December 2021
Data Tip of the Month
Did you know . . .
In the
International Classification of Diseases, 10th Revision, Clinical Modification (ICD-10-CM), which is used to code NAMCS and NHAMCS data, diagnosis codes can have a maximum of seven digits. For the NHAMCS ED public use file, only the first four digits of the diagnosis codes are included. There is an implied decimal between the third and fourth digits and inapplicable fourth digits are dash-filled. For example: F321 = F32.1 Major depressive disorder, single episode, moderate; I10- = I10 Essential (primary) hypertension. Since ICD-10-CM uses non-numeric characters extensively, we are not able to provide numeric recodes for the diagnosis codes.
November 2021
Data Tip of the Month
Did you know . . .
Prior to release of the public use data file, NCHS conducts extensive disclosure risk analysis to minimize the chance of any inadvertent disclosure. Based on research conducted by NCHS for 2018 NAMCS, certain variables were subject to masking in some cases and outlier values for certain variables (age, age of pregnant patient, height, weight, number of past visits in last 12 months, and time spent with physician) were top coded in accordance with NCHS confidentiality requirements. Masking was performed in such a way to cause minimal impact on the data; data users who wish to use unmasked data can apply to the
NCHS Research Data Center.
October 2021
Data Tip of the Month
Did you know . . .
State-based estimates can only be calculated using NAMCS data from 2012 through 2015. The NAMCS sampling design changed in 2012 in order to allow state-based estimates for the 34 most populous states. In 2013, 2014, and 2015, the numbers of targeted states were 22, 18, and 16, respectively. The state weighting variable is called PATWTST; it is only to be used for individual state estimates and will not sum to national or regional estimates.
September 2021
Data Tip of the Month
Did you know . . .
For 2018 NAMCS, the geographic variable REGIONOFF (region where majority of physician’s sampled visits occurred) was removed from the public use file. The only geographic variable remaining on the 2018 public use file is MSA (was physician’s interview location within a metropolitan statistical area). The physician specialty variable SPECR (a 14-category variable denoting broad specialty group), was also removed from the public use file. The remaining physician classification variables are MDDO (doctor of medicine or doctor of osteopathy) and SPECCAT (type of specialty: primary care, surgical care or medical care). The region and physician specialty variables were removed because low survey response in 2018 resulted in the inability to make reliable estimates using these variables.
August 2021
Data Tip of the Month
Did you know . . .
The emergency departments weights (EDWT) enable data users to calculate department-level estimates. There is generally one weight for each emergency department which appears on the first visit record only for that department. When running an analysis of facility-level characteristics using EDWT, it is recommended to select only those records where EDWT is greater than zero. This will result in correct sample counts of variables, which is useful for assessing reliability. Weighted estimates will be correct either way, because of the one weight per department format.
July 2021
Data Tip of the Month
Did you know . . .
When using SAS, to properly calculate the standard errors (SE) of your statistics (such as means and percentages), the Taylor series linearization method requires information on all records with a non-zero value for your weight variable.
You should never drop observations from a SAS dataset. It is recommended instead to use the DOMAIN statement or the TABLES statement to specify your population of interest. It is recommended to include the NOMCAR option (when using PROC SURVEYFREQ, SURVEYMEANS, SURVEYREG, and SURVEYLOGISTIC) so that records with missing data are included in SE computations. Without NOMCAR, these records are excluded by default and SEs are understated.
June 2021
Data Tip of the Month
Did you know . . .
For the first time in 2018, NAMCS weights were adjusted using Multipurpose Iterative Proportional Fitting (IPF). This is a calibration technique that simultaneously implements (
1) calibration in multiple specified domains; (2) nonresponse adjustment; and (
3) weight trimming, as part of a unified iteration cycle. As the method used to weight NAMCS data has changed with 2018 data, caution should be taken when interpreting differences between the 2016 and 2018 estimates. More details can be found in the
2018 NAMCS documentation.
May 2021
Data Tip of the Month
Did you know . . .
