Contents

Appendix I: Methods. 14-1

A Note about Race-specific Statistics. 14-1

Definition and Determination of Statistical Significance. 14-1

Health Planning Areas. 14-2

2005 Marion County Adult Obesity Needs Assessment Survey. 14-2

The Behavioral Risk Factor Surveillance System (BRFSS)14-2

The American Community Survey. 14-3

IHHA Hospital Discharge Data. 14-4

National Hospital Discharge Survey (NHDS)14-5

Marion County Mortality Rates. 14-6

Years of Potential Life Lost14-6

Big Cities Health Inventory. 14-7

Appendix I: Methods

A Note about Race-specific Statistics

Several data sources classify people by Hispanic status and, separately, by race.  In this report, Hispanic ethnicity is generally presented as a race category.  To convert datasets that coded both Hispanic ethnicity and race into the race classification used in this report, all persons of Hispanic ethnicity were classified as Hispanic, and only persons not of Hispanic ethnicity were classified as being of another race.

In the 2000 U.S. census, people could indicate that they were multiracial, choosing more than one race in response to the census’s race question.  In previous years, people were limited to indicating only one race on the census questionnaire.  Many statistics in this report use denominators from census data, and numerators from other sources that attribute only one race to a person.  To resolve the difference in how race was classified in the different data sources, multiracial persons were considered to be equal parts of each race they indicated.  They were counted as contributing equal portions of a person to each race group which they indicated, with those portions summing to one.  This approach avoids the misleading results that can arise if multiracial people are omitted from the denominator of such statistics.

Definition and Determination of Statistical Significance

In this report, the reader will sometimes see the term “statistically significant difference” when two groups or time periods are being compared.  A statistically significant difference is a difference that is likely to be due to an underlying difference between the groups being compared, rather than being due to chance or random fluctuations.  In this report, if a difference is called “statistically significant,” a difference in the same direction would be expected at least 95 out of 100 times were the population to be repeatedly re-sampled and compared.  In other words, this report used p < 0.05.

Summary statistics and their confidence intervals were often available without the detailed data from which the statistics was drawn.  In these cases, if an assessment of statistical significance was made, it was based on comparisons of statistics’ 95 percent confidence intervals (CIs).  This approach is very conservative.  Non-overlapping confidence intervals often indicate p < 0.01 if the confidence intervals are of similar width.[564]

Note that, where statistics’ 95 percent confidence intervals overlapped and the underlying data were not available, we could not determine whether or not the statistics were statistically significantly different at p < 0.05, unless one statistic’s confidence interval included the value of the other statistic.  When one statistic’s 95 percent confidence interval did include the value of the other statistic to be compared, it could be concluded that the statistics did not differ at p < 0.05, and the reader may see text saying that the difference in the statistics was “not statistically significant”.

Health Planning Areas

Health Planning Areas were developed by the Marion County Health Department with feedback from neighborhood associations to divide up Marion County into smaller, demographically cohesive areas of about 40,000 to 50,000 inhabitants, to better show the distribution of health statistics. The areas are based upon aggregations of census tracts.  Tracts were grouped based on their demographic similarities, and based on input from neighborhood liaisons from the city government, so the areas might be sensible and useful to community groups.  The 40,000 to 50,000 population range was selected because it provides an adequate denominator for fairly stable, reliable estimates of most health statistics.  These areas aid the Marion County Health Department in understanding the health status and health needs in the county.  When feasible, the health department uses these areas in reporting health statistics.    It is our hope that having sub-county level statistics based upon these areas will help target high-risk and high-prevalence communities for various diseases and to further the interest in health improvement by residents in each area.

