Validity and reliability of a new assessment of lower-extremity dysfunction.

Increased importance is being placed on the measurement of outcomes of rehabilitation programs. Rehabilitation is usually aimed at reducing not just organ function (impairment) but also the consequences for personal life (disability) and social life (handicap). in this process, the primary role of the physical therapist has been described as evaluation and treatment of dysfunction, whereas the role of the physician is to diagnose and treat diseases.[1] Related thoughts have been expressed by Sahrmann[2] and Jette.[3] These different roles demand different methods. In physical therapy, there is a need for practical evaluation methods, suitable for the everyday practice, that are easy to use and that can be used with equipment that is not too expensive. These methods, however, must provide valid and reliable measurements. They should be used to record the changes of functional status between admission and discharge, in follow-up studies, and for quality assurance.

Traditionally, physical status measures include goniometry, muscle force measurement, radiographic examination, and so forth. We believe, however, that the results obtained with such measures, to reflect relevant aspects of quality of life, should also be related to the physical and social ability of the patient. In our study, a 20-variable assessment system of the lower extremity was used, covering impairment, disability, and handicap according to the World Health Organization (WHO) classification.[4] The selection of items for the assessment system was based on the primary author's (UO) clinical experience. The result of the evaluation is a profile showing lower-extremity dysfunction (reduced functional capacity).

Various rating scales and evaluation systems have been described. All these scales cannot be reviewed here, but Table 1 shows the main properties of 25 such scales for evaluation of lower-extremity function. Some scales are very general, whereas others are constructed for specific tasks such as evaluation of knee function in sports, evaluation of social function in rheumatoid arthritis, and so on. The most frequent variables in these scales are gait, pain, stair climbing, gait aids, mobility, muscle strength, knee joint stability, activities of daily living (ADL) functions, and rising up from a chair. Other variables are varus/valgus deformity, flexion deformity, locking phenomena, swelling, range of movement in different joints, physical activities, social activities, capability to work, radiographic findings, capability to use common communications, sleep, energy expenditure, and so on. Several scales have been constructed for the needs of orthopedic surgeons in assessing range of movement, walking ability, and pain.[5-10] They have been constructed to help the surgeon decide whether to operate and to evaluate ate the results of surgery and other types of treatment. Other scales are designed for arthritis evaluation and diagnosis[11-15] and for the evaluation of patients with stroke.[16-18] Separate instruments have been made for evaluation of ADL[17-19] and quality of life.[19] Some of the scales have been tested for validity or reliability,[12,15,18-20] but most of them have not. There is a large variation of the number of items in the different evaluation system - from a few to 100. Some scales are based on questionnaires[15,21]; others are based on physical measurements.[9,22,23]

[TABULAR DATA OMITTED]

Our assessment system combines items on range of motion (ROM) and strength with physical and social ability and pain. The system is intended to fulfill the needs of physical therapists for valid and reliable measurements of lower-extremity dysfunction. With the scores plotted in a visual profile, a graphic representation of the functional status of the patient can be obtained. Pretraining and posttraining profiles can be plotted in the same diagram, and with only a quick glance the improvement can be estimated and compared with the goals. The aim of our study was to evaluate the content validity and intertester reliability of our assessment system for the lower extremity, applied on patients with esteoarthrosis of the hip or knee.

Method

Brief Description of the

Assessment System

The assessment system consists of the evaluation of 20 variables, subdivided into five groups: hip impairment, knee impairment, physical disability, social disability (handicap), and pain. Every variable is given a disability score on a five-point scale according to a key for each variable. Zero means no reduced function; a score of means severe dysfunction or total lack of function. The scores are plotted in a diagram, thus giving a profile that shows the functional reduction in every variable. The rating is done by a physical therapist in a standardized manner. The scoring form and a short key to the variables are presented in Tables 2 and 3.

