Scand J Caring Sci Vol. 4 No. 3 1990
Sensitivity, Specificity, and Predictive Value in Katz’s and Barthel’s ADL Indices Applied on Patients in Long Term Nursing Care Kristina Tornquist , Monika Lovgren and Bjorn Soderfeldt
ABSTRACT. There are propositions that two indices of ADL ability, Katz’s and Barthel’s indices, are exchangeable for each other. Here this is investigated with regard to clinical applicability. The indices are applied on a material of 23 long term care patients. Sensitivity, specificity, predictive value and interassociations are calculated between the indices, and between indices, nurses’ judgements of appropriate care level, an index of congnitive ability (PTS), and a physical examination. Results show clear differences between the indices as to specificity and predictive value, plus differences in sensitivity on dichotomised measurement level. Associations between measures are very high, between measures and nurses’ judgements and PTS medium high, and low between indices and physical status. It is concluded that the Katz’s ADL index should be preferred in a clinical decision-making for appropriate care level.
Key words: ADL-assessment, Katz’s index, Barthel’s index, self-care, long term care.
INTRODUCTION It is difficult to judge care quality for elderly people, since there is great variation of physical, psychological, and social status in that group. Methodological development is therefore necessary to create useful instruments for this judgement. ADL (Activities of Daily Living) assessment is one main direction in this development (Kane & Kane 1981). There, activities are described. They are then condensed to aggregate measures, so called indices, in order to compare the effect of treatment or care with different individuals (Brorsson 1980). Such ADL indices have been used as effect variables in studies where the aim has been to show a rehabilitation effect (Hamrin & Wohlin 1981, Lovgren 1985, Alenfelt, Norin-Jansson & Tornquist 1980, Tornquist 1984, Hamrin & Lindmark 1988, Soderback 1988). They are used either as single effect variables or in combination with other effect
measures. The function of the measures is to support clinical decision-making. Two frequently used ADL indices (cf. Gresham & Labi 1984, pp 65-85, Brorsson 1980) are Katz’s index of ADL (Katz et ai. 1963) and Barthel’s index (Mahoney & Barthel 1965). Both are based on individual actual performance in care situations. I
1
I
99
KRISTINA TORNQUIST, Occup. ther.“, MONIKA LOVGREN, RN, DMScb and BJORN SODERFELDT, PhD.‘ “Orebro College for the Health Professions, Department of Occupational Therapy, bDepartment of Long Term Care Medicine, Orebro County Council and ‘Department of Social Medicine, Orebro County Council, Sweden. Submitted for publication April 18, 1989, accepted November 6, 1989. Correspondence: K. Tornquist, Orebro College of the Health Professions, Box 1323, S-70234 Orebro, Sweden. Scand J
Caring Sci
100
K . Tornquist et al.
Katz’s index aims to measure the independence of the patient as to six nominal variables: 1. bathing, 2. dressing, 3. toilet, 4. transfer, 5. continence, 6. feeding. The index contains seven categories, A through G , ranked from least to greatest independence, plus a miscellaneous category for unclassifiable patients (Katz et al. 1963). Barthel’s index indicates the ability of the patient as to ten nominal variables: (1) feeding, (2) sitting up in bed, (3) moving from wheel-chair to bed and return, (4) personal toilet, (5) getting on and off toilet, (6) bathing, (7) walking/wheel-chair propelling, (8) ascending/descending stairs, (9) dressing, (10) continence. Each variable is assigned a value indicating independence. The values are summed in an interval scale ranging from 0 to 100. There are indications in earlier research that Katz’s and Barthel’s indices despite their very different construction are exchangeable for each other as to sensitivity. Both indices are considered equally valid for judgement of ADL capability (Labi & Gresham 1984, pp 86-98). Katz’s index aims to measure ADL activities, while Barthel’s index is more related to functional capability. The differing aims of the indices is a reason to inquire further into the question of exchangeability of the two indices. In decision-making concerning care of patients with low physical and psychological capacities, the need for decision support is high (Wulff 1976). An ADL measure must in this situation have good predictive value and high specificity, while sensitivity is less relevant (Hulter Asberg 1986). In the clinical decision situation, it is more important to be sure that the patient can manage a more demanding ADL situation. The desirable property of an ADL measure in such a situation is high predictive value and specificity. It is less important if one eventually would miss some minor functional improvement registered by a highly sensithe measure. The aim of this paper is from this background the following: To assess the two ADL measures (Katz and Barthel) as to predictive value, sensitivity, and specificty in a clinical decision situation. This assessment is done in two ways: (a) A comparison of the ADLindices to a measure of cognitive functional indices to judgements of appropriate care
level (PTS). (b) A comparison of the ADLlevel. It should be noted that our problem is not the validity of the measures, which we presume is adequate, but rather their usefulness in clinical decision-making.
