Changing Incidence and Survival for Heart Failure in a Well-Defined Older Population, 1970–1974 and 1990–1994
Background— An epidemic increase in heart failure (HF) mortality, hospitalization, and prevalence rates has been observed among older persons in recent years. It is unclear whether this reflects an increase in incidence or survival.
Methods and Results— We conducted a retrospective cohort study comparing HF in 1970 to 1974 and 1990 to 1994 among persons ≥65 years old belonging to a large, well-defined population with complete medical records available for research. Using Framingham clinical criteria, we identified incident cases of HF in the respective periods. Age-specific and age-adjusted incidence, mortality, and survival rates were compared. Cox proportional-hazards models were used to assess association of comorbidities and medications with survival. During 38 800 and 127 419 person-years for 1970 to 1974 and 1990 to 1994, respectively, 387 and 1555 confirmed incident cases were identified. When adjusted for age, incidence increased by 14% (95% CI 2% to 28%). Increased incidence tended to be greater for older persons and for men. Based on 5-year follow-up and adjustment for age and comorbidities, the mortality hazards decreased 33% (95% CI 14% to 48%) among men and 24% (95% CI −1% to 43%) among women.
Conclusions— The epidemic increase in HF among the older population between the 1970s and 1990s is associated with increased incidence and improved survival, with both of these effects being greater in men.
Received July 13, 2004; revision received November 15, 2005; accepted December 2, 2005.
Between the 1970s and early 1990s, heart failure (HF) mortality, hospitalization, and prevalence rates all increased dramatically in the United States.1 This “new epidemic” of HF occurred simultaneously with dramatic declines in coronary artery and cerebrovascular disease mortality.2–4 Importantly, increases in HF mortality and morbidity rates were confined to the population over 65 years of age, who account for more than 80% of deaths and prevalent cases.5 Similar trends have been reported in Europe.6
Several alternative explanations may account for the observed increase in HF, including population aging, changing diagnostic criteria, increasing incidence of HF, or increasing survival among incident cases treated with medications that reduce HF mortality.7–9 To gain insight into these competing explanations, a Special Emphasis Panel on “Cardiovascular Disease Community Surveillance,” convened in 1996 by the National Heart, Lung, and Blood Institute, identified the need for population studies of clinical data sets that could accurately ascertain, confirm, and follow up all incident HF cases over the period that the national increases were observed.10
The present study was conducted on a large, well-defined, older population in the state of Oregon, where increases in HF mortality and hospitalization rates parallel the national pattern.11 We specifically sought evidence for increased incidence and/or survival between the 1970s and 1990s while controlling for population aging and diagnostic criteria.
The study was based at the Kaiser Permanente Center for Health Research (KPCHR), which has conducted extensive population-based studies in conjunction with the Kaiser Permanente Northwest Region health plan, a large group model health maintenance organization with an enrollment of more than 300 000 persons. Health plan members are similar to the Portland, Ore, metropolitan population in terms of age, sex, race, and income characteristics and constituted 15% to 20% of that predominantly white population between 1965 and 1985.12 On reaching 65 years of age, ≈95% of members continue to receive care through the health plan, converting to Medicare as payer. The KPCHR has maintained access to comprehensive medical records for research purposes since 1966.12
We used a retrospective study design to compare 2 cohorts of older members, each of which was followed for up to 5 years. The cohorts consisted of all persons ≥65 years old who had been health plan members for 2 or more years at the beginning and at least 1 month during the following 5-year study periods: 1970 to 1974 (period 1) and 1990 to 1994 (period 2). Total persons and person-years of observation for the cohorts during the 2 five-year study periods increased over time as increasing numbers of members reached age 65: 9272 persons and 38 800 person-years in period 1 and 31 399 and 127 419 in period 2.
Case Finding and Confirmation
To identify and confirm incident HF cases that occurred in 1970 to 1974 and 1990 to 1994, computerized hospital discharge files were searched for any mention of International Classification of Diseases (ICD) codes for HF or HF-associated conditions among all listed diagnoses for hospitalizations that occurred within the respective 5-year study periods (see Appendix in the online-only Data Supplement). Beginning with the earliest such hospitalization, inpatient and outpatient records were reviewed for up to 2 years before and after to exclude prevalent HF cases. For confirmation, we used the Framingham Heart Study clinical criteria with minor modification.13 Date and clinical setting (inpatient or outpatient) of diagnosis were defined as that day and site of care at which these criteria were first documented.
