Missing not at random in end of life care studies: multiple imputation and sensitivity analysis on data from the ACTION study

Conceptual framework as a guide to choose appropriate imputation method for missing values in a clinical structured dataset

BACKGROUND

Missing data is a common challenge in structured datasets, and numerous methods are available for imputing these missing values. While all of these imputation methods address the issue of incomplete data, it is important to note that some methods perform better than others in terms of their effectiveness. A thorough data analysis can help a researcher identify a given dataset's most appropriate imputation approach, leading to more reliable analytical results. The primary objective of this study is to develop a conceptual framework that integrates various data imputation methods.

METHODS

This study was conducted in two main steps. First, we defined the conceptual components and their interrelationships by identifying and categorizing primary concepts through a secondary analysis of our previous systematic review, which examined 58 studies to uncover influential factors for selecting optimal imputation methods. Second, we analyzed the implementation process, focusing on the properties of missing values and selecting appropriate imputation techniques while verifying the underlying assumptions according to the estimand framework from the ICH E9(R1) Guideline to ensure unbiased estimates and enhance the credibility of our findings.

RESULTS

The findings from the secondary analysis suggest that the primary concepts of the developed conceptual framework directly influence the selection of appropriate imputation methods.

CONCLUSIONS

This integrated structure will enable researchers to select the most suitable imputation method based on the specific characteristics and conditions of the dataset under investigation. By employing the appropriate imputation method, the study aims to enhance the overall quality and trustworthiness of the analytical outcomes derived from the research dataset.

Social Determinants of Quality of Life in the Last Year of Life Among Community-Dwelling Older Adults with Multimorbidity

Quality of life (QOL) at the end of life often declines in relation to various determinants, yet the role of social determinants, including social capital, in end-of-life care is frequently overlooked. This study aims to examine the association between social determinants and QOL in the last year of life and to test the mediating role of social capital in the relationship between social determinants and QOL among older adults with multimorbidity (MM). We used secondary data from the National Health and Aging Trends Study (NHATS) in Rounds 10 and 11, involving 3,085 adults aged 65 and older. The final analysis comprised 230 participants. Multiple linear regression was conducted to assess the associations between social determinants and QOL, and path analysis was used to evaluate the mediating effect of social capital. The regression model showed that social capital was positively and significantly associated with QOL (β = 0.378, 95% CI [0.099, 0.657], SE = 0.139), as were mental conditions (β = 0.614, 95% CI [0.167, 1.062], SE = 0.194). The mediation analysis demonstrated that social capital functioned as a complementary mediator, partially mediating the relationship between mental conditions and QOL in the last year of life. These findings underscore the potential role of social capital in enhancing QOL at the end of life, particularly through its influence on mental health. The study highlights the need for healthcare practices and policies that promote social support systems and community-based care for older adults with MM. By addressing social capital, end-of-life care could be improved, resulting in better overall well-being for individuals facing the last stages of life.

Relationship between reasons for intermittent missing patient-reported outcomes data and missing data mechanisms

PURPOSE

Non-response (NR) to patient-reported outcome (PRO) questionnaires may cause bias if not handled appropriately. Collecting reasons for NR is recommended, but how reasons for NR are related to missing data mechanisms remains unexplored. We aimed to explore this relationship for intermittent NRs.

METHODS

Patients with multiple myeloma completed validated PRO questionnaires at enrolment and 12 follow-up time-points. NR was defined as non-completion of a follow-up assessment within seven days, which triggered contact with the patient, recording the reason for missingness and an invitation to complete the questionnaire (denoted "salvage response"). Mean differences between salvage and previous on-time scores were estimated for groups defined by reasons for NR using linear regression with clustered standard errors. Statistically significant mean differences larger than minimal important difference thresholds were interpreted as "missing not at random" (MNAR) mechanism (i.e. assumed to be related to declining health), and the remainder interpreted as aligned with "missing completely at random" (MCAR) mechanism (i.e. assumed unrelated to changes in health).

RESULTS

Most (7228/7534 (96%)) follow-up questionnaires were completed; 11% (802/7534) were salvage responses. Mean salvage scores were compared to previous on-time scores by reason: those due to hospital admission, mental or physical reasons were worse in 10/22 PRO domains; those due to technical difficulties/procedural errors were no different in 21/22 PRO domains; and those due to overlooked/forgotten or other/unspecified reasons were no different in any domains.

