Data & Analytics

Today, more than ever before, it is all about the numbers. But burgeoning data from basic research, clinical studies, and epidemiologic studies (including electronic medical records) can be challenging to manage. Data may exist in various formats and distributed across multiple organizations and systems. Integrating these data is a growing need. Interpreting these data using new tools and techniques is equally important. CSR offers Ph.D.-level, highly experienced staff to support these tasks.

N

Needs Assessment

N

Data Visualization

N

Qualitative Data & Analytics

N

HIPAA Privacy and Security

N

Primary and Secondary Data & Analytics

N

Data Cleansing

N

Environmental Scans

N

OMB/IRB Process Capability (added)

N

Data Harmonization and Management

N

Formative Research

N

Survey Design and Implementation

N

Standardized Reporting and Improved Data Collection Techniques

Surveillance Reports

Initiated by NIAAA since 1986, Surveillance Reports have presented trend data on per capita alcohol consumption, liver cirrhosis mortality, and underage drinking, as well as the discontinued topics of substance use among reproductive-age females, alcohol-related morbidity, and alcohol-related fatal traffic crashes. These reports have served researchers, policymakers, and others interested in tracking alcohol-related information over time.

CSR staff have a long history of supporting the development and updating of these Surveillance Reports. Therefore, we are highly experienced with the data sources, methodology, and production of these reports. CSR staff understand the special requirements for developing highly useful and effective Surveillance Reports. These requirements entail:

 

  • Consistent measures over time
  • Data representative of sizable populations (usually national or State populations)
  • Regular availability of current data

CSR staff have collaborated on approximately 235 conference presentations at professional conferences, including the American College of Neuropsychopharmacology, the Society for Epidemiologic Research, Research Society on Alcoholism, and the International AIDS Conference. This cumulative experience in presenting epidemiologic research findings at major conferences reflects CSR staff members’ skill in writing concise abstracts that describe well-designed methods and important preliminary findings within the word limits and in the required formats specified by the host organizations. Depending on conference requirements, CSR will typically produce abstracts organized in five standard sections: (1) Background, (2) Objectives, (3) Methods, (4) Results, and (5) Conclusions. The acceptance rate of abstracts that we have produced over the past two decades has been over 99 percent. Moreover, a number of these presentations received significant media attention and the praise of conference attendees, which has contributed to greater visibility for NIAAA among the public and the larger scientific research community.

CSR staff have extensive experience pooling and linking data from multiple disparate sources, including vitalstatistics and large national survey databases, and across their individual modules over multiple years. These activities have included:

  • Pooling time series data with cross-sectional data
  • Linking discharge records with hospital-level data across different modules within a singlehospitalization
  • Linking individual records in survey data with aggregated State-level data
  • Linking survey data (e.g., NHANES/National Health Interview Survey) with their public- use Linked Mortality File compiled by National Center for Health Statistics.

Statistical Consultation

CSR staff are experts in statistical data analysis and programming and have provided consultation to NIAAA staff on many statistical issues, including recommending appropriate analytic methods and preparing dataset documentation. The choice of statistical model will depend on the research questions, study design, sample size, and whether the data are censored or truncated, among other considerations. Because NIAAA’s epidemiologic studies often involve secondary analysis of complex survey data or administrative medical records the calculation of point estimates and their standard errors must account for the relevant design parameters. Techniques that are routinely used for analysis of randomized clinical trials are generally not applicable to observational studies. Examples include:

 

  • Many meta-analytic methods have been proposed for combining study results for binary or continuous data from two groups, but consumption patterns in alcohol epidemiologic research usually have more than two categories.
  • When age is used as the timeline in survival analysis, left truncation or delayed entry must be considered.
  • For longitudinal analysis, random versus fixed effects and time-varying covariates will be incorporated appropriately into the statistical models.

CSR staff are knowledgeable about these types of statistical issues and keep current with new developments in statistical methods. Whenever analytic issues and questions arise, we will provide statistical consultation for NIAAA researchers.

CSR staff have applied advanced statistical approaches in the conduct of many epidemiologic studies for NIAAA. We will use direct standardization, life table analysis, and calculation of hazard ratios in mortality research, and calculation of odds ratios, incidence ratios, relative risks, population-attributable fractions, and multiple imputation data in morbidity research. CSR used structural equation modeling, item response theory, latent variable, or finite mixture modeling, and bifactor modeling in work we performed for and in collaboration with staff from Division of Emergency Preparedness and Response and other NIAAA divisions. CSR staff are also familiar with the following:

 

  • Computing composite index scores
  • Multiple imputation
  • Pooled cross-sectional time series analysis
  • Interrupted time series analysis
  • Difference in difference
  • Fixed effect regression analysis
  • Generalized least squares with autoregressive errors
  • Various multilevel models according to the outcome type
  • Using software such as Mplus, MLwiN, and GLLAMM