[TABULAR DATA OMITTED]

Patients

One hundred five patients (37 men, 68 women) referred to the orthopedic clinic at the county hospital (Eksjo, Sweden) and accepted for total joint replacement arthroplasty due to osteoarthrosis of the hip and knee were evaluated preoperatively with the assessment system. The mean age of the patients was 69.0 years (SD = 9.0, range = 46-91), 66.9 years (SD = 10.0, range = 46-87) for the men and 70.2 years (SD = 8.3, range = 54-91) years for the women. The mean height was 175 cm (SD = 8.4, range = 155-197) for men and 163 cm (SD = 5.6, range = 150-178) for women. The mean body weight was 81 kg (SD = 11.0, range = 58-112) for men and 73 kg (SD = 11.9, range = 54-100) for women. Most of the patients selected for surgery had a radiographically confirmed osteoarthrosis of degree III (on a four-grade scale) or disabling pain at rest. A subgroup of 42 patients was tested for intertester reliability. Data obtained for localization of osteoarthrosis are presented in Tables 4 and 5.

Table 4. Sex Distribution and
Localization of Osteoarthrosis for
All Patients

           Localization
Sex        Hip   Knee   Total

Men        29     8      37
Women      32    36      68
Total      61    44     105

Table 5. Sex Distribution and
Localization of Osteoarthrosis for
Subgroup in the Reliability Study

           Localization
Sex        Hip   Knee   Total

Men         8     3      11
Women      13    18      31
Total      21    21      42

Test Procedure

All patients were tested by a physical therapist (UO). The first 42 patients were measured by two independent physical therapists. Active range of motion in the hip and knee was measured with a standard manual goniometer(*) with long telescopic shanks. The tests were performed according to routines recommended by the American Academy of Orthopaedic Surgeons.[24] Muscle strength, tested as isometric extension and flexion forces in the knee, was measured with a strain-gauge dynamometer [dagger] at 45 degrees of knee flexion and with the patient in a sitting position. A cuff was applied to the leg 16 cm distal to the knee joint space. The cuff was connected to the dynamometer via a cord running perpendicular to the leg. Maximum force was read from a display on the recording equipment. Force data, together with height and weight (for calculating the torque produced by the weight of the leg). were entered into a programmable calculator, [double dagger] and torque was calculated, compensating for the weight of the leg. The computations were based on anthropometric data and performed according to standard biomechanical procedures.[25] The subject was tested in a sitting position, with the knee in 45 degrees of flexion. Torque was calculated according to the following formula:

M = 9.81 (0.18 x P [+ or -] 0.06 x l x w x cos 45 [degrees]

where M = torque (in newton-meters), P = registered force (in kilograms), l = body height (in centimeters), and w = body weight (in kilograms). The other variables were evaluated according to standardized routines (Tabs. 2 and 3). Rising/sitting down was recorded as the lowest possible sitting height from a chair with adjustable height and without armrests. The heights were adjusted to correspond to sitting heights in ordinary daily life (eg, an ordinary chair, sofa, car seat). Rising from a half-standing position was measured as the maximum number of times the patient could rise from a high chair, with a hip angle of about 135 degrees, Step height was measured using a platform with different step heights, corresponding to ordinary stairs, bus and train stairs, and so on. The time standing on one leg was tested as the number of seconds the patient was able to stand on his or her affected leg. Gait speed was tested on a 65-m indoor walkway. The social variables were evaluated by a personal interview of the patient. Pain was evaluated in a manner related to standard clinical evaluation of the reason for operation. The test results were converted to scores according to a special key explaining how to grade the measurements (Tab. 4). The scores were plotted in a diagram, thus constituting a personal profile of lower-extremity dysfunction. The total time needed to complete the profile was about 30 minutes. All patients gave their informed

consent before testing.

Data Analysis

Statistical measures, such as mean, standard deviation, median, and quartiles tiles were calculated for each variable according to standard routines. Validity was tested by means of factor analysis. The subgrouping of the variables made by the authors (face validity) was compared with the subgrouping obtained by the statistical factor analysis. The type of factor analysis found to be most appropriate was principal components analysis with varimax rotation.[26-31] The number of factors was determined by the criterion "eigenvalue > 1.0."[26,27,29,30] Reliability testing was performed by calculating the Goodman-Kruskal gamma coefficient.[32] For the statistical analysis, a commercial statistics package for the personal computer (SYSTAT 5.0/SYGRAPH [1.0.sup.[sections]]) was used. The validity and reliability concepts and the statistical testing procedures will be discussed in more detail in the "Discussion" section.

