A quantitative analysis of research in Physical Therapy.

Research and interest in research have increased in physical therapy over the past two decades. Many articles and books on how to do research have been published.[1-5] Physical therapy databases have been developed (eg, Physiotherapy Index). Editorials and journal policy statements about research have been published in most physical therapy journals.[6-9] Articles encouraging the formation of special interest groups on research and on obtaining funding for research have been published.[10,11] In some countries, special funds for research in physical therapy have been instituted (eg, the Physiotherapy Research Foundation in Australia and in the United Kingdom, and the Foundation for Physical Therapy Inc in the United States). journal editors and ex-editors have highlighted an increased publication of research in the profession.[12] New English-language journals, including the Journal of Orthopaedic and Sports Physical Therapy and Physiotherapy Theory and Practice, have appeared. At the same time, established physical therapy journals--Australian Journal of Physiotherapy, Physical Therapy, Physiotherapy Canada, and Physiotherapy--have increased the volume of research contributions they publish. Together, these developments demonstrate the increasingly active interest in research within physical therapy over the past two decades.

Physical therapists appear to anticipate that research will benefit the profession and practice of physical therapy. Among the benefits expected from research are scientific credibility[13] and the growth and development of the profession.[14,15] Research is also expected by some to validate current practice[16] and to satisfy increasing demands for the accountability of professional groups.[17,18] Various sources expect treatments to be of high quality and cost effective,[6] and research is expected to demonstrate that physical therapy interventions are cost effective[19,20] and of high quality.[21] Research is also expected to demonstrate and improve the effectiveness of physical therapy techniques, interventions, and modalities.7,15,22,23 The emphasis on the benefits of research for clinical practice reflects the centrality of practice in physical therapy. An important question is whether research is providing physical therapy with the expected benefits.

Research outcomes are in large part a function of the suitability and productivity of the research method used. A research method most conspicuously recommended in physical therapy, both explicitly and implicitly, is group-experimental designs. This is an experimental method that compares outcomes for groups, rather than outcomes for individuals or series of individuals (ie, single-subject experimental designs), following the application of controlled interventions.[1(p184)] The method is also known as group-comparison experiments[1(p184)] and randomized clinical trials.[24-27] If some requirements of the group-experimental method are made less rigorous, it is then called a quasi-group-comparison design or a clinical trial. Explicit support is apparent in the positive advocacy of the investigatory power[24,27,28] and general acceptability of the group-experimental method.[26] Implicit Support for group-experimental designs is apparent in the extent to which articles and books aimed at the physical therapy market focus on these designs. Some books and articles that describe how to do research[1,2,4,28-33] and how to evaluate research34,35 focus almost exclusively on group-experimental designs.

Published criticisms of group-experimental designs suggest that the method might not be optimal for general use in the clinical context.[1,36-42] Among other things, criticisms of group-experimental designs have referred to difficulties in obtaining sufficiently large and appropriately homogeneous samples of subjects within the required time period,[36-39] in providing and maintaining the required experimental control within the clinical context,[1,39-41] in using statistical procedures correctly and appropriately (eg, in medicine[43] and orthopedics,[44] psychology and Sociology,[45] and social work[46]), in interpreting findings based on inferential statistics,[42,47-51] and in validating findings and cumulating knowledge from research using group-experimental designs.[36,42] Further, when used to investigate the relatively small effects typically obtained in clinical practice, group designs (experimental and quasi) produce findings with low power and a reduced probability of correctly rejecting null hypotheses.[52]

The relevance of many findings from group designs for practitioners is questionable. The findings are based on averaged group performance when clinicians focus on individuals.[38,53-55] Another criticism suggests that the choice of problems investigated is constrained by amenability to statistical analysis[48,56] rather than by clinical, theoretical, or practical significance and relevance. Other criticisms of group-experimental designs concern the ethical dilemmas raised by the requirements of the research method. Such ethical issues can arise in the random allocation of subjects to treatment and control groups[25,55] and in the replacement of a therapeutic situation responsive to the changing needs of the patient by one in which treatment is based on previously specified experimental procedures.[36] Some of these problems may be obviated by using the less rigorous quasi-group-comparison designs. Some of the practical problems inherent in clinical research, such as the recruitment of a sufficient number of appropriate subjects, however, still tend to remain for users of group designs.

The popularity of group-experimental designs in physical therapy research and the critiques of these designs in clinical research raise some questions. To what extent does research in physical therapy depend on group-experimental designs? Conversely, to what extent are alternative research methods used? Possible alternative research methods discussed in physical therapy journals include the following: case studies and series of cases,[7,57-59] single-subject experimental designs,[38,60-63] ethnographic studies,[64,65] sequential trials design,[36,66] and surveys.[67] Books intended for physical therapists discuss these and other research methods.1,68 To what extent does research in physical therapy investigate clinical treatment? Does research focusing on clinical treatments use methods different from those used in nonclinical research?

Answers to these questions about published physical therapy research are not available, although some relevant studies[69,70] have been conducted. These studies indicate that the relative frequency of studies using group-experimental designs ranges from 10.7% in Physiotherapy Catzada, Physical Therapy, and Physiotherapy combined (data reworked[70]) to 41% in Physiotherapy Canada.[69] Another study[71] showed that the relative frequency of "experimental trials" increased from 19% over the period 1980 to 1984 to 27% over the period 1985 to 1989 in articles from the journals Physiotherapy, Physical Therapy, Physiotherapy Canada, Australian Journal of Physiotherapy, and South African journal of Physiotherapy. The category definitions provided in that study do not make clear how clinical studies using group designs would be classified and apparently exclude methods such as single-subject experimental designs and series of case studies. The definitions provided in other studies[38,72,73] do not differentiate group-experimental designs from quasi-group-comparison designs and other data-based research methods. The conclusions of some studies of research methods are limited by the lack of details of the source or the method of selection of the articles examined[38 69,71] and by the reliance on a convenience sample of articles[69,70,72] published over a period of relatively short duration (10 years maximum) in a context in which the relative frequency of the different types and topics of articles can vary markedly from year to year.

My study had three specific aims: (1) to establish the frequency of use of group-experimental designs in a longitudinal sample of articles from the physical therapy literature, (2) to examine the extent to which the same sample of articles investigated clinical treatments, and (3) to investigate the extent to which clinical research was based on methods differing from those used to examine nonclinical issues.

Method

Literature Sample

In this study, I analyzed a sample of 272 articles published from 1954 to 1993. The sample was derived from Physical Therapy, a physical therapy journal described in its purpose statement as "a scholarly, refereed journal that contributes to and documents the evolution and expansion of the scientific and professional body of knowledge in physical therapy."[74]

Physical Therapy was for many years the sole physical therapy journal published monthly and indexed in Index Medicus and the Science Citation Index as well as in physical therapy indexes (eg, Physiotherapy Current Awareness Topic Search, Physiotherapy Index). Physical Therapy also has a growing impact factor[75] and is cited more frequently within the profession than any other physical therapy journal.[76,77]

Articles published in Physical Therapy prior to 1954 were not included in this sample for three reasons. First, the number of articles published per year prior to 1954 was small, especially during World War 2. Second, by 1954, the journal had changed to its current monthly publication rate (in 1949). Third, a four-decade period was expected to provide a sufficiently long time frame to reliably indicate trends in the aspects of the articles investigated in this study.