MATERIAL AND METHODS Study group The study group consisted of 23 patients in a somatic long term care ward at Orebro Regional Medical Centre Hospital. In the ward, an intervention study was done during 1987, aimed at activation of self care ability with the patients (Lovgren, Backstrom & Tornquist 1989). Out of 28 patients in the ward, 23 were judged to have assessable ADL ability. In the study group, there were 15 women and eight men with ages between. 71-94 and 69-86, respectively. All patients were medically examined for physical functioning during the investigation period. Eventual impairments were considered by the ward MD as to their effect on the self-care ability of the patient. The examinations found no severe somatic impairments in seven patients, while the rest had such impairments, i.e. 16 patients. The ADL ability of the patients in the study group was observed at three points in time, (1) April 1987, (2) eight weeks later, and (3) 20 weeks later. A registered nurse (M.L. of the present authors) observed and recorded ADL situations. Some of the situations were video recorded. Notes and videograms have been interpreted by an occupational therapist (K.T. of the present authors) who assigned values for the patients on the Katz’s (Katz, et al. 1963) and Barthel’s indices (Mahoney & Barthel 1965). The main study material thus consists of 138 ADL judgements (3 occasions x 23 patients x 2 indices) which are described in Table I. The practical cognitive function of the patients was further judged by the ward occupational therapist at all three measurement occasions. A screening instrument called PTS (S~rensen1983) was used, measuring memory, time, and space orientation, problem solving and communicative ability, cooperative ability, initiative, and mood. Maximum PTS points was 30. Five patients were not contactable at the third measurement occasion. This material is described in Table 11. Scand I Caring Sci
ADL-assessment in long term care
101
Table I . Results of ADL-measures and ADL-changes over time Barthels index (0-100)
Katz’s index (A-G)
Rating:
Rating:
.-
Patient number -
___
1
2
3
Changes
1
2
3
55 5 15 0 0 30 40 30 5 25 0 25
70 45 45 0 0 50 55 45 5 65 0 35 80 25 0 0 15 0 85 30 40 0 5
75 55 35 0 0 50 55 55 5 65 0 35 80 20 0 0 15 5 85 40 35 5 5
X X
D F F
C F F
A
G G
G G
G G
F C E F F G F E F
E B F F D
D B E F
G
G
F B F
F C F
G G G
G G G
A
G G G G A
F F
F F
G G
G G
F F F
Changes
~
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
60 15 0 0 0 0 85 5 25 0 0
x X X X X X X X X X
F
X X X X
F F
B
F A
G
Table 11. Results of PTS-measures, for three occasiom Patient number 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
PTS-rating : 1
2
3
4 20 11 16 7 4 20 5 4 7 4 19 5 8 2 2
6 27 12 13 6 5 26 5 4 18 4 20 5 6 2 2
6 -(a) 11 16 -(a) 4 29 5 -(a) 24 -(a) 21 6 6 2 2 4 3 -(a) 21 14 3 18
4 2 11 6 14 3 13 Note: (a): Non-contactable
4 3 21 14 15 3 14
Positive changes (>1) X X
X X X
X X
X Scand J Caring Sci
102
K . Tornquist et al.
Independently of this study, all patients were judged by the responsible ward nurses as to appropriate care level at the end of the investigation period. The judgements were recorded in the daily records. The various care levels available for choice were the following, ranked according to required work load in the care situation, from the lightest to the heaviest situation: 1. Self-subsistence in own housing. 2. Self-subsistence in category housing for the elderly. 3. Partial self-subsistence in category housing with full accommodation. 4. Subsistence in group housing for seniles with 24 hour service. 5. Long-term hospital care. At the end of the investigation period, the judgements were distributed according to Table 111. All patients were obviously on level 5 at the beginning of the observation period. Table 111. Nurses’ judgement of correctly care level Patient number
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
Correctly care level 4 1 3 5 5 4 1 4 5 2
Changes
X X X X X X X
5
5 2 5 5 5 5
X
5 2 5 5 5 2
X
X
1 = Self-subsistence in own housing, 2 = Selfsubsistence in category housing for the elderly, 3 = partial self-subsistence in category housing with full accommodation,4 = Subsistence in group housing for seniles with 24 hour service, 5 = Long term hospital care.
Methods We define sensitivity, specificity, and predictive value according to Foldspang etal. (1981). Sensitivity is the proportion of “true” positive cases in relation to estimated positive cases. Specificity is the proportion of “true” negative cases in relation to estimated negative cases. This can be illustrated from Fig. 1. “True observations”
Fig. 1 . Correspondence between “true” and estimated observations. Sensitivity is then the relation a / ( a + c), and specificity the relation d / ( d + d ) From this, one can also define a predictive value as a/(a b ) , i.e. the proportion of “true” positive cases in relation to all positive cases. When calculating the magnitudes of sensitivity and predictive value, the ADL-measures must obviously be related to some “true” observations. We have chosen PTS and nurses observations for such reference values. The choice of “true” observations is in principle arbitrary, as long as they are standardised for, and have some relation to the study variables. In practical clinical decision-making, both indices are simplified to dichotomies. They are divided at the point which is deemed most relevant for independence, i.e. 60 points on Barthel’s index (Richard el al. 1981) and D (toilet independence) on Katz’s index. This simplified measure is also evaluated. The associations between the ADL measures and between the measures and the medical examinations have also been investigated. Here, only ordinal measures can be used, since Katz’s index is on ordinal scale level. We have chosen two measures of rank order correlation, Spearman’s rho and Kendall’s tau (Blalock 1960).