Results

Functional Rating Scores in Hip and Knee Osteoarthrosis

Figure l illustrates the median scores found for the 20 variables in the hip joint replacement and knee joint replacement groups. Figures 2 and 3 illustrate the median scores, in addition to the first and third quartiles, of the groups separately. In the hip joint replacement group, there was medium-high scoring for the hip impairment variables (1-4), except for variable 2, extension deficit in the hip joint. in the knee joint replacement group, there was only high scoring for variable 7, quadriceps femoris muscle force. For both groups, the highest scores were for the physical disability (9-15), social disability (16-19), and pain (20) variables.

Factor Analysis (Validity Testy

The component loadings are shown in Table 6. A five-factor solution fulfilled the "eigenvalue > 1" criterion.

[TABULAR DATA OMITTED]

Factors with loadings of less than 0.30 were not considered.[26] Most of the factors were grouped according to the preliminary subgrouping that we chose. There was a discrepancy, however, for variables 2, 6, and 20 (ie, extension deficit of hip and knee groups together with pain instead of hip and knee impairment groups, as was expected).

Correlations Between Two

Independent Testers (Intertester

Reliability)

In Table 7, the Goodman-Kruskal gamma coefficients are given. The values of the coefficients for the different variables vary from .99 to 1.00, thus indicating almost perfect agreement between the two physical therapists.

Table 7. Correlation Between Two
Independent Observers (Intertester
Reliability) for Scores of Rating
Scale for the Lower Extremity

                             Goodman-
                             Kruskal
Variable                     Gamma

Hip impairment
  Hip flexion                 1.00
  Extension deficit, hip      1.00
  Abduction, hip              1.00
  Abduction, hip              1.00
Knee impairment
  Knee flexion                1.00
  Extension deficit, knee     1.00
  Quadriceps femoris
   muscle force               1.00
  Hamstring muscle force      1.00
Physical disabilit
  Rising from half-standing   1.00
  Rising/sitting down         1.00
  Step height                 1.00
  Standing on one leg         1.00
  Stair climbing              1.00
  Gait speed (m/s)            1.00
  Walking aid                 1.00
Social disability
  Communication/transport      .99
  ADL(a) functions, other     1.00
  Leisure time/hobbies        1.00
Pain
  Pain                        1.00

(a) ADL = activities of daily living.

Discussion

The Functional Assessment

System

Our assessment system was designed to evaluate dysfunction (reduced functional capacity) of the lower extremities. It is intended to be used for documentation of disability status preoperatively and postoperatively, for long-term follow-up, and for the design of individual training and rehabilitation plans.

Neither the choice of variables nor the monitoring of data is unique, but to our knowledge this type of profile is new in physical therapy. The assessment systems main advantages are its simplicity and its practicality. We have found the system to be an effective instrument for communicating functional status to the patient, the physician, and other health professionals.

The WHO classification of impairments, disabilities, and handicaps is often referred to for description of rehabilitation on different levels.[4,33] We used the WHO classification as a basis for our assessment system, but there is not a complete correspondence. in our system, some variables that have been grouped under impairment are classified as handicaps in the WHO classification. In the WHO classification, there is a sequence of underlying illness-related phenomena[4]:

Disease Impairment Disability

Handicap

Impairments (I-code) are concerned with abnormalities of body structure and appearance and with organ or system function, resulting from any cause; in principle, impairments represent disturbances at the organ level,

Disabilities (C-code) reflect the consequences of impairments in terms of functional performance and activity by the individual; disabilities thus represent disturbances at the level of the person.

Handicaps (H-code) are concerned with the disadvantages experienced by the individual as a result of impairments and disabilities; handicaps reflect interaction with and adaption to the individual's surroundings.

Validity of the Scale

Validity is neither easily defined nor directly testable. It is concerned with what is being measured and what relation the measurement has to the phenomenon being measured. There are many kinds of validity. If an assessment system has enough relevant items and adequately covers the domain under investigation, it is said to have content validity. If the assessment system properly reflects the theory behind the measurement, it is said to have construct validity. The broader the validity of a measurements the broader the inferences that can be validly drawn about what is being measured.[34-36]

Our variables were chosen to show major consequences of a reduced lower-extremity function, as reflected in daily life. The selection of variables was based on our many years of practical

experience. No variable has been taken directly from any other scale, but other authors have a similar experience and many of the variables can be found in other scales as well (Tab. 1). We believe many of the disability variables, such as stair climbing and step height, have a kind of intuitive common-sense validity (face validity).