2005 Marion County Adult Obesity Needs Assessment Survey

From February through June 2005, a stratified random sample telephone survey was conducted among 4,784 respondents by the Center for Survey Research at IUPUI.  Respondents were Marion County residents 18 years old or older.  The questionnaire and survey protocol was designed by the Marion County Health Department and the Indiana University Bowen Research Center.  In addition to a general random sample, Hispanics, Hispanic males, African Americans, and African American males were over-sampled to increase the precision of subpopulation-specific analyses.  Survey topics included self-reported height and weight, physical activity opportunities and habits, eating habits, and health status.  Responses were weighted in accord with the sampling design, and also weighted to match the gender, race, and age distribution of Marion County per the 2005 mid-year population estimate from the U.S. Census Bureau.  Analyses were performed using SAS® software procedures for analysis of data from stratified surveys.  More details regarding the questionnaire and survey protocol are available at http://www.mchd.com/obesitysurvey.htm.

To increase the comparability of the Adult Obesity survey results with those of other surveys, many questions in the questionnaire were taken from standard instruments, including the CDC’s Behavioral Risk Factor Surveillance System questionnaire and the National Health Interview Survey.  Results of different surveys may differ due to differences in the questionnaires or in how the survey was conducted.  The impact of such differences is difficult to determine.  Comparisons between the Marion County Adult Obesity Needs Assessment results and those of other surveys should be interpreted with this potential source of extraneous differences kept in mind.

The Behavioral Risk Factor Surveillance System (BRFSS)

The BRFSS is a cross-sectional telephone survey conducted by all 50 states’ health departments with technical assistance from the Centers for Disease Control and Prevention (CDC).[565] Each year states conduct monthly telephone surveys with standardized questionnaires to estimate the distribution of risk behaviors and health practices among community-living adults aged 18 and older. The surveys collect information on health behaviors, preventive health practices, health care access and activities primarily related to preventable infectious, chronic diseases, and injury in the population, and to identify emerging health problems. Only one adult is interviewed per household. BRFSS data are weighted for the characteristics of the sample design, which includes disproportionate sampling by geographic and density strata (where they exist), number of phones, and number of adults in the household.[566]  A final post-stratification adjustment is made for non-response rates and non-coverage of households without telephones. The weights for each relevant factor are multiplied together to get a final weight. Indiana’s 2005 BRFSS sample included 5219 completed interviews. In recent years, because there were over 500 BRFSS respondents from Marion County, and at least 19 sample members in each of the weighting classes based on age, sex, and in some states, the CDC has calculated county-level statistics for some items.  However, the sample sizes for population sub-groups have been too small to allow age-group or ethnic-group stratified analyses.

More detail on the CDCP’s Behavioral Risk Factor Surveillance Survey methods and theIndiana Statewide Survey Data, 2005 may be found at: http://www.in.gov/isdh/reports/brfss/2005/index.htm.

The American Community Survey

The U.S. Census Bureau conducts the American Community Survey (ACS).  The 2005 survey sample was interviewed throughout 2005. The ACS employs a two-stage, two-phase sample design. The ACS first-stage sample consists of two separate samples, Main and Supplemental, each chosen at different points in time. Together, these constitute the first-stage sample. Both the Main and the Supplemental samples are chosen in two phases referred to as first- and second-phase sampling. Subsequent to second-phase sampling, sample addresses are randomly assigned to one of the twelve months of the sample year. The sampling frame for the ACS is created from the Master Address File (MAF), which is a database maintained by the Census Bureau containing a listing of residential and commercial addresses in the U.S. The data are limited to the household population and exclude the population living in institutions, college dormitories, and other group quarters.

Estimates from the ACS are obtained from a ratio estimation procedure that results in the assignment of weights to each sample person record and to each sample housing unit record. The characteristics considered in the forming the weights include:

·        Percent in poverty

·        Percent renting

·        Percent in rural areas

·        Race, ethnicity, age, and sex distribution

·        Distance between the centroids of the counties

·        Core-based Statistical Area status

 

The ACS and PRCS employ three modes of data collection:

·        Mailout/Mailback

·        Computer Assisted Telephone Interview (CATI)

·        Computer Assisted Personal Interview (CAPI)

 

Data are based on a sample and are subject to sampling variability. The degree of uncertainty for estimates from sampling variability is represented through a margin of error, which in the ACS data is a 90 percent margin of error (also known as a 90% confidence interval). Thus a result can be interpreted having a 90 percent probability that the estimate plus or minus it’s margin of error (or the lower and upper confidence bounds) contains the true value. For more information about the accuracy of the estimates, see Accuracy of the Data.[567]  More information regarding the survey methods is available at http://www.census.gov/acs/www/SBasics/index.htm.

