Improving statistical analysis of matched case-control studies

A comparative cross-sectional assessment of statistical knowledge of faculty across five health science disciplines

Introduction:

The purpose of this study was to compare statistical knowledge of health science faculty across accredited schools of dentistry, medicine, nursing, pharmacy, and public health.

Methods:

A probability sample of schools was selected, and all faculty at each selected school were invited to participate in an online statistical knowledge assessment that covered fundamental topics including randomization, study design, statistical power, confidence intervals, multiple testing, standard error, regression outcome, and odds ratio.

Results:

A total of 708 faculty from 102 schools participated. The overall response rate was 6.5%. Most (94.2%) faculty reported reading the peer-reviewed health-related literature. Respondents answered 66.2% of questions correctly across all questions and disciplines. Public health had the highest performance (80.7%) and dentistry the lowest (53.3%).

Conclusions:

Knowledge of statistics is essential for critically evaluating evidence and understanding the health literature. These study results identify a gap in knowledge by educators tasked with training the next generation of health science professionals. Recommendations for addressing this gap are provided.

A study of statistics knowledge among nurse faculty in schools with research doctorate programs

Statistics knowledge is essential for nursing faculty in both teaching and research roles. In the teaching role when discussing nursing research, nursing faculty are confronted with statistical concepts and statistical methods applications. Knowledge of fundamental statistical concepts is needed so that nursing faculty can understand and critically evaluate the literature. The purpose of this study was to assess nursing faculty knowledge of fundamental statistical concepts. A probability sample with a 7.7% response rate yielded participation from 164 nursing faculty from 26 accredited schools. Results showed that most faculty members (91.5%) read peer-reviewed health-related scientific journal articles. On average, nursing faculty answered 5.1 (SD = 1.6) out of 8 statistical knowledge questions correctly. Problematic concepts included randomization (43.3% correct), and interpreting a confidence interval (42.7%) and odds ratio (33.5%). The results of this study may be used to improve statistics education and training for future nursing faculty and strengthen scholarship for nursing faculty conducting research.

The relationship between spinal pain and temporomandibular joint disorders in Korea: a nationwide propensity score-matched study

BACKGROUND

Patients with temporomandibular joint disorder (TMD) often complain of pain in other areas. Several studies have been conducted on spinal pain in TMD patients, but have contained only limited information. Therefore, this study analyzed the relationship between TMD and spinal pain in greater detail by using nationwide data.

METHODS

A total of 12,375 TMD patients from the Korean National Health Insurance Review and Assessment database were analyzed. Controls were selected using propensity score-matching. The McNemar test, chi-square test, and paired t-test were used to compare the prevalence and severity of spinal pain between cases and matched controls. Logistic regression and linear regression models were used to analyze factors affecting the prevalence and severity of spinal pain in patients with TMD.

RESULTS

The annual period prevalence of TMD was 1.1%. The prevalence was higher in younger individuals than in individuals of other ages and was higher in women than in men. The medical expenditure for TMD per person was $86. Among TMD patients, 2.5% underwent surgical procedures and 0.3% were hospitalized. The prevalence of spinal pain in patients with TMD was 48%, whereas that in the control group was 34%. Increased severity of TMD was associated with an increased probability of spinal pain. The medical expenditure, mean number of visits, and lengths of treatment for spinal pain were greater for patients with TMD than for controls ($136 vs. $81, 4.8 days vs. 2.7 days, 5.5 days vs. 3.3 days). Higher TMD grade was associated with greater differences in average medical expenditure, number of visits, and lengths of treatment for spinal pain between cases and controls. Additionally, for women, living in a rural area and having an older age and more severe TMD were associated with a greater probability of spinal pain and higher medical expenditure related to spinal pain.

CONCLUSION

A strong association was observed between the presence of TMD and the presence of spinal pain. The association became stronger as the severity of TMD increased, indicating a positive correlation between the severity of TMD and spinal pain.

Extending Classification Algorithms to Case-Control Studies

Classification is a common technique applied to 'omics data to build predictive models and identify potential markers of biomedical outcomes. Despite the prevalence of case-control studies, the number of classification methods available to analyze data generated by such studies is extremely limited. Conditional logistic regression is the most commonly used technique, but the associated modeling assumptions limit its ability to identify a large class of sufficiently complicated 'omic signatures. We propose a data preprocessing step which generalizes and makes any linear or nonlinear classification algorithm, even those typically not appropriate for matched design data, available to be used to model case-control data and identify relevant biomarkers in these study designs. We demonstrate on simulated case-control data that both the classification and variable selection accuracy of each method is improved after applying this processing step and that the proposed methods are comparable to or outperform existing variable selection methods. Finally, we demonstrate the impact of conditional classification algorithms on a large cohort study of children with islet autoimmunity.