In daily life, patients with reduced lower-extremity function will meet a number of obstacles related to furniture, layout of their homes, the traffic situation, and so forth. We have measured seat heights of different chairs; collected data on step heights of staircases and footsteps on cars, buses, and trains; and obtained green-light traffic time from the local street commission. mission. Our scales are based on these data. The greater the difficulty, the higher the score will be for a specific item. Consequently, our system reflects the degree of reduced lower-extremity function.

The social disability variables have been graded according to their influence on work, household activities, and hobbies/leisure-time activities. For example, sick leave or need for external help will result in high scores.

We arranged the variables into five separate subgroups according to clinical knowledge. By means of factor analysis, the variables of the scales were examined and could be grouped into factors with regard to clinical, functional, and social relationships. Factor analysis can be used for either confirmatory or exploratory purposes.

The factor solution we obtained from the factor analysis was very close to our preliminary subgroupings. We took this finding to indicate good content validity. Two variables, extension deficit in the hip and extension deficit in the knee, had a tendency to group together with pain. There was a negative factor loading (ie, less pain in patients with extension deficit). The extension deficit may be due to a flexion contracture, which may protect the joint from painful movement. Contracture of the hip often gives compensatory contracture of the knee, and vice versa. Pain was equally scored under social disability variables and as a single factor of its own. In spite of this result, we regard a grouping of these variables under the hip and knee variables, respectively, as more logical than grouping them together with pain, but the association is worth noting. We used the "eigenvalue > 1.0" criterion (Kaiser criterion) to decide the number of factor.[26,27] This criterion is equivalent to a Cronbach's alpha of > 0. The five factors chosen explain 69.0% of the total variation. Other types of factor analysis (common factors analysis) and rotations (equamax, quartimax) were also attempted. The best factor solution, however, was obtained with classical varimax rotation of principal components. Factor analysis has also been performed for the hip and knee data separately, but these analyses only confirmed the general picture obtained from the total data.

Concurrent validity (criterion validity), that is, how well a measurement agrees with a well-established and accepted method, was not evaluated in this study, but will be examined in future studies.

Reliability of the Scale

Reliability is concerned with the repeatability of measurements or the ability of a test to measure something in a reproducible and consistent fashion.[34-36] Intratester reliability is often examined with a test-retest procedure. Such a procedure was not performed in our study. A short interval between the tests would, in our opinion, allow examiners to remember their results. A long time interval in a group of elderly patients did not seem appropriate because the condition of the patients may not be the same on the two occasions. In our study, we examined intertester reliability. We used the Goodman-Kruskal gamma statistic to examine reliability. We believe this coefficient is appropriate for measuring the relation between two ordinally scaled variables, especially if there are ties (ie, many observations having the same rank). It measures the difference between agreements and disagreements divided by the sum of the agreements and disagreements. We found high values for all variables Goodman-Kruskal gamma = .99-1.00), indicating almost perfect agreement between the two observers (ie, very high intertester reliability).

Choice of Variables

For knee and hip ROM in our scale, we tested active range of motion, whereas most other systems measure passive range of motion.[37-39] Good reliability in manual goniometry has been reported by several authors[39-41] for passive range of motion, although how these estimates relate to our measurements is uncertain. Quadriceps femoris muscle force measurements, isometric as well as isokinetic, have been used extensively. In our scale, we have used isometric measurements because they are easy to obtain in a standardized position.

The disability variables chosen refer to activities of everyday life (eg, rising from a chair, stair climbing). The social disability variables refer to common social activities of the patient such as work and household activities. Stam[42] used similar variables to assess rehabilitation after sports injuries. He also examined a group of patients with Guillain-Barre syndrome and found that they were capable of performing all 10 functional tests in his test battery, even though quadriceps femoris muscle force was reduced 30% to 40%. Thus, the disability group variables probably give a better picture of the patient's functional ability than traditional strength measurements.

Gait is a complex motor activity demanding postural control as well as control of movement. Generally, gait is represented by three classical variables: speed, step length, and step frequency. In our scale, gait speed, a combination of step frequency and step length, is used as a global measure of walking ability. Lowering of the gait speed generally affects the step frequency as well as step length. Mattsson and Bostrom[43] reported that self-selected gait speed validly reflects the degree of impairment in patients with moderate osteoarthrosis of the knee.