CONCLUSION

Intermittent NRs due to hospital admission, mental or physical reasons were aligned with MNAR mechanism for nearly half of PRO domains, while intermittent NRs due to technical difficulties/procedural errors or other/unspecified reasons generally were aligned with MCAR mechanism.

Patterns in health care use and intensity for diagnosed and undiagnosed cognitive impairment in the older australian community: Implications for primary care management

Objectives

While the economic burden imposed by dementia is well-documented, findings are mixed on health care use for those with mild cognitive impairment (MCI). Our objective was to analyse annual, non-hospital medical and pharmaceutical use patterns for older people with undiagnosed MCI and diagnosed dementia, living in the Australian community.

Methods

We analysed panel data from a community sample, the Sydney Memory and Ageing Study (Australia), linked to administrative data on health care use, using two-part models to estimate the probability of using health care and the annual costs incurred by study participants.

Results

People with MCI, unaware of their diagnoses, were significantly less likely to incur annual pathology and diagnostic imaging costs relative to cognitively normal individuals. This effect was concentrated in individuals with MCI who had non-amnestic symptoms, lived alone, or had limited carer support. Compared to cognitively normal individuals, people with MCI were predicted to have slightly lower annual costs for broad medical care categories related to the management and diagnosis of cognitive impairment, and people with dementia, substantially higher professional attendances, and pharmaceutical costs. These findings were consistent across estimation models adjusting for attrition over the study.

Policy implications

Diagnosis and symptom management in primary care may enable individuals with MCI to improve their quality of life and prevent more costly future health care use. However, our study found potential gaps in medical service use for people with undiagnosed MCI in the community, especially when they had less support or did not have memory symptoms. Primary care services may need to better diagnose and target these individuals.

Identify the most appropriate imputation method for handling missing values in clinical structured datasets: a systematic review

BACKGROUND AND OBJECTIVES

Comprehending the research dataset is crucial for obtaining reliable and valid outcomes. Health analysts must have a deep comprehension of the data being analyzed. This comprehension allows them to suggest practical solutions for handling missing data, in a clinical data source. Accurate handling of missing values is critical for producing precise estimates and making informed decisions, especially in crucial areas like clinical research. With data's increasing diversity and complexity, numerous scholars have developed a range of imputation techniques. To address this, we conducted a systematic review to introduce various imputation techniques based on tabular dataset characteristics, including the mechanism, pattern, and ratio of missingness, to identify the most appropriate imputation methods in the healthcare field.

MATERIALS AND METHODS

We searched four information databases namely PubMed, Web of Science, Scopus, and IEEE Xplore, for articles published up to September 20, 2023, that discussed imputation methods for addressing missing values in a clinically structured dataset. Our investigation of selected articles focused on four key aspects: the mechanism, pattern, ratio of missingness, and various imputation strategies. By synthesizing insights from these perspectives, we constructed an evidence map to recommend suitable imputation methods for handling missing values in a tabular dataset.

RESULTS

Out of 2955 articles, 58 were included in the analysis. The findings from the development of the evidence map, based on the structure of the missing values and the types of imputation methods used in the extracted items from these studies, revealed that 45% of the studies employed conventional statistical methods, 31% utilized machine learning and deep learning methods, and 24% applied hybrid imputation techniques for handling missing values.

CONCLUSION

Considering the structure and characteristics of missing values in a clinical dataset is essential for choosing the most appropriate data imputation technique, especially within conventional statistical methods. Accurately estimating missing values to reflect reality enhances the likelihood of obtaining high-quality and reusable data, contributing significantly to precise medical decision-making processes. Performing this review study creates a guideline for choosing the most appropriate imputation methods in data preprocessing stages to perform analytical processes on structured clinical datasets.

Sociodemographic disparities and potential biases in persistent pain estimates: Findings from 5 waves of the Irish Longitudinal Study on Ageing (TILDA)

BACKGROUND

Pain is a prevalent, debilitating condition among older adults. Much evidence on this topic comes from cohort studies, which may be affected by attrition and measurement bias. Little is known about the impact of these biases on pain estimates for European older adults. Additionally, there is a lack of longitudinal research on pain and sociodemographic disparities in Irish older adults.

