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 12 spans the two calendar years 2007 and 2008 while MEPS Panel 13 spans 2008 and 2009. This file consists of the subset of data from the twelfth and thirteenth MEPS panels covering from January 1 through, roughly, the spring of calendar year 2008. More specifically, data from the 2008 portion of the third Round of data collection for the MEPS Panel 12 sample are pooled with data from the first Round of data collection for the MEPS Panel 13 sample (see illustration below). Panel 12 2007-2008 Xxx Xxx Round 1 Round 2 Round 3 Panel 00 0000-0000 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. 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, 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). The 2008 PIT file marks a return to having both MEPS Panels (12 and 13) being based on a single NHIS sample design, serving to reduce both precision (due to increased clustering) and degrees 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 same general vicinity tend to have similar responses to questionnaire items and may be somewhat more unstable for estimates with relatively few degrees of freedom for 2008 data compared to that for 2007.

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 '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 12 11 spans the two calendar years 2006 and 2007 and 2008 while MEPS Panel 13 12 spans 2008 2007 and 20092008. This file consists of the subset of data from the eleventh and twelfth and thirteenth MEPS panels covering from January 1 through, roughly, the spring of calendar year 20082007. More specifically, data from the 2008 2007 portion of the third Round of data collection for the MEPS Panel 12 11 sample are pooled with data from the first Round of data collection for the MEPS Panel 13 12 sample (see illustration below). Panel 12 2007-2008 Xxx Xxx Round 1 Round 2 Round 3 Panel 00 0000-0000 Panel 11 2006-2007 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, Since in the NHIS, 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. 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 is the first MEPS Panel based on the new NHIS sample design. There are several implications with respect to this design change that should be noted. With two independent samples a new set of variance strata and PSUs had to be developed for Panel 12 while the ones associated with the old design were retained for Panel 11. Thus, there will be more variance strata and PSUs available for estimation purposes for the 2007 PIT database than for previous PIT files and, consequently, more degrees of freedom. Also, the degree of clustering of the sample will be lessened for this database since the two MEPS panels were not sampled from the same set of PSUs and secondary sample units. As a result, with the 2007 PIT file consisted reduction of two independent samples, serving to increase precision (due to decreased the clustering since of 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). The 2008 PIT file marks a return to having both MEPS Panels (12 and 13) being based on a single NHIS sample design, serving to reduce both precision (due to increased clustering) and degrees of freedom (due to a reduction in variance strata) relative to the 2007 PIT data base. As a resultsample, standard errors are expected to be generally higher lower than they were for the 2007 PIT would otherwise be for estimates where people living in the same general vicinity tend to have similar responses to questionnaire items items. In addition, as with any change in sample or study design, MEPS estimates have been and may will continue to be somewhat more unstable for assessed to determine if any substantial change in the survey estimates might be associated with relatively few degrees a change in design (e.g., as could arise due to increased coverage of freedom for 2008 data compared to that for 2007the target populations).

