The Registry's methods have been described previously.^4–5 Briefly, rescue/recovery workers and volunteers; lower Manhattan area residents, workers, and school attendees and staff; and commuters or passersby on 9/11 were recruited through building or employee lists or encouraged to enroll via a website or toll‐free telephone number. Between September 12, 2003, and November 24, 2004, 71 434 persons completed a telephone (95%) or in‐person (5%) enrollment questionnaire on exposures occurring on and after 9/11, sociodemographic factors, and health status. All participants gave verbal informed consent to participate in the Registry. The US Centers for Disease Control and Prevention and the New York City Department of Health and Mental Hygiene institutional review boards approved the Registry protocol.

The New York State Department of Health's SPARCS program is an administrative reporting system that records data, including discharge diagnoses and dates, for all patients discharged from acute care hospitals in New York State, excluding psychiatric and federal hospitals. SPARCS data have been used for health research in the past, including studies of CVD hospitalizations.^6–9

Personal identifying data for Registry enrollees were electronically linked to SPARCS records from January 1, 2000, through December 31, 2010. Records that matched on parts of key identifying information, such as name, date of birth, Social Security number, or address, were considered matches. Deaths were identified through data linkages to New York City vital records (through December 31, 2010) and the National Death Index (through December 31, 2008), as described previously.¹

We defined cardiovascular hospitalizations based on International Classification of Disease, 9th Revision (ICD‐9) codes for the principal discharge diagnosis. CVD hospitalizations were categorized as either heart disease or cerebrovascular disease hospitalizations. Heart disease hospitalizations included those for hypertension (ICD codes 401 to 405), acute coronary artery disease (410, 411), chronic coronary artery disease (412 to 414), dysrhythmia (427), and congestive heart failure (428). We also studied a more narrowly defined subset of these outcomes, coronary artery disease hospitalizations (ICD codes 410 to 414). Cerebrovascular disease was defined as ICD codes 430 to 438.

Because objective measures of 9/11‐related exposures are not available for most persons exposed to the disaster, we defined exposure based on responses to the enrollment questionnaire, as summarized in Table 1. Dust cloud exposure was defined as answering “yes” to having been caught in the dust and debris cloud on 9/11. Injury was defined as reporting any of the following: eye injury or irritation; cut, abrasion, or puncture wound; sprain or strain; burn; broken or dislocated bone; or concussion or head injury. Rescue/recovery workers were asked questions about arrival time, location, type, and duration of work. Area residents and workers were asked about evacuation of their homes and workplaces. Because enrollees were queried about many different but overlapping exposures, we summarized the overall level of 9/11‐related exposure as high, intermediate, or low, separately for rescue/recovery and nonrescue/recovery enrollees, as described previously.¹

Study participants were hierarchically categorized as rescue/recovery workers (including volunteers) or nonrescue/recovery workers (including lower Manhattan area residents or workers and passersby or commuters). A history of self‐reported, physician‐diagnosed hypertension or diabetes was obtained from the enrollment questionnaire. Posttraumatic stress disorder (PTSD) was assessed with the PTSD Checklist (PCL), a validated, 17‐item scale that inquired about 9/11‐specific psychological symptoms within the 30 days preceding the enrollment questionnaire.^10–12 We used a score ≥44¹⁰ to define probable PTSD, which we subsequently refer to as PTSD for simplicity. If an enrollee was missing responses to PCL items but had a score from the remaining items that could only be compatible with a total score of less than or ≥44, he or she was categorized accordingly. Otherwise, PTSD was considered missing.

The observation period began at Registry enrollment and ended at the first cerebrovascular or heart disease hospitalization, death, or December 31, 2010, whichever was earliest. The analysis was limited to enrollees who lived in New York State on Registry enrollment (54 650, 76.5%) because SPARCS reflects only New York State hospitalizations. Participants who reported a history of stroke or heart disease before Registry enrollment (n=4443) or with a preenrollment cardiovascular or cerebrovascular disease hospitalization (n=338) were excluded. We also excluded enrollees who withdrew from the Registry (n=470), who were <18 years old on 9/11 (n=2740) or had missing age data (n=242), or were lower Manhattan area school students or staff but did not belong to other eligibility groups (n=71), due to small numbers. Enrollees with a first cerebrovascular disease hospitalization during the study period were excluded from analyses of heart disease hospitalizations, and vice versa. We stratified the analysis on sex due to well‐established differences in cardiovascular risk profiles for men and women.

Proportional hazards regression models were used to examine associations of 9/11‐related exposures and PTSD with heart disease and cerebrovascular hospitalizations accounting for factors that were theoretically and statistically associated with the outcomes in bivariate analysis: age, race/ethnicity, education, smoking, history of hypertension, and history of diabetes. We evaluated linear trends in the hazards ratios by including each exposure as a continuous variable in a multivariable analysis and testing whether its coefficient differed significantly from zero. The low, intermediate, and high levels were assigned values of 1, 2, and 3, respectively. The assumption of proportional hazards over time was tested for all models presented by using time‐dependent variables to assess interactions between each predictor of interest and a function of survival time and by using the Kolmogorov–Smirnov test.

We considered P‐values <0.05 or CIs that did not include 1 to be statistically significant. Analyses used SAS version 9.2 (SAS Institute Inc).

Table 3 shows multivariable adjusted hazard ratios (AHRs) for relationships between 9/11‐related exposures and PTSD and heart disease hospitalization. Accounting for other factors, women with PTSD at enrollment had an elevated risk of heart disease hospitalization (AHR 1.32, 95% CI 1.01 to 1.71). Among men, this relationship did not reach statistical significance (AHR 1.16, 95% CI 0.97 to 1.40). Male rescue/recovery workers with a high overall level of 9/11‐related exposure were at an 82% higher risk of heart disease hospitalization compared with those with a low level of exposure (P for trend 0.05). The hazard ratios for female rescue/recovery workers with high or intermediate levels of exposure were also elevated, but this was not statistically significant (P for trend 0.09).

We repeated this analysis restricting the outcome to coronary artery disease hospitalizations (Table S1). The hazard ratios for associations between the various predictor variables and coronary artery disease hospitalization were similar to those for the more inclusively defined heart disease hospitalizations.

An analysis of associations between 9/11‐related exposures and PTSD and cerebrovascular hospitalizations is shown in Table 4. AHRs for cerebrovascular disease hospitalization were elevated among both women and men with PTSD at enrollment, although the relationship was only statistically significant among men. None of the other exposures examined were significantly associated with cerebrovascular hospitalization.