Pre- and in-therapy predictive score models of adult OSAS patients with poor adherence pattern on nCPAP therapy

Figures & data

Figure 1 Distributions of adherence to nCPAP among newly diagnosed OSAS patients over time.

Notes: (A) Distribution of 76 individuals’ adherence by three patterns of adherence; (B) means and 95% CI of each pattern of adherence. Each subject is represented by a single line in (A). Solid line: good adherence pattern; dashed line: moderate adherence pattern; dotted line: poor adherence pattern.
Abbreviations: nCPAP, nasal continuous positive airway pressure; OSAS, obstructive sleep apnea/hypopnea syndrome; CI, confidence interval.

Table 1 Evaluation parameter of bss/tss from K-means cluster analysis and the number of subjects in the smallest cluster from K-means cluster analysis

Number of clusters	bss/tss (%)	Number of patients in the smallest cluster
5	92.1	6
4	90.1	8
3	86.3	14

Abbreviations: bss, between-cluster sum of squares; tss, total sum of squares.

Table 2 Baseline and 3rd week information and the average nightly adherence during the study according to the three patterns of adherence

Items	Poor, n=14	Moderate, n=19	Good, n=43	P-value
Age, yr#	53.1 (16.6)^a	48.2 (14.3)^a	46.5 (13.9)^a	0.343
Sex*
Male	5 (8.6)^a	15 (25.9)^b	38 (65.5)^b	<0.001
Female	9 (50.0)	4 (22.2)	5 (22.8)
BMI,# kg/m^2	24.9 (3.3)^a	27.1 (3.1)^a,b	27.6 (3.3)^b	0.032
WHR#	0.9 (0.1)^a	1.0 (0.1)^b	1.0 (0.1)^b	0.039
AHI,^ events/h	30 (18, 48)^a	49 (32, 70)^b	53 (33, 70)^b	0.009
SaOR^	0.2 (0.1, 0.3)^a	0.2 (0.1, 0.3)^a	0.2 (0.1, 0.3)^a	0.818
ESS, score#
Baseline	10.6 (4.8)^a	12.3 (5.0)^a	13.5 (4.8)^a	0.167
3rd week	6.2 (2)^a	6.2 (2.7)^a	5.6 (2.4)^a	0.538
Change	−4.4 (4.2)^a	−6.1 (4.6)^a,b	−7.9 (4.4)^b	0.034
Paired test P-value&	0.002	<0.001	<0.001
FSS, score#
Baseline	45.5 (12.2)^a	47.7 (12.7)^a	42.6 (11.6)^a	0.261
3rd week	47.4 (11.7)^a	43.2 (12.2)^a	33.0 (10.9)^b	0.001
Change	1.9 (3.4)^a	−4.5 (3.3)^b	−9.6 (3.2)^c	<0.001
Paired test P-value&	0.060	<0.001	<0.001
SDS, score#
Baseline	43.2 (3.1)^a	39.2 (4.5)^b	34.7 (4.0)^c	<0.001
3rd week	43.2 (2.9)^a	37.1 (4.7)^b	31.3 (4.3)^c	<0.001
Change	0 (1.6)^a	−2.1 (1.5)^b	−3.4 (1.6)^c	<0.001
Paired test P-value&	1	<0.001	<0.001
PSQI, score^
Baseline	10 (6, 11)^a	5 (3, 6)^b	4 (3, 5)^b	<0.001
3rd week	10 (6, 11)^a	5 (4, 7)^b	4 (3, 5)^c	<0.001
Change	0 (−1, 1)^a,b	1 (0, 1)^a	0 (−1, 0)^b	0.003
Paired test P-value€	0.746	0.037	0.004
First 3-week adherence,# h/night	4.5 (0.9)^a	5.5 (0.5)^b	5.8 (0.5)^c	<0.001
4–12 weeks’ adherence,# h/night	1.6 (1.1)^a	5.2 (0.7)^b	7.2 (0.3)^c	<0.001
Overall adherence,# h/night	2.3 (0.9)^a	5.3 (0.6)^b	6.8 (0.3)^c	<0.001

Notes: In each row, those means, medians, or proportions that share a common superscript (a–c) do not differ significantly. Those means, medians, or proportions that do not share a common superscript differ significantly from one another (P<0.05, t-test, Wilcoxon rank sum test, chi-square test). Given an example of BMI, poor pattern did not differ significantly from moderate pattern, but differed significantly from good pattern; moderate pattern did not differ significantly from both poor and good patterns.

