A statistical profile of physical therapists, 1980 and 1990.

In the January 1995 issue of Physical Therapy, professional themes related to human resources were explored in a series of four articles.[1-4] In each of these articles, one conclusion reached was that a database of information about the profession was needed. Michels[5] reaffirmed this position in his letter to the editor responding to these articles. This report contributes to the goal of developing that database by providing a statistical profile of the locations and characteristics of physical therapists (PTs) at two points in time.

A great deal of information is available about the locations and characteristics of some health care professionals -- physicians, nurses, speech-language pathologists, and occupational therapists.[6-9] On the basis of this information, to various degrees these professions have addressed such issues as human resource requirements within the various medical specialties, distributional disparities, and the need for educational facilities and programs. As a profession, physical therapy has not taken these steps, in part because a comprehensive and unbiased database to examine these issues has not been available. This situation arose because there existed no nationally organized source of information on PTs. In 1985, the American Occupational Therapy Association was assisted in preparing its human resources planning document because, at that time, an estimated 99% of new graduates joined that professional organization.[8] In contrast, the American Physical Therapy Association (APTA) membership department estimates that fewer than 70% of eligible new PT graduates join that national professional organization (telephone conversation, November 1996). In the absence of a national database, we turned to a viable alternative: US census data.

Existing Databases

Every 3 years, APTA conducts a survey to develop a demographic profile of its membership.[10] Although these surveys provide valuable information about characteristics such as gender, age, race, and employment, the data are limited by their sampling frame -- the membership lists that serve as the basis for the sample. APTA is careful to remind readers that the survey results reflect only the characteristics of its membership. We believe that this bias in sampling affects most characteristics of interest -- economic status, age, race, nativity (classification as native or foreign born according to place of birth), and employment status. Also affecting the quality of the APTA data is a return rate for the mailed questionnaires of slightly less than 50%. A further limitation in the use of the APTA data is that the data are pooled nationally and no regional or state information is available.

Licensure data are available on a state-by-state basis from the 50 state boards of physical therapy. Plans currently are under way by the Federation of State Boards of Physical Therapy to aggregate state licensure data to the national level, but these data currently are only available through contact with each state board. In a survey of state boards that we conducted, only 13 boards had the capability to provide data on PTs before 1990. The data available were limited to a count of currently licensed PTs. Although some states collected information on the characteristics of their licensees, this information was informal and far from uniform. Furthermore, licensure counts are subject to an unknown degree of overlap. APTA, estimates that 7 of every 10 members are licensed in only one state but that the remaining 3 in 10 are licensed in two or more states.[10] On average, each PT holds 1.37 licenses. When Gwyer[4] developed estimates of the number of PTs from licensure data, the degree of overlap was estimated to be 25%. Renewal requirements vary from state to state, further complicating attempts to estimate the number of PTs from this source. Potential error because of this problem is generated by individuals who remain licensed in a previous state of residence after changing their place of residence or after leaving the profession because of retirement or death.

Counts of the number of full-time PTs and physical therapist assistants (PTAs) employed in hospitals are available from the American Hospital Association (AHA) annual survey of hospitals.[6] These counts cover less than half of all PTs, and their usefulness is restricted to examining trends in hospital-based employment.

During the 1970s, the US Department of Health and Human Services developed a database for 50 allied health occupations covering 13 states. Wilson et al[11] used these data as the source for estimates of the supply of PTs. Thus, the rates and supply estimates developed in this research were not representative of the entire United States. Recently, APTA commissioned Vector Research Inc to project the supply and demand for PTs through the year 2005. This study is notable for its efforts to clarify the human resources in physical therapy. The study is as yet unpublished.

Tabulations of the number of PTs are available from the 1990 US Census of Population and Housing.[12] No distinction is made in these counts among PTs and PTAs or physical therapy aides, and no information on additional characteristics is available from this source.

It is evident that each data source either is limited or contains bias. Corrections for the biases may be biased themselves because they are based on assumptions that cannot be validated from any other data source.