The sample of 272 articles consisted of all 106 articles indexed in Physical Therapy on the topic "knee," all 79 articles indexed on the topic "back," and all 87 articles indexed on the topic "electrical stimulation" between 1954 and 1993. The topics were chosen by establishing which indexed topics included the largest numbers of articles, excluding commentaries, author responses to commentaries, and letters to the editor, published in Physical Therapy from 1954 to 1993. When the topics "professional issues," "professional education," and "neuro-physiology/neuroanatomy" were excluded, the topics of direct relevance to clinical practice and with the largest numbers of articles were the knee, the back, and electrical stimulation. Given the system of cross-referencing used by Physical Therapy, each of these topics included articles on subjects as diverse as anatomy, equipment, and evaluations of treatments. Having two automically defined areas and a treatment modality broadened the basis for the study and optimized clinical relevance. The prominence of an electrical modality among the different manual and electrical treatment techniques and types of exercise indexed was consistent with the findings of an earlier survey of the treatment most commonly provided by physical therapists.[78] If a particular article was indexed under two or more of these three topic areas (ie, back, knee, or electrical stimulation), it was included in each area for analysis of the topic.

Classification Systems

Two systems were used to categorize each of the 272 articles. The first classification system was based on the type of method used in each article. The eight method categories used, shown in Table 1, were (1) group experimental, (2) quasi-group comparison, (3) single-subject design, (4) practical, (5) descriptive, (6) survey, (7) case study, and (8) correlational.

[TABULAR DATA OMITTED]

The classification of the method used in each article was based on a system described by Ottenbacher and Short,[79] who surveyed 694 research articles published from 1970 to 1980 in the Occupational Therapy Journal of Research. These authors subsequently tested the reliability of their categorizations and obtained an interobserver agreement of .94 using a coefficient of agreement. The wording of their seven-category definitions was modified in my study in an attempt to increase the precision and the relevance to physical therapy. I also included an eighth category called "single-subject design" to enable the differentiation of single-subject experimental designs from case studies, group-experimental designs, and quasi-group-comparison designs.

The second classification system was developed to distinguish articles dealing with treatment from those that did not deal with treatment (Tab. 2). The criteria were based on existing distinctions between aspects of clinical and nonclinical practice and research drawn within the physical therapy and psychology literatures.[70,80,81] The starting point was categories used by Holliday[70] in her analysis of research contributions made by physical therapists. The common element in the categories Holliday called "clinical surveys" and "clinical trials" was "treatment in a clinical context."[70] The term "in a clinical context" was too general, however, to provide a sufficient distinction between clinical treatment and nonclinical studies, as it did not require that the treatment was used to effect a beneficial outcome for a patient. A treatment could equally well have been tested on asymptomatic subjects.

[TABULAR DATA OMITTED]

Based on distinctions drawn by Goldiamond,[80] clinical treatment studies were differentiated from nonclinical studies by the requirement that they reported an aim of maintaining or improving the function of patients receiving the treatment. Support for this distinction is found in a guest editorial in Physical Therapy entitled Reliability Discussion Required."[81] The editorial emphasized the importance of selecting a sample that represents the population that the treatment being investigated is intended to affect and that using unimpaired volunteers when clinically useful information is being sought is inappropriate. That is, subjects can only be said to be participating in clinical studies of treatment if they represent at least some part of the population that treatment is intended to benefit.

Category Assignment

I read each of the 272 articles in the sample and assigned each article to one of the eight method categories. Following a second reading, I then categorized each article as either clinical treatment or nonclinical.

Reliability

I checked the reliability of both category assignments of the part of the sample dated 1954 to 1990 at least 1 year after the initial analyses. Twenty-seven of these articles, slightly over 10% of that section of the sample, were checked. Using a table of random numbers, articles were selected from each topic subset in turn. The initial assignments of each article to a particular method category and to either the clinical category or the nonclinical category were then compared with the blinded reassignments of each article. A coeffecient of agreement was used. The resultant level of agreement was 1.00 for the classification of articles as clinical or nonclinical and .93 for the methods. This level of agreement for the methods is comparable to the .94 value obtained by Ottenbacher and Short,[79] the originators of the methods classification system used in this study, I categorized each article in the remainder of the sample (ie, articles published 1991-1993) and then recategorized each article after a 1-week interval to confirm the reliability of the classification.

Results

Figure 1 shows the percentage of articles describing studies using group designs, group-experimental and quasi-group-comparison articles combined, in each 10-year period examined (ie, 1954-1963, 1964-1973, 1974-1983, 1984-1993). Figure 1 shows that an increased percentage of articles in the sample described the use of group designs in each successive 10-year period. In the most recent period (ie, 1984-1993), group designs were used in the studies reported in 53% of all articles in the sample. By contrast, group designs were used in studies reported in only 12% of the articles from the period 1954 to 1963.

Figure 2 shows the number of clinical treatment articles (top graph) and nonclinical articles (lower graph) in each of the successive 10-year periods examined. Figure 2 also shows the methods used in both the clinical treatment and nonclinical categories. The figure shows that increases have occurred in the total numbers of articles published and indexed per decade in each topic area since 1954.

Figure 2 shows that of the 272 articles in the sample, 187 (69%) were categorized as nonclinical and 85 (310/o) as clinical treatment. In the initial period examined, 1954 to 1963, more than half (55%) of the articles described studies that investigated clinical treatment. Since then, increasingly more nontreatment articles have been published. In 1964 to 1973, 57% of the articles were nonclinical; in 1974 to 1983, 80% were nonclinical. In the most recent period examined, 1984 to 1993, 700/o of the articles were classified as nonclinical. In 1973, the cumulative number of nonclinical articles in the sample finally equaled that of the treatment articles. Since 1973, the cumulative number of nonclinical articles in the sample has always exceeded the cumulative number of clinical treatment articles.

Figure 2 also shows a marked increase has occurred in the use of both types of group designs (ie, group-experimental and quasi-group-comparison designs) in both the clinical treatment and nonclinical categories. In the initial period examined, 1954 to 1963, group designs were used in studies reported in 11% of the clinical articles and in 13% of the nonclinical articles. The use of group designs increased steadily in the following two decades, and, in the most recent period examined (ie, 1984-1993), 61% of the studies reported in the clinical treatment articles and 50% of the studies reported in the nonclinical articles used group designs.

The studies reported in the nonclinical articles, as shown in Figure 2, used both group-experimental and quasi-group-comparison designs in addition to descriptive, survey, practical, and correlational methods. Nonclinical articles did not report the use of case studies or single-subject experimental designs. By contrast, clinical articles described the use of case studies, single-subject experimental designs, or either of the two types of group designs.