METHODS

We analysed data from 8171 participants (aged ≥50 at baseline) across five waves of the Irish Longitudinal Study on Ageing. Longitudinal pain severity and sociodemographic disparities in pain were explored visually and using a latent growth curve model. Using multivariate logistic regression, we examined bias due to attrition at later waves associated with reported pain at Wave 1. Measurement biases due to reporting heterogeneity were assessed by investigating associations between sociodemographic factors and pain-related disability for given pain levels.

RESULTS

Wave 1 severe pain was associated with increased odds of attrition due to death by Wave 5 (AOR: 1.63, 95% CI: 1.20, 2.19). Not having private health insurance was associated with increased odds of pain-related disability at Wave 1, controlling for pain severity (AOR: 1.37, 95% CI: 1.15, 1.64). These results suggested mortality bias and reporting heterogeneity measurement bias, respectively. Sex, education level, and private health insurance status disparities in pain were observed longitudinally.

CONCLUSIONS

Mortality bias and reporting heterogeneity measurement bias must be accounted for to improve older adult pain estimates. There is a need for policymakers to address sociodemographic disparities in older adult pain levels.

SIGNIFICANCE

This study highlights a need to address bias in the estimation of pain in observational studies of older adults. Understanding the sources and extent of these biases is important so that health practices and policies to address pain disparities can be guided by accurate estimates. Women, those with lower educational attainment, and those without private health insurance were found to have the highest pain burden longitudinally, suggesting a need for targeted interventions for these groups in Ireland and internationally.

Frailty prevalence in 42 European countries by age and gender: development of the SHARE Frailty Atlas for Europe

Comparative frailty prevalence data across European countries is sparse due to heterogeneous measurement methods. The Survey of Health, Ageing and Retirement (SHARE) initiative conducted interviews with probability sampling of non-institutionalized elderly people in several European countries. Previous frailty analyses of SHARE datasets were limited to initial SHARE countries and did not provide age- and gender-stratified frailty prevalence. Our aim was to provide age- and gender-stratified frailty prevalence estimates in all European countries, with predictions where necessary. From 29 SHARE participating countries, 311,915 individual surveys were analyzed. Frailty prevalence was estimated by country and gender in 5-year age bands using the SHARE Frailty Instrument and a frailty index. Association of frailty prevalence with age, gender, and GDP per capita (country-specific economic indicator for predictions) was investigated in multivariate mixed logistic regression models with or without multiple imputation. Female gender and increasing age were significantly associated with higher frailty prevalence. Higher GDP per capita, with or without purchasing power parity adjustment, was significantly associated with lower frailty prevalence in the 65-79 age groups in all analyses. Observed and predicted data on frailty rates by country are provided in the interactive SHARE Frailty Atlas for Europe. Our study provides age- and gender-stratified frailty prevalence estimates for all European countries, revealing remarkable between-country heterogeneity. Higher frailty prevalence is strongly associated with lower GDP per capita, underlining the importance of investigating transferability of evidence across countries at different developmental levels and calling for improved policies to reduce inequity in risk of developing frailty across European countries.

Prevalence of loneliness and associations with health behaviours and body mass index in 5835 people living with and beyond cancer: a cross-sectional study

BACKGROUND

A cancer diagnosis and its treatment may be an especially isolating experience. Despite evidence that positive health behaviours can improve outcomes for people living with and beyond cancer (LWBC), no studies have examined associations between loneliness and different health behaviours in this population. This study aimed to describe the prevalence of loneliness in a large sample of UK adults LWBC and to explore whether loneliness was associated with multiple health behaviours.

METHODS

Participants were adults (aged ≥ 18 years) diagnosed with breast, prostate or colorectal cancer who completed the Health and Lifestyle After Cancer Survey. Loneliness was reported using the UCLA loneliness score, dichotomised into higher (≥ 6) versus lower (< 6) loneliness. Engagement in moderate-to-vigorous physical activity, dietary intake, smoking status, alcohol use, and self-reported height and weight were recorded. Behaviours were coded to reflect meeting or not meeting the World Cancer Research Fund recommendations for people LWBC. Logistic regression analyses explored associations between loneliness and health behaviours. Covariates were age, sex, ethnicity, education, marital status, living situation, cancer type, spread and treatment, time since treatment, time since diagnosis and number of comorbid conditions. Multiple imputation was used to account for missing data.