Major criteria included paroxysmal nocturnal dyspnea or orthopnea, abnormal jugular vein distention, pulmonary rales, cardiomegaly, third heart sound, and pulmonary edema. Minor criteria included bilateral pedal edema, dyspnea on exertion, hepatomegaly, tachycardia (>120 bpm), pleural effusion, and engorged pulmonary vasculature. HF was defined as confirmed if 2 or more major criteria or 1 major and 2 or more minor criteria were documented.13 To adjudicate the diagnosis in the presence of acute myocardial infarction, renal failure, or chronic obstructive pulmonary disease, which may mimic clinical manifestations of HF, all such cases were reviewed by a cardiologist (WG).
Because our case-finding strategy would miss new cases in which the patients were outpatients and were not hospitalized during our study periods, we reviewed random samples of persons without an HF hospitalization for both study populations (2121 for 1970 to 1974 and 2777 for 1990 to 1994) to identify such cases and determine whether this occurred more frequently in either study period. This preliminary study yielded very small and essentially identical percentages of such “outpatient-only” cases, 0.25% and 0.22%, respectively, of these random samples.
Age, sex, date of HF confirmation, and vital status for a 5-year follow-up period were used in calculating incidence rates, case fatality rates, and survival patterns. Death dates were ascertained from medical records and from matching study subjects’ names against vital status records for Oregon and Washington State, which would capture essentially all deaths because subjects who moved to another geographic area, hence leaving the health plan, were censored at that time. Detailed inpatient and outpatient record reviews were conducted on all confirmed cases in period 1 and a 25% sample of confirmed cases in period 2 for up to 2 years before HF onset to identify clinical and therapeutic factors that might affect survival. These covariates included place of residence (home, nursing home, or other), medical setting at confirmation of diagnosis (outpatient or inpatient), selected comorbidities, and initial drug treatment. Trained technicians conducted the reviews using explicit written instructions (available from author on request).
Definition of Comorbidities
Myocardial infarction and atrial fibrillation were based on physician history or ECG report. Hypertension was defined as systolic pressure ≥140 mm Hg and/or diastolic pressure ≥90 mm Hg or documented drug treatment for hypertension. Valvular heart disease was defined as physician documentation of systolic murmur. Left ventricular hypertrophy was based on documentation in the cardiologist’s ECG report. Angina, cerebrovascular disease, chronic renal disease, arterial peripheral vascular disease, diabetes, chronic obstructive pulmonary disease (emphysema or chronic asthma), and cancer (excluding skin cancer) were considered present if documented in the physician history. For analytical purposes, cerebrovascular disease, peripheral vascular disease, and chronic renal disease were grouped together as “combined vascular.”
Classes of drugs ascertained comprised diuretics, digitalis, ACE inhibitors, (ACEI), β-blockers, and nitroglycerin. These were included if documented in physician orders up to 1 month from date of diagnosis.
Age- and sex-specific incidence rates were computed for 5-year age intervals with the number of confirmed new-onset cases as numerators divided by age- and sex-specific person-years of observation as denominators. Poisson regression models including period, age, sex, period-by-age, and period-by-sex were used to test for incidence rate differences between the 2 periods.
Age-adjusted and sex- and period-specific cumulative case fatality rates and their confidence intervals at 30 days, 1 year, and 5 years were computed with 5-year age intervals and 1990 to 1994 age distribution as reference. Period-specific survival functions were estimated with the Kaplan-Meier product-limit method, and log-rank statistics were used to test for period differences in survival. Cox proportional hazards regression models were used to test for period effect on mortality, controlling for exact age in years and comorbidities.14 Cox regression models were also used to assess the extent to which period effects were associated with therapeutic factors.
As shown in Figure 1, among 535 and 2352 persons with potential HF identified from hospital discharge records, incident HF occurring within the 5-year study periods was confirmed in 387 (74%) and 1555 (67%) persons, respectively. Those cases confirmed on a date before the study periods (prevalent cases) or that occurred within the period but did not meet Framingham criteria were excluded. Incident HF was initially diagnosed in the outpatient setting in 60 (16%) and 336 (22%) of cases in the respective cohorts. Among confirmed cases in 1970 to 1974 and 1990 to 1994, 86% and 87%, respectively, had 2 or more major Framingham criteria, and 88% and 91% had 2 or more minor criteria.
