Data Collection and Processing

NAMCS Data Collection and Processing

The U.S. Bureau of the Census acts as the field data collection agent for the NAMCS. The data collection for NAMCS is expected to be carried out by the physician or the physician’s staff, but it is often performed by Census field representatives. More than half of the NAMCS Patient Record forms submitted in 2009 (51.5 percent) were abstracted by Census Bureau staff rather than by the physician or medical office personnel.

Physicians were instructed to keep a daily listing of all patient visits during the assigned reporting week using an arrival log, optional worksheet, or similar method. This list was the sampling frame to indicate the visits for which data were to be recorded. It was to include both scheduled and unscheduled patients, but not cancellations or no-shows. Visits were selected from the list using a random start and a predetermined sampling interval based on the physician’s estimated visits for the week and the number of days the physician was expected to see patients that week. In this way, a systematic random sample of visits was obtained. The sampling procedures were designed so that about 30 Patient Record forms were completed during the assigned reporting week. This minimized the data collection workload and maintained about equal reporting levels among sample physicians regardless of practice size.

Data for sampled visits were recorded on Patient Record forms. The Patient Record form was printed on one side of an 8 x 14 inch sheet (Figure 1). At the top of each form, physicians could record the patient’s name and record number. However, this information was detached by the physician prior to submitting the completed forms and was never given to Census personnel. It was used only to assist physicians in clarifying missing or ambiguous responses that may have been found during the data editing process.

In addition to the completeness checks made by field staff, clerical edits are performed upon receipt of the data for central processing. Detailed editing instructions are provided to manually review the forms and to reclassify or recode ambiguous entries. Computer edits for code ranges and inconsistencies are also performed. All medical and drug coding and keying operations are performed centrally by SRA International, Inc. Durham, North Carolina and are subject to quality control procedures. Keying and coding error rates generally range between 0-1 percent for various survey items

Item nonresponse rates are generally 5 percent or less for NAMCS data items, with some exceptions. These are described in detail in the complete NAMCS downloadable documentation. For 2009, imputations for missing data were performed for patient’s date of birth, sex, ethnicity, and race, whether patient was seen before by physician or anyone in practice, and, if yes, number of past visits in last 12 months, and time spent with physician. Imputations, in general, were based on physician specialty, geographic region, and 3-digit ICD-9-CM codes for primary diagnosis. But the method used to impute race and ethnicity was refined in recent years, first for use with 2006 data, and later for use with 2009 data. See the 2009 NAMCS Public Use File Documentation for more information.

Drug data are coded using a unique classification scheme developed at NCHS. Listings of drugs by entry name (the name used by the respondent to record the drug on the Patient Record form) and by generic substance are available. The therapeutic class of drugs was based on the National Drug Code Directory for data years prior to 2006. However, beginning with the 2006 data year, drugs are coded in terms of their generic components and therapeutic classifications using Lexicon Plus®, a proprietary database of Cerner Multum, Inc. The Multum Lexicon provides a 3-level nested category system that assigns a therapeutic classification to each drug and each ingredient of the drug. Each drug may have up to four therapeutic categories on the public use file. For more information, please see the annual public use data file documentation.

NHAMCS Data Collection and Processing

The U.S. Bureau of the Census was the data collection agent for the 2009 NHAMCS. Census Headquarters staff were responsible for overseeing the data collection process, training the Census Regional Office staff, and writing the field manual. Regional Office staff were responsible for training the field representatives and monitoring hospital data collection activities. Field representatives inducted the hospitals and trained the hospital staff on visit sampling and completion of the Patient Record forms.

Different encounter forms were developed for use in the ED and OPD. The top section of each form, which contains the patient’s name and record number, is separated from the bottom section by a perforation running across the page. The top section remains attached to the bottom until the entire Patient Record form is completed. To ensure confidentiality, before collecting the completed Patient Record forms, the top section is detached and given to hospital staff. They are instructed to keep this information for 4 weeks, in case it is necessary to retrieve missing information or clarify ambiguous entries.

In addition to the completeness checks made by field staff, clerical edits are performed upon receipt of the data for central processing. Detailed editing instructions are provided to manually review the forms and to reclassify or recode ambiguous entries. Computer edits for code ranges and inconsistencies are also performed. All medical and drug coding and keying operations are performed centrally by SRA International, Inc. Durham, North Carolina and are subject to quality control procedures. Keying and coding error rates generally range between 0-1 percent for various survey items.

Item nonresponse rates are generally 5 percent or less for NHAMCS data items, with some exceptions. These are described in detail in the complete NHAMCS downloadable documentation. For 2009, imputations for missing data were performed for these ED items: patient’s birth year, sex, race, ethnicity, and triage level. On the OPD file, six items were imputed: patient’s birth year, sex, race, ethnicity, whether patient had been seen in the clinic before, and, if yes, number of past visits in last 12 months. The imputation procedures, which changed for 2009, are described in the annual public use file documentation. See the NHAMCS Public Use File Documentation.

Drug data are coded using a unique classification scheme developed at NCHS. Listings of drugs by entry name (the name used by the respondent to record the drug on the Patient Record form) and by generic substance are available. The therapeutic class of drugs was based on the National Drug Code Directory for data years prior to 2006. However, beginning with the 2006 data year, drugs are coded in terms of their generic components and therapeutic classifications using Lexicon Plus®, a proprietary database of Cerner Multum, Inc. The Multum Lexicon provides a 3-level nested category system that assigns a therapeutic classification to each drug and each ingredient of the drug. Each drug may have up to four therapeutic categories on the public use file. For more information, please see the annual public use data file documentation.