Figure 2 shows changes over time in the number of articles reporting the use of alternative methods. The use of case studies was reported in 49% of an sampled articles in 1954 to 1963 and in 100/o of the articles from 1984 to 1993. The percentage of descriptive studies dropped steadily from 27% in 1954 to 1963 to 6% in 1984 to 1993, except during the decade 1974 to 1983 when it rose to 42%. Special issues on both the back (August 1979) and the knee (December 1980) were published during this decade and included a large number of articles classified as descriptive. These special issues resulted in a marked inflation for the decade of the number of descriptive articles, which are, by definition, nonclinical articles. Two methods, single-subject designs and correlational studies, were used only in the most recent decade surveyed, 1984 to 1993. Small numbers of articles were involved; for example, only two articles described the use of single-subject designs. No clear trends are apparent in the use of the remaining alternative methods.

Figures 3 and 4 both illustrate the contribution of the individual topics (ie, knee, back, and electrical stimulation) comprising the sample of articles examined in this study. Figure 3 shows that group designs (ie, group-experimental and quasi-group-comparison designs) were used more frequently than any other methods in the articles on electrical stimulation (56%) and second most frequently in articles on the knee (30%) and the back (27%). More articles on the back (29%) and the knee (27%) were descriptive. Figure 3 also shows that the articles on electrical stimulation reported the use of few surveys and more case studies than articles on either the back or the knee.

Figure 4 shows the percentage of all articles on each topic categorized as clinical treatment in each of the successive 10-year periods investigated. The figure shows that in only five instances (ie, for the knee and back in 1954-1963, for the back and electrical stimulation in 1964-1973, and for electrical stimulation in 1974-1983) has the percentage of clinical treatment articles in any topic area reached or exceeded 500/o. In the most recent period examined (ie, 1984-1993), fewer than 300/o of the articles on the knee and back and 40% of the articles on electrical stimulation were clinical treatment articles.

Discussion

Over the four-decade period examined (ie, 1954-1993), an increasing percentage of articles on all three topics analyzed used group designs, and the majority of articles on each topic were nonclinical. The context of this finding is the literature of a clinical profession that actively and positively promotes research. The profession expects research to validate[16] and to further develop practice, and to increase its own scientific credibility.[13] These findings, however, suggest that the direct validation of practice via research may not be occurring as frequently as many believe and would like. These findings also suggest that a concurrent narrowing, rather than a widening, has occurred in the choices of acceptable research methods.

These findings must be put in perspective. Three of a large number of indexed topics were analyzed (ie, knee, back, electrical stimulation), and all articles on these topics came from the same physical therapy journal. The results showed some differences among the three topics in both the methods used and the percentage of clinical treatment articles. The results, however, were similar across all three topics: a general increase in the use of group designs and a clearly lower percentage of clinical treatment articles than nonclinical articles. Further, the number of articles published in and since 1954 on each of these particular topics shows they are of major interest to physical therapists and of major importance in the profession. Each of these three topics has direct relevance for practitioners in what is essentially a clinical profession. Further, all articles came from Physical Therapy, arguably the most influential and enduring English-language journal in the profession as, among other things, it is the only journal in the profession that both is indexed in major medical, scientific, and physical therapy indexes and has been published on a monthly basis for decades.

Arguably, the broader relevance of these findings appears to rest on further analyses of other major topics within the different parts of the physical therapy literature. That is, the generalizability of the findings of this study would appear to be limited until replications of the same procedures on articles indexed on other topics and from other parts of the physical therapy literature are undertaken. If subsequent investigations show the findings of this study apply to only the three topics examined and to the journal Physical Therapy, these findings would appear to be of limited interest. These findings, however, would remain relevant and difficult to easily dismiss. The findings of this study are generally consistent with what is said about research in physical therapy. Further, the articles examined in this study comprise a central portion of an important and influential body of the physical therapy literature and concern topic areas of major importance in the profession. These three sets of articles, therefore, likely reflect the more general trends and changes in the research paradigms acceptable within the profession, and possibly also influence them. Articles on topics of lesser general interest and in some of the smaller bodies of physical therapy literature are less likely to reflect general trends, or to influence them. Further studies of both different topics in and different parts of the physical therapy literature would indicate the extent of generalizability of these findings. Further studies would also extend the degree of reliable reflection possible within the profession on trends in relevant published research.

The finding of an increasing reliance on group designs in all three topic areas is consistent with what I believe is the extent of support in the profession for the method. Discussions of research and research methods in the different journals and books produced for the physical therapy market tend to advocate use of group designs. Implicit support for group designs is found in the greater space given them in many of the books and articles that describe how to do[]1,2,4,28-33] and to evaluate[34,35] research and in the unstated presumptions in many discussions of research that a group design will be used. Whether these findings are primarily a function of editorial policies in the different physical therapy journals, or whether they directly reflect the beliefs and practices of readers of and contributors to those journals, is beyond the scope of this study. This finding is, however, consistent with an apparently high level of support for a particular approach to research in much of the physical therapy literature.

The finding of an increasing reliance on group designs in both clinical treatment and nonclinical studies is, however, paradoxical given existing criticisms of the method. Physical therapy is a clinical profession. Many criticisms of group designs apply particularly when the method, not originally designed for clinical research, is used in a clinical context.[1,25,26,36-42] For example, criticisms of group designs concern the difficulties in obtaining sufficiently large and appropriate patient samples and the risks of compromises that may be made.[1,36-41] other criticisms concern difficulties in maintaining experimental control in the clinical environment and in overcoming the ethical problems consequent upon the requirement that treatment is allocated according to the research design rather than to the patients' needs. Yet other criticisms question the relevance to practitioners, concerned with individuals, of findings based on averaged group data.36,42 The relevance for physical therapy of these criticisms of group designs becomes apparent when the clinical nature of the profession, the context of the research, and the heavily clinically oriented benefits expected from research are remembered.

Little attention has been paid in the physical therapy literature to the role a research method may play in research outcomes. Some contributors to the physical therapy literature have suggested that the choice of problems investigated might be constrained by amenability to statistical analysis.[48,56] Inferences based on existing criticisms of group designs would support this view. The difficulties inherent in investigating questions where sufficiently large and homogeneous patient treatment and nontreatment or alternative treatment groups are a requirement of the method quite likely do have implications for what is researched in physical therapy. That is, many questions from within the practice domain are possibly not amenable to research using group designs. Further, problems arise when the likely outcomes of the research using group designs in the clinical context are considered. A consequence of the demonstrated low power of many clinical studies using group designs may be the unwarranted acceptance of the null hypothesis investigated.[52] Research conducted with larger groups of more homogeneous patients, or research using different measures or even a quite different research method, might produce a quite different outcome. That is, the selection of a particular research method has implications for the findings that may be obtained as well as for the questions that can be investigated.

Alternative research methods to group designs are available. Among those methods explicitly designed for use in the clinical context are single-subject experimental designs and series of case studies. In this study, however, I found few single-subject experimental designs and increasingly fewer case studies in the topics examined. This finding is not explained by editorial policy. Editors of all major physical therapy journals have published requests for[57,59] or advice on how to use[60-63] these methods. A range of different methods have also been used in the physical therapy literature. The question of why proportionately so few studies in the sample used alternative methods remains. The importance of this question lies in the role a research method can play in establishing which questions are investigated, in the possible research outcomes, and in the expectations in physical therapy that research will benefit practice.