+
RESULTS In Table IV, the test results of the two ADLindices are related to PTS values as “true” values. The comparison is done for the change Scand J Caring Sci
ADL-assessment in long term care
103
Table IV. Relations between changes assessed with ADL-measures and changes assessed with PTS. n=25
Index results:
Barthel’s index: Positive None or PTS-changes negative PTS-changes
Katz’s index: Positive F’TS-changes
None or negative F’TS-changes
Positive ADL-changes
3
3
None or negative ADL-changes
5
12
Sensitivity =88% Specificity=40% Predictive value=44% Spearman’s rho=0.45 Kendall’s tau=0.40
in the respective measure between the beginning and the end of the study. The rank order correlations between the measures and PTS values at the last measurement point are also stated in the table. The correlations between ADL and PTS measures are rather low. The reason is obvious-ADL ability demands more than
Sensitivity=38% Specificity= 80% Predictive value=50% Spearman’s rho=0.33 Kendall’s tau=0.29
cognitive ability. One should however note that sensitivity for Barthel’s index is 88%, meaning that most of those with registered changes in ADL-ability according to Barthel’s index also have improved cognitive function. The predictive value is somewhat better for Katz’s index, although still low in absolute terms. In Table V the ADL-measures are
Table V. Relations between changes assessed with ADL-measures and changes assessed with nurses’ judgements n=23 Barthel’s index: Nurses’ judgements: Lighter Appropriate level level
Katz’s index: Nurses’ judgements: Lighter Appropriate level level
Sensitivity=90% Specificity=46 % Predictive value=56% Spearman’s rho=0.58 Kendall’s tau=0.43
Sensitivity=50% Specificity=92 % Predictive value=83% Spearman’s rho=0.65 Kendall’s tau=0.66
Index results: Positive ADL-changes None or negative ADL-changes
Scand J Caring Sci
104
K . Tornquist et al.
Table VI. Relations between dichotomised ADL-measures (find) and Nurses’ judgements n=23 Barthel’s index: Nurses’ judgements: Lighter Appropriate care-level care-level
Katz’s index: Nurses’ judgements: Lighter Appropriate care-level care-level
Sensitivity=40% Specificity= 100% Predictive value= 100%
Sensitivity=70% Specificity= 100% Predictive value= 100%
Index results:
compared to nurses’ judgements as “true” values. “Lighter level” here means any lighter care level in relation to the one where the patients are. “Appropriate level” means no change of care level. The associations increase for both indices. It is reasonable to believe that an ADL index has higher relevance for actual care judgements than for a simple measure of cognitive ability. One can note that Katz’s index here has a slightly higher association with nurses’ judgements than Barthel’s index. The general level of sensitivity, specificity, and predictive value is higher when using nurses judgements instead of PTS as true values. The multiple dimensions of the ADL measures are probably better mirrored in the nurses’ judgements. Sensitivity is still about 90% for Barthel but it has improved for Katz’s index. The predictive value has improved to 83% for Katz and 56% for Barthel. The simplified measures, the dichotomies, have an even closer relevance for nurses’ judgements. Results here are shown in Table VI. Here the measures are equal as to specificity and predictive value. Katz’s index shows however a considerably larger sensitivity, which is contrary to earlier findings. Associations are not meaningful on these dichotomies. The multidimensionality of the ADL measures is a probable explanation of these results. To control for an additional such dimension,
we have investigated the association between the ADL measures and the medical judgements of somatic status. In Table VII, rank order correlations by Spearman’s rho and Kendall’s tau are stated. Results in Table VII show low associations-the ADL measures comprise more dimensions than medical examinations. We have finally investigated the associations between the measures themselves in Table VIII for two measurement occasions, at the beginning and at the end of the study. As can be expected, the associations are high. It is however interesting to note that despite these Table VII. Associations between A DL-measures and medical examinations. n =23
Barthel’s index Katz’s index
Spearman’s rho
Kendall’s tau
0.25 0.26
0.15 0.24
Table VIII. Associations between the measurements with Barthel’s index and Katz’s index at the beginning and at the end of the study. n =23 ~
Rating I Rating I11
Spearman’s rho
Kendall’s tau
0.92 0.90
0.83 0.83 Scand J Caring Sci
ADL-assessment in long term care
105
