Detection of activities of daily living impairment in Alzheimer’s disease and mild cognitive impairment using information and communication technology

Figures & data

Table 1 Daily activity scenarios characteristics in protocols 1 (P1) and 2 (P2)

	Daily activity scenario associated with the protocol 1 (P1)	Daily activity scenario associated with the protocol 2 (P2)
Activities	“Your task is to perform this list of activities in the order given within a time frame of 20 minutes” Walk to the reading table and read something for 2 mn Walk to the coffee corner where the kettle is and boil some water Walk to the phone and compose this number: xxxxxx Take the watering can and water the plant. Walk to the television and turn it on with the remote control Walk to the reading table, take the playing cards and classify them by color (reds with reds, blacks with blacks) Take the green “ABCD” folder on the desk with the A, B, C, D sheets in it. Match the A, B, C, D sheets from the folder to the relevant folders dispersed all over the room; A with A, etc … Put the “ABCD” folder back on the desk Leave the room	“Your task is to perform this list of 10 activities in a logical manner within 15 min. These 15 minutes represent a typical morning period of everyday life” Read the newspaper Water the plant Answer the phone Call the taxi Prepare today’s medication Make the check for the Electricity Company Leave the room when you have finished all activities Watch the TV Prepare a hot tea Write a shopping list for lunch
Constraints	Participant has to perform the activities in the given order	watch the TV before the phone call water the plant just before leaving the room call the taxi, which will arrive in 10 minutes and ask the driver to drive you to the market
Duration	20 minutes	15 minutes
Parameters extracted from video used to compute DAS.,n scores	– REff € R[0,1] – Number of activities omitted, a1,P1 € [1,7] – Total number of repetitions, a2,P1 € N – Order error, a3,P1 € {0,1} with a3,1 = 1 if order error was done – Total number of attempts before completing a given activity, a4,P1 € N	– REff € R[0,1] – Number of activities omitted, a1,P2 € [0,10] – Total number of repetitions excluding the leisure activities (“watching TV”, and “reading”), a2,P2 € N – Number of activities performed but not achieved, a3,P2 € [0,10] – Respect of temporal execution constraints, a4,P2 € {0,1} with a4,P2 = 1 if more than two or three temporal constraints were not respected

Table 2 Observable criteria used to annotate video sequences of the protocol 2 (P2)

Activity	Activity period		Achievement
	Start	End
Read the newspaper	Taking for the first time the newspaper	Putting on the table the last newspaper taken	Opening a newspaper taken to read the content
Water the plant	Taking the watering can	Putting the watering can on the table	Making the gesture of watering the plant
Answer the phone	Taking the handset in the hand	Putting the handset on the phone base	Speaking on the phone
Call the taxi	Taking the handset in the hand	Putting the handset on the phone base	Dialing the correct phone number and speaking on the phone
Prepare the medication for today	Taking the pillbox located inside a basket with the medication prescription inside	Putting the pillbox inside the basket	Correct use of medication prescription (both dose and timetable)
Make the check for the Electricity Company	Taking the pen to write on the check	Putting the pen on the desk	Correct amount, date, signature and recipient name on the check
Leave the room when you are finished with all activities	Taking the handle of the exit door	Closing the door	Closing the door as the last activity undertaken in the scenario
Watch the TV	Taking the remote call	Returning for the last time the remote control	TV must have been switched off
Prepare a hot tea	Turning on the tea kettle	Returning the tea kettle after having poured the water in the cup of tea	Brewed tea
Write the shopping list for the lunch	Taking the pen to write on the shopping list	Replacing the pen	Write at least one item to eat or drink

Table 3 Characteristics of the participants in protocol 1 (P1)