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 12 19 spans the two calendar years 2007 2014 and 2008 2015 while MEPS Panel 13 20 spans 2008 2015 and 20092016. This year’s point-in-time file consists of contains the subset of data from respondents to the twelfth nineteenth and thirteenth twentieth MEPS panels covering the time period from January 1 1, 2015 through, roughly, the spring of calendar year 20082015. More specificallyTo achieve this, data from representing the 2008 2015 portion of the third Round round of data collection for the MEPS Panel 12 19 sample are have been pooled with data from the first Round round of data collection for the MEPS Panel 13 20 sample (see illustration below). Panel 12 200719 2014-2008 Xxx Xxx 2015 2014 Jan 2015 Jan Round 1 Round 2 Round 3 Panel 00 0000-0000 Traditionally, the 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. The 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, serving to increase precision (due to decreased clustering since samples with MEPS Panel 11 reflecting 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)old design. The 2008 PIT file marks a return to having Since then both MEPS Panels (12 and 13) being panels have been based on a single the new design, including Panels 19 and 20, the MEPS panels found in this PIT file. 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 NHIS sample design will have been implemented, serving and MEPS Panel 21 will be based on the new design. For both Panels 19 and 20, 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 both precision (due to increased clustering) and degrees of freedom (due to a reduction in variance strata) relative to survey costs, since 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 same general vicinity “partials” tend to have similar responses higher costs in gaining survey participation, but increased sample variability due to questionnaire items and may be somewhat more unstable for estimates with relatively few degrees of freedom for 2008 data compared to that for 2007the 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 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 12 17 spans the two calendar years 2007 2012 and 2008 2013 while MEPS Panel 13 18 spans 2008 2013 and 20092014. This year’s point-in-time file consists of contains the subset of data from respondents to the twelfth seventeenth and thirteenth eighteenth MEPS panels covering the time period from January 1 1, 2013 through, roughly, the spring of calendar year 20082013. More specificallyTo achieve this, data from representing the 2008 2013 portion of the third Round round of data collection for the MEPS Panel 12 17 sample are have been pooled with data from the first Round round of data collection for the MEPS Panel 13 18 sample (see illustration below). Panel 12 200717 2012-2008 Xxx Xxx 2013 2012 Jan 2013 Jan Round 1 Round 2 Round 3 Panel 00 0000-0000 Traditionally, the 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. The 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 samplessamples with MEPS Panel 11 reflecting the old design. Since then both MEPS panels have been based on the new design, serving including Panels 17 and 18, the MEPS panels found in this PIT file. For both Panels 17 and 18, 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 increase precision (due the NHIS; and those characterized as “partial completes”. “NHIS partial completes” typically have a lower response rate to decreased clustering MEPS and were sampled at a lower rate than “full completes” for both panels. This approach served to reduce survey costs, 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). The 2008 PIT file marks a return to having both MEPS Panels (12 and 13) being based on a single NHIS sample design, serving to reduce both precision (due to increased clustering) and degrees 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 same general vicinity “partials” tend to have similar responses higher costs in gaining survey participation, but increased sample variability due to questionnaire items and may be somewhat more unstable for estimates with relatively few degrees of freedom for 2008 data compared to that for 2007the 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 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 12 15 spans the two calendar years 2007 2010 and 2008 2011 while MEPS Panel 13 16 spans 2008 2011 and 20092012. This file consists of the subset of data from the twelfth fifteenth and thirteenth sixteenth MEPS panels covering from January 1 through, roughly, the spring of calendar year 20082011. More specifically, data from the 2008 2011 portion of the third Round of data collection for the MEPS Panel 12 15 sample are pooled with data from the first Round of data collection for the MEPS Panel 13 16 sample (see illustration below). Panel 12 200715 2010-2008 Xxx Xxx 2011 2010 Jan 2011 Jan Round 1 Round 2 Round 3 Panel 00 0000-0000 Traditionally, the 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. The 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 samplessamples with MEPS Panel 11 reflecting the old design. Since then both MEPS panels have been based on the new design, serving 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 precision 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 decreased clustering since the sample was spread out among more PSUs and secondary sampling units50 percent subsampling rate) and increased degrees of freedom (due to many more variance strata available for variance estimation purposes)while those not sampled were dropped from MEPS entirely. The 2008 PIT file marks a return to having both MEPS Panels (12 and 13) being Thus, the Panel 16 Round 1 response rates discussed in the next section will reflect counts based on a single NHIS sample design, serving to reduce both precision (due to increased clustering) doubling the contribution of the subsampled “interim nonrespondents” and degrees 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 treating those “interim nonrespondents” not subsampled as if they were had never been sampled for the 2007 PIT for estimates where people living in the same general vicinity tend to have similar responses to questionnaire items and may be somewhat more unstable for estimates with relatively few degrees of freedom for 2008 data compared to that for 2007MEPS.

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 12 13 spans the two calendar years 2007 2008 and 2008 2009 while MEPS Panel 13 14 spans 2008 2009 and 20092010. This file consists of the subset of data from the twelfth thirteenth and thirteenth fourteenth MEPS panels covering from January 1 through, roughly, the spring of calendar year 20082009. More specifically, data from the 2008 2009 portion of the third Round of data collection for the MEPS Panel 12 13 sample are pooled with data from the first Round of data collection for the MEPS Panel 13 14 sample (see illustration below). Panel 12 200713 2008-2009 2008 Xxx Xxx Jan 2009 Jan Round 1 Round 2 Round 3 Panel 00 0000-0000 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. 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, 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). The As with the 2008 PIT file marks a return to having file, both MEPS Panels (12 13 and 1314) being in the 2009 PIT file are based on a single NHIS sample design, serving to reduce both precision (due to increased clustering) and degrees 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 same general vicinity tend to have similar responses to questionnaire items and may be somewhat more unstable for estimates with relatively few degrees of freedom for 2008 data compared to that for 2007.

