressure monitoring and lumbar puncture for cerebrospinal fluid (CSF) collection. As designed, participant groups were comparable for age (p = 0.31), sex (p = 0.95), and race (p = 0.65). MCI participants had greater Framingham Stroke Risk Profile scores (p = 0.008), systolic blood pressure values (p = 0.008), and history of left ventricular hypertrophy (p = 0.04) than NC participants. As expected, MCI participants performed worse on all neuropsychological measures (p-values < 0.001), were more likely to be APOEɛ4 carriers (p = 0.02), and had enhanced CSF biomarkers, including lower Aβ42 (p = 0.02), higher total tau (p = 0.004), and higher p-tau (p = 0.02) compared to NC participants. Diverse sources of baseline and longitudinal data will provide rich opportunities to investigate pathways linking vascular and cerebrovascular health, clinical and pathological AD, and neurodegeneration contributing to novel strategies to delay or prevent cognitive decline.

The burden of clostridium difficile infection: estimates of the incidence of CDI from U.S. Administrative databases.

BackgroundMany administrative data sources are available to study the epidemiology of infectious diseases, including Clostridium difficile infection (CDI), but few publications have compared CDI event rates across databases using similar methodology. We used comparable methods with multiple administrative databases to compare the incidence of CDI in older and younger persons in the United States.MethodsWe performed a retrospective study using three longitudinal data sources (Medicare, OptumInsight LabRx, and Healthcare Cost and Utilization Project State Inpatient Database (SID)), and two hospital encounter-level data sources (Nationwide Inpatient Sample (NIS) and Premier Perspective database) to identify CDI in adults aged 18 and older with calculation of CDI incidence rates/100,000 person-years of observation (pyo) and CDI categorization (onset and association).ResultsThe incidence of CDI ranged from 66/100,000 in persons under 65 years (LabRx), 383/100,000 in elderly persons (SID), and 677/100,000 in elderly persons (Medicare). Ninety percent of CDI episodes in the LabRx population were characterized as community-onset compared to 41 % in the Medicare population. The majority of CDI episodes in the Medicare and LabRx databases were identified based on only a CDI diagnosis, whereas almost ¾ of encounters coded for CDI in the Premier hospital data were confirmed with a positive test result plus treatment with metronidazole or oral vancomycin. Using only the Medicare inpatient data to calculate encounter-level CDI events resulted in 553 CDI events/100,000 persons, virtually the same as the encounter proportion calculated using the NIS (544/100,000 persons).ConclusionsWe found that the incidence of CDI was 35 % higher in the Medicare data and fewer episodes were attributed to hospital acquisition when all medical claims were used to identify CDI, compared to only inpatient data lacking information on diagnosis and treatment in the outpatient setting. The incidence of CDI was 10-fold lower and the proportion of community-onset CDI was much higher in the privately insured younger LabRx population compared to the elderly Medicare population. The methods we developed to identify incident CDI can be used by other investigators to study the incidence of other infectious diseases and adverse events using large generalizable administrative datasets.Electronic supplementary materialThe online version of this article (doi:10.1186/s12879-016-1501-7) contains supplementary material, which is available to authorized users.

Population-based study of giant cell tumor of bone in Sweden (1983–2011)

