Common use of Sample Design and Response Rates Clause in Contracts

Sample Design and Response Rates. The MEPS HC is designed to produce estimates at the national and regional levels over time for the civilian, non-institutionalized population of the United States and some subpopulations of interest. The MEPS HC uses an overlapping panel design in which data for two calendar years are obtained through five rounds of data collection. A new sample (new Panel) of households for MEPS is selected each year from among household respondents to the previous year’s National Health Interview Survey (NHIS) conducted by the National Center for Health Statistics, Centers for Disease Control and Prevention. (The NHIS is an ongoing general health survey of the U.S. civilian non-institutionalized population.) MEPS Panel 15 spans the two calendar years 2010 and 2011 while MEPS Panel 16 spans 2011 and 2012. This file consists of the subset of data from the fifteenth and sixteenth MEPS panels covering from January 1 through, roughly, the spring of calendar year 2011. More specifically, data from the 2011 portion of the third Round of data collection for the MEPS Panel 15 sample are pooled with data from the first Round of data collection for the MEPS Panel 16 sample (see illustration below). Panel 15 2010-2011 2010 Jan 2011 Jan Round 1 Round 2 Round 3 Panel 00 0000-0000 Round 1 The sample for the NHIS is redesigned and redrawn about every ten years. From 1995 to 2005 the NHIS used the same sample design, and thus the MEPS, which began in 1996, has been based on a single NHIS design through MEPS Panel 11 initiated in 2006. Since, in the NHIS, the same PSUs and second stage sampling units are used each year, the MEPS sample from its inception has likewise been clustered within these same sampling units. However, a new sample design for the NHIS was implemented in 2006. The fundamental structure of the new 2006 NHIS sample design is very similar to the previous 1995-2005 NHIS sample design although the sample PSUs and second stage sampling units for the new NHIS design were selected independent of the sample selection process under the previous design. Of course, there is some overlap between the area populations covered by the sampled PSUs selected under the two designs, mostly the larger ones selected with certainty. As households selected for MEPS participation are selected from among the previous year’s NHIS respondents, the MEPS Panel 12, fielded in 2007, was the first MEPS Panel based on the new NHIS sample design. As a result, the 2007 PIT file consisted of two independent samples with MEPS Panel 11 reflecting the old design. Since then both MEPS panels have been based on the new design, including panels (15 and 16) in the PIT file. There were several new features to the MEPS sample design employed for Panel 16. First, in anticipation of a self-administered questionnaire to be given to MEPS participants with cancer, those NHIS households eligible for MEPS were selected for MEPS if they contained an NHIS respondent who had been randomly selected to complete the NHIS “sampled adult” questionnaire, and, in completing that questionnaire, indicated that s/he had some form of cancer. Second, the sample domain “Other”, which is the catchall stratum and generally consists mainly of households with “White” members, was partitioned into two sample domains: those households characterized as “complete household” respondents to the NHIS; and those characterized as “partial completes”. “NHIS partial completes” typically have a lower response rate to MEPS and for Panel 16 were sampled at a lower rate than “full completes”. This approach served to reduce survey costs, since the “partials” tend to have higher costs in gaining survey participation, but increased sample variability due to the resulting increased variance in sampling rates. Finally, a small experiment was conducted in 11 PSUs, exploring further the concept of trading off a cost reduction due to a reduced field load with an increase in variance due to subsampling. The experiment has been focused primarily on learning about the complexities of implementing such an effort as part of the MEPS data collection process across the full five rounds of MEPS. After roughly 10 weeks of data collection in these PSUs, those households characterized as “interim nonrespondents” and eligible for participation in this experiment were sampled at a 50 percent rate (some nonrespondents were not deemed eligible for this subsampling and were worked as usual). Those “interim nonrespondents” that were sampled had their contribution to the sample doubled (due to the 50 percent subsampling rate) while those not sampled were dropped from MEPS entirely. Thus, the Panel 16 Round 1 response rates discussed in the next section will reflect counts based on doubling the contribution of the subsampled “interim nonrespondents” and treating those “interim nonrespondents” not subsampled as if they had never been sampled for MEPS.

