Hidradenitis Suppurativa Prevalence in the United States
Hidradenitis Suppurativa Prevalence in the United States
NAMCS and NHAMCS are event-based, rather than population-based, surveys, and are designed to provide estimates of the number of patient-physician encounters, or visits. Detailed documentation for the NAMCS and NHAMCS surveys are provided by the NCHS. NAMCS microdata files are derived annually by the NCHS from patient record forms documenting visits by patients to physicians at private offices and clinics. Annual NHAMCS data are divided into OPD and ED files. NHAMCS OPD files are organized similarly to NAMCS files, with many identical or similar variables.
NAMCS and NHAMCS data beginning in 2002 include sample design variables intended for use with software using ultimate cluster variance methods. Visit weights are included in each record in the public microdata file, so that national estimates of patient visits can be obtained from the number of records for selected variables. The most recent available public data are for visits in 2010. For the analyses performed here, the annual NAMCS, NHAMCS OPD, and NHAMCS ED microdata files from 2002 to 2010 were merged to produce a single file with records from physician offices and hospital OPDs and EDs.
Patient visits for treatment of HS were identified within the merged file by selecting records in which any of the 3 listed diagnoses was for HS, as identified by the International Classification of Diseases, Ninth Revision, Clinical Modification, code 705.83. The first listed diagnosis is the physician's primary diagnosis for the visit.
The patient's primary expected payment source for a visit is recorded in the data file on the basis of a hierarchy of expected sources of payment, with Medicare or Medicaid being first, followed by workers compensation, private insurance, self-payment, and no charge. Here, the variables for the primary expected payment source listed for each visit were combined and recoded into the following 3 groups: public (combining Medicare and Medicaid), private (private insurance and workers compensation), and uninsured (self-pay, no charge, other, unknown, or blank). Thus, if a patient had both Medicare and private insurance, the visit would be coded to indicate the primary expected payment source was public insurance. Patient age was recoded into 4 groups. The original race categories of black, white, and other were recoded as black or white/other, because the number of records with HS diagnosis and race category of "other" was too small for analysis. These characteristics of patients at visits recording a diagnosis of HS were compared with those of patients in records of visits with any diagnosis within the merged data file.
The method of Burt and Hing was used to estimate the number of patients with HS. The visit weight variable in the microdata file was divided by the total number of visits within the previous 12 months to the same physician to yield a patient weight variable. Use of the method of Burt and Hing will overestimate patient numbers when patients see more than 1 physician. However, when using records for visits in which a single diagnosis (e.g., HS) is clinically relevant, overestimation is less likely because most patients will have a single physician treating a particular condition in a given year. Estimation of the number of patients with HS was performed using NAMCS and NHAMCS OPD data only, as well as with the inclusion of NHAMCS ED data. For ED records before 2007 that did not include numbers of past visits, it was assumed that there were no other ED visits during the past 12 months. Patient numbers estimated without ED data are lower limits, because they exclude those diagnosed with HS who were seen only in the ED. Numbers estimated with ED data are upper limits, because a patient who is seen in the ED and at an office in the same 12-month period may be counted twice. Also, assuming no prior ED visits when ED patient record forms do not include the past visit variable will lead to overestimation if a patient has made multiple ED visits. However, because EDs frequently see patients with HS, and because some uninsured patients may obtain treatment only at EDs, that setting is considered in the analysis.
When analyzing the merged file containing all records of patient visits with any diagnosis, the method of Burt and Hing is inaccurate because of the likelihood that patients will have more than 1 physician. To compare the HS patient population to the population of all persons who visited a physician within a 12-month period from 2002 to 2010, we estimated the latter using National Health Interview Survey databases. The use of NHIS databases is described in the Web Appendix available at http://aje.oxfordjournals.org/.
Treatments prescribed or performed at visits in which there was a diagnosis of HS were analyzed. In survey years 2006–2010, the medication variables included codes from the Cerner Multum Lexicon Plus therapeutic classification system (Cerner Multum, Inc., Denver, Colorado). For each medication, which may be a single- or multiple-ingredient drug, the Cerner Multum Lexicon lists up to 4 therapeutic categories with up to 3 levels of increasing detail per category. In survey years 2002–2005, medications had been coded using the US Food and Drug Administration's National Drug Code Directory. For the analysis performed here, the listed medications for those years were recoded according to the Cerner Multum system. Estimates were then made for systemic antibiotics (i.e., drug categories 008-018, 240, and 315), systemic analgesics (i.e, drug category 058), biological agents (i.e., drug category 028), and topical dermatological agents (i.e., drug category 136) among any of the 8 listed medications included in each visit record (6 listed medications in records in survey year 2002).
The number of visits with an HS diagnosis in which at least 1 procedure was performed was estimated by selecting records for which an excision or incision and drainage was indicated on the patient record form, or the International Classification of Diseases, Ninth Revision, Clinical Modification, code 85.21, 86.01, 86.04, 86.09, 86.22, or 86.30 was included among the listed procedures. Also included were records for which a wound care procedure or procedures coded 93.57 or 93.59 were indicated.
Analysis was performed using the Complex Samples Module of SPSS, version 21, software (IBM Corp., Somers, New York). Estimates are reported with standard errors. NCHS criteria (number of records n ≥ 30 and relative standard error ≤ 30%) for statistical reliability for estimates of visits were followed unless otherwise noted. Reweighting increases standard error; here, estimates of patient numbers were considered reliable if the number of records was 30 or more and relative standard error was 35% or less. P values were calculated from Rao-Scott Pearson χ test statistics using the Crosstabs function of SPSS when comparing within the merged NAMCS and NHAMCS file. When comparing those data to National Health Interview Survey data, we implemented the Pearson χ test as described in the Web Appendix. Significance was examined at the 0.05 level.