Table 1 shows person-years, incident cases, and incidence rates per 1000 person-years by sex and 5-year age intervals for the respective study periods, 1970 to 1974 and 1990 to 1994. Age- and sex-adjusted incidence increased from 10.0 (8.9 to 10.9) to 11.3 (10.7 to 11.9) per 1000 person-years. This increase occurred among men, from 11.7 (10.1 to 13.4) to 12.7 (11.8 to 13.7), and among women, from 8.6 (7.4 to 9.9) to 11.8 (11.0 to 12.6). The difference between genders in increase in incidence was not significant at P=0.05. Poisson regression analysis found a statistically significant main effect of period on HF incidence from 1970–1974 to 1990–1994, with a rate ratio of 1.14 (95% CI 1.02 to 1.28; P=0.021). Although period-by-age (P=0.27) and period-by-sex (P=0.22) interaction terms were not significant at the 5% level, there was a trend toward a larger period effect with increasing age and among men, as shown in Figure 2.
Case Fatality and Survival
Five-year follow-up of incident HF cases yielded totals of 253 and 1007 deaths for the 1970 to 1974 and 1990 to 1994 study periods respectively. (See the online-only Data Supplement Table.)
Table 2 lists characteristics present at time of HF onset that might have affected survival and compares the proportions of these present in 1970 to 1974 and 1990 to 1994. The proportion of case subjects diagnosed as outpatients increased similarly among both genders; presentation with acute pulmonary edema increased significantly among women; and chronic obstructive pulmonary disease and combined vascular disease increased significantly among both genders. The proportions with angina, myocardial infarction, atrial fibrillation, and hypertension changed little between periods. Digoxin use declined, whereas use of ACEIs, β-blockers, and nitroglycerin medications increased among both genders.
Age-adjusted cumulative mortality rates at 30 days, 1 year, and 5 years in 1990 to 1994 compared with 1970 to 1974 were consistently lower among men, whereas no apparent changes were noted among women.(Table 3) In both the 1970s and 1990s, at all follow-up intervals, mortality rates were higher among men than women. Five-year survival for men 65 years of age or above improved significantly between 1970 to 1974 and 1990 to 1994, whereas no change in survival was observed for women (Figure 3). Controlling for exact age in years, Cox regression analysis showed a significantly reduced hazard ratio for mortality in the 1990 to 1994 period among men (0.70, 95% CI 0.54 to 0.89) but not among women (1.01, 95% CI 0.77 to 1.3).
Predictors of Mortality
Cox regression models that incorporated age, location at diagnosis, and selected comorbidities and medications were conducted to elucidate factors associated with survival (Table 4). In addition to period and age (model 1), the second model controlled for nursing home residence, hospitalization at time of diagnosis, and selected clinical variables, all of which were found in univariate analysis to be associated with increased 5-year mortality among male or female cases. In this model, the hazard ratio decreased significantly by 33% (95% CI 14% to 48%) among men and showed borderline decrease of 24% (95% CI −1% to 43%) among women.
Model 3 assessed the extent to which period effects were associated with therapeutic factors. This analysis modified the 1990 to 1994 period hazard ratios for mortality for men and women from 0.67 to 0.77 and 0.76 to 0.86, respectively, which corresponds to 23% and 14% reductions in mortality. Although confounding between drug and period effects could not be controlled completely, our results suggest that these medications, particularly β-blockers, may have contributed to the observed decline in mortality in 1990 to 1994.
This retrospective study of a well-defined population provides insight into the epidemic increase in HF observed among older persons in the United States and elsewhere between the 1970s and 1990s. Specifically, comparing cohorts in 1970 to 1974 and 1990 to 1994, we found evidence of increased incidence in men and women, with trends toward larger period effects on incidence with increasing age and among men. Overall survival among men with HF improved significantly, and there was marginal evidence of improved survival among women when we controlled for comorbid conditions.