Sample Design and Response Rates. The MEPS HC is designed to produce estimates at the national and regional levels over time for the civilian, non-institutionalized population of the United States and some subpopulations of interest. The MEPS HC uses an overlapping panel design in which data for two calendar years are obtained through five rounds of data collection. A Each year a new sample (MEPS panel is formed. Each new Panel) panel consists of a randomly selected subsample of responding households for MEPS is selected each year from among household respondents to the previous year’s National Health Interview Survey (NHIS) conducted by the National Center for Health Statistics, Centers for Disease Control and Prevention. (The NHIS is an ongoing general health survey of the U.S. civilian non-non- institutionalized population.) MEPS Panel 15 19 spans the two calendar years 2010 2014 and 2011 2015 while MEPS Panel 16 20 spans 2011 2015 and 20122016. This year’s point-in-time file consists of contains the subset of data from respondents to the fifteenth nineteenth and sixteenth twentieth MEPS panels covering the time period from January 1 1, 2015 through, roughly, the spring of calendar year 20112015. More specificallyTo achieve this, data from representing the 2011 2015 portion of the third Round round of data collection for the MEPS Panel 15 19 sample are have been pooled with data from the first Round round of data collection for the MEPS Panel 16 20 sample (see illustration below). Panel 15 201019 2014-2011 2010 2015 2014 Jan 2011 2015 Jan Round 1 Round 2 Round 3 Panel 00 0000-0000 Round 1 The sample for the NHIS is redesigned and redrawn sample PSUs and secondary sampling units (SSUs) are newly selected about every ten years. These same sample PSUs and SSUs are then used each year for the NHIS until the next NHIS redesign. From 1995 to 2005 the NHIS used the same sample design, and thus . Thus the MEPS, which began in 1996, has been was based on a single that NHIS design through MEPS Panel 11 11, initiated in 2006. Since, in the NHIS, the same PSUs and second stage sampling units are used each year, clustering the MEPS sample in the same general geographic areas from its inception has likewise been clustered within these same sampling units1996-2006. However, With a new NHIS sample design in 2006, MEPS sampled households have been selected from a new set of areas for the NHIS was implemented in 20062007 MEPS onward. The It should be noted that the fundamental structure of the new 2006 NHIS sample design is very similar to the previous 1995-1995- 2005 NHIS sample design although the sample PSUs and second stage sampling units for the new newest NHIS design were selected independent of the sample selection process under the previous design. Of course, there is There has been some overlap between the area populations population areas covered by the sampled NHIS PSUs selected under the two designs, mostly the larger ones selected with certainty. As households selected for MEPS participation are selected from among the previous year’s NHIS respondents, the MEPS Panel 12, fielded in 2007, was the first MEPS Panel based on the new NHIS sample design. As a result, the 2007 PIT Point-in-Time (PIT) file consisted of two independent samples with MEPS Panel 11 reflecting the old design. Since then both MEPS panels have been based on the new design, including Panels 19 and 20, the MEPS panels (15 and 16) found in the this PIT file. There were several It may be noted that this is the last year that both panels will be based on the 2006 NHIS sample design. In 2016 a new features to the MEPS NHIS sample design employed for will have been implemented, and MEPS Panel 1621 will be based on the new design. First, in anticipation of a self-administered questionnaire to be given to MEPS participants with cancer, those NHIS households eligible for MEPS were selected for MEPS if they contained an NHIS respondent who had been randomly selected to complete the NHIS “sampled adult” questionnaire, and, in completing that questionnaire, indicated that s/he had some form of cancer. SecondFor both Panels 19 and 20, the sample domain “Other”, which is the catchall stratum and generally consists mainly of households with “White” members, was partitioned into two sample domains: those households characterized as “complete household” respondents to the NHIS; and those characterized as “partial completes”. “NHIS partial completes” typically have a lower response rate to MEPS and for Panel 16 were sampled at a lower rate than “full completes”” for both panels. This approach served to reduce survey costs, since the “partials” tend to have higher costs in gaining survey participation, but increased sample variability due to the resulting increased variance in sampling rates. Finally, a small experiment was conducted in 11 PSUs, exploring further the concept of trading off a cost reduction due to a reduced field load with an increase in variance due to subsampling. The experiment has been focused primarily on learning about the complexities of implementing such an effort as part of the MEPS data collection process across the full five rounds of MEPS. After roughly 10 weeks of data collection in these PSUs, those households characterized as “interim nonrespondents” and eligible for participation in this experiment were sampled at a 50 percent rate (some nonrespondents were not deemed eligible for this subsampling and were worked as usual). Those “interim nonrespondents” that were sampled had their contribution to the sample doubled (due to the 50 percent subsampling rate) while those not sampled were dropped from MEPS entirely. Thus, the Panel 16 Round 1 response rates discussed in the next section will reflect counts based on doubling the contribution of the subsampled “interim nonrespondents” and treating those “interim nonrespondents” not subsampled as if they had never been sampled for MEPS.