Methods
NAMCS and NHAMCS are event-based, rather than population-based, surveys, and are designed to provide estimates of the number of patient-physician encounters, or visits. Detailed documentation for the NAMCS and NHAMCS surveys are provided by the NCHS. NAMCS microdata files are derived annually by the NCHS from patient record forms documenting visits by patients to physicians at private offices and clinics. Annual NHAMCS data are divided into OPD and ED files. NHAMCS OPD files are organized similarly to NAMCS files, with many identical or similar variables.
NAMCS and NHAMCS data beginning in 2002 include sample design variables intended for use with software using ultimate cluster variance methods. Visit weights are included in each record in the public microdata file, so that national estimates of patient visits can be obtained from the number of records for selected variables. The most recent available public data are for visits in 2010. For the analyses performed here, the annual NAMCS, NHAMCS OPD, and NHAMCS ED microdata files from 2002 to 2010 were merged to produce a single file with records from physician offices and hospital OPDs and EDs.
Patient visits for treatment of HS were identified within the merged file by selecting records in which any of the 3 listed diagnoses was for HS, as identified by the International Classification of Diseases, Ninth Revision, Clinical Modification, code 705.83. The first listed diagnosis is the physician's primary diagnosis for the visit.
The patient's primary expected payment source for a visit is recorded in the data file on the basis of a hierarchy of expected sources of payment, with Medicare or Medicaid being first, followed by workers compensation, private insurance, self-payment, and no charge. Here, the variables for the primary expected payment source listed for each visit were combined and recoded into the following 3 groups: public (combining Medicare and Medicaid), private (private insurance and workers compensation), and uninsured (self-pay, no charge, other, unknown, or blank). Thus, if a patient had both Medicare and private insurance, the visit would be coded to indicate the primary expected payment source was public insurance. Patient age was recoded into 4 groups. The original race categories of black, white, and other were recoded as black or white/other, because the number of records with HS diagnosis and race category of "other" was too small for analysis. These characteristics of patients at visits recording a diagnosis of HS were compared with those of patients in records of visits with any diagnosis within the merged data file.
The method of Burt and Hing was used to estimate the number of patients with HS. The visit weight variable in the microdata file was divided by the total number of visits within the previous 12 months to the same physician to yield a patient weight variable. Use of the method of Burt and Hing will overestimate patient numbers when patients see more than 1 physician. However, when using records for visits in which a single diagnosis (e.g., HS) is clinically relevant, overestimation is less likely because most patients will have a single physician treating a particular condition in a given year. Estimation of the number of patients with HS was performed using NAMCS and NHAMCS OPD data only, as well as with the inclusion of NHAMCS ED data. For ED records before 2007 that did not include numbers of past visits, it was assumed that there were no other ED visits during the past 12 months. Patient numbers estimated without ED data are lower limits, because they exclude those diagnosed with HS who were seen only in the ED. Numbers estimated with ED data are upper limits, because a patient who is seen in the ED and at an office in the same 12-month period may be counted twice. Also, assuming no prior ED visits when ED patient record forms do not include the past visit variable will lead to overestimation if a patient has made multiple ED visits. However, because EDs frequently see patients with HS, and because some uninsured patients may obtain treatment only at EDs, that setting is considered in the analysis.
When analyzing the merged file containing all records of patient visits with any diagnosis, the method of Burt and Hing is inaccurate because of the likelihood that patients will have more than 1 physician. To compare the HS patient population to the population of all persons who visited a physician within a 12-month period from 2002 to 2010, we estimated the latter using National Health Interview Survey databases. The use of NHIS databases is described in the Web Appendix available at http://aje.oxfordjournals.org/.
Treatments prescribed or performed at visits in which there was a diagnosis of HS were analyzed. In survey years 2006–2010, the medication variables included codes from the Cerner Multum Lexicon Plus therapeutic classification system (Cerner Multum, Inc., Denver, Colorado). For each medication, which may be a single- or multiple-ingredient drug, the Cerner Multum Lexicon lists up to 4 therapeutic categories with up to 3 levels of increasing detail per category. In survey years 2002–2005, medications had been coded using the US Food and Drug Administration's National Drug Code Directory. For the analysis performed here, the listed medications for those years were recoded according to the Cerner Multum system. Estimates were then made for systemic antibiotics (i.e., drug categories 008-018, 240, and 315), systemic analgesics (i.e, drug category 058), biological agents (i.e., drug category 028), and topical dermatological agents (i.e., drug category 136) among any of the 8 listed medications included in each visit record (6 listed medications in records in survey year 2002).
The number of visits with an HS diagnosis in which at least 1 procedure was performed was estimated by selecting records for which an excision or incision and drainage was indicated on the patient record form, or the International Classification of Diseases, Ninth Revision, Clinical Modification, code 85.21, 86.01, 86.04, 86.09, 86.22, or 86.30 was included among the listed procedures. Also included were records for which a wound care procedure or procedures coded 93.57 or 93.59 were indicated.
Analysis was performed using the Complex Samples Module of SPSS, version 21, software (IBM Corp., Somers, New York). Estimates are reported with standard errors. NCHS criteria (number of records n ≥ 30 and relative standard error ≤ 30%) for statistical reliability for estimates of visits were followed unless otherwise noted. Reweighting increases standard error; here, estimates of patient numbers were considered reliable if the number of records was 30 or more and relative standard error was 35% or less. P values were calculated from Rao-Scott Pearson χ test statistics using the Crosstabs function of SPSS when comparing within the merged NAMCS and NHAMCS file. When comparing those data to National Health Interview Survey data, we implemented the Pearson χ test as described in the Web Appendix. Significance was examined at the 0.05 level.
Source...