Early Release Program Races to Publish Timely Survey Results
Posted on byFrom among the many challenges that the National Health Insurance Survey (NHIS) has faced over the past 56 years since it first went into the field, one of the most daunting has been how to publish valid results in a timely manner without compromising the quality of the data. For Dr. Jane Gentleman and her NHIS team, the answer lies in data processing.
The Early Release (ER) Program of the National Health Interview Survey provides analytic reports and preliminary microdata files on an expedited schedule. These reports and files are produced prior to final data editing and final weighting to provide early access to the most recent information from NHIS. For the last 9 years, NHIS public-use data files have been released about 6 months after the end of each data collection year, which is admirable timeliness. And thanks to the ER Program, which publishes analyses of preliminary NHIS data 6 months after each calendar quarter, NHIS data users have access to estimates up to 9 months before the corresponding data files are released.
It used to take much longer—even years—to process and release NHIS data and, therefore, to produce analytic reports. For policy makers and others who, Dr. Gentleman says, “want their data while the data are young,” this was longer than they would have liked. Then, in 1999, NHIS received funding to investigate ways to produce preliminary NHIS data of sufficient quality to produce analytic reports sooner.
Working with Dr. Jennifer Madans, NCHS Associate Director for Science, and a grant from CDC, NHIS staff members organized a conference of experts who developed a list of key topics and variables for which they would especially like to see early analyses, and with the remaining funds they hired an additional programmer. The first Early Release of Selected Estimates from the National Health Interview Survey was published in April 2001, and it provided estimates of NHIS data from January through June 2000 for seven topics.
In 2001–2003, NHIS data processing procedures were re-engineered, which further sped up data production. Today, the ER Program provides estimates updated quarterly for 15 key health topics: health insurance coverage, having a usual place to go for medical care, obtaining needed medical care, receipt of influenza vaccination, receipt of pneumococcal vaccination, obesity, leisure-time physical activity, smoking, alcohol consumption, HIV testing, general health status, personal care needs, serious psychological distress, diabetes, and asthma. All ER reports are available publicly for free and are downloadable from the NCHS website.
Releasing estimates “while the data are young” is achieved by expediting data processing and analyzing the data at the end of each quarter. Because NHIS is conducted throughout the year and the sample is designed to yield a nationally representative sample each month, data can be analyzed quarterly.
Dr. Gentleman says the estimates are preliminary because the data processing procedure is very complex, likening it to a long assembly line. Selected data are copied preliminarily from the assembly line and used to produce the ER reports. Dr. Gentleman says caveats regarding this are included in every report. The caveats note that users who perform their own analyses using final public-use data may see slight differences, but the differences will typically be less than 0.1 percentage point. With this in mind, Dr. Gentleman says, users are happy with getting estimates up to 9 months before the final data are released.
During the initial ER planning conference, one topic beat out all the others by a long shot: health insurance. In addition to being one of the key health topics covered in the first Early Release report, since data year 2001 health insurance has had its own quarterly ER report, Health Insurance Coverage: Early Release of Estimates From the National Health Interview Survey, 2012 , which contains expanded details on that topic.
A perhaps surprising by-product of the ER Program is an analytic report on landline phone and cell phone usage in respondents’ households. Since the telephone is a means of data collection for many surveys, demand is heavy for information regarding household phone usage. NHIS is particularly suitable for collecting such information because it is an in-person survey, so it reaches households even if there are no phones used by household members. Starting in 2006, the semi-annual Wireless Substitution: Early Release of Estimates From the National Health Interview Survey has been an extremely popular ER product. Dr. Gentleman says that its authors, Dr. Stephen Blumberg and Julian Luke (the programmer originally hired to help launch the ER Program), enjoy celebrity status among survey experts.
Another by-product of the Early Release Program is preliminary NHIS microdata files produced 6 months after each of the first 3 quarters of the data collection year. NHIS data users can analyze these files through the NCHS Research Data Center without having to wait for the final annual NHIS microdata files to be released near June following the end of the data collection year.
The preliminary ER microdata are also used by NHIS staff to produce timely reports and tabulations on special topics, such as changes in health care access and utilization related to the implementation of the Affordable Care Act.
Clearly, the NHIS Early Release Program is aptly named.