Sample Design and Response Rates. The MEPS HC is designed to produce estimates at the national and regional levels over time for the civilian, non-institutionalized population of the United States and some subpopulations of interest. The MEPS HC uses an overlapping panel design in which data for two calendar years are obtained through five rounds of data collection. A new sample (new Panel) of households for MEPS is selected each year from among household respondents to the previous year’s National Health Interview Survey (NHIS) conducted by the National Center for Health Statistics, Centers for Disease Control and Prevention. (The NHIS is an ongoing general health survey of the U.S. civilian non-institutionalized population.) MEPS Panel 15 17 spans the two calendar years 2010 2012 and 2011 2013 while MEPS Panel 16 18 spans 2011 2013 and 20122014. This year’s point-in-time file consists of contains the subset of data from respondents to the fifteenth seventeenth and sixteenth eighteenth MEPS panels covering the time period from January 1 1, 2013 through, roughly, the spring of calendar year 20112013. More specificallyTo achieve this, data from representing the 2011 2013 portion of the third Round round of data collection for the MEPS Panel 15 17 sample are have been pooled with data from the first Round round of data collection for the MEPS Panel 16 18 sample (see illustration below). Panel 15 201017 2012-2011 2010 2013 2012 Jan 2011 2013 Jan Round 1 Round 2 Round 3 Panel 00 0000-0000 Round 1 The sample for the NHIS is redesigned and redrawn sample PSUs and secondary sampling units (SSUs) are newly selected about every ten years. These same sample PSUs and SSUs are then used each year for the NHIS until the next NHIS redesign. From 1995 to 2005 the NHIS used the same sample design, and thus . Thus the MEPS, which began in 1996, has been was based on a single that NHIS design through MEPS Panel 11 11, initiated in 2006. Since, in the NHIS, the same PSUs and second stage sampling units are used each year, clustering the MEPS sample in the same general geographic areas from its inception has likewise been clustered within these same sampling units1996-2006. However, With a new NHIS sample design in 2006, MEPS sampled households have been selected from a new set of areas for the NHIS was implemented in 20062007 MEPS onward. The It should be noted that the fundamental structure of the new 2006 NHIS sample design is very similar to the previous 1995-1995- 2005 NHIS sample design although the sample PSUs and second stage sampling units for the new newest NHIS design were selected independent of the sample selection process under the previous design. Of course, there is There has been some overlap between the area populations population areas covered by the sampled NHIS PSUs selected under the two designs, mostly the larger ones selected with certainty. As households selected for MEPS participation are selected from among the previous year’s NHIS respondents, the MEPS Panel 12, fielded in 2007, was the first MEPS Panel based on the new NHIS sample design. As a result, the 2007 PIT Point-in-Time (PIT) file consisted of two independent samples with MEPS Panel 11 reflecting the old design. Since then both MEPS panels have been based on the new design, including Panels 17 and 18, the MEPS panels (15 and 16) found in the this PIT file. There were several new features to the MEPS sample design employed for Panel 16. First, in anticipation of a self-administered questionnaire to be given to MEPS participants with cancer, those NHIS households eligible for MEPS were selected for MEPS if they contained an NHIS respondent who had been randomly selected to complete the NHIS “sampled adult” questionnaire, and, in completing that questionnaire, indicated that s/he had some form of cancer. SecondFor both Panels 17 and 18, the sample domain “Other”, which is the catchall stratum and generally consists mainly of households with “White” members, was partitioned into two sample domains: those households characterized as “complete household” respondents to the NHIS; and those characterized as “partial completes”. “NHIS partial completes” typically have a lower response rate to MEPS and for Panel 16 were sampled at a lower rate than “full completes”” for both panels. This approach served to reduce survey costs, since the “partials” tend to have higher costs in gaining survey participation, but increased sample variability due to the resulting increased variance in sampling rates. Finally, a small experiment was conducted in 11 PSUs, exploring further the concept of trading off a cost reduction due to a reduced field load with an increase in variance due to subsampling. The experiment has been focused primarily on learning about the complexities of implementing such an effort as part of the MEPS data collection process across the full five rounds of MEPS. After roughly 10 weeks of data collection in these PSUs, those households characterized as “interim nonrespondents” and eligible for participation in this experiment were sampled at a 50 percent rate (some nonrespondents were not deemed eligible for this subsampling and were worked as usual). Those “interim nonrespondents” that were sampled had their contribution to the sample doubled (due to the 50 percent subsampling rate) while those not sampled were dropped from MEPS entirely. Thus, the Panel 16 Round 1 response rates discussed in the next section will reflect counts based on doubling the contribution of the subsampled “interim nonrespondents” and treating those “interim nonrespondents” not subsampled as if they had never been sampled for MEPS.

