The Community Health Survey is an annual, cross-sectional, neighborhood-stratified, random-digit–dial telephone survey that the New York City Department of Health and Mental Hygiene conducts (22). It is based on the BRFSS survey of the Centers for Disease Control and Prevention (CDC). Using a computer-assisted telephone interviewing system, the CHS randomly samples approximately 10,000 noninstitutionalized adults aged 18 years or older to obtain citywide and neighborhood-level estimates of a number of health behaviors, health care access indicators, and health conditions. Neighborhood designations are determined by a zip code–based classification system developed and used by the United Hospital Fund (23). When contact is made with a household, one adult is selected randomly to complete an interview. Interviews are conducted in the interviewee's native language. In the 2002 CHS, which was conducted from May 2002 to July 2002, 9674 interviews were completed, representing a cooperation rate of 64% (percentage of contacted adults who agreed to participate) (24). For the 2004 CHS, conducted from May 2004 to February 2005, 9585 people were interviewed, with a cooperation rate of 59%. A comparison of these samples to the U.S. Census 2000 population of New York City adults is presented in Table 1. The table shows that the CHS is representative of the adult population of New York City. Our analysis was restricted to adults who identified themselves as non-Hispanic white, non-Hispanic black, or Hispanic because the number of adults of Asian and other ethnicities was small and because the BMI guidelines for obesity may not be an appropriate measure of obesity among Asians (25). The final sample sizes used in these analyses were 8943 in 2002 and 8571 in 2004.

The BRFSS is a cross-sectional, random-digit–dial telephone survey that is stratified by state or territory. Local health departments conduct the survey, in which data from all states and territories are pooled, in collaboration with CDC (26,27). In the 2002 BRFSS the median cooperation rate across all states was 74.3% (28). For the 2004 BRFSS the median cooperation rate across all states was 76.7% (29). After restricting the data sets to include only non-Hispanic whites, non-Hispanic blacks, and Hispanics, the final 2002 BRFSS sample size was 229,848 and the 2004 BRFSS sample size was 286,738.

In both the CHS and BRFSS for 2002 and 2004, self-reported height and weight were assessed by asking, "About how tall are you without shoes?" and "About how much do you weigh without shoes?"  Obesity status was determined using the BMI, calculated by weight in kilograms divided by height in meters squared, and classified according to World Health Organization, National Institutes of Health, and CDC guidelines (BMI <18.5 kg/m² for underweight, BMI 18.5–24.9 kg/m² for normal weight, BMI 25–29.9 kg/m² for overweight, and BMI ≥30 kg/m² for obese) (30,31). Obesity is further categorized into three classes: BMI from 30 to 34.9 kg/m² is defined as Class I obesity, BMI from 35 to 39.9 kg/m² as Class II obesity, and BMI of 40 kg/m² or greater as Class III, or severe, obesity (2,31). Because of small numbers (n = 1001 in 2002 and n = 689 in 2004) of adults with a BMI less than 18.5 kg/m², the two lowest categories were collapsed into one (underweight/normal weight). In both the CHS and BRFSS, diabetes was assessed by asking "Have you ever been told by a doctor that you have diabetes?"  Women who reported having had gestational diabetes exclusively were not considered to have diabetes.

We obtained New York City neighborhood income levels from U.S. Census 2000 data; income level was defined as the percentage of each neighborhood living below 200% of the federal poverty level and was stratified into categories of high, medium, and low. Neighborhoods in which 45% to 90% of the population lived at or below 200% of the federal poverty level were considered low income, and neighborhoods in which less than 30% of the population lived at or below 200% of the federal poverty level were defined as middle- or high-income neighborhoods. Adults were defined as foreign born if they reported a birthplace outside of the United States, Puerto Rico, or other U.S. territories except in the Hispanic subgroup analysis, for which Hispanics were assessed as either U.S.-born, foreign-born, or born in Puerto Rico. The Hispanic subgroup analysis was performed because New York City Hispanic subgroups may be different from subgroups in other parts of the United States because of different birth and migration patterns.