Parent–Child Interaction Therapy for children with autism spectrum disorder and a matched case-control sample

Parent–Child Interaction Therapy is an empirically based, behavioral parent training program for young children exhibiting disruptive behaviors. Parent–Child Interaction Therapy shows promise for treating disruptive behaviors in children with autism spectrum disorder. Treatment processes (i.e. treatment length and homework compliance), parenting skills, parenting stress, and behavioral outcomes (i.e. disruptive and externalizing behaviors and executive functioning) were compared in 16 children with autism spectrum disorder and 16 children without autism spectrum disorder matched on gender, age, and initial intensity of disruptive behaviors. Samples were statistically similar in terms of child receptive language, child race and ethnicity, parent age, gender and education, and number of two-parent families in treatment. Families received standard, mastery-based Parent–Child Interaction Therapy. Both groups demonstrated significant and clinically meaningful improvements in child disruptive and externalizing behavior and executive functioning, parenting skills, and parenting stress. Length of treatment, homework compliance, and parent and child outcomes did not differ significantly between groups. A subset of children with autism spectrum disorder also showed significant improvements in social responsiveness, adaptive skills, and restricted/repetitive behaviors. This study replicates and extends prior research by demonstrating that children with and without autism spectrum disorder experience similar benefits following Parent–Child Interaction Therapy. Findings may expand the availability and dissemination of time-limited, evidence-based interventions for autism spectrum disorder and comorbid disruptive behaviors.

How Often Do Orthopaedic Matched Case-Control Studies Use Matched Methods? A Review of Methodological Quality

BACKGROUND

Case-control studies are a common method of analyzing associations between clinical outcomes and potential risk factors. Matching cases to controls based on known confounding variables can decrease bias and allow investigators to assess the association of interest with increased precision. However, the analysis of matched data generally requires matched statistical methods, and failure to use these methods can lead to imprecise or biased results. The appropriate use of matched statistical methods in orthopaedic case-control studies has not been documented.

QUESTIONS/PURPOSES

(1) What proportion of matched orthopaedic case-control studies use the appropriate matched statistical analyses? (2) What study factors are associated with the use of appropriate matched statistical tests?

METHODS

All matched case-control studies published in the top 10 orthopaedic journals according to impact factor from 2007 to 2016 were identified by literature review. Studies using appropriate statistical techniques were identified by two independent evaluators; discrepancies were settled by a third evaluator, all with advanced training in biostatistics. The number of studies using appropriate matched statistical methods was compared with the number of studies reviewed. Logistic regression was used to identify key study factors (including journal, publication year, rank according to impact factor, number of matching factors, number of controls per case, and the inclusion of a biostatistician coauthor) associated with the use of appropriate statistical methods. Three hundred nineteen articles that were initially classified as case-control studies were screened, yielding 83 matched case-control studies. One hundred two of the excluded articles were cohort or cross-sectional studies that were misclassified as case-control studies. The median number of matching factors was 3.0 (range, 1-10) and the median number of controls per case was 1.0 (range, 0.5-6.0). Thirty studies (36%) had a statistician coauthor.

RESULTS

Thirty of the 83 included studies (36%) used appropriately matched methods throughout, 11 (13%) used matched methods for multivariable but not univariable analyses, and 42 (51%) used only unmatched methods, which we considered inappropriate. After controlling for the number of controls per case and publication year, we found that the inclusion of a statistician coauthor (70% versus 38%; odds ratio, 3.6; 95% confidence interval, 1.4-20.3; p = 0.01) and journal were associated with the use of appropriate methods.

CONCLUSIONS

Although matched case-control studies can be statistically more efficient study designs, in that they are capable of generating more precise effect size estimates than other kinds of retrospective research, most orthopaedic case-control studies use inappropriate statistical methods in their analyses. Additionally, the high degree of study misclassification indicates a need to more rigorously define differences among case-control, cohort, and cross-sectional study designs.