Sample Design and Response Rates. The MEPS HC is designed to produce estimates at the national and regional levels over time for the civilian, non-institutionalized population of the United States and some subpopulations of interest. The MEPS HC uses an overlapping panel design in which data for two calendar years are obtained through five rounds of data collection. A new sample (new Panel) of households for MEPS is selected each year from among household respondents to the previous year’s National Health Interview Survey (NHIS) conducted by the National Center for Health Statistics, Centers for Disease Control and Prevention. (The NHIS is an ongoing general health survey of the U.S. civilian non-institutionalized population.) MEPS Panel 15 16 spans the two calendar years 2010 2011 and 2011 2012 while MEPS Panel 16 17 spans 2011 2012 and 20122013. This year’s Point-in-Time file consists of contains the subset of data from respondents to the fifteenth sixteenth and sixteenth seventeenth MEPS panels covering the time period from January 1 1, 2012 through, roughly, the spring of calendar year 20112012. More specificallyTo achieve this, data from representing the 2011 2012 portion of the third Round of data collection for the MEPS Panel 15 sample are pooled with data from the first Round round of data collection for the MEPS Panel 16 sample have been pooled with data from the first round of data collection for the MEPS Panel 17 sample (see illustration below). Panel 15 201016 2011-2012 2011 2010 Jan 2011 2012 Jan Round 1 Round 2 Round 3 Panel 00 0000-0000 Round 1 The sample for the NHIS is redesigned and redrawn sample PSUs and secondary sampling units (SSUs) are newly selected about every ten years. These same sample PSUs and SSUs are then used each year for the NHIS until the next NHIS redesign. From 1995 to 2005 the NHIS used the same sample design, and thus . Thus the MEPS, which began in 1996, has been was based on a single that NHIS design through MEPS Panel 11 11, initiated in 2006. Since, in the NHIS, the same PSUs and second stage sampling units are used each year, clustering the MEPS sample in the same general geographic areas from its inception has likewise been clustered within these same sampling units1996-2006. However, With a new NHIS sample design in 2006, MEPS sampled households have been selected from a new set of areas for the NHIS was implemented in 20062007 MEPS onward. The It should be noted that the fundamental structure of the new 2006 NHIS sample design is very similar to the previous 1995-2005 NHIS sample design although the sample PSUs and second stage sampling units for the new NHIS design were selected independent of the sample selection process under the previous design. Of course, there is There has been some overlap between the area populations population areas covered by the sampled NHIS PSUs selected under the two designs, mostly the larger ones selected with certainty. As households selected for MEPS participation are selected from among the previous year’s NHIS respondents, the MEPS Panel 12, fielded in 2007, was the first MEPS Panel based on the new NHIS sample design. As a result, the 2007 PIT file consisted of two independent samples with MEPS Panel 11 reflecting the old design. Since then both MEPS panels have been based on the new design, including Panels 16 and 17, the MEPS panels (15 and 16) found in the this PIT file. There were several new two features to of the MEPS sample design employed for unique to Panel 16. First, in anticipation of a self-administered questionnaire to be given to MEPS participants with cancer, those NHIS households eligible for MEPS were selected for MEPS if they contained an NHIS respondent who had been randomly selected to complete the NHIS “sampled adult” questionnaire, and, in completing that questionnaire, indicated that s/he had some form of cancer. Second, the sample domain “Other”, which is the catchall stratum and generally consists mainly of households with “White” members, was partitioned into two sample domains: those households characterized as “complete household” respondents to the NHIS; and those characterized as “partial completes”. “NHIS partial completes” typically have a lower response rate to MEPS and for Panel 16 were sampled at a lower rate than “full completes”. This approach served to reduce survey costs, since the “partials” tend to have higher costs in gaining survey participation, but increased sample variability due to the resulting increased variance in sampling rates. Finally, a small experiment was conducted in 11 PSUs, exploring further the concept of trading off a cost reduction due to a reduced field load with an increase in variance due to subsampling. The experiment has been focused primarily on learning about the complexities of implementing such an effort as part of the MEPS data collection process across the full five rounds of MEPS. After roughly 10 weeks of data collection in these PSUs, those households characterized as “interim nonrespondents” and eligible for participation in this experiment were sampled at a 50 percent rate (some nonrespondents were not deemed eligible for this subsampling and were worked as usual). Those “interim nonrespondents” that were sampled had their contribution to the sample doubled (due to the 50 percent subsampling rate) while those not sampled were dropped from MEPS entirely. Thus, the Panel 16 Round 1 MEPS response rates rate information discussed in the next section will reflect reflects counts based on doubling the contribution of the subsampled “interim nonrespondents” and treating those “interim nonrespondents” not subsampled as if they had never been sampled for MEPS. Finally, for both Panels 16 and 17, the sample domain “Other”, which is the catchall stratum and consists mainly of households with “White” members, was partitioned into two sample domains: those households characterized as “complete household” respondents to the NHIS; and those characterized as “partial completes”. “NHIS partial completes” typically have a lower response rate to MEPS and were sampled at a lower rate than “full completes” for both panels. This approach served to reduce survey costs, since the “partials” tend to have higher costs in gaining survey participation, but increased sample variability due to the resulting increased variance in sampling rates.