Each record in the CHS was assigned a primary weight for the probability of selection (i.e., number of adults in each household, number of residential telephone lines) and a post-stratification weight in order to adjust the sample estimates to the composition of each neighborhood in age, race/ethnicity, and sex (32). Similarly, each record in the BRFSS was assigned both a primary weight and a post-stratification weight in order to adjust the sample estimates to the composition of the states and country in age, race/ethnicity, and sex (33). We calculated the BMI distribution and the prevalence of obesity and diabetes in New York City and the United States in 2002 and 2004, stratified by demographic subgroup. Then, we examined the change in prevalence of obesity and diabetes from 2002 to 2004 in order to assess whether either had increased over time. Finally, we compared trends in New York City and the United States. To compare the prevalence of obesity and diabetes across years and demographic subgroups in the CHS and BRFSS, t tests were calculated; BMI quartiles in New York City in 2002 and 2004 were compared using 95% confidence intervals. In addition, logistic regression was performed to assess the independence of the within-year effects of the demographic variables on obesity and diabetes in New York City. Models were built in a forward stepwise manner, and variables were retained in the model if they were statistically significant at α = .05. National estimates were compared with those of New York City using 95% confidence intervals. SAS and SAS-callable SUDAAN 9 (Research Triangle Institute, Research Triangle Park, North Carolina) were used to perform these analyses.

The age-adjusted prevalence of obesity in New York City increased from 19.5% in 2002 to 22.8% in 2004 (P < .001), representing an additional 173,500 obese adults or a 17.0% increase in 2 years (Table 2). Nationally, the overall prevalence of obesity increased by 6% during the same time period (from 21.3% to 22.7%, P < .001) (Table 2). The 2004 increase in obesity prevalence among adult New Yorkers was the result of a shift in the entire BMI distribution to higher values (Figure 1), yielding a smaller proportion of adults in the underweight/normal weight category (P < .001) and a larger proportion in the obese category (P < .001), compared with 2002. The proportion of adults in the overweight category did not change significantly from 2002 to 2004 (from 36.1% to 35.8%, P = .78); however, the total population that was either overweight or obese increased from 55.6% to 58.7% (P < .001). Although the upward shifts of the 25th and 50th (median) percentiles of BMI were not significant (P = .05), the larger shift of the 75th percentile (by .6 BMI-unit) was significant at P < .05.

During the 2 years under study, the prevalence of obesity rose in all sociodemographic groups, but increases were statistically significant among only certain subgroups. Significant increases occurred in both sex and nativity subgroups, both U.S.-born and foreign-born, as well as among adults aged 25 to 44 years and adults aged 65 or older. Obesity levels also increased significantly among whites (14.3%–17.1%, P < .01) and among Hispanics (22.9%–26.2%, P <.05), but not among blacks (25.7%–29.0%, P = .052). When we examined the prevalence of obesity among subgroups of Hispanics, we observed significant increases among both U.S.-born (28.0%–35.8%, P < .05) and foreign-born (16.3%–23.1%, P < .05), but not among adults born in Puerto Rico (28.4%–29.7%, P = .78), despite overall high levels in the Hispanic group. At the national level, there were also significant increases in obesity in male and female subgroups (P < .05). In contrast to the CHS data, national obesity rates increased among all age groups except among those aged 65 or older. The largest increase in obesity occurred in those aged 18 to 24 years, for whom obesity increased by 17% (12%–14%, P < .05). Nationally, the only racial/ethnic group with a significant increase in obesity was whites (P < .01). Because the subgroups of Hispanics who live in New York City may be different from the subgroups that live in the United States overall, we performed a multivariate analysis of Hispanic New Yorkers to determine if the levels of obesity among subgroups held when accounting for other sociodemographic variables. In this analysis, the independent effect of the Hispanic subgroup was consistent with previous results (i.e., the association between Puerto Rican- and U.S.-born Hispanics and obesity held). In fact, all sociodemographic variables had a significant independent effect on obesity (P < .05). Because nativity data are not collected by the national BRFSS, a national comparison was not possible.