CLINICAL RELEVANCE

Failing to use matched statistical tests may lead to imprecise and/or biased effect estimates, which may lead to a tendency to overestimate or underestimate associations between possible risk factors and clinically relevant outcomes. Orthopaedic researchers should be cognizant of the risks and benefits of matching and should consult individuals with biostatistical expertise as needed to ensure that their statistical methods are appropriate and methodologically rigorous.

Does Dual-Task Gait Differ in those with Late-Life Depression versus Mild Cognitive Impairment?

OBJECTIVES

To compare the dual-task gait performance of older adults with Late-Life Depression (LLD) versus Mild Cognitive Impairment (MCI).

DESIGN

Cross-sectional study with three matched groups: LLD, MCI and non-depressed and cognitively intact (NDCI).

SETTING

LLD group participants were recruited from geriatric psychiatry clinics in London, Ontario. Matched participants meeting criteria for the MCI or NDCI groups were previously recruited for other research studies from geriatric clinics and the community.

PARTICIPANTS

Individuals aged 60-85 who met criteria for mild-moderate LLD (N=23) without a diagnosis of a neurocognitive disorder.

MEASUREMENTS

Participants completed questionnaires regarding mood, cognition and physical activity. Gait speed was recorded using an electronic walkway during simple and dual-task gait (walking while naming animals aloud). Dual-task cost (DTC) is the percentage change in gait speed between simple and dual-task gait. It is a clinically relevant indicator of fall risk and is strongly associated with cognitive decline. For comparison, 23 MCI and 23 NDCI participants, matched with respect to age, sex and comorbidities, were randomly selected from existing research databases.

RESULTS

Each group had 8 males and 15 females, with mean age of 69.0-69.6 years. The mean (±SD) DTC of the NDCI, LLD and MCI groups were statistically different at 2.4±11.4%, 11.8±9.9% and 22.2±16.7%, respectively.

CONCLUSION

Older adults with LLD perform worse on dual-task gait than NDCI; however, they are less impaired than those with MCI. The elevated DTC seen in LLD is likely because of underlying executive dysfunction that is less significant than in those with MCI.

Swallowing Function and Kinematics in Stroke Patients with Tracheostomies

The purpose of this study was to compare the swallowing function and kinematics in stroke patients with and without tracheostomies. In this retrospective matched case-control study, we compared stroke patients with (TRACH group, n = 24) and without (NO-TRACH group, n = 24) tracheostomies. Patients were matched for age, sex, and stroke-type. Swallowing function was evaluated using the videofluoroscopic dysphagia scale (VDS) and functional oral intake scale (FOIS) obtained from videofluoroscopic swallow study (VFSS) images. Swallowing kinematics were evaluated using a two-dimensional kinematic analysis of the VFSS images. Mean duration of tracheostomy was 132.38 ± 150.46 days in the TRACH group. There was no significant difference in the total VDS score between the TRACH (35.17 ± 15.30) and NO-TRACH groups (29.25 ± 16.66, p = 0.247). FOIS was significantly lower in the TRACH group (2.33 ± 1.40) than in the NO-TRACH group (4.33 ± 1.79, p = 0.001). The TRACH group had a significantly lower maximum vertical displacement (15.23 ± 7.39 mm, p = 0.011) and velocity (54.99 ± 29.59 mm/s, p = 0.011), and two-dimensional velocity (61.07 ± 24.89 mm/s, p = 0.013) of the larynx than the NO-TRACH group (20.18 ± 5.70 mm, 82.23 ± 37.30 mm/s, and 84.40 ± 36.05 mm/s, respectively). Maximum horizontal velocity of the hyoid bone in the TRACH group (36.77 ± 16.97 mm/s) was also significantly lower than that in the NO-TRACH group (47.49 ± 15.73 mm/s, p = 0.032). This study demonstrated that stroke patients with tracheostomies had inferior swallowing function and kinematics than those without tracheostomies. A prospective longitudinal study is needed to elucidate the effect of a tracheostomy on swallowing recovery in stroke patients.

Quantitative Alterations of Capillary Diameter Have a Predictive Value for Development of the Capillaroscopic Systemic Sclerosis Pattern

Objective.

To quantify earlier capillary diameter abnormalities, observed by nailfold videocapillaroscopy (NVC), in primary Raynaud phenomenon (PRP) subjects compared with RP subjects later evolved to systemic sclerosis (SSc)-associated secondary Raynaud phenomenon (SRP).