IntroductionGiant-cell tumor of bone (GCTB) is a locally aggressive histologically benign neoplasm with a less common malignant counterpart. Longitudinal data sources on GCTB are sparse, limited to single institution case series or surgical outcomes studies. The Swedish Cancer Registry is one of the few national population-based databases recording GCTB, representing a unique source to study GCTB epidemiology. We estimated incidence rate (IR) and overall mortality rates based on registry data. Materials and methodsWe identified patients with a GCTB diagnosis in the Swedish Cancer Registry from 1983 to 2011: benign (ICD-7 196.0–196.9; PAD 741) and malignant (PAD 746). Results were stratified by age at diagnosis, gender, and anatomical lesion location. ResultsThe cohort included 337 GCTB cases (IR of 1.3 per million persons per year). The majority (n=310) had primary benign GCTB (IR of 1.2 per million per year). Median age at diagnosis was 34 years (range 10–88) with 54% (n=183) females. Malignant to benign ratio for women was 0.095 (16/167) and for men 0.077 (11/143). Incidence was highest in the 20–39 years age group (IR of 2.1 per million per year). The most common lesion sites were distal femur and proximal tibia. Mortality at 20 years from diagnosis was 14% (n=48) and was slightly higher for axial (17%; n=6) and pelvic (17%; n=4) lesions. Recurrence occurred in 39% of primary benign cases and 75% of primary malignant cases. ConclusionsIn our modern population-based series primary malignant cases were uncommon (8%), peak incidence 20–39 years with slight predominance in women. Recurrence rates remain significant with overall 39% occurring in benign GCTB, and 75% in malignant form. The linkage between databases allowed the first population based estimates of the proportion of patients who received surgery at initial GCTB diagnosis, and those who also received subsequent surgeries.

Economic consequences of workplace injuries in the United States: Findings from the National Longitudinal Survey of Youth (NLSY79).

This study explored economic consequences of work-related injuries using a longitudinal data source. Data were from the National Longitudinal Survey of Youth, 1979 cohort (n = 12,686). Short-term consequences were measured when the injury was reported. "Difference-in-differences" approach was applied to estimate income and wealth disparities between injured and non-injured workers before and after injury. Fixed effects models were used to identify variations over time. The annual earnings growth was $3,715 (in 2000 dollars) less for workers with DAFW injury and $1,152 less for workers with NDAFW injury compared to non-injured workers during a 10-year follow-up. Lost wages and disability following injury contributed to income loss for injured workers, but the loss was moderated by union membership. After controlling for confounders, income disparities persisted, but family wealth differences did not. Occupational injuries exacerbate income inequality. Efforts to reduce such disparities should include workplace safety and health enforcement.

What Happened in Sandtown-Winchester? Understanding the Impacts of a Comprehensive Community Initiative

We use multiple longitudinal data sources and propensity score matching to assess the long-term outcomes of the Neighborhood Transformation Initiative (NTI) in the Sandtown-Winchester community of Baltimore. This comprehensive community initiative, implemented in the 1990s, remains one of the most well-known urban revitalization projects in the country, due to its significant funding (more than $100 million) and comprehensive approach to neighborhood redevelopment, including housing construction, education reform, and employment services. We find significant increases in homeownership and reductions in unemployment in Sandtown. However, there were limited gains elsewhere, as poverty remained high and local schools did not show sustained improvement. Our findings speak to the durability of social inequality in high-poverty and racially segregated neighborhoods, and underscore the need to further develop rigorous standards for research that evaluates community-level interventions.

Reproductive History and Later-Life Comorbidity Trajectories: A Medicare-Linked Cohort Study From the Utah Population Database.

Reproductive lives of men and women may provide significant insight into later-life morbidity and mortality. Sociological, biological, and evolutionary theories predict a relationship between reproductive history and later-life health; however, current research is lacking consensus on the direction of the relationship. Parity, early age at first birth and last birth, birth weight of offspring, having a child die as an infant, and having a preterm birth may have long-term effects on health for both men and women. In this study, the relationship between these measures of reproductive history and later-life health is examined using the Utah Population Database (a rich source of longitudinal data), and Medicare claims data from 1992-2009. Later-life health is measured using annual Charlson comorbidity index scores, a construct that summarizes most serious illnesses afflicting older individuals. Group-based trajectory modeling that accounts for nonrandom attrition due to death is used to identify the number and types of morbidity trajectories by sex and age for 52,924 individuals aged 65-84 in 1992. For females, early age at first birth, high parity, and having a preterm or high-birth-weight baby are associated with increased risks of comorbidity; later age at last birth is associated with a decreased risk of comorbidity. For males, early age at first birth and having a child with an abnormal birth weight leads to increased risk of comorbidity. The results suggest that both biological and social factors play important roles in the relationships between fertility and morbidity profiles at older ages.