Figure 2 illustrates the cumulative percentage of adult New Yorkers at or below each BMI point on the basis of pooled BMI distributions by race/ethnicity for 2002 and 2004. The distribution illustrates that the weight distribution among whites was concentrated at lower BMI levels compared with the distribution among blacks and Hispanics; the median BMI among white adults was 25.1 kg/m², significantly lower than among Hispanics (26.4 kg/m², P < .05) and blacks (26.6 kg/m², P < .05). In addition, whites were more likely than blacks and Hispanics to be in the underweight/normal weight category and less likely to be in the overweight and obese categories (P < .05).

In New York City, the prevalence of diabetes increased by 17% from 2002 to 2004 (8.1%–9.5%, P < .01) (Table 3), with an estimated 73,600 more adults reporting that they were diagnosed with diabetes in 2004. Diabetes increased across all racial and ethnic groups. Significant increases in diabetes occurred among several other demographic subgroups: prevalence increased by 24% among men (8.1%–10.0%, P < .05), by 23% among adults aged 65 and older (19.0%–23.4%, P < .05), and by 24% among whites (5.1%–6.3%, P < .05). Nationally, there was no statistically significant increase in the overall prevalence of diabetes from 2002 to 2004 (6.9%–7.1%, P = .48), although there was a significant increase in the prevalence of diabetes among men (7.2%–7.7%, P < .05) (Table 3). In New York City, the prevalence of diabetes increased significantly among individuals who were underweight/normal weight (3.7%–5.2%, P < .05) (Table 4), although nonsignificant increases were observed across the other BMI categories (overweight and obese).

Limitations of this analysis include those related to self-reported data. Specifically, because data from the CHS are self-reported, estimates of obesity are likely to be low; people typically overstate their height and understate their weight (35). Similarly, our estimates of diabetes are likely to be low because not all adults with diabetes will recall their diabetes status during an interview and because diabetes is often undiagnosed; about 30% of adults with diabetes do not know they have it (36). However, because the questions were identical in 2002 and 2004 and because these samples are highly comparable, underreporting is not expected to have varied between years and would thus not affect our analysis. Additional limitations of the study include its cross-sectional design, which limits our ability to assess temporality or track incident conditions. Institutionalized adults and those without telephones were not represented in the sampling frame, limiting the generalizability of our findings. In addition, bias may have been introduced as a result of perceived pressure to provide socially desirable answers; however, the anonymous nature of the survey may have limited this effect (22,37). Because local and national data were collected with the same survey method, these limitations should not affect our comparisons. The large difference in the size of the BRFSS and CHS samples is an additional limitation because larger population samples, such as the BRFSS sample, are more likely to yield statistically significant results than smaller samples, such as the CHS. Therefore, all else being equal, the BRFSS would yield more statistically significant differences than the CHS. However, there were actually more statistically significant differences between years in New York City than in the United States. Finally, this trend analysis is limited to only 2 years of data. Additional research examining data from the CHS and BRFSS should be ongoing in the future to assess longer-term trends. On the other hand, the smaller CHS sample means less precise estimation of trends compared to BRFSS. Strengths of the CHS include representativeness because it is conducted in multiple languages and its data characterizes the adult New York City population.

From 2002 to 2004, an additional 173,500 adult New Yorkers became obese, and an additional 73,600 were diagnosed with diabetes. Increases in obesity and diabetes were largest among some of the most rapidly growing subgroups in New York City and the United States, suggesting that the health impact and burden to the health care system related to these conditions may accelerate in coming years. Differences in obesity and diabetes between New York City and the United States underscore the need for local data. Understanding trends is important for local programming and policy making. Without immediate action, both New York City and the United States as a whole will experience increasingly urgent and damaging epidemics.