Method

Sample -- Public Use Database

Public Use Microdata Samples (PUMS) from the US censuses of population are available to researchers for most censuses dating back to the beginning of this century. Among the many items included in the PUMS are questions on occupation and industry. Beginning with the 1980 census and continuing with the 1990 census, PTs were identified as a separately coded occupational category. The PUMS are available for a 5% sample of the entire US population for both censuses.[13,14] These samples are unbiased 5% random samples not only of the US population but also of all subgroups within the population. Each case extracted from the data represents an average of 20 PTs. Census data are collected every decade. Since the data used in this study were collected, there have been changes in the health care environment and the number of PTs produced annually. The next census will undoubtedly reflect those changes. Nonetheless, these data provide a benchmark for the profession reaching back to 1980.

Many descriptive items are available from the PUMS, including gender, age, race, nativity, educational attainment, employment characteristics, and income. Additionally, the data may be examined by geographic distribution at the level of region, state, and metropolitan area. After selecting cases for labor force participation and licensure potential, we extracted a sample of 3,112 PTs from the 1990 PUMS. To lend historical depth by extending the study to a second point in time, we extracted a sample of 1,530 PTs from the 1980 PUMS. Unemployed PTs and PTs who were no longer working or seeking employment were excluded from all but the analysis of labor force participation. They numbered 288 cases in 1990 and 210 in 1980. The 1980 sample is self-weighting, with each case representing 20 PTs, whereas weights supplied with the 1990 PUMS are used to estimate the actual number of PTs for that year.

A comparison of the AHA count of full-time -- equivalent PTs, PTAs, and physical therapy aides practicing in hospitals with the count of all PTs from census data provides a partial external validation of census data identification. The AHA survey indicated 35,455 full-time -- equivalent PTs working in hospitals in 1990.[6] In comparison, the census count of PTs of all types working in hospitals was 37,818. This number was 35,625 when the census count was restricted to persons working 20 or more hours each week. Although census and AHA counts are not strictly comparable because AHA counts are expressed as full-time equivalents whereas census counts are of persons, there is a rough correspondence between the two independent counts.

A problem arises from the failure of the census to distinguish PTs from PTAs and physical therapy aides. Educational attainment was used in this study as a proxy for licensure because at a minimum, a bachelor's degree is a requirement for a license in all states. Persons identified as PTs and finishing 4 or more years of college were selected for the sample, whereas those with less than 4 years of college were rejected. The numbers of persons rejected by this procedure were 771 for 1980 and 1,412 for 1990. By 1980, most PTs trained before a baccalaureate degree was required would have been either retired or approaching retirement. It is possible that some graduates of 4-year colleges had training leading to employment as a PTA or physical therapy aide while or after completing their baccalaureate degree programs. Internal validation of the educational criteria was supplied by a comparison of the median incomes of those selected and those rejected. In 1989, the median incomes of PTs selected for and rejected from the 1990 sample were $31,013 and $15,000, respectively; comparable incomes in 1979 for the 1980 sample were $24,308 and $13,706, respectively.

Data Analysis

A statistical profile was generated by use of descriptive statistics to focus on four areas used to characterize PTs: geographic distribution, social characteristics, employment characteristics, and income. Statistical analyses were performed with Stata software.(*)

Rates of PTs per 10,000 persons in the population, for each census year, were developed with the number of employed PTs derived from the PUMS for 1980 and 1990 as the numerator; the denominator was the published population data for states and Metropolitan Statistical Areas (MSA) for the years under study. To examine geographic distributions of the demand in 1990, PTs were allocated to their area of employment on the basis of the census question on the location of work.

Approximately 2.7% of PTs worked outside their state of residence in 1990. A stability correlation, which is a measure of the stability of rates, was developed to examine rates of PTs by state by comparing 1980 and 1990 rates.

Census income data are quite detailed and break down self-reported income into income from various sources, including income from wages and income from self-employment in the previous year. By taking the sum of both wages and income from self-employment and ignoring other income sources, we focused the analysis specifically on income from work. Persons who did not work in the year before the census were excluded from the analysis. Income for 1979 was expressed in 1989 dollars by multiplying 1979 income by the ratio of the 1989 consumer price index to the 1979 consumer price index.[7]

To discern differences between regions, PTs and state populations were aggregated for the nine census divisions. Rates were formed for these aggregations, and the resultant rates were mapped with Lotus 1-2-3.([dagger])

A multivariate regression analysis was used to determine factors that influenced income levels. The earnings distribution was skewed, with a long tail to the right. To overcome this problem and to represent earnings with the normal distribution assumed by multiple regression, we transformed earnings by taking the square root of earnings. A square root transformation yielded better results than did a log transformation.

