Aging of the Population

1. Population Forecasts
2. Implications of an Aging Population for the Demand for Health Workers
1. Increasing Demand for Health Care Services
2. Increasing Demand for Health Workers
3. Implications of an Aging Population for the Supply of Health Workers
1. Physician Supply
2. Nurse Supply
4. Implications of an Aging Population for the Economics of the Health Care System

Major Findings:

If health care consumption patterns and physician productivity remained constant over time, the aging population would increase the demand for physicians per thousand population from 2.8 in 2000 to 3.1 in 2020. Demand for full-time-equivalent RNs per thousand population would increase from 7 to 7.5 during this same period. In 2000, physicians spent an estimated 32 percent of patient care hours providing services to the age 65 and older population. If current consumption patterns continue, this percentage could increase to 39 percent by 2020. The aging of the health workforce raises concerns that many health professionals will retire about the same time that demand for their services is increasing. Also, the elderly population will grow at a faster rate than the working-age population. The rise in health care expenditures associated with the rapid increase in the elderly population will likely place pressures on the Medicaid and Medicare programs to control health care costs. Such measures would likely decrease the demand for and supply of health professionals.

Increased longevity and the aging of the baby boom generation will contribute to a substantial increase in the size of the elderly population during the next few decades as well as the aging of the overall population. Four major implications of an aging population on the health workforce are the following.

One, because the elderly have both greater and different health care needs than the non-elderly, the rapid growth in size of the elderly population could substantially increase overall demand for health care services and consequently the derived demand for health workers. Occupations and settings that disproportionately serve the elderly will experience the largest growth. If health care consumption patterns and physician productivity remained constant over time, the aging population would increase the demand for physicians per thousand population from 2.8 in 2000 to 3.1 in 2020. Demand for full-time-equivalent (FTE) RNs per thousand population would increase from 7 to 7.5 during this same period.

Two, physicians will spend an increasing proportion of their time treating the elderly. Our analysis of multiple health care use databases suggests that in 2000 physicians spent an estimated 32 percent of total patient care hours providing services to the age 65 and older population. If current patterns continue, this percentage could increase to 39 percent by 2020.

Three, the health workforce is aging along with the general population. As health professionals in the baby boom generation retire and as the pool of potential entrants to the health workforce (i.e., the population age 18 to 30) declines as a percentage of the total population, there is concern that the future supply of health professionals will be inadequate to meet demand.

Four, the expected increase in health care expenditures attributed to the growing elderly population will likely place pressures on the Medicaid and Medicare programs to control health care costs. The ratio of working-to-retired Americans will likely decrease, placing budget pressures on other government programs that compete with funding for Medicaid and Medicare. Economic pressures to curb the growth in health care costs could result in policies to reduce the demand for and supply of health workers.

2.1 Population Forecasts

Census Bureau population projections show significant shifts in the age distribution (Exhibit 2.1) with the number of elderly increasing in absolute size and as a proportion of the total population (Exhibit 2.2). The number of elderly, defined as the "age 65 and over" population, will grow by over 50 percent between 2000 and 2020, and by an estimated 127 percent by 2050. Furthermore, the relative size of the elderly population is projected to increase from 12.6 percent of the population in 2000 to an estimated 16.5 percent in 2020. Between 2030 and 2050, one in five Americans will be elderly.

The most rapidly growing demographic group among age categories is the "oldest elderly." This group is sometimes defined differently by researchers, but the most common definitions are the population age 75 and over, age 80 and over, and age 85 and over. ^[3] Two factors that contribute to researchers using different age breaks to define the oldest elderly are (1) differences in use of health care services, and (2) small sample size among the oldest elderly when using survey data.

In 2000, there were approximately 16.6 million people age 75 and over, 9.2 million people age 80 and over, and 4.2 million people age 85 and over. By 2020, the number of people in these age groups could reach 22 million, 13 million, and 7 million, respectively.

Exhibit 2.1. Age Distribution of U.S. Population

Source: Analysis of the 1996 HCUP database with an adjustment so that rates applied to the population in 1996 equaled total inpatient days reported by the American Hospital Association (AHA). See Dall and Hogan (2002).

Exhibit 2.1. Age Distribution of U.S. Population (Text Only)

Source: U. S. Census Bureau middle series population projections (Day, 1996).

Exhibit 2.2. Projections of U.S. Elderly Population

2.2 Implications of an Aging Population for the Demand for Health workers

2.2.1 Increasing Demand for Health Care Services

The greater medical needs of the elderly, combined with access to health care services through Medicare and Medicaid, have resulted in much higher per capita use of health care services for the elderly compared to the non-elderly. On a per capita basis, the elderly have more hospital inpatient days, outpatient visits, and emergency department visits. Relative to the non-elderly, they also have more home health visits per capita and are more likely to be in a long-term care facility.

