Methods

Top Methods Results Discussion Author & Article Info References

We examined deaths among adults who had contact with the Boston Health Care for the Homeless Program between 1 July 1988 and 31 December 1993. Patients were homeless when first seen by primary care providers at shelters and clinics. The amount of time each patient was homeless during the subsequent observation period was unknown. Thus, persons in the cohort can be described as persons who had been homeless at any time and who had contact with a clinician while they were homeless.

Deaths were ascertained by comparing the program's patient database with the Massachusetts death registry for 1988 to 1993. Matching criteria adapted from the National Death Index [15] required agreement on 1) first name or first initial, 2) last name, 3) month of birth, and 4) day of birth or year of birth ± 1 year. Names were compared by the Soundex algorithm to allow for errors in spelling. Matches were reviewed manually and were confirmed if records agreed on social security number or full name and date of birth. The International Classification of Diseases, Ninth Revision (ICD-9) codes that indicated cause of death were obtained from death certificates.

The observation period for each person lasted from the time of first contact until 31 December 1993 or the date of death. Person-years of observation and cause-specific mortality rates were calculated by age group, sex, and race. Mortality rates were adjusted for race by direct standardization using Boston residents as the standard population. Rate ratios were calculated by dividing the race-adjusted mortality rate in the homeless cohort by the corresponding mortality rate in the general population of Boston. Because available sources of data were limited, homeless persons aged 18 to 24 years were compared with Boston residents aged 15 to 24 years.

Deaths were categorized by the season of the year and the week of the month in which they occurred. The chi-square goodness-of-fit test was used to compare the number of deaths seen during each period with the expected number of deaths if a constant mortality rate was assumed.

Results

Top Methods Results Discussion Author & Article Info References

The study cohort included 17 292 adults, of whom 2365, 3577, 3026, 2687, 2688, and 2949 persons were seen for the first time in 1988, 1989, 1990, 1991, 1992, and 1993, respectively. The cohort was observed for 50 348 person-years (average, 2.9 years per person). Characteristics of the cohort and the 606 decedents in the cohort are shown in Table 1. More deaths were seen in later years of the study because homeless persons joined the cohort over time; therefore, the total number of persons in the study increased with each passing year.

For persons aged 18 to 64 years, the crude mortality rate was 1114 per 100 000 person-years. The average age at death was 47 years (median, 44 years [range, 18 to 86 years]). Death most commonly occurred in a hospital or residential dwelling. Cause-specific mortality rates and rate ratios are shown in Table 2 and Table 3. Because few adults in the homeless cohort were elderly, mortality rates for persons older than 64 years of age were omitted.

The acquired immunodeficiency syndrome (AIDS) was the leading cause of death among persons who were 25 to 44 years of age. In this age group, AIDS-related mortality rates per 100 000 person-years were 481.9 in black men, 331.4 in white men, 232.4 in black women, and 65.6 in white women. In 18% of the cohort, AIDS was the cause of death; HIV infection was documented on the death certificate in an additional 7% of the cohort.

Homicide was a leading cause of death in persons who were 18 to 24 years of age and in women who were 25 to 44 years of age. Traumatic injury and poisoning were the second most common causes of death in men 18 to 24 years of age and 25 to 44 years of age. Poisoning caused by an overdose of drugs, most often opiates, accounted for 6% of the deaths in the cohort.

Heart disease was a major cause of death in homeless persons 45 to 64 years of age. For men 25 to 44 years of age, the rate of death from heart disease was more than threefold higher than in the general population. About 50% of deaths caused by heart disease were attributed to coronary artery disease or myocardial infarction; the remainder were caused by various other cardiac illnesses. Certain conditions that are generally associated with homelessness were not common causes of death; exposure to cold caused four deaths, and tuberculosis caused only one death.

Death was most likely to occur during the first week of each month. In the first, second, third, fourth, and fifth weeks, 162, 154, 131, 120, and 39 deaths were seen, respectively (chi-square goodness-of-fit test, P = 0.05). Because the fifth week of each month is truncated, 39 deaths during that week is equivalent to 113 deaths during a full week. The number of deaths caused by injuries was highest during the first week of the month. The number of deaths did not vary significantly with the seasons.

Discussion

Top Methods Results Discussion Author & Article Info References

Death occurred at an average age of 47 years in this cohort of homeless adults in Boston. Homicide was a leading cause of death in persons who were 18 to 24 years of age, and AIDS caused the most deaths in persons who were 25 to 44 years of age. Heart disease and cancer were the major causes of death in persons who were 45 to 64 years of age. The crude mortality rate was similar to the rate of 1035 per 100 000 person-years noted in a study of homeless persons aged 15 to 74 years in Philadelphia [14]. In contrast with earlier studies [11-14], however, our study shows the enormous effect of the AIDS epidemic on the homeless population.

Our study has some limitations. Only homeless persons who had contact with a health care program were included; mortality rates in the general homeless population may be higher or lower than our estimate because homeless adults who avoid contact with clinicians may have long-neglected medical conditions or may be healthier than average. In addition, although persons in our cohort were homeless when first seen, they may have ceased to be homeless during the observation period. The fact that 19% of deaths occurred in a residential dwelling suggests that this did occur. Thus, our findings apply to persons who died and were homeless at any time and may not be generalizable to persons who are continuously homeless. Deaths outside of Massachusetts were not identified. To the extent that such deaths were overlooked, our results underestimate the true mortality rate in the cohort.

Causes of death were obtained from death certificates. The accuracy of the data acquired from death certificates has been questioned because major discrepancies have been found between the information on death certificates and that in autopsy reports [16]. However, death certificates have been shown to reliably document deaths caused by coronary heart disease, AIDS, and traumatic injury [17-19]. Although data from death certificates should be interpreted with caution, they remain a valuable epidemiologic tool.

Our findings have serious implications for clinicians and policy makers. Deaths caused by AIDS are a major concern. Because a previous study [20] found that homeless and nonhomeless persons in Boston have similar survival rates after the diagnosis of AIDS, efforts to reduce the rate of AIDS-related deaths in homeless persons should probably focus on encouraging early treatment and preventing HIV infection.

The high risk for death from homicide and accidental injury is a predictable result of poverty, substance abuse, and living on the streets. Increasing the availability of adequate low-income housing could conceivably reduce this risk. Improving alcohol and drug treatment programs, however, may be a more important way to reduce injuries in the homeless population and might also help to decrease the high mortality rates that are attributed to cirrhosis and drug overdose. The increased number of deaths during the first week of each month may indicate that the arrival of disability checks at the beginning of the month leads to a flurry of substance abuse, injury, and death. If so, mortality in homeless persons might be reduced by carefully monitored payee programs that administer funds on behalf of disabled persons.

Pneumonia and influenza were frequent causes of death, even in younger age groups. Homeless persons may be at increased risk for these infections because of a high prevalence of alcoholism, smoking, HIV infection, and chronic disease. Efforts to vaccinate all homeless persons against pneumonia and influenza should be considered.

In conclusion, high mortality rates caused by treatable or preventable conditions were seen among homeless adults in Boston. Efforts to reduce mortality rates among homeless persons should focus on these conditions. The treatment of underlying problems, such as substance abuse, must also be addressed.

Dr. Orav: Section of Clinical Epidemiology, Brigham and Women's Hospital, 75 Francis Street, Boston, MA 02115.

Drs. O'Connell and Lebow: Boston Health Care for the Homeless Program, 729 Massachusetts Avenue, Boston, MA 02118.

Dr. Brennan: Department of Health Policy and Management, Harvard School of Public Health, 677 Huntington Avenue, Boston, MA 02115.