Sample Design and Response Rates. The MEPS HC is designed to produce estimates at the national and regional levels over time for the civilian, non-institutionalized population of the United States and some subpopulations of interest. The MEPS HC uses an overlapping panel design in which data for two calendar years are obtained through five rounds of data collection. A Each year a new sample (MEPS panel is formed. Each new Panel) panel consists of a randomly selected subsample of responding households for MEPS is selected each year from among household respondents to the previous year’s National Health Interview Survey (NHIS) conducted by the National Center for Health Statistics, Centers for Disease Control and Prevention. (The NHIS is an ongoing general health survey of the U.S. civilian non-non- institutionalized population.) MEPS Panel 15 18 spans the two calendar years 2010 2013 and 2011 2014 while MEPS Panel 16 19 spans 2011 2014 and 20122015. This year’s point-in-time file consists of contains the subset of data from respondents to the fifteenth eighteenth and sixteenth nineteenth MEPS panels covering the time period from January 1 1, 2014 through, roughly, the spring of calendar year 20112014. More specificallyTo achieve this, data from representing the 2011 2014 portion of the third Round round of data collection for the MEPS Panel 15 18 sample are have been pooled with data from the first Round round of data collection for the MEPS Panel 16 19 sample (see illustration below). Panel 15 201018 2013-2011 2010 2014 2013 Jan 2011 2014 Jan Round 1 Round 2 Round 3 Panel 00 0000-0000 Round 1 The sample for the NHIS is redesigned and redrawn sample PSUs and secondary sampling units (SSUs) are newly selected about every ten years. These same sample PSUs and SSUs are then used each year for the NHIS until the next NHIS redesign. From 1995 to 2005 the NHIS used the same sample design, and thus . Thus the MEPS, which began in 1996, has been was based on a single that NHIS design through MEPS Panel 11 11, initiated in 2006. Since, in the NHIS, the same PSUs and second stage sampling units are used each year, clustering the MEPS sample in the same general geographic areas from its inception has likewise been clustered within these same sampling units1996-2006. However, With a new NHIS sample design in 2006, MEPS sampled households have been selected from a new set of areas for the NHIS was implemented in 20062007 MEPS onward. The It should be noted that the fundamental structure of the new 2006 NHIS sample design is very similar to the previous 1995-1995- 2005 NHIS sample design although the sample PSUs and second stage sampling units for the new newest NHIS design were selected independent of the sample selection process under the previous design. Of course, there is There has been some overlap between the area populations population areas covered by the sampled NHIS PSUs selected under the two designs, mostly the larger ones selected with certainty. As households selected for MEPS participation are selected from among the previous year’s NHIS respondents, the MEPS Panel 12, fielded in 2007, was the first MEPS Panel based on the new NHIS sample design. As a result, the 2007 PIT Point-in-Time (PIT) file consisted of two independent samples with MEPS Panel 11 reflecting the old design. Since then both MEPS panels have been based on the new design, including Panels 18 and 19, the MEPS panels (15 and 16) found in the this PIT file. There were several new features to the MEPS sample design employed for Panel 16. First, in anticipation of a self-administered questionnaire to be given to MEPS participants with cancer, those NHIS households eligible for MEPS were selected for MEPS if they contained an NHIS respondent who had been randomly selected to complete the NHIS “sampled adult” questionnaire, and, in completing that questionnaire, indicated that s/he had some form of cancer. SecondFor both Panels 18 and 19, the sample domain “Other”, which is the catchall stratum and generally consists mainly of households with “White” members, was partitioned into two sample domains: those households characterized as “complete household” respondents to the NHIS; and those characterized as “partial completes”. “NHIS partial completes” typically have a lower response rate to MEPS and for Panel 16 were sampled at a lower rate than “full completes”” for both panels. This approach served to reduce survey costs, since the “partials” tend to have higher costs in gaining survey participation, but increased sample variability due to the resulting increased variance in sampling rates. Finally, a small experiment was conducted in 11 PSUs, exploring further the concept of trading off a cost reduction due to a reduced field load with an increase in variance due to subsampling. The experiment has been focused primarily on learning about the complexities of implementing such an effort as part of the MEPS data collection process across the full five rounds of MEPS. After roughly 10 weeks of data collection in these PSUs, those households characterized as “interim nonrespondents” and eligible for participation in this experiment were sampled at a 50 percent rate (some nonrespondents were not deemed eligible for this subsampling and were worked as usual). Those “interim nonrespondents” that were sampled had their contribution to the sample doubled (due to the 50 percent subsampling rate) while those not sampled were dropped from MEPS entirely. Thus, the Panel 16 Round 1 response rates discussed in the next section will reflect counts based on doubling the contribution of the subsampled “interim nonrespondents” and treating those “interim nonrespondents” not subsampled as if they had never been sampled for MEPS.