To illustrate these points, consider Exhibits 2.3 through 2.8 that contain estimates of per capita health care use by age, sex, and urban or rural location for six health care settings modeled in the NDM. The most profound differences in per capita utilization exist across age groups; however, there are also important differences in per capita utilization by sex and by urban or rural location. Many of the following estimates are for 1996, the base year in the NDM, although more recent data are available for some settings.

An analysis of the 1996 Health Cost Utilization Project (HCUP) database finds that with the exception of the age 0-4 population, the number of inpatient days in general, short-term hospitals per 1,000 population increases substantially with age for both men and women, in both rural and urban areas (Exhibit 2.3). Analyses of other patient-level databases such as the National Hospital Ambulatory Medical Care Survey (NHAMCS), the National Home and Hospice Care Survey (NHHCS), and the National Nursing Home Survey (NNHS) produced estimates of per capita health care utilization in different settings for the eight age groups used in the NDM, by sex, and by urban or rural location. These are shown in Exhibits 2.4 through 2.8.

Exhibit 2.3. Inpatient Days in General, Short-term Hospitals (per 1,000 population)

Source: Analysis of the 1996 HCUP database with an adjustment so that rates applied to the population in 1996 equaled total inpatient days reported by the American Hospital Association (AHA). See Dall and Hogan (2002).

Exhibit 2.4. Outpatient Visits in General, Short-term Hospitals (per 1,000 population)

Source: Analysis of the 1996 NHAMCS database with an adjustment so that rates applied to the population in 1996 equaled total non-emergency, outpatient visits reported by the AHA. See Dall and Hogan (2002).

Exhibit 2.5. Emergency Department Visits in General, Short-term Hospitals (per 1,000 population)

Source: Analysis of the 1996 NHAMCS database with an adjustment so that rates applied to the population in 1996 equaled total emergency visits reported by the AHA. See Dall and Hogan (2002).

Exhibit 2.6. Inpatient Days in Non-General and Long-term Hospitals (per 1,000 population)

Source: Analysis of the 1996 HCUP database with an adjustment so that rates applied to the population in 1996 equaled total inpatient days reported by the AHA. See Dall and Hogan (2002).

Exhibit 2.7. Home Health Visits (per 1,000 population)

Source: Analysis of the 1995 NHHCS database with an adjustment so that rates applied to the population in 1998 equaled estimates of total home health visits paid for by Medicare, Medicaid and other sources in 1998. See Dall and Hogan (2002).

Exhibit 2.8. Nursing Home Residents (Residents per 1,000 population)

Source: Analysis of the 1997 National Nursing Home Survey (NNHS). See Dall and Hogan (2002).

Not only does per capita use of health care services within a delivery setting increase with age, but also the type of services used by the elderly (and the mix of health professionals who provide these services) differs from those of the non-elderly. To capture these differences in type of services received, the PARM uses physician-patient encounters in hospital inpatient and outpatient settings, in non-hospital office settings, and in other settings (e.g., nursing homes and home health) to forecast future demand for physician services by medical specialty. ^[4] The nature of a physician-patient encounter, as well as the length of the encounter, can vary substantially by medical specialty and delivery setting. Physician surveys, reported in the annual AMA publication Physician Socioeconomic Statistics, reveal that physicians typically spend more time per encounter with patients in hospital-based visits versus office visits that are not hospital-based. Encounters that involve surgical procedures often last two to five times longer, on average, than visits that do not involve surgical procedures. Consequently, the PARM forecasts demand for each physician specialty by health care setting, and the hospital inpatient setting is subdivided by whether or not a surgical procedure was performed. Even within a specialty, the types of services demanded might differ by age. For example, eye diseases such as cataracts and glaucoma are much more prevalent in the older population (White et al., 2000). Consequently, as the population ages, optometrists will likely see a shift in the type of services provided.

An important question for modeling requirements for physicians and other health workers is whether these caregivers spend different amounts of time per encounter with the elderly relative to the non-elderly. Two databases used to update the PARM-the National Ambulatory Medical Care Survey (NAMCS) and the National Hospital Ambulatory Care Survey (NHAMCS) Outpatient File-contain information on the amount of time physicians spent with patients during each encounter. To increase sample size, we combined the 1997, 1998, and 1999 NAMCS, and we combined the 1997, 1998, and 1999 NHAMCS. We tested the hypothesis that patient demographic characteristics and insurance status are determinants of the amount of time physicians spend per visit with patients in doctors' offices and hospital outpatient settings. We tested this hypothesis by estimating a series of regressions, using the ordinary least squares (OLS) criterion, with length of time as the dependent variable and dummy variables that indicate patient characteristics and insurance status as the exogenous variables. The dummy variables take on the value of 1 if the patient has that characteristic, and take on the value of 0 if the patient does not have that characteristic. We estimated separate regressions for each medical specialty.