Categorical variables were entered as dichotomies (0 or 1) in the regression. For variables represented by more than two categories, an omitted category served as the base category against which the other categories were compared. Thus, the base census division was the Middle Atlantic division, and income in the other eight divisions was measured against income in that division. The base division for race-ethnicity was non-Hispanic white, for educational level it was a bachelor's degree, and for class of worker it was private employment. Variables for which the relationship to income was nonlinear, such as age and hours worked, were represented by additional terms, the square of the variable. Educational attainment was measured differently in the two censuses: as years of education completed in 1980 and as the highest degree attained in 1990. The 1990 census included four educational degrees that would cover PTs -- bachelor's, master's, professional (medicine, law, dentistry, osteopathy, chiropractic, and so forth), and doctorate. To make educational attainment approximately, comparable for the regression with both years combined. it was assumed that in 1980 a master's degree required 5 or 6 years of college and a professional or doctorate degree required more than 6 years of college. Place of work was determined for only part of the 1980 sample, and place of residence was used for both years in the regression to achieve comparability. The effect of metropolitan area location on income was determined by comparing PTs who resided in the 50 largest metropolitan areas with PTs residing in all other locations. The rate of PTs per 10,000 persons in the population within the state of residence was included to test the effect of the supply of PTs on income.

In addition to the regression coefficients, dollar estimates of the effects of statistically significant variables were computed. These estimated effects were calculated by use of the regression coefficients to predict income. Two estimates were calculated for each variable, and the estimated effect was the difference between these quantities. For example, the estimated effect for gender was found by first multiplying the 1990 mean of each independent variable (other than gender) by its corresponding regression coefficient and then summing these products over all variables. Gender is a dichotomous variable in which men are coded as 0 and women are coded as 1. Multiplying the coefficient for gender by 0 adds nothing to the sum; therefore, the square of this first sum represents the estimated income for men. Adding the coefficient for gender to the previous sum and taking the square of this value yields the income for women. The difference between the estimated incomes for women and men is the effect of gender on income with all other variables in the regression kept constant. For continuous variables, values of the 1990 first and third quartiles rather than 0 and 1 were used in calculating the two estimates to be subtracted from one another. To allow comparisons between years, we used 1990 means for both years, thereby controlling for changes in the variables between the years.

Results

The 1980s was a decade of remarkable growth in the profession of physical therapy. In 1990, the number of employed PTs was 66,270, more than double the 30,600 practicing in 1980. Slightly less growth was demonstrated in the rate, which increased from 1.4 PTs per 10,000 persons in the population in 1980 to 2.7 in 1990.

Geographic Distribution

Table 1 contains data on the relative availability of PTs in each state,and the 50 most populous MSA of the United States in 1990. Rates of PTs per 10,000 persons in the population are presented in rank order from highest to lowest.

[TABULAR DATA 1 NOT REPRODUCIBLE IN ASCII]

As shown in Table 1, PTs were distributed unevenly across the United States. The rates of PTs in different states ranged from 0.9 to 6.6, and the MSA rates ranged from 1.4 to 5.7. The top four states -- Vermont, New Hampshire, Maine, and Connecticut -- were all in New England. The other New England states, Rhode Island and Massachusetts, were also among the top 10 states. Most southern states appeared in the bottom half of the distribution. States tended to retain their rankings between 1980 and 1990, for although the rate of PTs was nearly doubling between 1980 and 1990, the stability correlation between 1980 and 1990 state rankings as .71.

More than half of all PTs practiced in the 50 largest metropolitan areas. Two thirds of the larger MSA had rates that exceeded the US rate. For the largest MSA as a whole, the rate was 45% larger than that for the remainder of the United States, or 3.2 versus 2.2. There was no apparent regional concentration of either high rate or low rate for MSA.

The map in the Figure highlights the regional distribution of PTs. Physical therapists were more available in the states bordering the Pacific Coast and the North Atlantic seaboard. They were relatively unavailable in the states of the lower Mississippi River valley in the East-South-central census division. The Plains states fell in the middle range among the states.