In addition to the standard drug data included on public use files each year, separate files are available that can be used to add ingredient information for combination drug products. Combination products such as cold preparations on the public use file may only have a therapeutic category of a respiratory combination, for example. By merging the drug ingredient file to the regular public use file, each combination drug can have up to 6 separate ingredients listed with up to 4 therapeutic categories for each one. Both a SAS program to add the drug ingredients and a data file for merging are produced annually and are available for downloading for use with
NAMCS and
NHAMCS data.
April 2021
Data Tip of the Month
Did you know . . .
Starting in 2011, NAMCS and NHAMCS data files have been converted into a zipped (compressed) format which must be uncompressed prior to use. To download and uncompress the zipped files: 1) Create a new folder on your location workstation, if desired, to house the files; 2) Click on the file name from the FTP server here:
https://ftp.cdc.gov/pub/Health_Statistics/NCHS/Datasets/NHAMCS or
https://ftp.cdc.gov/pub/Health_Statistics/NCHS/Datasets/NAMCS and follow the prompts to save the file to your desired location; 3) Right-click on the name of the compressed file from your directory screen. On the pop-up menu, there should be an option to extract the file to a location of your choosing.
Beginning with 1999, NHAMCS and NAMCS documentation has been produced as a single file in PDF format which can be viewed or downloaded
here. These files are not compressed.
March 2021
Data Tip of the Month
Did you know . . .
NAMCS and NHAMCS data for estimated numbers of visits and visit rates are considered to be reliable if two conditions are met. The first condition is that each estimate is based on at least 30 sample (unweighted) records. If an estimate is based on < 30 cases it is considered unreliable. The second condition is that the weighted data have a relative standard error (RSE) of 30 percent or less. The RSE is equal to the standard error divided by the estimate and expressed as a percent.
Proportion estimates are considered to be unreliable based on the procedure specified in “
National Center for Health Statistics Data Presentation Standards for Proportions [PDF - 1.43 MB]” . These standards are based on a minimum denominator sample size and on the absolute and relative widths of a confidence interval calculated using the Clopper-Pearson method.
February 2021
Data Tip of the Month
Did you know . . .
When performing trend analysis and/or combining multiple years of data, it is important to check that the variables of interest were collected the same way for all of the years contained in your analysis. For example, the number of drugs that are collected has increased over time: from 5 in 1994 to 6 in 1995 to 8 in 2003 to 10 in 2012 and to 30 in 2014. If an increase is observed in the number of medications prescribed, this could be due to the increase in the number of drugs collected in the survey and not an actual increase in prescribing patterns. Therefore, a trend examining prescribing from 2000-2018 should only include the first six listed drugs in all years. You can
review the survey instruments for NAMCS and NHAMCS.
December 2020
DATA TIP OF THE MONTH
Did you know . . .
With the release of the 2005 National Ambulatory Medical Care Survey (NAMCS) and National Hospital Ambulatory Medical Care Survey (NHAMCS) public use micro-data files, provider weights became available. The variable PHYSWT on the NAMCS file allows researchers to make physician-level estimates. EDWT on the NHAMCS file allows researchers to make estimates at the emergency department level. SAS code examples to produce aggregate visits statistics at the physician or facility level are available
here [PDF - 52.6 KB].
November 2020
DATA TIP OF THE MONTH
Did you know . . .
NAMCS and NHAMCS collect information on up to 30 drugs provided, prescribed or continued at ambulatory medical care visits. Drug entries are coded in two ways: as reported in the survey instrument using an internal NCHS classification, and also by their generic components and therapeutic classifications using Lexicon Plus®, a proprietary database of Cerner Multum, Inc. NCHS provides an online
look-up tool which allows you to search for drugs in recent years of NAMCS and NHAMCS data.
October 2020
DATA TIP OF THE MONTH
Did you know . . .