Sample Design and Response Rates. The MEPS HC is designed to produce estimates at the national and regional levels over time for the civilian, non-institutionalized population of the United States and some subpopulations of interest. The MEPS HC uses an overlapping panel design in which data for two calendar years are obtained through five rounds of data collection. A new sample (new Panel) of households for MEPS is selected each year from among household respondents to the previous year’s National Health Interview Survey (NHIS) conducted by the National Center for Health Statistics, Centers for Disease Control and Prevention. (The NHIS is an ongoing general health survey of the U.S. civilian non-institutionalized population.) MEPS Panel 15 13 spans the two calendar years 2010 2008 and 2011 2009 while MEPS Panel 16 14 spans 2011 2009 and 20122010. This file consists of the subset of data from the fifteenth thirteenth and sixteenth fourteenth MEPS panels covering from January 1 through, roughly, the spring of calendar year 20112009. More specifically, data from the 2011 2009 portion of the third Round of data collection for the MEPS Panel 15 13 sample are pooled with data from the first Round of data collection for the MEPS Panel 16 14 sample (see illustration below). Panel 15 201013 2008-2011 2010 2009 2008 Jan 2011 2009 Jan Round 1 Round 2 Round 3 Panel 00 0000-0000 Round 1 The Traditionally, the sample for the NHIS is redesigned and redrawn about every ten years. From 1995 to 2005 the NHIS used the same sample design, and thus the MEPS, which began in 1996, has been based on a single NHIS design through MEPS Panel 11 initiated in 2006. Since, in the NHIS, the same PSUs and second stage sampling units are used each year, the MEPS sample from its inception has likewise been clustered within these same sampling units. However, a new sample design for the NHIS was implemented in 2006. The fundamental structure of the new 2006 NHIS sample design is very similar to the previous 1995-2005 NHIS sample design although the design. The sample PSUs and second stage sampling units for the new NHIS design were selected independent of the sample selection process under the previous design. Of course, there is some overlap between the area populations covered by the sampled PSUs selected under the two designs, mostly the larger ones selected with certainty. As households selected for MEPS participation are selected from among the previous year’s NHIS respondents, the MEPS Panel 12, fielded in 2007, was the first MEPS Panel based on the new NHIS sample design. As a result, the 2007 PIT file consisted of two independent samples samples, serving to increase precision (due to decreased clustering since the sample was spread out among more PSUs and secondary sampling units) and increased degrees of freedom (due to many more variance strata available for variance estimation purposes). As with MEPS Panel 11 reflecting the old design. Since then 2008 PIT file, both MEPS panels have been based on the new design, including panels Panels (15 13 and 1614) in the 2009 PIT file. There were several new features to the MEPS sample design employed for Panel 16. First, in anticipation of a self-administered questionnaire to be given to MEPS participants with cancer, those NHIS households eligible for MEPS were selected for MEPS if they contained an NHIS respondent who had been randomly selected to complete the NHIS “sampled adult” questionnaire, and, in completing that questionnaire, indicated that s/he had some form of cancer. Second, the sample domain “Other”, which is the catchall stratum and generally consists mainly of households with “White” members, was partitioned into two sample domains: those households characterized as “complete household” respondents to the NHIS; and those characterized as “partial completes”. “NHIS partial completes” typically have a lower response rate to MEPS and for Panel 16 were sampled at a lower rate than “full completes”. This approach served to reduce survey costs, since the “partials” tend to have higher costs in gaining survey participation, but increased sample variability due to the resulting increased variance in sampling rates. Finally, a small experiment was conducted in 11 PSUs, exploring further the concept of trading off a cost reduction due to a reduced field load with an increase in variance due to subsampling. The experiment has been focused primarily on learning about the complexities of implementing such an effort as part of the MEPS data collection process across the full five rounds of MEPS. After roughly 10 weeks of data collection in these PSUs, those households characterized as “interim nonrespondents” and eligible for participation in this experiment were sampled at a 50 percent rate (some nonrespondents were not deemed eligible for this subsampling and were worked as usual). Those “interim nonrespondents” that were sampled had their contribution to the sample doubled (due to the 50 percent subsampling rate) while those not sampled were dropped from MEPS entirely. Thus, the Panel 16 Round 1 response rates discussed in the next section will reflect counts file are based on doubling the contribution of the subsampled “interim nonrespondents” and treating those “interim nonrespondents” not subsampled as if they had never been sampled for MEPSa single NHIS sample design.