In many studies, pain has been evaluated with the visual analog scale. However, we decided to evaluate pain in a manner to obtain a uniform scoring system (ie, the same range of scores for every item). This evaluation includes both quantitative and qualitative aspects, and it also relates to indications for operation.

Comparison With Other Scales

Many different scales have been described (Tab. 1). Many of these scales are very general, such as the Sickness Impact Profile.[19] Other scales are very specific, intended to be used, for example, in sports medicine.[20.43] Some scales consist of only a few variables,[14,17] whereas others have many variables, are very time consuming, and thus are not very practical to use.[20,23] Our assessment system has a moderate number of variables grouped in five categories, representing five different dimensions of lower-extremity function that we believe are important in physical therapy evaluation.

Relation to Health Concept

Nordenfelt,[45] in his theory on health, puts emphasis on vital goals. He defines health as a person's ability to fulfill his or her basic needs and as the person's ability to attain the goals set by himself or herself

Our assessment system can be related to the ideas of this theory. Our system permits an evaluation of the patient's present physical status (repertoire) and also an evaluation of a desired (expected) physical status, which can be seen as an individual, vital goal. The deficit between the present status and the desired status indicates a need for physical therapy (Fig. 4).

Conclusions

From this study, we conclude that our assessment system for the lower extremity has good validity and good intertester reliability. The scale can be used for functional evaluation of patients with joint impairments affecting the lower extemity. It is intended to be used for preoperative and postoperative evaluation as well as for follow-up and design of training programs. It is specially adapted for the needs of the physical therapist, but might be used by other health professions as well.