These findings may be readily compared with recent HF trend studies conducted in Framingham, Mass, and Rochester, Minn, both of which also used the widely recognized Framingham clinical criteria. The Framingham study reported declining age-adjusted incidence in women when comparing the broad periods 1950 to 1969 and 1970 to 1999; however, between 1970 and 1999, the period of national increase in HF, there was essentially no change in the age-adjusted incidence among women or men.15 In an earlier report from Framingham, age-specific incidence rates showed a tendency toward increased incidence among both men and women in older age groups, 70 to 79 and 80 to 89 years, but no trend among those 60 to 69 years old between the 1970s and 1980s, similar to trends we observed.16 The Rochester study reported no change in age-adjusted incidence among men and women between 1979 and 200017; however, in an earlier publication, age-specific data showed a pattern that suggested declining incidence in those aged 60 to 69 years and increasing incidence in those aged 70 to 79 years between 1981 and 1991, again similar to the present study.18 We suggest that age-adjusted rates in both the Framingham and Rochester studies may have obscured the observed countervailing patterns of increased HF incidence among persons over 70 years of age and the stable or decreased incidence among those under 70 years of age.
The pattern of increasing HF incidence selectively involving persons over 70 years of age observed in 3 separate population studies is consistent with national trends in hypertension and ischemic heart disease, the 2 leading causes of HF.19,20 On the one hand, control of hypertension increased significantly between the 1970s and 1990s, particularly among persons aged less than 70 years, thus potentially delaying its effect on the heart until later in life.21 Concurrently, the widespread introduction of lifesaving treatments for ischemic heart disease increased the pool of surviving older persons at high risk of developing HF.22
During the 1990s, the mortality rate declined significantly in both Framingham and Rochester.15,17 In both of these settings, age-adjusted mortality rates were consistently higher among men than women, which is very similar to our observations. Although authors of these studies hypothesized that the observed increased survival in the 1990s may be attributable to the increasing use of ACEI and β-blocker therapy, they did not provide data to explore this possibility.
We found statistical trends that suggested a modest association of ACEIs and a substantial association of β-blockers with improved survival. Our observed reduction in the mortality hazard associated with β-blockers is compatible with observations that a maladaptive increase in β-adrenergic cardiovascular activity constitutes one of the earliest potentially controllable pathophysiological manifestations in HF.9
Marginal improvement in survival among women in the 1990s was demonstrated when we controlled for the increasing prevalence of comorbidities, which were independently associated with an increased chance of mortality among women. The prognostic importance of comorbidities, specifically diabetes, chronic obstructive pulmonary disease, and cancer, has been observed in other studies of HF mortality.23–25
Our findings are based on a large, well-defined, older population that included incident cases identified from hospital records and confirmed with the widely utilized Framingham clinical criteria. The possibility that HF cases managed only as outpatients occurred more frequently in one study period, hence biasing the comparison of incidence rates, was refuted by our preliminary study, which showed equivalent, very small rates of such cases in both periods. Our confirmation rates of 74% and 67% are very similar to the 70% confirmation of suspected HF cases in the Cardiovascular Health Study of the population 65 years of age and older.26 The possibility that observed increases in incidence and survival resulted from systematic ascertainment of milder cases in the 1990s seems unlikely, given that such changes did not occur consistently among all age and gender subsets. Communitywide enrollment policies for the health plan in the Portland metropolitan area have remained consistent from 1960 to the present. Furthermore, because the Prospective Payment System introduced in the 1980s is not used for reimbursement in the Kaiser health plan, “DRG (diagnosis-related group) creep” would not have inflated the diagnosis of HF over time.27
Although these points support the validity of the present study, there are several limitations to the interpretation and generalizability of the findings. First, with respect to our incidence estimates, medical record review to detect HF with onset that occurred before the index hospitalization was limited to 2 years; hence, we would have misestimated incidence to the extent that there were incident cases in which HF onset occurred >2 years before the index episode. Second, echocardiography and other measures to identify HF with preserved systolic function were not available for the cases in the first study period and for approximately half of the cases in the second period; hence, we could not assess possible secular changes in this form of HF, which tends to be more common in older women and is an important predictor of HF survival and response to pharmocotherapy.28,29 Third, the precision with which comorbidities were ascertained for use in survival analyses was limited by our dependence on physician documentation in medical records. Although the presence of hypertension was based on conventional record review definitions, as used in the Framingham Study15 and others, and the presence of left ventricular hypertrophy and atrial fibrillation was based on cardiologist documentation in ECGs, the presence of other conditions, including angina, myocardial infarction, and valvular heart disease, was based on less precise, broadly inclusive operational definitions. Medical record details on diagnosis, duration, and severity of chronic renal disease were limited and variable; hence, we grouped all physician-diagnosed cases into one broad category. Although we used these comorbidity data to seek insight into mortality patterns in our regression models, we acknowledge that these findings must be interpreted with caution in light of the imprecise nature of the comorbidity data. Finally, because the present study population was elderly and drawn from 1 predominantly white metropolitan setting, the findings may not be generalizable to younger and nonwhite persons and to other settings.