RESULTS

5835 participants, mean age 67.4 (standard deviation = 11.8) years, completed the survey. 56% were female (n = 3266) and 44% (n = 2553) male, and 48% (n = 2786) were living with or beyond breast cancer, 32% (n = 1839) prostate, and 21% (n = 1210) colorectal. Of 5485 who completed the loneliness scale, 81% (n = 4423) of participants reported lower and 19% (n = 1035) higher loneliness. After adjustment for confounders, those reporting higher levels of loneliness had lower odds of meeting the WCRF recommendations for moderate-to-vigorous physical activity (Odds Ratio [OR] 0.78, 95% Confidence Internal [CI], 0.67, 0.97, p =.028), fruit and vegetable intake (OR 0.81, CI 0.67, 1.00, p =.046), and smoking (OR 0.62, 0.46, 0.84, p =.003). No association was observed between loneliness and the other dietary behaviours, alcohol, or body mass index.

CONCLUSIONS

Loneliness is relatively common in people LWBC and may represent an unmet need. People LWBC who experience higher levels of loneliness may need additional support to improve their health behaviours.

Impact of Short-Course Palliative Radiation Therapy on Pancreatic Cancer-Related Pain: Prospective Phase 2 Nonrandomized PAINPANC Trial

PURPOSE

Clinical evidence is limited regarding palliative radiation therapy for relieving pancreatic cancer-related pain. We prospectively investigated pain response after short-course palliative radiation therapy in patients with moderate-to-severe pancreatic cancer-related pain.

METHODS AND MATERIALS

In this prospective phase 2 single center nonrandomized trial, 30 patients with moderate-to-severe pain (5-10, on a 0-10 scale) of pancreatic cancer refractory to pain medication, were treated with a short-course palliative radiation therapy; 24 Gy in 3 weekly fractions (2015-2018). Primary endpoint was defined as a clinically relevant average decrease of ≥2 points in pain severity, compared with baseline, within 7 weeks after the start of treatment. Secondary endpoint was global quality of life (QoL), with a clinically relevant increase of 5 to 10 points (0-100 scale). Pain severity reduction and QoL were assessed 9 times using the Brief Pain Inventory and European Organization for Research and Treatment of Cancer Quality of Life Questionnaire-C15-PAL, respectively. Both outcomes were analyzed using joint modeling. In addition, acute toxicity based on clinician reporting and overall survival (OS) were assessed.

RESULTS

Overall, 29 of 30 patients (96.7%) received palliative radiation therapy. At baseline, the median oral morphine equivalent daily dose was 129.5 mg (range, 20.0-540.0 mg), which decreased to 75.0 mg (range, 15.0-360.0 mg) after radiation (P = .021). Pain decreased on average 3.15 points from baseline to 7 weeks (one-sided P = .045). Patients reported a clinically relevant mean pain severity reduction from 5.9 to 3.8 points (P = .011) during the first 3 weeks, which further decreased to 3.2 until week 11, ending at 3.4 (P = .006) in week 21 after the first radiation therapy fraction. Global QoL significantly improved from 50.5 to 60.8 during the follow-up period (P = .001). Grade 3 acute toxicity occurred in 3 patients and no grade 4 to 5 toxicity was observed. Median OS was 11.8 weeks, with a 13.3% 1-year actuarial OS rate.

CONCLUSIONS

Short-course palliative radiation therapy for pancreatic cancer-related pain was associated with rapid, clinically relevant reduction in pain severity, and clinically relevant improvement in global QoL, with mostly mild toxicity.

Perceiving a need for dietary change in adults living with and beyond cancer: A cross-sectional study

BACKGROUND

Many people living with and beyond cancer (LWBC) do not meet dietary recommendations. To implement a healthier diet, people LWBC must perceive a need to improve their diet.