Social Characteristics

The social characteristics of persons in a profession describe the background and qualities of individuals currently in the labor force. These characteristics are presented in Table 2.

Table 2. Social Characteristics of Physical Therapists in 1980 and 1990(a)

                                            Proportion in:
Characteristic and Group                    1980       1990

Age distribution (y)
  Total                                     100.0      100.0
  20-29                                      48.9       33.7
  30-39                                      30.2       41.9
  40-49                                      11.5       16.2
  50 and over                                 9.4        8.2
  Median                                     30.2       33.5
Sex ratio by age (y)
  Total                                      39         35
  20-29                                      26         26
  30-39                                      51         37
  40-49                                      47         44
  50 and over                                82         59
Race-ethnicity distribution
  Total                                     100.0      100.0
  Non-Hispanic white                         93.0       90.2
  Other                                       7.0        9.8
    African American                          3.6        2.8
    Hispanic                                  1.1        2.4
    Other                                     2.3        4.6

Marital status distribution
  Total                                     100.0       99.9
  Married                                    63.7       65.3
  Never married                              29.0       26.4
  Divorced, separated, or widowed             7.3        8.2

Nativity and citizenship distribution
  Total                                     100.0      100.0
  Native born                                93.8       90.5
  Foreign born                                6.2        9.5
    Naturalized                               2.6        2.9
    Not citizen                               2.8        5.7
    Born abroad of American citizens          0.8        0.2

Educational attainment distribution
  Total                                     100.0      100.0
  Bachelor's degree (16 or 17 y)             78.4       73.5
  Master's degree (18 y)                     14.5       16.6
  Professional degree (19+ y)                 7.1        9.9

(a) From the Public Use Microdata Samples of the 1980 and 1990 US censuses.

Physical therapists were found to be young. Less than 10% were more than 60 years of age in 1980 and 1990. However, there are indications that, as a group, PTs are aging. Between 1980 and 1990, the median age of PTs rose from 30.2 to 33.5 years and the modal age shifted from 20 through 29 years to 30 through 39 years. With the rapid expansion in the number of PTs over the last several decades, young PTs dominated the profession numerically, at least until 1980. By 1980 there was already a substantial number of PTs. The addition of a large cohort of young PTs during the 1980s was insufficient to overcome the aging of the large base of PTs already in the profession.

The sex ratio, defined as the number of men per 100 women, is presented by age in Table 2 to describe the balance between men and women. In 1990, the profession had about three women for every man. Between 1980 and 1990, there was actually a small decrease in the sex ratio, from 39 to 35. Although the sex ratio at the entry ages remained constant between 1980 and 1990, it declined at ages above 30 years, from 55 to 41. Nevertheless, in 1990, there were still relatively more older men rather than younger men in the profession. This fact had an impact on the age distributions of female and male PTs, for the median ages were 32.9 years for women and 35.2 years for men.

More than 90% of PTs were non-Hispanic whites in both years. There was a small increase in the proportion of all others during the 1980s. This growth was centered in the Hispanic and other categories (mostly Asian Americans), and there was a small decline in the representation of African Americans. Most PTs were married in both 1980 and 1990, with very little change in the marital status distribution between 1980 and 1990.

One effort to meet the demand for personnel in the United States has been through the recruitment of foreign-born PTs. Table 2 confirms the increased use of this source. In 1980, about 6% of PTs were foreign-born. In 1990, about 10% were foreign-born. Some foreign-born personnel remain in the United States and eventually become naturalized American citizens. Between 1980 and 1990, the proportion of foreign-born personnel remaining unnaturalized doubled from 2.8% of all PTs to 5.7%.

In 1979, the professional educational requirement recommended by the APTA House of Delegates became a postbaccalaureate degree. Although many schools had not yet changed the degree awarded, there was already evidence of an increase in the attainment of postbaccalaureate degrees by 1990. Between 1980 and 1990, the percentage of PTs with a postbaccalaureate degree rose from 21.6 to 26.5. Some of these advanced degrees were earned while people were working as PTs. In 1990, 9.1% of PTs were enrolled in school, up slightly from 8.1% in 1980. Physical therapy aides and PTAs were even more likely to be enrolled in school, 20.2% in 1990 and 23.5% in 1980.

