Table 5.

Latent Class Model Results

Attribute	Parameter (Level)	Class 1a (Predicted Share, 43%b)		Class 2 (Predicted Share, 47%)		Class 3 (Predicted Share, 10%)
		Marginal Effect	95% CI	Marginal Effect	95% CI	Marginal Effect	95% CI
	Opt-out constant	0.58	–0.81 to 1.98	–0.72	–1.7 to 0.25	4.30c	2.16 to 6.44
Lung function	Modest deterioration (–5%)	–0.11	–0.98 to 0.75	–1.11c	–1.64 to –0.57	–0.92	–1.91 to 0.06
	No change	Referent
	Modest improvement (+5%)	0.97d	0.11 to 1.83	0.97c	0.6 to 1.35	0.2	–0.78 to 1.19
	Excellent improvement (+15%)	1.85e	0.7 to 3	1.73c	1.16 to 2.31	0.83	–0.35 to 2.01
Need for IV antibiotics	No change	Referent
	One-half the number of IV courses	0.27	–0.24 to 0.78	0.52c	0.27 to 0.77	–0.21	–0.83 to 0.42
Abdominal symptoms	No change	Referent
	Improvement in symptoms	–0.09	–0.68 to 0.5	0.57e	0.23 to 0.91	0.17	–0.58 to 0.93
	Improvement in symptoms and a reduction in pancreatic enzymes	0.21	–0.35 to 0.77	0.67c	0.32 to 1.02	0.34	–0.47 to 1.15
Life expectancy	No change	Referent
	Increases by 5 y	1.85c	0.94 to 2.75	0.49	–0.02 to 1.01	1.79e	0.45 to 3.12
	Increases by 10 y	4.79c	3.57 to 6.02	1.49c	0.96 to 2.01	2.45e	1.04 to 3.86
	Increases by 15 y	6.62c	5.08 to 8.16	1.71c	1.1 to 2.32	3.05c	1.51 to 4.59
Overall quality of life	No change	Referent
	Good improvement (+10%)	0.89c	0.37 to 1.42	0.29	0 to 0.58	0.46	–0.35 to 1.27
	Excellent improvement (20%)	1.14c	0.57 to 1.71	0.9c	0.56 to 1.25	0.72	–0.08 to 1.52
Use of inhaled medicines	No change	Referent
	A modest reduction in time spent (25%)	0.24	–0.37 to 0.85	0.20	–0.11 to 0.5	1.13e	0.29 to 1.97
	A large reduction in time spent (50%)	0.95e	0.41 to 1.49	0.34e	0.03 to 0.65	0.99e	0.08 to 1.9
Physiotherapy/ACT	No change	Referent
	Time spent on physiotherapy is halved	0.53e	0.02 to 1.04	0.17	–0.1 to 0.45	0.45	–0.32 to 1.22
	Able to fully stop physiotherapy	0.71	–0.17 to 1.59	0.80c	0.47 to 1.13	0.91e	0.08 to 1.74
Model statistics	No. of observations, 3,708; McFadden R2, N/A; LL, –697; AIC, 1,501; BIC, 1,831

ACT = airway clearance therapy; AIC = Akaike information criterion; BIC = Bayesian information criterion; LL = log likelihood; N/A = not applicable.

^aAs marginal effects are calculated relative to the reference level for each class, they are not directly comparable across classes; the ratios between effects, however, may be directly compared.

^bProbability of class membership equates to percentage class share of the population.

^cP < .001.

^dP < .1.

^eP < .05.