METHODS

Participants included people diagnosed with breast, prostate or colorectal cancer in the UK. Two binary logistic regression models were conducted with perceived need for dietary change as the outcome (need to improve vs. no need). Predictor variables included demographic and clinical characteristics, receipt of dietary advice, and either body mass index (BMI) or adherence to seven relevant World Cancer Research Fund (WCRF) dietary recommendations.

RESULTS

The sample included 5835 responses. Only 31% perceived a need to improve their diet. Being younger (odds ratio [OR] 0.95, 95% confidence interval [CI] = 94-0.95), female (OR = 1.33, 95% CI = 1.15-1.53), not of white ethnicity (OR = 1.8, 95% CI = 1.48-2.27), not married/cohabiting (OR = 1.32, 95% CI = 1.16-1.52) and having received dietary advice (OR = 1.36, 95% CI = 1.43-1.86) was associated with an increased odds of perceiving a need to improve diet. This association was also seen for participants with two or more comorbidities (OR = 1.31, 95% CI = 1.09-1.57), those not meeting the recommendations for fruit and vegetables (OR = 0.47, 95% CI = 0.41-0.55), fat (OR = 0.67, 95% CI = 0.58-0.77), and sugar (OR = 0.86, 95% CI = 0.75-0.98) in the dietary components model and those who had a higher BMI (OR = 1.53, 95% CI = 1.32-1.77) in the BMI model.

CONCLUSIONS

Most of this sample of people LWBC did not perceive a need to improve their diet. More research is needed to understand the reasons for this and to target these reasons in dietary interventions.

The Joint Awareness Score: A Shortened, Simplified, Improved Alternative to the Forgotten Joint Score

Background

The Forgotten Joint Score (FJS) is a 12-question patient-reported outcomes measure created to measure a patient's awareness of their artificial joint. The FJS has attained wide usage, though it is not without weaknesses. Our patients report that the semantics of the English translation are flawed and that the 5 answer options for each question are poorly differentiated. Additionally, the FJS will result in no score if 3 or more questions are unanswered. This prompted the development of an alternative patient-reported outcomes measure, the Joint Awareness Score (JAS), that builds upon the core concept of joint awareness underlying the FJS, but that is easier to understand and shorter to complete. We completed an exploratory, pilot study to evaluate this outcomes instrument. Our hypothesis is that the JAS will correlate strongly with the FJS and could be used as a substitute.

Methods

Knee arthroplasty patients in a prospective registry were administered the FJS and the JAS. Internal consistency and correlation were calculated with Cronbach's alpha and Pearson's correlation coefficient, respectively.

Results

This study included 174 patients. Cronbach's alpha for FJS was 0.97 for 6 months and 0.97 for 12 months, whereas JAS was 0.89 at 6 months and 0.85 at 12 months. Pearson correlation comparing FJS and JAS at 6 months was 0.88 (95% confidence interval: 0.83, 0.92) and 0.86 (95% confidence interval: 0.78, 0.92) at 12 months.

Conclusions

The Joint Awareness Score is a new patient-reported outcomes measure that is a substitute for the FJS, with half the number of questions, improved semantics, and simplified answers.

Palliative clinical trials in advanced chronic liver disease: Challenges and opportunities

Patients with advanced chronic liver disease have a complex symptom burden and many are not candidates for curative therapy. Despite this, provision of palliative interventions remains woefully inadequate, with an insufficient evidence base being a contributory factor. Designing and conducting palliative interventional trials in advanced chronic liver disease remains challenging for a multitude of reasons. In this manuscript we review past and ongoing palliative interventional trials. We identify barriers and facilitators and offer guidance on addressing these challenges. We hope that this will reduce the inequity in palliative care provision in advanced chronic liver disease.