Characteristics	NC n = 10	AD n = 16	P values
Female, n (%)	5 (50%)	11 (68.8%)	0.425
Age-yrs, mean [CI95%]	73.9 [69.4, 78.4]	76.7 [72.7, 80.7]	0.182
Level of education, n (%)
Unknown	0 (0%)	0 (0%)	–
No formal education	0 (0%)	0 (0%)	–
Elementary school	1 (10%)	7 (43.8%)	0.099
Middle school	0 (0%)	4 (25.0%)	0.136
High school	2 (20.0%)	2 (12.5%)	0.625
Post-secondary education	7 (70.0%)	3 (18.8%)	0.015*
MMSEa, mean [CI95%]	28.1 [26.8, 29.4]	20.7 [18.7, 22.7]	<0.001**
IADL-Eb, mean [CI95%]	10.5 [9.7, 11.3]	14.3 [11.2, 17.5]	0.031*
MADRSc, mean [CI95%]	3.5 [0.3, 6.7]	4.3 [1.9, 6.6]	0.698
GDSd, mean [CI95%]	7.2 [1.5, 12.9]	7.9 [5.4, 10.5]e	0.411

Notes:

a Scores on the Mini Mental State Examination (MMSE) range from 0 to 30, with higher scores indicating better cognitive function;

b score on the Instrumental Activities of Daily Living for Elderly (IADL-E) range from 0 to 36, with lower score indicating a better functional independency;

c scores on the Montgomery Asberg Depression Rating Scale (MADRS) range from 0 to 60 (10 items range from 0 to 6), with higher scores indicating depressive state;

d scores on the Geriatric Depression Scale (GDS) range from 0 to 30, with higher scores indicating depressive state;

e 1 missing data.

* P < 0.05;

** P < 0.001.

Abbreviations: NC, normal control; AD, Alzheimer disease.

Table 4 Characteristics of the participants in protocol 2 (P2)

Characteristics	NC n = 19	MCI n = 19	P-values
Female, n (%)	15 (78.9%)	9 (47.4%)	0.091
Age-yrs, mean [CI95%]/(±SD)	71.7 [69.2, 74.3]	75.2 [73.2, 77.3]	0.070
Level of education, n (%)
Unknown	2 (10.5%)	2 (10.5%)	1.00
No formal education	0 (0%)	0 (0%)	–
Elementary school	1 (5.3%)	5 (26.3%)	0.405
Middle school	4 (21.0%)	7 (36.8%)	0.269
High school	4 (21.0%)	3 (15.8%)	1.00
Post-secondary education	8 (42.1%)	2 (10.5%)	0.062
MMSEa, mean [CI95%]	28.8 [28.3, 29.3]	25.8 [24.8, 26.8]	<0.001**
IADL-Eb, mean [CI95%]	9.6 [9.0, 10.1]	9.9 [9.1, 10.7]	0.488
MADRSc, mean [CI95%]	2.8 [1.1, 4.5]	4.6 [2.9, 6.4]e	0.075
GDSd, mean [CI95%]	7.1 [3.7, 10.4]e	8.4 [5.5, 11.3]	0.525

Notes:

a Scores on the Mini Mental State Examination (MMSE) range from 0 to 30, with higher scores indicating better cognitive function;

b score on the Instrumental Activities of Daily Living for Elderly (IADL-E) range from 0 to 36, with lower score indicating a better functional independency;

c scores on the Montgomery Asberg Depression Rating Scale (MADRS) range from 0 to 60 (10 items range from 0 to 6), with higher scores indicating depressive state;

d scores on the Geriatric Depression Scale (GDS) range from 0 to 30, with higher scores indicating depressive state.

e 1 missing data.

** P < 0.001.

Abbreviations: NC, normal control; MCI middle cognitive impairment.

Table 5 Evaluation of predictive performance of our assessment methodology based on the DAS scores: Leave-one-out cross-validation (LOOCV) resultsa

	DAS.,P1() score	DAS.,P2() score
K value of the K-folder for the cross-validation method, K	26	38
Number of observations of each training dataset used for classifying the new test data (ntest = 1), ntraining	25	37
Relationship between the DAS scores compared with current rating scales
Spearman’s correlation coefficient with MMSE, mean (±SD)	0.68 (±0.03)	0.81 (±0.01)
Spearman’s correlation with IADL-E, mean (±SD)	−0.64 (±0.03)	−0.05 (±0.03)
Predictive performance of the ecological assessment based on the DAS scores resultsb
Threshold of cut-off point, mean (±SD)c	0.62 (±0.01)	0.21 (±0.01)
Accuracy rate	88.5%	71.1%
Predictive positive value	93.3%	68.2%
Predictive negative value	81.8%	75%
False positive rate	3.9%	18.4%
False negative rate	7.7%	10.5%