IHHA Hospital Discharge Data

Hospital discharge data for 2000 to 2005 from all community acute care facilities in Marion County have been compiled by the Indiana Hospital and Health Association (IHHA) and include all inpatient stays and procedures occurring in these facilities. Discharge diagnoses are coded using the International Classification of Diseases, 9th Revision, Clinical Modification (ICD–9–CM).[568] Unless otherwise noted, all IHHA Hospital Discharge Data analyses in this report are based on the principal diagnosis (if so identified) or first-listed diagnosis in a patient’s medical record. The data does not include individual identifiers, but does include demographic information such as age, gender, and race.  Note that, because the dataset does not identify individuals, statistics based on the data are counts of hospital visits rather than of patients.  One patient may have multiple visits. The proportion of discharges for diagnoses that often involve multiple hospital admissions, such as cancer or other chronic conditions, may overestimate the proportion of persons hospitalized for that condition, relative to more transient conditions such as injuries.

Like the national hospital discharge data (below) the Marion County statistics only reflect discharges from short-stay hospitals, and not long-term institutions, or federal facilities such as Veterans Administration hospitals.

Discharges of newly born infants were excluded from the analysis of the hospital discharge data, so that total included discharges would be more specific to hospitalizations for health problems.

National Hospital Discharge Survey (NHDS)

The National Hospital Discharge Survey (NHDS) is conducted by the National Center for Health Statistics of the U.S. Centers for Disease Control and Prevention.  The NHDS collects data from a sample of inpatient records acquired from a complex national sample of hospitals.  In 2004, data were collected for approximately 371,000 discharges.  Estimates are for discharges, as individuals may have multiple discharges in a year. The sample includes hospitals with an average length of stay under 30 days (‘short-stay’ hospitals), general and children’s hospitals. Federal, military, veteran’s hospitals, and long-stay institutions such as prison and state hospitals are excluded. Of the 476 eligible hospitals, 439 responded to the 2004 survey.[569]

Data are collected by abstraction of hospital records by the hospital or NCHS trained staff, or as machine-readable medical records purchased from commercial organizations. In 2004, approximately 44 percent of respondent hospitals provided data through the purchased data system. Data include discharged patient’s age, sex, race, ethnicity, marital status, ZIP code, and expected sources of payment, admission and discharge dates, admission type and source, and discharge status. Medical information about patients includes up to seven diagnoses, as many as four surgical and non-surgical operations and procedures, and dates of surgery. Medical data, including discharge diagnoses and procedures, are coded according to the International Classification of Diseases, 9th Revision, Clinical Modification (ICD–9–CM). [570]^,[571]

Discharge diagnoses are reported by the first-listed diagnosis, or that specified as the principal diagnosis on the face sheet or discharge summary, usually the main cause of the hospitalization. The number of first-listed diagnoses is the same as the number of discharges. Procedures include surgical or non-surgical operations, diagnostic procedures, and special treatments reported on the medical record. Up to four procedures are coded for each discharge. Estimates for discharges and procedures do not include newborn infants in calculations of rates and counts.[572]

NDHS survey data are weighted to produce national estimates to produce unbiased national estimates, using probabilities of sample selection, non-response adjustments and population weighting adjustments[573]. The standard error, a measure of sampling variability, was calculated by CDC to account for the complex sample design.  Discharge and procedure rates are calculated using post-census estimates of the civilian population of the United States as of July 1, 2004.[574]

More details about the NDHS are available at http://www.cdc.gov/nchs/about/major/hdasd/nhds.htm. The NDHS methodology is available at http://www.cdc.gov/nchs/data/series/sr_01/sr01_039.pdf.

Marion County Mortality Rates

The infant mortality rate is expressed as the number of number of infant deaths in a population per 1,000 live births in that same population. Small changes in the numbers of infant deaths or live births, especially among minority populations or within small geographies, can greatly influence the rate.