For users who are interested in analyzing drug data, one method involves the isolation of those records with drugs, or drug mentions, and the creation of a separate data file of drug mentions. Each Patient Record for 2017 can have up to thirty drug mentions recorded, so whatever file is created would need to include all of them. This method can be used for obtaining estimates of drug mentions, but is not recommended for variance estimation. Etimates of drug mentions can be obtained by creating a new weight variable (called DRUGWT in this example). This variable is created by multiplying PATWT (the patient visit weight) by NUMMED (the number of medications recorded at the sampled visit) or DRUGWT=PATWT*NUMMED. DRUGWT can then be used in place of PATWT to weight one’s data; it produces the estimated number of drug mentions rather than visits.
September 2020
DATA TIP OF THE MONTH
Did you know . . .
When conducting a trend analysis, data users should note that, beginning with the 2002 public use files, two masked design variables were added to the file, for use with statistical software that assumes a single stage of sampling. For the 2003 public use files and beyond, only these two masked design variables are included for variance estimation. Therefore, when combining years of data from 2003 and beyond with years prior to 2002, data users will need to create these two variables for each file prior to 2002. A technical paper with instructions can be found
here [PDF - 33.9 KB].
August 2020
DATA TIP OF THE MONTH
Did you know . . .
A Diagnosis Master Category List (DMCL) which includes diagnosis categories appropriate for use across both inpatient and ambulatory settings is available
here [PDF - 443]. The DMCL uses all ICD–10–CM codes and classifies them into 339 mutually exclusive categories. The DMCL will slightly change each year to account for annual changes to the ICD–10–CM coding system. This list was based on the 2016 ICD–10–CM coding system.
July 2020
DATA TIP OF THE MONTH
Did you know . . .
The CSTRATM and CPSUM variables were created and first added to the NAMCS and NHAMCS 2002 public use files for running programs that use an ultimate cluster model. They can be used to estimate variance with SUDAAN’s with-replacement (WR) option, as well as with Stata, SPSS, SAS, and other statistical software packages utilizing an ultimate cluster model for variance estimation. These variables and their use are described in detail in the “Relative Standard Errors” section of the public use file documentation. The decision was made to include only these new variables, CSTRATM and CPSUM, and not the multi-stage design variables, beginning with the 2003 data release. For those who wish to combine data from 2003 forward with survey data from years prior to 2002 which do not contain CSTRATM and CPSUM, please see the technical paper:
Using Ultimate Cluster Models with NAMCS and NHAMCS Public Use Files [PDF - 33.9 KB].
June 2020
DATA TIP OF THE MONTH
Did you know . . .
NHAMCS public use files are provided in ASCII text format that allow the data to be used by many software products. NCHS provides sample SAS statements, SPSS and Stata code for use with NHAMCS data files from 1992 forward. The survey undergoes periodic redesign, which can make the process of combining multiple years of data challenging. NCHS staff make every effort to point out important changes in the annual public use file
documentation. Staff members are also available to answer technical questions 301-458-4600.
May 2020
DATA TIP OF THE MONTH
Did you know . . .
Missing values are imputed every year for some items. Starting with 2010 data, the imputation of patient race and ethnicity was performed using a model-based single, sequential regression method. The model used to impute race and ethnicity in 2017 NHAMCS included the following variables: Census variables for ZIP code level race and ethnicity population estimates and an indicator for whether it was patient or hospital ZIP (used when patient ZIP was not available); patient age, sex, race, and ethnicity; triage level; log of ED wait time; primary expected source of payment derived from a hierarchical recode of the expected source of payment question; grouped 3-digit ICD-10-CM codes for primary diagnosis; year of visit; type of ESA area; provider’s MSA status; and ED weighting and volume variables. The list of items with a nonresponse rate greater than 5% and the list of the imputed items can be found
here starting on page 13 [PDF - 14 KB].
April 2020
DATA TIP OF THE MONTH
Did you know . . .
For NHAMCS 2017, anomalies between 2017 and 2016 data were assessed. To accomplish this, more than 20 tables of 2017 estimates were compared with the same tables of 2016 estimates. The variables that were compared included most of those published in the annual survey web tables available
here. Significant changes between the two years were noted and investigated. More information about the changes can be found in the 2017 NHAMCS
Public use file documentation.
March 2020
DATA TIP OF THE MONTH
Did you know . . .