livets aktiviteter. Metodoversikt. Medicinska forskningsridet, Stockholm. Foldspang, A . , Juul, S., Olsen, J. & Sabroe, S. 1981. Epidemiologi. Sygdom og befolkning. Ch. 3 . Munksgaard, K~benhavn. G.E. & Labi, M.L.C. 1984. Functional DISCUSSION AND CONCLUSIONS Gresham, assessment instruments currently available for The results, as to the associations between the documenting outcomes in rehabilitation rnedicine. In: Granger, C.V. & Gresham, G.E. (eds) ADL-indices, show a seemingly large degree Functional assessment in rehabilitation medicine. of exchangeability between the measures. Williams & Wilkins, Baltimore, Md., 65-85. This also shows in the comparison of senHamrin, E. & Wohlin, A. 1981. Evaluation of the sitivity. This exchangeability does however functional capacity of stroke patients through an not hold on closer analysis. activity index. Scand. J. Rehab. Med., 14, We have thus found considerable dif93-100. ferences between the indices as to specificity Hamrin, E. & Lindmark, B. 1988. Evaluation of and predictive value, and also as to sensitivity functional capacity after stroke as a basis for actfor the clinically relevant dichotomised measive intervention. A comparison between an ures. The general picture, in our opinion, Activity Index and the Katz Index of ADL. Scand. J. Caring Sci., 2, 113-22. shows that Katz’s index should be preferred as a basis for clinical decision-making. It differs Hulter Asberg, K. 1986. Elderly patients in acute medical wards and home care. Functional assessfavourably as to specificity and predictive ment, prediction of outcome, and trial of early value; those properties that are most importactivation. Acta Universitatis Upsalisensis 25, ant in the clinical situation. Uppsala University, Uppsala (Academical Our results cannot however be regarded as Dissertation). conclusive and further research is needed. The Kane, R.A. & Kane, R.L. 1981. Assessing the material is limited to only 23 patients. Signifielderly. A practical guide to measurement, Ch. 2. cance tests have not been meaningful here, Lexington Books, Lexington. both because rank order correlations lack sim- Katz, S., Ford, A.B., Moskowitz, R.W., Jackson, B.A. & Jaffe, M. W. 1963 Studies of illness in the ple probability distributions and because of aged: The index of ADL, a standardized measure the small material. It has further not been of biological and psychosocial function. J . Am. possible to include inquiry into test-retest Med. Ass., 185, 914-19. and/or interobserver reliability, which of course is very important in clinical Labi, M.L.C. & Gresham, G.E. 1984. Some research applications of functional assessment applications. instruments used in rehabilitation medicine. In: With these reservations, our results howGranger, C.V. & Gresham, G.E. (eds) Funcever indicate that Katz’s index is preferable to tional assessment in rehabilitation medicine. Barthel’s when assessing adequate care level Williams & Wilkins, Baltimore, Md., 86-98. in long term somatic care. We also believe Lovgren, M. 1985. Individanpassad omvirdnad i samverkan med aldre patienter-kartlaggning from our results that Katz’s index may be och intervention inom geriatrisk v h d . (diss.), more suitable in evaluating situations, despite Medicinska fakulteten, Goteborgs Universitet, the higher sensitivity of Barthel’s index. It Goteborg. certainly seems to be more suitable when using dichotomised measures, which also Lovgren, M., Backstrom, U. & Tornquist, K. 1989. Egenomsorg-ett satt for aldre att leva-inte bara should be simpler, less costly, and more reliaatt overleva. Rapport, Lingvirds-och ble in practical clinical work. rehabiliteringsblocket, Orebro Lans Landsting, Orebro. Mahoney, F.I. & Barthel, D.W. 1965. Functional REFERENCES evaluation: The Barthel Index. Maryland State Alenfelt, G. Norin-Jansson M. & Tornquist K. Medical Journal 24,61-5. 1980. Omvdrdnadsprocessen. Hogskolan i Richard, H., Fortinsky, M.A., Granger, C.V. et al. Orebro, Skriftserie 4. 1981. The use of functional assessment in underBlalock, H. 1960. Social statistics. McGraw Hill, standing home care needs. Med. Care, 19,489New York. 97. Brorsson, B. 1980. ADL-index: Sammanfattande Soderback, I. 1988. The effectiveness of training mdtt pd individens formdga atf klara det dagliga intellectual functions in adults with acquired
high associations, the predicitive values of the indices differ considerably as has been shown in previous tables.
Scand J Caring Sci
106
K . Tornquist et al.
brain damage. An evaluation of occupational therapy methods. Scand. J . Rehab. Med. 20, 47-56. Skensen, L. 1983. Praktiske cognitive- og ADLunders#gelser. Kompendium, Ergoterapien, Tranehaven, Gentofte Kommunes geriatriske behandlingscenter, Charlottenlund.
Tornquist, K. 1984. Vardet av individuell vdrdplanering for sjukhemspatienter. Kompendium. Planerings- och utvecklingsavdelningen, Orebro Lans Landsting, Orebro. Wulff, H. 1976. Rational diagnosis and treatment. Blackwell Scientific Publications, Oxford.