Sample Design and Response Rates. The MEPS HC is designed to produce estimates at the national and regional levels over time for the civilian, non-institutionalized population of the United States and some subpopulations of interest. The MEPS HC uses an overlapping panel design in which data for two calendar years are obtained through five rounds of data collection. A new sample (new Panel) of households for MEPS is selected each year from among household respondents to the previous year’s National Health Interview Survey (NHIS) conducted by the National Center for Health Statistics, Centers for Disease Control and Prevention. (The NHIS is an ongoing general health survey of the U.S. civilian non-institutionalized population.) MEPS Panel 15 14 spans the two calendar years 2009 and 2010 and 2011 while MEPS Panel 16 15 spans 2011 2010 and 20122011. This file consists of the subset of data from the fourteenth and fifteenth and sixteenth MEPS panels covering from January 1 through, roughly, the spring of calendar year 20112010. More specifically, data from the 2011 2010 portion of the third Round of data collection for the MEPS Panel 15 14 sample are pooled with data from the first Round of data collection for the MEPS Panel 16 15 sample (see illustration below). 2010 Jan Round 1 Round 3 Panel 15 2010-2011 2010 2009 Jan 2011 Jan Round 1 Round 2 Round 3 1 Panel 00 000014 2009-0000 Round 1 2010 The sample for the NHIS is redesigned and redrawn about every ten years. From 1995 to 2005 the NHIS used the same sample design, and thus the MEPS, which began in 1996, has been based on a single NHIS design through MEPS Panel 11 initiated in 2006. Since, in the NHIS, the same PSUs and second stage sampling units are used each year, the MEPS sample from its inception has likewise been clustered within these same sampling units. However, a new sample design for the NHIS was implemented in 2006. The fundamental structure of the new 2006 NHIS sample design is very similar to the previous 1995-2005 NHIS sample design although the sample PSUs and second stage sampling units for the new NHIS design were selected independent of the sample selection process under the previous design. Of course, there is some overlap between the area populations covered by the sampled PSUs selected under the two designs, mostly the larger ones selected with certainty. As households selected for MEPS participation are selected from among the previous year’s NHIS respondents, the MEPS Panel 12, fielded in 2007, was the first MEPS Panel based on the new NHIS sample design. As a result, the 2007 PIT file consisted of two independent samples with MEPS Panel 11 reflecting the old design. Since then both MEPS panels have been based on the new design, including panels (15 14 and 1615) in the PIT file. There were several new features to the MEPS sample design employed for Panel 16. First, in anticipation of a self-administered questionnaire to be given to MEPS participants with cancer, those NHIS households eligible for MEPS were selected for MEPS if they contained an NHIS respondent who had been randomly selected to complete the NHIS “sampled adult” questionnaire, and, in completing that questionnaire, indicated that s/he had some form of cancer. Second, the sample domain “Other”, which is the catchall stratum and generally consists mainly of households with “White” members, was partitioned into two sample domains: those households characterized as “complete household” respondents to the NHIS; and those characterized as “partial completes”. “NHIS partial completes” typically have a lower response rate to MEPS and for Panel 16 were sampled at a lower rate than “full completes”. This approach served to reduce survey costs, since the “partials” tend to have higher costs in gaining survey participation, but increased sample variability due to the resulting increased variance in sampling rates. Finally, a small experiment was conducted in 11 PSUs, exploring further the concept of trading off a cost reduction due to a reduced field load with an increase in variance due to subsampling. The experiment has been focused primarily on learning about the complexities of implementing such an effort as part of the MEPS data collection process across the full five rounds of MEPS. After roughly 10 weeks of data collection in these PSUs, those households characterized as “interim nonrespondents” and eligible for participation in this experiment were sampled at a 50 percent rate (some nonrespondents were not deemed eligible for this subsampling and were worked as usual). Those “interim nonrespondents” that were sampled had their contribution to the sample doubled (due to the 50 percent subsampling rate) while those not sampled were dropped from MEPS entirely. Thus, the Panel 16 Round 1 response rates discussed in the next section will reflect counts based on doubling the contribution of the subsampled “interim nonrespondents” and treating those “interim nonrespondents” not subsampled as if they had never been sampled for MEPS.