A new template for producing Methods sections of journal articles using NHAMCS data covering data years 1992-2016 is available
here [PDF - 28 KB]. The document also provides a useful checklist for NHAMCS article submission.
February 2020
DATA TIP OF THE MONTH
Did you know . . .
To calculate department-level estimates an emergency department weight (EDWT) can be used. When running an analysis of facility-level characteristics using EDWT, it is recommended to select only those records where EDWT is greater than 0. This will result in correct sample counts of variables, which is useful for assessing reliability. Estimates at the ED level generated using EDWT reflect only the characteristics of facilities which participated in the survey.
December 2019
DATA TIP OF THE MONTH
Did you know…
The National Center for Health Statistics considers an estimated number of visits or a visit rate to be reliable if it has a relative standard error of 30 percent or less, and it is based on at least 30 sample records. NCHS recently released new guidelines for determining the reliability of proportions. These standards are based on a minimum denominator sample size and on the absolute and relative widths of a confidence interval calculated using the Clopper-Pearson method. The guidelines can be found
here [PDF - 144 MB].
November 2019
DATA TIP OF THE MONTH
Did you know…
Multiple years of data can be combined to improve reliability of estimates, which is necessary for rare estimates. When combining multiple years, it is easier for the reader to understand an average annual estimate rather than the total number of visits that occurred during the study period. For example, it is preferable to report that from 2012-2016, the average annual number of office-based physician visits for condition X was 5 million rather than from 2012-2016, there were 25 million visits for condition X.
OCTOBER 2019
DATA TIP OF THE MONTH
Did you know…
The
Research Data Center contains records for NAMCS physicians who are not included on the public use files for various reasons (did not see patients during reporting week or did not submit visit forms). Including these physicians in one’s analysis along with those who saw patients can provide national estimates of all office-based physicians.
September 2019
DATA TIP OF THE MONTH
Did you know…
A problem was found with the checkbox for CHF (Congestive Heart Failure) in the item: “Regardless of the diagnoses written above, does the patient now have:”. This checkbox is typically edited to ensure that any diagnoses of CHF reported in the survey items DIAG1-DIAG5 (Diagnosis #1 – Diagnosis #5) are also reflected in the CHF check box. It was recently discovered that the ICD-10-CM codes I50.1-I50.9 (Heart Failure) were not included in the edit. In all, 28 2016 NAMCS records and 45 2016 NHAMCS ED records with a diagnosis of Heart Failure reported in DIAG1-DIAG5 should have had the CHF checkbox checked but did not. More information is available
here.
August 2019
DATA TIP OF THE MONTH
Did you know…
To make
estimates at the physician level [PDF – 1.19 MB] the PHYSWT variable can be used. When running physician-level analysis, it is recommended that only records with a PHYSWT greater than zero be selected; this will give the correct sample counts and will not affect estimation of variance. Beside generating estimates for physician characteristics at the physician level, the addition of PHYSWT allows you to link visit data with physician data.
June 2019
DATA TIP OF THE MONTH
Did you know…
NAMCS and NHAMCS are national probability sample surveys that use stratified, multistage sampling designs. That is, visits are always clustered within providers and providers are sometimes clustered within geographic primary sampling units. The sampling design affects the standard errors of estimates based on data collected from the sample. Therefore, data users should always calculate standard errors for the survey estimates using statistical software that takes the complex sampling design into account. Such software includes SUDAAN, SAS (SURVEYFREQ and SURVEYMEANS procedures), Stata, and SPSS Complex Samples, among others. Make sure all of the sample records - and not just those in a subset of the sample - are included in the analysis in order to obtain the best variance estimates. This can be accomplished with a SUBPOPN statement in SUDAAN, or by using the DOMAIN statement in SAS, for example.
May 2019
DATA TIP OF THE MONTH
Did you know…
In 2005, data for the initial visit to ED (EPISODE=1) was not collected. For researchers interested in the episode of care for visits in 2005, an imputed variable (INITVIS; included in the 2005 public use file) was created based on a regression model-based imputation strategy using data from 2003-04. It should be noted that because this is an imputed variable, it is not comparable to the EPISODE variable from other years (starting in 2001) and should be used with caution when conducting any year-to-year trend analysis. More information on the methodology and limitations of the imputation model can be found
here [PDF - 37 KB].