Employment Characteristics

Table 3 contains data that describe labor force participation by PTs, their employees, places of work, and the hours and weeks of work in the previous year. Taken collectively, these topics describe the conditions of work for PTs.

Table 3. Work Characteristics of Physical Therapists in 1980 and 1990(a)

                                            Percent
                                            Distribution in:
Characteristic and Group                    1980         1990

Labor force participation
  Total                                      99.9        100.0
  Employed                                   87.9         91.9
  Unemployed                                  0.7          0.5
  Out of labor force                         11.3          7.6

Class of worker
  Total                                     100.1        100.0
  Private                                    67.1         73.7
  Government                                 20.8          9.9
  Self                                       12.2         16.4

Place of work
  Total                                     100.0         99.9
  Medical office                              4.6          8.3
  Hospital                                   56.7         43.8
  Nursing home                                6.5          3.0
  Physical therapist office                  18.0         35.3
  Other                                      14.2          9.5

Weeks worked in previous year
  Total                                      99.9         99.9
  Less than 26                                9.7          6.6
  26-44                                      14.4         13.8
  45 and over                                75.8         79.5

Usual weekly hours worked in
     previous year
  Total                                      10.1         99.9
  Less than 20                               11.6          9.0
  20-34                                      11.3         15.8
  35-44                                      56.5         49.0
  45 and over                                20.6         26.1

(a) From the Public Use Microdata Samples of the 1980 and 1990 US censuses.

Most person identified as PTs were found to be active in the labor force. For 1990, almost 92% were active, up from about 88% in 1980. Unemployment in both years was less than 1%. The increase in labor force participation was general and occurred for both genders. Although female PTs were more likely than male PTs to be inactive, more than 90% were practicing by 1990.

The censuses delineate three categories of employment: private, government, and self. Private employment is an omnibus category describing persons who do not work for the government or themselves and is the category to which most PTs belonged in both years. Government employment declined from 20.8% in 1980 to 9.9% in 1990, whereas private employment and self-employment were both gaining in importance.

Place of work goes beyond the location of employment and also addresses the question of how work is organized or structured in the profession. Place of work is available from census data and is derived from reports on industry of employment. In 1980, 57% of PTs worked in hospitals. Employment in a PT office almost doubled during the 1980s and, by 1990, was approaching the percentage of PTs working in hospitals. The remaining three places of work employed about one in five PTs in 1990, down from about one in four in 1980. Census data reflect work in the major place of employment and do not report work in multiple sites, a practice that may be common among PTs.[10]

Employment in a hospital appears to have been the entry point into the profession for most PTs. In 1980, 67% of PTs less than 30 years of age were employed in hospitals; that figure was 42% for PTs 40 or more years of age. Respective percentages for 1990 were 57 and 32. Work in hospitals and PT offices was related to gender and race. In 1990, but not in 1980, women were more likely than men to work in a hospital and less likely to work in a PT office. Similarly, all other PTs were more likely than non-Hispanic white PTs to work in a hospital and less likely to work in a PT office.

Hours and weeks of work in a year are two expressions of attachment to the labor force that may be gleaned from census data; these are shown in Table 3 for persons who worked in the year before the census. The majority of PTs worked for 45 or more weeks and for 35 or more hours per week. Between 1980 and 1990, there was a slight shift toward a longer work week and a longer work year. These shifts were related to the gender and marital compositions of the profession.

Many women spend more time than men on family and home responsibilities. Professional and family responsibilities often interact with and have an influence on one another. There was, however, virtually no difference between unmarried women and men in weeks and hours worked in 1979 and 1989 in our data; similar percentage worked 45 or more weeks and 35 or more hours per week. Among married women and men, male PTs were more likely to work 45 or more weeks and 35 or more hours per week. In 1989, 63% of married women and 95% of married men worked 35 or more hours per week. Among the unmarried persons, 88% of women and 87% of men worked 35 or more hours per week. More than half of the women with a child under 18 years of age in their families, 51%, worked less than 35 hours per week.