Scand J Caring Sci
Sensitivity, Specificity, and Predictive Value in Katz’s and Barthel’s ADL Indices Applied on Patients in Long Term Nursing Care Kristina Tornquist , Monika Lovgren and Bjorn Soderfeldt
ABSTRACT. There are propositions that two indices of ADL ability, Katz’s and Barthel’s indices, are exchangeable for each other. Here this is investigated with regard to clinical applicability. The indices are applied on a material of 23 long term care patients. Sensitivity, specificity, predictive value and interassociations are calculated between the indices, and between indices, nurses’ judgements of appropriate care level, an index of congnitive ability (PTS), and a physical examination. Results show clear differences between the indices as to specificity and predictive value, plus differences in sensitivity on dichotomised measurement level. Associations between measures are very high, between measures and nurses’ judgements and PTS medium high, and low between indices and physical status. It is concluded that the Katz’s ADL index should be preferred in a clinical decision-making for appropriate care level.
Key words: ADL-assessment, Katz’s index, Barthel’s index, self-care, long term care.
INTRODUCTION It is difficult to judge care quality for elderly people, since there is great variation of physical, psychological, and social status in that group. Methodological development is therefore necessary to create useful instruments for this judgement. ADL (Activities of Daily Living) assessment is one main direction in this development (Kane & Kane 1981). There, activities are described. They are then condensed to aggregate measures, so called indices, in order to compare the effect of treatment or care with different individuals (Brorsson 1980). Such ADL indices have been used as effect variables in studies where the aim has been to show a rehabilitation effect (Hamrin & Wohlin 1981, Lovgren 1985, Alenfelt, Norin-Jansson & Tornquist 1980, Tornquist 1984, Hamrin & Lindmark 1988, Soderback 1988). They are used either as single effect variables or in combination with other effect
measures. The function of the measures is to support clinical decision-making. Two frequently used ADL indices (cf. Gresham & Labi 1984, pp 65-85, Brorsson 1980) are Katz’s index of ADL (Katz et ai. 1963) and Barthel’s index (Mahoney & Barthel 1965). Both are based on individual actual performance in care situations. I
1
I
99
KRISTINA TORNQUIST, Occup. ther.“, MONIKA LOVGREN, RN, DMScb and BJORN SODERFELDT, PhD.‘ “Orebro College for the Health Professions, Department of Occupational Therapy, bDepartment of Long Term Care Medicine, Orebro County Council and ‘Department of Social Medicine, Orebro County Council, Sweden. Submitted for publication April 18, 1989, accepted November 6, 1989. Correspondence: K. Tornquist, Orebro College of the Health Professions, Box 1323, S-70234 Orebro, Sweden. Scand J
Caring Sci
100
K . Tornquist et al.
Katz’s index aims to measure the independence of the patient as to six nominal variables: 1. bathing, 2. dressing, 3. toilet, 4. transfer, 5. continence, 6. feeding. The index contains seven categories, A through G , ranked from least to greatest independence, plus a miscellaneous category for unclassifiable patients (Katz et al. 1963). Barthel’s index indicates the ability of the patient as to ten nominal variables: (1) feeding, (2) sitting up in bed, (3) moving from wheel-chair to bed and return, (4) personal toilet, (5) getting on and off toilet, (6) bathing, (7) walking/wheel-chair propelling, (8) ascending/descending stairs, (9) dressing, (10) continence. Each variable is assigned a value indicating independence. The values are summed in an interval scale ranging from 0 to 100. There are indications in earlier research that Katz’s and Barthel’s indices despite their very different construction are exchangeable for each other as to sensitivity. Both indices are considered equally valid for judgement of ADL capability (Labi & Gresham 1984, pp 86-98). Katz’s index aims to measure ADL activities, while Barthel’s index is more related to functional capability. The differing aims of the indices is a reason to inquire further into the question of exchangeability of the two indices. In decision-making concerning care of patients with low physical and psychological capacities, the need for decision support is high (Wulff 1976). An ADL measure must in this situation have good predictive value and high specificity, while sensitivity is less relevant (Hulter Asberg 1986). In the clinical decision situation, it is more important to be sure that the patient can manage a more demanding ADL situation. The desirable property of an ADL measure in such a situation is high predictive value and specificity. It is less important if one eventually would miss some minor functional improvement registered by a highly sensithe measure. The aim of this paper is from this background the following: To assess the two ADL measures (Katz and Barthel) as to predictive value, sensitivity, and specificty in a clinical decision situation. This assessment is done in two ways: (a) A comparison of the ADLindices to a measure of cognitive functional indices to judgements of appropriate care
level (PTS). (b) A comparison of the ADLlevel. It should be noted that our problem is not the validity of the measures, which we presume is adequate, but rather their usefulness in clinical decision-making.