April 2019
DATA TIP OF THE MONTH
Did you know…
Beginning with the 2002 public use files, two new masked design variables were added to the file, for use with statistical software that assumes a single stage of sampling. Data users who wish to combine years of data from 2002 and beyond with years prior to 2002 need to create these two masked design variables for data file years prior to 2002. A technical paper gives instructions on how to do this, and is available
here [PDF - 34 KB].
March 2019
DATA TIP OF THE MONTH
Did you know…
The method of collecting payment data in NAMCS/NHAMCS 2005-2007 was different than survey years 2008-present. In 2005-2007 (2005 was the first year that allowed collection of multiple expected sources of payment per visit), beneficiaries of both Medicare and Medicaid (dual eligibles) were assigned Medicaid as the primary expected source of payment (PAYTYPE); from 2008-present, dual eligibles were assigned Medicare as the primary source (PAYTYPER) based on research and analysis which determined that the Medicaid-dominant hierarchy is inconsistent with insurance industry practices. Therefore, when analyzing trends across these two time periods, researchers should keep in mind that two different methods were used to collect payment data. More information can be found in
this document [PDF - 43 KB] and the
2008 public use file documentation [PDF - 1.4 MB].
February 2019
DATA TIP OF THE MONTH
Did you know…
A template for producing the Methods section of a journal article using NHAMCS data is available on the Ambulatory Health Care Data website and is linked
here [PDF - 27 KB].
January 2019
DATA TIP OF THE MONTH
Did you know…
When examining trends and/or combining multiple years of data, it is important to check that the variables of interest were collected in all of the years contained in your analysis and that the wording of the item and/or answer categories did not change in a way that might impact the analysis. NCHS staff make every effort to point out important changes in the annual
public use file documentation, both in the summary of changes at the beginning of the document and in the codebook section.
December 2018
DATA TIP OF THE MONTH
Did you know…
When examining trends and/or combining multiple years of data, it is important to check that the variables of interest were collected in all of the years contained in your analysis and that the wording of the item and/or answer categories did not change in a way that might impact the analysis. NCHS staff make every effort to point out important changes in the annual public use file documentation, both in the summary of changes at the beginning of the document and in the codebook section.
November 2018
DATA TIP OF THE MONTH
Did you know…
For 2012, estimates of visits to all providers (traditional NAMCS physicians and providers sampled within Community Health Center [CHC] strata) can be derived by combining physician data from the 2012 traditional NAMCS public use file with provider data from the 2012 NAMCS CHC public use data file. [PDF - 982 KB] New variables available only on the CHC public use file include type of CHC sampled provider (physician, nurse practitioner, physician assistant, or nurse midwife) and imputed time spent with non-physician clinician.
October 2018
DATA TIP OF THE MONTH
Did you know…
NAMCS and NHAMCS data can be used to estimate annual visit rates per unit of population (e.g., per 1,000 population) by dividing the number of visits by the population denominator of interest. Using rates removes the influence of different population sizes (e.g., for adults aged 65-74 and 75+) so that data can be compared across these groups. (
2015 NAMCS Public Use File Documentation [PDF - 1.6 MB])
September 2018
DATA TIP OF THE MONTH
Did you know…
Estimates and standard errors of drug mentions can be obtained by creating a new weight variable (e.g., DRUGWT) by multiplying PATWT (the patient visit weight) by NUMMED (the number of medications recorded at the visit) or DRUGWT=PATWT*NUMMED. DRUGWT can then be used in place of PATWT to produce the estimated number of drug mentions rather than visits. (
2015 NAMCS Public Use File Documentation [PDF - 1.6 MB])
August 2018
DATA TIP OF THE MONTH
Did you know…
The NAMCS and NHAMCS datasets that are available to approved users of the NCHS
Research Data Center contain additional variables not included on the public use datasets.