Income

Income is determined by a combination of the valuation placed on the work that PTs do and market conditions -- the demand and supply of PTs.[4] Many of the variables previously discussed influence the earnings of individual PTs. Income is reported in Table 4 as the median income from work in 1979 and 1989 for the geographic, social, and work categories used in Tables 1 through 3. Median incomes were $24,308 in 1979 and $31,013 in 1989, an increase of about 28% between these years. There were considerable differences in income between categories. All variables, with the exception of nativity in 1989, appeared to some extent to influence PT income. For example, gender appeared to be related to income. In 1979, women earned 73 cents for every dollar earned by men; in 1989, they earned a bit more, 74 cents. Many factors other than gender and year of observation determine the income of PTs. These factors are interrelated. Gender, as previously shown, is related to age. As noted, male PTS are, on average, years older than female PTs. One reason that older PTs earn more than younger PTs is that age represents years on the job and years of experience. Hence, the gender gap income in part reflects the age gap in income. As a result of this and other relationships, the simple differences between categories in Table 4 are not true indicators of the effect of a variable. A multiple regression analysis presented in Table 5 assists in appraising the influence of each variable in Table 4. Income is analyzed for both years combined and for each year separately.

Table 4. Median Income from Work in 1979 and 1989(a)

                                            Income
                                            (Dollars) in:
Category                                    1979(b)     1989

All physical therapists                     24,308      31,013

Gender
  Women                                     22,786      29,224
  Men                                       31,371      39,386

Age (y)
  20-29                                     23,268      29,029
  30-39                                     26,403      32,178
  40-49                                     29,455      33,317
  50 and over                               30,618      35,369

Nativity
  Native born                               24,248      31,013
  Foreign born                              25,731      31,241

Race-ethnicity
  Non-Hispanic white                        24,240      31,001
  African American                          24,804      33,668
  Hispanic                                  26,799      32,329
  Other                                     29,079      33,126

Marital status
  Married                                   25,663      30,983
  Never Married                             22,275      30,013
  Divorced, separated, or widowed           26,326      35,026

Educational attainment
  Bachelor's degree                         23,923      30,014
  Master's degree                           25,660      32,111
  Professional degree                       27,560      35,098

Class of worker
  Private                                   23,997      30,001
  Government                                24,864      30,007
  Self                                      31,127      40,001

Place of work
  Medical office                            28,221      33,051
  Hospital                                  24,359      30,014
  Nursing home                              22,239      30,116
  Physical therapist office                 26,155      33,602
  Other                                     23,951      28,177

Weeks worked in previous year
  Less than 26                               6,867       6,086
  26-44                                     14,646      19,020
  45 and over                               26,617      33,222

Weekly hours worked in previous year
  Less than 20                              11,273       9,743
  20-34                                     17,643      20,037
  35-44                                     25,326      32,004
  45 and over                               30,797      41,008

Division
  New England                               22,252      29,600
  Middle Atlantic                           25,783      33,788
  East-North-Central                        24,633      30,034
  West-North-Central                        24,599      28,547
  South Atlantic                            24,006      30,832
  East-South-Central                        24,804      33,131
  West-South-Central                        25,236      31,427
  Mountain                                  25,232      30,859
  Pacific                                   25,574      32,012

Metropolitan area residence
  In large metropolitan area                24,633      31,274
  Not in large metropolitan area            24,035      30,001

(a) From the Public Use Microdata Samples of the 1980 and 1990 US censuses.

(b) In 1989 dollars.

[TABULAR DATA 5 NOT REPRODUCIBLE IN ASCII]

Nativity was the only variable that did not affect income. Self-employment had the strongest effect on PT income in both years. In 1989, there was a $9,565 income gap between self-employed PTs and PTs working in other types of employment; this difference was more than $3,000 greater than the difference in 1979. Self-employed PTs, who represented 16% of all PTs in 1989, comprised 73% of those earning $100,000 or more. However, self-employment was not a guarantee of a high income because many part-time PTs were self-employed and their income was no higher than that of other part-time employees.

The effects on income of weeks worked in 1979 and 1989 were $3,855 and $2,645, respectively, or the difference in income between PTs at the first quartile (48 weeks) and PTs at the third quartile (52 weeks).

