Methodologies and Resources

Notes on Methods

For the projections two methods were implemented within the R Statistical Language. First a Holt-Winters projection (Holt, Charles C. (1957). "Forecasting Trends and Seasonal by Exponentially Weighted Averages". Office of Naval Research Memorandum 52. reprinted in Holt, Charles C. (January–March 2004). "Forecasting Seasonals and Trends by Exponentially Weighted Averages"International Journal of Forecasting 20(1): 5–10.) was produced. Secondly a Exponential smoothing state space model (Hyndman et al, International Journal of Forecasting (2002), 18(3), 439-454) was produced. The resulting output included the graphical reflection of the trend with confidence intervals. The damped technique was applied to better reflect the uncertainty of the trend at both 10 and especially 20 years into the future. The ultimate projection value was obtained from these methods and reported as either a number of people or a population per 100,000 value. Each of the following disease burden projections was produced using either these methods or a crude percentage transfer in the event of uncertain or unavailable trend data.

Access to Care

Preventable Hospitalzations

Data from 2001 to 2009 for Preventable Hospitalzations was aggregated for both County and HSA by gender and age group for specific disease breakdowns. These data were then consumed into the R Statistical program. Once in R both Holt-Winters and ETS time series analyses were conducted at the 95% and 80% confidence intervals. Each condition produced estimates by age/gender and these data were then aggregated and normalized to population per 100,000. Aggregating the output from these predicted conditions provided predicted populations by age/gender discharges for ambulatory care sensitive condition discharges projected in 2020 and 2030.

Emergency Department Discharges for Mental Health

Data from 2001 to 2009 for ED Discharges was aggregated for both County and HSA by gender and age group for specific disease breakdowns. These data were then consumed into the R Statistical program. Once in R both Holt-Winters and ETS time series analyses were conducted at the 95% and 80% confidence intervals. Estimates were produced by age group/gender and these data were then normalized to population per 100,000. The resulting output provided predicted populations by age/gender discharges for emergency department discharges projected in 2020 and 2030. 

Primary Care Demand 

These projections were created by JSI/Community Health Institute, and involved creating a matrix of primary care utilization rates, stratified by specified age and gender groupings, based on observed primary care utilization patterns.  The analysis focused on a 'Barrier Free' sub-sample extracted from 2 years (2009-2010) of the Agency for Healthcare Research and Quality’s (AHRQ’s) Medical Expenditure Panel Survey (MEPS), comprised of those individuals most likely to have unfettered access to the health care system.

This data is combined with an analysis of the New Hampshire Behavioral Risk Factor Surveillance Survey (BRFSS) covering the same timeframe (2009-2010), to adjust the results to reflect health status in the state within each age/gender group.

The results were then applied to population projections within the age/gender strata for years 2010, 2020, and 2030.

Click here for the full projections methodology document. 

For more information:  Healthcare Resources and Services Administration (HRSA)  

Health Behaviors

Obesity 

Using BRFSS data from 2002 to 2010 for Obesity, a state level trend was produced for the percent of the state’s population exhibiting each condition. This time series was then applied to both a Holt-Winters and ets function within R to produce a projected high and low for the state in 2020 and 2030. County level projections were also produced, though the numerical values were deemed too unstable to rely on the value - hence only the comparison to the state value is presented.

Tobacco Use 

Using BRFSS data from 2002 to 2010 for Tobacco Use, a state level trend was produced for the percent of the state’s population exhibiting each condition. This time series was then applied to both a Holt-Winters and ets function within R to produce a projected high and low for the state in 2020 and 2030. County level projections were also produced, though the numerical values were deemed too unstable to rely on the value - hence only the comparison to the state value is presented.

Cost of Care

Chronic Disease Cost

Baseline data from calendar year 2011 for commercial data, and a three-year average (2009-2011) for Medicaid data were used to construct a Per Member per Year cost by age/gender for each county and HSA. Concurrently, the percent of the population by age/gender making up the current year cost was calculated. These percentages of potential users were applied to the 2020 and 2030 data. At that point per member per year cost values were multiplied by the projected number of users to produce a projected cost in 2011 dollars for the years 2020 and 2030.