The observed increase in incidence and survival for HF among a well-defined sample of the rapidly growing older population30 portends an accelerating rise in prevalence of this disabling, costly condition. To address this growing public health problem, in the short run, we need to strive for optimal provision of effective management for newly diagnosed HF to enhance survival and reduce hospitalizations,31 and in the long run, we need to gain better understanding of how to modify factors that contribute to its pathogenesis.
This study was funded by the National Heart, Lung, and Blood Institute (HL5880). The authors acknowledge the lead technical support roles in data management, collection, and analysis of Carol Sullivan, Jill Mesa, and Weiming Hu. The authors had full access to the data and take full responsibility for its integrity. All authors have read and agree to the manuscript as written.
The online-only Data Supplement can be found at http://circ.ahajournals.org/cgi/content/full/CIRCULATIONAHA.104.492033.
Thom TJ, Kannel WB. Congestive heart failure: epidemiology and cost of illness. Dis Manage Health Outcomes. 1997; I: 75–83.
National Heart, Lung, and Blood Institute. Data Fact Sheet: Congestive Heart Failure in the United States: A New Epidemic. Bethesda, Md: US Dept of Health and Human Services, National Institutes of Health, National Heart, Lung, and Blood Institute; 1996.
American Heart Association. 2000 Heart and Stroke Statistical Update. Dallas, Tex: American Heart Association; 2000.
Rich MW. Heart failure in the 21st century: a cardiogeriatric syndrome. J. Gerontol. 2001; 56A: M88–M96.
Armstrong PW, Moe GW. Medical advances in the treatment of congestive heart failure. Circulation. 1994; 88: 2941–2952.
Special Emphasis Panel on Cardiovascular Disease Community Surveillance. Bethesda, Md: National Heart, Lung, and Blood Institute; 1996.
Centers for Disease Control and Prevention, Compressed Mortality File, 1979–2000 [database online]. CDC Wonder. Atlanta, Ga; Centers for Disease Control and Prevention; 2001. Available at: http://wonder.cdc.gov/mortSQL.html. Accessed March 24, 2003.
Greenlick MR, Freeborn DK, Pope CR, eds. Health Care Research in an HMO. Baltimore, Md: John Hopkins University Press; 1988.
Cox DR, Oakes D. Analysis of Survival Data. London, UK: Chapman & Hall; 1984: 1–201.
Levy D, Larson M, Vasan R, Ho KK. No evidence of an increase in incidence of heart failure from 1950–1990. Circulation. 1996; 94: 1691. Abstract.
Yusuf S, Thom T, Abbott RD. Changes in hypertension treatment and in congestive heart failure mortality in the United States. Hypertension. 1989; 13(suppl 5): I-74–I-79.
Kannel WB, Ho K, Thom T. Changing epidemiologic features of cardiac failure. Br Heart J. 1994; 72: S3–S9.
Burt VL, Cutler JA, Higgins M, Horan MJ, Labarthe D, Whelton P, Brown C, Roccella EJ. Trends in the prevalence, awareness, treatment, and control of hypertension in the adult US population: data from the health examination surveys, 1960 to 1991. Hypertension. 1995; 26: 60–69.
Ho KK, Anderson KM, Kannel WB, Grossman W, Levy D. Survival after the onset of congestive heart failure in Framingham Heart Study subjects. Circulation. 1993; 88: 107–115.
MacIntyre K, Capewell S, Stewart S, Chalmers JWT, Boyd J, Finlayson A, Redpath A, Pell JP, McMurray JJV. Evidence of improving prognosis in heart failure: trends in case fatality in 66,547 patients hospitalized between 1986 and 1995. Circulation. 2000; 102: 1126–1131.
US Bureau of the Census. Current Population Reports, Special Studies, P23–178, Sixty-Five Plus in America. Washington, DC: US Government Printing Office; 1992.