A Bayesian Approach for Estimating the Survivor Average Causal Effect When Outcomes Are Truncated by Death in Cluster-Randomized Trials

Many studies encounter clustering due to multicenter enrollment and nonmortality outcomes, such as quality of life, that are truncated due to death-that is, missing not at random and nonignorable. Traditional missing-data methods and target causal estimands are suboptimal for statistical inference in the presence of these combined issues, which are especially common in multicenter studies and cluster-randomized trials (CRTs) carried out among the elderly or seriously ill. Using principal stratification, we developed a Bayesian estimator that jointly identifies the always-survivor principal stratum in a clustered/hierarchical data setting and estimates the average treatment effect among them (i.e., the survivor average causal effect (SACE)). In simulations, we observed low bias and good coverage with our method. In a motivating CRT, the SACE and the estimate from complete-case analysis differed in magnitude, but both were small, and neither was incompatible with a null effect. However, the SACE estimate has a clear causal interpretation. The option to assess the rigorously defined SACE estimand in studies with informative truncation and clustering can provide additional insight into an important subset of study participants. Based on the simulation study and CRT reanalysis, we provide practical recommendations for using the SACE in CRTs and software code to support future research.

Data quality considerations for evaluating COVID-19 treatments using real world data: learnings from the National COVID Cohort Collaborative (N3C)

BACKGROUND

Multi-institution electronic health records (EHR) are a rich source of real world data (RWD) for generating real world evidence (RWE) regarding the utilization, benefits and harms of medical interventions. They provide access to clinical data from large pooled patient populations in addition to laboratory measurements unavailable in insurance claims-based data. However, secondary use of these data for research requires specialized knowledge and careful evaluation of data quality and completeness. We discuss data quality assessments undertaken during the conduct of prep-to-research, focusing on the investigation of treatment safety and effectiveness.

METHODS

Using the National COVID Cohort Collaborative (N3C) enclave, we defined a patient population using criteria typical in non-interventional inpatient drug effectiveness studies. We present the challenges encountered when constructing this dataset, beginning with an examination of data quality across data partners. We then discuss the methods and best practices used to operationalize several important study elements: exposure to treatment, baseline health comorbidities, and key outcomes of interest.

RESULTS

We share our experiences and lessons learned when working with heterogeneous EHR data from over 65 healthcare institutions and 4 common data models. We discuss six key areas of data variability and quality. (1) The specific EHR data elements captured from a site can vary depending on source data model and practice. (2) Data missingness remains a significant issue. (3) Drug exposures can be recorded at different levels and may not contain route of administration or dosage information. (4) Reconstruction of continuous drug exposure intervals may not always be possible. (5) EHR discontinuity is a major concern for capturing history of prior treatment and comorbidities. Lastly, (6) access to EHR data alone limits the potential outcomes which can be used in studies.

CONCLUSIONS

The creation of large scale centralized multi-site EHR databases such as N3C enables a wide range of research aimed at better understanding treatments and health impacts of many conditions including COVID-19. As with all observational research, it is important that research teams engage with appropriate domain experts to understand the data in order to define research questions that are both clinically important and feasible to address using these real world data.

Principal component analysis of texture features for grading of meningioma: not effective from the peritumoral area but effective from the tumor area

PURPOSE

To investigate whether texture features from tumor and peritumoral areas based on sequence combinations can differentiate between low- and non-low-grade meningiomas.

METHODS

Consecutive patients diagnosed with meningioma by surgery (77 low-grade and 28 non-low-grade meningiomas) underwent preoperative magnetic resonance imaging including T1-weighted imaging (T1WI), T2-weighted imaging (T2WI), diffusion-weighted imaging (DWI), and contrast-enhanced T1WI (CE-T1WI). Manual segmentation of the tumor area was performed to extract texture features. Segmentation of the peritumoral area was performed for peritumoral high-signal intensity (PHSI) on T2WI. Principal component analysis was performed to fuse the texture features to principal components (PCs), and PCs of each sequence of the tumor and peritumoral areas were compared between low- and non-low-grade meningiomas. Only PCs with statistical significance were used for the model construction using a support vector machine algorithm. k-fold cross-validation with receiver operating characteristic curve analysis was used to evaluate diagnostic performance.

RESULTS

Two, one, and three PCs of T1WI, apparent diffusion coefficient (ADC), and CE-T1WI, respectively, for the tumor area, were significantly different between low- and non-low-grade meningiomas, while PCs of T2WI for the tumor area and PCs for the peritumoral area were not. No significant differences were observed in PHSI. Among models of sequence combination, the model with PCs of ADC and CE-T1WI for the tumor area showed the highest area under the curve (0.84).

