Clinical characteristics do not reliably identify non-adherence in patients with uncontrolled hypertension

Figures & data

Table 1. Baseline characteristics of the derivation set and validation set.

	University Medical Centre Utrecht		University Hospital Birmingham
	n = 495	Missings	n = 240	Missings
Clinical characteristics
Male sex	250 (51%)	0 (0%)	123 (51%)	0 (0%)
Age (years)	57 ± 14	0 (0%)	57 ± 14	0 (0%)
Current smoker	53 (11%)	0 (0%)	39 (24%)	75 (31%)
Body mass index (kg/m2)	28.2 ± 5.3	5 (1%)	32.2 ± 5.9	65 (27%)
History of cardiovascular disease	133 (27%)	0 (0%)	58 (24%)	0 (0%)
Laboratory values
Estimated GFR (ml/min/1.73m2)	81 ± 27	22 (4%)	75 ± 30	17 (7%)
HbA1c (mmol/mol)	39 ± 9	22 (4%)	46 ± 16	115 (48%)
Total cholesterol (mmol/L)	5.2 ± 1.2	22 (4%)	5.0 ± 1.3	58 (24%)
LDL-cholesterol (mmol/L)	3.1 ± 1.0	13 (3%)	3.0 ± 1.1	66 (28%)
HDL-cholesterol (mmol/L)	1.4 ± 0.4	22 (4%)	1.2 ± 0.4	60 (25%)
Triglycerides (mmol/L)	1.7 ± 1.2	22 (4%)	1.9 ± 2.0	58 (24%)
Blood pressure
Office systolic blood pressure (mmHg)	165 ± 28	1 (0%)	165 ± 27	1 (0%)
Office diastolic blood pressure (mmHg)	94 ± 14	1 (0%)	93 ± 18	3 (1%)
Office heart rate (bpm)	73 ± 13	3 (1%)	NA	240 (100%)
Ambulatory systolic blood pressure (mmHg)	138 ± 19	18 (4%)	153 ± 20	199 (83%)
Ambulatory diastolic blood pressure (mmHg)	82 ± 11	18 (4%)	88 ± 13	199 (83%)
Medication use
Antihypertensive medication tablets (n/day)	3 (1-12)	0 (0%)	3 (1-10)	0 (0%)
class A	336 (68%)	0 (0%)	210 (88%)	0 (0%)
class B	204 (41%)	0 (0%)	104 (43%)	0 (0%)
class C	314 (63%)	0 (0%)	194 (81%)	0 (0%)
class D	253 (51%)	0 (0%)	147 (61%)	0 (0%)
class E	122 (25%)	0 (0%)	139 (58%)	0 (0%)
Non-adherence	93 (19%)	0 (0%)	105 (44%)	0 (0%)
Partial non-adherence	73 (15%)	0 (0%)	37 (15%)	0 (0%)
Full non-adherence	20 (4%)	0 (0%)	67 (28%)	0 (0%)

All data in n (%) or mean ± standard deviation. GFR, glomerular filtration rate (calculated with Chronic Kidney Disease Epidemiology Collaboration [CKDEPI] formula), LDL, low-density lipoprotein, HDL, high-density lipoprotein, class A, ACE-inhibitors, angiotensin receptor blockers and direct renin inhibitors, class B, beta-blockers, class C, calcium channel blockers, class D, diuretics, class E, others.

Table 2. Model coefficients and odds ratios.

	Coëfficiënt	Odds ratio (95% CI)
Intercept	−6.5909
Age (per year)	−0.0269	0.97 (0.95–0.99)
Sex (female)	−0.0359	0.96 (0.59–1.57)
Office systolic blood pressure (mmHg)	0.0215	1.02 (1.01–1.03)
Office heart rate (beats/min)	0.0053	1.01 (0.99–1.02)
Number of tablets (n/day)	0.3028	1.35 (1.16–1.58)
History of CVD (yes)	−0.1977	0.82 (0.46–1.47)
BMI (kg/m^2)	0.0614	1.06 (1.01–1.12)

CVD, cardiovascular disease; BMI, body mass index.

Figure 1. External validation of the clinical screening tool for non-adherence. (A) Plot of external calibration of clinical screening tool in Heartlands Hospital population showing the agreement between predicted and observed probabilities of non-adherence after recalibration. (B) Receiver operating characteristics (ROC) curve showing the discriminative performance of the diagnostic tool.

Table 3. Test characteristics and proportion of patients spared chemical adherence testing.

	Cut-off value of predicted probability
	10.0%	12.5%	15.0%	17.5%	20.0%	22.5%	25.0%
Sensitivity	0.99 (0.53–1.00)	0.98 (0.53–1.00)	0.97 (0.64–1.00)	0.95 (0.73–0.99)	0.93 (0.82–0.97)	0.90 (0.80–0.95)	0.86 (0.77–0.92)
Specificity	0.00 (0.00–1.00)	0.03 (0.00–0.74)	0.06 (0.02–0.18)	0.10 (0.05–0.18)	0.15 (0.09–0.23)	0.19 (0.12–0.27)	0.24 (0.17–0.34)
Positive predictive value	0.44 (0.37–0.50)	0.44 (0.37–0.50)	0.44 (0.38–0.51)	0.45 (0.38–0.51)	0.46 (0.39–0.52)	0.46 (0.39–0.53)	0.47 (0.40–0.54)
Negative predictive value	0.18 (0.00–1.00)	0.65 (0.00–1.00)	0.74 (0.10–0.99)	0.72 (0.23–0.96)	0.72 (0.45–0.89)	0.70 (0.50–0.85)	0.70 (0.53–0.83)
Proportion of patients spared testing (%)	1 (0–3)	3 (0–5)	5 (2–8)	8 (4–12)	12 (7–17)	15 (10–20)	20 (14–25)

The proportion of patients spared intensive testing is the proportion of patients with a predicted probability equal to or below the cut-off value. Estimates and corresponding Bootstrap-based 95% confidence intervals are presented for different cut-off values of the predicted probability by the diagnostic tool.

Data availability statement

The data that support the findings of this study are available from the corresponding author, WS, upon reasonable request.