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 12 16 spans the two calendar years 2007 2011 and 2008 2012 while MEPS Panel 13 17 spans 2008 2012 and 20092013. This year’s Point-in-Time file consists of contains the subset of data from respondents to the twelfth sixteenth and thirteenth seventeenth MEPS panels covering the time period from January 1 1, 2012 through, roughly, the spring of calendar year 20082012. More specificallyTo achieve this, data from representing the 2008 2012 portion of the third Round round of data collection for the MEPS Panel 12 16 sample are have been pooled with data from the first Round round of data collection for the MEPS Panel 13 17 sample (see illustration below). Panel 12 200716 2011-2008 Xxx Xxx 2012 2011 Jan 2012 Jan Round 1 Round 2 Round 3 Panel 00 0000-0000 Traditionally, the 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. The 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 samplessamples with MEPS Panel 11 reflecting the old design. Since then both MEPS panels have been based on the new design, serving including Panels 16 and 17, the MEPS panels found in this PIT file. There were two features of the MEPS sample design 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, 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 precision 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 decreased clustering the 50 percent subsampling rate) while those not sampled were dropped from MEPS entirely. Thus, the Panel 16 MEPS response rate information discussed in the next section 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 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). The 2008 PIT file marks a return to having both MEPS Panels (12 and 13) being based on a single NHIS sample design, serving to reduce both precision (due to increased clustering) and degrees 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 same general vicinity “partials” tend to have similar responses higher costs in gaining survey participation, but increased sample variability due to questionnaire items and may be somewhat more unstable for estimates with relatively few degrees of freedom for 2008 data compared to that for 2007the 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 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 12 14 spans the two calendar years 2007 2009 and 2008 2010 while MEPS Panel 13 15 spans 2008 2010 and 20092011. This file consists of the subset of data from the twelfth fourteenth and thirteenth fifteenth MEPS panels covering from January 1 through, roughly, the spring of calendar year 20082010. More specifically, data from the 2008 2010 portion of the third Round of data collection for the MEPS Panel 12 14 sample are pooled with data from the first Round of data collection for the MEPS Panel 13 15 sample (see illustration below). Jan Panel 12 200715 2010-2008 Xxx Xxx Round 1 Round 2 Round 3 2011 Jan Panel 00 000014 2009-0000 Traditionally, the 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. The 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, serving to increase precision (due to decreased clustering since samples with MEPS Panel 11 reflecting 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)old design. The 2008 PIT file marks a return to having Since then both MEPS Panels (12 and 13) being panels have been based on a single NHIS sample the new design, serving to reduce both precision including panels (due to increased clustering14 and 15) and degrees 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 same general vicinity tend to have similar responses to questionnaire items and may be somewhat more unstable for estimates with relatively few degrees of freedom for 2008 data compared to that for 2007PIT file.

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 12 18 spans the two calendar years 2007 2013 and 2008 2014 while MEPS Panel 13 19 spans 2008 2014 and 20092015. This year’s point-in-time file consists of contains the subset of data from respondents to the twelfth eighteenth and thirteenth nineteenth MEPS panels covering the time period from January 1 1, 2014 through, roughly, the spring of calendar year 20082014. More specificallyTo achieve this, data from representing the 2008 2014 portion of the third Round round of data collection for the MEPS Panel 12 18 sample are have been pooled with data from the first Round round of data collection for the MEPS Panel 13 19 sample (see illustration below). Panel 12 200718 2013-2008 Xxx Xxx 2014 2013 Jan 2014 Jan Round 1 Round 2 Round 3 Panel 00 0000-0000 Traditionally, the 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. The 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 samplessamples with MEPS Panel 11 reflecting the old design. Since then both MEPS panels have been based on the new design, serving including Panels 18 and 19, the MEPS panels found in this PIT file. For both Panels 18 and 19, 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 increase precision (due the NHIS; and those characterized as “partial completes”. “NHIS partial completes” typically have a lower response rate to decreased clustering MEPS and were sampled at a lower rate than “full completes” for both panels. This approach served to reduce survey costs, 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). The 2008 PIT file marks a return to having both MEPS Panels (12 and 13) being based on a single NHIS sample design, serving to reduce both precision (due to increased clustering) and degrees 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 same general vicinity “partials” tend to have similar responses higher costs in gaining survey participation, but increased sample variability due to questionnaire items and may be somewhat more unstable for estimates with relatively few degrees of freedom for 2008 data compared to that for 2007the resulting increased variance in sampling rates.