The second major finding of this study was that most studies reported in the sample of articles examined did not investigate clinical treatments. Clinical practice is central to physical therapy, and physical therapists expect research to benefit clinical practice. Most surveyed articles, however, did not report or evaluate the treatment techniques or regimens used in physical therapy to maintain or improve patient function. Instead, most articles described procedures, apparatuses, or techniques used on unimpaired subjects who explicitly did not require treatment to improve their functioning or provided information or discussed issues of interest to physical therapists.

The prevalence of nonclinical articles is not explained by the sample selected. All three constituent topics (ie, knee, back, and electrical stimulation) are highly relevant to practice. All articles are cross-indexed and include either areas often treated or methods often used for treating patients, Further, Physical Therapy actively solicits articles on topics of clinical relevance and interest. For example, in the period investigated, the journal's highest priorities included the publication of clinically relevant research[57] and studies on the efficacy of therapeutic interventions.[81] Additionally, during that period, the current editor of the journal on more than one occasion explicitly solicited research articles and case reports relevant to clinicians.[7] Although requests by the current editor for articles could not have affected most of the articles analyzed here, they are indicative of the emphasis the journal has placed over time on clinical articles.

The trend toward relatively fewer clinical treatment articles cannot be explained by the advent of "new" journals. The newer physical therapy journals, those introduced since 1970, commenced publication on a relatively infrequent basis, and the earliest of them, the Journal of Oribopaedic and Sports Physical Therapy, was not published until 1979. By contrast, the trend toward a higher relative frequency of nonclinical articles was well established by the 1980s, to the extent that their cumulative frequency in the sample always exceeded that of the clinical articles published after 1973. That is, the trend away from clinical treatment research was well established prior to the advent of the spate of "new" physical therapy journals.

Questions arise about the clinical treatment and nonclinical classification system. In particular, given the lack of precedents for such an analysis in physical therapy, does the classification system used enable meaningful and reliable distinctions? Distinctions between clinical treatment and nonclinical articles were based on those existing within the physical therapy and psychological literatures, appear valid, and were shown to be reliable. Care, however, must be taken in interpreting the findings. In particular, the division between clinical treatment and nonclinical articles does not imply that nonclinical articles have no relevance for clinical practice. Clearly, many nonclinical articles describe techniques used by clinicians or the anatomy of a region, or they report

Figures 1 to 4 [ILLUSTRATION OMITTED] the testing of measurement or treatment procedures subsequently used and even investigated clinically. These articles, however, do not report on the implementation of treatments. Their subsequent or potential clinical usefulness would be very difficult to establish and would involve a number of very questionable assumptions. The distinction between nonclinical and clinical treatment articles was specifically made in this study to establish the extent to which a carefully selected sample of physical therapy literature explicitly investigated clinical treatments. As reported, even though the profession is primarily clinical and most expected outcomes from research relate to improving and validating practice, most articles examined were nonclinical.

The findings of this study, together with the increasing body of Criticism of group designs and the awareness in the profession of the need for research into clinical treatments, suggest more extensive consideration of and reflections on the research published within the physical therapy literature are needed. For example, is the reliance on group designs found in this study also found in the more extended physical therapy literature Why are apparently so few researchers using alternatives to group designs, or having them published Do the findings in this study of apparently so few studies of clinical treatments also apply to other topic areas? If so, why, and how can more research into clinical treatments be actively promoted and published within the profession's literature?

Conclusions

The findings of this study question whether the benefits expected to accrue from research to physical therapy can be realized, given current research practices and emphases. More physical therapists need to be aware of the importance of researching our clinical treatments and the range of research methods available.