MATERIAL AND METHODS Study group The study group consisted of 23 patients in a somatic long term care ward at Orebro Regional Medical Centre Hospital. In the ward, an intervention study was done during 1987, aimed at activation of self care ability with the patients (Lovgren, Backstrom & Tornquist 1989). Out of 28 patients in the ward, 23 were judged to have assessable ADL ability. In the study group, there were 15 women and eight men with ages between. 71-94 and 69-86, respectively. All patients were medically examined for physical functioning during the investigation period. Eventual impairments were considered by the ward MD as to their effect on the self-care ability of the patient. The examinations found no severe somatic impairments in seven patients, while the rest had such impairments, i.e. 16 patients. The ADL ability of the patients in the study group was observed at three points in time, (1) April 1987, (2) eight weeks later, and (3) 20 weeks later. A registered nurse (M.L. of the present authors) observed and recorded ADL situations. Some of the situations were video recorded. Notes and videograms have been interpreted by an occupational therapist (K.T. of the present authors) who assigned values for the patients on the Katz’s (Katz, et al. 1963) and Barthel’s indices (Mahoney & Barthel 1965). The main study material thus consists of 138 ADL judgements (3 occasions x 23 patients x 2 indices) which are described in Table I. The practical cognitive function of the patients was further judged by the ward occupational therapist at all three measurement occasions. A screening instrument called PTS (S~rensen1983) was used, measuring memory, time, and space orientation, problem solving and communicative ability, cooperative ability, initiative, and mood. Maximum PTS points was 30. Five patients were not contactable at the third measurement occasion. This material is described in Table 11. Scand I Caring Sci
ADL-assessment in long term care
101
Table I . Results of ADL-measures and ADL-changes over time Barthels index (0-100)
Katz’s index (A-G)
Rating:
Rating:
.-
Patient number -
___
1
2
3
Changes
1
2
3
55 5 15 0 0 30 40 30 5 25 0 25
70 45 45 0 0 50 55 45 5 65 0 35 80 25 0 0 15 0 85 30 40 0 5
75 55 35 0 0 50 55 55 5 65 0 35 80 20 0 0 15 5 85 40 35 5 5
X X
D F F
C F F
A
G G
G G
G G
F C E F F G F E F
E B F F D
D B E F
G
G
F B F
F C F
G G G
G G G
A
G G G G A
F F
F F
G G
G G
F F F
Changes
~
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
60 15 0 0 0 0 85 5 25 0 0
x X X X X X X X X X
F
X X X X
F F
B
F A
G
Table 11. Results of PTS-measures, for three occasiom Patient number 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
PTS-rating : 1
2
3
4 20 11 16 7 4 20 5 4 7 4 19 5 8 2 2
6 27 12 13 6 5 26 5 4 18 4 20 5 6 2 2
6 -(a) 11 16 -(a) 4 29 5 -(a) 24 -(a) 21 6 6 2 2 4 3 -(a) 21 14 3 18
4 2 11 6 14 3 13 Note: (a): Non-contactable
4 3 21 14 15 3 14
Positive changes (>1) X X
X X X
X X
X Scand J Caring Sci
102
K . Tornquist et al.
Independently of this study, all patients were judged by the responsible ward nurses as to appropriate care level at the end of the investigation period. The judgements were recorded in the daily records. The various care levels available for choice were the following, ranked according to required work load in the care situation, from the lightest to the heaviest situation: 1. Self-subsistence in own housing. 2. Self-subsistence in category housing for the elderly. 3. Partial self-subsistence in category housing with full accommodation. 4. Subsistence in group housing for seniles with 24 hour service. 5. Long-term hospital care. At the end of the investigation period, the judgements were distributed according to Table 111. All patients were obviously on level 5 at the beginning of the observation period. Table 111. Nurses’ judgement of correctly care level Patient number
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
Correctly care level 4 1 3 5 5 4 1 4 5 2
Changes
X X X X X X X
5
5 2 5 5 5 5
X
5 2 5 5 5 2
X
X
1 = Self-subsistence in own housing, 2 = Selfsubsistence in category housing for the elderly, 3 = partial self-subsistence in category housing with full accommodation,4 = Subsistence in group housing for seniles with 24 hour service, 5 = Long term hospital care.
Methods We define sensitivity, specificity, and predictive value according to Foldspang etal. (1981). Sensitivity is the proportion of “true” positive cases in relation to estimated positive cases. Specificity is the proportion of “true” negative cases in relation to estimated negative cases. This can be illustrated from Fig. 1. “True observations”
Fig. 1 . Correspondence between “true” and estimated observations. Sensitivity is then the relation a / ( a + c), and specificity the relation d / ( d + d ) From this, one can also define a predictive value as a/(a b ) , i.e. the proportion of “true” positive cases in relation to all positive cases. When calculating the magnitudes of sensitivity and predictive value, the ADL-measures must obviously be related to some “true” observations. We have chosen PTS and nurses observations for such reference values. The choice of “true” observations is in principle arbitrary, as long as they are standardised for, and have some relation to the study variables. In practical clinical decision-making, both indices are simplified to dichotomies. They are divided at the point which is deemed most relevant for independence, i.e. 60 points on Barthel’s index (Richard el al. 1981) and D (toilet independence) on Katz’s index. This simplified measure is also evaluated. The associations between the ADL measures and between the measures and the medical examinations have also been investigated. Here, only ordinal measures can be used, since Katz’s index is on ordinal scale level. We have chosen two measures of rank order correlation, Spearman’s rho and Kendall’s tau (Blalock 1960).