# Mean (SD) with ANOVA F-test;

* frequency (%) with Fisher’s exact test;

^ median (IQR) with Kruskal–Wallis test;

& paired t-test (baseline vs 3rd week);

€ Wilcoxon paired signed rank test (baseline vs 3rd week). Change = 3rd week score – baseline score: negative value means improved.

Abbreviations: BMI, body mass index; WHR, waist–hip ratio; AHI, apnea/hypopnea index; ANOVA, analysis of variance; IQR, interquartile range; SaOR, (average degree of blood oxygen saturation – lowest degree of blood oxygen saturation)/average degree of blood oxygen saturation; ESS, Epworth Sleepiness Scale (0–24, higher score indicated sleepier); FSS, Fatigue Severity Scale (9–63, higher score indicated more fatigue); SDS, Zung’s Self-Rating Depression Scale (20–80, higher score indicated more depressed); PSQI, The Pittsburgh Sleep Quality Index (0–21, higher score indicated lower sleep quality).

Table 3 Cumulative logit regression results of baseline predictors of 12-week adherence pattern

Items	Coefficient	Ordinal OR (95% CI)	Weight
ESS, score	−0.271	0.763 (0.651, 0.893)	−3
SDS, score	0.474	1.461 (1.238, 1.724)	5
PSQI, score	0.816	2.261 (1.427, 3.584)	9

Notes: Adherence order: good < moderate < poor. Independent variables of baseline initially included in the model but subsequently removed as nonsignificant: age, sex, BMI, WHR, AHI, FSS score.

Abbreviations: OR, odds ratio; 95% CI, 95% of confidence interval; ESS, Epworth Sleepiness Scale (0–24, higher score indicates sleepier); SDS, Zung’s Self-Rating Depression Scale (20–80, higher score indicates more depressed); PSQI, The Pittsburgh Sleep Quality Index (0–21, higher score indicates lower sleep quality); BMI, body mass index; WHR, waist -hip ratio; AHI, apnea/hypopnea index; FSS, Fatigue Severity Scale (9–63, higher score indicated more fatigue).

Table 4 Cumulative logit regression results of first-3-week therapy predictors of 12-week adherence pattern

Items	Coefficient	Ordinal OR (95% CI)	Weight
ESS, score	−0.591	0.554 (0.331, 0.926)	−4
ESS change, score	−0.779	0.459 (0.243, 0.868)	−5
PSQI, score	0.935	2.548 (1.454, 4.465)	6
PSQI change, score	1.077	2.937 (1.273, 6.773)	7
FSS change, score	1.268	3.556 (1.788, 7.070)	8

Notes: Dependent variable order: good adherence < moderate adherence < poor adherence. Independent variables initially included in the model, but subsequently removed as nonsignificant: baseline information – age, sex, BMI, WHR, AHI, FSS score, SDS score; the change in SDS score; adherence in first therapy phase. Change = 3rd week score – baseline score: negative value means improved.

Abbreviations: OR, odds ratio; 95% CI, 95% of confidence interval; ESS, Epworth Sleepiness Scale (0–24, higher score indicates sleepier); PSQI, The Pittsburgh Sleep Quality Index (0–21, higher score indicates lower sleep quality); FSS, Fatigue Severity Scale (7–63, higher score indicates more fatigue); BMI, body mass index; WHR, waist-hip ratio; AHI, apnea/hypopnea index; FSS, Fatigue Severity Scale (9–63, higher score indicated more fatigue); SDS, Zung’s self-rating Depression scale (20–80, higher score indicated more depressed).

Figure 2 Distributions of baseline and first 3 weeks predictive score among three adherence patterns.

Notes: Each score is represented by a single dot in the figure. Reference line marks the selected cut-off point.

Figure 3 ROC curves of baseline and first-3-weeks predictive scores. Each symbol represents a different score.

Abbreviations: AUC, area under the curve; ROC, receiver operating characteristics.

Table 5 Bootstrap evaluation of predictive ability of baseline and first-3-week predictive scores using the selected cut-off pointsa (mean and 95% bootstrap confidence interval)

Items	Resample (n=500)
	Baseline score cut-off point: <221/≥221	3-week score cut-off point: <10/≥10
PPV	0.824 (0.815, 0.832)	0.936 (0.930, 0.942)
NPV	1 (1, 1)	1 (1, 1)

Notes:

a Cut-off points corresponding to minimum false positive proportion consistent with 100% true positive proportion.

Abbreviations: PPV, positive predictive value; NPV, negative predictive value.