References

[1] Currier DP. Elements of Research in Physical Therapy. 3rd ed. Baltimore, Md: Williams Wilkins; 1990. [2] Makrides L, Richman J. Research methodology and applied statistics, part 1: general principles and basic concepts. Physiotherapy Canada. 1980;32:135-139. [3] Michels E. Design of Research and Analysis of Data in the Clinic. Alexandria, Va: American can Physical Therapy Association; 1985. [4] Payton OD. Research: The Validation of Practice. 2nd ed. Philadelphia, Pa: FA Davis Co; 1988. [5] Schlapbach PB, Gerber NJ, eds. Physiotherapy: Controlled Trails and Facts. Basel, Switzerland: S Karger AG, Medical and Scientific Publishers; 1991. [6] Twomey L. A growing commitment to research and evaluation. Australian Journal of Physiotherapy. 1990;36:83. [7] Rothstein JM. Editor's note: The Journal--past, present, and future. Phys Ther. 1989;69: 795-796. [8] Physiotherapy research. Physiotherapy. 1992;78:356-357. Policy statement. [9] Dean E. A vision of Physiotherapy Canada: All things to all people. Physiotherapy Canada. 1993;45:5-6. [10] The Association of Chartered Physiotherapists in Research. Physiotherapy. 1983;69:317. [11] Miles-Tapping C, Dyck A, Brunham S, et al. Canadian therapists' priorities for clinical research: a Delphi study. Phys Ther. 1990;70:448-454. [12] Rose SJ. Editor's note: Gathering storms. Phys Ther. 1989;69:354-355. [13] Gielen F. Discussion of placebo effect in physiotherapy based on a noncritical review of die literature. Physiotherapy Canada. 1989;41:210-216. [14] List M. Physiotherapy: a mobile profession in health care. Physiotherapy. 1986;72: 122-124. [15] Morrissey MC, Kanda LT, Brewster CE. Development of a clinical physical therapy research program. Phys Ther. 1987;67: 1110-1114. [16] Cooper BA, Saarinen-Rahikka HK. Interrelationships of theory, clinical models and research. Physiotherapy Canada. 1986; 38:97-100. [17] Ottenbacher KJ. Evaluating Clinical Change: Strategies for Occupational and Physical Therapists. Baltimore, Md: Williams Wilkins; 1986. [18] Zimny NJ, Tandy CJ. Problem-knowledge coupling: a tool for physical therapy clinical practice. Phys Ther. 1989;69:155-161. [19] Mathews JS. The 1987 Presidential Address: Towards a profession of substance. Phys Ther. 1987;67:1723-1725. [20] Wood R. Twenty-third Mary McMillan Lecture: Footprints. Phys Ther. 1989;69: 975-980. [21] Stockham MJ. Are physiotherapists necessary. Physiotherapy. 1988;74:499-501. [22] Grieve E. A fair trial for physiotherapy: the controlled trial. Physiotherapy. 1993;79:304. [23] Jackson DA. Where is the physiotherapy profession going? Physiotherapy. 1987;73: 590-591. [24] Norton BJ, Strube MJ. Some cautionary comments on the use of retrospective designs to evaluate treatment efficacy. Phys Ther. 1988;68:1374-1377. [25] Norton BJ, Strube MJ. Making decisions based on group designs and meta-analysis. Phys Ther, 1989;69:594-600. [26] Sim J. Methodology and morality in physiotherapy research. Physiotherapy. 1989;75: 237-243. [27] Van der Linden S, Bouter L, Tugwell P. What are the minimal methodological and statistical requirements of a good trial? In: Schlapbach PB, Gerber NJ, eds. Physiotherapy: Controlled Trials and Facts. Basel, Switzerland: S Karger AG, Medical and Scientific Publishers; 1991:2. [28] Makrides L, Richman J. Research methodology and applied statistics, part 6: ethics in human research. Physiotherapy Canada. 1981;33:89-94. [29] Prince B, Makrides L, Richman J. Research methodology and applied statistics, part 2: the literature search. Physiotherapy Canada. 1980;32:201-206. [30] Richman J, Makrides L, Prince B. Research methodology and applied statistics, part 3: measurement procedures in research. Physiotherapy Canada. 1980;32:253-257. [31] Richman J, Makrides L. Research methodology and applied statistics, part 4: overcoming statistics anxiety. Physiotherapy Canada. 1980;32:321-329. [32] Makrides L, Richman J. Research methodology and applied statistics, part 5: experimental design. Physiotherapy Canada. 1981;33:6-14. [33] Makrides L, Richman J. Research methodology and applied statistics, part 7: writing the research proposal. Physiotherapy Canada. 1981;33:163-168. [34] Domholdt EA, Malone TR. Evaluating research literature: the educated clinician. Phys Ther. 1985;65:487-491. [35] APTA Committee on Research: Guidelines on critically considering research papers. Phys Ther. 1988;68:555-556. [36] Gonnella C. Designs for clinical research. Phys Ther. 1973;53:1276-1283. [37] Lehmkuhl LD. Mixing one part common sense with each part statistics in planning the design and reporting the results of clinical research in Physical Therapy. Phys Ther. 1987;67:1851-1853. [38] Riddoch J, Lennon S. Evaluation of practice: the single-case study approach. Physiotherapy Theory and Practice. 1991;7:3-11. [39] Shepard KF. Qualitative and quantitative research in clinical practice. Phys Ther. 1987;67:1891-1894. [40] Deyo RA, Walsh NE, Schoenfeld LS, Ramamurthy S. Can trials of physical treatments be blinded? Am J Phys Med Rehabil. 1990; 69:6-9. [41] Partridge C. The myth of the medical model. Physiotherapy. 1982;68:329. [42] Rosnow RL, Rosenthal R. Statistical procedures and the justification of knowledge in psychological science. Am Psychol. 1989; 44:1276-1284. [43] Gore SM, Jones IG, Rytter EC. Misuse of statistical methods: critical assessment of articles in BMJ from January to March 1976. BMJ. 1977;1:85-87. [44] Vrbos LA, Lorenz MA, Peabody EH, McGregor M. Clinical methodologies and incidence of appropriate statistical testing in orthopaedic spine literature. Spine. 1993;18: 1021-1029. [45] Grichting WL. Psychology and sociology in Australia: the published evidence. Australian Psychologist. 1989;24:115-126. [46] Huxley P. Statistical errors in papers in the British Journal of social Work. Soc Work. 1986;16:645-658. [47] Carver RP. The case against statistical significance testing. Harvard Educational Review. 1978;48:378-399. [48] French S. How significant is statistical significance? A critique of the use of statistics in research. Physiotherapy. 1988;74:266-268. [49] Harcum ER. The highly inappropriate calibrations of statistical significance. Am Psychol. 1989;44:964-965. [50] Oakes M. Statistical Inference: A Commentary for the Social and Behavioural Sciences. Chichester, England: John Wiley & Sons Ltd; 1986. [51] Rothstein JM. Editor's note: Much ado about probability. Phys Ther. 1990;70:535-536. [52] Matyas TA, Ottenbacher KJ. Confounds of insensitivity and blind luck: statistical conclusion validity in stroke rehabilitation clinical trials. Arch Phys Med Rehabil 1993;74: 559-565. [53] Barlow DH. On the relation of clinical research to clinical practice: current issues, new directions. J Consult Clin Psychol. 1981;2: 147-155. [54] Ottenbacher KJ. The significance of power and the power of significance: recommendations for occupational therapy research. Occupational Therapy Journal of Research. 1984; 4:37-50. [55] Ottenbacher KJ. Clinically relevant designs for rehabilitation research: the idiographic model. Am J Phys Med Rehabil. 1990;69: 286-292. [56] Tammivaara J, Shepard KF. Theory: the guide to clinical practice and research. Phys Ther. 1990;70:578-582. [57] Lister MJ. Editor's note: Farewell. Phys Ther. 1988;68:183-184. [58] Rothstein JM. Editor's note: Clinical literature. Phys Ther. 1989;69:895-896. [59] Rothstein JM. Editor's note: The case for case reports. Phys Ther. 1993;73:492-293. [60] Cole P. How to ... single-subject designs for clinical and research purposes. Australian Journal of Physiotherapy. 1991;37:127-128. [61] Gonnella C. Single-subject experimental paradigm as a clinical decision tool. Phys Ther. 1989;69:601-609. [62] Riddoch J, Lennon S. Single-case research designs. Physiotherapy Theory and Practice. 1991;7:1. Editorial. [63] Robertson VJ, Lee VL. Some misconceptions about single-subject designs in physiotherapy. Physiotherapy. 1994;80:762-766. [64] Jensen GM. Qualitative methods in physical therapy research: a form of disciplined inquiry. Phys Ther. 1989;69:492-500. [65] Schmoll BJ. Ethnographic inquiry in clinical settings. Phys Ther. 1987;67:1895-1897. [66] Stonecipher DR, Catlin PA. The effect of a forearm strap on wrist extensor strength. J Orthop Sports Phys Ther. 1984;6:184 -189. [67] Michels E. Using and Understanding Surveys: An Introductory Manual. Alexandria, Va: American Physical Therapy Association; 1985. [68] Portney LG, Watkins MP. Foundations of Clinical Research. East Norwalk, Conn: Appleton Lange; 1993. [69] Aufdemkampe G. Some comments on single-case studies. Physiotherapy Theory and Practice. 1991;7:63-71. [70] Holliday PJ. A survey of research contributions of physiotherapists. Physiotherapy Canada. 1981;33:372-376. [71] Mitchell J. Research in the 1980s: world trends in physiotherapy research. Physiotherapy Theory and Practice. 1993;9:171-176. [72] Dean E. Physiotherapy Canada survey (1973-1982): a trend to research. Physiotherapy Canada. 1985;37:158-161. [73] Jones D, Minichiello V. A survey of gerontology articles in physiotherapy journals. Australian Journal of Physiotherapy. 1991;37: 155-161. [74] Instructions for authors, Phys Ther, 1993; 73:58. [75] Bohannon RW. Importance of Physical Therapy grows. Phys Ther. 1988;68:584. Letter to the Editor. [76] Bohannon RW. Physiotherapy Practice. 1989;5:99-100. Letters page. [77] Bohannon RW. Core journals of physiotherapy. Physiotherapy Practice. 1987;3: 126-128. [78] Dennis JK. Decisions made by physiotherapists: a study of private practitioners in Victoria. Australian Journal of Physiotherapy. 1987; 33:181-191. [79] Ottenbacher KJ, Short MA. Publication trends in occupational therapy. Occupational Therapy Journal of Research. 1982;2:80-88. [80] Goldiamond I. Protection of human subjects and patients. Behaviorism. 1976;1:1-41. [81] Lamb RL, Bohannon RW, Craik RL, et al: Guest editorial: Reliability discussion required. Phys Ther. 1987;67:501.