Causes of death in this report are classified according to categories developed by the National Center for Health Statistics using the International Classification of Diseases, version 10 (ICD-10) coding scheme.[575]  All death rates (excluding infant mortality rates) in this report are per 100,000 persons, and are age-adjusted to the year 2000 U.S. standard population.[576]  Death rates for injury at work are age-adjusted using the U.S. 2000 standard population for the population aged 15 years and older, and are based on the number of deaths aged 15 years and older.

Population denominators for death rates are U.S. Census Bureau estimates of the population as of July 1 for the year(s) in which the deaths occurred. 

Years of Potential Life Lost

Years of Potential Life Lost (YPLL) approximates how many more years of life people might have lived if a certain cause of death were eliminated.  It emphasizes mortality due to causes of death that tend to be more predominant among younger persons, such as accidents, congenital anomalies, and AIDS. YPLL is calculated by subtracting the age at death of each decedent from an age representing the average life expectancy in the United States, and then summing all those differences for a total YPLL. If each decedent's age is not readily available for computation, an estimated YPLL can be computed using the midpoints of reported age groups.[577]  MCHD calculates YPLL using the age of individual decedents and a life expectancy of 75 years, the current standard used by NCHS.[578]

Big Cities Health Inventory

 Big Cities Health Inventory (BCHI) is a compilation of health status indicators produced in a comparative format for the 54 largest metropolitan areas in the United States.  The National Association of County and City Health Officials (NACCHO)  The purpose of the document is to focus specifically on the health of people living in large cities in the United States. In so doing, the report attempts to increase knowledge of the issues large cities face and stimulate dialogue that will lead to a healthier city population.  In addition to improving our understanding of the health in large cities, the data in the report serves as a reference point to monitor cities’ progress in reaching the nation’s Healthy People 2010 objectives.

 