Methods.

There were 6112 NVC images of 191 subjects analyzed at baseline and after a mean followup of 42.77 ± 35.80 months. We selected 48 patients affected by SRP and 143 matched controls confirmed with PRP. The diameter of the most dilated limbs (arterial, venous, and apical) was measured in 16 images per subject. Statistical analysis was performed using nonparametric tests. The threshold values for capillary diameters associated with the development of SSc-associated SRP were determined through receiver-operating characteristic curves.

Results.

Mean capillary diameter values were significantly different for arterial, venous, and average diameters (mean value of arterial, venous, and apical) between patients with PRP and SRP (p < 0.0001). These alterations were found to be independent predictors for disease development (p = 0.015). Threshold values of 30 µm (area under the curve = 0.802, sensitivity/specificity = 0.85/0.63) to 31 µm were identified for average, arterial, and venous diameters, with a shortening effect on time to disease development.

Conclusion.

The study showed that capillary diameter is an independent predictor for development of SSc-associated SRP. Progression to SRP is unlikely for subjects affected by RP when average capillary diameter is under 30 μm. Subsequently, the execution of the qualitative/quantitative integrated analysis should be part of the NVC followup of RP subjects.

Characterizing self-reported sleep disturbance after mild traumatic brain injury

Sleep disturbance after mild traumatic brain injury (mTBI) is commonly reported as debilitating and persistent. However, the nature of this disturbance is poorly understood. This study sought to characterize sleep after mTBI compared with a control group. A cross-sectional matched case control design was used. Thirty-three persons with recent mTBI (1-6 months ago) and 33 age, sex, and ethnicity matched controls completed established questionnaires of sleep quality, quantity, timing, and sleep-related daytime impairment. The mTBI participants were compared with an independent sample of close-matched controls (CMCs; n = 33) to allow partial internal replication. Compared with controls, persons with mTBI reported significantly greater sleep disturbance, more severe insomnia symptoms, a longer duration of wake after sleep onset, and greater sleep-related impairment (all medium to large effects, Cohen's d > 0.5). No differences were found in sleep quantity, timing, sleep onset latency, sleep efficiency, or daytime sleepiness. All findings except a measure of sleep timing (i.e., sleep midpoint) were replicated for CMCs. These results indicate a difference in the magnitude and nature of perceived sleep disturbance after mTBI compared with controls, where persons with mTBI report poorer sleep quality and greater sleep-related impairment. Sleep quantity and timing did not differ between the groups. These preliminary findings should guide the provision of clearer advice to patients about the aspects of their sleep that may change after mTBI and could inform treatment selection.

Urine neutrophil gelatinase-associated lipocalin to predict acute kidney injury in preterm neonates. A pilot study

BACKGROUND

The efficacy of urine neutrophil gelatinase-associated lipocalin (uNGAL) as an early acute kidney injury (AKI) biomarker in preterm neonates was evaluated.

METHODS

Thirty-five preterm neonates were prospectively evaluated for serum creatinine (sCre)-documented AKI during the first 14 days of life. Urine samples were collected daily throughout the study period. Of the neonates evaluated, we analyzed 11 who developed AKI (cases) and an equal number of neonates without AKI (controls) matched for gestational and postnatal age (case-control study). uNGAL was measured on the day of AKI occurrence (day 0) and on the 2 days preceding the event (day -1 and day -2, respectively) using an enzyme-linked immunosorbent assay.

RESULTS

Cases had significantly higher sCre levels than controls on day 0 (1.21 ± 0.48 vs. 0.83 ± 0.16 mg/dL, p =0.031) but not on days -1 and -2. Similarly, uNGAL levels (ng/mL) were significantly higher in cases than in controls only on day 0 (19.1 ± 3.5 vs. 13.3 ± 7.3, p=0.017) and not on days -1 (18.8 ± 3.4 vs. 16.3 ± 5.9, p=0.118) and -2 (19.3 ± 1.8 vs. 19.4 ± 0.8, p =0.979). The receiver operating characteristic curve analysis showed no significant ability of uNGAL to predict AKI on days -2 and -1.

CONCLUSIONS

In this pilot study in preterm neonates, although uNGAL detected sCre-based AKI upon its documentation, it failed to predict its development 1-2 days earlier.