In addition to self-employment, categories of variables that significantly affected the earnings of PTs in both years included gender, age, marital status, weeks and hours worked, working in a large metropolitan area, and working in the West-South-Central and Mountain divisions of the United States. Several categories of variables, including being African American, having a professional degree, working in a PT office, the rate of PTs in the state of residence, and working in the East-North-Central, West-North-Central, South Atlantic, and Pacific divisions, influenced income only in 1989. Working in New England was the only category of variable that affected income in 1979 but not in 1989.

Married PTs earned in excess of $2,000 more than unmarried PTs in both years. This result masks some striking differences between women and men. In both years, married women earned less than unmarried women earned, whereas the reverse was true for men. The positive coefficient for marital status in Table 5 resulted from the very large difference in earnings between unmarried and married men, more than $12,000 in both years, which more than compensated for the smaller difference between unmarried and married women.

Although Table 4 indicates a higher income among African Americans in 1989, when all other variables were controlled for, African-American PTs earned $3,500 less than did non-Hispanic white PTs. There was no difference in earnings in 1979.

In 1979, PTs in three regional divisions had lower incomes than did those in the Middle Atlantic states. By 1989, PTs in six of the eight regional divisions had lower incomes than did those in the Middle Atlantic states. Lowest of all were incomes for PTs in the West-North-Central states, where in 1989 earnings were $5,600 lower than those for PTs in the Middle Atlantic states.

A relative shortage of PTs raised income, and a relative excess of PTs lowered income. However, when separate regression analyses were carried out for female and male PTs in 1989, the coefficient for the proportion of PTs was negative and statistically significant in the regression for women but was not significant in the regression for men. An oversupply of PTs lowered the income of women, but the demand for men was such that their incomes were unaffected by the supply of PTs. Marital status also interacted with gender. It appeared to affect the income of men but not the income of women. However, the marital status relationship appeared for women when weeks worked and hours of work were omitted from the regression, indicating that for women these two variables acted as intervening variables in the relationship of marital status to income. When the number of children born was added to the regression for female PTs, incomes for women with three or more children were lower than those for other women by more than $2,100.

To establish the change in income between 1979 and 1989, a regression was performed with data for both years. In addition to the variables described in Table 5, year was entered as a separate dichotomous variable. Year had an impact on income in this regression. In dollar terms, adjusted PT income grew by $5,565 between 1979 and 1989.

Discussion

Physical therapy underwent great expansion during the 1980s. This report documents that growth by use of a previously untapped source, the US Census of Population and Housing. Although the numbers of PTs found with these data are slightly lower than those reported from other sources,[4,15] there is reason to believe that the counts reported here are more accurate than those previously estimated. However, the data have potential problems. Although the sample size is substantial for each census, statistics for smaller subgroups, such as some states and metropolitan areas, are subject to sampling error. The census relies on self-identification for all questions, and no independent checks are made of the responses. It is possible for persons to misidentify their occupations. Despite these limitations, these data provide a more detailed description of PTs than was previously available in terms of geographic distribution and pertinent social characteristics.

Although many of the data speak for themselves in describing who and where PTs are, a few salient points related to initiatives within the profession warrant discussion. There were 2.7 PTs per 10,000 persons in the population in 1990, nearly double the 1980 rate of 1.4. No standard relating PTs to population currently exists; therefore, no statements can be made about the oversupply or undersupply of PTs. Recruitment of foreign-born persons has been used to meet the increased demand for PTs in the United States. If domestic growth does not meet the demand for PTs and immigration laws allow, these trends may continue.

Physical therapists are far from randomly distributed across the states and metropolitan areas of the United States. This distribution may be related to the distribution of ther health professionals (particularly physicians), the number of health care facilities in metropolitan areas, or the location of schools of physical therapy. Our finding of a high representation of PTs in northeastern and West Coast states is consistent with previous work indicating that PTs are in greater demand in areas of high income and high physician presence.[5]

Even as PTs grew more numerous during the 1980s, members of the profession as a whole grew older. A similar aging of the PT ranks is expected to occur during the current decade.