+
RESULTS In Table IV, the test results of the two ADLindices are related to PTS values as “true” values. The comparison is done for the change Scand J Caring Sci
ADL-assessment in long term care
103
Table IV. Relations between changes assessed with ADL-measures and changes assessed with PTS. n=25
Index results:
Barthel’s index: Positive None or PTS-changes negative PTS-changes
Katz’s index: Positive F’TS-changes
None or negative F’TS-changes
Positive ADL-changes
3
3
None or negative ADL-changes
5
12
Sensitivity =88% Specificity=40% Predictive value=44% Spearman’s rho=0.45 Kendall’s tau=0.40
in the respective measure between the beginning and the end of the study. The rank order correlations between the measures and PTS values at the last measurement point are also stated in the table. The correlations between ADL and PTS measures are rather low. The reason is obvious-ADL ability demands more than
Sensitivity=38% Specificity= 80% Predictive value=50% Spearman’s rho=0.33 Kendall’s tau=0.29
cognitive ability. One should however note that sensitivity for Barthel’s index is 88%, meaning that most of those with registered changes in ADL-ability according to Barthel’s index also have improved cognitive function. The predictive value is somewhat better for Katz’s index, although still low in absolute terms. In Table V the ADL-measures are
Table V. Relations between changes assessed with ADL-measures and changes assessed with nurses’ judgements n=23 Barthel’s index: Nurses’ judgements: Lighter Appropriate level level
Katz’s index: Nurses’ judgements: Lighter Appropriate level level
Sensitivity=90% Specificity=46 % Predictive value=56% Spearman’s rho=0.58 Kendall’s tau=0.43
Sensitivity=50% Specificity=92 % Predictive value=83% Spearman’s rho=0.65 Kendall’s tau=0.66
Index results: Positive ADL-changes None or negative ADL-changes
Scand J Caring Sci
104
K . Tornquist et al.
Table VI. Relations between dichotomised ADL-measures (find) and Nurses’ judgements n=23 Barthel’s index: Nurses’ judgements: Lighter Appropriate care-level care-level
Katz’s index: Nurses’ judgements: Lighter Appropriate care-level care-level
Sensitivity=40% Specificity= 100% Predictive value= 100%
Sensitivity=70% Specificity= 100% Predictive value= 100%
Index results:
compared to nurses’ judgements as “true” values. “Lighter level” here means any lighter care level in relation to the one where the patients are. “Appropriate level” means no change of care level. The associations increase for both indices. It is reasonable to believe that an ADL index has higher relevance for actual care judgements than for a simple measure of cognitive ability. One can note that Katz’s index here has a slightly higher association with nurses’ judgements than Barthel’s index. The general level of sensitivity, specificity, and predictive value is higher when using nurses judgements instead of PTS as true values. The multiple dimensions of the ADL measures are probably better mirrored in the nurses’ judgements. Sensitivity is still about 90% for Barthel but it has improved for Katz’s index. The predictive value has improved to 83% for Katz and 56% for Barthel. The simplified measures, the dichotomies, have an even closer relevance for nurses’ judgements. Results here are shown in Table VI. Here the measures are equal as to specificity and predictive value. Katz’s index shows however a considerably larger sensitivity, which is contrary to earlier findings. Associations are not meaningful on these dichotomies. The multidimensionality of the ADL measures is a probable explanation of these results. To control for an additional such dimension,
we have investigated the association between the ADL measures and the medical judgements of somatic status. In Table VII, rank order correlations by Spearman’s rho and Kendall’s tau are stated. Results in Table VII show low associations-the ADL measures comprise more dimensions than medical examinations. We have finally investigated the associations between the measures themselves in Table VIII for two measurement occasions, at the beginning and at the end of the study. As can be expected, the associations are high. It is however interesting to note that despite these Table VII. Associations between A DL-measures and medical examinations. n =23
Barthel’s index Katz’s index
Spearman’s rho
Kendall’s tau
0.25 0.26
0.15 0.24
Table VIII. Associations between the measurements with Barthel’s index and Katz’s index at the beginning and at the end of the study. n =23 ~
Rating I Rating I11
Spearman’s rho
Kendall’s tau
0.92 0.90
0.83 0.83 Scand J Caring Sci
ADL-assessment in long term care
105