References

[1] Paris SV. Clinical decision making: orthopaedic physical therapy. In: Wolf SL, ed. Clinical Decision Making, in Physical Therapy. Philadelphia, Pa: FA Davis Co; 1985:215-254. [2] Sahrmann SA. Diagnosis by the physical therapist - prerequisite for treatment: a special communication. Phys Ther. 1988;68:1703-1706. [3] Jette AM. Diagnosis and classification by physical therapists: a special communication. Phys Ther. 1989;69:967-969. [4] International Classification of Impairments, Disabilities, and Handicaps. Geneva, Switzerland: World Health Organization; 1980. [5] Hutchinson TT, Boyd NF, Feinstein AF, Scientific problems in clinical scales, as demonstrated in the Karnofsky index of performance status. J Chronic Dis. 1979;32:661-666. [6] Insall JN, Ranawat CS, Aglietti P, Shine J. A comparison of four models of total knee replacement prostheses. J Bone Joint Surg [Am]. 1976;58:754-765. [7] Kettelkamp DB, Thomson C. Development of a knee scoring scale. Clin Orthop. 1975;107: 93-99. [8] Larsson CB. Rating scale for hip disabilities. Clin Orthop. 1963;31:85-95. [9] Laskin RS. Total knee replacement. Orthop Clin North Am. 1979;10:223-247. [10] MacKinnon J, Young S, Baily RA. The St George sledge for unicompartimental replacement of the knee. J Bone Joint Surg [Br]. 1988; [11] D'aubigne RM, Postel M. Functional results of hip arthroplasty with acrylic prosthesis. J Bone Joint Surg [Am]. 1954;36:451-475. [12] Larsson S-E, Jonsson B. Locomotion score in rheumatoid arthritis. Acta Orthop Scand. 1989;60:271-277. [13] Katz S, Vignos PJ, Moskowitz RW, et al. Comprehensive outpatient care in rheumatoid arthritis. JAMA. 1968;206:1249-1254. [14] Steinbrocker O, Traeger CH, Batterman RC. Therapeutic criteria in rheumatoid arthritis. JAMA. 1949;140:659-662. [15] Jette AM. Functional status index: reliability of a chronic disease evaluation instrument. Arch Phys Med Rehabil. 1980;61:395-401. [16] Brorsson B, Hulter-Asberg K. Katz index of independence in ADL: reliability and validity in short-term care. Scand J Rehabil Med. 1984;16: 125-132. [17] Katz S, Ford AB, Moskowitz RW, et al. Studies of illness in aged: the index of ADL - a standardized measure of the biological and psychosocial function. JAMA. 1963;185:914-919. [18] Viitanen M, Fugl-Meyer KS, Bernspang B, Fugl-Meyer AR. Life satisfaction in long-term survivors after stroke. Scand J Rehabil Med. 1988;20:17-29. [19] Sullivan M. Livskvalitetsmatning i kliniska utvarderingsforsok (quality of life measurement in clinical trials) [in Swedish, English summary], Scand J Behav Ther. 1988;17 (suppl 8):29-47. [20] Tegner Y, Lysholm J. Rating systems in the evaluation of knee ligament injuries. Clin Orthop. 1985;198:43-49. [21] Meenan RS, Gertman PM, Mason JH. Measuring health status in arthritis. Arthritis Rheum. 1980;23:146-152. [22] Harris WH. Traumatic arthritis of the hip after dislocation and acetabular fractures: treatment by mold arthroplasty. J Bone Joint Surg [Am]. 1969;51:738-745. [23] Geens S, Clayton ML, Leidholt JD, et al, Synovectomy and debridement of the knee in rheumatoid arthritis. J Bone Joint Surg [Am]. 1969;51:626-642. [24] American Academy of Orthopaedic Surgeons. Joint Motion: Method of Measuring and Recording. Edinburgh, Scotland: Churchill Livingstone; 1965. [25] LeVeau B. Williams and Lissners Biomechanics of Human Motion. 3rd ed. Philadelphia, Pa: WB Saunders Co; 1992. [26] Gorsuch RL. Factor Analysis. 2nd ed. London, England: Lawrence Erlbaum Associates Inc; 1983. [27] Harman HH. Modern Factor Analysis. 3rd ed. Chicago, Ill: University of Chicago Press; 1976. [28] Johnson RA, Wichern DW. Applied Multivariate Statistical Analysis. 2nd ed. London, England: Prentice-Hall; 1988. [29] Kim J-O, Mueller CW. Introduction to Factor Analysis, Newbury Park, Calif. Sage Publications Inc; 1988. [30] Kim J-O, Mueller CW. Factor Analysis: Statistical Methods and Practical Issues. Newbury Park, Calif: Sage Publications Inc; 1988. [31] Kleinbaum DG, Kupper LL, Muller KE. Applied Regression Analysis and Other Multivariate methods. 2nd ed, Boston, Mass: PWS-Kent Publishing Co; 1988. [32] Siegel S, Castellan NJ. Nonparamettic Statistics for the Behavioral Sciences. 2nd ed. New York, NY: McGraw-Hill Inc. 1988. [33] Grimby G, Finnstam J, Jette A. On the application of the WHO handicap classification in rehabilitation. Scand J Rehabil Med. 1988;20: 93-98. [34] Currier DP. Elements of research in Physical Therapy. 2nd ed. Baltimore: William & Wilkins; 1984. [35] Payton OD. Research: The Validation of Clinical Practice. 2nd ed. Philadelphia, Pa: FA Davis Co; 1988. [36] Streiner DL, Norman GR. Health Measurement Scales: A Practical Guide to Their Development and Use. Oxford, England: Oxford University Press; 1989. [37] Boone DL, Azen SP, Lin C-M, et al. Reliability of goniometric measurements. Phys Ther. 1978;58:1355-1360. [38] Gajdosik RL, Bohannon RW. Clinical measurements of range of motion: review of goniometry emphasizing reliability and validity. Phys Ther. 1987;67:1867-1872. [39] Rothstein JM, Miller PJ, Roettger RF. Goniometric metric reliability in a clinical setting, Phys Ther. 1983;63:1611-1615. [40] Enwemeka CS. Radiographic verification of knee goniometry. Scand J Rehabil Med. 1986; 18:47-49. [41] Gogia PP, Braatz JH, Rose SJ, Norton BJ. Reliability and validity of goniometric measurements at the knee. Phys Ther. 1987;67:192-195. [42] Stam HJ. Dynamometry of the Knee Extensors: Isometric and Isokinetic Testing in Healthy Subjects and Patients. The Hague, the Netherlands: University of Rotterdam: 1990. Thesis. [43] Mattsson E, Bostrom L-A. The physical and psychosocial effect of moderate oteoarthrosis of the knee, Scand J Rehabil Med. 1991;23: 215-218. [44] Feagin JA, Blake WP. Postoperative evaluation and result recording in the anterior cruciate ligament reconstructed knee. Clin Orthop. 1983;172:143-147. [45] Nordenfelt L. On the Nature of health. Dordrecht, the Netherlands: D Reidel Publishing Co; 1987:35-104.