The regression results showed each of the exogenous variables (age, sex, race/ethnicity, and insurance status) to have a significant impact on the dependent variable (time per encounter) for some specialties but not for others. Even when statistically significant, the impact was in many cases quite small, less than two minutes per encounter. One caution when interpreting the regression results is that the R-squared statistic for every regression is extremely low, indicating that the exogenous variables in the model explain only a small proportion of the overall variation in length of time physicians spend with patients. The relatively large residual variance makes it more difficult to find a statistically significant relationship. Also, for some specialties the number of patients in a particular demographic group is small which reduces the precision of the estimates for those demographic groups.

Exhibit 2.9 contains the regression results for encounters in doctors' offices, and Exhibit 2.10 contains the results for encounters in hospital outpatient settings. The column labeled AVG reports the average minutes per encounter for the reference group (non-Hispanic, white females age 55-64, insured in a fee-for-service arrangement). The other columns represent the marginal impact of the demographic characteristic or insurance status on minutes of physician time per encounter. Shaded boxes indicate marginal impacts, relative to the reference category, that are statistically different from zero at the 0.05 level of significance.

To illustrate, consider the first specialty: general and family practitioners. The average time spent with the reference group is 18.36 minutes per encounter in doctors' offices (Exhibit 2.9). Time spent with men is just 6 seconds shorter than time spent with women, on average, after controlling for age, race/ethnicity, and insurance status. General and family practitioners spend, on average, 2.43 fewer minutes per encounter with patients age 0-17 and 1.08 fewer minutes per encounter with patients age 18-34 compared to the reference group of patients age 55-64. Both of these differences in average minutes per encounter are statistically different from zero at the 0.05 level of significance. General and family practitioners also spend 0.91 fewer minutes per encounter with African Americans and 0.53 fewer minutes per encounter with other minorities, relative to non-Hispanic whites, although only the estimate for African Americans is statistically different from zero. Time spent with patients in a health maintenance organization (HMO) is 0.81 minutes less than time spent with patients insured in a fee-for-service arrangement, while the time spent with uninsured patients is 0.74 minutes greater than that spent with patients covered under fee-for-service. Neither of these differences is large, however, and of the two, only the former is statistically different from zero.

With respect to the other specialties shown in Exhibit 2.9, major regression effects noted are as follows:

Sex. - Only orthopedic surgery and other surgical specialties show statistically significant differences for men and women. The time per encounter is in both cases greater for men than it is for women: an additional 0.66 minutes, on average, for orthopedic surgery, an additional 3.86 minutes for other surgical specialties.

Age. - Of the sixteen specialties shown, ten display significant age effects with respect to at least one age group. General and family practitioners, for example, spend significantly fewer minutes per encounter with patients under 35; internal medicine (IM) subspecialists spend significantly fewer minutes per encounter with patients over 74; etc. Most of these effects, however, although statistically significant, are no more than a minute or two, with the following exceptions: physicians in other medical specialties spend over three minutes more per encounter with children under 18 while physicians in other surgical specialties spend almost seven minutes less per encounter with patients in that age group.

Race/ethnicity. - Significant race/ethnicity effects are evident for ten of the specialties shown. African Americans spend significantly fewer minutes per encounter with physicians in four specialties (general and family practice, internal medicine subspecialties, cardiovascular disease, and other patient care) and significantly more minutes per encounter with ob/gyn's. Patients in the "other" minority category spend significantly fewer minutes per encounter with physicians in three specialties (general internal medicine, pediatrics, and psychiatry) and significantly more minutes per encounter with physicians in another three (other medical specialties, emergency medicine, and other patient care). The added 14.51 minutes per encounter for "other" minority patients seen by emergency medicine physicians is particularly noteworthy.

Insurance status. - A marked insurance effect is also evident. HMO patients spend significantly fewer minutes per encounter with physicians in four specialties (general and family practice, pediatrics, orthopedic surgery, and other patient care) and significantly more minutes per encounter with physicians in four other specialties (IM subspecialties, cardiovascular disease, other surgical specialties, and psychiatry). Of these differences, only those for other surgical specialties (plus 3.82 minutes) and other patient care (minus 2.61) exceed 2 minutes. Somewhat surprisingly, there are no specialties for which uninsured patients receive fewer minutes per encounter, on average, than the reference group, whereas there are six specialties for which they receive more minutes on average. Those six are pediatrics, other medical specialties, general surgery,