livets aktiviteter. Metodoversikt. Medicinska forskningsridet, Stockholm. Foldspang, A . , Juul, S., Olsen, J. & Sabroe, S. 1981. Epidemiologi. Sygdom og befolkning. Ch. 3 . Munksgaard, K~benhavn. G.E. & Labi, M.L.C. 1984. Functional DISCUSSION AND CONCLUSIONS Gresham, assessment instruments currently available for The results, as to the associations between the documenting outcomes in rehabilitation rnedicine. In: Granger, C.V. & Gresham, G.E. (eds) ADL-indices, show a seemingly large degree Functional assessment in rehabilitation medicine. of exchangeability between the measures. Williams & Wilkins, Baltimore, Md., 65-85. This also shows in the comparison of senHamrin, E. & Wohlin, A. 1981. Evaluation of the sitivity. This exchangeability does however functional capacity of stroke patients through an not hold on closer analysis. activity index. Scand. J. Rehab. Med., 14, We have thus found considerable dif93-100. ferences between the indices as to specificity Hamrin, E. & Lindmark, B. 1988. Evaluation of and predictive value, and also as to sensitivity functional capacity after stroke as a basis for actfor the clinically relevant dichotomised measive intervention. A comparison between an ures. The general picture, in our opinion, Activity Index and the Katz Index of ADL. Scand. J. Caring Sci., 2, 113-22. shows that Katz’s index should be preferred as a basis for clinical decision-making. It differs Hulter Asberg, K. 1986. Elderly patients in acute medical wards and home care. Functional assessfavourably as to specificity and predictive ment, prediction of outcome, and trial of early value; those properties that are most importactivation. Acta Universitatis Upsalisensis 25, ant in the clinical situation. Uppsala University, Uppsala (Academical Our results cannot however be regarded as Dissertation). conclusive and further research is needed. The Kane, R.A. & Kane, R.L. 1981. Assessing the material is limited to only 23 patients. Signifielderly. A practical guide to measurement, Ch. 2. cance tests have not been meaningful here, Lexington Books, Lexington. both because rank order correlations lack sim- Katz, S., Ford, A.B., Moskowitz, R.W., Jackson, B.A. & Jaffe, M. W. 1963 Studies of illness in the ple probability distributions and because of aged: The index of ADL, a standardized measure the small material. It has further not been of biological and psychosocial function. J . Am. possible to include inquiry into test-retest Med. Ass., 185, 914-19. and/or interobserver reliability, which of course is very important in clinical Labi, M.L.C. & Gresham, G.E. 1984. Some research applications of functional assessment applications. instruments used in rehabilitation medicine. In: With these reservations, our results howGranger, C.V. & Gresham, G.E. (eds) Funcever indicate that Katz’s index is preferable to tional assessment in rehabilitation medicine. Barthel’s when assessing adequate care level Williams & Wilkins, Baltimore, Md., 86-98. in long term somatic care. We also believe Lovgren, M. 1985. Individanpassad omvirdnad i samverkan med aldre patienter-kartlaggning from our results that Katz’s index may be och intervention inom geriatrisk v h d . (diss.), more suitable in evaluating situations, despite Medicinska fakulteten, Goteborgs Universitet, the higher sensitivity of Barthel’s index. It Goteborg. certainly seems to be more suitable when using dichotomised measures, which also Lovgren, M., Backstrom, U. & Tornquist, K. 1989. Egenomsorg-ett satt for aldre att leva-inte bara should be simpler, less costly, and more reliaatt overleva. Rapport, Lingvirds-och ble in practical clinical work. rehabiliteringsblocket, Orebro Lans Landsting, Orebro. Mahoney, F.I. & Barthel, D.W. 1965. Functional REFERENCES evaluation: The Barthel Index. Maryland State Alenfelt, G. Norin-Jansson M. & Tornquist K. Medical Journal 24,61-5. 1980. Omvdrdnadsprocessen. Hogskolan i Richard, H., Fortinsky, M.A., Granger, C.V. et al. Orebro, Skriftserie 4. 1981. The use of functional assessment in underBlalock, H. 1960. Social statistics. McGraw Hill, standing home care needs. Med. Care, 19,489New York. 97. Brorsson, B. 1980. ADL-index: Sammanfattande Soderback, I. 1988. The effectiveness of training mdtt pd individens formdga atf klara det dagliga intellectual functions in adults with acquired
high associations, the predicitive values of the indices differ considerably as has been shown in previous tables.
Scand J Caring Sci
106
K . Tornquist et al.
brain damage. An evaluation of occupational therapy methods. Scand. J . Rehab. Med. 20, 47-56. Skensen, L. 1983. Praktiske cognitive- og ADLunders#gelser. Kompendium, Ergoterapien, Tranehaven, Gentofte Kommunes geriatriske behandlingscenter, Charlottenlund.
Tornquist, K. 1984. Vardet av individuell vdrdplanering for sjukhemspatienter. Kompendium. Planerings- och utvecklingsavdelningen, Orebro Lans Landsting, Orebro. Wulff, H. 1976. Rational diagnosis and treatment. Blackwell Scientific Publications, Oxford.
Scand J Caring Sci