Sample Design and Response Rates. The MEPS HC is designed to produce estimates at the national and regional levels over time for the civilian, non-institutionalized population of the United States and some subpopulations of interest. The MEPS HC uses an overlapping panel design in which data for two calendar years are obtained through five rounds of data collection. A new sample (new Panel) of households for MEPS is selected each year from among household respondents to the previous year’s National Health Interview Survey (NHIS) conducted by the National Center for Health Statistics, Centers for Disease Control and Prevention. (The NHIS is an ongoing general health survey of the U.S. civilian non-institutionalized population.) MEPS Panel 15 12 spans the two calendar years 2010 2007 and 2011 2008 while MEPS Panel 16 13 spans 2011 2008 and 20122009. This file consists of the subset of data from the fifteenth twelfth and sixteenth thirteenth MEPS panels covering from January 1 through, roughly, the spring of calendar year 20112008. More specifically, data from the 2011 2008 portion of the third Round of data collection for the MEPS Panel 15 12 sample are pooled with data from the first Round of data collection for the MEPS Panel 16 13 sample (see illustration below). 2007 2008 Panel 15 201012 2007-2011 2010 Jan 2011 Jan 2008 Xxx Xxx Round 1 Round 2 Round 3 Panel 00 0000-0000 Round 1 The Traditionally, the sample for the NHIS is redesigned and redrawn about every ten years. From 1995 to 2005 the NHIS used the same sample design, and thus the MEPS, which began in 1996, has been based on a single NHIS design through MEPS Panel 11 initiated in 2006. Since, in the NHIS, the same PSUs and second stage sampling units are used each year, the MEPS sample from its inception has likewise been clustered within these same sampling units. However, a new sample design for the NHIS was implemented in 2006. The fundamental structure of the new 2006 NHIS sample design is very similar to the previous 1995-2005 NHIS sample design although the design. The sample PSUs and second stage sampling units for the new NHIS design were selected independent of the sample selection process under the previous design. Of course, there is some overlap between the area populations covered by the sampled PSUs selected under the two designs, mostly the larger ones selected with certainty. As households selected for MEPS participation are selected from among the previous year’s NHIS respondents, the MEPS Panel 12, fielded in 2007, was the first MEPS Panel based on the new NHIS sample design. As a result, the 2007 PIT file consisted of two independent samples with MEPS Panel 11 reflecting samples, serving to increase precision (due to decreased clustering since the old designsample was spread out among more PSUs and secondary sampling units) and increased degrees of freedom (due to many more variance strata available for variance estimation purposes). Since then The 2008 PIT file marks a return to having both MEPS panels have been Panels (12 and 13) being based on the new a single NHIS sample design, including panels serving to reduce both precision (15 due to increased clustering) and 16degrees of freedom (due to a reduction in variance strata) relative to the 2007 PIT data base. As a result, standard errors are expected to be generally higher than they were for the 2007 PIT for estimates where people living in the PIT file. There were several new features to the MEPS sample design employed for Panel 16. First, in anticipation of a self-administered questionnaire to be given to MEPS participants with cancer, those NHIS households eligible for MEPS were selected for MEPS if they contained an NHIS respondent who had been randomly selected to complete the NHIS “sampled adult” questionnaire, and, in completing that questionnaire, indicated that s/he had some form of cancer. Second, the sample domain “Other”, which is the catchall stratum and generally consists mainly of households with “White” members, was partitioned into two sample domains: those households characterized as “complete household” respondents to the NHIS; and those characterized as “partial completes”. “NHIS partial completes” typically have a lower response rate to MEPS and for Panel 16 were sampled at a lower rate than “full completes”. This approach served to reduce survey costs, since the “partials” same general vicinity tend to have higher costs in gaining survey participation, but increased sample variability due similar responses to the resulting increased variance in sampling rates. Finally, a small experiment was conducted in 11 PSUs, exploring further the concept questionnaire items and may be somewhat more unstable for estimates with relatively few degrees of trading off a cost reduction due freedom for 2008 data compared to a reduced field load with an increase in variance due to subsampling. The experiment has been focused primarily on learning about the complexities of implementing such an effort as part of the MEPS data collection process across the full five rounds of MEPS. After roughly 10 weeks of data collection in these PSUs, those households characterized as “interim nonrespondents” and eligible that for participation in this experiment were sampled at a 50 percent rate (some nonrespondents were not deemed eligible for this subsampling and were worked as usual). Those “interim nonrespondents” that were sampled had their contribution to the sample doubled (due to the 50 percent subsampling rate) while those not sampled were dropped from MEPS entirely. Thus, the Panel 16 Round 1 response rates discussed in the next section will reflect counts based on doubling the contribution of the subsampled “interim nonrespondents” and treating those “interim nonrespondents” not subsampled as if they had never been sampled for MEPS2007.