Notes:

a The leave-one-out cross-validation was used to provide an un-biased estimation of the prevision errors of our models (the DAS scores), and therefore to assess how the results of our models will generalize to an independent data (data not used during the fitting procedure). In our context, this methodology consists to estimate the parameter set (k_1,n, k_2,n, k_3,n, k_4,n), n € {P1, P2} with the training dataset (n_training) and to define the threshold of cut-off point from the DAS scores computed from the training dataset, and then to validate this assessment methodology to classify the test data. For the LOOCV, this procedure was repeated K times: each observation in the initial sample is used once as the test data;

b predictive performance was computed according to the LOOCV methodology;

c the threshold was defined for each training dataset as the cut-off that minimizes the absolute value between the specificity and the sensitivity to optimize both.

Table 6 Spearman’s correlation coefficients between DAS_.,n() scores with Diagnosis, MMSE and IADL-E

	Diagnosis		MMSE		IADL-E
	P1a	P2b	P1	P2	P1	P2
REff	−0.50**	0.11	0.55**	−0.15	−0.53**	−0.01
Index 1	−0.63**	−0.20	0.59**	0.32	−0.62**	−0.01
Index 2	−0.69**	0.34*	0.70**	0.34*	−0.64**	0.07
Index 3	−0.76**	0.77**	0.77**	0.77**	−0.65**	−0.08
Index 4	−0.76**	0.81**	0.81**	0.81**	−0.64**	−0.06

Notes:

a For P1, Diagnosis = 0 refers to NC group, Diagnosis = 1 refers to AD group;

b for P2, Diagnosis = 0 refers to NC group, Diagnosis = 1 refers to MCI group.

* P < 0.05;

** P < 0.001.

Figure 1 DAS_.,P1 scores for AD patients and NC participants for protocol 1 (P1). The first line (A) represents the ratio of efficacy for the two groups (Grey dots for NC and Black dots for AD). Then the next lines (B–E) show the evolution of the index from ratio of efficacy (index 0) to DAS_.,P1 score (index 4) including omission (B), repetition (C), order error (D), and bad completion at the first attempt (E). Measurements represented for each participant j: (A) R_Eff = Index_0,P1(j) (Percentage of time spent in the room to behave directed to perform a listed activities). (B) ${\text{Index}}_{\text{1},\text{P1}}(\text{j})=[{\text{R}}_{\text{Eff}}(\text{j})]\times {\prod }_{\text{i}=1}^4{k}_{1,P1}^{{\text{a}}_{1,F}}$ (impact of omission mistakes on the R_Eff). (C) ${\text{Index}}_{2,\text{P1}}(\text{j})=[{\text{R}}_{\text{Eff}}(\text{j})]\times {\prod }_{\text{i}=1}^4{k}_{1,P1}^{{\text{a}}_{1,F1}(\text{j})}$ (cumulative impact of omission and repetition mistakes on the R_Eff). (D) ${\text{Index}}_{3,\text{P1}}(\text{j})=[{\text{R}}_{\text{Eff}}(\text{j})]\times {\prod }_{\text{i}=1}^2{k}_{1,P2}^{{\text{a}}_{1,F2}(j)}$ (cumulative impact of omission, repetition, and order mistakes on the R_Eff). (E) Final DAS score ${\text{S}}_{\text{j},\text{P}1}({\text{k}}_{1,\text{P}1,}\hspace{0.17em}{\text{k}}_{2,\text{P}1,}\hspace{0.17em}{\text{k}}_{3,\text{P1},}\hspace{0.17em}{\text{k}}_{4,\text{P1},})\hspace{0.17em}(\text{j})=[{\text{R}}_{\text{Eff}}(\text{j})]\times {\prod }_{\text{i}=1}^4{k}_{1,P1}^{{\text{a}}_{1,F1}\hspace{0.17em}(\text{j})}$ (cumulative impact of omission, repetition, order mistakes, and bad completion at the first attempt on the R_Eff).

Figure 2 Performance evaluation of the assessment methodology to classify observation data as AD group or NC group. (A) Specificity and sensitivity in function of the cutoff point of DAS_.,P1() scores. (B) Receiver operating curve associated with protocol 1.