Invited Commentary

Dr Robertson provides us with much useful information about selected publications--including their methods and clinical content--that have appeared in Physical Therapy over the last 40 years. She and I differ, however, in our views of the framework into which these data should he placed and in the conclusions that we draw from them. In this commentary, then, I will place a different spin on the data that Dr Robertson has carefully collected and reported.

Clinical/nonclinical Conclusions

One of Dr Robertson's major findings is that only 31% of the articles in her sample investigated clinical treatments. My first point related to the clinical/ nonclinical conclusions drawn in this study concerns Dr Rohertson's view of the purpose of research. The finding of only 31% of clinical articles was viewed as problematic by Dr Robertson because such a low proportion of clinical treatment articles seems unlikely to fulfill a mission of research that is "expected to demonstrate and improve the effectiveness of physical therapy techniques, interventions, and modalities." I find this to be a limited and somewhat unrealistic view of research. Because I have not examined the citations that Dr Robertson used to substantiate the above quote, I am uncertain whether my debate is with Dr Robertson's interpretation of these sources or with the sources themselves. In any event, this view that research will "demonstrate effectiveness" presupposes that physical therapy is effective and that a major purpose of research is to document for others what we, as physical therapists already convinced of our worth to individuals and the health care system, already know. If this, indeed, is the purpose of research, then a higher proportion of clinical treatment studies would be needed to achieve its aims.

I believe, however, that the purpose of research is broader, that its purpose is to "inform" the practice of physical therapy. This view does not presuppose that research findings will demonstrate the effectiveness of our techniques; it allows for the possibility that some of our techniques will not stand up to the scrutiny of the research process. In addition, a goal of "informing" practice gives legitimacy to research articles that do not relate directly to clinical treatments as applied to individual patients or groups of patients--articles, for example, that present treatment models for low back pain,[1] that examine electrical stimulation variables in asymptomatic subjects,[2] or that document knee flexion contractures in institutionalized elderly individuals.[3] These three articles were published in Physical Therapy in either 1992 or 1993 and presumably were included in Dr Robertson's study and classified as "nonclinical" articles, even though they clearly have something to add to our clinical view of back care, electrical stimulation, and knee dysfunction.

A second point related to the clinical/ nonclinical conclusion in this study concerns the purpose of Physical Therapy. The mission statement published in each issue of Physical Therapy includes a description of Physical Therapy as "a scholarly, refereed journal that contributes to and documents the evolution and expansion of the scientific and professional body of knowledge related to physical therapy." This statement is very broad. According to Haynes, a physician who writes widely about practitioner use of medical literature, journals with such broad purposes "impede the dissemination of validated advances to practitioners by mixing a few rigorous studies (communications from scientists to practitioners) with many preliminary investigations (communications from scientists to scientists)." [4(p724)] He sees the major audiences for clinical journals to be scientists and practitioners, and the possible forms of communication as (1) from scientist to scientist, (2) from scientist to practitioner, (3) from practitioner to practitioner, and (4) from practitioner to scientist. He argues that reports of preliminary studies are communications from scientist to scientist and that "this messy process of reporting findings that will be invalidated later is essential to the advance of science, but is a hazard for clinicians and their patients." [4(PP724 -725)]

In Haynes' framework,4 then, preliminary studies (probably the majority of clinical treatment studies published in Physical Therapy) are essential to the scientific process, but are not really meant for use and application by clinicians. In addition, Haynes views case reports as communications from practitioners to scientists because "their methods do not permit discrimination of the valid from the interesting but erroneous, and they cannot provide a sound basis for clinical action. Case reports are, however, a fertile source of hypotheses that could lead to systematic observations. Their publication in journals dedicated to communications to scientists is therefore warranted."[4(p726)] It is interesting that two forms of inquiry that are seen by Dr Robertson as "clinical" are seen by Haynes as existing to inform scientists, rather than clinicians. In addition, Haynes views reports of rigorous trials (scientist-to-clinician communications) and review articles (practitioner-to-practitioner communications) as those that should form the basis for journals with a clinical audience. Only one of these two types of articles would be considered a "clinical treatment" article by Dr Robertson. Although I am not completely convinced of the validity of Haynes' arguments, his work shows that the question of whether our literature provides enough of a clinical focus is indeed a complex issue that cannot be answered by a single percentage of studies that meet certain criteria.

A third aspect of the clinical/nonclinical controversy relates to the development of physical therapy education and research in relation to the practice of physical therapy in the United States. Physical therapy education programs in the United States are located in virtually every type of post-secondary academic institution, many of which have no formal connections with research-oriented health care centers. This frequently leads to poor access to patient populations for researchers and to limited networks with clinicians who could assist with the implementation of clinical treatment studies. Because scholars in physical therapy are concentrated in the academic setting, and because many academic settings are isolated from clinical settings, it is not surprising that the proportion of clinical treatment studies is low. Contrast the situation in physical therapy with that of medicine in the United States. The typical medical school is located within an academic health center that includes a major teaching hospital and a concentration of biomedical research resources. This provides medical researchers with convenient access to practitioners and patients for the study of questions of a clinical nature. I am not suggesting that physical therapy education necessarily follow the medical center model, but it does seem that physical therapy educators, scholars, and practitioners should consider ways to strengthen ties between the clinic, the laboratory, and the academy. I would be interested in hearing from Dr Robertson how these ties are developed within the physical therapy community in Australia.

Group/Nongroup Conclusions

Dr Robertson found that group designs (including both group-experimental articles and quasi-group comparison designs) were used in 53% of the articles in her sample. She implies that this percentage represents an overreliance on one method, an overreliance consistent with an overemphasis on group designs in our literature about research.

My first point about the group/nongroup conclusions relates to Dr Robertson's citations from the literature about physical therapy research. She cites numerous references to support her supposition that the literature of the profession "focus[es] almost exclusively on group-experimental designs." In support of this contention, she lists 24 physical therapy-based references with explicit or implicit support for group comparisons. Of these, 10 are from the early 1980s or earlier and 14 are from the late 1980s or early 1990s. Next, she lists many published criticisms of group designs, many of which have appeared in the physical therapy literature. Of the 16 physical therapy-based references, 13 are from the late 1980s and early 1990s and only 3 are from the early 1980s or earlier. Thus, if Dr Robertson's own references are representative, it would appear that the physical therapy literature provides a balanced view of the strengths and weaknesses of many different research designs. In addition, her citations about less traditional designs tend to have recent publication dates, indicating an evolution of methodological thinking over the last 20 years.

My second point about the group/ nongroup conclusions relates to the rate at which scientific advances are reflected in the research literature. Altman and Goodman[5] have investigated the rate at which statistical innovations are incorporated into the literature of medicine. They found that it often took 10 to 25 years for new statistical techniques to reach an appreciable mass in the medical literature. Although this example of transfer of statistical technology may not be directly generalizable to the transfer of research design methods, Altman and Goodman's work still provides a basis for judging whether new research methods are appearing in our literature at a reasonable pace. Dr Robertson's references about single-subject research methods, for example, have publication dates from 1989 to 1994. In addition, a major rehabilitation text on single-subject designs was published in 1986.[6] The presence of single-subject designs at all in Physical Therapy during the 1984 to 1993 period of study suggests to me that there has been an extraordinarily rapid transfer of technology from articles that discuss the method to articles that report on studies using the method.