CONCLUSION

The model with PCs of ADC and CE-T1WI for the tumor area showed the highest diagnostic performance for differentiating between low- and non-low-grade meningiomas. Neither PHSI nor PCs in the peritumoral area showed added value.

Sensitivity analyses for data missing at random versus missing not at random using latent growth modelling: a practical guide for randomised controlled trials

BACKGROUND

Missing data are ubiquitous in randomised controlled trials. Although sensitivity analyses for different missing data mechanisms (missing at random vs. missing not at random) are widely recommended, they are rarely conducted in practice. The aim of the present study was to demonstrate sensitivity analyses for different assumptions regarding the missing data mechanism for randomised controlled trials using latent growth modelling (LGM).

METHODS

Data from a randomised controlled brief alcohol intervention trial was used. The sample included 1646 adults (56% female; mean age = 31.0 years) from the general population who had received up to three individualized alcohol feedback letters or assessment-only. Follow-up interviews were conducted after 12 and 36 months via telephone. The main outcome for the analysis was change in alcohol use over time. A three-step LGM approach was used. First, evidence about the process that generated the missing data was accumulated by analysing the extent of missing values in both study conditions, missing data patterns, and baseline variables that predicted participation in the two follow-up assessments using logistic regression. Second, growth models were calculated to analyse intervention effects over time. These models assumed that data were missing at random and applied full-information maximum likelihood estimation. Third, the findings were safeguarded by incorporating model components to account for the possibility that data were missing not at random. For that purpose, Diggle-Kenward selection, Wu-Carroll shared parameter and pattern mixture models were implemented.

RESULTS

Although the true data generating process remained unknown, the evidence was unequivocal: both the intervention and control group reduced their alcohol use over time, but no significant group differences emerged. There was no clear evidence for intervention efficacy, neither in the growth models that assumed the missing data to be at random nor those that assumed the missing data to be not at random.

CONCLUSION

The illustrated approach allows the assessment of how sensitive conclusions about the efficacy of an intervention are to different assumptions regarding the missing data mechanism. For researchers familiar with LGM, it is a valuable statistical supplement to safeguard their findings against the possibility of nonignorable missingness.

TRIAL REGISTRATION

The PRINT trial was prospectively registered at the German Clinical Trials Register (DRKS00014274, date of registration: 12th March 2018).

Sequential Multiple Imputation for Real-World Health-Related Quality of Life Missing Data after Bariatric Surgery

One of the main challenges for the successful implementation of health-related quality of life (HRQoL) assessments is missing data. The current study examined the feasibility and validity of a sequential multiple imputation (MI) method to deal with missing values in the longitudinal HRQoL data from the Scandinavian Obesity Surgery Registry. All patients in the SOReg who received bariatric surgery between 1 January 2011 and 31 March 2019 (n = 47,653) were included for the descriptive analysis and missingness pattern exploration. The patients who had completed the short-form 36 (SF-36) at baseline (year 0), and one-, two-, and five-year follow-ups were included (n = 3957) for the missingness pattern simulation and the sequential MI analysis. Eleven items of the SF-36 were selected to create the six domains of SF-6D, and the SF-6D utility index of each patient was calculated accordingly. The multiply-imputed variables in previous year were used as input to impute the missing values in later years. The performance of the sequential MI was evaluated by comparing the actual values with the imputed values of the selected SF-36 items and index at all four time points. At the baseline and year 1, where missing proportions were about 20% and 40%, respectively, there were no statistically significant discrepancies between the distributions of the actual and imputed responses (all p-values > 0.05). In year 2, where the missing proportion was about 60%, distributions of the actual and imputed responses were consistent in 9 of the 11 SF-36 items. However, in year 5, where the missing proportion was about 80%, no consistency was found between the actual and imputed responses in any of the SF-36 items. Relatively high missing proportions in HRQoL data are common in clinical registries, which brings a challenge to analyzing the HRQoL of longitudinal cohorts. The experimental sequential multiple imputation method adopted in the current study might be an ideal strategy for handling missing data (even though the follow-up survey had a missing proportion of 60%), avoiding significant information waste in the multivariate analysis. However, the imputations for data with higher missing proportions warrant more research.