The analysis of dynamic travel mode choice: a heterogeneous hidden Markov approach

Discrete choices are often analyzed statically. The limitations of static models become more obvious when employing them in more long-term travel demand forecasting. The research gap lies in a theoretical model which is dynamically formulated, and in readily available longitudinal data sources. To address this, a heterogeneous hidden Markov modeling approach (HMM) is proposed in this paper to model dynamic discrete choices. Both longitudinal and cross-sectional heterogeneity are considered. The approach is demonstrated on a travel mode choice application using ten-wave Puget Sound Transport Panel data coupled with some other supplementary data sources. Results indicate that travelers’ long-term life-cycle stages have an enduring impact when shifted to different mode choice states, wherein sensitivities to travel time and cost vary. Empirical results are put in line with static discrete choice models. The paper demonstrates that the family of HMM models provide the best fitting model. The dynamic model has superior explanatory power in fitting longitudinal data and thus shall provide more accurate estimates for planning and policy analyses. The proposed approach can be generalized to study other short/mid-term travel behavior. The estimated model can be easily calibrated and transferred for applications elsewhere.

Swapping bricks for clicks: Crowdsourcing longitudinal data on Amazon Turk

Locating reliable sources of generalizable longitudinal data is an extremely important issue for business research. The aim of this paper was to empirically verify that crowdsourcing can be used to source longitudinal samples. Specifically, three studies assess reliability of the Amazon Mechanical Turk Marketplace (MTurk). All three studies demonstrate that MTurk is a reliable, inexpensive source for generalizable longitudinal data. Study 1 (n=752) examines the two-month re-response rate (study 1, n=752; 75%) of a US MTurk sample. Study 2 (n=373) investigates the four- and eight-month re-response rate (56 and 38%, respectively) of a US immigrant sample. Study 3 examines the thirteen-month re-response rate (47%). Each study demonstrates minimal non-response biases and longitudinal response consistency, in terms of both demographics and personality traits. This study also independently verifies the accuracy of self-report state of residence for 94% of the participants.

Alcohol problem recognition and help seeking in adolescents and young adults at varying genetic and environmental risk

IntroductionAlcohol use disorder symptoms frequently occur in adolescents and younger adults who seldom acknowledge a need for help. We identified sociodemographic, clinical, and familial predictors of alcohol problem recognition and help seeking in an offspring of twin sample. MethodWe analyzed longitudinal data from the Children of Alcoholics and Twins as Parents studies, which are combinable longitudinal data sources due to their equivalent design. We analyzed respondents (n=1073, 56.0% of the total sample) with alcohol use disorder symptoms at the baseline interview. Familial characteristics included perceptions of alcohol problems and help seeking for alcohol problems within the immediate family and a categorical variable indicating genetic and environmental risk. We used logistic regression to examine predictors of alcohol problem recognition and help seeking. ResultsApproximately 25.9% recognized their alcohol problems and 26.7% sought help for drinking. In covariate-adjusted analyses, help seeking among family members predicted problem recognition, several clinical characteristics predicted both problem recognition and help seeking, and familial risk predicted help seeking. Alcohol problem recognition mediated the association between alcohol use disorder symptoms and incident help seeking. ConclusionsFacilitating the self-recognition of alcohol use disorder symptoms, and perhaps the awareness of family members’ help seeking for alcohol problems, may be potentially promising methods to facilitate help seeking.

65. Bridging The Gap: A New Model of Transition Care for Young Adults With Serious Chronic Conditions

65. Bridging The Gap: A New Model of Transition Care for Young Adults With Serious Chronic Conditions

Reinventing the Role of the University Researcher

This study examines the structuring of university–community research partnerships that facilitate theoretically grounded research while also generating findings that community partners find actionable. We analyze one partnership that positions university-based researchers as members of a team working to create, maintain, and use a longitudinal multiagency data source. Through our focus on the evolution of this university–community collaboration, we show how researchers established their commitment to a mutually beneficial exchange and how data-driven action emerged when community agencies assumed ownership and prioritized action throughout the research process.