My third point related to the group/ nongroup conclusions concerns the nature of scientific change. Kuhn's classic work on scientific change, The Structure of Scientific Revolutions,[7] provides some insight for physical therapists. Although Kuhn's work examines great theoretical shifts in scientific thought (including such revolutions as the change from a geocentric model of the universe, in which the sun was thought to revolve around the earth, to our current heliocentric model, in which we believe that the earth revolves around the sun), his model may also be used to view methodological changes within a discipline.

In brief, Kuhn's model of scientific change can be viewed as having four stages.[8] The first, pre-paradigm, stage is characterized by little agreement on research methods and disagreement as to which problems are important to the discipline. The preponderance of case reports and nongroup designs in the physical therapy literature in the 1950s and 1960s may represent a form of pre-paradigm science. Kuhn's second stage, normal science, is characterized by widespread agreement on methods, instrumentation, and design. The increasing reliance on group designs in the 1970s, 1980s, and into the early 1990s may represent a normalization of science within physical therapy, realized in the 1980s and early 1990s by widespread use of group methodologies. Kuhn's third stage, the crisis period, is characterized by the recognition that the normal science of the discipline has failed to answer all the important questions. In this stage, scientists begin to criticize the usual way of doing things and propose new ways of developing knowledge within the discipline. This may be happening in physical therapy research now, as shown by the citations of the late 1980s and early 1990s that have become critical of group designs. The appearance of single-subject designs and the increase in the number of case reports in the 1984 to 1993 time period (compared with the previous two time periods) show that researchers are willing to experiment with ways of knowing that are outside of the traditional group design paradigm. The final stage in Kuhn's model of scientific change is acceptance of a new paradigm for the discipline. Although it is my opinion that the science of physical therapy is in the early stages of embracing challenges to the group design methodology, it seems clear that we are a long way from fashioning a consensus on a new research paradigm.

I appreciate the opportunity to comment on this study. I hope that Dr Robertson's data are discussed widely among physical therapy researchers and by the editors and editorial boards of journals within the profession. I believe that the meaning of her results can be enhanced by careful consideration of the purposes of research, examination of the purpose of various journals within physical therapy, consideration of the connections (or lack of connections) between physical therapy practice and research, and by taking a longitudinal view of methodological research citations within the profession, understanding the time lag that is typical for transfer of literature about research methods to actual research reports using the methods and placing the data from Physical Therapy into the context of established models of scientific change.

References

[1] Derosa CP, Porterfield JA. A physical therapy model for the treatment of low back pain. Phys Ther. 1992;72:261-269. [2] Rooney JG, Currier DP, Nitz AJ. Effect of variation in the burst and carrier frequency modes of neuromuscular electrical stimulation on pain perception of healthy subjects. Phys Ther. 1992;72:800-806. [3] Mollinger LA, Steffen TM. Knee flexion contractures in institutionalized elderly: prevalence, severity, stability, and related variables. Phys Ther. 1993;73:437-444. [4] Haynes RB. Loose connections between peer-reviewed clinical journals and clinical practice. Ann Intern Med. 1990;113:724-728. [5] Altman DG, Goodman SN. Transfer of technology from statistical journals to the biomedical literature. JAMA. 1994;272:129-132. [6] Ottenbacher KJ. Evaluating Clinical Change: Strategies for Occupational and Physical Therapists. Baltimore, Md: Williams Wilkins; 1986. [7] Kuhn TS. The Structure of scientific Revolutions. Chicago, Ill: University of Chicago Press; 1962. [8] Kent T. The Kuhnian model of scientific change. In: Langdon JH, McGann ME. The Natural History of Paradigms: Science and the Process of intellectual Evolution. Indianapolis, Ind: University of Indianapolis Press; 1993.

Author Response

Dr Domholdt provided an alternative interpretation of the data reported in my article. Several of the points she raised in her interpretation need clarification in relation to the findings, scope, and purpose of my article. Also, several key elements in Domholdt's interpretation of my findings are, I will suggest, mistaken.

A substantial part of Domholdt's commentary concerned the purpose of research in physical therapy. This part of her commentary is problematic on two grounds. First, contrary to what Domholdt suggested, it was not my view that the purpose of research in physical therapy is to "demonstrate and improve the effectiveness of physical therapy techniques, interventions, and modalities." This view was only one of seven views that I listed as discussed in the physical therapy literature. Second, Domholdt's proposal that the purpose of research "is to inform' the practice of physical therapy" is too vague to be useful. At its simplest, Domholdt's proposal could mean only that research should be available for practitioners to use. Practitioners in many professions, however, are notoriously poor at using research findings.[1] The extent to which physical therapists rely on others as sources of information[2] Suggests similar problems in this profession, At a more complex level, Domholdt's proposal does not address the implications for practice of research that produces contradictory, ambiguous, or possibly unreliable findings or that does not directly concern current practice issues.

Ultimately, physical therapists must research their own clinical practices, no matter what other kinds of research they might also do. Physical therapists do use and should continue to use research from various other disciplines, including medicine and physiology.[3] Research from these disciplines, as with nonclinical physical therapy research, can, and often should, have a direct bearing on, or lead to, developments in clinical practice. Unless clinical practice is subsequently investigated directly and systematically, the effects of practice will remain unknown. This view is not a narrow one that presupposes the effectiveness of clinical practice. This view says simply that the direct and systematic investigation of physical therapy practice should have a central place among the purposes that physical therapists might have for doing research.

Two points Domholdt raised in her commentary canvas possible reasons for the small number of clinical studies in my sample. Her first point concerned the breadth of the statement of purpose published in the masthead of Physical Therapy. Domholdt implied that relatively few clinical articles could be expected in a journal that has such a broad statement of purpose. This point is irrelevant on two grounds. First, irrespective of the breadth of the statement of purpose of the journal, as the most influential journal in the profession, Physical Therapy should be carefully and critically scrutinized by a range of users. Second, editors and guest editors of Physical Therapy explicitly requested more clinically relevant research during the period I examined.[4-8] In her second point about the small number of clinical studies, Dr Domholdt discussed the prevalence of physical therapy education programs in nonclinical settings in the United States. This point is beyond the scope of my study. I agree with Domholdt's implication, though, that a finding of relatively few clinical studies requires serious attention. I suggest that factors other than educational settings or a journal's statement of purpose might prove pivotal when considering the type of physical therapy research pursued and published. Such factors would include the types of research methods generally accepted in physical therapy.

In the second section of her commentary, Domholdt addressed my findings concerning the relative incidence of various research methods, including group designs. My data showed an increasing use of group designs in the articles sampled. The context of this finding, as my article explained, is that in physical therapy, "[s]ome books and articles that describe how to do research and how to evaluate research focus almost exclusively on group-experimental designs [emphasis added]." By replacing all the words prior to "focus" with "the literature of the profession," Domholdt changed the meaning of my claim.