	Age-adjusted deaths per 100,000 population, 2004				Infant Deaths per 1000 births, 2003	Smoking During Pregnan-cy, 2003	New cases per 100,000 persons, 2005
	Overall	Lung Cancer	Breast Cancer	Homi-cide			Gonor-rhea	Chla-mydia	Syphilis
Indianapolis rank  (1 would be worst)	15 of 54	6 of 54	52 of 54	26 of 49	17 of 53	3 of 44	7 of 43	10 of 43	32 of 32
Albuquerque, NM	976.2	48.9	29.6	8.2	4.5	11.0%	118.6	662.1	5
Arlington, TX	1004.5	72.4	26.4	---	6.2	5.6%	124.9	332.5	9
Atlanta, GA	974.3	51.5	29.4	21.9	.	3.0%	318.1	464.1	38
Austin, TX	826.5	51	27.4	4	6.2	3.6%	---	---	---
Baltimore, MD	1139.6	70.6	33.2	39.5	12.8	13.1%	534	979.8	29.9
Boston, MA	820.9	53.3	26.7	8.3	6.1	4.0%	143.8	641.4	10.9
Charlotte, NC	973	65.5	30.3	11.7	7.2	6.2%	367.6	591.7	14.6
Chicago, IL	868.1	54.1	29.9	14.5	10	6.2%	341.5	789.2	14.4
Cincinnati, OH	1072.9	75	36.4	15.8	10.3	15.6%	829.5	1386.7	---
Cleveland, OH	1131	81	38.3	15.9	10	17.3%	683.3	972.6	---
Colorado Springs, CO	911.7	51.8	36.1	---	7.2	9.5%	116.3	464.3	---
Columbus, OH	1051	83.7	28.1	12.3	9.8	15.7%	293.3	589.2	13.5
Dallas, TX	881.6	50.3	24.6	16.1	8.1	2.8%	264.3	444.3	10.9
Denver, CO	841.9	44.3	23.5	14.2	6.9	7.0%	277.3	889.2	6.3
Detroit, MI	1103.4	67	30.7	36.8	16.4	13.8%	---	---	---
El Paso, TX	854.4	33.8	29	---	3.9	1.9%	51.3	414.1	---
Fort Worth, TX	969.2	65.1	28.4	9.8	8.8	4.4%	336.3	585.6	10.8
Fresno, CA	1135	66.2	30.7	12	7	---	162.6	607.5	---
Honolulu, HI	669.9	37.4	14.9	---	8	4.1%	---	---	---
Houston, TX	997.9	59.4	29.9	14	6.9	2.3%	202.3	481.6	11.8
Indianapolis, IN	1009.1	77.6	20.1	12.7	9.4	17.6%	455.1	840	3.6
Jacksonville, FL	1061.2	73.6	30.6	14.3	10.2	9.6%	318.5	628.7	6.7
Kansas City, MO	912.6	68.7	27.2	17.5	8.3	13.0%	550.6	959.8	13.8
Las Vegas, NV	1660.3	106.9	53.3	19.9	3.7	8.3%	---	---	---
Long Beach, CA	853.5	47.2	25.3	12.5	7	---	---	---	---
Los Angeles, CA	806.5	40.1	28.4	13.5	5.5	---	109.9	382.9	10.9
Louisville, KY	942.7	76.4	29.2	9.8	6.4	18.2%	200	359.6	5.6
Memphis, TN	1045.7	62.7	35.3	17.6	17.3	6.9%	449.9	1040.9	19.5
Mesa, AZ	943.1	51.7	34.7	6	6	6.6%	81.7	394.6	---
Miami, FL	1288.4	55.7	41.8	22.5	5	0.7%	458.2	1073.7	46.3
Milwaukee, WI	965.4	65.1	22.4	13.4	11.4	12.4%	643.6	1565.9	---
Minneapolis, MN	799.5	54.9	19.1	10.5	5.4	6.2%	313.9	669.6	11
Nashville-Davidson, TN	899.7	66.5	22.9	9.6	7.4	9.2%	195.8	494.7	4.2
New Orleans, LA	1095.1	67.5	35.3	47.9	11.3	2.5%	---	---	---
New York, NY	748.9	38.6	26.8	6.7	6.2	1.9%	132.3	489.7	7.7
Oakland, CA	822.8	44	24	17.6	6.5	---	294.6	604.8	5.8
Oklahoma City, OK	936.2	62.7	24.8	10.2	10.3	15.1%	345	517.1	4.1
Omaha, NB	859.5	61.6	29.7	---	7.3	13.2%	193.3	711	---
Philadelphia, PA	1006.1	67.4	31.7	21.4	10.6	17.7%	333.2	1027.3	5.7
Phoenix, AZ	852.2	49.1	22.4	13.9	6.8	5.8%	152.5	509.4	5.9
Pittsburgh, PA	950.5	71.9	27.5	12.1	14.5	---	---	---	---
Portland, OR	902.7	59.6	23.1	4.7	3.9	11.1%	134.1	434.4	4.1
Sacramento, CA	1449.6	79.3	42.3	17.6	5.9	---	186.3	569.1	---
San Antonio, TX	983.5	49	27.3	8.2	7.5	3.2%	183.7	618.7	10.7
San Diego, CA	762	45.5	23.1	5.4	4.5	---	92.6	391	6.9
San Francisco, CA	699.3	43.4	24.5	8.6	3.1	---	310.7	477.4	32.1
San Jose, CA	708.9	41.8	21.5	3.5	4	---	63.9	343.5	4.6
Seattle, WA	729.5	46.8	23.3	4.4	5.6	---	---	---	---
St. Louis, MO	1013.2	75.5	28.8	25.4	12.7	16.0%	---	---	---
Tucson, AZ	1398.2	89.5	41.7	16	12.1	11.1%	---	---	---
Tulsa, OK	933.6	60.2	24	12.9	4.5	4.9%	---	---	---
Virginia Beach, VA	888.2	73.1	24.6	4.4	6.7	6.5%	111.7	373.9	---
Washington, DC	966.4	50.6	27.1	27.7	10.5	3.7%	375.1	643.3	19.9
Wichita, KS	950.5	71.9	31.2	8.7	6.3	14.7%	144.9	424.6	---