Abbreviations: AD, Alzheimer’s disease; DAS, daily activity scenario; NC, normal control.

Figure 3 DAS_.,P2 scores for MCI patients and NC participants for protocol 2 (P2). The first line (A) represents the ratio of efficacy for the two groups (Grey dots for NC and Black dots for MCI). Then the next lines (B–E) show the evolution of the index from ratio of efficacy (index 0) to DAS_.,P2 score (index 4) including omission (B), repetition (C), bad achievement of activities (D), and planning errors (E). Measurements represented for each participant j: (A) R_Eff = Index_0,P2 (j) (Percentage of time spent in the room to behave directed to perform a listed activities). (B) ${\text{Index}}_{1,\text{P}2}(\text{j})=[{\text{R}}_{\text{Eff}}(\text{j})]\times k_{1,P2}^{{\alpha }_{1,F2}\hspace{0.17em}(j)}$ (impact of omission mistakes on the R_Eff). (C) ${\text{Index}}_{2,\text{P}2}(\text{j})=[{\text{R}}_{\text{Eff}}(\text{j})]\times {\prod }_{\text{i}=1}^2{k}_{1,P2}^{a_{1,F2}\hspace{0.17em}(j)}$ (cumulative impact of omission and repetition (excluding repetition of leisure activities) mistakes on the R_Eff). (D) ${\text{Index}}_{3,\text{P}2}(\text{j})=[{\text{R}}_{\text{Eff}}(\text{j})]\times {\prod }_{\text{i}=1}^3{k}_{1,P2}^{a_{1,F2}\hspace{0.17em}(j)}$ (cumulative impact of omission, repetition (excluding repetition of leisure activities) mistakes and bad achievement of activities on the R_Eff). (E) Final DAS score ${\text{S}}_{\text{j},\text{P}2}\hspace{0.17em}({\text{k}}_{1,\text{P}2,}\hspace{0.17em}{\text{k}}_{2,\text{P}2,}\hspace{0.17em}{\text{k}}_{3,\text{P}2,}\hspace{0.17em}{\text{k}}_{4,\text{P}2,})\hspace{0.17em}(\text{j})=[\text{REff}(\text{j})]\times {\prod }_{\text{i}=1}^4{k}_{1,P2}^{{\text{a}}_{1,F2}\hspace{0.17em}(j)}$ (cumulative impact of omission, repetition (excluding repetition of leisure activities) mistakes, bad achievements of activities, and planification mistakes on the R_Eff).

Figure 4 Performance evaluation of the assessment methodology to classify observation data as MCI group or NC group. (A) Specificity and sensitivity in function of the cutoff point of DAS_.,P2() scores. (B) Receiver operating curve associated with the protocol 2.

Abbreviations: DAS, daily activity scenario; MCI, mild cognitive impairment; NC, normal control.

Table 7 Ratio of efficacy and qualitative parameters in the two protocols for each diagnosis group

Raw parameters	Participants
Raw parameters for the protocol 1 (P1)	NC group (N = 10)	AD group (N = 16)
Ratio of efficacy, mean [CI(95%)]*	0.71 [0.68, 0.74]	0.61 [0.54, 0.68]
Omission of at least one activity, n (%)	0 (0%)	2 (12.5%)
Repetition of at least one activity, n (%)	0 (0%)	6 (37.5%)
Incorrect order, n (%)	0 (0%)	4 (25%)
At least one failure to complete one activity at the first time, n (%)	1 (10%)	7 (43.75%)
Raw parameters for the protocol 2 (P2)	NC group (N = 19)	MCI group (N = 19)
Ratio of efficacy, mean [CI(95%)]	0.54 [0.49, 0.59]	0.56 [0.52, 0.61]
Omission of at least one activity, n (%)	0 (0%)	3 (15.8%)
Repetition of at least one activity excluding leisure activity, n (%)	3 (15.8%)	6 (31.6%)
No respect of 2 or 3 constraints order, n (%)	6 (31.6%)	8 (42.1%)
At least one activity not achieved, n (%)**	3 (15.8%)	13 (68.4%)

Notes:

* P < 0.05;

** P < 0.001.

Abbreviations: NC, normal control; AD, Alzheimer disease; MCI, middle cognitive impairment.