Domholdt stated that my findings provide evidence both of "an evolution of methodological thinking over the last 20 years" and of a willingness of researchers to experiment with methods other than group designs. In support of this, Domholdt referred to the relative recency of the publication dates of articles I cited on "less traditional designs," the short time between the publication of a particular text on single-subject designs and articles using such designs in my sample, and the changes I reported in the number of case studies in the sample over time.

Domholdt's suggestion that methodological thinking has evolved over the past 20 years is problematic. First, my article indicated that group designs were becoming more often used and alternative designs were becoming less Often used. These changes have occurred during a period of increasing interest in research in physical therapy. As this interest has increased, more articles on how to do research have been published. Although articles on how to do research have covered a range of methods, increasingly fewer of the sampled articles used anything other than group designs. Instead, the trend toward more group designs has continued to increase.

Second, the physical therapy literature has included examples of and discussions about single-subject designs for nearly 20 years. The first applications of this method in the physical therapy literature were published in 1976,[9,10] 10 years before the book Domholdt mentioned was published and the same year that a subsequently highly influential book on the topic was published.[11] In the 10 years to 1986, articles were published either describing the method or using it.[12-18] Even so, my study found little use of single-subject designs in articles on the three selected topics (ie, the knee, the back, and electrical stimulation). This finding, together with the history of nearly 20 years of use and discussion of single-subject designs in the physical therapy literature, suggests two things. First, the eventual duration of the lag between the introduction of single subject designs into the physical therapy literature and evidence of their general acceptance remains unknown. The current length of the lag is not consistent with Domholdt's interpretation of "a rapid transfer of technology." Second, little substantial change has occurred in methodological thinking in physical therapy during the past 20 years or so. The persistently small number of studies using single-subject designs suggests little real change has yet occurred, at least with respect to the use of that kind of research design. The reasons for the continuing limited use of single-subject designs are unknown but might relate to misconceptions about the method that still exist in physical therapy.[19]

Third, my article reported a marked drop in the percentage of case studies. As Figure 2 in my article showed, case studies dropped from 490/o of all articles sampled from 1954 to 1963 to only 10% of all articles between 1984 to 1993. Dr Domholdt is correct in saying that the number of case studies had increased. But, as Figure 2 showed, the absolute number of occurrences of most of the eight methods investigated rose in the period 1984 to 1993. That is, as more articles were published, the number of case studies using most of the methods rose. In the context of increasing absolute numbers of articles, the use of particular methods needs to be expressed as a relative frequency (ie, percentage). My results demonstrated a marked drop over time in the relative frequency (percentage) of case studies.

In summary, contrary to Domholdt's inference, my data do not support conclusions that change is occurring in methodological thinking or that researchers are willing to experiment with methods other than group designs. Instead, the data suggested an entrenchment of the existing trend toward using fewer case studies (and most other alternative methods) and more group designs. The data also suggested a continuing lack of acceptance of single-subject designs, a method introduced into the physical therapy literature nearly 20 years ago.

The preceding points indicate serious difficulties in the alternative interpretation of my data offered by Dr Domholdt. A further difficulty derives from Domholdt's reliance on Kuhn's view of the changes that occur in the sciences. Domholdt argued that physical therapy might be experiencing the third in a sequence of four stages described by Kuhn as occurring during a paradigmatic shift. Problems in Kuhn's account have long been discussed. For example, Kuhn provided many different, ambiguous, and often contradictory explanations of what is meant by a paradigm, and his view of change in science is that of a relativist, with all the inherent weaknesses of that position.[20] Also, few physical therapists would agree that physical therapy is a science or that there is a science of physical therapy. Physical therapy is a clinical profession. Kuhn's view of paradigm shifts was not intended to apply to a clinical profession that has not, and is not expected to, develop the laws, theories, or models of a science. Although possibly intuitively appealing, comments on developments in physical therapy that are based on Kuhn's approach are fundamentally misconceived.

I thank Dr Domholdt and the Editor of Physical Therapy for this opportunity to discuss my findings and the interpretation I presented in my article. My findings imply a need for more reflection within physical therapy on research in the profession and on whether current research practices are, or are capable of, providing their expected benefits. My hope is that the necessary process of self-reflection win occur and will not be discouraged by attempts to insist that all is well in physical therapy research.

References

[1] Bohannon RW, LeVeau BF. Clinicians' use of research findings. Phys Ther. 1986;66:45-50. [2] Bohannon RW. Information accessing behaviour of physical therapists. Physiotherapy Theory and Practice. 1990;6:215-225. [3] Robertson VJ. Research and the cumulation of knowledge in Physical Therapy. Phys Ther. 1995;75:313-322. [4] Rothstein JM. Editor's note: Clinical literature. Phys Ther. 1989;69:895-896. [5] Rothstein JM. Editor's note: The Journal--past, present, and future. Phys Ther. 1989;69: 795-796. [6] Lamb RL, Bohannon RW, Craik RL, et al. Guest editorial: Reliability discussion required. Phys Ther. 1987;67:501. [7] Lister MJ. Editor's note: Farewell. Phys Ther. 1988;68:183-184. [8] Lamb RL, Bohannon RW, Craik RL, et al. Guest editorial: Priorities for publication. Phys Ther. 1987;67:1053. [9] Martin JE, Epstein LH. Evaluating treatment effectiveness in cerebral palsy. Phys Ther. 1976;56:285-294. [10] Fowler SA, Rowbury TG, Nordyke N, et al. Color matching technique to train children in the correct use of stairs. Phys Ther. 1976;56: 903-910. [11] Hersen M, Barlow DH. Single-case Experimental Designs. New York, NY: Pergamon Press; 1976. [12] Connolly BH, Craik RL, Krebs DE. Guest editorial: Single-case research: when is it valid? Phys Ther. 1983;63:1767-1768. [13] Goodisman LD. A manipulation-free design for single-subject cerebral palsy research. Phys Ther. 1982;62:284-289. [14] Harris SR, Riffle K. Effects of inhibitive ankle-foot orthoses on standing balance in a child with cerebral palsy: a single-subject design. Phys Ther. 1986;66:663-667. [15] Hill LD. Contributions of behavior modification to cerebral palsy habilitation. Phys Ther. 1985;65:341-345. [16] Laskas C, Mullen S, Nelson DL, et al. Enhancement of two motor functions of the lower extremity in a child with spastic quadriplegia. Phys Ther. 1985;65:11-16. [17] Leonard EL, Trykowski LE, Kirkpatrick BV, et al. Nutritive sucking in high-risk neonates after perioral stimulation. Phys Ther. 1980;60: 299-302. [18] Wolery M, Harris SR. Interpreting results of single-subject research designs. Phys Ther. 1982;62:445-451. [19] Robertson VJ, Lee VL. Some misconceptions about single-subject designs in physiotherapy, Physiotherapy. 1994;80:762-766. [20] Shapere D. The structure of scientific revolutions. Philosophical Review. 1964;73:383-394.

VJ Robertson, PhD, BAppSc(Phty), BA(Hons), is Senior Lecturer, School of Physiotherapy, La Trobe University, Locked Bag 12, Carlton South, Victoria, Australia 3053.

This article was submitted January 26, 1994, and was accepted November 23,1994.