Physical therapy, along with nursing and other health professions, has been attractive to women. In large measure, this situation stems from the openness of the profession to women from its inception shortly after the turn of the century. Career paths for female and male PTs differ,[4] and several gender-based differences became apparent in the analysis. Retention has been a major concern for the profession. There are hints of a small increase in retention, but the continued existence of sex ratios that increase with age provides evidence that the problem has not disappeared completely. Unless men enter physical therapy at a later age, the difference by age in the sex ratio must be attributable to the larger numbers of women leaving the profession. The retention problem appears to be centered around female PTs, particularly those employed in hospitals. Although more research is warranted to examine the career paths of women in physical therapy, there is evidence in these data that efforts to accommodate women need to be made to continue to retain more experienced, older PTs. After controlling for many relevant variables, we found that a gender gap of $6,600 remained between the incomes of female and male PTs. This point should be addressed, primarily to achieve parity between men and women but also because it may have a bearing on the issue of retention.

An ongoing effort exists to recruit minorities to the field of physical therapy. These data provide evidence that minority PTs are underrepresented and that their representation does not approximate the percentage of minorities in the general population. It is hoped that continued recruitment efforts will yield higher percentages in the future. The negative coefficient for African-American income in the 1989 regression analysis raises disturbing questions that are not easily answered by these data. Equally disturbing is the lack of significance for this variable in 1979.

There are several indicators in these data that point to increased professionalization and maturity of the field of physical therapy. Educational levels shifted toward graduate education. After adjustment for inflation, income increased by more than 25%. This increase occurred concurrent with a doubling in the number of PTs, by itself a deterrent to higher earnings. Physical therapists moved from hospital-based practices to more autonomous settings, a mark of professionalization. Finally, several employment-related changes occurred among PTs -- increased labor force participation, longer work week, and longer work year -- indicating a greater commitment of PTs to the profession.

Conclusion

This report presents a data source not previously used to investigate demographic characteristics of PTs. The information should prove useful for health planners, administrators, and educators. Ideally, it would be better for the profession if a report of this nature did not have to turn to census sources. The data in this study are already 7 yeas old. Data generated by the profession should be readily available to analysts and health planners investigating professional issues. Such is not currently the case, and this report is an initial step toward generating a profile of the profession. In the future, studies based on census data could be used to triangulate data generated by planners within the profession. Efforts under way by the Federation of State Boards of Physical Therapy and APTA should dovetail with the results reported here and hopefully take the profession beyond these findings.

(*) State Corp, 702 University Dr E, College Station, TX 77840.

([dagger]) Lotus Development Corp, 55 Cambridge Pkwy, Cambridge, MA 02142.

References

[1] Selker LG. Human resources in physical therapy: opportunities for service in a rapidly changing system. Phys Ther. 1995;75:31-37.

[2] Jacoby I. Forecasting requirements for physical therapists. Phys Ther. 1995;75:38-44.

[3] Hack LM, Konrad TR. Determination of supply and requirements in physical therapy: some considerations and examples. Phys Ther. 1995; 75:47-53.

[4] Gwyer J. Personnel resources in physical therapy: an analysis of supply, career patterns, and methods to enhance availability. Phys Ther. 1995;75:56-65.

[5] Michels E. Human resources. Phys Ther. 1995;75:444-445. Letter to the editor.

[6] Health United States: 1995. Hyattsville, Md: National Center for Health Statistics; 1996.

[7] Statistical Abstract of the United States: 1995. Washington, DC: US Bureau of the Census; 1995.

[8] Occupational Therapy Manpower: A Plan for Progress. Rockville, Md: American Occupational Therapy Association; 1985.

[9] Bello JM. Speech-language pathologists and audiologists: how many and where? ASHA. 1994;36(12):42-43.

[10] 1993 Active Membership Profile Report. Alexandria, Va: American Physical Therapy Association; 1993.

[11] Wilson Sd, Langwell KM, Deane RT, et al. Identification of physical therapists shortage areas: a study of thirteen states. Phys Ther. 1982-62: 315-323.

[12] Occupation and Industry: 1990 Census of Population and Housing [subject summary tape file]. Washington, DC: US Bureau of the Census; 1995.

[13] Census of Population and Housing, 1980: Public Use Microdata Sample [machine-readable data files]. Washington, DC: US Bureau of the Census; 1983.

[14] Census of Population and Housing, 1990: Public Use Microdata Sample [machine-readable data files]. Washington, DC: US Bureau of the Census; 1992.

[15] Ninth Report to Congress: Health Personnel in the United States. Rockville, Md: Department of Health and Human Services, Health Resources and Services Administration; 1993.