Health Outcomes

Cancer Incidence

Preliminary data was obtained from NH State Cancer Registry aggregated for the years 2001-2011. The ten year totals for each age group and geography were then calculated for the ten year average period. The cancer rate per age group over the period was then applied forward to the 2020 and 2030 projected populations to produce a 2010-2020 Cancer value and 2020-2030 ten year cancer average.

For more information about cancer incidence in New Hampshire, visit NH Department of Health and Human Services.

Alzheimer’s Disease

Estimates of Alzheimer’s prevalence percentages were obtained from the publication “National estimates of the prevalence of Alzheimer’s disease in the United States” published in Alzheimer Dement. Jan 2011; 7(1):61-73. These figures provide prevalence estimates for men and women age 70-79, 80-89 and 90+. These age categories were then re-produced in the projected populations for 2020 and 2030 and the resulting Alzheimer’s population numbers were calculated using the current year ratio applied to the projected population.

For further information: Alzheimer’s Association

Diabetes 

Using BRFSS data from 2002 to 2010 for Diabetes, a state level trend was produced for the percent of the state’s population exhibiting each condition. This time series was then applied to both a Holt-Winters and ets function within R to produce a projected high and low for the state in 2020 and 2030. County level projections were also produced, though the numerical values were deemed too unstable to rely on the value - hence only the comparison to the state value is presented.

Cardiovascular Disease 

Using BRFSS data from 2002 to 2010 for Cardiovascular Disease, a state level trend was produced for the percent of the state’s population exhibiting each condition. This time series was then applied to both a Holt-Winters and ets function within R to produce a projected high and low for the state in 2020 and 2030. County level projections were also produced, though the numerical values were deemed too unstable to rely on the value - hence only the comparison to the state value is presented.

For more information on Health Outcomes and Behaviors:

County Health Rankings

NH State Health Improvement Plan

Demographics

Population

  • 2010: U.S. Census Bureau
  • Projections: NH Office of Energy and Planning
  • Health Service Area level projections were produced by the Applied Population Laboratory at the University of Wisconsin. In developing the projections methodology the goal was to produce the most robust age-sex projections for New Hampshire's Hospital Service Areas (HSAs). Using a model that best fit recent age-sex patterns—1990 through 2010—the demographer disaggregated the NHOffice of Energy and Planning's county-level age-sex projections for 2010-2040 to the minor civil division (MCD) level to generate MCD age-sex projections. Then, these MCD-level age-sex projections were aggregated to HSAs. While checks were run on the total projected MCD populations for reasonableness with recent Census results, the small size of certain age-sex cells at the MCD level (not to mention the small population size of some MCDs) means that greater unreliability exists in the projections at the MCD level. Thus, caution is warranted in using the MCD age-sex projections for low-population municipalities.

For more information:

Poverty

Poverty percentages obtained from the 2000 and 2010 US Census were used to produce a small area demographic projections of poverty in 2020 and 2030. Poverty was determined as the percentage of the population living below 200% of the poverty level at the time of the Census tabulation. In 2000, these data were obtained from the Long Form Census or SF3 file. In 2010 these data were obtained from the 5-year American Community Survey Data Sample spanning 2008-2012, centered on 2010. The 2000 variable used was P088 Ratio of Income to Poverty Level in 1999 , the 2010 variable used was C017002 Ratio of Income to Poverty Level for Families. Poverty projections were performed by the Applied Population Laboratory at the University of Wisconsin, Madison. 

For more information:

Population Age Ratio

Also referred to as Age Dependency Ratio, this is the ratio of dependents (under 20 and over 64) to the working age population. The data is shown as a percentage; for example, if the ratio displayed is 120% it means that there are 120 dependents per 100 working age population. Some analyses separate the under 20 and over 64 age groups to show the distinct ratio for each; however MapNH Health has displayed the overall ratio. While this ratio is a demographically based measure, it is used in this project to highlight the regions in which there is or will potentially be a disproportionate number of dependents to the working age population. 

For more information: 

Your browser is out-of-date

Update your browser to view this website correctly. Update my browser now

×